1 /* 2 * CDDL HEADER START 3 * 4 * The contents of this file are subject to the terms of the 5 * Common Development and Distribution License (the "License"). 6 * You may not use this file except in compliance with the License. 7 * 8 * You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE 9 * or http://www.opensolaris.org/os/licensing. 10 * See the License for the specific language governing permissions 11 * and limitations under the License. 12 * 13 * When distributing Covered Code, include this CDDL HEADER in each 14 * file and include the License file at usr/src/OPENSOLARIS.LICENSE. 15 * If applicable, add the following below this CDDL HEADER, with the 16 * fields enclosed by brackets "[]" replaced with your own identifying 17 * information: Portions Copyright [yyyy] [name of copyright owner] 18 * 19 * CDDL HEADER END 20 */ 21 22 /* 23 * Copyright (c) 2003, 2010, Oracle and/or its affiliates. All rights reserved. 24 */ 25 26 /* 27 * Copyright (c) 1983,1984,1985,1986,1987,1988,1989 AT&T. 28 * All Rights Reserved 29 */ 30 31 /* 32 * Copyright (c) 2012, 2016 by Delphix. All rights reserved. 33 * Copyright 2019 Nexenta Systems, Inc. 34 * Copyright 2019 Nexenta by DDN, Inc. 35 * Copyright 2021 Racktop Systems, Inc. 36 */ 37 38 #include <sys/param.h> 39 #include <sys/types.h> 40 #include <sys/systm.h> 41 #include <sys/cred.h> 42 #include <sys/buf.h> 43 #include <sys/vfs.h> 44 #include <sys/vfs_opreg.h> 45 #include <sys/vnode.h> 46 #include <sys/uio.h> 47 #include <sys/errno.h> 48 #include <sys/sysmacros.h> 49 #include <sys/statvfs.h> 50 #include <sys/kmem.h> 51 #include <sys/dirent.h> 52 #include <sys/cmn_err.h> 53 #include <sys/debug.h> 54 #include <sys/systeminfo.h> 55 #include <sys/flock.h> 56 #include <sys/pathname.h> 57 #include <sys/nbmlock.h> 58 #include <sys/share.h> 59 #include <sys/atomic.h> 60 #include <sys/policy.h> 61 #include <sys/fem.h> 62 #include <sys/sdt.h> 63 #include <sys/ddi.h> 64 #include <sys/zone.h> 65 66 #include <fs/fs_reparse.h> 67 68 #include <rpc/types.h> 69 #include <rpc/auth.h> 70 #include <rpc/rpcsec_gss.h> 71 #include <rpc/svc.h> 72 73 #include <nfs/nfs.h> 74 #include <nfs/nfssys.h> 75 #include <nfs/export.h> 76 #include <nfs/nfs_cmd.h> 77 #include <nfs/lm.h> 78 #include <nfs/nfs4.h> 79 #include <nfs/nfs4_drc.h> 80 81 #include <sys/strsubr.h> 82 #include <sys/strsun.h> 83 84 #include <inet/common.h> 85 #include <inet/ip.h> 86 #include <inet/ip6.h> 87 88 #include <sys/tsol/label.h> 89 #include <sys/tsol/tndb.h> 90 91 #define RFS4_MAXLOCK_TRIES 4 /* Try to get the lock this many times */ 92 static int rfs4_maxlock_tries = RFS4_MAXLOCK_TRIES; 93 #define RFS4_LOCK_DELAY 10 /* Milliseconds */ 94 static clock_t rfs4_lock_delay = RFS4_LOCK_DELAY; 95 extern struct svc_ops rdma_svc_ops; 96 extern int nfs_loaned_buffers; 97 /* End of Tunables */ 98 99 static int rdma_setup_read_data4(READ4args *, READ4res *); 100 101 /* 102 * Used to bump the stateid4.seqid value and show changes in the stateid 103 */ 104 #define next_stateid(sp) (++(sp)->bits.chgseq) 105 106 /* 107 * RFS4_MINLEN_ENTRY4: XDR-encoded size of smallest possible dirent. 108 * This is used to return NFS4ERR_TOOSMALL when clients specify 109 * maxcount that isn't large enough to hold the smallest possible 110 * XDR encoded dirent. 111 * 112 * sizeof cookie (8 bytes) + 113 * sizeof name_len (4 bytes) + 114 * sizeof smallest (padded) name (4 bytes) + 115 * sizeof bitmap4_len (12 bytes) + NOTE: we always encode len=2 bm4 116 * sizeof attrlist4_len (4 bytes) + 117 * sizeof next boolean (4 bytes) 118 * 119 * RFS4_MINLEN_RDDIR4: XDR-encoded size of READDIR op reply containing 120 * the smallest possible entry4 (assumes no attrs requested). 121 * sizeof nfsstat4 (4 bytes) + 122 * sizeof verifier4 (8 bytes) + 123 * sizeof entry4list bool (4 bytes) + 124 * sizeof entry4 (36 bytes) + 125 * sizeof eof bool (4 bytes) 126 * 127 * RFS4_MINLEN_RDDIR_BUF: minimum length of buffer server will provide to 128 * VOP_READDIR. Its value is the size of the maximum possible dirent 129 * for solaris. The DIRENT64_RECLEN macro returns the size of dirent 130 * required for a given name length. MAXNAMELEN is the maximum 131 * filename length allowed in Solaris. The first two DIRENT64_RECLEN() 132 * macros are to allow for . and .. entries -- just a minor tweak to try 133 * and guarantee that buffer we give to VOP_READDIR will be large enough 134 * to hold ., .., and the largest possible solaris dirent64. 135 */ 136 #define RFS4_MINLEN_ENTRY4 36 137 #define RFS4_MINLEN_RDDIR4 (4 + NFS4_VERIFIER_SIZE + 4 + RFS4_MINLEN_ENTRY4 + 4) 138 #define RFS4_MINLEN_RDDIR_BUF \ 139 (DIRENT64_RECLEN(1) + DIRENT64_RECLEN(2) + DIRENT64_RECLEN(MAXNAMELEN)) 140 141 /* 142 * It would be better to pad to 4 bytes since that's what XDR would do, 143 * but the dirents UFS gives us are already padded to 8, so just take 144 * what we're given. Dircount is only a hint anyway. Currently the 145 * solaris kernel is ASCII only, so there's no point in calling the 146 * UTF8 functions. 147 * 148 * dirent64: named padded to provide 8 byte struct alignment 149 * d_ino(8) + d_off(8) + d_reclen(2) + d_name(namelen + null(1) + pad) 150 * 151 * cookie: uint64_t + utf8namelen: uint_t + utf8name padded to 8 bytes 152 * 153 */ 154 #define DIRENT64_TO_DIRCOUNT(dp) \ 155 (3 * BYTES_PER_XDR_UNIT + DIRENT64_NAMELEN((dp)->d_reclen)) 156 157 158 static sysid_t lockt_sysid; /* dummy sysid for all LOCKT calls */ 159 160 u_longlong_t nfs4_srv_caller_id; 161 uint_t nfs4_srv_vkey = 0; 162 163 void rfs4_init_compound_state(struct compound_state *); 164 165 static void nullfree(caddr_t); 166 static void rfs4_op_inval(nfs_argop4 *, nfs_resop4 *, struct svc_req *, 167 struct compound_state *); 168 static void rfs4_op_access(nfs_argop4 *, nfs_resop4 *, struct svc_req *, 169 struct compound_state *); 170 static void rfs4_op_close(nfs_argop4 *, nfs_resop4 *, struct svc_req *, 171 struct compound_state *); 172 static void rfs4_op_commit(nfs_argop4 *, nfs_resop4 *, struct svc_req *, 173 struct compound_state *); 174 static void rfs4_op_create(nfs_argop4 *, nfs_resop4 *, struct svc_req *, 175 struct compound_state *); 176 static void rfs4_op_create_free(nfs_resop4 *resop); 177 static void rfs4_op_delegreturn(nfs_argop4 *, nfs_resop4 *, 178 struct svc_req *, struct compound_state *); 179 static void rfs4_op_delegpurge(nfs_argop4 *, nfs_resop4 *, 180 struct svc_req *, struct compound_state *); 181 static void rfs4_op_getattr(nfs_argop4 *, nfs_resop4 *, struct svc_req *, 182 struct compound_state *); 183 static void rfs4_op_getattr_free(nfs_resop4 *); 184 static void rfs4_op_getfh(nfs_argop4 *, nfs_resop4 *, struct svc_req *, 185 struct compound_state *); 186 static void rfs4_op_getfh_free(nfs_resop4 *); 187 static void rfs4_op_illegal(nfs_argop4 *, nfs_resop4 *, struct svc_req *, 188 struct compound_state *); 189 static void rfs4_op_link(nfs_argop4 *, nfs_resop4 *, struct svc_req *, 190 struct compound_state *); 191 static void rfs4_op_lock(nfs_argop4 *, nfs_resop4 *, struct svc_req *, 192 struct compound_state *); 193 static void lock_denied_free(nfs_resop4 *); 194 static void rfs4_op_locku(nfs_argop4 *, nfs_resop4 *, struct svc_req *, 195 struct compound_state *); 196 static void rfs4_op_lockt(nfs_argop4 *, nfs_resop4 *, struct svc_req *, 197 struct compound_state *); 198 static void rfs4_op_lookup(nfs_argop4 *, nfs_resop4 *, struct svc_req *, 199 struct compound_state *); 200 static void rfs4_op_lookupp(nfs_argop4 *, nfs_resop4 *, struct svc_req *, 201 struct compound_state *); 202 static void rfs4_op_openattr(nfs_argop4 *argop, nfs_resop4 *resop, 203 struct svc_req *req, struct compound_state *cs); 204 static void rfs4_op_nverify(nfs_argop4 *, nfs_resop4 *, struct svc_req *, 205 struct compound_state *); 206 static void rfs4_op_open(nfs_argop4 *, nfs_resop4 *, struct svc_req *, 207 struct compound_state *); 208 static void rfs4_op_open_confirm(nfs_argop4 *, nfs_resop4 *, 209 struct svc_req *, struct compound_state *); 210 static void rfs4_op_open_downgrade(nfs_argop4 *, nfs_resop4 *, 211 struct svc_req *, struct compound_state *); 212 static void rfs4_op_putfh(nfs_argop4 *, nfs_resop4 *, struct svc_req *, 213 struct compound_state *); 214 static void rfs4_op_putpubfh(nfs_argop4 *, nfs_resop4 *, struct svc_req *, 215 struct compound_state *); 216 static void rfs4_op_putrootfh(nfs_argop4 *, nfs_resop4 *, struct svc_req *, 217 struct compound_state *); 218 static void rfs4_op_read(nfs_argop4 *, nfs_resop4 *, struct svc_req *, 219 struct compound_state *); 220 static void rfs4_op_read_free(nfs_resop4 *); 221 static void rfs4_op_readdir_free(nfs_resop4 *resop); 222 static void rfs4_op_readlink(nfs_argop4 *, nfs_resop4 *, struct svc_req *, 223 struct compound_state *); 224 static void rfs4_op_readlink_free(nfs_resop4 *); 225 static void rfs4_op_release_lockowner(nfs_argop4 *, nfs_resop4 *, 226 struct svc_req *, struct compound_state *); 227 static void rfs4_op_remove(nfs_argop4 *, nfs_resop4 *, struct svc_req *, 228 struct compound_state *); 229 static void rfs4_op_rename(nfs_argop4 *, nfs_resop4 *, struct svc_req *, 230 struct compound_state *); 231 static void rfs4_op_renew(nfs_argop4 *, nfs_resop4 *, struct svc_req *, 232 struct compound_state *); 233 static void rfs4_op_restorefh(nfs_argop4 *, nfs_resop4 *, struct svc_req *, 234 struct compound_state *); 235 static void rfs4_op_savefh(nfs_argop4 *, nfs_resop4 *, struct svc_req *, 236 struct compound_state *); 237 static void rfs4_op_setattr(nfs_argop4 *, nfs_resop4 *, struct svc_req *, 238 struct compound_state *); 239 static void rfs4_op_verify(nfs_argop4 *, nfs_resop4 *, struct svc_req *, 240 struct compound_state *); 241 static void rfs4_op_write(nfs_argop4 *, nfs_resop4 *, struct svc_req *, 242 struct compound_state *); 243 static void rfs4_op_setclientid(nfs_argop4 *, nfs_resop4 *, 244 struct svc_req *, struct compound_state *); 245 static void rfs4_op_setclientid_confirm(nfs_argop4 *, nfs_resop4 *, 246 struct svc_req *req, struct compound_state *); 247 static void rfs4_op_secinfo(nfs_argop4 *, nfs_resop4 *, struct svc_req *, 248 struct compound_state *); 249 static void rfs4_op_secinfo_free(nfs_resop4 *); 250 251 static nfsstat4 check_open_access(uint32_t, struct compound_state *, 252 struct svc_req *); 253 nfsstat4 rfs4_client_sysid(rfs4_client_t *, sysid_t *); 254 void rfs4_ss_clid(nfs4_srv_t *, rfs4_client_t *); 255 256 257 /* 258 * translation table for attrs 259 */ 260 struct nfs4_ntov_table { 261 union nfs4_attr_u *na; 262 uint8_t amap[NFS4_MAXNUM_ATTRS]; 263 int attrcnt; 264 bool_t vfsstat; 265 }; 266 267 static void nfs4_ntov_table_init(struct nfs4_ntov_table *ntovp); 268 static void nfs4_ntov_table_free(struct nfs4_ntov_table *ntovp, 269 struct nfs4_svgetit_arg *sargp); 270 271 static nfsstat4 do_rfs4_set_attrs(bitmap4 *resp, fattr4 *fattrp, 272 struct compound_state *cs, struct nfs4_svgetit_arg *sargp, 273 struct nfs4_ntov_table *ntovp, nfs4_attr_cmd_t cmd); 274 275 static void hanfsv4_failover(nfs4_srv_t *); 276 277 fem_t *deleg_rdops; 278 fem_t *deleg_wrops; 279 280 /* 281 * NFS4 op dispatch table 282 */ 283 284 struct rfsv4disp { 285 void (*dis_proc)(); /* proc to call */ 286 void (*dis_resfree)(); /* frees space allocated by proc */ 287 int dis_flags; /* RPC_IDEMPOTENT, etc... */ 288 }; 289 290 static struct rfsv4disp rfsv4disptab[] = { 291 /* 292 * NFS VERSION 4 293 */ 294 295 /* RFS_NULL = 0 */ 296 {rfs4_op_illegal, nullfree, 0}, 297 298 /* UNUSED = 1 */ 299 {rfs4_op_illegal, nullfree, 0}, 300 301 /* UNUSED = 2 */ 302 {rfs4_op_illegal, nullfree, 0}, 303 304 /* OP_ACCESS = 3 */ 305 {rfs4_op_access, nullfree, RPC_IDEMPOTENT}, 306 307 /* OP_CLOSE = 4 */ 308 {rfs4_op_close, nullfree, 0}, 309 310 /* OP_COMMIT = 5 */ 311 {rfs4_op_commit, nullfree, RPC_IDEMPOTENT}, 312 313 /* OP_CREATE = 6 */ 314 {rfs4_op_create, nullfree, 0}, 315 316 /* OP_DELEGPURGE = 7 */ 317 {rfs4_op_delegpurge, nullfree, 0}, 318 319 /* OP_DELEGRETURN = 8 */ 320 {rfs4_op_delegreturn, nullfree, 0}, 321 322 /* OP_GETATTR = 9 */ 323 {rfs4_op_getattr, rfs4_op_getattr_free, RPC_IDEMPOTENT}, 324 325 /* OP_GETFH = 10 */ 326 {rfs4_op_getfh, rfs4_op_getfh_free, RPC_ALL}, 327 328 /* OP_LINK = 11 */ 329 {rfs4_op_link, nullfree, 0}, 330 331 /* OP_LOCK = 12 */ 332 {rfs4_op_lock, lock_denied_free, 0}, 333 334 /* OP_LOCKT = 13 */ 335 {rfs4_op_lockt, lock_denied_free, 0}, 336 337 /* OP_LOCKU = 14 */ 338 {rfs4_op_locku, nullfree, 0}, 339 340 /* OP_LOOKUP = 15 */ 341 {rfs4_op_lookup, nullfree, (RPC_IDEMPOTENT | RPC_PUBLICFH_OK)}, 342 343 /* OP_LOOKUPP = 16 */ 344 {rfs4_op_lookupp, nullfree, (RPC_IDEMPOTENT | RPC_PUBLICFH_OK)}, 345 346 /* OP_NVERIFY = 17 */ 347 {rfs4_op_nverify, nullfree, RPC_IDEMPOTENT}, 348 349 /* OP_OPEN = 18 */ 350 {rfs4_op_open, rfs4_free_reply, 0}, 351 352 /* OP_OPENATTR = 19 */ 353 {rfs4_op_openattr, nullfree, 0}, 354 355 /* OP_OPEN_CONFIRM = 20 */ 356 {rfs4_op_open_confirm, nullfree, 0}, 357 358 /* OP_OPEN_DOWNGRADE = 21 */ 359 {rfs4_op_open_downgrade, nullfree, 0}, 360 361 /* OP_OPEN_PUTFH = 22 */ 362 {rfs4_op_putfh, nullfree, RPC_ALL}, 363 364 /* OP_PUTPUBFH = 23 */ 365 {rfs4_op_putpubfh, nullfree, RPC_ALL}, 366 367 /* OP_PUTROOTFH = 24 */ 368 {rfs4_op_putrootfh, nullfree, RPC_ALL}, 369 370 /* OP_READ = 25 */ 371 {rfs4_op_read, rfs4_op_read_free, RPC_IDEMPOTENT}, 372 373 /* OP_READDIR = 26 */ 374 {rfs4_op_readdir, rfs4_op_readdir_free, RPC_IDEMPOTENT}, 375 376 /* OP_READLINK = 27 */ 377 {rfs4_op_readlink, rfs4_op_readlink_free, RPC_IDEMPOTENT}, 378 379 /* OP_REMOVE = 28 */ 380 {rfs4_op_remove, nullfree, 0}, 381 382 /* OP_RENAME = 29 */ 383 {rfs4_op_rename, nullfree, 0}, 384 385 /* OP_RENEW = 30 */ 386 {rfs4_op_renew, nullfree, 0}, 387 388 /* OP_RESTOREFH = 31 */ 389 {rfs4_op_restorefh, nullfree, RPC_ALL}, 390 391 /* OP_SAVEFH = 32 */ 392 {rfs4_op_savefh, nullfree, RPC_ALL}, 393 394 /* OP_SECINFO = 33 */ 395 {rfs4_op_secinfo, rfs4_op_secinfo_free, 0}, 396 397 /* OP_SETATTR = 34 */ 398 {rfs4_op_setattr, nullfree, 0}, 399 400 /* OP_SETCLIENTID = 35 */ 401 {rfs4_op_setclientid, nullfree, 0}, 402 403 /* OP_SETCLIENTID_CONFIRM = 36 */ 404 {rfs4_op_setclientid_confirm, nullfree, 0}, 405 406 /* OP_VERIFY = 37 */ 407 {rfs4_op_verify, nullfree, RPC_IDEMPOTENT}, 408 409 /* OP_WRITE = 38 */ 410 {rfs4_op_write, nullfree, 0}, 411 412 /* OP_RELEASE_LOCKOWNER = 39 */ 413 {rfs4_op_release_lockowner, nullfree, 0}, 414 }; 415 416 static uint_t rfsv4disp_cnt = sizeof (rfsv4disptab) / sizeof (rfsv4disptab[0]); 417 418 #define OP_ILLEGAL_IDX (rfsv4disp_cnt) 419 420 #ifdef DEBUG 421 422 int rfs4_fillone_debug = 0; 423 int rfs4_no_stub_access = 1; 424 int rfs4_rddir_debug = 0; 425 426 static char *rfs4_op_string[] = { 427 "rfs4_op_null", 428 "rfs4_op_1 unused", 429 "rfs4_op_2 unused", 430 "rfs4_op_access", 431 "rfs4_op_close", 432 "rfs4_op_commit", 433 "rfs4_op_create", 434 "rfs4_op_delegpurge", 435 "rfs4_op_delegreturn", 436 "rfs4_op_getattr", 437 "rfs4_op_getfh", 438 "rfs4_op_link", 439 "rfs4_op_lock", 440 "rfs4_op_lockt", 441 "rfs4_op_locku", 442 "rfs4_op_lookup", 443 "rfs4_op_lookupp", 444 "rfs4_op_nverify", 445 "rfs4_op_open", 446 "rfs4_op_openattr", 447 "rfs4_op_open_confirm", 448 "rfs4_op_open_downgrade", 449 "rfs4_op_putfh", 450 "rfs4_op_putpubfh", 451 "rfs4_op_putrootfh", 452 "rfs4_op_read", 453 "rfs4_op_readdir", 454 "rfs4_op_readlink", 455 "rfs4_op_remove", 456 "rfs4_op_rename", 457 "rfs4_op_renew", 458 "rfs4_op_restorefh", 459 "rfs4_op_savefh", 460 "rfs4_op_secinfo", 461 "rfs4_op_setattr", 462 "rfs4_op_setclientid", 463 "rfs4_op_setclient_confirm", 464 "rfs4_op_verify", 465 "rfs4_op_write", 466 "rfs4_op_release_lockowner", 467 "rfs4_op_illegal" 468 }; 469 #endif 470 471 void rfs4_ss_chkclid(nfs4_srv_t *, rfs4_client_t *); 472 473 extern size_t strlcpy(char *dst, const char *src, size_t dstsize); 474 475 extern void rfs4_free_fs_locations4(fs_locations4 *); 476 477 #ifdef nextdp 478 #undef nextdp 479 #endif 480 #define nextdp(dp) ((struct dirent64 *)((char *)(dp) + (dp)->d_reclen)) 481 482 static const fs_operation_def_t nfs4_rd_deleg_tmpl[] = { 483 VOPNAME_OPEN, { .femop_open = deleg_rd_open }, 484 VOPNAME_WRITE, { .femop_write = deleg_rd_write }, 485 VOPNAME_SETATTR, { .femop_setattr = deleg_rd_setattr }, 486 VOPNAME_RWLOCK, { .femop_rwlock = deleg_rd_rwlock }, 487 VOPNAME_SPACE, { .femop_space = deleg_rd_space }, 488 VOPNAME_SETSECATTR, { .femop_setsecattr = deleg_rd_setsecattr }, 489 VOPNAME_VNEVENT, { .femop_vnevent = deleg_rd_vnevent }, 490 NULL, NULL 491 }; 492 static const fs_operation_def_t nfs4_wr_deleg_tmpl[] = { 493 VOPNAME_OPEN, { .femop_open = deleg_wr_open }, 494 VOPNAME_READ, { .femop_read = deleg_wr_read }, 495 VOPNAME_WRITE, { .femop_write = deleg_wr_write }, 496 VOPNAME_SETATTR, { .femop_setattr = deleg_wr_setattr }, 497 VOPNAME_RWLOCK, { .femop_rwlock = deleg_wr_rwlock }, 498 VOPNAME_SPACE, { .femop_space = deleg_wr_space }, 499 VOPNAME_SETSECATTR, { .femop_setsecattr = deleg_wr_setsecattr }, 500 VOPNAME_VNEVENT, { .femop_vnevent = deleg_wr_vnevent }, 501 NULL, NULL 502 }; 503 504 nfs4_srv_t * 505 nfs4_get_srv(void) 506 { 507 nfs_globals_t *ng = nfs_srv_getzg(); 508 nfs4_srv_t *srv = ng->nfs4_srv; 509 ASSERT(srv != NULL); 510 return (srv); 511 } 512 513 void 514 rfs4_srv_zone_init(nfs_globals_t *ng) 515 { 516 nfs4_srv_t *nsrv4; 517 timespec32_t verf; 518 519 nsrv4 = kmem_zalloc(sizeof (*nsrv4), KM_SLEEP); 520 521 /* 522 * The following algorithm attempts to find a unique verifier 523 * to be used as the write verifier returned from the server 524 * to the client. It is important that this verifier change 525 * whenever the server reboots. Of secondary importance, it 526 * is important for the verifier to be unique between two 527 * different servers. 528 * 529 * Thus, an attempt is made to use the system hostid and the 530 * current time in seconds when the nfssrv kernel module is 531 * loaded. It is assumed that an NFS server will not be able 532 * to boot and then to reboot in less than a second. If the 533 * hostid has not been set, then the current high resolution 534 * time is used. This will ensure different verifiers each 535 * time the server reboots and minimize the chances that two 536 * different servers will have the same verifier. 537 * XXX - this is broken on LP64 kernels. 538 */ 539 verf.tv_sec = (time_t)zone_get_hostid(NULL); 540 if (verf.tv_sec != 0) { 541 verf.tv_nsec = gethrestime_sec(); 542 } else { 543 timespec_t tverf; 544 545 gethrestime(&tverf); 546 verf.tv_sec = (time_t)tverf.tv_sec; 547 verf.tv_nsec = tverf.tv_nsec; 548 } 549 nsrv4->write4verf = *(uint64_t *)&verf; 550 551 /* Used to manage create/destroy of server state */ 552 nsrv4->nfs4_server_state = NULL; 553 nsrv4->nfs4_cur_servinst = NULL; 554 nsrv4->nfs4_deleg_policy = SRV_NEVER_DELEGATE; 555 mutex_init(&nsrv4->deleg_lock, NULL, MUTEX_DEFAULT, NULL); 556 mutex_init(&nsrv4->state_lock, NULL, MUTEX_DEFAULT, NULL); 557 mutex_init(&nsrv4->servinst_lock, NULL, MUTEX_DEFAULT, NULL); 558 rw_init(&nsrv4->deleg_policy_lock, NULL, RW_DEFAULT, NULL); 559 560 ng->nfs4_srv = nsrv4; 561 } 562 563 void 564 rfs4_srv_zone_fini(nfs_globals_t *ng) 565 { 566 nfs4_srv_t *nsrv4 = ng->nfs4_srv; 567 568 ng->nfs4_srv = NULL; 569 570 mutex_destroy(&nsrv4->deleg_lock); 571 mutex_destroy(&nsrv4->state_lock); 572 mutex_destroy(&nsrv4->servinst_lock); 573 rw_destroy(&nsrv4->deleg_policy_lock); 574 575 kmem_free(nsrv4, sizeof (*nsrv4)); 576 } 577 578 void 579 rfs4_srvrinit(void) 580 { 581 extern void rfs4_attr_init(); 582 583 rfs4_attr_init(); 584 585 if (fem_create("deleg_rdops", nfs4_rd_deleg_tmpl, &deleg_rdops) != 0) { 586 rfs4_disable_delegation(); 587 } else if (fem_create("deleg_wrops", nfs4_wr_deleg_tmpl, 588 &deleg_wrops) != 0) { 589 rfs4_disable_delegation(); 590 fem_free(deleg_rdops); 591 } 592 593 nfs4_srv_caller_id = fs_new_caller_id(); 594 lockt_sysid = lm_alloc_sysidt(); 595 vsd_create(&nfs4_srv_vkey, NULL); 596 rfs4_state_g_init(); 597 } 598 599 void 600 rfs4_srvrfini(void) 601 { 602 if (lockt_sysid != LM_NOSYSID) { 603 lm_free_sysidt(lockt_sysid); 604 lockt_sysid = LM_NOSYSID; 605 } 606 607 rfs4_state_g_fini(); 608 609 fem_free(deleg_rdops); 610 fem_free(deleg_wrops); 611 } 612 613 void 614 rfs4_do_server_start(int server_upordown, 615 int srv_delegation, int cluster_booted) 616 { 617 nfs4_srv_t *nsrv4 = nfs4_get_srv(); 618 619 /* Is this a warm start? */ 620 if (server_upordown == NFS_SERVER_QUIESCED) { 621 cmn_err(CE_NOTE, "nfs4_srv: " 622 "server was previously quiesced; " 623 "existing NFSv4 state will be re-used"); 624 625 /* 626 * HA-NFSv4: this is also the signal 627 * that a Resource Group failover has 628 * occurred. 629 */ 630 if (cluster_booted) 631 hanfsv4_failover(nsrv4); 632 } else { 633 /* Cold start */ 634 nsrv4->rfs4_start_time = 0; 635 rfs4_state_zone_init(nsrv4); 636 nsrv4->nfs4_drc = rfs4_init_drc(nfs4_drc_max, 637 nfs4_drc_hash); 638 639 /* 640 * The nfsd service was started with the -s option 641 * we need to pull in any state from the paths indicated. 642 */ 643 if (curzone == global_zone && rfs4_dss_numnewpaths > 0) { 644 /* read in the stable storage state from these paths */ 645 rfs4_dss_readstate(nsrv4, rfs4_dss_numnewpaths, 646 rfs4_dss_newpaths); 647 } 648 } 649 650 /* Check if delegation is to be enabled */ 651 if (srv_delegation != FALSE) 652 rfs4_set_deleg_policy(nsrv4, SRV_NORMAL_DELEGATE); 653 } 654 655 void 656 rfs4_init_compound_state(struct compound_state *cs) 657 { 658 bzero(cs, sizeof (*cs)); 659 cs->cont = TRUE; 660 cs->access = CS_ACCESS_DENIED; 661 cs->deleg = FALSE; 662 cs->mandlock = FALSE; 663 cs->fh.nfs_fh4_val = cs->fhbuf; 664 } 665 666 void 667 rfs4_grace_start(rfs4_servinst_t *sip) 668 { 669 rw_enter(&sip->rwlock, RW_WRITER); 670 sip->start_time = (time_t)TICK_TO_SEC(ddi_get_lbolt()); 671 sip->grace_period = rfs4_grace_period; 672 rw_exit(&sip->rwlock); 673 } 674 675 /* 676 * returns true if the instance's grace period has never been started 677 */ 678 int 679 rfs4_servinst_grace_new(rfs4_servinst_t *sip) 680 { 681 time_t start_time; 682 683 rw_enter(&sip->rwlock, RW_READER); 684 start_time = sip->start_time; 685 rw_exit(&sip->rwlock); 686 687 return (start_time == 0); 688 } 689 690 /* 691 * Indicates if server instance is within the 692 * grace period. 693 */ 694 int 695 rfs4_servinst_in_grace(rfs4_servinst_t *sip) 696 { 697 time_t grace_expiry; 698 699 rw_enter(&sip->rwlock, RW_READER); 700 grace_expiry = sip->start_time + sip->grace_period; 701 rw_exit(&sip->rwlock); 702 703 return (((time_t)TICK_TO_SEC(ddi_get_lbolt())) < grace_expiry); 704 } 705 706 int 707 rfs4_clnt_in_grace(rfs4_client_t *cp) 708 { 709 ASSERT(rfs4_dbe_refcnt(cp->rc_dbe) > 0); 710 711 return (rfs4_servinst_in_grace(cp->rc_server_instance)); 712 } 713 714 /* 715 * reset all currently active grace periods 716 */ 717 void 718 rfs4_grace_reset_all(nfs4_srv_t *nsrv4) 719 { 720 rfs4_servinst_t *sip; 721 722 mutex_enter(&nsrv4->servinst_lock); 723 for (sip = nsrv4->nfs4_cur_servinst; sip != NULL; sip = sip->prev) 724 if (rfs4_servinst_in_grace(sip)) 725 rfs4_grace_start(sip); 726 mutex_exit(&nsrv4->servinst_lock); 727 } 728 729 /* 730 * start any new instances' grace periods 731 */ 732 void 733 rfs4_grace_start_new(nfs4_srv_t *nsrv4) 734 { 735 rfs4_servinst_t *sip; 736 737 mutex_enter(&nsrv4->servinst_lock); 738 for (sip = nsrv4->nfs4_cur_servinst; sip != NULL; sip = sip->prev) 739 if (rfs4_servinst_grace_new(sip)) 740 rfs4_grace_start(sip); 741 mutex_exit(&nsrv4->servinst_lock); 742 } 743 744 static rfs4_dss_path_t * 745 rfs4_dss_newpath(nfs4_srv_t *nsrv4, rfs4_servinst_t *sip, 746 char *path, unsigned index) 747 { 748 size_t len; 749 rfs4_dss_path_t *dss_path; 750 751 dss_path = kmem_alloc(sizeof (rfs4_dss_path_t), KM_SLEEP); 752 753 /* 754 * Take a copy of the string, since the original may be overwritten. 755 * Sadly, no strdup() in the kernel. 756 */ 757 /* allow for NUL */ 758 len = strlen(path) + 1; 759 dss_path->path = kmem_alloc(len, KM_SLEEP); 760 (void) strlcpy(dss_path->path, path, len); 761 762 /* associate with servinst */ 763 dss_path->sip = sip; 764 dss_path->index = index; 765 766 /* 767 * Add to list of served paths. 768 * No locking required, as we're only ever called at startup. 769 */ 770 if (nsrv4->dss_pathlist == NULL) { 771 /* this is the first dss_path_t */ 772 773 /* needed for insque/remque */ 774 dss_path->next = dss_path->prev = dss_path; 775 776 nsrv4->dss_pathlist = dss_path; 777 } else { 778 insque(dss_path, nsrv4->dss_pathlist); 779 } 780 781 return (dss_path); 782 } 783 784 /* 785 * Create a new server instance, and make it the currently active instance. 786 * Note that starting the grace period too early will reduce the clients' 787 * recovery window. 788 */ 789 void 790 rfs4_servinst_create(nfs4_srv_t *nsrv4, int start_grace, 791 int dss_npaths, char **dss_paths) 792 { 793 unsigned i; 794 rfs4_servinst_t *sip; 795 rfs4_oldstate_t *oldstate; 796 797 sip = kmem_alloc(sizeof (rfs4_servinst_t), KM_SLEEP); 798 rw_init(&sip->rwlock, NULL, RW_DEFAULT, NULL); 799 800 sip->start_time = (time_t)0; 801 sip->grace_period = (time_t)0; 802 sip->next = NULL; 803 sip->prev = NULL; 804 805 rw_init(&sip->oldstate_lock, NULL, RW_DEFAULT, NULL); 806 /* 807 * This initial dummy entry is required to setup for insque/remque. 808 * It must be skipped over whenever the list is traversed. 809 */ 810 oldstate = kmem_alloc(sizeof (rfs4_oldstate_t), KM_SLEEP); 811 /* insque/remque require initial list entry to be self-terminated */ 812 oldstate->next = oldstate; 813 oldstate->prev = oldstate; 814 sip->oldstate = oldstate; 815 816 817 sip->dss_npaths = dss_npaths; 818 sip->dss_paths = kmem_alloc(dss_npaths * 819 sizeof (rfs4_dss_path_t *), KM_SLEEP); 820 821 for (i = 0; i < dss_npaths; i++) { 822 sip->dss_paths[i] = 823 rfs4_dss_newpath(nsrv4, sip, dss_paths[i], i); 824 } 825 826 mutex_enter(&nsrv4->servinst_lock); 827 if (nsrv4->nfs4_cur_servinst != NULL) { 828 /* add to linked list */ 829 sip->prev = nsrv4->nfs4_cur_servinst; 830 nsrv4->nfs4_cur_servinst->next = sip; 831 } 832 if (start_grace) 833 rfs4_grace_start(sip); 834 /* make the new instance "current" */ 835 nsrv4->nfs4_cur_servinst = sip; 836 837 mutex_exit(&nsrv4->servinst_lock); 838 } 839 840 /* 841 * In future, we might add a rfs4_servinst_destroy(sip) but, for now, destroy 842 * all instances directly. 843 */ 844 void 845 rfs4_servinst_destroy_all(nfs4_srv_t *nsrv4) 846 { 847 rfs4_servinst_t *sip, *prev, *current; 848 #ifdef DEBUG 849 int n = 0; 850 #endif 851 852 mutex_enter(&nsrv4->servinst_lock); 853 ASSERT(nsrv4->nfs4_cur_servinst != NULL); 854 current = nsrv4->nfs4_cur_servinst; 855 nsrv4->nfs4_cur_servinst = NULL; 856 for (sip = current; sip != NULL; sip = prev) { 857 prev = sip->prev; 858 rw_destroy(&sip->rwlock); 859 if (sip->oldstate) 860 kmem_free(sip->oldstate, sizeof (rfs4_oldstate_t)); 861 if (sip->dss_paths) { 862 int i = sip->dss_npaths; 863 864 while (i > 0) { 865 i--; 866 if (sip->dss_paths[i] != NULL) { 867 char *path = sip->dss_paths[i]->path; 868 869 if (path != NULL) { 870 kmem_free(path, 871 strlen(path) + 1); 872 } 873 kmem_free(sip->dss_paths[i], 874 sizeof (rfs4_dss_path_t)); 875 } 876 } 877 kmem_free(sip->dss_paths, 878 sip->dss_npaths * sizeof (rfs4_dss_path_t *)); 879 } 880 kmem_free(sip, sizeof (rfs4_servinst_t)); 881 #ifdef DEBUG 882 n++; 883 #endif 884 } 885 mutex_exit(&nsrv4->servinst_lock); 886 } 887 888 /* 889 * Assign the current server instance to a client_t. 890 * Should be called with cp->rc_dbe held. 891 */ 892 void 893 rfs4_servinst_assign(nfs4_srv_t *nsrv4, rfs4_client_t *cp, 894 rfs4_servinst_t *sip) 895 { 896 ASSERT(rfs4_dbe_refcnt(cp->rc_dbe) > 0); 897 898 /* 899 * The lock ensures that if the current instance is in the process 900 * of changing, we will see the new one. 901 */ 902 mutex_enter(&nsrv4->servinst_lock); 903 cp->rc_server_instance = sip; 904 mutex_exit(&nsrv4->servinst_lock); 905 } 906 907 rfs4_servinst_t * 908 rfs4_servinst(rfs4_client_t *cp) 909 { 910 ASSERT(rfs4_dbe_refcnt(cp->rc_dbe) > 0); 911 912 return (cp->rc_server_instance); 913 } 914 915 /* ARGSUSED */ 916 static void 917 nullfree(caddr_t resop) 918 { 919 } 920 921 /* 922 * This is a fall-through for invalid or not implemented (yet) ops 923 */ 924 /* ARGSUSED */ 925 static void 926 rfs4_op_inval(nfs_argop4 *argop, nfs_resop4 *resop, struct svc_req *req, 927 struct compound_state *cs) 928 { 929 *cs->statusp = *((nfsstat4 *)&(resop)->nfs_resop4_u) = NFS4ERR_INVAL; 930 } 931 932 /* 933 * Check if the security flavor, nfsnum, is in the flavor_list. 934 */ 935 bool_t 936 in_flavor_list(int nfsnum, int *flavor_list, int count) 937 { 938 int i; 939 940 for (i = 0; i < count; i++) { 941 if (nfsnum == flavor_list[i]) 942 return (TRUE); 943 } 944 return (FALSE); 945 } 946 947 /* 948 * Used by rfs4_op_secinfo to get the security information from the 949 * export structure associated with the component. 950 */ 951 /* ARGSUSED */ 952 static nfsstat4 953 do_rfs4_op_secinfo(struct compound_state *cs, char *nm, SECINFO4res *resp) 954 { 955 int error, different_export = 0; 956 vnode_t *dvp, *vp; 957 struct exportinfo *exi; 958 fid_t fid; 959 uint_t count, i; 960 secinfo4 *resok_val; 961 struct secinfo *secp; 962 seconfig_t *si; 963 bool_t did_traverse = FALSE; 964 int dotdot, walk; 965 nfs_export_t *ne = nfs_get_export(); 966 967 dvp = cs->vp; 968 exi = cs->exi; 969 ASSERT(exi != NULL); 970 dotdot = (nm[0] == '.' && nm[1] == '.' && nm[2] == '\0'); 971 972 /* 973 * If dotdotting, then need to check whether it's above the 974 * root of a filesystem, or above an export point. 975 */ 976 if (dotdot) { 977 vnode_t *zone_rootvp = ne->exi_root->exi_vp; 978 979 ASSERT3U(exi->exi_zoneid, ==, ne->exi_root->exi_zoneid); 980 /* 981 * If dotdotting at the root of a filesystem, then 982 * need to traverse back to the mounted-on filesystem 983 * and do the dotdot lookup there. 984 */ 985 if ((dvp->v_flag & VROOT) || VN_CMP(dvp, zone_rootvp)) { 986 987 /* 988 * If at the system root, then can 989 * go up no further. 990 */ 991 if (VN_CMP(dvp, zone_rootvp)) 992 return (puterrno4(ENOENT)); 993 994 /* 995 * Traverse back to the mounted-on filesystem 996 */ 997 dvp = untraverse(dvp, zone_rootvp); 998 999 /* 1000 * Set the different_export flag so we remember 1001 * to pick up a new exportinfo entry for 1002 * this new filesystem. 1003 */ 1004 different_export = 1; 1005 } else { 1006 1007 /* 1008 * If dotdotting above an export point then set 1009 * the different_export to get new export info. 1010 */ 1011 different_export = nfs_exported(exi, dvp); 1012 } 1013 } 1014 1015 /* 1016 * Get the vnode for the component "nm". 1017 */ 1018 error = VOP_LOOKUP(dvp, nm, &vp, NULL, 0, NULL, cs->cr, 1019 NULL, NULL, NULL); 1020 if (error) 1021 return (puterrno4(error)); 1022 1023 /* 1024 * If the vnode is in a pseudo filesystem, or if the security flavor 1025 * used in the request is valid but not an explicitly shared flavor, 1026 * or the access bit indicates that this is a limited access, 1027 * check whether this vnode is visible. 1028 */ 1029 if (!different_export && 1030 (PSEUDO(exi) || !is_exported_sec(cs->nfsflavor, exi) || 1031 cs->access & CS_ACCESS_LIMITED)) { 1032 if (! nfs_visible(exi, vp, &different_export)) { 1033 VN_RELE(vp); 1034 return (puterrno4(ENOENT)); 1035 } 1036 } 1037 1038 /* 1039 * If it's a mountpoint, then traverse it. 1040 */ 1041 if (vn_ismntpt(vp)) { 1042 if ((error = traverse(&vp)) != 0) { 1043 VN_RELE(vp); 1044 return (puterrno4(error)); 1045 } 1046 /* remember that we had to traverse mountpoint */ 1047 did_traverse = TRUE; 1048 different_export = 1; 1049 } else if (vp->v_vfsp != dvp->v_vfsp) { 1050 /* 1051 * If vp isn't a mountpoint and the vfs ptrs aren't the same, 1052 * then vp is probably an LOFS object. We don't need the 1053 * realvp, we just need to know that we might have crossed 1054 * a server fs boundary and need to call checkexport4. 1055 * (LOFS lookup hides server fs mountpoints, and actually calls 1056 * traverse) 1057 */ 1058 different_export = 1; 1059 } 1060 1061 /* 1062 * Get the export information for it. 1063 */ 1064 if (different_export) { 1065 1066 bzero(&fid, sizeof (fid)); 1067 fid.fid_len = MAXFIDSZ; 1068 error = vop_fid_pseudo(vp, &fid); 1069 if (error) { 1070 VN_RELE(vp); 1071 return (puterrno4(error)); 1072 } 1073 1074 /* We'll need to reassign "exi". */ 1075 if (dotdot) 1076 exi = nfs_vptoexi(NULL, vp, cs->cr, &walk, NULL, TRUE); 1077 else 1078 exi = checkexport4(&vp->v_vfsp->vfs_fsid, &fid, vp); 1079 1080 if (exi == NULL) { 1081 if (did_traverse == TRUE) { 1082 /* 1083 * If this vnode is a mounted-on vnode, 1084 * but the mounted-on file system is not 1085 * exported, send back the secinfo for 1086 * the exported node that the mounted-on 1087 * vnode lives in. 1088 */ 1089 exi = cs->exi; 1090 } else { 1091 VN_RELE(vp); 1092 return (puterrno4(EACCES)); 1093 } 1094 } 1095 } 1096 ASSERT(exi != NULL); 1097 1098 1099 /* 1100 * Create the secinfo result based on the security information 1101 * from the exportinfo structure (exi). 1102 * 1103 * Return all flavors for a pseudo node. 1104 * For a real export node, return the flavor that the client 1105 * has access with. 1106 */ 1107 ASSERT(RW_LOCK_HELD(&ne->exported_lock)); 1108 if (PSEUDO(exi)) { 1109 count = exi->exi_export.ex_seccnt; /* total sec count */ 1110 resok_val = kmem_alloc(count * sizeof (secinfo4), KM_SLEEP); 1111 secp = exi->exi_export.ex_secinfo; 1112 1113 for (i = 0; i < count; i++) { 1114 si = &secp[i].s_secinfo; 1115 resok_val[i].flavor = si->sc_rpcnum; 1116 if (resok_val[i].flavor == RPCSEC_GSS) { 1117 rpcsec_gss_info *info; 1118 1119 info = &resok_val[i].flavor_info; 1120 info->qop = si->sc_qop; 1121 info->service = (rpc_gss_svc_t)si->sc_service; 1122 1123 /* get oid opaque data */ 1124 info->oid.sec_oid4_len = 1125 si->sc_gss_mech_type->length; 1126 info->oid.sec_oid4_val = kmem_alloc( 1127 si->sc_gss_mech_type->length, KM_SLEEP); 1128 bcopy( 1129 si->sc_gss_mech_type->elements, 1130 info->oid.sec_oid4_val, 1131 info->oid.sec_oid4_len); 1132 } 1133 } 1134 resp->SECINFO4resok_len = count; 1135 resp->SECINFO4resok_val = resok_val; 1136 } else { 1137 int ret_cnt = 0, k = 0; 1138 int *flavor_list; 1139 1140 count = exi->exi_export.ex_seccnt; /* total sec count */ 1141 secp = exi->exi_export.ex_secinfo; 1142 1143 flavor_list = kmem_alloc(count * sizeof (int), KM_SLEEP); 1144 /* find out which flavors to return */ 1145 for (i = 0; i < count; i ++) { 1146 int access, flavor, perm; 1147 1148 flavor = secp[i].s_secinfo.sc_nfsnum; 1149 perm = secp[i].s_flags; 1150 1151 access = nfsauth4_secinfo_access(exi, cs->req, 1152 flavor, perm, cs->basecr); 1153 1154 if (! (access & NFSAUTH_DENIED) && 1155 ! (access & NFSAUTH_WRONGSEC)) { 1156 flavor_list[ret_cnt] = flavor; 1157 ret_cnt++; 1158 } 1159 } 1160 1161 /* Create the returning SECINFO value */ 1162 resok_val = kmem_alloc(ret_cnt * sizeof (secinfo4), KM_SLEEP); 1163 1164 for (i = 0; i < count; i++) { 1165 /* 1166 * If the flavor is in the flavor list, 1167 * fill in resok_val. 1168 */ 1169 si = &secp[i].s_secinfo; 1170 if (in_flavor_list(si->sc_nfsnum, 1171 flavor_list, ret_cnt)) { 1172 resok_val[k].flavor = si->sc_rpcnum; 1173 if (resok_val[k].flavor == RPCSEC_GSS) { 1174 rpcsec_gss_info *info; 1175 1176 info = &resok_val[k].flavor_info; 1177 info->qop = si->sc_qop; 1178 info->service = (rpc_gss_svc_t) 1179 si->sc_service; 1180 1181 /* get oid opaque data */ 1182 info->oid.sec_oid4_len = 1183 si->sc_gss_mech_type->length; 1184 info->oid.sec_oid4_val = kmem_alloc( 1185 si->sc_gss_mech_type->length, 1186 KM_SLEEP); 1187 bcopy(si->sc_gss_mech_type->elements, 1188 info->oid.sec_oid4_val, 1189 info->oid.sec_oid4_len); 1190 } 1191 k++; 1192 } 1193 if (k >= ret_cnt) 1194 break; 1195 } 1196 resp->SECINFO4resok_len = ret_cnt; 1197 resp->SECINFO4resok_val = resok_val; 1198 kmem_free(flavor_list, count * sizeof (int)); 1199 } 1200 1201 VN_RELE(vp); 1202 return (NFS4_OK); 1203 } 1204 1205 /* 1206 * SECINFO (Operation 33): Obtain required security information on 1207 * the component name in the format of (security-mechanism-oid, qop, service) 1208 * triplets. 1209 */ 1210 /* ARGSUSED */ 1211 static void 1212 rfs4_op_secinfo(nfs_argop4 *argop, nfs_resop4 *resop, struct svc_req *req, 1213 struct compound_state *cs) 1214 { 1215 SECINFO4args *args = &argop->nfs_argop4_u.opsecinfo; 1216 SECINFO4res *resp = &resop->nfs_resop4_u.opsecinfo; 1217 utf8string *utfnm = &args->name; 1218 uint_t len; 1219 char *nm; 1220 struct sockaddr *ca; 1221 char *name = NULL; 1222 nfsstat4 status = NFS4_OK; 1223 1224 DTRACE_NFSV4_2(op__secinfo__start, struct compound_state *, cs, 1225 SECINFO4args *, args); 1226 1227 /* 1228 * Current file handle (cfh) should have been set before getting 1229 * into this function. If not, return error. 1230 */ 1231 if (cs->vp == NULL) { 1232 *cs->statusp = resp->status = NFS4ERR_NOFILEHANDLE; 1233 goto out; 1234 } 1235 1236 if (cs->vp->v_type != VDIR) { 1237 *cs->statusp = resp->status = NFS4ERR_NOTDIR; 1238 goto out; 1239 } 1240 1241 /* 1242 * Verify the component name. If failed, error out, but 1243 * do not error out if the component name is a "..". 1244 * SECINFO will return its parents secinfo data for SECINFO "..". 1245 */ 1246 status = utf8_dir_verify(utfnm); 1247 if (status != NFS4_OK) { 1248 if (utfnm->utf8string_len != 2 || 1249 utfnm->utf8string_val[0] != '.' || 1250 utfnm->utf8string_val[1] != '.') { 1251 *cs->statusp = resp->status = status; 1252 goto out; 1253 } 1254 } 1255 1256 nm = utf8_to_str(utfnm, &len, NULL); 1257 if (nm == NULL) { 1258 *cs->statusp = resp->status = NFS4ERR_INVAL; 1259 goto out; 1260 } 1261 1262 if (len > MAXNAMELEN) { 1263 *cs->statusp = resp->status = NFS4ERR_NAMETOOLONG; 1264 kmem_free(nm, len); 1265 goto out; 1266 } 1267 1268 ca = (struct sockaddr *)svc_getrpccaller(req->rq_xprt)->buf; 1269 name = nfscmd_convname(ca, cs->exi, nm, NFSCMD_CONV_INBOUND, 1270 MAXPATHLEN + 1); 1271 1272 if (name == NULL) { 1273 *cs->statusp = resp->status = NFS4ERR_INVAL; 1274 kmem_free(nm, len); 1275 goto out; 1276 } 1277 1278 1279 *cs->statusp = resp->status = do_rfs4_op_secinfo(cs, name, resp); 1280 1281 if (name != nm) 1282 kmem_free(name, MAXPATHLEN + 1); 1283 kmem_free(nm, len); 1284 1285 out: 1286 DTRACE_NFSV4_2(op__secinfo__done, struct compound_state *, cs, 1287 SECINFO4res *, resp); 1288 } 1289 1290 /* 1291 * Free SECINFO result. 1292 */ 1293 /* ARGSUSED */ 1294 static void 1295 rfs4_op_secinfo_free(nfs_resop4 *resop) 1296 { 1297 SECINFO4res *resp = &resop->nfs_resop4_u.opsecinfo; 1298 int count, i; 1299 secinfo4 *resok_val; 1300 1301 /* If this is not an Ok result, nothing to free. */ 1302 if (resp->status != NFS4_OK) { 1303 return; 1304 } 1305 1306 count = resp->SECINFO4resok_len; 1307 resok_val = resp->SECINFO4resok_val; 1308 1309 for (i = 0; i < count; i++) { 1310 if (resok_val[i].flavor == RPCSEC_GSS) { 1311 rpcsec_gss_info *info; 1312 1313 info = &resok_val[i].flavor_info; 1314 kmem_free(info->oid.sec_oid4_val, 1315 info->oid.sec_oid4_len); 1316 } 1317 } 1318 kmem_free(resok_val, count * sizeof (secinfo4)); 1319 resp->SECINFO4resok_len = 0; 1320 resp->SECINFO4resok_val = NULL; 1321 } 1322 1323 /* ARGSUSED */ 1324 static void 1325 rfs4_op_access(nfs_argop4 *argop, nfs_resop4 *resop, struct svc_req *req, 1326 struct compound_state *cs) 1327 { 1328 ACCESS4args *args = &argop->nfs_argop4_u.opaccess; 1329 ACCESS4res *resp = &resop->nfs_resop4_u.opaccess; 1330 int error; 1331 vnode_t *vp; 1332 struct vattr va; 1333 int checkwriteperm; 1334 cred_t *cr = cs->cr; 1335 bslabel_t *clabel, *slabel; 1336 ts_label_t *tslabel; 1337 boolean_t admin_low_client; 1338 1339 DTRACE_NFSV4_2(op__access__start, struct compound_state *, cs, 1340 ACCESS4args *, args); 1341 1342 #if 0 /* XXX allow access even if !cs->access. Eventually only pseudo fs */ 1343 if (cs->access == CS_ACCESS_DENIED) { 1344 *cs->statusp = resp->status = NFS4ERR_ACCESS; 1345 goto out; 1346 } 1347 #endif 1348 if (cs->vp == NULL) { 1349 *cs->statusp = resp->status = NFS4ERR_NOFILEHANDLE; 1350 goto out; 1351 } 1352 1353 ASSERT(cr != NULL); 1354 1355 vp = cs->vp; 1356 1357 /* 1358 * If the file system is exported read only, it is not appropriate 1359 * to check write permissions for regular files and directories. 1360 * Special files are interpreted by the client, so the underlying 1361 * permissions are sent back to the client for interpretation. 1362 */ 1363 if (rdonly4(req, cs) && 1364 (vp->v_type == VREG || vp->v_type == VDIR)) 1365 checkwriteperm = 0; 1366 else 1367 checkwriteperm = 1; 1368 1369 /* 1370 * XXX 1371 * We need the mode so that we can correctly determine access 1372 * permissions relative to a mandatory lock file. Access to 1373 * mandatory lock files is denied on the server, so it might 1374 * as well be reflected to the server during the open. 1375 */ 1376 va.va_mask = AT_MODE; 1377 error = VOP_GETATTR(vp, &va, 0, cr, NULL); 1378 if (error) { 1379 *cs->statusp = resp->status = puterrno4(error); 1380 goto out; 1381 } 1382 resp->access = 0; 1383 resp->supported = 0; 1384 1385 if (is_system_labeled()) { 1386 ASSERT(req->rq_label != NULL); 1387 clabel = req->rq_label; 1388 DTRACE_PROBE2(tx__rfs4__log__info__opaccess__clabel, char *, 1389 "got client label from request(1)", 1390 struct svc_req *, req); 1391 if (!blequal(&l_admin_low->tsl_label, clabel)) { 1392 if ((tslabel = nfs_getflabel(vp, cs->exi)) == NULL) { 1393 *cs->statusp = resp->status = puterrno4(EACCES); 1394 goto out; 1395 } 1396 slabel = label2bslabel(tslabel); 1397 DTRACE_PROBE3(tx__rfs4__log__info__opaccess__slabel, 1398 char *, "got server label(1) for vp(2)", 1399 bslabel_t *, slabel, vnode_t *, vp); 1400 1401 admin_low_client = B_FALSE; 1402 } else 1403 admin_low_client = B_TRUE; 1404 } 1405 1406 if (args->access & ACCESS4_READ) { 1407 error = VOP_ACCESS(vp, VREAD, 0, cr, NULL); 1408 if (!error && !MANDLOCK(vp, va.va_mode) && 1409 (!is_system_labeled() || admin_low_client || 1410 bldominates(clabel, slabel))) 1411 resp->access |= ACCESS4_READ; 1412 resp->supported |= ACCESS4_READ; 1413 } 1414 if ((args->access & ACCESS4_LOOKUP) && vp->v_type == VDIR) { 1415 error = VOP_ACCESS(vp, VEXEC, 0, cr, NULL); 1416 if (!error && (!is_system_labeled() || admin_low_client || 1417 bldominates(clabel, slabel))) 1418 resp->access |= ACCESS4_LOOKUP; 1419 resp->supported |= ACCESS4_LOOKUP; 1420 } 1421 if (checkwriteperm && 1422 (args->access & (ACCESS4_MODIFY|ACCESS4_EXTEND))) { 1423 error = VOP_ACCESS(vp, VWRITE, 0, cr, NULL); 1424 if (!error && !MANDLOCK(vp, va.va_mode) && 1425 (!is_system_labeled() || admin_low_client || 1426 blequal(clabel, slabel))) 1427 resp->access |= 1428 (args->access & (ACCESS4_MODIFY | ACCESS4_EXTEND)); 1429 resp->supported |= 1430 resp->access & (ACCESS4_MODIFY | ACCESS4_EXTEND); 1431 } 1432 1433 if (checkwriteperm && 1434 (args->access & ACCESS4_DELETE) && vp->v_type == VDIR) { 1435 error = VOP_ACCESS(vp, VWRITE, 0, cr, NULL); 1436 if (!error && (!is_system_labeled() || admin_low_client || 1437 blequal(clabel, slabel))) 1438 resp->access |= ACCESS4_DELETE; 1439 resp->supported |= ACCESS4_DELETE; 1440 } 1441 if (args->access & ACCESS4_EXECUTE && vp->v_type != VDIR) { 1442 error = VOP_ACCESS(vp, VEXEC, 0, cr, NULL); 1443 if (!error && !MANDLOCK(vp, va.va_mode) && 1444 (!is_system_labeled() || admin_low_client || 1445 bldominates(clabel, slabel))) 1446 resp->access |= ACCESS4_EXECUTE; 1447 resp->supported |= ACCESS4_EXECUTE; 1448 } 1449 1450 if (is_system_labeled() && !admin_low_client) 1451 label_rele(tslabel); 1452 1453 *cs->statusp = resp->status = NFS4_OK; 1454 out: 1455 DTRACE_NFSV4_2(op__access__done, struct compound_state *, cs, 1456 ACCESS4res *, resp); 1457 } 1458 1459 /* ARGSUSED */ 1460 static void 1461 rfs4_op_commit(nfs_argop4 *argop, nfs_resop4 *resop, struct svc_req *req, 1462 struct compound_state *cs) 1463 { 1464 COMMIT4args *args = &argop->nfs_argop4_u.opcommit; 1465 COMMIT4res *resp = &resop->nfs_resop4_u.opcommit; 1466 int error; 1467 vnode_t *vp = cs->vp; 1468 cred_t *cr = cs->cr; 1469 vattr_t va; 1470 nfs4_srv_t *nsrv4; 1471 1472 DTRACE_NFSV4_2(op__commit__start, struct compound_state *, cs, 1473 COMMIT4args *, args); 1474 1475 if (vp == NULL) { 1476 *cs->statusp = resp->status = NFS4ERR_NOFILEHANDLE; 1477 goto out; 1478 } 1479 if (cs->access == CS_ACCESS_DENIED) { 1480 *cs->statusp = resp->status = NFS4ERR_ACCESS; 1481 goto out; 1482 } 1483 1484 if (args->offset + args->count < args->offset) { 1485 *cs->statusp = resp->status = NFS4ERR_INVAL; 1486 goto out; 1487 } 1488 1489 va.va_mask = AT_UID; 1490 error = VOP_GETATTR(vp, &va, 0, cr, NULL); 1491 1492 /* 1493 * If we can't get the attributes, then we can't do the 1494 * right access checking. So, we'll fail the request. 1495 */ 1496 if (error) { 1497 *cs->statusp = resp->status = puterrno4(error); 1498 goto out; 1499 } 1500 if (rdonly4(req, cs)) { 1501 *cs->statusp = resp->status = NFS4ERR_ROFS; 1502 goto out; 1503 } 1504 1505 if (vp->v_type != VREG) { 1506 if (vp->v_type == VDIR) 1507 resp->status = NFS4ERR_ISDIR; 1508 else 1509 resp->status = NFS4ERR_INVAL; 1510 *cs->statusp = resp->status; 1511 goto out; 1512 } 1513 1514 if (crgetuid(cr) != va.va_uid && 1515 (error = VOP_ACCESS(vp, VWRITE, 0, cs->cr, NULL))) { 1516 *cs->statusp = resp->status = puterrno4(error); 1517 goto out; 1518 } 1519 1520 error = VOP_FSYNC(vp, FSYNC, cr, NULL); 1521 1522 if (error) { 1523 *cs->statusp = resp->status = puterrno4(error); 1524 goto out; 1525 } 1526 1527 nsrv4 = nfs4_get_srv(); 1528 *cs->statusp = resp->status = NFS4_OK; 1529 resp->writeverf = nsrv4->write4verf; 1530 out: 1531 DTRACE_NFSV4_2(op__commit__done, struct compound_state *, cs, 1532 COMMIT4res *, resp); 1533 } 1534 1535 /* 1536 * rfs4_op_mknod is called from rfs4_op_create after all initial verification 1537 * was completed. It does the nfsv4 create for special files. 1538 */ 1539 /* ARGSUSED */ 1540 static vnode_t * 1541 do_rfs4_op_mknod(CREATE4args *args, CREATE4res *resp, struct svc_req *req, 1542 struct compound_state *cs, vattr_t *vap, char *nm) 1543 { 1544 int error; 1545 cred_t *cr = cs->cr; 1546 vnode_t *dvp = cs->vp; 1547 vnode_t *vp = NULL; 1548 int mode; 1549 enum vcexcl excl; 1550 1551 switch (args->type) { 1552 case NF4CHR: 1553 case NF4BLK: 1554 if (secpolicy_sys_devices(cr) != 0) { 1555 *cs->statusp = resp->status = NFS4ERR_PERM; 1556 return (NULL); 1557 } 1558 if (args->type == NF4CHR) 1559 vap->va_type = VCHR; 1560 else 1561 vap->va_type = VBLK; 1562 vap->va_rdev = makedevice(args->ftype4_u.devdata.specdata1, 1563 args->ftype4_u.devdata.specdata2); 1564 vap->va_mask |= AT_RDEV; 1565 break; 1566 case NF4SOCK: 1567 vap->va_type = VSOCK; 1568 break; 1569 case NF4FIFO: 1570 vap->va_type = VFIFO; 1571 break; 1572 default: 1573 *cs->statusp = resp->status = NFS4ERR_BADTYPE; 1574 return (NULL); 1575 } 1576 1577 /* 1578 * Must specify the mode. 1579 */ 1580 if (!(vap->va_mask & AT_MODE)) { 1581 *cs->statusp = resp->status = NFS4ERR_INVAL; 1582 return (NULL); 1583 } 1584 1585 excl = EXCL; 1586 1587 mode = 0; 1588 1589 error = VOP_CREATE(dvp, nm, vap, excl, mode, &vp, cr, 0, NULL, NULL); 1590 if (error) { 1591 *cs->statusp = resp->status = puterrno4(error); 1592 return (NULL); 1593 } 1594 return (vp); 1595 } 1596 1597 /* 1598 * nfsv4 create is used to create non-regular files. For regular files, 1599 * use nfsv4 open. 1600 */ 1601 /* ARGSUSED */ 1602 static void 1603 rfs4_op_create(nfs_argop4 *argop, nfs_resop4 *resop, struct svc_req *req, 1604 struct compound_state *cs) 1605 { 1606 CREATE4args *args = &argop->nfs_argop4_u.opcreate; 1607 CREATE4res *resp = &resop->nfs_resop4_u.opcreate; 1608 int error; 1609 struct vattr bva, iva, iva2, ava, *vap; 1610 cred_t *cr = cs->cr; 1611 vnode_t *dvp = cs->vp; 1612 vnode_t *vp = NULL; 1613 vnode_t *realvp; 1614 char *nm, *lnm; 1615 uint_t len, llen; 1616 int syncval = 0; 1617 struct nfs4_svgetit_arg sarg; 1618 struct nfs4_ntov_table ntov; 1619 struct statvfs64 sb; 1620 nfsstat4 status; 1621 struct sockaddr *ca; 1622 char *name = NULL; 1623 char *lname = NULL; 1624 1625 DTRACE_NFSV4_2(op__create__start, struct compound_state *, cs, 1626 CREATE4args *, args); 1627 1628 resp->attrset = 0; 1629 1630 if (dvp == NULL) { 1631 *cs->statusp = resp->status = NFS4ERR_NOFILEHANDLE; 1632 goto out; 1633 } 1634 1635 /* 1636 * If there is an unshared filesystem mounted on this vnode, 1637 * do not allow to create an object in this directory. 1638 */ 1639 if (vn_ismntpt(dvp)) { 1640 *cs->statusp = resp->status = NFS4ERR_ACCESS; 1641 goto out; 1642 } 1643 1644 /* Verify that type is correct */ 1645 switch (args->type) { 1646 case NF4LNK: 1647 case NF4BLK: 1648 case NF4CHR: 1649 case NF4SOCK: 1650 case NF4FIFO: 1651 case NF4DIR: 1652 break; 1653 default: 1654 *cs->statusp = resp->status = NFS4ERR_BADTYPE; 1655 goto out; 1656 }; 1657 1658 if (cs->access == CS_ACCESS_DENIED) { 1659 *cs->statusp = resp->status = NFS4ERR_ACCESS; 1660 goto out; 1661 } 1662 if (dvp->v_type != VDIR) { 1663 *cs->statusp = resp->status = NFS4ERR_NOTDIR; 1664 goto out; 1665 } 1666 status = utf8_dir_verify(&args->objname); 1667 if (status != NFS4_OK) { 1668 *cs->statusp = resp->status = status; 1669 goto out; 1670 } 1671 1672 if (rdonly4(req, cs)) { 1673 *cs->statusp = resp->status = NFS4ERR_ROFS; 1674 goto out; 1675 } 1676 1677 /* 1678 * Name of newly created object 1679 */ 1680 nm = utf8_to_fn(&args->objname, &len, NULL); 1681 if (nm == NULL) { 1682 *cs->statusp = resp->status = NFS4ERR_INVAL; 1683 goto out; 1684 } 1685 1686 if (len > MAXNAMELEN) { 1687 *cs->statusp = resp->status = NFS4ERR_NAMETOOLONG; 1688 kmem_free(nm, len); 1689 goto out; 1690 } 1691 1692 ca = (struct sockaddr *)svc_getrpccaller(req->rq_xprt)->buf; 1693 name = nfscmd_convname(ca, cs->exi, nm, NFSCMD_CONV_INBOUND, 1694 MAXPATHLEN + 1); 1695 1696 if (name == NULL) { 1697 *cs->statusp = resp->status = NFS4ERR_INVAL; 1698 kmem_free(nm, len); 1699 goto out; 1700 } 1701 1702 resp->attrset = 0; 1703 1704 sarg.sbp = &sb; 1705 sarg.is_referral = B_FALSE; 1706 nfs4_ntov_table_init(&ntov); 1707 1708 status = do_rfs4_set_attrs(&resp->attrset, 1709 &args->createattrs, cs, &sarg, &ntov, NFS4ATTR_SETIT); 1710 1711 if (sarg.vap->va_mask == 0 && status == NFS4_OK) 1712 status = NFS4ERR_INVAL; 1713 1714 if (status != NFS4_OK) { 1715 *cs->statusp = resp->status = status; 1716 if (name != nm) 1717 kmem_free(name, MAXPATHLEN + 1); 1718 kmem_free(nm, len); 1719 nfs4_ntov_table_free(&ntov, &sarg); 1720 resp->attrset = 0; 1721 goto out; 1722 } 1723 1724 /* Get "before" change value */ 1725 bva.va_mask = AT_CTIME|AT_SEQ|AT_MODE; 1726 error = VOP_GETATTR(dvp, &bva, 0, cr, NULL); 1727 if (error) { 1728 *cs->statusp = resp->status = puterrno4(error); 1729 if (name != nm) 1730 kmem_free(name, MAXPATHLEN + 1); 1731 kmem_free(nm, len); 1732 nfs4_ntov_table_free(&ntov, &sarg); 1733 resp->attrset = 0; 1734 goto out; 1735 } 1736 NFS4_SET_FATTR4_CHANGE(resp->cinfo.before, bva.va_ctime) 1737 1738 vap = sarg.vap; 1739 1740 /* 1741 * Set the default initial values for attributes when the parent 1742 * directory does not have the VSUID/VSGID bit set and they have 1743 * not been specified in createattrs. 1744 */ 1745 if (!(bva.va_mode & VSUID) && (vap->va_mask & AT_UID) == 0) { 1746 vap->va_uid = crgetuid(cr); 1747 vap->va_mask |= AT_UID; 1748 } 1749 if (!(bva.va_mode & VSGID) && (vap->va_mask & AT_GID) == 0) { 1750 vap->va_gid = crgetgid(cr); 1751 vap->va_mask |= AT_GID; 1752 } 1753 1754 vap->va_mask |= AT_TYPE; 1755 switch (args->type) { 1756 case NF4DIR: 1757 vap->va_type = VDIR; 1758 if ((vap->va_mask & AT_MODE) == 0) { 1759 vap->va_mode = 0700; /* default: owner rwx only */ 1760 vap->va_mask |= AT_MODE; 1761 } 1762 error = VOP_MKDIR(dvp, name, vap, &vp, cr, NULL, 0, NULL); 1763 if (error) 1764 break; 1765 1766 /* 1767 * Get the initial "after" sequence number, if it fails, 1768 * set to zero 1769 */ 1770 iva.va_mask = AT_SEQ; 1771 if (VOP_GETATTR(dvp, &iva, 0, cs->cr, NULL)) 1772 iva.va_seq = 0; 1773 break; 1774 case NF4LNK: 1775 vap->va_type = VLNK; 1776 if ((vap->va_mask & AT_MODE) == 0) { 1777 vap->va_mode = 0700; /* default: owner rwx only */ 1778 vap->va_mask |= AT_MODE; 1779 } 1780 1781 /* 1782 * symlink names must be treated as data 1783 */ 1784 lnm = utf8_to_str((utf8string *)&args->ftype4_u.linkdata, 1785 &llen, NULL); 1786 1787 if (lnm == NULL) { 1788 *cs->statusp = resp->status = NFS4ERR_INVAL; 1789 if (name != nm) 1790 kmem_free(name, MAXPATHLEN + 1); 1791 kmem_free(nm, len); 1792 nfs4_ntov_table_free(&ntov, &sarg); 1793 resp->attrset = 0; 1794 goto out; 1795 } 1796 1797 if (llen > MAXPATHLEN) { 1798 *cs->statusp = resp->status = NFS4ERR_NAMETOOLONG; 1799 if (name != nm) 1800 kmem_free(name, MAXPATHLEN + 1); 1801 kmem_free(nm, len); 1802 kmem_free(lnm, llen); 1803 nfs4_ntov_table_free(&ntov, &sarg); 1804 resp->attrset = 0; 1805 goto out; 1806 } 1807 1808 lname = nfscmd_convname(ca, cs->exi, lnm, 1809 NFSCMD_CONV_INBOUND, MAXPATHLEN + 1); 1810 1811 if (lname == NULL) { 1812 *cs->statusp = resp->status = NFS4ERR_SERVERFAULT; 1813 if (name != nm) 1814 kmem_free(name, MAXPATHLEN + 1); 1815 kmem_free(nm, len); 1816 kmem_free(lnm, llen); 1817 nfs4_ntov_table_free(&ntov, &sarg); 1818 resp->attrset = 0; 1819 goto out; 1820 } 1821 1822 error = VOP_SYMLINK(dvp, name, vap, lname, cr, NULL, 0); 1823 if (lname != lnm) 1824 kmem_free(lname, MAXPATHLEN + 1); 1825 kmem_free(lnm, llen); 1826 if (error) 1827 break; 1828 1829 /* 1830 * Get the initial "after" sequence number, if it fails, 1831 * set to zero 1832 */ 1833 iva.va_mask = AT_SEQ; 1834 if (VOP_GETATTR(dvp, &iva, 0, cs->cr, NULL)) 1835 iva.va_seq = 0; 1836 1837 error = VOP_LOOKUP(dvp, name, &vp, NULL, 0, NULL, cr, 1838 NULL, NULL, NULL); 1839 if (error) 1840 break; 1841 1842 /* 1843 * va_seq is not safe over VOP calls, check it again 1844 * if it has changed zero out iva to force atomic = FALSE. 1845 */ 1846 iva2.va_mask = AT_SEQ; 1847 if (VOP_GETATTR(dvp, &iva2, 0, cs->cr, NULL) || 1848 iva2.va_seq != iva.va_seq) 1849 iva.va_seq = 0; 1850 break; 1851 default: 1852 /* 1853 * probably a special file. 1854 */ 1855 if ((vap->va_mask & AT_MODE) == 0) { 1856 vap->va_mode = 0600; /* default: owner rw only */ 1857 vap->va_mask |= AT_MODE; 1858 } 1859 syncval = FNODSYNC; 1860 /* 1861 * We know this will only generate one VOP call 1862 */ 1863 vp = do_rfs4_op_mknod(args, resp, req, cs, vap, name); 1864 1865 if (vp == NULL) { 1866 if (name != nm) 1867 kmem_free(name, MAXPATHLEN + 1); 1868 kmem_free(nm, len); 1869 nfs4_ntov_table_free(&ntov, &sarg); 1870 resp->attrset = 0; 1871 goto out; 1872 } 1873 1874 /* 1875 * Get the initial "after" sequence number, if it fails, 1876 * set to zero 1877 */ 1878 iva.va_mask = AT_SEQ; 1879 if (VOP_GETATTR(dvp, &iva, 0, cs->cr, NULL)) 1880 iva.va_seq = 0; 1881 1882 break; 1883 } 1884 if (name != nm) 1885 kmem_free(name, MAXPATHLEN + 1); 1886 kmem_free(nm, len); 1887 1888 if (error) { 1889 *cs->statusp = resp->status = puterrno4(error); 1890 } 1891 1892 /* 1893 * Force modified data and metadata out to stable storage. 1894 */ 1895 (void) VOP_FSYNC(dvp, 0, cr, NULL); 1896 1897 if (resp->status != NFS4_OK) { 1898 if (vp != NULL) 1899 VN_RELE(vp); 1900 nfs4_ntov_table_free(&ntov, &sarg); 1901 resp->attrset = 0; 1902 goto out; 1903 } 1904 1905 /* 1906 * Finish setup of cinfo response, "before" value already set. 1907 * Get "after" change value, if it fails, simply return the 1908 * before value. 1909 */ 1910 ava.va_mask = AT_CTIME|AT_SEQ; 1911 if (VOP_GETATTR(dvp, &ava, 0, cr, NULL)) { 1912 ava.va_ctime = bva.va_ctime; 1913 ava.va_seq = 0; 1914 } 1915 NFS4_SET_FATTR4_CHANGE(resp->cinfo.after, ava.va_ctime); 1916 1917 /* 1918 * True verification that object was created with correct 1919 * attrs is impossible. The attrs could have been changed 1920 * immediately after object creation. If attributes did 1921 * not verify, the only recourse for the server is to 1922 * destroy the object. Maybe if some attrs (like gid) 1923 * are set incorrectly, the object should be destroyed; 1924 * however, seems bad as a default policy. Do we really 1925 * want to destroy an object over one of the times not 1926 * verifying correctly? For these reasons, the server 1927 * currently sets bits in attrset for createattrs 1928 * that were set; however, no verification is done. 1929 * 1930 * vmask_to_nmask accounts for vattr bits set on create 1931 * [do_rfs4_set_attrs() only sets resp bits for 1932 * non-vattr/vfs bits.] 1933 * Mask off any bits set by default so as not to return 1934 * more attrset bits than were requested in createattrs 1935 */ 1936 nfs4_vmask_to_nmask(sarg.vap->va_mask, &resp->attrset); 1937 resp->attrset &= args->createattrs.attrmask; 1938 nfs4_ntov_table_free(&ntov, &sarg); 1939 1940 error = makefh4(&cs->fh, vp, cs->exi); 1941 if (error) { 1942 *cs->statusp = resp->status = puterrno4(error); 1943 } 1944 1945 /* 1946 * The cinfo.atomic = TRUE only if we got no errors, we have 1947 * non-zero va_seq's, and it has incremented by exactly one 1948 * during the creation and it didn't change during the VOP_LOOKUP 1949 * or VOP_FSYNC. 1950 */ 1951 if (!error && bva.va_seq && iva.va_seq && ava.va_seq && 1952 iva.va_seq == (bva.va_seq + 1) && iva.va_seq == ava.va_seq) 1953 resp->cinfo.atomic = TRUE; 1954 else 1955 resp->cinfo.atomic = FALSE; 1956 1957 /* 1958 * Force modified metadata out to stable storage. 1959 * 1960 * if a underlying vp exists, pass it to VOP_FSYNC 1961 */ 1962 if (VOP_REALVP(vp, &realvp, NULL) == 0) 1963 (void) VOP_FSYNC(realvp, syncval, cr, NULL); 1964 else 1965 (void) VOP_FSYNC(vp, syncval, cr, NULL); 1966 1967 if (resp->status != NFS4_OK) { 1968 VN_RELE(vp); 1969 goto out; 1970 } 1971 if (cs->vp) 1972 VN_RELE(cs->vp); 1973 1974 cs->vp = vp; 1975 *cs->statusp = resp->status = NFS4_OK; 1976 out: 1977 DTRACE_NFSV4_2(op__create__done, struct compound_state *, cs, 1978 CREATE4res *, resp); 1979 } 1980 1981 /*ARGSUSED*/ 1982 static void 1983 rfs4_op_delegpurge(nfs_argop4 *argop, nfs_resop4 *resop, struct svc_req *req, 1984 struct compound_state *cs) 1985 { 1986 DTRACE_NFSV4_2(op__delegpurge__start, struct compound_state *, cs, 1987 DELEGPURGE4args *, &argop->nfs_argop4_u.opdelegpurge); 1988 1989 rfs4_op_inval(argop, resop, req, cs); 1990 1991 DTRACE_NFSV4_2(op__delegpurge__done, struct compound_state *, cs, 1992 DELEGPURGE4res *, &resop->nfs_resop4_u.opdelegpurge); 1993 } 1994 1995 /*ARGSUSED*/ 1996 static void 1997 rfs4_op_delegreturn(nfs_argop4 *argop, nfs_resop4 *resop, struct svc_req *req, 1998 struct compound_state *cs) 1999 { 2000 DELEGRETURN4args *args = &argop->nfs_argop4_u.opdelegreturn; 2001 DELEGRETURN4res *resp = &resop->nfs_resop4_u.opdelegreturn; 2002 rfs4_deleg_state_t *dsp; 2003 nfsstat4 status; 2004 2005 DTRACE_NFSV4_2(op__delegreturn__start, struct compound_state *, cs, 2006 DELEGRETURN4args *, args); 2007 2008 status = rfs4_get_deleg_state(&args->deleg_stateid, &dsp); 2009 resp->status = *cs->statusp = status; 2010 if (status != NFS4_OK) 2011 goto out; 2012 2013 /* Ensure specified filehandle matches */ 2014 if (cs->vp != dsp->rds_finfo->rf_vp) { 2015 resp->status = *cs->statusp = NFS4ERR_BAD_STATEID; 2016 } else 2017 rfs4_return_deleg(dsp, FALSE); 2018 2019 rfs4_update_lease(dsp->rds_client); 2020 2021 rfs4_deleg_state_rele(dsp); 2022 out: 2023 DTRACE_NFSV4_2(op__delegreturn__done, struct compound_state *, cs, 2024 DELEGRETURN4res *, resp); 2025 } 2026 2027 /* 2028 * Check to see if a given "flavor" is an explicitly shared flavor. 2029 * The assumption of this routine is the "flavor" is already a valid 2030 * flavor in the secinfo list of "exi". 2031 * 2032 * e.g. 2033 * # share -o sec=flavor1 /export 2034 * # share -o sec=flavor2 /export/home 2035 * 2036 * flavor2 is not an explicitly shared flavor for /export, 2037 * however it is in the secinfo list for /export thru the 2038 * server namespace setup. 2039 */ 2040 int 2041 is_exported_sec(int flavor, struct exportinfo *exi) 2042 { 2043 int i; 2044 struct secinfo *sp; 2045 2046 sp = exi->exi_export.ex_secinfo; 2047 for (i = 0; i < exi->exi_export.ex_seccnt; i++) { 2048 if (flavor == sp[i].s_secinfo.sc_nfsnum || 2049 sp[i].s_secinfo.sc_nfsnum == AUTH_NONE) { 2050 return (SEC_REF_EXPORTED(&sp[i])); 2051 } 2052 } 2053 2054 /* Should not reach this point based on the assumption */ 2055 return (0); 2056 } 2057 2058 /* 2059 * Check if the security flavor used in the request matches what is 2060 * required at the export point or at the root pseudo node (exi_root). 2061 * 2062 * returns 1 if there's a match or if exported with AUTH_NONE; 0 otherwise. 2063 * 2064 */ 2065 static int 2066 secinfo_match_or_authnone(struct compound_state *cs) 2067 { 2068 int i; 2069 struct secinfo *sp; 2070 2071 /* 2072 * Check cs->nfsflavor (from the request) against 2073 * the current export data in cs->exi. 2074 */ 2075 sp = cs->exi->exi_export.ex_secinfo; 2076 for (i = 0; i < cs->exi->exi_export.ex_seccnt; i++) { 2077 if (cs->nfsflavor == sp[i].s_secinfo.sc_nfsnum || 2078 sp[i].s_secinfo.sc_nfsnum == AUTH_NONE) 2079 return (1); 2080 } 2081 2082 return (0); 2083 } 2084 2085 /* 2086 * Check the access authority for the client and return the correct error. 2087 */ 2088 nfsstat4 2089 call_checkauth4(struct compound_state *cs, struct svc_req *req) 2090 { 2091 int authres; 2092 2093 /* 2094 * First, check if the security flavor used in the request 2095 * are among the flavors set in the server namespace. 2096 */ 2097 if (!secinfo_match_or_authnone(cs)) { 2098 *cs->statusp = NFS4ERR_WRONGSEC; 2099 return (*cs->statusp); 2100 } 2101 2102 authres = checkauth4(cs, req); 2103 2104 if (authres > 0) { 2105 *cs->statusp = NFS4_OK; 2106 if (! (cs->access & CS_ACCESS_LIMITED)) 2107 cs->access = CS_ACCESS_OK; 2108 } else if (authres == 0) { 2109 *cs->statusp = NFS4ERR_ACCESS; 2110 } else if (authres == -2) { 2111 *cs->statusp = NFS4ERR_WRONGSEC; 2112 } else { 2113 *cs->statusp = NFS4ERR_DELAY; 2114 } 2115 return (*cs->statusp); 2116 } 2117 2118 /* 2119 * bitmap4_to_attrmask is called by getattr and readdir. 2120 * It sets up the vattr mask and determines whether vfsstat call is needed 2121 * based on the input bitmap. 2122 * Returns nfsv4 status. 2123 */ 2124 static nfsstat4 2125 bitmap4_to_attrmask(bitmap4 breq, struct nfs4_svgetit_arg *sargp) 2126 { 2127 int i; 2128 uint_t va_mask; 2129 struct statvfs64 *sbp = sargp->sbp; 2130 2131 sargp->sbp = NULL; 2132 sargp->flag = 0; 2133 sargp->rdattr_error = NFS4_OK; 2134 sargp->mntdfid_set = FALSE; 2135 if (sargp->cs->vp) 2136 sargp->xattr = get_fh4_flag(&sargp->cs->fh, 2137 FH4_ATTRDIR | FH4_NAMEDATTR); 2138 else 2139 sargp->xattr = 0; 2140 2141 /* 2142 * Set rdattr_error_req to true if return error per 2143 * failed entry rather than fail the readdir. 2144 */ 2145 if (breq & FATTR4_RDATTR_ERROR_MASK) 2146 sargp->rdattr_error_req = 1; 2147 else 2148 sargp->rdattr_error_req = 0; 2149 2150 /* 2151 * generate the va_mask 2152 * Handle the easy cases first 2153 */ 2154 switch (breq) { 2155 case NFS4_NTOV_ATTR_MASK: 2156 sargp->vap->va_mask = NFS4_NTOV_ATTR_AT_MASK; 2157 return (NFS4_OK); 2158 2159 case NFS4_FS_ATTR_MASK: 2160 sargp->vap->va_mask = NFS4_FS_ATTR_AT_MASK; 2161 sargp->sbp = sbp; 2162 return (NFS4_OK); 2163 2164 case NFS4_NTOV_ATTR_CACHE_MASK: 2165 sargp->vap->va_mask = NFS4_NTOV_ATTR_CACHE_AT_MASK; 2166 return (NFS4_OK); 2167 2168 case FATTR4_LEASE_TIME_MASK: 2169 sargp->vap->va_mask = 0; 2170 return (NFS4_OK); 2171 2172 default: 2173 va_mask = 0; 2174 for (i = 0; i < nfs4_ntov_map_size; i++) { 2175 if ((breq & nfs4_ntov_map[i].fbit) && 2176 nfs4_ntov_map[i].vbit) 2177 va_mask |= nfs4_ntov_map[i].vbit; 2178 } 2179 2180 /* 2181 * Check is vfsstat is needed 2182 */ 2183 if (breq & NFS4_FS_ATTR_MASK) 2184 sargp->sbp = sbp; 2185 2186 sargp->vap->va_mask = va_mask; 2187 return (NFS4_OK); 2188 } 2189 /* NOTREACHED */ 2190 } 2191 2192 /* 2193 * bitmap4_get_sysattrs is called by getattr and readdir. 2194 * It calls both VOP_GETATTR and VFS_STATVFS calls to get the attrs. 2195 * Returns nfsv4 status. 2196 */ 2197 static nfsstat4 2198 bitmap4_get_sysattrs(struct nfs4_svgetit_arg *sargp) 2199 { 2200 int error; 2201 struct compound_state *cs = sargp->cs; 2202 vnode_t *vp = cs->vp; 2203 2204 if (sargp->sbp != NULL) { 2205 if (error = VFS_STATVFS(vp->v_vfsp, sargp->sbp)) { 2206 sargp->sbp = NULL; /* to identify error */ 2207 return (puterrno4(error)); 2208 } 2209 } 2210 2211 return (rfs4_vop_getattr(vp, sargp->vap, 0, cs->cr)); 2212 } 2213 2214 static void 2215 nfs4_ntov_table_init(struct nfs4_ntov_table *ntovp) 2216 { 2217 ntovp->na = kmem_zalloc(sizeof (union nfs4_attr_u) * nfs4_ntov_map_size, 2218 KM_SLEEP); 2219 ntovp->attrcnt = 0; 2220 ntovp->vfsstat = FALSE; 2221 } 2222 2223 static void 2224 nfs4_ntov_table_free(struct nfs4_ntov_table *ntovp, 2225 struct nfs4_svgetit_arg *sargp) 2226 { 2227 int i; 2228 union nfs4_attr_u *na; 2229 uint8_t *amap; 2230 2231 /* 2232 * XXX Should do the same checks for whether the bit is set 2233 */ 2234 for (i = 0, na = ntovp->na, amap = ntovp->amap; 2235 i < ntovp->attrcnt; i++, na++, amap++) { 2236 (void) (*nfs4_ntov_map[*amap].sv_getit)( 2237 NFS4ATTR_FREEIT, sargp, na); 2238 } 2239 if ((sargp->op == NFS4ATTR_SETIT) || (sargp->op == NFS4ATTR_VERIT)) { 2240 /* 2241 * xdr_free for getattr will be done later 2242 */ 2243 for (i = 0, na = ntovp->na, amap = ntovp->amap; 2244 i < ntovp->attrcnt; i++, na++, amap++) { 2245 xdr_free(nfs4_ntov_map[*amap].xfunc, (caddr_t)na); 2246 } 2247 } 2248 kmem_free(ntovp->na, sizeof (union nfs4_attr_u) * nfs4_ntov_map_size); 2249 } 2250 2251 /* 2252 * do_rfs4_op_getattr gets the system attrs and converts into fattr4. 2253 */ 2254 static nfsstat4 2255 do_rfs4_op_getattr(bitmap4 breq, fattr4 *fattrp, 2256 struct nfs4_svgetit_arg *sargp) 2257 { 2258 int error = 0; 2259 int i, k; 2260 struct nfs4_ntov_table ntov; 2261 XDR xdr; 2262 ulong_t xdr_size; 2263 char *xdr_attrs; 2264 nfsstat4 status = NFS4_OK; 2265 nfsstat4 prev_rdattr_error = sargp->rdattr_error; 2266 union nfs4_attr_u *na; 2267 uint8_t *amap; 2268 2269 sargp->op = NFS4ATTR_GETIT; 2270 sargp->flag = 0; 2271 2272 fattrp->attrmask = 0; 2273 /* if no bits requested, then return empty fattr4 */ 2274 if (breq == 0) { 2275 fattrp->attrlist4_len = 0; 2276 fattrp->attrlist4 = NULL; 2277 return (NFS4_OK); 2278 } 2279 2280 /* 2281 * return NFS4ERR_INVAL when client requests write-only attrs 2282 */ 2283 if (breq & (FATTR4_TIME_ACCESS_SET_MASK | FATTR4_TIME_MODIFY_SET_MASK)) 2284 return (NFS4ERR_INVAL); 2285 2286 nfs4_ntov_table_init(&ntov); 2287 na = ntov.na; 2288 amap = ntov.amap; 2289 2290 /* 2291 * Now loop to get or verify the attrs 2292 */ 2293 for (i = 0; i < nfs4_ntov_map_size; i++) { 2294 if (breq & nfs4_ntov_map[i].fbit) { 2295 if ((*nfs4_ntov_map[i].sv_getit)( 2296 NFS4ATTR_SUPPORTED, sargp, NULL) == 0) { 2297 2298 error = (*nfs4_ntov_map[i].sv_getit)( 2299 NFS4ATTR_GETIT, sargp, na); 2300 2301 /* 2302 * Possible error values: 2303 * >0 if sv_getit failed to 2304 * get the attr; 0 if succeeded; 2305 * <0 if rdattr_error and the 2306 * attribute cannot be returned. 2307 */ 2308 if (error && !(sargp->rdattr_error_req)) 2309 goto done; 2310 /* 2311 * If error then just for entry 2312 */ 2313 if (error == 0) { 2314 fattrp->attrmask |= 2315 nfs4_ntov_map[i].fbit; 2316 *amap++ = 2317 (uint8_t)nfs4_ntov_map[i].nval; 2318 na++; 2319 (ntov.attrcnt)++; 2320 } else if ((error > 0) && 2321 (sargp->rdattr_error == NFS4_OK)) { 2322 sargp->rdattr_error = puterrno4(error); 2323 } 2324 error = 0; 2325 } 2326 } 2327 } 2328 2329 /* 2330 * If rdattr_error was set after the return value for it was assigned, 2331 * update it. 2332 */ 2333 if (prev_rdattr_error != sargp->rdattr_error) { 2334 na = ntov.na; 2335 amap = ntov.amap; 2336 for (i = 0; i < ntov.attrcnt; i++, na++, amap++) { 2337 k = *amap; 2338 if (k < FATTR4_RDATTR_ERROR) { 2339 continue; 2340 } 2341 if ((k == FATTR4_RDATTR_ERROR) && 2342 ((*nfs4_ntov_map[k].sv_getit)( 2343 NFS4ATTR_SUPPORTED, sargp, NULL) == 0)) { 2344 2345 (void) (*nfs4_ntov_map[k].sv_getit)( 2346 NFS4ATTR_GETIT, sargp, na); 2347 } 2348 break; 2349 } 2350 } 2351 2352 xdr_size = 0; 2353 na = ntov.na; 2354 amap = ntov.amap; 2355 for (i = 0; i < ntov.attrcnt; i++, na++, amap++) { 2356 xdr_size += xdr_sizeof(nfs4_ntov_map[*amap].xfunc, na); 2357 } 2358 2359 fattrp->attrlist4_len = xdr_size; 2360 if (xdr_size) { 2361 /* freed by rfs4_op_getattr_free() */ 2362 fattrp->attrlist4 = xdr_attrs = kmem_zalloc(xdr_size, KM_SLEEP); 2363 2364 xdrmem_create(&xdr, xdr_attrs, xdr_size, XDR_ENCODE); 2365 2366 na = ntov.na; 2367 amap = ntov.amap; 2368 for (i = 0; i < ntov.attrcnt; i++, na++, amap++) { 2369 if (!(*nfs4_ntov_map[*amap].xfunc)(&xdr, na)) { 2370 DTRACE_PROBE1(nfss__e__getattr4_encfail, 2371 int, *amap); 2372 status = NFS4ERR_SERVERFAULT; 2373 break; 2374 } 2375 } 2376 /* xdrmem_destroy(&xdrs); */ /* NO-OP */ 2377 } else { 2378 fattrp->attrlist4 = NULL; 2379 } 2380 done: 2381 2382 nfs4_ntov_table_free(&ntov, sargp); 2383 2384 if (error != 0) 2385 status = puterrno4(error); 2386 2387 return (status); 2388 } 2389 2390 /* ARGSUSED */ 2391 static void 2392 rfs4_op_getattr(nfs_argop4 *argop, nfs_resop4 *resop, struct svc_req *req, 2393 struct compound_state *cs) 2394 { 2395 GETATTR4args *args = &argop->nfs_argop4_u.opgetattr; 2396 GETATTR4res *resp = &resop->nfs_resop4_u.opgetattr; 2397 struct nfs4_svgetit_arg sarg; 2398 struct statvfs64 sb; 2399 nfsstat4 status; 2400 2401 DTRACE_NFSV4_2(op__getattr__start, struct compound_state *, cs, 2402 GETATTR4args *, args); 2403 2404 if (cs->vp == NULL) { 2405 *cs->statusp = resp->status = NFS4ERR_NOFILEHANDLE; 2406 goto out; 2407 } 2408 2409 if (cs->access == CS_ACCESS_DENIED) { 2410 *cs->statusp = resp->status = NFS4ERR_ACCESS; 2411 goto out; 2412 } 2413 2414 sarg.sbp = &sb; 2415 sarg.cs = cs; 2416 sarg.is_referral = B_FALSE; 2417 2418 status = bitmap4_to_attrmask(args->attr_request, &sarg); 2419 if (status == NFS4_OK) { 2420 2421 status = bitmap4_get_sysattrs(&sarg); 2422 if (status == NFS4_OK) { 2423 2424 /* Is this a referral? */ 2425 if (vn_is_nfs_reparse(cs->vp, cs->cr)) { 2426 /* Older V4 Solaris client sees a link */ 2427 if (client_is_downrev(req)) 2428 sarg.vap->va_type = VLNK; 2429 else 2430 sarg.is_referral = B_TRUE; 2431 } 2432 2433 status = do_rfs4_op_getattr(args->attr_request, 2434 &resp->obj_attributes, &sarg); 2435 } 2436 } 2437 *cs->statusp = resp->status = status; 2438 out: 2439 DTRACE_NFSV4_2(op__getattr__done, struct compound_state *, cs, 2440 GETATTR4res *, resp); 2441 } 2442 2443 static void 2444 rfs4_op_getattr_free(nfs_resop4 *resop) 2445 { 2446 GETATTR4res *resp = &resop->nfs_resop4_u.opgetattr; 2447 2448 nfs4_fattr4_free(&resp->obj_attributes); 2449 } 2450 2451 /* ARGSUSED */ 2452 static void 2453 rfs4_op_getfh(nfs_argop4 *argop, nfs_resop4 *resop, struct svc_req *req, 2454 struct compound_state *cs) 2455 { 2456 GETFH4res *resp = &resop->nfs_resop4_u.opgetfh; 2457 2458 DTRACE_NFSV4_1(op__getfh__start, struct compound_state *, cs); 2459 2460 if (cs->vp == NULL) { 2461 *cs->statusp = resp->status = NFS4ERR_NOFILEHANDLE; 2462 goto out; 2463 } 2464 if (cs->access == CS_ACCESS_DENIED) { 2465 *cs->statusp = resp->status = NFS4ERR_ACCESS; 2466 goto out; 2467 } 2468 2469 /* check for reparse point at the share point */ 2470 if (cs->exi->exi_moved || vn_is_nfs_reparse(cs->exi->exi_vp, cs->cr)) { 2471 /* it's all bad */ 2472 cs->exi->exi_moved = 1; 2473 *cs->statusp = resp->status = NFS4ERR_MOVED; 2474 DTRACE_PROBE2(nfs4serv__func__referral__shared__moved, 2475 vnode_t *, cs->vp, char *, "rfs4_op_getfh"); 2476 return; 2477 } 2478 2479 /* check for reparse point at vp */ 2480 if (vn_is_nfs_reparse(cs->vp, cs->cr) && !client_is_downrev(req)) { 2481 /* it's not all bad */ 2482 *cs->statusp = resp->status = NFS4ERR_MOVED; 2483 DTRACE_PROBE2(nfs4serv__func__referral__moved, 2484 vnode_t *, cs->vp, char *, "rfs4_op_getfh"); 2485 return; 2486 } 2487 2488 resp->object.nfs_fh4_val = 2489 kmem_alloc(cs->fh.nfs_fh4_len, KM_SLEEP); 2490 nfs_fh4_copy(&cs->fh, &resp->object); 2491 *cs->statusp = resp->status = NFS4_OK; 2492 out: 2493 DTRACE_NFSV4_2(op__getfh__done, struct compound_state *, cs, 2494 GETFH4res *, resp); 2495 } 2496 2497 static void 2498 rfs4_op_getfh_free(nfs_resop4 *resop) 2499 { 2500 GETFH4res *resp = &resop->nfs_resop4_u.opgetfh; 2501 2502 if (resp->status == NFS4_OK && 2503 resp->object.nfs_fh4_val != NULL) { 2504 kmem_free(resp->object.nfs_fh4_val, resp->object.nfs_fh4_len); 2505 resp->object.nfs_fh4_val = NULL; 2506 resp->object.nfs_fh4_len = 0; 2507 } 2508 } 2509 2510 /* 2511 * illegal: args: void 2512 * res : status (NFS4ERR_OP_ILLEGAL) 2513 */ 2514 /* ARGSUSED */ 2515 static void 2516 rfs4_op_illegal(nfs_argop4 *argop, nfs_resop4 *resop, 2517 struct svc_req *req, struct compound_state *cs) 2518 { 2519 ILLEGAL4res *resp = &resop->nfs_resop4_u.opillegal; 2520 2521 resop->resop = OP_ILLEGAL; 2522 *cs->statusp = resp->status = NFS4ERR_OP_ILLEGAL; 2523 } 2524 2525 /* 2526 * link: args: SAVED_FH: file, CURRENT_FH: target directory 2527 * res: status. If success - CURRENT_FH unchanged, return change_info 2528 */ 2529 /* ARGSUSED */ 2530 static void 2531 rfs4_op_link(nfs_argop4 *argop, nfs_resop4 *resop, struct svc_req *req, 2532 struct compound_state *cs) 2533 { 2534 LINK4args *args = &argop->nfs_argop4_u.oplink; 2535 LINK4res *resp = &resop->nfs_resop4_u.oplink; 2536 int error; 2537 vnode_t *vp; 2538 vnode_t *dvp; 2539 struct vattr bdva, idva, adva; 2540 char *nm; 2541 uint_t len; 2542 struct sockaddr *ca; 2543 char *name = NULL; 2544 nfsstat4 status; 2545 2546 DTRACE_NFSV4_2(op__link__start, struct compound_state *, cs, 2547 LINK4args *, args); 2548 2549 /* SAVED_FH: source object */ 2550 vp = cs->saved_vp; 2551 if (vp == NULL) { 2552 *cs->statusp = resp->status = NFS4ERR_NOFILEHANDLE; 2553 goto out; 2554 } 2555 2556 /* CURRENT_FH: target directory */ 2557 dvp = cs->vp; 2558 if (dvp == NULL) { 2559 *cs->statusp = resp->status = NFS4ERR_NOFILEHANDLE; 2560 goto out; 2561 } 2562 2563 /* 2564 * If there is a non-shared filesystem mounted on this vnode, 2565 * do not allow to link any file in this directory. 2566 */ 2567 if (vn_ismntpt(dvp)) { 2568 *cs->statusp = resp->status = NFS4ERR_ACCESS; 2569 goto out; 2570 } 2571 2572 if (cs->access == CS_ACCESS_DENIED) { 2573 *cs->statusp = resp->status = NFS4ERR_ACCESS; 2574 goto out; 2575 } 2576 2577 /* Check source object's type validity */ 2578 if (vp->v_type == VDIR) { 2579 *cs->statusp = resp->status = NFS4ERR_ISDIR; 2580 goto out; 2581 } 2582 2583 /* Check target directory's type */ 2584 if (dvp->v_type != VDIR) { 2585 *cs->statusp = resp->status = NFS4ERR_NOTDIR; 2586 goto out; 2587 } 2588 2589 if (cs->saved_exi != cs->exi) { 2590 *cs->statusp = resp->status = NFS4ERR_XDEV; 2591 goto out; 2592 } 2593 2594 status = utf8_dir_verify(&args->newname); 2595 if (status != NFS4_OK) { 2596 *cs->statusp = resp->status = status; 2597 goto out; 2598 } 2599 2600 nm = utf8_to_fn(&args->newname, &len, NULL); 2601 if (nm == NULL) { 2602 *cs->statusp = resp->status = NFS4ERR_INVAL; 2603 goto out; 2604 } 2605 2606 if (len > MAXNAMELEN) { 2607 *cs->statusp = resp->status = NFS4ERR_NAMETOOLONG; 2608 kmem_free(nm, len); 2609 goto out; 2610 } 2611 2612 if (rdonly4(req, cs)) { 2613 *cs->statusp = resp->status = NFS4ERR_ROFS; 2614 kmem_free(nm, len); 2615 goto out; 2616 } 2617 2618 /* Get "before" change value */ 2619 bdva.va_mask = AT_CTIME|AT_SEQ; 2620 error = VOP_GETATTR(dvp, &bdva, 0, cs->cr, NULL); 2621 if (error) { 2622 *cs->statusp = resp->status = puterrno4(error); 2623 kmem_free(nm, len); 2624 goto out; 2625 } 2626 2627 ca = (struct sockaddr *)svc_getrpccaller(req->rq_xprt)->buf; 2628 name = nfscmd_convname(ca, cs->exi, nm, NFSCMD_CONV_INBOUND, 2629 MAXPATHLEN + 1); 2630 2631 if (name == NULL) { 2632 *cs->statusp = resp->status = NFS4ERR_INVAL; 2633 kmem_free(nm, len); 2634 goto out; 2635 } 2636 2637 NFS4_SET_FATTR4_CHANGE(resp->cinfo.before, bdva.va_ctime) 2638 2639 error = VOP_LINK(dvp, vp, name, cs->cr, NULL, 0); 2640 2641 if (nm != name) 2642 kmem_free(name, MAXPATHLEN + 1); 2643 kmem_free(nm, len); 2644 2645 /* 2646 * Get the initial "after" sequence number, if it fails, set to zero 2647 */ 2648 idva.va_mask = AT_SEQ; 2649 if (VOP_GETATTR(dvp, &idva, 0, cs->cr, NULL)) 2650 idva.va_seq = 0; 2651 2652 /* 2653 * Force modified data and metadata out to stable storage. 2654 */ 2655 (void) VOP_FSYNC(vp, FNODSYNC, cs->cr, NULL); 2656 (void) VOP_FSYNC(dvp, 0, cs->cr, NULL); 2657 2658 if (error) { 2659 *cs->statusp = resp->status = puterrno4(error); 2660 goto out; 2661 } 2662 2663 /* 2664 * Get "after" change value, if it fails, simply return the 2665 * before value. 2666 */ 2667 adva.va_mask = AT_CTIME|AT_SEQ; 2668 if (VOP_GETATTR(dvp, &adva, 0, cs->cr, NULL)) { 2669 adva.va_ctime = bdva.va_ctime; 2670 adva.va_seq = 0; 2671 } 2672 2673 NFS4_SET_FATTR4_CHANGE(resp->cinfo.after, adva.va_ctime) 2674 2675 /* 2676 * The cinfo.atomic = TRUE only if we have 2677 * non-zero va_seq's, and it has incremented by exactly one 2678 * during the VOP_LINK and it didn't change during the VOP_FSYNC. 2679 */ 2680 if (bdva.va_seq && idva.va_seq && adva.va_seq && 2681 idva.va_seq == (bdva.va_seq + 1) && idva.va_seq == adva.va_seq) 2682 resp->cinfo.atomic = TRUE; 2683 else 2684 resp->cinfo.atomic = FALSE; 2685 2686 *cs->statusp = resp->status = NFS4_OK; 2687 out: 2688 DTRACE_NFSV4_2(op__link__done, struct compound_state *, cs, 2689 LINK4res *, resp); 2690 } 2691 2692 /* 2693 * Used by rfs4_op_lookup and rfs4_op_lookupp to do the actual work. 2694 */ 2695 2696 /* ARGSUSED */ 2697 static nfsstat4 2698 do_rfs4_op_lookup(char *nm, struct svc_req *req, struct compound_state *cs) 2699 { 2700 int error; 2701 int different_export = 0; 2702 vnode_t *vp, *pre_tvp = NULL, *oldvp = NULL; 2703 struct exportinfo *exi = NULL, *pre_exi = NULL; 2704 nfsstat4 stat; 2705 fid_t fid; 2706 int attrdir, dotdot, walk; 2707 bool_t is_newvp = FALSE; 2708 2709 if (cs->vp->v_flag & V_XATTRDIR) { 2710 attrdir = 1; 2711 ASSERT(get_fh4_flag(&cs->fh, FH4_ATTRDIR)); 2712 } else { 2713 attrdir = 0; 2714 ASSERT(! get_fh4_flag(&cs->fh, FH4_ATTRDIR)); 2715 } 2716 2717 dotdot = (nm[0] == '.' && nm[1] == '.' && nm[2] == '\0'); 2718 2719 /* 2720 * If dotdotting, then need to check whether it's 2721 * above the root of a filesystem, or above an 2722 * export point. 2723 */ 2724 if (dotdot) { 2725 vnode_t *zone_rootvp; 2726 2727 ASSERT(cs->exi != NULL); 2728 zone_rootvp = cs->exi->exi_ne->exi_root->exi_vp; 2729 /* 2730 * If dotdotting at the root of a filesystem, then 2731 * need to traverse back to the mounted-on filesystem 2732 * and do the dotdot lookup there. 2733 */ 2734 if ((cs->vp->v_flag & VROOT) || VN_CMP(cs->vp, zone_rootvp)) { 2735 2736 /* 2737 * If at the system root, then can 2738 * go up no further. 2739 */ 2740 if (VN_CMP(cs->vp, zone_rootvp)) 2741 return (puterrno4(ENOENT)); 2742 2743 /* 2744 * Traverse back to the mounted-on filesystem 2745 */ 2746 cs->vp = untraverse(cs->vp, zone_rootvp); 2747 2748 /* 2749 * Set the different_export flag so we remember 2750 * to pick up a new exportinfo entry for 2751 * this new filesystem. 2752 */ 2753 different_export = 1; 2754 } else { 2755 2756 /* 2757 * If dotdotting above an export point then set 2758 * the different_export to get new export info. 2759 */ 2760 different_export = nfs_exported(cs->exi, cs->vp); 2761 } 2762 } 2763 2764 error = VOP_LOOKUP(cs->vp, nm, &vp, NULL, 0, NULL, cs->cr, 2765 NULL, NULL, NULL); 2766 if (error) 2767 return (puterrno4(error)); 2768 2769 /* 2770 * If the vnode is in a pseudo filesystem, check whether it is visible. 2771 * 2772 * XXX if the vnode is a symlink and it is not visible in 2773 * a pseudo filesystem, return ENOENT (not following symlink). 2774 * V4 client can not mount such symlink. This is a regression 2775 * from V2/V3. 2776 * 2777 * In the same exported filesystem, if the security flavor used 2778 * is not an explicitly shared flavor, limit the view to the visible 2779 * list entries only. This is not a WRONGSEC case because it's already 2780 * checked via PUTROOTFH/PUTPUBFH or PUTFH. 2781 */ 2782 if (!different_export && 2783 (PSEUDO(cs->exi) || ! is_exported_sec(cs->nfsflavor, cs->exi) || 2784 cs->access & CS_ACCESS_LIMITED)) { 2785 if (! nfs_visible(cs->exi, vp, &different_export)) { 2786 VN_RELE(vp); 2787 return (puterrno4(ENOENT)); 2788 } 2789 } 2790 2791 /* 2792 * If it's a mountpoint, then traverse it. 2793 */ 2794 if (vn_ismntpt(vp)) { 2795 pre_exi = cs->exi; /* save pre-traversed exportinfo */ 2796 pre_tvp = vp; /* save pre-traversed vnode */ 2797 2798 /* 2799 * hold pre_tvp to counteract rele by traverse. We will 2800 * need pre_tvp below if checkexport4 fails 2801 */ 2802 VN_HOLD(pre_tvp); 2803 if ((error = traverse(&vp)) != 0) { 2804 VN_RELE(vp); 2805 VN_RELE(pre_tvp); 2806 return (puterrno4(error)); 2807 } 2808 different_export = 1; 2809 } else if (vp->v_vfsp != cs->vp->v_vfsp) { 2810 /* 2811 * The vfsp comparison is to handle the case where 2812 * a LOFS mount is shared. lo_lookup traverses mount points, 2813 * and NFS is unaware of local fs transistions because 2814 * v_vfsmountedhere isn't set. For this special LOFS case, 2815 * the dir and the obj returned by lookup will have different 2816 * vfs ptrs. 2817 */ 2818 different_export = 1; 2819 } 2820 2821 if (different_export) { 2822 2823 bzero(&fid, sizeof (fid)); 2824 fid.fid_len = MAXFIDSZ; 2825 error = vop_fid_pseudo(vp, &fid); 2826 if (error) { 2827 VN_RELE(vp); 2828 if (pre_tvp) 2829 VN_RELE(pre_tvp); 2830 return (puterrno4(error)); 2831 } 2832 2833 if (dotdot) 2834 exi = nfs_vptoexi(NULL, vp, cs->cr, &walk, NULL, TRUE); 2835 else 2836 exi = checkexport4(&vp->v_vfsp->vfs_fsid, &fid, vp); 2837 2838 if (exi == NULL) { 2839 if (pre_tvp) { 2840 /* 2841 * If this vnode is a mounted-on vnode, 2842 * but the mounted-on file system is not 2843 * exported, send back the filehandle for 2844 * the mounted-on vnode, not the root of 2845 * the mounted-on file system. 2846 */ 2847 VN_RELE(vp); 2848 vp = pre_tvp; 2849 exi = pre_exi; 2850 } else { 2851 VN_RELE(vp); 2852 return (puterrno4(EACCES)); 2853 } 2854 } else if (pre_tvp) { 2855 /* we're done with pre_tvp now. release extra hold */ 2856 VN_RELE(pre_tvp); 2857 } 2858 2859 cs->exi = exi; 2860 2861 /* 2862 * Now we do a checkauth4. The reason is that 2863 * this client/user may not have access to the new 2864 * exported file system, and if they do, 2865 * the client/user may be mapped to a different uid. 2866 * 2867 * We start with a new cr, because the checkauth4 done 2868 * in the PUT*FH operation over wrote the cred's uid, 2869 * gid, etc, and we want the real thing before calling 2870 * checkauth4() 2871 */ 2872 crfree(cs->cr); 2873 cs->cr = crdup(cs->basecr); 2874 2875 oldvp = cs->vp; 2876 cs->vp = vp; 2877 is_newvp = TRUE; 2878 2879 stat = call_checkauth4(cs, req); 2880 if (stat != NFS4_OK) { 2881 VN_RELE(cs->vp); 2882 cs->vp = oldvp; 2883 return (stat); 2884 } 2885 } 2886 2887 /* 2888 * After various NFS checks, do a label check on the path 2889 * component. The label on this path should either be the 2890 * global zone's label or a zone's label. We are only 2891 * interested in the zone's label because exported files 2892 * in global zone is accessible (though read-only) to 2893 * clients. The exportability/visibility check is already 2894 * done before reaching this code. 2895 */ 2896 if (is_system_labeled()) { 2897 bslabel_t *clabel; 2898 2899 ASSERT(req->rq_label != NULL); 2900 clabel = req->rq_label; 2901 DTRACE_PROBE2(tx__rfs4__log__info__oplookup__clabel, char *, 2902 "got client label from request(1)", struct svc_req *, req); 2903 2904 if (!blequal(&l_admin_low->tsl_label, clabel)) { 2905 if (!do_rfs_label_check(clabel, vp, DOMINANCE_CHECK, 2906 cs->exi)) { 2907 error = EACCES; 2908 goto err_out; 2909 } 2910 } else { 2911 /* 2912 * We grant access to admin_low label clients 2913 * only if the client is trusted, i.e. also 2914 * running Solaris Trusted Extension. 2915 */ 2916 struct sockaddr *ca; 2917 int addr_type; 2918 void *ipaddr; 2919 tsol_tpc_t *tp; 2920 2921 ca = (struct sockaddr *)svc_getrpccaller( 2922 req->rq_xprt)->buf; 2923 if (ca->sa_family == AF_INET) { 2924 addr_type = IPV4_VERSION; 2925 ipaddr = &((struct sockaddr_in *)ca)->sin_addr; 2926 } else if (ca->sa_family == AF_INET6) { 2927 addr_type = IPV6_VERSION; 2928 ipaddr = &((struct sockaddr_in6 *) 2929 ca)->sin6_addr; 2930 } 2931 tp = find_tpc(ipaddr, addr_type, B_FALSE); 2932 if (tp == NULL || tp->tpc_tp.tp_doi != 2933 l_admin_low->tsl_doi || tp->tpc_tp.host_type != 2934 SUN_CIPSO) { 2935 if (tp != NULL) 2936 TPC_RELE(tp); 2937 error = EACCES; 2938 goto err_out; 2939 } 2940 TPC_RELE(tp); 2941 } 2942 } 2943 2944 error = makefh4(&cs->fh, vp, cs->exi); 2945 2946 err_out: 2947 if (error) { 2948 if (is_newvp) { 2949 VN_RELE(cs->vp); 2950 cs->vp = oldvp; 2951 } else 2952 VN_RELE(vp); 2953 return (puterrno4(error)); 2954 } 2955 2956 if (!is_newvp) { 2957 if (cs->vp) 2958 VN_RELE(cs->vp); 2959 cs->vp = vp; 2960 } else if (oldvp) 2961 VN_RELE(oldvp); 2962 2963 /* 2964 * if did lookup on attrdir and didn't lookup .., set named 2965 * attr fh flag 2966 */ 2967 if (attrdir && ! dotdot) 2968 set_fh4_flag(&cs->fh, FH4_NAMEDATTR); 2969 2970 /* Assume false for now, open proc will set this */ 2971 cs->mandlock = FALSE; 2972 2973 return (NFS4_OK); 2974 } 2975 2976 /* ARGSUSED */ 2977 static void 2978 rfs4_op_lookup(nfs_argop4 *argop, nfs_resop4 *resop, struct svc_req *req, 2979 struct compound_state *cs) 2980 { 2981 LOOKUP4args *args = &argop->nfs_argop4_u.oplookup; 2982 LOOKUP4res *resp = &resop->nfs_resop4_u.oplookup; 2983 char *nm; 2984 uint_t len; 2985 struct sockaddr *ca; 2986 char *name = NULL; 2987 nfsstat4 status; 2988 2989 DTRACE_NFSV4_2(op__lookup__start, struct compound_state *, cs, 2990 LOOKUP4args *, args); 2991 2992 if (cs->vp == NULL) { 2993 *cs->statusp = resp->status = NFS4ERR_NOFILEHANDLE; 2994 goto out; 2995 } 2996 2997 if (cs->vp->v_type == VLNK) { 2998 *cs->statusp = resp->status = NFS4ERR_SYMLINK; 2999 goto out; 3000 } 3001 3002 if (cs->vp->v_type != VDIR) { 3003 *cs->statusp = resp->status = NFS4ERR_NOTDIR; 3004 goto out; 3005 } 3006 3007 status = utf8_dir_verify(&args->objname); 3008 if (status != NFS4_OK) { 3009 *cs->statusp = resp->status = status; 3010 goto out; 3011 } 3012 3013 nm = utf8_to_str(&args->objname, &len, NULL); 3014 if (nm == NULL) { 3015 *cs->statusp = resp->status = NFS4ERR_INVAL; 3016 goto out; 3017 } 3018 3019 if (len > MAXNAMELEN) { 3020 *cs->statusp = resp->status = NFS4ERR_NAMETOOLONG; 3021 kmem_free(nm, len); 3022 goto out; 3023 } 3024 3025 ca = (struct sockaddr *)svc_getrpccaller(req->rq_xprt)->buf; 3026 name = nfscmd_convname(ca, cs->exi, nm, NFSCMD_CONV_INBOUND, 3027 MAXPATHLEN + 1); 3028 3029 if (name == NULL) { 3030 *cs->statusp = resp->status = NFS4ERR_INVAL; 3031 kmem_free(nm, len); 3032 goto out; 3033 } 3034 3035 *cs->statusp = resp->status = do_rfs4_op_lookup(name, req, cs); 3036 3037 if (name != nm) 3038 kmem_free(name, MAXPATHLEN + 1); 3039 kmem_free(nm, len); 3040 3041 out: 3042 DTRACE_NFSV4_2(op__lookup__done, struct compound_state *, cs, 3043 LOOKUP4res *, resp); 3044 } 3045 3046 /* ARGSUSED */ 3047 static void 3048 rfs4_op_lookupp(nfs_argop4 *args, nfs_resop4 *resop, struct svc_req *req, 3049 struct compound_state *cs) 3050 { 3051 LOOKUPP4res *resp = &resop->nfs_resop4_u.oplookupp; 3052 3053 DTRACE_NFSV4_1(op__lookupp__start, struct compound_state *, cs); 3054 3055 if (cs->vp == NULL) { 3056 *cs->statusp = resp->status = NFS4ERR_NOFILEHANDLE; 3057 goto out; 3058 } 3059 3060 if (cs->vp->v_type != VDIR) { 3061 *cs->statusp = resp->status = NFS4ERR_NOTDIR; 3062 goto out; 3063 } 3064 3065 *cs->statusp = resp->status = do_rfs4_op_lookup("..", req, cs); 3066 3067 /* 3068 * From NFSV4 Specification, LOOKUPP should not check for 3069 * NFS4ERR_WRONGSEC. Retrun NFS4_OK instead. 3070 */ 3071 if (resp->status == NFS4ERR_WRONGSEC) { 3072 *cs->statusp = resp->status = NFS4_OK; 3073 } 3074 3075 out: 3076 DTRACE_NFSV4_2(op__lookupp__done, struct compound_state *, cs, 3077 LOOKUPP4res *, resp); 3078 } 3079 3080 3081 /*ARGSUSED2*/ 3082 static void 3083 rfs4_op_openattr(nfs_argop4 *argop, nfs_resop4 *resop, struct svc_req *req, 3084 struct compound_state *cs) 3085 { 3086 OPENATTR4args *args = &argop->nfs_argop4_u.opopenattr; 3087 OPENATTR4res *resp = &resop->nfs_resop4_u.opopenattr; 3088 vnode_t *avp = NULL; 3089 int lookup_flags = LOOKUP_XATTR, error; 3090 int exp_ro = 0; 3091 3092 DTRACE_NFSV4_2(op__openattr__start, struct compound_state *, cs, 3093 OPENATTR4args *, args); 3094 3095 if (cs->vp == NULL) { 3096 *cs->statusp = resp->status = NFS4ERR_NOFILEHANDLE; 3097 goto out; 3098 } 3099 3100 if ((cs->vp->v_vfsp->vfs_flag & VFS_XATTR) == 0 && 3101 !vfs_has_feature(cs->vp->v_vfsp, VFSFT_SYSATTR_VIEWS)) { 3102 *cs->statusp = resp->status = puterrno4(ENOTSUP); 3103 goto out; 3104 } 3105 3106 /* 3107 * If file system supports passing ACE mask to VOP_ACCESS then 3108 * check for ACE_READ_NAMED_ATTRS, otherwise do legacy checks 3109 */ 3110 3111 if (vfs_has_feature(cs->vp->v_vfsp, VFSFT_ACEMASKONACCESS)) 3112 error = VOP_ACCESS(cs->vp, ACE_READ_NAMED_ATTRS, 3113 V_ACE_MASK, cs->cr, NULL); 3114 else 3115 error = ((VOP_ACCESS(cs->vp, VREAD, 0, cs->cr, NULL) != 0) && 3116 (VOP_ACCESS(cs->vp, VWRITE, 0, cs->cr, NULL) != 0) && 3117 (VOP_ACCESS(cs->vp, VEXEC, 0, cs->cr, NULL) != 0)); 3118 3119 if (error) { 3120 *cs->statusp = resp->status = puterrno4(EACCES); 3121 goto out; 3122 } 3123 3124 /* 3125 * The CREATE_XATTR_DIR VOP flag cannot be specified if 3126 * the file system is exported read-only -- regardless of 3127 * createdir flag. Otherwise the attrdir would be created 3128 * (assuming server fs isn't mounted readonly locally). If 3129 * VOP_LOOKUP returns ENOENT in this case, the error will 3130 * be translated into EROFS. ENOSYS is mapped to ENOTSUP 3131 * because specfs has no VOP_LOOKUP op, so the macro would 3132 * return ENOSYS. EINVAL is returned by all (current) 3133 * Solaris file system implementations when any of their 3134 * restrictions are violated (xattr(dir) can't have xattrdir). 3135 * Returning NOTSUPP is more appropriate in this case 3136 * because the object will never be able to have an attrdir. 3137 */ 3138 if (args->createdir && ! (exp_ro = rdonly4(req, cs))) 3139 lookup_flags |= CREATE_XATTR_DIR; 3140 3141 error = VOP_LOOKUP(cs->vp, "", &avp, NULL, lookup_flags, NULL, cs->cr, 3142 NULL, NULL, NULL); 3143 3144 if (error) { 3145 if (error == ENOENT && args->createdir && exp_ro) 3146 *cs->statusp = resp->status = puterrno4(EROFS); 3147 else if (error == EINVAL || error == ENOSYS) 3148 *cs->statusp = resp->status = puterrno4(ENOTSUP); 3149 else 3150 *cs->statusp = resp->status = puterrno4(error); 3151 goto out; 3152 } 3153 3154 ASSERT(avp->v_flag & V_XATTRDIR); 3155 3156 error = makefh4(&cs->fh, avp, cs->exi); 3157 3158 if (error) { 3159 VN_RELE(avp); 3160 *cs->statusp = resp->status = puterrno4(error); 3161 goto out; 3162 } 3163 3164 VN_RELE(cs->vp); 3165 cs->vp = avp; 3166 3167 /* 3168 * There is no requirement for an attrdir fh flag 3169 * because the attrdir has a vnode flag to distinguish 3170 * it from regular (non-xattr) directories. The 3171 * FH4_ATTRDIR flag is set for future sanity checks. 3172 */ 3173 set_fh4_flag(&cs->fh, FH4_ATTRDIR); 3174 *cs->statusp = resp->status = NFS4_OK; 3175 3176 out: 3177 DTRACE_NFSV4_2(op__openattr__done, struct compound_state *, cs, 3178 OPENATTR4res *, resp); 3179 } 3180 3181 static int 3182 do_io(int direction, vnode_t *vp, struct uio *uio, int ioflag, cred_t *cred, 3183 caller_context_t *ct) 3184 { 3185 int error; 3186 int i; 3187 clock_t delaytime; 3188 3189 delaytime = MSEC_TO_TICK_ROUNDUP(rfs4_lock_delay); 3190 3191 /* 3192 * Don't block on mandatory locks. If this routine returns 3193 * EAGAIN, the caller should return NFS4ERR_LOCKED. 3194 */ 3195 uio->uio_fmode = FNONBLOCK; 3196 3197 for (i = 0; i < rfs4_maxlock_tries; i++) { 3198 3199 3200 if (direction == FREAD) { 3201 (void) VOP_RWLOCK(vp, V_WRITELOCK_FALSE, ct); 3202 error = VOP_READ(vp, uio, ioflag, cred, ct); 3203 VOP_RWUNLOCK(vp, V_WRITELOCK_FALSE, ct); 3204 } else { 3205 (void) VOP_RWLOCK(vp, V_WRITELOCK_TRUE, ct); 3206 error = VOP_WRITE(vp, uio, ioflag, cred, ct); 3207 VOP_RWUNLOCK(vp, V_WRITELOCK_TRUE, ct); 3208 } 3209 3210 if (error != EAGAIN) 3211 break; 3212 3213 if (i < rfs4_maxlock_tries - 1) { 3214 delay(delaytime); 3215 delaytime *= 2; 3216 } 3217 } 3218 3219 return (error); 3220 } 3221 3222 /* ARGSUSED */ 3223 static void 3224 rfs4_op_read(nfs_argop4 *argop, nfs_resop4 *resop, struct svc_req *req, 3225 struct compound_state *cs) 3226 { 3227 READ4args *args = &argop->nfs_argop4_u.opread; 3228 READ4res *resp = &resop->nfs_resop4_u.opread; 3229 int error; 3230 int verror; 3231 vnode_t *vp; 3232 struct vattr va; 3233 struct iovec iov, *iovp = NULL; 3234 int iovcnt; 3235 struct uio uio; 3236 u_offset_t offset; 3237 bool_t *deleg = &cs->deleg; 3238 nfsstat4 stat; 3239 int in_crit = 0; 3240 mblk_t *mp = NULL; 3241 int alloc_err = 0; 3242 int rdma_used = 0; 3243 int loaned_buffers; 3244 caller_context_t ct; 3245 struct uio *uiop; 3246 3247 DTRACE_NFSV4_2(op__read__start, struct compound_state *, cs, 3248 READ4args, args); 3249 3250 vp = cs->vp; 3251 if (vp == NULL) { 3252 *cs->statusp = resp->status = NFS4ERR_NOFILEHANDLE; 3253 goto out; 3254 } 3255 if (cs->access == CS_ACCESS_DENIED) { 3256 *cs->statusp = resp->status = NFS4ERR_ACCESS; 3257 goto out; 3258 } 3259 3260 if ((stat = rfs4_check_stateid(FREAD, vp, &args->stateid, FALSE, 3261 deleg, TRUE, &ct)) != NFS4_OK) { 3262 *cs->statusp = resp->status = stat; 3263 goto out; 3264 } 3265 3266 /* 3267 * Enter the critical region before calling VOP_RWLOCK 3268 * to avoid a deadlock with write requests. 3269 */ 3270 if (nbl_need_check(vp)) { 3271 nbl_start_crit(vp, RW_READER); 3272 in_crit = 1; 3273 if (nbl_conflict(vp, NBL_READ, args->offset, args->count, 0, 3274 &ct)) { 3275 *cs->statusp = resp->status = NFS4ERR_LOCKED; 3276 goto out; 3277 } 3278 } 3279 3280 if (args->wlist) { 3281 if (args->count > clist_len(args->wlist)) { 3282 *cs->statusp = resp->status = NFS4ERR_INVAL; 3283 goto out; 3284 } 3285 rdma_used = 1; 3286 } 3287 3288 /* use loaned buffers for TCP */ 3289 loaned_buffers = (nfs_loaned_buffers && !rdma_used) ? 1 : 0; 3290 3291 va.va_mask = AT_MODE|AT_SIZE|AT_UID; 3292 verror = VOP_GETATTR(vp, &va, 0, cs->cr, &ct); 3293 3294 /* 3295 * If we can't get the attributes, then we can't do the 3296 * right access checking. So, we'll fail the request. 3297 */ 3298 if (verror) { 3299 *cs->statusp = resp->status = puterrno4(verror); 3300 goto out; 3301 } 3302 3303 if (vp->v_type != VREG) { 3304 *cs->statusp = resp->status = 3305 ((vp->v_type == VDIR) ? NFS4ERR_ISDIR : NFS4ERR_INVAL); 3306 goto out; 3307 } 3308 3309 if (crgetuid(cs->cr) != va.va_uid && 3310 (error = VOP_ACCESS(vp, VREAD, 0, cs->cr, &ct)) && 3311 (error = VOP_ACCESS(vp, VEXEC, 0, cs->cr, &ct))) { 3312 *cs->statusp = resp->status = puterrno4(error); 3313 goto out; 3314 } 3315 3316 if (MANDLOCK(vp, va.va_mode)) { /* XXX - V4 supports mand locking */ 3317 *cs->statusp = resp->status = NFS4ERR_ACCESS; 3318 goto out; 3319 } 3320 3321 offset = args->offset; 3322 if (offset >= va.va_size) { 3323 *cs->statusp = resp->status = NFS4_OK; 3324 resp->eof = TRUE; 3325 resp->data_len = 0; 3326 resp->data_val = NULL; 3327 resp->mblk = NULL; 3328 /* RDMA */ 3329 resp->wlist = args->wlist; 3330 resp->wlist_len = resp->data_len; 3331 *cs->statusp = resp->status = NFS4_OK; 3332 if (resp->wlist) 3333 clist_zero_len(resp->wlist); 3334 goto out; 3335 } 3336 3337 if (args->count == 0) { 3338 *cs->statusp = resp->status = NFS4_OK; 3339 resp->eof = FALSE; 3340 resp->data_len = 0; 3341 resp->data_val = NULL; 3342 resp->mblk = NULL; 3343 /* RDMA */ 3344 resp->wlist = args->wlist; 3345 resp->wlist_len = resp->data_len; 3346 if (resp->wlist) 3347 clist_zero_len(resp->wlist); 3348 goto out; 3349 } 3350 3351 /* 3352 * Do not allocate memory more than maximum allowed 3353 * transfer size 3354 */ 3355 if (args->count > rfs4_tsize(req)) 3356 args->count = rfs4_tsize(req); 3357 3358 if (loaned_buffers) { 3359 uiop = (uio_t *)rfs_setup_xuio(vp); 3360 ASSERT(uiop != NULL); 3361 uiop->uio_segflg = UIO_SYSSPACE; 3362 uiop->uio_loffset = args->offset; 3363 uiop->uio_resid = args->count; 3364 3365 /* Jump to do the read if successful */ 3366 if (!VOP_REQZCBUF(vp, UIO_READ, (xuio_t *)uiop, cs->cr, &ct)) { 3367 /* 3368 * Need to hold the vnode until after VOP_RETZCBUF() 3369 * is called. 3370 */ 3371 VN_HOLD(vp); 3372 goto doio_read; 3373 } 3374 3375 DTRACE_PROBE2(nfss__i__reqzcbuf_failed, int, 3376 uiop->uio_loffset, int, uiop->uio_resid); 3377 3378 uiop->uio_extflg = 0; 3379 3380 /* failure to setup for zero copy */ 3381 rfs_free_xuio((void *)uiop); 3382 loaned_buffers = 0; 3383 } 3384 3385 /* 3386 * If returning data via RDMA Write, then grab the chunk list. If we 3387 * aren't returning READ data w/RDMA_WRITE, then grab a mblk. 3388 */ 3389 if (rdma_used) { 3390 mp = NULL; 3391 (void) rdma_get_wchunk(req, &iov, args->wlist); 3392 uio.uio_iov = &iov; 3393 uio.uio_iovcnt = 1; 3394 } else { 3395 /* 3396 * mp will contain the data to be sent out in the read reply. 3397 * It will be freed after the reply has been sent. 3398 */ 3399 mp = rfs_read_alloc(args->count, &iovp, &iovcnt); 3400 ASSERT(mp != NULL); 3401 ASSERT(alloc_err == 0); 3402 uio.uio_iov = iovp; 3403 uio.uio_iovcnt = iovcnt; 3404 } 3405 3406 uio.uio_segflg = UIO_SYSSPACE; 3407 uio.uio_extflg = UIO_COPY_CACHED; 3408 uio.uio_loffset = args->offset; 3409 uio.uio_resid = args->count; 3410 uiop = &uio; 3411 3412 doio_read: 3413 error = do_io(FREAD, vp, uiop, 0, cs->cr, &ct); 3414 3415 va.va_mask = AT_SIZE; 3416 verror = VOP_GETATTR(vp, &va, 0, cs->cr, &ct); 3417 3418 if (error) { 3419 if (mp) 3420 freemsg(mp); 3421 *cs->statusp = resp->status = puterrno4(error); 3422 goto out; 3423 } 3424 3425 /* make mblk using zc buffers */ 3426 if (loaned_buffers) { 3427 mp = uio_to_mblk(uiop); 3428 ASSERT(mp != NULL); 3429 } 3430 3431 *cs->statusp = resp->status = NFS4_OK; 3432 3433 ASSERT(uiop->uio_resid >= 0); 3434 resp->data_len = args->count - uiop->uio_resid; 3435 if (mp) { 3436 resp->data_val = (char *)mp->b_datap->db_base; 3437 rfs_rndup_mblks(mp, resp->data_len, loaned_buffers); 3438 } else { 3439 resp->data_val = (caddr_t)iov.iov_base; 3440 } 3441 3442 resp->mblk = mp; 3443 3444 if (!verror && offset + resp->data_len == va.va_size) 3445 resp->eof = TRUE; 3446 else 3447 resp->eof = FALSE; 3448 3449 if (rdma_used) { 3450 if (!rdma_setup_read_data4(args, resp)) { 3451 *cs->statusp = resp->status = NFS4ERR_INVAL; 3452 } 3453 } else { 3454 resp->wlist = NULL; 3455 } 3456 3457 out: 3458 if (in_crit) 3459 nbl_end_crit(vp); 3460 3461 if (iovp != NULL) 3462 kmem_free(iovp, iovcnt * sizeof (struct iovec)); 3463 3464 DTRACE_NFSV4_2(op__read__done, struct compound_state *, cs, 3465 READ4res *, resp); 3466 } 3467 3468 static void 3469 rfs4_op_read_free(nfs_resop4 *resop) 3470 { 3471 READ4res *resp = &resop->nfs_resop4_u.opread; 3472 3473 if (resp->status == NFS4_OK && resp->mblk != NULL) { 3474 freemsg(resp->mblk); 3475 resp->mblk = NULL; 3476 resp->data_val = NULL; 3477 resp->data_len = 0; 3478 } 3479 } 3480 3481 static void 3482 rfs4_op_readdir_free(nfs_resop4 * resop) 3483 { 3484 READDIR4res *resp = &resop->nfs_resop4_u.opreaddir; 3485 3486 if (resp->status == NFS4_OK && resp->mblk != NULL) { 3487 freeb(resp->mblk); 3488 resp->mblk = NULL; 3489 resp->data_len = 0; 3490 } 3491 } 3492 3493 3494 /* ARGSUSED */ 3495 static void 3496 rfs4_op_putpubfh(nfs_argop4 *args, nfs_resop4 *resop, struct svc_req *req, 3497 struct compound_state *cs) 3498 { 3499 PUTPUBFH4res *resp = &resop->nfs_resop4_u.opputpubfh; 3500 int error; 3501 vnode_t *vp; 3502 struct exportinfo *exi, *sav_exi; 3503 nfs_fh4_fmt_t *fh_fmtp; 3504 nfs_export_t *ne = nfs_get_export(); 3505 3506 DTRACE_NFSV4_1(op__putpubfh__start, struct compound_state *, cs); 3507 3508 if (cs->vp) { 3509 VN_RELE(cs->vp); 3510 cs->vp = NULL; 3511 } 3512 3513 if (cs->cr) 3514 crfree(cs->cr); 3515 3516 cs->cr = crdup(cs->basecr); 3517 3518 vp = ne->exi_public->exi_vp; 3519 if (vp == NULL) { 3520 *cs->statusp = resp->status = NFS4ERR_SERVERFAULT; 3521 goto out; 3522 } 3523 3524 error = makefh4(&cs->fh, vp, ne->exi_public); 3525 if (error != 0) { 3526 *cs->statusp = resp->status = puterrno4(error); 3527 goto out; 3528 } 3529 sav_exi = cs->exi; 3530 if (ne->exi_public == ne->exi_root) { 3531 /* 3532 * No filesystem is actually shared public, so we default 3533 * to exi_root. In this case, we must check whether root 3534 * is exported. 3535 */ 3536 fh_fmtp = (nfs_fh4_fmt_t *)cs->fh.nfs_fh4_val; 3537 3538 /* 3539 * if root filesystem is exported, the exportinfo struct that we 3540 * should use is what checkexport4 returns, because root_exi is 3541 * actually a mostly empty struct. 3542 */ 3543 exi = checkexport4(&fh_fmtp->fh4_fsid, 3544 (fid_t *)&fh_fmtp->fh4_xlen, NULL); 3545 cs->exi = ((exi != NULL) ? exi : ne->exi_public); 3546 } else { 3547 /* 3548 * it's a properly shared filesystem 3549 */ 3550 cs->exi = ne->exi_public; 3551 } 3552 3553 if (is_system_labeled()) { 3554 bslabel_t *clabel; 3555 3556 ASSERT(req->rq_label != NULL); 3557 clabel = req->rq_label; 3558 DTRACE_PROBE2(tx__rfs4__log__info__opputpubfh__clabel, char *, 3559 "got client label from request(1)", 3560 struct svc_req *, req); 3561 if (!blequal(&l_admin_low->tsl_label, clabel)) { 3562 if (!do_rfs_label_check(clabel, vp, DOMINANCE_CHECK, 3563 cs->exi)) { 3564 *cs->statusp = resp->status = 3565 NFS4ERR_SERVERFAULT; 3566 goto out; 3567 } 3568 } 3569 } 3570 3571 VN_HOLD(vp); 3572 cs->vp = vp; 3573 3574 if ((resp->status = call_checkauth4(cs, req)) != NFS4_OK) { 3575 VN_RELE(cs->vp); 3576 cs->vp = NULL; 3577 cs->exi = sav_exi; 3578 goto out; 3579 } 3580 3581 *cs->statusp = resp->status = NFS4_OK; 3582 out: 3583 DTRACE_NFSV4_2(op__putpubfh__done, struct compound_state *, cs, 3584 PUTPUBFH4res *, resp); 3585 } 3586 3587 /* 3588 * XXX - issue with put*fh operations. Suppose /export/home is exported. 3589 * Suppose an NFS client goes to mount /export/home/joe. If /export, home, 3590 * or joe have restrictive search permissions, then we shouldn't let 3591 * the client get a file handle. This is easy to enforce. However, we 3592 * don't know what security flavor should be used until we resolve the 3593 * path name. Another complication is uid mapping. If root is 3594 * the user, then it will be mapped to the anonymous user by default, 3595 * but we won't know that till we've resolved the path name. And we won't 3596 * know what the anonymous user is. 3597 * Luckily, SECINFO is specified to take a full filename. 3598 * So what we will have to in rfs4_op_lookup is check that flavor of 3599 * the target object matches that of the request, and if root was the 3600 * caller, check for the root= and anon= options, and if necessary, 3601 * repeat the lookup using the right cred_t. But that's not done yet. 3602 */ 3603 /* ARGSUSED */ 3604 static void 3605 rfs4_op_putfh(nfs_argop4 *argop, nfs_resop4 *resop, struct svc_req *req, 3606 struct compound_state *cs) 3607 { 3608 PUTFH4args *args = &argop->nfs_argop4_u.opputfh; 3609 PUTFH4res *resp = &resop->nfs_resop4_u.opputfh; 3610 nfs_fh4_fmt_t *fh_fmtp; 3611 3612 DTRACE_NFSV4_2(op__putfh__start, struct compound_state *, cs, 3613 PUTFH4args *, args); 3614 3615 if (cs->vp) { 3616 VN_RELE(cs->vp); 3617 cs->vp = NULL; 3618 } 3619 3620 if (cs->cr) { 3621 crfree(cs->cr); 3622 cs->cr = NULL; 3623 } 3624 3625 3626 if (args->object.nfs_fh4_len < NFS_FH4_LEN) { 3627 *cs->statusp = resp->status = NFS4ERR_BADHANDLE; 3628 goto out; 3629 } 3630 3631 fh_fmtp = (nfs_fh4_fmt_t *)args->object.nfs_fh4_val; 3632 cs->exi = checkexport4(&fh_fmtp->fh4_fsid, (fid_t *)&fh_fmtp->fh4_xlen, 3633 NULL); 3634 3635 if (cs->exi == NULL) { 3636 *cs->statusp = resp->status = NFS4ERR_STALE; 3637 goto out; 3638 } 3639 3640 cs->cr = crdup(cs->basecr); 3641 3642 ASSERT(cs->cr != NULL); 3643 3644 if (! (cs->vp = nfs4_fhtovp(&args->object, cs->exi, &resp->status))) { 3645 *cs->statusp = resp->status; 3646 goto out; 3647 } 3648 3649 if ((resp->status = call_checkauth4(cs, req)) != NFS4_OK) { 3650 VN_RELE(cs->vp); 3651 cs->vp = NULL; 3652 goto out; 3653 } 3654 3655 nfs_fh4_copy(&args->object, &cs->fh); 3656 *cs->statusp = resp->status = NFS4_OK; 3657 cs->deleg = FALSE; 3658 3659 out: 3660 DTRACE_NFSV4_2(op__putfh__done, struct compound_state *, cs, 3661 PUTFH4res *, resp); 3662 } 3663 3664 /* ARGSUSED */ 3665 static void 3666 rfs4_op_putrootfh(nfs_argop4 *argop, nfs_resop4 *resop, struct svc_req *req, 3667 struct compound_state *cs) 3668 { 3669 PUTROOTFH4res *resp = &resop->nfs_resop4_u.opputrootfh; 3670 int error; 3671 fid_t fid; 3672 struct exportinfo *exi, *sav_exi; 3673 3674 DTRACE_NFSV4_1(op__putrootfh__start, struct compound_state *, cs); 3675 3676 if (cs->vp) { 3677 VN_RELE(cs->vp); 3678 cs->vp = NULL; 3679 } 3680 3681 if (cs->cr) 3682 crfree(cs->cr); 3683 3684 cs->cr = crdup(cs->basecr); 3685 3686 /* 3687 * Using rootdir, the system root vnode, 3688 * get its fid. 3689 */ 3690 bzero(&fid, sizeof (fid)); 3691 fid.fid_len = MAXFIDSZ; 3692 error = vop_fid_pseudo(ZONE_ROOTVP(), &fid); 3693 if (error != 0) { 3694 *cs->statusp = resp->status = puterrno4(error); 3695 goto out; 3696 } 3697 3698 /* 3699 * Then use the root fsid & fid it to find out if it's exported 3700 * 3701 * If the server root isn't exported directly, then 3702 * it should at least be a pseudo export based on 3703 * one or more exports further down in the server's 3704 * file tree. 3705 */ 3706 exi = checkexport4(&ZONE_ROOTVP()->v_vfsp->vfs_fsid, &fid, NULL); 3707 if (exi == NULL || exi->exi_export.ex_flags & EX_PUBLIC) { 3708 NFS4_DEBUG(rfs4_debug, 3709 (CE_WARN, "rfs4_op_putrootfh: export check failure")); 3710 *cs->statusp = resp->status = NFS4ERR_SERVERFAULT; 3711 goto out; 3712 } 3713 3714 /* 3715 * Now make a filehandle based on the root 3716 * export and root vnode. 3717 */ 3718 error = makefh4(&cs->fh, ZONE_ROOTVP(), exi); 3719 if (error != 0) { 3720 *cs->statusp = resp->status = puterrno4(error); 3721 goto out; 3722 } 3723 3724 sav_exi = cs->exi; 3725 cs->exi = exi; 3726 3727 VN_HOLD(ZONE_ROOTVP()); 3728 cs->vp = ZONE_ROOTVP(); 3729 3730 if ((resp->status = call_checkauth4(cs, req)) != NFS4_OK) { 3731 VN_RELE(cs->vp); 3732 cs->vp = NULL; 3733 cs->exi = sav_exi; 3734 goto out; 3735 } 3736 3737 *cs->statusp = resp->status = NFS4_OK; 3738 cs->deleg = FALSE; 3739 out: 3740 DTRACE_NFSV4_2(op__putrootfh__done, struct compound_state *, cs, 3741 PUTROOTFH4res *, resp); 3742 } 3743 3744 /* 3745 * readlink: args: CURRENT_FH. 3746 * res: status. If success - CURRENT_FH unchanged, return linktext. 3747 */ 3748 3749 /* ARGSUSED */ 3750 static void 3751 rfs4_op_readlink(nfs_argop4 *argop, nfs_resop4 *resop, struct svc_req *req, 3752 struct compound_state *cs) 3753 { 3754 READLINK4res *resp = &resop->nfs_resop4_u.opreadlink; 3755 int error; 3756 vnode_t *vp; 3757 struct iovec iov; 3758 struct vattr va; 3759 struct uio uio; 3760 char *data; 3761 struct sockaddr *ca; 3762 char *name = NULL; 3763 int is_referral; 3764 3765 DTRACE_NFSV4_1(op__readlink__start, struct compound_state *, cs); 3766 3767 /* CURRENT_FH: directory */ 3768 vp = cs->vp; 3769 if (vp == NULL) { 3770 *cs->statusp = resp->status = NFS4ERR_NOFILEHANDLE; 3771 goto out; 3772 } 3773 3774 if (cs->access == CS_ACCESS_DENIED) { 3775 *cs->statusp = resp->status = NFS4ERR_ACCESS; 3776 goto out; 3777 } 3778 3779 /* Is it a referral? */ 3780 if (vn_is_nfs_reparse(vp, cs->cr) && client_is_downrev(req)) { 3781 3782 is_referral = 1; 3783 3784 } else { 3785 3786 is_referral = 0; 3787 3788 if (vp->v_type == VDIR) { 3789 *cs->statusp = resp->status = NFS4ERR_ISDIR; 3790 goto out; 3791 } 3792 3793 if (vp->v_type != VLNK) { 3794 *cs->statusp = resp->status = NFS4ERR_INVAL; 3795 goto out; 3796 } 3797 3798 } 3799 3800 va.va_mask = AT_MODE; 3801 error = VOP_GETATTR(vp, &va, 0, cs->cr, NULL); 3802 if (error) { 3803 *cs->statusp = resp->status = puterrno4(error); 3804 goto out; 3805 } 3806 3807 if (MANDLOCK(vp, va.va_mode)) { 3808 *cs->statusp = resp->status = NFS4ERR_ACCESS; 3809 goto out; 3810 } 3811 3812 data = kmem_alloc(MAXPATHLEN + 1, KM_SLEEP); 3813 3814 if (is_referral) { 3815 char *s; 3816 size_t strsz; 3817 kstat_named_t *stat = 3818 cs->exi->exi_ne->ne_globals->svstat[NFS_V4]; 3819 3820 /* Get an artificial symlink based on a referral */ 3821 s = build_symlink(vp, cs->cr, &strsz); 3822 stat[NFS_REFERLINKS].value.ui64++; 3823 DTRACE_PROBE2(nfs4serv__func__referral__reflink, 3824 vnode_t *, vp, char *, s); 3825 if (s == NULL) 3826 error = EINVAL; 3827 else { 3828 error = 0; 3829 (void) strlcpy(data, s, MAXPATHLEN + 1); 3830 kmem_free(s, strsz); 3831 } 3832 3833 } else { 3834 3835 iov.iov_base = data; 3836 iov.iov_len = MAXPATHLEN; 3837 uio.uio_iov = &iov; 3838 uio.uio_iovcnt = 1; 3839 uio.uio_segflg = UIO_SYSSPACE; 3840 uio.uio_extflg = UIO_COPY_CACHED; 3841 uio.uio_loffset = 0; 3842 uio.uio_resid = MAXPATHLEN; 3843 3844 error = VOP_READLINK(vp, &uio, cs->cr, NULL); 3845 3846 if (!error) 3847 *(data + MAXPATHLEN - uio.uio_resid) = '\0'; 3848 } 3849 3850 if (error) { 3851 kmem_free((caddr_t)data, (uint_t)MAXPATHLEN + 1); 3852 *cs->statusp = resp->status = puterrno4(error); 3853 goto out; 3854 } 3855 3856 ca = (struct sockaddr *)svc_getrpccaller(req->rq_xprt)->buf; 3857 name = nfscmd_convname(ca, cs->exi, data, NFSCMD_CONV_OUTBOUND, 3858 MAXPATHLEN + 1); 3859 3860 if (name == NULL) { 3861 /* 3862 * Even though the conversion failed, we return 3863 * something. We just don't translate it. 3864 */ 3865 name = data; 3866 } 3867 3868 /* 3869 * treat link name as data 3870 */ 3871 (void) str_to_utf8(name, (utf8string *)&resp->link); 3872 3873 if (name != data) 3874 kmem_free(name, MAXPATHLEN + 1); 3875 kmem_free((caddr_t)data, (uint_t)MAXPATHLEN + 1); 3876 *cs->statusp = resp->status = NFS4_OK; 3877 3878 out: 3879 DTRACE_NFSV4_2(op__readlink__done, struct compound_state *, cs, 3880 READLINK4res *, resp); 3881 } 3882 3883 static void 3884 rfs4_op_readlink_free(nfs_resop4 *resop) 3885 { 3886 READLINK4res *resp = &resop->nfs_resop4_u.opreadlink; 3887 utf8string *symlink = (utf8string *)&resp->link; 3888 3889 if (symlink->utf8string_val) { 3890 UTF8STRING_FREE(*symlink) 3891 } 3892 } 3893 3894 /* 3895 * release_lockowner: 3896 * Release any state associated with the supplied 3897 * lockowner. Note if any lo_state is holding locks we will not 3898 * rele that lo_state and thus the lockowner will not be destroyed. 3899 * A client using lock after the lock owner stateid has been released 3900 * will suffer the consequence of NFS4ERR_BAD_STATEID and would have 3901 * to reissue the lock with new_lock_owner set to TRUE. 3902 * args: lock_owner 3903 * res: status 3904 */ 3905 /* ARGSUSED */ 3906 static void 3907 rfs4_op_release_lockowner(nfs_argop4 *argop, nfs_resop4 *resop, 3908 struct svc_req *req, struct compound_state *cs) 3909 { 3910 RELEASE_LOCKOWNER4args *ap = &argop->nfs_argop4_u.oprelease_lockowner; 3911 RELEASE_LOCKOWNER4res *resp = &resop->nfs_resop4_u.oprelease_lockowner; 3912 rfs4_lockowner_t *lo; 3913 rfs4_openowner_t *oo; 3914 rfs4_state_t *sp; 3915 rfs4_lo_state_t *lsp; 3916 rfs4_client_t *cp; 3917 bool_t create = FALSE; 3918 locklist_t *llist; 3919 sysid_t sysid; 3920 3921 DTRACE_NFSV4_2(op__release__lockowner__start, struct compound_state *, 3922 cs, RELEASE_LOCKOWNER4args *, ap); 3923 3924 /* Make sure there is a clientid around for this request */ 3925 cp = rfs4_findclient_by_id(ap->lock_owner.clientid, FALSE); 3926 3927 if (cp == NULL) { 3928 *cs->statusp = resp->status = 3929 rfs4_check_clientid(&ap->lock_owner.clientid, 0); 3930 goto out; 3931 } 3932 rfs4_client_rele(cp); 3933 3934 lo = rfs4_findlockowner(&ap->lock_owner, &create); 3935 if (lo == NULL) { 3936 *cs->statusp = resp->status = NFS4_OK; 3937 goto out; 3938 } 3939 ASSERT(lo->rl_client != NULL); 3940 3941 /* 3942 * Check for EXPIRED client. If so will reap state with in a lease 3943 * period or on next set_clientid_confirm step 3944 */ 3945 if (rfs4_lease_expired(lo->rl_client)) { 3946 rfs4_lockowner_rele(lo); 3947 *cs->statusp = resp->status = NFS4ERR_EXPIRED; 3948 goto out; 3949 } 3950 3951 /* 3952 * If no sysid has been assigned, then no locks exist; just return. 3953 */ 3954 rfs4_dbe_lock(lo->rl_client->rc_dbe); 3955 if (lo->rl_client->rc_sysidt == LM_NOSYSID) { 3956 rfs4_lockowner_rele(lo); 3957 rfs4_dbe_unlock(lo->rl_client->rc_dbe); 3958 goto out; 3959 } 3960 3961 sysid = lo->rl_client->rc_sysidt; 3962 rfs4_dbe_unlock(lo->rl_client->rc_dbe); 3963 3964 /* 3965 * Mark the lockowner invalid. 3966 */ 3967 rfs4_dbe_hide(lo->rl_dbe); 3968 3969 /* 3970 * sysid-pid pair should now not be used since the lockowner is 3971 * invalid. If the client were to instantiate the lockowner again 3972 * it would be assigned a new pid. Thus we can get the list of 3973 * current locks. 3974 */ 3975 3976 llist = flk_get_active_locks(sysid, lo->rl_pid); 3977 /* If we are still holding locks fail */ 3978 if (llist != NULL) { 3979 3980 *cs->statusp = resp->status = NFS4ERR_LOCKS_HELD; 3981 3982 flk_free_locklist(llist); 3983 /* 3984 * We need to unhide the lockowner so the client can 3985 * try it again. The bad thing here is if the client 3986 * has a logic error that took it here in the first place 3987 * they probably have lost accounting of the locks that it 3988 * is holding. So we may have dangling state until the 3989 * open owner state is reaped via close. One scenario 3990 * that could possibly occur is that the client has 3991 * sent the unlock request(s) in separate threads 3992 * and has not waited for the replies before sending the 3993 * RELEASE_LOCKOWNER request. Presumably, it would expect 3994 * and deal appropriately with NFS4ERR_LOCKS_HELD, by 3995 * reissuing the request. 3996 */ 3997 rfs4_dbe_unhide(lo->rl_dbe); 3998 rfs4_lockowner_rele(lo); 3999 goto out; 4000 } 4001 4002 /* 4003 * For the corresponding client we need to check each open 4004 * owner for any opens that have lockowner state associated 4005 * with this lockowner. 4006 */ 4007 4008 rfs4_dbe_lock(lo->rl_client->rc_dbe); 4009 for (oo = list_head(&lo->rl_client->rc_openownerlist); oo != NULL; 4010 oo = list_next(&lo->rl_client->rc_openownerlist, oo)) { 4011 4012 rfs4_dbe_lock(oo->ro_dbe); 4013 for (sp = list_head(&oo->ro_statelist); sp != NULL; 4014 sp = list_next(&oo->ro_statelist, sp)) { 4015 4016 rfs4_dbe_lock(sp->rs_dbe); 4017 for (lsp = list_head(&sp->rs_lostatelist); 4018 lsp != NULL; 4019 lsp = list_next(&sp->rs_lostatelist, lsp)) { 4020 if (lsp->rls_locker == lo) { 4021 rfs4_dbe_lock(lsp->rls_dbe); 4022 rfs4_dbe_invalidate(lsp->rls_dbe); 4023 rfs4_dbe_unlock(lsp->rls_dbe); 4024 } 4025 } 4026 rfs4_dbe_unlock(sp->rs_dbe); 4027 } 4028 rfs4_dbe_unlock(oo->ro_dbe); 4029 } 4030 rfs4_dbe_unlock(lo->rl_client->rc_dbe); 4031 4032 rfs4_lockowner_rele(lo); 4033 4034 *cs->statusp = resp->status = NFS4_OK; 4035 4036 out: 4037 DTRACE_NFSV4_2(op__release__lockowner__done, struct compound_state *, 4038 cs, RELEASE_LOCKOWNER4res *, resp); 4039 } 4040 4041 /* 4042 * short utility function to lookup a file and recall the delegation 4043 */ 4044 static rfs4_file_t * 4045 rfs4_lookup_and_findfile(vnode_t *dvp, char *nm, vnode_t **vpp, 4046 int *lkup_error, cred_t *cr) 4047 { 4048 vnode_t *vp; 4049 rfs4_file_t *fp = NULL; 4050 bool_t fcreate = FALSE; 4051 int error; 4052 4053 if (vpp) 4054 *vpp = NULL; 4055 4056 if ((error = VOP_LOOKUP(dvp, nm, &vp, NULL, 0, NULL, cr, NULL, NULL, 4057 NULL)) == 0) { 4058 if (vp->v_type == VREG) 4059 fp = rfs4_findfile(vp, NULL, &fcreate); 4060 if (vpp) 4061 *vpp = vp; 4062 else 4063 VN_RELE(vp); 4064 } 4065 4066 if (lkup_error) 4067 *lkup_error = error; 4068 4069 return (fp); 4070 } 4071 4072 /* 4073 * remove: args: CURRENT_FH: directory; name. 4074 * res: status. If success - CURRENT_FH unchanged, return change_info 4075 * for directory. 4076 */ 4077 /* ARGSUSED */ 4078 static void 4079 rfs4_op_remove(nfs_argop4 *argop, nfs_resop4 *resop, struct svc_req *req, 4080 struct compound_state *cs) 4081 { 4082 REMOVE4args *args = &argop->nfs_argop4_u.opremove; 4083 REMOVE4res *resp = &resop->nfs_resop4_u.opremove; 4084 int error; 4085 vnode_t *dvp, *vp; 4086 struct vattr bdva, idva, adva; 4087 char *nm; 4088 uint_t len; 4089 rfs4_file_t *fp; 4090 int in_crit = 0; 4091 bslabel_t *clabel; 4092 struct sockaddr *ca; 4093 char *name = NULL; 4094 nfsstat4 status; 4095 4096 DTRACE_NFSV4_2(op__remove__start, struct compound_state *, cs, 4097 REMOVE4args *, args); 4098 4099 /* CURRENT_FH: directory */ 4100 dvp = cs->vp; 4101 if (dvp == NULL) { 4102 *cs->statusp = resp->status = NFS4ERR_NOFILEHANDLE; 4103 goto out; 4104 } 4105 4106 if (cs->access == CS_ACCESS_DENIED) { 4107 *cs->statusp = resp->status = NFS4ERR_ACCESS; 4108 goto out; 4109 } 4110 4111 /* 4112 * If there is an unshared filesystem mounted on this vnode, 4113 * Do not allow to remove anything in this directory. 4114 */ 4115 if (vn_ismntpt(dvp)) { 4116 *cs->statusp = resp->status = NFS4ERR_ACCESS; 4117 goto out; 4118 } 4119 4120 if (dvp->v_type != VDIR) { 4121 *cs->statusp = resp->status = NFS4ERR_NOTDIR; 4122 goto out; 4123 } 4124 4125 status = utf8_dir_verify(&args->target); 4126 if (status != NFS4_OK) { 4127 *cs->statusp = resp->status = status; 4128 goto out; 4129 } 4130 4131 /* 4132 * Lookup the file so that we can check if it's a directory 4133 */ 4134 nm = utf8_to_fn(&args->target, &len, NULL); 4135 if (nm == NULL) { 4136 *cs->statusp = resp->status = NFS4ERR_INVAL; 4137 goto out; 4138 } 4139 4140 if (len > MAXNAMELEN) { 4141 *cs->statusp = resp->status = NFS4ERR_NAMETOOLONG; 4142 kmem_free(nm, len); 4143 goto out; 4144 } 4145 4146 if (rdonly4(req, cs)) { 4147 *cs->statusp = resp->status = NFS4ERR_ROFS; 4148 kmem_free(nm, len); 4149 goto out; 4150 } 4151 4152 ca = (struct sockaddr *)svc_getrpccaller(req->rq_xprt)->buf; 4153 name = nfscmd_convname(ca, cs->exi, nm, NFSCMD_CONV_INBOUND, 4154 MAXPATHLEN + 1); 4155 4156 if (name == NULL) { 4157 *cs->statusp = resp->status = NFS4ERR_INVAL; 4158 kmem_free(nm, len); 4159 goto out; 4160 } 4161 4162 /* 4163 * Lookup the file to determine type and while we are see if 4164 * there is a file struct around and check for delegation. 4165 * We don't need to acquire va_seq before this lookup, if 4166 * it causes an update, cinfo.before will not match, which will 4167 * trigger a cache flush even if atomic is TRUE. 4168 */ 4169 if (fp = rfs4_lookup_and_findfile(dvp, name, &vp, &error, cs->cr)) { 4170 if (rfs4_check_delegated_byfp(FWRITE, fp, TRUE, TRUE, TRUE, 4171 NULL)) { 4172 VN_RELE(vp); 4173 rfs4_file_rele(fp); 4174 *cs->statusp = resp->status = NFS4ERR_DELAY; 4175 if (nm != name) 4176 kmem_free(name, MAXPATHLEN + 1); 4177 kmem_free(nm, len); 4178 goto out; 4179 } 4180 } 4181 4182 /* Didn't find anything to remove */ 4183 if (vp == NULL) { 4184 *cs->statusp = resp->status = error; 4185 if (nm != name) 4186 kmem_free(name, MAXPATHLEN + 1); 4187 kmem_free(nm, len); 4188 goto out; 4189 } 4190 4191 if (nbl_need_check(vp)) { 4192 nbl_start_crit(vp, RW_READER); 4193 in_crit = 1; 4194 if (nbl_conflict(vp, NBL_REMOVE, 0, 0, 0, NULL)) { 4195 *cs->statusp = resp->status = NFS4ERR_FILE_OPEN; 4196 if (nm != name) 4197 kmem_free(name, MAXPATHLEN + 1); 4198 kmem_free(nm, len); 4199 nbl_end_crit(vp); 4200 VN_RELE(vp); 4201 if (fp) { 4202 rfs4_clear_dont_grant(fp); 4203 rfs4_file_rele(fp); 4204 } 4205 goto out; 4206 } 4207 } 4208 4209 /* check label before allowing removal */ 4210 if (is_system_labeled()) { 4211 ASSERT(req->rq_label != NULL); 4212 clabel = req->rq_label; 4213 DTRACE_PROBE2(tx__rfs4__log__info__opremove__clabel, char *, 4214 "got client label from request(1)", 4215 struct svc_req *, req); 4216 if (!blequal(&l_admin_low->tsl_label, clabel)) { 4217 if (!do_rfs_label_check(clabel, vp, EQUALITY_CHECK, 4218 cs->exi)) { 4219 *cs->statusp = resp->status = NFS4ERR_ACCESS; 4220 if (name != nm) 4221 kmem_free(name, MAXPATHLEN + 1); 4222 kmem_free(nm, len); 4223 if (in_crit) 4224 nbl_end_crit(vp); 4225 VN_RELE(vp); 4226 if (fp) { 4227 rfs4_clear_dont_grant(fp); 4228 rfs4_file_rele(fp); 4229 } 4230 goto out; 4231 } 4232 } 4233 } 4234 4235 /* Get dir "before" change value */ 4236 bdva.va_mask = AT_CTIME|AT_SEQ; 4237 error = VOP_GETATTR(dvp, &bdva, 0, cs->cr, NULL); 4238 if (error) { 4239 *cs->statusp = resp->status = puterrno4(error); 4240 if (nm != name) 4241 kmem_free(name, MAXPATHLEN + 1); 4242 kmem_free(nm, len); 4243 if (in_crit) 4244 nbl_end_crit(vp); 4245 VN_RELE(vp); 4246 if (fp) { 4247 rfs4_clear_dont_grant(fp); 4248 rfs4_file_rele(fp); 4249 } 4250 goto out; 4251 } 4252 NFS4_SET_FATTR4_CHANGE(resp->cinfo.before, bdva.va_ctime) 4253 4254 /* Actually do the REMOVE operation */ 4255 if (vp->v_type == VDIR) { 4256 /* 4257 * Can't remove a directory that has a mounted-on filesystem. 4258 */ 4259 if (vn_ismntpt(vp)) { 4260 error = EACCES; 4261 } else { 4262 /* 4263 * System V defines rmdir to return EEXIST, 4264 * not ENOTEMPTY, if the directory is not 4265 * empty. A System V NFS server needs to map 4266 * NFS4ERR_EXIST to NFS4ERR_NOTEMPTY to 4267 * transmit over the wire. 4268 */ 4269 if ((error = VOP_RMDIR(dvp, name, ZONE_ROOTVP(), cs->cr, 4270 NULL, 0)) == EEXIST) 4271 error = ENOTEMPTY; 4272 } 4273 } else { 4274 if ((error = VOP_REMOVE(dvp, name, cs->cr, NULL, 0)) == 0 && 4275 fp != NULL) { 4276 struct vattr va; 4277 vnode_t *tvp; 4278 4279 rfs4_dbe_lock(fp->rf_dbe); 4280 tvp = fp->rf_vp; 4281 if (tvp) 4282 VN_HOLD(tvp); 4283 rfs4_dbe_unlock(fp->rf_dbe); 4284 4285 if (tvp) { 4286 /* 4287 * This is va_seq safe because we are not 4288 * manipulating dvp. 4289 */ 4290 va.va_mask = AT_NLINK; 4291 if (!VOP_GETATTR(tvp, &va, 0, cs->cr, NULL) && 4292 va.va_nlink == 0) { 4293 /* Remove state on file remove */ 4294 if (in_crit) { 4295 nbl_end_crit(vp); 4296 in_crit = 0; 4297 } 4298 rfs4_close_all_state(fp); 4299 } 4300 VN_RELE(tvp); 4301 } 4302 } 4303 } 4304 4305 if (in_crit) 4306 nbl_end_crit(vp); 4307 VN_RELE(vp); 4308 4309 if (fp) { 4310 rfs4_clear_dont_grant(fp); 4311 rfs4_file_rele(fp); 4312 } 4313 if (nm != name) 4314 kmem_free(name, MAXPATHLEN + 1); 4315 kmem_free(nm, len); 4316 4317 if (error) { 4318 *cs->statusp = resp->status = puterrno4(error); 4319 goto out; 4320 } 4321 4322 /* 4323 * Get the initial "after" sequence number, if it fails, set to zero 4324 */ 4325 idva.va_mask = AT_SEQ; 4326 if (VOP_GETATTR(dvp, &idva, 0, cs->cr, NULL)) 4327 idva.va_seq = 0; 4328 4329 /* 4330 * Force modified data and metadata out to stable storage. 4331 */ 4332 (void) VOP_FSYNC(dvp, 0, cs->cr, NULL); 4333 4334 /* 4335 * Get "after" change value, if it fails, simply return the 4336 * before value. 4337 */ 4338 adva.va_mask = AT_CTIME|AT_SEQ; 4339 if (VOP_GETATTR(dvp, &adva, 0, cs->cr, NULL)) { 4340 adva.va_ctime = bdva.va_ctime; 4341 adva.va_seq = 0; 4342 } 4343 4344 NFS4_SET_FATTR4_CHANGE(resp->cinfo.after, adva.va_ctime) 4345 4346 /* 4347 * The cinfo.atomic = TRUE only if we have 4348 * non-zero va_seq's, and it has incremented by exactly one 4349 * during the VOP_REMOVE/RMDIR and it didn't change during 4350 * the VOP_FSYNC. 4351 */ 4352 if (bdva.va_seq && idva.va_seq && adva.va_seq && 4353 idva.va_seq == (bdva.va_seq + 1) && idva.va_seq == adva.va_seq) 4354 resp->cinfo.atomic = TRUE; 4355 else 4356 resp->cinfo.atomic = FALSE; 4357 4358 *cs->statusp = resp->status = NFS4_OK; 4359 4360 out: 4361 DTRACE_NFSV4_2(op__remove__done, struct compound_state *, cs, 4362 REMOVE4res *, resp); 4363 } 4364 4365 /* 4366 * rename: args: SAVED_FH: from directory, CURRENT_FH: target directory, 4367 * oldname and newname. 4368 * res: status. If success - CURRENT_FH unchanged, return change_info 4369 * for both from and target directories. 4370 */ 4371 /* ARGSUSED */ 4372 static void 4373 rfs4_op_rename(nfs_argop4 *argop, nfs_resop4 *resop, struct svc_req *req, 4374 struct compound_state *cs) 4375 { 4376 RENAME4args *args = &argop->nfs_argop4_u.oprename; 4377 RENAME4res *resp = &resop->nfs_resop4_u.oprename; 4378 int error; 4379 vnode_t *odvp; 4380 vnode_t *ndvp; 4381 vnode_t *srcvp, *targvp, *tvp; 4382 struct vattr obdva, oidva, oadva; 4383 struct vattr nbdva, nidva, nadva; 4384 char *onm, *nnm; 4385 uint_t olen, nlen; 4386 rfs4_file_t *fp, *sfp; 4387 int in_crit_src, in_crit_targ; 4388 int fp_rele_grant_hold, sfp_rele_grant_hold; 4389 int unlinked; 4390 bslabel_t *clabel; 4391 struct sockaddr *ca; 4392 char *converted_onm = NULL; 4393 char *converted_nnm = NULL; 4394 nfsstat4 status; 4395 4396 DTRACE_NFSV4_2(op__rename__start, struct compound_state *, cs, 4397 RENAME4args *, args); 4398 4399 fp = sfp = NULL; 4400 srcvp = targvp = tvp = NULL; 4401 in_crit_src = in_crit_targ = 0; 4402 fp_rele_grant_hold = sfp_rele_grant_hold = 0; 4403 unlinked = 0; 4404 4405 /* CURRENT_FH: target directory */ 4406 ndvp = cs->vp; 4407 if (ndvp == NULL) { 4408 *cs->statusp = resp->status = NFS4ERR_NOFILEHANDLE; 4409 goto out; 4410 } 4411 4412 /* SAVED_FH: from directory */ 4413 odvp = cs->saved_vp; 4414 if (odvp == NULL) { 4415 *cs->statusp = resp->status = NFS4ERR_NOFILEHANDLE; 4416 goto out; 4417 } 4418 4419 if (cs->access == CS_ACCESS_DENIED) { 4420 *cs->statusp = resp->status = NFS4ERR_ACCESS; 4421 goto out; 4422 } 4423 4424 /* 4425 * If there is an unshared filesystem mounted on this vnode, 4426 * do not allow to rename objects in this directory. 4427 */ 4428 if (vn_ismntpt(odvp)) { 4429 *cs->statusp = resp->status = NFS4ERR_ACCESS; 4430 goto out; 4431 } 4432 4433 /* 4434 * If there is an unshared filesystem mounted on this vnode, 4435 * do not allow to rename to this directory. 4436 */ 4437 if (vn_ismntpt(ndvp)) { 4438 *cs->statusp = resp->status = NFS4ERR_ACCESS; 4439 goto out; 4440 } 4441 4442 if (odvp->v_type != VDIR || ndvp->v_type != VDIR) { 4443 *cs->statusp = resp->status = NFS4ERR_NOTDIR; 4444 goto out; 4445 } 4446 4447 if (cs->saved_exi != cs->exi) { 4448 *cs->statusp = resp->status = NFS4ERR_XDEV; 4449 goto out; 4450 } 4451 4452 status = utf8_dir_verify(&args->oldname); 4453 if (status != NFS4_OK) { 4454 *cs->statusp = resp->status = status; 4455 goto out; 4456 } 4457 4458 status = utf8_dir_verify(&args->newname); 4459 if (status != NFS4_OK) { 4460 *cs->statusp = resp->status = status; 4461 goto out; 4462 } 4463 4464 onm = utf8_to_fn(&args->oldname, &olen, NULL); 4465 if (onm == NULL) { 4466 *cs->statusp = resp->status = NFS4ERR_INVAL; 4467 goto out; 4468 } 4469 ca = (struct sockaddr *)svc_getrpccaller(req->rq_xprt)->buf; 4470 nlen = MAXPATHLEN + 1; 4471 converted_onm = nfscmd_convname(ca, cs->exi, onm, NFSCMD_CONV_INBOUND, 4472 nlen); 4473 4474 if (converted_onm == NULL) { 4475 *cs->statusp = resp->status = NFS4ERR_INVAL; 4476 kmem_free(onm, olen); 4477 goto out; 4478 } 4479 4480 nnm = utf8_to_fn(&args->newname, &nlen, NULL); 4481 if (nnm == NULL) { 4482 *cs->statusp = resp->status = NFS4ERR_INVAL; 4483 if (onm != converted_onm) 4484 kmem_free(converted_onm, MAXPATHLEN + 1); 4485 kmem_free(onm, olen); 4486 goto out; 4487 } 4488 converted_nnm = nfscmd_convname(ca, cs->exi, nnm, NFSCMD_CONV_INBOUND, 4489 MAXPATHLEN + 1); 4490 4491 if (converted_nnm == NULL) { 4492 *cs->statusp = resp->status = NFS4ERR_INVAL; 4493 kmem_free(nnm, nlen); 4494 nnm = NULL; 4495 if (onm != converted_onm) 4496 kmem_free(converted_onm, MAXPATHLEN + 1); 4497 kmem_free(onm, olen); 4498 goto out; 4499 } 4500 4501 4502 if (olen > MAXNAMELEN || nlen > MAXNAMELEN) { 4503 *cs->statusp = resp->status = NFS4ERR_NAMETOOLONG; 4504 kmem_free(onm, olen); 4505 kmem_free(nnm, nlen); 4506 goto out; 4507 } 4508 4509 4510 if (rdonly4(req, cs)) { 4511 *cs->statusp = resp->status = NFS4ERR_ROFS; 4512 if (onm != converted_onm) 4513 kmem_free(converted_onm, MAXPATHLEN + 1); 4514 kmem_free(onm, olen); 4515 if (nnm != converted_nnm) 4516 kmem_free(converted_nnm, MAXPATHLEN + 1); 4517 kmem_free(nnm, nlen); 4518 goto out; 4519 } 4520 4521 /* check label of the target dir */ 4522 if (is_system_labeled()) { 4523 ASSERT(req->rq_label != NULL); 4524 clabel = req->rq_label; 4525 DTRACE_PROBE2(tx__rfs4__log__info__oprename__clabel, char *, 4526 "got client label from request(1)", 4527 struct svc_req *, req); 4528 if (!blequal(&l_admin_low->tsl_label, clabel)) { 4529 if (!do_rfs_label_check(clabel, ndvp, 4530 EQUALITY_CHECK, cs->exi)) { 4531 *cs->statusp = resp->status = NFS4ERR_ACCESS; 4532 goto err_out; 4533 } 4534 } 4535 } 4536 4537 /* 4538 * Is the source a file and have a delegation? 4539 * We don't need to acquire va_seq before these lookups, if 4540 * it causes an update, cinfo.before will not match, which will 4541 * trigger a cache flush even if atomic is TRUE. 4542 */ 4543 if (sfp = rfs4_lookup_and_findfile(odvp, converted_onm, &srcvp, 4544 &error, cs->cr)) { 4545 if (rfs4_check_delegated_byfp(FWRITE, sfp, TRUE, TRUE, TRUE, 4546 NULL)) { 4547 *cs->statusp = resp->status = NFS4ERR_DELAY; 4548 goto err_out; 4549 } 4550 } 4551 4552 if (srcvp == NULL) { 4553 *cs->statusp = resp->status = puterrno4(error); 4554 if (onm != converted_onm) 4555 kmem_free(converted_onm, MAXPATHLEN + 1); 4556 kmem_free(onm, olen); 4557 if (nnm != converted_nnm) 4558 kmem_free(converted_nnm, MAXPATHLEN + 1); 4559 kmem_free(nnm, nlen); 4560 goto out; 4561 } 4562 4563 sfp_rele_grant_hold = 1; 4564 4565 /* Does the destination exist and a file and have a delegation? */ 4566 if (fp = rfs4_lookup_and_findfile(ndvp, converted_nnm, &targvp, 4567 NULL, cs->cr)) { 4568 if (rfs4_check_delegated_byfp(FWRITE, fp, TRUE, TRUE, TRUE, 4569 NULL)) { 4570 *cs->statusp = resp->status = NFS4ERR_DELAY; 4571 goto err_out; 4572 } 4573 } 4574 fp_rele_grant_hold = 1; 4575 4576 /* Check for NBMAND lock on both source and target */ 4577 if (nbl_need_check(srcvp)) { 4578 nbl_start_crit(srcvp, RW_READER); 4579 in_crit_src = 1; 4580 if (nbl_conflict(srcvp, NBL_RENAME, 0, 0, 0, NULL)) { 4581 *cs->statusp = resp->status = NFS4ERR_FILE_OPEN; 4582 goto err_out; 4583 } 4584 } 4585 4586 if (targvp && nbl_need_check(targvp)) { 4587 nbl_start_crit(targvp, RW_READER); 4588 in_crit_targ = 1; 4589 if (nbl_conflict(targvp, NBL_REMOVE, 0, 0, 0, NULL)) { 4590 *cs->statusp = resp->status = NFS4ERR_FILE_OPEN; 4591 goto err_out; 4592 } 4593 } 4594 4595 /* Get source "before" change value */ 4596 obdva.va_mask = AT_CTIME|AT_SEQ; 4597 error = VOP_GETATTR(odvp, &obdva, 0, cs->cr, NULL); 4598 if (!error) { 4599 nbdva.va_mask = AT_CTIME|AT_SEQ; 4600 error = VOP_GETATTR(ndvp, &nbdva, 0, cs->cr, NULL); 4601 } 4602 if (error) { 4603 *cs->statusp = resp->status = puterrno4(error); 4604 goto err_out; 4605 } 4606 4607 NFS4_SET_FATTR4_CHANGE(resp->source_cinfo.before, obdva.va_ctime) 4608 NFS4_SET_FATTR4_CHANGE(resp->target_cinfo.before, nbdva.va_ctime) 4609 4610 error = VOP_RENAME(odvp, converted_onm, ndvp, converted_nnm, cs->cr, 4611 NULL, 0); 4612 4613 /* 4614 * If target existed and was unlinked by VOP_RENAME, state will need 4615 * closed. To avoid deadlock, rfs4_close_all_state will be done after 4616 * any necessary nbl_end_crit on srcvp and tgtvp. 4617 */ 4618 if (error == 0 && fp != NULL) { 4619 rfs4_dbe_lock(fp->rf_dbe); 4620 tvp = fp->rf_vp; 4621 if (tvp) 4622 VN_HOLD(tvp); 4623 rfs4_dbe_unlock(fp->rf_dbe); 4624 4625 if (tvp) { 4626 struct vattr va; 4627 va.va_mask = AT_NLINK; 4628 4629 if (!VOP_GETATTR(tvp, &va, 0, cs->cr, NULL) && 4630 va.va_nlink == 0) { 4631 unlinked = 1; 4632 4633 /* DEBUG data */ 4634 if ((srcvp == targvp) || (tvp != targvp)) { 4635 cmn_err(CE_WARN, "rfs4_op_rename: " 4636 "srcvp %p, targvp: %p, tvp: %p", 4637 (void *)srcvp, (void *)targvp, 4638 (void *)tvp); 4639 } 4640 } else { 4641 VN_RELE(tvp); 4642 } 4643 } 4644 } 4645 if (error == 0) 4646 vn_renamepath(ndvp, srcvp, nnm, nlen - 1); 4647 4648 if (in_crit_src) 4649 nbl_end_crit(srcvp); 4650 if (srcvp) 4651 VN_RELE(srcvp); 4652 if (in_crit_targ) 4653 nbl_end_crit(targvp); 4654 if (targvp) 4655 VN_RELE(targvp); 4656 4657 if (unlinked) { 4658 ASSERT(fp != NULL); 4659 ASSERT(tvp != NULL); 4660 4661 /* DEBUG data */ 4662 if (RW_READ_HELD(&tvp->v_nbllock)) { 4663 cmn_err(CE_WARN, "rfs4_op_rename: " 4664 "RW_READ_HELD(%p)", (void *)tvp); 4665 } 4666 4667 /* The file is gone and so should the state */ 4668 rfs4_close_all_state(fp); 4669 VN_RELE(tvp); 4670 } 4671 4672 if (sfp) { 4673 rfs4_clear_dont_grant(sfp); 4674 rfs4_file_rele(sfp); 4675 } 4676 if (fp) { 4677 rfs4_clear_dont_grant(fp); 4678 rfs4_file_rele(fp); 4679 } 4680 4681 if (converted_onm != onm) 4682 kmem_free(converted_onm, MAXPATHLEN + 1); 4683 kmem_free(onm, olen); 4684 if (converted_nnm != nnm) 4685 kmem_free(converted_nnm, MAXPATHLEN + 1); 4686 kmem_free(nnm, nlen); 4687 4688 /* 4689 * Get the initial "after" sequence number, if it fails, set to zero 4690 */ 4691 oidva.va_mask = AT_SEQ; 4692 if (VOP_GETATTR(odvp, &oidva, 0, cs->cr, NULL)) 4693 oidva.va_seq = 0; 4694 4695 nidva.va_mask = AT_SEQ; 4696 if (VOP_GETATTR(ndvp, &nidva, 0, cs->cr, NULL)) 4697 nidva.va_seq = 0; 4698 4699 /* 4700 * Force modified data and metadata out to stable storage. 4701 */ 4702 (void) VOP_FSYNC(odvp, 0, cs->cr, NULL); 4703 (void) VOP_FSYNC(ndvp, 0, cs->cr, NULL); 4704 4705 if (error) { 4706 *cs->statusp = resp->status = puterrno4(error); 4707 goto out; 4708 } 4709 4710 /* 4711 * Get "after" change values, if it fails, simply return the 4712 * before value. 4713 */ 4714 oadva.va_mask = AT_CTIME|AT_SEQ; 4715 if (VOP_GETATTR(odvp, &oadva, 0, cs->cr, NULL)) { 4716 oadva.va_ctime = obdva.va_ctime; 4717 oadva.va_seq = 0; 4718 } 4719 4720 nadva.va_mask = AT_CTIME|AT_SEQ; 4721 if (VOP_GETATTR(odvp, &nadva, 0, cs->cr, NULL)) { 4722 nadva.va_ctime = nbdva.va_ctime; 4723 nadva.va_seq = 0; 4724 } 4725 4726 NFS4_SET_FATTR4_CHANGE(resp->source_cinfo.after, oadva.va_ctime) 4727 NFS4_SET_FATTR4_CHANGE(resp->target_cinfo.after, nadva.va_ctime) 4728 4729 /* 4730 * The cinfo.atomic = TRUE only if we have 4731 * non-zero va_seq's, and it has incremented by exactly one 4732 * during the VOP_RENAME and it didn't change during the VOP_FSYNC. 4733 */ 4734 if (obdva.va_seq && oidva.va_seq && oadva.va_seq && 4735 oidva.va_seq == (obdva.va_seq + 1) && oidva.va_seq == oadva.va_seq) 4736 resp->source_cinfo.atomic = TRUE; 4737 else 4738 resp->source_cinfo.atomic = FALSE; 4739 4740 if (nbdva.va_seq && nidva.va_seq && nadva.va_seq && 4741 nidva.va_seq == (nbdva.va_seq + 1) && nidva.va_seq == nadva.va_seq) 4742 resp->target_cinfo.atomic = TRUE; 4743 else 4744 resp->target_cinfo.atomic = FALSE; 4745 4746 #ifdef VOLATILE_FH_TEST 4747 { 4748 extern void add_volrnm_fh(struct exportinfo *, vnode_t *); 4749 4750 /* 4751 * Add the renamed file handle to the volatile rename list 4752 */ 4753 if (cs->exi->exi_export.ex_flags & EX_VOLRNM) { 4754 /* file handles may expire on rename */ 4755 vnode_t *vp; 4756 4757 nnm = utf8_to_fn(&args->newname, &nlen, NULL); 4758 /* 4759 * Already know that nnm will be a valid string 4760 */ 4761 error = VOP_LOOKUP(ndvp, nnm, &vp, NULL, 0, NULL, cs->cr, 4762 NULL, NULL, NULL); 4763 kmem_free(nnm, nlen); 4764 if (!error) { 4765 add_volrnm_fh(cs->exi, vp); 4766 VN_RELE(vp); 4767 } 4768 } 4769 } 4770 #endif /* VOLATILE_FH_TEST */ 4771 4772 *cs->statusp = resp->status = NFS4_OK; 4773 out: 4774 DTRACE_NFSV4_2(op__rename__done, struct compound_state *, cs, 4775 RENAME4res *, resp); 4776 return; 4777 4778 err_out: 4779 if (onm != converted_onm) 4780 kmem_free(converted_onm, MAXPATHLEN + 1); 4781 if (onm != NULL) 4782 kmem_free(onm, olen); 4783 if (nnm != converted_nnm) 4784 kmem_free(converted_nnm, MAXPATHLEN + 1); 4785 if (nnm != NULL) 4786 kmem_free(nnm, nlen); 4787 4788 if (in_crit_src) nbl_end_crit(srcvp); 4789 if (in_crit_targ) nbl_end_crit(targvp); 4790 if (targvp) VN_RELE(targvp); 4791 if (srcvp) VN_RELE(srcvp); 4792 if (sfp) { 4793 if (sfp_rele_grant_hold) rfs4_clear_dont_grant(sfp); 4794 rfs4_file_rele(sfp); 4795 } 4796 if (fp) { 4797 if (fp_rele_grant_hold) rfs4_clear_dont_grant(fp); 4798 rfs4_file_rele(fp); 4799 } 4800 4801 DTRACE_NFSV4_2(op__rename__done, struct compound_state *, cs, 4802 RENAME4res *, resp); 4803 } 4804 4805 /* ARGSUSED */ 4806 static void 4807 rfs4_op_renew(nfs_argop4 *argop, nfs_resop4 *resop, struct svc_req *req, 4808 struct compound_state *cs) 4809 { 4810 RENEW4args *args = &argop->nfs_argop4_u.oprenew; 4811 RENEW4res *resp = &resop->nfs_resop4_u.oprenew; 4812 rfs4_client_t *cp; 4813 4814 DTRACE_NFSV4_2(op__renew__start, struct compound_state *, cs, 4815 RENEW4args *, args); 4816 4817 if ((cp = rfs4_findclient_by_id(args->clientid, FALSE)) == NULL) { 4818 *cs->statusp = resp->status = 4819 rfs4_check_clientid(&args->clientid, 0); 4820 goto out; 4821 } 4822 4823 if (rfs4_lease_expired(cp)) { 4824 rfs4_client_rele(cp); 4825 *cs->statusp = resp->status = NFS4ERR_EXPIRED; 4826 goto out; 4827 } 4828 4829 rfs4_update_lease(cp); 4830 4831 mutex_enter(cp->rc_cbinfo.cb_lock); 4832 if (cp->rc_cbinfo.cb_notified_of_cb_path_down == FALSE) { 4833 cp->rc_cbinfo.cb_notified_of_cb_path_down = TRUE; 4834 *cs->statusp = resp->status = NFS4ERR_CB_PATH_DOWN; 4835 } else { 4836 *cs->statusp = resp->status = NFS4_OK; 4837 } 4838 mutex_exit(cp->rc_cbinfo.cb_lock); 4839 4840 rfs4_client_rele(cp); 4841 4842 out: 4843 DTRACE_NFSV4_2(op__renew__done, struct compound_state *, cs, 4844 RENEW4res *, resp); 4845 } 4846 4847 /* ARGSUSED */ 4848 static void 4849 rfs4_op_restorefh(nfs_argop4 *args, nfs_resop4 *resop, struct svc_req *req, 4850 struct compound_state *cs) 4851 { 4852 RESTOREFH4res *resp = &resop->nfs_resop4_u.oprestorefh; 4853 4854 DTRACE_NFSV4_1(op__restorefh__start, struct compound_state *, cs); 4855 4856 /* No need to check cs->access - we are not accessing any object */ 4857 if ((cs->saved_vp == NULL) || (cs->saved_fh.nfs_fh4_val == NULL)) { 4858 *cs->statusp = resp->status = NFS4ERR_RESTOREFH; 4859 goto out; 4860 } 4861 if (cs->vp != NULL) { 4862 VN_RELE(cs->vp); 4863 } 4864 cs->vp = cs->saved_vp; 4865 cs->saved_vp = NULL; 4866 cs->exi = cs->saved_exi; 4867 nfs_fh4_copy(&cs->saved_fh, &cs->fh); 4868 *cs->statusp = resp->status = NFS4_OK; 4869 cs->deleg = FALSE; 4870 4871 out: 4872 DTRACE_NFSV4_2(op__restorefh__done, struct compound_state *, cs, 4873 RESTOREFH4res *, resp); 4874 } 4875 4876 /* ARGSUSED */ 4877 static void 4878 rfs4_op_savefh(nfs_argop4 *argop, nfs_resop4 *resop, struct svc_req *req, 4879 struct compound_state *cs) 4880 { 4881 SAVEFH4res *resp = &resop->nfs_resop4_u.opsavefh; 4882 4883 DTRACE_NFSV4_1(op__savefh__start, struct compound_state *, cs); 4884 4885 /* No need to check cs->access - we are not accessing any object */ 4886 if (cs->vp == NULL) { 4887 *cs->statusp = resp->status = NFS4ERR_NOFILEHANDLE; 4888 goto out; 4889 } 4890 if (cs->saved_vp != NULL) { 4891 VN_RELE(cs->saved_vp); 4892 } 4893 cs->saved_vp = cs->vp; 4894 VN_HOLD(cs->saved_vp); 4895 cs->saved_exi = cs->exi; 4896 /* 4897 * since SAVEFH is fairly rare, don't alloc space for its fh 4898 * unless necessary. 4899 */ 4900 if (cs->saved_fh.nfs_fh4_val == NULL) { 4901 cs->saved_fh.nfs_fh4_val = kmem_alloc(NFS4_FHSIZE, KM_SLEEP); 4902 } 4903 nfs_fh4_copy(&cs->fh, &cs->saved_fh); 4904 *cs->statusp = resp->status = NFS4_OK; 4905 4906 out: 4907 DTRACE_NFSV4_2(op__savefh__done, struct compound_state *, cs, 4908 SAVEFH4res *, resp); 4909 } 4910 4911 /* 4912 * rfs4_verify_attr is called when nfsv4 Setattr failed, but we wish to 4913 * return the bitmap of attrs that were set successfully. It is also 4914 * called by Verify/Nverify to test the vattr/vfsstat attrs. It should 4915 * always be called only after rfs4_do_set_attrs(). 4916 * 4917 * Verify that the attributes are same as the expected ones. sargp->vap 4918 * and sargp->sbp contain the input attributes as translated from fattr4. 4919 * 4920 * This function verifies only the attrs that correspond to a vattr or 4921 * vfsstat struct. That is because of the extra step needed to get the 4922 * corresponding system structs. Other attributes have already been set or 4923 * verified by do_rfs4_set_attrs. 4924 * 4925 * Return 0 if all attrs match, -1 if some don't, error if error processing. 4926 */ 4927 static int 4928 rfs4_verify_attr(struct nfs4_svgetit_arg *sargp, 4929 bitmap4 *resp, struct nfs4_ntov_table *ntovp) 4930 { 4931 int error, ret_error = 0; 4932 int i, k; 4933 uint_t sva_mask = sargp->vap->va_mask; 4934 uint_t vbit; 4935 union nfs4_attr_u *na; 4936 uint8_t *amap; 4937 bool_t getsb = ntovp->vfsstat; 4938 4939 if (sva_mask != 0) { 4940 /* 4941 * Okay to overwrite sargp->vap because we verify based 4942 * on the incoming values. 4943 */ 4944 ret_error = VOP_GETATTR(sargp->cs->vp, sargp->vap, 0, 4945 sargp->cs->cr, NULL); 4946 if (ret_error) { 4947 if (resp == NULL) 4948 return (ret_error); 4949 /* 4950 * Must return bitmap of successful attrs 4951 */ 4952 sva_mask = 0; /* to prevent checking vap later */ 4953 } else { 4954 /* 4955 * Some file systems clobber va_mask. it is probably 4956 * wrong of them to do so, nonethless we practice 4957 * defensive coding. 4958 * See bug id 4276830. 4959 */ 4960 sargp->vap->va_mask = sva_mask; 4961 } 4962 } 4963 4964 if (getsb) { 4965 /* 4966 * Now get the superblock and loop on the bitmap, as there is 4967 * no simple way of translating from superblock to bitmap4. 4968 */ 4969 ret_error = VFS_STATVFS(sargp->cs->vp->v_vfsp, sargp->sbp); 4970 if (ret_error) { 4971 if (resp == NULL) 4972 goto errout; 4973 getsb = FALSE; 4974 } 4975 } 4976 4977 /* 4978 * Now loop and verify each attribute which getattr returned 4979 * whether it's the same as the input. 4980 */ 4981 if (resp == NULL && !getsb && (sva_mask == 0)) 4982 goto errout; 4983 4984 na = ntovp->na; 4985 amap = ntovp->amap; 4986 k = 0; 4987 for (i = 0; i < ntovp->attrcnt; i++, na++, amap++) { 4988 k = *amap; 4989 ASSERT(nfs4_ntov_map[k].nval == k); 4990 vbit = nfs4_ntov_map[k].vbit; 4991 4992 /* 4993 * If vattr attribute but VOP_GETATTR failed, or it's 4994 * superblock attribute but VFS_STATVFS failed, skip 4995 */ 4996 if (vbit) { 4997 if ((vbit & sva_mask) == 0) 4998 continue; 4999 } else if (!(getsb && nfs4_ntov_map[k].vfsstat)) { 5000 continue; 5001 } 5002 error = (*nfs4_ntov_map[k].sv_getit)(NFS4ATTR_VERIT, sargp, na); 5003 if (resp != NULL) { 5004 if (error) 5005 ret_error = -1; /* not all match */ 5006 else /* update response bitmap */ 5007 *resp |= nfs4_ntov_map[k].fbit; 5008 continue; 5009 } 5010 if (error) { 5011 ret_error = -1; /* not all match */ 5012 break; 5013 } 5014 } 5015 errout: 5016 return (ret_error); 5017 } 5018 5019 /* 5020 * Decode the attribute to be set/verified. If the attr requires a sys op 5021 * (VOP_GETATTR, VFS_VFSSTAT), and the request is to verify, then don't 5022 * call the sv_getit function for it, because the sys op hasn't yet been done. 5023 * Return 0 for success, error code if failed. 5024 * 5025 * Note: the decoded arg is not freed here but in nfs4_ntov_table_free. 5026 */ 5027 static int 5028 decode_fattr4_attr(nfs4_attr_cmd_t cmd, struct nfs4_svgetit_arg *sargp, 5029 int k, XDR *xdrp, bitmap4 *resp_bval, union nfs4_attr_u *nap) 5030 { 5031 int error = 0; 5032 bool_t set_later; 5033 5034 sargp->vap->va_mask |= nfs4_ntov_map[k].vbit; 5035 5036 if ((*nfs4_ntov_map[k].xfunc)(xdrp, nap)) { 5037 set_later = nfs4_ntov_map[k].vbit || nfs4_ntov_map[k].vfsstat; 5038 /* 5039 * don't verify yet if a vattr or sb dependent attr, 5040 * because we don't have their sys values yet. 5041 * Will be done later. 5042 */ 5043 if (! (set_later && (cmd == NFS4ATTR_VERIT))) { 5044 /* 5045 * ACLs are a special case, since setting the MODE 5046 * conflicts with setting the ACL. We delay setting 5047 * the ACL until all other attributes have been set. 5048 * The ACL gets set in do_rfs4_op_setattr(). 5049 */ 5050 if (nfs4_ntov_map[k].fbit != FATTR4_ACL_MASK) { 5051 error = (*nfs4_ntov_map[k].sv_getit)(cmd, 5052 sargp, nap); 5053 if (error) { 5054 xdr_free(nfs4_ntov_map[k].xfunc, 5055 (caddr_t)nap); 5056 } 5057 } 5058 } 5059 } else { 5060 #ifdef DEBUG 5061 cmn_err(CE_NOTE, "decode_fattr4_attr: error " 5062 "decoding attribute %d\n", k); 5063 #endif 5064 error = EINVAL; 5065 } 5066 if (!error && resp_bval && !set_later) { 5067 *resp_bval |= nfs4_ntov_map[k].fbit; 5068 } 5069 5070 return (error); 5071 } 5072 5073 /* 5074 * Set vattr based on incoming fattr4 attrs - used by setattr. 5075 * Set response mask. Ignore any values that are not writable vattr attrs. 5076 */ 5077 static nfsstat4 5078 do_rfs4_set_attrs(bitmap4 *resp, fattr4 *fattrp, struct compound_state *cs, 5079 struct nfs4_svgetit_arg *sargp, struct nfs4_ntov_table *ntovp, 5080 nfs4_attr_cmd_t cmd) 5081 { 5082 int error = 0; 5083 int i; 5084 char *attrs = fattrp->attrlist4; 5085 uint32_t attrslen = fattrp->attrlist4_len; 5086 XDR xdr; 5087 nfsstat4 status = NFS4_OK; 5088 vnode_t *vp = cs->vp; 5089 union nfs4_attr_u *na; 5090 uint8_t *amap; 5091 5092 #ifndef lint 5093 /* 5094 * Make sure that maximum attribute number can be expressed as an 5095 * 8 bit quantity. 5096 */ 5097 ASSERT(NFS4_MAXNUM_ATTRS <= (UINT8_MAX + 1)); 5098 #endif 5099 5100 if (vp == NULL) { 5101 if (resp) 5102 *resp = 0; 5103 return (NFS4ERR_NOFILEHANDLE); 5104 } 5105 if (cs->access == CS_ACCESS_DENIED) { 5106 if (resp) 5107 *resp = 0; 5108 return (NFS4ERR_ACCESS); 5109 } 5110 5111 sargp->op = cmd; 5112 sargp->cs = cs; 5113 sargp->flag = 0; /* may be set later */ 5114 sargp->vap->va_mask = 0; 5115 sargp->rdattr_error = NFS4_OK; 5116 sargp->rdattr_error_req = FALSE; 5117 /* sargp->sbp is set by the caller */ 5118 5119 xdrmem_create(&xdr, attrs, attrslen, XDR_DECODE); 5120 5121 na = ntovp->na; 5122 amap = ntovp->amap; 5123 5124 /* 5125 * The following loop iterates on the nfs4_ntov_map checking 5126 * if the fbit is set in the requested bitmap. 5127 * If set then we process the arguments using the 5128 * rfs4_fattr4 conversion functions to populate the setattr 5129 * vattr and va_mask. Any settable attrs that are not using vattr 5130 * will be set in this loop. 5131 */ 5132 for (i = 0; i < nfs4_ntov_map_size; i++) { 5133 if (!(fattrp->attrmask & nfs4_ntov_map[i].fbit)) { 5134 continue; 5135 } 5136 /* 5137 * If setattr, must be a writable attr. 5138 * If verify/nverify, must be a readable attr. 5139 */ 5140 if ((error = (*nfs4_ntov_map[i].sv_getit)( 5141 NFS4ATTR_SUPPORTED, sargp, NULL)) != 0) { 5142 /* 5143 * Client tries to set/verify an 5144 * unsupported attribute, tries to set 5145 * a read only attr or verify a write 5146 * only one - error! 5147 */ 5148 break; 5149 } 5150 /* 5151 * Decode the attribute to set/verify 5152 */ 5153 error = decode_fattr4_attr(cmd, sargp, nfs4_ntov_map[i].nval, 5154 &xdr, resp ? resp : NULL, na); 5155 if (error) 5156 break; 5157 *amap++ = (uint8_t)nfs4_ntov_map[i].nval; 5158 na++; 5159 (ntovp->attrcnt)++; 5160 if (nfs4_ntov_map[i].vfsstat) 5161 ntovp->vfsstat = TRUE; 5162 } 5163 5164 if (error != 0) 5165 status = (error == ENOTSUP ? NFS4ERR_ATTRNOTSUPP : 5166 puterrno4(error)); 5167 /* xdrmem_destroy(&xdrs); */ /* NO-OP */ 5168 return (status); 5169 } 5170 5171 static nfsstat4 5172 do_rfs4_op_setattr(bitmap4 *resp, fattr4 *fattrp, struct compound_state *cs, 5173 stateid4 *stateid) 5174 { 5175 int error = 0; 5176 struct nfs4_svgetit_arg sarg; 5177 bool_t trunc; 5178 5179 nfsstat4 status = NFS4_OK; 5180 cred_t *cr = cs->cr; 5181 vnode_t *vp = cs->vp; 5182 struct nfs4_ntov_table ntov; 5183 struct statvfs64 sb; 5184 struct vattr bva; 5185 struct flock64 bf; 5186 int in_crit = 0; 5187 uint_t saved_mask = 0; 5188 caller_context_t ct; 5189 5190 *resp = 0; 5191 sarg.sbp = &sb; 5192 sarg.is_referral = B_FALSE; 5193 nfs4_ntov_table_init(&ntov); 5194 status = do_rfs4_set_attrs(resp, fattrp, cs, &sarg, &ntov, 5195 NFS4ATTR_SETIT); 5196 if (status != NFS4_OK) { 5197 /* 5198 * failed set attrs 5199 */ 5200 goto done; 5201 } 5202 if ((sarg.vap->va_mask == 0) && 5203 (! (fattrp->attrmask & FATTR4_ACL_MASK))) { 5204 /* 5205 * no further work to be done 5206 */ 5207 goto done; 5208 } 5209 5210 /* 5211 * If we got a request to set the ACL and the MODE, only 5212 * allow changing VSUID, VSGID, and VSVTX. Attempting 5213 * to change any other bits, along with setting an ACL, 5214 * gives NFS4ERR_INVAL. 5215 */ 5216 if ((fattrp->attrmask & FATTR4_ACL_MASK) && 5217 (fattrp->attrmask & FATTR4_MODE_MASK)) { 5218 vattr_t va; 5219 5220 va.va_mask = AT_MODE; 5221 error = VOP_GETATTR(vp, &va, 0, cs->cr, NULL); 5222 if (error) { 5223 status = puterrno4(error); 5224 goto done; 5225 } 5226 if ((sarg.vap->va_mode ^ va.va_mode) & 5227 ~(VSUID | VSGID | VSVTX)) { 5228 status = NFS4ERR_INVAL; 5229 goto done; 5230 } 5231 } 5232 5233 /* Check stateid only if size has been set */ 5234 if (sarg.vap->va_mask & AT_SIZE) { 5235 trunc = (sarg.vap->va_size == 0); 5236 status = rfs4_check_stateid(FWRITE, cs->vp, stateid, 5237 trunc, &cs->deleg, sarg.vap->va_mask & AT_SIZE, &ct); 5238 if (status != NFS4_OK) 5239 goto done; 5240 } else { 5241 ct.cc_sysid = 0; 5242 ct.cc_pid = 0; 5243 ct.cc_caller_id = nfs4_srv_caller_id; 5244 ct.cc_flags = CC_DONTBLOCK; 5245 } 5246 5247 /* XXX start of possible race with delegations */ 5248 5249 /* 5250 * We need to specially handle size changes because it is 5251 * possible for the client to create a file with read-only 5252 * modes, but with the file opened for writing. If the client 5253 * then tries to set the file size, e.g. ftruncate(3C), 5254 * fcntl(F_FREESP), the normal access checking done in 5255 * VOP_SETATTR would prevent the client from doing it even though 5256 * it should be allowed to do so. To get around this, we do the 5257 * access checking for ourselves and use VOP_SPACE which doesn't 5258 * do the access checking. 5259 * Also the client should not be allowed to change the file 5260 * size if there is a conflicting non-blocking mandatory lock in 5261 * the region of the change. 5262 */ 5263 if (vp->v_type == VREG && (sarg.vap->va_mask & AT_SIZE)) { 5264 u_offset_t offset; 5265 ssize_t length; 5266 5267 /* 5268 * ufs_setattr clears AT_SIZE from vap->va_mask, but 5269 * before returning, sarg.vap->va_mask is used to 5270 * generate the setattr reply bitmap. We also clear 5271 * AT_SIZE below before calling VOP_SPACE. For both 5272 * of these cases, the va_mask needs to be saved here 5273 * and restored after calling VOP_SETATTR. 5274 */ 5275 saved_mask = sarg.vap->va_mask; 5276 5277 /* 5278 * Check any possible conflict due to NBMAND locks. 5279 * Get into critical region before VOP_GETATTR, so the 5280 * size attribute is valid when checking conflicts. 5281 */ 5282 if (nbl_need_check(vp)) { 5283 nbl_start_crit(vp, RW_READER); 5284 in_crit = 1; 5285 } 5286 5287 bva.va_mask = AT_UID|AT_SIZE; 5288 if (error = VOP_GETATTR(vp, &bva, 0, cr, &ct)) { 5289 status = puterrno4(error); 5290 goto done; 5291 } 5292 5293 if (in_crit) { 5294 if (sarg.vap->va_size < bva.va_size) { 5295 offset = sarg.vap->va_size; 5296 length = bva.va_size - sarg.vap->va_size; 5297 } else { 5298 offset = bva.va_size; 5299 length = sarg.vap->va_size - bva.va_size; 5300 } 5301 if (nbl_conflict(vp, NBL_WRITE, offset, length, 0, 5302 &ct)) { 5303 status = NFS4ERR_LOCKED; 5304 goto done; 5305 } 5306 } 5307 5308 if (crgetuid(cr) == bva.va_uid) { 5309 sarg.vap->va_mask &= ~AT_SIZE; 5310 bf.l_type = F_WRLCK; 5311 bf.l_whence = 0; 5312 bf.l_start = (off64_t)sarg.vap->va_size; 5313 bf.l_len = 0; 5314 bf.l_sysid = 0; 5315 bf.l_pid = 0; 5316 error = VOP_SPACE(vp, F_FREESP, &bf, FWRITE, 5317 (offset_t)sarg.vap->va_size, cr, &ct); 5318 } 5319 } 5320 5321 if (!error && sarg.vap->va_mask != 0) 5322 error = VOP_SETATTR(vp, sarg.vap, sarg.flag, cr, &ct); 5323 5324 /* restore va_mask -- ufs_setattr clears AT_SIZE */ 5325 if (saved_mask & AT_SIZE) 5326 sarg.vap->va_mask |= AT_SIZE; 5327 5328 /* 5329 * If an ACL was being set, it has been delayed until now, 5330 * in order to set the mode (via the VOP_SETATTR() above) first. 5331 */ 5332 if ((! error) && (fattrp->attrmask & FATTR4_ACL_MASK)) { 5333 int i; 5334 5335 for (i = 0; i < NFS4_MAXNUM_ATTRS; i++) 5336 if (ntov.amap[i] == FATTR4_ACL) 5337 break; 5338 if (i < NFS4_MAXNUM_ATTRS) { 5339 error = (*nfs4_ntov_map[FATTR4_ACL].sv_getit)( 5340 NFS4ATTR_SETIT, &sarg, &ntov.na[i]); 5341 if (error == 0) { 5342 *resp |= FATTR4_ACL_MASK; 5343 } else if (error == ENOTSUP) { 5344 (void) rfs4_verify_attr(&sarg, resp, &ntov); 5345 status = NFS4ERR_ATTRNOTSUPP; 5346 goto done; 5347 } 5348 } else { 5349 NFS4_DEBUG(rfs4_debug, 5350 (CE_NOTE, "do_rfs4_op_setattr: " 5351 "unable to find ACL in fattr4")); 5352 error = EINVAL; 5353 } 5354 } 5355 5356 if (error) { 5357 /* check if a monitor detected a delegation conflict */ 5358 if (error == EAGAIN && (ct.cc_flags & CC_WOULDBLOCK)) 5359 status = NFS4ERR_DELAY; 5360 else 5361 status = puterrno4(error); 5362 5363 /* 5364 * Set the response bitmap when setattr failed. 5365 * If VOP_SETATTR partially succeeded, test by doing a 5366 * VOP_GETATTR on the object and comparing the data 5367 * to the setattr arguments. 5368 */ 5369 (void) rfs4_verify_attr(&sarg, resp, &ntov); 5370 } else { 5371 /* 5372 * Force modified metadata out to stable storage. 5373 */ 5374 (void) VOP_FSYNC(vp, FNODSYNC, cr, &ct); 5375 /* 5376 * Set response bitmap 5377 */ 5378 nfs4_vmask_to_nmask_set(sarg.vap->va_mask, resp); 5379 } 5380 5381 /* Return early and already have a NFSv4 error */ 5382 done: 5383 /* 5384 * Except for nfs4_vmask_to_nmask_set(), vattr --> fattr 5385 * conversion sets both readable and writeable NFS4 attrs 5386 * for AT_MTIME and AT_ATIME. The line below masks out 5387 * unrequested attrs from the setattr result bitmap. This 5388 * is placed after the done: label to catch the ATTRNOTSUP 5389 * case. 5390 */ 5391 *resp &= fattrp->attrmask; 5392 5393 if (in_crit) 5394 nbl_end_crit(vp); 5395 5396 nfs4_ntov_table_free(&ntov, &sarg); 5397 5398 return (status); 5399 } 5400 5401 /* ARGSUSED */ 5402 static void 5403 rfs4_op_setattr(nfs_argop4 *argop, nfs_resop4 *resop, struct svc_req *req, 5404 struct compound_state *cs) 5405 { 5406 SETATTR4args *args = &argop->nfs_argop4_u.opsetattr; 5407 SETATTR4res *resp = &resop->nfs_resop4_u.opsetattr; 5408 bslabel_t *clabel; 5409 5410 DTRACE_NFSV4_2(op__setattr__start, struct compound_state *, cs, 5411 SETATTR4args *, args); 5412 5413 if (cs->vp == NULL) { 5414 *cs->statusp = resp->status = NFS4ERR_NOFILEHANDLE; 5415 goto out; 5416 } 5417 5418 /* 5419 * If there is an unshared filesystem mounted on this vnode, 5420 * do not allow to setattr on this vnode. 5421 */ 5422 if (vn_ismntpt(cs->vp)) { 5423 *cs->statusp = resp->status = NFS4ERR_ACCESS; 5424 goto out; 5425 } 5426 5427 resp->attrsset = 0; 5428 5429 if (rdonly4(req, cs)) { 5430 *cs->statusp = resp->status = NFS4ERR_ROFS; 5431 goto out; 5432 } 5433 5434 /* check label before setting attributes */ 5435 if (is_system_labeled()) { 5436 ASSERT(req->rq_label != NULL); 5437 clabel = req->rq_label; 5438 DTRACE_PROBE2(tx__rfs4__log__info__opsetattr__clabel, char *, 5439 "got client label from request(1)", 5440 struct svc_req *, req); 5441 if (!blequal(&l_admin_low->tsl_label, clabel)) { 5442 if (!do_rfs_label_check(clabel, cs->vp, 5443 EQUALITY_CHECK, cs->exi)) { 5444 *cs->statusp = resp->status = NFS4ERR_ACCESS; 5445 goto out; 5446 } 5447 } 5448 } 5449 5450 *cs->statusp = resp->status = 5451 do_rfs4_op_setattr(&resp->attrsset, &args->obj_attributes, cs, 5452 &args->stateid); 5453 5454 out: 5455 DTRACE_NFSV4_2(op__setattr__done, struct compound_state *, cs, 5456 SETATTR4res *, resp); 5457 } 5458 5459 /* ARGSUSED */ 5460 static void 5461 rfs4_op_verify(nfs_argop4 *argop, nfs_resop4 *resop, struct svc_req *req, 5462 struct compound_state *cs) 5463 { 5464 /* 5465 * verify and nverify are exactly the same, except that nverify 5466 * succeeds when some argument changed, and verify succeeds when 5467 * when none changed. 5468 */ 5469 5470 VERIFY4args *args = &argop->nfs_argop4_u.opverify; 5471 VERIFY4res *resp = &resop->nfs_resop4_u.opverify; 5472 5473 int error; 5474 struct nfs4_svgetit_arg sarg; 5475 struct statvfs64 sb; 5476 struct nfs4_ntov_table ntov; 5477 5478 DTRACE_NFSV4_2(op__verify__start, struct compound_state *, cs, 5479 VERIFY4args *, args); 5480 5481 if (cs->vp == NULL) { 5482 *cs->statusp = resp->status = NFS4ERR_NOFILEHANDLE; 5483 goto out; 5484 } 5485 5486 sarg.sbp = &sb; 5487 sarg.is_referral = B_FALSE; 5488 nfs4_ntov_table_init(&ntov); 5489 resp->status = do_rfs4_set_attrs(NULL, &args->obj_attributes, cs, 5490 &sarg, &ntov, NFS4ATTR_VERIT); 5491 if (resp->status != NFS4_OK) { 5492 /* 5493 * do_rfs4_set_attrs will try to verify systemwide attrs, 5494 * so could return -1 for "no match". 5495 */ 5496 if (resp->status == -1) 5497 resp->status = NFS4ERR_NOT_SAME; 5498 goto done; 5499 } 5500 error = rfs4_verify_attr(&sarg, NULL, &ntov); 5501 switch (error) { 5502 case 0: 5503 resp->status = NFS4_OK; 5504 break; 5505 case -1: 5506 resp->status = NFS4ERR_NOT_SAME; 5507 break; 5508 default: 5509 resp->status = puterrno4(error); 5510 break; 5511 } 5512 done: 5513 *cs->statusp = resp->status; 5514 nfs4_ntov_table_free(&ntov, &sarg); 5515 out: 5516 DTRACE_NFSV4_2(op__verify__done, struct compound_state *, cs, 5517 VERIFY4res *, resp); 5518 } 5519 5520 /* ARGSUSED */ 5521 static void 5522 rfs4_op_nverify(nfs_argop4 *argop, nfs_resop4 *resop, struct svc_req *req, 5523 struct compound_state *cs) 5524 { 5525 /* 5526 * verify and nverify are exactly the same, except that nverify 5527 * succeeds when some argument changed, and verify succeeds when 5528 * when none changed. 5529 */ 5530 5531 NVERIFY4args *args = &argop->nfs_argop4_u.opnverify; 5532 NVERIFY4res *resp = &resop->nfs_resop4_u.opnverify; 5533 5534 int error; 5535 struct nfs4_svgetit_arg sarg; 5536 struct statvfs64 sb; 5537 struct nfs4_ntov_table ntov; 5538 5539 DTRACE_NFSV4_2(op__nverify__start, struct compound_state *, cs, 5540 NVERIFY4args *, args); 5541 5542 if (cs->vp == NULL) { 5543 *cs->statusp = resp->status = NFS4ERR_NOFILEHANDLE; 5544 DTRACE_NFSV4_2(op__nverify__done, struct compound_state *, cs, 5545 NVERIFY4res *, resp); 5546 return; 5547 } 5548 sarg.sbp = &sb; 5549 sarg.is_referral = B_FALSE; 5550 nfs4_ntov_table_init(&ntov); 5551 resp->status = do_rfs4_set_attrs(NULL, &args->obj_attributes, cs, 5552 &sarg, &ntov, NFS4ATTR_VERIT); 5553 if (resp->status != NFS4_OK) { 5554 /* 5555 * do_rfs4_set_attrs will try to verify systemwide attrs, 5556 * so could return -1 for "no match". 5557 */ 5558 if (resp->status == -1) 5559 resp->status = NFS4_OK; 5560 goto done; 5561 } 5562 error = rfs4_verify_attr(&sarg, NULL, &ntov); 5563 switch (error) { 5564 case 0: 5565 resp->status = NFS4ERR_SAME; 5566 break; 5567 case -1: 5568 resp->status = NFS4_OK; 5569 break; 5570 default: 5571 resp->status = puterrno4(error); 5572 break; 5573 } 5574 done: 5575 *cs->statusp = resp->status; 5576 nfs4_ntov_table_free(&ntov, &sarg); 5577 5578 DTRACE_NFSV4_2(op__nverify__done, struct compound_state *, cs, 5579 NVERIFY4res *, resp); 5580 } 5581 5582 /* 5583 * XXX - This should live in an NFS header file. 5584 */ 5585 #define MAX_IOVECS 12 5586 5587 /* ARGSUSED */ 5588 static void 5589 rfs4_op_write(nfs_argop4 *argop, nfs_resop4 *resop, struct svc_req *req, 5590 struct compound_state *cs) 5591 { 5592 WRITE4args *args = &argop->nfs_argop4_u.opwrite; 5593 WRITE4res *resp = &resop->nfs_resop4_u.opwrite; 5594 int error; 5595 vnode_t *vp; 5596 struct vattr bva; 5597 u_offset_t rlimit; 5598 struct uio uio; 5599 struct iovec iov[MAX_IOVECS]; 5600 struct iovec *iovp; 5601 int iovcnt; 5602 int ioflag; 5603 cred_t *savecred, *cr; 5604 bool_t *deleg = &cs->deleg; 5605 nfsstat4 stat; 5606 int in_crit = 0; 5607 caller_context_t ct; 5608 nfs4_srv_t *nsrv4; 5609 5610 DTRACE_NFSV4_2(op__write__start, struct compound_state *, cs, 5611 WRITE4args *, args); 5612 5613 vp = cs->vp; 5614 if (vp == NULL) { 5615 *cs->statusp = resp->status = NFS4ERR_NOFILEHANDLE; 5616 goto out; 5617 } 5618 if (cs->access == CS_ACCESS_DENIED) { 5619 *cs->statusp = resp->status = NFS4ERR_ACCESS; 5620 goto out; 5621 } 5622 5623 cr = cs->cr; 5624 5625 if ((stat = rfs4_check_stateid(FWRITE, vp, &args->stateid, FALSE, 5626 deleg, TRUE, &ct)) != NFS4_OK) { 5627 *cs->statusp = resp->status = stat; 5628 goto out; 5629 } 5630 5631 /* 5632 * We have to enter the critical region before calling VOP_RWLOCK 5633 * to avoid a deadlock with ufs. 5634 */ 5635 if (nbl_need_check(vp)) { 5636 nbl_start_crit(vp, RW_READER); 5637 in_crit = 1; 5638 if (nbl_conflict(vp, NBL_WRITE, 5639 args->offset, args->data_len, 0, &ct)) { 5640 *cs->statusp = resp->status = NFS4ERR_LOCKED; 5641 goto out; 5642 } 5643 } 5644 5645 bva.va_mask = AT_MODE | AT_UID; 5646 error = VOP_GETATTR(vp, &bva, 0, cr, &ct); 5647 5648 /* 5649 * If we can't get the attributes, then we can't do the 5650 * right access checking. So, we'll fail the request. 5651 */ 5652 if (error) { 5653 *cs->statusp = resp->status = puterrno4(error); 5654 goto out; 5655 } 5656 5657 if (rdonly4(req, cs)) { 5658 *cs->statusp = resp->status = NFS4ERR_ROFS; 5659 goto out; 5660 } 5661 5662 if (vp->v_type != VREG) { 5663 *cs->statusp = resp->status = 5664 ((vp->v_type == VDIR) ? NFS4ERR_ISDIR : NFS4ERR_INVAL); 5665 goto out; 5666 } 5667 5668 if (crgetuid(cr) != bva.va_uid && 5669 (error = VOP_ACCESS(vp, VWRITE, 0, cr, &ct))) { 5670 *cs->statusp = resp->status = puterrno4(error); 5671 goto out; 5672 } 5673 5674 if (MANDLOCK(vp, bva.va_mode)) { 5675 *cs->statusp = resp->status = NFS4ERR_ACCESS; 5676 goto out; 5677 } 5678 5679 nsrv4 = nfs4_get_srv(); 5680 if (args->data_len == 0) { 5681 *cs->statusp = resp->status = NFS4_OK; 5682 resp->count = 0; 5683 resp->committed = args->stable; 5684 resp->writeverf = nsrv4->write4verf; 5685 goto out; 5686 } 5687 5688 if (args->mblk != NULL) { 5689 mblk_t *m; 5690 uint_t bytes, round_len; 5691 5692 iovcnt = 0; 5693 bytes = 0; 5694 round_len = roundup(args->data_len, BYTES_PER_XDR_UNIT); 5695 for (m = args->mblk; 5696 m != NULL && bytes < round_len; 5697 m = m->b_cont) { 5698 iovcnt++; 5699 bytes += MBLKL(m); 5700 } 5701 #ifdef DEBUG 5702 /* should have ended on an mblk boundary */ 5703 if (bytes != round_len) { 5704 printf("bytes=0x%x, round_len=0x%x, req len=0x%x\n", 5705 bytes, round_len, args->data_len); 5706 printf("args=%p, args->mblk=%p, m=%p", (void *)args, 5707 (void *)args->mblk, (void *)m); 5708 ASSERT(bytes == round_len); 5709 } 5710 #endif 5711 if (iovcnt <= MAX_IOVECS) { 5712 iovp = iov; 5713 } else { 5714 iovp = kmem_alloc(sizeof (*iovp) * iovcnt, KM_SLEEP); 5715 } 5716 mblk_to_iov(args->mblk, iovcnt, iovp); 5717 } else if (args->rlist != NULL) { 5718 iovcnt = 1; 5719 iovp = iov; 5720 iovp->iov_base = (char *)((args->rlist)->u.c_daddr3); 5721 iovp->iov_len = args->data_len; 5722 } else { 5723 iovcnt = 1; 5724 iovp = iov; 5725 iovp->iov_base = args->data_val; 5726 iovp->iov_len = args->data_len; 5727 } 5728 5729 uio.uio_iov = iovp; 5730 uio.uio_iovcnt = iovcnt; 5731 5732 uio.uio_segflg = UIO_SYSSPACE; 5733 uio.uio_extflg = UIO_COPY_DEFAULT; 5734 uio.uio_loffset = args->offset; 5735 uio.uio_resid = args->data_len; 5736 uio.uio_llimit = curproc->p_fsz_ctl; 5737 rlimit = uio.uio_llimit - args->offset; 5738 if (rlimit < (u_offset_t)uio.uio_resid) 5739 uio.uio_resid = (int)rlimit; 5740 5741 if (args->stable == UNSTABLE4) 5742 ioflag = 0; 5743 else if (args->stable == FILE_SYNC4) 5744 ioflag = FSYNC; 5745 else if (args->stable == DATA_SYNC4) 5746 ioflag = FDSYNC; 5747 else { 5748 if (iovp != iov) 5749 kmem_free(iovp, sizeof (*iovp) * iovcnt); 5750 *cs->statusp = resp->status = NFS4ERR_INVAL; 5751 goto out; 5752 } 5753 5754 /* 5755 * We're changing creds because VM may fault and we need 5756 * the cred of the current thread to be used if quota 5757 * checking is enabled. 5758 */ 5759 savecred = curthread->t_cred; 5760 curthread->t_cred = cr; 5761 error = do_io(FWRITE, vp, &uio, ioflag, cr, &ct); 5762 curthread->t_cred = savecred; 5763 5764 if (iovp != iov) 5765 kmem_free(iovp, sizeof (*iovp) * iovcnt); 5766 5767 if (error) { 5768 *cs->statusp = resp->status = puterrno4(error); 5769 goto out; 5770 } 5771 5772 *cs->statusp = resp->status = NFS4_OK; 5773 resp->count = args->data_len - uio.uio_resid; 5774 5775 if (ioflag == 0) 5776 resp->committed = UNSTABLE4; 5777 else 5778 resp->committed = FILE_SYNC4; 5779 5780 resp->writeverf = nsrv4->write4verf; 5781 5782 out: 5783 if (in_crit) 5784 nbl_end_crit(vp); 5785 5786 DTRACE_NFSV4_2(op__write__done, struct compound_state *, cs, 5787 WRITE4res *, resp); 5788 } 5789 5790 void 5791 rfs4_compound(COMPOUND4args *args, COMPOUND4res *resp, struct exportinfo *exi, 5792 struct svc_req *req, cred_t *cr, int *rv) 5793 { 5794 uint_t i; 5795 struct compound_state cs; 5796 nfs4_srv_t *nsrv4; 5797 nfs_export_t *ne = nfs_get_export(); 5798 5799 if (rv != NULL) 5800 *rv = 0; 5801 rfs4_init_compound_state(&cs); 5802 /* 5803 * Form a reply tag by copying over the request tag. 5804 */ 5805 resp->tag.utf8string_len = args->tag.utf8string_len; 5806 if (args->tag.utf8string_len != 0) { 5807 resp->tag.utf8string_val = 5808 kmem_alloc(args->tag.utf8string_len, KM_SLEEP); 5809 bcopy(args->tag.utf8string_val, resp->tag.utf8string_val, 5810 resp->tag.utf8string_len); 5811 } else { 5812 resp->tag.utf8string_val = NULL; 5813 } 5814 5815 cs.statusp = &resp->status; 5816 cs.req = req; 5817 resp->array = NULL; 5818 resp->array_len = 0; 5819 5820 /* 5821 * XXX for now, minorversion should be zero 5822 */ 5823 if (args->minorversion != NFS4_MINORVERSION) { 5824 DTRACE_NFSV4_2(compound__start, struct compound_state *, 5825 &cs, COMPOUND4args *, args); 5826 resp->status = NFS4ERR_MINOR_VERS_MISMATCH; 5827 DTRACE_NFSV4_2(compound__done, struct compound_state *, 5828 &cs, COMPOUND4res *, resp); 5829 return; 5830 } 5831 5832 if (args->array_len == 0) { 5833 resp->status = NFS4_OK; 5834 return; 5835 } 5836 5837 ASSERT(exi == NULL); 5838 ASSERT(cr == NULL); 5839 5840 cr = svc_xprt_cred(req->rq_xprt); 5841 ASSERT(cr != NULL); 5842 5843 if (sec_svc_getcred(req, cr, &cs.principal, &cs.nfsflavor) == 0) { 5844 DTRACE_NFSV4_2(compound__start, struct compound_state *, 5845 &cs, COMPOUND4args *, args); 5846 DTRACE_NFSV4_2(compound__done, struct compound_state *, 5847 &cs, COMPOUND4res *, resp); 5848 svcerr_badcred(req->rq_xprt); 5849 if (rv != NULL) 5850 *rv = 1; 5851 return; 5852 } 5853 resp->array_len = args->array_len; 5854 resp->array = kmem_zalloc(args->array_len * sizeof (nfs_resop4), 5855 KM_SLEEP); 5856 5857 cs.basecr = cr; 5858 nsrv4 = nfs4_get_srv(); 5859 5860 DTRACE_NFSV4_2(compound__start, struct compound_state *, &cs, 5861 COMPOUND4args *, args); 5862 5863 /* 5864 * For now, NFS4 compound processing must be protected by 5865 * exported_lock because it can access more than one exportinfo 5866 * per compound and share/unshare can now change multiple 5867 * exinfo structs. The NFS2/3 code only refs 1 exportinfo 5868 * per proc (excluding public exinfo), and exi_count design 5869 * is sufficient to protect concurrent execution of NFS2/3 5870 * ops along with unexport. This lock will be removed as 5871 * part of the NFSv4 phase 2 namespace redesign work. 5872 */ 5873 rw_enter(&ne->exported_lock, RW_READER); 5874 5875 /* 5876 * If this is the first compound we've seen, we need to start all 5877 * new instances' grace periods. 5878 */ 5879 if (nsrv4->seen_first_compound == 0) { 5880 rfs4_grace_start_new(nsrv4); 5881 /* 5882 * This must be set after rfs4_grace_start_new(), otherwise 5883 * another thread could proceed past here before the former 5884 * is finished. 5885 */ 5886 nsrv4->seen_first_compound = 1; 5887 } 5888 5889 for (i = 0; i < args->array_len && cs.cont; i++) { 5890 nfs_argop4 *argop; 5891 nfs_resop4 *resop; 5892 uint_t op; 5893 kstat_named_t *stat = ne->ne_globals->rfsproccnt[NFS_V4]; 5894 5895 argop = &args->array[i]; 5896 resop = &resp->array[i]; 5897 resop->resop = argop->argop; 5898 op = (uint_t)resop->resop; 5899 5900 if (op < rfsv4disp_cnt) { 5901 /* 5902 * Count the individual ops here; NULL and COMPOUND 5903 * are counted in common_dispatch() 5904 */ 5905 stat[op].value.ui64++; 5906 5907 NFS4_DEBUG(rfs4_debug > 1, 5908 (CE_NOTE, "Executing %s", rfs4_op_string[op])); 5909 (*rfsv4disptab[op].dis_proc)(argop, resop, req, &cs); 5910 NFS4_DEBUG(rfs4_debug > 1, (CE_NOTE, "%s returned %d", 5911 rfs4_op_string[op], *cs.statusp)); 5912 if (*cs.statusp != NFS4_OK) 5913 cs.cont = FALSE; 5914 } else { 5915 /* 5916 * This is effectively dead code since XDR code 5917 * will have already returned BADXDR if op doesn't 5918 * decode to legal value. This only done for a 5919 * day when XDR code doesn't verify v4 opcodes. 5920 */ 5921 op = OP_ILLEGAL; 5922 stat[OP_ILLEGAL_IDX].value.ui64++; 5923 5924 rfs4_op_illegal(argop, resop, req, &cs); 5925 cs.cont = FALSE; 5926 } 5927 5928 /* 5929 * If not at last op, and if we are to stop, then 5930 * compact the results array. 5931 */ 5932 if ((i + 1) < args->array_len && !cs.cont) { 5933 nfs_resop4 *new_res = kmem_alloc( 5934 (i+1) * sizeof (nfs_resop4), KM_SLEEP); 5935 bcopy(resp->array, 5936 new_res, (i+1) * sizeof (nfs_resop4)); 5937 kmem_free(resp->array, 5938 args->array_len * sizeof (nfs_resop4)); 5939 5940 resp->array_len = i + 1; 5941 resp->array = new_res; 5942 } 5943 } 5944 5945 rw_exit(&ne->exported_lock); 5946 5947 /* 5948 * clear exportinfo and vnode fields from compound_state before dtrace 5949 * probe, to avoid tracing residual values for path and share path. 5950 */ 5951 if (cs.vp) 5952 VN_RELE(cs.vp); 5953 if (cs.saved_vp) 5954 VN_RELE(cs.saved_vp); 5955 cs.exi = cs.saved_exi = NULL; 5956 cs.vp = cs.saved_vp = NULL; 5957 5958 DTRACE_NFSV4_2(compound__done, struct compound_state *, &cs, 5959 COMPOUND4res *, resp); 5960 5961 if (cs.saved_fh.nfs_fh4_val) 5962 kmem_free(cs.saved_fh.nfs_fh4_val, NFS4_FHSIZE); 5963 5964 if (cs.cr) 5965 crfree(cs.cr); 5966 /* 5967 * done with this compound request, free the label 5968 */ 5969 5970 if (req->rq_label != NULL) { 5971 kmem_free(req->rq_label, sizeof (bslabel_t)); 5972 req->rq_label = NULL; 5973 } 5974 } 5975 5976 /* 5977 * XXX because of what appears to be duplicate calls to rfs4_compound_free 5978 * XXX zero out the tag and array values. Need to investigate why the 5979 * XXX calls occur, but at least prevent the panic for now. 5980 */ 5981 void 5982 rfs4_compound_free(COMPOUND4res *resp) 5983 { 5984 uint_t i; 5985 5986 if (resp->tag.utf8string_val) { 5987 UTF8STRING_FREE(resp->tag) 5988 } 5989 5990 for (i = 0; i < resp->array_len; i++) { 5991 nfs_resop4 *resop; 5992 uint_t op; 5993 5994 resop = &resp->array[i]; 5995 op = (uint_t)resop->resop; 5996 if (op < rfsv4disp_cnt) { 5997 (*rfsv4disptab[op].dis_resfree)(resop); 5998 } 5999 } 6000 if (resp->array != NULL) { 6001 kmem_free(resp->array, resp->array_len * sizeof (nfs_resop4)); 6002 } 6003 } 6004 6005 /* 6006 * Process the value of the compound request rpc flags, as a bit-AND 6007 * of the individual per-op flags (idempotent, allowork, publicfh_ok) 6008 */ 6009 void 6010 rfs4_compound_flagproc(COMPOUND4args *args, int *flagp) 6011 { 6012 int i; 6013 int flag = RPC_ALL; 6014 6015 for (i = 0; flag && i < args->array_len; i++) { 6016 uint_t op; 6017 6018 op = (uint_t)args->array[i].argop; 6019 6020 if (op < rfsv4disp_cnt) 6021 flag &= rfsv4disptab[op].dis_flags; 6022 else 6023 flag = 0; 6024 } 6025 *flagp = flag; 6026 } 6027 6028 nfsstat4 6029 rfs4_client_sysid(rfs4_client_t *cp, sysid_t *sp) 6030 { 6031 nfsstat4 e; 6032 6033 rfs4_dbe_lock(cp->rc_dbe); 6034 6035 if (cp->rc_sysidt != LM_NOSYSID) { 6036 *sp = cp->rc_sysidt; 6037 e = NFS4_OK; 6038 6039 } else if ((cp->rc_sysidt = lm_alloc_sysidt()) != LM_NOSYSID) { 6040 *sp = cp->rc_sysidt; 6041 e = NFS4_OK; 6042 6043 NFS4_DEBUG(rfs4_debug, (CE_NOTE, 6044 "rfs4_client_sysid: allocated 0x%x\n", *sp)); 6045 } else 6046 e = NFS4ERR_DELAY; 6047 6048 rfs4_dbe_unlock(cp->rc_dbe); 6049 return (e); 6050 } 6051 6052 #if defined(DEBUG) && ! defined(lint) 6053 static void lock_print(char *str, int operation, struct flock64 *flk) 6054 { 6055 char *op, *type; 6056 6057 switch (operation) { 6058 case F_GETLK: op = "F_GETLK"; 6059 break; 6060 case F_SETLK: op = "F_SETLK"; 6061 break; 6062 case F_SETLK_NBMAND: op = "F_SETLK_NBMAND"; 6063 break; 6064 default: op = "F_UNKNOWN"; 6065 break; 6066 } 6067 switch (flk->l_type) { 6068 case F_UNLCK: type = "F_UNLCK"; 6069 break; 6070 case F_RDLCK: type = "F_RDLCK"; 6071 break; 6072 case F_WRLCK: type = "F_WRLCK"; 6073 break; 6074 default: type = "F_UNKNOWN"; 6075 break; 6076 } 6077 6078 ASSERT(flk->l_whence == 0); 6079 cmn_err(CE_NOTE, "%s: %s, type = %s, off = %llx len = %llx pid = %d", 6080 str, op, type, (longlong_t)flk->l_start, 6081 flk->l_len ? (longlong_t)flk->l_len : ~0LL, flk->l_pid); 6082 } 6083 6084 #define LOCK_PRINT(d, s, t, f) if (d) lock_print(s, t, f) 6085 #else 6086 #define LOCK_PRINT(d, s, t, f) 6087 #endif 6088 6089 /*ARGSUSED*/ 6090 static bool_t 6091 creds_ok(cred_set_t cr_set, struct svc_req *req, struct compound_state *cs) 6092 { 6093 return (TRUE); 6094 } 6095 6096 /* 6097 * Look up the pathname using the vp in cs as the directory vnode. 6098 * cs->vp will be the vnode for the file on success 6099 */ 6100 6101 static nfsstat4 6102 rfs4_lookup(component4 *component, struct svc_req *req, 6103 struct compound_state *cs) 6104 { 6105 char *nm; 6106 uint32_t len; 6107 nfsstat4 status; 6108 struct sockaddr *ca; 6109 char *name; 6110 6111 if (cs->vp == NULL) { 6112 return (NFS4ERR_NOFILEHANDLE); 6113 } 6114 if (cs->vp->v_type != VDIR) { 6115 return (NFS4ERR_NOTDIR); 6116 } 6117 6118 status = utf8_dir_verify(component); 6119 if (status != NFS4_OK) 6120 return (status); 6121 6122 nm = utf8_to_fn(component, &len, NULL); 6123 if (nm == NULL) { 6124 return (NFS4ERR_INVAL); 6125 } 6126 6127 if (len > MAXNAMELEN) { 6128 kmem_free(nm, len); 6129 return (NFS4ERR_NAMETOOLONG); 6130 } 6131 6132 ca = (struct sockaddr *)svc_getrpccaller(req->rq_xprt)->buf; 6133 name = nfscmd_convname(ca, cs->exi, nm, NFSCMD_CONV_INBOUND, 6134 MAXPATHLEN + 1); 6135 6136 if (name == NULL) { 6137 kmem_free(nm, len); 6138 return (NFS4ERR_INVAL); 6139 } 6140 6141 status = do_rfs4_op_lookup(name, req, cs); 6142 6143 if (name != nm) 6144 kmem_free(name, MAXPATHLEN + 1); 6145 6146 kmem_free(nm, len); 6147 6148 return (status); 6149 } 6150 6151 static nfsstat4 6152 rfs4_lookupfile(component4 *component, struct svc_req *req, 6153 struct compound_state *cs, uint32_t access, change_info4 *cinfo) 6154 { 6155 nfsstat4 status; 6156 vnode_t *dvp = cs->vp; 6157 vattr_t bva, ava, fva; 6158 int error; 6159 6160 /* Get "before" change value */ 6161 bva.va_mask = AT_CTIME|AT_SEQ; 6162 error = VOP_GETATTR(dvp, &bva, 0, cs->cr, NULL); 6163 if (error) 6164 return (puterrno4(error)); 6165 6166 /* rfs4_lookup may VN_RELE directory */ 6167 VN_HOLD(dvp); 6168 6169 status = rfs4_lookup(component, req, cs); 6170 if (status != NFS4_OK) { 6171 VN_RELE(dvp); 6172 return (status); 6173 } 6174 6175 /* 6176 * Get "after" change value, if it fails, simply return the 6177 * before value. 6178 */ 6179 ava.va_mask = AT_CTIME|AT_SEQ; 6180 if (VOP_GETATTR(dvp, &ava, 0, cs->cr, NULL)) { 6181 ava.va_ctime = bva.va_ctime; 6182 ava.va_seq = 0; 6183 } 6184 VN_RELE(dvp); 6185 6186 /* 6187 * Validate the file is a file 6188 */ 6189 fva.va_mask = AT_TYPE|AT_MODE; 6190 error = VOP_GETATTR(cs->vp, &fva, 0, cs->cr, NULL); 6191 if (error) 6192 return (puterrno4(error)); 6193 6194 if (fva.va_type != VREG) { 6195 if (fva.va_type == VDIR) 6196 return (NFS4ERR_ISDIR); 6197 if (fva.va_type == VLNK) 6198 return (NFS4ERR_SYMLINK); 6199 return (NFS4ERR_INVAL); 6200 } 6201 6202 NFS4_SET_FATTR4_CHANGE(cinfo->before, bva.va_ctime); 6203 NFS4_SET_FATTR4_CHANGE(cinfo->after, ava.va_ctime); 6204 6205 /* 6206 * It is undefined if VOP_LOOKUP will change va_seq, so 6207 * cinfo.atomic = TRUE only if we have 6208 * non-zero va_seq's, and they have not changed. 6209 */ 6210 if (bva.va_seq && ava.va_seq && ava.va_seq == bva.va_seq) 6211 cinfo->atomic = TRUE; 6212 else 6213 cinfo->atomic = FALSE; 6214 6215 /* Check for mandatory locking */ 6216 cs->mandlock = MANDLOCK(cs->vp, fva.va_mode); 6217 return (check_open_access(access, cs, req)); 6218 } 6219 6220 static nfsstat4 6221 create_vnode(vnode_t *dvp, char *nm, vattr_t *vap, createmode4 mode, 6222 cred_t *cr, vnode_t **vpp, bool_t *created) 6223 { 6224 int error; 6225 nfsstat4 status = NFS4_OK; 6226 vattr_t va; 6227 6228 tryagain: 6229 6230 /* 6231 * The file open mode used is VWRITE. If the client needs 6232 * some other semantic, then it should do the access checking 6233 * itself. It would have been nice to have the file open mode 6234 * passed as part of the arguments. 6235 */ 6236 6237 *created = TRUE; 6238 error = VOP_CREATE(dvp, nm, vap, EXCL, VWRITE, vpp, cr, 0, NULL, NULL); 6239 6240 if (error) { 6241 *created = FALSE; 6242 6243 /* 6244 * If we got something other than file already exists 6245 * then just return this error. Otherwise, we got 6246 * EEXIST. If we were doing a GUARDED create, then 6247 * just return this error. Otherwise, we need to 6248 * make sure that this wasn't a duplicate of an 6249 * exclusive create request. 6250 * 6251 * The assumption is made that a non-exclusive create 6252 * request will never return EEXIST. 6253 */ 6254 6255 if (error != EEXIST || mode == GUARDED4) { 6256 status = puterrno4(error); 6257 return (status); 6258 } 6259 error = VOP_LOOKUP(dvp, nm, vpp, NULL, 0, NULL, cr, 6260 NULL, NULL, NULL); 6261 6262 if (error) { 6263 /* 6264 * We couldn't find the file that we thought that 6265 * we just created. So, we'll just try creating 6266 * it again. 6267 */ 6268 if (error == ENOENT) 6269 goto tryagain; 6270 6271 status = puterrno4(error); 6272 return (status); 6273 } 6274 6275 if (mode == UNCHECKED4) { 6276 /* existing object must be regular file */ 6277 if ((*vpp)->v_type != VREG) { 6278 if ((*vpp)->v_type == VDIR) 6279 status = NFS4ERR_ISDIR; 6280 else if ((*vpp)->v_type == VLNK) 6281 status = NFS4ERR_SYMLINK; 6282 else 6283 status = NFS4ERR_INVAL; 6284 VN_RELE(*vpp); 6285 return (status); 6286 } 6287 6288 return (NFS4_OK); 6289 } 6290 6291 /* Check for duplicate request */ 6292 va.va_mask = AT_MTIME; 6293 error = VOP_GETATTR(*vpp, &va, 0, cr, NULL); 6294 if (!error) { 6295 /* We found the file */ 6296 const timestruc_t *mtime = &vap->va_mtime; 6297 6298 if (va.va_mtime.tv_sec != mtime->tv_sec || 6299 va.va_mtime.tv_nsec != mtime->tv_nsec) { 6300 /* but its not our creation */ 6301 VN_RELE(*vpp); 6302 return (NFS4ERR_EXIST); 6303 } 6304 *created = TRUE; /* retrans of create == created */ 6305 return (NFS4_OK); 6306 } 6307 VN_RELE(*vpp); 6308 return (NFS4ERR_EXIST); 6309 } 6310 6311 return (NFS4_OK); 6312 } 6313 6314 static nfsstat4 6315 check_open_access(uint32_t access, struct compound_state *cs, 6316 struct svc_req *req) 6317 { 6318 int error; 6319 vnode_t *vp; 6320 bool_t readonly; 6321 cred_t *cr = cs->cr; 6322 6323 /* For now we don't allow mandatory locking as per V2/V3 */ 6324 if (cs->access == CS_ACCESS_DENIED || cs->mandlock) { 6325 return (NFS4ERR_ACCESS); 6326 } 6327 6328 vp = cs->vp; 6329 ASSERT(cr != NULL && vp->v_type == VREG); 6330 6331 /* 6332 * If the file system is exported read only and we are trying 6333 * to open for write, then return NFS4ERR_ROFS 6334 */ 6335 6336 readonly = rdonly4(req, cs); 6337 6338 if ((access & OPEN4_SHARE_ACCESS_WRITE) && readonly) 6339 return (NFS4ERR_ROFS); 6340 6341 if (access & OPEN4_SHARE_ACCESS_READ) { 6342 if ((VOP_ACCESS(vp, VREAD, 0, cr, NULL) != 0) && 6343 (VOP_ACCESS(vp, VEXEC, 0, cr, NULL) != 0)) { 6344 return (NFS4ERR_ACCESS); 6345 } 6346 } 6347 6348 if (access & OPEN4_SHARE_ACCESS_WRITE) { 6349 error = VOP_ACCESS(vp, VWRITE, 0, cr, NULL); 6350 if (error) 6351 return (NFS4ERR_ACCESS); 6352 } 6353 6354 return (NFS4_OK); 6355 } 6356 6357 static nfsstat4 6358 rfs4_createfile(OPEN4args *args, struct svc_req *req, struct compound_state *cs, 6359 change_info4 *cinfo, bitmap4 *attrset, clientid4 clientid) 6360 { 6361 struct nfs4_svgetit_arg sarg; 6362 struct nfs4_ntov_table ntov; 6363 6364 bool_t ntov_table_init = FALSE; 6365 struct statvfs64 sb; 6366 nfsstat4 status; 6367 vnode_t *vp; 6368 vattr_t bva, ava, iva, cva, *vap; 6369 vnode_t *dvp; 6370 timespec32_t *mtime; 6371 char *nm = NULL; 6372 uint_t buflen; 6373 bool_t created; 6374 bool_t setsize = FALSE; 6375 len_t reqsize; 6376 int error; 6377 bool_t trunc; 6378 caller_context_t ct; 6379 component4 *component; 6380 bslabel_t *clabel; 6381 struct sockaddr *ca; 6382 char *name = NULL; 6383 6384 sarg.sbp = &sb; 6385 sarg.is_referral = B_FALSE; 6386 6387 dvp = cs->vp; 6388 6389 /* Check if the file system is read only */ 6390 if (rdonly4(req, cs)) 6391 return (NFS4ERR_ROFS); 6392 6393 /* check the label of including directory */ 6394 if (is_system_labeled()) { 6395 ASSERT(req->rq_label != NULL); 6396 clabel = req->rq_label; 6397 DTRACE_PROBE2(tx__rfs4__log__info__opremove__clabel, char *, 6398 "got client label from request(1)", 6399 struct svc_req *, req); 6400 if (!blequal(&l_admin_low->tsl_label, clabel)) { 6401 if (!do_rfs_label_check(clabel, dvp, EQUALITY_CHECK, 6402 cs->exi)) { 6403 return (NFS4ERR_ACCESS); 6404 } 6405 } 6406 } 6407 6408 /* 6409 * Get the last component of path name in nm. cs will reference 6410 * the including directory on success. 6411 */ 6412 component = &args->open_claim4_u.file; 6413 status = utf8_dir_verify(component); 6414 if (status != NFS4_OK) 6415 return (status); 6416 6417 nm = utf8_to_fn(component, &buflen, NULL); 6418 6419 if (nm == NULL) 6420 return (NFS4ERR_RESOURCE); 6421 6422 if (buflen > MAXNAMELEN) { 6423 kmem_free(nm, buflen); 6424 return (NFS4ERR_NAMETOOLONG); 6425 } 6426 6427 bva.va_mask = AT_TYPE|AT_CTIME|AT_SEQ; 6428 error = VOP_GETATTR(dvp, &bva, 0, cs->cr, NULL); 6429 if (error) { 6430 kmem_free(nm, buflen); 6431 return (puterrno4(error)); 6432 } 6433 6434 if (bva.va_type != VDIR) { 6435 kmem_free(nm, buflen); 6436 return (NFS4ERR_NOTDIR); 6437 } 6438 6439 NFS4_SET_FATTR4_CHANGE(cinfo->before, bva.va_ctime) 6440 6441 switch (args->mode) { 6442 case GUARDED4: 6443 /*FALLTHROUGH*/ 6444 case UNCHECKED4: 6445 nfs4_ntov_table_init(&ntov); 6446 ntov_table_init = TRUE; 6447 6448 *attrset = 0; 6449 status = do_rfs4_set_attrs(attrset, 6450 &args->createhow4_u.createattrs, 6451 cs, &sarg, &ntov, NFS4ATTR_SETIT); 6452 6453 if (status == NFS4_OK && (sarg.vap->va_mask & AT_TYPE) && 6454 sarg.vap->va_type != VREG) { 6455 if (sarg.vap->va_type == VDIR) 6456 status = NFS4ERR_ISDIR; 6457 else if (sarg.vap->va_type == VLNK) 6458 status = NFS4ERR_SYMLINK; 6459 else 6460 status = NFS4ERR_INVAL; 6461 } 6462 6463 if (status != NFS4_OK) { 6464 kmem_free(nm, buflen); 6465 nfs4_ntov_table_free(&ntov, &sarg); 6466 *attrset = 0; 6467 return (status); 6468 } 6469 6470 vap = sarg.vap; 6471 vap->va_type = VREG; 6472 vap->va_mask |= AT_TYPE; 6473 6474 if ((vap->va_mask & AT_MODE) == 0) { 6475 vap->va_mask |= AT_MODE; 6476 vap->va_mode = (mode_t)0600; 6477 } 6478 6479 if (vap->va_mask & AT_SIZE) { 6480 6481 /* Disallow create with a non-zero size */ 6482 6483 if ((reqsize = sarg.vap->va_size) != 0) { 6484 kmem_free(nm, buflen); 6485 nfs4_ntov_table_free(&ntov, &sarg); 6486 *attrset = 0; 6487 return (NFS4ERR_INVAL); 6488 } 6489 setsize = TRUE; 6490 } 6491 break; 6492 6493 case EXCLUSIVE4: 6494 /* prohibit EXCL create of named attributes */ 6495 if (dvp->v_flag & V_XATTRDIR) { 6496 kmem_free(nm, buflen); 6497 *attrset = 0; 6498 return (NFS4ERR_INVAL); 6499 } 6500 6501 cva.va_mask = AT_TYPE | AT_MTIME | AT_MODE; 6502 cva.va_type = VREG; 6503 /* 6504 * Ensure no time overflows. Assumes underlying 6505 * filesystem supports at least 32 bits. 6506 * Truncate nsec to usec resolution to allow valid 6507 * compares even if the underlying filesystem truncates. 6508 */ 6509 mtime = (timespec32_t *)&args->createhow4_u.createverf; 6510 cva.va_mtime.tv_sec = mtime->tv_sec % TIME32_MAX; 6511 cva.va_mtime.tv_nsec = (mtime->tv_nsec / 1000) * 1000; 6512 cva.va_mode = (mode_t)0; 6513 vap = &cva; 6514 6515 /* 6516 * For EXCL create, attrset is set to the server attr 6517 * used to cache the client's verifier. 6518 */ 6519 *attrset = FATTR4_TIME_MODIFY_MASK; 6520 break; 6521 } 6522 6523 ca = (struct sockaddr *)svc_getrpccaller(req->rq_xprt)->buf; 6524 name = nfscmd_convname(ca, cs->exi, nm, NFSCMD_CONV_INBOUND, 6525 MAXPATHLEN + 1); 6526 6527 if (name == NULL) { 6528 kmem_free(nm, buflen); 6529 return (NFS4ERR_SERVERFAULT); 6530 } 6531 6532 status = create_vnode(dvp, name, vap, args->mode, 6533 cs->cr, &vp, &created); 6534 if (nm != name) 6535 kmem_free(name, MAXPATHLEN + 1); 6536 kmem_free(nm, buflen); 6537 6538 if (status != NFS4_OK) { 6539 if (ntov_table_init) 6540 nfs4_ntov_table_free(&ntov, &sarg); 6541 *attrset = 0; 6542 return (status); 6543 } 6544 6545 trunc = (setsize && !created); 6546 6547 if (args->mode != EXCLUSIVE4) { 6548 bitmap4 createmask = args->createhow4_u.createattrs.attrmask; 6549 6550 /* 6551 * True verification that object was created with correct 6552 * attrs is impossible. The attrs could have been changed 6553 * immediately after object creation. If attributes did 6554 * not verify, the only recourse for the server is to 6555 * destroy the object. Maybe if some attrs (like gid) 6556 * are set incorrectly, the object should be destroyed; 6557 * however, seems bad as a default policy. Do we really 6558 * want to destroy an object over one of the times not 6559 * verifying correctly? For these reasons, the server 6560 * currently sets bits in attrset for createattrs 6561 * that were set; however, no verification is done. 6562 * 6563 * vmask_to_nmask accounts for vattr bits set on create 6564 * [do_rfs4_set_attrs() only sets resp bits for 6565 * non-vattr/vfs bits.] 6566 * Mask off any bits we set by default so as not to return 6567 * more attrset bits than were requested in createattrs 6568 */ 6569 if (created) { 6570 nfs4_vmask_to_nmask(sarg.vap->va_mask, attrset); 6571 *attrset &= createmask; 6572 } else { 6573 /* 6574 * We did not create the vnode (we tried but it 6575 * already existed). In this case, the only createattr 6576 * that the spec allows the server to set is size, 6577 * and even then, it can only be set if it is 0. 6578 */ 6579 *attrset = 0; 6580 if (trunc) 6581 *attrset = FATTR4_SIZE_MASK; 6582 } 6583 } 6584 if (ntov_table_init) 6585 nfs4_ntov_table_free(&ntov, &sarg); 6586 6587 /* 6588 * Get the initial "after" sequence number, if it fails, 6589 * set to zero, time to before. 6590 */ 6591 iva.va_mask = AT_CTIME|AT_SEQ; 6592 if (VOP_GETATTR(dvp, &iva, 0, cs->cr, NULL)) { 6593 iva.va_seq = 0; 6594 iva.va_ctime = bva.va_ctime; 6595 } 6596 6597 /* 6598 * create_vnode attempts to create the file exclusive, 6599 * if it already exists the VOP_CREATE will fail and 6600 * may not increase va_seq. It is atomic if 6601 * we haven't changed the directory, but if it has changed 6602 * we don't know what changed it. 6603 */ 6604 if (!created) { 6605 if (bva.va_seq && iva.va_seq && 6606 bva.va_seq == iva.va_seq) 6607 cinfo->atomic = TRUE; 6608 else 6609 cinfo->atomic = FALSE; 6610 NFS4_SET_FATTR4_CHANGE(cinfo->after, iva.va_ctime); 6611 } else { 6612 /* 6613 * The entry was created, we need to sync the 6614 * directory metadata. 6615 */ 6616 (void) VOP_FSYNC(dvp, 0, cs->cr, NULL); 6617 6618 /* 6619 * Get "after" change value, if it fails, simply return the 6620 * before value. 6621 */ 6622 ava.va_mask = AT_CTIME|AT_SEQ; 6623 if (VOP_GETATTR(dvp, &ava, 0, cs->cr, NULL)) { 6624 ava.va_ctime = bva.va_ctime; 6625 ava.va_seq = 0; 6626 } 6627 6628 NFS4_SET_FATTR4_CHANGE(cinfo->after, ava.va_ctime); 6629 6630 /* 6631 * The cinfo->atomic = TRUE only if we have 6632 * non-zero va_seq's, and it has incremented by exactly one 6633 * during the create_vnode and it didn't 6634 * change during the VOP_FSYNC. 6635 */ 6636 if (bva.va_seq && iva.va_seq && ava.va_seq && 6637 iva.va_seq == (bva.va_seq + 1) && iva.va_seq == ava.va_seq) 6638 cinfo->atomic = TRUE; 6639 else 6640 cinfo->atomic = FALSE; 6641 } 6642 6643 /* Check for mandatory locking and that the size gets set. */ 6644 cva.va_mask = AT_MODE; 6645 if (setsize) 6646 cva.va_mask |= AT_SIZE; 6647 6648 /* Assume the worst */ 6649 cs->mandlock = TRUE; 6650 6651 if (VOP_GETATTR(vp, &cva, 0, cs->cr, NULL) == 0) { 6652 cs->mandlock = MANDLOCK(cs->vp, cva.va_mode); 6653 6654 /* 6655 * Truncate the file if necessary; this would be 6656 * the case for create over an existing file. 6657 */ 6658 6659 if (trunc) { 6660 int in_crit = 0; 6661 rfs4_file_t *fp; 6662 nfs4_srv_t *nsrv4; 6663 bool_t create = FALSE; 6664 6665 /* 6666 * We are writing over an existing file. 6667 * Check to see if we need to recall a delegation. 6668 */ 6669 nsrv4 = nfs4_get_srv(); 6670 rfs4_hold_deleg_policy(nsrv4); 6671 if ((fp = rfs4_findfile(vp, NULL, &create)) != NULL) { 6672 if (rfs4_check_delegated_byfp(FWRITE, fp, 6673 (reqsize == 0), FALSE, FALSE, &clientid)) { 6674 rfs4_file_rele(fp); 6675 rfs4_rele_deleg_policy(nsrv4); 6676 VN_RELE(vp); 6677 *attrset = 0; 6678 return (NFS4ERR_DELAY); 6679 } 6680 rfs4_file_rele(fp); 6681 } 6682 rfs4_rele_deleg_policy(nsrv4); 6683 6684 if (nbl_need_check(vp)) { 6685 in_crit = 1; 6686 6687 ASSERT(reqsize == 0); 6688 6689 nbl_start_crit(vp, RW_READER); 6690 if (nbl_conflict(vp, NBL_WRITE, 0, 6691 cva.va_size, 0, NULL)) { 6692 in_crit = 0; 6693 nbl_end_crit(vp); 6694 VN_RELE(vp); 6695 *attrset = 0; 6696 return (NFS4ERR_ACCESS); 6697 } 6698 } 6699 ct.cc_sysid = 0; 6700 ct.cc_pid = 0; 6701 ct.cc_caller_id = nfs4_srv_caller_id; 6702 ct.cc_flags = CC_DONTBLOCK; 6703 6704 cva.va_mask = AT_SIZE; 6705 cva.va_size = reqsize; 6706 (void) VOP_SETATTR(vp, &cva, 0, cs->cr, &ct); 6707 if (in_crit) 6708 nbl_end_crit(vp); 6709 } 6710 } 6711 6712 error = makefh4(&cs->fh, vp, cs->exi); 6713 6714 /* 6715 * Force modified data and metadata out to stable storage. 6716 */ 6717 (void) VOP_FSYNC(vp, FNODSYNC, cs->cr, NULL); 6718 6719 if (error) { 6720 VN_RELE(vp); 6721 *attrset = 0; 6722 return (puterrno4(error)); 6723 } 6724 6725 /* if parent dir is attrdir, set namedattr fh flag */ 6726 if (dvp->v_flag & V_XATTRDIR) 6727 set_fh4_flag(&cs->fh, FH4_NAMEDATTR); 6728 6729 if (cs->vp) 6730 VN_RELE(cs->vp); 6731 6732 cs->vp = vp; 6733 6734 /* 6735 * if we did not create the file, we will need to check 6736 * the access bits on the file 6737 */ 6738 6739 if (!created) { 6740 if (setsize) 6741 args->share_access |= OPEN4_SHARE_ACCESS_WRITE; 6742 status = check_open_access(args->share_access, cs, req); 6743 if (status != NFS4_OK) 6744 *attrset = 0; 6745 } 6746 return (status); 6747 } 6748 6749 /*ARGSUSED*/ 6750 static void 6751 rfs4_do_open(struct compound_state *cs, struct svc_req *req, 6752 rfs4_openowner_t *oo, delegreq_t deleg, 6753 uint32_t access, uint32_t deny, 6754 OPEN4res *resp, int deleg_cur) 6755 { 6756 /* XXX Currently not using req */ 6757 rfs4_state_t *sp; 6758 rfs4_file_t *fp; 6759 bool_t screate = TRUE; 6760 bool_t fcreate = TRUE; 6761 uint32_t open_a, share_a; 6762 uint32_t open_d, share_d; 6763 rfs4_deleg_state_t *dsp; 6764 sysid_t sysid; 6765 nfsstat4 status; 6766 caller_context_t ct; 6767 int fflags = 0; 6768 int recall = 0; 6769 int err; 6770 int first_open; 6771 6772 /* get the file struct and hold a lock on it during initial open */ 6773 fp = rfs4_findfile_withlock(cs->vp, &cs->fh, &fcreate); 6774 if (fp == NULL) { 6775 resp->status = NFS4ERR_RESOURCE; 6776 DTRACE_PROBE1(nfss__e__do__open1, nfsstat4, resp->status); 6777 return; 6778 } 6779 6780 sp = rfs4_findstate_by_owner_file(oo, fp, &screate); 6781 if (sp == NULL) { 6782 resp->status = NFS4ERR_RESOURCE; 6783 DTRACE_PROBE1(nfss__e__do__open2, nfsstat4, resp->status); 6784 /* No need to keep any reference */ 6785 rw_exit(&fp->rf_file_rwlock); 6786 rfs4_file_rele(fp); 6787 return; 6788 } 6789 6790 /* try to get the sysid before continuing */ 6791 if ((status = rfs4_client_sysid(oo->ro_client, &sysid)) != NFS4_OK) { 6792 resp->status = status; 6793 rfs4_file_rele(fp); 6794 /* Not a fully formed open; "close" it */ 6795 if (screate == TRUE) 6796 rfs4_state_close(sp, FALSE, FALSE, cs->cr); 6797 rfs4_state_rele(sp); 6798 return; 6799 } 6800 6801 /* Calculate the fflags for this OPEN. */ 6802 if (access & OPEN4_SHARE_ACCESS_READ) 6803 fflags |= FREAD; 6804 if (access & OPEN4_SHARE_ACCESS_WRITE) 6805 fflags |= FWRITE; 6806 6807 rfs4_dbe_lock(sp->rs_dbe); 6808 6809 /* 6810 * Calculate the new deny and access mode that this open is adding to 6811 * the file for this open owner; 6812 */ 6813 open_d = (deny & ~sp->rs_open_deny); 6814 open_a = (access & ~sp->rs_open_access); 6815 6816 /* 6817 * Calculate the new share access and share deny modes that this open 6818 * is adding to the file for this open owner; 6819 */ 6820 share_a = (access & ~sp->rs_share_access); 6821 share_d = (deny & ~sp->rs_share_deny); 6822 6823 first_open = (sp->rs_open_access & OPEN4_SHARE_ACCESS_BOTH) == 0; 6824 6825 /* 6826 * Check to see the client has already sent an open for this 6827 * open owner on this file with the same share/deny modes. 6828 * If so, we don't need to check for a conflict and we don't 6829 * need to add another shrlock. If not, then we need to 6830 * check for conflicts in deny and access before checking for 6831 * conflicts in delegation. We don't want to recall a 6832 * delegation based on an open that will eventually fail based 6833 * on shares modes. 6834 */ 6835 6836 if (share_a || share_d) { 6837 if ((err = rfs4_share(sp, access, deny)) != 0) { 6838 rfs4_dbe_unlock(sp->rs_dbe); 6839 resp->status = err; 6840 6841 rfs4_file_rele(fp); 6842 /* Not a fully formed open; "close" it */ 6843 if (screate == TRUE) 6844 rfs4_state_close(sp, FALSE, FALSE, cs->cr); 6845 rfs4_state_rele(sp); 6846 return; 6847 } 6848 } 6849 6850 rfs4_dbe_lock(fp->rf_dbe); 6851 6852 /* 6853 * Check to see if this file is delegated and if so, if a 6854 * recall needs to be done. 6855 */ 6856 if (rfs4_check_recall(sp, access)) { 6857 rfs4_dbe_unlock(fp->rf_dbe); 6858 rfs4_dbe_unlock(sp->rs_dbe); 6859 rfs4_recall_deleg(fp, FALSE, sp->rs_owner->ro_client); 6860 delay(NFS4_DELEGATION_CONFLICT_DELAY); 6861 rfs4_dbe_lock(sp->rs_dbe); 6862 6863 /* if state closed while lock was dropped */ 6864 if (sp->rs_closed) { 6865 if (share_a || share_d) 6866 (void) rfs4_unshare(sp); 6867 rfs4_dbe_unlock(sp->rs_dbe); 6868 rfs4_file_rele(fp); 6869 /* Not a fully formed open; "close" it */ 6870 if (screate == TRUE) 6871 rfs4_state_close(sp, FALSE, FALSE, cs->cr); 6872 rfs4_state_rele(sp); 6873 resp->status = NFS4ERR_OLD_STATEID; 6874 return; 6875 } 6876 6877 rfs4_dbe_lock(fp->rf_dbe); 6878 /* Let's see if the delegation was returned */ 6879 if (rfs4_check_recall(sp, access)) { 6880 rfs4_dbe_unlock(fp->rf_dbe); 6881 if (share_a || share_d) 6882 (void) rfs4_unshare(sp); 6883 rfs4_dbe_unlock(sp->rs_dbe); 6884 rfs4_file_rele(fp); 6885 rfs4_update_lease(sp->rs_owner->ro_client); 6886 6887 /* Not a fully formed open; "close" it */ 6888 if (screate == TRUE) 6889 rfs4_state_close(sp, FALSE, FALSE, cs->cr); 6890 rfs4_state_rele(sp); 6891 resp->status = NFS4ERR_DELAY; 6892 return; 6893 } 6894 } 6895 /* 6896 * the share check passed and any delegation conflict has been 6897 * taken care of, now call vop_open. 6898 * if this is the first open then call vop_open with fflags. 6899 * if not, call vn_open_upgrade with just the upgrade flags. 6900 * 6901 * if the file has been opened already, it will have the current 6902 * access mode in the state struct. if it has no share access, then 6903 * this is a new open. 6904 * 6905 * However, if this is open with CLAIM_DLEGATE_CUR, then don't 6906 * call VOP_OPEN(), just do the open upgrade. 6907 */ 6908 if (first_open && !deleg_cur) { 6909 ct.cc_sysid = sysid; 6910 ct.cc_pid = rfs4_dbe_getid(sp->rs_owner->ro_dbe); 6911 ct.cc_caller_id = nfs4_srv_caller_id; 6912 ct.cc_flags = CC_DONTBLOCK; 6913 err = VOP_OPEN(&cs->vp, fflags, cs->cr, &ct); 6914 if (err) { 6915 rfs4_dbe_unlock(fp->rf_dbe); 6916 if (share_a || share_d) 6917 (void) rfs4_unshare(sp); 6918 rfs4_dbe_unlock(sp->rs_dbe); 6919 rfs4_file_rele(fp); 6920 6921 /* Not a fully formed open; "close" it */ 6922 if (screate == TRUE) 6923 rfs4_state_close(sp, FALSE, FALSE, cs->cr); 6924 rfs4_state_rele(sp); 6925 /* check if a monitor detected a delegation conflict */ 6926 if (err == EAGAIN && (ct.cc_flags & CC_WOULDBLOCK)) 6927 resp->status = NFS4ERR_DELAY; 6928 else 6929 resp->status = NFS4ERR_SERVERFAULT; 6930 return; 6931 } 6932 } else { /* open upgrade */ 6933 /* 6934 * calculate the fflags for the new mode that is being added 6935 * by this upgrade. 6936 */ 6937 fflags = 0; 6938 if (open_a & OPEN4_SHARE_ACCESS_READ) 6939 fflags |= FREAD; 6940 if (open_a & OPEN4_SHARE_ACCESS_WRITE) 6941 fflags |= FWRITE; 6942 vn_open_upgrade(cs->vp, fflags); 6943 } 6944 sp->rs_open_access |= access; 6945 sp->rs_open_deny |= deny; 6946 6947 if (open_d & OPEN4_SHARE_DENY_READ) 6948 fp->rf_deny_read++; 6949 if (open_d & OPEN4_SHARE_DENY_WRITE) 6950 fp->rf_deny_write++; 6951 fp->rf_share_deny |= deny; 6952 6953 if (open_a & OPEN4_SHARE_ACCESS_READ) 6954 fp->rf_access_read++; 6955 if (open_a & OPEN4_SHARE_ACCESS_WRITE) 6956 fp->rf_access_write++; 6957 fp->rf_share_access |= access; 6958 6959 /* 6960 * Check for delegation here. if the deleg argument is not 6961 * DELEG_ANY, then this is a reclaim from a client and 6962 * we must honor the delegation requested. If necessary we can 6963 * set the recall flag. 6964 */ 6965 6966 dsp = rfs4_grant_delegation(deleg, sp, &recall); 6967 6968 cs->deleg = (fp->rf_dinfo.rd_dtype == OPEN_DELEGATE_WRITE); 6969 6970 next_stateid(&sp->rs_stateid); 6971 6972 resp->stateid = sp->rs_stateid.stateid; 6973 6974 rfs4_dbe_unlock(fp->rf_dbe); 6975 rfs4_dbe_unlock(sp->rs_dbe); 6976 6977 if (dsp) { 6978 rfs4_set_deleg_response(dsp, &resp->delegation, NULL, recall); 6979 rfs4_deleg_state_rele(dsp); 6980 } 6981 6982 rfs4_file_rele(fp); 6983 rfs4_state_rele(sp); 6984 6985 resp->status = NFS4_OK; 6986 } 6987 6988 /*ARGSUSED*/ 6989 static void 6990 rfs4_do_opennull(struct compound_state *cs, struct svc_req *req, 6991 OPEN4args *args, rfs4_openowner_t *oo, OPEN4res *resp) 6992 { 6993 change_info4 *cinfo = &resp->cinfo; 6994 bitmap4 *attrset = &resp->attrset; 6995 6996 if (args->opentype == OPEN4_NOCREATE) 6997 resp->status = rfs4_lookupfile(&args->open_claim4_u.file, 6998 req, cs, args->share_access, cinfo); 6999 else { 7000 /* inhibit delegation grants during exclusive create */ 7001 7002 if (args->mode == EXCLUSIVE4) 7003 rfs4_disable_delegation(); 7004 7005 resp->status = rfs4_createfile(args, req, cs, cinfo, attrset, 7006 oo->ro_client->rc_clientid); 7007 } 7008 7009 if (resp->status == NFS4_OK) { 7010 7011 /* cs->vp cs->fh now reference the desired file */ 7012 7013 rfs4_do_open(cs, req, oo, 7014 oo->ro_need_confirm ? DELEG_NONE : DELEG_ANY, 7015 args->share_access, args->share_deny, resp, 0); 7016 7017 /* 7018 * If rfs4_createfile set attrset, we must 7019 * clear this attrset before the response is copied. 7020 */ 7021 if (resp->status != NFS4_OK && resp->attrset) { 7022 resp->attrset = 0; 7023 } 7024 } 7025 else 7026 *cs->statusp = resp->status; 7027 7028 if (args->mode == EXCLUSIVE4) 7029 rfs4_enable_delegation(); 7030 } 7031 7032 /*ARGSUSED*/ 7033 static void 7034 rfs4_do_openprev(struct compound_state *cs, struct svc_req *req, 7035 OPEN4args *args, rfs4_openowner_t *oo, OPEN4res *resp) 7036 { 7037 change_info4 *cinfo = &resp->cinfo; 7038 vattr_t va; 7039 vtype_t v_type = cs->vp->v_type; 7040 int error = 0; 7041 7042 /* Verify that we have a regular file */ 7043 if (v_type != VREG) { 7044 if (v_type == VDIR) 7045 resp->status = NFS4ERR_ISDIR; 7046 else if (v_type == VLNK) 7047 resp->status = NFS4ERR_SYMLINK; 7048 else 7049 resp->status = NFS4ERR_INVAL; 7050 return; 7051 } 7052 7053 va.va_mask = AT_MODE|AT_UID; 7054 error = VOP_GETATTR(cs->vp, &va, 0, cs->cr, NULL); 7055 if (error) { 7056 resp->status = puterrno4(error); 7057 return; 7058 } 7059 7060 cs->mandlock = MANDLOCK(cs->vp, va.va_mode); 7061 7062 /* 7063 * Check if we have access to the file, Note the the file 7064 * could have originally been open UNCHECKED or GUARDED 7065 * with mode bits that will now fail, but there is nothing 7066 * we can really do about that except in the case that the 7067 * owner of the file is the one requesting the open. 7068 */ 7069 if (crgetuid(cs->cr) != va.va_uid) { 7070 resp->status = check_open_access(args->share_access, cs, req); 7071 if (resp->status != NFS4_OK) { 7072 return; 7073 } 7074 } 7075 7076 /* 7077 * cinfo on a CLAIM_PREVIOUS is undefined, initialize to zero 7078 */ 7079 cinfo->before = 0; 7080 cinfo->after = 0; 7081 cinfo->atomic = FALSE; 7082 7083 rfs4_do_open(cs, req, oo, 7084 NFS4_DELEG4TYPE2REQTYPE(args->open_claim4_u.delegate_type), 7085 args->share_access, args->share_deny, resp, 0); 7086 } 7087 7088 static void 7089 rfs4_do_opendelcur(struct compound_state *cs, struct svc_req *req, 7090 OPEN4args *args, rfs4_openowner_t *oo, OPEN4res *resp) 7091 { 7092 int error; 7093 nfsstat4 status; 7094 stateid4 stateid = 7095 args->open_claim4_u.delegate_cur_info.delegate_stateid; 7096 rfs4_deleg_state_t *dsp; 7097 7098 /* 7099 * Find the state info from the stateid and confirm that the 7100 * file is delegated. If the state openowner is the same as 7101 * the supplied openowner we're done. If not, get the file 7102 * info from the found state info. Use that file info to 7103 * create the state for this lock owner. Note solaris doen't 7104 * really need the pathname to find the file. We may want to 7105 * lookup the pathname and make sure that the vp exist and 7106 * matches the vp in the file structure. However it is 7107 * possible that the pathname nolonger exists (local process 7108 * unlinks the file), so this may not be that useful. 7109 */ 7110 7111 status = rfs4_get_deleg_state(&stateid, &dsp); 7112 if (status != NFS4_OK) { 7113 resp->status = status; 7114 return; 7115 } 7116 7117 ASSERT(dsp->rds_finfo->rf_dinfo.rd_dtype != OPEN_DELEGATE_NONE); 7118 7119 /* 7120 * New lock owner, create state. Since this was probably called 7121 * in response to a CB_RECALL we set deleg to DELEG_NONE 7122 */ 7123 7124 ASSERT(cs->vp != NULL); 7125 VN_RELE(cs->vp); 7126 VN_HOLD(dsp->rds_finfo->rf_vp); 7127 cs->vp = dsp->rds_finfo->rf_vp; 7128 7129 if (error = makefh4(&cs->fh, cs->vp, cs->exi)) { 7130 rfs4_deleg_state_rele(dsp); 7131 *cs->statusp = resp->status = puterrno4(error); 7132 return; 7133 } 7134 7135 /* Mark progress for delegation returns */ 7136 dsp->rds_finfo->rf_dinfo.rd_time_lastwrite = gethrestime_sec(); 7137 rfs4_deleg_state_rele(dsp); 7138 rfs4_do_open(cs, req, oo, DELEG_NONE, 7139 args->share_access, args->share_deny, resp, 1); 7140 } 7141 7142 /*ARGSUSED*/ 7143 static void 7144 rfs4_do_opendelprev(struct compound_state *cs, struct svc_req *req, 7145 OPEN4args *args, rfs4_openowner_t *oo, OPEN4res *resp) 7146 { 7147 /* 7148 * Lookup the pathname, it must already exist since this file 7149 * was delegated. 7150 * 7151 * Find the file and state info for this vp and open owner pair. 7152 * check that they are in fact delegated. 7153 * check that the state access and deny modes are the same. 7154 * 7155 * Return the delgation possibly seting the recall flag. 7156 */ 7157 rfs4_file_t *fp; 7158 rfs4_state_t *sp; 7159 bool_t create = FALSE; 7160 bool_t dcreate = FALSE; 7161 rfs4_deleg_state_t *dsp; 7162 nfsace4 *ace; 7163 7164 /* Note we ignore oflags */ 7165 resp->status = rfs4_lookupfile(&args->open_claim4_u.file_delegate_prev, 7166 req, cs, args->share_access, &resp->cinfo); 7167 7168 if (resp->status != NFS4_OK) { 7169 return; 7170 } 7171 7172 /* get the file struct and hold a lock on it during initial open */ 7173 fp = rfs4_findfile_withlock(cs->vp, NULL, &create); 7174 if (fp == NULL) { 7175 resp->status = NFS4ERR_RESOURCE; 7176 DTRACE_PROBE1(nfss__e__do_opendelprev1, nfsstat4, resp->status); 7177 return; 7178 } 7179 7180 sp = rfs4_findstate_by_owner_file(oo, fp, &create); 7181 if (sp == NULL) { 7182 resp->status = NFS4ERR_SERVERFAULT; 7183 DTRACE_PROBE1(nfss__e__do_opendelprev2, nfsstat4, resp->status); 7184 rw_exit(&fp->rf_file_rwlock); 7185 rfs4_file_rele(fp); 7186 return; 7187 } 7188 7189 rfs4_dbe_lock(sp->rs_dbe); 7190 rfs4_dbe_lock(fp->rf_dbe); 7191 if (args->share_access != sp->rs_share_access || 7192 args->share_deny != sp->rs_share_deny || 7193 sp->rs_finfo->rf_dinfo.rd_dtype == OPEN_DELEGATE_NONE) { 7194 NFS4_DEBUG(rfs4_debug, 7195 (CE_NOTE, "rfs4_do_opendelprev: state mixup")); 7196 rfs4_dbe_unlock(fp->rf_dbe); 7197 rfs4_dbe_unlock(sp->rs_dbe); 7198 rfs4_file_rele(fp); 7199 rfs4_state_rele(sp); 7200 resp->status = NFS4ERR_SERVERFAULT; 7201 return; 7202 } 7203 rfs4_dbe_unlock(fp->rf_dbe); 7204 rfs4_dbe_unlock(sp->rs_dbe); 7205 7206 dsp = rfs4_finddeleg(sp, &dcreate); 7207 if (dsp == NULL) { 7208 rfs4_state_rele(sp); 7209 rfs4_file_rele(fp); 7210 resp->status = NFS4ERR_SERVERFAULT; 7211 return; 7212 } 7213 7214 next_stateid(&sp->rs_stateid); 7215 7216 resp->stateid = sp->rs_stateid.stateid; 7217 7218 resp->delegation.delegation_type = dsp->rds_dtype; 7219 7220 if (dsp->rds_dtype == OPEN_DELEGATE_READ) { 7221 open_read_delegation4 *rv = 7222 &resp->delegation.open_delegation4_u.read; 7223 7224 rv->stateid = dsp->rds_delegid.stateid; 7225 rv->recall = FALSE; /* no policy in place to set to TRUE */ 7226 ace = &rv->permissions; 7227 } else { 7228 open_write_delegation4 *rv = 7229 &resp->delegation.open_delegation4_u.write; 7230 7231 rv->stateid = dsp->rds_delegid.stateid; 7232 rv->recall = FALSE; /* no policy in place to set to TRUE */ 7233 ace = &rv->permissions; 7234 rv->space_limit.limitby = NFS_LIMIT_SIZE; 7235 rv->space_limit.nfs_space_limit4_u.filesize = UINT64_MAX; 7236 } 7237 7238 /* XXX For now */ 7239 ace->type = ACE4_ACCESS_ALLOWED_ACE_TYPE; 7240 ace->flag = 0; 7241 ace->access_mask = 0; 7242 ace->who.utf8string_len = 0; 7243 ace->who.utf8string_val = 0; 7244 7245 rfs4_deleg_state_rele(dsp); 7246 rfs4_state_rele(sp); 7247 rfs4_file_rele(fp); 7248 } 7249 7250 typedef enum { 7251 NFS4_CHKSEQ_OKAY = 0, 7252 NFS4_CHKSEQ_REPLAY = 1, 7253 NFS4_CHKSEQ_BAD = 2 7254 } rfs4_chkseq_t; 7255 7256 /* 7257 * Generic function for sequence number checks. 7258 */ 7259 static rfs4_chkseq_t 7260 rfs4_check_seqid(seqid4 seqid, nfs_resop4 *lastop, 7261 seqid4 rqst_seq, nfs_resop4 *resop, bool_t copyres) 7262 { 7263 /* Same sequence ids and matching operations? */ 7264 if (seqid == rqst_seq && resop->resop == lastop->resop) { 7265 if (copyres == TRUE) { 7266 rfs4_free_reply(resop); 7267 rfs4_copy_reply(resop, lastop); 7268 } 7269 NFS4_DEBUG(rfs4_debug, (CE_NOTE, 7270 "Replayed SEQID %d\n", seqid)); 7271 return (NFS4_CHKSEQ_REPLAY); 7272 } 7273 7274 /* If the incoming sequence is not the next expected then it is bad */ 7275 if (rqst_seq != seqid + 1) { 7276 if (rqst_seq == seqid) { 7277 NFS4_DEBUG(rfs4_debug, 7278 (CE_NOTE, "BAD SEQID: Replayed sequence id " 7279 "but last op was %d current op is %d\n", 7280 lastop->resop, resop->resop)); 7281 return (NFS4_CHKSEQ_BAD); 7282 } 7283 NFS4_DEBUG(rfs4_debug, 7284 (CE_NOTE, "BAD SEQID: got %u expecting %u\n", 7285 rqst_seq, seqid)); 7286 return (NFS4_CHKSEQ_BAD); 7287 } 7288 7289 /* Everything okay -- next expected */ 7290 return (NFS4_CHKSEQ_OKAY); 7291 } 7292 7293 7294 static rfs4_chkseq_t 7295 rfs4_check_open_seqid(seqid4 seqid, rfs4_openowner_t *op, nfs_resop4 *resop) 7296 { 7297 rfs4_chkseq_t rc; 7298 7299 rfs4_dbe_lock(op->ro_dbe); 7300 rc = rfs4_check_seqid(op->ro_open_seqid, &op->ro_reply, seqid, resop, 7301 TRUE); 7302 rfs4_dbe_unlock(op->ro_dbe); 7303 7304 if (rc == NFS4_CHKSEQ_OKAY) 7305 rfs4_update_lease(op->ro_client); 7306 7307 return (rc); 7308 } 7309 7310 static rfs4_chkseq_t 7311 rfs4_check_olo_seqid(seqid4 olo_seqid, rfs4_openowner_t *op, nfs_resop4 *resop) 7312 { 7313 rfs4_chkseq_t rc; 7314 7315 rfs4_dbe_lock(op->ro_dbe); 7316 rc = rfs4_check_seqid(op->ro_open_seqid, &op->ro_reply, 7317 olo_seqid, resop, FALSE); 7318 rfs4_dbe_unlock(op->ro_dbe); 7319 7320 return (rc); 7321 } 7322 7323 static rfs4_chkseq_t 7324 rfs4_check_lock_seqid(seqid4 seqid, rfs4_lo_state_t *lsp, nfs_resop4 *resop) 7325 { 7326 rfs4_chkseq_t rc = NFS4_CHKSEQ_OKAY; 7327 7328 rfs4_dbe_lock(lsp->rls_dbe); 7329 if (!lsp->rls_skip_seqid_check) 7330 rc = rfs4_check_seqid(lsp->rls_seqid, &lsp->rls_reply, seqid, 7331 resop, TRUE); 7332 rfs4_dbe_unlock(lsp->rls_dbe); 7333 7334 return (rc); 7335 } 7336 7337 static void 7338 rfs4_op_open(nfs_argop4 *argop, nfs_resop4 *resop, 7339 struct svc_req *req, struct compound_state *cs) 7340 { 7341 OPEN4args *args = &argop->nfs_argop4_u.opopen; 7342 OPEN4res *resp = &resop->nfs_resop4_u.opopen; 7343 open_owner4 *owner = &args->owner; 7344 open_claim_type4 claim = args->claim; 7345 rfs4_client_t *cp; 7346 rfs4_openowner_t *oo; 7347 bool_t create; 7348 bool_t replay = FALSE; 7349 int can_reclaim; 7350 7351 DTRACE_NFSV4_2(op__open__start, struct compound_state *, cs, 7352 OPEN4args *, args); 7353 7354 if (cs->vp == NULL) { 7355 *cs->statusp = resp->status = NFS4ERR_NOFILEHANDLE; 7356 goto end; 7357 } 7358 7359 /* 7360 * Need to check clientid and lease expiration first based on 7361 * error ordering and incrementing sequence id. 7362 */ 7363 cp = rfs4_findclient_by_id(owner->clientid, FALSE); 7364 if (cp == NULL) { 7365 *cs->statusp = resp->status = 7366 rfs4_check_clientid(&owner->clientid, 0); 7367 goto end; 7368 } 7369 7370 if (rfs4_lease_expired(cp)) { 7371 rfs4_client_close(cp); 7372 *cs->statusp = resp->status = NFS4ERR_EXPIRED; 7373 goto end; 7374 } 7375 can_reclaim = cp->rc_can_reclaim; 7376 7377 /* 7378 * Find the open_owner for use from this point forward. Take 7379 * care in updating the sequence id based on the type of error 7380 * being returned. 7381 */ 7382 retry: 7383 create = TRUE; 7384 oo = rfs4_findopenowner(owner, &create, args->seqid); 7385 if (oo == NULL) { 7386 *cs->statusp = resp->status = NFS4ERR_RESOURCE; 7387 rfs4_client_rele(cp); 7388 goto end; 7389 } 7390 7391 /* Hold off access to the sequence space while the open is done */ 7392 rfs4_sw_enter(&oo->ro_sw); 7393 7394 /* 7395 * If the open_owner existed before at the server, then check 7396 * the sequence id. 7397 */ 7398 if (!create && !oo->ro_postpone_confirm) { 7399 switch (rfs4_check_open_seqid(args->seqid, oo, resop)) { 7400 case NFS4_CHKSEQ_BAD: 7401 if ((args->seqid > oo->ro_open_seqid) && 7402 oo->ro_need_confirm) { 7403 rfs4_free_opens(oo, TRUE, FALSE); 7404 rfs4_sw_exit(&oo->ro_sw); 7405 rfs4_openowner_rele(oo); 7406 goto retry; 7407 } 7408 resp->status = NFS4ERR_BAD_SEQID; 7409 goto out; 7410 case NFS4_CHKSEQ_REPLAY: /* replay of previous request */ 7411 replay = TRUE; 7412 goto out; 7413 default: 7414 break; 7415 } 7416 7417 /* 7418 * Sequence was ok and open owner exists 7419 * check to see if we have yet to see an 7420 * open_confirm. 7421 */ 7422 if (oo->ro_need_confirm) { 7423 rfs4_free_opens(oo, TRUE, FALSE); 7424 rfs4_sw_exit(&oo->ro_sw); 7425 rfs4_openowner_rele(oo); 7426 goto retry; 7427 } 7428 } 7429 /* Grace only applies to regular-type OPENs */ 7430 if (rfs4_clnt_in_grace(cp) && 7431 (claim == CLAIM_NULL || claim == CLAIM_DELEGATE_CUR)) { 7432 *cs->statusp = resp->status = NFS4ERR_GRACE; 7433 goto out; 7434 } 7435 7436 /* 7437 * If previous state at the server existed then can_reclaim 7438 * will be set. If not reply NFS4ERR_NO_GRACE to the 7439 * client. 7440 */ 7441 if (rfs4_clnt_in_grace(cp) && claim == CLAIM_PREVIOUS && !can_reclaim) { 7442 *cs->statusp = resp->status = NFS4ERR_NO_GRACE; 7443 goto out; 7444 } 7445 7446 7447 /* 7448 * Reject the open if the client has missed the grace period 7449 */ 7450 if (!rfs4_clnt_in_grace(cp) && claim == CLAIM_PREVIOUS) { 7451 *cs->statusp = resp->status = NFS4ERR_NO_GRACE; 7452 goto out; 7453 } 7454 7455 /* Couple of up-front bookkeeping items */ 7456 if (oo->ro_need_confirm) { 7457 /* 7458 * If this is a reclaim OPEN then we should not ask 7459 * for a confirmation of the open_owner per the 7460 * protocol specification. 7461 */ 7462 if (claim == CLAIM_PREVIOUS) 7463 oo->ro_need_confirm = FALSE; 7464 else 7465 resp->rflags |= OPEN4_RESULT_CONFIRM; 7466 } 7467 resp->rflags |= OPEN4_RESULT_LOCKTYPE_POSIX; 7468 7469 /* 7470 * If there is an unshared filesystem mounted on this vnode, 7471 * do not allow to open/create in this directory. 7472 */ 7473 if (vn_ismntpt(cs->vp)) { 7474 *cs->statusp = resp->status = NFS4ERR_ACCESS; 7475 goto out; 7476 } 7477 7478 /* 7479 * access must READ, WRITE, or BOTH. No access is invalid. 7480 * deny can be READ, WRITE, BOTH, or NONE. 7481 * bits not defined for access/deny are invalid. 7482 */ 7483 if (! (args->share_access & OPEN4_SHARE_ACCESS_BOTH) || 7484 (args->share_access & ~OPEN4_SHARE_ACCESS_BOTH) || 7485 (args->share_deny & ~OPEN4_SHARE_DENY_BOTH)) { 7486 *cs->statusp = resp->status = NFS4ERR_INVAL; 7487 goto out; 7488 } 7489 7490 7491 /* 7492 * make sure attrset is zero before response is built. 7493 */ 7494 resp->attrset = 0; 7495 7496 switch (claim) { 7497 case CLAIM_NULL: 7498 rfs4_do_opennull(cs, req, args, oo, resp); 7499 break; 7500 case CLAIM_PREVIOUS: 7501 rfs4_do_openprev(cs, req, args, oo, resp); 7502 break; 7503 case CLAIM_DELEGATE_CUR: 7504 rfs4_do_opendelcur(cs, req, args, oo, resp); 7505 break; 7506 case CLAIM_DELEGATE_PREV: 7507 rfs4_do_opendelprev(cs, req, args, oo, resp); 7508 break; 7509 default: 7510 resp->status = NFS4ERR_INVAL; 7511 break; 7512 } 7513 7514 out: 7515 rfs4_client_rele(cp); 7516 7517 /* Catch sequence id handling here to make it a little easier */ 7518 switch (resp->status) { 7519 case NFS4ERR_BADXDR: 7520 case NFS4ERR_BAD_SEQID: 7521 case NFS4ERR_BAD_STATEID: 7522 case NFS4ERR_NOFILEHANDLE: 7523 case NFS4ERR_RESOURCE: 7524 case NFS4ERR_STALE_CLIENTID: 7525 case NFS4ERR_STALE_STATEID: 7526 /* 7527 * The protocol states that if any of these errors are 7528 * being returned, the sequence id should not be 7529 * incremented. Any other return requires an 7530 * increment. 7531 */ 7532 break; 7533 default: 7534 /* Always update the lease in this case */ 7535 rfs4_update_lease(oo->ro_client); 7536 7537 /* Regular response - copy the result */ 7538 if (!replay) 7539 rfs4_update_open_resp(oo, resop, &cs->fh); 7540 7541 /* 7542 * REPLAY case: Only if the previous response was OK 7543 * do we copy the filehandle. If not OK, no 7544 * filehandle to copy. 7545 */ 7546 if (replay == TRUE && 7547 resp->status == NFS4_OK && 7548 oo->ro_reply_fh.nfs_fh4_val) { 7549 /* 7550 * If this is a replay, we must restore the 7551 * current filehandle/vp to that of what was 7552 * returned originally. Try our best to do 7553 * it. 7554 */ 7555 nfs_fh4_fmt_t *fh_fmtp = 7556 (nfs_fh4_fmt_t *)oo->ro_reply_fh.nfs_fh4_val; 7557 7558 cs->exi = checkexport4(&fh_fmtp->fh4_fsid, 7559 (fid_t *)&fh_fmtp->fh4_xlen, NULL); 7560 7561 if (cs->exi == NULL) { 7562 resp->status = NFS4ERR_STALE; 7563 goto finish; 7564 } 7565 7566 VN_RELE(cs->vp); 7567 7568 cs->vp = nfs4_fhtovp(&oo->ro_reply_fh, cs->exi, 7569 &resp->status); 7570 7571 if (cs->vp == NULL) 7572 goto finish; 7573 7574 nfs_fh4_copy(&oo->ro_reply_fh, &cs->fh); 7575 } 7576 7577 /* 7578 * If this was a replay, no need to update the 7579 * sequence id. If the open_owner was not created on 7580 * this pass, then update. The first use of an 7581 * open_owner will not bump the sequence id. 7582 */ 7583 if (replay == FALSE && !create) 7584 rfs4_update_open_sequence(oo); 7585 /* 7586 * If the client is receiving an error and the 7587 * open_owner needs to be confirmed, there is no way 7588 * to notify the client of this fact ignoring the fact 7589 * that the server has no method of returning a 7590 * stateid to confirm. Therefore, the server needs to 7591 * mark this open_owner in a way as to avoid the 7592 * sequence id checking the next time the client uses 7593 * this open_owner. 7594 */ 7595 if (resp->status != NFS4_OK && oo->ro_need_confirm) 7596 oo->ro_postpone_confirm = TRUE; 7597 /* 7598 * If OK response then clear the postpone flag and 7599 * reset the sequence id to keep in sync with the 7600 * client. 7601 */ 7602 if (resp->status == NFS4_OK && oo->ro_postpone_confirm) { 7603 oo->ro_postpone_confirm = FALSE; 7604 oo->ro_open_seqid = args->seqid; 7605 } 7606 break; 7607 } 7608 7609 finish: 7610 *cs->statusp = resp->status; 7611 7612 rfs4_sw_exit(&oo->ro_sw); 7613 rfs4_openowner_rele(oo); 7614 7615 end: 7616 DTRACE_NFSV4_2(op__open__done, struct compound_state *, cs, 7617 OPEN4res *, resp); 7618 } 7619 7620 /*ARGSUSED*/ 7621 void 7622 rfs4_op_open_confirm(nfs_argop4 *argop, nfs_resop4 *resop, 7623 struct svc_req *req, struct compound_state *cs) 7624 { 7625 OPEN_CONFIRM4args *args = &argop->nfs_argop4_u.opopen_confirm; 7626 OPEN_CONFIRM4res *resp = &resop->nfs_resop4_u.opopen_confirm; 7627 rfs4_state_t *sp; 7628 nfsstat4 status; 7629 7630 DTRACE_NFSV4_2(op__open__confirm__start, struct compound_state *, cs, 7631 OPEN_CONFIRM4args *, args); 7632 7633 if (cs->vp == NULL) { 7634 *cs->statusp = resp->status = NFS4ERR_NOFILEHANDLE; 7635 goto out; 7636 } 7637 7638 if (cs->vp->v_type != VREG) { 7639 *cs->statusp = resp->status = 7640 cs->vp->v_type == VDIR ? NFS4ERR_ISDIR : NFS4ERR_INVAL; 7641 return; 7642 } 7643 7644 status = rfs4_get_state(&args->open_stateid, &sp, RFS4_DBS_VALID); 7645 if (status != NFS4_OK) { 7646 *cs->statusp = resp->status = status; 7647 goto out; 7648 } 7649 7650 /* Ensure specified filehandle matches */ 7651 if (cs->vp != sp->rs_finfo->rf_vp) { 7652 rfs4_state_rele(sp); 7653 *cs->statusp = resp->status = NFS4ERR_BAD_STATEID; 7654 goto out; 7655 } 7656 7657 /* hold off other access to open_owner while we tinker */ 7658 rfs4_sw_enter(&sp->rs_owner->ro_sw); 7659 7660 switch (rfs4_check_stateid_seqid(sp, &args->open_stateid)) { 7661 case NFS4_CHECK_STATEID_OKAY: 7662 if (rfs4_check_open_seqid(args->seqid, sp->rs_owner, 7663 resop) != 0) { 7664 *cs->statusp = resp->status = NFS4ERR_BAD_SEQID; 7665 break; 7666 } 7667 /* 7668 * If it is the appropriate stateid and determined to 7669 * be "OKAY" then this means that the stateid does not 7670 * need to be confirmed and the client is in error for 7671 * sending an OPEN_CONFIRM. 7672 */ 7673 *cs->statusp = resp->status = NFS4ERR_BAD_STATEID; 7674 break; 7675 case NFS4_CHECK_STATEID_OLD: 7676 *cs->statusp = resp->status = NFS4ERR_OLD_STATEID; 7677 break; 7678 case NFS4_CHECK_STATEID_BAD: 7679 *cs->statusp = resp->status = NFS4ERR_BAD_STATEID; 7680 break; 7681 case NFS4_CHECK_STATEID_EXPIRED: 7682 *cs->statusp = resp->status = NFS4ERR_EXPIRED; 7683 break; 7684 case NFS4_CHECK_STATEID_CLOSED: 7685 *cs->statusp = resp->status = NFS4ERR_OLD_STATEID; 7686 break; 7687 case NFS4_CHECK_STATEID_REPLAY: 7688 switch (rfs4_check_open_seqid(args->seqid, sp->rs_owner, 7689 resop)) { 7690 case NFS4_CHKSEQ_OKAY: 7691 /* 7692 * This is replayed stateid; if seqid matches 7693 * next expected, then client is using wrong seqid. 7694 */ 7695 /* fall through */ 7696 case NFS4_CHKSEQ_BAD: 7697 *cs->statusp = resp->status = NFS4ERR_BAD_SEQID; 7698 break; 7699 case NFS4_CHKSEQ_REPLAY: 7700 /* 7701 * Note this case is the duplicate case so 7702 * resp->status is already set. 7703 */ 7704 *cs->statusp = resp->status; 7705 rfs4_update_lease(sp->rs_owner->ro_client); 7706 break; 7707 } 7708 break; 7709 case NFS4_CHECK_STATEID_UNCONFIRMED: 7710 if (rfs4_check_open_seqid(args->seqid, sp->rs_owner, 7711 resop) != NFS4_CHKSEQ_OKAY) { 7712 *cs->statusp = resp->status = NFS4ERR_BAD_SEQID; 7713 break; 7714 } 7715 *cs->statusp = resp->status = NFS4_OK; 7716 7717 next_stateid(&sp->rs_stateid); 7718 resp->open_stateid = sp->rs_stateid.stateid; 7719 sp->rs_owner->ro_need_confirm = FALSE; 7720 rfs4_update_lease(sp->rs_owner->ro_client); 7721 rfs4_update_open_sequence(sp->rs_owner); 7722 rfs4_update_open_resp(sp->rs_owner, resop, NULL); 7723 break; 7724 default: 7725 ASSERT(FALSE); 7726 *cs->statusp = resp->status = NFS4ERR_SERVERFAULT; 7727 break; 7728 } 7729 rfs4_sw_exit(&sp->rs_owner->ro_sw); 7730 rfs4_state_rele(sp); 7731 7732 out: 7733 DTRACE_NFSV4_2(op__open__confirm__done, struct compound_state *, cs, 7734 OPEN_CONFIRM4res *, resp); 7735 } 7736 7737 /*ARGSUSED*/ 7738 void 7739 rfs4_op_open_downgrade(nfs_argop4 *argop, nfs_resop4 *resop, 7740 struct svc_req *req, struct compound_state *cs) 7741 { 7742 OPEN_DOWNGRADE4args *args = &argop->nfs_argop4_u.opopen_downgrade; 7743 OPEN_DOWNGRADE4res *resp = &resop->nfs_resop4_u.opopen_downgrade; 7744 uint32_t access = args->share_access; 7745 uint32_t deny = args->share_deny; 7746 nfsstat4 status; 7747 rfs4_state_t *sp; 7748 rfs4_file_t *fp; 7749 int fflags = 0; 7750 7751 DTRACE_NFSV4_2(op__open__downgrade__start, struct compound_state *, cs, 7752 OPEN_DOWNGRADE4args *, args); 7753 7754 if (cs->vp == NULL) { 7755 *cs->statusp = resp->status = NFS4ERR_NOFILEHANDLE; 7756 goto out; 7757 } 7758 7759 if (cs->vp->v_type != VREG) { 7760 *cs->statusp = resp->status = NFS4ERR_INVAL; 7761 return; 7762 } 7763 7764 status = rfs4_get_state(&args->open_stateid, &sp, RFS4_DBS_VALID); 7765 if (status != NFS4_OK) { 7766 *cs->statusp = resp->status = status; 7767 goto out; 7768 } 7769 7770 /* Ensure specified filehandle matches */ 7771 if (cs->vp != sp->rs_finfo->rf_vp) { 7772 rfs4_state_rele(sp); 7773 *cs->statusp = resp->status = NFS4ERR_BAD_STATEID; 7774 goto out; 7775 } 7776 7777 /* hold off other access to open_owner while we tinker */ 7778 rfs4_sw_enter(&sp->rs_owner->ro_sw); 7779 7780 switch (rfs4_check_stateid_seqid(sp, &args->open_stateid)) { 7781 case NFS4_CHECK_STATEID_OKAY: 7782 if (rfs4_check_open_seqid(args->seqid, sp->rs_owner, 7783 resop) != NFS4_CHKSEQ_OKAY) { 7784 *cs->statusp = resp->status = NFS4ERR_BAD_SEQID; 7785 goto end; 7786 } 7787 break; 7788 case NFS4_CHECK_STATEID_OLD: 7789 *cs->statusp = resp->status = NFS4ERR_OLD_STATEID; 7790 goto end; 7791 case NFS4_CHECK_STATEID_BAD: 7792 *cs->statusp = resp->status = NFS4ERR_BAD_STATEID; 7793 goto end; 7794 case NFS4_CHECK_STATEID_EXPIRED: 7795 *cs->statusp = resp->status = NFS4ERR_EXPIRED; 7796 goto end; 7797 case NFS4_CHECK_STATEID_CLOSED: 7798 *cs->statusp = resp->status = NFS4ERR_OLD_STATEID; 7799 goto end; 7800 case NFS4_CHECK_STATEID_UNCONFIRMED: 7801 *cs->statusp = resp->status = NFS4ERR_BAD_STATEID; 7802 goto end; 7803 case NFS4_CHECK_STATEID_REPLAY: 7804 /* Check the sequence id for the open owner */ 7805 switch (rfs4_check_open_seqid(args->seqid, sp->rs_owner, 7806 resop)) { 7807 case NFS4_CHKSEQ_OKAY: 7808 /* 7809 * This is replayed stateid; if seqid matches 7810 * next expected, then client is using wrong seqid. 7811 */ 7812 /* fall through */ 7813 case NFS4_CHKSEQ_BAD: 7814 *cs->statusp = resp->status = NFS4ERR_BAD_SEQID; 7815 goto end; 7816 case NFS4_CHKSEQ_REPLAY: 7817 /* 7818 * Note this case is the duplicate case so 7819 * resp->status is already set. 7820 */ 7821 *cs->statusp = resp->status; 7822 rfs4_update_lease(sp->rs_owner->ro_client); 7823 goto end; 7824 } 7825 break; 7826 default: 7827 ASSERT(FALSE); 7828 break; 7829 } 7830 7831 rfs4_dbe_lock(sp->rs_dbe); 7832 /* 7833 * Check that the new access modes and deny modes are valid. 7834 * Check that no invalid bits are set. 7835 */ 7836 if ((access & ~(OPEN4_SHARE_ACCESS_READ | OPEN4_SHARE_ACCESS_WRITE)) || 7837 (deny & ~(OPEN4_SHARE_DENY_READ | OPEN4_SHARE_DENY_WRITE))) { 7838 *cs->statusp = resp->status = NFS4ERR_INVAL; 7839 rfs4_update_open_sequence(sp->rs_owner); 7840 rfs4_dbe_unlock(sp->rs_dbe); 7841 goto end; 7842 } 7843 7844 /* 7845 * The new modes must be a subset of the current modes and 7846 * the access must specify at least one mode. To test that 7847 * the new mode is a subset of the current modes we bitwise 7848 * AND them together and check that the result equals the new 7849 * mode. For example: 7850 * New mode, access == R and current mode, sp->rs_open_access == RW 7851 * access & sp->rs_open_access == R == access, so the new access mode 7852 * is valid. Consider access == RW, sp->rs_open_access = R 7853 * access & sp->rs_open_access == R != access, so the new access mode 7854 * is invalid. 7855 */ 7856 if ((access & sp->rs_open_access) != access || 7857 (deny & sp->rs_open_deny) != deny || 7858 (access & 7859 (OPEN4_SHARE_ACCESS_READ | OPEN4_SHARE_ACCESS_WRITE)) == 0) { 7860 *cs->statusp = resp->status = NFS4ERR_INVAL; 7861 rfs4_update_open_sequence(sp->rs_owner); 7862 rfs4_dbe_unlock(sp->rs_dbe); 7863 goto end; 7864 } 7865 7866 /* 7867 * Release any share locks associated with this stateID. 7868 * Strictly speaking, this violates the spec because the 7869 * spec effectively requires that open downgrade be atomic. 7870 * At present, fs_shrlock does not have this capability. 7871 */ 7872 (void) rfs4_unshare(sp); 7873 7874 status = rfs4_share(sp, access, deny); 7875 if (status != NFS4_OK) { 7876 *cs->statusp = resp->status = NFS4ERR_SERVERFAULT; 7877 rfs4_update_open_sequence(sp->rs_owner); 7878 rfs4_dbe_unlock(sp->rs_dbe); 7879 goto end; 7880 } 7881 7882 fp = sp->rs_finfo; 7883 rfs4_dbe_lock(fp->rf_dbe); 7884 7885 /* 7886 * If the current mode has deny read and the new mode 7887 * does not, decrement the number of deny read mode bits 7888 * and if it goes to zero turn off the deny read bit 7889 * on the file. 7890 */ 7891 if ((sp->rs_open_deny & OPEN4_SHARE_DENY_READ) && 7892 (deny & OPEN4_SHARE_DENY_READ) == 0) { 7893 fp->rf_deny_read--; 7894 if (fp->rf_deny_read == 0) 7895 fp->rf_share_deny &= ~OPEN4_SHARE_DENY_READ; 7896 } 7897 7898 /* 7899 * If the current mode has deny write and the new mode 7900 * does not, decrement the number of deny write mode bits 7901 * and if it goes to zero turn off the deny write bit 7902 * on the file. 7903 */ 7904 if ((sp->rs_open_deny & OPEN4_SHARE_DENY_WRITE) && 7905 (deny & OPEN4_SHARE_DENY_WRITE) == 0) { 7906 fp->rf_deny_write--; 7907 if (fp->rf_deny_write == 0) 7908 fp->rf_share_deny &= ~OPEN4_SHARE_DENY_WRITE; 7909 } 7910 7911 /* 7912 * If the current mode has access read and the new mode 7913 * does not, decrement the number of access read mode bits 7914 * and if it goes to zero turn off the access read bit 7915 * on the file. set fflags to FREAD for the call to 7916 * vn_open_downgrade(). 7917 */ 7918 if ((sp->rs_open_access & OPEN4_SHARE_ACCESS_READ) && 7919 (access & OPEN4_SHARE_ACCESS_READ) == 0) { 7920 fp->rf_access_read--; 7921 if (fp->rf_access_read == 0) 7922 fp->rf_share_access &= ~OPEN4_SHARE_ACCESS_READ; 7923 fflags |= FREAD; 7924 } 7925 7926 /* 7927 * If the current mode has access write and the new mode 7928 * does not, decrement the number of access write mode bits 7929 * and if it goes to zero turn off the access write bit 7930 * on the file. set fflags to FWRITE for the call to 7931 * vn_open_downgrade(). 7932 */ 7933 if ((sp->rs_open_access & OPEN4_SHARE_ACCESS_WRITE) && 7934 (access & OPEN4_SHARE_ACCESS_WRITE) == 0) { 7935 fp->rf_access_write--; 7936 if (fp->rf_access_write == 0) 7937 fp->rf_share_deny &= ~OPEN4_SHARE_ACCESS_WRITE; 7938 fflags |= FWRITE; 7939 } 7940 7941 /* Check that the file is still accessible */ 7942 ASSERT(fp->rf_share_access); 7943 7944 rfs4_dbe_unlock(fp->rf_dbe); 7945 7946 /* now set the new open access and deny modes */ 7947 sp->rs_open_access = access; 7948 sp->rs_open_deny = deny; 7949 7950 /* 7951 * we successfully downgraded the share lock, now we need to downgrade 7952 * the open. it is possible that the downgrade was only for a deny 7953 * mode and we have nothing else to do. 7954 */ 7955 if ((fflags & (FREAD|FWRITE)) != 0) 7956 vn_open_downgrade(cs->vp, fflags); 7957 7958 /* Update the stateid */ 7959 next_stateid(&sp->rs_stateid); 7960 resp->open_stateid = sp->rs_stateid.stateid; 7961 7962 rfs4_dbe_unlock(sp->rs_dbe); 7963 7964 *cs->statusp = resp->status = NFS4_OK; 7965 /* Update the lease */ 7966 rfs4_update_lease(sp->rs_owner->ro_client); 7967 /* And the sequence */ 7968 rfs4_update_open_sequence(sp->rs_owner); 7969 rfs4_update_open_resp(sp->rs_owner, resop, NULL); 7970 7971 end: 7972 rfs4_sw_exit(&sp->rs_owner->ro_sw); 7973 rfs4_state_rele(sp); 7974 out: 7975 DTRACE_NFSV4_2(op__open__downgrade__done, struct compound_state *, cs, 7976 OPEN_DOWNGRADE4res *, resp); 7977 } 7978 7979 static void * 7980 memstr(const void *s1, const char *s2, size_t n) 7981 { 7982 size_t l = strlen(s2); 7983 char *p = (char *)s1; 7984 7985 while (n >= l) { 7986 if (bcmp(p, s2, l) == 0) 7987 return (p); 7988 p++; 7989 n--; 7990 } 7991 7992 return (NULL); 7993 } 7994 7995 /* 7996 * The logic behind this function is detailed in the NFSv4 RFC in the 7997 * SETCLIENTID operation description under IMPLEMENTATION. Refer to 7998 * that section for explicit guidance to server behavior for 7999 * SETCLIENTID. 8000 */ 8001 void 8002 rfs4_op_setclientid(nfs_argop4 *argop, nfs_resop4 *resop, 8003 struct svc_req *req, struct compound_state *cs) 8004 { 8005 SETCLIENTID4args *args = &argop->nfs_argop4_u.opsetclientid; 8006 SETCLIENTID4res *res = &resop->nfs_resop4_u.opsetclientid; 8007 rfs4_client_t *cp, *newcp, *cp_confirmed, *cp_unconfirmed; 8008 rfs4_clntip_t *ci; 8009 bool_t create; 8010 char *addr, *netid; 8011 int len; 8012 8013 DTRACE_NFSV4_2(op__setclientid__start, struct compound_state *, cs, 8014 SETCLIENTID4args *, args); 8015 retry: 8016 newcp = cp_confirmed = cp_unconfirmed = NULL; 8017 8018 /* 8019 * Save the caller's IP address 8020 */ 8021 args->client.cl_addr = 8022 (struct sockaddr *)svc_getrpccaller(req->rq_xprt)->buf; 8023 8024 /* 8025 * Record if it is a Solaris client that cannot handle referrals. 8026 */ 8027 if (memstr(args->client.id_val, "Solaris", args->client.id_len) && 8028 !memstr(args->client.id_val, "+referrals", args->client.id_len)) { 8029 /* Add a "yes, it's downrev" record */ 8030 create = TRUE; 8031 ci = rfs4_find_clntip(args->client.cl_addr, &create); 8032 ASSERT(ci != NULL); 8033 rfs4_dbe_rele(ci->ri_dbe); 8034 } else { 8035 /* Remove any previous record */ 8036 rfs4_invalidate_clntip(args->client.cl_addr); 8037 } 8038 8039 /* 8040 * In search of an EXISTING client matching the incoming 8041 * request to establish a new client identifier at the server 8042 */ 8043 create = TRUE; 8044 cp = rfs4_findclient(&args->client, &create, NULL); 8045 8046 /* Should never happen */ 8047 ASSERT(cp != NULL); 8048 8049 if (cp == NULL) { 8050 *cs->statusp = res->status = NFS4ERR_SERVERFAULT; 8051 goto out; 8052 } 8053 8054 /* 8055 * Easiest case. Client identifier is newly created and is 8056 * unconfirmed. Also note that for this case, no other 8057 * entries exist for the client identifier. Nothing else to 8058 * check. Just setup the response and respond. 8059 */ 8060 if (create) { 8061 *cs->statusp = res->status = NFS4_OK; 8062 res->SETCLIENTID4res_u.resok4.clientid = cp->rc_clientid; 8063 res->SETCLIENTID4res_u.resok4.setclientid_confirm = 8064 cp->rc_confirm_verf; 8065 /* Setup callback information; CB_NULL confirmation later */ 8066 rfs4_client_setcb(cp, &args->callback, args->callback_ident); 8067 8068 rfs4_client_rele(cp); 8069 goto out; 8070 } 8071 8072 /* 8073 * An existing, confirmed client may exist but it may not have 8074 * been active for at least one lease period. If so, then 8075 * "close" the client and create a new client identifier 8076 */ 8077 if (rfs4_lease_expired(cp)) { 8078 rfs4_client_close(cp); 8079 goto retry; 8080 } 8081 8082 if (cp->rc_need_confirm == TRUE) 8083 cp_unconfirmed = cp; 8084 else 8085 cp_confirmed = cp; 8086 8087 cp = NULL; 8088 8089 /* 8090 * We have a confirmed client, now check for an 8091 * unconfimred entry 8092 */ 8093 if (cp_confirmed) { 8094 /* If creds don't match then client identifier is inuse */ 8095 if (!creds_ok(cp_confirmed->rc_cr_set, req, cs)) { 8096 rfs4_cbinfo_t *cbp; 8097 /* 8098 * Some one else has established this client 8099 * id. Try and say * who they are. We will use 8100 * the call back address supplied by * the 8101 * first client. 8102 */ 8103 *cs->statusp = res->status = NFS4ERR_CLID_INUSE; 8104 8105 addr = netid = NULL; 8106 8107 cbp = &cp_confirmed->rc_cbinfo; 8108 if (cbp->cb_callback.cb_location.r_addr && 8109 cbp->cb_callback.cb_location.r_netid) { 8110 cb_client4 *cbcp = &cbp->cb_callback; 8111 8112 len = strlen(cbcp->cb_location.r_addr)+1; 8113 addr = kmem_alloc(len, KM_SLEEP); 8114 bcopy(cbcp->cb_location.r_addr, addr, len); 8115 len = strlen(cbcp->cb_location.r_netid)+1; 8116 netid = kmem_alloc(len, KM_SLEEP); 8117 bcopy(cbcp->cb_location.r_netid, netid, len); 8118 } 8119 8120 res->SETCLIENTID4res_u.client_using.r_addr = addr; 8121 res->SETCLIENTID4res_u.client_using.r_netid = netid; 8122 8123 rfs4_client_rele(cp_confirmed); 8124 } 8125 8126 /* 8127 * Confirmed, creds match, and verifier matches; must 8128 * be an update of the callback info 8129 */ 8130 if (cp_confirmed->rc_nfs_client.verifier == 8131 args->client.verifier) { 8132 /* Setup callback information */ 8133 rfs4_client_setcb(cp_confirmed, &args->callback, 8134 args->callback_ident); 8135 8136 /* everything okay -- move ahead */ 8137 *cs->statusp = res->status = NFS4_OK; 8138 res->SETCLIENTID4res_u.resok4.clientid = 8139 cp_confirmed->rc_clientid; 8140 8141 /* update the confirm_verifier and return it */ 8142 rfs4_client_scv_next(cp_confirmed); 8143 res->SETCLIENTID4res_u.resok4.setclientid_confirm = 8144 cp_confirmed->rc_confirm_verf; 8145 8146 rfs4_client_rele(cp_confirmed); 8147 goto out; 8148 } 8149 8150 /* 8151 * Creds match but the verifier doesn't. Must search 8152 * for an unconfirmed client that would be replaced by 8153 * this request. 8154 */ 8155 create = FALSE; 8156 cp_unconfirmed = rfs4_findclient(&args->client, &create, 8157 cp_confirmed); 8158 } 8159 8160 /* 8161 * At this point, we have taken care of the brand new client 8162 * struct, INUSE case, update of an existing, and confirmed 8163 * client struct. 8164 */ 8165 8166 /* 8167 * check to see if things have changed while we originally 8168 * picked up the client struct. If they have, then return and 8169 * retry the processing of this SETCLIENTID request. 8170 */ 8171 if (cp_unconfirmed) { 8172 rfs4_dbe_lock(cp_unconfirmed->rc_dbe); 8173 if (!cp_unconfirmed->rc_need_confirm) { 8174 rfs4_dbe_unlock(cp_unconfirmed->rc_dbe); 8175 rfs4_client_rele(cp_unconfirmed); 8176 if (cp_confirmed) 8177 rfs4_client_rele(cp_confirmed); 8178 goto retry; 8179 } 8180 /* do away with the old unconfirmed one */ 8181 rfs4_dbe_invalidate(cp_unconfirmed->rc_dbe); 8182 rfs4_dbe_unlock(cp_unconfirmed->rc_dbe); 8183 rfs4_client_rele(cp_unconfirmed); 8184 cp_unconfirmed = NULL; 8185 } 8186 8187 /* 8188 * This search will temporarily hide the confirmed client 8189 * struct while a new client struct is created as the 8190 * unconfirmed one. 8191 */ 8192 create = TRUE; 8193 newcp = rfs4_findclient(&args->client, &create, cp_confirmed); 8194 8195 ASSERT(newcp != NULL); 8196 8197 if (newcp == NULL) { 8198 *cs->statusp = res->status = NFS4ERR_SERVERFAULT; 8199 rfs4_client_rele(cp_confirmed); 8200 goto out; 8201 } 8202 8203 /* 8204 * If one was not created, then a similar request must be in 8205 * process so release and start over with this one 8206 */ 8207 if (create != TRUE) { 8208 rfs4_client_rele(newcp); 8209 if (cp_confirmed) 8210 rfs4_client_rele(cp_confirmed); 8211 goto retry; 8212 } 8213 8214 *cs->statusp = res->status = NFS4_OK; 8215 res->SETCLIENTID4res_u.resok4.clientid = newcp->rc_clientid; 8216 res->SETCLIENTID4res_u.resok4.setclientid_confirm = 8217 newcp->rc_confirm_verf; 8218 /* Setup callback information; CB_NULL confirmation later */ 8219 rfs4_client_setcb(newcp, &args->callback, args->callback_ident); 8220 8221 newcp->rc_cp_confirmed = cp_confirmed; 8222 8223 rfs4_client_rele(newcp); 8224 8225 out: 8226 DTRACE_NFSV4_2(op__setclientid__done, struct compound_state *, cs, 8227 SETCLIENTID4res *, res); 8228 } 8229 8230 /*ARGSUSED*/ 8231 void 8232 rfs4_op_setclientid_confirm(nfs_argop4 *argop, nfs_resop4 *resop, 8233 struct svc_req *req, struct compound_state *cs) 8234 { 8235 SETCLIENTID_CONFIRM4args *args = 8236 &argop->nfs_argop4_u.opsetclientid_confirm; 8237 SETCLIENTID_CONFIRM4res *res = 8238 &resop->nfs_resop4_u.opsetclientid_confirm; 8239 rfs4_client_t *cp, *cptoclose = NULL; 8240 nfs4_srv_t *nsrv4; 8241 8242 DTRACE_NFSV4_2(op__setclientid__confirm__start, 8243 struct compound_state *, cs, 8244 SETCLIENTID_CONFIRM4args *, args); 8245 8246 nsrv4 = nfs4_get_srv(); 8247 *cs->statusp = res->status = NFS4_OK; 8248 8249 cp = rfs4_findclient_by_id(args->clientid, TRUE); 8250 8251 if (cp == NULL) { 8252 *cs->statusp = res->status = 8253 rfs4_check_clientid(&args->clientid, 1); 8254 goto out; 8255 } 8256 8257 if (!creds_ok(cp, req, cs)) { 8258 *cs->statusp = res->status = NFS4ERR_CLID_INUSE; 8259 rfs4_client_rele(cp); 8260 goto out; 8261 } 8262 8263 /* If the verifier doesn't match, the record doesn't match */ 8264 if (cp->rc_confirm_verf != args->setclientid_confirm) { 8265 *cs->statusp = res->status = NFS4ERR_STALE_CLIENTID; 8266 rfs4_client_rele(cp); 8267 goto out; 8268 } 8269 8270 rfs4_dbe_lock(cp->rc_dbe); 8271 cp->rc_need_confirm = FALSE; 8272 if (cp->rc_cp_confirmed) { 8273 cptoclose = cp->rc_cp_confirmed; 8274 cptoclose->rc_ss_remove = 1; 8275 cp->rc_cp_confirmed = NULL; 8276 } 8277 8278 /* 8279 * Update the client's associated server instance, if it's changed 8280 * since the client was created. 8281 */ 8282 if (rfs4_servinst(cp) != nsrv4->nfs4_cur_servinst) 8283 rfs4_servinst_assign(nsrv4, cp, nsrv4->nfs4_cur_servinst); 8284 8285 /* 8286 * Record clientid in stable storage. 8287 * Must be done after server instance has been assigned. 8288 */ 8289 rfs4_ss_clid(nsrv4, cp); 8290 8291 rfs4_dbe_unlock(cp->rc_dbe); 8292 8293 if (cptoclose) 8294 /* don't need to rele, client_close does it */ 8295 rfs4_client_close(cptoclose); 8296 8297 /* If needed, initiate CB_NULL call for callback path */ 8298 rfs4_deleg_cb_check(cp); 8299 rfs4_update_lease(cp); 8300 8301 /* 8302 * Check to see if client can perform reclaims 8303 */ 8304 rfs4_ss_chkclid(nsrv4, cp); 8305 8306 rfs4_client_rele(cp); 8307 8308 out: 8309 DTRACE_NFSV4_2(op__setclientid__confirm__done, 8310 struct compound_state *, cs, 8311 SETCLIENTID_CONFIRM4 *, res); 8312 } 8313 8314 8315 /*ARGSUSED*/ 8316 void 8317 rfs4_op_close(nfs_argop4 *argop, nfs_resop4 *resop, 8318 struct svc_req *req, struct compound_state *cs) 8319 { 8320 CLOSE4args *args = &argop->nfs_argop4_u.opclose; 8321 CLOSE4res *resp = &resop->nfs_resop4_u.opclose; 8322 rfs4_state_t *sp; 8323 nfsstat4 status; 8324 8325 DTRACE_NFSV4_2(op__close__start, struct compound_state *, cs, 8326 CLOSE4args *, args); 8327 8328 if (cs->vp == NULL) { 8329 *cs->statusp = resp->status = NFS4ERR_NOFILEHANDLE; 8330 goto out; 8331 } 8332 8333 status = rfs4_get_state(&args->open_stateid, &sp, RFS4_DBS_INVALID); 8334 if (status != NFS4_OK) { 8335 *cs->statusp = resp->status = status; 8336 goto out; 8337 } 8338 8339 /* Ensure specified filehandle matches */ 8340 if (cs->vp != sp->rs_finfo->rf_vp) { 8341 rfs4_state_rele(sp); 8342 *cs->statusp = resp->status = NFS4ERR_BAD_STATEID; 8343 goto out; 8344 } 8345 8346 /* hold off other access to open_owner while we tinker */ 8347 rfs4_sw_enter(&sp->rs_owner->ro_sw); 8348 8349 switch (rfs4_check_stateid_seqid(sp, &args->open_stateid)) { 8350 case NFS4_CHECK_STATEID_OKAY: 8351 if (rfs4_check_open_seqid(args->seqid, sp->rs_owner, 8352 resop) != NFS4_CHKSEQ_OKAY) { 8353 *cs->statusp = resp->status = NFS4ERR_BAD_SEQID; 8354 goto end; 8355 } 8356 break; 8357 case NFS4_CHECK_STATEID_OLD: 8358 *cs->statusp = resp->status = NFS4ERR_OLD_STATEID; 8359 goto end; 8360 case NFS4_CHECK_STATEID_BAD: 8361 *cs->statusp = resp->status = NFS4ERR_BAD_STATEID; 8362 goto end; 8363 case NFS4_CHECK_STATEID_EXPIRED: 8364 *cs->statusp = resp->status = NFS4ERR_EXPIRED; 8365 goto end; 8366 case NFS4_CHECK_STATEID_CLOSED: 8367 *cs->statusp = resp->status = NFS4ERR_OLD_STATEID; 8368 goto end; 8369 case NFS4_CHECK_STATEID_UNCONFIRMED: 8370 *cs->statusp = resp->status = NFS4ERR_BAD_STATEID; 8371 goto end; 8372 case NFS4_CHECK_STATEID_REPLAY: 8373 /* Check the sequence id for the open owner */ 8374 switch (rfs4_check_open_seqid(args->seqid, sp->rs_owner, 8375 resop)) { 8376 case NFS4_CHKSEQ_OKAY: 8377 /* 8378 * This is replayed stateid; if seqid matches 8379 * next expected, then client is using wrong seqid. 8380 */ 8381 /* FALL THROUGH */ 8382 case NFS4_CHKSEQ_BAD: 8383 *cs->statusp = resp->status = NFS4ERR_BAD_SEQID; 8384 goto end; 8385 case NFS4_CHKSEQ_REPLAY: 8386 /* 8387 * Note this case is the duplicate case so 8388 * resp->status is already set. 8389 */ 8390 *cs->statusp = resp->status; 8391 rfs4_update_lease(sp->rs_owner->ro_client); 8392 goto end; 8393 } 8394 break; 8395 default: 8396 ASSERT(FALSE); 8397 break; 8398 } 8399 8400 rfs4_dbe_lock(sp->rs_dbe); 8401 8402 /* Update the stateid. */ 8403 next_stateid(&sp->rs_stateid); 8404 resp->open_stateid = sp->rs_stateid.stateid; 8405 8406 rfs4_dbe_unlock(sp->rs_dbe); 8407 8408 rfs4_update_lease(sp->rs_owner->ro_client); 8409 rfs4_update_open_sequence(sp->rs_owner); 8410 rfs4_update_open_resp(sp->rs_owner, resop, NULL); 8411 8412 rfs4_state_close(sp, FALSE, FALSE, cs->cr); 8413 8414 *cs->statusp = resp->status = status; 8415 8416 end: 8417 rfs4_sw_exit(&sp->rs_owner->ro_sw); 8418 rfs4_state_rele(sp); 8419 out: 8420 DTRACE_NFSV4_2(op__close__done, struct compound_state *, cs, 8421 CLOSE4res *, resp); 8422 } 8423 8424 /* 8425 * Manage the counts on the file struct and close all file locks 8426 */ 8427 /*ARGSUSED*/ 8428 void 8429 rfs4_release_share_lock_state(rfs4_state_t *sp, cred_t *cr, 8430 bool_t close_of_client) 8431 { 8432 rfs4_file_t *fp = sp->rs_finfo; 8433 rfs4_lo_state_t *lsp; 8434 int fflags = 0; 8435 8436 /* 8437 * If this call is part of the larger closing down of client 8438 * state then it is just easier to release all locks 8439 * associated with this client instead of going through each 8440 * individual file and cleaning locks there. 8441 */ 8442 if (close_of_client) { 8443 if (sp->rs_owner->ro_client->rc_unlksys_completed == FALSE && 8444 !list_is_empty(&sp->rs_lostatelist) && 8445 sp->rs_owner->ro_client->rc_sysidt != LM_NOSYSID) { 8446 /* Is the PxFS kernel module loaded? */ 8447 if (lm_remove_file_locks != NULL) { 8448 int new_sysid; 8449 8450 /* Encode the cluster nodeid in new sysid */ 8451 new_sysid = sp->rs_owner->ro_client->rc_sysidt; 8452 lm_set_nlmid_flk(&new_sysid); 8453 8454 /* 8455 * This PxFS routine removes file locks for a 8456 * client over all nodes of a cluster. 8457 */ 8458 NFS4_DEBUG(rfs4_debug, (CE_NOTE, 8459 "lm_remove_file_locks(sysid=0x%x)\n", 8460 new_sysid)); 8461 (*lm_remove_file_locks)(new_sysid); 8462 } else { 8463 struct flock64 flk; 8464 8465 /* Release all locks for this client */ 8466 flk.l_type = F_UNLKSYS; 8467 flk.l_whence = 0; 8468 flk.l_start = 0; 8469 flk.l_len = 0; 8470 flk.l_sysid = 8471 sp->rs_owner->ro_client->rc_sysidt; 8472 flk.l_pid = 0; 8473 (void) VOP_FRLOCK(sp->rs_finfo->rf_vp, F_SETLK, 8474 &flk, F_REMOTELOCK | FREAD | FWRITE, 8475 (u_offset_t)0, NULL, CRED(), NULL); 8476 } 8477 8478 sp->rs_owner->ro_client->rc_unlksys_completed = TRUE; 8479 } 8480 } 8481 8482 /* 8483 * Release all locks on this file by this lock owner or at 8484 * least mark the locks as having been released 8485 */ 8486 for (lsp = list_head(&sp->rs_lostatelist); lsp != NULL; 8487 lsp = list_next(&sp->rs_lostatelist, lsp)) { 8488 lsp->rls_locks_cleaned = TRUE; 8489 8490 /* Was this already taken care of above? */ 8491 if (!close_of_client && 8492 sp->rs_owner->ro_client->rc_sysidt != LM_NOSYSID) 8493 (void) cleanlocks(sp->rs_finfo->rf_vp, 8494 lsp->rls_locker->rl_pid, 8495 lsp->rls_locker->rl_client->rc_sysidt); 8496 } 8497 8498 /* 8499 * Release any shrlocks associated with this open state ID. 8500 * This must be done before the rfs4_state gets marked closed. 8501 */ 8502 if (sp->rs_owner->ro_client->rc_sysidt != LM_NOSYSID) 8503 (void) rfs4_unshare(sp); 8504 8505 if (sp->rs_open_access) { 8506 rfs4_dbe_lock(fp->rf_dbe); 8507 8508 /* 8509 * Decrement the count for each access and deny bit that this 8510 * state has contributed to the file. 8511 * If the file counts go to zero 8512 * clear the appropriate bit in the appropriate mask. 8513 */ 8514 if (sp->rs_open_access & OPEN4_SHARE_ACCESS_READ) { 8515 fp->rf_access_read--; 8516 fflags |= FREAD; 8517 if (fp->rf_access_read == 0) 8518 fp->rf_share_access &= ~OPEN4_SHARE_ACCESS_READ; 8519 } 8520 if (sp->rs_open_access & OPEN4_SHARE_ACCESS_WRITE) { 8521 fp->rf_access_write--; 8522 fflags |= FWRITE; 8523 if (fp->rf_access_write == 0) 8524 fp->rf_share_access &= 8525 ~OPEN4_SHARE_ACCESS_WRITE; 8526 } 8527 if (sp->rs_open_deny & OPEN4_SHARE_DENY_READ) { 8528 fp->rf_deny_read--; 8529 if (fp->rf_deny_read == 0) 8530 fp->rf_share_deny &= ~OPEN4_SHARE_DENY_READ; 8531 } 8532 if (sp->rs_open_deny & OPEN4_SHARE_DENY_WRITE) { 8533 fp->rf_deny_write--; 8534 if (fp->rf_deny_write == 0) 8535 fp->rf_share_deny &= ~OPEN4_SHARE_DENY_WRITE; 8536 } 8537 8538 (void) VOP_CLOSE(fp->rf_vp, fflags, 1, (offset_t)0, cr, NULL); 8539 8540 rfs4_dbe_unlock(fp->rf_dbe); 8541 8542 sp->rs_open_access = 0; 8543 sp->rs_open_deny = 0; 8544 } 8545 } 8546 8547 /* 8548 * lock_denied: Fill in a LOCK4deneid structure given an flock64 structure. 8549 */ 8550 static nfsstat4 8551 lock_denied(LOCK4denied *dp, struct flock64 *flk) 8552 { 8553 rfs4_lockowner_t *lo; 8554 rfs4_client_t *cp; 8555 uint32_t len; 8556 8557 lo = rfs4_findlockowner_by_pid(flk->l_pid); 8558 if (lo != NULL) { 8559 cp = lo->rl_client; 8560 if (rfs4_lease_expired(cp)) { 8561 rfs4_lockowner_rele(lo); 8562 rfs4_dbe_hold(cp->rc_dbe); 8563 rfs4_client_close(cp); 8564 return (NFS4ERR_EXPIRED); 8565 } 8566 dp->owner.clientid = lo->rl_owner.clientid; 8567 len = lo->rl_owner.owner_len; 8568 dp->owner.owner_val = kmem_alloc(len, KM_SLEEP); 8569 bcopy(lo->rl_owner.owner_val, dp->owner.owner_val, len); 8570 dp->owner.owner_len = len; 8571 rfs4_lockowner_rele(lo); 8572 goto finish; 8573 } 8574 8575 /* 8576 * Its not a NFS4 lock. We take advantage that the upper 32 bits 8577 * of the client id contain the boot time for a NFS4 lock. So we 8578 * fabricate and identity by setting clientid to the sysid, and 8579 * the lock owner to the pid. 8580 */ 8581 dp->owner.clientid = flk->l_sysid; 8582 len = sizeof (pid_t); 8583 dp->owner.owner_len = len; 8584 dp->owner.owner_val = kmem_alloc(len, KM_SLEEP); 8585 bcopy(&flk->l_pid, dp->owner.owner_val, len); 8586 finish: 8587 dp->offset = flk->l_start; 8588 dp->length = flk->l_len; 8589 8590 if (flk->l_type == F_RDLCK) 8591 dp->locktype = READ_LT; 8592 else if (flk->l_type == F_WRLCK) 8593 dp->locktype = WRITE_LT; 8594 else 8595 return (NFS4ERR_INVAL); /* no mapping from POSIX ltype to v4 */ 8596 8597 return (NFS4_OK); 8598 } 8599 8600 /* 8601 * The NFSv4.0 LOCK operation does not support the blocking lock (at the 8602 * NFSv4.0 protocol level) so the client needs to resend the LOCK request in a 8603 * case the lock is denied by the NFSv4.0 server. NFSv4.0 clients are prepared 8604 * for that (obviously); they are sending the LOCK requests with some delays 8605 * between the attempts. See nfs4frlock() and nfs4_block_and_wait() for the 8606 * locking and delay implementation at the client side. 8607 * 8608 * To make the life of the clients easier, the NFSv4.0 server tries to do some 8609 * fast retries on its own (the for loop below) in a hope the lock will be 8610 * available soon. And if not, the client won't need to resend the LOCK 8611 * requests so fast to check the lock availability. This basically saves some 8612 * network traffic and tries to make sure the client gets the lock ASAP. 8613 */ 8614 static int 8615 setlock(vnode_t *vp, struct flock64 *flock, int flag, cred_t *cred) 8616 { 8617 int error; 8618 struct flock64 flk; 8619 int i; 8620 clock_t delaytime; 8621 int cmd; 8622 int spin_cnt = 0; 8623 8624 cmd = nbl_need_check(vp) ? F_SETLK_NBMAND : F_SETLK; 8625 retry: 8626 delaytime = MSEC_TO_TICK_ROUNDUP(rfs4_lock_delay); 8627 8628 for (i = 0; i < rfs4_maxlock_tries; i++) { 8629 LOCK_PRINT(rfs4_debug, "setlock", cmd, flock); 8630 error = VOP_FRLOCK(vp, cmd, 8631 flock, flag, (u_offset_t)0, NULL, cred, NULL); 8632 8633 if (error != EAGAIN && error != EACCES) 8634 break; 8635 8636 if (i < rfs4_maxlock_tries - 1) { 8637 delay(delaytime); 8638 delaytime *= 2; 8639 } 8640 } 8641 8642 if (error == EAGAIN || error == EACCES) { 8643 /* Get the owner of the lock */ 8644 flk = *flock; 8645 LOCK_PRINT(rfs4_debug, "setlock", F_GETLK, &flk); 8646 if (VOP_FRLOCK(vp, F_GETLK, &flk, flag, 0, NULL, cred, 8647 NULL) == 0) { 8648 /* 8649 * There's a race inherent in the current VOP_FRLOCK 8650 * design where: 8651 * a: "other guy" takes a lock that conflicts with a 8652 * lock we want 8653 * b: we attempt to take our lock (non-blocking) and 8654 * the attempt fails. 8655 * c: "other guy" releases the conflicting lock 8656 * d: we ask what lock conflicts with the lock we want, 8657 * getting F_UNLCK (no lock blocks us) 8658 * 8659 * If we retry the non-blocking lock attempt in this 8660 * case (restart at step 'b') there's some possibility 8661 * that many such attempts might fail. However a test 8662 * designed to actually provoke this race shows that 8663 * the vast majority of cases require no retry, and 8664 * only a few took as many as three retries. Here's 8665 * the test outcome: 8666 * 8667 * number of retries how many times we needed 8668 * that many retries 8669 * 0 79461 8670 * 1 862 8671 * 2 49 8672 * 3 5 8673 * 8674 * Given those empirical results, we arbitrarily limit 8675 * the retry count to ten. 8676 * 8677 * If we actually make to ten retries and give up, 8678 * nothing catastrophic happens, but we're unable to 8679 * return the information about the conflicting lock to 8680 * the NFS client. That's an acceptable trade off vs. 8681 * letting this retry loop run forever. 8682 */ 8683 if (flk.l_type == F_UNLCK) { 8684 if (spin_cnt++ < 10) { 8685 /* No longer locked, retry */ 8686 goto retry; 8687 } 8688 } else { 8689 *flock = flk; 8690 LOCK_PRINT(rfs4_debug, "setlock(blocking lock)", 8691 F_GETLK, &flk); 8692 } 8693 } 8694 } 8695 8696 return (error); 8697 } 8698 8699 /*ARGSUSED*/ 8700 static nfsstat4 8701 rfs4_do_lock(rfs4_lo_state_t *lsp, nfs_lock_type4 locktype, 8702 offset4 offset, length4 length, cred_t *cred, nfs_resop4 *resop) 8703 { 8704 nfsstat4 status; 8705 rfs4_lockowner_t *lo = lsp->rls_locker; 8706 rfs4_state_t *sp = lsp->rls_state; 8707 struct flock64 flock; 8708 int16_t ltype; 8709 int flag; 8710 int error; 8711 sysid_t sysid; 8712 LOCK4res *lres; 8713 vnode_t *vp; 8714 8715 if (rfs4_lease_expired(lo->rl_client)) { 8716 return (NFS4ERR_EXPIRED); 8717 } 8718 8719 if ((status = rfs4_client_sysid(lo->rl_client, &sysid)) != NFS4_OK) 8720 return (status); 8721 8722 /* Check for zero length. To lock to end of file use all ones for V4 */ 8723 if (length == 0) 8724 return (NFS4ERR_INVAL); 8725 else if (length == (length4)(~0)) 8726 length = 0; /* Posix to end of file */ 8727 8728 retry: 8729 rfs4_dbe_lock(sp->rs_dbe); 8730 if (sp->rs_closed == TRUE) { 8731 rfs4_dbe_unlock(sp->rs_dbe); 8732 return (NFS4ERR_OLD_STATEID); 8733 } 8734 8735 if (resop->resop != OP_LOCKU) { 8736 switch (locktype) { 8737 case READ_LT: 8738 case READW_LT: 8739 if ((sp->rs_share_access 8740 & OPEN4_SHARE_ACCESS_READ) == 0) { 8741 rfs4_dbe_unlock(sp->rs_dbe); 8742 8743 return (NFS4ERR_OPENMODE); 8744 } 8745 ltype = F_RDLCK; 8746 break; 8747 case WRITE_LT: 8748 case WRITEW_LT: 8749 if ((sp->rs_share_access 8750 & OPEN4_SHARE_ACCESS_WRITE) == 0) { 8751 rfs4_dbe_unlock(sp->rs_dbe); 8752 8753 return (NFS4ERR_OPENMODE); 8754 } 8755 ltype = F_WRLCK; 8756 break; 8757 } 8758 } else 8759 ltype = F_UNLCK; 8760 8761 flock.l_type = ltype; 8762 flock.l_whence = 0; /* SEEK_SET */ 8763 flock.l_start = offset; 8764 flock.l_len = length; 8765 flock.l_sysid = sysid; 8766 flock.l_pid = lsp->rls_locker->rl_pid; 8767 8768 /* Note that length4 is uint64_t but l_len and l_start are off64_t */ 8769 if (flock.l_len < 0 || flock.l_start < 0) { 8770 rfs4_dbe_unlock(sp->rs_dbe); 8771 return (NFS4ERR_INVAL); 8772 } 8773 8774 /* 8775 * N.B. FREAD has the same value as OPEN4_SHARE_ACCESS_READ and 8776 * FWRITE has the same value as OPEN4_SHARE_ACCESS_WRITE. 8777 */ 8778 flag = (int)sp->rs_share_access | F_REMOTELOCK; 8779 8780 vp = sp->rs_finfo->rf_vp; 8781 VN_HOLD(vp); 8782 8783 /* 8784 * We need to unlock sp before we call the underlying filesystem to 8785 * acquire the file lock. 8786 */ 8787 rfs4_dbe_unlock(sp->rs_dbe); 8788 8789 error = setlock(vp, &flock, flag, cred); 8790 8791 /* 8792 * Make sure the file is still open. In a case the file was closed in 8793 * the meantime, clean the lock we acquired using the setlock() call 8794 * above, and return the appropriate error. 8795 */ 8796 rfs4_dbe_lock(sp->rs_dbe); 8797 if (sp->rs_closed == TRUE) { 8798 cleanlocks(vp, lsp->rls_locker->rl_pid, sysid); 8799 rfs4_dbe_unlock(sp->rs_dbe); 8800 8801 VN_RELE(vp); 8802 8803 return (NFS4ERR_OLD_STATEID); 8804 } 8805 rfs4_dbe_unlock(sp->rs_dbe); 8806 8807 VN_RELE(vp); 8808 8809 if (error == 0) { 8810 rfs4_dbe_lock(lsp->rls_dbe); 8811 next_stateid(&lsp->rls_lockid); 8812 rfs4_dbe_unlock(lsp->rls_dbe); 8813 } 8814 8815 /* 8816 * N.B. We map error values to nfsv4 errors. This is differrent 8817 * than puterrno4 routine. 8818 */ 8819 switch (error) { 8820 case 0: 8821 status = NFS4_OK; 8822 break; 8823 case EAGAIN: 8824 case EACCES: /* Old value */ 8825 /* Can only get here if op is OP_LOCK */ 8826 ASSERT(resop->resop == OP_LOCK); 8827 lres = &resop->nfs_resop4_u.oplock; 8828 status = NFS4ERR_DENIED; 8829 if (lock_denied(&lres->LOCK4res_u.denied, &flock) 8830 == NFS4ERR_EXPIRED) 8831 goto retry; 8832 break; 8833 case ENOLCK: 8834 status = NFS4ERR_DELAY; 8835 break; 8836 case EOVERFLOW: 8837 status = NFS4ERR_INVAL; 8838 break; 8839 case EINVAL: 8840 status = NFS4ERR_NOTSUPP; 8841 break; 8842 default: 8843 status = NFS4ERR_SERVERFAULT; 8844 break; 8845 } 8846 8847 return (status); 8848 } 8849 8850 /*ARGSUSED*/ 8851 void 8852 rfs4_op_lock(nfs_argop4 *argop, nfs_resop4 *resop, 8853 struct svc_req *req, struct compound_state *cs) 8854 { 8855 LOCK4args *args = &argop->nfs_argop4_u.oplock; 8856 LOCK4res *resp = &resop->nfs_resop4_u.oplock; 8857 nfsstat4 status; 8858 stateid4 *stateid; 8859 rfs4_lockowner_t *lo; 8860 rfs4_client_t *cp; 8861 rfs4_state_t *sp = NULL; 8862 rfs4_lo_state_t *lsp = NULL; 8863 bool_t ls_sw_held = FALSE; 8864 bool_t create = TRUE; 8865 bool_t lcreate = TRUE; 8866 bool_t dup_lock = FALSE; 8867 int rc; 8868 8869 DTRACE_NFSV4_2(op__lock__start, struct compound_state *, cs, 8870 LOCK4args *, args); 8871 8872 if (cs->vp == NULL) { 8873 *cs->statusp = resp->status = NFS4ERR_NOFILEHANDLE; 8874 DTRACE_NFSV4_2(op__lock__done, struct compound_state *, 8875 cs, LOCK4res *, resp); 8876 return; 8877 } 8878 8879 if (args->locker.new_lock_owner) { 8880 /* Create a new lockowner for this instance */ 8881 open_to_lock_owner4 *olo = &args->locker.locker4_u.open_owner; 8882 8883 NFS4_DEBUG(rfs4_debug, (CE_NOTE, "Creating new lock owner")); 8884 8885 stateid = &olo->open_stateid; 8886 status = rfs4_get_state(stateid, &sp, RFS4_DBS_VALID); 8887 if (status != NFS4_OK) { 8888 NFS4_DEBUG(rfs4_debug, 8889 (CE_NOTE, "Get state failed in lock %d", status)); 8890 *cs->statusp = resp->status = status; 8891 DTRACE_NFSV4_2(op__lock__done, struct compound_state *, 8892 cs, LOCK4res *, resp); 8893 return; 8894 } 8895 8896 /* Ensure specified filehandle matches */ 8897 if (cs->vp != sp->rs_finfo->rf_vp) { 8898 rfs4_state_rele(sp); 8899 *cs->statusp = resp->status = NFS4ERR_BAD_STATEID; 8900 DTRACE_NFSV4_2(op__lock__done, struct compound_state *, 8901 cs, LOCK4res *, resp); 8902 return; 8903 } 8904 8905 /* hold off other access to open_owner while we tinker */ 8906 rfs4_sw_enter(&sp->rs_owner->ro_sw); 8907 8908 switch (rc = rfs4_check_stateid_seqid(sp, stateid)) { 8909 case NFS4_CHECK_STATEID_OLD: 8910 *cs->statusp = resp->status = NFS4ERR_OLD_STATEID; 8911 goto end; 8912 case NFS4_CHECK_STATEID_BAD: 8913 *cs->statusp = resp->status = NFS4ERR_BAD_STATEID; 8914 goto end; 8915 case NFS4_CHECK_STATEID_EXPIRED: 8916 *cs->statusp = resp->status = NFS4ERR_EXPIRED; 8917 goto end; 8918 case NFS4_CHECK_STATEID_UNCONFIRMED: 8919 *cs->statusp = resp->status = NFS4ERR_BAD_STATEID; 8920 goto end; 8921 case NFS4_CHECK_STATEID_CLOSED: 8922 *cs->statusp = resp->status = NFS4ERR_OLD_STATEID; 8923 goto end; 8924 case NFS4_CHECK_STATEID_OKAY: 8925 case NFS4_CHECK_STATEID_REPLAY: 8926 switch (rfs4_check_olo_seqid(olo->open_seqid, 8927 sp->rs_owner, resop)) { 8928 case NFS4_CHKSEQ_OKAY: 8929 if (rc == NFS4_CHECK_STATEID_OKAY) 8930 break; 8931 /* 8932 * This is replayed stateid; if seqid 8933 * matches next expected, then client 8934 * is using wrong seqid. 8935 */ 8936 /* FALLTHROUGH */ 8937 case NFS4_CHKSEQ_BAD: 8938 *cs->statusp = resp->status = NFS4ERR_BAD_SEQID; 8939 goto end; 8940 case NFS4_CHKSEQ_REPLAY: 8941 /* This is a duplicate LOCK request */ 8942 dup_lock = TRUE; 8943 8944 /* 8945 * For a duplicate we do not want to 8946 * create a new lockowner as it should 8947 * already exist. 8948 * Turn off the lockowner create flag. 8949 */ 8950 lcreate = FALSE; 8951 } 8952 break; 8953 } 8954 8955 lo = rfs4_findlockowner(&olo->lock_owner, &lcreate); 8956 if (lo == NULL) { 8957 NFS4_DEBUG(rfs4_debug, 8958 (CE_NOTE, "rfs4_op_lock: no lock owner")); 8959 *cs->statusp = resp->status = NFS4ERR_RESOURCE; 8960 goto end; 8961 } 8962 8963 lsp = rfs4_findlo_state_by_owner(lo, sp, &create); 8964 if (lsp == NULL) { 8965 rfs4_update_lease(sp->rs_owner->ro_client); 8966 /* 8967 * Only update theh open_seqid if this is not 8968 * a duplicate request 8969 */ 8970 if (dup_lock == FALSE) { 8971 rfs4_update_open_sequence(sp->rs_owner); 8972 } 8973 8974 NFS4_DEBUG(rfs4_debug, 8975 (CE_NOTE, "rfs4_op_lock: no state")); 8976 *cs->statusp = resp->status = NFS4ERR_SERVERFAULT; 8977 rfs4_update_open_resp(sp->rs_owner, resop, NULL); 8978 rfs4_lockowner_rele(lo); 8979 goto end; 8980 } 8981 8982 /* 8983 * This is the new_lock_owner branch and the client is 8984 * supposed to be associating a new lock_owner with 8985 * the open file at this point. If we find that a 8986 * lock_owner/state association already exists and a 8987 * successful LOCK request was returned to the client, 8988 * an error is returned to the client since this is 8989 * not appropriate. The client should be using the 8990 * existing lock_owner branch. 8991 */ 8992 if (dup_lock == FALSE && create == FALSE) { 8993 if (lsp->rls_lock_completed == TRUE) { 8994 *cs->statusp = 8995 resp->status = NFS4ERR_BAD_SEQID; 8996 rfs4_lockowner_rele(lo); 8997 goto end; 8998 } 8999 } 9000 9001 rfs4_update_lease(sp->rs_owner->ro_client); 9002 9003 /* 9004 * Only update theh open_seqid if this is not 9005 * a duplicate request 9006 */ 9007 if (dup_lock == FALSE) { 9008 rfs4_update_open_sequence(sp->rs_owner); 9009 } 9010 9011 /* 9012 * If this is a duplicate lock request, just copy the 9013 * previously saved reply and return. 9014 */ 9015 if (dup_lock == TRUE) { 9016 /* verify that lock_seqid's match */ 9017 if (lsp->rls_seqid != olo->lock_seqid) { 9018 NFS4_DEBUG(rfs4_debug, 9019 (CE_NOTE, "rfs4_op_lock: Dup-Lock seqid bad" 9020 "lsp->seqid=%d old->seqid=%d", 9021 lsp->rls_seqid, olo->lock_seqid)); 9022 *cs->statusp = resp->status = NFS4ERR_BAD_SEQID; 9023 } else { 9024 rfs4_copy_reply(resop, &lsp->rls_reply); 9025 /* 9026 * Make sure to copy the just 9027 * retrieved reply status into the 9028 * overall compound status 9029 */ 9030 *cs->statusp = resp->status; 9031 } 9032 rfs4_lockowner_rele(lo); 9033 goto end; 9034 } 9035 9036 rfs4_dbe_lock(lsp->rls_dbe); 9037 9038 /* Make sure to update the lock sequence id */ 9039 lsp->rls_seqid = olo->lock_seqid; 9040 9041 NFS4_DEBUG(rfs4_debug, 9042 (CE_NOTE, "Lock seqid established as %d", lsp->rls_seqid)); 9043 9044 /* 9045 * This is used to signify the newly created lockowner 9046 * stateid and its sequence number. The checks for 9047 * sequence number and increment don't occur on the 9048 * very first lock request for a lockowner. 9049 */ 9050 lsp->rls_skip_seqid_check = TRUE; 9051 9052 /* hold off other access to lsp while we tinker */ 9053 rfs4_sw_enter(&lsp->rls_sw); 9054 ls_sw_held = TRUE; 9055 9056 rfs4_dbe_unlock(lsp->rls_dbe); 9057 9058 rfs4_lockowner_rele(lo); 9059 } else { 9060 stateid = &args->locker.locker4_u.lock_owner.lock_stateid; 9061 /* get lsp and hold the lock on the underlying file struct */ 9062 if ((status = rfs4_get_lo_state(stateid, &lsp, TRUE)) 9063 != NFS4_OK) { 9064 *cs->statusp = resp->status = status; 9065 DTRACE_NFSV4_2(op__lock__done, struct compound_state *, 9066 cs, LOCK4res *, resp); 9067 return; 9068 } 9069 create = FALSE; /* We didn't create lsp */ 9070 9071 /* Ensure specified filehandle matches */ 9072 if (cs->vp != lsp->rls_state->rs_finfo->rf_vp) { 9073 rfs4_lo_state_rele(lsp, TRUE); 9074 *cs->statusp = resp->status = NFS4ERR_BAD_STATEID; 9075 DTRACE_NFSV4_2(op__lock__done, struct compound_state *, 9076 cs, LOCK4res *, resp); 9077 return; 9078 } 9079 9080 /* hold off other access to lsp while we tinker */ 9081 rfs4_sw_enter(&lsp->rls_sw); 9082 ls_sw_held = TRUE; 9083 9084 switch (rfs4_check_lo_stateid_seqid(lsp, stateid)) { 9085 /* 9086 * The stateid looks like it was okay (expected to be 9087 * the next one) 9088 */ 9089 case NFS4_CHECK_STATEID_OKAY: 9090 /* 9091 * The sequence id is now checked. Determine 9092 * if this is a replay or if it is in the 9093 * expected (next) sequence. In the case of a 9094 * replay, there are two replay conditions 9095 * that may occur. The first is the normal 9096 * condition where a LOCK is done with a 9097 * NFS4_OK response and the stateid is 9098 * updated. That case is handled below when 9099 * the stateid is identified as a REPLAY. The 9100 * second is the case where an error is 9101 * returned, like NFS4ERR_DENIED, and the 9102 * sequence number is updated but the stateid 9103 * is not updated. This second case is dealt 9104 * with here. So it may seem odd that the 9105 * stateid is okay but the sequence id is a 9106 * replay but it is okay. 9107 */ 9108 switch (rfs4_check_lock_seqid( 9109 args->locker.locker4_u.lock_owner.lock_seqid, 9110 lsp, resop)) { 9111 case NFS4_CHKSEQ_REPLAY: 9112 if (resp->status != NFS4_OK) { 9113 /* 9114 * Here is our replay and need 9115 * to verify that the last 9116 * response was an error. 9117 */ 9118 *cs->statusp = resp->status; 9119 goto end; 9120 } 9121 /* 9122 * This is done since the sequence id 9123 * looked like a replay but it didn't 9124 * pass our check so a BAD_SEQID is 9125 * returned as a result. 9126 */ 9127 /*FALLTHROUGH*/ 9128 case NFS4_CHKSEQ_BAD: 9129 *cs->statusp = resp->status = NFS4ERR_BAD_SEQID; 9130 goto end; 9131 case NFS4_CHKSEQ_OKAY: 9132 /* Everything looks okay move ahead */ 9133 break; 9134 } 9135 break; 9136 case NFS4_CHECK_STATEID_OLD: 9137 *cs->statusp = resp->status = NFS4ERR_OLD_STATEID; 9138 goto end; 9139 case NFS4_CHECK_STATEID_BAD: 9140 *cs->statusp = resp->status = NFS4ERR_BAD_STATEID; 9141 goto end; 9142 case NFS4_CHECK_STATEID_EXPIRED: 9143 *cs->statusp = resp->status = NFS4ERR_EXPIRED; 9144 goto end; 9145 case NFS4_CHECK_STATEID_CLOSED: 9146 *cs->statusp = resp->status = NFS4ERR_OLD_STATEID; 9147 goto end; 9148 case NFS4_CHECK_STATEID_REPLAY: 9149 switch (rfs4_check_lock_seqid( 9150 args->locker.locker4_u.lock_owner.lock_seqid, 9151 lsp, resop)) { 9152 case NFS4_CHKSEQ_OKAY: 9153 /* 9154 * This is a replayed stateid; if 9155 * seqid matches the next expected, 9156 * then client is using wrong seqid. 9157 */ 9158 case NFS4_CHKSEQ_BAD: 9159 *cs->statusp = resp->status = NFS4ERR_BAD_SEQID; 9160 goto end; 9161 case NFS4_CHKSEQ_REPLAY: 9162 rfs4_update_lease(lsp->rls_locker->rl_client); 9163 *cs->statusp = status = resp->status; 9164 goto end; 9165 } 9166 break; 9167 default: 9168 ASSERT(FALSE); 9169 break; 9170 } 9171 9172 rfs4_update_lock_sequence(lsp); 9173 rfs4_update_lease(lsp->rls_locker->rl_client); 9174 } 9175 9176 /* 9177 * NFS4 only allows locking on regular files, so 9178 * verify type of object. 9179 */ 9180 if (cs->vp->v_type != VREG) { 9181 if (cs->vp->v_type == VDIR) 9182 status = NFS4ERR_ISDIR; 9183 else 9184 status = NFS4ERR_INVAL; 9185 goto out; 9186 } 9187 9188 cp = lsp->rls_state->rs_owner->ro_client; 9189 9190 if (rfs4_clnt_in_grace(cp) && !args->reclaim) { 9191 status = NFS4ERR_GRACE; 9192 goto out; 9193 } 9194 9195 if (rfs4_clnt_in_grace(cp) && args->reclaim && !cp->rc_can_reclaim) { 9196 status = NFS4ERR_NO_GRACE; 9197 goto out; 9198 } 9199 9200 if (!rfs4_clnt_in_grace(cp) && args->reclaim) { 9201 status = NFS4ERR_NO_GRACE; 9202 goto out; 9203 } 9204 9205 if (lsp->rls_state->rs_finfo->rf_dinfo.rd_dtype == OPEN_DELEGATE_WRITE) 9206 cs->deleg = TRUE; 9207 9208 status = rfs4_do_lock(lsp, args->locktype, 9209 args->offset, args->length, cs->cr, resop); 9210 9211 out: 9212 lsp->rls_skip_seqid_check = FALSE; 9213 9214 *cs->statusp = resp->status = status; 9215 9216 if (status == NFS4_OK) { 9217 resp->LOCK4res_u.lock_stateid = lsp->rls_lockid.stateid; 9218 lsp->rls_lock_completed = TRUE; 9219 } 9220 /* 9221 * Only update the "OPEN" response here if this was a new 9222 * lock_owner 9223 */ 9224 if (sp) 9225 rfs4_update_open_resp(sp->rs_owner, resop, NULL); 9226 9227 rfs4_update_lock_resp(lsp, resop); 9228 9229 end: 9230 if (lsp) { 9231 if (ls_sw_held) 9232 rfs4_sw_exit(&lsp->rls_sw); 9233 /* 9234 * If an sp obtained, then the lsp does not represent 9235 * a lock on the file struct. 9236 */ 9237 if (sp != NULL) 9238 rfs4_lo_state_rele(lsp, FALSE); 9239 else 9240 rfs4_lo_state_rele(lsp, TRUE); 9241 } 9242 if (sp) { 9243 rfs4_sw_exit(&sp->rs_owner->ro_sw); 9244 rfs4_state_rele(sp); 9245 } 9246 9247 DTRACE_NFSV4_2(op__lock__done, struct compound_state *, cs, 9248 LOCK4res *, resp); 9249 } 9250 9251 /* free function for LOCK/LOCKT */ 9252 static void 9253 lock_denied_free(nfs_resop4 *resop) 9254 { 9255 LOCK4denied *dp = NULL; 9256 9257 switch (resop->resop) { 9258 case OP_LOCK: 9259 if (resop->nfs_resop4_u.oplock.status == NFS4ERR_DENIED) 9260 dp = &resop->nfs_resop4_u.oplock.LOCK4res_u.denied; 9261 break; 9262 case OP_LOCKT: 9263 if (resop->nfs_resop4_u.oplockt.status == NFS4ERR_DENIED) 9264 dp = &resop->nfs_resop4_u.oplockt.denied; 9265 break; 9266 default: 9267 break; 9268 } 9269 9270 if (dp) 9271 kmem_free(dp->owner.owner_val, dp->owner.owner_len); 9272 } 9273 9274 /*ARGSUSED*/ 9275 void 9276 rfs4_op_locku(nfs_argop4 *argop, nfs_resop4 *resop, 9277 struct svc_req *req, struct compound_state *cs) 9278 { 9279 LOCKU4args *args = &argop->nfs_argop4_u.oplocku; 9280 LOCKU4res *resp = &resop->nfs_resop4_u.oplocku; 9281 nfsstat4 status; 9282 stateid4 *stateid = &args->lock_stateid; 9283 rfs4_lo_state_t *lsp; 9284 9285 DTRACE_NFSV4_2(op__locku__start, struct compound_state *, cs, 9286 LOCKU4args *, args); 9287 9288 if (cs->vp == NULL) { 9289 *cs->statusp = resp->status = NFS4ERR_NOFILEHANDLE; 9290 DTRACE_NFSV4_2(op__locku__done, struct compound_state *, cs, 9291 LOCKU4res *, resp); 9292 return; 9293 } 9294 9295 if ((status = rfs4_get_lo_state(stateid, &lsp, TRUE)) != NFS4_OK) { 9296 *cs->statusp = resp->status = status; 9297 DTRACE_NFSV4_2(op__locku__done, struct compound_state *, cs, 9298 LOCKU4res *, resp); 9299 return; 9300 } 9301 9302 /* Ensure specified filehandle matches */ 9303 if (cs->vp != lsp->rls_state->rs_finfo->rf_vp) { 9304 rfs4_lo_state_rele(lsp, TRUE); 9305 *cs->statusp = resp->status = NFS4ERR_BAD_STATEID; 9306 DTRACE_NFSV4_2(op__locku__done, struct compound_state *, cs, 9307 LOCKU4res *, resp); 9308 return; 9309 } 9310 9311 /* hold off other access to lsp while we tinker */ 9312 rfs4_sw_enter(&lsp->rls_sw); 9313 9314 switch (rfs4_check_lo_stateid_seqid(lsp, stateid)) { 9315 case NFS4_CHECK_STATEID_OKAY: 9316 if (rfs4_check_lock_seqid(args->seqid, lsp, resop) 9317 != NFS4_CHKSEQ_OKAY) { 9318 *cs->statusp = resp->status = NFS4ERR_BAD_SEQID; 9319 goto end; 9320 } 9321 break; 9322 case NFS4_CHECK_STATEID_OLD: 9323 *cs->statusp = resp->status = NFS4ERR_OLD_STATEID; 9324 goto end; 9325 case NFS4_CHECK_STATEID_BAD: 9326 *cs->statusp = resp->status = NFS4ERR_BAD_STATEID; 9327 goto end; 9328 case NFS4_CHECK_STATEID_EXPIRED: 9329 *cs->statusp = resp->status = NFS4ERR_EXPIRED; 9330 goto end; 9331 case NFS4_CHECK_STATEID_CLOSED: 9332 *cs->statusp = resp->status = NFS4ERR_OLD_STATEID; 9333 goto end; 9334 case NFS4_CHECK_STATEID_REPLAY: 9335 switch (rfs4_check_lock_seqid(args->seqid, lsp, resop)) { 9336 case NFS4_CHKSEQ_OKAY: 9337 /* 9338 * This is a replayed stateid; if 9339 * seqid matches the next expected, 9340 * then client is using wrong seqid. 9341 */ 9342 case NFS4_CHKSEQ_BAD: 9343 *cs->statusp = resp->status = NFS4ERR_BAD_SEQID; 9344 goto end; 9345 case NFS4_CHKSEQ_REPLAY: 9346 rfs4_update_lease(lsp->rls_locker->rl_client); 9347 *cs->statusp = status = resp->status; 9348 goto end; 9349 } 9350 break; 9351 default: 9352 ASSERT(FALSE); 9353 break; 9354 } 9355 9356 rfs4_update_lock_sequence(lsp); 9357 rfs4_update_lease(lsp->rls_locker->rl_client); 9358 9359 /* 9360 * NFS4 only allows locking on regular files, so 9361 * verify type of object. 9362 */ 9363 if (cs->vp->v_type != VREG) { 9364 if (cs->vp->v_type == VDIR) 9365 status = NFS4ERR_ISDIR; 9366 else 9367 status = NFS4ERR_INVAL; 9368 goto out; 9369 } 9370 9371 if (rfs4_clnt_in_grace(lsp->rls_state->rs_owner->ro_client)) { 9372 status = NFS4ERR_GRACE; 9373 goto out; 9374 } 9375 9376 status = rfs4_do_lock(lsp, args->locktype, 9377 args->offset, args->length, cs->cr, resop); 9378 9379 out: 9380 *cs->statusp = resp->status = status; 9381 9382 if (status == NFS4_OK) 9383 resp->lock_stateid = lsp->rls_lockid.stateid; 9384 9385 rfs4_update_lock_resp(lsp, resop); 9386 9387 end: 9388 rfs4_sw_exit(&lsp->rls_sw); 9389 rfs4_lo_state_rele(lsp, TRUE); 9390 9391 DTRACE_NFSV4_2(op__locku__done, struct compound_state *, cs, 9392 LOCKU4res *, resp); 9393 } 9394 9395 /* 9396 * LOCKT is a best effort routine, the client can not be guaranteed that 9397 * the status return is still in effect by the time the reply is received. 9398 * They are numerous race conditions in this routine, but we are not required 9399 * and can not be accurate. 9400 */ 9401 /*ARGSUSED*/ 9402 void 9403 rfs4_op_lockt(nfs_argop4 *argop, nfs_resop4 *resop, 9404 struct svc_req *req, struct compound_state *cs) 9405 { 9406 LOCKT4args *args = &argop->nfs_argop4_u.oplockt; 9407 LOCKT4res *resp = &resop->nfs_resop4_u.oplockt; 9408 rfs4_lockowner_t *lo; 9409 rfs4_client_t *cp; 9410 bool_t create = FALSE; 9411 struct flock64 flk; 9412 int error; 9413 int flag = FREAD | FWRITE; 9414 int ltype; 9415 length4 posix_length; 9416 sysid_t sysid; 9417 pid_t pid; 9418 9419 DTRACE_NFSV4_2(op__lockt__start, struct compound_state *, cs, 9420 LOCKT4args *, args); 9421 9422 if (cs->vp == NULL) { 9423 *cs->statusp = resp->status = NFS4ERR_NOFILEHANDLE; 9424 goto out; 9425 } 9426 9427 /* 9428 * NFS4 only allows locking on regular files, so 9429 * verify type of object. 9430 */ 9431 if (cs->vp->v_type != VREG) { 9432 if (cs->vp->v_type == VDIR) 9433 *cs->statusp = resp->status = NFS4ERR_ISDIR; 9434 else 9435 *cs->statusp = resp->status = NFS4ERR_INVAL; 9436 goto out; 9437 } 9438 9439 /* 9440 * Check out the clientid to ensure the server knows about it 9441 * so that we correctly inform the client of a server reboot. 9442 */ 9443 if ((cp = rfs4_findclient_by_id(args->owner.clientid, FALSE)) 9444 == NULL) { 9445 *cs->statusp = resp->status = 9446 rfs4_check_clientid(&args->owner.clientid, 0); 9447 goto out; 9448 } 9449 if (rfs4_lease_expired(cp)) { 9450 rfs4_client_close(cp); 9451 /* 9452 * Protocol doesn't allow returning NFS4ERR_STALE as 9453 * other operations do on this check so STALE_CLIENTID 9454 * is returned instead 9455 */ 9456 *cs->statusp = resp->status = NFS4ERR_STALE_CLIENTID; 9457 goto out; 9458 } 9459 9460 if (rfs4_clnt_in_grace(cp) && !(cp->rc_can_reclaim)) { 9461 *cs->statusp = resp->status = NFS4ERR_GRACE; 9462 rfs4_client_rele(cp); 9463 goto out; 9464 } 9465 rfs4_client_rele(cp); 9466 9467 resp->status = NFS4_OK; 9468 9469 switch (args->locktype) { 9470 case READ_LT: 9471 case READW_LT: 9472 ltype = F_RDLCK; 9473 break; 9474 case WRITE_LT: 9475 case WRITEW_LT: 9476 ltype = F_WRLCK; 9477 break; 9478 } 9479 9480 posix_length = args->length; 9481 /* Check for zero length. To lock to end of file use all ones for V4 */ 9482 if (posix_length == 0) { 9483 *cs->statusp = resp->status = NFS4ERR_INVAL; 9484 goto out; 9485 } else if (posix_length == (length4)(~0)) { 9486 posix_length = 0; /* Posix to end of file */ 9487 } 9488 9489 /* Find or create a lockowner */ 9490 lo = rfs4_findlockowner(&args->owner, &create); 9491 9492 if (lo) { 9493 pid = lo->rl_pid; 9494 if ((resp->status = 9495 rfs4_client_sysid(lo->rl_client, &sysid)) != NFS4_OK) 9496 goto err; 9497 } else { 9498 pid = 0; 9499 sysid = lockt_sysid; 9500 } 9501 retry: 9502 flk.l_type = ltype; 9503 flk.l_whence = 0; /* SEEK_SET */ 9504 flk.l_start = args->offset; 9505 flk.l_len = posix_length; 9506 flk.l_sysid = sysid; 9507 flk.l_pid = pid; 9508 flag |= F_REMOTELOCK; 9509 9510 LOCK_PRINT(rfs4_debug, "rfs4_op_lockt", F_GETLK, &flk); 9511 9512 /* Note that length4 is uint64_t but l_len and l_start are off64_t */ 9513 if (flk.l_len < 0 || flk.l_start < 0) { 9514 resp->status = NFS4ERR_INVAL; 9515 goto err; 9516 } 9517 error = VOP_FRLOCK(cs->vp, F_GETLK, &flk, flag, (u_offset_t)0, 9518 NULL, cs->cr, NULL); 9519 9520 /* 9521 * N.B. We map error values to nfsv4 errors. This is differrent 9522 * than puterrno4 routine. 9523 */ 9524 switch (error) { 9525 case 0: 9526 if (flk.l_type == F_UNLCK) 9527 resp->status = NFS4_OK; 9528 else { 9529 if (lock_denied(&resp->denied, &flk) == NFS4ERR_EXPIRED) 9530 goto retry; 9531 resp->status = NFS4ERR_DENIED; 9532 } 9533 break; 9534 case EOVERFLOW: 9535 resp->status = NFS4ERR_INVAL; 9536 break; 9537 case EINVAL: 9538 resp->status = NFS4ERR_NOTSUPP; 9539 break; 9540 default: 9541 cmn_err(CE_WARN, "rfs4_op_lockt: unexpected errno (%d)", 9542 error); 9543 resp->status = NFS4ERR_SERVERFAULT; 9544 break; 9545 } 9546 9547 err: 9548 if (lo) 9549 rfs4_lockowner_rele(lo); 9550 *cs->statusp = resp->status; 9551 out: 9552 DTRACE_NFSV4_2(op__lockt__done, struct compound_state *, cs, 9553 LOCKT4res *, resp); 9554 } 9555 9556 int 9557 rfs4_share(rfs4_state_t *sp, uint32_t access, uint32_t deny) 9558 { 9559 int err; 9560 int cmd; 9561 vnode_t *vp; 9562 struct shrlock shr; 9563 struct shr_locowner shr_loco; 9564 int fflags = 0; 9565 9566 ASSERT(rfs4_dbe_islocked(sp->rs_dbe)); 9567 ASSERT(sp->rs_owner->ro_client->rc_sysidt != LM_NOSYSID); 9568 9569 if (sp->rs_closed) 9570 return (NFS4ERR_OLD_STATEID); 9571 9572 vp = sp->rs_finfo->rf_vp; 9573 ASSERT(vp); 9574 9575 shr.s_access = shr.s_deny = 0; 9576 9577 if (access & OPEN4_SHARE_ACCESS_READ) { 9578 fflags |= FREAD; 9579 shr.s_access |= F_RDACC; 9580 } 9581 if (access & OPEN4_SHARE_ACCESS_WRITE) { 9582 fflags |= FWRITE; 9583 shr.s_access |= F_WRACC; 9584 } 9585 ASSERT(shr.s_access); 9586 9587 if (deny & OPEN4_SHARE_DENY_READ) 9588 shr.s_deny |= F_RDDNY; 9589 if (deny & OPEN4_SHARE_DENY_WRITE) 9590 shr.s_deny |= F_WRDNY; 9591 9592 shr.s_pid = rfs4_dbe_getid(sp->rs_owner->ro_dbe); 9593 shr.s_sysid = sp->rs_owner->ro_client->rc_sysidt; 9594 shr_loco.sl_pid = shr.s_pid; 9595 shr_loco.sl_id = shr.s_sysid; 9596 shr.s_owner = (caddr_t)&shr_loco; 9597 shr.s_own_len = sizeof (shr_loco); 9598 9599 cmd = nbl_need_check(vp) ? F_SHARE_NBMAND : F_SHARE; 9600 9601 err = VOP_SHRLOCK(vp, cmd, &shr, fflags, CRED(), NULL); 9602 if (err != 0) { 9603 if (err == EAGAIN) 9604 err = NFS4ERR_SHARE_DENIED; 9605 else 9606 err = puterrno4(err); 9607 return (err); 9608 } 9609 9610 sp->rs_share_access |= access; 9611 sp->rs_share_deny |= deny; 9612 9613 return (0); 9614 } 9615 9616 int 9617 rfs4_unshare(rfs4_state_t *sp) 9618 { 9619 int err; 9620 struct shrlock shr; 9621 struct shr_locowner shr_loco; 9622 9623 ASSERT(rfs4_dbe_islocked(sp->rs_dbe)); 9624 9625 if (sp->rs_closed || sp->rs_share_access == 0) 9626 return (0); 9627 9628 ASSERT(sp->rs_owner->ro_client->rc_sysidt != LM_NOSYSID); 9629 ASSERT(sp->rs_finfo->rf_vp); 9630 9631 shr.s_access = shr.s_deny = 0; 9632 shr.s_pid = rfs4_dbe_getid(sp->rs_owner->ro_dbe); 9633 shr.s_sysid = sp->rs_owner->ro_client->rc_sysidt; 9634 shr_loco.sl_pid = shr.s_pid; 9635 shr_loco.sl_id = shr.s_sysid; 9636 shr.s_owner = (caddr_t)&shr_loco; 9637 shr.s_own_len = sizeof (shr_loco); 9638 9639 err = VOP_SHRLOCK(sp->rs_finfo->rf_vp, F_UNSHARE, &shr, 0, CRED(), 9640 NULL); 9641 if (err != 0) { 9642 err = puterrno4(err); 9643 return (err); 9644 } 9645 9646 sp->rs_share_access = 0; 9647 sp->rs_share_deny = 0; 9648 9649 return (0); 9650 9651 } 9652 9653 static int 9654 rdma_setup_read_data4(READ4args *args, READ4res *rok) 9655 { 9656 struct clist *wcl; 9657 count4 count = rok->data_len; 9658 int wlist_len; 9659 9660 wcl = args->wlist; 9661 if (rdma_setup_read_chunks(wcl, count, &wlist_len) == FALSE) { 9662 return (FALSE); 9663 } 9664 wcl = args->wlist; 9665 rok->wlist_len = wlist_len; 9666 rok->wlist = wcl; 9667 return (TRUE); 9668 } 9669 9670 /* tunable to disable server referrals */ 9671 int rfs4_no_referrals = 0; 9672 9673 /* 9674 * Find an NFS record in reparse point data. 9675 * Returns 0 for success and <0 or an errno value on failure. 9676 */ 9677 int 9678 vn_find_nfs_record(vnode_t *vp, nvlist_t **nvlp, char **svcp, char **datap) 9679 { 9680 int err; 9681 char *stype, *val; 9682 nvlist_t *nvl; 9683 nvpair_t *curr; 9684 9685 if ((nvl = reparse_init()) == NULL) 9686 return (-1); 9687 9688 if ((err = reparse_vnode_parse(vp, nvl)) != 0) { 9689 reparse_free(nvl); 9690 return (err); 9691 } 9692 9693 curr = NULL; 9694 while ((curr = nvlist_next_nvpair(nvl, curr)) != NULL) { 9695 if ((stype = nvpair_name(curr)) == NULL) { 9696 reparse_free(nvl); 9697 return (-2); 9698 } 9699 if (strncasecmp(stype, "NFS", 3) == 0) 9700 break; 9701 } 9702 9703 if ((curr == NULL) || 9704 (nvpair_value_string(curr, &val))) { 9705 reparse_free(nvl); 9706 return (-3); 9707 } 9708 *nvlp = nvl; 9709 *svcp = stype; 9710 *datap = val; 9711 return (0); 9712 } 9713 9714 int 9715 vn_is_nfs_reparse(vnode_t *vp, cred_t *cr) 9716 { 9717 nvlist_t *nvl; 9718 char *s, *d; 9719 9720 if (rfs4_no_referrals != 0) 9721 return (B_FALSE); 9722 9723 if (vn_is_reparse(vp, cr, NULL) == B_FALSE) 9724 return (B_FALSE); 9725 9726 if (vn_find_nfs_record(vp, &nvl, &s, &d) != 0) 9727 return (B_FALSE); 9728 9729 reparse_free(nvl); 9730 9731 return (B_TRUE); 9732 } 9733 9734 /* 9735 * There is a user-level copy of this routine in ref_subr.c. 9736 * Changes should be kept in sync. 9737 */ 9738 static int 9739 nfs4_create_components(char *path, component4 *comp4) 9740 { 9741 int slen, plen, ncomp; 9742 char *ori_path, *nxtc, buf[MAXNAMELEN]; 9743 9744 if (path == NULL) 9745 return (0); 9746 9747 plen = strlen(path) + 1; /* include the terminator */ 9748 ori_path = path; 9749 ncomp = 0; 9750 9751 /* count number of components in the path */ 9752 for (nxtc = path; nxtc < ori_path + plen; nxtc++) { 9753 if (*nxtc == '/' || *nxtc == '\0' || *nxtc == '\n') { 9754 if ((slen = nxtc - path) == 0) { 9755 path = nxtc + 1; 9756 continue; 9757 } 9758 9759 if (comp4 != NULL) { 9760 bcopy(path, buf, slen); 9761 buf[slen] = '\0'; 9762 (void) str_to_utf8(buf, &comp4[ncomp]); 9763 } 9764 9765 ncomp++; /* 1 valid component */ 9766 path = nxtc + 1; 9767 } 9768 if (*nxtc == '\0' || *nxtc == '\n') 9769 break; 9770 } 9771 9772 return (ncomp); 9773 } 9774 9775 /* 9776 * There is a user-level copy of this routine in ref_subr.c. 9777 * Changes should be kept in sync. 9778 */ 9779 static int 9780 make_pathname4(char *path, pathname4 *pathname) 9781 { 9782 int ncomp; 9783 component4 *comp4; 9784 9785 if (pathname == NULL) 9786 return (0); 9787 9788 if (path == NULL) { 9789 pathname->pathname4_val = NULL; 9790 pathname->pathname4_len = 0; 9791 return (0); 9792 } 9793 9794 /* count number of components to alloc buffer */ 9795 if ((ncomp = nfs4_create_components(path, NULL)) == 0) { 9796 pathname->pathname4_val = NULL; 9797 pathname->pathname4_len = 0; 9798 return (0); 9799 } 9800 comp4 = kmem_zalloc(ncomp * sizeof (component4), KM_SLEEP); 9801 9802 /* copy components into allocated buffer */ 9803 ncomp = nfs4_create_components(path, comp4); 9804 9805 pathname->pathname4_val = comp4; 9806 pathname->pathname4_len = ncomp; 9807 9808 return (ncomp); 9809 } 9810 9811 #define xdr_fs_locations4 xdr_fattr4_fs_locations 9812 9813 fs_locations4 * 9814 fetch_referral(vnode_t *vp, cred_t *cr) 9815 { 9816 nvlist_t *nvl; 9817 char *stype, *sdata; 9818 fs_locations4 *result; 9819 char buf[1024]; 9820 size_t bufsize; 9821 XDR xdr; 9822 int err; 9823 9824 /* 9825 * Check attrs to ensure it's a reparse point 9826 */ 9827 if (vn_is_reparse(vp, cr, NULL) == B_FALSE) 9828 return (NULL); 9829 9830 /* 9831 * Look for an NFS record and get the type and data 9832 */ 9833 if (vn_find_nfs_record(vp, &nvl, &stype, &sdata) != 0) 9834 return (NULL); 9835 9836 /* 9837 * With the type and data, upcall to get the referral 9838 */ 9839 bufsize = sizeof (buf); 9840 bzero(buf, sizeof (buf)); 9841 err = reparse_kderef((const char *)stype, (const char *)sdata, 9842 buf, &bufsize); 9843 reparse_free(nvl); 9844 9845 DTRACE_PROBE4(nfs4serv__func__referral__upcall, 9846 char *, stype, char *, sdata, char *, buf, int, err); 9847 if (err) { 9848 cmn_err(CE_NOTE, 9849 "reparsed daemon not running: unable to get referral (%d)", 9850 err); 9851 return (NULL); 9852 } 9853 9854 /* 9855 * We get an XDR'ed record back from the kderef call 9856 */ 9857 xdrmem_create(&xdr, buf, bufsize, XDR_DECODE); 9858 result = kmem_alloc(sizeof (fs_locations4), KM_SLEEP); 9859 err = xdr_fs_locations4(&xdr, result); 9860 XDR_DESTROY(&xdr); 9861 if (err != TRUE) { 9862 DTRACE_PROBE1(nfs4serv__func__referral__upcall__xdrfail, 9863 int, err); 9864 return (NULL); 9865 } 9866 9867 /* 9868 * Look at path to recover fs_root, ignoring the leading '/' 9869 */ 9870 (void) make_pathname4(vp->v_path, &result->fs_root); 9871 9872 return (result); 9873 } 9874 9875 char * 9876 build_symlink(vnode_t *vp, cred_t *cr, size_t *strsz) 9877 { 9878 fs_locations4 *fsl; 9879 fs_location4 *fs; 9880 char *server, *path, *symbuf; 9881 static char *prefix = "/net/"; 9882 int i, size, npaths; 9883 uint_t len; 9884 9885 /* Get the referral */ 9886 if ((fsl = fetch_referral(vp, cr)) == NULL) 9887 return (NULL); 9888 9889 /* Deal with only the first location and first server */ 9890 fs = &fsl->locations_val[0]; 9891 server = utf8_to_str(&fs->server_val[0], &len, NULL); 9892 if (server == NULL) { 9893 rfs4_free_fs_locations4(fsl); 9894 kmem_free(fsl, sizeof (fs_locations4)); 9895 return (NULL); 9896 } 9897 9898 /* Figure out size for "/net/" + host + /path/path/path + NULL */ 9899 size = strlen(prefix) + len; 9900 for (i = 0; i < fs->rootpath.pathname4_len; i++) 9901 size += fs->rootpath.pathname4_val[i].utf8string_len + 1; 9902 9903 /* Allocate the symlink buffer and fill it */ 9904 symbuf = kmem_zalloc(size, KM_SLEEP); 9905 (void) strcat(symbuf, prefix); 9906 (void) strcat(symbuf, server); 9907 kmem_free(server, len); 9908 9909 npaths = 0; 9910 for (i = 0; i < fs->rootpath.pathname4_len; i++) { 9911 path = utf8_to_str(&fs->rootpath.pathname4_val[i], &len, NULL); 9912 if (path == NULL) 9913 continue; 9914 (void) strcat(symbuf, "/"); 9915 (void) strcat(symbuf, path); 9916 npaths++; 9917 kmem_free(path, len); 9918 } 9919 9920 rfs4_free_fs_locations4(fsl); 9921 kmem_free(fsl, sizeof (fs_locations4)); 9922 9923 if (strsz != NULL) 9924 *strsz = size; 9925 return (symbuf); 9926 } 9927 9928 /* 9929 * Check to see if we have a downrev Solaris client, so that we 9930 * can send it a symlink instead of a referral. 9931 */ 9932 int 9933 client_is_downrev(struct svc_req *req) 9934 { 9935 struct sockaddr *ca; 9936 rfs4_clntip_t *ci; 9937 bool_t create = FALSE; 9938 int is_downrev; 9939 9940 ca = (struct sockaddr *)svc_getrpccaller(req->rq_xprt)->buf; 9941 ASSERT(ca); 9942 ci = rfs4_find_clntip(ca, &create); 9943 if (ci == NULL) 9944 return (0); 9945 is_downrev = ci->ri_no_referrals; 9946 rfs4_dbe_rele(ci->ri_dbe); 9947 return (is_downrev); 9948 } 9949 9950 /* 9951 * Do the main work of handling HA-NFSv4 Resource Group failover on 9952 * Sun Cluster. 9953 * We need to detect whether any RG admin paths have been added or removed, 9954 * and adjust resources accordingly. 9955 * Currently we're using a very inefficient algorithm, ~ 2 * O(n**2). In 9956 * order to scale, the list and array of paths need to be held in more 9957 * suitable data structures. 9958 */ 9959 static void 9960 hanfsv4_failover(nfs4_srv_t *nsrv4) 9961 { 9962 int i, start_grace, numadded_paths = 0; 9963 char **added_paths = NULL; 9964 rfs4_dss_path_t *dss_path; 9965 9966 /* 9967 * Note: currently, dss_pathlist cannot be NULL, since 9968 * it will always include an entry for NFS4_DSS_VAR_DIR. If we 9969 * make the latter dynamically specified too, the following will 9970 * need to be adjusted. 9971 */ 9972 9973 /* 9974 * First, look for removed paths: RGs that have been failed-over 9975 * away from this node. 9976 * Walk the "currently-serving" dss_pathlist and, for each 9977 * path, check if it is on the "passed-in" rfs4_dss_newpaths array 9978 * from nfsd. If not, that RG path has been removed. 9979 * 9980 * Note that nfsd has sorted rfs4_dss_newpaths for us, and removed 9981 * any duplicates. 9982 */ 9983 dss_path = nsrv4->dss_pathlist; 9984 do { 9985 int found = 0; 9986 char *path = dss_path->path; 9987 9988 /* used only for non-HA so may not be removed */ 9989 if (strcmp(path, NFS4_DSS_VAR_DIR) == 0) { 9990 dss_path = dss_path->next; 9991 continue; 9992 } 9993 9994 for (i = 0; i < rfs4_dss_numnewpaths; i++) { 9995 int cmpret; 9996 char *newpath = rfs4_dss_newpaths[i]; 9997 9998 /* 9999 * Since nfsd has sorted rfs4_dss_newpaths for us, 10000 * once the return from strcmp is negative we know 10001 * we've passed the point where "path" should be, 10002 * and can stop searching: "path" has been removed. 10003 */ 10004 cmpret = strcmp(path, newpath); 10005 if (cmpret < 0) 10006 break; 10007 if (cmpret == 0) { 10008 found = 1; 10009 break; 10010 } 10011 } 10012 10013 if (found == 0) { 10014 unsigned index = dss_path->index; 10015 rfs4_servinst_t *sip = dss_path->sip; 10016 rfs4_dss_path_t *path_next = dss_path->next; 10017 10018 /* 10019 * This path has been removed. 10020 * We must clear out the servinst reference to 10021 * it, since it's now owned by another 10022 * node: we should not attempt to touch it. 10023 */ 10024 ASSERT(dss_path == sip->dss_paths[index]); 10025 sip->dss_paths[index] = NULL; 10026 10027 /* remove from "currently-serving" list, and destroy */ 10028 remque(dss_path); 10029 /* allow for NUL */ 10030 kmem_free(dss_path->path, strlen(dss_path->path) + 1); 10031 kmem_free(dss_path, sizeof (rfs4_dss_path_t)); 10032 10033 dss_path = path_next; 10034 } else { 10035 /* path was found; not removed */ 10036 dss_path = dss_path->next; 10037 } 10038 } while (dss_path != nsrv4->dss_pathlist); 10039 10040 /* 10041 * Now, look for added paths: RGs that have been failed-over 10042 * to this node. 10043 * Walk the "passed-in" rfs4_dss_newpaths array from nfsd and, 10044 * for each path, check if it is on the "currently-serving" 10045 * dss_pathlist. If not, that RG path has been added. 10046 * 10047 * Note: we don't do duplicate detection here; nfsd does that for us. 10048 * 10049 * Note: numadded_paths <= rfs4_dss_numnewpaths, which gives us 10050 * an upper bound for the size needed for added_paths[numadded_paths]. 10051 */ 10052 10053 /* probably more space than we need, but guaranteed to be enough */ 10054 if (rfs4_dss_numnewpaths > 0) { 10055 size_t sz = rfs4_dss_numnewpaths * sizeof (char *); 10056 added_paths = kmem_zalloc(sz, KM_SLEEP); 10057 } 10058 10059 /* walk the "passed-in" rfs4_dss_newpaths array from nfsd */ 10060 for (i = 0; i < rfs4_dss_numnewpaths; i++) { 10061 int found = 0; 10062 char *newpath = rfs4_dss_newpaths[i]; 10063 10064 dss_path = nsrv4->dss_pathlist; 10065 do { 10066 char *path = dss_path->path; 10067 10068 /* used only for non-HA */ 10069 if (strcmp(path, NFS4_DSS_VAR_DIR) == 0) { 10070 dss_path = dss_path->next; 10071 continue; 10072 } 10073 10074 if (strncmp(path, newpath, strlen(path)) == 0) { 10075 found = 1; 10076 break; 10077 } 10078 10079 dss_path = dss_path->next; 10080 } while (dss_path != nsrv4->dss_pathlist); 10081 10082 if (found == 0) { 10083 added_paths[numadded_paths] = newpath; 10084 numadded_paths++; 10085 } 10086 } 10087 10088 /* did we find any added paths? */ 10089 if (numadded_paths > 0) { 10090 10091 /* create a new server instance, and start its grace period */ 10092 start_grace = 1; 10093 /* CSTYLED */ 10094 rfs4_servinst_create(nsrv4, start_grace, numadded_paths, added_paths); 10095 10096 /* read in the stable storage state from these paths */ 10097 rfs4_dss_readstate(nsrv4, numadded_paths, added_paths); 10098 10099 /* 10100 * Multiple failovers during a grace period will cause 10101 * clients of the same resource group to be partitioned 10102 * into different server instances, with different 10103 * grace periods. Since clients of the same resource 10104 * group must be subject to the same grace period, 10105 * we need to reset all currently active grace periods. 10106 */ 10107 rfs4_grace_reset_all(nsrv4); 10108 } 10109 10110 if (rfs4_dss_numnewpaths > 0) 10111 kmem_free(added_paths, rfs4_dss_numnewpaths * sizeof (char *)); 10112 } 10113