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 vnode_t *realvp; 1509 char *nm, *lnm; 1510 uint_t len, llen; 1511 int syncval = 0; 1512 struct nfs4_svgetit_arg sarg; 1513 struct nfs4_ntov_table ntov; 1514 struct statvfs64 sb; 1515 nfsstat4 status; 1516 1517 DTRACE_NFSV4_2(op__create__start, struct compound_state *, cs, 1518 CREATE4args *, args); 1519 1520 resp->attrset = 0; 1521 1522 if (dvp == NULL) { 1523 *cs->statusp = resp->status = NFS4ERR_NOFILEHANDLE; 1524 goto out; 1525 } 1526 1527 /* 1528 * If there is an unshared filesystem mounted on this vnode, 1529 * do not allow to create an object in this directory. 1530 */ 1531 if (vn_ismntpt(dvp)) { 1532 *cs->statusp = resp->status = NFS4ERR_ACCESS; 1533 goto out; 1534 } 1535 1536 /* Verify that type is correct */ 1537 switch (args->type) { 1538 case NF4LNK: 1539 case NF4BLK: 1540 case NF4CHR: 1541 case NF4SOCK: 1542 case NF4FIFO: 1543 case NF4DIR: 1544 break; 1545 default: 1546 *cs->statusp = resp->status = NFS4ERR_BADTYPE; 1547 goto out; 1548 }; 1549 1550 if (cs->access == CS_ACCESS_DENIED) { 1551 *cs->statusp = resp->status = NFS4ERR_ACCESS; 1552 goto out; 1553 } 1554 if (dvp->v_type != VDIR) { 1555 *cs->statusp = resp->status = NFS4ERR_NOTDIR; 1556 goto out; 1557 } 1558 if (!utf8_dir_verify(&args->objname)) { 1559 *cs->statusp = resp->status = NFS4ERR_INVAL; 1560 goto out; 1561 } 1562 1563 if (rdonly4(cs->exi, cs->vp, req)) { 1564 *cs->statusp = resp->status = NFS4ERR_ROFS; 1565 goto out; 1566 } 1567 1568 /* 1569 * Name of newly created object 1570 */ 1571 nm = utf8_to_fn(&args->objname, &len, NULL); 1572 if (nm == NULL) { 1573 *cs->statusp = resp->status = NFS4ERR_INVAL; 1574 goto out; 1575 } 1576 1577 if (len > MAXNAMELEN) { 1578 *cs->statusp = resp->status = NFS4ERR_NAMETOOLONG; 1579 kmem_free(nm, len); 1580 goto out; 1581 } 1582 1583 resp->attrset = 0; 1584 1585 sarg.sbp = &sb; 1586 nfs4_ntov_table_init(&ntov); 1587 1588 status = do_rfs4_set_attrs(&resp->attrset, 1589 &args->createattrs, cs, &sarg, &ntov, NFS4ATTR_SETIT); 1590 1591 if (sarg.vap->va_mask == 0 && status == NFS4_OK) 1592 status = NFS4ERR_INVAL; 1593 1594 if (status != NFS4_OK) { 1595 *cs->statusp = resp->status = status; 1596 kmem_free(nm, len); 1597 nfs4_ntov_table_free(&ntov, &sarg); 1598 resp->attrset = 0; 1599 goto out; 1600 } 1601 1602 /* Get "before" change value */ 1603 bva.va_mask = AT_CTIME|AT_SEQ; 1604 error = VOP_GETATTR(dvp, &bva, 0, cr, NULL); 1605 if (error) { 1606 *cs->statusp = resp->status = puterrno4(error); 1607 kmem_free(nm, len); 1608 nfs4_ntov_table_free(&ntov, &sarg); 1609 resp->attrset = 0; 1610 goto out; 1611 } 1612 NFS4_SET_FATTR4_CHANGE(resp->cinfo.before, bva.va_ctime) 1613 1614 vap = sarg.vap; 1615 1616 /* 1617 * Set default initial values for attributes when not specified 1618 * in createattrs. 1619 */ 1620 if ((vap->va_mask & AT_UID) == 0) { 1621 vap->va_uid = crgetuid(cr); 1622 vap->va_mask |= AT_UID; 1623 } 1624 if ((vap->va_mask & AT_GID) == 0) { 1625 vap->va_gid = crgetgid(cr); 1626 vap->va_mask |= AT_GID; 1627 } 1628 1629 vap->va_mask |= AT_TYPE; 1630 switch (args->type) { 1631 case NF4DIR: 1632 vap->va_type = VDIR; 1633 if ((vap->va_mask & AT_MODE) == 0) { 1634 vap->va_mode = 0700; /* default: owner rwx only */ 1635 vap->va_mask |= AT_MODE; 1636 } 1637 error = VOP_MKDIR(dvp, nm, vap, &vp, cr, NULL, 0, NULL); 1638 if (error) 1639 break; 1640 1641 /* 1642 * Get the initial "after" sequence number, if it fails, 1643 * set to zero 1644 */ 1645 iva.va_mask = AT_SEQ; 1646 if (VOP_GETATTR(dvp, &iva, 0, cs->cr, NULL)) 1647 iva.va_seq = 0; 1648 break; 1649 case NF4LNK: 1650 vap->va_type = VLNK; 1651 if ((vap->va_mask & AT_MODE) == 0) { 1652 vap->va_mode = 0700; /* default: owner rwx only */ 1653 vap->va_mask |= AT_MODE; 1654 } 1655 1656 /* 1657 * symlink names must be treated as data 1658 */ 1659 lnm = utf8_to_str(&args->ftype4_u.linkdata, &llen, NULL); 1660 1661 if (lnm == NULL) { 1662 *cs->statusp = resp->status = NFS4ERR_INVAL; 1663 kmem_free(nm, len); 1664 nfs4_ntov_table_free(&ntov, &sarg); 1665 resp->attrset = 0; 1666 goto out; 1667 } 1668 1669 if (llen > MAXPATHLEN) { 1670 *cs->statusp = resp->status = NFS4ERR_NAMETOOLONG; 1671 kmem_free(nm, len); 1672 kmem_free(lnm, llen); 1673 nfs4_ntov_table_free(&ntov, &sarg); 1674 resp->attrset = 0; 1675 goto out; 1676 } 1677 1678 error = VOP_SYMLINK(dvp, nm, vap, lnm, cr, NULL, 0); 1679 if (lnm != NULL) 1680 kmem_free(lnm, llen); 1681 if (error) 1682 break; 1683 1684 /* 1685 * Get the initial "after" sequence number, if it fails, 1686 * set to zero 1687 */ 1688 iva.va_mask = AT_SEQ; 1689 if (VOP_GETATTR(dvp, &iva, 0, cs->cr, NULL)) 1690 iva.va_seq = 0; 1691 1692 error = VOP_LOOKUP(dvp, nm, &vp, NULL, 0, NULL, cr, 1693 NULL, NULL, NULL); 1694 if (error) 1695 break; 1696 1697 /* 1698 * va_seq is not safe over VOP calls, check it again 1699 * if it has changed zero out iva to force atomic = FALSE. 1700 */ 1701 iva2.va_mask = AT_SEQ; 1702 if (VOP_GETATTR(dvp, &iva2, 0, cs->cr, NULL) || 1703 iva2.va_seq != iva.va_seq) 1704 iva.va_seq = 0; 1705 break; 1706 default: 1707 /* 1708 * probably a special file. 1709 */ 1710 if ((vap->va_mask & AT_MODE) == 0) { 1711 vap->va_mode = 0600; /* default: owner rw only */ 1712 vap->va_mask |= AT_MODE; 1713 } 1714 syncval = FNODSYNC; 1715 /* 1716 * We know this will only generate one VOP call 1717 */ 1718 vp = do_rfs4_op_mknod(args, resp, req, cs, vap, nm); 1719 1720 if (vp == NULL) { 1721 kmem_free(nm, len); 1722 nfs4_ntov_table_free(&ntov, &sarg); 1723 resp->attrset = 0; 1724 goto out; 1725 } 1726 1727 /* 1728 * Get the initial "after" sequence number, if it fails, 1729 * set to zero 1730 */ 1731 iva.va_mask = AT_SEQ; 1732 if (VOP_GETATTR(dvp, &iva, 0, cs->cr, NULL)) 1733 iva.va_seq = 0; 1734 1735 break; 1736 } 1737 kmem_free(nm, len); 1738 1739 if (error) { 1740 *cs->statusp = resp->status = puterrno4(error); 1741 } 1742 1743 /* 1744 * Force modified data and metadata out to stable storage. 1745 */ 1746 (void) VOP_FSYNC(dvp, 0, cr, NULL); 1747 1748 if (resp->status != NFS4_OK) { 1749 if (vp != NULL) 1750 VN_RELE(vp); 1751 nfs4_ntov_table_free(&ntov, &sarg); 1752 resp->attrset = 0; 1753 goto out; 1754 } 1755 1756 /* 1757 * Finish setup of cinfo response, "before" value already set. 1758 * Get "after" change value, if it fails, simply return the 1759 * before value. 1760 */ 1761 ava.va_mask = AT_CTIME|AT_SEQ; 1762 if (VOP_GETATTR(dvp, &ava, 0, cr, NULL)) { 1763 ava.va_ctime = bva.va_ctime; 1764 ava.va_seq = 0; 1765 } 1766 NFS4_SET_FATTR4_CHANGE(resp->cinfo.after, ava.va_ctime); 1767 1768 /* 1769 * True verification that object was created with correct 1770 * attrs is impossible. The attrs could have been changed 1771 * immediately after object creation. If attributes did 1772 * not verify, the only recourse for the server is to 1773 * destroy the object. Maybe if some attrs (like gid) 1774 * are set incorrectly, the object should be destroyed; 1775 * however, seems bad as a default policy. Do we really 1776 * want to destroy an object over one of the times not 1777 * verifying correctly? For these reasons, the server 1778 * currently sets bits in attrset for createattrs 1779 * that were set; however, no verification is done. 1780 * 1781 * vmask_to_nmask accounts for vattr bits set on create 1782 * [do_rfs4_set_attrs() only sets resp bits for 1783 * non-vattr/vfs bits.] 1784 * Mask off any bits set by default so as not to return 1785 * more attrset bits than were requested in createattrs 1786 */ 1787 nfs4_vmask_to_nmask(sarg.vap->va_mask, &resp->attrset); 1788 resp->attrset &= args->createattrs.attrmask; 1789 nfs4_ntov_table_free(&ntov, &sarg); 1790 1791 error = makefh4(&cs->fh, vp, cs->exi); 1792 if (error) { 1793 *cs->statusp = resp->status = puterrno4(error); 1794 } 1795 1796 /* 1797 * The cinfo.atomic = TRUE only if we got no errors, we have 1798 * non-zero va_seq's, and it has incremented by exactly one 1799 * during the creation and it didn't change during the VOP_LOOKUP 1800 * or VOP_FSYNC. 1801 */ 1802 if (!error && bva.va_seq && iva.va_seq && ava.va_seq && 1803 iva.va_seq == (bva.va_seq + 1) && iva.va_seq == ava.va_seq) 1804 resp->cinfo.atomic = TRUE; 1805 else 1806 resp->cinfo.atomic = FALSE; 1807 1808 /* 1809 * Force modified metadata out to stable storage. 1810 * 1811 * if a underlying vp exists, pass it to VOP_FSYNC 1812 */ 1813 if (VOP_REALVP(vp, &realvp, NULL) == 0) 1814 (void) VOP_FSYNC(realvp, syncval, cr, NULL); 1815 else 1816 (void) VOP_FSYNC(vp, syncval, cr, NULL); 1817 1818 if (resp->status != NFS4_OK) { 1819 VN_RELE(vp); 1820 goto out; 1821 } 1822 if (cs->vp) 1823 VN_RELE(cs->vp); 1824 1825 cs->vp = vp; 1826 *cs->statusp = resp->status = NFS4_OK; 1827 out: 1828 DTRACE_NFSV4_2(op__create__done, struct compound_state *, cs, 1829 CREATE4res *, resp); 1830 } 1831 1832 /*ARGSUSED*/ 1833 static void 1834 rfs4_op_delegpurge(nfs_argop4 *argop, nfs_resop4 *resop, struct svc_req *req, 1835 struct compound_state *cs) 1836 { 1837 DTRACE_NFSV4_2(op__delegpurge__start, struct compound_state *, cs, 1838 DELEGPURGE4args *, &argop->nfs_argop4_u.opdelegpurge); 1839 1840 rfs4_op_inval(argop, resop, req, cs); 1841 1842 DTRACE_NFSV4_2(op__delegpurge__done, struct compound_state *, cs, 1843 DELEGPURGE4res *, &resop->nfs_resop4_u.opdelegpurge); 1844 } 1845 1846 /*ARGSUSED*/ 1847 static void 1848 rfs4_op_delegreturn(nfs_argop4 *argop, nfs_resop4 *resop, struct svc_req *req, 1849 struct compound_state *cs) 1850 { 1851 DELEGRETURN4args *args = &argop->nfs_argop4_u.opdelegreturn; 1852 DELEGRETURN4res *resp = &resop->nfs_resop4_u.opdelegreturn; 1853 rfs4_deleg_state_t *dsp; 1854 nfsstat4 status; 1855 1856 DTRACE_NFSV4_2(op__delegreturn__start, struct compound_state *, cs, 1857 DELEGRETURN4args *, args); 1858 1859 status = rfs4_get_deleg_state(&args->deleg_stateid, &dsp); 1860 resp->status = *cs->statusp = status; 1861 if (status != NFS4_OK) 1862 goto out; 1863 1864 /* Ensure specified filehandle matches */ 1865 if (cs->vp != dsp->finfo->vp) { 1866 resp->status = *cs->statusp = NFS4ERR_BAD_STATEID; 1867 } else 1868 rfs4_return_deleg(dsp, FALSE); 1869 1870 rfs4_update_lease(dsp->client); 1871 1872 rfs4_deleg_state_rele(dsp); 1873 out: 1874 DTRACE_NFSV4_2(op__delegreturn__done, struct compound_state *, cs, 1875 DELEGRETURN4res *, resp); 1876 } 1877 1878 /* 1879 * Check to see if a given "flavor" is an explicitly shared flavor. 1880 * The assumption of this routine is the "flavor" is already a valid 1881 * flavor in the secinfo list of "exi". 1882 * 1883 * e.g. 1884 * # share -o sec=flavor1 /export 1885 * # share -o sec=flavor2 /export/home 1886 * 1887 * flavor2 is not an explicitly shared flavor for /export, 1888 * however it is in the secinfo list for /export thru the 1889 * server namespace setup. 1890 */ 1891 int 1892 is_exported_sec(int flavor, struct exportinfo *exi) 1893 { 1894 int i; 1895 struct secinfo *sp; 1896 1897 sp = exi->exi_export.ex_secinfo; 1898 for (i = 0; i < exi->exi_export.ex_seccnt; i++) { 1899 if (flavor == sp[i].s_secinfo.sc_nfsnum || 1900 sp[i].s_secinfo.sc_nfsnum == AUTH_NONE) { 1901 return (SEC_REF_EXPORTED(&sp[i])); 1902 } 1903 } 1904 1905 /* Should not reach this point based on the assumption */ 1906 return (0); 1907 } 1908 1909 /* 1910 * Check if the security flavor used in the request matches what is 1911 * required at the export point or at the root pseudo node (exi_root). 1912 * 1913 * returns 1 if there's a match or if exported with AUTH_NONE; 0 otherwise. 1914 * 1915 */ 1916 static int 1917 secinfo_match_or_authnone(struct compound_state *cs) 1918 { 1919 int i; 1920 struct secinfo *sp; 1921 1922 /* 1923 * Check cs->nfsflavor (from the request) against 1924 * the current export data in cs->exi. 1925 */ 1926 sp = cs->exi->exi_export.ex_secinfo; 1927 for (i = 0; i < cs->exi->exi_export.ex_seccnt; i++) { 1928 if (cs->nfsflavor == sp[i].s_secinfo.sc_nfsnum || 1929 sp[i].s_secinfo.sc_nfsnum == AUTH_NONE) 1930 return (1); 1931 } 1932 1933 return (0); 1934 } 1935 1936 /* 1937 * Check the access authority for the client and return the correct error. 1938 */ 1939 nfsstat4 1940 call_checkauth4(struct compound_state *cs, struct svc_req *req) 1941 { 1942 int authres; 1943 1944 /* 1945 * First, check if the security flavor used in the request 1946 * are among the flavors set in the server namespace. 1947 */ 1948 if (!secinfo_match_or_authnone(cs)) { 1949 *cs->statusp = NFS4ERR_WRONGSEC; 1950 return (*cs->statusp); 1951 } 1952 1953 authres = checkauth4(cs, req); 1954 1955 if (authres > 0) { 1956 *cs->statusp = NFS4_OK; 1957 if (! (cs->access & CS_ACCESS_LIMITED)) 1958 cs->access = CS_ACCESS_OK; 1959 } else if (authres == 0) { 1960 *cs->statusp = NFS4ERR_ACCESS; 1961 } else if (authres == -2) { 1962 *cs->statusp = NFS4ERR_WRONGSEC; 1963 } else { 1964 *cs->statusp = NFS4ERR_DELAY; 1965 } 1966 return (*cs->statusp); 1967 } 1968 1969 /* 1970 * bitmap4_to_attrmask is called by getattr and readdir. 1971 * It sets up the vattr mask and determines whether vfsstat call is needed 1972 * based on the input bitmap. 1973 * Returns nfsv4 status. 1974 */ 1975 static nfsstat4 1976 bitmap4_to_attrmask(bitmap4 breq, struct nfs4_svgetit_arg *sargp) 1977 { 1978 int i; 1979 uint_t va_mask; 1980 struct statvfs64 *sbp = sargp->sbp; 1981 1982 sargp->sbp = NULL; 1983 sargp->flag = 0; 1984 sargp->rdattr_error = NFS4_OK; 1985 sargp->mntdfid_set = FALSE; 1986 if (sargp->cs->vp) 1987 sargp->xattr = get_fh4_flag(&sargp->cs->fh, 1988 FH4_ATTRDIR | FH4_NAMEDATTR); 1989 else 1990 sargp->xattr = 0; 1991 1992 /* 1993 * Set rdattr_error_req to true if return error per 1994 * failed entry rather than fail the readdir. 1995 */ 1996 if (breq & FATTR4_RDATTR_ERROR_MASK) 1997 sargp->rdattr_error_req = 1; 1998 else 1999 sargp->rdattr_error_req = 0; 2000 2001 /* 2002 * generate the va_mask 2003 * Handle the easy cases first 2004 */ 2005 switch (breq) { 2006 case NFS4_NTOV_ATTR_MASK: 2007 sargp->vap->va_mask = NFS4_NTOV_ATTR_AT_MASK; 2008 return (NFS4_OK); 2009 2010 case NFS4_FS_ATTR_MASK: 2011 sargp->vap->va_mask = NFS4_FS_ATTR_AT_MASK; 2012 sargp->sbp = sbp; 2013 return (NFS4_OK); 2014 2015 case NFS4_NTOV_ATTR_CACHE_MASK: 2016 sargp->vap->va_mask = NFS4_NTOV_ATTR_CACHE_AT_MASK; 2017 return (NFS4_OK); 2018 2019 case FATTR4_LEASE_TIME_MASK: 2020 sargp->vap->va_mask = 0; 2021 return (NFS4_OK); 2022 2023 default: 2024 va_mask = 0; 2025 for (i = 0; i < nfs4_ntov_map_size; i++) { 2026 if ((breq & nfs4_ntov_map[i].fbit) && 2027 nfs4_ntov_map[i].vbit) 2028 va_mask |= nfs4_ntov_map[i].vbit; 2029 } 2030 2031 /* 2032 * Check is vfsstat is needed 2033 */ 2034 if (breq & NFS4_FS_ATTR_MASK) 2035 sargp->sbp = sbp; 2036 2037 sargp->vap->va_mask = va_mask; 2038 return (NFS4_OK); 2039 } 2040 /* NOTREACHED */ 2041 } 2042 2043 /* 2044 * bitmap4_get_sysattrs is called by getattr and readdir. 2045 * It calls both VOP_GETATTR and VFS_STATVFS calls to get the attrs. 2046 * Returns nfsv4 status. 2047 */ 2048 static nfsstat4 2049 bitmap4_get_sysattrs(struct nfs4_svgetit_arg *sargp) 2050 { 2051 int error; 2052 struct compound_state *cs = sargp->cs; 2053 vnode_t *vp = cs->vp; 2054 2055 if (sargp->sbp != NULL) { 2056 if (error = VFS_STATVFS(vp->v_vfsp, sargp->sbp)) { 2057 sargp->sbp = NULL; /* to identify error */ 2058 return (puterrno4(error)); 2059 } 2060 } 2061 2062 return (rfs4_vop_getattr(vp, sargp->vap, 0, cs->cr)); 2063 } 2064 2065 static void 2066 nfs4_ntov_table_init(struct nfs4_ntov_table *ntovp) 2067 { 2068 ntovp->na = kmem_zalloc(sizeof (union nfs4_attr_u) * nfs4_ntov_map_size, 2069 KM_SLEEP); 2070 ntovp->attrcnt = 0; 2071 ntovp->vfsstat = FALSE; 2072 } 2073 2074 static void 2075 nfs4_ntov_table_free(struct nfs4_ntov_table *ntovp, 2076 struct nfs4_svgetit_arg *sargp) 2077 { 2078 int i; 2079 union nfs4_attr_u *na; 2080 uint8_t *amap; 2081 2082 /* 2083 * XXX Should do the same checks for whether the bit is set 2084 */ 2085 for (i = 0, na = ntovp->na, amap = ntovp->amap; 2086 i < ntovp->attrcnt; i++, na++, amap++) { 2087 (void) (*nfs4_ntov_map[*amap].sv_getit)( 2088 NFS4ATTR_FREEIT, sargp, na); 2089 } 2090 if ((sargp->op == NFS4ATTR_SETIT) || (sargp->op == NFS4ATTR_VERIT)) { 2091 /* 2092 * xdr_free for getattr will be done later 2093 */ 2094 for (i = 0, na = ntovp->na, amap = ntovp->amap; 2095 i < ntovp->attrcnt; i++, na++, amap++) { 2096 xdr_free(nfs4_ntov_map[*amap].xfunc, (caddr_t)na); 2097 } 2098 } 2099 kmem_free(ntovp->na, sizeof (union nfs4_attr_u) * nfs4_ntov_map_size); 2100 } 2101 2102 /* 2103 * do_rfs4_op_getattr gets the system attrs and converts into fattr4. 2104 */ 2105 static nfsstat4 2106 do_rfs4_op_getattr(bitmap4 breq, fattr4 *fattrp, 2107 struct nfs4_svgetit_arg *sargp) 2108 { 2109 int error = 0; 2110 int i, k; 2111 struct nfs4_ntov_table ntov; 2112 XDR xdr; 2113 ulong_t xdr_size; 2114 char *xdr_attrs; 2115 nfsstat4 status = NFS4_OK; 2116 nfsstat4 prev_rdattr_error = sargp->rdattr_error; 2117 union nfs4_attr_u *na; 2118 uint8_t *amap; 2119 2120 sargp->op = NFS4ATTR_GETIT; 2121 sargp->flag = 0; 2122 2123 fattrp->attrmask = 0; 2124 /* if no bits requested, then return empty fattr4 */ 2125 if (breq == 0) { 2126 fattrp->attrlist4_len = 0; 2127 fattrp->attrlist4 = NULL; 2128 return (NFS4_OK); 2129 } 2130 2131 /* 2132 * return NFS4ERR_INVAL when client requests write-only attrs 2133 */ 2134 if (breq & (FATTR4_TIME_ACCESS_SET_MASK | FATTR4_TIME_MODIFY_SET_MASK)) 2135 return (NFS4ERR_INVAL); 2136 2137 nfs4_ntov_table_init(&ntov); 2138 na = ntov.na; 2139 amap = ntov.amap; 2140 2141 /* 2142 * Now loop to get or verify the attrs 2143 */ 2144 for (i = 0; i < nfs4_ntov_map_size; i++) { 2145 if (breq & nfs4_ntov_map[i].fbit) { 2146 if ((*nfs4_ntov_map[i].sv_getit)( 2147 NFS4ATTR_SUPPORTED, sargp, NULL) == 0) { 2148 2149 error = (*nfs4_ntov_map[i].sv_getit)( 2150 NFS4ATTR_GETIT, sargp, na); 2151 2152 /* 2153 * Possible error values: 2154 * >0 if sv_getit failed to 2155 * get the attr; 0 if succeeded; 2156 * <0 if rdattr_error and the 2157 * attribute cannot be returned. 2158 */ 2159 if (error && !(sargp->rdattr_error_req)) 2160 goto done; 2161 /* 2162 * If error then just for entry 2163 */ 2164 if (error == 0) { 2165 fattrp->attrmask |= 2166 nfs4_ntov_map[i].fbit; 2167 *amap++ = 2168 (uint8_t)nfs4_ntov_map[i].nval; 2169 na++; 2170 (ntov.attrcnt)++; 2171 } else if ((error > 0) && 2172 (sargp->rdattr_error == NFS4_OK)) { 2173 sargp->rdattr_error = puterrno4(error); 2174 } 2175 error = 0; 2176 } 2177 } 2178 } 2179 2180 /* 2181 * If rdattr_error was set after the return value for it was assigned, 2182 * update it. 2183 */ 2184 if (prev_rdattr_error != sargp->rdattr_error) { 2185 na = ntov.na; 2186 amap = ntov.amap; 2187 for (i = 0; i < ntov.attrcnt; i++, na++, amap++) { 2188 k = *amap; 2189 if (k < FATTR4_RDATTR_ERROR) { 2190 continue; 2191 } 2192 if ((k == FATTR4_RDATTR_ERROR) && 2193 ((*nfs4_ntov_map[k].sv_getit)( 2194 NFS4ATTR_SUPPORTED, sargp, NULL) == 0)) { 2195 2196 (void) (*nfs4_ntov_map[k].sv_getit)( 2197 NFS4ATTR_GETIT, sargp, na); 2198 } 2199 break; 2200 } 2201 } 2202 2203 xdr_size = 0; 2204 na = ntov.na; 2205 amap = ntov.amap; 2206 for (i = 0; i < ntov.attrcnt; i++, na++, amap++) { 2207 xdr_size += xdr_sizeof(nfs4_ntov_map[*amap].xfunc, na); 2208 } 2209 2210 fattrp->attrlist4_len = xdr_size; 2211 if (xdr_size) { 2212 /* freed by rfs4_op_getattr_free() */ 2213 fattrp->attrlist4 = xdr_attrs = kmem_zalloc(xdr_size, KM_SLEEP); 2214 2215 xdrmem_create(&xdr, xdr_attrs, xdr_size, XDR_ENCODE); 2216 2217 na = ntov.na; 2218 amap = ntov.amap; 2219 for (i = 0; i < ntov.attrcnt; i++, na++, amap++) { 2220 if (!(*nfs4_ntov_map[*amap].xfunc)(&xdr, na)) { 2221 cmn_err(CE_WARN, "do_rfs4_op_getattr: xdr " 2222 "encode of attribute %d failed\n", *amap); 2223 status = NFS4ERR_SERVERFAULT; 2224 break; 2225 } 2226 } 2227 /* xdrmem_destroy(&xdrs); */ /* NO-OP */ 2228 } else { 2229 fattrp->attrlist4 = NULL; 2230 } 2231 done: 2232 2233 nfs4_ntov_table_free(&ntov, sargp); 2234 2235 if (error != 0) 2236 status = puterrno4(error); 2237 2238 return (status); 2239 } 2240 2241 /* ARGSUSED */ 2242 static void 2243 rfs4_op_getattr(nfs_argop4 *argop, nfs_resop4 *resop, struct svc_req *req, 2244 struct compound_state *cs) 2245 { 2246 GETATTR4args *args = &argop->nfs_argop4_u.opgetattr; 2247 GETATTR4res *resp = &resop->nfs_resop4_u.opgetattr; 2248 struct nfs4_svgetit_arg sarg; 2249 struct statvfs64 sb; 2250 nfsstat4 status; 2251 2252 DTRACE_NFSV4_2(op__getattr__start, struct compound_state *, cs, 2253 GETATTR4args *, args); 2254 2255 if (cs->vp == NULL) { 2256 *cs->statusp = resp->status = NFS4ERR_NOFILEHANDLE; 2257 goto out; 2258 } 2259 2260 if (cs->access == CS_ACCESS_DENIED) { 2261 *cs->statusp = resp->status = NFS4ERR_ACCESS; 2262 goto out; 2263 } 2264 2265 sarg.sbp = &sb; 2266 sarg.cs = cs; 2267 2268 status = bitmap4_to_attrmask(args->attr_request, &sarg); 2269 if (status == NFS4_OK) { 2270 status = bitmap4_get_sysattrs(&sarg); 2271 if (status == NFS4_OK) 2272 status = do_rfs4_op_getattr(args->attr_request, 2273 &resp->obj_attributes, &sarg); 2274 } 2275 *cs->statusp = resp->status = status; 2276 out: 2277 DTRACE_NFSV4_2(op__getattr__done, struct compound_state *, cs, 2278 GETATTR4res *, resp); 2279 } 2280 2281 static void 2282 rfs4_op_getattr_free(nfs_resop4 *resop) 2283 { 2284 GETATTR4res *resp = &resop->nfs_resop4_u.opgetattr; 2285 2286 nfs4_fattr4_free(&resp->obj_attributes); 2287 } 2288 2289 /* ARGSUSED */ 2290 static void 2291 rfs4_op_getfh(nfs_argop4 *argop, nfs_resop4 *resop, struct svc_req *req, 2292 struct compound_state *cs) 2293 { 2294 GETFH4res *resp = &resop->nfs_resop4_u.opgetfh; 2295 2296 DTRACE_NFSV4_1(op__getfh__start, struct compound_state *, cs); 2297 2298 if (cs->vp == NULL) { 2299 *cs->statusp = resp->status = NFS4ERR_NOFILEHANDLE; 2300 goto out; 2301 } 2302 if (cs->access == CS_ACCESS_DENIED) { 2303 *cs->statusp = resp->status = NFS4ERR_ACCESS; 2304 goto out; 2305 } 2306 2307 resp->object.nfs_fh4_val = 2308 kmem_alloc(cs->fh.nfs_fh4_len, KM_SLEEP); 2309 nfs_fh4_copy(&cs->fh, &resp->object); 2310 *cs->statusp = resp->status = NFS4_OK; 2311 out: 2312 DTRACE_NFSV4_2(op__getfh__done, struct compound_state *, cs, 2313 GETFH4res *, resp); 2314 } 2315 2316 static void 2317 rfs4_op_getfh_free(nfs_resop4 *resop) 2318 { 2319 GETFH4res *resp = &resop->nfs_resop4_u.opgetfh; 2320 2321 if (resp->status == NFS4_OK && 2322 resp->object.nfs_fh4_val != NULL) { 2323 kmem_free(resp->object.nfs_fh4_val, resp->object.nfs_fh4_len); 2324 resp->object.nfs_fh4_val = NULL; 2325 resp->object.nfs_fh4_len = 0; 2326 } 2327 } 2328 2329 /* 2330 * illegal: args: void 2331 * res : status (NFS4ERR_OP_ILLEGAL) 2332 */ 2333 /* ARGSUSED */ 2334 static void 2335 rfs4_op_illegal(nfs_argop4 *argop, nfs_resop4 *resop, 2336 struct svc_req *req, struct compound_state *cs) 2337 { 2338 ILLEGAL4res *resp = &resop->nfs_resop4_u.opillegal; 2339 2340 resop->resop = OP_ILLEGAL; 2341 *cs->statusp = resp->status = NFS4ERR_OP_ILLEGAL; 2342 } 2343 2344 /* 2345 * link: args: SAVED_FH: file, CURRENT_FH: target directory 2346 * res: status. If success - CURRENT_FH unchanged, return change_info 2347 */ 2348 /* ARGSUSED */ 2349 static void 2350 rfs4_op_link(nfs_argop4 *argop, nfs_resop4 *resop, struct svc_req *req, 2351 struct compound_state *cs) 2352 { 2353 LINK4args *args = &argop->nfs_argop4_u.oplink; 2354 LINK4res *resp = &resop->nfs_resop4_u.oplink; 2355 int error; 2356 vnode_t *vp; 2357 vnode_t *dvp; 2358 struct vattr bdva, idva, adva; 2359 char *nm; 2360 uint_t len; 2361 2362 DTRACE_NFSV4_2(op__link__start, struct compound_state *, cs, 2363 LINK4args *, args); 2364 2365 /* SAVED_FH: source object */ 2366 vp = cs->saved_vp; 2367 if (vp == NULL) { 2368 *cs->statusp = resp->status = NFS4ERR_NOFILEHANDLE; 2369 goto out; 2370 } 2371 2372 /* CURRENT_FH: target directory */ 2373 dvp = cs->vp; 2374 if (dvp == NULL) { 2375 *cs->statusp = resp->status = NFS4ERR_NOFILEHANDLE; 2376 goto out; 2377 } 2378 2379 /* 2380 * If there is a non-shared filesystem mounted on this vnode, 2381 * do not allow to link any file in this directory. 2382 */ 2383 if (vn_ismntpt(dvp)) { 2384 *cs->statusp = resp->status = NFS4ERR_ACCESS; 2385 goto out; 2386 } 2387 2388 if (cs->access == CS_ACCESS_DENIED) { 2389 *cs->statusp = resp->status = NFS4ERR_ACCESS; 2390 goto out; 2391 } 2392 2393 /* Check source object's type validity */ 2394 if (vp->v_type == VDIR) { 2395 *cs->statusp = resp->status = NFS4ERR_ISDIR; 2396 goto out; 2397 } 2398 2399 /* Check target directory's type */ 2400 if (dvp->v_type != VDIR) { 2401 *cs->statusp = resp->status = NFS4ERR_NOTDIR; 2402 goto out; 2403 } 2404 2405 if (cs->saved_exi != cs->exi) { 2406 *cs->statusp = resp->status = NFS4ERR_XDEV; 2407 goto out; 2408 } 2409 2410 if (!utf8_dir_verify(&args->newname)) { 2411 *cs->statusp = resp->status = NFS4ERR_INVAL; 2412 goto out; 2413 } 2414 2415 nm = utf8_to_fn(&args->newname, &len, NULL); 2416 if (nm == NULL) { 2417 *cs->statusp = resp->status = NFS4ERR_INVAL; 2418 goto out; 2419 } 2420 2421 if (len > MAXNAMELEN) { 2422 *cs->statusp = resp->status = NFS4ERR_NAMETOOLONG; 2423 kmem_free(nm, len); 2424 goto out; 2425 } 2426 2427 if (rdonly4(cs->exi, cs->vp, req)) { 2428 *cs->statusp = resp->status = NFS4ERR_ROFS; 2429 kmem_free(nm, len); 2430 goto out; 2431 } 2432 2433 /* Get "before" change value */ 2434 bdva.va_mask = AT_CTIME|AT_SEQ; 2435 error = VOP_GETATTR(dvp, &bdva, 0, cs->cr, NULL); 2436 if (error) { 2437 *cs->statusp = resp->status = puterrno4(error); 2438 kmem_free(nm, len); 2439 goto out; 2440 } 2441 2442 NFS4_SET_FATTR4_CHANGE(resp->cinfo.before, bdva.va_ctime) 2443 2444 error = VOP_LINK(dvp, vp, nm, cs->cr, NULL, 0); 2445 2446 kmem_free(nm, len); 2447 2448 /* 2449 * Get the initial "after" sequence number, if it fails, set to zero 2450 */ 2451 idva.va_mask = AT_SEQ; 2452 if (VOP_GETATTR(dvp, &idva, 0, cs->cr, NULL)) 2453 idva.va_seq = 0; 2454 2455 /* 2456 * Force modified data and metadata out to stable storage. 2457 */ 2458 (void) VOP_FSYNC(vp, FNODSYNC, cs->cr, NULL); 2459 (void) VOP_FSYNC(dvp, 0, cs->cr, NULL); 2460 2461 if (error) { 2462 *cs->statusp = resp->status = puterrno4(error); 2463 goto out; 2464 } 2465 2466 /* 2467 * Get "after" change value, if it fails, simply return the 2468 * before value. 2469 */ 2470 adva.va_mask = AT_CTIME|AT_SEQ; 2471 if (VOP_GETATTR(dvp, &adva, 0, cs->cr, NULL)) { 2472 adva.va_ctime = bdva.va_ctime; 2473 adva.va_seq = 0; 2474 } 2475 2476 NFS4_SET_FATTR4_CHANGE(resp->cinfo.after, adva.va_ctime) 2477 2478 /* 2479 * The cinfo.atomic = TRUE only if we have 2480 * non-zero va_seq's, and it has incremented by exactly one 2481 * during the VOP_LINK and it didn't change during the VOP_FSYNC. 2482 */ 2483 if (bdva.va_seq && idva.va_seq && adva.va_seq && 2484 idva.va_seq == (bdva.va_seq + 1) && idva.va_seq == adva.va_seq) 2485 resp->cinfo.atomic = TRUE; 2486 else 2487 resp->cinfo.atomic = FALSE; 2488 2489 *cs->statusp = resp->status = NFS4_OK; 2490 out: 2491 DTRACE_NFSV4_2(op__link__done, struct compound_state *, cs, 2492 LINK4res *, resp); 2493 } 2494 2495 /* 2496 * Used by rfs4_op_lookup and rfs4_op_lookupp to do the actual work. 2497 */ 2498 2499 /* ARGSUSED */ 2500 static nfsstat4 2501 do_rfs4_op_lookup(char *nm, uint_t buflen, struct svc_req *req, 2502 struct compound_state *cs) 2503 { 2504 int error; 2505 int different_export = 0; 2506 vnode_t *vp, *tvp, *pre_tvp = NULL, *oldvp = NULL; 2507 struct exportinfo *exi = NULL, *pre_exi = NULL; 2508 nfsstat4 stat; 2509 fid_t fid; 2510 int attrdir, dotdot, walk; 2511 bool_t is_newvp = FALSE; 2512 2513 if (cs->vp->v_flag & V_XATTRDIR) { 2514 attrdir = 1; 2515 ASSERT(get_fh4_flag(&cs->fh, FH4_ATTRDIR)); 2516 } else { 2517 attrdir = 0; 2518 ASSERT(! get_fh4_flag(&cs->fh, FH4_ATTRDIR)); 2519 } 2520 2521 dotdot = (nm[0] == '.' && nm[1] == '.' && nm[2] == '\0'); 2522 2523 /* 2524 * If dotdotting, then need to check whether it's 2525 * above the root of a filesystem, or above an 2526 * export point. 2527 */ 2528 if (dotdot) { 2529 2530 /* 2531 * If dotdotting at the root of a filesystem, then 2532 * need to traverse back to the mounted-on filesystem 2533 * and do the dotdot lookup there. 2534 */ 2535 if (cs->vp->v_flag & VROOT) { 2536 2537 /* 2538 * If at the system root, then can 2539 * go up no further. 2540 */ 2541 if (VN_CMP(cs->vp, rootdir)) 2542 return (puterrno4(ENOENT)); 2543 2544 /* 2545 * Traverse back to the mounted-on filesystem 2546 */ 2547 cs->vp = untraverse(cs->vp); 2548 2549 /* 2550 * Set the different_export flag so we remember 2551 * to pick up a new exportinfo entry for 2552 * this new filesystem. 2553 */ 2554 different_export = 1; 2555 } else { 2556 2557 /* 2558 * If dotdotting above an export point then set 2559 * the different_export to get new export info. 2560 */ 2561 different_export = nfs_exported(cs->exi, cs->vp); 2562 } 2563 } 2564 2565 error = VOP_LOOKUP(cs->vp, nm, &vp, NULL, 0, NULL, cs->cr, 2566 NULL, NULL, NULL); 2567 if (error) 2568 return (puterrno4(error)); 2569 2570 /* 2571 * If the vnode is in a pseudo filesystem, check whether it is visible. 2572 * 2573 * XXX if the vnode is a symlink and it is not visible in 2574 * a pseudo filesystem, return ENOENT (not following symlink). 2575 * V4 client can not mount such symlink. This is a regression 2576 * from V2/V3. 2577 * 2578 * In the same exported filesystem, if the security flavor used 2579 * is not an explicitly shared flavor, limit the view to the visible 2580 * list entries only. This is not a WRONGSEC case because it's already 2581 * checked via PUTROOTFH/PUTPUBFH or PUTFH. 2582 */ 2583 if (!different_export && 2584 (PSEUDO(cs->exi) || ! is_exported_sec(cs->nfsflavor, cs->exi) || 2585 cs->access & CS_ACCESS_LIMITED)) { 2586 if (! nfs_visible(cs->exi, vp, &different_export)) { 2587 VN_RELE(vp); 2588 return (puterrno4(ENOENT)); 2589 } 2590 } 2591 2592 /* 2593 * If it's a mountpoint, then traverse it. 2594 */ 2595 if (vn_ismntpt(vp)) { 2596 pre_exi = cs->exi; /* save pre-traversed exportinfo */ 2597 pre_tvp = vp; /* save pre-traversed vnode */ 2598 2599 /* 2600 * hold pre_tvp to counteract rele by traverse. We will 2601 * need pre_tvp below if checkexport4 fails 2602 */ 2603 VN_HOLD(pre_tvp); 2604 tvp = vp; 2605 if ((error = traverse(&tvp)) != 0) { 2606 VN_RELE(vp); 2607 VN_RELE(pre_tvp); 2608 return (puterrno4(error)); 2609 } 2610 vp = tvp; 2611 different_export = 1; 2612 } else if (vp->v_vfsp != cs->vp->v_vfsp) { 2613 /* 2614 * The vfsp comparison is to handle the case where 2615 * a LOFS mount is shared. lo_lookup traverses mount points, 2616 * and NFS is unaware of local fs transistions because 2617 * v_vfsmountedhere isn't set. For this special LOFS case, 2618 * the dir and the obj returned by lookup will have different 2619 * vfs ptrs. 2620 */ 2621 different_export = 1; 2622 } 2623 2624 if (different_export) { 2625 2626 bzero(&fid, sizeof (fid)); 2627 fid.fid_len = MAXFIDSZ; 2628 error = vop_fid_pseudo(vp, &fid); 2629 if (error) { 2630 VN_RELE(vp); 2631 if (pre_tvp) 2632 VN_RELE(pre_tvp); 2633 return (puterrno4(error)); 2634 } 2635 2636 if (dotdot) 2637 exi = nfs_vptoexi(NULL, vp, cs->cr, &walk, NULL, TRUE); 2638 else 2639 exi = checkexport4(&vp->v_vfsp->vfs_fsid, &fid, vp); 2640 2641 if (exi == NULL) { 2642 if (pre_tvp) { 2643 /* 2644 * If this vnode is a mounted-on vnode, 2645 * but the mounted-on file system is not 2646 * exported, send back the filehandle for 2647 * the mounted-on vnode, not the root of 2648 * the mounted-on file system. 2649 */ 2650 VN_RELE(vp); 2651 vp = pre_tvp; 2652 exi = pre_exi; 2653 } else { 2654 VN_RELE(vp); 2655 return (puterrno4(EACCES)); 2656 } 2657 } else if (pre_tvp) { 2658 /* we're done with pre_tvp now. release extra hold */ 2659 VN_RELE(pre_tvp); 2660 } 2661 2662 cs->exi = exi; 2663 2664 /* 2665 * Now we do a checkauth4. The reason is that 2666 * this client/user may not have access to the new 2667 * exported file system, and if he does, 2668 * the client/user may be mapped to a different uid. 2669 * 2670 * We start with a new cr, because the checkauth4 done 2671 * in the PUT*FH operation over wrote the cred's uid, 2672 * gid, etc, and we want the real thing before calling 2673 * checkauth4() 2674 */ 2675 crfree(cs->cr); 2676 cs->cr = crdup(cs->basecr); 2677 2678 if (cs->vp) 2679 oldvp = cs->vp; 2680 cs->vp = vp; 2681 is_newvp = TRUE; 2682 2683 stat = call_checkauth4(cs, req); 2684 if (stat != NFS4_OK) { 2685 VN_RELE(cs->vp); 2686 cs->vp = oldvp; 2687 return (stat); 2688 } 2689 } 2690 2691 /* 2692 * After various NFS checks, do a label check on the path 2693 * component. The label on this path should either be the 2694 * global zone's label or a zone's label. We are only 2695 * interested in the zone's label because exported files 2696 * in global zone is accessible (though read-only) to 2697 * clients. The exportability/visibility check is already 2698 * done before reaching this code. 2699 */ 2700 if (is_system_labeled()) { 2701 bslabel_t *clabel; 2702 2703 ASSERT(req->rq_label != NULL); 2704 clabel = req->rq_label; 2705 DTRACE_PROBE2(tx__rfs4__log__info__oplookup__clabel, char *, 2706 "got client label from request(1)", struct svc_req *, req); 2707 2708 if (!blequal(&l_admin_low->tsl_label, clabel)) { 2709 if (!do_rfs_label_check(clabel, vp, DOMINANCE_CHECK)) { 2710 error = EACCES; 2711 goto err_out; 2712 } 2713 } else { 2714 /* 2715 * We grant access to admin_low label clients 2716 * only if the client is trusted, i.e. also 2717 * running Solaris Trusted Extension. 2718 */ 2719 struct sockaddr *ca; 2720 int addr_type; 2721 void *ipaddr; 2722 tsol_tpc_t *tp; 2723 2724 ca = (struct sockaddr *)svc_getrpccaller( 2725 req->rq_xprt)->buf; 2726 if (ca->sa_family == AF_INET) { 2727 addr_type = IPV4_VERSION; 2728 ipaddr = &((struct sockaddr_in *)ca)->sin_addr; 2729 } else if (ca->sa_family == AF_INET6) { 2730 addr_type = IPV6_VERSION; 2731 ipaddr = &((struct sockaddr_in6 *) 2732 ca)->sin6_addr; 2733 } 2734 tp = find_tpc(ipaddr, addr_type, B_FALSE); 2735 if (tp == NULL || tp->tpc_tp.tp_doi != 2736 l_admin_low->tsl_doi || tp->tpc_tp.host_type != 2737 SUN_CIPSO) { 2738 if (tp != NULL) 2739 TPC_RELE(tp); 2740 error = EACCES; 2741 goto err_out; 2742 } 2743 TPC_RELE(tp); 2744 } 2745 } 2746 2747 error = makefh4(&cs->fh, vp, cs->exi); 2748 2749 err_out: 2750 if (error) { 2751 if (is_newvp) { 2752 VN_RELE(cs->vp); 2753 cs->vp = oldvp; 2754 } else 2755 VN_RELE(vp); 2756 return (puterrno4(error)); 2757 } 2758 2759 if (!is_newvp) { 2760 if (cs->vp) 2761 VN_RELE(cs->vp); 2762 cs->vp = vp; 2763 } else if (oldvp) 2764 VN_RELE(oldvp); 2765 2766 /* 2767 * if did lookup on attrdir and didn't lookup .., set named 2768 * attr fh flag 2769 */ 2770 if (attrdir && ! dotdot) 2771 set_fh4_flag(&cs->fh, FH4_NAMEDATTR); 2772 2773 /* Assume false for now, open proc will set this */ 2774 cs->mandlock = FALSE; 2775 2776 return (NFS4_OK); 2777 } 2778 2779 /* ARGSUSED */ 2780 static void 2781 rfs4_op_lookup(nfs_argop4 *argop, nfs_resop4 *resop, struct svc_req *req, 2782 struct compound_state *cs) 2783 { 2784 LOOKUP4args *args = &argop->nfs_argop4_u.oplookup; 2785 LOOKUP4res *resp = &resop->nfs_resop4_u.oplookup; 2786 char *nm; 2787 uint_t len; 2788 2789 DTRACE_NFSV4_2(op__lookup__start, struct compound_state *, cs, 2790 LOOKUP4args *, args); 2791 2792 if (cs->vp == NULL) { 2793 *cs->statusp = resp->status = NFS4ERR_NOFILEHANDLE; 2794 goto out; 2795 } 2796 2797 if (cs->vp->v_type == VLNK) { 2798 *cs->statusp = resp->status = NFS4ERR_SYMLINK; 2799 goto out; 2800 } 2801 2802 if (cs->vp->v_type != VDIR) { 2803 *cs->statusp = resp->status = NFS4ERR_NOTDIR; 2804 goto out; 2805 } 2806 2807 if (!utf8_dir_verify(&args->objname)) { 2808 *cs->statusp = resp->status = NFS4ERR_INVAL; 2809 goto out; 2810 } 2811 2812 nm = utf8_to_str(&args->objname, &len, NULL); 2813 if (nm == NULL) { 2814 *cs->statusp = resp->status = NFS4ERR_INVAL; 2815 goto out; 2816 } 2817 2818 if (len > MAXNAMELEN) { 2819 *cs->statusp = resp->status = NFS4ERR_NAMETOOLONG; 2820 kmem_free(nm, len); 2821 goto out; 2822 } 2823 2824 *cs->statusp = resp->status = do_rfs4_op_lookup(nm, len, req, cs); 2825 2826 kmem_free(nm, len); 2827 2828 out: 2829 DTRACE_NFSV4_2(op__lookup__done, struct compound_state *, cs, 2830 LOOKUP4res *, resp); 2831 } 2832 2833 /* ARGSUSED */ 2834 static void 2835 rfs4_op_lookupp(nfs_argop4 *args, nfs_resop4 *resop, struct svc_req *req, 2836 struct compound_state *cs) 2837 { 2838 LOOKUPP4res *resp = &resop->nfs_resop4_u.oplookupp; 2839 2840 DTRACE_NFSV4_1(op__lookupp__start, struct compound_state *, cs); 2841 2842 if (cs->vp == NULL) { 2843 *cs->statusp = resp->status = NFS4ERR_NOFILEHANDLE; 2844 goto out; 2845 } 2846 2847 if (cs->vp->v_type != VDIR) { 2848 *cs->statusp = resp->status = NFS4ERR_NOTDIR; 2849 goto out; 2850 } 2851 2852 *cs->statusp = resp->status = do_rfs4_op_lookup("..", 3, req, cs); 2853 2854 /* 2855 * From NFSV4 Specification, LOOKUPP should not check for 2856 * NFS4ERR_WRONGSEC. Retrun NFS4_OK instead. 2857 */ 2858 if (resp->status == NFS4ERR_WRONGSEC) { 2859 *cs->statusp = resp->status = NFS4_OK; 2860 } 2861 2862 out: 2863 DTRACE_NFSV4_2(op__lookupp__done, struct compound_state *, cs, 2864 LOOKUPP4res *, resp); 2865 } 2866 2867 2868 /*ARGSUSED2*/ 2869 static void 2870 rfs4_op_openattr(nfs_argop4 *argop, nfs_resop4 *resop, struct svc_req *req, 2871 struct compound_state *cs) 2872 { 2873 OPENATTR4args *args = &argop->nfs_argop4_u.opopenattr; 2874 OPENATTR4res *resp = &resop->nfs_resop4_u.opopenattr; 2875 vnode_t *avp = NULL; 2876 int lookup_flags = LOOKUP_XATTR, error; 2877 int exp_ro = 0; 2878 2879 DTRACE_NFSV4_2(op__openattr__start, struct compound_state *, cs, 2880 OPENATTR4args *, args); 2881 2882 if (cs->vp == NULL) { 2883 *cs->statusp = resp->status = NFS4ERR_NOFILEHANDLE; 2884 goto out; 2885 } 2886 2887 /* 2888 * Make a couple of checks made by copen() 2889 * 2890 * Check to make sure underlying fs supports xattrs. This 2891 * is required because solaris filesystem implementations 2892 * (UFS/TMPFS) don't enforce the noxattr mount option 2893 * in VOP_LOOKUP(LOOKUP_XATTR). If fs doesn't support this 2894 * pathconf cmd or if fs supports cmd but doesn't claim 2895 * support for xattr, return NOTSUPP. It would be better 2896 * to use VOP_PATHCONF( _PC_XATTR_ENABLED) for this; however, 2897 * that cmd is not available to VOP_PATHCONF interface 2898 * (it's only implemented inside pathconf syscall)... 2899 * 2900 * Verify permission to put attributes on files (access 2901 * checks from copen). 2902 */ 2903 2904 if ((cs->vp->v_vfsp->vfs_flag & VFS_XATTR) == 0 && 2905 !vfs_has_feature(cs->vp->v_vfsp, VFSFT_XVATTR)) { 2906 *cs->statusp = resp->status = puterrno4(ENOTSUP); 2907 goto out; 2908 } 2909 2910 if ((VOP_ACCESS(cs->vp, VREAD, 0, cs->cr, NULL) != 0) && 2911 (VOP_ACCESS(cs->vp, VWRITE, 0, cs->cr, NULL) != 0) && 2912 (VOP_ACCESS(cs->vp, VEXEC, 0, cs->cr, NULL) != 0)) { 2913 *cs->statusp = resp->status = puterrno4(EACCES); 2914 goto out; 2915 } 2916 2917 /* 2918 * The CREATE_XATTR_DIR VOP flag cannot be specified if 2919 * the file system is exported read-only -- regardless of 2920 * createdir flag. Otherwise the attrdir would be created 2921 * (assuming server fs isn't mounted readonly locally). If 2922 * VOP_LOOKUP returns ENOENT in this case, the error will 2923 * be translated into EROFS. ENOSYS is mapped to ENOTSUP 2924 * because specfs has no VOP_LOOKUP op, so the macro would 2925 * return ENOSYS. EINVAL is returned by all (current) 2926 * Solaris file system implementations when any of their 2927 * restrictions are violated (xattr(dir) can't have xattrdir). 2928 * Returning NOTSUPP is more appropriate in this case 2929 * because the object will never be able to have an attrdir. 2930 */ 2931 if (args->createdir && ! (exp_ro = rdonly4(cs->exi, cs->vp, req))) 2932 lookup_flags |= CREATE_XATTR_DIR; 2933 2934 error = VOP_LOOKUP(cs->vp, "", &avp, NULL, lookup_flags, NULL, cs->cr, 2935 NULL, NULL, NULL); 2936 2937 if (error) { 2938 if (error == ENOENT && args->createdir && exp_ro) 2939 *cs->statusp = resp->status = puterrno4(EROFS); 2940 else if (error == EINVAL || error == ENOSYS) 2941 *cs->statusp = resp->status = puterrno4(ENOTSUP); 2942 else 2943 *cs->statusp = resp->status = puterrno4(error); 2944 goto out; 2945 } 2946 2947 ASSERT(avp->v_flag & V_XATTRDIR); 2948 2949 error = makefh4(&cs->fh, avp, cs->exi); 2950 2951 if (error) { 2952 VN_RELE(avp); 2953 *cs->statusp = resp->status = puterrno4(error); 2954 goto out; 2955 } 2956 2957 VN_RELE(cs->vp); 2958 cs->vp = avp; 2959 2960 /* 2961 * There is no requirement for an attrdir fh flag 2962 * because the attrdir has a vnode flag to distinguish 2963 * it from regular (non-xattr) directories. The 2964 * FH4_ATTRDIR flag is set for future sanity checks. 2965 */ 2966 set_fh4_flag(&cs->fh, FH4_ATTRDIR); 2967 *cs->statusp = resp->status = NFS4_OK; 2968 2969 out: 2970 DTRACE_NFSV4_2(op__openattr__done, struct compound_state *, cs, 2971 OPENATTR4res *, resp); 2972 } 2973 2974 static int 2975 do_io(int direction, vnode_t *vp, struct uio *uio, int ioflag, cred_t *cred, 2976 caller_context_t *ct) 2977 { 2978 int error; 2979 int i; 2980 clock_t delaytime; 2981 2982 delaytime = MSEC_TO_TICK_ROUNDUP(rfs4_lock_delay); 2983 2984 /* 2985 * Don't block on mandatory locks. If this routine returns 2986 * EAGAIN, the caller should return NFS4ERR_LOCKED. 2987 */ 2988 uio->uio_fmode = FNONBLOCK; 2989 2990 for (i = 0; i < rfs4_maxlock_tries; i++) { 2991 2992 2993 if (direction == FREAD) { 2994 (void) VOP_RWLOCK(vp, V_WRITELOCK_FALSE, ct); 2995 error = VOP_READ(vp, uio, ioflag, cred, ct); 2996 VOP_RWUNLOCK(vp, V_WRITELOCK_FALSE, ct); 2997 } else { 2998 (void) VOP_RWLOCK(vp, V_WRITELOCK_TRUE, ct); 2999 error = VOP_WRITE(vp, uio, ioflag, cred, ct); 3000 VOP_RWUNLOCK(vp, V_WRITELOCK_TRUE, ct); 3001 } 3002 3003 if (error != EAGAIN) 3004 break; 3005 3006 if (i < rfs4_maxlock_tries - 1) { 3007 delay(delaytime); 3008 delaytime *= 2; 3009 } 3010 } 3011 3012 return (error); 3013 } 3014 3015 /* ARGSUSED */ 3016 static void 3017 rfs4_op_read(nfs_argop4 *argop, nfs_resop4 *resop, struct svc_req *req, 3018 struct compound_state *cs) 3019 { 3020 READ4args *args = &argop->nfs_argop4_u.opread; 3021 READ4res *resp = &resop->nfs_resop4_u.opread; 3022 int error; 3023 int verror; 3024 vnode_t *vp; 3025 struct vattr va; 3026 struct iovec iov; 3027 struct uio uio; 3028 u_offset_t offset; 3029 bool_t *deleg = &cs->deleg; 3030 nfsstat4 stat; 3031 int in_crit = 0; 3032 mblk_t *mp; 3033 int alloc_err = 0; 3034 caller_context_t ct; 3035 3036 DTRACE_NFSV4_2(op__read__start, struct compound_state *, cs, 3037 READ4args, args); 3038 3039 vp = cs->vp; 3040 if (vp == NULL) { 3041 *cs->statusp = resp->status = NFS4ERR_NOFILEHANDLE; 3042 goto out; 3043 } 3044 if (cs->access == CS_ACCESS_DENIED) { 3045 *cs->statusp = resp->status = NFS4ERR_ACCESS; 3046 goto out; 3047 } 3048 3049 if ((stat = rfs4_check_stateid(FREAD, vp, &args->stateid, FALSE, 3050 deleg, TRUE, &ct)) != NFS4_OK) { 3051 *cs->statusp = resp->status = stat; 3052 goto out; 3053 } 3054 3055 /* 3056 * Enter the critical region before calling VOP_RWLOCK 3057 * to avoid a deadlock with write requests. 3058 */ 3059 if (nbl_need_check(vp)) { 3060 nbl_start_crit(vp, RW_READER); 3061 in_crit = 1; 3062 if (nbl_conflict(vp, NBL_READ, args->offset, args->count, 0, 3063 &ct)) { 3064 *cs->statusp = resp->status = NFS4ERR_LOCKED; 3065 goto out; 3066 } 3067 } 3068 3069 if ((stat = rfs4_check_stateid(FREAD, vp, &args->stateid, FALSE, 3070 deleg, TRUE, &ct)) != NFS4_OK) { 3071 *cs->statusp = resp->status = stat; 3072 goto out; 3073 } 3074 3075 va.va_mask = AT_MODE|AT_SIZE|AT_UID; 3076 verror = VOP_GETATTR(vp, &va, 0, cs->cr, &ct); 3077 3078 /* 3079 * If we can't get the attributes, then we can't do the 3080 * right access checking. So, we'll fail the request. 3081 */ 3082 if (verror) { 3083 *cs->statusp = resp->status = puterrno4(verror); 3084 goto out; 3085 } 3086 3087 if (vp->v_type != VREG) { 3088 *cs->statusp = resp->status = 3089 ((vp->v_type == VDIR) ? NFS4ERR_ISDIR : NFS4ERR_INVAL); 3090 goto out; 3091 } 3092 3093 if (crgetuid(cs->cr) != va.va_uid && 3094 (error = VOP_ACCESS(vp, VREAD, 0, cs->cr, &ct)) && 3095 (error = VOP_ACCESS(vp, VEXEC, 0, cs->cr, &ct))) { 3096 *cs->statusp = resp->status = puterrno4(error); 3097 goto out; 3098 } 3099 3100 if (MANDLOCK(vp, va.va_mode)) { /* XXX - V4 supports mand locking */ 3101 *cs->statusp = resp->status = NFS4ERR_ACCESS; 3102 goto out; 3103 } 3104 3105 offset = args->offset; 3106 if (offset >= va.va_size) { 3107 *cs->statusp = resp->status = NFS4_OK; 3108 resp->eof = TRUE; 3109 resp->data_len = 0; 3110 resp->data_val = NULL; 3111 resp->mblk = NULL; 3112 *cs->statusp = resp->status = NFS4_OK; 3113 goto out; 3114 } 3115 3116 if (args->count == 0) { 3117 *cs->statusp = resp->status = NFS4_OK; 3118 resp->eof = FALSE; 3119 resp->data_len = 0; 3120 resp->data_val = NULL; 3121 resp->mblk = NULL; 3122 goto out; 3123 } 3124 3125 /* 3126 * Do not allocate memory more than maximum allowed 3127 * transfer size 3128 */ 3129 if (args->count > rfs4_tsize(req)) 3130 args->count = rfs4_tsize(req); 3131 3132 /* 3133 * mp will contain the data to be sent out in the read reply. 3134 * It will be freed after the reply has been sent. 3135 * Let's roundup the data to a BYTES_PER_XDR_UNIT multiple, 3136 * so that the call to xdrmblk_putmblk() never fails. 3137 * If the first alloc of the requested size fails, then 3138 * decrease the size to something more reasonable and wait 3139 * for the allocation to occur. 3140 */ 3141 mp = allocb(RNDUP(args->count), BPRI_MED); 3142 if (mp == NULL) { 3143 if (args->count > MAXBSIZE) 3144 args->count = MAXBSIZE; 3145 mp = allocb_wait(RNDUP(args->count), BPRI_MED, 3146 STR_NOSIG, &alloc_err); 3147 } 3148 ASSERT(mp != NULL); 3149 ASSERT(alloc_err == 0); 3150 3151 iov.iov_base = (caddr_t)mp->b_datap->db_base; 3152 iov.iov_len = args->count; 3153 uio.uio_iov = &iov; 3154 uio.uio_iovcnt = 1; 3155 uio.uio_segflg = UIO_SYSSPACE; 3156 uio.uio_extflg = UIO_COPY_CACHED; 3157 uio.uio_loffset = args->offset; 3158 uio.uio_resid = args->count; 3159 3160 error = do_io(FREAD, vp, &uio, 0, cs->cr, &ct); 3161 3162 va.va_mask = AT_SIZE; 3163 verror = VOP_GETATTR(vp, &va, 0, cs->cr, &ct); 3164 3165 if (error) { 3166 freeb(mp); 3167 *cs->statusp = resp->status = puterrno4(error); 3168 goto out; 3169 } 3170 3171 *cs->statusp = resp->status = NFS4_OK; 3172 3173 ASSERT(uio.uio_resid >= 0); 3174 resp->data_len = args->count - uio.uio_resid; 3175 resp->data_val = (char *)mp->b_datap->db_base; 3176 resp->mblk = mp; 3177 3178 if (!verror && offset + resp->data_len == va.va_size) 3179 resp->eof = TRUE; 3180 else 3181 resp->eof = FALSE; 3182 3183 out: 3184 if (in_crit) 3185 nbl_end_crit(vp); 3186 3187 DTRACE_NFSV4_2(op__read__done, struct compound_state *, cs, 3188 READ4res *, resp); 3189 } 3190 3191 static void 3192 rfs4_op_read_free(nfs_resop4 *resop) 3193 { 3194 READ4res *resp = &resop->nfs_resop4_u.opread; 3195 3196 if (resp->status == NFS4_OK && resp->mblk != NULL) { 3197 freeb(resp->mblk); 3198 resp->mblk = NULL; 3199 resp->data_val = NULL; 3200 resp->data_len = 0; 3201 } 3202 } 3203 3204 static void 3205 rfs4_op_readdir_free(nfs_resop4 *resop) 3206 { 3207 READDIR4res *resp = &resop->nfs_resop4_u.opreaddir; 3208 3209 if (resp->status == NFS4_OK && resp->mblk != NULL) { 3210 freeb(resp->mblk); 3211 resp->mblk = NULL; 3212 resp->data_len = 0; 3213 } 3214 } 3215 3216 3217 /* ARGSUSED */ 3218 static void 3219 rfs4_op_putpubfh(nfs_argop4 *args, nfs_resop4 *resop, struct svc_req *req, 3220 struct compound_state *cs) 3221 { 3222 PUTPUBFH4res *resp = &resop->nfs_resop4_u.opputpubfh; 3223 int error; 3224 vnode_t *vp; 3225 struct exportinfo *exi, *sav_exi; 3226 nfs_fh4_fmt_t *fh_fmtp; 3227 3228 DTRACE_NFSV4_1(op__putpubfh__start, struct compound_state *, cs); 3229 3230 if (cs->vp) { 3231 VN_RELE(cs->vp); 3232 cs->vp = NULL; 3233 } 3234 3235 if (cs->cr) 3236 crfree(cs->cr); 3237 3238 cs->cr = crdup(cs->basecr); 3239 3240 vp = exi_public->exi_vp; 3241 if (vp == NULL) { 3242 *cs->statusp = resp->status = NFS4ERR_SERVERFAULT; 3243 goto out; 3244 } 3245 3246 if (is_system_labeled()) { 3247 bslabel_t *clabel; 3248 3249 ASSERT(req->rq_label != NULL); 3250 clabel = req->rq_label; 3251 DTRACE_PROBE2(tx__rfs4__log__info__opputpubfh__clabel, char *, 3252 "got client label from request(1)", 3253 struct svc_req *, req); 3254 if (!blequal(&l_admin_low->tsl_label, clabel)) { 3255 if (!do_rfs_label_check(clabel, vp, DOMINANCE_CHECK)) { 3256 *cs->statusp = resp->status = 3257 NFS4ERR_SERVERFAULT; 3258 return; 3259 } 3260 } 3261 } 3262 3263 error = makefh4(&cs->fh, vp, exi_public); 3264 if (error != 0) { 3265 *cs->statusp = resp->status = puterrno4(error); 3266 goto out; 3267 } 3268 sav_exi = cs->exi; 3269 if (exi_public == exi_root) { 3270 /* 3271 * No filesystem is actually shared public, so we default 3272 * to exi_root. In this case, we must check whether root 3273 * is exported. 3274 */ 3275 fh_fmtp = (nfs_fh4_fmt_t *)cs->fh.nfs_fh4_val; 3276 3277 /* 3278 * if root filesystem is exported, the exportinfo struct that we 3279 * should use is what checkexport4 returns, because root_exi is 3280 * actually a mostly empty struct. 3281 */ 3282 exi = checkexport4(&fh_fmtp->fh4_fsid, 3283 (fid_t *)&fh_fmtp->fh4_xlen, NULL); 3284 cs->exi = ((exi != NULL) ? exi : exi_public); 3285 } else { 3286 /* 3287 * it's a properly shared filesystem 3288 */ 3289 cs->exi = exi_public; 3290 } 3291 3292 VN_HOLD(vp); 3293 cs->vp = vp; 3294 3295 if ((resp->status = call_checkauth4(cs, req)) != NFS4_OK) { 3296 VN_RELE(cs->vp); 3297 cs->vp = NULL; 3298 cs->exi = sav_exi; 3299 goto out; 3300 } 3301 3302 *cs->statusp = resp->status = NFS4_OK; 3303 out: 3304 DTRACE_NFSV4_2(op__putpubfh__done, struct compound_state *, cs, 3305 PUTPUBFH4res *, resp); 3306 } 3307 3308 /* 3309 * XXX - issue with put*fh operations. Suppose /export/home is exported. 3310 * Suppose an NFS client goes to mount /export/home/joe. If /export, home, 3311 * or joe have restrictive search permissions, then we shouldn't let 3312 * the client get a file handle. This is easy to enforce. However, we 3313 * don't know what security flavor should be used until we resolve the 3314 * path name. Another complication is uid mapping. If root is 3315 * the user, then it will be mapped to the anonymous user by default, 3316 * but we won't know that till we've resolved the path name. And we won't 3317 * know what the anonymous user is. 3318 * Luckily, SECINFO is specified to take a full filename. 3319 * So what we will have to in rfs4_op_lookup is check that flavor of 3320 * the target object matches that of the request, and if root was the 3321 * caller, check for the root= and anon= options, and if necessary, 3322 * repeat the lookup using the right cred_t. But that's not done yet. 3323 */ 3324 /* ARGSUSED */ 3325 static void 3326 rfs4_op_putfh(nfs_argop4 *argop, nfs_resop4 *resop, struct svc_req *req, 3327 struct compound_state *cs) 3328 { 3329 PUTFH4args *args = &argop->nfs_argop4_u.opputfh; 3330 PUTFH4res *resp = &resop->nfs_resop4_u.opputfh; 3331 nfs_fh4_fmt_t *fh_fmtp; 3332 3333 DTRACE_NFSV4_2(op__putfh__start, struct compound_state *, cs, 3334 PUTFH4args *, args); 3335 3336 if (cs->vp) { 3337 VN_RELE(cs->vp); 3338 cs->vp = NULL; 3339 } 3340 3341 if (cs->cr) { 3342 crfree(cs->cr); 3343 cs->cr = NULL; 3344 } 3345 3346 3347 if (args->object.nfs_fh4_len < NFS_FH4_LEN) { 3348 *cs->statusp = resp->status = NFS4ERR_BADHANDLE; 3349 goto out; 3350 } 3351 3352 fh_fmtp = (nfs_fh4_fmt_t *)args->object.nfs_fh4_val; 3353 cs->exi = checkexport4(&fh_fmtp->fh4_fsid, (fid_t *)&fh_fmtp->fh4_xlen, 3354 NULL); 3355 3356 if (cs->exi == NULL) { 3357 *cs->statusp = resp->status = NFS4ERR_STALE; 3358 goto out; 3359 } 3360 3361 cs->cr = crdup(cs->basecr); 3362 3363 ASSERT(cs->cr != NULL); 3364 3365 if (! (cs->vp = nfs4_fhtovp(&args->object, cs->exi, &resp->status))) { 3366 *cs->statusp = resp->status; 3367 goto out; 3368 } 3369 3370 if ((resp->status = call_checkauth4(cs, req)) != NFS4_OK) { 3371 VN_RELE(cs->vp); 3372 cs->vp = NULL; 3373 goto out; 3374 } 3375 3376 nfs_fh4_copy(&args->object, &cs->fh); 3377 *cs->statusp = resp->status = NFS4_OK; 3378 cs->deleg = FALSE; 3379 3380 out: 3381 DTRACE_NFSV4_2(op__putfh__done, struct compound_state *, cs, 3382 PUTFH4res *, resp); 3383 } 3384 3385 /* ARGSUSED */ 3386 static void 3387 rfs4_op_putrootfh(nfs_argop4 *argop, nfs_resop4 *resop, struct svc_req *req, 3388 struct compound_state *cs) 3389 3390 { 3391 PUTROOTFH4res *resp = &resop->nfs_resop4_u.opputrootfh; 3392 int error; 3393 fid_t fid; 3394 struct exportinfo *exi, *sav_exi; 3395 3396 DTRACE_NFSV4_1(op__putrootfh__start, struct compound_state *, cs); 3397 3398 if (cs->vp) { 3399 VN_RELE(cs->vp); 3400 cs->vp = NULL; 3401 } 3402 3403 if (cs->cr) 3404 crfree(cs->cr); 3405 3406 cs->cr = crdup(cs->basecr); 3407 3408 /* 3409 * Using rootdir, the system root vnode, 3410 * get its fid. 3411 */ 3412 bzero(&fid, sizeof (fid)); 3413 fid.fid_len = MAXFIDSZ; 3414 error = vop_fid_pseudo(rootdir, &fid); 3415 if (error != 0) { 3416 *cs->statusp = resp->status = puterrno4(error); 3417 goto out; 3418 } 3419 3420 /* 3421 * Then use the root fsid & fid it to find out if it's exported 3422 * 3423 * If the server root isn't exported directly, then 3424 * it should at least be a pseudo export based on 3425 * one or more exports further down in the server's 3426 * file tree. 3427 */ 3428 exi = checkexport4(&rootdir->v_vfsp->vfs_fsid, &fid, NULL); 3429 if (exi == NULL || exi->exi_export.ex_flags & EX_PUBLIC) { 3430 NFS4_DEBUG(rfs4_debug, 3431 (CE_WARN, "rfs4_op_putrootfh: export check failure")); 3432 *cs->statusp = resp->status = NFS4ERR_SERVERFAULT; 3433 goto out; 3434 } 3435 3436 /* 3437 * Now make a filehandle based on the root 3438 * export and root vnode. 3439 */ 3440 error = makefh4(&cs->fh, rootdir, exi); 3441 if (error != 0) { 3442 *cs->statusp = resp->status = puterrno4(error); 3443 goto out; 3444 } 3445 3446 sav_exi = cs->exi; 3447 cs->exi = exi; 3448 3449 VN_HOLD(rootdir); 3450 cs->vp = rootdir; 3451 3452 if ((resp->status = call_checkauth4(cs, req)) != NFS4_OK) { 3453 VN_RELE(rootdir); 3454 cs->vp = NULL; 3455 cs->exi = sav_exi; 3456 goto out; 3457 } 3458 3459 *cs->statusp = resp->status = NFS4_OK; 3460 cs->deleg = FALSE; 3461 out: 3462 DTRACE_NFSV4_2(op__putrootfh__done, struct compound_state *, cs, 3463 PUTROOTFH4res *, resp); 3464 } 3465 3466 /* 3467 * A directory entry is a valid nfsv4 entry if 3468 * - it has a non-zero ino 3469 * - it is not a dot or dotdot name 3470 * - it is visible in a pseudo export or in a real export that can 3471 * only have a limited view. 3472 */ 3473 static bool_t 3474 valid_nfs4_entry(struct exportinfo *exi, struct dirent64 *dp, 3475 int *expseudo, int check_visible) 3476 { 3477 if (dp->d_ino == 0 || NFS_IS_DOTNAME(dp->d_name)) { 3478 *expseudo = 0; 3479 return (FALSE); 3480 } 3481 3482 if (! check_visible) { 3483 *expseudo = 0; 3484 return (TRUE); 3485 } 3486 3487 return (nfs_visible_inode(exi, dp->d_ino, expseudo)); 3488 } 3489 3490 /* 3491 * set_rdattr_params sets up the variables used to manage what information 3492 * to get for each directory entry. 3493 */ 3494 static nfsstat4 3495 set_rdattr_params(struct nfs4_svgetit_arg *sargp, 3496 bitmap4 attrs, bool_t *need_to_lookup) 3497 { 3498 uint_t va_mask; 3499 nfsstat4 status; 3500 bitmap4 objbits; 3501 3502 status = bitmap4_to_attrmask(attrs, sargp); 3503 if (status != NFS4_OK) { 3504 /* 3505 * could not even figure attr mask 3506 */ 3507 return (status); 3508 } 3509 va_mask = sargp->vap->va_mask; 3510 3511 /* 3512 * dirent's d_ino is always correct value for mounted_on_fileid. 3513 * mntdfid_set is set once here, but mounted_on_fileid is 3514 * set in main dirent processing loop for each dirent. 3515 * The mntdfid_set is a simple optimization that lets the 3516 * server attr code avoid work when caller is readdir. 3517 */ 3518 sargp->mntdfid_set = TRUE; 3519 3520 /* 3521 * Lookup entry only if client asked for any of the following: 3522 * a) vattr attrs 3523 * b) vfs attrs 3524 * c) attrs w/per-object scope requested (change, filehandle, etc) 3525 * other than mounted_on_fileid (which we can take from dirent) 3526 */ 3527 objbits = attrs ? attrs & NFS4_VP_ATTR_MASK : 0; 3528 3529 if (va_mask || sargp->sbp || (objbits & ~FATTR4_MOUNTED_ON_FILEID_MASK)) 3530 *need_to_lookup = TRUE; 3531 else 3532 *need_to_lookup = FALSE; 3533 3534 if (sargp->sbp == NULL) 3535 return (NFS4_OK); 3536 3537 /* 3538 * If filesystem attrs are requested, get them now from the 3539 * directory vp, as most entries will have same filesystem. The only 3540 * exception are mounted over entries but we handle 3541 * those as we go (XXX mounted over detection not yet implemented). 3542 */ 3543 sargp->vap->va_mask = 0; /* to avoid VOP_GETATTR */ 3544 status = bitmap4_get_sysattrs(sargp); 3545 sargp->vap->va_mask = va_mask; 3546 3547 if ((status != NFS4_OK) && sargp->rdattr_error_req) { 3548 /* 3549 * Failed to get filesystem attributes. 3550 * Return a rdattr_error for each entry, but don't fail. 3551 * However, don't get any obj-dependent attrs. 3552 */ 3553 sargp->rdattr_error = status; /* for rdattr_error */ 3554 *need_to_lookup = FALSE; 3555 /* 3556 * At least get fileid for regular readdir output 3557 */ 3558 sargp->vap->va_mask &= AT_NODEID; 3559 status = NFS4_OK; 3560 } 3561 3562 return (status); 3563 } 3564 3565 /* 3566 * readlink: args: CURRENT_FH. 3567 * res: status. If success - CURRENT_FH unchanged, return linktext. 3568 */ 3569 3570 /* ARGSUSED */ 3571 static void 3572 rfs4_op_readlink(nfs_argop4 *argop, nfs_resop4 *resop, struct svc_req *req, 3573 struct compound_state *cs) 3574 { 3575 READLINK4res *resp = &resop->nfs_resop4_u.opreadlink; 3576 int error; 3577 vnode_t *vp; 3578 struct iovec iov; 3579 struct vattr va; 3580 struct uio uio; 3581 char *data; 3582 3583 DTRACE_NFSV4_1(op__readlink__start, struct compound_state *, cs); 3584 3585 /* CURRENT_FH: directory */ 3586 vp = cs->vp; 3587 if (vp == NULL) { 3588 *cs->statusp = resp->status = NFS4ERR_NOFILEHANDLE; 3589 goto out; 3590 } 3591 3592 if (cs->access == CS_ACCESS_DENIED) { 3593 *cs->statusp = resp->status = NFS4ERR_ACCESS; 3594 goto out; 3595 } 3596 3597 if (vp->v_type == VDIR) { 3598 *cs->statusp = resp->status = NFS4ERR_ISDIR; 3599 goto out; 3600 } 3601 3602 if (vp->v_type != VLNK) { 3603 *cs->statusp = resp->status = NFS4ERR_INVAL; 3604 goto out; 3605 } 3606 3607 va.va_mask = AT_MODE; 3608 error = VOP_GETATTR(vp, &va, 0, cs->cr, NULL); 3609 if (error) { 3610 *cs->statusp = resp->status = puterrno4(error); 3611 goto out; 3612 } 3613 3614 if (MANDLOCK(vp, va.va_mode)) { 3615 *cs->statusp = resp->status = NFS4ERR_ACCESS; 3616 goto out; 3617 } 3618 3619 data = kmem_alloc(MAXPATHLEN + 1, KM_SLEEP); 3620 3621 iov.iov_base = data; 3622 iov.iov_len = MAXPATHLEN; 3623 uio.uio_iov = &iov; 3624 uio.uio_iovcnt = 1; 3625 uio.uio_segflg = UIO_SYSSPACE; 3626 uio.uio_extflg = UIO_COPY_CACHED; 3627 uio.uio_loffset = 0; 3628 uio.uio_resid = MAXPATHLEN; 3629 3630 error = VOP_READLINK(vp, &uio, cs->cr, NULL); 3631 3632 if (error) { 3633 kmem_free((caddr_t)data, (uint_t)MAXPATHLEN + 1); 3634 *cs->statusp = resp->status = puterrno4(error); 3635 goto out; 3636 } 3637 3638 *(data + MAXPATHLEN - uio.uio_resid) = '\0'; 3639 3640 /* 3641 * treat link name as data 3642 */ 3643 (void) str_to_utf8(data, &resp->link); 3644 3645 kmem_free((caddr_t)data, (uint_t)MAXPATHLEN + 1); 3646 *cs->statusp = resp->status = NFS4_OK; 3647 3648 out: 3649 DTRACE_NFSV4_2(op__readlink__done, struct compound_state *, cs, 3650 READLINK4res *, resp); 3651 } 3652 3653 static void 3654 rfs4_op_readlink_free(nfs_resop4 *resop) 3655 { 3656 READLINK4res *resp = &resop->nfs_resop4_u.opreadlink; 3657 utf8string *symlink = &resp->link; 3658 3659 if (symlink->utf8string_val) { 3660 UTF8STRING_FREE(*symlink) 3661 } 3662 } 3663 3664 /* 3665 * release_lockowner: 3666 * Release any state associated with the supplied 3667 * lockowner. Note if any lo_state is holding locks we will not 3668 * rele that lo_state and thus the lockowner will not be destroyed. 3669 * A client using lock after the lock owner stateid has been released 3670 * will suffer the consequence of NFS4ERR_BAD_STATEID and would have 3671 * to reissue the lock with new_lock_owner set to TRUE. 3672 * args: lock_owner 3673 * res: status 3674 */ 3675 /* ARGSUSED */ 3676 static void 3677 rfs4_op_release_lockowner(nfs_argop4 *argop, nfs_resop4 *resop, 3678 struct svc_req *req, struct compound_state *cs) 3679 { 3680 RELEASE_LOCKOWNER4args *ap = &argop->nfs_argop4_u.oprelease_lockowner; 3681 RELEASE_LOCKOWNER4res *resp = &resop->nfs_resop4_u.oprelease_lockowner; 3682 rfs4_lockowner_t *lo; 3683 rfs4_openowner_t *oop; 3684 rfs4_state_t *sp; 3685 rfs4_lo_state_t *lsp; 3686 rfs4_client_t *cp; 3687 bool_t create = FALSE; 3688 locklist_t *llist; 3689 sysid_t sysid; 3690 3691 DTRACE_NFSV4_2(op__release__lockowner__start, struct compound_state *, 3692 cs, RELEASE_LOCKOWNER4args *, ap); 3693 3694 /* Make sure there is a clientid around for this request */ 3695 cp = rfs4_findclient_by_id(ap->lock_owner.clientid, FALSE); 3696 3697 if (cp == NULL) { 3698 *cs->statusp = resp->status = 3699 rfs4_check_clientid(&ap->lock_owner.clientid, 0); 3700 goto out; 3701 } 3702 rfs4_client_rele(cp); 3703 3704 lo = rfs4_findlockowner(&ap->lock_owner, &create); 3705 if (lo == NULL) { 3706 *cs->statusp = resp->status = NFS4_OK; 3707 goto out; 3708 } 3709 ASSERT(lo->client != NULL); 3710 3711 /* 3712 * Check for EXPIRED client. If so will reap state with in a lease 3713 * period or on next set_clientid_confirm step 3714 */ 3715 if (rfs4_lease_expired(lo->client)) { 3716 rfs4_lockowner_rele(lo); 3717 *cs->statusp = resp->status = NFS4ERR_EXPIRED; 3718 goto out; 3719 } 3720 3721 /* 3722 * If no sysid has been assigned, then no locks exist; just return. 3723 */ 3724 rfs4_dbe_lock(lo->client->dbe); 3725 if (lo->client->sysidt == LM_NOSYSID) { 3726 rfs4_lockowner_rele(lo); 3727 rfs4_dbe_unlock(lo->client->dbe); 3728 goto out; 3729 } 3730 3731 sysid = lo->client->sysidt; 3732 rfs4_dbe_unlock(lo->client->dbe); 3733 3734 /* 3735 * Mark the lockowner invalid. 3736 */ 3737 rfs4_dbe_hide(lo->dbe); 3738 3739 /* 3740 * sysid-pid pair should now not be used since the lockowner is 3741 * invalid. If the client were to instantiate the lockowner again 3742 * it would be assigned a new pid. Thus we can get the list of 3743 * current locks. 3744 */ 3745 3746 llist = flk_get_active_locks(sysid, lo->pid); 3747 /* If we are still holding locks fail */ 3748 if (llist != NULL) { 3749 3750 *cs->statusp = resp->status = NFS4ERR_LOCKS_HELD; 3751 3752 flk_free_locklist(llist); 3753 /* 3754 * We need to unhide the lockowner so the client can 3755 * try it again. The bad thing here is if the client 3756 * has a logic error that took it here in the first place 3757 * he probably has lost accounting of the locks that it 3758 * is holding. So we may have dangling state until the 3759 * open owner state is reaped via close. One scenario 3760 * that could possibly occur is that the client has 3761 * sent the unlock request(s) in separate threads 3762 * and has not waited for the replies before sending the 3763 * RELEASE_LOCKOWNER request. Presumably, it would expect 3764 * and deal appropriately with NFS4ERR_LOCKS_HELD, by 3765 * reissuing the request. 3766 */ 3767 rfs4_dbe_unhide(lo->dbe); 3768 rfs4_lockowner_rele(lo); 3769 goto out; 3770 } 3771 3772 /* 3773 * For the corresponding client we need to check each open 3774 * owner for any opens that have lockowner state associated 3775 * with this lockowner. 3776 */ 3777 3778 rfs4_dbe_lock(lo->client->dbe); 3779 for (oop = lo->client->openownerlist.next->oop; oop != NULL; 3780 oop = oop->openownerlist.next->oop) { 3781 3782 rfs4_dbe_lock(oop->dbe); 3783 for (sp = oop->ownerstateids.next->sp; sp != NULL; 3784 sp = sp->ownerstateids.next->sp) { 3785 3786 rfs4_dbe_lock(sp->dbe); 3787 for (lsp = sp->lockownerlist.next->lsp; 3788 lsp != NULL; lsp = lsp->lockownerlist.next->lsp) { 3789 if (lsp->locker == lo) { 3790 rfs4_dbe_lock(lsp->dbe); 3791 rfs4_dbe_invalidate(lsp->dbe); 3792 rfs4_dbe_unlock(lsp->dbe); 3793 } 3794 } 3795 rfs4_dbe_unlock(sp->dbe); 3796 } 3797 rfs4_dbe_unlock(oop->dbe); 3798 } 3799 rfs4_dbe_unlock(lo->client->dbe); 3800 3801 rfs4_lockowner_rele(lo); 3802 3803 *cs->statusp = resp->status = NFS4_OK; 3804 3805 out: 3806 DTRACE_NFSV4_2(op__release__lockowner__done, struct compound_state *, 3807 cs, RELEASE_LOCKOWNER4res *, resp); 3808 } 3809 3810 /* 3811 * short utility function to lookup a file and recall the delegation 3812 */ 3813 static rfs4_file_t * 3814 rfs4_lookup_and_findfile(vnode_t *dvp, char *nm, vnode_t **vpp, 3815 int *lkup_error, cred_t *cr) 3816 { 3817 vnode_t *vp; 3818 rfs4_file_t *fp = NULL; 3819 bool_t fcreate = FALSE; 3820 int error; 3821 3822 if (vpp) 3823 *vpp = NULL; 3824 3825 if ((error = VOP_LOOKUP(dvp, nm, &vp, NULL, 0, NULL, cr, NULL, NULL, 3826 NULL)) == 0) { 3827 if (vp->v_type == VREG) 3828 fp = rfs4_findfile(vp, NULL, &fcreate); 3829 if (vpp) 3830 *vpp = vp; 3831 else 3832 VN_RELE(vp); 3833 } 3834 3835 if (lkup_error) 3836 *lkup_error = error; 3837 3838 return (fp); 3839 } 3840 3841 /* 3842 * remove: args: CURRENT_FH: directory; name. 3843 * res: status. If success - CURRENT_FH unchanged, return change_info 3844 * for directory. 3845 */ 3846 /* ARGSUSED */ 3847 static void 3848 rfs4_op_remove(nfs_argop4 *argop, nfs_resop4 *resop, struct svc_req *req, 3849 struct compound_state *cs) 3850 { 3851 REMOVE4args *args = &argop->nfs_argop4_u.opremove; 3852 REMOVE4res *resp = &resop->nfs_resop4_u.opremove; 3853 int error; 3854 vnode_t *dvp, *vp; 3855 struct vattr bdva, idva, adva; 3856 char *nm; 3857 uint_t len; 3858 rfs4_file_t *fp; 3859 int in_crit = 0; 3860 bslabel_t *clabel; 3861 3862 DTRACE_NFSV4_2(op__remove__start, struct compound_state *, cs, 3863 REMOVE4args *, args); 3864 3865 /* CURRENT_FH: directory */ 3866 dvp = cs->vp; 3867 if (dvp == NULL) { 3868 *cs->statusp = resp->status = NFS4ERR_NOFILEHANDLE; 3869 goto out; 3870 } 3871 3872 if (cs->access == CS_ACCESS_DENIED) { 3873 *cs->statusp = resp->status = NFS4ERR_ACCESS; 3874 goto out; 3875 } 3876 3877 /* 3878 * If there is an unshared filesystem mounted on this vnode, 3879 * Do not allow to remove anything in this directory. 3880 */ 3881 if (vn_ismntpt(dvp)) { 3882 *cs->statusp = resp->status = NFS4ERR_ACCESS; 3883 goto out; 3884 } 3885 3886 if (dvp->v_type != VDIR) { 3887 *cs->statusp = resp->status = NFS4ERR_NOTDIR; 3888 goto out; 3889 } 3890 3891 if (!utf8_dir_verify(&args->target)) { 3892 *cs->statusp = resp->status = NFS4ERR_INVAL; 3893 goto out; 3894 } 3895 3896 /* 3897 * Lookup the file so that we can check if it's a directory 3898 */ 3899 nm = utf8_to_fn(&args->target, &len, NULL); 3900 if (nm == NULL) { 3901 *cs->statusp = resp->status = NFS4ERR_INVAL; 3902 goto out; 3903 } 3904 3905 if (len > MAXNAMELEN) { 3906 *cs->statusp = resp->status = NFS4ERR_NAMETOOLONG; 3907 kmem_free(nm, len); 3908 goto out; 3909 } 3910 3911 if (rdonly4(cs->exi, cs->vp, req)) { 3912 *cs->statusp = resp->status = NFS4ERR_ROFS; 3913 kmem_free(nm, len); 3914 goto out; 3915 } 3916 3917 /* 3918 * Lookup the file to determine type and while we are see if 3919 * there is a file struct around and check for delegation. 3920 * We don't need to acquire va_seq before this lookup, if 3921 * it causes an update, cinfo.before will not match, which will 3922 * trigger a cache flush even if atomic is TRUE. 3923 */ 3924 if (fp = rfs4_lookup_and_findfile(dvp, nm, &vp, &error, cs->cr)) { 3925 if (rfs4_check_delegated_byfp(FWRITE, fp, TRUE, TRUE, TRUE, 3926 NULL)) { 3927 VN_RELE(vp); 3928 rfs4_file_rele(fp); 3929 *cs->statusp = resp->status = NFS4ERR_DELAY; 3930 kmem_free(nm, len); 3931 goto out; 3932 } 3933 } 3934 3935 /* Didn't find anything to remove */ 3936 if (vp == NULL) { 3937 *cs->statusp = resp->status = error; 3938 kmem_free(nm, len); 3939 goto out; 3940 } 3941 3942 if (nbl_need_check(vp)) { 3943 nbl_start_crit(vp, RW_READER); 3944 in_crit = 1; 3945 if (nbl_conflict(vp, NBL_REMOVE, 0, 0, 0, NULL)) { 3946 *cs->statusp = resp->status = NFS4ERR_FILE_OPEN; 3947 kmem_free(nm, len); 3948 nbl_end_crit(vp); 3949 VN_RELE(vp); 3950 if (fp) { 3951 rfs4_clear_dont_grant(fp); 3952 rfs4_file_rele(fp); 3953 } 3954 goto out; 3955 } 3956 } 3957 3958 /* check label before allowing removal */ 3959 if (is_system_labeled()) { 3960 ASSERT(req->rq_label != NULL); 3961 clabel = req->rq_label; 3962 DTRACE_PROBE2(tx__rfs4__log__info__opremove__clabel, char *, 3963 "got client label from request(1)", 3964 struct svc_req *, req); 3965 if (!blequal(&l_admin_low->tsl_label, clabel)) { 3966 if (!do_rfs_label_check(clabel, vp, EQUALITY_CHECK)) { 3967 *cs->statusp = resp->status = NFS4ERR_ACCESS; 3968 kmem_free(nm, len); 3969 if (in_crit) 3970 nbl_end_crit(vp); 3971 VN_RELE(vp); 3972 if (fp) { 3973 rfs4_clear_dont_grant(fp); 3974 rfs4_file_rele(fp); 3975 } 3976 goto out; 3977 } 3978 } 3979 } 3980 3981 /* Get dir "before" change value */ 3982 bdva.va_mask = AT_CTIME|AT_SEQ; 3983 error = VOP_GETATTR(dvp, &bdva, 0, cs->cr, NULL); 3984 if (error) { 3985 *cs->statusp = resp->status = puterrno4(error); 3986 kmem_free(nm, len); 3987 goto out; 3988 } 3989 NFS4_SET_FATTR4_CHANGE(resp->cinfo.before, bdva.va_ctime) 3990 3991 /* Actually do the REMOVE operation */ 3992 if (vp->v_type == VDIR) { 3993 /* 3994 * Can't remove a directory that has a mounted-on filesystem. 3995 */ 3996 if (vn_ismntpt(vp)) { 3997 error = EACCES; 3998 } else { 3999 /* 4000 * System V defines rmdir to return EEXIST, 4001 * not * ENOTEMPTY, if the directory is not 4002 * empty. A System V NFS server needs to map 4003 * NFS4ERR_EXIST to NFS4ERR_NOTEMPTY to 4004 * transmit over the wire. 4005 */ 4006 if ((error = VOP_RMDIR(dvp, nm, rootdir, cs->cr, 4007 NULL, 0)) == EEXIST) 4008 error = ENOTEMPTY; 4009 } 4010 } else { 4011 if ((error = VOP_REMOVE(dvp, nm, cs->cr, NULL, 0)) == 0 && 4012 fp != NULL) { 4013 struct vattr va; 4014 vnode_t *tvp; 4015 4016 rfs4_dbe_lock(fp->dbe); 4017 tvp = fp->vp; 4018 if (tvp) 4019 VN_HOLD(tvp); 4020 rfs4_dbe_unlock(fp->dbe); 4021 4022 if (tvp) { 4023 /* 4024 * This is va_seq safe because we are not 4025 * manipulating dvp. 4026 */ 4027 va.va_mask = AT_NLINK; 4028 if (!VOP_GETATTR(tvp, &va, 0, cs->cr, NULL) && 4029 va.va_nlink == 0) { 4030 /* Remove state on file remove */ 4031 if (in_crit) { 4032 nbl_end_crit(vp); 4033 in_crit = 0; 4034 } 4035 rfs4_close_all_state(fp); 4036 } 4037 VN_RELE(tvp); 4038 } 4039 } 4040 } 4041 4042 if (in_crit) 4043 nbl_end_crit(vp); 4044 VN_RELE(vp); 4045 4046 if (fp) { 4047 rfs4_clear_dont_grant(fp); 4048 rfs4_file_rele(fp); 4049 } 4050 kmem_free(nm, len); 4051 4052 if (error) { 4053 *cs->statusp = resp->status = puterrno4(error); 4054 goto out; 4055 } 4056 4057 /* 4058 * Get the initial "after" sequence number, if it fails, set to zero 4059 */ 4060 idva.va_mask = AT_SEQ; 4061 if (VOP_GETATTR(dvp, &idva, 0, cs->cr, NULL)) 4062 idva.va_seq = 0; 4063 4064 /* 4065 * Force modified data and metadata out to stable storage. 4066 */ 4067 (void) VOP_FSYNC(dvp, 0, cs->cr, NULL); 4068 4069 /* 4070 * Get "after" change value, if it fails, simply return the 4071 * before value. 4072 */ 4073 adva.va_mask = AT_CTIME|AT_SEQ; 4074 if (VOP_GETATTR(dvp, &adva, 0, cs->cr, NULL)) { 4075 adva.va_ctime = bdva.va_ctime; 4076 adva.va_seq = 0; 4077 } 4078 4079 NFS4_SET_FATTR4_CHANGE(resp->cinfo.after, adva.va_ctime) 4080 4081 /* 4082 * The cinfo.atomic = TRUE only if we have 4083 * non-zero va_seq's, and it has incremented by exactly one 4084 * during the VOP_REMOVE/RMDIR and it didn't change during 4085 * the VOP_FSYNC. 4086 */ 4087 if (bdva.va_seq && idva.va_seq && adva.va_seq && 4088 idva.va_seq == (bdva.va_seq + 1) && idva.va_seq == adva.va_seq) 4089 resp->cinfo.atomic = TRUE; 4090 else 4091 resp->cinfo.atomic = FALSE; 4092 4093 *cs->statusp = resp->status = NFS4_OK; 4094 4095 out: 4096 DTRACE_NFSV4_2(op__remove__done, struct compound_state *, cs, 4097 REMOVE4res *, resp); 4098 } 4099 4100 /* 4101 * rename: args: SAVED_FH: from directory, CURRENT_FH: target directory, 4102 * oldname and newname. 4103 * res: status. If success - CURRENT_FH unchanged, return change_info 4104 * for both from and target directories. 4105 */ 4106 /* ARGSUSED */ 4107 static void 4108 rfs4_op_rename(nfs_argop4 *argop, nfs_resop4 *resop, struct svc_req *req, 4109 struct compound_state *cs) 4110 { 4111 RENAME4args *args = &argop->nfs_argop4_u.oprename; 4112 RENAME4res *resp = &resop->nfs_resop4_u.oprename; 4113 int error; 4114 vnode_t *odvp; 4115 vnode_t *ndvp; 4116 vnode_t *srcvp, *targvp; 4117 struct vattr obdva, oidva, oadva; 4118 struct vattr nbdva, nidva, nadva; 4119 char *onm, *nnm; 4120 uint_t olen, nlen; 4121 rfs4_file_t *fp, *sfp; 4122 int in_crit_src, in_crit_targ; 4123 int fp_rele_grant_hold, sfp_rele_grant_hold; 4124 bslabel_t *clabel; 4125 4126 DTRACE_NFSV4_2(op__rename__start, struct compound_state *, cs, 4127 RENAME4args *, args); 4128 4129 fp = sfp = NULL; 4130 srcvp = targvp = NULL; 4131 in_crit_src = in_crit_targ = 0; 4132 fp_rele_grant_hold = sfp_rele_grant_hold = 0; 4133 4134 /* CURRENT_FH: target directory */ 4135 ndvp = cs->vp; 4136 if (ndvp == NULL) { 4137 *cs->statusp = resp->status = NFS4ERR_NOFILEHANDLE; 4138 goto out; 4139 } 4140 4141 /* SAVED_FH: from directory */ 4142 odvp = cs->saved_vp; 4143 if (odvp == NULL) { 4144 *cs->statusp = resp->status = NFS4ERR_NOFILEHANDLE; 4145 goto out; 4146 } 4147 4148 if (cs->access == CS_ACCESS_DENIED) { 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 objects in this directory. 4156 */ 4157 if (vn_ismntpt(odvp)) { 4158 *cs->statusp = resp->status = NFS4ERR_ACCESS; 4159 goto out; 4160 } 4161 4162 /* 4163 * If there is an unshared filesystem mounted on this vnode, 4164 * do not allow to rename to this directory. 4165 */ 4166 if (vn_ismntpt(ndvp)) { 4167 *cs->statusp = resp->status = NFS4ERR_ACCESS; 4168 goto out; 4169 } 4170 4171 if (odvp->v_type != VDIR || ndvp->v_type != VDIR) { 4172 *cs->statusp = resp->status = NFS4ERR_NOTDIR; 4173 goto out; 4174 } 4175 4176 if (cs->saved_exi != cs->exi) { 4177 *cs->statusp = resp->status = NFS4ERR_XDEV; 4178 goto out; 4179 } 4180 4181 if (!utf8_dir_verify(&args->oldname)) { 4182 *cs->statusp = resp->status = NFS4ERR_INVAL; 4183 goto out; 4184 } 4185 4186 if (!utf8_dir_verify(&args->newname)) { 4187 *cs->statusp = resp->status = NFS4ERR_INVAL; 4188 goto out; 4189 } 4190 4191 onm = utf8_to_fn(&args->oldname, &olen, NULL); 4192 if (onm == NULL) { 4193 *cs->statusp = resp->status = NFS4ERR_INVAL; 4194 goto out; 4195 } 4196 4197 nnm = utf8_to_fn(&args->newname, &nlen, NULL); 4198 if (nnm == NULL) { 4199 *cs->statusp = resp->status = NFS4ERR_INVAL; 4200 kmem_free(onm, olen); 4201 goto out; 4202 } 4203 4204 if (olen > MAXNAMELEN || nlen > MAXNAMELEN) { 4205 *cs->statusp = resp->status = NFS4ERR_NAMETOOLONG; 4206 kmem_free(onm, olen); 4207 kmem_free(nnm, nlen); 4208 goto out; 4209 } 4210 4211 4212 if (rdonly4(cs->exi, cs->vp, req)) { 4213 *cs->statusp = resp->status = NFS4ERR_ROFS; 4214 kmem_free(onm, olen); 4215 kmem_free(nnm, nlen); 4216 goto out; 4217 } 4218 4219 /* check label of the target dir */ 4220 if (is_system_labeled()) { 4221 ASSERT(req->rq_label != NULL); 4222 clabel = req->rq_label; 4223 DTRACE_PROBE2(tx__rfs4__log__info__oprename__clabel, char *, 4224 "got client label from request(1)", 4225 struct svc_req *, req); 4226 if (!blequal(&l_admin_low->tsl_label, clabel)) { 4227 if (!do_rfs_label_check(clabel, ndvp, 4228 EQUALITY_CHECK)) { 4229 *cs->statusp = resp->status = NFS4ERR_ACCESS; 4230 goto out; 4231 } 4232 } 4233 } 4234 4235 /* 4236 * Is the source a file and have a delegation? 4237 * We don't need to acquire va_seq before these lookups, if 4238 * it causes an update, cinfo.before will not match, which will 4239 * trigger a cache flush even if atomic is TRUE. 4240 */ 4241 if (sfp = rfs4_lookup_and_findfile(odvp, onm, &srcvp, &error, cs->cr)) { 4242 if (rfs4_check_delegated_byfp(FWRITE, sfp, TRUE, TRUE, TRUE, 4243 NULL)) { 4244 *cs->statusp = resp->status = NFS4ERR_DELAY; 4245 goto err_out; 4246 } 4247 } 4248 4249 if (srcvp == NULL) { 4250 *cs->statusp = resp->status = puterrno4(error); 4251 kmem_free(onm, olen); 4252 kmem_free(nnm, nlen); 4253 goto out; 4254 } 4255 4256 sfp_rele_grant_hold = 1; 4257 4258 /* Does the destination exist and a file and have a delegation? */ 4259 if (fp = rfs4_lookup_and_findfile(ndvp, nnm, &targvp, NULL, cs->cr)) { 4260 if (rfs4_check_delegated_byfp(FWRITE, fp, TRUE, TRUE, TRUE, 4261 NULL)) { 4262 *cs->statusp = resp->status = NFS4ERR_DELAY; 4263 goto err_out; 4264 } 4265 } 4266 fp_rele_grant_hold = 1; 4267 4268 4269 /* Check for NBMAND lock on both source and target */ 4270 if (nbl_need_check(srcvp)) { 4271 nbl_start_crit(srcvp, RW_READER); 4272 in_crit_src = 1; 4273 if (nbl_conflict(srcvp, NBL_RENAME, 0, 0, 0, NULL)) { 4274 *cs->statusp = resp->status = NFS4ERR_FILE_OPEN; 4275 goto err_out; 4276 } 4277 } 4278 4279 if (targvp && nbl_need_check(targvp)) { 4280 nbl_start_crit(targvp, RW_READER); 4281 in_crit_targ = 1; 4282 if (nbl_conflict(targvp, NBL_REMOVE, 0, 0, 0, NULL)) { 4283 *cs->statusp = resp->status = NFS4ERR_FILE_OPEN; 4284 goto err_out; 4285 } 4286 } 4287 4288 /* Get source "before" change value */ 4289 obdva.va_mask = AT_CTIME|AT_SEQ; 4290 error = VOP_GETATTR(odvp, &obdva, 0, cs->cr, NULL); 4291 if (!error) { 4292 nbdva.va_mask = AT_CTIME|AT_SEQ; 4293 error = VOP_GETATTR(ndvp, &nbdva, 0, cs->cr, NULL); 4294 } 4295 if (error) { 4296 *cs->statusp = resp->status = puterrno4(error); 4297 goto err_out; 4298 } 4299 4300 NFS4_SET_FATTR4_CHANGE(resp->source_cinfo.before, obdva.va_ctime) 4301 NFS4_SET_FATTR4_CHANGE(resp->target_cinfo.before, nbdva.va_ctime) 4302 4303 if ((error = VOP_RENAME(odvp, onm, ndvp, nnm, cs->cr, NULL, 0)) == 0 && 4304 fp != NULL) { 4305 struct vattr va; 4306 vnode_t *tvp; 4307 4308 rfs4_dbe_lock(fp->dbe); 4309 tvp = fp->vp; 4310 if (tvp) 4311 VN_HOLD(tvp); 4312 rfs4_dbe_unlock(fp->dbe); 4313 4314 if (tvp) { 4315 va.va_mask = AT_NLINK; 4316 if (!VOP_GETATTR(tvp, &va, 0, cs->cr, NULL) && 4317 va.va_nlink == 0) { 4318 /* The file is gone and so should the state */ 4319 if (in_crit_targ) { 4320 nbl_end_crit(targvp); 4321 in_crit_targ = 0; 4322 } 4323 rfs4_close_all_state(fp); 4324 } 4325 VN_RELE(tvp); 4326 } 4327 } 4328 if (error == 0) 4329 vn_renamepath(ndvp, srcvp, nnm, nlen - 1); 4330 4331 if (in_crit_src) 4332 nbl_end_crit(srcvp); 4333 if (srcvp) 4334 VN_RELE(srcvp); 4335 if (in_crit_targ) 4336 nbl_end_crit(targvp); 4337 if (targvp) 4338 VN_RELE(targvp); 4339 4340 if (sfp) { 4341 rfs4_clear_dont_grant(sfp); 4342 rfs4_file_rele(sfp); 4343 } 4344 if (fp) { 4345 rfs4_clear_dont_grant(fp); 4346 rfs4_file_rele(fp); 4347 } 4348 4349 kmem_free(onm, olen); 4350 kmem_free(nnm, nlen); 4351 4352 /* 4353 * Get the initial "after" sequence number, if it fails, set to zero 4354 */ 4355 oidva.va_mask = AT_SEQ; 4356 if (VOP_GETATTR(odvp, &oidva, 0, cs->cr, NULL)) 4357 oidva.va_seq = 0; 4358 4359 nidva.va_mask = AT_SEQ; 4360 if (VOP_GETATTR(ndvp, &nidva, 0, cs->cr, NULL)) 4361 nidva.va_seq = 0; 4362 4363 /* 4364 * Force modified data and metadata out to stable storage. 4365 */ 4366 (void) VOP_FSYNC(odvp, 0, cs->cr, NULL); 4367 (void) VOP_FSYNC(ndvp, 0, cs->cr, NULL); 4368 4369 if (error) { 4370 *cs->statusp = resp->status = puterrno4(error); 4371 goto out; 4372 } 4373 4374 /* 4375 * Get "after" change values, if it fails, simply return the 4376 * before value. 4377 */ 4378 oadva.va_mask = AT_CTIME|AT_SEQ; 4379 if (VOP_GETATTR(odvp, &oadva, 0, cs->cr, NULL)) { 4380 oadva.va_ctime = obdva.va_ctime; 4381 oadva.va_seq = 0; 4382 } 4383 4384 nadva.va_mask = AT_CTIME|AT_SEQ; 4385 if (VOP_GETATTR(odvp, &nadva, 0, cs->cr, NULL)) { 4386 nadva.va_ctime = nbdva.va_ctime; 4387 nadva.va_seq = 0; 4388 } 4389 4390 NFS4_SET_FATTR4_CHANGE(resp->source_cinfo.after, oadva.va_ctime) 4391 NFS4_SET_FATTR4_CHANGE(resp->target_cinfo.after, nadva.va_ctime) 4392 4393 /* 4394 * The cinfo.atomic = TRUE only if we have 4395 * non-zero va_seq's, and it has incremented by exactly one 4396 * during the VOP_RENAME and it didn't change during the VOP_FSYNC. 4397 */ 4398 if (obdva.va_seq && oidva.va_seq && oadva.va_seq && 4399 oidva.va_seq == (obdva.va_seq + 1) && oidva.va_seq == oadva.va_seq) 4400 resp->source_cinfo.atomic = TRUE; 4401 else 4402 resp->source_cinfo.atomic = FALSE; 4403 4404 if (nbdva.va_seq && nidva.va_seq && nadva.va_seq && 4405 nidva.va_seq == (nbdva.va_seq + 1) && nidva.va_seq == nadva.va_seq) 4406 resp->target_cinfo.atomic = TRUE; 4407 else 4408 resp->target_cinfo.atomic = FALSE; 4409 4410 #ifdef VOLATILE_FH_TEST 4411 { 4412 extern void add_volrnm_fh(struct exportinfo *, vnode_t *); 4413 4414 /* 4415 * Add the renamed file handle to the volatile rename list 4416 */ 4417 if (cs->exi->exi_export.ex_flags & EX_VOLRNM) { 4418 /* file handles may expire on rename */ 4419 vnode_t *vp; 4420 4421 nnm = utf8_to_fn(&args->newname, &nlen, NULL); 4422 /* 4423 * Already know that nnm will be a valid string 4424 */ 4425 error = VOP_LOOKUP(ndvp, nnm, &vp, NULL, 0, NULL, cs->cr, 4426 NULL, NULL, NULL); 4427 kmem_free(nnm, nlen); 4428 if (!error) { 4429 add_volrnm_fh(cs->exi, vp); 4430 VN_RELE(vp); 4431 } 4432 } 4433 } 4434 #endif /* VOLATILE_FH_TEST */ 4435 4436 *cs->statusp = resp->status = NFS4_OK; 4437 out: 4438 DTRACE_NFSV4_2(op__rename__done, struct compound_state *, cs, 4439 RENAME4res *, resp); 4440 return; 4441 4442 err_out: 4443 kmem_free(onm, olen); 4444 kmem_free(nnm, nlen); 4445 4446 if (in_crit_src) nbl_end_crit(srcvp); 4447 if (in_crit_targ) nbl_end_crit(targvp); 4448 if (targvp) VN_RELE(targvp); 4449 if (srcvp) VN_RELE(srcvp); 4450 if (sfp) { 4451 if (sfp_rele_grant_hold) rfs4_clear_dont_grant(sfp); 4452 rfs4_file_rele(sfp); 4453 } 4454 if (fp) { 4455 if (fp_rele_grant_hold) rfs4_clear_dont_grant(fp); 4456 rfs4_file_rele(fp); 4457 } 4458 4459 DTRACE_NFSV4_2(op__rename__done, struct compound_state *, cs, 4460 RENAME4res *, resp); 4461 } 4462 4463 /* ARGSUSED */ 4464 static void 4465 rfs4_op_renew(nfs_argop4 *argop, nfs_resop4 *resop, struct svc_req *req, 4466 struct compound_state *cs) 4467 { 4468 RENEW4args *args = &argop->nfs_argop4_u.oprenew; 4469 RENEW4res *resp = &resop->nfs_resop4_u.oprenew; 4470 rfs4_client_t *cp; 4471 4472 DTRACE_NFSV4_2(op__renew__start, struct compound_state *, cs, 4473 RENEW4args *, args); 4474 4475 if ((cp = rfs4_findclient_by_id(args->clientid, FALSE)) == NULL) { 4476 *cs->statusp = resp->status = 4477 rfs4_check_clientid(&args->clientid, 0); 4478 goto out; 4479 } 4480 4481 if (rfs4_lease_expired(cp)) { 4482 rfs4_client_rele(cp); 4483 *cs->statusp = resp->status = NFS4ERR_EXPIRED; 4484 goto out; 4485 } 4486 4487 rfs4_update_lease(cp); 4488 4489 mutex_enter(cp->cbinfo.cb_lock); 4490 if (cp->cbinfo.cb_notified_of_cb_path_down == FALSE) { 4491 cp->cbinfo.cb_notified_of_cb_path_down = TRUE; 4492 *cs->statusp = resp->status = NFS4ERR_CB_PATH_DOWN; 4493 } else { 4494 *cs->statusp = resp->status = NFS4_OK; 4495 } 4496 mutex_exit(cp->cbinfo.cb_lock); 4497 4498 rfs4_client_rele(cp); 4499 4500 out: 4501 DTRACE_NFSV4_2(op__renew__done, struct compound_state *, cs, 4502 RENEW4res *, resp); 4503 } 4504 4505 /* ARGSUSED */ 4506 static void 4507 rfs4_op_restorefh(nfs_argop4 *args, nfs_resop4 *resop, struct svc_req *req, 4508 struct compound_state *cs) 4509 { 4510 RESTOREFH4res *resp = &resop->nfs_resop4_u.oprestorefh; 4511 4512 DTRACE_NFSV4_1(op__restorefh__start, struct compound_state *, cs); 4513 4514 /* No need to check cs->access - we are not accessing any object */ 4515 if ((cs->saved_vp == NULL) || (cs->saved_fh.nfs_fh4_val == NULL)) { 4516 *cs->statusp = resp->status = NFS4ERR_RESTOREFH; 4517 goto out; 4518 } 4519 if (cs->vp != NULL) { 4520 VN_RELE(cs->vp); 4521 } 4522 cs->vp = cs->saved_vp; 4523 cs->saved_vp = NULL; 4524 cs->exi = cs->saved_exi; 4525 nfs_fh4_copy(&cs->saved_fh, &cs->fh); 4526 *cs->statusp = resp->status = NFS4_OK; 4527 cs->deleg = FALSE; 4528 4529 out: 4530 DTRACE_NFSV4_2(op__restorefh__done, struct compound_state *, cs, 4531 RESTOREFH4res *, resp); 4532 } 4533 4534 /* ARGSUSED */ 4535 static void 4536 rfs4_op_savefh(nfs_argop4 *argop, nfs_resop4 *resop, struct svc_req *req, 4537 struct compound_state *cs) 4538 { 4539 SAVEFH4res *resp = &resop->nfs_resop4_u.opsavefh; 4540 4541 DTRACE_NFSV4_1(op__savefh__start, struct compound_state *, cs); 4542 4543 /* No need to check cs->access - we are not accessing any object */ 4544 if (cs->vp == NULL) { 4545 *cs->statusp = resp->status = NFS4ERR_NOFILEHANDLE; 4546 goto out; 4547 } 4548 if (cs->saved_vp != NULL) { 4549 VN_RELE(cs->saved_vp); 4550 } 4551 cs->saved_vp = cs->vp; 4552 VN_HOLD(cs->saved_vp); 4553 cs->saved_exi = cs->exi; 4554 /* 4555 * since SAVEFH is fairly rare, don't alloc space for its fh 4556 * unless necessary. 4557 */ 4558 if (cs->saved_fh.nfs_fh4_val == NULL) { 4559 cs->saved_fh.nfs_fh4_val = kmem_alloc(NFS4_FHSIZE, KM_SLEEP); 4560 } 4561 nfs_fh4_copy(&cs->fh, &cs->saved_fh); 4562 *cs->statusp = resp->status = NFS4_OK; 4563 4564 out: 4565 DTRACE_NFSV4_2(op__savefh__done, struct compound_state *, cs, 4566 SAVEFH4res *, resp); 4567 } 4568 4569 /* 4570 * rfs4_verify_attr is called when nfsv4 Setattr failed, but we wish to 4571 * return the bitmap of attrs that were set successfully. It is also 4572 * called by Verify/Nverify to test the vattr/vfsstat attrs. It should 4573 * always be called only after rfs4_do_set_attrs(). 4574 * 4575 * Verify that the attributes are same as the expected ones. sargp->vap 4576 * and sargp->sbp contain the input attributes as translated from fattr4. 4577 * 4578 * This function verifies only the attrs that correspond to a vattr or 4579 * vfsstat struct. That is because of the extra step needed to get the 4580 * corresponding system structs. Other attributes have already been set or 4581 * verified by do_rfs4_set_attrs. 4582 * 4583 * Return 0 if all attrs match, -1 if some don't, error if error processing. 4584 */ 4585 static int 4586 rfs4_verify_attr(struct nfs4_svgetit_arg *sargp, 4587 bitmap4 *resp, struct nfs4_ntov_table *ntovp) 4588 { 4589 int error, ret_error = 0; 4590 int i, k; 4591 uint_t sva_mask = sargp->vap->va_mask; 4592 uint_t vbit; 4593 union nfs4_attr_u *na; 4594 uint8_t *amap; 4595 bool_t getsb = ntovp->vfsstat; 4596 4597 if (sva_mask != 0) { 4598 /* 4599 * Okay to overwrite sargp->vap because we verify based 4600 * on the incoming values. 4601 */ 4602 ret_error = VOP_GETATTR(sargp->cs->vp, sargp->vap, 0, 4603 sargp->cs->cr, NULL); 4604 if (ret_error) { 4605 if (resp == NULL) 4606 return (ret_error); 4607 /* 4608 * Must return bitmap of successful attrs 4609 */ 4610 sva_mask = 0; /* to prevent checking vap later */ 4611 } else { 4612 /* 4613 * Some file systems clobber va_mask. it is probably 4614 * wrong of them to do so, nonethless we practice 4615 * defensive coding. 4616 * See bug id 4276830. 4617 */ 4618 sargp->vap->va_mask = sva_mask; 4619 } 4620 } 4621 4622 if (getsb) { 4623 /* 4624 * Now get the superblock and loop on the bitmap, as there is 4625 * no simple way of translating from superblock to bitmap4. 4626 */ 4627 ret_error = VFS_STATVFS(sargp->cs->vp->v_vfsp, sargp->sbp); 4628 if (ret_error) { 4629 if (resp == NULL) 4630 goto errout; 4631 getsb = FALSE; 4632 } 4633 } 4634 4635 /* 4636 * Now loop and verify each attribute which getattr returned 4637 * whether it's the same as the input. 4638 */ 4639 if (resp == NULL && !getsb && (sva_mask == 0)) 4640 goto errout; 4641 4642 na = ntovp->na; 4643 amap = ntovp->amap; 4644 k = 0; 4645 for (i = 0; i < ntovp->attrcnt; i++, na++, amap++) { 4646 k = *amap; 4647 ASSERT(nfs4_ntov_map[k].nval == k); 4648 vbit = nfs4_ntov_map[k].vbit; 4649 4650 /* 4651 * If vattr attribute but VOP_GETATTR failed, or it's 4652 * superblock attribute but VFS_STATVFS failed, skip 4653 */ 4654 if (vbit) { 4655 if ((vbit & sva_mask) == 0) 4656 continue; 4657 } else if (!(getsb && nfs4_ntov_map[k].vfsstat)) { 4658 continue; 4659 } 4660 error = (*nfs4_ntov_map[k].sv_getit)(NFS4ATTR_VERIT, sargp, na); 4661 if (resp != NULL) { 4662 if (error) 4663 ret_error = -1; /* not all match */ 4664 else /* update response bitmap */ 4665 *resp |= nfs4_ntov_map[k].fbit; 4666 continue; 4667 } 4668 if (error) { 4669 ret_error = -1; /* not all match */ 4670 break; 4671 } 4672 } 4673 errout: 4674 return (ret_error); 4675 } 4676 4677 /* 4678 * Decode the attribute to be set/verified. If the attr requires a sys op 4679 * (VOP_GETATTR, VFS_VFSSTAT), and the request is to verify, then don't 4680 * call the sv_getit function for it, because the sys op hasn't yet been done. 4681 * Return 0 for success, error code if failed. 4682 * 4683 * Note: the decoded arg is not freed here but in nfs4_ntov_table_free. 4684 */ 4685 static int 4686 decode_fattr4_attr(nfs4_attr_cmd_t cmd, struct nfs4_svgetit_arg *sargp, 4687 int k, XDR *xdrp, bitmap4 *resp_bval, union nfs4_attr_u *nap) 4688 { 4689 int error = 0; 4690 bool_t set_later; 4691 4692 sargp->vap->va_mask |= nfs4_ntov_map[k].vbit; 4693 4694 if ((*nfs4_ntov_map[k].xfunc)(xdrp, nap)) { 4695 set_later = nfs4_ntov_map[k].vbit || nfs4_ntov_map[k].vfsstat; 4696 /* 4697 * don't verify yet if a vattr or sb dependent attr, 4698 * because we don't have their sys values yet. 4699 * Will be done later. 4700 */ 4701 if (! (set_later && (cmd == NFS4ATTR_VERIT))) { 4702 /* 4703 * ACLs are a special case, since setting the MODE 4704 * conflicts with setting the ACL. We delay setting 4705 * the ACL until all other attributes have been set. 4706 * The ACL gets set in do_rfs4_op_setattr(). 4707 */ 4708 if (nfs4_ntov_map[k].fbit != FATTR4_ACL_MASK) { 4709 error = (*nfs4_ntov_map[k].sv_getit)(cmd, 4710 sargp, nap); 4711 if (error) { 4712 xdr_free(nfs4_ntov_map[k].xfunc, 4713 (caddr_t)nap); 4714 } 4715 } 4716 } 4717 } else { 4718 #ifdef DEBUG 4719 cmn_err(CE_NOTE, "decode_fattr4_attr: error " 4720 "decoding attribute %d\n", k); 4721 #endif 4722 error = EINVAL; 4723 } 4724 if (!error && resp_bval && !set_later) { 4725 *resp_bval |= nfs4_ntov_map[k].fbit; 4726 } 4727 4728 return (error); 4729 } 4730 4731 /* 4732 * Set vattr based on incoming fattr4 attrs - used by setattr. 4733 * Set response mask. Ignore any values that are not writable vattr attrs. 4734 */ 4735 static nfsstat4 4736 do_rfs4_set_attrs(bitmap4 *resp, fattr4 *fattrp, struct compound_state *cs, 4737 struct nfs4_svgetit_arg *sargp, struct nfs4_ntov_table *ntovp, 4738 nfs4_attr_cmd_t cmd) 4739 { 4740 int error = 0; 4741 int i; 4742 char *attrs = fattrp->attrlist4; 4743 uint32_t attrslen = fattrp->attrlist4_len; 4744 XDR xdr; 4745 nfsstat4 status = NFS4_OK; 4746 vnode_t *vp = cs->vp; 4747 union nfs4_attr_u *na; 4748 uint8_t *amap; 4749 4750 #ifndef lint 4751 /* 4752 * Make sure that maximum attribute number can be expressed as an 4753 * 8 bit quantity. 4754 */ 4755 ASSERT(NFS4_MAXNUM_ATTRS <= (UINT8_MAX + 1)); 4756 #endif 4757 4758 if (vp == NULL) { 4759 if (resp) 4760 *resp = 0; 4761 return (NFS4ERR_NOFILEHANDLE); 4762 } 4763 if (cs->access == CS_ACCESS_DENIED) { 4764 if (resp) 4765 *resp = 0; 4766 return (NFS4ERR_ACCESS); 4767 } 4768 4769 sargp->op = cmd; 4770 sargp->cs = cs; 4771 sargp->flag = 0; /* may be set later */ 4772 sargp->vap->va_mask = 0; 4773 sargp->rdattr_error = NFS4_OK; 4774 sargp->rdattr_error_req = FALSE; 4775 /* sargp->sbp is set by the caller */ 4776 4777 xdrmem_create(&xdr, attrs, attrslen, XDR_DECODE); 4778 4779 na = ntovp->na; 4780 amap = ntovp->amap; 4781 4782 /* 4783 * The following loop iterates on the nfs4_ntov_map checking 4784 * if the fbit is set in the requested bitmap. 4785 * If set then we process the arguments using the 4786 * rfs4_fattr4 conversion functions to populate the setattr 4787 * vattr and va_mask. Any settable attrs that are not using vattr 4788 * will be set in this loop. 4789 */ 4790 for (i = 0; i < nfs4_ntov_map_size; i++) { 4791 if (!(fattrp->attrmask & nfs4_ntov_map[i].fbit)) { 4792 continue; 4793 } 4794 /* 4795 * If setattr, must be a writable attr. 4796 * If verify/nverify, must be a readable attr. 4797 */ 4798 if ((error = (*nfs4_ntov_map[i].sv_getit)( 4799 NFS4ATTR_SUPPORTED, sargp, NULL)) != 0) { 4800 /* 4801 * Client tries to set/verify an 4802 * unsupported attribute, tries to set 4803 * a read only attr or verify a write 4804 * only one - error! 4805 */ 4806 break; 4807 } 4808 /* 4809 * Decode the attribute to set/verify 4810 */ 4811 error = decode_fattr4_attr(cmd, sargp, nfs4_ntov_map[i].nval, 4812 &xdr, resp ? resp : NULL, na); 4813 if (error) 4814 break; 4815 *amap++ = (uint8_t)nfs4_ntov_map[i].nval; 4816 na++; 4817 (ntovp->attrcnt)++; 4818 if (nfs4_ntov_map[i].vfsstat) 4819 ntovp->vfsstat = TRUE; 4820 } 4821 4822 if (error != 0) 4823 status = (error == ENOTSUP ? NFS4ERR_ATTRNOTSUPP : 4824 puterrno4(error)); 4825 /* xdrmem_destroy(&xdrs); */ /* NO-OP */ 4826 return (status); 4827 } 4828 4829 static nfsstat4 4830 do_rfs4_op_setattr(bitmap4 *resp, fattr4 *fattrp, struct compound_state *cs, 4831 stateid4 *stateid) 4832 { 4833 int error = 0; 4834 struct nfs4_svgetit_arg sarg; 4835 bool_t trunc; 4836 4837 nfsstat4 status = NFS4_OK; 4838 cred_t *cr = cs->cr; 4839 vnode_t *vp = cs->vp; 4840 struct nfs4_ntov_table ntov; 4841 struct statvfs64 sb; 4842 struct vattr bva; 4843 struct flock64 bf; 4844 int in_crit = 0; 4845 uint_t saved_mask = 0; 4846 caller_context_t ct; 4847 4848 *resp = 0; 4849 sarg.sbp = &sb; 4850 nfs4_ntov_table_init(&ntov); 4851 status = do_rfs4_set_attrs(resp, fattrp, cs, &sarg, &ntov, 4852 NFS4ATTR_SETIT); 4853 if (status != NFS4_OK) { 4854 /* 4855 * failed set attrs 4856 */ 4857 goto done; 4858 } 4859 if ((sarg.vap->va_mask == 0) && 4860 (! (fattrp->attrmask & FATTR4_ACL_MASK))) { 4861 /* 4862 * no further work to be done 4863 */ 4864 goto done; 4865 } 4866 4867 /* 4868 * If we got a request to set the ACL and the MODE, only 4869 * allow changing VSUID, VSGID, and VSVTX. Attempting 4870 * to change any other bits, along with setting an ACL, 4871 * gives NFS4ERR_INVAL. 4872 */ 4873 if ((fattrp->attrmask & FATTR4_ACL_MASK) && 4874 (fattrp->attrmask & FATTR4_MODE_MASK)) { 4875 vattr_t va; 4876 4877 va.va_mask = AT_MODE; 4878 error = VOP_GETATTR(vp, &va, 0, cs->cr, NULL); 4879 if (error) { 4880 status = puterrno4(error); 4881 goto done; 4882 } 4883 if ((sarg.vap->va_mode ^ va.va_mode) & 4884 ~(VSUID | VSGID | VSVTX)) { 4885 status = NFS4ERR_INVAL; 4886 goto done; 4887 } 4888 } 4889 4890 /* Check stateid only if size has been set */ 4891 if (sarg.vap->va_mask & AT_SIZE) { 4892 trunc = (sarg.vap->va_size == 0); 4893 status = rfs4_check_stateid(FWRITE, cs->vp, stateid, 4894 trunc, &cs->deleg, sarg.vap->va_mask & AT_SIZE, &ct); 4895 if (status != NFS4_OK) 4896 goto done; 4897 } else { 4898 ct.cc_sysid = 0; 4899 ct.cc_pid = 0; 4900 ct.cc_caller_id = nfs4_srv_caller_id; 4901 ct.cc_flags = CC_DONTBLOCK; 4902 } 4903 4904 /* XXX start of possible race with delegations */ 4905 4906 /* 4907 * We need to specially handle size changes because it is 4908 * possible for the client to create a file with read-only 4909 * modes, but with the file opened for writing. If the client 4910 * then tries to set the file size, e.g. ftruncate(3C), 4911 * fcntl(F_FREESP), the normal access checking done in 4912 * VOP_SETATTR would prevent the client from doing it even though 4913 * it should be allowed to do so. To get around this, we do the 4914 * access checking for ourselves and use VOP_SPACE which doesn't 4915 * do the access checking. 4916 * Also the client should not be allowed to change the file 4917 * size if there is a conflicting non-blocking mandatory lock in 4918 * the region of the change. 4919 */ 4920 if (vp->v_type == VREG && (sarg.vap->va_mask & AT_SIZE)) { 4921 u_offset_t offset; 4922 ssize_t length; 4923 4924 /* 4925 * ufs_setattr clears AT_SIZE from vap->va_mask, but 4926 * before returning, sarg.vap->va_mask is used to 4927 * generate the setattr reply bitmap. We also clear 4928 * AT_SIZE below before calling VOP_SPACE. For both 4929 * of these cases, the va_mask needs to be saved here 4930 * and restored after calling VOP_SETATTR. 4931 */ 4932 saved_mask = sarg.vap->va_mask; 4933 4934 /* 4935 * Check any possible conflict due to NBMAND locks. 4936 * Get into critical region before VOP_GETATTR, so the 4937 * size attribute is valid when checking conflicts. 4938 */ 4939 if (nbl_need_check(vp)) { 4940 nbl_start_crit(vp, RW_READER); 4941 in_crit = 1; 4942 } 4943 4944 bva.va_mask = AT_UID|AT_SIZE; 4945 if (error = VOP_GETATTR(vp, &bva, 0, cr, &ct)) { 4946 status = puterrno4(error); 4947 goto done; 4948 } 4949 4950 if (in_crit) { 4951 if (sarg.vap->va_size < bva.va_size) { 4952 offset = sarg.vap->va_size; 4953 length = bva.va_size - sarg.vap->va_size; 4954 } else { 4955 offset = bva.va_size; 4956 length = sarg.vap->va_size - bva.va_size; 4957 } 4958 if (nbl_conflict(vp, NBL_WRITE, offset, length, 0, 4959 &ct)) { 4960 status = NFS4ERR_LOCKED; 4961 goto done; 4962 } 4963 } 4964 4965 if (crgetuid(cr) == bva.va_uid) { 4966 sarg.vap->va_mask &= ~AT_SIZE; 4967 bf.l_type = F_WRLCK; 4968 bf.l_whence = 0; 4969 bf.l_start = (off64_t)sarg.vap->va_size; 4970 bf.l_len = 0; 4971 bf.l_sysid = 0; 4972 bf.l_pid = 0; 4973 error = VOP_SPACE(vp, F_FREESP, &bf, FWRITE, 4974 (offset_t)sarg.vap->va_size, cr, &ct); 4975 } 4976 } 4977 4978 if (!error && sarg.vap->va_mask != 0) 4979 error = VOP_SETATTR(vp, sarg.vap, sarg.flag, cr, &ct); 4980 4981 /* restore va_mask -- ufs_setattr clears AT_SIZE */ 4982 if (saved_mask & AT_SIZE) 4983 sarg.vap->va_mask |= AT_SIZE; 4984 4985 /* 4986 * If an ACL was being set, it has been delayed until now, 4987 * in order to set the mode (via the VOP_SETATTR() above) first. 4988 */ 4989 if ((! error) && (fattrp->attrmask & FATTR4_ACL_MASK)) { 4990 int i; 4991 4992 for (i = 0; i < NFS4_MAXNUM_ATTRS; i++) 4993 if (ntov.amap[i] == FATTR4_ACL) 4994 break; 4995 if (i < NFS4_MAXNUM_ATTRS) { 4996 error = (*nfs4_ntov_map[FATTR4_ACL].sv_getit)( 4997 NFS4ATTR_SETIT, &sarg, &ntov.na[i]); 4998 if (error == 0) { 4999 *resp |= FATTR4_ACL_MASK; 5000 } else if (error == ENOTSUP) { 5001 (void) rfs4_verify_attr(&sarg, resp, &ntov); 5002 status = NFS4ERR_ATTRNOTSUPP; 5003 goto done; 5004 } 5005 } else { 5006 NFS4_DEBUG(rfs4_debug, 5007 (CE_NOTE, "do_rfs4_op_setattr: " 5008 "unable to find ACL in fattr4")); 5009 error = EINVAL; 5010 } 5011 } 5012 5013 if (error) { 5014 /* check if a monitor detected a delegation conflict */ 5015 if (error == EAGAIN && (ct.cc_flags & CC_WOULDBLOCK)) 5016 status = NFS4ERR_DELAY; 5017 else 5018 status = puterrno4(error); 5019 5020 /* 5021 * Set the response bitmap when setattr failed. 5022 * If VOP_SETATTR partially succeeded, test by doing a 5023 * VOP_GETATTR on the object and comparing the data 5024 * to the setattr arguments. 5025 */ 5026 (void) rfs4_verify_attr(&sarg, resp, &ntov); 5027 } else { 5028 /* 5029 * Force modified metadata out to stable storage. 5030 */ 5031 (void) VOP_FSYNC(vp, FNODSYNC, cr, &ct); 5032 /* 5033 * Set response bitmap 5034 */ 5035 nfs4_vmask_to_nmask_set(sarg.vap->va_mask, resp); 5036 } 5037 5038 /* Return early and already have a NFSv4 error */ 5039 done: 5040 /* 5041 * Except for nfs4_vmask_to_nmask_set(), vattr --> fattr 5042 * conversion sets both readable and writeable NFS4 attrs 5043 * for AT_MTIME and AT_ATIME. The line below masks out 5044 * unrequested attrs from the setattr result bitmap. This 5045 * is placed after the done: label to catch the ATTRNOTSUP 5046 * case. 5047 */ 5048 *resp &= fattrp->attrmask; 5049 5050 if (in_crit) 5051 nbl_end_crit(vp); 5052 5053 nfs4_ntov_table_free(&ntov, &sarg); 5054 5055 return (status); 5056 } 5057 5058 /* ARGSUSED */ 5059 static void 5060 rfs4_op_setattr(nfs_argop4 *argop, nfs_resop4 *resop, struct svc_req *req, 5061 struct compound_state *cs) 5062 { 5063 SETATTR4args *args = &argop->nfs_argop4_u.opsetattr; 5064 SETATTR4res *resp = &resop->nfs_resop4_u.opsetattr; 5065 bslabel_t *clabel; 5066 5067 DTRACE_NFSV4_2(op__setattr__start, struct compound_state *, cs, 5068 SETATTR4args *, args); 5069 5070 if (cs->vp == NULL) { 5071 *cs->statusp = resp->status = NFS4ERR_NOFILEHANDLE; 5072 goto out; 5073 } 5074 5075 /* 5076 * If there is an unshared filesystem mounted on this vnode, 5077 * do not allow to setattr on this vnode. 5078 */ 5079 if (vn_ismntpt(cs->vp)) { 5080 *cs->statusp = resp->status = NFS4ERR_ACCESS; 5081 goto out; 5082 } 5083 5084 resp->attrsset = 0; 5085 5086 if (rdonly4(cs->exi, cs->vp, req)) { 5087 *cs->statusp = resp->status = NFS4ERR_ROFS; 5088 goto out; 5089 } 5090 5091 /* check label before setting attributes */ 5092 if (is_system_labeled()) { 5093 ASSERT(req->rq_label != NULL); 5094 clabel = req->rq_label; 5095 DTRACE_PROBE2(tx__rfs4__log__info__opsetattr__clabel, char *, 5096 "got client label from request(1)", 5097 struct svc_req *, req); 5098 if (!blequal(&l_admin_low->tsl_label, clabel)) { 5099 if (!do_rfs_label_check(clabel, cs->vp, 5100 EQUALITY_CHECK)) { 5101 *cs->statusp = resp->status = NFS4ERR_ACCESS; 5102 goto out; 5103 } 5104 } 5105 } 5106 5107 *cs->statusp = resp->status = 5108 do_rfs4_op_setattr(&resp->attrsset, &args->obj_attributes, cs, 5109 &args->stateid); 5110 5111 out: 5112 DTRACE_NFSV4_2(op__setattr__done, struct compound_state *, cs, 5113 SETATTR4res *, resp); 5114 } 5115 5116 /* ARGSUSED */ 5117 static void 5118 rfs4_op_verify(nfs_argop4 *argop, nfs_resop4 *resop, struct svc_req *req, 5119 struct compound_state *cs) 5120 { 5121 /* 5122 * verify and nverify are exactly the same, except that nverify 5123 * succeeds when some argument changed, and verify succeeds when 5124 * when none changed. 5125 */ 5126 5127 VERIFY4args *args = &argop->nfs_argop4_u.opverify; 5128 VERIFY4res *resp = &resop->nfs_resop4_u.opverify; 5129 5130 int error; 5131 struct nfs4_svgetit_arg sarg; 5132 struct statvfs64 sb; 5133 struct nfs4_ntov_table ntov; 5134 5135 DTRACE_NFSV4_2(op__verify__start, struct compound_state *, cs, 5136 VERIFY4args *, args); 5137 5138 if (cs->vp == NULL) { 5139 *cs->statusp = resp->status = NFS4ERR_NOFILEHANDLE; 5140 goto out; 5141 } 5142 5143 sarg.sbp = &sb; 5144 nfs4_ntov_table_init(&ntov); 5145 resp->status = do_rfs4_set_attrs(NULL, &args->obj_attributes, cs, 5146 &sarg, &ntov, NFS4ATTR_VERIT); 5147 if (resp->status != NFS4_OK) { 5148 /* 5149 * do_rfs4_set_attrs will try to verify systemwide attrs, 5150 * so could return -1 for "no match". 5151 */ 5152 if (resp->status == -1) 5153 resp->status = NFS4ERR_NOT_SAME; 5154 goto done; 5155 } 5156 error = rfs4_verify_attr(&sarg, NULL, &ntov); 5157 switch (error) { 5158 case 0: 5159 resp->status = NFS4_OK; 5160 break; 5161 case -1: 5162 resp->status = NFS4ERR_NOT_SAME; 5163 break; 5164 default: 5165 resp->status = puterrno4(error); 5166 break; 5167 } 5168 done: 5169 *cs->statusp = resp->status; 5170 nfs4_ntov_table_free(&ntov, &sarg); 5171 out: 5172 DTRACE_NFSV4_2(op__verify__done, struct compound_state *, cs, 5173 VERIFY4res *, resp); 5174 } 5175 5176 /* ARGSUSED */ 5177 static void 5178 rfs4_op_nverify(nfs_argop4 *argop, nfs_resop4 *resop, struct svc_req *req, 5179 struct compound_state *cs) 5180 { 5181 /* 5182 * verify and nverify are exactly the same, except that nverify 5183 * succeeds when some argument changed, and verify succeeds when 5184 * when none changed. 5185 */ 5186 5187 NVERIFY4args *args = &argop->nfs_argop4_u.opnverify; 5188 NVERIFY4res *resp = &resop->nfs_resop4_u.opnverify; 5189 5190 int error; 5191 struct nfs4_svgetit_arg sarg; 5192 struct statvfs64 sb; 5193 struct nfs4_ntov_table ntov; 5194 5195 DTRACE_NFSV4_2(op__nverify__start, struct compound_state *, cs, 5196 NVERIFY4args *, args); 5197 5198 if (cs->vp == NULL) { 5199 *cs->statusp = resp->status = NFS4ERR_NOFILEHANDLE; 5200 DTRACE_NFSV4_2(op__nverify__done, struct compound_state *, cs, 5201 NVERIFY4res *, resp); 5202 return; 5203 } 5204 sarg.sbp = &sb; 5205 nfs4_ntov_table_init(&ntov); 5206 resp->status = do_rfs4_set_attrs(NULL, &args->obj_attributes, cs, 5207 &sarg, &ntov, NFS4ATTR_VERIT); 5208 if (resp->status != NFS4_OK) { 5209 /* 5210 * do_rfs4_set_attrs will try to verify systemwide attrs, 5211 * so could return -1 for "no match". 5212 */ 5213 if (resp->status == -1) 5214 resp->status = NFS4_OK; 5215 goto done; 5216 } 5217 error = rfs4_verify_attr(&sarg, NULL, &ntov); 5218 switch (error) { 5219 case 0: 5220 resp->status = NFS4ERR_SAME; 5221 break; 5222 case -1: 5223 resp->status = NFS4_OK; 5224 break; 5225 default: 5226 resp->status = puterrno4(error); 5227 break; 5228 } 5229 done: 5230 *cs->statusp = resp->status; 5231 nfs4_ntov_table_free(&ntov, &sarg); 5232 5233 DTRACE_NFSV4_2(op__nverify__done, struct compound_state *, cs, 5234 NVERIFY4res *, resp); 5235 } 5236 5237 /* 5238 * XXX - This should live in an NFS header file. 5239 */ 5240 #define MAX_IOVECS 12 5241 5242 /* ARGSUSED */ 5243 static void 5244 rfs4_op_write(nfs_argop4 *argop, nfs_resop4 *resop, struct svc_req *req, 5245 struct compound_state *cs) 5246 { 5247 WRITE4args *args = &argop->nfs_argop4_u.opwrite; 5248 WRITE4res *resp = &resop->nfs_resop4_u.opwrite; 5249 int error; 5250 vnode_t *vp; 5251 struct vattr bva; 5252 u_offset_t rlimit; 5253 struct uio uio; 5254 struct iovec iov[MAX_IOVECS]; 5255 struct iovec *iovp; 5256 int iovcnt; 5257 int ioflag; 5258 cred_t *savecred, *cr; 5259 bool_t *deleg = &cs->deleg; 5260 nfsstat4 stat; 5261 int in_crit = 0; 5262 caller_context_t ct; 5263 5264 DTRACE_NFSV4_2(op__write__start, struct compound_state *, cs, 5265 WRITE4args *, args); 5266 5267 vp = cs->vp; 5268 if (vp == NULL) { 5269 *cs->statusp = resp->status = NFS4ERR_NOFILEHANDLE; 5270 goto out; 5271 } 5272 if (cs->access == CS_ACCESS_DENIED) { 5273 *cs->statusp = resp->status = NFS4ERR_ACCESS; 5274 goto out; 5275 } 5276 5277 cr = cs->cr; 5278 5279 if ((stat = rfs4_check_stateid(FWRITE, vp, &args->stateid, FALSE, 5280 deleg, TRUE, &ct)) != NFS4_OK) { 5281 *cs->statusp = resp->status = stat; 5282 goto out; 5283 } 5284 5285 /* 5286 * We have to enter the critical region before calling VOP_RWLOCK 5287 * to avoid a deadlock with ufs. 5288 */ 5289 if (nbl_need_check(vp)) { 5290 nbl_start_crit(vp, RW_READER); 5291 in_crit = 1; 5292 if (nbl_conflict(vp, NBL_WRITE, 5293 args->offset, args->data_len, 0, &ct)) { 5294 *cs->statusp = resp->status = NFS4ERR_LOCKED; 5295 goto out; 5296 } 5297 } 5298 5299 bva.va_mask = AT_MODE | AT_UID; 5300 error = VOP_GETATTR(vp, &bva, 0, cr, &ct); 5301 5302 /* 5303 * If we can't get the attributes, then we can't do the 5304 * right access checking. So, we'll fail the request. 5305 */ 5306 if (error) { 5307 *cs->statusp = resp->status = puterrno4(error); 5308 goto out; 5309 } 5310 5311 if (rdonly4(cs->exi, cs->vp, req)) { 5312 *cs->statusp = resp->status = NFS4ERR_ROFS; 5313 goto out; 5314 } 5315 5316 if (vp->v_type != VREG) { 5317 *cs->statusp = resp->status = 5318 ((vp->v_type == VDIR) ? NFS4ERR_ISDIR : NFS4ERR_INVAL); 5319 goto out; 5320 } 5321 5322 if (crgetuid(cr) != bva.va_uid && 5323 (error = VOP_ACCESS(vp, VWRITE, 0, cr, &ct))) { 5324 *cs->statusp = resp->status = puterrno4(error); 5325 goto out; 5326 } 5327 5328 if (MANDLOCK(vp, bva.va_mode)) { 5329 *cs->statusp = resp->status = NFS4ERR_ACCESS; 5330 goto out; 5331 } 5332 5333 if (args->data_len == 0) { 5334 *cs->statusp = resp->status = NFS4_OK; 5335 resp->count = 0; 5336 resp->committed = args->stable; 5337 resp->writeverf = Write4verf; 5338 goto out; 5339 } 5340 5341 if (args->mblk != NULL) { 5342 mblk_t *m; 5343 uint_t bytes, round_len; 5344 5345 iovcnt = 0; 5346 bytes = 0; 5347 round_len = roundup(args->data_len, BYTES_PER_XDR_UNIT); 5348 for (m = args->mblk; 5349 m != NULL && bytes < round_len; 5350 m = m->b_cont) { 5351 iovcnt++; 5352 bytes += MBLKL(m); 5353 } 5354 #ifdef DEBUG 5355 /* should have ended on an mblk boundary */ 5356 if (bytes != round_len) { 5357 printf("bytes=0x%x, round_len=0x%x, req len=0x%x\n", 5358 bytes, round_len, args->data_len); 5359 printf("args=%p, args->mblk=%p, m=%p", (void *)args, 5360 (void *)args->mblk, (void *)m); 5361 ASSERT(bytes == round_len); 5362 } 5363 #endif 5364 if (iovcnt <= MAX_IOVECS) { 5365 iovp = iov; 5366 } else { 5367 iovp = kmem_alloc(sizeof (*iovp) * iovcnt, KM_SLEEP); 5368 } 5369 mblk_to_iov(args->mblk, iovcnt, iovp); 5370 } else { 5371 iovcnt = 1; 5372 iovp = iov; 5373 iovp->iov_base = args->data_val; 5374 iovp->iov_len = args->data_len; 5375 } 5376 5377 uio.uio_iov = iovp; 5378 uio.uio_iovcnt = iovcnt; 5379 5380 uio.uio_segflg = UIO_SYSSPACE; 5381 uio.uio_extflg = UIO_COPY_DEFAULT; 5382 uio.uio_loffset = args->offset; 5383 uio.uio_resid = args->data_len; 5384 uio.uio_llimit = curproc->p_fsz_ctl; 5385 rlimit = uio.uio_llimit - args->offset; 5386 if (rlimit < (u_offset_t)uio.uio_resid) 5387 uio.uio_resid = (int)rlimit; 5388 5389 if (args->stable == UNSTABLE4) 5390 ioflag = 0; 5391 else if (args->stable == FILE_SYNC4) 5392 ioflag = FSYNC; 5393 else if (args->stable == DATA_SYNC4) 5394 ioflag = FDSYNC; 5395 else { 5396 if (iovp != iov) 5397 kmem_free(iovp, sizeof (*iovp) * iovcnt); 5398 *cs->statusp = resp->status = NFS4ERR_INVAL; 5399 goto out; 5400 } 5401 5402 /* 5403 * We're changing creds because VM may fault and we need 5404 * the cred of the current thread to be used if quota 5405 * checking is enabled. 5406 */ 5407 savecred = curthread->t_cred; 5408 curthread->t_cred = cr; 5409 error = do_io(FWRITE, vp, &uio, ioflag, cr, &ct); 5410 curthread->t_cred = savecred; 5411 5412 if (iovp != iov) 5413 kmem_free(iovp, sizeof (*iovp) * iovcnt); 5414 5415 if (error) { 5416 *cs->statusp = resp->status = puterrno4(error); 5417 goto out; 5418 } 5419 5420 *cs->statusp = resp->status = NFS4_OK; 5421 resp->count = args->data_len - uio.uio_resid; 5422 5423 if (ioflag == 0) 5424 resp->committed = UNSTABLE4; 5425 else 5426 resp->committed = FILE_SYNC4; 5427 5428 resp->writeverf = Write4verf; 5429 5430 out: 5431 if (in_crit) 5432 nbl_end_crit(vp); 5433 5434 DTRACE_NFSV4_2(op__write__done, struct compound_state *, cs, 5435 WRITE4res *, resp); 5436 } 5437 5438 5439 /* XXX put in a header file */ 5440 extern int sec_svc_getcred(struct svc_req *, cred_t *, caddr_t *, int *); 5441 5442 void 5443 rfs4_compound(COMPOUND4args *args, COMPOUND4res *resp, struct exportinfo *exi, 5444 struct svc_req *req, cred_t *cr, int *rv) 5445 { 5446 uint_t i; 5447 struct compound_state cs; 5448 5449 if (rv != NULL) 5450 *rv = 0; 5451 rfs4_init_compound_state(&cs); 5452 /* 5453 * Form a reply tag by copying over the reqeuest tag. 5454 */ 5455 resp->tag.utf8string_val = 5456 kmem_alloc(args->tag.utf8string_len, KM_SLEEP); 5457 resp->tag.utf8string_len = args->tag.utf8string_len; 5458 bcopy(args->tag.utf8string_val, resp->tag.utf8string_val, 5459 resp->tag.utf8string_len); 5460 5461 cs.statusp = &resp->status; 5462 cs.req = req; 5463 5464 /* 5465 * XXX for now, minorversion should be zero 5466 */ 5467 if (args->minorversion != NFS4_MINORVERSION) { 5468 DTRACE_NFSV4_2(compound__start, struct compound_state *, 5469 &cs, COMPOUND4args *, args); 5470 resp->array_len = 0; 5471 resp->array = NULL; 5472 resp->status = NFS4ERR_MINOR_VERS_MISMATCH; 5473 DTRACE_NFSV4_2(compound__done, struct compound_state *, 5474 &cs, COMPOUND4res *, resp); 5475 return; 5476 } 5477 5478 ASSERT(exi == NULL); 5479 ASSERT(cr == NULL); 5480 5481 cr = crget(); 5482 ASSERT(cr != NULL); 5483 5484 if (sec_svc_getcred(req, cr, &cs.principal, &cs.nfsflavor) == 0) { 5485 DTRACE_NFSV4_2(compound__start, struct compound_state *, 5486 &cs, COMPOUND4args *, args); 5487 crfree(cr); 5488 DTRACE_NFSV4_2(compound__done, struct compound_state *, 5489 &cs, COMPOUND4res *, resp); 5490 svcerr_badcred(req->rq_xprt); 5491 if (rv != NULL) 5492 *rv = 1; 5493 return; 5494 } 5495 resp->array_len = args->array_len; 5496 resp->array = kmem_zalloc(args->array_len * sizeof (nfs_resop4), 5497 KM_SLEEP); 5498 5499 cs.basecr = cr; 5500 5501 DTRACE_NFSV4_2(compound__start, struct compound_state *, &cs, 5502 COMPOUND4args *, args); 5503 5504 /* 5505 * For now, NFS4 compound processing must be protected by 5506 * exported_lock because it can access more than one exportinfo 5507 * per compound and share/unshare can now change multiple 5508 * exinfo structs. The NFS2/3 code only refs 1 exportinfo 5509 * per proc (excluding public exinfo), and exi_count design 5510 * is sufficient to protect concurrent execution of NFS2/3 5511 * ops along with unexport. This lock will be removed as 5512 * part of the NFSv4 phase 2 namespace redesign work. 5513 */ 5514 rw_enter(&exported_lock, RW_READER); 5515 5516 /* 5517 * If this is the first compound we've seen, we need to start all 5518 * new instances' grace periods. 5519 */ 5520 if (rfs4_seen_first_compound == 0) { 5521 rfs4_grace_start_new(); 5522 /* 5523 * This must be set after rfs4_grace_start_new(), otherwise 5524 * another thread could proceed past here before the former 5525 * is finished. 5526 */ 5527 rfs4_seen_first_compound = 1; 5528 } 5529 5530 for (i = 0; i < args->array_len && cs.cont; i++) { 5531 nfs_argop4 *argop; 5532 nfs_resop4 *resop; 5533 uint_t op; 5534 5535 argop = &args->array[i]; 5536 resop = &resp->array[i]; 5537 resop->resop = argop->argop; 5538 op = (uint_t)resop->resop; 5539 5540 if (op < rfsv4disp_cnt) { 5541 /* 5542 * Count the individual ops here; NULL and COMPOUND 5543 * are counted in common_dispatch() 5544 */ 5545 rfsproccnt_v4_ptr[op].value.ui64++; 5546 5547 NFS4_DEBUG(rfs4_debug > 1, 5548 (CE_NOTE, "Executing %s", rfs4_op_string[op])); 5549 (*rfsv4disptab[op].dis_proc)(argop, resop, req, &cs); 5550 NFS4_DEBUG(rfs4_debug > 1, (CE_NOTE, "%s returned %d", 5551 rfs4_op_string[op], *cs.statusp)); 5552 if (*cs.statusp != NFS4_OK) 5553 cs.cont = FALSE; 5554 } else { 5555 /* 5556 * This is effectively dead code since XDR code 5557 * will have already returned BADXDR if op doesn't 5558 * decode to legal value. This only done for a 5559 * day when XDR code doesn't verify v4 opcodes. 5560 */ 5561 op = OP_ILLEGAL; 5562 rfsproccnt_v4_ptr[OP_ILLEGAL_IDX].value.ui64++; 5563 5564 rfs4_op_illegal(argop, resop, req, &cs); 5565 cs.cont = FALSE; 5566 } 5567 5568 /* 5569 * If not at last op, and if we are to stop, then 5570 * compact the results array. 5571 */ 5572 if ((i + 1) < args->array_len && !cs.cont) { 5573 nfs_resop4 *new_res = kmem_alloc( 5574 (i+1) * sizeof (nfs_resop4), KM_SLEEP); 5575 bcopy(resp->array, 5576 new_res, (i+1) * sizeof (nfs_resop4)); 5577 kmem_free(resp->array, 5578 args->array_len * sizeof (nfs_resop4)); 5579 5580 resp->array_len = i + 1; 5581 resp->array = new_res; 5582 } 5583 } 5584 5585 rw_exit(&exported_lock); 5586 5587 DTRACE_NFSV4_2(compound__done, struct compound_state *, &cs, 5588 COMPOUND4res *, resp); 5589 5590 if (cs.vp) 5591 VN_RELE(cs.vp); 5592 if (cs.saved_vp) 5593 VN_RELE(cs.saved_vp); 5594 if (cs.saved_fh.nfs_fh4_val) 5595 kmem_free(cs.saved_fh.nfs_fh4_val, NFS4_FHSIZE); 5596 5597 if (cs.basecr) 5598 crfree(cs.basecr); 5599 if (cs.cr) 5600 crfree(cs.cr); 5601 /* 5602 * done with this compound request, free the label 5603 */ 5604 5605 if (req->rq_label != NULL) { 5606 kmem_free(req->rq_label, sizeof (bslabel_t)); 5607 req->rq_label = NULL; 5608 } 5609 } 5610 5611 /* 5612 * XXX because of what appears to be duplicate calls to rfs4_compound_free 5613 * XXX zero out the tag and array values. Need to investigate why the 5614 * XXX calls occur, but at least prevent the panic for now. 5615 */ 5616 void 5617 rfs4_compound_free(COMPOUND4res *resp) 5618 { 5619 uint_t i; 5620 5621 if (resp->tag.utf8string_val) { 5622 UTF8STRING_FREE(resp->tag) 5623 } 5624 5625 for (i = 0; i < resp->array_len; i++) { 5626 nfs_resop4 *resop; 5627 uint_t op; 5628 5629 resop = &resp->array[i]; 5630 op = (uint_t)resop->resop; 5631 if (op < rfsv4disp_cnt) { 5632 (*rfsv4disptab[op].dis_resfree)(resop); 5633 } 5634 } 5635 if (resp->array != NULL) { 5636 kmem_free(resp->array, resp->array_len * sizeof (nfs_resop4)); 5637 } 5638 } 5639 5640 /* 5641 * Process the value of the compound request rpc flags, as a bit-AND 5642 * of the individual per-op flags (idempotent, allowork, publicfh_ok) 5643 */ 5644 void 5645 rfs4_compound_flagproc(COMPOUND4args *args, int *flagp) 5646 { 5647 int i; 5648 int flag = RPC_ALL; 5649 5650 for (i = 0; flag && i < args->array_len; i++) { 5651 uint_t op; 5652 5653 op = (uint_t)args->array[i].argop; 5654 5655 if (op < rfsv4disp_cnt) 5656 flag &= rfsv4disptab[op].dis_flags; 5657 else 5658 flag = 0; 5659 } 5660 *flagp = flag; 5661 } 5662 5663 nfsstat4 5664 rfs4_client_sysid(rfs4_client_t *cp, sysid_t *sp) 5665 { 5666 nfsstat4 e; 5667 5668 rfs4_dbe_lock(cp->dbe); 5669 5670 if (cp->sysidt != LM_NOSYSID) { 5671 *sp = cp->sysidt; 5672 e = NFS4_OK; 5673 5674 } else if ((cp->sysidt = lm_alloc_sysidt()) != LM_NOSYSID) { 5675 *sp = cp->sysidt; 5676 e = NFS4_OK; 5677 5678 NFS4_DEBUG(rfs4_debug, (CE_NOTE, 5679 "rfs4_client_sysid: allocated 0x%x\n", *sp)); 5680 } else 5681 e = NFS4ERR_DELAY; 5682 5683 rfs4_dbe_unlock(cp->dbe); 5684 return (e); 5685 } 5686 5687 #if defined(DEBUG) && ! defined(lint) 5688 static void lock_print(char *str, int operation, struct flock64 *flk) 5689 { 5690 char *op, *type; 5691 5692 switch (operation) { 5693 case F_GETLK: op = "F_GETLK"; 5694 break; 5695 case F_SETLK: op = "F_SETLK"; 5696 break; 5697 case F_SETLK_NBMAND: op = "F_SETLK_NBMAND"; 5698 break; 5699 default: op = "F_UNKNOWN"; 5700 break; 5701 } 5702 switch (flk->l_type) { 5703 case F_UNLCK: type = "F_UNLCK"; 5704 break; 5705 case F_RDLCK: type = "F_RDLCK"; 5706 break; 5707 case F_WRLCK: type = "F_WRLCK"; 5708 break; 5709 default: type = "F_UNKNOWN"; 5710 break; 5711 } 5712 5713 ASSERT(flk->l_whence == 0); 5714 cmn_err(CE_NOTE, "%s: %s, type = %s, off = %llx len = %llx pid = %d", 5715 str, op, type, (longlong_t)flk->l_start, 5716 flk->l_len ? (longlong_t)flk->l_len : ~0LL, flk->l_pid); 5717 } 5718 5719 #define LOCK_PRINT(d, s, t, f) if (d) lock_print(s, t, f) 5720 #else 5721 #define LOCK_PRINT(d, s, t, f) 5722 #endif 5723 5724 /*ARGSUSED*/ 5725 static bool_t 5726 creds_ok(cred_set_t cr_set, struct svc_req *req, struct compound_state *cs) 5727 { 5728 return (TRUE); 5729 } 5730 5731 /* 5732 * Look up the pathname using the vp in cs as the directory vnode. 5733 * cs->vp will be the vnode for the file on success 5734 */ 5735 5736 static nfsstat4 5737 rfs4_lookup(component4 *component, struct svc_req *req, 5738 struct compound_state *cs) 5739 { 5740 char *nm; 5741 uint32_t len; 5742 nfsstat4 status; 5743 5744 if (cs->vp == NULL) { 5745 return (NFS4ERR_NOFILEHANDLE); 5746 } 5747 if (cs->vp->v_type != VDIR) { 5748 return (NFS4ERR_NOTDIR); 5749 } 5750 5751 if (!utf8_dir_verify(component)) 5752 return (NFS4ERR_INVAL); 5753 5754 nm = utf8_to_fn(component, &len, NULL); 5755 if (nm == NULL) { 5756 return (NFS4ERR_INVAL); 5757 } 5758 5759 if (len > MAXNAMELEN) { 5760 kmem_free(nm, len); 5761 return (NFS4ERR_NAMETOOLONG); 5762 } 5763 5764 status = do_rfs4_op_lookup(nm, len, req, cs); 5765 5766 kmem_free(nm, len); 5767 5768 return (status); 5769 } 5770 5771 static nfsstat4 5772 rfs4_lookupfile(component4 *component, struct svc_req *req, 5773 struct compound_state *cs, uint32_t access, 5774 change_info4 *cinfo) 5775 { 5776 nfsstat4 status; 5777 vnode_t *dvp = cs->vp; 5778 vattr_t bva, ava, fva; 5779 int error; 5780 5781 /* Get "before" change value */ 5782 bva.va_mask = AT_CTIME|AT_SEQ; 5783 error = VOP_GETATTR(dvp, &bva, 0, cs->cr, NULL); 5784 if (error) 5785 return (puterrno4(error)); 5786 5787 /* rfs4_lookup may VN_RELE directory */ 5788 VN_HOLD(dvp); 5789 5790 status = rfs4_lookup(component, req, cs); 5791 if (status != NFS4_OK) { 5792 VN_RELE(dvp); 5793 return (status); 5794 } 5795 5796 /* 5797 * Get "after" change value, if it fails, simply return the 5798 * before value. 5799 */ 5800 ava.va_mask = AT_CTIME|AT_SEQ; 5801 if (VOP_GETATTR(dvp, &ava, 0, cs->cr, NULL)) { 5802 ava.va_ctime = bva.va_ctime; 5803 ava.va_seq = 0; 5804 } 5805 VN_RELE(dvp); 5806 5807 /* 5808 * Validate the file is a file 5809 */ 5810 fva.va_mask = AT_TYPE|AT_MODE; 5811 error = VOP_GETATTR(cs->vp, &fva, 0, cs->cr, NULL); 5812 if (error) 5813 return (puterrno4(error)); 5814 5815 if (fva.va_type != VREG) { 5816 if (fva.va_type == VDIR) 5817 return (NFS4ERR_ISDIR); 5818 if (fva.va_type == VLNK) 5819 return (NFS4ERR_SYMLINK); 5820 return (NFS4ERR_INVAL); 5821 } 5822 5823 NFS4_SET_FATTR4_CHANGE(cinfo->before, bva.va_ctime); 5824 NFS4_SET_FATTR4_CHANGE(cinfo->after, ava.va_ctime); 5825 5826 /* 5827 * It is undefined if VOP_LOOKUP will change va_seq, so 5828 * cinfo.atomic = TRUE only if we have 5829 * non-zero va_seq's, and they have not changed. 5830 */ 5831 if (bva.va_seq && ava.va_seq && ava.va_seq == bva.va_seq) 5832 cinfo->atomic = TRUE; 5833 else 5834 cinfo->atomic = FALSE; 5835 5836 /* Check for mandatory locking */ 5837 cs->mandlock = MANDLOCK(cs->vp, fva.va_mode); 5838 return (check_open_access(access, cs, req)); 5839 } 5840 5841 static nfsstat4 5842 create_vnode(vnode_t *dvp, char *nm, vattr_t *vap, createmode4 mode, 5843 timespec32_t *mtime, cred_t *cr, vnode_t **vpp, bool_t *created) 5844 { 5845 int error; 5846 nfsstat4 status = NFS4_OK; 5847 vattr_t va; 5848 5849 tryagain: 5850 5851 /* 5852 * The file open mode used is VWRITE. If the client needs 5853 * some other semantic, then it should do the access checking 5854 * itself. It would have been nice to have the file open mode 5855 * passed as part of the arguments. 5856 */ 5857 5858 *created = TRUE; 5859 error = VOP_CREATE(dvp, nm, vap, EXCL, VWRITE, vpp, cr, 0, NULL, NULL); 5860 5861 if (error) { 5862 *created = FALSE; 5863 5864 /* 5865 * If we got something other than file already exists 5866 * then just return this error. Otherwise, we got 5867 * EEXIST. If we were doing a GUARDED create, then 5868 * just return this error. Otherwise, we need to 5869 * make sure that this wasn't a duplicate of an 5870 * exclusive create request. 5871 * 5872 * The assumption is made that a non-exclusive create 5873 * request will never return EEXIST. 5874 */ 5875 5876 if (error != EEXIST || mode == GUARDED4) { 5877 status = puterrno4(error); 5878 return (status); 5879 } 5880 error = VOP_LOOKUP(dvp, nm, vpp, NULL, 0, NULL, cr, 5881 NULL, NULL, NULL); 5882 5883 if (error) { 5884 /* 5885 * We couldn't find the file that we thought that 5886 * we just created. So, we'll just try creating 5887 * it again. 5888 */ 5889 if (error == ENOENT) 5890 goto tryagain; 5891 5892 status = puterrno4(error); 5893 return (status); 5894 } 5895 5896 if (mode == UNCHECKED4) { 5897 /* existing object must be regular file */ 5898 if ((*vpp)->v_type != VREG) { 5899 if ((*vpp)->v_type == VDIR) 5900 status = NFS4ERR_ISDIR; 5901 else if ((*vpp)->v_type == VLNK) 5902 status = NFS4ERR_SYMLINK; 5903 else 5904 status = NFS4ERR_INVAL; 5905 VN_RELE(*vpp); 5906 return (status); 5907 } 5908 5909 return (NFS4_OK); 5910 } 5911 5912 /* Check for duplicate request */ 5913 ASSERT(mtime != 0); 5914 va.va_mask = AT_MTIME; 5915 error = VOP_GETATTR(*vpp, &va, 0, cr, NULL); 5916 if (!error) { 5917 /* We found the file */ 5918 if (va.va_mtime.tv_sec != mtime->tv_sec || 5919 va.va_mtime.tv_nsec != mtime->tv_nsec) { 5920 /* but its not our creation */ 5921 VN_RELE(*vpp); 5922 return (NFS4ERR_EXIST); 5923 } 5924 *created = TRUE; /* retrans of create == created */ 5925 return (NFS4_OK); 5926 } 5927 VN_RELE(*vpp); 5928 return (NFS4ERR_EXIST); 5929 } 5930 5931 return (NFS4_OK); 5932 } 5933 5934 static nfsstat4 5935 check_open_access(uint32_t access, 5936 struct compound_state *cs, struct svc_req *req) 5937 { 5938 int error; 5939 vnode_t *vp; 5940 bool_t readonly; 5941 cred_t *cr = cs->cr; 5942 5943 /* For now we don't allow mandatory locking as per V2/V3 */ 5944 if (cs->access == CS_ACCESS_DENIED || cs->mandlock) { 5945 return (NFS4ERR_ACCESS); 5946 } 5947 5948 vp = cs->vp; 5949 ASSERT(cr != NULL && vp->v_type == VREG); 5950 5951 /* 5952 * If the file system is exported read only and we are trying 5953 * to open for write, then return NFS4ERR_ROFS 5954 */ 5955 5956 readonly = rdonly4(cs->exi, cs->vp, req); 5957 5958 if ((access & OPEN4_SHARE_ACCESS_WRITE) && readonly) 5959 return (NFS4ERR_ROFS); 5960 5961 if (access & OPEN4_SHARE_ACCESS_READ) { 5962 if ((VOP_ACCESS(vp, VREAD, 0, cr, NULL) != 0) && 5963 (VOP_ACCESS(vp, VEXEC, 0, cr, NULL) != 0)) { 5964 return (NFS4ERR_ACCESS); 5965 } 5966 } 5967 5968 if (access & OPEN4_SHARE_ACCESS_WRITE) { 5969 error = VOP_ACCESS(vp, VWRITE, 0, cr, NULL); 5970 if (error) 5971 return (NFS4ERR_ACCESS); 5972 } 5973 5974 return (NFS4_OK); 5975 } 5976 5977 static nfsstat4 5978 rfs4_createfile(OPEN4args *args, struct svc_req *req, struct compound_state *cs, 5979 change_info4 *cinfo, bitmap4 *attrset, clientid4 clientid) 5980 { 5981 struct nfs4_svgetit_arg sarg; 5982 struct nfs4_ntov_table ntov; 5983 5984 bool_t ntov_table_init = FALSE; 5985 struct statvfs64 sb; 5986 nfsstat4 status; 5987 vnode_t *vp; 5988 vattr_t bva, ava, iva, cva, *vap; 5989 vnode_t *dvp; 5990 timespec32_t *mtime; 5991 char *nm = NULL; 5992 uint_t buflen; 5993 bool_t created; 5994 bool_t setsize = FALSE; 5995 len_t reqsize; 5996 int error; 5997 bool_t trunc; 5998 caller_context_t ct; 5999 component4 *component; 6000 bslabel_t *clabel; 6001 6002 sarg.sbp = &sb; 6003 6004 dvp = cs->vp; 6005 6006 /* Check if the file system is read only */ 6007 if (rdonly4(cs->exi, dvp, req)) 6008 return (NFS4ERR_ROFS); 6009 6010 /* check the label of including directory */ 6011 if (is_system_labeled()) { 6012 ASSERT(req->rq_label != NULL); 6013 clabel = req->rq_label; 6014 DTRACE_PROBE2(tx__rfs4__log__info__opremove__clabel, char *, 6015 "got client label from request(1)", 6016 struct svc_req *, req); 6017 if (!blequal(&l_admin_low->tsl_label, clabel)) { 6018 if (!do_rfs_label_check(clabel, dvp, EQUALITY_CHECK)) { 6019 return (NFS4ERR_ACCESS); 6020 } 6021 } 6022 } 6023 6024 /* 6025 * Get the last component of path name in nm. cs will reference 6026 * the including directory on success. 6027 */ 6028 component = &args->open_claim4_u.file; 6029 if (!utf8_dir_verify(component)) 6030 return (NFS4ERR_INVAL); 6031 6032 nm = utf8_to_fn(component, &buflen, NULL); 6033 6034 if (nm == NULL) 6035 return (NFS4ERR_RESOURCE); 6036 6037 if (buflen > MAXNAMELEN) { 6038 kmem_free(nm, buflen); 6039 return (NFS4ERR_NAMETOOLONG); 6040 } 6041 6042 bva.va_mask = AT_TYPE|AT_CTIME|AT_SEQ; 6043 error = VOP_GETATTR(dvp, &bva, 0, cs->cr, NULL); 6044 if (error) { 6045 kmem_free(nm, buflen); 6046 return (puterrno4(error)); 6047 } 6048 6049 if (bva.va_type != VDIR) { 6050 kmem_free(nm, buflen); 6051 return (NFS4ERR_NOTDIR); 6052 } 6053 6054 NFS4_SET_FATTR4_CHANGE(cinfo->before, bva.va_ctime) 6055 6056 switch (args->mode) { 6057 case GUARDED4: 6058 /*FALLTHROUGH*/ 6059 case UNCHECKED4: 6060 nfs4_ntov_table_init(&ntov); 6061 ntov_table_init = TRUE; 6062 6063 *attrset = 0; 6064 status = do_rfs4_set_attrs(attrset, 6065 &args->createhow4_u.createattrs, 6066 cs, &sarg, &ntov, NFS4ATTR_SETIT); 6067 6068 if (status == NFS4_OK && (sarg.vap->va_mask & AT_TYPE) && 6069 sarg.vap->va_type != VREG) { 6070 if (sarg.vap->va_type == VDIR) 6071 status = NFS4ERR_ISDIR; 6072 else if (sarg.vap->va_type == VLNK) 6073 status = NFS4ERR_SYMLINK; 6074 else 6075 status = NFS4ERR_INVAL; 6076 } 6077 6078 if (status != NFS4_OK) { 6079 kmem_free(nm, buflen); 6080 nfs4_ntov_table_free(&ntov, &sarg); 6081 *attrset = 0; 6082 return (status); 6083 } 6084 6085 vap = sarg.vap; 6086 vap->va_type = VREG; 6087 vap->va_mask |= AT_TYPE; 6088 6089 if ((vap->va_mask & AT_MODE) == 0) { 6090 vap->va_mask |= AT_MODE; 6091 vap->va_mode = (mode_t)0600; 6092 } 6093 6094 if (vap->va_mask & AT_SIZE) { 6095 6096 /* Disallow create with a non-zero size */ 6097 6098 if ((reqsize = sarg.vap->va_size) != 0) { 6099 kmem_free(nm, buflen); 6100 nfs4_ntov_table_free(&ntov, &sarg); 6101 *attrset = 0; 6102 return (NFS4ERR_INVAL); 6103 } 6104 setsize = TRUE; 6105 } 6106 break; 6107 6108 case EXCLUSIVE4: 6109 /* prohibit EXCL create of named attributes */ 6110 if (dvp->v_flag & V_XATTRDIR) { 6111 kmem_free(nm, buflen); 6112 *attrset = 0; 6113 return (NFS4ERR_INVAL); 6114 } 6115 6116 cva.va_mask = AT_TYPE | AT_MTIME | AT_MODE; 6117 cva.va_type = VREG; 6118 /* 6119 * Ensure no time overflows. Assumes underlying 6120 * filesystem supports at least 32 bits. 6121 * Truncate nsec to usec resolution to allow valid 6122 * compares even if the underlying filesystem truncates. 6123 */ 6124 mtime = (timespec32_t *)&args->createhow4_u.createverf; 6125 cva.va_mtime.tv_sec = mtime->tv_sec % TIME32_MAX; 6126 cva.va_mtime.tv_nsec = (mtime->tv_nsec / 1000) * 1000; 6127 cva.va_mode = (mode_t)0; 6128 vap = &cva; 6129 break; 6130 } 6131 6132 status = create_vnode(dvp, nm, vap, args->mode, mtime, 6133 cs->cr, &vp, &created); 6134 kmem_free(nm, buflen); 6135 6136 if (status != NFS4_OK) { 6137 if (ntov_table_init) 6138 nfs4_ntov_table_free(&ntov, &sarg); 6139 *attrset = 0; 6140 return (status); 6141 } 6142 6143 trunc = (setsize && !created); 6144 6145 if (args->mode != EXCLUSIVE4) { 6146 bitmap4 createmask = args->createhow4_u.createattrs.attrmask; 6147 6148 /* 6149 * True verification that object was created with correct 6150 * attrs is impossible. The attrs could have been changed 6151 * immediately after object creation. If attributes did 6152 * not verify, the only recourse for the server is to 6153 * destroy the object. Maybe if some attrs (like gid) 6154 * are set incorrectly, the object should be destroyed; 6155 * however, seems bad as a default policy. Do we really 6156 * want to destroy an object over one of the times not 6157 * verifying correctly? For these reasons, the server 6158 * currently sets bits in attrset for createattrs 6159 * that were set; however, no verification is done. 6160 * 6161 * vmask_to_nmask accounts for vattr bits set on create 6162 * [do_rfs4_set_attrs() only sets resp bits for 6163 * non-vattr/vfs bits.] 6164 * Mask off any bits we set by default so as not to return 6165 * more attrset bits than were requested in createattrs 6166 */ 6167 if (created) { 6168 nfs4_vmask_to_nmask(sarg.vap->va_mask, attrset); 6169 *attrset &= createmask; 6170 } else { 6171 /* 6172 * We did not create the vnode (we tried but it 6173 * already existed). In this case, the only createattr 6174 * that the spec allows the server to set is size, 6175 * and even then, it can only be set if it is 0. 6176 */ 6177 *attrset = 0; 6178 if (trunc) 6179 *attrset = FATTR4_SIZE_MASK; 6180 } 6181 } 6182 if (ntov_table_init) 6183 nfs4_ntov_table_free(&ntov, &sarg); 6184 6185 /* 6186 * Get the initial "after" sequence number, if it fails, 6187 * set to zero, time to before. 6188 */ 6189 iva.va_mask = AT_CTIME|AT_SEQ; 6190 if (VOP_GETATTR(dvp, &iva, 0, cs->cr, NULL)) { 6191 iva.va_seq = 0; 6192 iva.va_ctime = bva.va_ctime; 6193 } 6194 6195 /* 6196 * create_vnode attempts to create the file exclusive, 6197 * if it already exists the VOP_CREATE will fail and 6198 * may not increase va_seq. It is atomic if 6199 * we haven't changed the directory, but if it has changed 6200 * we don't know what changed it. 6201 */ 6202 if (!created) { 6203 if (bva.va_seq && iva.va_seq && 6204 bva.va_seq == iva.va_seq) 6205 cinfo->atomic = TRUE; 6206 else 6207 cinfo->atomic = FALSE; 6208 NFS4_SET_FATTR4_CHANGE(cinfo->after, iva.va_ctime); 6209 } else { 6210 /* 6211 * The entry was created, we need to sync the 6212 * directory metadata. 6213 */ 6214 (void) VOP_FSYNC(dvp, 0, cs->cr, NULL); 6215 6216 /* 6217 * Get "after" change value, if it fails, simply return the 6218 * before value. 6219 */ 6220 ava.va_mask = AT_CTIME|AT_SEQ; 6221 if (VOP_GETATTR(dvp, &ava, 0, cs->cr, NULL)) { 6222 ava.va_ctime = bva.va_ctime; 6223 ava.va_seq = 0; 6224 } 6225 6226 NFS4_SET_FATTR4_CHANGE(cinfo->after, ava.va_ctime); 6227 6228 /* 6229 * The cinfo->atomic = TRUE only if we have 6230 * non-zero va_seq's, and it has incremented by exactly one 6231 * during the create_vnode and it didn't 6232 * change during the VOP_FSYNC. 6233 */ 6234 if (bva.va_seq && iva.va_seq && ava.va_seq && 6235 iva.va_seq == (bva.va_seq + 1) && iva.va_seq == ava.va_seq) 6236 cinfo->atomic = TRUE; 6237 else 6238 cinfo->atomic = FALSE; 6239 } 6240 6241 /* Check for mandatory locking and that the size gets set. */ 6242 cva.va_mask = AT_MODE; 6243 if (setsize) 6244 cva.va_mask |= AT_SIZE; 6245 6246 /* Assume the worst */ 6247 cs->mandlock = TRUE; 6248 6249 if (VOP_GETATTR(vp, &cva, 0, cs->cr, NULL) == 0) { 6250 cs->mandlock = MANDLOCK(cs->vp, cva.va_mode); 6251 6252 /* 6253 * Truncate the file if necessary; this would be 6254 * the case for create over an existing file. 6255 */ 6256 6257 if (trunc) { 6258 int in_crit = 0; 6259 rfs4_file_t *fp; 6260 bool_t create = FALSE; 6261 6262 /* 6263 * We are writing over an existing file. 6264 * Check to see if we need to recall a delegation. 6265 */ 6266 rfs4_hold_deleg_policy(); 6267 if ((fp = rfs4_findfile(vp, NULL, &create)) != NULL) { 6268 if (rfs4_check_delegated_byfp(FWRITE, fp, 6269 (reqsize == 0), FALSE, FALSE, &clientid)) { 6270 rfs4_file_rele(fp); 6271 rfs4_rele_deleg_policy(); 6272 VN_RELE(vp); 6273 *attrset = 0; 6274 return (NFS4ERR_DELAY); 6275 } 6276 rfs4_file_rele(fp); 6277 } 6278 rfs4_rele_deleg_policy(); 6279 6280 if (nbl_need_check(vp)) { 6281 in_crit = 1; 6282 6283 ASSERT(reqsize == 0); 6284 6285 nbl_start_crit(vp, RW_READER); 6286 if (nbl_conflict(vp, NBL_WRITE, 0, 6287 cva.va_size, 0, NULL)) { 6288 in_crit = 0; 6289 nbl_end_crit(vp); 6290 VN_RELE(vp); 6291 *attrset = 0; 6292 return (NFS4ERR_ACCESS); 6293 } 6294 } 6295 ct.cc_sysid = 0; 6296 ct.cc_pid = 0; 6297 ct.cc_caller_id = nfs4_srv_caller_id; 6298 ct.cc_flags = CC_DONTBLOCK; 6299 6300 cva.va_mask = AT_SIZE; 6301 cva.va_size = reqsize; 6302 (void) VOP_SETATTR(vp, &cva, 0, cs->cr, &ct); 6303 if (in_crit) 6304 nbl_end_crit(vp); 6305 } 6306 } 6307 6308 error = makefh4(&cs->fh, vp, cs->exi); 6309 6310 /* 6311 * Force modified data and metadata out to stable storage. 6312 */ 6313 (void) VOP_FSYNC(vp, FNODSYNC, cs->cr, NULL); 6314 6315 if (error) { 6316 VN_RELE(vp); 6317 *attrset = 0; 6318 return (puterrno4(error)); 6319 } 6320 6321 /* if parent dir is attrdir, set namedattr fh flag */ 6322 if (dvp->v_flag & V_XATTRDIR) 6323 set_fh4_flag(&cs->fh, FH4_NAMEDATTR); 6324 6325 if (cs->vp) 6326 VN_RELE(cs->vp); 6327 6328 cs->vp = vp; 6329 6330 /* 6331 * if we did not create the file, we will need to check 6332 * the access bits on the file 6333 */ 6334 6335 if (!created) { 6336 if (setsize) 6337 args->share_access |= OPEN4_SHARE_ACCESS_WRITE; 6338 status = check_open_access(args->share_access, cs, req); 6339 if (status != NFS4_OK) 6340 *attrset = 0; 6341 } 6342 return (status); 6343 } 6344 6345 /*ARGSUSED*/ 6346 static void 6347 rfs4_do_open(struct compound_state *cs, struct svc_req *req, 6348 rfs4_openowner_t *oo, delegreq_t deleg, 6349 uint32_t access, uint32_t deny, 6350 OPEN4res *resp, int deleg_cur) 6351 { 6352 /* XXX Currently not using req */ 6353 rfs4_state_t *state; 6354 rfs4_file_t *file; 6355 bool_t screate = TRUE; 6356 bool_t fcreate = TRUE; 6357 uint32_t amodes; 6358 uint32_t dmodes; 6359 rfs4_deleg_state_t *dsp; 6360 struct shrlock shr; 6361 struct shr_locowner shr_loco; 6362 sysid_t sysid; 6363 nfsstat4 status; 6364 caller_context_t ct; 6365 int fflags = 0; 6366 int recall = 0; 6367 int err; 6368 int cmd; 6369 6370 /* get the file struct and hold a lock on it during initial open */ 6371 file = rfs4_findfile_withlock(cs->vp, &cs->fh, &fcreate); 6372 if (file == NULL) { 6373 NFS4_DEBUG(rfs4_debug, 6374 (CE_NOTE, "rfs4_do_open: can't find file")); 6375 resp->status = NFS4ERR_SERVERFAULT; 6376 return; 6377 } 6378 6379 state = rfs4_findstate_by_owner_file(oo, file, &screate); 6380 if (state == NULL) { 6381 NFS4_DEBUG(rfs4_debug, 6382 (CE_NOTE, "rfs4_do_open: can't find state")); 6383 resp->status = NFS4ERR_RESOURCE; 6384 /* No need to keep any reference */ 6385 rfs4_file_rele_withunlock(file); 6386 return; 6387 } 6388 6389 /* try to get the sysid before continuing */ 6390 if ((status = rfs4_client_sysid(oo->client, &sysid)) != NFS4_OK) { 6391 resp->status = status; 6392 rfs4_file_rele(file); 6393 /* Not a fully formed open; "close" it */ 6394 if (screate == TRUE) 6395 rfs4_state_close(state, FALSE, FALSE, cs->cr); 6396 rfs4_state_rele(state); 6397 return; 6398 } 6399 6400 /* Calculate the fflags for this OPEN. */ 6401 if (access & OPEN4_SHARE_ACCESS_READ) 6402 fflags |= FREAD; 6403 if (access & OPEN4_SHARE_ACCESS_WRITE) 6404 fflags |= FWRITE; 6405 6406 /* 6407 * Calculate the new deny and access mode that this open is adding to 6408 * the file for this open owner; 6409 */ 6410 dmodes = (deny & ~state->share_deny); 6411 amodes = (access & ~state->share_access); 6412 6413 /* 6414 * Check to see the client has already sent an open for this 6415 * open owner on this file with the same share/deny modes. 6416 * If so, we don't need to check for a conflict and we don't 6417 * need to add another shrlock. If not, then we need to 6418 * check for conflicts in deny and access before checking for 6419 * conflicts in delegation. We don't want to recall a 6420 * delegation based on an open that will eventually fail based 6421 * on shares modes. 6422 */ 6423 6424 if (dmodes || amodes) { 6425 shr.s_access = (short)access; 6426 shr.s_deny = (short)deny; 6427 shr.s_pid = rfs4_dbe_getid(oo->dbe); 6428 shr.s_sysid = sysid; 6429 shr_loco.sl_pid = shr.s_pid; 6430 shr_loco.sl_id = shr.s_sysid; 6431 shr.s_owner = (caddr_t)&shr_loco; 6432 shr.s_own_len = sizeof (shr_loco); 6433 6434 cmd = nbl_need_check(cs->vp) ? F_SHARE_NBMAND : F_SHARE; 6435 if ((err = vop_shrlock(cs->vp, cmd, &shr, fflags)) != 0) { 6436 6437 resp->status = err == EAGAIN ? 6438 NFS4ERR_SHARE_DENIED : puterrno4(err); 6439 6440 rfs4_file_rele(file); 6441 /* Not a fully formed open; "close" it */ 6442 if (screate == TRUE) 6443 rfs4_state_close(state, FALSE, FALSE, cs->cr); 6444 rfs4_state_rele(state); 6445 return; 6446 } 6447 } 6448 6449 rfs4_dbe_lock(state->dbe); 6450 rfs4_dbe_lock(file->dbe); 6451 6452 /* 6453 * Check to see if this file is delegated and if so, if a 6454 * recall needs to be done. 6455 */ 6456 if (rfs4_check_recall(state, access)) { 6457 rfs4_dbe_unlock(file->dbe); 6458 rfs4_dbe_unlock(state->dbe); 6459 rfs4_recall_deleg(file, FALSE, state->owner->client); 6460 delay(NFS4_DELEGATION_CONFLICT_DELAY); 6461 rfs4_dbe_lock(state->dbe); 6462 rfs4_dbe_lock(file->dbe); 6463 /* Let's see if the delegation was returned */ 6464 if (rfs4_check_recall(state, access)) { 6465 rfs4_dbe_unlock(file->dbe); 6466 rfs4_dbe_unlock(state->dbe); 6467 rfs4_file_rele(file); 6468 rfs4_update_lease(state->owner->client); 6469 (void) vop_shrlock(cs->vp, F_UNSHARE, &shr, fflags); 6470 /* Not a fully formed open; "close" it */ 6471 if (screate == TRUE) 6472 rfs4_state_close(state, FALSE, FALSE, cs->cr); 6473 rfs4_state_rele(state); 6474 resp->status = NFS4ERR_DELAY; 6475 return; 6476 } 6477 } 6478 /* 6479 * the share check passed and any delegation conflict has been 6480 * taken care of, now call vop_open. 6481 * if this is the first open then call vop_open with fflags. 6482 * if not, call vn_open_upgrade with just the upgrade flags. 6483 * 6484 * if the file has been opened already, it will have the current 6485 * access mode in the state struct. if it has no share access, then 6486 * this is a new open. 6487 * 6488 * However, if this is open with CLAIM_DLEGATE_CUR, then don't 6489 * call VOP_OPEN(), just do the open upgrade. 6490 */ 6491 if (((state->share_access & OPEN4_SHARE_ACCESS_BOTH) == 0) && 6492 !deleg_cur) { 6493 ct.cc_sysid = sysid; 6494 ct.cc_pid = shr.s_pid; 6495 ct.cc_caller_id = nfs4_srv_caller_id; 6496 ct.cc_flags = CC_DONTBLOCK; 6497 err = VOP_OPEN(&cs->vp, fflags, cs->cr, &ct); 6498 if (err) { 6499 rfs4_dbe_unlock(file->dbe); 6500 rfs4_dbe_unlock(state->dbe); 6501 rfs4_file_rele(file); 6502 (void) vop_shrlock(cs->vp, F_UNSHARE, &shr, fflags); 6503 /* Not a fully formed open; "close" it */ 6504 if (screate == TRUE) 6505 rfs4_state_close(state, FALSE, FALSE, cs->cr); 6506 rfs4_state_rele(state); 6507 /* check if a monitor detected a delegation conflict */ 6508 if (err == EAGAIN && (ct.cc_flags & CC_WOULDBLOCK)) 6509 resp->status = NFS4ERR_DELAY; 6510 else 6511 resp->status = NFS4ERR_SERVERFAULT; 6512 return; 6513 } 6514 } else { /* open upgrade */ 6515 /* 6516 * calculate the fflags for the new mode that is being added 6517 * by this upgrade. 6518 */ 6519 fflags = 0; 6520 if (amodes & OPEN4_SHARE_ACCESS_READ) 6521 fflags |= FREAD; 6522 if (amodes & OPEN4_SHARE_ACCESS_WRITE) 6523 fflags |= FWRITE; 6524 vn_open_upgrade(cs->vp, fflags); 6525 } 6526 6527 if (dmodes & OPEN4_SHARE_DENY_READ) 6528 file->deny_read++; 6529 if (dmodes & OPEN4_SHARE_DENY_WRITE) 6530 file->deny_write++; 6531 file->share_deny |= deny; 6532 state->share_deny |= deny; 6533 6534 if (amodes & OPEN4_SHARE_ACCESS_READ) 6535 file->access_read++; 6536 if (amodes & OPEN4_SHARE_ACCESS_WRITE) 6537 file->access_write++; 6538 file->share_access |= access; 6539 state->share_access |= access; 6540 6541 /* 6542 * Check for delegation here. if the deleg argument is not 6543 * DELEG_ANY, then this is a reclaim from a client and 6544 * we must honor the delegation requested. If necessary we can 6545 * set the recall flag. 6546 */ 6547 6548 dsp = rfs4_grant_delegation(deleg, state, &recall); 6549 6550 cs->deleg = (file->dinfo->dtype == OPEN_DELEGATE_WRITE); 6551 6552 next_stateid(&state->stateid); 6553 6554 resp->stateid = state->stateid.stateid; 6555 6556 rfs4_dbe_unlock(file->dbe); 6557 rfs4_dbe_unlock(state->dbe); 6558 6559 if (dsp) { 6560 rfs4_set_deleg_response(dsp, &resp->delegation, NULL, recall); 6561 rfs4_deleg_state_rele(dsp); 6562 } 6563 6564 rfs4_file_rele(file); 6565 rfs4_state_rele(state); 6566 6567 resp->status = NFS4_OK; 6568 } 6569 6570 /*ARGSUSED*/ 6571 static void 6572 rfs4_do_opennull(struct compound_state *cs, struct svc_req *req, 6573 OPEN4args *args, rfs4_openowner_t *oo, OPEN4res *resp) 6574 { 6575 change_info4 *cinfo = &resp->cinfo; 6576 bitmap4 *attrset = &resp->attrset; 6577 6578 if (args->opentype == OPEN4_NOCREATE) 6579 resp->status = rfs4_lookupfile(&args->open_claim4_u.file, 6580 req, cs, args->share_access, cinfo); 6581 else { 6582 /* inhibit delegation grants during exclusive create */ 6583 6584 if (args->mode == EXCLUSIVE4) 6585 rfs4_disable_delegation(); 6586 6587 resp->status = rfs4_createfile(args, req, cs, cinfo, attrset, 6588 oo->client->clientid); 6589 } 6590 6591 if (resp->status == NFS4_OK) { 6592 6593 /* cs->vp cs->fh now reference the desired file */ 6594 6595 rfs4_do_open(cs, req, oo, DELEG_ANY, args->share_access, 6596 args->share_deny, resp, 0); 6597 6598 /* 6599 * If rfs4_createfile set attrset, we must 6600 * clear this attrset before the response is copied. 6601 */ 6602 if (resp->status != NFS4_OK && resp->attrset) { 6603 resp->attrset = 0; 6604 } 6605 } 6606 else 6607 *cs->statusp = resp->status; 6608 6609 if (args->mode == EXCLUSIVE4) 6610 rfs4_enable_delegation(); 6611 } 6612 6613 /*ARGSUSED*/ 6614 static void 6615 rfs4_do_openprev(struct compound_state *cs, struct svc_req *req, 6616 OPEN4args *args, rfs4_openowner_t *oo, OPEN4res *resp) 6617 { 6618 change_info4 *cinfo = &resp->cinfo; 6619 vattr_t va; 6620 vtype_t v_type = cs->vp->v_type; 6621 int error = 0; 6622 6623 /* Verify that we have a regular file */ 6624 if (v_type != VREG) { 6625 if (v_type == VDIR) 6626 resp->status = NFS4ERR_ISDIR; 6627 else if (v_type == VLNK) 6628 resp->status = NFS4ERR_SYMLINK; 6629 else 6630 resp->status = NFS4ERR_INVAL; 6631 return; 6632 } 6633 6634 va.va_mask = AT_MODE|AT_UID; 6635 error = VOP_GETATTR(cs->vp, &va, 0, cs->cr, NULL); 6636 if (error) { 6637 resp->status = puterrno4(error); 6638 return; 6639 } 6640 6641 cs->mandlock = MANDLOCK(cs->vp, va.va_mode); 6642 6643 /* 6644 * Check if we have access to the file, Note the the file 6645 * could have originally been open UNCHECKED or GUARDED 6646 * with mode bits that will now fail, but there is nothing 6647 * we can really do about that except in the case that the 6648 * owner of the file is the one requesting the open. 6649 */ 6650 if (crgetuid(cs->cr) != va.va_uid) { 6651 resp->status = check_open_access(args->share_access, cs, req); 6652 if (resp->status != NFS4_OK) { 6653 return; 6654 } 6655 } 6656 6657 /* 6658 * cinfo on a CLAIM_PREVIOUS is undefined, initialize to zero 6659 */ 6660 cinfo->before = 0; 6661 cinfo->after = 0; 6662 cinfo->atomic = FALSE; 6663 6664 rfs4_do_open(cs, req, oo, 6665 NFS4_DELEG4TYPE2REQTYPE(args->open_claim4_u.delegate_type), 6666 args->share_access, args->share_deny, resp, 0); 6667 } 6668 6669 static void 6670 rfs4_do_opendelcur(struct compound_state *cs, struct svc_req *req, 6671 OPEN4args *args, rfs4_openowner_t *oo, OPEN4res *resp) 6672 { 6673 int error; 6674 nfsstat4 status; 6675 stateid4 stateid = 6676 args->open_claim4_u.delegate_cur_info.delegate_stateid; 6677 rfs4_deleg_state_t *dsp; 6678 6679 /* 6680 * Find the state info from the stateid and confirm that the 6681 * file is delegated. If the state openowner is the same as 6682 * the supplied openowner we're done. If not, get the file 6683 * info from the found state info. Use that file info to 6684 * create the state for this lock owner. Note solaris doen't 6685 * really need the pathname to find the file. We may want to 6686 * lookup the pathname and make sure that the vp exist and 6687 * matches the vp in the file structure. However it is 6688 * possible that the pathname nolonger exists (local process 6689 * unlinks the file), so this may not be that useful. 6690 */ 6691 6692 status = rfs4_get_deleg_state(&stateid, &dsp); 6693 if (status != NFS4_OK) { 6694 resp->status = status; 6695 return; 6696 } 6697 6698 ASSERT(dsp->finfo->dinfo->dtype != OPEN_DELEGATE_NONE); 6699 6700 /* 6701 * New lock owner, create state. Since this was probably called 6702 * in response to a CB_RECALL we set deleg to DELEG_NONE 6703 */ 6704 6705 ASSERT(cs->vp != NULL); 6706 VN_RELE(cs->vp); 6707 VN_HOLD(dsp->finfo->vp); 6708 cs->vp = dsp->finfo->vp; 6709 6710 if (error = makefh4(&cs->fh, cs->vp, cs->exi)) { 6711 rfs4_deleg_state_rele(dsp); 6712 *cs->statusp = resp->status = puterrno4(error); 6713 return; 6714 } 6715 6716 /* Mark progress for delegation returns */ 6717 dsp->finfo->dinfo->time_lastwrite = gethrestime_sec(); 6718 rfs4_deleg_state_rele(dsp); 6719 rfs4_do_open(cs, req, oo, DELEG_NONE, 6720 args->share_access, args->share_deny, resp, 1); 6721 } 6722 6723 /*ARGSUSED*/ 6724 static void 6725 rfs4_do_opendelprev(struct compound_state *cs, struct svc_req *req, 6726 OPEN4args *args, rfs4_openowner_t *oo, OPEN4res *resp) 6727 { 6728 /* 6729 * Lookup the pathname, it must already exist since this file 6730 * was delegated. 6731 * 6732 * Find the file and state info for this vp and open owner pair. 6733 * check that they are in fact delegated. 6734 * check that the state access and deny modes are the same. 6735 * 6736 * Return the delgation possibly seting the recall flag. 6737 */ 6738 rfs4_file_t *file; 6739 rfs4_state_t *state; 6740 bool_t create = FALSE; 6741 bool_t dcreate = FALSE; 6742 rfs4_deleg_state_t *dsp; 6743 nfsace4 *ace; 6744 6745 6746 /* Note we ignore oflags */ 6747 resp->status = rfs4_lookupfile(&args->open_claim4_u.file_delegate_prev, 6748 req, cs, args->share_access, &resp->cinfo); 6749 6750 if (resp->status != NFS4_OK) { 6751 return; 6752 } 6753 6754 /* get the file struct and hold a lock on it during initial open */ 6755 file = rfs4_findfile_withlock(cs->vp, NULL, &create); 6756 if (file == NULL) { 6757 NFS4_DEBUG(rfs4_debug, 6758 (CE_NOTE, "rfs4_do_opendelprev: can't find file")); 6759 resp->status = NFS4ERR_SERVERFAULT; 6760 return; 6761 } 6762 6763 state = rfs4_findstate_by_owner_file(oo, file, &create); 6764 if (state == NULL) { 6765 NFS4_DEBUG(rfs4_debug, 6766 (CE_NOTE, "rfs4_do_opendelprev: can't find state")); 6767 resp->status = NFS4ERR_SERVERFAULT; 6768 rfs4_file_rele_withunlock(file); 6769 return; 6770 } 6771 6772 rfs4_dbe_lock(state->dbe); 6773 rfs4_dbe_lock(file->dbe); 6774 if (args->share_access != state->share_access || 6775 args->share_deny != state->share_deny || 6776 state->finfo->dinfo->dtype == OPEN_DELEGATE_NONE) { 6777 NFS4_DEBUG(rfs4_debug, 6778 (CE_NOTE, "rfs4_do_opendelprev: state mixup")); 6779 rfs4_dbe_unlock(file->dbe); 6780 rfs4_dbe_unlock(state->dbe); 6781 rfs4_file_rele(file); 6782 rfs4_state_rele(state); 6783 resp->status = NFS4ERR_SERVERFAULT; 6784 return; 6785 } 6786 rfs4_dbe_unlock(file->dbe); 6787 rfs4_dbe_unlock(state->dbe); 6788 6789 dsp = rfs4_finddeleg(state, &dcreate); 6790 if (dsp == NULL) { 6791 rfs4_state_rele(state); 6792 rfs4_file_rele(file); 6793 resp->status = NFS4ERR_SERVERFAULT; 6794 return; 6795 } 6796 6797 next_stateid(&state->stateid); 6798 6799 resp->stateid = state->stateid.stateid; 6800 6801 resp->delegation.delegation_type = dsp->dtype; 6802 6803 if (dsp->dtype == OPEN_DELEGATE_READ) { 6804 open_read_delegation4 *rv = 6805 &resp->delegation.open_delegation4_u.read; 6806 6807 rv->stateid = dsp->delegid.stateid; 6808 rv->recall = FALSE; /* no policy in place to set to TRUE */ 6809 ace = &rv->permissions; 6810 } else { 6811 open_write_delegation4 *rv = 6812 &resp->delegation.open_delegation4_u.write; 6813 6814 rv->stateid = dsp->delegid.stateid; 6815 rv->recall = FALSE; /* no policy in place to set to TRUE */ 6816 ace = &rv->permissions; 6817 rv->space_limit.limitby = NFS_LIMIT_SIZE; 6818 rv->space_limit.nfs_space_limit4_u.filesize = UINT64_MAX; 6819 } 6820 6821 /* XXX For now */ 6822 ace->type = ACE4_ACCESS_ALLOWED_ACE_TYPE; 6823 ace->flag = 0; 6824 ace->access_mask = 0; 6825 ace->who.utf8string_len = 0; 6826 ace->who.utf8string_val = 0; 6827 6828 rfs4_deleg_state_rele(dsp); 6829 rfs4_state_rele(state); 6830 rfs4_file_rele(file); 6831 } 6832 6833 typedef enum { 6834 NFS4_CHKSEQ_OKAY = 0, 6835 NFS4_CHKSEQ_REPLAY = 1, 6836 NFS4_CHKSEQ_BAD = 2 6837 } rfs4_chkseq_t; 6838 6839 /* 6840 * Generic function for sequence number checks. 6841 */ 6842 static rfs4_chkseq_t 6843 rfs4_check_seqid(seqid4 seqid, nfs_resop4 *lastop, 6844 seqid4 rqst_seq, nfs_resop4 *resop, bool_t copyres) 6845 { 6846 /* Same sequence ids and matching operations? */ 6847 if (seqid == rqst_seq && resop->resop == lastop->resop) { 6848 if (copyres == TRUE) { 6849 rfs4_free_reply(resop); 6850 rfs4_copy_reply(resop, lastop); 6851 } 6852 NFS4_DEBUG(rfs4_debug, (CE_NOTE, 6853 "Replayed SEQID %d\n", seqid)); 6854 return (NFS4_CHKSEQ_REPLAY); 6855 } 6856 6857 /* If the incoming sequence is not the next expected then it is bad */ 6858 if (rqst_seq != seqid + 1) { 6859 if (rqst_seq == seqid) { 6860 NFS4_DEBUG(rfs4_debug, 6861 (CE_NOTE, "BAD SEQID: Replayed sequence id " 6862 "but last op was %d current op is %d\n", 6863 lastop->resop, resop->resop)); 6864 return (NFS4_CHKSEQ_BAD); 6865 } 6866 NFS4_DEBUG(rfs4_debug, 6867 (CE_NOTE, "BAD SEQID: got %u expecting %u\n", 6868 rqst_seq, seqid)); 6869 return (NFS4_CHKSEQ_BAD); 6870 } 6871 6872 /* Everything okay -- next expected */ 6873 return (NFS4_CHKSEQ_OKAY); 6874 } 6875 6876 6877 static rfs4_chkseq_t 6878 rfs4_check_open_seqid(seqid4 seqid, rfs4_openowner_t *op, nfs_resop4 *resop) 6879 { 6880 rfs4_chkseq_t rc; 6881 6882 rfs4_dbe_lock(op->dbe); 6883 rc = rfs4_check_seqid(op->open_seqid, op->reply, seqid, resop, TRUE); 6884 rfs4_dbe_unlock(op->dbe); 6885 6886 if (rc == NFS4_CHKSEQ_OKAY) 6887 rfs4_update_lease(op->client); 6888 6889 return (rc); 6890 } 6891 6892 static rfs4_chkseq_t 6893 rfs4_check_olo_seqid(seqid4 olo_seqid, rfs4_openowner_t *op, 6894 nfs_resop4 *resop) 6895 { 6896 rfs4_chkseq_t rc; 6897 6898 rfs4_dbe_lock(op->dbe); 6899 rc = rfs4_check_seqid(op->open_seqid, op->reply, 6900 olo_seqid, resop, FALSE); 6901 rfs4_dbe_unlock(op->dbe); 6902 6903 return (rc); 6904 } 6905 6906 static rfs4_chkseq_t 6907 rfs4_check_lock_seqid(seqid4 seqid, rfs4_lo_state_t *lp, nfs_resop4 *resop) 6908 { 6909 rfs4_chkseq_t rc = NFS4_CHKSEQ_OKAY; 6910 6911 rfs4_dbe_lock(lp->dbe); 6912 if (!lp->skip_seqid_check) 6913 rc = rfs4_check_seqid(lp->seqid, lp->reply, seqid, resop, TRUE); 6914 rfs4_dbe_unlock(lp->dbe); 6915 6916 return (rc); 6917 } 6918 6919 static void 6920 rfs4_op_open(nfs_argop4 *argop, nfs_resop4 *resop, 6921 struct svc_req *req, struct compound_state *cs) 6922 { 6923 OPEN4args *args = &argop->nfs_argop4_u.opopen; 6924 OPEN4res *resp = &resop->nfs_resop4_u.opopen; 6925 open_owner4 *owner = &args->owner; 6926 open_claim_type4 claim = args->claim; 6927 rfs4_client_t *cp; 6928 rfs4_openowner_t *oo; 6929 bool_t create; 6930 bool_t replay = FALSE; 6931 int can_reclaim; 6932 6933 DTRACE_NFSV4_2(op__open__start, struct compound_state *, cs, 6934 OPEN4args *, args); 6935 6936 if (cs->vp == NULL) { 6937 *cs->statusp = resp->status = NFS4ERR_NOFILEHANDLE; 6938 goto end; 6939 } 6940 6941 /* 6942 * Need to check clientid and lease expiration first based on 6943 * error ordering and incrementing sequence id. 6944 */ 6945 cp = rfs4_findclient_by_id(owner->clientid, FALSE); 6946 if (cp == NULL) { 6947 *cs->statusp = resp->status = 6948 rfs4_check_clientid(&owner->clientid, 0); 6949 goto end; 6950 } 6951 6952 if (rfs4_lease_expired(cp)) { 6953 rfs4_client_close(cp); 6954 *cs->statusp = resp->status = NFS4ERR_EXPIRED; 6955 goto end; 6956 } 6957 can_reclaim = cp->can_reclaim; 6958 6959 /* 6960 * Find the open_owner for use from this point forward. Take 6961 * care in updating the sequence id based on the type of error 6962 * being returned. 6963 */ 6964 retry: 6965 create = TRUE; 6966 oo = rfs4_findopenowner(owner, &create, args->seqid); 6967 if (oo == NULL) { 6968 *cs->statusp = resp->status = NFS4ERR_STALE_CLIENTID; 6969 rfs4_client_rele(cp); 6970 goto end; 6971 } 6972 6973 /* Hold off access to the sequence space while the open is done */ 6974 rfs4_sw_enter(&oo->oo_sw); 6975 6976 /* 6977 * If the open_owner existed before at the server, then check 6978 * the sequence id. 6979 */ 6980 if (!create && !oo->postpone_confirm) { 6981 switch (rfs4_check_open_seqid(args->seqid, oo, resop)) { 6982 case NFS4_CHKSEQ_BAD: 6983 if ((args->seqid > oo->open_seqid) && 6984 oo->need_confirm) { 6985 rfs4_free_opens(oo, TRUE, FALSE); 6986 rfs4_sw_exit(&oo->oo_sw); 6987 rfs4_openowner_rele(oo); 6988 goto retry; 6989 } 6990 resp->status = NFS4ERR_BAD_SEQID; 6991 goto out; 6992 case NFS4_CHKSEQ_REPLAY: /* replay of previous request */ 6993 replay = TRUE; 6994 goto out; 6995 default: 6996 break; 6997 } 6998 6999 /* 7000 * Sequence was ok and open owner exists 7001 * check to see if we have yet to see an 7002 * open_confirm. 7003 */ 7004 if (oo->need_confirm) { 7005 rfs4_free_opens(oo, TRUE, FALSE); 7006 rfs4_sw_exit(&oo->oo_sw); 7007 rfs4_openowner_rele(oo); 7008 goto retry; 7009 } 7010 } 7011 /* Grace only applies to regular-type OPENs */ 7012 if (rfs4_clnt_in_grace(cp) && 7013 (claim == CLAIM_NULL || claim == CLAIM_DELEGATE_CUR)) { 7014 *cs->statusp = resp->status = NFS4ERR_GRACE; 7015 goto out; 7016 } 7017 7018 /* 7019 * If previous state at the server existed then can_reclaim 7020 * will be set. If not reply NFS4ERR_NO_GRACE to the 7021 * client. 7022 */ 7023 if (rfs4_clnt_in_grace(cp) && claim == CLAIM_PREVIOUS && !can_reclaim) { 7024 *cs->statusp = resp->status = NFS4ERR_NO_GRACE; 7025 goto out; 7026 } 7027 7028 7029 /* 7030 * Reject the open if the client has missed the grace period 7031 */ 7032 if (!rfs4_clnt_in_grace(cp) && claim == CLAIM_PREVIOUS) { 7033 *cs->statusp = resp->status = NFS4ERR_NO_GRACE; 7034 goto out; 7035 } 7036 7037 /* Couple of up-front bookkeeping items */ 7038 if (oo->need_confirm) { 7039 /* 7040 * If this is a reclaim OPEN then we should not ask 7041 * for a confirmation of the open_owner per the 7042 * protocol specification. 7043 */ 7044 if (claim == CLAIM_PREVIOUS) 7045 oo->need_confirm = FALSE; 7046 else 7047 resp->rflags |= OPEN4_RESULT_CONFIRM; 7048 } 7049 resp->rflags |= OPEN4_RESULT_LOCKTYPE_POSIX; 7050 7051 /* 7052 * If there is an unshared filesystem mounted on this vnode, 7053 * do not allow to open/create in this directory. 7054 */ 7055 if (vn_ismntpt(cs->vp)) { 7056 *cs->statusp = resp->status = NFS4ERR_ACCESS; 7057 goto out; 7058 } 7059 7060 /* 7061 * access must READ, WRITE, or BOTH. No access is invalid. 7062 * deny can be READ, WRITE, BOTH, or NONE. 7063 * bits not defined for access/deny are invalid. 7064 */ 7065 if (! (args->share_access & OPEN4_SHARE_ACCESS_BOTH) || 7066 (args->share_access & ~OPEN4_SHARE_ACCESS_BOTH) || 7067 (args->share_deny & ~OPEN4_SHARE_DENY_BOTH)) { 7068 *cs->statusp = resp->status = NFS4ERR_INVAL; 7069 goto out; 7070 } 7071 7072 7073 /* 7074 * make sure attrset is zero before response is built. 7075 */ 7076 resp->attrset = 0; 7077 7078 switch (claim) { 7079 case CLAIM_NULL: 7080 rfs4_do_opennull(cs, req, args, oo, resp); 7081 break; 7082 case CLAIM_PREVIOUS: 7083 rfs4_do_openprev(cs, req, args, oo, resp); 7084 break; 7085 case CLAIM_DELEGATE_CUR: 7086 rfs4_do_opendelcur(cs, req, args, oo, resp); 7087 break; 7088 case CLAIM_DELEGATE_PREV: 7089 rfs4_do_opendelprev(cs, req, args, oo, resp); 7090 break; 7091 default: 7092 resp->status = NFS4ERR_INVAL; 7093 break; 7094 } 7095 7096 out: 7097 rfs4_client_rele(cp); 7098 7099 /* Catch sequence id handling here to make it a little easier */ 7100 switch (resp->status) { 7101 case NFS4ERR_BADXDR: 7102 case NFS4ERR_BAD_SEQID: 7103 case NFS4ERR_BAD_STATEID: 7104 case NFS4ERR_NOFILEHANDLE: 7105 case NFS4ERR_RESOURCE: 7106 case NFS4ERR_STALE_CLIENTID: 7107 case NFS4ERR_STALE_STATEID: 7108 /* 7109 * The protocol states that if any of these errors are 7110 * being returned, the sequence id should not be 7111 * incremented. Any other return requires an 7112 * increment. 7113 */ 7114 break; 7115 default: 7116 /* Always update the lease in this case */ 7117 rfs4_update_lease(oo->client); 7118 7119 /* Regular response - copy the result */ 7120 if (!replay) 7121 rfs4_update_open_resp(oo, resop, &cs->fh); 7122 7123 /* 7124 * REPLAY case: Only if the previous response was OK 7125 * do we copy the filehandle. If not OK, no 7126 * filehandle to copy. 7127 */ 7128 if (replay == TRUE && 7129 resp->status == NFS4_OK && 7130 oo->reply_fh.nfs_fh4_val) { 7131 /* 7132 * If this is a replay, we must restore the 7133 * current filehandle/vp to that of what was 7134 * returned originally. Try our best to do 7135 * it. 7136 */ 7137 nfs_fh4_fmt_t *fh_fmtp = 7138 (nfs_fh4_fmt_t *)oo->reply_fh.nfs_fh4_val; 7139 7140 cs->exi = checkexport4(&fh_fmtp->fh4_fsid, 7141 (fid_t *)&fh_fmtp->fh4_xlen, NULL); 7142 7143 if (cs->exi == NULL) { 7144 resp->status = NFS4ERR_STALE; 7145 goto finish; 7146 } 7147 7148 VN_RELE(cs->vp); 7149 7150 cs->vp = nfs4_fhtovp(&oo->reply_fh, cs->exi, 7151 &resp->status); 7152 7153 if (cs->vp == NULL) 7154 goto finish; 7155 7156 nfs_fh4_copy(&oo->reply_fh, &cs->fh); 7157 } 7158 7159 /* 7160 * If this was a replay, no need to update the 7161 * sequence id. If the open_owner was not created on 7162 * this pass, then update. The first use of an 7163 * open_owner will not bump the sequence id. 7164 */ 7165 if (replay == FALSE && !create) 7166 rfs4_update_open_sequence(oo); 7167 /* 7168 * If the client is receiving an error and the 7169 * open_owner needs to be confirmed, there is no way 7170 * to notify the client of this fact ignoring the fact 7171 * that the server has no method of returning a 7172 * stateid to confirm. Therefore, the server needs to 7173 * mark this open_owner in a way as to avoid the 7174 * sequence id checking the next time the client uses 7175 * this open_owner. 7176 */ 7177 if (resp->status != NFS4_OK && oo->need_confirm) 7178 oo->postpone_confirm = TRUE; 7179 /* 7180 * If OK response then clear the postpone flag and 7181 * reset the sequence id to keep in sync with the 7182 * client. 7183 */ 7184 if (resp->status == NFS4_OK && oo->postpone_confirm) { 7185 oo->postpone_confirm = FALSE; 7186 oo->open_seqid = args->seqid; 7187 } 7188 break; 7189 } 7190 7191 finish: 7192 *cs->statusp = resp->status; 7193 7194 rfs4_sw_exit(&oo->oo_sw); 7195 rfs4_openowner_rele(oo); 7196 7197 end: 7198 DTRACE_NFSV4_2(op__open__done, struct compound_state *, cs, 7199 OPEN4res *, resp); 7200 } 7201 7202 /*ARGSUSED*/ 7203 void 7204 rfs4_op_open_confirm(nfs_argop4 *argop, nfs_resop4 *resop, 7205 struct svc_req *req, struct compound_state *cs) 7206 { 7207 OPEN_CONFIRM4args *args = &argop->nfs_argop4_u.opopen_confirm; 7208 OPEN_CONFIRM4res *resp = &resop->nfs_resop4_u.opopen_confirm; 7209 rfs4_state_t *sp; 7210 nfsstat4 status; 7211 7212 DTRACE_NFSV4_2(op__open__confirm__start, struct compound_state *, cs, 7213 OPEN_CONFIRM4args *, args); 7214 7215 if (cs->vp == NULL) { 7216 *cs->statusp = resp->status = NFS4ERR_NOFILEHANDLE; 7217 goto out; 7218 } 7219 7220 status = rfs4_get_state(&args->open_stateid, &sp, RFS4_DBS_VALID); 7221 if (status != NFS4_OK) { 7222 *cs->statusp = resp->status = status; 7223 goto out; 7224 } 7225 7226 /* Ensure specified filehandle matches */ 7227 if (cs->vp != sp->finfo->vp) { 7228 rfs4_state_rele(sp); 7229 *cs->statusp = resp->status = NFS4ERR_BAD_STATEID; 7230 goto out; 7231 } 7232 7233 /* hold off other access to open_owner while we tinker */ 7234 rfs4_sw_enter(&sp->owner->oo_sw); 7235 7236 switch (rfs4_check_stateid_seqid(sp, &args->open_stateid)) { 7237 case NFS4_CHECK_STATEID_OKAY: 7238 if (rfs4_check_open_seqid(args->seqid, sp->owner, 7239 resop) != 0) { 7240 *cs->statusp = resp->status = NFS4ERR_BAD_SEQID; 7241 break; 7242 } 7243 /* 7244 * If it is the appropriate stateid and determined to 7245 * be "OKAY" then this means that the stateid does not 7246 * need to be confirmed and the client is in error for 7247 * sending an OPEN_CONFIRM. 7248 */ 7249 *cs->statusp = resp->status = NFS4ERR_BAD_STATEID; 7250 break; 7251 case NFS4_CHECK_STATEID_OLD: 7252 *cs->statusp = resp->status = NFS4ERR_OLD_STATEID; 7253 break; 7254 case NFS4_CHECK_STATEID_BAD: 7255 *cs->statusp = resp->status = NFS4ERR_BAD_STATEID; 7256 break; 7257 case NFS4_CHECK_STATEID_EXPIRED: 7258 *cs->statusp = resp->status = NFS4ERR_EXPIRED; 7259 break; 7260 case NFS4_CHECK_STATEID_CLOSED: 7261 *cs->statusp = resp->status = NFS4ERR_OLD_STATEID; 7262 break; 7263 case NFS4_CHECK_STATEID_REPLAY: 7264 switch (rfs4_check_open_seqid(args->seqid, sp->owner, resop)) { 7265 case NFS4_CHKSEQ_OKAY: 7266 /* 7267 * This is replayed stateid; if seqid matches 7268 * next expected, then client is using wrong seqid. 7269 */ 7270 /* fall through */ 7271 case NFS4_CHKSEQ_BAD: 7272 *cs->statusp = resp->status = NFS4ERR_BAD_SEQID; 7273 break; 7274 case NFS4_CHKSEQ_REPLAY: 7275 /* 7276 * Note this case is the duplicate case so 7277 * resp->status is already set. 7278 */ 7279 *cs->statusp = resp->status; 7280 rfs4_update_lease(sp->owner->client); 7281 break; 7282 } 7283 break; 7284 case NFS4_CHECK_STATEID_UNCONFIRMED: 7285 if (rfs4_check_open_seqid(args->seqid, sp->owner, 7286 resop) != NFS4_CHKSEQ_OKAY) { 7287 *cs->statusp = resp->status = NFS4ERR_BAD_SEQID; 7288 break; 7289 } 7290 *cs->statusp = resp->status = NFS4_OK; 7291 7292 next_stateid(&sp->stateid); 7293 resp->open_stateid = sp->stateid.stateid; 7294 sp->owner->need_confirm = FALSE; 7295 rfs4_update_lease(sp->owner->client); 7296 rfs4_update_open_sequence(sp->owner); 7297 rfs4_update_open_resp(sp->owner, resop, NULL); 7298 break; 7299 default: 7300 ASSERT(FALSE); 7301 *cs->statusp = resp->status = NFS4ERR_SERVERFAULT; 7302 break; 7303 } 7304 rfs4_sw_exit(&sp->owner->oo_sw); 7305 rfs4_state_rele(sp); 7306 7307 out: 7308 DTRACE_NFSV4_2(op__open__confirm__done, struct compound_state *, cs, 7309 OPEN_CONFIRM4res *, resp); 7310 } 7311 7312 /*ARGSUSED*/ 7313 void 7314 rfs4_op_open_downgrade(nfs_argop4 *argop, nfs_resop4 *resop, 7315 struct svc_req *req, struct compound_state *cs) 7316 { 7317 OPEN_DOWNGRADE4args *args = &argop->nfs_argop4_u.opopen_downgrade; 7318 OPEN_DOWNGRADE4res *resp = &resop->nfs_resop4_u.opopen_downgrade; 7319 uint32_t access = args->share_access; 7320 uint32_t deny = args->share_deny; 7321 nfsstat4 status; 7322 rfs4_state_t *sp; 7323 rfs4_file_t *fp; 7324 int fflags = 0; 7325 7326 DTRACE_NFSV4_2(op__open__downgrade__start, struct compound_state *, cs, 7327 OPEN_DOWNGRADE4args *, args); 7328 7329 if (cs->vp == NULL) { 7330 *cs->statusp = resp->status = NFS4ERR_NOFILEHANDLE; 7331 goto out; 7332 } 7333 7334 status = rfs4_get_state(&args->open_stateid, &sp, RFS4_DBS_VALID); 7335 if (status != NFS4_OK) { 7336 *cs->statusp = resp->status = status; 7337 goto out; 7338 } 7339 7340 /* Ensure specified filehandle matches */ 7341 if (cs->vp != sp->finfo->vp) { 7342 rfs4_state_rele(sp); 7343 *cs->statusp = resp->status = NFS4ERR_BAD_STATEID; 7344 goto out; 7345 } 7346 7347 /* hold off other access to open_owner while we tinker */ 7348 rfs4_sw_enter(&sp->owner->oo_sw); 7349 7350 switch (rfs4_check_stateid_seqid(sp, &args->open_stateid)) { 7351 case NFS4_CHECK_STATEID_OKAY: 7352 if (rfs4_check_open_seqid(args->seqid, sp->owner, 7353 resop) != NFS4_CHKSEQ_OKAY) { 7354 *cs->statusp = resp->status = NFS4ERR_BAD_SEQID; 7355 goto end; 7356 } 7357 break; 7358 case NFS4_CHECK_STATEID_OLD: 7359 *cs->statusp = resp->status = NFS4ERR_OLD_STATEID; 7360 goto end; 7361 case NFS4_CHECK_STATEID_BAD: 7362 *cs->statusp = resp->status = NFS4ERR_BAD_STATEID; 7363 goto end; 7364 case NFS4_CHECK_STATEID_EXPIRED: 7365 *cs->statusp = resp->status = NFS4ERR_EXPIRED; 7366 goto end; 7367 case NFS4_CHECK_STATEID_CLOSED: 7368 *cs->statusp = resp->status = NFS4ERR_OLD_STATEID; 7369 goto end; 7370 case NFS4_CHECK_STATEID_UNCONFIRMED: 7371 *cs->statusp = resp->status = NFS4ERR_BAD_STATEID; 7372 goto end; 7373 case NFS4_CHECK_STATEID_REPLAY: 7374 /* Check the sequence id for the open owner */ 7375 switch (rfs4_check_open_seqid(args->seqid, sp->owner, resop)) { 7376 case NFS4_CHKSEQ_OKAY: 7377 /* 7378 * This is replayed stateid; if seqid matches 7379 * next expected, then client is using wrong seqid. 7380 */ 7381 /* fall through */ 7382 case NFS4_CHKSEQ_BAD: 7383 *cs->statusp = resp->status = NFS4ERR_BAD_SEQID; 7384 goto end; 7385 case NFS4_CHKSEQ_REPLAY: 7386 /* 7387 * Note this case is the duplicate case so 7388 * resp->status is already set. 7389 */ 7390 *cs->statusp = resp->status; 7391 rfs4_update_lease(sp->owner->client); 7392 goto end; 7393 } 7394 break; 7395 default: 7396 ASSERT(FALSE); 7397 break; 7398 } 7399 7400 rfs4_dbe_lock(sp->dbe); 7401 /* 7402 * Check that the new access modes and deny modes are valid. 7403 * Check that no invalid bits are set. 7404 */ 7405 if ((access & ~(OPEN4_SHARE_ACCESS_READ | OPEN4_SHARE_ACCESS_WRITE)) || 7406 (deny & ~(OPEN4_SHARE_DENY_READ | OPEN4_SHARE_DENY_WRITE))) { 7407 *cs->statusp = resp->status = NFS4ERR_INVAL; 7408 rfs4_update_open_sequence(sp->owner); 7409 rfs4_dbe_unlock(sp->dbe); 7410 goto end; 7411 } 7412 7413 /* 7414 * The new modes must be a subset of the current modes and 7415 * the access must specify at least one mode. To test that 7416 * the new mode is a subset of the current modes we bitwise 7417 * AND them together and check that the result equals the new 7418 * mode. For example: 7419 * New mode, access == R and current mode, sp->share_access == RW 7420 * access & sp->share_access == R == access, so the new access mode 7421 * is valid. Consider access == RW, sp->share_access = R 7422 * access & sp->share_access == R != access, so the new access mode 7423 * is invalid. 7424 */ 7425 if ((access & sp->share_access) != access || 7426 (deny & sp->share_deny) != deny || 7427 (access & 7428 (OPEN4_SHARE_ACCESS_READ | OPEN4_SHARE_ACCESS_WRITE)) == 0) { 7429 *cs->statusp = resp->status = NFS4ERR_INVAL; 7430 rfs4_update_open_sequence(sp->owner); 7431 rfs4_dbe_unlock(sp->dbe); 7432 goto end; 7433 } 7434 7435 /* 7436 * Release any share locks associated with this stateID. 7437 * Strictly speaking, this violates the spec because the 7438 * spec effectively requires that open downgrade be atomic. 7439 * At present, fs_shrlock does not have this capability. 7440 */ 7441 rfs4_dbe_unlock(sp->dbe); 7442 rfs4_unshare(sp); 7443 rfs4_dbe_lock(sp->dbe); 7444 7445 fp = sp->finfo; 7446 rfs4_dbe_lock(fp->dbe); 7447 7448 /* 7449 * If the current mode has deny read and the new mode 7450 * does not, decrement the number of deny read mode bits 7451 * and if it goes to zero turn off the deny read bit 7452 * on the file. 7453 */ 7454 if ((sp->share_deny & OPEN4_SHARE_DENY_READ) && 7455 (deny & OPEN4_SHARE_DENY_READ) == 0) { 7456 fp->deny_read--; 7457 if (fp->deny_read == 0) 7458 fp->share_deny &= ~OPEN4_SHARE_DENY_READ; 7459 } 7460 7461 /* 7462 * If the current mode has deny write and the new mode 7463 * does not, decrement the number of deny write mode bits 7464 * and if it goes to zero turn off the deny write bit 7465 * on the file. 7466 */ 7467 if ((sp->share_deny & OPEN4_SHARE_DENY_WRITE) && 7468 (deny & OPEN4_SHARE_DENY_WRITE) == 0) { 7469 fp->deny_write--; 7470 if (fp->deny_write == 0) 7471 fp->share_deny &= ~OPEN4_SHARE_DENY_WRITE; 7472 } 7473 7474 /* 7475 * If the current mode has access read and the new mode 7476 * does not, decrement the number of access read mode bits 7477 * and if it goes to zero turn off the access read bit 7478 * on the file. set fflags to FREAD for the call to 7479 * vn_open_downgrade(). 7480 */ 7481 if ((sp->share_access & OPEN4_SHARE_ACCESS_READ) && 7482 (access & OPEN4_SHARE_ACCESS_READ) == 0) { 7483 fp->access_read--; 7484 if (fp->access_read == 0) 7485 fp->share_access &= ~OPEN4_SHARE_ACCESS_READ; 7486 fflags |= FREAD; 7487 } 7488 7489 /* 7490 * If the current mode has access write and the new mode 7491 * does not, decrement the number of access write mode bits 7492 * and if it goes to zero turn off the access write bit 7493 * on the file. set fflags to FWRITE for the call to 7494 * vn_open_downgrade(). 7495 */ 7496 if ((sp->share_access & OPEN4_SHARE_ACCESS_WRITE) && 7497 (access & OPEN4_SHARE_ACCESS_WRITE) == 0) { 7498 fp->access_write--; 7499 if (fp->access_write == 0) 7500 fp->share_deny &= ~OPEN4_SHARE_ACCESS_WRITE; 7501 fflags |= FWRITE; 7502 } 7503 7504 /* Set the new access and deny modes */ 7505 sp->share_access = access; 7506 sp->share_deny = deny; 7507 /* Check that the file is still accessible */ 7508 ASSERT(fp->share_access); 7509 7510 rfs4_dbe_unlock(fp->dbe); 7511 7512 rfs4_dbe_unlock(sp->dbe); 7513 7514 if ((status = rfs4_share(sp)) != NFS4_OK) { 7515 *cs->statusp = resp->status = NFS4ERR_SERVERFAULT; 7516 rfs4_update_open_sequence(sp->owner); 7517 goto end; 7518 } 7519 7520 /* 7521 * we successfully downgraded the share lock, now we need to downgrade 7522 * the open. it is possible that the downgrade was only for a deny 7523 * mode and we have nothing else to do. 7524 */ 7525 if ((fflags & (FREAD|FWRITE)) != 0) 7526 vn_open_downgrade(cs->vp, fflags); 7527 7528 rfs4_dbe_lock(sp->dbe); 7529 7530 /* Update the stateid */ 7531 next_stateid(&sp->stateid); 7532 resp->open_stateid = sp->stateid.stateid; 7533 7534 rfs4_dbe_unlock(sp->dbe); 7535 7536 *cs->statusp = resp->status = NFS4_OK; 7537 /* Update the lease */ 7538 rfs4_update_lease(sp->owner->client); 7539 /* And the sequence */ 7540 rfs4_update_open_sequence(sp->owner); 7541 rfs4_update_open_resp(sp->owner, resop, NULL); 7542 7543 end: 7544 rfs4_sw_exit(&sp->owner->oo_sw); 7545 rfs4_state_rele(sp); 7546 out: 7547 DTRACE_NFSV4_2(op__open__downgrade__done, struct compound_state *, cs, 7548 OPEN_DOWNGRADE4res *, resp); 7549 } 7550 7551 /* 7552 * The logic behind this function is detailed in the NFSv4 RFC in the 7553 * SETCLIENTID operation description under IMPLEMENTATION. Refer to 7554 * that section for explicit guidance to server behavior for 7555 * SETCLIENTID. 7556 */ 7557 void 7558 rfs4_op_setclientid(nfs_argop4 *argop, nfs_resop4 *resop, 7559 struct svc_req *req, struct compound_state *cs) 7560 { 7561 SETCLIENTID4args *args = &argop->nfs_argop4_u.opsetclientid; 7562 SETCLIENTID4res *res = &resop->nfs_resop4_u.opsetclientid; 7563 rfs4_client_t *cp, *newcp, *cp_confirmed, *cp_unconfirmed; 7564 bool_t create = TRUE; 7565 char *addr, *netid; 7566 int len; 7567 7568 DTRACE_NFSV4_2(op__setclientid__start, struct compound_state *, cs, 7569 SETCLIENTID4args *, args); 7570 retry: 7571 newcp = cp_confirmed = cp_unconfirmed = NULL; 7572 7573 /* 7574 * In search of an EXISTING client matching the incoming 7575 * request to establish a new client identifier at the server 7576 */ 7577 create = TRUE; 7578 cp = rfs4_findclient(&args->client, &create, NULL); 7579 7580 /* Should never happen */ 7581 ASSERT(cp != NULL); 7582 7583 if (cp == NULL) { 7584 *cs->statusp = res->status = NFS4ERR_SERVERFAULT; 7585 goto out; 7586 } 7587 7588 /* 7589 * Easiest case. Client identifier is newly created and is 7590 * unconfirmed. Also note that for this case, no other 7591 * entries exist for the client identifier. Nothing else to 7592 * check. Just setup the response and respond. 7593 */ 7594 if (create) { 7595 *cs->statusp = res->status = NFS4_OK; 7596 res->SETCLIENTID4res_u.resok4.clientid = cp->clientid; 7597 res->SETCLIENTID4res_u.resok4.setclientid_confirm = 7598 cp->confirm_verf; 7599 /* Setup callback information; CB_NULL confirmation later */ 7600 rfs4_client_setcb(cp, &args->callback, args->callback_ident); 7601 7602 rfs4_client_rele(cp); 7603 goto out; 7604 } 7605 7606 /* 7607 * An existing, confirmed client may exist but it may not have 7608 * been active for at least one lease period. If so, then 7609 * "close" the client and create a new client identifier 7610 */ 7611 if (rfs4_lease_expired(cp)) { 7612 rfs4_client_close(cp); 7613 goto retry; 7614 } 7615 7616 if (cp->need_confirm == TRUE) 7617 cp_unconfirmed = cp; 7618 else 7619 cp_confirmed = cp; 7620 7621 cp = NULL; 7622 7623 /* 7624 * We have a confirmed client, now check for an 7625 * unconfimred entry 7626 */ 7627 if (cp_confirmed) { 7628 /* If creds don't match then client identifier is inuse */ 7629 if (!creds_ok(cp_confirmed->cr_set, req, cs)) { 7630 rfs4_cbinfo_t *cbp; 7631 /* 7632 * Some one else has established this client 7633 * id. Try and say * who they are. We will use 7634 * the call back address supplied by * the 7635 * first client. 7636 */ 7637 *cs->statusp = res->status = NFS4ERR_CLID_INUSE; 7638 7639 addr = netid = NULL; 7640 7641 cbp = &cp_confirmed->cbinfo; 7642 if (cbp->cb_callback.cb_location.r_addr && 7643 cbp->cb_callback.cb_location.r_netid) { 7644 cb_client4 *cbcp = &cbp->cb_callback; 7645 7646 len = strlen(cbcp->cb_location.r_addr)+1; 7647 addr = kmem_alloc(len, KM_SLEEP); 7648 bcopy(cbcp->cb_location.r_addr, addr, len); 7649 len = strlen(cbcp->cb_location.r_netid)+1; 7650 netid = kmem_alloc(len, KM_SLEEP); 7651 bcopy(cbcp->cb_location.r_netid, netid, len); 7652 } 7653 7654 res->SETCLIENTID4res_u.client_using.r_addr = addr; 7655 res->SETCLIENTID4res_u.client_using.r_netid = netid; 7656 7657 rfs4_client_rele(cp_confirmed); 7658 } 7659 7660 /* 7661 * Confirmed, creds match, and verifier matches; must 7662 * be an update of the callback info 7663 */ 7664 if (cp_confirmed->nfs_client.verifier == 7665 args->client.verifier) { 7666 /* Setup callback information */ 7667 rfs4_client_setcb(cp_confirmed, &args->callback, 7668 args->callback_ident); 7669 7670 /* everything okay -- move ahead */ 7671 *cs->statusp = res->status = NFS4_OK; 7672 res->SETCLIENTID4res_u.resok4.clientid = 7673 cp_confirmed->clientid; 7674 7675 /* update the confirm_verifier and return it */ 7676 rfs4_client_scv_next(cp_confirmed); 7677 res->SETCLIENTID4res_u.resok4.setclientid_confirm = 7678 cp_confirmed->confirm_verf; 7679 7680 rfs4_client_rele(cp_confirmed); 7681 goto out; 7682 } 7683 7684 /* 7685 * Creds match but the verifier doesn't. Must search 7686 * for an unconfirmed client that would be replaced by 7687 * this request. 7688 */ 7689 create = FALSE; 7690 cp_unconfirmed = rfs4_findclient(&args->client, &create, 7691 cp_confirmed); 7692 } 7693 7694 /* 7695 * At this point, we have taken care of the brand new client 7696 * struct, INUSE case, update of an existing, and confirmed 7697 * client struct. 7698 */ 7699 7700 /* 7701 * check to see if things have changed while we originally 7702 * picked up the client struct. If they have, then return and 7703 * retry the processing of this SETCLIENTID request. 7704 */ 7705 if (cp_unconfirmed) { 7706 rfs4_dbe_lock(cp_unconfirmed->dbe); 7707 if (!cp_unconfirmed->need_confirm) { 7708 rfs4_dbe_unlock(cp_unconfirmed->dbe); 7709 rfs4_client_rele(cp_unconfirmed); 7710 if (cp_confirmed) 7711 rfs4_client_rele(cp_confirmed); 7712 goto retry; 7713 } 7714 /* do away with the old unconfirmed one */ 7715 rfs4_dbe_invalidate(cp_unconfirmed->dbe); 7716 rfs4_dbe_unlock(cp_unconfirmed->dbe); 7717 rfs4_client_rele(cp_unconfirmed); 7718 cp_unconfirmed = NULL; 7719 } 7720 7721 /* 7722 * This search will temporarily hide the confirmed client 7723 * struct while a new client struct is created as the 7724 * unconfirmed one. 7725 */ 7726 create = TRUE; 7727 newcp = rfs4_findclient(&args->client, &create, cp_confirmed); 7728 7729 ASSERT(newcp != NULL); 7730 7731 if (newcp == NULL) { 7732 *cs->statusp = res->status = NFS4ERR_SERVERFAULT; 7733 rfs4_client_rele(cp_confirmed); 7734 goto out; 7735 } 7736 7737 /* 7738 * If one was not created, then a similar request must be in 7739 * process so release and start over with this one 7740 */ 7741 if (create != TRUE) { 7742 rfs4_client_rele(newcp); 7743 if (cp_confirmed) 7744 rfs4_client_rele(cp_confirmed); 7745 goto retry; 7746 } 7747 7748 *cs->statusp = res->status = NFS4_OK; 7749 res->SETCLIENTID4res_u.resok4.clientid = newcp->clientid; 7750 res->SETCLIENTID4res_u.resok4.setclientid_confirm = newcp->confirm_verf; 7751 /* Setup callback information; CB_NULL confirmation later */ 7752 rfs4_client_setcb(newcp, &args->callback, args->callback_ident); 7753 7754 newcp->cp_confirmed = cp_confirmed; 7755 7756 rfs4_client_rele(newcp); 7757 7758 out: 7759 DTRACE_NFSV4_2(op__setclientid__done, struct compound_state *, cs, 7760 SETCLIENTID4res *, res); 7761 } 7762 7763 /*ARGSUSED*/ 7764 void 7765 rfs4_op_setclientid_confirm(nfs_argop4 *argop, nfs_resop4 *resop, 7766 struct svc_req *req, struct compound_state *cs) 7767 { 7768 SETCLIENTID_CONFIRM4args *args = 7769 &argop->nfs_argop4_u.opsetclientid_confirm; 7770 SETCLIENTID_CONFIRM4res *res = 7771 &resop->nfs_resop4_u.opsetclientid_confirm; 7772 rfs4_client_t *cp, *cptoclose = NULL; 7773 7774 DTRACE_NFSV4_2(op__setclientid__confirm__start, 7775 struct compound_state *, cs, 7776 SETCLIENTID_CONFIRM4args *, args); 7777 7778 *cs->statusp = res->status = NFS4_OK; 7779 7780 cp = rfs4_findclient_by_id(args->clientid, TRUE); 7781 7782 if (cp == NULL) { 7783 *cs->statusp = res->status = 7784 rfs4_check_clientid(&args->clientid, 1); 7785 goto out; 7786 } 7787 7788 if (!creds_ok(cp, req, cs)) { 7789 *cs->statusp = res->status = NFS4ERR_CLID_INUSE; 7790 rfs4_client_rele(cp); 7791 goto out; 7792 } 7793 7794 /* If the verifier doesn't match, the record doesn't match */ 7795 if (cp->confirm_verf != args->setclientid_confirm) { 7796 *cs->statusp = res->status = NFS4ERR_STALE_CLIENTID; 7797 rfs4_client_rele(cp); 7798 goto out; 7799 } 7800 7801 rfs4_dbe_lock(cp->dbe); 7802 cp->need_confirm = FALSE; 7803 if (cp->cp_confirmed) { 7804 cptoclose = cp->cp_confirmed; 7805 cptoclose->ss_remove = 1; 7806 cp->cp_confirmed = NULL; 7807 } 7808 7809 /* 7810 * Update the client's associated server instance, if it's changed 7811 * since the client was created. 7812 */ 7813 if (rfs4_servinst(cp) != rfs4_cur_servinst) 7814 rfs4_servinst_assign(cp, rfs4_cur_servinst); 7815 7816 /* 7817 * Record clientid in stable storage. 7818 * Must be done after server instance has been assigned. 7819 */ 7820 rfs4_ss_clid(cp, req); 7821 7822 rfs4_dbe_unlock(cp->dbe); 7823 7824 if (cptoclose) 7825 /* don't need to rele, client_close does it */ 7826 rfs4_client_close(cptoclose); 7827 7828 /* If needed, initiate CB_NULL call for callback path */ 7829 rfs4_deleg_cb_check(cp); 7830 rfs4_update_lease(cp); 7831 7832 /* 7833 * Check to see if client can perform reclaims 7834 */ 7835 rfs4_ss_chkclid(cp); 7836 7837 rfs4_client_rele(cp); 7838 7839 out: 7840 DTRACE_NFSV4_2(op__setclientid__confirm__done, 7841 struct compound_state *, cs, 7842 SETCLIENTID_CONFIRM4 *, res); 7843 } 7844 7845 7846 /*ARGSUSED*/ 7847 void 7848 rfs4_op_close(nfs_argop4 *argop, nfs_resop4 *resop, 7849 struct svc_req *req, struct compound_state *cs) 7850 { 7851 CLOSE4args *args = &argop->nfs_argop4_u.opclose; 7852 CLOSE4res *resp = &resop->nfs_resop4_u.opclose; 7853 rfs4_state_t *sp; 7854 nfsstat4 status; 7855 7856 DTRACE_NFSV4_2(op__close__start, struct compound_state *, cs, 7857 CLOSE4args *, args); 7858 7859 if (cs->vp == NULL) { 7860 *cs->statusp = resp->status = NFS4ERR_NOFILEHANDLE; 7861 goto out; 7862 } 7863 7864 status = rfs4_get_state(&args->open_stateid, &sp, RFS4_DBS_INVALID); 7865 if (status != NFS4_OK) { 7866 *cs->statusp = resp->status = status; 7867 goto out; 7868 } 7869 7870 /* Ensure specified filehandle matches */ 7871 if (cs->vp != sp->finfo->vp) { 7872 rfs4_state_rele(sp); 7873 *cs->statusp = resp->status = NFS4ERR_BAD_STATEID; 7874 goto out; 7875 } 7876 7877 /* hold off other access to open_owner while we tinker */ 7878 rfs4_sw_enter(&sp->owner->oo_sw); 7879 7880 switch (rfs4_check_stateid_seqid(sp, &args->open_stateid)) { 7881 case NFS4_CHECK_STATEID_OKAY: 7882 if (rfs4_check_open_seqid(args->seqid, sp->owner, 7883 resop) != NFS4_CHKSEQ_OKAY) { 7884 *cs->statusp = resp->status = NFS4ERR_BAD_SEQID; 7885 goto end; 7886 } 7887 break; 7888 case NFS4_CHECK_STATEID_OLD: 7889 *cs->statusp = resp->status = NFS4ERR_OLD_STATEID; 7890 goto end; 7891 case NFS4_CHECK_STATEID_BAD: 7892 *cs->statusp = resp->status = NFS4ERR_BAD_STATEID; 7893 goto end; 7894 case NFS4_CHECK_STATEID_EXPIRED: 7895 *cs->statusp = resp->status = NFS4ERR_EXPIRED; 7896 goto end; 7897 case NFS4_CHECK_STATEID_CLOSED: 7898 *cs->statusp = resp->status = NFS4ERR_OLD_STATEID; 7899 goto end; 7900 case NFS4_CHECK_STATEID_UNCONFIRMED: 7901 *cs->statusp = resp->status = NFS4ERR_BAD_STATEID; 7902 goto end; 7903 case NFS4_CHECK_STATEID_REPLAY: 7904 /* Check the sequence id for the open owner */ 7905 switch (rfs4_check_open_seqid(args->seqid, sp->owner, resop)) { 7906 case NFS4_CHKSEQ_OKAY: 7907 /* 7908 * This is replayed stateid; if seqid matches 7909 * next expected, then client is using wrong seqid. 7910 */ 7911 /* FALL THROUGH */ 7912 case NFS4_CHKSEQ_BAD: 7913 *cs->statusp = resp->status = NFS4ERR_BAD_SEQID; 7914 goto end; 7915 case NFS4_CHKSEQ_REPLAY: 7916 /* 7917 * Note this case is the duplicate case so 7918 * resp->status is already set. 7919 */ 7920 *cs->statusp = resp->status; 7921 rfs4_update_lease(sp->owner->client); 7922 goto end; 7923 } 7924 break; 7925 default: 7926 ASSERT(FALSE); 7927 break; 7928 } 7929 7930 rfs4_dbe_lock(sp->dbe); 7931 7932 /* Update the stateid. */ 7933 next_stateid(&sp->stateid); 7934 resp->open_stateid = sp->stateid.stateid; 7935 7936 rfs4_dbe_unlock(sp->dbe); 7937 7938 rfs4_update_lease(sp->owner->client); 7939 rfs4_update_open_sequence(sp->owner); 7940 rfs4_update_open_resp(sp->owner, resop, NULL); 7941 7942 rfs4_state_close(sp, FALSE, FALSE, cs->cr); 7943 7944 *cs->statusp = resp->status = status; 7945 7946 end: 7947 rfs4_sw_exit(&sp->owner->oo_sw); 7948 rfs4_state_rele(sp); 7949 out: 7950 DTRACE_NFSV4_2(op__close__done, struct compound_state *, cs, 7951 CLOSE4res *, resp); 7952 } 7953 7954 /* 7955 * Manage the counts on the file struct and close all file locks 7956 */ 7957 /*ARGSUSED*/ 7958 void 7959 rfs4_release_share_lock_state(rfs4_state_t *sp, cred_t *cr, 7960 bool_t close_of_client) 7961 { 7962 rfs4_file_t *fp = sp->finfo; 7963 rfs4_lo_state_t *lsp; 7964 struct shrlock shr; 7965 struct shr_locowner shr_loco; 7966 int fflags, s_access, s_deny; 7967 7968 fflags = s_access = s_deny = 0; 7969 /* 7970 * Decrement the count for each access and deny bit that this 7971 * state has contributed to the file. If the file counts go to zero 7972 * clear the appropriate bit in the appropriate mask. 7973 */ 7974 7975 if (sp->share_access & OPEN4_SHARE_ACCESS_READ) { 7976 fp->access_read--; 7977 fflags |= FREAD; 7978 s_access |= F_RDACC; 7979 if (fp->access_read == 0) 7980 fp->share_access &= ~OPEN4_SHARE_ACCESS_READ; 7981 } 7982 if (sp->share_access & OPEN4_SHARE_ACCESS_WRITE) { 7983 fp->access_write--; 7984 fflags |= FWRITE; 7985 s_access |= F_WRACC; 7986 if (fp->access_write == 0) 7987 fp->share_access &= ~OPEN4_SHARE_ACCESS_WRITE; 7988 } 7989 if (sp->share_deny & OPEN4_SHARE_DENY_READ) { 7990 fp->deny_read--; 7991 s_deny |= F_RDDNY; 7992 if (fp->deny_read == 0) 7993 fp->share_deny &= ~OPEN4_SHARE_DENY_READ; 7994 } 7995 if (sp->share_deny & OPEN4_SHARE_DENY_WRITE) { 7996 fp->deny_write--; 7997 s_deny |= F_WRDNY; 7998 if (fp->deny_write == 0) 7999 fp->share_deny &= ~OPEN4_SHARE_DENY_WRITE; 8000 } 8001 8002 /* 8003 * If this call is part of the larger closing down of client 8004 * state then it is just easier to release all locks 8005 * associated with this client instead of going through each 8006 * individual file and cleaning locks there. 8007 */ 8008 if (close_of_client) { 8009 if (sp->owner->client->unlksys_completed == FALSE && 8010 sp->lockownerlist.next->lsp != NULL && 8011 sp->owner->client->sysidt != LM_NOSYSID) { 8012 /* Is the PxFS kernel module loaded? */ 8013 if (lm_remove_file_locks != NULL) { 8014 int new_sysid; 8015 8016 /* Encode the cluster nodeid in new sysid */ 8017 new_sysid = sp->owner->client->sysidt; 8018 lm_set_nlmid_flk(&new_sysid); 8019 8020 /* 8021 * This PxFS routine removes file locks for a 8022 * client over all nodes of a cluster. 8023 */ 8024 NFS4_DEBUG(rfs4_debug, (CE_NOTE, 8025 "lm_remove_file_locks(sysid=0x%x)\n", 8026 new_sysid)); 8027 (*lm_remove_file_locks)(new_sysid); 8028 } else { 8029 struct flock64 flk; 8030 8031 /* Release all locks for this client */ 8032 flk.l_type = F_UNLKSYS; 8033 flk.l_whence = 0; 8034 flk.l_start = 0; 8035 flk.l_len = 0; 8036 flk.l_sysid = sp->owner->client->sysidt; 8037 flk.l_pid = 0; 8038 (void) VOP_FRLOCK(sp->finfo->vp, F_SETLK, &flk, 8039 F_REMOTELOCK | FREAD | FWRITE, 8040 (u_offset_t)0, NULL, CRED(), NULL); 8041 } 8042 8043 sp->owner->client->unlksys_completed = TRUE; 8044 } 8045 } 8046 8047 /* 8048 * Release all locks on this file by this lock owner or at 8049 * least mark the locks as having been released 8050 */ 8051 for (lsp = sp->lockownerlist.next->lsp; lsp != NULL; 8052 lsp = lsp->lockownerlist.next->lsp) { 8053 8054 lsp->locks_cleaned = TRUE; 8055 8056 /* Was this already taken care of above? */ 8057 if (!close_of_client && 8058 sp->owner->client->sysidt != LM_NOSYSID) 8059 (void) cleanlocks(sp->finfo->vp, lsp->locker->pid, 8060 lsp->locker->client->sysidt); 8061 } 8062 8063 /* 8064 * Release any shrlocks associated with this open state ID. 8065 * This must be done before the rfs4_state gets marked closed. 8066 */ 8067 if (sp->owner->client->sysidt != LM_NOSYSID) { 8068 shr.s_access = s_access; 8069 shr.s_deny = s_deny; 8070 shr.s_pid = rfs4_dbe_getid(sp->owner->dbe); 8071 shr.s_sysid = sp->owner->client->sysidt; 8072 shr_loco.sl_pid = shr.s_pid; 8073 shr_loco.sl_id = shr.s_sysid; 8074 shr.s_owner = (caddr_t)&shr_loco; 8075 shr.s_own_len = sizeof (shr_loco); 8076 (void) vop_shrlock(sp->finfo->vp, F_UNSHARE, &shr, fflags); 8077 } 8078 8079 (void) VOP_CLOSE(fp->vp, fflags, 1, (offset_t)0, cr, NULL); 8080 } 8081 8082 /* 8083 * lock_denied: Fill in a LOCK4deneid structure given an flock64 structure. 8084 */ 8085 static nfsstat4 8086 lock_denied(LOCK4denied *dp, struct flock64 *flk) 8087 { 8088 rfs4_lockowner_t *lo; 8089 rfs4_client_t *cp; 8090 uint32_t len; 8091 8092 lo = rfs4_findlockowner_by_pid(flk->l_pid); 8093 if (lo != NULL) { 8094 cp = lo->client; 8095 if (rfs4_lease_expired(cp)) { 8096 rfs4_lockowner_rele(lo); 8097 rfs4_dbe_hold(cp->dbe); 8098 rfs4_client_close(cp); 8099 return (NFS4ERR_EXPIRED); 8100 } 8101 dp->owner.clientid = lo->owner.clientid; 8102 len = lo->owner.owner_len; 8103 dp->owner.owner_val = kmem_alloc(len, KM_SLEEP); 8104 bcopy(lo->owner.owner_val, dp->owner.owner_val, len); 8105 dp->owner.owner_len = len; 8106 rfs4_lockowner_rele(lo); 8107 goto finish; 8108 } 8109 8110 /* 8111 * Its not a NFS4 lock. We take advantage that the upper 32 bits 8112 * of the client id contain the boot time for a NFS4 lock. So we 8113 * fabricate and identity by setting clientid to the sysid, and 8114 * the lock owner to the pid. 8115 */ 8116 dp->owner.clientid = flk->l_sysid; 8117 len = sizeof (pid_t); 8118 dp->owner.owner_len = len; 8119 dp->owner.owner_val = kmem_alloc(len, KM_SLEEP); 8120 bcopy(&flk->l_pid, dp->owner.owner_val, len); 8121 finish: 8122 dp->offset = flk->l_start; 8123 dp->length = flk->l_len; 8124 8125 if (flk->l_type == F_RDLCK) 8126 dp->locktype = READ_LT; 8127 else if (flk->l_type == F_WRLCK) 8128 dp->locktype = WRITE_LT; 8129 else 8130 return (NFS4ERR_INVAL); /* no mapping from POSIX ltype to v4 */ 8131 8132 return (NFS4_OK); 8133 } 8134 8135 static int 8136 setlock(vnode_t *vp, struct flock64 *flock, int flag, cred_t *cred) 8137 { 8138 int error; 8139 struct flock64 flk; 8140 int i; 8141 clock_t delaytime; 8142 int cmd; 8143 8144 cmd = nbl_need_check(vp) ? F_SETLK_NBMAND : F_SETLK; 8145 retry: 8146 delaytime = MSEC_TO_TICK_ROUNDUP(rfs4_lock_delay); 8147 8148 for (i = 0; i < rfs4_maxlock_tries; i++) { 8149 LOCK_PRINT(rfs4_debug, "setlock", cmd, flock); 8150 error = VOP_FRLOCK(vp, cmd, 8151 flock, flag, (u_offset_t)0, NULL, cred, NULL); 8152 8153 if (error != EAGAIN && error != EACCES) 8154 break; 8155 8156 if (i < rfs4_maxlock_tries - 1) { 8157 delay(delaytime); 8158 delaytime *= 2; 8159 } 8160 } 8161 8162 if (error == EAGAIN || error == EACCES) { 8163 /* Get the owner of the lock */ 8164 flk = *flock; 8165 LOCK_PRINT(rfs4_debug, "setlock", F_GETLK, &flk); 8166 if (VOP_FRLOCK(vp, F_GETLK, &flk, flag, 8167 (u_offset_t)0, NULL, cred, NULL) == 0) { 8168 if (flk.l_type == F_UNLCK) { 8169 /* No longer locked, retry */ 8170 goto retry; 8171 } 8172 *flock = flk; 8173 LOCK_PRINT(rfs4_debug, "setlock(blocking lock)", 8174 F_GETLK, &flk); 8175 } 8176 } 8177 8178 return (error); 8179 } 8180 8181 /*ARGSUSED*/ 8182 static nfsstat4 8183 rfs4_do_lock(rfs4_lo_state_t *lp, nfs_lock_type4 locktype, 8184 seqid4 seqid, offset4 offset, 8185 length4 length, cred_t *cred, nfs_resop4 *resop) 8186 { 8187 nfsstat4 status; 8188 rfs4_lockowner_t *lo = lp->locker; 8189 rfs4_state_t *sp = lp->state; 8190 struct flock64 flock; 8191 int16_t ltype; 8192 int flag; 8193 int error; 8194 sysid_t sysid; 8195 LOCK4res *lres; 8196 8197 if (rfs4_lease_expired(lo->client)) { 8198 return (NFS4ERR_EXPIRED); 8199 } 8200 8201 if ((status = rfs4_client_sysid(lo->client, &sysid)) != NFS4_OK) 8202 return (status); 8203 8204 /* Check for zero length. To lock to end of file use all ones for V4 */ 8205 if (length == 0) 8206 return (NFS4ERR_INVAL); 8207 else if (length == (length4)(~0)) 8208 length = 0; /* Posix to end of file */ 8209 8210 retry: 8211 rfs4_dbe_lock(sp->dbe); 8212 8213 8214 if (resop->resop != OP_LOCKU) { 8215 switch (locktype) { 8216 case READ_LT: 8217 case READW_LT: 8218 if ((sp->share_access 8219 & OPEN4_SHARE_ACCESS_READ) == 0) { 8220 rfs4_dbe_unlock(sp->dbe); 8221 8222 return (NFS4ERR_OPENMODE); 8223 } 8224 ltype = F_RDLCK; 8225 break; 8226 case WRITE_LT: 8227 case WRITEW_LT: 8228 if ((sp->share_access 8229 & OPEN4_SHARE_ACCESS_WRITE) == 0) { 8230 rfs4_dbe_unlock(sp->dbe); 8231 8232 return (NFS4ERR_OPENMODE); 8233 } 8234 ltype = F_WRLCK; 8235 break; 8236 } 8237 } else 8238 ltype = F_UNLCK; 8239 8240 flock.l_type = ltype; 8241 flock.l_whence = 0; /* SEEK_SET */ 8242 flock.l_start = offset; 8243 flock.l_len = length; 8244 flock.l_sysid = sysid; 8245 flock.l_pid = lp->locker->pid; 8246 8247 /* Note that length4 is uint64_t but l_len and l_start are off64_t */ 8248 if (flock.l_len < 0 || flock.l_start < 0) { 8249 rfs4_dbe_unlock(sp->dbe); 8250 return (NFS4ERR_INVAL); 8251 } 8252 8253 /* 8254 * N.B. FREAD has the same value as OPEN4_SHARE_ACCESS_READ and 8255 * FWRITE has the same value as OPEN4_SHARE_ACCESS_WRITE. 8256 */ 8257 flag = (int)sp->share_access | F_REMOTELOCK; 8258 8259 error = setlock(sp->finfo->vp, &flock, flag, cred); 8260 if (error == 0) { 8261 rfs4_dbe_lock(lp->dbe); 8262 next_stateid(&lp->lockid); 8263 rfs4_dbe_unlock(lp->dbe); 8264 } 8265 8266 rfs4_dbe_unlock(sp->dbe); 8267 8268 /* 8269 * N.B. We map error values to nfsv4 errors. This is differrent 8270 * than puterrno4 routine. 8271 */ 8272 switch (error) { 8273 case 0: 8274 status = NFS4_OK; 8275 break; 8276 case EAGAIN: 8277 case EACCES: /* Old value */ 8278 /* Can only get here if op is OP_LOCK */ 8279 ASSERT(resop->resop == OP_LOCK); 8280 lres = &resop->nfs_resop4_u.oplock; 8281 status = NFS4ERR_DENIED; 8282 if (lock_denied(&lres->LOCK4res_u.denied, &flock) 8283 == NFS4ERR_EXPIRED) 8284 goto retry; 8285 break; 8286 case ENOLCK: 8287 status = NFS4ERR_DELAY; 8288 break; 8289 case EOVERFLOW: 8290 status = NFS4ERR_INVAL; 8291 break; 8292 case EINVAL: 8293 status = NFS4ERR_NOTSUPP; 8294 break; 8295 default: 8296 cmn_err(CE_WARN, "rfs4_do_lock: unexpected errno (%d)", error); 8297 status = NFS4ERR_SERVERFAULT; 8298 break; 8299 } 8300 8301 return (status); 8302 } 8303 8304 /*ARGSUSED*/ 8305 void 8306 rfs4_op_lock(nfs_argop4 *argop, nfs_resop4 *resop, 8307 struct svc_req *req, struct compound_state *cs) 8308 { 8309 LOCK4args *args = &argop->nfs_argop4_u.oplock; 8310 LOCK4res *resp = &resop->nfs_resop4_u.oplock; 8311 nfsstat4 status; 8312 stateid4 *stateid; 8313 rfs4_lockowner_t *lo; 8314 rfs4_client_t *cp; 8315 rfs4_state_t *sp = NULL; 8316 rfs4_lo_state_t *lsp = NULL; 8317 bool_t ls_sw_held = FALSE; 8318 bool_t create = TRUE; 8319 bool_t lcreate = TRUE; 8320 bool_t dup_lock = FALSE; 8321 int rc; 8322 8323 DTRACE_NFSV4_2(op__lock__start, struct compound_state *, cs, 8324 LOCK4args *, args); 8325 8326 if (cs->vp == NULL) { 8327 *cs->statusp = resp->status = NFS4ERR_NOFILEHANDLE; 8328 DTRACE_NFSV4_2(op__lock__done, struct compound_state *, 8329 cs, LOCK4res *, resp); 8330 return; 8331 } 8332 8333 if (args->locker.new_lock_owner) { 8334 /* Create a new lockowner for this instance */ 8335 open_to_lock_owner4 *olo = &args->locker.locker4_u.open_owner; 8336 8337 NFS4_DEBUG(rfs4_debug, (CE_NOTE, "Creating new lock owner")); 8338 8339 stateid = &olo->open_stateid; 8340 status = rfs4_get_state(stateid, &sp, RFS4_DBS_VALID); 8341 if (status != NFS4_OK) { 8342 NFS4_DEBUG(rfs4_debug, 8343 (CE_NOTE, "Get state failed in lock %d", status)); 8344 *cs->statusp = resp->status = status; 8345 DTRACE_NFSV4_2(op__lock__done, struct compound_state *, 8346 cs, LOCK4res *, resp); 8347 return; 8348 } 8349 8350 /* Ensure specified filehandle matches */ 8351 if (cs->vp != sp->finfo->vp) { 8352 rfs4_state_rele(sp); 8353 *cs->statusp = resp->status = NFS4ERR_BAD_STATEID; 8354 DTRACE_NFSV4_2(op__lock__done, struct compound_state *, 8355 cs, LOCK4res *, resp); 8356 return; 8357 } 8358 8359 /* hold off other access to open_owner while we tinker */ 8360 rfs4_sw_enter(&sp->owner->oo_sw); 8361 8362 switch (rc = rfs4_check_stateid_seqid(sp, stateid)) { 8363 case NFS4_CHECK_STATEID_OLD: 8364 *cs->statusp = resp->status = NFS4ERR_OLD_STATEID; 8365 goto end; 8366 case NFS4_CHECK_STATEID_BAD: 8367 *cs->statusp = resp->status = NFS4ERR_BAD_STATEID; 8368 goto end; 8369 case NFS4_CHECK_STATEID_EXPIRED: 8370 *cs->statusp = resp->status = NFS4ERR_EXPIRED; 8371 goto end; 8372 case NFS4_CHECK_STATEID_UNCONFIRMED: 8373 *cs->statusp = resp->status = NFS4ERR_BAD_STATEID; 8374 goto end; 8375 case NFS4_CHECK_STATEID_CLOSED: 8376 *cs->statusp = resp->status = NFS4ERR_OLD_STATEID; 8377 goto end; 8378 case NFS4_CHECK_STATEID_OKAY: 8379 case NFS4_CHECK_STATEID_REPLAY: 8380 switch (rfs4_check_olo_seqid(olo->open_seqid, 8381 sp->owner, resop)) { 8382 case NFS4_CHKSEQ_OKAY: 8383 if (rc == NFS4_CHECK_STATEID_OKAY) 8384 break; 8385 /* 8386 * This is replayed stateid; if seqid 8387 * matches next expected, then client 8388 * is using wrong seqid. 8389 */ 8390 /* FALLTHROUGH */ 8391 case NFS4_CHKSEQ_BAD: 8392 *cs->statusp = resp->status = NFS4ERR_BAD_SEQID; 8393 goto end; 8394 case NFS4_CHKSEQ_REPLAY: 8395 /* This is a duplicate LOCK request */ 8396 dup_lock = TRUE; 8397 8398 /* 8399 * For a duplicate we do not want to 8400 * create a new lockowner as it should 8401 * already exist. 8402 * Turn off the lockowner create flag. 8403 */ 8404 lcreate = FALSE; 8405 } 8406 break; 8407 } 8408 8409 lo = rfs4_findlockowner(&olo->lock_owner, &lcreate); 8410 if (lo == NULL) { 8411 NFS4_DEBUG(rfs4_debug, 8412 (CE_NOTE, "rfs4_op_lock: no lock owner")); 8413 *cs->statusp = resp->status = NFS4ERR_RESOURCE; 8414 goto end; 8415 } 8416 8417 lsp = rfs4_findlo_state_by_owner(lo, sp, &create); 8418 if (lsp == NULL) { 8419 rfs4_update_lease(sp->owner->client); 8420 /* 8421 * Only update theh open_seqid if this is not 8422 * a duplicate request 8423 */ 8424 if (dup_lock == FALSE) { 8425 rfs4_update_open_sequence(sp->owner); 8426 } 8427 8428 NFS4_DEBUG(rfs4_debug, 8429 (CE_NOTE, "rfs4_op_lock: no state")); 8430 *cs->statusp = resp->status = NFS4ERR_SERVERFAULT; 8431 rfs4_update_open_resp(sp->owner, resop, NULL); 8432 rfs4_lockowner_rele(lo); 8433 goto end; 8434 } 8435 8436 /* 8437 * This is the new_lock_owner branch and the client is 8438 * supposed to be associating a new lock_owner with 8439 * the open file at this point. If we find that a 8440 * lock_owner/state association already exists and a 8441 * successful LOCK request was returned to the client, 8442 * an error is returned to the client since this is 8443 * not appropriate. The client should be using the 8444 * existing lock_owner branch. 8445 */ 8446 if (dup_lock == FALSE && create == FALSE) { 8447 if (lsp->lock_completed == TRUE) { 8448 *cs->statusp = 8449 resp->status = NFS4ERR_BAD_SEQID; 8450 rfs4_lockowner_rele(lo); 8451 goto end; 8452 } 8453 } 8454 8455 rfs4_update_lease(sp->owner->client); 8456 8457 /* 8458 * Only update theh open_seqid if this is not 8459 * a duplicate request 8460 */ 8461 if (dup_lock == FALSE) { 8462 rfs4_update_open_sequence(sp->owner); 8463 } 8464 8465 /* 8466 * If this is a duplicate lock request, just copy the 8467 * previously saved reply and return. 8468 */ 8469 if (dup_lock == TRUE) { 8470 /* verify that lock_seqid's match */ 8471 if (lsp->seqid != olo->lock_seqid) { 8472 NFS4_DEBUG(rfs4_debug, 8473 (CE_NOTE, "rfs4_op_lock: Dup-Lock seqid bad" 8474 "lsp->seqid=%d old->seqid=%d", 8475 lsp->seqid, olo->lock_seqid)); 8476 *cs->statusp = resp->status = NFS4ERR_BAD_SEQID; 8477 } else { 8478 rfs4_copy_reply(resop, lsp->reply); 8479 /* 8480 * Make sure to copy the just 8481 * retrieved reply status into the 8482 * overall compound status 8483 */ 8484 *cs->statusp = resp->status; 8485 } 8486 rfs4_lockowner_rele(lo); 8487 goto end; 8488 } 8489 8490 rfs4_dbe_lock(lsp->dbe); 8491 8492 /* Make sure to update the lock sequence id */ 8493 lsp->seqid = olo->lock_seqid; 8494 8495 NFS4_DEBUG(rfs4_debug, 8496 (CE_NOTE, "Lock seqid established as %d", lsp->seqid)); 8497 8498 /* 8499 * This is used to signify the newly created lockowner 8500 * stateid and its sequence number. The checks for 8501 * sequence number and increment don't occur on the 8502 * very first lock request for a lockowner. 8503 */ 8504 lsp->skip_seqid_check = TRUE; 8505 8506 /* hold off other access to lsp while we tinker */ 8507 rfs4_sw_enter(&lsp->ls_sw); 8508 ls_sw_held = TRUE; 8509 8510 rfs4_dbe_unlock(lsp->dbe); 8511 8512 rfs4_lockowner_rele(lo); 8513 } else { 8514 stateid = &args->locker.locker4_u.lock_owner.lock_stateid; 8515 /* get lsp and hold the lock on the underlying file struct */ 8516 if ((status = rfs4_get_lo_state(stateid, &lsp, TRUE)) 8517 != NFS4_OK) { 8518 *cs->statusp = resp->status = status; 8519 DTRACE_NFSV4_2(op__lock__done, struct compound_state *, 8520 cs, LOCK4res *, resp); 8521 return; 8522 } 8523 create = FALSE; /* We didn't create lsp */ 8524 8525 /* Ensure specified filehandle matches */ 8526 if (cs->vp != lsp->state->finfo->vp) { 8527 rfs4_lo_state_rele(lsp, TRUE); 8528 *cs->statusp = resp->status = NFS4ERR_BAD_STATEID; 8529 DTRACE_NFSV4_2(op__lock__done, struct compound_state *, 8530 cs, LOCK4res *, resp); 8531 return; 8532 } 8533 8534 /* hold off other access to lsp while we tinker */ 8535 rfs4_sw_enter(&lsp->ls_sw); 8536 ls_sw_held = TRUE; 8537 8538 switch (rfs4_check_lo_stateid_seqid(lsp, stateid)) { 8539 /* 8540 * The stateid looks like it was okay (expected to be 8541 * the next one) 8542 */ 8543 case NFS4_CHECK_STATEID_OKAY: 8544 /* 8545 * The sequence id is now checked. Determine 8546 * if this is a replay or if it is in the 8547 * expected (next) sequence. In the case of a 8548 * replay, there are two replay conditions 8549 * that may occur. The first is the normal 8550 * condition where a LOCK is done with a 8551 * NFS4_OK response and the stateid is 8552 * updated. That case is handled below when 8553 * the stateid is identified as a REPLAY. The 8554 * second is the case where an error is 8555 * returned, like NFS4ERR_DENIED, and the 8556 * sequence number is updated but the stateid 8557 * is not updated. This second case is dealt 8558 * with here. So it may seem odd that the 8559 * stateid is okay but the sequence id is a 8560 * replay but it is okay. 8561 */ 8562 switch (rfs4_check_lock_seqid( 8563 args->locker.locker4_u.lock_owner.lock_seqid, 8564 lsp, resop)) { 8565 case NFS4_CHKSEQ_REPLAY: 8566 if (resp->status != NFS4_OK) { 8567 /* 8568 * Here is our replay and need 8569 * to verify that the last 8570 * response was an error. 8571 */ 8572 *cs->statusp = resp->status; 8573 goto end; 8574 } 8575 /* 8576 * This is done since the sequence id 8577 * looked like a replay but it didn't 8578 * pass our check so a BAD_SEQID is 8579 * returned as a result. 8580 */ 8581 /*FALLTHROUGH*/ 8582 case NFS4_CHKSEQ_BAD: 8583 *cs->statusp = resp->status = NFS4ERR_BAD_SEQID; 8584 goto end; 8585 case NFS4_CHKSEQ_OKAY: 8586 /* Everything looks okay move ahead */ 8587 break; 8588 } 8589 break; 8590 case NFS4_CHECK_STATEID_OLD: 8591 *cs->statusp = resp->status = NFS4ERR_OLD_STATEID; 8592 goto end; 8593 case NFS4_CHECK_STATEID_BAD: 8594 *cs->statusp = resp->status = NFS4ERR_BAD_STATEID; 8595 goto end; 8596 case NFS4_CHECK_STATEID_EXPIRED: 8597 *cs->statusp = resp->status = NFS4ERR_EXPIRED; 8598 goto end; 8599 case NFS4_CHECK_STATEID_CLOSED: 8600 *cs->statusp = resp->status = NFS4ERR_OLD_STATEID; 8601 goto end; 8602 case NFS4_CHECK_STATEID_REPLAY: 8603 switch (rfs4_check_lock_seqid( 8604 args->locker.locker4_u.lock_owner.lock_seqid, 8605 lsp, resop)) { 8606 case NFS4_CHKSEQ_OKAY: 8607 /* 8608 * This is a replayed stateid; if 8609 * seqid matches the next expected, 8610 * then client is using wrong seqid. 8611 */ 8612 case NFS4_CHKSEQ_BAD: 8613 *cs->statusp = resp->status = NFS4ERR_BAD_SEQID; 8614 goto end; 8615 case NFS4_CHKSEQ_REPLAY: 8616 rfs4_update_lease(lsp->locker->client); 8617 *cs->statusp = status = resp->status; 8618 goto end; 8619 } 8620 break; 8621 default: 8622 ASSERT(FALSE); 8623 break; 8624 } 8625 8626 rfs4_update_lock_sequence(lsp); 8627 rfs4_update_lease(lsp->locker->client); 8628 } 8629 8630 /* 8631 * NFS4 only allows locking on regular files, so 8632 * verify type of object. 8633 */ 8634 if (cs->vp->v_type != VREG) { 8635 if (cs->vp->v_type == VDIR) 8636 status = NFS4ERR_ISDIR; 8637 else 8638 status = NFS4ERR_INVAL; 8639 goto out; 8640 } 8641 8642 cp = lsp->state->owner->client; 8643 8644 if (rfs4_clnt_in_grace(cp) && !args->reclaim) { 8645 status = NFS4ERR_GRACE; 8646 goto out; 8647 } 8648 8649 if (rfs4_clnt_in_grace(cp) && args->reclaim && !cp->can_reclaim) { 8650 status = NFS4ERR_NO_GRACE; 8651 goto out; 8652 } 8653 8654 if (!rfs4_clnt_in_grace(cp) && args->reclaim) { 8655 status = NFS4ERR_NO_GRACE; 8656 goto out; 8657 } 8658 8659 if (lsp->state->finfo->dinfo->dtype == OPEN_DELEGATE_WRITE) 8660 cs->deleg = TRUE; 8661 8662 status = rfs4_do_lock(lsp, args->locktype, 8663 args->locker.locker4_u.lock_owner.lock_seqid, 8664 args->offset, args->length, cs->cr, resop); 8665 8666 out: 8667 lsp->skip_seqid_check = FALSE; 8668 8669 *cs->statusp = resp->status = status; 8670 8671 if (status == NFS4_OK) { 8672 resp->LOCK4res_u.lock_stateid = lsp->lockid.stateid; 8673 lsp->lock_completed = TRUE; 8674 } 8675 /* 8676 * Only update the "OPEN" response here if this was a new 8677 * lock_owner 8678 */ 8679 if (sp) 8680 rfs4_update_open_resp(sp->owner, resop, NULL); 8681 8682 rfs4_update_lock_resp(lsp, resop); 8683 8684 end: 8685 if (lsp) { 8686 if (ls_sw_held) 8687 rfs4_sw_exit(&lsp->ls_sw); 8688 /* 8689 * If an sp obtained, then the lsp does not represent 8690 * a lock on the file struct. 8691 */ 8692 if (sp != NULL) 8693 rfs4_lo_state_rele(lsp, FALSE); 8694 else 8695 rfs4_lo_state_rele(lsp, TRUE); 8696 } 8697 if (sp) { 8698 rfs4_sw_exit(&sp->owner->oo_sw); 8699 rfs4_state_rele(sp); 8700 } 8701 8702 DTRACE_NFSV4_2(op__lock__done, struct compound_state *, cs, 8703 LOCK4res *, resp); 8704 } 8705 8706 /* free function for LOCK/LOCKT */ 8707 static void 8708 lock_denied_free(nfs_resop4 *resop) 8709 { 8710 LOCK4denied *dp = NULL; 8711 8712 switch (resop->resop) { 8713 case OP_LOCK: 8714 if (resop->nfs_resop4_u.oplock.status == NFS4ERR_DENIED) 8715 dp = &resop->nfs_resop4_u.oplock.LOCK4res_u.denied; 8716 break; 8717 case OP_LOCKT: 8718 if (resop->nfs_resop4_u.oplockt.status == NFS4ERR_DENIED) 8719 dp = &resop->nfs_resop4_u.oplockt.denied; 8720 break; 8721 default: 8722 break; 8723 } 8724 8725 if (dp) 8726 kmem_free(dp->owner.owner_val, dp->owner.owner_len); 8727 } 8728 8729 /*ARGSUSED*/ 8730 void 8731 rfs4_op_locku(nfs_argop4 *argop, nfs_resop4 *resop, 8732 struct svc_req *req, struct compound_state *cs) 8733 { 8734 LOCKU4args *args = &argop->nfs_argop4_u.oplocku; 8735 LOCKU4res *resp = &resop->nfs_resop4_u.oplocku; 8736 nfsstat4 status; 8737 stateid4 *stateid = &args->lock_stateid; 8738 rfs4_lo_state_t *lsp; 8739 8740 DTRACE_NFSV4_2(op__locku__start, struct compound_state *, cs, 8741 LOCKU4args *, args); 8742 8743 if (cs->vp == NULL) { 8744 *cs->statusp = resp->status = NFS4ERR_NOFILEHANDLE; 8745 DTRACE_NFSV4_2(op__locku__done, struct compound_state *, cs, 8746 LOCKU4res *, resp); 8747 return; 8748 } 8749 8750 if ((status = rfs4_get_lo_state(stateid, &lsp, TRUE)) != NFS4_OK) { 8751 *cs->statusp = resp->status = status; 8752 DTRACE_NFSV4_2(op__locku__done, struct compound_state *, cs, 8753 LOCKU4res *, resp); 8754 return; 8755 } 8756 8757 /* Ensure specified filehandle matches */ 8758 if (cs->vp != lsp->state->finfo->vp) { 8759 rfs4_lo_state_rele(lsp, TRUE); 8760 *cs->statusp = resp->status = NFS4ERR_BAD_STATEID; 8761 DTRACE_NFSV4_2(op__locku__done, struct compound_state *, cs, 8762 LOCKU4res *, resp); 8763 return; 8764 } 8765 8766 /* hold off other access to lsp while we tinker */ 8767 rfs4_sw_enter(&lsp->ls_sw); 8768 8769 switch (rfs4_check_lo_stateid_seqid(lsp, stateid)) { 8770 case NFS4_CHECK_STATEID_OKAY: 8771 if (rfs4_check_lock_seqid(args->seqid, lsp, resop) 8772 != NFS4_CHKSEQ_OKAY) { 8773 *cs->statusp = resp->status = NFS4ERR_BAD_SEQID; 8774 goto end; 8775 } 8776 break; 8777 case NFS4_CHECK_STATEID_OLD: 8778 *cs->statusp = resp->status = NFS4ERR_OLD_STATEID; 8779 goto end; 8780 case NFS4_CHECK_STATEID_BAD: 8781 *cs->statusp = resp->status = NFS4ERR_BAD_STATEID; 8782 goto end; 8783 case NFS4_CHECK_STATEID_EXPIRED: 8784 *cs->statusp = resp->status = NFS4ERR_EXPIRED; 8785 goto end; 8786 case NFS4_CHECK_STATEID_CLOSED: 8787 *cs->statusp = resp->status = NFS4ERR_OLD_STATEID; 8788 goto end; 8789 case NFS4_CHECK_STATEID_REPLAY: 8790 switch (rfs4_check_lock_seqid(args->seqid, lsp, resop)) { 8791 case NFS4_CHKSEQ_OKAY: 8792 /* 8793 * This is a replayed stateid; if 8794 * seqid matches the next expected, 8795 * then client is using wrong seqid. 8796 */ 8797 case NFS4_CHKSEQ_BAD: 8798 *cs->statusp = resp->status = NFS4ERR_BAD_SEQID; 8799 goto end; 8800 case NFS4_CHKSEQ_REPLAY: 8801 rfs4_update_lease(lsp->locker->client); 8802 *cs->statusp = status = resp->status; 8803 goto end; 8804 } 8805 break; 8806 default: 8807 ASSERT(FALSE); 8808 break; 8809 } 8810 8811 rfs4_update_lock_sequence(lsp); 8812 rfs4_update_lease(lsp->locker->client); 8813 8814 /* 8815 * NFS4 only allows locking on regular files, so 8816 * verify type of object. 8817 */ 8818 if (cs->vp->v_type != VREG) { 8819 if (cs->vp->v_type == VDIR) 8820 status = NFS4ERR_ISDIR; 8821 else 8822 status = NFS4ERR_INVAL; 8823 goto out; 8824 } 8825 8826 if (rfs4_clnt_in_grace(lsp->state->owner->client)) { 8827 status = NFS4ERR_GRACE; 8828 goto out; 8829 } 8830 8831 status = rfs4_do_lock(lsp, args->locktype, 8832 args->seqid, args->offset, args->length, cs->cr, resop); 8833 8834 out: 8835 *cs->statusp = resp->status = status; 8836 8837 if (status == NFS4_OK) 8838 resp->lock_stateid = lsp->lockid.stateid; 8839 8840 rfs4_update_lock_resp(lsp, resop); 8841 8842 end: 8843 rfs4_sw_exit(&lsp->ls_sw); 8844 rfs4_lo_state_rele(lsp, TRUE); 8845 8846 DTRACE_NFSV4_2(op__locku__done, struct compound_state *, cs, 8847 LOCKU4res *, resp); 8848 } 8849 8850 /* 8851 * LOCKT is a best effort routine, the client can not be guaranteed that 8852 * the status return is still in effect by the time the reply is received. 8853 * They are numerous race conditions in this routine, but we are not required 8854 * and can not be accurate. 8855 */ 8856 /*ARGSUSED*/ 8857 void 8858 rfs4_op_lockt(nfs_argop4 *argop, nfs_resop4 *resop, 8859 struct svc_req *req, struct compound_state *cs) 8860 { 8861 LOCKT4args *args = &argop->nfs_argop4_u.oplockt; 8862 LOCKT4res *resp = &resop->nfs_resop4_u.oplockt; 8863 rfs4_lockowner_t *lo; 8864 rfs4_client_t *cp; 8865 bool_t create = FALSE; 8866 struct flock64 flk; 8867 int error; 8868 int flag = FREAD | FWRITE; 8869 int ltype; 8870 length4 posix_length; 8871 sysid_t sysid; 8872 pid_t pid; 8873 8874 DTRACE_NFSV4_2(op__lockt__start, struct compound_state *, cs, 8875 LOCKT4args *, args); 8876 8877 if (cs->vp == NULL) { 8878 *cs->statusp = resp->status = NFS4ERR_NOFILEHANDLE; 8879 goto out; 8880 } 8881 8882 /* 8883 * NFS4 only allows locking on regular files, so 8884 * verify type of object. 8885 */ 8886 if (cs->vp->v_type != VREG) { 8887 if (cs->vp->v_type == VDIR) 8888 *cs->statusp = resp->status = NFS4ERR_ISDIR; 8889 else 8890 *cs->statusp = resp->status = NFS4ERR_INVAL; 8891 goto out; 8892 } 8893 8894 /* 8895 * Check out the clientid to ensure the server knows about it 8896 * so that we correctly inform the client of a server reboot. 8897 */ 8898 if ((cp = rfs4_findclient_by_id(args->owner.clientid, FALSE)) 8899 == NULL) { 8900 *cs->statusp = resp->status = 8901 rfs4_check_clientid(&args->owner.clientid, 0); 8902 goto out; 8903 } 8904 if (rfs4_lease_expired(cp)) { 8905 rfs4_client_close(cp); 8906 /* 8907 * Protocol doesn't allow returning NFS4ERR_STALE as 8908 * other operations do on this check so STALE_CLIENTID 8909 * is returned instead 8910 */ 8911 *cs->statusp = resp->status = NFS4ERR_STALE_CLIENTID; 8912 goto out; 8913 } 8914 8915 if (rfs4_clnt_in_grace(cp) && !(cp->can_reclaim)) { 8916 *cs->statusp = resp->status = NFS4ERR_GRACE; 8917 rfs4_client_rele(cp); 8918 goto out; 8919 } 8920 rfs4_client_rele(cp); 8921 8922 resp->status = NFS4_OK; 8923 8924 switch (args->locktype) { 8925 case READ_LT: 8926 case READW_LT: 8927 ltype = F_RDLCK; 8928 break; 8929 case WRITE_LT: 8930 case WRITEW_LT: 8931 ltype = F_WRLCK; 8932 break; 8933 } 8934 8935 posix_length = args->length; 8936 /* Check for zero length. To lock to end of file use all ones for V4 */ 8937 if (posix_length == 0) { 8938 *cs->statusp = resp->status = NFS4ERR_INVAL; 8939 goto out; 8940 } else if (posix_length == (length4)(~0)) { 8941 posix_length = 0; /* Posix to end of file */ 8942 } 8943 8944 /* Find or create a lockowner */ 8945 lo = rfs4_findlockowner(&args->owner, &create); 8946 8947 if (lo) { 8948 pid = lo->pid; 8949 if ((resp->status = 8950 rfs4_client_sysid(lo->client, &sysid)) != NFS4_OK) 8951 goto err; 8952 } else { 8953 pid = 0; 8954 sysid = lockt_sysid; 8955 } 8956 retry: 8957 flk.l_type = ltype; 8958 flk.l_whence = 0; /* SEEK_SET */ 8959 flk.l_start = args->offset; 8960 flk.l_len = posix_length; 8961 flk.l_sysid = sysid; 8962 flk.l_pid = pid; 8963 flag |= F_REMOTELOCK; 8964 8965 LOCK_PRINT(rfs4_debug, "rfs4_op_lockt", F_GETLK, &flk); 8966 8967 /* Note that length4 is uint64_t but l_len and l_start are off64_t */ 8968 if (flk.l_len < 0 || flk.l_start < 0) { 8969 resp->status = NFS4ERR_INVAL; 8970 goto err; 8971 } 8972 error = VOP_FRLOCK(cs->vp, F_GETLK, &flk, flag, (u_offset_t)0, 8973 NULL, cs->cr, NULL); 8974 8975 /* 8976 * N.B. We map error values to nfsv4 errors. This is differrent 8977 * than puterrno4 routine. 8978 */ 8979 switch (error) { 8980 case 0: 8981 if (flk.l_type == F_UNLCK) 8982 resp->status = NFS4_OK; 8983 else { 8984 if (lock_denied(&resp->denied, &flk) == NFS4ERR_EXPIRED) 8985 goto retry; 8986 resp->status = NFS4ERR_DENIED; 8987 } 8988 break; 8989 case EOVERFLOW: 8990 resp->status = NFS4ERR_INVAL; 8991 break; 8992 case EINVAL: 8993 resp->status = NFS4ERR_NOTSUPP; 8994 break; 8995 default: 8996 cmn_err(CE_WARN, "rfs4_op_lockt: unexpected errno (%d)", 8997 error); 8998 resp->status = NFS4ERR_SERVERFAULT; 8999 break; 9000 } 9001 9002 err: 9003 if (lo) 9004 rfs4_lockowner_rele(lo); 9005 *cs->statusp = resp->status; 9006 out: 9007 DTRACE_NFSV4_2(op__lockt__done, struct compound_state *, cs, 9008 LOCKT4res *, resp); 9009 } 9010 9011 static int 9012 vop_shrlock(vnode_t *vp, int cmd, struct shrlock *sp, int fflags) 9013 { 9014 int err; 9015 9016 if (cmd == F_UNSHARE && sp->s_deny == 0 && sp->s_access == 0) 9017 return (0); 9018 9019 err = VOP_SHRLOCK(vp, cmd, sp, fflags, CRED(), NULL); 9020 9021 NFS4_DEBUG(rfs4_shrlock_debug, 9022 (CE_NOTE, "rfs4_shrlock %s vp=%p acc=%d dny=%d sysid=%d " 9023 "pid=%d err=%d\n", cmd == F_UNSHARE ? "UNSHR" : "SHARE", 9024 (void *) vp, sp->s_access, sp->s_deny, sp->s_sysid, sp->s_pid, 9025 err)); 9026 9027 return (err); 9028 } 9029 9030 static int 9031 rfs4_shrlock(rfs4_state_t *sp, int cmd) 9032 { 9033 struct shrlock shr; 9034 struct shr_locowner shr_loco; 9035 int fflags; 9036 9037 fflags = shr.s_access = shr.s_deny = 0; 9038 9039 if (sp->share_access & OPEN4_SHARE_ACCESS_READ) { 9040 fflags |= FREAD; 9041 shr.s_access |= F_RDACC; 9042 } 9043 if (sp->share_access & OPEN4_SHARE_ACCESS_WRITE) { 9044 fflags |= FWRITE; 9045 shr.s_access |= F_WRACC; 9046 } 9047 if (sp->share_deny & OPEN4_SHARE_DENY_READ) 9048 shr.s_deny |= F_RDDNY; 9049 if (sp->share_deny & OPEN4_SHARE_DENY_WRITE) 9050 shr.s_deny |= F_WRDNY; 9051 9052 shr.s_pid = rfs4_dbe_getid(sp->owner->dbe); 9053 shr.s_sysid = sp->owner->client->sysidt; 9054 shr_loco.sl_pid = shr.s_pid; 9055 shr_loco.sl_id = shr.s_sysid; 9056 shr.s_owner = (caddr_t)&shr_loco; 9057 shr.s_own_len = sizeof (shr_loco); 9058 return (vop_shrlock(sp->finfo->vp, cmd, &shr, fflags)); 9059 } 9060 9061 static int 9062 rfs4_share(rfs4_state_t *sp) 9063 { 9064 int cmd; 9065 cmd = nbl_need_check(sp->finfo->vp) ? F_SHARE_NBMAND : F_SHARE; 9066 return (rfs4_shrlock(sp, cmd)); 9067 } 9068 9069 void 9070 rfs4_unshare(rfs4_state_t *sp) 9071 { 9072 (void) rfs4_shrlock(sp, F_UNSHARE); 9073 } 9074