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 error = VFS_STATVFS(vp->v_vfsp, sargp->sbp); 2206 if (error != 0) { 2207 sargp->sbp = NULL; /* to identify error */ 2208 return (puterrno4(error)); 2209 } 2210 } 2211 2212 return (rfs4_vop_getattr(vp, sargp->vap, 0, cs->cr)); 2213 } 2214 2215 static void 2216 nfs4_ntov_table_init(struct nfs4_ntov_table *ntovp) 2217 { 2218 ntovp->na = kmem_zalloc(sizeof (union nfs4_attr_u) * nfs4_ntov_map_size, 2219 KM_SLEEP); 2220 ntovp->attrcnt = 0; 2221 ntovp->vfsstat = FALSE; 2222 } 2223 2224 static void 2225 nfs4_ntov_table_free(struct nfs4_ntov_table *ntovp, 2226 struct nfs4_svgetit_arg *sargp) 2227 { 2228 int i; 2229 union nfs4_attr_u *na; 2230 uint8_t *amap; 2231 2232 /* 2233 * XXX Should do the same checks for whether the bit is set 2234 */ 2235 for (i = 0, na = ntovp->na, amap = ntovp->amap; 2236 i < ntovp->attrcnt; i++, na++, amap++) { 2237 (void) (*nfs4_ntov_map[*amap].sv_getit)( 2238 NFS4ATTR_FREEIT, sargp, na); 2239 } 2240 if ((sargp->op == NFS4ATTR_SETIT) || (sargp->op == NFS4ATTR_VERIT)) { 2241 /* 2242 * xdr_free for getattr will be done later 2243 */ 2244 for (i = 0, na = ntovp->na, amap = ntovp->amap; 2245 i < ntovp->attrcnt; i++, na++, amap++) { 2246 xdr_free(nfs4_ntov_map[*amap].xfunc, (caddr_t)na); 2247 } 2248 } 2249 kmem_free(ntovp->na, sizeof (union nfs4_attr_u) * nfs4_ntov_map_size); 2250 } 2251 2252 /* 2253 * do_rfs4_op_getattr gets the system attrs and converts into fattr4. 2254 */ 2255 static nfsstat4 2256 do_rfs4_op_getattr(bitmap4 breq, fattr4 *fattrp, 2257 struct nfs4_svgetit_arg *sargp) 2258 { 2259 int error = 0; 2260 int i, k; 2261 struct nfs4_ntov_table ntov; 2262 XDR xdr; 2263 ulong_t xdr_size; 2264 char *xdr_attrs; 2265 nfsstat4 status = NFS4_OK; 2266 nfsstat4 prev_rdattr_error = sargp->rdattr_error; 2267 union nfs4_attr_u *na; 2268 uint8_t *amap; 2269 2270 sargp->op = NFS4ATTR_GETIT; 2271 sargp->flag = 0; 2272 2273 fattrp->attrmask = 0; 2274 /* if no bits requested, then return empty fattr4 */ 2275 if (breq == 0) { 2276 fattrp->attrlist4_len = 0; 2277 fattrp->attrlist4 = NULL; 2278 return (NFS4_OK); 2279 } 2280 2281 /* 2282 * return NFS4ERR_INVAL when client requests write-only attrs 2283 */ 2284 if (breq & (FATTR4_TIME_ACCESS_SET_MASK | FATTR4_TIME_MODIFY_SET_MASK)) 2285 return (NFS4ERR_INVAL); 2286 2287 nfs4_ntov_table_init(&ntov); 2288 na = ntov.na; 2289 amap = ntov.amap; 2290 2291 /* 2292 * Now loop to get or verify the attrs 2293 */ 2294 for (i = 0; i < nfs4_ntov_map_size; i++) { 2295 if (breq & nfs4_ntov_map[i].fbit) { 2296 if ((*nfs4_ntov_map[i].sv_getit)( 2297 NFS4ATTR_SUPPORTED, sargp, NULL) == 0) { 2298 2299 error = (*nfs4_ntov_map[i].sv_getit)( 2300 NFS4ATTR_GETIT, sargp, na); 2301 2302 /* 2303 * Possible error values: 2304 * >0 if sv_getit failed to 2305 * get the attr; 0 if succeeded; 2306 * <0 if rdattr_error and the 2307 * attribute cannot be returned. 2308 */ 2309 if (error && !(sargp->rdattr_error_req)) 2310 goto done; 2311 /* 2312 * If error then just for entry 2313 */ 2314 if (error == 0) { 2315 fattrp->attrmask |= 2316 nfs4_ntov_map[i].fbit; 2317 *amap++ = 2318 (uint8_t)nfs4_ntov_map[i].nval; 2319 na++; 2320 (ntov.attrcnt)++; 2321 } else if ((error > 0) && 2322 (sargp->rdattr_error == NFS4_OK)) { 2323 sargp->rdattr_error = puterrno4(error); 2324 } 2325 error = 0; 2326 } 2327 } 2328 } 2329 2330 /* 2331 * If rdattr_error was set after the return value for it was assigned, 2332 * update it. 2333 */ 2334 if (prev_rdattr_error != sargp->rdattr_error) { 2335 na = ntov.na; 2336 amap = ntov.amap; 2337 for (i = 0; i < ntov.attrcnt; i++, na++, amap++) { 2338 k = *amap; 2339 if (k < FATTR4_RDATTR_ERROR) { 2340 continue; 2341 } 2342 if ((k == FATTR4_RDATTR_ERROR) && 2343 ((*nfs4_ntov_map[k].sv_getit)( 2344 NFS4ATTR_SUPPORTED, sargp, NULL) == 0)) { 2345 2346 (void) (*nfs4_ntov_map[k].sv_getit)( 2347 NFS4ATTR_GETIT, sargp, na); 2348 } 2349 break; 2350 } 2351 } 2352 2353 xdr_size = 0; 2354 na = ntov.na; 2355 amap = ntov.amap; 2356 for (i = 0; i < ntov.attrcnt; i++, na++, amap++) { 2357 xdr_size += xdr_sizeof(nfs4_ntov_map[*amap].xfunc, na); 2358 } 2359 2360 fattrp->attrlist4_len = xdr_size; 2361 if (xdr_size) { 2362 /* freed by rfs4_op_getattr_free() */ 2363 fattrp->attrlist4 = xdr_attrs = kmem_zalloc(xdr_size, KM_SLEEP); 2364 2365 xdrmem_create(&xdr, xdr_attrs, xdr_size, XDR_ENCODE); 2366 2367 na = ntov.na; 2368 amap = ntov.amap; 2369 for (i = 0; i < ntov.attrcnt; i++, na++, amap++) { 2370 if (!(*nfs4_ntov_map[*amap].xfunc)(&xdr, na)) { 2371 DTRACE_PROBE1(nfss__e__getattr4_encfail, 2372 int, *amap); 2373 status = NFS4ERR_SERVERFAULT; 2374 break; 2375 } 2376 } 2377 /* xdrmem_destroy(&xdrs); */ /* NO-OP */ 2378 } else { 2379 fattrp->attrlist4 = NULL; 2380 } 2381 done: 2382 2383 nfs4_ntov_table_free(&ntov, sargp); 2384 2385 if (error != 0) 2386 status = puterrno4(error); 2387 2388 return (status); 2389 } 2390 2391 /* ARGSUSED */ 2392 static void 2393 rfs4_op_getattr(nfs_argop4 *argop, nfs_resop4 *resop, struct svc_req *req, 2394 struct compound_state *cs) 2395 { 2396 GETATTR4args *args = &argop->nfs_argop4_u.opgetattr; 2397 GETATTR4res *resp = &resop->nfs_resop4_u.opgetattr; 2398 struct nfs4_svgetit_arg sarg; 2399 struct statvfs64 sb; 2400 nfsstat4 status; 2401 2402 DTRACE_NFSV4_2(op__getattr__start, struct compound_state *, cs, 2403 GETATTR4args *, args); 2404 2405 if (cs->vp == NULL) { 2406 *cs->statusp = resp->status = NFS4ERR_NOFILEHANDLE; 2407 goto out; 2408 } 2409 2410 if (cs->access == CS_ACCESS_DENIED) { 2411 *cs->statusp = resp->status = NFS4ERR_ACCESS; 2412 goto out; 2413 } 2414 2415 sarg.sbp = &sb; 2416 sarg.cs = cs; 2417 sarg.is_referral = B_FALSE; 2418 2419 status = bitmap4_to_attrmask(args->attr_request, &sarg); 2420 if (status == NFS4_OK) { 2421 2422 status = bitmap4_get_sysattrs(&sarg); 2423 if (status == NFS4_OK) { 2424 2425 /* Is this a referral? */ 2426 if (vn_is_nfs_reparse(cs->vp, cs->cr)) { 2427 /* Older V4 Solaris client sees a link */ 2428 if (client_is_downrev(req)) 2429 sarg.vap->va_type = VLNK; 2430 else 2431 sarg.is_referral = B_TRUE; 2432 } 2433 2434 status = do_rfs4_op_getattr(args->attr_request, 2435 &resp->obj_attributes, &sarg); 2436 } 2437 } 2438 *cs->statusp = resp->status = status; 2439 out: 2440 DTRACE_NFSV4_2(op__getattr__done, struct compound_state *, cs, 2441 GETATTR4res *, resp); 2442 } 2443 2444 static void 2445 rfs4_op_getattr_free(nfs_resop4 *resop) 2446 { 2447 GETATTR4res *resp = &resop->nfs_resop4_u.opgetattr; 2448 2449 nfs4_fattr4_free(&resp->obj_attributes); 2450 } 2451 2452 /* ARGSUSED */ 2453 static void 2454 rfs4_op_getfh(nfs_argop4 *argop, nfs_resop4 *resop, struct svc_req *req, 2455 struct compound_state *cs) 2456 { 2457 GETFH4res *resp = &resop->nfs_resop4_u.opgetfh; 2458 2459 DTRACE_NFSV4_1(op__getfh__start, struct compound_state *, cs); 2460 2461 if (cs->vp == NULL) { 2462 *cs->statusp = resp->status = NFS4ERR_NOFILEHANDLE; 2463 goto out; 2464 } 2465 if (cs->access == CS_ACCESS_DENIED) { 2466 *cs->statusp = resp->status = NFS4ERR_ACCESS; 2467 goto out; 2468 } 2469 2470 /* check for reparse point at the share point */ 2471 if (cs->exi->exi_moved || vn_is_nfs_reparse(cs->exi->exi_vp, cs->cr)) { 2472 /* it's all bad */ 2473 cs->exi->exi_moved = 1; 2474 *cs->statusp = resp->status = NFS4ERR_MOVED; 2475 DTRACE_PROBE2(nfs4serv__func__referral__shared__moved, 2476 vnode_t *, cs->vp, char *, "rfs4_op_getfh"); 2477 return; 2478 } 2479 2480 /* check for reparse point at vp */ 2481 if (vn_is_nfs_reparse(cs->vp, cs->cr) && !client_is_downrev(req)) { 2482 /* it's not all bad */ 2483 *cs->statusp = resp->status = NFS4ERR_MOVED; 2484 DTRACE_PROBE2(nfs4serv__func__referral__moved, 2485 vnode_t *, cs->vp, char *, "rfs4_op_getfh"); 2486 return; 2487 } 2488 2489 resp->object.nfs_fh4_val = 2490 kmem_alloc(cs->fh.nfs_fh4_len, KM_SLEEP); 2491 nfs_fh4_copy(&cs->fh, &resp->object); 2492 *cs->statusp = resp->status = NFS4_OK; 2493 out: 2494 DTRACE_NFSV4_2(op__getfh__done, struct compound_state *, cs, 2495 GETFH4res *, resp); 2496 } 2497 2498 static void 2499 rfs4_op_getfh_free(nfs_resop4 *resop) 2500 { 2501 GETFH4res *resp = &resop->nfs_resop4_u.opgetfh; 2502 2503 if (resp->status == NFS4_OK && 2504 resp->object.nfs_fh4_val != NULL) { 2505 kmem_free(resp->object.nfs_fh4_val, resp->object.nfs_fh4_len); 2506 resp->object.nfs_fh4_val = NULL; 2507 resp->object.nfs_fh4_len = 0; 2508 } 2509 } 2510 2511 /* 2512 * illegal: args: void 2513 * res : status (NFS4ERR_OP_ILLEGAL) 2514 */ 2515 /* ARGSUSED */ 2516 static void 2517 rfs4_op_illegal(nfs_argop4 *argop, nfs_resop4 *resop, 2518 struct svc_req *req, struct compound_state *cs) 2519 { 2520 ILLEGAL4res *resp = &resop->nfs_resop4_u.opillegal; 2521 2522 resop->resop = OP_ILLEGAL; 2523 *cs->statusp = resp->status = NFS4ERR_OP_ILLEGAL; 2524 } 2525 2526 /* 2527 * link: args: SAVED_FH: file, CURRENT_FH: target directory 2528 * res: status. If success - CURRENT_FH unchanged, return change_info 2529 */ 2530 /* ARGSUSED */ 2531 static void 2532 rfs4_op_link(nfs_argop4 *argop, nfs_resop4 *resop, struct svc_req *req, 2533 struct compound_state *cs) 2534 { 2535 LINK4args *args = &argop->nfs_argop4_u.oplink; 2536 LINK4res *resp = &resop->nfs_resop4_u.oplink; 2537 int error; 2538 vnode_t *vp; 2539 vnode_t *dvp; 2540 struct vattr bdva, idva, adva; 2541 char *nm; 2542 uint_t len; 2543 struct sockaddr *ca; 2544 char *name = NULL; 2545 nfsstat4 status; 2546 2547 DTRACE_NFSV4_2(op__link__start, struct compound_state *, cs, 2548 LINK4args *, args); 2549 2550 /* SAVED_FH: source object */ 2551 vp = cs->saved_vp; 2552 if (vp == NULL) { 2553 *cs->statusp = resp->status = NFS4ERR_NOFILEHANDLE; 2554 goto out; 2555 } 2556 2557 /* CURRENT_FH: target directory */ 2558 dvp = cs->vp; 2559 if (dvp == NULL) { 2560 *cs->statusp = resp->status = NFS4ERR_NOFILEHANDLE; 2561 goto out; 2562 } 2563 2564 /* 2565 * If there is a non-shared filesystem mounted on this vnode, 2566 * do not allow to link any file in this directory. 2567 */ 2568 if (vn_ismntpt(dvp)) { 2569 *cs->statusp = resp->status = NFS4ERR_ACCESS; 2570 goto out; 2571 } 2572 2573 if (cs->access == CS_ACCESS_DENIED) { 2574 *cs->statusp = resp->status = NFS4ERR_ACCESS; 2575 goto out; 2576 } 2577 2578 /* Check source object's type validity */ 2579 if (vp->v_type == VDIR) { 2580 *cs->statusp = resp->status = NFS4ERR_ISDIR; 2581 goto out; 2582 } 2583 2584 /* Check target directory's type */ 2585 if (dvp->v_type != VDIR) { 2586 *cs->statusp = resp->status = NFS4ERR_NOTDIR; 2587 goto out; 2588 } 2589 2590 if (cs->saved_exi != cs->exi) { 2591 *cs->statusp = resp->status = NFS4ERR_XDEV; 2592 goto out; 2593 } 2594 2595 status = utf8_dir_verify(&args->newname); 2596 if (status != NFS4_OK) { 2597 *cs->statusp = resp->status = status; 2598 goto out; 2599 } 2600 2601 nm = utf8_to_fn(&args->newname, &len, NULL); 2602 if (nm == NULL) { 2603 *cs->statusp = resp->status = NFS4ERR_INVAL; 2604 goto out; 2605 } 2606 2607 if (len > MAXNAMELEN) { 2608 *cs->statusp = resp->status = NFS4ERR_NAMETOOLONG; 2609 kmem_free(nm, len); 2610 goto out; 2611 } 2612 2613 if (rdonly4(req, cs)) { 2614 *cs->statusp = resp->status = NFS4ERR_ROFS; 2615 kmem_free(nm, len); 2616 goto out; 2617 } 2618 2619 /* Get "before" change value */ 2620 bdva.va_mask = AT_CTIME|AT_SEQ; 2621 error = VOP_GETATTR(dvp, &bdva, 0, cs->cr, NULL); 2622 if (error) { 2623 *cs->statusp = resp->status = puterrno4(error); 2624 kmem_free(nm, len); 2625 goto out; 2626 } 2627 2628 ca = (struct sockaddr *)svc_getrpccaller(req->rq_xprt)->buf; 2629 name = nfscmd_convname(ca, cs->exi, nm, NFSCMD_CONV_INBOUND, 2630 MAXPATHLEN + 1); 2631 2632 if (name == NULL) { 2633 *cs->statusp = resp->status = NFS4ERR_INVAL; 2634 kmem_free(nm, len); 2635 goto out; 2636 } 2637 2638 NFS4_SET_FATTR4_CHANGE(resp->cinfo.before, bdva.va_ctime) 2639 2640 error = VOP_LINK(dvp, vp, name, cs->cr, NULL, 0); 2641 2642 if (nm != name) 2643 kmem_free(name, MAXPATHLEN + 1); 2644 kmem_free(nm, len); 2645 2646 /* 2647 * Get the initial "after" sequence number, if it fails, set to zero 2648 */ 2649 idva.va_mask = AT_SEQ; 2650 if (VOP_GETATTR(dvp, &idva, 0, cs->cr, NULL)) 2651 idva.va_seq = 0; 2652 2653 /* 2654 * Force modified data and metadata out to stable storage. 2655 */ 2656 (void) VOP_FSYNC(vp, FNODSYNC, cs->cr, NULL); 2657 (void) VOP_FSYNC(dvp, 0, cs->cr, NULL); 2658 2659 if (error) { 2660 *cs->statusp = resp->status = puterrno4(error); 2661 goto out; 2662 } 2663 2664 /* 2665 * Get "after" change value, if it fails, simply return the 2666 * before value. 2667 */ 2668 adva.va_mask = AT_CTIME|AT_SEQ; 2669 if (VOP_GETATTR(dvp, &adva, 0, cs->cr, NULL)) { 2670 adva.va_ctime = bdva.va_ctime; 2671 adva.va_seq = 0; 2672 } 2673 2674 NFS4_SET_FATTR4_CHANGE(resp->cinfo.after, adva.va_ctime) 2675 2676 /* 2677 * The cinfo.atomic = TRUE only if we have 2678 * non-zero va_seq's, and it has incremented by exactly one 2679 * during the VOP_LINK and it didn't change during the VOP_FSYNC. 2680 */ 2681 if (bdva.va_seq && idva.va_seq && adva.va_seq && 2682 idva.va_seq == (bdva.va_seq + 1) && idva.va_seq == adva.va_seq) 2683 resp->cinfo.atomic = TRUE; 2684 else 2685 resp->cinfo.atomic = FALSE; 2686 2687 *cs->statusp = resp->status = NFS4_OK; 2688 out: 2689 DTRACE_NFSV4_2(op__link__done, struct compound_state *, cs, 2690 LINK4res *, resp); 2691 } 2692 2693 /* 2694 * Used by rfs4_op_lookup and rfs4_op_lookupp to do the actual work. 2695 */ 2696 2697 /* ARGSUSED */ 2698 static nfsstat4 2699 do_rfs4_op_lookup(char *nm, struct svc_req *req, struct compound_state *cs) 2700 { 2701 int error; 2702 int different_export = 0; 2703 vnode_t *vp, *pre_tvp = NULL, *oldvp = NULL; 2704 struct exportinfo *exi = NULL, *pre_exi = NULL; 2705 nfsstat4 stat; 2706 fid_t fid; 2707 int attrdir, dotdot, walk; 2708 bool_t is_newvp = FALSE; 2709 2710 if (cs->vp->v_flag & V_XATTRDIR) { 2711 attrdir = 1; 2712 ASSERT(get_fh4_flag(&cs->fh, FH4_ATTRDIR)); 2713 } else { 2714 attrdir = 0; 2715 ASSERT(! get_fh4_flag(&cs->fh, FH4_ATTRDIR)); 2716 } 2717 2718 dotdot = (nm[0] == '.' && nm[1] == '.' && nm[2] == '\0'); 2719 2720 /* 2721 * If dotdotting, then need to check whether it's 2722 * above the root of a filesystem, or above an 2723 * export point. 2724 */ 2725 if (dotdot) { 2726 vnode_t *zone_rootvp; 2727 2728 ASSERT(cs->exi != NULL); 2729 zone_rootvp = cs->exi->exi_ne->exi_root->exi_vp; 2730 /* 2731 * If dotdotting at the root of a filesystem, then 2732 * need to traverse back to the mounted-on filesystem 2733 * and do the dotdot lookup there. 2734 */ 2735 if ((cs->vp->v_flag & VROOT) || VN_CMP(cs->vp, zone_rootvp)) { 2736 2737 /* 2738 * If at the system root, then can 2739 * go up no further. 2740 */ 2741 if (VN_CMP(cs->vp, zone_rootvp)) 2742 return (puterrno4(ENOENT)); 2743 2744 /* 2745 * Traverse back to the mounted-on filesystem 2746 */ 2747 cs->vp = untraverse(cs->vp, zone_rootvp); 2748 2749 /* 2750 * Set the different_export flag so we remember 2751 * to pick up a new exportinfo entry for 2752 * this new filesystem. 2753 */ 2754 different_export = 1; 2755 } else { 2756 2757 /* 2758 * If dotdotting above an export point then set 2759 * the different_export to get new export info. 2760 */ 2761 different_export = nfs_exported(cs->exi, cs->vp); 2762 } 2763 } 2764 2765 error = VOP_LOOKUP(cs->vp, nm, &vp, NULL, 0, NULL, cs->cr, 2766 NULL, NULL, NULL); 2767 if (error) 2768 return (puterrno4(error)); 2769 2770 /* 2771 * If the vnode is in a pseudo filesystem, check whether it is visible. 2772 * 2773 * XXX if the vnode is a symlink and it is not visible in 2774 * a pseudo filesystem, return ENOENT (not following symlink). 2775 * V4 client can not mount such symlink. This is a regression 2776 * from V2/V3. 2777 * 2778 * In the same exported filesystem, if the security flavor used 2779 * is not an explicitly shared flavor, limit the view to the visible 2780 * list entries only. This is not a WRONGSEC case because it's already 2781 * checked via PUTROOTFH/PUTPUBFH or PUTFH. 2782 */ 2783 if (!different_export && 2784 (PSEUDO(cs->exi) || ! is_exported_sec(cs->nfsflavor, cs->exi) || 2785 cs->access & CS_ACCESS_LIMITED)) { 2786 if (! nfs_visible(cs->exi, vp, &different_export)) { 2787 VN_RELE(vp); 2788 return (puterrno4(ENOENT)); 2789 } 2790 } 2791 2792 /* 2793 * If it's a mountpoint, then traverse it. 2794 */ 2795 if (vn_ismntpt(vp)) { 2796 pre_exi = cs->exi; /* save pre-traversed exportinfo */ 2797 pre_tvp = vp; /* save pre-traversed vnode */ 2798 2799 /* 2800 * hold pre_tvp to counteract rele by traverse. We will 2801 * need pre_tvp below if checkexport4 fails 2802 */ 2803 VN_HOLD(pre_tvp); 2804 if ((error = traverse(&vp)) != 0) { 2805 VN_RELE(vp); 2806 VN_RELE(pre_tvp); 2807 return (puterrno4(error)); 2808 } 2809 different_export = 1; 2810 } else if (vp->v_vfsp != cs->vp->v_vfsp) { 2811 /* 2812 * The vfsp comparison is to handle the case where 2813 * a LOFS mount is shared. lo_lookup traverses mount points, 2814 * and NFS is unaware of local fs transistions because 2815 * v_vfsmountedhere isn't set. For this special LOFS case, 2816 * the dir and the obj returned by lookup will have different 2817 * vfs ptrs. 2818 */ 2819 different_export = 1; 2820 } 2821 2822 if (different_export) { 2823 2824 bzero(&fid, sizeof (fid)); 2825 fid.fid_len = MAXFIDSZ; 2826 error = vop_fid_pseudo(vp, &fid); 2827 if (error) { 2828 VN_RELE(vp); 2829 if (pre_tvp) 2830 VN_RELE(pre_tvp); 2831 return (puterrno4(error)); 2832 } 2833 2834 if (dotdot) 2835 exi = nfs_vptoexi(NULL, vp, cs->cr, &walk, NULL, TRUE); 2836 else 2837 exi = checkexport4(&vp->v_vfsp->vfs_fsid, &fid, vp); 2838 2839 if (exi == NULL) { 2840 if (pre_tvp) { 2841 /* 2842 * If this vnode is a mounted-on vnode, 2843 * but the mounted-on file system is not 2844 * exported, send back the filehandle for 2845 * the mounted-on vnode, not the root of 2846 * the mounted-on file system. 2847 */ 2848 VN_RELE(vp); 2849 vp = pre_tvp; 2850 exi = pre_exi; 2851 } else { 2852 VN_RELE(vp); 2853 return (puterrno4(EACCES)); 2854 } 2855 } else if (pre_tvp) { 2856 /* we're done with pre_tvp now. release extra hold */ 2857 VN_RELE(pre_tvp); 2858 } 2859 2860 cs->exi = exi; 2861 2862 /* 2863 * Now we do a checkauth4. The reason is that 2864 * this client/user may not have access to the new 2865 * exported file system, and if they do, 2866 * the client/user may be mapped to a different uid. 2867 * 2868 * We start with a new cr, because the checkauth4 done 2869 * in the PUT*FH operation over wrote the cred's uid, 2870 * gid, etc, and we want the real thing before calling 2871 * checkauth4() 2872 */ 2873 crfree(cs->cr); 2874 cs->cr = crdup(cs->basecr); 2875 2876 oldvp = cs->vp; 2877 cs->vp = vp; 2878 is_newvp = TRUE; 2879 2880 stat = call_checkauth4(cs, req); 2881 if (stat != NFS4_OK) { 2882 VN_RELE(cs->vp); 2883 cs->vp = oldvp; 2884 return (stat); 2885 } 2886 } 2887 2888 /* 2889 * After various NFS checks, do a label check on the path 2890 * component. The label on this path should either be the 2891 * global zone's label or a zone's label. We are only 2892 * interested in the zone's label because exported files 2893 * in global zone is accessible (though read-only) to 2894 * clients. The exportability/visibility check is already 2895 * done before reaching this code. 2896 */ 2897 if (is_system_labeled()) { 2898 bslabel_t *clabel; 2899 2900 ASSERT(req->rq_label != NULL); 2901 clabel = req->rq_label; 2902 DTRACE_PROBE2(tx__rfs4__log__info__oplookup__clabel, char *, 2903 "got client label from request(1)", struct svc_req *, req); 2904 2905 if (!blequal(&l_admin_low->tsl_label, clabel)) { 2906 if (!do_rfs_label_check(clabel, vp, DOMINANCE_CHECK, 2907 cs->exi)) { 2908 error = EACCES; 2909 goto err_out; 2910 } 2911 } else { 2912 /* 2913 * We grant access to admin_low label clients 2914 * only if the client is trusted, i.e. also 2915 * running Solaris Trusted Extension. 2916 */ 2917 struct sockaddr *ca; 2918 int addr_type; 2919 void *ipaddr; 2920 tsol_tpc_t *tp; 2921 2922 ca = (struct sockaddr *)svc_getrpccaller( 2923 req->rq_xprt)->buf; 2924 if (ca->sa_family == AF_INET) { 2925 addr_type = IPV4_VERSION; 2926 ipaddr = &((struct sockaddr_in *)ca)->sin_addr; 2927 } else if (ca->sa_family == AF_INET6) { 2928 addr_type = IPV6_VERSION; 2929 ipaddr = &((struct sockaddr_in6 *) 2930 ca)->sin6_addr; 2931 } 2932 tp = find_tpc(ipaddr, addr_type, B_FALSE); 2933 if (tp == NULL || tp->tpc_tp.tp_doi != 2934 l_admin_low->tsl_doi || tp->tpc_tp.host_type != 2935 SUN_CIPSO) { 2936 if (tp != NULL) 2937 TPC_RELE(tp); 2938 error = EACCES; 2939 goto err_out; 2940 } 2941 TPC_RELE(tp); 2942 } 2943 } 2944 2945 error = makefh4(&cs->fh, vp, cs->exi); 2946 2947 err_out: 2948 if (error) { 2949 if (is_newvp) { 2950 VN_RELE(cs->vp); 2951 cs->vp = oldvp; 2952 } else 2953 VN_RELE(vp); 2954 return (puterrno4(error)); 2955 } 2956 2957 if (!is_newvp) { 2958 if (cs->vp) 2959 VN_RELE(cs->vp); 2960 cs->vp = vp; 2961 } else if (oldvp) 2962 VN_RELE(oldvp); 2963 2964 /* 2965 * if did lookup on attrdir and didn't lookup .., set named 2966 * attr fh flag 2967 */ 2968 if (attrdir && ! dotdot) 2969 set_fh4_flag(&cs->fh, FH4_NAMEDATTR); 2970 2971 /* Assume false for now, open proc will set this */ 2972 cs->mandlock = FALSE; 2973 2974 return (NFS4_OK); 2975 } 2976 2977 /* ARGSUSED */ 2978 static void 2979 rfs4_op_lookup(nfs_argop4 *argop, nfs_resop4 *resop, struct svc_req *req, 2980 struct compound_state *cs) 2981 { 2982 LOOKUP4args *args = &argop->nfs_argop4_u.oplookup; 2983 LOOKUP4res *resp = &resop->nfs_resop4_u.oplookup; 2984 char *nm; 2985 uint_t len; 2986 struct sockaddr *ca; 2987 char *name = NULL; 2988 nfsstat4 status; 2989 2990 DTRACE_NFSV4_2(op__lookup__start, struct compound_state *, cs, 2991 LOOKUP4args *, args); 2992 2993 if (cs->vp == NULL) { 2994 *cs->statusp = resp->status = NFS4ERR_NOFILEHANDLE; 2995 goto out; 2996 } 2997 2998 if (cs->vp->v_type == VLNK) { 2999 *cs->statusp = resp->status = NFS4ERR_SYMLINK; 3000 goto out; 3001 } 3002 3003 if (cs->vp->v_type != VDIR) { 3004 *cs->statusp = resp->status = NFS4ERR_NOTDIR; 3005 goto out; 3006 } 3007 3008 status = utf8_dir_verify(&args->objname); 3009 if (status != NFS4_OK) { 3010 *cs->statusp = resp->status = status; 3011 goto out; 3012 } 3013 3014 nm = utf8_to_str(&args->objname, &len, NULL); 3015 if (nm == NULL) { 3016 *cs->statusp = resp->status = NFS4ERR_INVAL; 3017 goto out; 3018 } 3019 3020 if (len > MAXNAMELEN) { 3021 *cs->statusp = resp->status = NFS4ERR_NAMETOOLONG; 3022 kmem_free(nm, len); 3023 goto out; 3024 } 3025 3026 ca = (struct sockaddr *)svc_getrpccaller(req->rq_xprt)->buf; 3027 name = nfscmd_convname(ca, cs->exi, nm, NFSCMD_CONV_INBOUND, 3028 MAXPATHLEN + 1); 3029 3030 if (name == NULL) { 3031 *cs->statusp = resp->status = NFS4ERR_INVAL; 3032 kmem_free(nm, len); 3033 goto out; 3034 } 3035 3036 *cs->statusp = resp->status = do_rfs4_op_lookup(name, req, cs); 3037 3038 if (name != nm) 3039 kmem_free(name, MAXPATHLEN + 1); 3040 kmem_free(nm, len); 3041 3042 out: 3043 DTRACE_NFSV4_2(op__lookup__done, struct compound_state *, cs, 3044 LOOKUP4res *, resp); 3045 } 3046 3047 /* ARGSUSED */ 3048 static void 3049 rfs4_op_lookupp(nfs_argop4 *args, nfs_resop4 *resop, struct svc_req *req, 3050 struct compound_state *cs) 3051 { 3052 LOOKUPP4res *resp = &resop->nfs_resop4_u.oplookupp; 3053 3054 DTRACE_NFSV4_1(op__lookupp__start, struct compound_state *, cs); 3055 3056 if (cs->vp == NULL) { 3057 *cs->statusp = resp->status = NFS4ERR_NOFILEHANDLE; 3058 goto out; 3059 } 3060 3061 if (cs->vp->v_type != VDIR) { 3062 *cs->statusp = resp->status = NFS4ERR_NOTDIR; 3063 goto out; 3064 } 3065 3066 *cs->statusp = resp->status = do_rfs4_op_lookup("..", req, cs); 3067 3068 /* 3069 * From NFSV4 Specification, LOOKUPP should not check for 3070 * NFS4ERR_WRONGSEC. Retrun NFS4_OK instead. 3071 */ 3072 if (resp->status == NFS4ERR_WRONGSEC) { 3073 *cs->statusp = resp->status = NFS4_OK; 3074 } 3075 3076 out: 3077 DTRACE_NFSV4_2(op__lookupp__done, struct compound_state *, cs, 3078 LOOKUPP4res *, resp); 3079 } 3080 3081 3082 /*ARGSUSED2*/ 3083 static void 3084 rfs4_op_openattr(nfs_argop4 *argop, nfs_resop4 *resop, struct svc_req *req, 3085 struct compound_state *cs) 3086 { 3087 OPENATTR4args *args = &argop->nfs_argop4_u.opopenattr; 3088 OPENATTR4res *resp = &resop->nfs_resop4_u.opopenattr; 3089 vnode_t *avp = NULL; 3090 int lookup_flags = LOOKUP_XATTR, error; 3091 int exp_ro = 0; 3092 3093 DTRACE_NFSV4_2(op__openattr__start, struct compound_state *, cs, 3094 OPENATTR4args *, args); 3095 3096 if (cs->vp == NULL) { 3097 *cs->statusp = resp->status = NFS4ERR_NOFILEHANDLE; 3098 goto out; 3099 } 3100 3101 if ((cs->vp->v_vfsp->vfs_flag & VFS_XATTR) == 0 && 3102 !vfs_has_feature(cs->vp->v_vfsp, VFSFT_SYSATTR_VIEWS)) { 3103 *cs->statusp = resp->status = puterrno4(ENOTSUP); 3104 goto out; 3105 } 3106 3107 /* 3108 * If file system supports passing ACE mask to VOP_ACCESS then 3109 * check for ACE_READ_NAMED_ATTRS, otherwise do legacy checks 3110 */ 3111 3112 if (vfs_has_feature(cs->vp->v_vfsp, VFSFT_ACEMASKONACCESS)) 3113 error = VOP_ACCESS(cs->vp, ACE_READ_NAMED_ATTRS, 3114 V_ACE_MASK, cs->cr, NULL); 3115 else 3116 error = ((VOP_ACCESS(cs->vp, VREAD, 0, cs->cr, NULL) != 0) && 3117 (VOP_ACCESS(cs->vp, VWRITE, 0, cs->cr, NULL) != 0) && 3118 (VOP_ACCESS(cs->vp, VEXEC, 0, cs->cr, NULL) != 0)); 3119 3120 if (error) { 3121 *cs->statusp = resp->status = puterrno4(EACCES); 3122 goto out; 3123 } 3124 3125 /* 3126 * The CREATE_XATTR_DIR VOP flag cannot be specified if 3127 * the file system is exported read-only -- regardless of 3128 * createdir flag. Otherwise the attrdir would be created 3129 * (assuming server fs isn't mounted readonly locally). If 3130 * VOP_LOOKUP returns ENOENT in this case, the error will 3131 * be translated into EROFS. ENOSYS is mapped to ENOTSUP 3132 * because specfs has no VOP_LOOKUP op, so the macro would 3133 * return ENOSYS. EINVAL is returned by all (current) 3134 * Solaris file system implementations when any of their 3135 * restrictions are violated (xattr(dir) can't have xattrdir). 3136 * Returning NOTSUPP is more appropriate in this case 3137 * because the object will never be able to have an attrdir. 3138 */ 3139 if (args->createdir && ! (exp_ro = rdonly4(req, cs))) 3140 lookup_flags |= CREATE_XATTR_DIR; 3141 3142 error = VOP_LOOKUP(cs->vp, "", &avp, NULL, lookup_flags, NULL, cs->cr, 3143 NULL, NULL, NULL); 3144 3145 if (error) { 3146 if (error == ENOENT && args->createdir && exp_ro) 3147 *cs->statusp = resp->status = puterrno4(EROFS); 3148 else if (error == EINVAL || error == ENOSYS) 3149 *cs->statusp = resp->status = puterrno4(ENOTSUP); 3150 else 3151 *cs->statusp = resp->status = puterrno4(error); 3152 goto out; 3153 } 3154 3155 ASSERT(avp->v_flag & V_XATTRDIR); 3156 3157 error = makefh4(&cs->fh, avp, cs->exi); 3158 3159 if (error) { 3160 VN_RELE(avp); 3161 *cs->statusp = resp->status = puterrno4(error); 3162 goto out; 3163 } 3164 3165 VN_RELE(cs->vp); 3166 cs->vp = avp; 3167 3168 /* 3169 * There is no requirement for an attrdir fh flag 3170 * because the attrdir has a vnode flag to distinguish 3171 * it from regular (non-xattr) directories. The 3172 * FH4_ATTRDIR flag is set for future sanity checks. 3173 */ 3174 set_fh4_flag(&cs->fh, FH4_ATTRDIR); 3175 *cs->statusp = resp->status = NFS4_OK; 3176 3177 out: 3178 DTRACE_NFSV4_2(op__openattr__done, struct compound_state *, cs, 3179 OPENATTR4res *, resp); 3180 } 3181 3182 static int 3183 do_io(int direction, vnode_t *vp, struct uio *uio, int ioflag, cred_t *cred, 3184 caller_context_t *ct) 3185 { 3186 int error; 3187 int i; 3188 clock_t delaytime; 3189 3190 delaytime = MSEC_TO_TICK_ROUNDUP(rfs4_lock_delay); 3191 3192 /* 3193 * Don't block on mandatory locks. If this routine returns 3194 * EAGAIN, the caller should return NFS4ERR_LOCKED. 3195 */ 3196 uio->uio_fmode = FNONBLOCK; 3197 3198 for (i = 0; i < rfs4_maxlock_tries; i++) { 3199 3200 3201 if (direction == FREAD) { 3202 (void) VOP_RWLOCK(vp, V_WRITELOCK_FALSE, ct); 3203 error = VOP_READ(vp, uio, ioflag, cred, ct); 3204 VOP_RWUNLOCK(vp, V_WRITELOCK_FALSE, ct); 3205 } else { 3206 (void) VOP_RWLOCK(vp, V_WRITELOCK_TRUE, ct); 3207 error = VOP_WRITE(vp, uio, ioflag, cred, ct); 3208 VOP_RWUNLOCK(vp, V_WRITELOCK_TRUE, ct); 3209 } 3210 3211 if (error != EAGAIN) 3212 break; 3213 3214 if (i < rfs4_maxlock_tries - 1) { 3215 delay(delaytime); 3216 delaytime *= 2; 3217 } 3218 } 3219 3220 return (error); 3221 } 3222 3223 /* ARGSUSED */ 3224 static void 3225 rfs4_op_read(nfs_argop4 *argop, nfs_resop4 *resop, struct svc_req *req, 3226 struct compound_state *cs) 3227 { 3228 READ4args *args = &argop->nfs_argop4_u.opread; 3229 READ4res *resp = &resop->nfs_resop4_u.opread; 3230 int error; 3231 int verror; 3232 vnode_t *vp; 3233 struct vattr va; 3234 struct iovec iov, *iovp = NULL; 3235 int iovcnt; 3236 struct uio uio; 3237 u_offset_t offset; 3238 bool_t *deleg = &cs->deleg; 3239 nfsstat4 stat; 3240 int in_crit = 0; 3241 mblk_t *mp = NULL; 3242 int alloc_err = 0; 3243 int rdma_used = 0; 3244 int loaned_buffers; 3245 caller_context_t ct; 3246 struct uio *uiop; 3247 3248 DTRACE_NFSV4_2(op__read__start, struct compound_state *, cs, 3249 READ4args, args); 3250 3251 vp = cs->vp; 3252 if (vp == NULL) { 3253 *cs->statusp = resp->status = NFS4ERR_NOFILEHANDLE; 3254 goto out; 3255 } 3256 if (cs->access == CS_ACCESS_DENIED) { 3257 *cs->statusp = resp->status = NFS4ERR_ACCESS; 3258 goto out; 3259 } 3260 3261 if ((stat = rfs4_check_stateid(FREAD, vp, &args->stateid, FALSE, 3262 deleg, TRUE, &ct)) != NFS4_OK) { 3263 *cs->statusp = resp->status = stat; 3264 goto out; 3265 } 3266 3267 /* 3268 * Enter the critical region before calling VOP_RWLOCK 3269 * to avoid a deadlock with write requests. 3270 */ 3271 if (nbl_need_check(vp)) { 3272 nbl_start_crit(vp, RW_READER); 3273 in_crit = 1; 3274 if (nbl_conflict(vp, NBL_READ, args->offset, args->count, 0, 3275 &ct)) { 3276 *cs->statusp = resp->status = NFS4ERR_LOCKED; 3277 goto out; 3278 } 3279 } 3280 3281 if (args->wlist) { 3282 if (args->count > clist_len(args->wlist)) { 3283 *cs->statusp = resp->status = NFS4ERR_INVAL; 3284 goto out; 3285 } 3286 rdma_used = 1; 3287 } 3288 3289 /* use loaned buffers for TCP */ 3290 loaned_buffers = (nfs_loaned_buffers && !rdma_used) ? 1 : 0; 3291 3292 va.va_mask = AT_MODE|AT_SIZE|AT_UID; 3293 verror = VOP_GETATTR(vp, &va, 0, cs->cr, &ct); 3294 3295 /* 3296 * If we can't get the attributes, then we can't do the 3297 * right access checking. So, we'll fail the request. 3298 */ 3299 if (verror) { 3300 *cs->statusp = resp->status = puterrno4(verror); 3301 goto out; 3302 } 3303 3304 if (vp->v_type != VREG) { 3305 *cs->statusp = resp->status = 3306 ((vp->v_type == VDIR) ? NFS4ERR_ISDIR : NFS4ERR_INVAL); 3307 goto out; 3308 } 3309 3310 if (crgetuid(cs->cr) != va.va_uid && 3311 (error = VOP_ACCESS(vp, VREAD, 0, cs->cr, &ct)) && 3312 (error = VOP_ACCESS(vp, VEXEC, 0, cs->cr, &ct))) { 3313 *cs->statusp = resp->status = puterrno4(error); 3314 goto out; 3315 } 3316 3317 if (MANDLOCK(vp, va.va_mode)) { /* XXX - V4 supports mand locking */ 3318 *cs->statusp = resp->status = NFS4ERR_ACCESS; 3319 goto out; 3320 } 3321 3322 offset = args->offset; 3323 if (offset >= va.va_size) { 3324 *cs->statusp = resp->status = NFS4_OK; 3325 resp->eof = TRUE; 3326 resp->data_len = 0; 3327 resp->data_val = NULL; 3328 resp->mblk = NULL; 3329 /* RDMA */ 3330 resp->wlist = args->wlist; 3331 resp->wlist_len = resp->data_len; 3332 *cs->statusp = resp->status = NFS4_OK; 3333 if (resp->wlist) 3334 clist_zero_len(resp->wlist); 3335 goto out; 3336 } 3337 3338 if (args->count == 0) { 3339 *cs->statusp = resp->status = NFS4_OK; 3340 resp->eof = FALSE; 3341 resp->data_len = 0; 3342 resp->data_val = NULL; 3343 resp->mblk = NULL; 3344 /* RDMA */ 3345 resp->wlist = args->wlist; 3346 resp->wlist_len = resp->data_len; 3347 if (resp->wlist) 3348 clist_zero_len(resp->wlist); 3349 goto out; 3350 } 3351 3352 /* 3353 * Do not allocate memory more than maximum allowed 3354 * transfer size 3355 */ 3356 if (args->count > rfs4_tsize(req)) 3357 args->count = rfs4_tsize(req); 3358 3359 if (loaned_buffers) { 3360 uiop = (uio_t *)rfs_setup_xuio(vp); 3361 ASSERT(uiop != NULL); 3362 uiop->uio_segflg = UIO_SYSSPACE; 3363 uiop->uio_loffset = args->offset; 3364 uiop->uio_resid = args->count; 3365 3366 /* Jump to do the read if successful */ 3367 if (!VOP_REQZCBUF(vp, UIO_READ, (xuio_t *)uiop, cs->cr, &ct)) { 3368 /* 3369 * Need to hold the vnode until after VOP_RETZCBUF() 3370 * is called. 3371 */ 3372 VN_HOLD(vp); 3373 goto doio_read; 3374 } 3375 3376 DTRACE_PROBE2(nfss__i__reqzcbuf_failed, int, 3377 uiop->uio_loffset, int, uiop->uio_resid); 3378 3379 uiop->uio_extflg = 0; 3380 3381 /* failure to setup for zero copy */ 3382 rfs_free_xuio((void *)uiop); 3383 loaned_buffers = 0; 3384 } 3385 3386 /* 3387 * If returning data via RDMA Write, then grab the chunk list. If we 3388 * aren't returning READ data w/RDMA_WRITE, then grab a mblk. 3389 */ 3390 if (rdma_used) { 3391 mp = NULL; 3392 (void) rdma_get_wchunk(req, &iov, args->wlist); 3393 uio.uio_iov = &iov; 3394 uio.uio_iovcnt = 1; 3395 } else { 3396 /* 3397 * mp will contain the data to be sent out in the read reply. 3398 * It will be freed after the reply has been sent. 3399 */ 3400 mp = rfs_read_alloc(args->count, &iovp, &iovcnt); 3401 ASSERT(mp != NULL); 3402 ASSERT(alloc_err == 0); 3403 uio.uio_iov = iovp; 3404 uio.uio_iovcnt = iovcnt; 3405 } 3406 3407 uio.uio_segflg = UIO_SYSSPACE; 3408 uio.uio_extflg = UIO_COPY_CACHED; 3409 uio.uio_loffset = args->offset; 3410 uio.uio_resid = args->count; 3411 uiop = &uio; 3412 3413 doio_read: 3414 error = do_io(FREAD, vp, uiop, 0, cs->cr, &ct); 3415 3416 va.va_mask = AT_SIZE; 3417 verror = VOP_GETATTR(vp, &va, 0, cs->cr, &ct); 3418 3419 if (error) { 3420 if (mp) 3421 freemsg(mp); 3422 *cs->statusp = resp->status = puterrno4(error); 3423 goto out; 3424 } 3425 3426 /* make mblk using zc buffers */ 3427 if (loaned_buffers) { 3428 mp = uio_to_mblk(uiop); 3429 ASSERT(mp != NULL); 3430 } 3431 3432 *cs->statusp = resp->status = NFS4_OK; 3433 3434 ASSERT(uiop->uio_resid >= 0); 3435 resp->data_len = args->count - uiop->uio_resid; 3436 if (mp) { 3437 resp->data_val = (char *)mp->b_datap->db_base; 3438 rfs_rndup_mblks(mp, resp->data_len, loaned_buffers); 3439 } else { 3440 resp->data_val = (caddr_t)iov.iov_base; 3441 } 3442 3443 resp->mblk = mp; 3444 3445 if (!verror && offset + resp->data_len == va.va_size) 3446 resp->eof = TRUE; 3447 else 3448 resp->eof = FALSE; 3449 3450 if (rdma_used) { 3451 if (!rdma_setup_read_data4(args, resp)) { 3452 *cs->statusp = resp->status = NFS4ERR_INVAL; 3453 } 3454 } else { 3455 resp->wlist = NULL; 3456 } 3457 3458 out: 3459 if (in_crit) 3460 nbl_end_crit(vp); 3461 3462 if (iovp != NULL) 3463 kmem_free(iovp, iovcnt * sizeof (struct iovec)); 3464 3465 DTRACE_NFSV4_2(op__read__done, struct compound_state *, cs, 3466 READ4res *, resp); 3467 } 3468 3469 static void 3470 rfs4_op_read_free(nfs_resop4 *resop) 3471 { 3472 READ4res *resp = &resop->nfs_resop4_u.opread; 3473 3474 if (resp->status == NFS4_OK && resp->mblk != NULL) { 3475 freemsg(resp->mblk); 3476 resp->mblk = NULL; 3477 resp->data_val = NULL; 3478 resp->data_len = 0; 3479 } 3480 } 3481 3482 static void 3483 rfs4_op_readdir_free(nfs_resop4 * resop) 3484 { 3485 READDIR4res *resp = &resop->nfs_resop4_u.opreaddir; 3486 3487 if (resp->status == NFS4_OK && resp->mblk != NULL) { 3488 freeb(resp->mblk); 3489 resp->mblk = NULL; 3490 resp->data_len = 0; 3491 } 3492 } 3493 3494 3495 /* ARGSUSED */ 3496 static void 3497 rfs4_op_putpubfh(nfs_argop4 *args, nfs_resop4 *resop, struct svc_req *req, 3498 struct compound_state *cs) 3499 { 3500 PUTPUBFH4res *resp = &resop->nfs_resop4_u.opputpubfh; 3501 int error; 3502 vnode_t *vp; 3503 struct exportinfo *exi, *sav_exi; 3504 nfs_fh4_fmt_t *fh_fmtp; 3505 nfs_export_t *ne = nfs_get_export(); 3506 3507 DTRACE_NFSV4_1(op__putpubfh__start, struct compound_state *, cs); 3508 3509 if (cs->vp) { 3510 VN_RELE(cs->vp); 3511 cs->vp = NULL; 3512 } 3513 3514 if (cs->cr) 3515 crfree(cs->cr); 3516 3517 cs->cr = crdup(cs->basecr); 3518 3519 vp = ne->exi_public->exi_vp; 3520 if (vp == NULL) { 3521 *cs->statusp = resp->status = NFS4ERR_SERVERFAULT; 3522 goto out; 3523 } 3524 3525 error = makefh4(&cs->fh, vp, ne->exi_public); 3526 if (error != 0) { 3527 *cs->statusp = resp->status = puterrno4(error); 3528 goto out; 3529 } 3530 sav_exi = cs->exi; 3531 if (ne->exi_public == ne->exi_root) { 3532 /* 3533 * No filesystem is actually shared public, so we default 3534 * to exi_root. In this case, we must check whether root 3535 * is exported. 3536 */ 3537 fh_fmtp = (nfs_fh4_fmt_t *)cs->fh.nfs_fh4_val; 3538 3539 /* 3540 * if root filesystem is exported, the exportinfo struct that we 3541 * should use is what checkexport4 returns, because root_exi is 3542 * actually a mostly empty struct. 3543 */ 3544 exi = checkexport4(&fh_fmtp->fh4_fsid, 3545 (fid_t *)&fh_fmtp->fh4_xlen, NULL); 3546 cs->exi = ((exi != NULL) ? exi : ne->exi_public); 3547 } else { 3548 /* 3549 * it's a properly shared filesystem 3550 */ 3551 cs->exi = ne->exi_public; 3552 } 3553 3554 if (is_system_labeled()) { 3555 bslabel_t *clabel; 3556 3557 ASSERT(req->rq_label != NULL); 3558 clabel = req->rq_label; 3559 DTRACE_PROBE2(tx__rfs4__log__info__opputpubfh__clabel, char *, 3560 "got client label from request(1)", 3561 struct svc_req *, req); 3562 if (!blequal(&l_admin_low->tsl_label, clabel)) { 3563 if (!do_rfs_label_check(clabel, vp, DOMINANCE_CHECK, 3564 cs->exi)) { 3565 *cs->statusp = resp->status = 3566 NFS4ERR_SERVERFAULT; 3567 goto out; 3568 } 3569 } 3570 } 3571 3572 VN_HOLD(vp); 3573 cs->vp = vp; 3574 3575 if ((resp->status = call_checkauth4(cs, req)) != NFS4_OK) { 3576 VN_RELE(cs->vp); 3577 cs->vp = NULL; 3578 cs->exi = sav_exi; 3579 goto out; 3580 } 3581 3582 *cs->statusp = resp->status = NFS4_OK; 3583 out: 3584 DTRACE_NFSV4_2(op__putpubfh__done, struct compound_state *, cs, 3585 PUTPUBFH4res *, resp); 3586 } 3587 3588 /* 3589 * XXX - issue with put*fh operations. Suppose /export/home is exported. 3590 * Suppose an NFS client goes to mount /export/home/joe. If /export, home, 3591 * or joe have restrictive search permissions, then we shouldn't let 3592 * the client get a file handle. This is easy to enforce. However, we 3593 * don't know what security flavor should be used until we resolve the 3594 * path name. Another complication is uid mapping. If root is 3595 * the user, then it will be mapped to the anonymous user by default, 3596 * but we won't know that till we've resolved the path name. And we won't 3597 * know what the anonymous user is. 3598 * Luckily, SECINFO is specified to take a full filename. 3599 * So what we will have to in rfs4_op_lookup is check that flavor of 3600 * the target object matches that of the request, and if root was the 3601 * caller, check for the root= and anon= options, and if necessary, 3602 * repeat the lookup using the right cred_t. But that's not done yet. 3603 */ 3604 /* ARGSUSED */ 3605 static void 3606 rfs4_op_putfh(nfs_argop4 *argop, nfs_resop4 *resop, struct svc_req *req, 3607 struct compound_state *cs) 3608 { 3609 PUTFH4args *args = &argop->nfs_argop4_u.opputfh; 3610 PUTFH4res *resp = &resop->nfs_resop4_u.opputfh; 3611 nfs_fh4_fmt_t *fh_fmtp; 3612 3613 DTRACE_NFSV4_2(op__putfh__start, struct compound_state *, cs, 3614 PUTFH4args *, args); 3615 3616 if (cs->vp) { 3617 VN_RELE(cs->vp); 3618 cs->vp = NULL; 3619 } 3620 3621 if (cs->cr) { 3622 crfree(cs->cr); 3623 cs->cr = NULL; 3624 } 3625 3626 3627 if (args->object.nfs_fh4_len < NFS_FH4_LEN) { 3628 *cs->statusp = resp->status = NFS4ERR_BADHANDLE; 3629 goto out; 3630 } 3631 3632 fh_fmtp = (nfs_fh4_fmt_t *)args->object.nfs_fh4_val; 3633 cs->exi = checkexport4(&fh_fmtp->fh4_fsid, (fid_t *)&fh_fmtp->fh4_xlen, 3634 NULL); 3635 3636 if (cs->exi == NULL) { 3637 *cs->statusp = resp->status = NFS4ERR_STALE; 3638 goto out; 3639 } 3640 3641 cs->cr = crdup(cs->basecr); 3642 3643 ASSERT(cs->cr != NULL); 3644 3645 if (! (cs->vp = nfs4_fhtovp(&args->object, cs->exi, &resp->status))) { 3646 *cs->statusp = resp->status; 3647 goto out; 3648 } 3649 3650 if ((resp->status = call_checkauth4(cs, req)) != NFS4_OK) { 3651 VN_RELE(cs->vp); 3652 cs->vp = NULL; 3653 goto out; 3654 } 3655 3656 nfs_fh4_copy(&args->object, &cs->fh); 3657 *cs->statusp = resp->status = NFS4_OK; 3658 cs->deleg = FALSE; 3659 3660 out: 3661 DTRACE_NFSV4_2(op__putfh__done, struct compound_state *, cs, 3662 PUTFH4res *, resp); 3663 } 3664 3665 /* ARGSUSED */ 3666 static void 3667 rfs4_op_putrootfh(nfs_argop4 *argop, nfs_resop4 *resop, struct svc_req *req, 3668 struct compound_state *cs) 3669 { 3670 PUTROOTFH4res *resp = &resop->nfs_resop4_u.opputrootfh; 3671 int error; 3672 fid_t fid; 3673 struct exportinfo *exi, *sav_exi; 3674 3675 DTRACE_NFSV4_1(op__putrootfh__start, struct compound_state *, cs); 3676 3677 if (cs->vp) { 3678 VN_RELE(cs->vp); 3679 cs->vp = NULL; 3680 } 3681 3682 if (cs->cr) 3683 crfree(cs->cr); 3684 3685 cs->cr = crdup(cs->basecr); 3686 3687 /* 3688 * Using rootdir, the system root vnode, 3689 * get its fid. 3690 */ 3691 bzero(&fid, sizeof (fid)); 3692 fid.fid_len = MAXFIDSZ; 3693 error = vop_fid_pseudo(ZONE_ROOTVP(), &fid); 3694 if (error != 0) { 3695 *cs->statusp = resp->status = puterrno4(error); 3696 goto out; 3697 } 3698 3699 /* 3700 * Then use the root fsid & fid it to find out if it's exported 3701 * 3702 * If the server root isn't exported directly, then 3703 * it should at least be a pseudo export based on 3704 * one or more exports further down in the server's 3705 * file tree. 3706 */ 3707 exi = checkexport4(&ZONE_ROOTVP()->v_vfsp->vfs_fsid, &fid, NULL); 3708 if (exi == NULL || exi->exi_export.ex_flags & EX_PUBLIC) { 3709 NFS4_DEBUG(rfs4_debug, 3710 (CE_WARN, "rfs4_op_putrootfh: export check failure")); 3711 *cs->statusp = resp->status = NFS4ERR_SERVERFAULT; 3712 goto out; 3713 } 3714 3715 /* 3716 * Now make a filehandle based on the root 3717 * export and root vnode. 3718 */ 3719 error = makefh4(&cs->fh, ZONE_ROOTVP(), exi); 3720 if (error != 0) { 3721 *cs->statusp = resp->status = puterrno4(error); 3722 goto out; 3723 } 3724 3725 sav_exi = cs->exi; 3726 cs->exi = exi; 3727 3728 VN_HOLD(ZONE_ROOTVP()); 3729 cs->vp = ZONE_ROOTVP(); 3730 3731 if ((resp->status = call_checkauth4(cs, req)) != NFS4_OK) { 3732 VN_RELE(cs->vp); 3733 cs->vp = NULL; 3734 cs->exi = sav_exi; 3735 goto out; 3736 } 3737 3738 *cs->statusp = resp->status = NFS4_OK; 3739 cs->deleg = FALSE; 3740 out: 3741 DTRACE_NFSV4_2(op__putrootfh__done, struct compound_state *, cs, 3742 PUTROOTFH4res *, resp); 3743 } 3744 3745 /* 3746 * readlink: args: CURRENT_FH. 3747 * res: status. If success - CURRENT_FH unchanged, return linktext. 3748 */ 3749 3750 /* ARGSUSED */ 3751 static void 3752 rfs4_op_readlink(nfs_argop4 *argop, nfs_resop4 *resop, struct svc_req *req, 3753 struct compound_state *cs) 3754 { 3755 READLINK4res *resp = &resop->nfs_resop4_u.opreadlink; 3756 int error; 3757 vnode_t *vp; 3758 struct iovec iov; 3759 struct vattr va; 3760 struct uio uio; 3761 char *data; 3762 struct sockaddr *ca; 3763 char *name = NULL; 3764 int is_referral; 3765 3766 DTRACE_NFSV4_1(op__readlink__start, struct compound_state *, cs); 3767 3768 /* CURRENT_FH: directory */ 3769 vp = cs->vp; 3770 if (vp == NULL) { 3771 *cs->statusp = resp->status = NFS4ERR_NOFILEHANDLE; 3772 goto out; 3773 } 3774 3775 if (cs->access == CS_ACCESS_DENIED) { 3776 *cs->statusp = resp->status = NFS4ERR_ACCESS; 3777 goto out; 3778 } 3779 3780 /* Is it a referral? */ 3781 if (vn_is_nfs_reparse(vp, cs->cr) && client_is_downrev(req)) { 3782 3783 is_referral = 1; 3784 3785 } else { 3786 3787 is_referral = 0; 3788 3789 if (vp->v_type == VDIR) { 3790 *cs->statusp = resp->status = NFS4ERR_ISDIR; 3791 goto out; 3792 } 3793 3794 if (vp->v_type != VLNK) { 3795 *cs->statusp = resp->status = NFS4ERR_INVAL; 3796 goto out; 3797 } 3798 3799 } 3800 3801 va.va_mask = AT_MODE; 3802 error = VOP_GETATTR(vp, &va, 0, cs->cr, NULL); 3803 if (error) { 3804 *cs->statusp = resp->status = puterrno4(error); 3805 goto out; 3806 } 3807 3808 if (MANDLOCK(vp, va.va_mode)) { 3809 *cs->statusp = resp->status = NFS4ERR_ACCESS; 3810 goto out; 3811 } 3812 3813 data = kmem_alloc(MAXPATHLEN + 1, KM_SLEEP); 3814 3815 if (is_referral) { 3816 char *s; 3817 size_t strsz; 3818 kstat_named_t *stat = 3819 cs->exi->exi_ne->ne_globals->svstat[NFS_V4]; 3820 3821 /* Get an artificial symlink based on a referral */ 3822 s = build_symlink(vp, cs->cr, &strsz); 3823 stat[NFS_REFERLINKS].value.ui64++; 3824 DTRACE_PROBE2(nfs4serv__func__referral__reflink, 3825 vnode_t *, vp, char *, s); 3826 if (s == NULL) 3827 error = EINVAL; 3828 else { 3829 error = 0; 3830 (void) strlcpy(data, s, MAXPATHLEN + 1); 3831 kmem_free(s, strsz); 3832 } 3833 3834 } else { 3835 3836 iov.iov_base = data; 3837 iov.iov_len = MAXPATHLEN; 3838 uio.uio_iov = &iov; 3839 uio.uio_iovcnt = 1; 3840 uio.uio_segflg = UIO_SYSSPACE; 3841 uio.uio_extflg = UIO_COPY_CACHED; 3842 uio.uio_loffset = 0; 3843 uio.uio_resid = MAXPATHLEN; 3844 3845 error = VOP_READLINK(vp, &uio, cs->cr, NULL); 3846 3847 if (!error) 3848 *(data + MAXPATHLEN - uio.uio_resid) = '\0'; 3849 } 3850 3851 if (error) { 3852 kmem_free((caddr_t)data, (uint_t)MAXPATHLEN + 1); 3853 *cs->statusp = resp->status = puterrno4(error); 3854 goto out; 3855 } 3856 3857 ca = (struct sockaddr *)svc_getrpccaller(req->rq_xprt)->buf; 3858 name = nfscmd_convname(ca, cs->exi, data, NFSCMD_CONV_OUTBOUND, 3859 MAXPATHLEN + 1); 3860 3861 if (name == NULL) { 3862 /* 3863 * Even though the conversion failed, we return 3864 * something. We just don't translate it. 3865 */ 3866 name = data; 3867 } 3868 3869 /* 3870 * treat link name as data 3871 */ 3872 (void) str_to_utf8(name, (utf8string *)&resp->link); 3873 3874 if (name != data) 3875 kmem_free(name, MAXPATHLEN + 1); 3876 kmem_free((caddr_t)data, (uint_t)MAXPATHLEN + 1); 3877 *cs->statusp = resp->status = NFS4_OK; 3878 3879 out: 3880 DTRACE_NFSV4_2(op__readlink__done, struct compound_state *, cs, 3881 READLINK4res *, resp); 3882 } 3883 3884 static void 3885 rfs4_op_readlink_free(nfs_resop4 *resop) 3886 { 3887 READLINK4res *resp = &resop->nfs_resop4_u.opreadlink; 3888 utf8string *symlink = (utf8string *)&resp->link; 3889 3890 if (symlink->utf8string_val) { 3891 UTF8STRING_FREE(*symlink) 3892 } 3893 } 3894 3895 /* 3896 * release_lockowner: 3897 * Release any state associated with the supplied 3898 * lockowner. Note if any lo_state is holding locks we will not 3899 * rele that lo_state and thus the lockowner will not be destroyed. 3900 * A client using lock after the lock owner stateid has been released 3901 * will suffer the consequence of NFS4ERR_BAD_STATEID and would have 3902 * to reissue the lock with new_lock_owner set to TRUE. 3903 * args: lock_owner 3904 * res: status 3905 */ 3906 /* ARGSUSED */ 3907 static void 3908 rfs4_op_release_lockowner(nfs_argop4 *argop, nfs_resop4 *resop, 3909 struct svc_req *req, struct compound_state *cs) 3910 { 3911 RELEASE_LOCKOWNER4args *ap = &argop->nfs_argop4_u.oprelease_lockowner; 3912 RELEASE_LOCKOWNER4res *resp = &resop->nfs_resop4_u.oprelease_lockowner; 3913 rfs4_lockowner_t *lo; 3914 rfs4_openowner_t *oo; 3915 rfs4_state_t *sp; 3916 rfs4_lo_state_t *lsp; 3917 rfs4_client_t *cp; 3918 bool_t create = FALSE; 3919 locklist_t *llist; 3920 sysid_t sysid; 3921 3922 DTRACE_NFSV4_2(op__release__lockowner__start, struct compound_state *, 3923 cs, RELEASE_LOCKOWNER4args *, ap); 3924 3925 /* Make sure there is a clientid around for this request */ 3926 cp = rfs4_findclient_by_id(ap->lock_owner.clientid, FALSE); 3927 3928 if (cp == NULL) { 3929 *cs->statusp = resp->status = 3930 rfs4_check_clientid(&ap->lock_owner.clientid, 0); 3931 goto out; 3932 } 3933 rfs4_client_rele(cp); 3934 3935 lo = rfs4_findlockowner(&ap->lock_owner, &create); 3936 if (lo == NULL) { 3937 *cs->statusp = resp->status = NFS4_OK; 3938 goto out; 3939 } 3940 ASSERT(lo->rl_client != NULL); 3941 3942 /* 3943 * Check for EXPIRED client. If so will reap state with in a lease 3944 * period or on next set_clientid_confirm step 3945 */ 3946 if (rfs4_lease_expired(lo->rl_client)) { 3947 rfs4_lockowner_rele(lo); 3948 *cs->statusp = resp->status = NFS4ERR_EXPIRED; 3949 goto out; 3950 } 3951 3952 /* 3953 * If no sysid has been assigned, then no locks exist; just return. 3954 */ 3955 rfs4_dbe_lock(lo->rl_client->rc_dbe); 3956 if (lo->rl_client->rc_sysidt == LM_NOSYSID) { 3957 rfs4_lockowner_rele(lo); 3958 rfs4_dbe_unlock(lo->rl_client->rc_dbe); 3959 goto out; 3960 } 3961 3962 sysid = lo->rl_client->rc_sysidt; 3963 rfs4_dbe_unlock(lo->rl_client->rc_dbe); 3964 3965 /* 3966 * Mark the lockowner invalid. 3967 */ 3968 rfs4_dbe_hide(lo->rl_dbe); 3969 3970 /* 3971 * sysid-pid pair should now not be used since the lockowner is 3972 * invalid. If the client were to instantiate the lockowner again 3973 * it would be assigned a new pid. Thus we can get the list of 3974 * current locks. 3975 */ 3976 3977 llist = flk_get_active_locks(sysid, lo->rl_pid); 3978 /* If we are still holding locks fail */ 3979 if (llist != NULL) { 3980 3981 *cs->statusp = resp->status = NFS4ERR_LOCKS_HELD; 3982 3983 flk_free_locklist(llist); 3984 /* 3985 * We need to unhide the lockowner so the client can 3986 * try it again. The bad thing here is if the client 3987 * has a logic error that took it here in the first place 3988 * they probably have lost accounting of the locks that it 3989 * is holding. So we may have dangling state until the 3990 * open owner state is reaped via close. One scenario 3991 * that could possibly occur is that the client has 3992 * sent the unlock request(s) in separate threads 3993 * and has not waited for the replies before sending the 3994 * RELEASE_LOCKOWNER request. Presumably, it would expect 3995 * and deal appropriately with NFS4ERR_LOCKS_HELD, by 3996 * reissuing the request. 3997 */ 3998 rfs4_dbe_unhide(lo->rl_dbe); 3999 rfs4_lockowner_rele(lo); 4000 goto out; 4001 } 4002 4003 /* 4004 * For the corresponding client we need to check each open 4005 * owner for any opens that have lockowner state associated 4006 * with this lockowner. 4007 */ 4008 4009 rfs4_dbe_lock(lo->rl_client->rc_dbe); 4010 for (oo = list_head(&lo->rl_client->rc_openownerlist); oo != NULL; 4011 oo = list_next(&lo->rl_client->rc_openownerlist, oo)) { 4012 4013 rfs4_dbe_lock(oo->ro_dbe); 4014 for (sp = list_head(&oo->ro_statelist); sp != NULL; 4015 sp = list_next(&oo->ro_statelist, sp)) { 4016 4017 rfs4_dbe_lock(sp->rs_dbe); 4018 for (lsp = list_head(&sp->rs_lostatelist); 4019 lsp != NULL; 4020 lsp = list_next(&sp->rs_lostatelist, lsp)) { 4021 if (lsp->rls_locker == lo) { 4022 rfs4_dbe_lock(lsp->rls_dbe); 4023 rfs4_dbe_invalidate(lsp->rls_dbe); 4024 rfs4_dbe_unlock(lsp->rls_dbe); 4025 } 4026 } 4027 rfs4_dbe_unlock(sp->rs_dbe); 4028 } 4029 rfs4_dbe_unlock(oo->ro_dbe); 4030 } 4031 rfs4_dbe_unlock(lo->rl_client->rc_dbe); 4032 4033 rfs4_lockowner_rele(lo); 4034 4035 *cs->statusp = resp->status = NFS4_OK; 4036 4037 out: 4038 DTRACE_NFSV4_2(op__release__lockowner__done, struct compound_state *, 4039 cs, RELEASE_LOCKOWNER4res *, resp); 4040 } 4041 4042 /* 4043 * short utility function to lookup a file and recall the delegation 4044 */ 4045 static rfs4_file_t * 4046 rfs4_lookup_and_findfile(vnode_t *dvp, char *nm, vnode_t **vpp, 4047 int *lkup_error, cred_t *cr) 4048 { 4049 vnode_t *vp; 4050 rfs4_file_t *fp = NULL; 4051 bool_t fcreate = FALSE; 4052 int error; 4053 4054 if (vpp) 4055 *vpp = NULL; 4056 4057 if ((error = VOP_LOOKUP(dvp, nm, &vp, NULL, 0, NULL, cr, NULL, NULL, 4058 NULL)) == 0) { 4059 if (vp->v_type == VREG) 4060 fp = rfs4_findfile(vp, NULL, &fcreate); 4061 if (vpp) 4062 *vpp = vp; 4063 else 4064 VN_RELE(vp); 4065 } 4066 4067 if (lkup_error) 4068 *lkup_error = error; 4069 4070 return (fp); 4071 } 4072 4073 /* 4074 * remove: args: CURRENT_FH: directory; name. 4075 * res: status. If success - CURRENT_FH unchanged, return change_info 4076 * for directory. 4077 */ 4078 /* ARGSUSED */ 4079 static void 4080 rfs4_op_remove(nfs_argop4 *argop, nfs_resop4 *resop, struct svc_req *req, 4081 struct compound_state *cs) 4082 { 4083 REMOVE4args *args = &argop->nfs_argop4_u.opremove; 4084 REMOVE4res *resp = &resop->nfs_resop4_u.opremove; 4085 int error; 4086 vnode_t *dvp, *vp; 4087 struct vattr bdva, idva, adva; 4088 char *nm; 4089 uint_t len; 4090 rfs4_file_t *fp; 4091 int in_crit = 0; 4092 bslabel_t *clabel; 4093 struct sockaddr *ca; 4094 char *name = NULL; 4095 nfsstat4 status; 4096 4097 DTRACE_NFSV4_2(op__remove__start, struct compound_state *, cs, 4098 REMOVE4args *, args); 4099 4100 /* CURRENT_FH: directory */ 4101 dvp = cs->vp; 4102 if (dvp == NULL) { 4103 *cs->statusp = resp->status = NFS4ERR_NOFILEHANDLE; 4104 goto out; 4105 } 4106 4107 if (cs->access == CS_ACCESS_DENIED) { 4108 *cs->statusp = resp->status = NFS4ERR_ACCESS; 4109 goto out; 4110 } 4111 4112 /* 4113 * If there is an unshared filesystem mounted on this vnode, 4114 * Do not allow to remove anything in this directory. 4115 */ 4116 if (vn_ismntpt(dvp)) { 4117 *cs->statusp = resp->status = NFS4ERR_ACCESS; 4118 goto out; 4119 } 4120 4121 if (dvp->v_type != VDIR) { 4122 *cs->statusp = resp->status = NFS4ERR_NOTDIR; 4123 goto out; 4124 } 4125 4126 status = utf8_dir_verify(&args->target); 4127 if (status != NFS4_OK) { 4128 *cs->statusp = resp->status = status; 4129 goto out; 4130 } 4131 4132 /* 4133 * Lookup the file so that we can check if it's a directory 4134 */ 4135 nm = utf8_to_fn(&args->target, &len, NULL); 4136 if (nm == NULL) { 4137 *cs->statusp = resp->status = NFS4ERR_INVAL; 4138 goto out; 4139 } 4140 4141 if (len > MAXNAMELEN) { 4142 *cs->statusp = resp->status = NFS4ERR_NAMETOOLONG; 4143 kmem_free(nm, len); 4144 goto out; 4145 } 4146 4147 if (rdonly4(req, cs)) { 4148 *cs->statusp = resp->status = NFS4ERR_ROFS; 4149 kmem_free(nm, len); 4150 goto out; 4151 } 4152 4153 ca = (struct sockaddr *)svc_getrpccaller(req->rq_xprt)->buf; 4154 name = nfscmd_convname(ca, cs->exi, nm, NFSCMD_CONV_INBOUND, 4155 MAXPATHLEN + 1); 4156 4157 if (name == NULL) { 4158 *cs->statusp = resp->status = NFS4ERR_INVAL; 4159 kmem_free(nm, len); 4160 goto out; 4161 } 4162 4163 /* 4164 * Lookup the file to determine type and while we are see if 4165 * there is a file struct around and check for delegation. 4166 * We don't need to acquire va_seq before this lookup, if 4167 * it causes an update, cinfo.before will not match, which will 4168 * trigger a cache flush even if atomic is TRUE. 4169 */ 4170 fp = rfs4_lookup_and_findfile(dvp, name, &vp, &error, cs->cr); 4171 if (fp != NULL) { 4172 if (rfs4_check_delegated_byfp(FWRITE, fp, TRUE, TRUE, TRUE, 4173 NULL)) { 4174 VN_RELE(vp); 4175 rfs4_file_rele(fp); 4176 *cs->statusp = resp->status = NFS4ERR_DELAY; 4177 if (nm != name) 4178 kmem_free(name, MAXPATHLEN + 1); 4179 kmem_free(nm, len); 4180 goto out; 4181 } 4182 } 4183 4184 /* Didn't find anything to remove */ 4185 if (vp == NULL) { 4186 *cs->statusp = resp->status = error; 4187 if (nm != name) 4188 kmem_free(name, MAXPATHLEN + 1); 4189 kmem_free(nm, len); 4190 goto out; 4191 } 4192 4193 if (nbl_need_check(vp)) { 4194 nbl_start_crit(vp, RW_READER); 4195 in_crit = 1; 4196 if (nbl_conflict(vp, NBL_REMOVE, 0, 0, 0, NULL)) { 4197 *cs->statusp = resp->status = NFS4ERR_FILE_OPEN; 4198 if (nm != name) 4199 kmem_free(name, MAXPATHLEN + 1); 4200 kmem_free(nm, len); 4201 nbl_end_crit(vp); 4202 VN_RELE(vp); 4203 if (fp) { 4204 rfs4_clear_dont_grant(fp); 4205 rfs4_file_rele(fp); 4206 } 4207 goto out; 4208 } 4209 } 4210 4211 /* check label before allowing removal */ 4212 if (is_system_labeled()) { 4213 ASSERT(req->rq_label != NULL); 4214 clabel = req->rq_label; 4215 DTRACE_PROBE2(tx__rfs4__log__info__opremove__clabel, char *, 4216 "got client label from request(1)", 4217 struct svc_req *, req); 4218 if (!blequal(&l_admin_low->tsl_label, clabel)) { 4219 if (!do_rfs_label_check(clabel, vp, EQUALITY_CHECK, 4220 cs->exi)) { 4221 *cs->statusp = resp->status = NFS4ERR_ACCESS; 4222 if (name != nm) 4223 kmem_free(name, MAXPATHLEN + 1); 4224 kmem_free(nm, len); 4225 if (in_crit) 4226 nbl_end_crit(vp); 4227 VN_RELE(vp); 4228 if (fp) { 4229 rfs4_clear_dont_grant(fp); 4230 rfs4_file_rele(fp); 4231 } 4232 goto out; 4233 } 4234 } 4235 } 4236 4237 /* Get dir "before" change value */ 4238 bdva.va_mask = AT_CTIME|AT_SEQ; 4239 error = VOP_GETATTR(dvp, &bdva, 0, cs->cr, NULL); 4240 if (error) { 4241 *cs->statusp = resp->status = puterrno4(error); 4242 if (nm != name) 4243 kmem_free(name, MAXPATHLEN + 1); 4244 kmem_free(nm, len); 4245 if (in_crit) 4246 nbl_end_crit(vp); 4247 VN_RELE(vp); 4248 if (fp) { 4249 rfs4_clear_dont_grant(fp); 4250 rfs4_file_rele(fp); 4251 } 4252 goto out; 4253 } 4254 NFS4_SET_FATTR4_CHANGE(resp->cinfo.before, bdva.va_ctime) 4255 4256 /* Actually do the REMOVE operation */ 4257 if (vp->v_type == VDIR) { 4258 /* 4259 * Can't remove a directory that has a mounted-on filesystem. 4260 */ 4261 if (vn_ismntpt(vp)) { 4262 error = EACCES; 4263 } else { 4264 /* 4265 * System V defines rmdir to return EEXIST, 4266 * not ENOTEMPTY, if the directory is not 4267 * empty. A System V NFS server needs to map 4268 * NFS4ERR_EXIST to NFS4ERR_NOTEMPTY to 4269 * transmit over the wire. 4270 */ 4271 if ((error = VOP_RMDIR(dvp, name, ZONE_ROOTVP(), cs->cr, 4272 NULL, 0)) == EEXIST) 4273 error = ENOTEMPTY; 4274 } 4275 } else { 4276 if ((error = VOP_REMOVE(dvp, name, cs->cr, NULL, 0)) == 0 && 4277 fp != NULL) { 4278 struct vattr va; 4279 vnode_t *tvp; 4280 4281 rfs4_dbe_lock(fp->rf_dbe); 4282 tvp = fp->rf_vp; 4283 if (tvp) 4284 VN_HOLD(tvp); 4285 rfs4_dbe_unlock(fp->rf_dbe); 4286 4287 if (tvp) { 4288 /* 4289 * This is va_seq safe because we are not 4290 * manipulating dvp. 4291 */ 4292 va.va_mask = AT_NLINK; 4293 if (!VOP_GETATTR(tvp, &va, 0, cs->cr, NULL) && 4294 va.va_nlink == 0) { 4295 /* Remove state on file remove */ 4296 if (in_crit) { 4297 nbl_end_crit(vp); 4298 in_crit = 0; 4299 } 4300 rfs4_close_all_state(fp); 4301 } 4302 VN_RELE(tvp); 4303 } 4304 } 4305 } 4306 4307 if (in_crit) 4308 nbl_end_crit(vp); 4309 VN_RELE(vp); 4310 4311 if (fp) { 4312 rfs4_clear_dont_grant(fp); 4313 rfs4_file_rele(fp); 4314 } 4315 if (nm != name) 4316 kmem_free(name, MAXPATHLEN + 1); 4317 kmem_free(nm, len); 4318 4319 if (error) { 4320 *cs->statusp = resp->status = puterrno4(error); 4321 goto out; 4322 } 4323 4324 /* 4325 * Get the initial "after" sequence number, if it fails, set to zero 4326 */ 4327 idva.va_mask = AT_SEQ; 4328 if (VOP_GETATTR(dvp, &idva, 0, cs->cr, NULL)) 4329 idva.va_seq = 0; 4330 4331 /* 4332 * Force modified data and metadata out to stable storage. 4333 */ 4334 (void) VOP_FSYNC(dvp, 0, cs->cr, NULL); 4335 4336 /* 4337 * Get "after" change value, if it fails, simply return the 4338 * before value. 4339 */ 4340 adva.va_mask = AT_CTIME|AT_SEQ; 4341 if (VOP_GETATTR(dvp, &adva, 0, cs->cr, NULL)) { 4342 adva.va_ctime = bdva.va_ctime; 4343 adva.va_seq = 0; 4344 } 4345 4346 NFS4_SET_FATTR4_CHANGE(resp->cinfo.after, adva.va_ctime) 4347 4348 /* 4349 * The cinfo.atomic = TRUE only if we have 4350 * non-zero va_seq's, and it has incremented by exactly one 4351 * during the VOP_REMOVE/RMDIR and it didn't change during 4352 * the VOP_FSYNC. 4353 */ 4354 if (bdva.va_seq && idva.va_seq && adva.va_seq && 4355 idva.va_seq == (bdva.va_seq + 1) && idva.va_seq == adva.va_seq) 4356 resp->cinfo.atomic = TRUE; 4357 else 4358 resp->cinfo.atomic = FALSE; 4359 4360 *cs->statusp = resp->status = NFS4_OK; 4361 4362 out: 4363 DTRACE_NFSV4_2(op__remove__done, struct compound_state *, cs, 4364 REMOVE4res *, resp); 4365 } 4366 4367 /* 4368 * rename: args: SAVED_FH: from directory, CURRENT_FH: target directory, 4369 * oldname and newname. 4370 * res: status. If success - CURRENT_FH unchanged, return change_info 4371 * for both from and target directories. 4372 */ 4373 /* ARGSUSED */ 4374 static void 4375 rfs4_op_rename(nfs_argop4 *argop, nfs_resop4 *resop, struct svc_req *req, 4376 struct compound_state *cs) 4377 { 4378 RENAME4args *args = &argop->nfs_argop4_u.oprename; 4379 RENAME4res *resp = &resop->nfs_resop4_u.oprename; 4380 int error; 4381 vnode_t *odvp; 4382 vnode_t *ndvp; 4383 vnode_t *srcvp, *targvp, *tvp; 4384 struct vattr obdva, oidva, oadva; 4385 struct vattr nbdva, nidva, nadva; 4386 char *onm, *nnm; 4387 uint_t olen, nlen; 4388 rfs4_file_t *fp, *sfp; 4389 int in_crit_src, in_crit_targ; 4390 int fp_rele_grant_hold, sfp_rele_grant_hold; 4391 int unlinked; 4392 bslabel_t *clabel; 4393 struct sockaddr *ca; 4394 char *converted_onm = NULL; 4395 char *converted_nnm = NULL; 4396 nfsstat4 status; 4397 4398 DTRACE_NFSV4_2(op__rename__start, struct compound_state *, cs, 4399 RENAME4args *, args); 4400 4401 fp = sfp = NULL; 4402 srcvp = targvp = tvp = NULL; 4403 in_crit_src = in_crit_targ = 0; 4404 fp_rele_grant_hold = sfp_rele_grant_hold = 0; 4405 unlinked = 0; 4406 4407 /* CURRENT_FH: target directory */ 4408 ndvp = cs->vp; 4409 if (ndvp == NULL) { 4410 *cs->statusp = resp->status = NFS4ERR_NOFILEHANDLE; 4411 goto out; 4412 } 4413 4414 /* SAVED_FH: from directory */ 4415 odvp = cs->saved_vp; 4416 if (odvp == NULL) { 4417 *cs->statusp = resp->status = NFS4ERR_NOFILEHANDLE; 4418 goto out; 4419 } 4420 4421 if (cs->access == CS_ACCESS_DENIED) { 4422 *cs->statusp = resp->status = NFS4ERR_ACCESS; 4423 goto out; 4424 } 4425 4426 /* 4427 * If there is an unshared filesystem mounted on this vnode, 4428 * do not allow to rename objects in this directory. 4429 */ 4430 if (vn_ismntpt(odvp)) { 4431 *cs->statusp = resp->status = NFS4ERR_ACCESS; 4432 goto out; 4433 } 4434 4435 /* 4436 * If there is an unshared filesystem mounted on this vnode, 4437 * do not allow to rename to this directory. 4438 */ 4439 if (vn_ismntpt(ndvp)) { 4440 *cs->statusp = resp->status = NFS4ERR_ACCESS; 4441 goto out; 4442 } 4443 4444 if (odvp->v_type != VDIR || ndvp->v_type != VDIR) { 4445 *cs->statusp = resp->status = NFS4ERR_NOTDIR; 4446 goto out; 4447 } 4448 4449 if (cs->saved_exi != cs->exi) { 4450 *cs->statusp = resp->status = NFS4ERR_XDEV; 4451 goto out; 4452 } 4453 4454 status = utf8_dir_verify(&args->oldname); 4455 if (status != NFS4_OK) { 4456 *cs->statusp = resp->status = status; 4457 goto out; 4458 } 4459 4460 status = utf8_dir_verify(&args->newname); 4461 if (status != NFS4_OK) { 4462 *cs->statusp = resp->status = status; 4463 goto out; 4464 } 4465 4466 onm = utf8_to_fn(&args->oldname, &olen, NULL); 4467 if (onm == NULL) { 4468 *cs->statusp = resp->status = NFS4ERR_INVAL; 4469 goto out; 4470 } 4471 ca = (struct sockaddr *)svc_getrpccaller(req->rq_xprt)->buf; 4472 nlen = MAXPATHLEN + 1; 4473 converted_onm = nfscmd_convname(ca, cs->exi, onm, NFSCMD_CONV_INBOUND, 4474 nlen); 4475 4476 if (converted_onm == NULL) { 4477 *cs->statusp = resp->status = NFS4ERR_INVAL; 4478 kmem_free(onm, olen); 4479 goto out; 4480 } 4481 4482 nnm = utf8_to_fn(&args->newname, &nlen, NULL); 4483 if (nnm == NULL) { 4484 *cs->statusp = resp->status = NFS4ERR_INVAL; 4485 if (onm != converted_onm) 4486 kmem_free(converted_onm, MAXPATHLEN + 1); 4487 kmem_free(onm, olen); 4488 goto out; 4489 } 4490 converted_nnm = nfscmd_convname(ca, cs->exi, nnm, NFSCMD_CONV_INBOUND, 4491 MAXPATHLEN + 1); 4492 4493 if (converted_nnm == NULL) { 4494 *cs->statusp = resp->status = NFS4ERR_INVAL; 4495 kmem_free(nnm, nlen); 4496 nnm = NULL; 4497 if (onm != converted_onm) 4498 kmem_free(converted_onm, MAXPATHLEN + 1); 4499 kmem_free(onm, olen); 4500 goto out; 4501 } 4502 4503 4504 if (olen > MAXNAMELEN || nlen > MAXNAMELEN) { 4505 *cs->statusp = resp->status = NFS4ERR_NAMETOOLONG; 4506 kmem_free(onm, olen); 4507 kmem_free(nnm, nlen); 4508 goto out; 4509 } 4510 4511 4512 if (rdonly4(req, cs)) { 4513 *cs->statusp = resp->status = NFS4ERR_ROFS; 4514 if (onm != converted_onm) 4515 kmem_free(converted_onm, MAXPATHLEN + 1); 4516 kmem_free(onm, olen); 4517 if (nnm != converted_nnm) 4518 kmem_free(converted_nnm, MAXPATHLEN + 1); 4519 kmem_free(nnm, nlen); 4520 goto out; 4521 } 4522 4523 /* check label of the target dir */ 4524 if (is_system_labeled()) { 4525 ASSERT(req->rq_label != NULL); 4526 clabel = req->rq_label; 4527 DTRACE_PROBE2(tx__rfs4__log__info__oprename__clabel, char *, 4528 "got client label from request(1)", 4529 struct svc_req *, req); 4530 if (!blequal(&l_admin_low->tsl_label, clabel)) { 4531 if (!do_rfs_label_check(clabel, ndvp, 4532 EQUALITY_CHECK, cs->exi)) { 4533 *cs->statusp = resp->status = NFS4ERR_ACCESS; 4534 goto err_out; 4535 } 4536 } 4537 } 4538 4539 /* 4540 * Is the source a file and have a delegation? 4541 * We don't need to acquire va_seq before these lookups, if 4542 * it causes an update, cinfo.before will not match, which will 4543 * trigger a cache flush even if atomic is TRUE. 4544 */ 4545 sfp = rfs4_lookup_and_findfile(odvp, converted_onm, &srcvp, 4546 &error, cs->cr); 4547 if (sfp != NULL) { 4548 if (rfs4_check_delegated_byfp(FWRITE, sfp, TRUE, TRUE, TRUE, 4549 NULL)) { 4550 *cs->statusp = resp->status = NFS4ERR_DELAY; 4551 goto err_out; 4552 } 4553 } 4554 4555 if (srcvp == NULL) { 4556 *cs->statusp = resp->status = puterrno4(error); 4557 if (onm != converted_onm) 4558 kmem_free(converted_onm, MAXPATHLEN + 1); 4559 kmem_free(onm, olen); 4560 if (nnm != converted_nnm) 4561 kmem_free(converted_nnm, MAXPATHLEN + 1); 4562 kmem_free(nnm, nlen); 4563 goto out; 4564 } 4565 4566 sfp_rele_grant_hold = 1; 4567 4568 /* Does the destination exist and a file and have a delegation? */ 4569 fp = rfs4_lookup_and_findfile(ndvp, converted_nnm, &targvp, NULL, 4570 cs->cr); 4571 if (fp != NULL) { 4572 if (rfs4_check_delegated_byfp(FWRITE, fp, TRUE, TRUE, TRUE, 4573 NULL)) { 4574 *cs->statusp = resp->status = NFS4ERR_DELAY; 4575 goto err_out; 4576 } 4577 } 4578 fp_rele_grant_hold = 1; 4579 4580 /* Check for NBMAND lock on both source and target */ 4581 if (nbl_need_check(srcvp)) { 4582 nbl_start_crit(srcvp, RW_READER); 4583 in_crit_src = 1; 4584 if (nbl_conflict(srcvp, NBL_RENAME, 0, 0, 0, NULL)) { 4585 *cs->statusp = resp->status = NFS4ERR_FILE_OPEN; 4586 goto err_out; 4587 } 4588 } 4589 4590 if (targvp && nbl_need_check(targvp)) { 4591 nbl_start_crit(targvp, RW_READER); 4592 in_crit_targ = 1; 4593 if (nbl_conflict(targvp, NBL_REMOVE, 0, 0, 0, NULL)) { 4594 *cs->statusp = resp->status = NFS4ERR_FILE_OPEN; 4595 goto err_out; 4596 } 4597 } 4598 4599 /* Get source "before" change value */ 4600 obdva.va_mask = AT_CTIME|AT_SEQ; 4601 error = VOP_GETATTR(odvp, &obdva, 0, cs->cr, NULL); 4602 if (!error) { 4603 nbdva.va_mask = AT_CTIME|AT_SEQ; 4604 error = VOP_GETATTR(ndvp, &nbdva, 0, cs->cr, NULL); 4605 } 4606 if (error) { 4607 *cs->statusp = resp->status = puterrno4(error); 4608 goto err_out; 4609 } 4610 4611 NFS4_SET_FATTR4_CHANGE(resp->source_cinfo.before, obdva.va_ctime) 4612 NFS4_SET_FATTR4_CHANGE(resp->target_cinfo.before, nbdva.va_ctime) 4613 4614 error = VOP_RENAME(odvp, converted_onm, ndvp, converted_nnm, cs->cr, 4615 NULL, 0); 4616 4617 /* 4618 * If target existed and was unlinked by VOP_RENAME, state will need 4619 * closed. To avoid deadlock, rfs4_close_all_state will be done after 4620 * any necessary nbl_end_crit on srcvp and tgtvp. 4621 */ 4622 if (error == 0 && fp != NULL) { 4623 rfs4_dbe_lock(fp->rf_dbe); 4624 tvp = fp->rf_vp; 4625 if (tvp) 4626 VN_HOLD(tvp); 4627 rfs4_dbe_unlock(fp->rf_dbe); 4628 4629 if (tvp) { 4630 struct vattr va; 4631 va.va_mask = AT_NLINK; 4632 4633 if (!VOP_GETATTR(tvp, &va, 0, cs->cr, NULL) && 4634 va.va_nlink == 0) { 4635 unlinked = 1; 4636 4637 /* DEBUG data */ 4638 if ((srcvp == targvp) || (tvp != targvp)) { 4639 cmn_err(CE_WARN, "rfs4_op_rename: " 4640 "srcvp %p, targvp: %p, tvp: %p", 4641 (void *)srcvp, (void *)targvp, 4642 (void *)tvp); 4643 } 4644 } else { 4645 VN_RELE(tvp); 4646 } 4647 } 4648 } 4649 if (error == 0) 4650 vn_renamepath(ndvp, srcvp, nnm, nlen - 1); 4651 4652 if (in_crit_src) 4653 nbl_end_crit(srcvp); 4654 if (srcvp) 4655 VN_RELE(srcvp); 4656 if (in_crit_targ) 4657 nbl_end_crit(targvp); 4658 if (targvp) 4659 VN_RELE(targvp); 4660 4661 if (unlinked) { 4662 ASSERT(fp != NULL); 4663 ASSERT(tvp != NULL); 4664 4665 /* DEBUG data */ 4666 if (RW_READ_HELD(&tvp->v_nbllock)) { 4667 cmn_err(CE_WARN, "rfs4_op_rename: " 4668 "RW_READ_HELD(%p)", (void *)tvp); 4669 } 4670 4671 /* The file is gone and so should the state */ 4672 rfs4_close_all_state(fp); 4673 VN_RELE(tvp); 4674 } 4675 4676 if (sfp) { 4677 rfs4_clear_dont_grant(sfp); 4678 rfs4_file_rele(sfp); 4679 } 4680 if (fp) { 4681 rfs4_clear_dont_grant(fp); 4682 rfs4_file_rele(fp); 4683 } 4684 4685 if (converted_onm != onm) 4686 kmem_free(converted_onm, MAXPATHLEN + 1); 4687 kmem_free(onm, olen); 4688 if (converted_nnm != nnm) 4689 kmem_free(converted_nnm, MAXPATHLEN + 1); 4690 kmem_free(nnm, nlen); 4691 4692 /* 4693 * Get the initial "after" sequence number, if it fails, set to zero 4694 */ 4695 oidva.va_mask = AT_SEQ; 4696 if (VOP_GETATTR(odvp, &oidva, 0, cs->cr, NULL)) 4697 oidva.va_seq = 0; 4698 4699 nidva.va_mask = AT_SEQ; 4700 if (VOP_GETATTR(ndvp, &nidva, 0, cs->cr, NULL)) 4701 nidva.va_seq = 0; 4702 4703 /* 4704 * Force modified data and metadata out to stable storage. 4705 */ 4706 (void) VOP_FSYNC(odvp, 0, cs->cr, NULL); 4707 (void) VOP_FSYNC(ndvp, 0, cs->cr, NULL); 4708 4709 if (error) { 4710 *cs->statusp = resp->status = puterrno4(error); 4711 goto out; 4712 } 4713 4714 /* 4715 * Get "after" change values, if it fails, simply return the 4716 * before value. 4717 */ 4718 oadva.va_mask = AT_CTIME|AT_SEQ; 4719 if (VOP_GETATTR(odvp, &oadva, 0, cs->cr, NULL)) { 4720 oadva.va_ctime = obdva.va_ctime; 4721 oadva.va_seq = 0; 4722 } 4723 4724 nadva.va_mask = AT_CTIME|AT_SEQ; 4725 if (VOP_GETATTR(odvp, &nadva, 0, cs->cr, NULL)) { 4726 nadva.va_ctime = nbdva.va_ctime; 4727 nadva.va_seq = 0; 4728 } 4729 4730 NFS4_SET_FATTR4_CHANGE(resp->source_cinfo.after, oadva.va_ctime) 4731 NFS4_SET_FATTR4_CHANGE(resp->target_cinfo.after, nadva.va_ctime) 4732 4733 /* 4734 * The cinfo.atomic = TRUE only if we have 4735 * non-zero va_seq's, and it has incremented by exactly one 4736 * during the VOP_RENAME and it didn't change during the VOP_FSYNC. 4737 */ 4738 if (obdva.va_seq && oidva.va_seq && oadva.va_seq && 4739 oidva.va_seq == (obdva.va_seq + 1) && oidva.va_seq == oadva.va_seq) 4740 resp->source_cinfo.atomic = TRUE; 4741 else 4742 resp->source_cinfo.atomic = FALSE; 4743 4744 if (nbdva.va_seq && nidva.va_seq && nadva.va_seq && 4745 nidva.va_seq == (nbdva.va_seq + 1) && nidva.va_seq == nadva.va_seq) 4746 resp->target_cinfo.atomic = TRUE; 4747 else 4748 resp->target_cinfo.atomic = FALSE; 4749 4750 #ifdef VOLATILE_FH_TEST 4751 { 4752 extern void add_volrnm_fh(struct exportinfo *, vnode_t *); 4753 4754 /* 4755 * Add the renamed file handle to the volatile rename list 4756 */ 4757 if (cs->exi->exi_export.ex_flags & EX_VOLRNM) { 4758 /* file handles may expire on rename */ 4759 vnode_t *vp; 4760 4761 nnm = utf8_to_fn(&args->newname, &nlen, NULL); 4762 /* 4763 * Already know that nnm will be a valid string 4764 */ 4765 error = VOP_LOOKUP(ndvp, nnm, &vp, NULL, 0, NULL, cs->cr, 4766 NULL, NULL, NULL); 4767 kmem_free(nnm, nlen); 4768 if (!error) { 4769 add_volrnm_fh(cs->exi, vp); 4770 VN_RELE(vp); 4771 } 4772 } 4773 } 4774 #endif /* VOLATILE_FH_TEST */ 4775 4776 *cs->statusp = resp->status = NFS4_OK; 4777 out: 4778 DTRACE_NFSV4_2(op__rename__done, struct compound_state *, cs, 4779 RENAME4res *, resp); 4780 return; 4781 4782 err_out: 4783 if (onm != converted_onm) 4784 kmem_free(converted_onm, MAXPATHLEN + 1); 4785 if (onm != NULL) 4786 kmem_free(onm, olen); 4787 if (nnm != converted_nnm) 4788 kmem_free(converted_nnm, MAXPATHLEN + 1); 4789 if (nnm != NULL) 4790 kmem_free(nnm, nlen); 4791 4792 if (in_crit_src) nbl_end_crit(srcvp); 4793 if (in_crit_targ) nbl_end_crit(targvp); 4794 if (targvp) VN_RELE(targvp); 4795 if (srcvp) VN_RELE(srcvp); 4796 if (sfp) { 4797 if (sfp_rele_grant_hold) rfs4_clear_dont_grant(sfp); 4798 rfs4_file_rele(sfp); 4799 } 4800 if (fp) { 4801 if (fp_rele_grant_hold) rfs4_clear_dont_grant(fp); 4802 rfs4_file_rele(fp); 4803 } 4804 4805 DTRACE_NFSV4_2(op__rename__done, struct compound_state *, cs, 4806 RENAME4res *, resp); 4807 } 4808 4809 /* ARGSUSED */ 4810 static void 4811 rfs4_op_renew(nfs_argop4 *argop, nfs_resop4 *resop, struct svc_req *req, 4812 struct compound_state *cs) 4813 { 4814 RENEW4args *args = &argop->nfs_argop4_u.oprenew; 4815 RENEW4res *resp = &resop->nfs_resop4_u.oprenew; 4816 rfs4_client_t *cp; 4817 4818 DTRACE_NFSV4_2(op__renew__start, struct compound_state *, cs, 4819 RENEW4args *, args); 4820 4821 if ((cp = rfs4_findclient_by_id(args->clientid, FALSE)) == NULL) { 4822 *cs->statusp = resp->status = 4823 rfs4_check_clientid(&args->clientid, 0); 4824 goto out; 4825 } 4826 4827 if (rfs4_lease_expired(cp)) { 4828 rfs4_client_rele(cp); 4829 *cs->statusp = resp->status = NFS4ERR_EXPIRED; 4830 goto out; 4831 } 4832 4833 rfs4_update_lease(cp); 4834 4835 mutex_enter(cp->rc_cbinfo.cb_lock); 4836 if (cp->rc_cbinfo.cb_notified_of_cb_path_down == FALSE) { 4837 cp->rc_cbinfo.cb_notified_of_cb_path_down = TRUE; 4838 *cs->statusp = resp->status = NFS4ERR_CB_PATH_DOWN; 4839 } else { 4840 *cs->statusp = resp->status = NFS4_OK; 4841 } 4842 mutex_exit(cp->rc_cbinfo.cb_lock); 4843 4844 rfs4_client_rele(cp); 4845 4846 out: 4847 DTRACE_NFSV4_2(op__renew__done, struct compound_state *, cs, 4848 RENEW4res *, resp); 4849 } 4850 4851 /* ARGSUSED */ 4852 static void 4853 rfs4_op_restorefh(nfs_argop4 *args, nfs_resop4 *resop, struct svc_req *req, 4854 struct compound_state *cs) 4855 { 4856 RESTOREFH4res *resp = &resop->nfs_resop4_u.oprestorefh; 4857 4858 DTRACE_NFSV4_1(op__restorefh__start, struct compound_state *, cs); 4859 4860 /* No need to check cs->access - we are not accessing any object */ 4861 if ((cs->saved_vp == NULL) || (cs->saved_fh.nfs_fh4_val == NULL)) { 4862 *cs->statusp = resp->status = NFS4ERR_RESTOREFH; 4863 goto out; 4864 } 4865 if (cs->vp != NULL) { 4866 VN_RELE(cs->vp); 4867 } 4868 cs->vp = cs->saved_vp; 4869 cs->saved_vp = NULL; 4870 cs->exi = cs->saved_exi; 4871 nfs_fh4_copy(&cs->saved_fh, &cs->fh); 4872 *cs->statusp = resp->status = NFS4_OK; 4873 cs->deleg = FALSE; 4874 4875 out: 4876 DTRACE_NFSV4_2(op__restorefh__done, struct compound_state *, cs, 4877 RESTOREFH4res *, resp); 4878 } 4879 4880 /* ARGSUSED */ 4881 static void 4882 rfs4_op_savefh(nfs_argop4 *argop, nfs_resop4 *resop, struct svc_req *req, 4883 struct compound_state *cs) 4884 { 4885 SAVEFH4res *resp = &resop->nfs_resop4_u.opsavefh; 4886 4887 DTRACE_NFSV4_1(op__savefh__start, struct compound_state *, cs); 4888 4889 /* No need to check cs->access - we are not accessing any object */ 4890 if (cs->vp == NULL) { 4891 *cs->statusp = resp->status = NFS4ERR_NOFILEHANDLE; 4892 goto out; 4893 } 4894 if (cs->saved_vp != NULL) { 4895 VN_RELE(cs->saved_vp); 4896 } 4897 cs->saved_vp = cs->vp; 4898 VN_HOLD(cs->saved_vp); 4899 cs->saved_exi = cs->exi; 4900 /* 4901 * since SAVEFH is fairly rare, don't alloc space for its fh 4902 * unless necessary. 4903 */ 4904 if (cs->saved_fh.nfs_fh4_val == NULL) { 4905 cs->saved_fh.nfs_fh4_val = kmem_alloc(NFS4_FHSIZE, KM_SLEEP); 4906 } 4907 nfs_fh4_copy(&cs->fh, &cs->saved_fh); 4908 *cs->statusp = resp->status = NFS4_OK; 4909 4910 out: 4911 DTRACE_NFSV4_2(op__savefh__done, struct compound_state *, cs, 4912 SAVEFH4res *, resp); 4913 } 4914 4915 /* 4916 * rfs4_verify_attr is called when nfsv4 Setattr failed, but we wish to 4917 * return the bitmap of attrs that were set successfully. It is also 4918 * called by Verify/Nverify to test the vattr/vfsstat attrs. It should 4919 * always be called only after rfs4_do_set_attrs(). 4920 * 4921 * Verify that the attributes are same as the expected ones. sargp->vap 4922 * and sargp->sbp contain the input attributes as translated from fattr4. 4923 * 4924 * This function verifies only the attrs that correspond to a vattr or 4925 * vfsstat struct. That is because of the extra step needed to get the 4926 * corresponding system structs. Other attributes have already been set or 4927 * verified by do_rfs4_set_attrs. 4928 * 4929 * Return 0 if all attrs match, -1 if some don't, error if error processing. 4930 */ 4931 static int 4932 rfs4_verify_attr(struct nfs4_svgetit_arg *sargp, 4933 bitmap4 *resp, struct nfs4_ntov_table *ntovp) 4934 { 4935 int error, ret_error = 0; 4936 int i, k; 4937 uint_t sva_mask = sargp->vap->va_mask; 4938 uint_t vbit; 4939 union nfs4_attr_u *na; 4940 uint8_t *amap; 4941 bool_t getsb = ntovp->vfsstat; 4942 4943 if (sva_mask != 0) { 4944 /* 4945 * Okay to overwrite sargp->vap because we verify based 4946 * on the incoming values. 4947 */ 4948 ret_error = VOP_GETATTR(sargp->cs->vp, sargp->vap, 0, 4949 sargp->cs->cr, NULL); 4950 if (ret_error) { 4951 if (resp == NULL) 4952 return (ret_error); 4953 /* 4954 * Must return bitmap of successful attrs 4955 */ 4956 sva_mask = 0; /* to prevent checking vap later */ 4957 } else { 4958 /* 4959 * Some file systems clobber va_mask. it is probably 4960 * wrong of them to do so, nonethless we practice 4961 * defensive coding. 4962 * See bug id 4276830. 4963 */ 4964 sargp->vap->va_mask = sva_mask; 4965 } 4966 } 4967 4968 if (getsb) { 4969 /* 4970 * Now get the superblock and loop on the bitmap, as there is 4971 * no simple way of translating from superblock to bitmap4. 4972 */ 4973 ret_error = VFS_STATVFS(sargp->cs->vp->v_vfsp, sargp->sbp); 4974 if (ret_error) { 4975 if (resp == NULL) 4976 goto errout; 4977 getsb = FALSE; 4978 } 4979 } 4980 4981 /* 4982 * Now loop and verify each attribute which getattr returned 4983 * whether it's the same as the input. 4984 */ 4985 if (resp == NULL && !getsb && (sva_mask == 0)) 4986 goto errout; 4987 4988 na = ntovp->na; 4989 amap = ntovp->amap; 4990 k = 0; 4991 for (i = 0; i < ntovp->attrcnt; i++, na++, amap++) { 4992 k = *amap; 4993 ASSERT(nfs4_ntov_map[k].nval == k); 4994 vbit = nfs4_ntov_map[k].vbit; 4995 4996 /* 4997 * If vattr attribute but VOP_GETATTR failed, or it's 4998 * superblock attribute but VFS_STATVFS failed, skip 4999 */ 5000 if (vbit) { 5001 if ((vbit & sva_mask) == 0) 5002 continue; 5003 } else if (!(getsb && nfs4_ntov_map[k].vfsstat)) { 5004 continue; 5005 } 5006 error = (*nfs4_ntov_map[k].sv_getit)(NFS4ATTR_VERIT, sargp, na); 5007 if (resp != NULL) { 5008 if (error) 5009 ret_error = -1; /* not all match */ 5010 else /* update response bitmap */ 5011 *resp |= nfs4_ntov_map[k].fbit; 5012 continue; 5013 } 5014 if (error) { 5015 ret_error = -1; /* not all match */ 5016 break; 5017 } 5018 } 5019 errout: 5020 return (ret_error); 5021 } 5022 5023 /* 5024 * Decode the attribute to be set/verified. If the attr requires a sys op 5025 * (VOP_GETATTR, VFS_VFSSTAT), and the request is to verify, then don't 5026 * call the sv_getit function for it, because the sys op hasn't yet been done. 5027 * Return 0 for success, error code if failed. 5028 * 5029 * Note: the decoded arg is not freed here but in nfs4_ntov_table_free. 5030 */ 5031 static int 5032 decode_fattr4_attr(nfs4_attr_cmd_t cmd, struct nfs4_svgetit_arg *sargp, 5033 int k, XDR *xdrp, bitmap4 *resp_bval, union nfs4_attr_u *nap) 5034 { 5035 int error = 0; 5036 bool_t set_later; 5037 5038 sargp->vap->va_mask |= nfs4_ntov_map[k].vbit; 5039 5040 if ((*nfs4_ntov_map[k].xfunc)(xdrp, nap)) { 5041 set_later = nfs4_ntov_map[k].vbit || nfs4_ntov_map[k].vfsstat; 5042 /* 5043 * don't verify yet if a vattr or sb dependent attr, 5044 * because we don't have their sys values yet. 5045 * Will be done later. 5046 */ 5047 if (! (set_later && (cmd == NFS4ATTR_VERIT))) { 5048 /* 5049 * ACLs are a special case, since setting the MODE 5050 * conflicts with setting the ACL. We delay setting 5051 * the ACL until all other attributes have been set. 5052 * The ACL gets set in do_rfs4_op_setattr(). 5053 */ 5054 if (nfs4_ntov_map[k].fbit != FATTR4_ACL_MASK) { 5055 error = (*nfs4_ntov_map[k].sv_getit)(cmd, 5056 sargp, nap); 5057 if (error) { 5058 xdr_free(nfs4_ntov_map[k].xfunc, 5059 (caddr_t)nap); 5060 } 5061 } 5062 } 5063 } else { 5064 #ifdef DEBUG 5065 cmn_err(CE_NOTE, "decode_fattr4_attr: error " 5066 "decoding attribute %d\n", k); 5067 #endif 5068 error = EINVAL; 5069 } 5070 if (!error && resp_bval && !set_later) { 5071 *resp_bval |= nfs4_ntov_map[k].fbit; 5072 } 5073 5074 return (error); 5075 } 5076 5077 /* 5078 * Set vattr based on incoming fattr4 attrs - used by setattr. 5079 * Set response mask. Ignore any values that are not writable vattr attrs. 5080 */ 5081 static nfsstat4 5082 do_rfs4_set_attrs(bitmap4 *resp, fattr4 *fattrp, struct compound_state *cs, 5083 struct nfs4_svgetit_arg *sargp, struct nfs4_ntov_table *ntovp, 5084 nfs4_attr_cmd_t cmd) 5085 { 5086 int error = 0; 5087 int i; 5088 char *attrs = fattrp->attrlist4; 5089 uint32_t attrslen = fattrp->attrlist4_len; 5090 XDR xdr; 5091 nfsstat4 status = NFS4_OK; 5092 vnode_t *vp = cs->vp; 5093 union nfs4_attr_u *na; 5094 uint8_t *amap; 5095 5096 #ifndef lint 5097 /* 5098 * Make sure that maximum attribute number can be expressed as an 5099 * 8 bit quantity. 5100 */ 5101 ASSERT(NFS4_MAXNUM_ATTRS <= (UINT8_MAX + 1)); 5102 #endif 5103 5104 if (vp == NULL) { 5105 if (resp) 5106 *resp = 0; 5107 return (NFS4ERR_NOFILEHANDLE); 5108 } 5109 if (cs->access == CS_ACCESS_DENIED) { 5110 if (resp) 5111 *resp = 0; 5112 return (NFS4ERR_ACCESS); 5113 } 5114 5115 sargp->op = cmd; 5116 sargp->cs = cs; 5117 sargp->flag = 0; /* may be set later */ 5118 sargp->vap->va_mask = 0; 5119 sargp->rdattr_error = NFS4_OK; 5120 sargp->rdattr_error_req = FALSE; 5121 /* sargp->sbp is set by the caller */ 5122 5123 xdrmem_create(&xdr, attrs, attrslen, XDR_DECODE); 5124 5125 na = ntovp->na; 5126 amap = ntovp->amap; 5127 5128 /* 5129 * The following loop iterates on the nfs4_ntov_map checking 5130 * if the fbit is set in the requested bitmap. 5131 * If set then we process the arguments using the 5132 * rfs4_fattr4 conversion functions to populate the setattr 5133 * vattr and va_mask. Any settable attrs that are not using vattr 5134 * will be set in this loop. 5135 */ 5136 for (i = 0; i < nfs4_ntov_map_size; i++) { 5137 if (!(fattrp->attrmask & nfs4_ntov_map[i].fbit)) { 5138 continue; 5139 } 5140 /* 5141 * If setattr, must be a writable attr. 5142 * If verify/nverify, must be a readable attr. 5143 */ 5144 if ((error = (*nfs4_ntov_map[i].sv_getit)( 5145 NFS4ATTR_SUPPORTED, sargp, NULL)) != 0) { 5146 /* 5147 * Client tries to set/verify an 5148 * unsupported attribute, tries to set 5149 * a read only attr or verify a write 5150 * only one - error! 5151 */ 5152 break; 5153 } 5154 /* 5155 * Decode the attribute to set/verify 5156 */ 5157 error = decode_fattr4_attr(cmd, sargp, nfs4_ntov_map[i].nval, 5158 &xdr, resp ? resp : NULL, na); 5159 if (error) 5160 break; 5161 *amap++ = (uint8_t)nfs4_ntov_map[i].nval; 5162 na++; 5163 (ntovp->attrcnt)++; 5164 if (nfs4_ntov_map[i].vfsstat) 5165 ntovp->vfsstat = TRUE; 5166 } 5167 5168 if (error != 0) 5169 status = (error == ENOTSUP ? NFS4ERR_ATTRNOTSUPP : 5170 puterrno4(error)); 5171 /* xdrmem_destroy(&xdrs); */ /* NO-OP */ 5172 return (status); 5173 } 5174 5175 static nfsstat4 5176 do_rfs4_op_setattr(bitmap4 *resp, fattr4 *fattrp, struct compound_state *cs, 5177 stateid4 *stateid) 5178 { 5179 int error = 0; 5180 struct nfs4_svgetit_arg sarg; 5181 bool_t trunc; 5182 5183 nfsstat4 status = NFS4_OK; 5184 cred_t *cr = cs->cr; 5185 vnode_t *vp = cs->vp; 5186 struct nfs4_ntov_table ntov; 5187 struct statvfs64 sb; 5188 struct vattr bva; 5189 struct flock64 bf; 5190 int in_crit = 0; 5191 uint_t saved_mask = 0; 5192 caller_context_t ct; 5193 5194 *resp = 0; 5195 sarg.sbp = &sb; 5196 sarg.is_referral = B_FALSE; 5197 nfs4_ntov_table_init(&ntov); 5198 status = do_rfs4_set_attrs(resp, fattrp, cs, &sarg, &ntov, 5199 NFS4ATTR_SETIT); 5200 if (status != NFS4_OK) { 5201 /* 5202 * failed set attrs 5203 */ 5204 goto done; 5205 } 5206 if ((sarg.vap->va_mask == 0) && 5207 (! (fattrp->attrmask & FATTR4_ACL_MASK))) { 5208 /* 5209 * no further work to be done 5210 */ 5211 goto done; 5212 } 5213 5214 /* 5215 * If we got a request to set the ACL and the MODE, only 5216 * allow changing VSUID, VSGID, and VSVTX. Attempting 5217 * to change any other bits, along with setting an ACL, 5218 * gives NFS4ERR_INVAL. 5219 */ 5220 if ((fattrp->attrmask & FATTR4_ACL_MASK) && 5221 (fattrp->attrmask & FATTR4_MODE_MASK)) { 5222 vattr_t va; 5223 5224 va.va_mask = AT_MODE; 5225 error = VOP_GETATTR(vp, &va, 0, cs->cr, NULL); 5226 if (error) { 5227 status = puterrno4(error); 5228 goto done; 5229 } 5230 if ((sarg.vap->va_mode ^ va.va_mode) & 5231 ~(VSUID | VSGID | VSVTX)) { 5232 status = NFS4ERR_INVAL; 5233 goto done; 5234 } 5235 } 5236 5237 /* Check stateid only if size has been set */ 5238 if (sarg.vap->va_mask & AT_SIZE) { 5239 trunc = (sarg.vap->va_size == 0); 5240 status = rfs4_check_stateid(FWRITE, cs->vp, stateid, 5241 trunc, &cs->deleg, sarg.vap->va_mask & AT_SIZE, &ct); 5242 if (status != NFS4_OK) 5243 goto done; 5244 } else { 5245 ct.cc_sysid = 0; 5246 ct.cc_pid = 0; 5247 ct.cc_caller_id = nfs4_srv_caller_id; 5248 ct.cc_flags = CC_DONTBLOCK; 5249 } 5250 5251 /* XXX start of possible race with delegations */ 5252 5253 /* 5254 * We need to specially handle size changes because it is 5255 * possible for the client to create a file with read-only 5256 * modes, but with the file opened for writing. If the client 5257 * then tries to set the file size, e.g. ftruncate(3C), 5258 * fcntl(F_FREESP), the normal access checking done in 5259 * VOP_SETATTR would prevent the client from doing it even though 5260 * it should be allowed to do so. To get around this, we do the 5261 * access checking for ourselves and use VOP_SPACE which doesn't 5262 * do the access checking. 5263 * Also the client should not be allowed to change the file 5264 * size if there is a conflicting non-blocking mandatory lock in 5265 * the region of the change. 5266 */ 5267 if (vp->v_type == VREG && (sarg.vap->va_mask & AT_SIZE)) { 5268 u_offset_t offset; 5269 ssize_t length; 5270 5271 /* 5272 * ufs_setattr clears AT_SIZE from vap->va_mask, but 5273 * before returning, sarg.vap->va_mask is used to 5274 * generate the setattr reply bitmap. We also clear 5275 * AT_SIZE below before calling VOP_SPACE. For both 5276 * of these cases, the va_mask needs to be saved here 5277 * and restored after calling VOP_SETATTR. 5278 */ 5279 saved_mask = sarg.vap->va_mask; 5280 5281 /* 5282 * Check any possible conflict due to NBMAND locks. 5283 * Get into critical region before VOP_GETATTR, so the 5284 * size attribute is valid when checking conflicts. 5285 */ 5286 if (nbl_need_check(vp)) { 5287 nbl_start_crit(vp, RW_READER); 5288 in_crit = 1; 5289 } 5290 5291 bva.va_mask = AT_UID|AT_SIZE; 5292 error = VOP_GETATTR(vp, &bva, 0, cr, &ct); 5293 if (error != 0) { 5294 status = puterrno4(error); 5295 goto done; 5296 } 5297 5298 if (in_crit) { 5299 if (sarg.vap->va_size < bva.va_size) { 5300 offset = sarg.vap->va_size; 5301 length = bva.va_size - sarg.vap->va_size; 5302 } else { 5303 offset = bva.va_size; 5304 length = sarg.vap->va_size - bva.va_size; 5305 } 5306 if (nbl_conflict(vp, NBL_WRITE, offset, length, 0, 5307 &ct)) { 5308 status = NFS4ERR_LOCKED; 5309 goto done; 5310 } 5311 } 5312 5313 if (crgetuid(cr) == bva.va_uid) { 5314 sarg.vap->va_mask &= ~AT_SIZE; 5315 bf.l_type = F_WRLCK; 5316 bf.l_whence = 0; 5317 bf.l_start = (off64_t)sarg.vap->va_size; 5318 bf.l_len = 0; 5319 bf.l_sysid = 0; 5320 bf.l_pid = 0; 5321 error = VOP_SPACE(vp, F_FREESP, &bf, FWRITE, 5322 (offset_t)sarg.vap->va_size, cr, &ct); 5323 } 5324 } 5325 5326 if (!error && sarg.vap->va_mask != 0) 5327 error = VOP_SETATTR(vp, sarg.vap, sarg.flag, cr, &ct); 5328 5329 /* restore va_mask -- ufs_setattr clears AT_SIZE */ 5330 if (saved_mask & AT_SIZE) 5331 sarg.vap->va_mask |= AT_SIZE; 5332 5333 /* 5334 * If an ACL was being set, it has been delayed until now, 5335 * in order to set the mode (via the VOP_SETATTR() above) first. 5336 */ 5337 if ((! error) && (fattrp->attrmask & FATTR4_ACL_MASK)) { 5338 int i; 5339 5340 for (i = 0; i < NFS4_MAXNUM_ATTRS; i++) 5341 if (ntov.amap[i] == FATTR4_ACL) 5342 break; 5343 if (i < NFS4_MAXNUM_ATTRS) { 5344 error = (*nfs4_ntov_map[FATTR4_ACL].sv_getit)( 5345 NFS4ATTR_SETIT, &sarg, &ntov.na[i]); 5346 if (error == 0) { 5347 *resp |= FATTR4_ACL_MASK; 5348 } else if (error == ENOTSUP) { 5349 (void) rfs4_verify_attr(&sarg, resp, &ntov); 5350 status = NFS4ERR_ATTRNOTSUPP; 5351 goto done; 5352 } 5353 } else { 5354 NFS4_DEBUG(rfs4_debug, 5355 (CE_NOTE, "do_rfs4_op_setattr: " 5356 "unable to find ACL in fattr4")); 5357 error = EINVAL; 5358 } 5359 } 5360 5361 if (error) { 5362 /* check if a monitor detected a delegation conflict */ 5363 if (error == EAGAIN && (ct.cc_flags & CC_WOULDBLOCK)) 5364 status = NFS4ERR_DELAY; 5365 else 5366 status = puterrno4(error); 5367 5368 /* 5369 * Set the response bitmap when setattr failed. 5370 * If VOP_SETATTR partially succeeded, test by doing a 5371 * VOP_GETATTR on the object and comparing the data 5372 * to the setattr arguments. 5373 */ 5374 (void) rfs4_verify_attr(&sarg, resp, &ntov); 5375 } else { 5376 /* 5377 * Force modified metadata out to stable storage. 5378 */ 5379 (void) VOP_FSYNC(vp, FNODSYNC, cr, &ct); 5380 /* 5381 * Set response bitmap 5382 */ 5383 nfs4_vmask_to_nmask_set(sarg.vap->va_mask, resp); 5384 } 5385 5386 /* Return early and already have a NFSv4 error */ 5387 done: 5388 /* 5389 * Except for nfs4_vmask_to_nmask_set(), vattr --> fattr 5390 * conversion sets both readable and writeable NFS4 attrs 5391 * for AT_MTIME and AT_ATIME. The line below masks out 5392 * unrequested attrs from the setattr result bitmap. This 5393 * is placed after the done: label to catch the ATTRNOTSUP 5394 * case. 5395 */ 5396 *resp &= fattrp->attrmask; 5397 5398 if (in_crit) 5399 nbl_end_crit(vp); 5400 5401 nfs4_ntov_table_free(&ntov, &sarg); 5402 5403 return (status); 5404 } 5405 5406 /* ARGSUSED */ 5407 static void 5408 rfs4_op_setattr(nfs_argop4 *argop, nfs_resop4 *resop, struct svc_req *req, 5409 struct compound_state *cs) 5410 { 5411 SETATTR4args *args = &argop->nfs_argop4_u.opsetattr; 5412 SETATTR4res *resp = &resop->nfs_resop4_u.opsetattr; 5413 bslabel_t *clabel; 5414 5415 DTRACE_NFSV4_2(op__setattr__start, struct compound_state *, cs, 5416 SETATTR4args *, args); 5417 5418 if (cs->vp == NULL) { 5419 *cs->statusp = resp->status = NFS4ERR_NOFILEHANDLE; 5420 goto out; 5421 } 5422 5423 /* 5424 * If there is an unshared filesystem mounted on this vnode, 5425 * do not allow to setattr on this vnode. 5426 */ 5427 if (vn_ismntpt(cs->vp)) { 5428 *cs->statusp = resp->status = NFS4ERR_ACCESS; 5429 goto out; 5430 } 5431 5432 resp->attrsset = 0; 5433 5434 if (rdonly4(req, cs)) { 5435 *cs->statusp = resp->status = NFS4ERR_ROFS; 5436 goto out; 5437 } 5438 5439 /* check label before setting attributes */ 5440 if (is_system_labeled()) { 5441 ASSERT(req->rq_label != NULL); 5442 clabel = req->rq_label; 5443 DTRACE_PROBE2(tx__rfs4__log__info__opsetattr__clabel, char *, 5444 "got client label from request(1)", 5445 struct svc_req *, req); 5446 if (!blequal(&l_admin_low->tsl_label, clabel)) { 5447 if (!do_rfs_label_check(clabel, cs->vp, 5448 EQUALITY_CHECK, cs->exi)) { 5449 *cs->statusp = resp->status = NFS4ERR_ACCESS; 5450 goto out; 5451 } 5452 } 5453 } 5454 5455 *cs->statusp = resp->status = 5456 do_rfs4_op_setattr(&resp->attrsset, &args->obj_attributes, cs, 5457 &args->stateid); 5458 5459 out: 5460 DTRACE_NFSV4_2(op__setattr__done, struct compound_state *, cs, 5461 SETATTR4res *, resp); 5462 } 5463 5464 /* ARGSUSED */ 5465 static void 5466 rfs4_op_verify(nfs_argop4 *argop, nfs_resop4 *resop, struct svc_req *req, 5467 struct compound_state *cs) 5468 { 5469 /* 5470 * verify and nverify are exactly the same, except that nverify 5471 * succeeds when some argument changed, and verify succeeds when 5472 * when none changed. 5473 */ 5474 5475 VERIFY4args *args = &argop->nfs_argop4_u.opverify; 5476 VERIFY4res *resp = &resop->nfs_resop4_u.opverify; 5477 5478 int error; 5479 struct nfs4_svgetit_arg sarg; 5480 struct statvfs64 sb; 5481 struct nfs4_ntov_table ntov; 5482 5483 DTRACE_NFSV4_2(op__verify__start, struct compound_state *, cs, 5484 VERIFY4args *, args); 5485 5486 if (cs->vp == NULL) { 5487 *cs->statusp = resp->status = NFS4ERR_NOFILEHANDLE; 5488 goto out; 5489 } 5490 5491 sarg.sbp = &sb; 5492 sarg.is_referral = B_FALSE; 5493 nfs4_ntov_table_init(&ntov); 5494 resp->status = do_rfs4_set_attrs(NULL, &args->obj_attributes, cs, 5495 &sarg, &ntov, NFS4ATTR_VERIT); 5496 if (resp->status != NFS4_OK) { 5497 /* 5498 * do_rfs4_set_attrs will try to verify systemwide attrs, 5499 * so could return -1 for "no match". 5500 */ 5501 if (resp->status == -1) 5502 resp->status = NFS4ERR_NOT_SAME; 5503 goto done; 5504 } 5505 error = rfs4_verify_attr(&sarg, NULL, &ntov); 5506 switch (error) { 5507 case 0: 5508 resp->status = NFS4_OK; 5509 break; 5510 case -1: 5511 resp->status = NFS4ERR_NOT_SAME; 5512 break; 5513 default: 5514 resp->status = puterrno4(error); 5515 break; 5516 } 5517 done: 5518 *cs->statusp = resp->status; 5519 nfs4_ntov_table_free(&ntov, &sarg); 5520 out: 5521 DTRACE_NFSV4_2(op__verify__done, struct compound_state *, cs, 5522 VERIFY4res *, resp); 5523 } 5524 5525 /* ARGSUSED */ 5526 static void 5527 rfs4_op_nverify(nfs_argop4 *argop, nfs_resop4 *resop, struct svc_req *req, 5528 struct compound_state *cs) 5529 { 5530 /* 5531 * verify and nverify are exactly the same, except that nverify 5532 * succeeds when some argument changed, and verify succeeds when 5533 * when none changed. 5534 */ 5535 5536 NVERIFY4args *args = &argop->nfs_argop4_u.opnverify; 5537 NVERIFY4res *resp = &resop->nfs_resop4_u.opnverify; 5538 5539 int error; 5540 struct nfs4_svgetit_arg sarg; 5541 struct statvfs64 sb; 5542 struct nfs4_ntov_table ntov; 5543 5544 DTRACE_NFSV4_2(op__nverify__start, struct compound_state *, cs, 5545 NVERIFY4args *, args); 5546 5547 if (cs->vp == NULL) { 5548 *cs->statusp = resp->status = NFS4ERR_NOFILEHANDLE; 5549 DTRACE_NFSV4_2(op__nverify__done, struct compound_state *, cs, 5550 NVERIFY4res *, resp); 5551 return; 5552 } 5553 sarg.sbp = &sb; 5554 sarg.is_referral = B_FALSE; 5555 nfs4_ntov_table_init(&ntov); 5556 resp->status = do_rfs4_set_attrs(NULL, &args->obj_attributes, cs, 5557 &sarg, &ntov, NFS4ATTR_VERIT); 5558 if (resp->status != NFS4_OK) { 5559 /* 5560 * do_rfs4_set_attrs will try to verify systemwide attrs, 5561 * so could return -1 for "no match". 5562 */ 5563 if (resp->status == -1) 5564 resp->status = NFS4_OK; 5565 goto done; 5566 } 5567 error = rfs4_verify_attr(&sarg, NULL, &ntov); 5568 switch (error) { 5569 case 0: 5570 resp->status = NFS4ERR_SAME; 5571 break; 5572 case -1: 5573 resp->status = NFS4_OK; 5574 break; 5575 default: 5576 resp->status = puterrno4(error); 5577 break; 5578 } 5579 done: 5580 *cs->statusp = resp->status; 5581 nfs4_ntov_table_free(&ntov, &sarg); 5582 5583 DTRACE_NFSV4_2(op__nverify__done, struct compound_state *, cs, 5584 NVERIFY4res *, resp); 5585 } 5586 5587 /* 5588 * XXX - This should live in an NFS header file. 5589 */ 5590 #define MAX_IOVECS 12 5591 5592 /* ARGSUSED */ 5593 static void 5594 rfs4_op_write(nfs_argop4 *argop, nfs_resop4 *resop, struct svc_req *req, 5595 struct compound_state *cs) 5596 { 5597 WRITE4args *args = &argop->nfs_argop4_u.opwrite; 5598 WRITE4res *resp = &resop->nfs_resop4_u.opwrite; 5599 int error; 5600 vnode_t *vp; 5601 struct vattr bva; 5602 u_offset_t rlimit; 5603 struct uio uio; 5604 struct iovec iov[MAX_IOVECS]; 5605 struct iovec *iovp; 5606 int iovcnt; 5607 int ioflag; 5608 cred_t *savecred, *cr; 5609 bool_t *deleg = &cs->deleg; 5610 nfsstat4 stat; 5611 int in_crit = 0; 5612 caller_context_t ct; 5613 nfs4_srv_t *nsrv4; 5614 5615 DTRACE_NFSV4_2(op__write__start, struct compound_state *, cs, 5616 WRITE4args *, args); 5617 5618 vp = cs->vp; 5619 if (vp == NULL) { 5620 *cs->statusp = resp->status = NFS4ERR_NOFILEHANDLE; 5621 goto out; 5622 } 5623 if (cs->access == CS_ACCESS_DENIED) { 5624 *cs->statusp = resp->status = NFS4ERR_ACCESS; 5625 goto out; 5626 } 5627 5628 cr = cs->cr; 5629 5630 if ((stat = rfs4_check_stateid(FWRITE, vp, &args->stateid, FALSE, 5631 deleg, TRUE, &ct)) != NFS4_OK) { 5632 *cs->statusp = resp->status = stat; 5633 goto out; 5634 } 5635 5636 /* 5637 * We have to enter the critical region before calling VOP_RWLOCK 5638 * to avoid a deadlock with ufs. 5639 */ 5640 if (nbl_need_check(vp)) { 5641 nbl_start_crit(vp, RW_READER); 5642 in_crit = 1; 5643 if (nbl_conflict(vp, NBL_WRITE, 5644 args->offset, args->data_len, 0, &ct)) { 5645 *cs->statusp = resp->status = NFS4ERR_LOCKED; 5646 goto out; 5647 } 5648 } 5649 5650 bva.va_mask = AT_MODE | AT_UID; 5651 error = VOP_GETATTR(vp, &bva, 0, cr, &ct); 5652 5653 /* 5654 * If we can't get the attributes, then we can't do the 5655 * right access checking. So, we'll fail the request. 5656 */ 5657 if (error) { 5658 *cs->statusp = resp->status = puterrno4(error); 5659 goto out; 5660 } 5661 5662 if (rdonly4(req, cs)) { 5663 *cs->statusp = resp->status = NFS4ERR_ROFS; 5664 goto out; 5665 } 5666 5667 if (vp->v_type != VREG) { 5668 *cs->statusp = resp->status = 5669 ((vp->v_type == VDIR) ? NFS4ERR_ISDIR : NFS4ERR_INVAL); 5670 goto out; 5671 } 5672 5673 if (crgetuid(cr) != bva.va_uid && 5674 (error = VOP_ACCESS(vp, VWRITE, 0, cr, &ct))) { 5675 *cs->statusp = resp->status = puterrno4(error); 5676 goto out; 5677 } 5678 5679 if (MANDLOCK(vp, bva.va_mode)) { 5680 *cs->statusp = resp->status = NFS4ERR_ACCESS; 5681 goto out; 5682 } 5683 5684 nsrv4 = nfs4_get_srv(); 5685 if (args->data_len == 0) { 5686 *cs->statusp = resp->status = NFS4_OK; 5687 resp->count = 0; 5688 resp->committed = args->stable; 5689 resp->writeverf = nsrv4->write4verf; 5690 goto out; 5691 } 5692 5693 if (args->mblk != NULL) { 5694 mblk_t *m; 5695 uint_t bytes, round_len; 5696 5697 iovcnt = 0; 5698 bytes = 0; 5699 round_len = roundup(args->data_len, BYTES_PER_XDR_UNIT); 5700 for (m = args->mblk; 5701 m != NULL && bytes < round_len; 5702 m = m->b_cont) { 5703 iovcnt++; 5704 bytes += MBLKL(m); 5705 } 5706 #ifdef DEBUG 5707 /* should have ended on an mblk boundary */ 5708 if (bytes != round_len) { 5709 printf("bytes=0x%x, round_len=0x%x, req len=0x%x\n", 5710 bytes, round_len, args->data_len); 5711 printf("args=%p, args->mblk=%p, m=%p", (void *)args, 5712 (void *)args->mblk, (void *)m); 5713 ASSERT(bytes == round_len); 5714 } 5715 #endif 5716 if (iovcnt <= MAX_IOVECS) { 5717 iovp = iov; 5718 } else { 5719 iovp = kmem_alloc(sizeof (*iovp) * iovcnt, KM_SLEEP); 5720 } 5721 mblk_to_iov(args->mblk, iovcnt, iovp); 5722 } else if (args->rlist != NULL) { 5723 iovcnt = 1; 5724 iovp = iov; 5725 iovp->iov_base = (char *)((args->rlist)->u.c_daddr3); 5726 iovp->iov_len = args->data_len; 5727 } else { 5728 iovcnt = 1; 5729 iovp = iov; 5730 iovp->iov_base = args->data_val; 5731 iovp->iov_len = args->data_len; 5732 } 5733 5734 uio.uio_iov = iovp; 5735 uio.uio_iovcnt = iovcnt; 5736 5737 uio.uio_segflg = UIO_SYSSPACE; 5738 uio.uio_extflg = UIO_COPY_DEFAULT; 5739 uio.uio_loffset = args->offset; 5740 uio.uio_resid = args->data_len; 5741 uio.uio_llimit = curproc->p_fsz_ctl; 5742 rlimit = uio.uio_llimit - args->offset; 5743 if (rlimit < (u_offset_t)uio.uio_resid) 5744 uio.uio_resid = (int)rlimit; 5745 5746 if (args->stable == UNSTABLE4) 5747 ioflag = 0; 5748 else if (args->stable == FILE_SYNC4) 5749 ioflag = FSYNC; 5750 else if (args->stable == DATA_SYNC4) 5751 ioflag = FDSYNC; 5752 else { 5753 if (iovp != iov) 5754 kmem_free(iovp, sizeof (*iovp) * iovcnt); 5755 *cs->statusp = resp->status = NFS4ERR_INVAL; 5756 goto out; 5757 } 5758 5759 /* 5760 * We're changing creds because VM may fault and we need 5761 * the cred of the current thread to be used if quota 5762 * checking is enabled. 5763 */ 5764 savecred = curthread->t_cred; 5765 curthread->t_cred = cr; 5766 error = do_io(FWRITE, vp, &uio, ioflag, cr, &ct); 5767 curthread->t_cred = savecred; 5768 5769 if (iovp != iov) 5770 kmem_free(iovp, sizeof (*iovp) * iovcnt); 5771 5772 if (error) { 5773 *cs->statusp = resp->status = puterrno4(error); 5774 goto out; 5775 } 5776 5777 *cs->statusp = resp->status = NFS4_OK; 5778 resp->count = args->data_len - uio.uio_resid; 5779 5780 if (ioflag == 0) 5781 resp->committed = UNSTABLE4; 5782 else 5783 resp->committed = FILE_SYNC4; 5784 5785 resp->writeverf = nsrv4->write4verf; 5786 5787 out: 5788 if (in_crit) 5789 nbl_end_crit(vp); 5790 5791 DTRACE_NFSV4_2(op__write__done, struct compound_state *, cs, 5792 WRITE4res *, resp); 5793 } 5794 5795 void 5796 rfs4_compound(COMPOUND4args *args, COMPOUND4res *resp, struct exportinfo *exi, 5797 struct svc_req *req, cred_t *cr, int *rv) 5798 { 5799 uint_t i; 5800 struct compound_state cs; 5801 nfs4_srv_t *nsrv4; 5802 nfs_export_t *ne = nfs_get_export(); 5803 5804 if (rv != NULL) 5805 *rv = 0; 5806 rfs4_init_compound_state(&cs); 5807 /* 5808 * Form a reply tag by copying over the request tag. 5809 */ 5810 resp->tag.utf8string_len = args->tag.utf8string_len; 5811 if (args->tag.utf8string_len != 0) { 5812 resp->tag.utf8string_val = 5813 kmem_alloc(args->tag.utf8string_len, KM_SLEEP); 5814 bcopy(args->tag.utf8string_val, resp->tag.utf8string_val, 5815 resp->tag.utf8string_len); 5816 } else { 5817 resp->tag.utf8string_val = NULL; 5818 } 5819 5820 cs.statusp = &resp->status; 5821 cs.req = req; 5822 resp->array = NULL; 5823 resp->array_len = 0; 5824 5825 /* 5826 * XXX for now, minorversion should be zero 5827 */ 5828 if (args->minorversion != NFS4_MINORVERSION) { 5829 DTRACE_NFSV4_2(compound__start, struct compound_state *, 5830 &cs, COMPOUND4args *, args); 5831 resp->status = NFS4ERR_MINOR_VERS_MISMATCH; 5832 DTRACE_NFSV4_2(compound__done, struct compound_state *, 5833 &cs, COMPOUND4res *, resp); 5834 return; 5835 } 5836 5837 if (args->array_len == 0) { 5838 resp->status = NFS4_OK; 5839 return; 5840 } 5841 5842 ASSERT(exi == NULL); 5843 ASSERT(cr == NULL); 5844 5845 cr = svc_xprt_cred(req->rq_xprt); 5846 ASSERT(cr != NULL); 5847 5848 if (sec_svc_getcred(req, cr, &cs.principal, &cs.nfsflavor) == 0) { 5849 DTRACE_NFSV4_2(compound__start, struct compound_state *, 5850 &cs, COMPOUND4args *, args); 5851 DTRACE_NFSV4_2(compound__done, struct compound_state *, 5852 &cs, COMPOUND4res *, resp); 5853 svcerr_badcred(req->rq_xprt); 5854 if (rv != NULL) 5855 *rv = 1; 5856 return; 5857 } 5858 resp->array_len = args->array_len; 5859 resp->array = kmem_zalloc(args->array_len * sizeof (nfs_resop4), 5860 KM_SLEEP); 5861 5862 cs.basecr = cr; 5863 nsrv4 = nfs4_get_srv(); 5864 5865 DTRACE_NFSV4_2(compound__start, struct compound_state *, &cs, 5866 COMPOUND4args *, args); 5867 5868 /* 5869 * For now, NFS4 compound processing must be protected by 5870 * exported_lock because it can access more than one exportinfo 5871 * per compound and share/unshare can now change multiple 5872 * exinfo structs. The NFS2/3 code only refs 1 exportinfo 5873 * per proc (excluding public exinfo), and exi_count design 5874 * is sufficient to protect concurrent execution of NFS2/3 5875 * ops along with unexport. This lock will be removed as 5876 * part of the NFSv4 phase 2 namespace redesign work. 5877 */ 5878 rw_enter(&ne->exported_lock, RW_READER); 5879 5880 /* 5881 * If this is the first compound we've seen, we need to start all 5882 * new instances' grace periods. 5883 */ 5884 if (nsrv4->seen_first_compound == 0) { 5885 rfs4_grace_start_new(nsrv4); 5886 /* 5887 * This must be set after rfs4_grace_start_new(), otherwise 5888 * another thread could proceed past here before the former 5889 * is finished. 5890 */ 5891 nsrv4->seen_first_compound = 1; 5892 } 5893 5894 for (i = 0; i < args->array_len && cs.cont; i++) { 5895 nfs_argop4 *argop; 5896 nfs_resop4 *resop; 5897 uint_t op; 5898 kstat_named_t *stat = ne->ne_globals->rfsproccnt[NFS_V4]; 5899 5900 argop = &args->array[i]; 5901 resop = &resp->array[i]; 5902 resop->resop = argop->argop; 5903 op = (uint_t)resop->resop; 5904 5905 if (op < rfsv4disp_cnt) { 5906 /* 5907 * Count the individual ops here; NULL and COMPOUND 5908 * are counted in common_dispatch() 5909 */ 5910 stat[op].value.ui64++; 5911 5912 NFS4_DEBUG(rfs4_debug > 1, 5913 (CE_NOTE, "Executing %s", rfs4_op_string[op])); 5914 (*rfsv4disptab[op].dis_proc)(argop, resop, req, &cs); 5915 NFS4_DEBUG(rfs4_debug > 1, (CE_NOTE, "%s returned %d", 5916 rfs4_op_string[op], *cs.statusp)); 5917 if (*cs.statusp != NFS4_OK) 5918 cs.cont = FALSE; 5919 } else { 5920 /* 5921 * This is effectively dead code since XDR code 5922 * will have already returned BADXDR if op doesn't 5923 * decode to legal value. This only done for a 5924 * day when XDR code doesn't verify v4 opcodes. 5925 */ 5926 op = OP_ILLEGAL; 5927 stat[OP_ILLEGAL_IDX].value.ui64++; 5928 5929 rfs4_op_illegal(argop, resop, req, &cs); 5930 cs.cont = FALSE; 5931 } 5932 5933 /* 5934 * If not at last op, and if we are to stop, then 5935 * compact the results array. 5936 */ 5937 if ((i + 1) < args->array_len && !cs.cont) { 5938 nfs_resop4 *new_res = kmem_alloc( 5939 (i+1) * sizeof (nfs_resop4), KM_SLEEP); 5940 bcopy(resp->array, 5941 new_res, (i+1) * sizeof (nfs_resop4)); 5942 kmem_free(resp->array, 5943 args->array_len * sizeof (nfs_resop4)); 5944 5945 resp->array_len = i + 1; 5946 resp->array = new_res; 5947 } 5948 } 5949 5950 rw_exit(&ne->exported_lock); 5951 5952 /* 5953 * clear exportinfo and vnode fields from compound_state before dtrace 5954 * probe, to avoid tracing residual values for path and share path. 5955 */ 5956 if (cs.vp) 5957 VN_RELE(cs.vp); 5958 if (cs.saved_vp) 5959 VN_RELE(cs.saved_vp); 5960 cs.exi = cs.saved_exi = NULL; 5961 cs.vp = cs.saved_vp = NULL; 5962 5963 DTRACE_NFSV4_2(compound__done, struct compound_state *, &cs, 5964 COMPOUND4res *, resp); 5965 5966 if (cs.saved_fh.nfs_fh4_val) 5967 kmem_free(cs.saved_fh.nfs_fh4_val, NFS4_FHSIZE); 5968 5969 if (cs.cr) 5970 crfree(cs.cr); 5971 /* 5972 * done with this compound request, free the label 5973 */ 5974 5975 if (req->rq_label != NULL) { 5976 kmem_free(req->rq_label, sizeof (bslabel_t)); 5977 req->rq_label = NULL; 5978 } 5979 } 5980 5981 /* 5982 * XXX because of what appears to be duplicate calls to rfs4_compound_free 5983 * XXX zero out the tag and array values. Need to investigate why the 5984 * XXX calls occur, but at least prevent the panic for now. 5985 */ 5986 void 5987 rfs4_compound_free(COMPOUND4res *resp) 5988 { 5989 uint_t i; 5990 5991 if (resp->tag.utf8string_val) { 5992 UTF8STRING_FREE(resp->tag) 5993 } 5994 5995 for (i = 0; i < resp->array_len; i++) { 5996 nfs_resop4 *resop; 5997 uint_t op; 5998 5999 resop = &resp->array[i]; 6000 op = (uint_t)resop->resop; 6001 if (op < rfsv4disp_cnt) { 6002 (*rfsv4disptab[op].dis_resfree)(resop); 6003 } 6004 } 6005 if (resp->array != NULL) { 6006 kmem_free(resp->array, resp->array_len * sizeof (nfs_resop4)); 6007 } 6008 } 6009 6010 /* 6011 * Process the value of the compound request rpc flags, as a bit-AND 6012 * of the individual per-op flags (idempotent, allowork, publicfh_ok) 6013 */ 6014 void 6015 rfs4_compound_flagproc(COMPOUND4args *args, int *flagp) 6016 { 6017 int i; 6018 int flag = RPC_ALL; 6019 6020 for (i = 0; flag && i < args->array_len; i++) { 6021 uint_t op; 6022 6023 op = (uint_t)args->array[i].argop; 6024 6025 if (op < rfsv4disp_cnt) 6026 flag &= rfsv4disptab[op].dis_flags; 6027 else 6028 flag = 0; 6029 } 6030 *flagp = flag; 6031 } 6032 6033 nfsstat4 6034 rfs4_client_sysid(rfs4_client_t *cp, sysid_t *sp) 6035 { 6036 nfsstat4 e; 6037 6038 rfs4_dbe_lock(cp->rc_dbe); 6039 6040 if (cp->rc_sysidt != LM_NOSYSID) { 6041 *sp = cp->rc_sysidt; 6042 e = NFS4_OK; 6043 6044 } else if ((cp->rc_sysidt = lm_alloc_sysidt()) != LM_NOSYSID) { 6045 *sp = cp->rc_sysidt; 6046 e = NFS4_OK; 6047 6048 NFS4_DEBUG(rfs4_debug, (CE_NOTE, 6049 "rfs4_client_sysid: allocated 0x%x\n", *sp)); 6050 } else 6051 e = NFS4ERR_DELAY; 6052 6053 rfs4_dbe_unlock(cp->rc_dbe); 6054 return (e); 6055 } 6056 6057 #if defined(DEBUG) && ! defined(lint) 6058 static void lock_print(char *str, int operation, struct flock64 *flk) 6059 { 6060 char *op, *type; 6061 6062 switch (operation) { 6063 case F_GETLK: op = "F_GETLK"; 6064 break; 6065 case F_SETLK: op = "F_SETLK"; 6066 break; 6067 case F_SETLK_NBMAND: op = "F_SETLK_NBMAND"; 6068 break; 6069 default: op = "F_UNKNOWN"; 6070 break; 6071 } 6072 switch (flk->l_type) { 6073 case F_UNLCK: type = "F_UNLCK"; 6074 break; 6075 case F_RDLCK: type = "F_RDLCK"; 6076 break; 6077 case F_WRLCK: type = "F_WRLCK"; 6078 break; 6079 default: type = "F_UNKNOWN"; 6080 break; 6081 } 6082 6083 ASSERT(flk->l_whence == 0); 6084 cmn_err(CE_NOTE, "%s: %s, type = %s, off = %llx len = %llx pid = %d", 6085 str, op, type, (longlong_t)flk->l_start, 6086 flk->l_len ? (longlong_t)flk->l_len : ~0LL, flk->l_pid); 6087 } 6088 6089 #define LOCK_PRINT(d, s, t, f) if (d) lock_print(s, t, f) 6090 #else 6091 #define LOCK_PRINT(d, s, t, f) 6092 #endif 6093 6094 /*ARGSUSED*/ 6095 static bool_t 6096 creds_ok(cred_set_t cr_set, struct svc_req *req, struct compound_state *cs) 6097 { 6098 return (TRUE); 6099 } 6100 6101 /* 6102 * Look up the pathname using the vp in cs as the directory vnode. 6103 * cs->vp will be the vnode for the file on success 6104 */ 6105 6106 static nfsstat4 6107 rfs4_lookup(component4 *component, struct svc_req *req, 6108 struct compound_state *cs) 6109 { 6110 char *nm; 6111 uint32_t len; 6112 nfsstat4 status; 6113 struct sockaddr *ca; 6114 char *name; 6115 6116 if (cs->vp == NULL) { 6117 return (NFS4ERR_NOFILEHANDLE); 6118 } 6119 if (cs->vp->v_type != VDIR) { 6120 return (NFS4ERR_NOTDIR); 6121 } 6122 6123 status = utf8_dir_verify(component); 6124 if (status != NFS4_OK) 6125 return (status); 6126 6127 nm = utf8_to_fn(component, &len, NULL); 6128 if (nm == NULL) { 6129 return (NFS4ERR_INVAL); 6130 } 6131 6132 if (len > MAXNAMELEN) { 6133 kmem_free(nm, len); 6134 return (NFS4ERR_NAMETOOLONG); 6135 } 6136 6137 ca = (struct sockaddr *)svc_getrpccaller(req->rq_xprt)->buf; 6138 name = nfscmd_convname(ca, cs->exi, nm, NFSCMD_CONV_INBOUND, 6139 MAXPATHLEN + 1); 6140 6141 if (name == NULL) { 6142 kmem_free(nm, len); 6143 return (NFS4ERR_INVAL); 6144 } 6145 6146 status = do_rfs4_op_lookup(name, req, cs); 6147 6148 if (name != nm) 6149 kmem_free(name, MAXPATHLEN + 1); 6150 6151 kmem_free(nm, len); 6152 6153 return (status); 6154 } 6155 6156 static nfsstat4 6157 rfs4_lookupfile(component4 *component, struct svc_req *req, 6158 struct compound_state *cs, uint32_t access, change_info4 *cinfo) 6159 { 6160 nfsstat4 status; 6161 vnode_t *dvp = cs->vp; 6162 vattr_t bva, ava, fva; 6163 int error; 6164 6165 /* Get "before" change value */ 6166 bva.va_mask = AT_CTIME|AT_SEQ; 6167 error = VOP_GETATTR(dvp, &bva, 0, cs->cr, NULL); 6168 if (error) 6169 return (puterrno4(error)); 6170 6171 /* rfs4_lookup may VN_RELE directory */ 6172 VN_HOLD(dvp); 6173 6174 status = rfs4_lookup(component, req, cs); 6175 if (status != NFS4_OK) { 6176 VN_RELE(dvp); 6177 return (status); 6178 } 6179 6180 /* 6181 * Get "after" change value, if it fails, simply return the 6182 * before value. 6183 */ 6184 ava.va_mask = AT_CTIME|AT_SEQ; 6185 if (VOP_GETATTR(dvp, &ava, 0, cs->cr, NULL)) { 6186 ava.va_ctime = bva.va_ctime; 6187 ava.va_seq = 0; 6188 } 6189 VN_RELE(dvp); 6190 6191 /* 6192 * Validate the file is a file 6193 */ 6194 fva.va_mask = AT_TYPE|AT_MODE; 6195 error = VOP_GETATTR(cs->vp, &fva, 0, cs->cr, NULL); 6196 if (error) 6197 return (puterrno4(error)); 6198 6199 if (fva.va_type != VREG) { 6200 if (fva.va_type == VDIR) 6201 return (NFS4ERR_ISDIR); 6202 if (fva.va_type == VLNK) 6203 return (NFS4ERR_SYMLINK); 6204 return (NFS4ERR_INVAL); 6205 } 6206 6207 NFS4_SET_FATTR4_CHANGE(cinfo->before, bva.va_ctime); 6208 NFS4_SET_FATTR4_CHANGE(cinfo->after, ava.va_ctime); 6209 6210 /* 6211 * It is undefined if VOP_LOOKUP will change va_seq, so 6212 * cinfo.atomic = TRUE only if we have 6213 * non-zero va_seq's, and they have not changed. 6214 */ 6215 if (bva.va_seq && ava.va_seq && ava.va_seq == bva.va_seq) 6216 cinfo->atomic = TRUE; 6217 else 6218 cinfo->atomic = FALSE; 6219 6220 /* Check for mandatory locking */ 6221 cs->mandlock = MANDLOCK(cs->vp, fva.va_mode); 6222 return (check_open_access(access, cs, req)); 6223 } 6224 6225 static nfsstat4 6226 create_vnode(vnode_t *dvp, char *nm, vattr_t *vap, createmode4 mode, 6227 cred_t *cr, vnode_t **vpp, bool_t *created) 6228 { 6229 int error; 6230 nfsstat4 status = NFS4_OK; 6231 vattr_t va; 6232 6233 tryagain: 6234 6235 /* 6236 * The file open mode used is VWRITE. If the client needs 6237 * some other semantic, then it should do the access checking 6238 * itself. It would have been nice to have the file open mode 6239 * passed as part of the arguments. 6240 */ 6241 6242 *created = TRUE; 6243 error = VOP_CREATE(dvp, nm, vap, EXCL, VWRITE, vpp, cr, 0, NULL, NULL); 6244 6245 if (error) { 6246 *created = FALSE; 6247 6248 /* 6249 * If we got something other than file already exists 6250 * then just return this error. Otherwise, we got 6251 * EEXIST. If we were doing a GUARDED create, then 6252 * just return this error. Otherwise, we need to 6253 * make sure that this wasn't a duplicate of an 6254 * exclusive create request. 6255 * 6256 * The assumption is made that a non-exclusive create 6257 * request will never return EEXIST. 6258 */ 6259 6260 if (error != EEXIST || mode == GUARDED4) { 6261 status = puterrno4(error); 6262 return (status); 6263 } 6264 error = VOP_LOOKUP(dvp, nm, vpp, NULL, 0, NULL, cr, 6265 NULL, NULL, NULL); 6266 6267 if (error) { 6268 /* 6269 * We couldn't find the file that we thought that 6270 * we just created. So, we'll just try creating 6271 * it again. 6272 */ 6273 if (error == ENOENT) 6274 goto tryagain; 6275 6276 status = puterrno4(error); 6277 return (status); 6278 } 6279 6280 if (mode == UNCHECKED4) { 6281 /* existing object must be regular file */ 6282 if ((*vpp)->v_type != VREG) { 6283 if ((*vpp)->v_type == VDIR) 6284 status = NFS4ERR_ISDIR; 6285 else if ((*vpp)->v_type == VLNK) 6286 status = NFS4ERR_SYMLINK; 6287 else 6288 status = NFS4ERR_INVAL; 6289 VN_RELE(*vpp); 6290 return (status); 6291 } 6292 6293 return (NFS4_OK); 6294 } 6295 6296 /* Check for duplicate request */ 6297 va.va_mask = AT_MTIME; 6298 error = VOP_GETATTR(*vpp, &va, 0, cr, NULL); 6299 if (!error) { 6300 /* We found the file */ 6301 const timestruc_t *mtime = &vap->va_mtime; 6302 6303 if (va.va_mtime.tv_sec != mtime->tv_sec || 6304 va.va_mtime.tv_nsec != mtime->tv_nsec) { 6305 /* but its not our creation */ 6306 VN_RELE(*vpp); 6307 return (NFS4ERR_EXIST); 6308 } 6309 *created = TRUE; /* retrans of create == created */ 6310 return (NFS4_OK); 6311 } 6312 VN_RELE(*vpp); 6313 return (NFS4ERR_EXIST); 6314 } 6315 6316 return (NFS4_OK); 6317 } 6318 6319 static nfsstat4 6320 check_open_access(uint32_t access, struct compound_state *cs, 6321 struct svc_req *req) 6322 { 6323 int error; 6324 vnode_t *vp; 6325 bool_t readonly; 6326 cred_t *cr = cs->cr; 6327 6328 /* For now we don't allow mandatory locking as per V2/V3 */ 6329 if (cs->access == CS_ACCESS_DENIED || cs->mandlock) { 6330 return (NFS4ERR_ACCESS); 6331 } 6332 6333 vp = cs->vp; 6334 ASSERT(cr != NULL && vp->v_type == VREG); 6335 6336 /* 6337 * If the file system is exported read only and we are trying 6338 * to open for write, then return NFS4ERR_ROFS 6339 */ 6340 6341 readonly = rdonly4(req, cs); 6342 6343 if ((access & OPEN4_SHARE_ACCESS_WRITE) && readonly) 6344 return (NFS4ERR_ROFS); 6345 6346 if (access & OPEN4_SHARE_ACCESS_READ) { 6347 if ((VOP_ACCESS(vp, VREAD, 0, cr, NULL) != 0) && 6348 (VOP_ACCESS(vp, VEXEC, 0, cr, NULL) != 0)) { 6349 return (NFS4ERR_ACCESS); 6350 } 6351 } 6352 6353 if (access & OPEN4_SHARE_ACCESS_WRITE) { 6354 error = VOP_ACCESS(vp, VWRITE, 0, cr, NULL); 6355 if (error) 6356 return (NFS4ERR_ACCESS); 6357 } 6358 6359 return (NFS4_OK); 6360 } 6361 6362 static nfsstat4 6363 rfs4_createfile(OPEN4args *args, struct svc_req *req, struct compound_state *cs, 6364 change_info4 *cinfo, bitmap4 *attrset, clientid4 clientid) 6365 { 6366 struct nfs4_svgetit_arg sarg; 6367 struct nfs4_ntov_table ntov; 6368 6369 bool_t ntov_table_init = FALSE; 6370 struct statvfs64 sb; 6371 nfsstat4 status; 6372 vnode_t *vp; 6373 vattr_t bva, ava, iva, cva, *vap; 6374 vnode_t *dvp; 6375 timespec32_t *mtime; 6376 char *nm = NULL; 6377 uint_t buflen; 6378 bool_t created; 6379 bool_t setsize = FALSE; 6380 len_t reqsize; 6381 int error; 6382 bool_t trunc; 6383 caller_context_t ct; 6384 component4 *component; 6385 bslabel_t *clabel; 6386 struct sockaddr *ca; 6387 char *name = NULL; 6388 6389 sarg.sbp = &sb; 6390 sarg.is_referral = B_FALSE; 6391 6392 dvp = cs->vp; 6393 6394 /* Check if the file system is read only */ 6395 if (rdonly4(req, cs)) 6396 return (NFS4ERR_ROFS); 6397 6398 /* check the label of including directory */ 6399 if (is_system_labeled()) { 6400 ASSERT(req->rq_label != NULL); 6401 clabel = req->rq_label; 6402 DTRACE_PROBE2(tx__rfs4__log__info__opremove__clabel, char *, 6403 "got client label from request(1)", 6404 struct svc_req *, req); 6405 if (!blequal(&l_admin_low->tsl_label, clabel)) { 6406 if (!do_rfs_label_check(clabel, dvp, EQUALITY_CHECK, 6407 cs->exi)) { 6408 return (NFS4ERR_ACCESS); 6409 } 6410 } 6411 } 6412 6413 /* 6414 * Get the last component of path name in nm. cs will reference 6415 * the including directory on success. 6416 */ 6417 component = &args->open_claim4_u.file; 6418 status = utf8_dir_verify(component); 6419 if (status != NFS4_OK) 6420 return (status); 6421 6422 nm = utf8_to_fn(component, &buflen, NULL); 6423 6424 if (nm == NULL) 6425 return (NFS4ERR_RESOURCE); 6426 6427 if (buflen > MAXNAMELEN) { 6428 kmem_free(nm, buflen); 6429 return (NFS4ERR_NAMETOOLONG); 6430 } 6431 6432 bva.va_mask = AT_TYPE|AT_CTIME|AT_SEQ; 6433 error = VOP_GETATTR(dvp, &bva, 0, cs->cr, NULL); 6434 if (error) { 6435 kmem_free(nm, buflen); 6436 return (puterrno4(error)); 6437 } 6438 6439 if (bva.va_type != VDIR) { 6440 kmem_free(nm, buflen); 6441 return (NFS4ERR_NOTDIR); 6442 } 6443 6444 NFS4_SET_FATTR4_CHANGE(cinfo->before, bva.va_ctime) 6445 6446 switch (args->mode) { 6447 case GUARDED4: 6448 /*FALLTHROUGH*/ 6449 case UNCHECKED4: 6450 nfs4_ntov_table_init(&ntov); 6451 ntov_table_init = TRUE; 6452 6453 *attrset = 0; 6454 status = do_rfs4_set_attrs(attrset, 6455 &args->createhow4_u.createattrs, 6456 cs, &sarg, &ntov, NFS4ATTR_SETIT); 6457 6458 if (status == NFS4_OK && (sarg.vap->va_mask & AT_TYPE) && 6459 sarg.vap->va_type != VREG) { 6460 if (sarg.vap->va_type == VDIR) 6461 status = NFS4ERR_ISDIR; 6462 else if (sarg.vap->va_type == VLNK) 6463 status = NFS4ERR_SYMLINK; 6464 else 6465 status = NFS4ERR_INVAL; 6466 } 6467 6468 if (status != NFS4_OK) { 6469 kmem_free(nm, buflen); 6470 nfs4_ntov_table_free(&ntov, &sarg); 6471 *attrset = 0; 6472 return (status); 6473 } 6474 6475 vap = sarg.vap; 6476 vap->va_type = VREG; 6477 vap->va_mask |= AT_TYPE; 6478 6479 if ((vap->va_mask & AT_MODE) == 0) { 6480 vap->va_mask |= AT_MODE; 6481 vap->va_mode = (mode_t)0600; 6482 } 6483 6484 if (vap->va_mask & AT_SIZE) { 6485 6486 /* Disallow create with a non-zero size */ 6487 6488 if ((reqsize = sarg.vap->va_size) != 0) { 6489 kmem_free(nm, buflen); 6490 nfs4_ntov_table_free(&ntov, &sarg); 6491 *attrset = 0; 6492 return (NFS4ERR_INVAL); 6493 } 6494 setsize = TRUE; 6495 } 6496 break; 6497 6498 case EXCLUSIVE4: 6499 /* prohibit EXCL create of named attributes */ 6500 if (dvp->v_flag & V_XATTRDIR) { 6501 kmem_free(nm, buflen); 6502 *attrset = 0; 6503 return (NFS4ERR_INVAL); 6504 } 6505 6506 cva.va_mask = AT_TYPE | AT_MTIME | AT_MODE; 6507 cva.va_type = VREG; 6508 /* 6509 * Ensure no time overflows. Assumes underlying 6510 * filesystem supports at least 32 bits. 6511 * Truncate nsec to usec resolution to allow valid 6512 * compares even if the underlying filesystem truncates. 6513 */ 6514 mtime = (timespec32_t *)&args->createhow4_u.createverf; 6515 cva.va_mtime.tv_sec = mtime->tv_sec % TIME32_MAX; 6516 cva.va_mtime.tv_nsec = (mtime->tv_nsec / 1000) * 1000; 6517 cva.va_mode = (mode_t)0; 6518 vap = &cva; 6519 6520 /* 6521 * For EXCL create, attrset is set to the server attr 6522 * used to cache the client's verifier. 6523 */ 6524 *attrset = FATTR4_TIME_MODIFY_MASK; 6525 break; 6526 } 6527 6528 ca = (struct sockaddr *)svc_getrpccaller(req->rq_xprt)->buf; 6529 name = nfscmd_convname(ca, cs->exi, nm, NFSCMD_CONV_INBOUND, 6530 MAXPATHLEN + 1); 6531 6532 if (name == NULL) { 6533 kmem_free(nm, buflen); 6534 return (NFS4ERR_SERVERFAULT); 6535 } 6536 6537 status = create_vnode(dvp, name, vap, args->mode, 6538 cs->cr, &vp, &created); 6539 if (nm != name) 6540 kmem_free(name, MAXPATHLEN + 1); 6541 kmem_free(nm, buflen); 6542 6543 if (status != NFS4_OK) { 6544 if (ntov_table_init) 6545 nfs4_ntov_table_free(&ntov, &sarg); 6546 *attrset = 0; 6547 return (status); 6548 } 6549 6550 trunc = (setsize && !created); 6551 6552 if (args->mode != EXCLUSIVE4) { 6553 bitmap4 createmask = args->createhow4_u.createattrs.attrmask; 6554 6555 /* 6556 * True verification that object was created with correct 6557 * attrs is impossible. The attrs could have been changed 6558 * immediately after object creation. If attributes did 6559 * not verify, the only recourse for the server is to 6560 * destroy the object. Maybe if some attrs (like gid) 6561 * are set incorrectly, the object should be destroyed; 6562 * however, seems bad as a default policy. Do we really 6563 * want to destroy an object over one of the times not 6564 * verifying correctly? For these reasons, the server 6565 * currently sets bits in attrset for createattrs 6566 * that were set; however, no verification is done. 6567 * 6568 * vmask_to_nmask accounts for vattr bits set on create 6569 * [do_rfs4_set_attrs() only sets resp bits for 6570 * non-vattr/vfs bits.] 6571 * Mask off any bits we set by default so as not to return 6572 * more attrset bits than were requested in createattrs 6573 */ 6574 if (created) { 6575 nfs4_vmask_to_nmask(sarg.vap->va_mask, attrset); 6576 *attrset &= createmask; 6577 } else { 6578 /* 6579 * We did not create the vnode (we tried but it 6580 * already existed). In this case, the only createattr 6581 * that the spec allows the server to set is size, 6582 * and even then, it can only be set if it is 0. 6583 */ 6584 *attrset = 0; 6585 if (trunc) 6586 *attrset = FATTR4_SIZE_MASK; 6587 } 6588 } 6589 if (ntov_table_init) 6590 nfs4_ntov_table_free(&ntov, &sarg); 6591 6592 /* 6593 * Get the initial "after" sequence number, if it fails, 6594 * set to zero, time to before. 6595 */ 6596 iva.va_mask = AT_CTIME|AT_SEQ; 6597 if (VOP_GETATTR(dvp, &iva, 0, cs->cr, NULL)) { 6598 iva.va_seq = 0; 6599 iva.va_ctime = bva.va_ctime; 6600 } 6601 6602 /* 6603 * create_vnode attempts to create the file exclusive, 6604 * if it already exists the VOP_CREATE will fail and 6605 * may not increase va_seq. It is atomic if 6606 * we haven't changed the directory, but if it has changed 6607 * we don't know what changed it. 6608 */ 6609 if (!created) { 6610 if (bva.va_seq && iva.va_seq && 6611 bva.va_seq == iva.va_seq) 6612 cinfo->atomic = TRUE; 6613 else 6614 cinfo->atomic = FALSE; 6615 NFS4_SET_FATTR4_CHANGE(cinfo->after, iva.va_ctime); 6616 } else { 6617 /* 6618 * The entry was created, we need to sync the 6619 * directory metadata. 6620 */ 6621 (void) VOP_FSYNC(dvp, 0, cs->cr, NULL); 6622 6623 /* 6624 * Get "after" change value, if it fails, simply return the 6625 * before value. 6626 */ 6627 ava.va_mask = AT_CTIME|AT_SEQ; 6628 if (VOP_GETATTR(dvp, &ava, 0, cs->cr, NULL)) { 6629 ava.va_ctime = bva.va_ctime; 6630 ava.va_seq = 0; 6631 } 6632 6633 NFS4_SET_FATTR4_CHANGE(cinfo->after, ava.va_ctime); 6634 6635 /* 6636 * The cinfo->atomic = TRUE only if we have 6637 * non-zero va_seq's, and it has incremented by exactly one 6638 * during the create_vnode and it didn't 6639 * change during the VOP_FSYNC. 6640 */ 6641 if (bva.va_seq && iva.va_seq && ava.va_seq && 6642 iva.va_seq == (bva.va_seq + 1) && iva.va_seq == ava.va_seq) 6643 cinfo->atomic = TRUE; 6644 else 6645 cinfo->atomic = FALSE; 6646 } 6647 6648 /* Check for mandatory locking and that the size gets set. */ 6649 cva.va_mask = AT_MODE; 6650 if (setsize) 6651 cva.va_mask |= AT_SIZE; 6652 6653 /* Assume the worst */ 6654 cs->mandlock = TRUE; 6655 6656 if (VOP_GETATTR(vp, &cva, 0, cs->cr, NULL) == 0) { 6657 cs->mandlock = MANDLOCK(cs->vp, cva.va_mode); 6658 6659 /* 6660 * Truncate the file if necessary; this would be 6661 * the case for create over an existing file. 6662 */ 6663 6664 if (trunc) { 6665 int in_crit = 0; 6666 rfs4_file_t *fp; 6667 nfs4_srv_t *nsrv4; 6668 bool_t create = FALSE; 6669 6670 /* 6671 * We are writing over an existing file. 6672 * Check to see if we need to recall a delegation. 6673 */ 6674 nsrv4 = nfs4_get_srv(); 6675 rfs4_hold_deleg_policy(nsrv4); 6676 if ((fp = rfs4_findfile(vp, NULL, &create)) != NULL) { 6677 if (rfs4_check_delegated_byfp(FWRITE, fp, 6678 (reqsize == 0), FALSE, FALSE, &clientid)) { 6679 rfs4_file_rele(fp); 6680 rfs4_rele_deleg_policy(nsrv4); 6681 VN_RELE(vp); 6682 *attrset = 0; 6683 return (NFS4ERR_DELAY); 6684 } 6685 rfs4_file_rele(fp); 6686 } 6687 rfs4_rele_deleg_policy(nsrv4); 6688 6689 if (nbl_need_check(vp)) { 6690 in_crit = 1; 6691 6692 ASSERT(reqsize == 0); 6693 6694 nbl_start_crit(vp, RW_READER); 6695 if (nbl_conflict(vp, NBL_WRITE, 0, 6696 cva.va_size, 0, NULL)) { 6697 in_crit = 0; 6698 nbl_end_crit(vp); 6699 VN_RELE(vp); 6700 *attrset = 0; 6701 return (NFS4ERR_ACCESS); 6702 } 6703 } 6704 ct.cc_sysid = 0; 6705 ct.cc_pid = 0; 6706 ct.cc_caller_id = nfs4_srv_caller_id; 6707 ct.cc_flags = CC_DONTBLOCK; 6708 6709 cva.va_mask = AT_SIZE; 6710 cva.va_size = reqsize; 6711 (void) VOP_SETATTR(vp, &cva, 0, cs->cr, &ct); 6712 if (in_crit) 6713 nbl_end_crit(vp); 6714 } 6715 } 6716 6717 error = makefh4(&cs->fh, vp, cs->exi); 6718 6719 /* 6720 * Force modified data and metadata out to stable storage. 6721 */ 6722 (void) VOP_FSYNC(vp, FNODSYNC, cs->cr, NULL); 6723 6724 if (error) { 6725 VN_RELE(vp); 6726 *attrset = 0; 6727 return (puterrno4(error)); 6728 } 6729 6730 /* if parent dir is attrdir, set namedattr fh flag */ 6731 if (dvp->v_flag & V_XATTRDIR) 6732 set_fh4_flag(&cs->fh, FH4_NAMEDATTR); 6733 6734 if (cs->vp) 6735 VN_RELE(cs->vp); 6736 6737 cs->vp = vp; 6738 6739 /* 6740 * if we did not create the file, we will need to check 6741 * the access bits on the file 6742 */ 6743 6744 if (!created) { 6745 if (setsize) 6746 args->share_access |= OPEN4_SHARE_ACCESS_WRITE; 6747 status = check_open_access(args->share_access, cs, req); 6748 if (status != NFS4_OK) 6749 *attrset = 0; 6750 } 6751 return (status); 6752 } 6753 6754 /*ARGSUSED*/ 6755 static void 6756 rfs4_do_open(struct compound_state *cs, struct svc_req *req, 6757 rfs4_openowner_t *oo, delegreq_t deleg, 6758 uint32_t access, uint32_t deny, 6759 OPEN4res *resp, int deleg_cur) 6760 { 6761 /* XXX Currently not using req */ 6762 rfs4_state_t *sp; 6763 rfs4_file_t *fp; 6764 bool_t screate = TRUE; 6765 bool_t fcreate = TRUE; 6766 uint32_t open_a, share_a; 6767 uint32_t open_d, share_d; 6768 rfs4_deleg_state_t *dsp; 6769 sysid_t sysid; 6770 nfsstat4 status; 6771 caller_context_t ct; 6772 int fflags = 0; 6773 int recall = 0; 6774 int err; 6775 int first_open; 6776 6777 /* get the file struct and hold a lock on it during initial open */ 6778 fp = rfs4_findfile_withlock(cs->vp, &cs->fh, &fcreate); 6779 if (fp == NULL) { 6780 resp->status = NFS4ERR_RESOURCE; 6781 DTRACE_PROBE1(nfss__e__do__open1, nfsstat4, resp->status); 6782 return; 6783 } 6784 6785 sp = rfs4_findstate_by_owner_file(oo, fp, &screate); 6786 if (sp == NULL) { 6787 resp->status = NFS4ERR_RESOURCE; 6788 DTRACE_PROBE1(nfss__e__do__open2, nfsstat4, resp->status); 6789 /* No need to keep any reference */ 6790 rw_exit(&fp->rf_file_rwlock); 6791 rfs4_file_rele(fp); 6792 return; 6793 } 6794 6795 /* try to get the sysid before continuing */ 6796 if ((status = rfs4_client_sysid(oo->ro_client, &sysid)) != NFS4_OK) { 6797 resp->status = status; 6798 rfs4_file_rele(fp); 6799 /* Not a fully formed open; "close" it */ 6800 if (screate == TRUE) 6801 rfs4_state_close(sp, FALSE, FALSE, cs->cr); 6802 rfs4_state_rele(sp); 6803 return; 6804 } 6805 6806 /* Calculate the fflags for this OPEN. */ 6807 if (access & OPEN4_SHARE_ACCESS_READ) 6808 fflags |= FREAD; 6809 if (access & OPEN4_SHARE_ACCESS_WRITE) 6810 fflags |= FWRITE; 6811 6812 rfs4_dbe_lock(sp->rs_dbe); 6813 6814 /* 6815 * Calculate the new deny and access mode that this open is adding to 6816 * the file for this open owner; 6817 */ 6818 open_d = (deny & ~sp->rs_open_deny); 6819 open_a = (access & ~sp->rs_open_access); 6820 6821 /* 6822 * Calculate the new share access and share deny modes that this open 6823 * is adding to the file for this open owner; 6824 */ 6825 share_a = (access & ~sp->rs_share_access); 6826 share_d = (deny & ~sp->rs_share_deny); 6827 6828 first_open = (sp->rs_open_access & OPEN4_SHARE_ACCESS_BOTH) == 0; 6829 6830 /* 6831 * Check to see the client has already sent an open for this 6832 * open owner on this file with the same share/deny modes. 6833 * If so, we don't need to check for a conflict and we don't 6834 * need to add another shrlock. If not, then we need to 6835 * check for conflicts in deny and access before checking for 6836 * conflicts in delegation. We don't want to recall a 6837 * delegation based on an open that will eventually fail based 6838 * on shares modes. 6839 */ 6840 6841 if (share_a || share_d) { 6842 if ((err = rfs4_share(sp, access, deny)) != 0) { 6843 rfs4_dbe_unlock(sp->rs_dbe); 6844 resp->status = err; 6845 6846 rfs4_file_rele(fp); 6847 /* Not a fully formed open; "close" it */ 6848 if (screate == TRUE) 6849 rfs4_state_close(sp, FALSE, FALSE, cs->cr); 6850 rfs4_state_rele(sp); 6851 return; 6852 } 6853 } 6854 6855 rfs4_dbe_lock(fp->rf_dbe); 6856 6857 /* 6858 * Check to see if this file is delegated and if so, if a 6859 * recall needs to be done. 6860 */ 6861 if (rfs4_check_recall(sp, access)) { 6862 rfs4_dbe_unlock(fp->rf_dbe); 6863 rfs4_dbe_unlock(sp->rs_dbe); 6864 rfs4_recall_deleg(fp, FALSE, sp->rs_owner->ro_client); 6865 delay(NFS4_DELEGATION_CONFLICT_DELAY); 6866 rfs4_dbe_lock(sp->rs_dbe); 6867 6868 /* if state closed while lock was dropped */ 6869 if (sp->rs_closed) { 6870 if (share_a || share_d) 6871 (void) rfs4_unshare(sp); 6872 rfs4_dbe_unlock(sp->rs_dbe); 6873 rfs4_file_rele(fp); 6874 /* Not a fully formed open; "close" it */ 6875 if (screate == TRUE) 6876 rfs4_state_close(sp, FALSE, FALSE, cs->cr); 6877 rfs4_state_rele(sp); 6878 resp->status = NFS4ERR_OLD_STATEID; 6879 return; 6880 } 6881 6882 rfs4_dbe_lock(fp->rf_dbe); 6883 /* Let's see if the delegation was returned */ 6884 if (rfs4_check_recall(sp, access)) { 6885 rfs4_dbe_unlock(fp->rf_dbe); 6886 if (share_a || share_d) 6887 (void) rfs4_unshare(sp); 6888 rfs4_dbe_unlock(sp->rs_dbe); 6889 rfs4_file_rele(fp); 6890 rfs4_update_lease(sp->rs_owner->ro_client); 6891 6892 /* Not a fully formed open; "close" it */ 6893 if (screate == TRUE) 6894 rfs4_state_close(sp, FALSE, FALSE, cs->cr); 6895 rfs4_state_rele(sp); 6896 resp->status = NFS4ERR_DELAY; 6897 return; 6898 } 6899 } 6900 /* 6901 * the share check passed and any delegation conflict has been 6902 * taken care of, now call vop_open. 6903 * if this is the first open then call vop_open with fflags. 6904 * if not, call vn_open_upgrade with just the upgrade flags. 6905 * 6906 * if the file has been opened already, it will have the current 6907 * access mode in the state struct. if it has no share access, then 6908 * this is a new open. 6909 * 6910 * However, if this is open with CLAIM_DLEGATE_CUR, then don't 6911 * call VOP_OPEN(), just do the open upgrade. 6912 */ 6913 if (first_open && !deleg_cur) { 6914 ct.cc_sysid = sysid; 6915 ct.cc_pid = rfs4_dbe_getid(sp->rs_owner->ro_dbe); 6916 ct.cc_caller_id = nfs4_srv_caller_id; 6917 ct.cc_flags = CC_DONTBLOCK; 6918 err = VOP_OPEN(&cs->vp, fflags, cs->cr, &ct); 6919 if (err) { 6920 rfs4_dbe_unlock(fp->rf_dbe); 6921 if (share_a || share_d) 6922 (void) rfs4_unshare(sp); 6923 rfs4_dbe_unlock(sp->rs_dbe); 6924 rfs4_file_rele(fp); 6925 6926 /* Not a fully formed open; "close" it */ 6927 if (screate == TRUE) 6928 rfs4_state_close(sp, FALSE, FALSE, cs->cr); 6929 rfs4_state_rele(sp); 6930 /* check if a monitor detected a delegation conflict */ 6931 if (err == EAGAIN && (ct.cc_flags & CC_WOULDBLOCK)) 6932 resp->status = NFS4ERR_DELAY; 6933 else 6934 resp->status = NFS4ERR_SERVERFAULT; 6935 return; 6936 } 6937 } else { /* open upgrade */ 6938 /* 6939 * calculate the fflags for the new mode that is being added 6940 * by this upgrade. 6941 */ 6942 fflags = 0; 6943 if (open_a & OPEN4_SHARE_ACCESS_READ) 6944 fflags |= FREAD; 6945 if (open_a & OPEN4_SHARE_ACCESS_WRITE) 6946 fflags |= FWRITE; 6947 vn_open_upgrade(cs->vp, fflags); 6948 } 6949 sp->rs_open_access |= access; 6950 sp->rs_open_deny |= deny; 6951 6952 if (open_d & OPEN4_SHARE_DENY_READ) 6953 fp->rf_deny_read++; 6954 if (open_d & OPEN4_SHARE_DENY_WRITE) 6955 fp->rf_deny_write++; 6956 fp->rf_share_deny |= deny; 6957 6958 if (open_a & OPEN4_SHARE_ACCESS_READ) 6959 fp->rf_access_read++; 6960 if (open_a & OPEN4_SHARE_ACCESS_WRITE) 6961 fp->rf_access_write++; 6962 fp->rf_share_access |= access; 6963 6964 /* 6965 * Check for delegation here. if the deleg argument is not 6966 * DELEG_ANY, then this is a reclaim from a client and 6967 * we must honor the delegation requested. If necessary we can 6968 * set the recall flag. 6969 */ 6970 6971 dsp = rfs4_grant_delegation(deleg, sp, &recall); 6972 6973 cs->deleg = (fp->rf_dinfo.rd_dtype == OPEN_DELEGATE_WRITE); 6974 6975 next_stateid(&sp->rs_stateid); 6976 6977 resp->stateid = sp->rs_stateid.stateid; 6978 6979 rfs4_dbe_unlock(fp->rf_dbe); 6980 rfs4_dbe_unlock(sp->rs_dbe); 6981 6982 if (dsp) { 6983 rfs4_set_deleg_response(dsp, &resp->delegation, NULL, recall); 6984 rfs4_deleg_state_rele(dsp); 6985 } 6986 6987 rfs4_file_rele(fp); 6988 rfs4_state_rele(sp); 6989 6990 resp->status = NFS4_OK; 6991 } 6992 6993 /*ARGSUSED*/ 6994 static void 6995 rfs4_do_opennull(struct compound_state *cs, struct svc_req *req, 6996 OPEN4args *args, rfs4_openowner_t *oo, OPEN4res *resp) 6997 { 6998 change_info4 *cinfo = &resp->cinfo; 6999 bitmap4 *attrset = &resp->attrset; 7000 7001 if (args->opentype == OPEN4_NOCREATE) 7002 resp->status = rfs4_lookupfile(&args->open_claim4_u.file, 7003 req, cs, args->share_access, cinfo); 7004 else { 7005 /* inhibit delegation grants during exclusive create */ 7006 7007 if (args->mode == EXCLUSIVE4) 7008 rfs4_disable_delegation(); 7009 7010 resp->status = rfs4_createfile(args, req, cs, cinfo, attrset, 7011 oo->ro_client->rc_clientid); 7012 } 7013 7014 if (resp->status == NFS4_OK) { 7015 7016 /* cs->vp cs->fh now reference the desired file */ 7017 7018 rfs4_do_open(cs, req, oo, 7019 oo->ro_need_confirm ? DELEG_NONE : DELEG_ANY, 7020 args->share_access, args->share_deny, resp, 0); 7021 7022 /* 7023 * If rfs4_createfile set attrset, we must 7024 * clear this attrset before the response is copied. 7025 */ 7026 if (resp->status != NFS4_OK && resp->attrset) { 7027 resp->attrset = 0; 7028 } 7029 } 7030 else 7031 *cs->statusp = resp->status; 7032 7033 if (args->mode == EXCLUSIVE4) 7034 rfs4_enable_delegation(); 7035 } 7036 7037 /*ARGSUSED*/ 7038 static void 7039 rfs4_do_openprev(struct compound_state *cs, struct svc_req *req, 7040 OPEN4args *args, rfs4_openowner_t *oo, OPEN4res *resp) 7041 { 7042 change_info4 *cinfo = &resp->cinfo; 7043 vattr_t va; 7044 vtype_t v_type = cs->vp->v_type; 7045 int error = 0; 7046 7047 /* Verify that we have a regular file */ 7048 if (v_type != VREG) { 7049 if (v_type == VDIR) 7050 resp->status = NFS4ERR_ISDIR; 7051 else if (v_type == VLNK) 7052 resp->status = NFS4ERR_SYMLINK; 7053 else 7054 resp->status = NFS4ERR_INVAL; 7055 return; 7056 } 7057 7058 va.va_mask = AT_MODE|AT_UID; 7059 error = VOP_GETATTR(cs->vp, &va, 0, cs->cr, NULL); 7060 if (error) { 7061 resp->status = puterrno4(error); 7062 return; 7063 } 7064 7065 cs->mandlock = MANDLOCK(cs->vp, va.va_mode); 7066 7067 /* 7068 * Check if we have access to the file, Note the the file 7069 * could have originally been open UNCHECKED or GUARDED 7070 * with mode bits that will now fail, but there is nothing 7071 * we can really do about that except in the case that the 7072 * owner of the file is the one requesting the open. 7073 */ 7074 if (crgetuid(cs->cr) != va.va_uid) { 7075 resp->status = check_open_access(args->share_access, cs, req); 7076 if (resp->status != NFS4_OK) { 7077 return; 7078 } 7079 } 7080 7081 /* 7082 * cinfo on a CLAIM_PREVIOUS is undefined, initialize to zero 7083 */ 7084 cinfo->before = 0; 7085 cinfo->after = 0; 7086 cinfo->atomic = FALSE; 7087 7088 rfs4_do_open(cs, req, oo, 7089 NFS4_DELEG4TYPE2REQTYPE(args->open_claim4_u.delegate_type), 7090 args->share_access, args->share_deny, resp, 0); 7091 } 7092 7093 static void 7094 rfs4_do_opendelcur(struct compound_state *cs, struct svc_req *req, 7095 OPEN4args *args, rfs4_openowner_t *oo, OPEN4res *resp) 7096 { 7097 int error; 7098 nfsstat4 status; 7099 stateid4 stateid = 7100 args->open_claim4_u.delegate_cur_info.delegate_stateid; 7101 rfs4_deleg_state_t *dsp; 7102 7103 /* 7104 * Find the state info from the stateid and confirm that the 7105 * file is delegated. If the state openowner is the same as 7106 * the supplied openowner we're done. If not, get the file 7107 * info from the found state info. Use that file info to 7108 * create the state for this lock owner. Note solaris doen't 7109 * really need the pathname to find the file. We may want to 7110 * lookup the pathname and make sure that the vp exist and 7111 * matches the vp in the file structure. However it is 7112 * possible that the pathname nolonger exists (local process 7113 * unlinks the file), so this may not be that useful. 7114 */ 7115 7116 status = rfs4_get_deleg_state(&stateid, &dsp); 7117 if (status != NFS4_OK) { 7118 resp->status = status; 7119 return; 7120 } 7121 7122 ASSERT(dsp->rds_finfo->rf_dinfo.rd_dtype != OPEN_DELEGATE_NONE); 7123 7124 /* 7125 * New lock owner, create state. Since this was probably called 7126 * in response to a CB_RECALL we set deleg to DELEG_NONE 7127 */ 7128 7129 ASSERT(cs->vp != NULL); 7130 VN_RELE(cs->vp); 7131 VN_HOLD(dsp->rds_finfo->rf_vp); 7132 cs->vp = dsp->rds_finfo->rf_vp; 7133 7134 error = makefh4(&cs->fh, cs->vp, cs->exi); 7135 if (error != 0) { 7136 rfs4_deleg_state_rele(dsp); 7137 *cs->statusp = resp->status = puterrno4(error); 7138 return; 7139 } 7140 7141 /* Mark progress for delegation returns */ 7142 dsp->rds_finfo->rf_dinfo.rd_time_lastwrite = gethrestime_sec(); 7143 rfs4_deleg_state_rele(dsp); 7144 rfs4_do_open(cs, req, oo, DELEG_NONE, 7145 args->share_access, args->share_deny, resp, 1); 7146 } 7147 7148 /*ARGSUSED*/ 7149 static void 7150 rfs4_do_opendelprev(struct compound_state *cs, struct svc_req *req, 7151 OPEN4args *args, rfs4_openowner_t *oo, OPEN4res *resp) 7152 { 7153 /* 7154 * Lookup the pathname, it must already exist since this file 7155 * was delegated. 7156 * 7157 * Find the file and state info for this vp and open owner pair. 7158 * check that they are in fact delegated. 7159 * check that the state access and deny modes are the same. 7160 * 7161 * Return the delgation possibly seting the recall flag. 7162 */ 7163 rfs4_file_t *fp; 7164 rfs4_state_t *sp; 7165 bool_t create = FALSE; 7166 bool_t dcreate = FALSE; 7167 rfs4_deleg_state_t *dsp; 7168 nfsace4 *ace; 7169 7170 /* Note we ignore oflags */ 7171 resp->status = rfs4_lookupfile(&args->open_claim4_u.file_delegate_prev, 7172 req, cs, args->share_access, &resp->cinfo); 7173 7174 if (resp->status != NFS4_OK) { 7175 return; 7176 } 7177 7178 /* get the file struct and hold a lock on it during initial open */ 7179 fp = rfs4_findfile_withlock(cs->vp, NULL, &create); 7180 if (fp == NULL) { 7181 resp->status = NFS4ERR_RESOURCE; 7182 DTRACE_PROBE1(nfss__e__do_opendelprev1, nfsstat4, resp->status); 7183 return; 7184 } 7185 7186 sp = rfs4_findstate_by_owner_file(oo, fp, &create); 7187 if (sp == NULL) { 7188 resp->status = NFS4ERR_SERVERFAULT; 7189 DTRACE_PROBE1(nfss__e__do_opendelprev2, nfsstat4, resp->status); 7190 rw_exit(&fp->rf_file_rwlock); 7191 rfs4_file_rele(fp); 7192 return; 7193 } 7194 7195 rfs4_dbe_lock(sp->rs_dbe); 7196 rfs4_dbe_lock(fp->rf_dbe); 7197 if (args->share_access != sp->rs_share_access || 7198 args->share_deny != sp->rs_share_deny || 7199 sp->rs_finfo->rf_dinfo.rd_dtype == OPEN_DELEGATE_NONE) { 7200 NFS4_DEBUG(rfs4_debug, 7201 (CE_NOTE, "rfs4_do_opendelprev: state mixup")); 7202 rfs4_dbe_unlock(fp->rf_dbe); 7203 rfs4_dbe_unlock(sp->rs_dbe); 7204 rfs4_file_rele(fp); 7205 rfs4_state_rele(sp); 7206 resp->status = NFS4ERR_SERVERFAULT; 7207 return; 7208 } 7209 rfs4_dbe_unlock(fp->rf_dbe); 7210 rfs4_dbe_unlock(sp->rs_dbe); 7211 7212 dsp = rfs4_finddeleg(sp, &dcreate); 7213 if (dsp == NULL) { 7214 rfs4_state_rele(sp); 7215 rfs4_file_rele(fp); 7216 resp->status = NFS4ERR_SERVERFAULT; 7217 return; 7218 } 7219 7220 next_stateid(&sp->rs_stateid); 7221 7222 resp->stateid = sp->rs_stateid.stateid; 7223 7224 resp->delegation.delegation_type = dsp->rds_dtype; 7225 7226 if (dsp->rds_dtype == OPEN_DELEGATE_READ) { 7227 open_read_delegation4 *rv = 7228 &resp->delegation.open_delegation4_u.read; 7229 7230 rv->stateid = dsp->rds_delegid.stateid; 7231 rv->recall = FALSE; /* no policy in place to set to TRUE */ 7232 ace = &rv->permissions; 7233 } else { 7234 open_write_delegation4 *rv = 7235 &resp->delegation.open_delegation4_u.write; 7236 7237 rv->stateid = dsp->rds_delegid.stateid; 7238 rv->recall = FALSE; /* no policy in place to set to TRUE */ 7239 ace = &rv->permissions; 7240 rv->space_limit.limitby = NFS_LIMIT_SIZE; 7241 rv->space_limit.nfs_space_limit4_u.filesize = UINT64_MAX; 7242 } 7243 7244 /* XXX For now */ 7245 ace->type = ACE4_ACCESS_ALLOWED_ACE_TYPE; 7246 ace->flag = 0; 7247 ace->access_mask = 0; 7248 ace->who.utf8string_len = 0; 7249 ace->who.utf8string_val = 0; 7250 7251 rfs4_deleg_state_rele(dsp); 7252 rfs4_state_rele(sp); 7253 rfs4_file_rele(fp); 7254 } 7255 7256 typedef enum { 7257 NFS4_CHKSEQ_OKAY = 0, 7258 NFS4_CHKSEQ_REPLAY = 1, 7259 NFS4_CHKSEQ_BAD = 2 7260 } rfs4_chkseq_t; 7261 7262 /* 7263 * Generic function for sequence number checks. 7264 */ 7265 static rfs4_chkseq_t 7266 rfs4_check_seqid(seqid4 seqid, nfs_resop4 *lastop, 7267 seqid4 rqst_seq, nfs_resop4 *resop, bool_t copyres) 7268 { 7269 /* Same sequence ids and matching operations? */ 7270 if (seqid == rqst_seq && resop->resop == lastop->resop) { 7271 if (copyres == TRUE) { 7272 rfs4_free_reply(resop); 7273 rfs4_copy_reply(resop, lastop); 7274 } 7275 NFS4_DEBUG(rfs4_debug, (CE_NOTE, 7276 "Replayed SEQID %d\n", seqid)); 7277 return (NFS4_CHKSEQ_REPLAY); 7278 } 7279 7280 /* If the incoming sequence is not the next expected then it is bad */ 7281 if (rqst_seq != seqid + 1) { 7282 if (rqst_seq == seqid) { 7283 NFS4_DEBUG(rfs4_debug, 7284 (CE_NOTE, "BAD SEQID: Replayed sequence id " 7285 "but last op was %d current op is %d\n", 7286 lastop->resop, resop->resop)); 7287 return (NFS4_CHKSEQ_BAD); 7288 } 7289 NFS4_DEBUG(rfs4_debug, 7290 (CE_NOTE, "BAD SEQID: got %u expecting %u\n", 7291 rqst_seq, seqid)); 7292 return (NFS4_CHKSEQ_BAD); 7293 } 7294 7295 /* Everything okay -- next expected */ 7296 return (NFS4_CHKSEQ_OKAY); 7297 } 7298 7299 7300 static rfs4_chkseq_t 7301 rfs4_check_open_seqid(seqid4 seqid, rfs4_openowner_t *op, nfs_resop4 *resop) 7302 { 7303 rfs4_chkseq_t rc; 7304 7305 rfs4_dbe_lock(op->ro_dbe); 7306 rc = rfs4_check_seqid(op->ro_open_seqid, &op->ro_reply, seqid, resop, 7307 TRUE); 7308 rfs4_dbe_unlock(op->ro_dbe); 7309 7310 if (rc == NFS4_CHKSEQ_OKAY) 7311 rfs4_update_lease(op->ro_client); 7312 7313 return (rc); 7314 } 7315 7316 static rfs4_chkseq_t 7317 rfs4_check_olo_seqid(seqid4 olo_seqid, rfs4_openowner_t *op, nfs_resop4 *resop) 7318 { 7319 rfs4_chkseq_t rc; 7320 7321 rfs4_dbe_lock(op->ro_dbe); 7322 rc = rfs4_check_seqid(op->ro_open_seqid, &op->ro_reply, 7323 olo_seqid, resop, FALSE); 7324 rfs4_dbe_unlock(op->ro_dbe); 7325 7326 return (rc); 7327 } 7328 7329 static rfs4_chkseq_t 7330 rfs4_check_lock_seqid(seqid4 seqid, rfs4_lo_state_t *lsp, nfs_resop4 *resop) 7331 { 7332 rfs4_chkseq_t rc = NFS4_CHKSEQ_OKAY; 7333 7334 rfs4_dbe_lock(lsp->rls_dbe); 7335 if (!lsp->rls_skip_seqid_check) 7336 rc = rfs4_check_seqid(lsp->rls_seqid, &lsp->rls_reply, seqid, 7337 resop, TRUE); 7338 rfs4_dbe_unlock(lsp->rls_dbe); 7339 7340 return (rc); 7341 } 7342 7343 static void 7344 rfs4_op_open(nfs_argop4 *argop, nfs_resop4 *resop, 7345 struct svc_req *req, struct compound_state *cs) 7346 { 7347 OPEN4args *args = &argop->nfs_argop4_u.opopen; 7348 OPEN4res *resp = &resop->nfs_resop4_u.opopen; 7349 open_owner4 *owner = &args->owner; 7350 open_claim_type4 claim = args->claim; 7351 rfs4_client_t *cp; 7352 rfs4_openowner_t *oo; 7353 bool_t create; 7354 bool_t replay = FALSE; 7355 int can_reclaim; 7356 7357 DTRACE_NFSV4_2(op__open__start, struct compound_state *, cs, 7358 OPEN4args *, args); 7359 7360 if (cs->vp == NULL) { 7361 *cs->statusp = resp->status = NFS4ERR_NOFILEHANDLE; 7362 goto end; 7363 } 7364 7365 /* 7366 * Need to check clientid and lease expiration first based on 7367 * error ordering and incrementing sequence id. 7368 */ 7369 cp = rfs4_findclient_by_id(owner->clientid, FALSE); 7370 if (cp == NULL) { 7371 *cs->statusp = resp->status = 7372 rfs4_check_clientid(&owner->clientid, 0); 7373 goto end; 7374 } 7375 7376 if (rfs4_lease_expired(cp)) { 7377 rfs4_client_close(cp); 7378 *cs->statusp = resp->status = NFS4ERR_EXPIRED; 7379 goto end; 7380 } 7381 can_reclaim = cp->rc_can_reclaim; 7382 7383 /* 7384 * Find the open_owner for use from this point forward. Take 7385 * care in updating the sequence id based on the type of error 7386 * being returned. 7387 */ 7388 retry: 7389 create = TRUE; 7390 oo = rfs4_findopenowner(owner, &create, args->seqid); 7391 if (oo == NULL) { 7392 *cs->statusp = resp->status = NFS4ERR_RESOURCE; 7393 rfs4_client_rele(cp); 7394 goto end; 7395 } 7396 7397 /* Hold off access to the sequence space while the open is done */ 7398 rfs4_sw_enter(&oo->ro_sw); 7399 7400 /* 7401 * If the open_owner existed before at the server, then check 7402 * the sequence id. 7403 */ 7404 if (!create && !oo->ro_postpone_confirm) { 7405 switch (rfs4_check_open_seqid(args->seqid, oo, resop)) { 7406 case NFS4_CHKSEQ_BAD: 7407 if ((args->seqid > oo->ro_open_seqid) && 7408 oo->ro_need_confirm) { 7409 rfs4_free_opens(oo, TRUE, FALSE); 7410 rfs4_sw_exit(&oo->ro_sw); 7411 rfs4_openowner_rele(oo); 7412 goto retry; 7413 } 7414 resp->status = NFS4ERR_BAD_SEQID; 7415 goto out; 7416 case NFS4_CHKSEQ_REPLAY: /* replay of previous request */ 7417 replay = TRUE; 7418 goto out; 7419 default: 7420 break; 7421 } 7422 7423 /* 7424 * Sequence was ok and open owner exists 7425 * check to see if we have yet to see an 7426 * open_confirm. 7427 */ 7428 if (oo->ro_need_confirm) { 7429 rfs4_free_opens(oo, TRUE, FALSE); 7430 rfs4_sw_exit(&oo->ro_sw); 7431 rfs4_openowner_rele(oo); 7432 goto retry; 7433 } 7434 } 7435 /* Grace only applies to regular-type OPENs */ 7436 if (rfs4_clnt_in_grace(cp) && 7437 (claim == CLAIM_NULL || claim == CLAIM_DELEGATE_CUR)) { 7438 *cs->statusp = resp->status = NFS4ERR_GRACE; 7439 goto out; 7440 } 7441 7442 /* 7443 * If previous state at the server existed then can_reclaim 7444 * will be set. If not reply NFS4ERR_NO_GRACE to the 7445 * client. 7446 */ 7447 if (rfs4_clnt_in_grace(cp) && claim == CLAIM_PREVIOUS && !can_reclaim) { 7448 *cs->statusp = resp->status = NFS4ERR_NO_GRACE; 7449 goto out; 7450 } 7451 7452 7453 /* 7454 * Reject the open if the client has missed the grace period 7455 */ 7456 if (!rfs4_clnt_in_grace(cp) && claim == CLAIM_PREVIOUS) { 7457 *cs->statusp = resp->status = NFS4ERR_NO_GRACE; 7458 goto out; 7459 } 7460 7461 /* Couple of up-front bookkeeping items */ 7462 if (oo->ro_need_confirm) { 7463 /* 7464 * If this is a reclaim OPEN then we should not ask 7465 * for a confirmation of the open_owner per the 7466 * protocol specification. 7467 */ 7468 if (claim == CLAIM_PREVIOUS) 7469 oo->ro_need_confirm = FALSE; 7470 else 7471 resp->rflags |= OPEN4_RESULT_CONFIRM; 7472 } 7473 resp->rflags |= OPEN4_RESULT_LOCKTYPE_POSIX; 7474 7475 /* 7476 * If there is an unshared filesystem mounted on this vnode, 7477 * do not allow to open/create in this directory. 7478 */ 7479 if (vn_ismntpt(cs->vp)) { 7480 *cs->statusp = resp->status = NFS4ERR_ACCESS; 7481 goto out; 7482 } 7483 7484 /* 7485 * access must READ, WRITE, or BOTH. No access is invalid. 7486 * deny can be READ, WRITE, BOTH, or NONE. 7487 * bits not defined for access/deny are invalid. 7488 */ 7489 if (! (args->share_access & OPEN4_SHARE_ACCESS_BOTH) || 7490 (args->share_access & ~OPEN4_SHARE_ACCESS_BOTH) || 7491 (args->share_deny & ~OPEN4_SHARE_DENY_BOTH)) { 7492 *cs->statusp = resp->status = NFS4ERR_INVAL; 7493 goto out; 7494 } 7495 7496 7497 /* 7498 * make sure attrset is zero before response is built. 7499 */ 7500 resp->attrset = 0; 7501 7502 switch (claim) { 7503 case CLAIM_NULL: 7504 rfs4_do_opennull(cs, req, args, oo, resp); 7505 break; 7506 case CLAIM_PREVIOUS: 7507 rfs4_do_openprev(cs, req, args, oo, resp); 7508 break; 7509 case CLAIM_DELEGATE_CUR: 7510 rfs4_do_opendelcur(cs, req, args, oo, resp); 7511 break; 7512 case CLAIM_DELEGATE_PREV: 7513 rfs4_do_opendelprev(cs, req, args, oo, resp); 7514 break; 7515 default: 7516 resp->status = NFS4ERR_INVAL; 7517 break; 7518 } 7519 7520 out: 7521 rfs4_client_rele(cp); 7522 7523 /* Catch sequence id handling here to make it a little easier */ 7524 switch (resp->status) { 7525 case NFS4ERR_BADXDR: 7526 case NFS4ERR_BAD_SEQID: 7527 case NFS4ERR_BAD_STATEID: 7528 case NFS4ERR_NOFILEHANDLE: 7529 case NFS4ERR_RESOURCE: 7530 case NFS4ERR_STALE_CLIENTID: 7531 case NFS4ERR_STALE_STATEID: 7532 /* 7533 * The protocol states that if any of these errors are 7534 * being returned, the sequence id should not be 7535 * incremented. Any other return requires an 7536 * increment. 7537 */ 7538 break; 7539 default: 7540 /* Always update the lease in this case */ 7541 rfs4_update_lease(oo->ro_client); 7542 7543 /* Regular response - copy the result */ 7544 if (!replay) 7545 rfs4_update_open_resp(oo, resop, &cs->fh); 7546 7547 /* 7548 * REPLAY case: Only if the previous response was OK 7549 * do we copy the filehandle. If not OK, no 7550 * filehandle to copy. 7551 */ 7552 if (replay == TRUE && 7553 resp->status == NFS4_OK && 7554 oo->ro_reply_fh.nfs_fh4_val) { 7555 /* 7556 * If this is a replay, we must restore the 7557 * current filehandle/vp to that of what was 7558 * returned originally. Try our best to do 7559 * it. 7560 */ 7561 nfs_fh4_fmt_t *fh_fmtp = 7562 (nfs_fh4_fmt_t *)oo->ro_reply_fh.nfs_fh4_val; 7563 7564 cs->exi = checkexport4(&fh_fmtp->fh4_fsid, 7565 (fid_t *)&fh_fmtp->fh4_xlen, NULL); 7566 7567 if (cs->exi == NULL) { 7568 resp->status = NFS4ERR_STALE; 7569 goto finish; 7570 } 7571 7572 VN_RELE(cs->vp); 7573 7574 cs->vp = nfs4_fhtovp(&oo->ro_reply_fh, cs->exi, 7575 &resp->status); 7576 7577 if (cs->vp == NULL) 7578 goto finish; 7579 7580 nfs_fh4_copy(&oo->ro_reply_fh, &cs->fh); 7581 } 7582 7583 /* 7584 * If this was a replay, no need to update the 7585 * sequence id. If the open_owner was not created on 7586 * this pass, then update. The first use of an 7587 * open_owner will not bump the sequence id. 7588 */ 7589 if (replay == FALSE && !create) 7590 rfs4_update_open_sequence(oo); 7591 /* 7592 * If the client is receiving an error and the 7593 * open_owner needs to be confirmed, there is no way 7594 * to notify the client of this fact ignoring the fact 7595 * that the server has no method of returning a 7596 * stateid to confirm. Therefore, the server needs to 7597 * mark this open_owner in a way as to avoid the 7598 * sequence id checking the next time the client uses 7599 * this open_owner. 7600 */ 7601 if (resp->status != NFS4_OK && oo->ro_need_confirm) 7602 oo->ro_postpone_confirm = TRUE; 7603 /* 7604 * If OK response then clear the postpone flag and 7605 * reset the sequence id to keep in sync with the 7606 * client. 7607 */ 7608 if (resp->status == NFS4_OK && oo->ro_postpone_confirm) { 7609 oo->ro_postpone_confirm = FALSE; 7610 oo->ro_open_seqid = args->seqid; 7611 } 7612 break; 7613 } 7614 7615 finish: 7616 *cs->statusp = resp->status; 7617 7618 rfs4_sw_exit(&oo->ro_sw); 7619 rfs4_openowner_rele(oo); 7620 7621 end: 7622 DTRACE_NFSV4_2(op__open__done, struct compound_state *, cs, 7623 OPEN4res *, resp); 7624 } 7625 7626 /*ARGSUSED*/ 7627 void 7628 rfs4_op_open_confirm(nfs_argop4 *argop, nfs_resop4 *resop, 7629 struct svc_req *req, struct compound_state *cs) 7630 { 7631 OPEN_CONFIRM4args *args = &argop->nfs_argop4_u.opopen_confirm; 7632 OPEN_CONFIRM4res *resp = &resop->nfs_resop4_u.opopen_confirm; 7633 rfs4_state_t *sp; 7634 nfsstat4 status; 7635 7636 DTRACE_NFSV4_2(op__open__confirm__start, struct compound_state *, cs, 7637 OPEN_CONFIRM4args *, args); 7638 7639 if (cs->vp == NULL) { 7640 *cs->statusp = resp->status = NFS4ERR_NOFILEHANDLE; 7641 goto out; 7642 } 7643 7644 if (cs->vp->v_type != VREG) { 7645 *cs->statusp = resp->status = 7646 cs->vp->v_type == VDIR ? NFS4ERR_ISDIR : NFS4ERR_INVAL; 7647 return; 7648 } 7649 7650 status = rfs4_get_state(&args->open_stateid, &sp, RFS4_DBS_VALID); 7651 if (status != NFS4_OK) { 7652 *cs->statusp = resp->status = status; 7653 goto out; 7654 } 7655 7656 /* Ensure specified filehandle matches */ 7657 if (cs->vp != sp->rs_finfo->rf_vp) { 7658 rfs4_state_rele(sp); 7659 *cs->statusp = resp->status = NFS4ERR_BAD_STATEID; 7660 goto out; 7661 } 7662 7663 /* hold off other access to open_owner while we tinker */ 7664 rfs4_sw_enter(&sp->rs_owner->ro_sw); 7665 7666 switch (rfs4_check_stateid_seqid(sp, &args->open_stateid)) { 7667 case NFS4_CHECK_STATEID_OKAY: 7668 if (rfs4_check_open_seqid(args->seqid, sp->rs_owner, 7669 resop) != 0) { 7670 *cs->statusp = resp->status = NFS4ERR_BAD_SEQID; 7671 break; 7672 } 7673 /* 7674 * If it is the appropriate stateid and determined to 7675 * be "OKAY" then this means that the stateid does not 7676 * need to be confirmed and the client is in error for 7677 * sending an OPEN_CONFIRM. 7678 */ 7679 *cs->statusp = resp->status = NFS4ERR_BAD_STATEID; 7680 break; 7681 case NFS4_CHECK_STATEID_OLD: 7682 *cs->statusp = resp->status = NFS4ERR_OLD_STATEID; 7683 break; 7684 case NFS4_CHECK_STATEID_BAD: 7685 *cs->statusp = resp->status = NFS4ERR_BAD_STATEID; 7686 break; 7687 case NFS4_CHECK_STATEID_EXPIRED: 7688 *cs->statusp = resp->status = NFS4ERR_EXPIRED; 7689 break; 7690 case NFS4_CHECK_STATEID_CLOSED: 7691 *cs->statusp = resp->status = NFS4ERR_OLD_STATEID; 7692 break; 7693 case NFS4_CHECK_STATEID_REPLAY: 7694 switch (rfs4_check_open_seqid(args->seqid, sp->rs_owner, 7695 resop)) { 7696 case NFS4_CHKSEQ_OKAY: 7697 /* 7698 * This is replayed stateid; if seqid matches 7699 * next expected, then client is using wrong seqid. 7700 */ 7701 /* fall through */ 7702 case NFS4_CHKSEQ_BAD: 7703 *cs->statusp = resp->status = NFS4ERR_BAD_SEQID; 7704 break; 7705 case NFS4_CHKSEQ_REPLAY: 7706 /* 7707 * Note this case is the duplicate case so 7708 * resp->status is already set. 7709 */ 7710 *cs->statusp = resp->status; 7711 rfs4_update_lease(sp->rs_owner->ro_client); 7712 break; 7713 } 7714 break; 7715 case NFS4_CHECK_STATEID_UNCONFIRMED: 7716 if (rfs4_check_open_seqid(args->seqid, sp->rs_owner, 7717 resop) != NFS4_CHKSEQ_OKAY) { 7718 *cs->statusp = resp->status = NFS4ERR_BAD_SEQID; 7719 break; 7720 } 7721 *cs->statusp = resp->status = NFS4_OK; 7722 7723 next_stateid(&sp->rs_stateid); 7724 resp->open_stateid = sp->rs_stateid.stateid; 7725 sp->rs_owner->ro_need_confirm = FALSE; 7726 rfs4_update_lease(sp->rs_owner->ro_client); 7727 rfs4_update_open_sequence(sp->rs_owner); 7728 rfs4_update_open_resp(sp->rs_owner, resop, NULL); 7729 break; 7730 default: 7731 ASSERT(FALSE); 7732 *cs->statusp = resp->status = NFS4ERR_SERVERFAULT; 7733 break; 7734 } 7735 rfs4_sw_exit(&sp->rs_owner->ro_sw); 7736 rfs4_state_rele(sp); 7737 7738 out: 7739 DTRACE_NFSV4_2(op__open__confirm__done, struct compound_state *, cs, 7740 OPEN_CONFIRM4res *, resp); 7741 } 7742 7743 /*ARGSUSED*/ 7744 void 7745 rfs4_op_open_downgrade(nfs_argop4 *argop, nfs_resop4 *resop, 7746 struct svc_req *req, struct compound_state *cs) 7747 { 7748 OPEN_DOWNGRADE4args *args = &argop->nfs_argop4_u.opopen_downgrade; 7749 OPEN_DOWNGRADE4res *resp = &resop->nfs_resop4_u.opopen_downgrade; 7750 uint32_t access = args->share_access; 7751 uint32_t deny = args->share_deny; 7752 nfsstat4 status; 7753 rfs4_state_t *sp; 7754 rfs4_file_t *fp; 7755 int fflags = 0; 7756 7757 DTRACE_NFSV4_2(op__open__downgrade__start, struct compound_state *, cs, 7758 OPEN_DOWNGRADE4args *, args); 7759 7760 if (cs->vp == NULL) { 7761 *cs->statusp = resp->status = NFS4ERR_NOFILEHANDLE; 7762 goto out; 7763 } 7764 7765 if (cs->vp->v_type != VREG) { 7766 *cs->statusp = resp->status = NFS4ERR_INVAL; 7767 return; 7768 } 7769 7770 status = rfs4_get_state(&args->open_stateid, &sp, RFS4_DBS_VALID); 7771 if (status != NFS4_OK) { 7772 *cs->statusp = resp->status = status; 7773 goto out; 7774 } 7775 7776 /* Ensure specified filehandle matches */ 7777 if (cs->vp != sp->rs_finfo->rf_vp) { 7778 rfs4_state_rele(sp); 7779 *cs->statusp = resp->status = NFS4ERR_BAD_STATEID; 7780 goto out; 7781 } 7782 7783 /* hold off other access to open_owner while we tinker */ 7784 rfs4_sw_enter(&sp->rs_owner->ro_sw); 7785 7786 switch (rfs4_check_stateid_seqid(sp, &args->open_stateid)) { 7787 case NFS4_CHECK_STATEID_OKAY: 7788 if (rfs4_check_open_seqid(args->seqid, sp->rs_owner, 7789 resop) != NFS4_CHKSEQ_OKAY) { 7790 *cs->statusp = resp->status = NFS4ERR_BAD_SEQID; 7791 goto end; 7792 } 7793 break; 7794 case NFS4_CHECK_STATEID_OLD: 7795 *cs->statusp = resp->status = NFS4ERR_OLD_STATEID; 7796 goto end; 7797 case NFS4_CHECK_STATEID_BAD: 7798 *cs->statusp = resp->status = NFS4ERR_BAD_STATEID; 7799 goto end; 7800 case NFS4_CHECK_STATEID_EXPIRED: 7801 *cs->statusp = resp->status = NFS4ERR_EXPIRED; 7802 goto end; 7803 case NFS4_CHECK_STATEID_CLOSED: 7804 *cs->statusp = resp->status = NFS4ERR_OLD_STATEID; 7805 goto end; 7806 case NFS4_CHECK_STATEID_UNCONFIRMED: 7807 *cs->statusp = resp->status = NFS4ERR_BAD_STATEID; 7808 goto end; 7809 case NFS4_CHECK_STATEID_REPLAY: 7810 /* Check the sequence id for the open owner */ 7811 switch (rfs4_check_open_seqid(args->seqid, sp->rs_owner, 7812 resop)) { 7813 case NFS4_CHKSEQ_OKAY: 7814 /* 7815 * This is replayed stateid; if seqid matches 7816 * next expected, then client is using wrong seqid. 7817 */ 7818 /* fall through */ 7819 case NFS4_CHKSEQ_BAD: 7820 *cs->statusp = resp->status = NFS4ERR_BAD_SEQID; 7821 goto end; 7822 case NFS4_CHKSEQ_REPLAY: 7823 /* 7824 * Note this case is the duplicate case so 7825 * resp->status is already set. 7826 */ 7827 *cs->statusp = resp->status; 7828 rfs4_update_lease(sp->rs_owner->ro_client); 7829 goto end; 7830 } 7831 break; 7832 default: 7833 ASSERT(FALSE); 7834 break; 7835 } 7836 7837 rfs4_dbe_lock(sp->rs_dbe); 7838 /* 7839 * Check that the new access modes and deny modes are valid. 7840 * Check that no invalid bits are set. 7841 */ 7842 if ((access & ~(OPEN4_SHARE_ACCESS_READ | OPEN4_SHARE_ACCESS_WRITE)) || 7843 (deny & ~(OPEN4_SHARE_DENY_READ | OPEN4_SHARE_DENY_WRITE))) { 7844 *cs->statusp = resp->status = NFS4ERR_INVAL; 7845 rfs4_update_open_sequence(sp->rs_owner); 7846 rfs4_dbe_unlock(sp->rs_dbe); 7847 goto end; 7848 } 7849 7850 /* 7851 * The new modes must be a subset of the current modes and 7852 * the access must specify at least one mode. To test that 7853 * the new mode is a subset of the current modes we bitwise 7854 * AND them together and check that the result equals the new 7855 * mode. For example: 7856 * New mode, access == R and current mode, sp->rs_open_access == RW 7857 * access & sp->rs_open_access == R == access, so the new access mode 7858 * is valid. Consider access == RW, sp->rs_open_access = R 7859 * access & sp->rs_open_access == R != access, so the new access mode 7860 * is invalid. 7861 */ 7862 if ((access & sp->rs_open_access) != access || 7863 (deny & sp->rs_open_deny) != deny || 7864 (access & 7865 (OPEN4_SHARE_ACCESS_READ | OPEN4_SHARE_ACCESS_WRITE)) == 0) { 7866 *cs->statusp = resp->status = NFS4ERR_INVAL; 7867 rfs4_update_open_sequence(sp->rs_owner); 7868 rfs4_dbe_unlock(sp->rs_dbe); 7869 goto end; 7870 } 7871 7872 /* 7873 * Release any share locks associated with this stateID. 7874 * Strictly speaking, this violates the spec because the 7875 * spec effectively requires that open downgrade be atomic. 7876 * At present, fs_shrlock does not have this capability. 7877 */ 7878 (void) rfs4_unshare(sp); 7879 7880 status = rfs4_share(sp, access, deny); 7881 if (status != NFS4_OK) { 7882 *cs->statusp = resp->status = NFS4ERR_SERVERFAULT; 7883 rfs4_update_open_sequence(sp->rs_owner); 7884 rfs4_dbe_unlock(sp->rs_dbe); 7885 goto end; 7886 } 7887 7888 fp = sp->rs_finfo; 7889 rfs4_dbe_lock(fp->rf_dbe); 7890 7891 /* 7892 * If the current mode has deny read and the new mode 7893 * does not, decrement the number of deny read mode bits 7894 * and if it goes to zero turn off the deny read bit 7895 * on the file. 7896 */ 7897 if ((sp->rs_open_deny & OPEN4_SHARE_DENY_READ) && 7898 (deny & OPEN4_SHARE_DENY_READ) == 0) { 7899 fp->rf_deny_read--; 7900 if (fp->rf_deny_read == 0) 7901 fp->rf_share_deny &= ~OPEN4_SHARE_DENY_READ; 7902 } 7903 7904 /* 7905 * If the current mode has deny write and the new mode 7906 * does not, decrement the number of deny write mode bits 7907 * and if it goes to zero turn off the deny write bit 7908 * on the file. 7909 */ 7910 if ((sp->rs_open_deny & OPEN4_SHARE_DENY_WRITE) && 7911 (deny & OPEN4_SHARE_DENY_WRITE) == 0) { 7912 fp->rf_deny_write--; 7913 if (fp->rf_deny_write == 0) 7914 fp->rf_share_deny &= ~OPEN4_SHARE_DENY_WRITE; 7915 } 7916 7917 /* 7918 * If the current mode has access read and the new mode 7919 * does not, decrement the number of access read mode bits 7920 * and if it goes to zero turn off the access read bit 7921 * on the file. set fflags to FREAD for the call to 7922 * vn_open_downgrade(). 7923 */ 7924 if ((sp->rs_open_access & OPEN4_SHARE_ACCESS_READ) && 7925 (access & OPEN4_SHARE_ACCESS_READ) == 0) { 7926 fp->rf_access_read--; 7927 if (fp->rf_access_read == 0) 7928 fp->rf_share_access &= ~OPEN4_SHARE_ACCESS_READ; 7929 fflags |= FREAD; 7930 } 7931 7932 /* 7933 * If the current mode has access write and the new mode 7934 * does not, decrement the number of access write mode bits 7935 * and if it goes to zero turn off the access write bit 7936 * on the file. set fflags to FWRITE for the call to 7937 * vn_open_downgrade(). 7938 */ 7939 if ((sp->rs_open_access & OPEN4_SHARE_ACCESS_WRITE) && 7940 (access & OPEN4_SHARE_ACCESS_WRITE) == 0) { 7941 fp->rf_access_write--; 7942 if (fp->rf_access_write == 0) 7943 fp->rf_share_deny &= ~OPEN4_SHARE_ACCESS_WRITE; 7944 fflags |= FWRITE; 7945 } 7946 7947 /* Check that the file is still accessible */ 7948 ASSERT(fp->rf_share_access); 7949 7950 rfs4_dbe_unlock(fp->rf_dbe); 7951 7952 /* now set the new open access and deny modes */ 7953 sp->rs_open_access = access; 7954 sp->rs_open_deny = deny; 7955 7956 /* 7957 * we successfully downgraded the share lock, now we need to downgrade 7958 * the open. it is possible that the downgrade was only for a deny 7959 * mode and we have nothing else to do. 7960 */ 7961 if ((fflags & (FREAD|FWRITE)) != 0) 7962 vn_open_downgrade(cs->vp, fflags); 7963 7964 /* Update the stateid */ 7965 next_stateid(&sp->rs_stateid); 7966 resp->open_stateid = sp->rs_stateid.stateid; 7967 7968 rfs4_dbe_unlock(sp->rs_dbe); 7969 7970 *cs->statusp = resp->status = NFS4_OK; 7971 /* Update the lease */ 7972 rfs4_update_lease(sp->rs_owner->ro_client); 7973 /* And the sequence */ 7974 rfs4_update_open_sequence(sp->rs_owner); 7975 rfs4_update_open_resp(sp->rs_owner, resop, NULL); 7976 7977 end: 7978 rfs4_sw_exit(&sp->rs_owner->ro_sw); 7979 rfs4_state_rele(sp); 7980 out: 7981 DTRACE_NFSV4_2(op__open__downgrade__done, struct compound_state *, cs, 7982 OPEN_DOWNGRADE4res *, resp); 7983 } 7984 7985 static void * 7986 memstr(const void *s1, const char *s2, size_t n) 7987 { 7988 size_t l = strlen(s2); 7989 char *p = (char *)s1; 7990 7991 while (n >= l) { 7992 if (bcmp(p, s2, l) == 0) 7993 return (p); 7994 p++; 7995 n--; 7996 } 7997 7998 return (NULL); 7999 } 8000 8001 /* 8002 * The logic behind this function is detailed in the NFSv4 RFC in the 8003 * SETCLIENTID operation description under IMPLEMENTATION. Refer to 8004 * that section for explicit guidance to server behavior for 8005 * SETCLIENTID. 8006 */ 8007 void 8008 rfs4_op_setclientid(nfs_argop4 *argop, nfs_resop4 *resop, 8009 struct svc_req *req, struct compound_state *cs) 8010 { 8011 SETCLIENTID4args *args = &argop->nfs_argop4_u.opsetclientid; 8012 SETCLIENTID4res *res = &resop->nfs_resop4_u.opsetclientid; 8013 rfs4_client_t *cp, *newcp, *cp_confirmed, *cp_unconfirmed; 8014 rfs4_clntip_t *ci; 8015 bool_t create; 8016 char *addr, *netid; 8017 int len; 8018 8019 DTRACE_NFSV4_2(op__setclientid__start, struct compound_state *, cs, 8020 SETCLIENTID4args *, args); 8021 retry: 8022 newcp = cp_confirmed = cp_unconfirmed = NULL; 8023 8024 /* 8025 * Save the caller's IP address 8026 */ 8027 args->client.cl_addr = 8028 (struct sockaddr *)svc_getrpccaller(req->rq_xprt)->buf; 8029 8030 /* 8031 * Record if it is a Solaris client that cannot handle referrals. 8032 */ 8033 if (memstr(args->client.id_val, "Solaris", args->client.id_len) && 8034 !memstr(args->client.id_val, "+referrals", args->client.id_len)) { 8035 /* Add a "yes, it's downrev" record */ 8036 create = TRUE; 8037 ci = rfs4_find_clntip(args->client.cl_addr, &create); 8038 ASSERT(ci != NULL); 8039 rfs4_dbe_rele(ci->ri_dbe); 8040 } else { 8041 /* Remove any previous record */ 8042 rfs4_invalidate_clntip(args->client.cl_addr); 8043 } 8044 8045 /* 8046 * In search of an EXISTING client matching the incoming 8047 * request to establish a new client identifier at the server 8048 */ 8049 create = TRUE; 8050 cp = rfs4_findclient(&args->client, &create, NULL); 8051 8052 /* Should never happen */ 8053 ASSERT(cp != NULL); 8054 8055 if (cp == NULL) { 8056 *cs->statusp = res->status = NFS4ERR_SERVERFAULT; 8057 goto out; 8058 } 8059 8060 /* 8061 * Easiest case. Client identifier is newly created and is 8062 * unconfirmed. Also note that for this case, no other 8063 * entries exist for the client identifier. Nothing else to 8064 * check. Just setup the response and respond. 8065 */ 8066 if (create) { 8067 *cs->statusp = res->status = NFS4_OK; 8068 res->SETCLIENTID4res_u.resok4.clientid = cp->rc_clientid; 8069 res->SETCLIENTID4res_u.resok4.setclientid_confirm = 8070 cp->rc_confirm_verf; 8071 /* Setup callback information; CB_NULL confirmation later */ 8072 rfs4_client_setcb(cp, &args->callback, args->callback_ident); 8073 8074 rfs4_client_rele(cp); 8075 goto out; 8076 } 8077 8078 /* 8079 * An existing, confirmed client may exist but it may not have 8080 * been active for at least one lease period. If so, then 8081 * "close" the client and create a new client identifier 8082 */ 8083 if (rfs4_lease_expired(cp)) { 8084 rfs4_client_close(cp); 8085 goto retry; 8086 } 8087 8088 if (cp->rc_need_confirm == TRUE) 8089 cp_unconfirmed = cp; 8090 else 8091 cp_confirmed = cp; 8092 8093 cp = NULL; 8094 8095 /* 8096 * We have a confirmed client, now check for an 8097 * unconfimred entry 8098 */ 8099 if (cp_confirmed) { 8100 /* If creds don't match then client identifier is inuse */ 8101 if (!creds_ok(cp_confirmed->rc_cr_set, req, cs)) { 8102 rfs4_cbinfo_t *cbp; 8103 /* 8104 * Some one else has established this client 8105 * id. Try and say * who they are. We will use 8106 * the call back address supplied by * the 8107 * first client. 8108 */ 8109 *cs->statusp = res->status = NFS4ERR_CLID_INUSE; 8110 8111 addr = netid = NULL; 8112 8113 cbp = &cp_confirmed->rc_cbinfo; 8114 if (cbp->cb_callback.cb_location.r_addr && 8115 cbp->cb_callback.cb_location.r_netid) { 8116 cb_client4 *cbcp = &cbp->cb_callback; 8117 8118 len = strlen(cbcp->cb_location.r_addr)+1; 8119 addr = kmem_alloc(len, KM_SLEEP); 8120 bcopy(cbcp->cb_location.r_addr, addr, len); 8121 len = strlen(cbcp->cb_location.r_netid)+1; 8122 netid = kmem_alloc(len, KM_SLEEP); 8123 bcopy(cbcp->cb_location.r_netid, netid, len); 8124 } 8125 8126 res->SETCLIENTID4res_u.client_using.r_addr = addr; 8127 res->SETCLIENTID4res_u.client_using.r_netid = netid; 8128 8129 rfs4_client_rele(cp_confirmed); 8130 } 8131 8132 /* 8133 * Confirmed, creds match, and verifier matches; must 8134 * be an update of the callback info 8135 */ 8136 if (cp_confirmed->rc_nfs_client.verifier == 8137 args->client.verifier) { 8138 /* Setup callback information */ 8139 rfs4_client_setcb(cp_confirmed, &args->callback, 8140 args->callback_ident); 8141 8142 /* everything okay -- move ahead */ 8143 *cs->statusp = res->status = NFS4_OK; 8144 res->SETCLIENTID4res_u.resok4.clientid = 8145 cp_confirmed->rc_clientid; 8146 8147 /* update the confirm_verifier and return it */ 8148 rfs4_client_scv_next(cp_confirmed); 8149 res->SETCLIENTID4res_u.resok4.setclientid_confirm = 8150 cp_confirmed->rc_confirm_verf; 8151 8152 rfs4_client_rele(cp_confirmed); 8153 goto out; 8154 } 8155 8156 /* 8157 * Creds match but the verifier doesn't. Must search 8158 * for an unconfirmed client that would be replaced by 8159 * this request. 8160 */ 8161 create = FALSE; 8162 cp_unconfirmed = rfs4_findclient(&args->client, &create, 8163 cp_confirmed); 8164 } 8165 8166 /* 8167 * At this point, we have taken care of the brand new client 8168 * struct, INUSE case, update of an existing, and confirmed 8169 * client struct. 8170 */ 8171 8172 /* 8173 * check to see if things have changed while we originally 8174 * picked up the client struct. If they have, then return and 8175 * retry the processing of this SETCLIENTID request. 8176 */ 8177 if (cp_unconfirmed) { 8178 rfs4_dbe_lock(cp_unconfirmed->rc_dbe); 8179 if (!cp_unconfirmed->rc_need_confirm) { 8180 rfs4_dbe_unlock(cp_unconfirmed->rc_dbe); 8181 rfs4_client_rele(cp_unconfirmed); 8182 if (cp_confirmed) 8183 rfs4_client_rele(cp_confirmed); 8184 goto retry; 8185 } 8186 /* do away with the old unconfirmed one */ 8187 rfs4_dbe_invalidate(cp_unconfirmed->rc_dbe); 8188 rfs4_dbe_unlock(cp_unconfirmed->rc_dbe); 8189 rfs4_client_rele(cp_unconfirmed); 8190 cp_unconfirmed = NULL; 8191 } 8192 8193 /* 8194 * This search will temporarily hide the confirmed client 8195 * struct while a new client struct is created as the 8196 * unconfirmed one. 8197 */ 8198 create = TRUE; 8199 newcp = rfs4_findclient(&args->client, &create, cp_confirmed); 8200 8201 ASSERT(newcp != NULL); 8202 8203 if (newcp == NULL) { 8204 *cs->statusp = res->status = NFS4ERR_SERVERFAULT; 8205 rfs4_client_rele(cp_confirmed); 8206 goto out; 8207 } 8208 8209 /* 8210 * If one was not created, then a similar request must be in 8211 * process so release and start over with this one 8212 */ 8213 if (create != TRUE) { 8214 rfs4_client_rele(newcp); 8215 if (cp_confirmed) 8216 rfs4_client_rele(cp_confirmed); 8217 goto retry; 8218 } 8219 8220 *cs->statusp = res->status = NFS4_OK; 8221 res->SETCLIENTID4res_u.resok4.clientid = newcp->rc_clientid; 8222 res->SETCLIENTID4res_u.resok4.setclientid_confirm = 8223 newcp->rc_confirm_verf; 8224 /* Setup callback information; CB_NULL confirmation later */ 8225 rfs4_client_setcb(newcp, &args->callback, args->callback_ident); 8226 8227 newcp->rc_cp_confirmed = cp_confirmed; 8228 8229 rfs4_client_rele(newcp); 8230 8231 out: 8232 DTRACE_NFSV4_2(op__setclientid__done, struct compound_state *, cs, 8233 SETCLIENTID4res *, res); 8234 } 8235 8236 /*ARGSUSED*/ 8237 void 8238 rfs4_op_setclientid_confirm(nfs_argop4 *argop, nfs_resop4 *resop, 8239 struct svc_req *req, struct compound_state *cs) 8240 { 8241 SETCLIENTID_CONFIRM4args *args = 8242 &argop->nfs_argop4_u.opsetclientid_confirm; 8243 SETCLIENTID_CONFIRM4res *res = 8244 &resop->nfs_resop4_u.opsetclientid_confirm; 8245 rfs4_client_t *cp, *cptoclose = NULL; 8246 nfs4_srv_t *nsrv4; 8247 8248 DTRACE_NFSV4_2(op__setclientid__confirm__start, 8249 struct compound_state *, cs, 8250 SETCLIENTID_CONFIRM4args *, args); 8251 8252 nsrv4 = nfs4_get_srv(); 8253 *cs->statusp = res->status = NFS4_OK; 8254 8255 cp = rfs4_findclient_by_id(args->clientid, TRUE); 8256 8257 if (cp == NULL) { 8258 *cs->statusp = res->status = 8259 rfs4_check_clientid(&args->clientid, 1); 8260 goto out; 8261 } 8262 8263 if (!creds_ok(cp, req, cs)) { 8264 *cs->statusp = res->status = NFS4ERR_CLID_INUSE; 8265 rfs4_client_rele(cp); 8266 goto out; 8267 } 8268 8269 /* If the verifier doesn't match, the record doesn't match */ 8270 if (cp->rc_confirm_verf != args->setclientid_confirm) { 8271 *cs->statusp = res->status = NFS4ERR_STALE_CLIENTID; 8272 rfs4_client_rele(cp); 8273 goto out; 8274 } 8275 8276 rfs4_dbe_lock(cp->rc_dbe); 8277 cp->rc_need_confirm = FALSE; 8278 if (cp->rc_cp_confirmed) { 8279 cptoclose = cp->rc_cp_confirmed; 8280 cptoclose->rc_ss_remove = 1; 8281 cp->rc_cp_confirmed = NULL; 8282 } 8283 8284 /* 8285 * Update the client's associated server instance, if it's changed 8286 * since the client was created. 8287 */ 8288 if (rfs4_servinst(cp) != nsrv4->nfs4_cur_servinst) 8289 rfs4_servinst_assign(nsrv4, cp, nsrv4->nfs4_cur_servinst); 8290 8291 /* 8292 * Record clientid in stable storage. 8293 * Must be done after server instance has been assigned. 8294 */ 8295 rfs4_ss_clid(nsrv4, cp); 8296 8297 rfs4_dbe_unlock(cp->rc_dbe); 8298 8299 if (cptoclose) 8300 /* don't need to rele, client_close does it */ 8301 rfs4_client_close(cptoclose); 8302 8303 /* If needed, initiate CB_NULL call for callback path */ 8304 rfs4_deleg_cb_check(cp); 8305 rfs4_update_lease(cp); 8306 8307 /* 8308 * Check to see if client can perform reclaims 8309 */ 8310 rfs4_ss_chkclid(nsrv4, cp); 8311 8312 rfs4_client_rele(cp); 8313 8314 out: 8315 DTRACE_NFSV4_2(op__setclientid__confirm__done, 8316 struct compound_state *, cs, 8317 SETCLIENTID_CONFIRM4 *, res); 8318 } 8319 8320 8321 /*ARGSUSED*/ 8322 void 8323 rfs4_op_close(nfs_argop4 *argop, nfs_resop4 *resop, 8324 struct svc_req *req, struct compound_state *cs) 8325 { 8326 CLOSE4args *args = &argop->nfs_argop4_u.opclose; 8327 CLOSE4res *resp = &resop->nfs_resop4_u.opclose; 8328 rfs4_state_t *sp; 8329 nfsstat4 status; 8330 8331 DTRACE_NFSV4_2(op__close__start, struct compound_state *, cs, 8332 CLOSE4args *, args); 8333 8334 if (cs->vp == NULL) { 8335 *cs->statusp = resp->status = NFS4ERR_NOFILEHANDLE; 8336 goto out; 8337 } 8338 8339 status = rfs4_get_state(&args->open_stateid, &sp, RFS4_DBS_INVALID); 8340 if (status != NFS4_OK) { 8341 *cs->statusp = resp->status = status; 8342 goto out; 8343 } 8344 8345 /* Ensure specified filehandle matches */ 8346 if (cs->vp != sp->rs_finfo->rf_vp) { 8347 rfs4_state_rele(sp); 8348 *cs->statusp = resp->status = NFS4ERR_BAD_STATEID; 8349 goto out; 8350 } 8351 8352 /* hold off other access to open_owner while we tinker */ 8353 rfs4_sw_enter(&sp->rs_owner->ro_sw); 8354 8355 switch (rfs4_check_stateid_seqid(sp, &args->open_stateid)) { 8356 case NFS4_CHECK_STATEID_OKAY: 8357 if (rfs4_check_open_seqid(args->seqid, sp->rs_owner, 8358 resop) != NFS4_CHKSEQ_OKAY) { 8359 *cs->statusp = resp->status = NFS4ERR_BAD_SEQID; 8360 goto end; 8361 } 8362 break; 8363 case NFS4_CHECK_STATEID_OLD: 8364 *cs->statusp = resp->status = NFS4ERR_OLD_STATEID; 8365 goto end; 8366 case NFS4_CHECK_STATEID_BAD: 8367 *cs->statusp = resp->status = NFS4ERR_BAD_STATEID; 8368 goto end; 8369 case NFS4_CHECK_STATEID_EXPIRED: 8370 *cs->statusp = resp->status = NFS4ERR_EXPIRED; 8371 goto end; 8372 case NFS4_CHECK_STATEID_CLOSED: 8373 *cs->statusp = resp->status = NFS4ERR_OLD_STATEID; 8374 goto end; 8375 case NFS4_CHECK_STATEID_UNCONFIRMED: 8376 *cs->statusp = resp->status = NFS4ERR_BAD_STATEID; 8377 goto end; 8378 case NFS4_CHECK_STATEID_REPLAY: 8379 /* Check the sequence id for the open owner */ 8380 switch (rfs4_check_open_seqid(args->seqid, sp->rs_owner, 8381 resop)) { 8382 case NFS4_CHKSEQ_OKAY: 8383 /* 8384 * This is replayed stateid; if seqid matches 8385 * next expected, then client is using wrong seqid. 8386 */ 8387 /* FALL THROUGH */ 8388 case NFS4_CHKSEQ_BAD: 8389 *cs->statusp = resp->status = NFS4ERR_BAD_SEQID; 8390 goto end; 8391 case NFS4_CHKSEQ_REPLAY: 8392 /* 8393 * Note this case is the duplicate case so 8394 * resp->status is already set. 8395 */ 8396 *cs->statusp = resp->status; 8397 rfs4_update_lease(sp->rs_owner->ro_client); 8398 goto end; 8399 } 8400 break; 8401 default: 8402 ASSERT(FALSE); 8403 break; 8404 } 8405 8406 rfs4_dbe_lock(sp->rs_dbe); 8407 8408 /* Update the stateid. */ 8409 next_stateid(&sp->rs_stateid); 8410 resp->open_stateid = sp->rs_stateid.stateid; 8411 8412 rfs4_dbe_unlock(sp->rs_dbe); 8413 8414 rfs4_update_lease(sp->rs_owner->ro_client); 8415 rfs4_update_open_sequence(sp->rs_owner); 8416 rfs4_update_open_resp(sp->rs_owner, resop, NULL); 8417 8418 rfs4_state_close(sp, FALSE, FALSE, cs->cr); 8419 8420 *cs->statusp = resp->status = status; 8421 8422 end: 8423 rfs4_sw_exit(&sp->rs_owner->ro_sw); 8424 rfs4_state_rele(sp); 8425 out: 8426 DTRACE_NFSV4_2(op__close__done, struct compound_state *, cs, 8427 CLOSE4res *, resp); 8428 } 8429 8430 /* 8431 * Manage the counts on the file struct and close all file locks 8432 */ 8433 /*ARGSUSED*/ 8434 void 8435 rfs4_release_share_lock_state(rfs4_state_t *sp, cred_t *cr, 8436 bool_t close_of_client) 8437 { 8438 rfs4_file_t *fp = sp->rs_finfo; 8439 rfs4_lo_state_t *lsp; 8440 int fflags = 0; 8441 8442 /* 8443 * If this call is part of the larger closing down of client 8444 * state then it is just easier to release all locks 8445 * associated with this client instead of going through each 8446 * individual file and cleaning locks there. 8447 */ 8448 if (close_of_client) { 8449 if (sp->rs_owner->ro_client->rc_unlksys_completed == FALSE && 8450 !list_is_empty(&sp->rs_lostatelist) && 8451 sp->rs_owner->ro_client->rc_sysidt != LM_NOSYSID) { 8452 /* Is the PxFS kernel module loaded? */ 8453 if (lm_remove_file_locks != NULL) { 8454 int new_sysid; 8455 8456 /* Encode the cluster nodeid in new sysid */ 8457 new_sysid = sp->rs_owner->ro_client->rc_sysidt; 8458 lm_set_nlmid_flk(&new_sysid); 8459 8460 /* 8461 * This PxFS routine removes file locks for a 8462 * client over all nodes of a cluster. 8463 */ 8464 NFS4_DEBUG(rfs4_debug, (CE_NOTE, 8465 "lm_remove_file_locks(sysid=0x%x)\n", 8466 new_sysid)); 8467 (*lm_remove_file_locks)(new_sysid); 8468 } else { 8469 struct flock64 flk; 8470 8471 /* Release all locks for this client */ 8472 flk.l_type = F_UNLKSYS; 8473 flk.l_whence = 0; 8474 flk.l_start = 0; 8475 flk.l_len = 0; 8476 flk.l_sysid = 8477 sp->rs_owner->ro_client->rc_sysidt; 8478 flk.l_pid = 0; 8479 (void) VOP_FRLOCK(sp->rs_finfo->rf_vp, F_SETLK, 8480 &flk, F_REMOTELOCK | FREAD | FWRITE, 8481 (u_offset_t)0, NULL, CRED(), NULL); 8482 } 8483 8484 sp->rs_owner->ro_client->rc_unlksys_completed = TRUE; 8485 } 8486 } 8487 8488 /* 8489 * Release all locks on this file by this lock owner or at 8490 * least mark the locks as having been released 8491 */ 8492 for (lsp = list_head(&sp->rs_lostatelist); lsp != NULL; 8493 lsp = list_next(&sp->rs_lostatelist, lsp)) { 8494 lsp->rls_locks_cleaned = TRUE; 8495 8496 /* Was this already taken care of above? */ 8497 if (!close_of_client && 8498 sp->rs_owner->ro_client->rc_sysidt != LM_NOSYSID) 8499 (void) cleanlocks(sp->rs_finfo->rf_vp, 8500 lsp->rls_locker->rl_pid, 8501 lsp->rls_locker->rl_client->rc_sysidt); 8502 } 8503 8504 /* 8505 * Release any shrlocks associated with this open state ID. 8506 * This must be done before the rfs4_state gets marked closed. 8507 */ 8508 if (sp->rs_owner->ro_client->rc_sysidt != LM_NOSYSID) 8509 (void) rfs4_unshare(sp); 8510 8511 if (sp->rs_open_access) { 8512 rfs4_dbe_lock(fp->rf_dbe); 8513 8514 /* 8515 * Decrement the count for each access and deny bit that this 8516 * state has contributed to the file. 8517 * If the file counts go to zero 8518 * clear the appropriate bit in the appropriate mask. 8519 */ 8520 if (sp->rs_open_access & OPEN4_SHARE_ACCESS_READ) { 8521 fp->rf_access_read--; 8522 fflags |= FREAD; 8523 if (fp->rf_access_read == 0) 8524 fp->rf_share_access &= ~OPEN4_SHARE_ACCESS_READ; 8525 } 8526 if (sp->rs_open_access & OPEN4_SHARE_ACCESS_WRITE) { 8527 fp->rf_access_write--; 8528 fflags |= FWRITE; 8529 if (fp->rf_access_write == 0) 8530 fp->rf_share_access &= 8531 ~OPEN4_SHARE_ACCESS_WRITE; 8532 } 8533 if (sp->rs_open_deny & OPEN4_SHARE_DENY_READ) { 8534 fp->rf_deny_read--; 8535 if (fp->rf_deny_read == 0) 8536 fp->rf_share_deny &= ~OPEN4_SHARE_DENY_READ; 8537 } 8538 if (sp->rs_open_deny & OPEN4_SHARE_DENY_WRITE) { 8539 fp->rf_deny_write--; 8540 if (fp->rf_deny_write == 0) 8541 fp->rf_share_deny &= ~OPEN4_SHARE_DENY_WRITE; 8542 } 8543 8544 (void) VOP_CLOSE(fp->rf_vp, fflags, 1, (offset_t)0, cr, NULL); 8545 8546 rfs4_dbe_unlock(fp->rf_dbe); 8547 8548 sp->rs_open_access = 0; 8549 sp->rs_open_deny = 0; 8550 } 8551 } 8552 8553 /* 8554 * lock_denied: Fill in a LOCK4deneid structure given an flock64 structure. 8555 */ 8556 static nfsstat4 8557 lock_denied(LOCK4denied *dp, struct flock64 *flk) 8558 { 8559 rfs4_lockowner_t *lo; 8560 rfs4_client_t *cp; 8561 uint32_t len; 8562 8563 lo = rfs4_findlockowner_by_pid(flk->l_pid); 8564 if (lo != NULL) { 8565 cp = lo->rl_client; 8566 if (rfs4_lease_expired(cp)) { 8567 rfs4_lockowner_rele(lo); 8568 rfs4_dbe_hold(cp->rc_dbe); 8569 rfs4_client_close(cp); 8570 return (NFS4ERR_EXPIRED); 8571 } 8572 dp->owner.clientid = lo->rl_owner.clientid; 8573 len = lo->rl_owner.owner_len; 8574 dp->owner.owner_val = kmem_alloc(len, KM_SLEEP); 8575 bcopy(lo->rl_owner.owner_val, dp->owner.owner_val, len); 8576 dp->owner.owner_len = len; 8577 rfs4_lockowner_rele(lo); 8578 goto finish; 8579 } 8580 8581 /* 8582 * Its not a NFS4 lock. We take advantage that the upper 32 bits 8583 * of the client id contain the boot time for a NFS4 lock. So we 8584 * fabricate and identity by setting clientid to the sysid, and 8585 * the lock owner to the pid. 8586 */ 8587 dp->owner.clientid = flk->l_sysid; 8588 len = sizeof (pid_t); 8589 dp->owner.owner_len = len; 8590 dp->owner.owner_val = kmem_alloc(len, KM_SLEEP); 8591 bcopy(&flk->l_pid, dp->owner.owner_val, len); 8592 finish: 8593 dp->offset = flk->l_start; 8594 dp->length = flk->l_len; 8595 8596 if (flk->l_type == F_RDLCK) 8597 dp->locktype = READ_LT; 8598 else if (flk->l_type == F_WRLCK) 8599 dp->locktype = WRITE_LT; 8600 else 8601 return (NFS4ERR_INVAL); /* no mapping from POSIX ltype to v4 */ 8602 8603 return (NFS4_OK); 8604 } 8605 8606 /* 8607 * The NFSv4.0 LOCK operation does not support the blocking lock (at the 8608 * NFSv4.0 protocol level) so the client needs to resend the LOCK request in a 8609 * case the lock is denied by the NFSv4.0 server. NFSv4.0 clients are prepared 8610 * for that (obviously); they are sending the LOCK requests with some delays 8611 * between the attempts. See nfs4frlock() and nfs4_block_and_wait() for the 8612 * locking and delay implementation at the client side. 8613 * 8614 * To make the life of the clients easier, the NFSv4.0 server tries to do some 8615 * fast retries on its own (the for loop below) in a hope the lock will be 8616 * available soon. And if not, the client won't need to resend the LOCK 8617 * requests so fast to check the lock availability. This basically saves some 8618 * network traffic and tries to make sure the client gets the lock ASAP. 8619 */ 8620 static int 8621 setlock(vnode_t *vp, struct flock64 *flock, int flag, cred_t *cred) 8622 { 8623 int error; 8624 struct flock64 flk; 8625 int i; 8626 clock_t delaytime; 8627 int cmd; 8628 int spin_cnt = 0; 8629 8630 cmd = nbl_need_check(vp) ? F_SETLK_NBMAND : F_SETLK; 8631 retry: 8632 delaytime = MSEC_TO_TICK_ROUNDUP(rfs4_lock_delay); 8633 8634 for (i = 0; i < rfs4_maxlock_tries; i++) { 8635 LOCK_PRINT(rfs4_debug, "setlock", cmd, flock); 8636 error = VOP_FRLOCK(vp, cmd, 8637 flock, flag, (u_offset_t)0, NULL, cred, NULL); 8638 8639 if (error != EAGAIN && error != EACCES) 8640 break; 8641 8642 if (i < rfs4_maxlock_tries - 1) { 8643 delay(delaytime); 8644 delaytime *= 2; 8645 } 8646 } 8647 8648 if (error == EAGAIN || error == EACCES) { 8649 /* Get the owner of the lock */ 8650 flk = *flock; 8651 LOCK_PRINT(rfs4_debug, "setlock", F_GETLK, &flk); 8652 if (VOP_FRLOCK(vp, F_GETLK, &flk, flag, 0, NULL, cred, 8653 NULL) == 0) { 8654 /* 8655 * There's a race inherent in the current VOP_FRLOCK 8656 * design where: 8657 * a: "other guy" takes a lock that conflicts with a 8658 * lock we want 8659 * b: we attempt to take our lock (non-blocking) and 8660 * the attempt fails. 8661 * c: "other guy" releases the conflicting lock 8662 * d: we ask what lock conflicts with the lock we want, 8663 * getting F_UNLCK (no lock blocks us) 8664 * 8665 * If we retry the non-blocking lock attempt in this 8666 * case (restart at step 'b') there's some possibility 8667 * that many such attempts might fail. However a test 8668 * designed to actually provoke this race shows that 8669 * the vast majority of cases require no retry, and 8670 * only a few took as many as three retries. Here's 8671 * the test outcome: 8672 * 8673 * number of retries how many times we needed 8674 * that many retries 8675 * 0 79461 8676 * 1 862 8677 * 2 49 8678 * 3 5 8679 * 8680 * Given those empirical results, we arbitrarily limit 8681 * the retry count to ten. 8682 * 8683 * If we actually make to ten retries and give up, 8684 * nothing catastrophic happens, but we're unable to 8685 * return the information about the conflicting lock to 8686 * the NFS client. That's an acceptable trade off vs. 8687 * letting this retry loop run forever. 8688 */ 8689 if (flk.l_type == F_UNLCK) { 8690 if (spin_cnt++ < 10) { 8691 /* No longer locked, retry */ 8692 goto retry; 8693 } 8694 } else { 8695 *flock = flk; 8696 LOCK_PRINT(rfs4_debug, "setlock(blocking lock)", 8697 F_GETLK, &flk); 8698 } 8699 } 8700 } 8701 8702 return (error); 8703 } 8704 8705 /*ARGSUSED*/ 8706 static nfsstat4 8707 rfs4_do_lock(rfs4_lo_state_t *lsp, nfs_lock_type4 locktype, 8708 offset4 offset, length4 length, cred_t *cred, nfs_resop4 *resop) 8709 { 8710 nfsstat4 status; 8711 rfs4_lockowner_t *lo = lsp->rls_locker; 8712 rfs4_state_t *sp = lsp->rls_state; 8713 struct flock64 flock; 8714 int16_t ltype; 8715 int flag; 8716 int error; 8717 sysid_t sysid; 8718 LOCK4res *lres; 8719 vnode_t *vp; 8720 8721 if (rfs4_lease_expired(lo->rl_client)) { 8722 return (NFS4ERR_EXPIRED); 8723 } 8724 8725 if ((status = rfs4_client_sysid(lo->rl_client, &sysid)) != NFS4_OK) 8726 return (status); 8727 8728 /* Check for zero length. To lock to end of file use all ones for V4 */ 8729 if (length == 0) 8730 return (NFS4ERR_INVAL); 8731 else if (length == (length4)(~0)) 8732 length = 0; /* Posix to end of file */ 8733 8734 retry: 8735 rfs4_dbe_lock(sp->rs_dbe); 8736 if (sp->rs_closed == TRUE) { 8737 rfs4_dbe_unlock(sp->rs_dbe); 8738 return (NFS4ERR_OLD_STATEID); 8739 } 8740 8741 if (resop->resop != OP_LOCKU) { 8742 switch (locktype) { 8743 case READ_LT: 8744 case READW_LT: 8745 if ((sp->rs_share_access 8746 & OPEN4_SHARE_ACCESS_READ) == 0) { 8747 rfs4_dbe_unlock(sp->rs_dbe); 8748 8749 return (NFS4ERR_OPENMODE); 8750 } 8751 ltype = F_RDLCK; 8752 break; 8753 case WRITE_LT: 8754 case WRITEW_LT: 8755 if ((sp->rs_share_access 8756 & OPEN4_SHARE_ACCESS_WRITE) == 0) { 8757 rfs4_dbe_unlock(sp->rs_dbe); 8758 8759 return (NFS4ERR_OPENMODE); 8760 } 8761 ltype = F_WRLCK; 8762 break; 8763 } 8764 } else 8765 ltype = F_UNLCK; 8766 8767 flock.l_type = ltype; 8768 flock.l_whence = 0; /* SEEK_SET */ 8769 flock.l_start = offset; 8770 flock.l_len = length; 8771 flock.l_sysid = sysid; 8772 flock.l_pid = lsp->rls_locker->rl_pid; 8773 8774 /* Note that length4 is uint64_t but l_len and l_start are off64_t */ 8775 if (flock.l_len < 0 || flock.l_start < 0) { 8776 rfs4_dbe_unlock(sp->rs_dbe); 8777 return (NFS4ERR_INVAL); 8778 } 8779 8780 /* 8781 * N.B. FREAD has the same value as OPEN4_SHARE_ACCESS_READ and 8782 * FWRITE has the same value as OPEN4_SHARE_ACCESS_WRITE. 8783 */ 8784 flag = (int)sp->rs_share_access | F_REMOTELOCK; 8785 8786 vp = sp->rs_finfo->rf_vp; 8787 VN_HOLD(vp); 8788 8789 /* 8790 * We need to unlock sp before we call the underlying filesystem to 8791 * acquire the file lock. 8792 */ 8793 rfs4_dbe_unlock(sp->rs_dbe); 8794 8795 error = setlock(vp, &flock, flag, cred); 8796 8797 /* 8798 * Make sure the file is still open. In a case the file was closed in 8799 * the meantime, clean the lock we acquired using the setlock() call 8800 * above, and return the appropriate error. 8801 */ 8802 rfs4_dbe_lock(sp->rs_dbe); 8803 if (sp->rs_closed == TRUE) { 8804 cleanlocks(vp, lsp->rls_locker->rl_pid, sysid); 8805 rfs4_dbe_unlock(sp->rs_dbe); 8806 8807 VN_RELE(vp); 8808 8809 return (NFS4ERR_OLD_STATEID); 8810 } 8811 rfs4_dbe_unlock(sp->rs_dbe); 8812 8813 VN_RELE(vp); 8814 8815 if (error == 0) { 8816 rfs4_dbe_lock(lsp->rls_dbe); 8817 next_stateid(&lsp->rls_lockid); 8818 rfs4_dbe_unlock(lsp->rls_dbe); 8819 } 8820 8821 /* 8822 * N.B. We map error values to nfsv4 errors. This is differrent 8823 * than puterrno4 routine. 8824 */ 8825 switch (error) { 8826 case 0: 8827 status = NFS4_OK; 8828 break; 8829 case EAGAIN: 8830 case EACCES: /* Old value */ 8831 /* Can only get here if op is OP_LOCK */ 8832 ASSERT(resop->resop == OP_LOCK); 8833 lres = &resop->nfs_resop4_u.oplock; 8834 status = NFS4ERR_DENIED; 8835 if (lock_denied(&lres->LOCK4res_u.denied, &flock) 8836 == NFS4ERR_EXPIRED) 8837 goto retry; 8838 break; 8839 case ENOLCK: 8840 status = NFS4ERR_DELAY; 8841 break; 8842 case EOVERFLOW: 8843 status = NFS4ERR_INVAL; 8844 break; 8845 case EINVAL: 8846 status = NFS4ERR_NOTSUPP; 8847 break; 8848 default: 8849 status = NFS4ERR_SERVERFAULT; 8850 break; 8851 } 8852 8853 return (status); 8854 } 8855 8856 /*ARGSUSED*/ 8857 void 8858 rfs4_op_lock(nfs_argop4 *argop, nfs_resop4 *resop, 8859 struct svc_req *req, struct compound_state *cs) 8860 { 8861 LOCK4args *args = &argop->nfs_argop4_u.oplock; 8862 LOCK4res *resp = &resop->nfs_resop4_u.oplock; 8863 nfsstat4 status; 8864 stateid4 *stateid; 8865 rfs4_lockowner_t *lo; 8866 rfs4_client_t *cp; 8867 rfs4_state_t *sp = NULL; 8868 rfs4_lo_state_t *lsp = NULL; 8869 bool_t ls_sw_held = FALSE; 8870 bool_t create = TRUE; 8871 bool_t lcreate = TRUE; 8872 bool_t dup_lock = FALSE; 8873 int rc; 8874 8875 DTRACE_NFSV4_2(op__lock__start, struct compound_state *, cs, 8876 LOCK4args *, args); 8877 8878 if (cs->vp == NULL) { 8879 *cs->statusp = resp->status = NFS4ERR_NOFILEHANDLE; 8880 DTRACE_NFSV4_2(op__lock__done, struct compound_state *, 8881 cs, LOCK4res *, resp); 8882 return; 8883 } 8884 8885 if (args->locker.new_lock_owner) { 8886 /* Create a new lockowner for this instance */ 8887 open_to_lock_owner4 *olo = &args->locker.locker4_u.open_owner; 8888 8889 NFS4_DEBUG(rfs4_debug, (CE_NOTE, "Creating new lock owner")); 8890 8891 stateid = &olo->open_stateid; 8892 status = rfs4_get_state(stateid, &sp, RFS4_DBS_VALID); 8893 if (status != NFS4_OK) { 8894 NFS4_DEBUG(rfs4_debug, 8895 (CE_NOTE, "Get state failed in lock %d", status)); 8896 *cs->statusp = resp->status = status; 8897 DTRACE_NFSV4_2(op__lock__done, struct compound_state *, 8898 cs, LOCK4res *, resp); 8899 return; 8900 } 8901 8902 /* Ensure specified filehandle matches */ 8903 if (cs->vp != sp->rs_finfo->rf_vp) { 8904 rfs4_state_rele(sp); 8905 *cs->statusp = resp->status = NFS4ERR_BAD_STATEID; 8906 DTRACE_NFSV4_2(op__lock__done, struct compound_state *, 8907 cs, LOCK4res *, resp); 8908 return; 8909 } 8910 8911 /* hold off other access to open_owner while we tinker */ 8912 rfs4_sw_enter(&sp->rs_owner->ro_sw); 8913 8914 switch (rc = rfs4_check_stateid_seqid(sp, stateid)) { 8915 case NFS4_CHECK_STATEID_OLD: 8916 *cs->statusp = resp->status = NFS4ERR_OLD_STATEID; 8917 goto end; 8918 case NFS4_CHECK_STATEID_BAD: 8919 *cs->statusp = resp->status = NFS4ERR_BAD_STATEID; 8920 goto end; 8921 case NFS4_CHECK_STATEID_EXPIRED: 8922 *cs->statusp = resp->status = NFS4ERR_EXPIRED; 8923 goto end; 8924 case NFS4_CHECK_STATEID_UNCONFIRMED: 8925 *cs->statusp = resp->status = NFS4ERR_BAD_STATEID; 8926 goto end; 8927 case NFS4_CHECK_STATEID_CLOSED: 8928 *cs->statusp = resp->status = NFS4ERR_OLD_STATEID; 8929 goto end; 8930 case NFS4_CHECK_STATEID_OKAY: 8931 case NFS4_CHECK_STATEID_REPLAY: 8932 switch (rfs4_check_olo_seqid(olo->open_seqid, 8933 sp->rs_owner, resop)) { 8934 case NFS4_CHKSEQ_OKAY: 8935 if (rc == NFS4_CHECK_STATEID_OKAY) 8936 break; 8937 /* 8938 * This is replayed stateid; if seqid 8939 * matches next expected, then client 8940 * is using wrong seqid. 8941 */ 8942 /* FALLTHROUGH */ 8943 case NFS4_CHKSEQ_BAD: 8944 *cs->statusp = resp->status = NFS4ERR_BAD_SEQID; 8945 goto end; 8946 case NFS4_CHKSEQ_REPLAY: 8947 /* This is a duplicate LOCK request */ 8948 dup_lock = TRUE; 8949 8950 /* 8951 * For a duplicate we do not want to 8952 * create a new lockowner as it should 8953 * already exist. 8954 * Turn off the lockowner create flag. 8955 */ 8956 lcreate = FALSE; 8957 } 8958 break; 8959 } 8960 8961 lo = rfs4_findlockowner(&olo->lock_owner, &lcreate); 8962 if (lo == NULL) { 8963 NFS4_DEBUG(rfs4_debug, 8964 (CE_NOTE, "rfs4_op_lock: no lock owner")); 8965 *cs->statusp = resp->status = NFS4ERR_RESOURCE; 8966 goto end; 8967 } 8968 8969 lsp = rfs4_findlo_state_by_owner(lo, sp, &create); 8970 if (lsp == NULL) { 8971 rfs4_update_lease(sp->rs_owner->ro_client); 8972 /* 8973 * Only update theh open_seqid if this is not 8974 * a duplicate request 8975 */ 8976 if (dup_lock == FALSE) { 8977 rfs4_update_open_sequence(sp->rs_owner); 8978 } 8979 8980 NFS4_DEBUG(rfs4_debug, 8981 (CE_NOTE, "rfs4_op_lock: no state")); 8982 *cs->statusp = resp->status = NFS4ERR_SERVERFAULT; 8983 rfs4_update_open_resp(sp->rs_owner, resop, NULL); 8984 rfs4_lockowner_rele(lo); 8985 goto end; 8986 } 8987 8988 /* 8989 * This is the new_lock_owner branch and the client is 8990 * supposed to be associating a new lock_owner with 8991 * the open file at this point. If we find that a 8992 * lock_owner/state association already exists and a 8993 * successful LOCK request was returned to the client, 8994 * an error is returned to the client since this is 8995 * not appropriate. The client should be using the 8996 * existing lock_owner branch. 8997 */ 8998 if (dup_lock == FALSE && create == FALSE) { 8999 if (lsp->rls_lock_completed == TRUE) { 9000 *cs->statusp = 9001 resp->status = NFS4ERR_BAD_SEQID; 9002 rfs4_lockowner_rele(lo); 9003 goto end; 9004 } 9005 } 9006 9007 rfs4_update_lease(sp->rs_owner->ro_client); 9008 9009 /* 9010 * Only update theh open_seqid if this is not 9011 * a duplicate request 9012 */ 9013 if (dup_lock == FALSE) { 9014 rfs4_update_open_sequence(sp->rs_owner); 9015 } 9016 9017 /* 9018 * If this is a duplicate lock request, just copy the 9019 * previously saved reply and return. 9020 */ 9021 if (dup_lock == TRUE) { 9022 /* verify that lock_seqid's match */ 9023 if (lsp->rls_seqid != olo->lock_seqid) { 9024 NFS4_DEBUG(rfs4_debug, 9025 (CE_NOTE, "rfs4_op_lock: Dup-Lock seqid bad" 9026 "lsp->seqid=%d old->seqid=%d", 9027 lsp->rls_seqid, olo->lock_seqid)); 9028 *cs->statusp = resp->status = NFS4ERR_BAD_SEQID; 9029 } else { 9030 rfs4_copy_reply(resop, &lsp->rls_reply); 9031 /* 9032 * Make sure to copy the just 9033 * retrieved reply status into the 9034 * overall compound status 9035 */ 9036 *cs->statusp = resp->status; 9037 } 9038 rfs4_lockowner_rele(lo); 9039 goto end; 9040 } 9041 9042 rfs4_dbe_lock(lsp->rls_dbe); 9043 9044 /* Make sure to update the lock sequence id */ 9045 lsp->rls_seqid = olo->lock_seqid; 9046 9047 NFS4_DEBUG(rfs4_debug, 9048 (CE_NOTE, "Lock seqid established as %d", lsp->rls_seqid)); 9049 9050 /* 9051 * This is used to signify the newly created lockowner 9052 * stateid and its sequence number. The checks for 9053 * sequence number and increment don't occur on the 9054 * very first lock request for a lockowner. 9055 */ 9056 lsp->rls_skip_seqid_check = TRUE; 9057 9058 /* hold off other access to lsp while we tinker */ 9059 rfs4_sw_enter(&lsp->rls_sw); 9060 ls_sw_held = TRUE; 9061 9062 rfs4_dbe_unlock(lsp->rls_dbe); 9063 9064 rfs4_lockowner_rele(lo); 9065 } else { 9066 stateid = &args->locker.locker4_u.lock_owner.lock_stateid; 9067 /* get lsp and hold the lock on the underlying file struct */ 9068 if ((status = rfs4_get_lo_state(stateid, &lsp, TRUE)) 9069 != NFS4_OK) { 9070 *cs->statusp = resp->status = status; 9071 DTRACE_NFSV4_2(op__lock__done, struct compound_state *, 9072 cs, LOCK4res *, resp); 9073 return; 9074 } 9075 create = FALSE; /* We didn't create lsp */ 9076 9077 /* Ensure specified filehandle matches */ 9078 if (cs->vp != lsp->rls_state->rs_finfo->rf_vp) { 9079 rfs4_lo_state_rele(lsp, TRUE); 9080 *cs->statusp = resp->status = NFS4ERR_BAD_STATEID; 9081 DTRACE_NFSV4_2(op__lock__done, struct compound_state *, 9082 cs, LOCK4res *, resp); 9083 return; 9084 } 9085 9086 /* hold off other access to lsp while we tinker */ 9087 rfs4_sw_enter(&lsp->rls_sw); 9088 ls_sw_held = TRUE; 9089 9090 switch (rfs4_check_lo_stateid_seqid(lsp, stateid)) { 9091 /* 9092 * The stateid looks like it was okay (expected to be 9093 * the next one) 9094 */ 9095 case NFS4_CHECK_STATEID_OKAY: 9096 /* 9097 * The sequence id is now checked. Determine 9098 * if this is a replay or if it is in the 9099 * expected (next) sequence. In the case of a 9100 * replay, there are two replay conditions 9101 * that may occur. The first is the normal 9102 * condition where a LOCK is done with a 9103 * NFS4_OK response and the stateid is 9104 * updated. That case is handled below when 9105 * the stateid is identified as a REPLAY. The 9106 * second is the case where an error is 9107 * returned, like NFS4ERR_DENIED, and the 9108 * sequence number is updated but the stateid 9109 * is not updated. This second case is dealt 9110 * with here. So it may seem odd that the 9111 * stateid is okay but the sequence id is a 9112 * replay but it is okay. 9113 */ 9114 switch (rfs4_check_lock_seqid( 9115 args->locker.locker4_u.lock_owner.lock_seqid, 9116 lsp, resop)) { 9117 case NFS4_CHKSEQ_REPLAY: 9118 if (resp->status != NFS4_OK) { 9119 /* 9120 * Here is our replay and need 9121 * to verify that the last 9122 * response was an error. 9123 */ 9124 *cs->statusp = resp->status; 9125 goto end; 9126 } 9127 /* 9128 * This is done since the sequence id 9129 * looked like a replay but it didn't 9130 * pass our check so a BAD_SEQID is 9131 * returned as a result. 9132 */ 9133 /*FALLTHROUGH*/ 9134 case NFS4_CHKSEQ_BAD: 9135 *cs->statusp = resp->status = NFS4ERR_BAD_SEQID; 9136 goto end; 9137 case NFS4_CHKSEQ_OKAY: 9138 /* Everything looks okay move ahead */ 9139 break; 9140 } 9141 break; 9142 case NFS4_CHECK_STATEID_OLD: 9143 *cs->statusp = resp->status = NFS4ERR_OLD_STATEID; 9144 goto end; 9145 case NFS4_CHECK_STATEID_BAD: 9146 *cs->statusp = resp->status = NFS4ERR_BAD_STATEID; 9147 goto end; 9148 case NFS4_CHECK_STATEID_EXPIRED: 9149 *cs->statusp = resp->status = NFS4ERR_EXPIRED; 9150 goto end; 9151 case NFS4_CHECK_STATEID_CLOSED: 9152 *cs->statusp = resp->status = NFS4ERR_OLD_STATEID; 9153 goto end; 9154 case NFS4_CHECK_STATEID_REPLAY: 9155 switch (rfs4_check_lock_seqid( 9156 args->locker.locker4_u.lock_owner.lock_seqid, 9157 lsp, resop)) { 9158 case NFS4_CHKSEQ_OKAY: 9159 /* 9160 * This is a replayed stateid; if 9161 * seqid matches the next expected, 9162 * then client is using wrong seqid. 9163 */ 9164 case NFS4_CHKSEQ_BAD: 9165 *cs->statusp = resp->status = NFS4ERR_BAD_SEQID; 9166 goto end; 9167 case NFS4_CHKSEQ_REPLAY: 9168 rfs4_update_lease(lsp->rls_locker->rl_client); 9169 *cs->statusp = status = resp->status; 9170 goto end; 9171 } 9172 break; 9173 default: 9174 ASSERT(FALSE); 9175 break; 9176 } 9177 9178 rfs4_update_lock_sequence(lsp); 9179 rfs4_update_lease(lsp->rls_locker->rl_client); 9180 } 9181 9182 /* 9183 * NFS4 only allows locking on regular files, so 9184 * verify type of object. 9185 */ 9186 if (cs->vp->v_type != VREG) { 9187 if (cs->vp->v_type == VDIR) 9188 status = NFS4ERR_ISDIR; 9189 else 9190 status = NFS4ERR_INVAL; 9191 goto out; 9192 } 9193 9194 cp = lsp->rls_state->rs_owner->ro_client; 9195 9196 if (rfs4_clnt_in_grace(cp) && !args->reclaim) { 9197 status = NFS4ERR_GRACE; 9198 goto out; 9199 } 9200 9201 if (rfs4_clnt_in_grace(cp) && args->reclaim && !cp->rc_can_reclaim) { 9202 status = NFS4ERR_NO_GRACE; 9203 goto out; 9204 } 9205 9206 if (!rfs4_clnt_in_grace(cp) && args->reclaim) { 9207 status = NFS4ERR_NO_GRACE; 9208 goto out; 9209 } 9210 9211 if (lsp->rls_state->rs_finfo->rf_dinfo.rd_dtype == OPEN_DELEGATE_WRITE) 9212 cs->deleg = TRUE; 9213 9214 status = rfs4_do_lock(lsp, args->locktype, 9215 args->offset, args->length, cs->cr, resop); 9216 9217 out: 9218 lsp->rls_skip_seqid_check = FALSE; 9219 9220 *cs->statusp = resp->status = status; 9221 9222 if (status == NFS4_OK) { 9223 resp->LOCK4res_u.lock_stateid = lsp->rls_lockid.stateid; 9224 lsp->rls_lock_completed = TRUE; 9225 } 9226 /* 9227 * Only update the "OPEN" response here if this was a new 9228 * lock_owner 9229 */ 9230 if (sp) 9231 rfs4_update_open_resp(sp->rs_owner, resop, NULL); 9232 9233 rfs4_update_lock_resp(lsp, resop); 9234 9235 end: 9236 if (lsp) { 9237 if (ls_sw_held) 9238 rfs4_sw_exit(&lsp->rls_sw); 9239 /* 9240 * If an sp obtained, then the lsp does not represent 9241 * a lock on the file struct. 9242 */ 9243 if (sp != NULL) 9244 rfs4_lo_state_rele(lsp, FALSE); 9245 else 9246 rfs4_lo_state_rele(lsp, TRUE); 9247 } 9248 if (sp) { 9249 rfs4_sw_exit(&sp->rs_owner->ro_sw); 9250 rfs4_state_rele(sp); 9251 } 9252 9253 DTRACE_NFSV4_2(op__lock__done, struct compound_state *, cs, 9254 LOCK4res *, resp); 9255 } 9256 9257 /* free function for LOCK/LOCKT */ 9258 static void 9259 lock_denied_free(nfs_resop4 *resop) 9260 { 9261 LOCK4denied *dp = NULL; 9262 9263 switch (resop->resop) { 9264 case OP_LOCK: 9265 if (resop->nfs_resop4_u.oplock.status == NFS4ERR_DENIED) 9266 dp = &resop->nfs_resop4_u.oplock.LOCK4res_u.denied; 9267 break; 9268 case OP_LOCKT: 9269 if (resop->nfs_resop4_u.oplockt.status == NFS4ERR_DENIED) 9270 dp = &resop->nfs_resop4_u.oplockt.denied; 9271 break; 9272 default: 9273 break; 9274 } 9275 9276 if (dp) 9277 kmem_free(dp->owner.owner_val, dp->owner.owner_len); 9278 } 9279 9280 /*ARGSUSED*/ 9281 void 9282 rfs4_op_locku(nfs_argop4 *argop, nfs_resop4 *resop, 9283 struct svc_req *req, struct compound_state *cs) 9284 { 9285 LOCKU4args *args = &argop->nfs_argop4_u.oplocku; 9286 LOCKU4res *resp = &resop->nfs_resop4_u.oplocku; 9287 nfsstat4 status; 9288 stateid4 *stateid = &args->lock_stateid; 9289 rfs4_lo_state_t *lsp; 9290 9291 DTRACE_NFSV4_2(op__locku__start, struct compound_state *, cs, 9292 LOCKU4args *, args); 9293 9294 if (cs->vp == NULL) { 9295 *cs->statusp = resp->status = NFS4ERR_NOFILEHANDLE; 9296 DTRACE_NFSV4_2(op__locku__done, struct compound_state *, cs, 9297 LOCKU4res *, resp); 9298 return; 9299 } 9300 9301 if ((status = rfs4_get_lo_state(stateid, &lsp, TRUE)) != NFS4_OK) { 9302 *cs->statusp = resp->status = status; 9303 DTRACE_NFSV4_2(op__locku__done, struct compound_state *, cs, 9304 LOCKU4res *, resp); 9305 return; 9306 } 9307 9308 /* Ensure specified filehandle matches */ 9309 if (cs->vp != lsp->rls_state->rs_finfo->rf_vp) { 9310 rfs4_lo_state_rele(lsp, TRUE); 9311 *cs->statusp = resp->status = NFS4ERR_BAD_STATEID; 9312 DTRACE_NFSV4_2(op__locku__done, struct compound_state *, cs, 9313 LOCKU4res *, resp); 9314 return; 9315 } 9316 9317 /* hold off other access to lsp while we tinker */ 9318 rfs4_sw_enter(&lsp->rls_sw); 9319 9320 switch (rfs4_check_lo_stateid_seqid(lsp, stateid)) { 9321 case NFS4_CHECK_STATEID_OKAY: 9322 if (rfs4_check_lock_seqid(args->seqid, lsp, resop) 9323 != NFS4_CHKSEQ_OKAY) { 9324 *cs->statusp = resp->status = NFS4ERR_BAD_SEQID; 9325 goto end; 9326 } 9327 break; 9328 case NFS4_CHECK_STATEID_OLD: 9329 *cs->statusp = resp->status = NFS4ERR_OLD_STATEID; 9330 goto end; 9331 case NFS4_CHECK_STATEID_BAD: 9332 *cs->statusp = resp->status = NFS4ERR_BAD_STATEID; 9333 goto end; 9334 case NFS4_CHECK_STATEID_EXPIRED: 9335 *cs->statusp = resp->status = NFS4ERR_EXPIRED; 9336 goto end; 9337 case NFS4_CHECK_STATEID_CLOSED: 9338 *cs->statusp = resp->status = NFS4ERR_OLD_STATEID; 9339 goto end; 9340 case NFS4_CHECK_STATEID_REPLAY: 9341 switch (rfs4_check_lock_seqid(args->seqid, lsp, resop)) { 9342 case NFS4_CHKSEQ_OKAY: 9343 /* 9344 * This is a replayed stateid; if 9345 * seqid matches the next expected, 9346 * then client is using wrong seqid. 9347 */ 9348 case NFS4_CHKSEQ_BAD: 9349 *cs->statusp = resp->status = NFS4ERR_BAD_SEQID; 9350 goto end; 9351 case NFS4_CHKSEQ_REPLAY: 9352 rfs4_update_lease(lsp->rls_locker->rl_client); 9353 *cs->statusp = status = resp->status; 9354 goto end; 9355 } 9356 break; 9357 default: 9358 ASSERT(FALSE); 9359 break; 9360 } 9361 9362 rfs4_update_lock_sequence(lsp); 9363 rfs4_update_lease(lsp->rls_locker->rl_client); 9364 9365 /* 9366 * NFS4 only allows locking on regular files, so 9367 * verify type of object. 9368 */ 9369 if (cs->vp->v_type != VREG) { 9370 if (cs->vp->v_type == VDIR) 9371 status = NFS4ERR_ISDIR; 9372 else 9373 status = NFS4ERR_INVAL; 9374 goto out; 9375 } 9376 9377 if (rfs4_clnt_in_grace(lsp->rls_state->rs_owner->ro_client)) { 9378 status = NFS4ERR_GRACE; 9379 goto out; 9380 } 9381 9382 status = rfs4_do_lock(lsp, args->locktype, 9383 args->offset, args->length, cs->cr, resop); 9384 9385 out: 9386 *cs->statusp = resp->status = status; 9387 9388 if (status == NFS4_OK) 9389 resp->lock_stateid = lsp->rls_lockid.stateid; 9390 9391 rfs4_update_lock_resp(lsp, resop); 9392 9393 end: 9394 rfs4_sw_exit(&lsp->rls_sw); 9395 rfs4_lo_state_rele(lsp, TRUE); 9396 9397 DTRACE_NFSV4_2(op__locku__done, struct compound_state *, cs, 9398 LOCKU4res *, resp); 9399 } 9400 9401 /* 9402 * LOCKT is a best effort routine, the client can not be guaranteed that 9403 * the status return is still in effect by the time the reply is received. 9404 * They are numerous race conditions in this routine, but we are not required 9405 * and can not be accurate. 9406 */ 9407 /*ARGSUSED*/ 9408 void 9409 rfs4_op_lockt(nfs_argop4 *argop, nfs_resop4 *resop, 9410 struct svc_req *req, struct compound_state *cs) 9411 { 9412 LOCKT4args *args = &argop->nfs_argop4_u.oplockt; 9413 LOCKT4res *resp = &resop->nfs_resop4_u.oplockt; 9414 rfs4_lockowner_t *lo; 9415 rfs4_client_t *cp; 9416 bool_t create = FALSE; 9417 struct flock64 flk; 9418 int error; 9419 int flag = FREAD | FWRITE; 9420 int ltype; 9421 length4 posix_length; 9422 sysid_t sysid; 9423 pid_t pid; 9424 9425 DTRACE_NFSV4_2(op__lockt__start, struct compound_state *, cs, 9426 LOCKT4args *, args); 9427 9428 if (cs->vp == NULL) { 9429 *cs->statusp = resp->status = NFS4ERR_NOFILEHANDLE; 9430 goto out; 9431 } 9432 9433 /* 9434 * NFS4 only allows locking on regular files, so 9435 * verify type of object. 9436 */ 9437 if (cs->vp->v_type != VREG) { 9438 if (cs->vp->v_type == VDIR) 9439 *cs->statusp = resp->status = NFS4ERR_ISDIR; 9440 else 9441 *cs->statusp = resp->status = NFS4ERR_INVAL; 9442 goto out; 9443 } 9444 9445 /* 9446 * Check out the clientid to ensure the server knows about it 9447 * so that we correctly inform the client of a server reboot. 9448 */ 9449 if ((cp = rfs4_findclient_by_id(args->owner.clientid, FALSE)) 9450 == NULL) { 9451 *cs->statusp = resp->status = 9452 rfs4_check_clientid(&args->owner.clientid, 0); 9453 goto out; 9454 } 9455 if (rfs4_lease_expired(cp)) { 9456 rfs4_client_close(cp); 9457 /* 9458 * Protocol doesn't allow returning NFS4ERR_STALE as 9459 * other operations do on this check so STALE_CLIENTID 9460 * is returned instead 9461 */ 9462 *cs->statusp = resp->status = NFS4ERR_STALE_CLIENTID; 9463 goto out; 9464 } 9465 9466 if (rfs4_clnt_in_grace(cp) && !(cp->rc_can_reclaim)) { 9467 *cs->statusp = resp->status = NFS4ERR_GRACE; 9468 rfs4_client_rele(cp); 9469 goto out; 9470 } 9471 rfs4_client_rele(cp); 9472 9473 resp->status = NFS4_OK; 9474 9475 switch (args->locktype) { 9476 case READ_LT: 9477 case READW_LT: 9478 ltype = F_RDLCK; 9479 break; 9480 case WRITE_LT: 9481 case WRITEW_LT: 9482 ltype = F_WRLCK; 9483 break; 9484 } 9485 9486 posix_length = args->length; 9487 /* Check for zero length. To lock to end of file use all ones for V4 */ 9488 if (posix_length == 0) { 9489 *cs->statusp = resp->status = NFS4ERR_INVAL; 9490 goto out; 9491 } else if (posix_length == (length4)(~0)) { 9492 posix_length = 0; /* Posix to end of file */ 9493 } 9494 9495 /* Find or create a lockowner */ 9496 lo = rfs4_findlockowner(&args->owner, &create); 9497 9498 if (lo) { 9499 pid = lo->rl_pid; 9500 if ((resp->status = 9501 rfs4_client_sysid(lo->rl_client, &sysid)) != NFS4_OK) 9502 goto err; 9503 } else { 9504 pid = 0; 9505 sysid = lockt_sysid; 9506 } 9507 retry: 9508 flk.l_type = ltype; 9509 flk.l_whence = 0; /* SEEK_SET */ 9510 flk.l_start = args->offset; 9511 flk.l_len = posix_length; 9512 flk.l_sysid = sysid; 9513 flk.l_pid = pid; 9514 flag |= F_REMOTELOCK; 9515 9516 LOCK_PRINT(rfs4_debug, "rfs4_op_lockt", F_GETLK, &flk); 9517 9518 /* Note that length4 is uint64_t but l_len and l_start are off64_t */ 9519 if (flk.l_len < 0 || flk.l_start < 0) { 9520 resp->status = NFS4ERR_INVAL; 9521 goto err; 9522 } 9523 error = VOP_FRLOCK(cs->vp, F_GETLK, &flk, flag, (u_offset_t)0, 9524 NULL, cs->cr, NULL); 9525 9526 /* 9527 * N.B. We map error values to nfsv4 errors. This is differrent 9528 * than puterrno4 routine. 9529 */ 9530 switch (error) { 9531 case 0: 9532 if (flk.l_type == F_UNLCK) 9533 resp->status = NFS4_OK; 9534 else { 9535 if (lock_denied(&resp->denied, &flk) == NFS4ERR_EXPIRED) 9536 goto retry; 9537 resp->status = NFS4ERR_DENIED; 9538 } 9539 break; 9540 case EOVERFLOW: 9541 resp->status = NFS4ERR_INVAL; 9542 break; 9543 case EINVAL: 9544 resp->status = NFS4ERR_NOTSUPP; 9545 break; 9546 default: 9547 cmn_err(CE_WARN, "rfs4_op_lockt: unexpected errno (%d)", 9548 error); 9549 resp->status = NFS4ERR_SERVERFAULT; 9550 break; 9551 } 9552 9553 err: 9554 if (lo) 9555 rfs4_lockowner_rele(lo); 9556 *cs->statusp = resp->status; 9557 out: 9558 DTRACE_NFSV4_2(op__lockt__done, struct compound_state *, cs, 9559 LOCKT4res *, resp); 9560 } 9561 9562 int 9563 rfs4_share(rfs4_state_t *sp, uint32_t access, uint32_t deny) 9564 { 9565 int err; 9566 int cmd; 9567 vnode_t *vp; 9568 struct shrlock shr; 9569 struct shr_locowner shr_loco; 9570 int fflags = 0; 9571 9572 ASSERT(rfs4_dbe_islocked(sp->rs_dbe)); 9573 ASSERT(sp->rs_owner->ro_client->rc_sysidt != LM_NOSYSID); 9574 9575 if (sp->rs_closed) 9576 return (NFS4ERR_OLD_STATEID); 9577 9578 vp = sp->rs_finfo->rf_vp; 9579 ASSERT(vp); 9580 9581 shr.s_access = shr.s_deny = 0; 9582 9583 if (access & OPEN4_SHARE_ACCESS_READ) { 9584 fflags |= FREAD; 9585 shr.s_access |= F_RDACC; 9586 } 9587 if (access & OPEN4_SHARE_ACCESS_WRITE) { 9588 fflags |= FWRITE; 9589 shr.s_access |= F_WRACC; 9590 } 9591 ASSERT(shr.s_access); 9592 9593 if (deny & OPEN4_SHARE_DENY_READ) 9594 shr.s_deny |= F_RDDNY; 9595 if (deny & OPEN4_SHARE_DENY_WRITE) 9596 shr.s_deny |= F_WRDNY; 9597 9598 shr.s_pid = rfs4_dbe_getid(sp->rs_owner->ro_dbe); 9599 shr.s_sysid = sp->rs_owner->ro_client->rc_sysidt; 9600 shr_loco.sl_pid = shr.s_pid; 9601 shr_loco.sl_id = shr.s_sysid; 9602 shr.s_owner = (caddr_t)&shr_loco; 9603 shr.s_own_len = sizeof (shr_loco); 9604 9605 cmd = nbl_need_check(vp) ? F_SHARE_NBMAND : F_SHARE; 9606 9607 err = VOP_SHRLOCK(vp, cmd, &shr, fflags, CRED(), NULL); 9608 if (err != 0) { 9609 if (err == EAGAIN) 9610 err = NFS4ERR_SHARE_DENIED; 9611 else 9612 err = puterrno4(err); 9613 return (err); 9614 } 9615 9616 sp->rs_share_access |= access; 9617 sp->rs_share_deny |= deny; 9618 9619 return (0); 9620 } 9621 9622 int 9623 rfs4_unshare(rfs4_state_t *sp) 9624 { 9625 int err; 9626 struct shrlock shr; 9627 struct shr_locowner shr_loco; 9628 9629 ASSERT(rfs4_dbe_islocked(sp->rs_dbe)); 9630 9631 if (sp->rs_closed || sp->rs_share_access == 0) 9632 return (0); 9633 9634 ASSERT(sp->rs_owner->ro_client->rc_sysidt != LM_NOSYSID); 9635 ASSERT(sp->rs_finfo->rf_vp); 9636 9637 shr.s_access = shr.s_deny = 0; 9638 shr.s_pid = rfs4_dbe_getid(sp->rs_owner->ro_dbe); 9639 shr.s_sysid = sp->rs_owner->ro_client->rc_sysidt; 9640 shr_loco.sl_pid = shr.s_pid; 9641 shr_loco.sl_id = shr.s_sysid; 9642 shr.s_owner = (caddr_t)&shr_loco; 9643 shr.s_own_len = sizeof (shr_loco); 9644 9645 err = VOP_SHRLOCK(sp->rs_finfo->rf_vp, F_UNSHARE, &shr, 0, CRED(), 9646 NULL); 9647 if (err != 0) { 9648 err = puterrno4(err); 9649 return (err); 9650 } 9651 9652 sp->rs_share_access = 0; 9653 sp->rs_share_deny = 0; 9654 9655 return (0); 9656 9657 } 9658 9659 static int 9660 rdma_setup_read_data4(READ4args *args, READ4res *rok) 9661 { 9662 struct clist *wcl; 9663 count4 count = rok->data_len; 9664 int wlist_len; 9665 9666 wcl = args->wlist; 9667 if (rdma_setup_read_chunks(wcl, count, &wlist_len) == FALSE) { 9668 return (FALSE); 9669 } 9670 wcl = args->wlist; 9671 rok->wlist_len = wlist_len; 9672 rok->wlist = wcl; 9673 return (TRUE); 9674 } 9675 9676 /* tunable to disable server referrals */ 9677 int rfs4_no_referrals = 0; 9678 9679 /* 9680 * Find an NFS record in reparse point data. 9681 * Returns 0 for success and <0 or an errno value on failure. 9682 */ 9683 int 9684 vn_find_nfs_record(vnode_t *vp, nvlist_t **nvlp, char **svcp, char **datap) 9685 { 9686 int err; 9687 char *stype, *val; 9688 nvlist_t *nvl; 9689 nvpair_t *curr; 9690 9691 if ((nvl = reparse_init()) == NULL) 9692 return (-1); 9693 9694 if ((err = reparse_vnode_parse(vp, nvl)) != 0) { 9695 reparse_free(nvl); 9696 return (err); 9697 } 9698 9699 curr = NULL; 9700 while ((curr = nvlist_next_nvpair(nvl, curr)) != NULL) { 9701 if ((stype = nvpair_name(curr)) == NULL) { 9702 reparse_free(nvl); 9703 return (-2); 9704 } 9705 if (strncasecmp(stype, "NFS", 3) == 0) 9706 break; 9707 } 9708 9709 if ((curr == NULL) || 9710 (nvpair_value_string(curr, &val))) { 9711 reparse_free(nvl); 9712 return (-3); 9713 } 9714 *nvlp = nvl; 9715 *svcp = stype; 9716 *datap = val; 9717 return (0); 9718 } 9719 9720 int 9721 vn_is_nfs_reparse(vnode_t *vp, cred_t *cr) 9722 { 9723 nvlist_t *nvl; 9724 char *s, *d; 9725 9726 if (rfs4_no_referrals != 0) 9727 return (B_FALSE); 9728 9729 if (vn_is_reparse(vp, cr, NULL) == B_FALSE) 9730 return (B_FALSE); 9731 9732 if (vn_find_nfs_record(vp, &nvl, &s, &d) != 0) 9733 return (B_FALSE); 9734 9735 reparse_free(nvl); 9736 9737 return (B_TRUE); 9738 } 9739 9740 /* 9741 * There is a user-level copy of this routine in ref_subr.c. 9742 * Changes should be kept in sync. 9743 */ 9744 static int 9745 nfs4_create_components(char *path, component4 *comp4) 9746 { 9747 int slen, plen, ncomp; 9748 char *ori_path, *nxtc, buf[MAXNAMELEN]; 9749 9750 if (path == NULL) 9751 return (0); 9752 9753 plen = strlen(path) + 1; /* include the terminator */ 9754 ori_path = path; 9755 ncomp = 0; 9756 9757 /* count number of components in the path */ 9758 for (nxtc = path; nxtc < ori_path + plen; nxtc++) { 9759 if (*nxtc == '/' || *nxtc == '\0' || *nxtc == '\n') { 9760 if ((slen = nxtc - path) == 0) { 9761 path = nxtc + 1; 9762 continue; 9763 } 9764 9765 if (comp4 != NULL) { 9766 bcopy(path, buf, slen); 9767 buf[slen] = '\0'; 9768 (void) str_to_utf8(buf, &comp4[ncomp]); 9769 } 9770 9771 ncomp++; /* 1 valid component */ 9772 path = nxtc + 1; 9773 } 9774 if (*nxtc == '\0' || *nxtc == '\n') 9775 break; 9776 } 9777 9778 return (ncomp); 9779 } 9780 9781 /* 9782 * There is a user-level copy of this routine in ref_subr.c. 9783 * Changes should be kept in sync. 9784 */ 9785 static int 9786 make_pathname4(char *path, pathname4 *pathname) 9787 { 9788 int ncomp; 9789 component4 *comp4; 9790 9791 if (pathname == NULL) 9792 return (0); 9793 9794 if (path == NULL) { 9795 pathname->pathname4_val = NULL; 9796 pathname->pathname4_len = 0; 9797 return (0); 9798 } 9799 9800 /* count number of components to alloc buffer */ 9801 if ((ncomp = nfs4_create_components(path, NULL)) == 0) { 9802 pathname->pathname4_val = NULL; 9803 pathname->pathname4_len = 0; 9804 return (0); 9805 } 9806 comp4 = kmem_zalloc(ncomp * sizeof (component4), KM_SLEEP); 9807 9808 /* copy components into allocated buffer */ 9809 ncomp = nfs4_create_components(path, comp4); 9810 9811 pathname->pathname4_val = comp4; 9812 pathname->pathname4_len = ncomp; 9813 9814 return (ncomp); 9815 } 9816 9817 #define xdr_fs_locations4 xdr_fattr4_fs_locations 9818 9819 fs_locations4 * 9820 fetch_referral(vnode_t *vp, cred_t *cr) 9821 { 9822 nvlist_t *nvl; 9823 char *stype, *sdata; 9824 fs_locations4 *result; 9825 char buf[1024]; 9826 size_t bufsize; 9827 XDR xdr; 9828 int err; 9829 9830 /* 9831 * Check attrs to ensure it's a reparse point 9832 */ 9833 if (vn_is_reparse(vp, cr, NULL) == B_FALSE) 9834 return (NULL); 9835 9836 /* 9837 * Look for an NFS record and get the type and data 9838 */ 9839 if (vn_find_nfs_record(vp, &nvl, &stype, &sdata) != 0) 9840 return (NULL); 9841 9842 /* 9843 * With the type and data, upcall to get the referral 9844 */ 9845 bufsize = sizeof (buf); 9846 bzero(buf, sizeof (buf)); 9847 err = reparse_kderef((const char *)stype, (const char *)sdata, 9848 buf, &bufsize); 9849 reparse_free(nvl); 9850 9851 DTRACE_PROBE4(nfs4serv__func__referral__upcall, 9852 char *, stype, char *, sdata, char *, buf, int, err); 9853 if (err) { 9854 cmn_err(CE_NOTE, 9855 "reparsed daemon not running: unable to get referral (%d)", 9856 err); 9857 return (NULL); 9858 } 9859 9860 /* 9861 * We get an XDR'ed record back from the kderef call 9862 */ 9863 xdrmem_create(&xdr, buf, bufsize, XDR_DECODE); 9864 result = kmem_alloc(sizeof (fs_locations4), KM_SLEEP); 9865 err = xdr_fs_locations4(&xdr, result); 9866 XDR_DESTROY(&xdr); 9867 if (err != TRUE) { 9868 DTRACE_PROBE1(nfs4serv__func__referral__upcall__xdrfail, 9869 int, err); 9870 return (NULL); 9871 } 9872 9873 /* 9874 * Look at path to recover fs_root, ignoring the leading '/' 9875 */ 9876 (void) make_pathname4(vp->v_path, &result->fs_root); 9877 9878 return (result); 9879 } 9880 9881 char * 9882 build_symlink(vnode_t *vp, cred_t *cr, size_t *strsz) 9883 { 9884 fs_locations4 *fsl; 9885 fs_location4 *fs; 9886 char *server, *path, *symbuf; 9887 static char *prefix = "/net/"; 9888 int i, size, npaths; 9889 uint_t len; 9890 9891 /* Get the referral */ 9892 if ((fsl = fetch_referral(vp, cr)) == NULL) 9893 return (NULL); 9894 9895 /* Deal with only the first location and first server */ 9896 fs = &fsl->locations_val[0]; 9897 server = utf8_to_str(&fs->server_val[0], &len, NULL); 9898 if (server == NULL) { 9899 rfs4_free_fs_locations4(fsl); 9900 kmem_free(fsl, sizeof (fs_locations4)); 9901 return (NULL); 9902 } 9903 9904 /* Figure out size for "/net/" + host + /path/path/path + NULL */ 9905 size = strlen(prefix) + len; 9906 for (i = 0; i < fs->rootpath.pathname4_len; i++) 9907 size += fs->rootpath.pathname4_val[i].utf8string_len + 1; 9908 9909 /* Allocate the symlink buffer and fill it */ 9910 symbuf = kmem_zalloc(size, KM_SLEEP); 9911 (void) strcat(symbuf, prefix); 9912 (void) strcat(symbuf, server); 9913 kmem_free(server, len); 9914 9915 npaths = 0; 9916 for (i = 0; i < fs->rootpath.pathname4_len; i++) { 9917 path = utf8_to_str(&fs->rootpath.pathname4_val[i], &len, NULL); 9918 if (path == NULL) 9919 continue; 9920 (void) strcat(symbuf, "/"); 9921 (void) strcat(symbuf, path); 9922 npaths++; 9923 kmem_free(path, len); 9924 } 9925 9926 rfs4_free_fs_locations4(fsl); 9927 kmem_free(fsl, sizeof (fs_locations4)); 9928 9929 if (strsz != NULL) 9930 *strsz = size; 9931 return (symbuf); 9932 } 9933 9934 /* 9935 * Check to see if we have a downrev Solaris client, so that we 9936 * can send it a symlink instead of a referral. 9937 */ 9938 int 9939 client_is_downrev(struct svc_req *req) 9940 { 9941 struct sockaddr *ca; 9942 rfs4_clntip_t *ci; 9943 bool_t create = FALSE; 9944 int is_downrev; 9945 9946 ca = (struct sockaddr *)svc_getrpccaller(req->rq_xprt)->buf; 9947 ASSERT(ca); 9948 ci = rfs4_find_clntip(ca, &create); 9949 if (ci == NULL) 9950 return (0); 9951 is_downrev = ci->ri_no_referrals; 9952 rfs4_dbe_rele(ci->ri_dbe); 9953 return (is_downrev); 9954 } 9955 9956 /* 9957 * Do the main work of handling HA-NFSv4 Resource Group failover on 9958 * Sun Cluster. 9959 * We need to detect whether any RG admin paths have been added or removed, 9960 * and adjust resources accordingly. 9961 * Currently we're using a very inefficient algorithm, ~ 2 * O(n**2). In 9962 * order to scale, the list and array of paths need to be held in more 9963 * suitable data structures. 9964 */ 9965 static void 9966 hanfsv4_failover(nfs4_srv_t *nsrv4) 9967 { 9968 int i, start_grace, numadded_paths = 0; 9969 char **added_paths = NULL; 9970 rfs4_dss_path_t *dss_path; 9971 9972 /* 9973 * Note: currently, dss_pathlist cannot be NULL, since 9974 * it will always include an entry for NFS4_DSS_VAR_DIR. If we 9975 * make the latter dynamically specified too, the following will 9976 * need to be adjusted. 9977 */ 9978 9979 /* 9980 * First, look for removed paths: RGs that have been failed-over 9981 * away from this node. 9982 * Walk the "currently-serving" dss_pathlist and, for each 9983 * path, check if it is on the "passed-in" rfs4_dss_newpaths array 9984 * from nfsd. If not, that RG path has been removed. 9985 * 9986 * Note that nfsd has sorted rfs4_dss_newpaths for us, and removed 9987 * any duplicates. 9988 */ 9989 dss_path = nsrv4->dss_pathlist; 9990 do { 9991 int found = 0; 9992 char *path = dss_path->path; 9993 9994 /* used only for non-HA so may not be removed */ 9995 if (strcmp(path, NFS4_DSS_VAR_DIR) == 0) { 9996 dss_path = dss_path->next; 9997 continue; 9998 } 9999 10000 for (i = 0; i < rfs4_dss_numnewpaths; i++) { 10001 int cmpret; 10002 char *newpath = rfs4_dss_newpaths[i]; 10003 10004 /* 10005 * Since nfsd has sorted rfs4_dss_newpaths for us, 10006 * once the return from strcmp is negative we know 10007 * we've passed the point where "path" should be, 10008 * and can stop searching: "path" has been removed. 10009 */ 10010 cmpret = strcmp(path, newpath); 10011 if (cmpret < 0) 10012 break; 10013 if (cmpret == 0) { 10014 found = 1; 10015 break; 10016 } 10017 } 10018 10019 if (found == 0) { 10020 unsigned index = dss_path->index; 10021 rfs4_servinst_t *sip = dss_path->sip; 10022 rfs4_dss_path_t *path_next = dss_path->next; 10023 10024 /* 10025 * This path has been removed. 10026 * We must clear out the servinst reference to 10027 * it, since it's now owned by another 10028 * node: we should not attempt to touch it. 10029 */ 10030 ASSERT(dss_path == sip->dss_paths[index]); 10031 sip->dss_paths[index] = NULL; 10032 10033 /* remove from "currently-serving" list, and destroy */ 10034 remque(dss_path); 10035 /* allow for NUL */ 10036 kmem_free(dss_path->path, strlen(dss_path->path) + 1); 10037 kmem_free(dss_path, sizeof (rfs4_dss_path_t)); 10038 10039 dss_path = path_next; 10040 } else { 10041 /* path was found; not removed */ 10042 dss_path = dss_path->next; 10043 } 10044 } while (dss_path != nsrv4->dss_pathlist); 10045 10046 /* 10047 * Now, look for added paths: RGs that have been failed-over 10048 * to this node. 10049 * Walk the "passed-in" rfs4_dss_newpaths array from nfsd and, 10050 * for each path, check if it is on the "currently-serving" 10051 * dss_pathlist. If not, that RG path has been added. 10052 * 10053 * Note: we don't do duplicate detection here; nfsd does that for us. 10054 * 10055 * Note: numadded_paths <= rfs4_dss_numnewpaths, which gives us 10056 * an upper bound for the size needed for added_paths[numadded_paths]. 10057 */ 10058 10059 /* probably more space than we need, but guaranteed to be enough */ 10060 if (rfs4_dss_numnewpaths > 0) { 10061 size_t sz = rfs4_dss_numnewpaths * sizeof (char *); 10062 added_paths = kmem_zalloc(sz, KM_SLEEP); 10063 } 10064 10065 /* walk the "passed-in" rfs4_dss_newpaths array from nfsd */ 10066 for (i = 0; i < rfs4_dss_numnewpaths; i++) { 10067 int found = 0; 10068 char *newpath = rfs4_dss_newpaths[i]; 10069 10070 dss_path = nsrv4->dss_pathlist; 10071 do { 10072 char *path = dss_path->path; 10073 10074 /* used only for non-HA */ 10075 if (strcmp(path, NFS4_DSS_VAR_DIR) == 0) { 10076 dss_path = dss_path->next; 10077 continue; 10078 } 10079 10080 if (strncmp(path, newpath, strlen(path)) == 0) { 10081 found = 1; 10082 break; 10083 } 10084 10085 dss_path = dss_path->next; 10086 } while (dss_path != nsrv4->dss_pathlist); 10087 10088 if (found == 0) { 10089 added_paths[numadded_paths] = newpath; 10090 numadded_paths++; 10091 } 10092 } 10093 10094 /* did we find any added paths? */ 10095 if (numadded_paths > 0) { 10096 10097 /* create a new server instance, and start its grace period */ 10098 start_grace = 1; 10099 /* CSTYLED */ 10100 rfs4_servinst_create(nsrv4, start_grace, numadded_paths, added_paths); 10101 10102 /* read in the stable storage state from these paths */ 10103 rfs4_dss_readstate(nsrv4, numadded_paths, added_paths); 10104 10105 /* 10106 * Multiple failovers during a grace period will cause 10107 * clients of the same resource group to be partitioned 10108 * into different server instances, with different 10109 * grace periods. Since clients of the same resource 10110 * group must be subject to the same grace period, 10111 * we need to reset all currently active grace periods. 10112 */ 10113 rfs4_grace_reset_all(nsrv4); 10114 } 10115 10116 if (rfs4_dss_numnewpaths > 0) 10117 kmem_free(added_paths, rfs4_dss_numnewpaths * sizeof (char *)); 10118 } 10119