1 /* 2 * CDDL HEADER START 3 * 4 * The contents of this file are subject to the terms of the 5 * Common Development and Distribution License (the "License"). 6 * You may not use this file except in compliance with the License. 7 * 8 * You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE 9 * or http://www.opensolaris.org/os/licensing. 10 * See the License for the specific language governing permissions 11 * and limitations under the License. 12 * 13 * When distributing Covered Code, include this CDDL HEADER in each 14 * file and include the License file at usr/src/OPENSOLARIS.LICENSE. 15 * If applicable, add the following below this CDDL HEADER, with the 16 * fields enclosed by brackets "[]" replaced with your own identifying 17 * information: Portions Copyright [yyyy] [name of copyright owner] 18 * 19 * CDDL HEADER END 20 */ 21 /* 22 * Copyright (c) 2003, 2010, Oracle and/or its affiliates. All rights reserved. 23 * Copyright 2012 Nexenta Systems, Inc. All rights reserved. 24 */ 25 26 27 /* 28 * Copyright (c) 1983,1984,1985,1986,1987,1988,1989 AT&T. 29 * All Rights Reserved 30 */ 31 32 #include <sys/param.h> 33 #include <sys/types.h> 34 #include <sys/systm.h> 35 #include <sys/cred.h> 36 #include <sys/buf.h> 37 #include <sys/vfs.h> 38 #include <sys/vfs_opreg.h> 39 #include <sys/vnode.h> 40 #include <sys/uio.h> 41 #include <sys/errno.h> 42 #include <sys/sysmacros.h> 43 #include <sys/statvfs.h> 44 #include <sys/kmem.h> 45 #include <sys/dirent.h> 46 #include <sys/cmn_err.h> 47 #include <sys/debug.h> 48 #include <sys/systeminfo.h> 49 #include <sys/flock.h> 50 #include <sys/pathname.h> 51 #include <sys/nbmlock.h> 52 #include <sys/share.h> 53 #include <sys/atomic.h> 54 #include <sys/policy.h> 55 #include <sys/fem.h> 56 #include <sys/sdt.h> 57 #include <sys/ddi.h> 58 #include <sys/zone.h> 59 60 #include <fs/fs_reparse.h> 61 62 #include <rpc/types.h> 63 #include <rpc/auth.h> 64 #include <rpc/rpcsec_gss.h> 65 #include <rpc/svc.h> 66 67 #include <nfs/nfs.h> 68 #include <nfs/export.h> 69 #include <nfs/nfs_cmd.h> 70 #include <nfs/lm.h> 71 #include <nfs/nfs4.h> 72 73 #include <sys/strsubr.h> 74 #include <sys/strsun.h> 75 76 #include <inet/common.h> 77 #include <inet/ip.h> 78 #include <inet/ip6.h> 79 80 #include <sys/tsol/label.h> 81 #include <sys/tsol/tndb.h> 82 83 #define RFS4_MAXLOCK_TRIES 4 /* Try to get the lock this many times */ 84 static int rfs4_maxlock_tries = RFS4_MAXLOCK_TRIES; 85 #define RFS4_LOCK_DELAY 10 /* Milliseconds */ 86 static clock_t rfs4_lock_delay = RFS4_LOCK_DELAY; 87 extern struct svc_ops rdma_svc_ops; 88 extern int nfs_loaned_buffers; 89 /* End of Tunables */ 90 91 static int rdma_setup_read_data4(READ4args *, READ4res *); 92 93 /* 94 * Used to bump the stateid4.seqid value and show changes in the stateid 95 */ 96 #define next_stateid(sp) (++(sp)->bits.chgseq) 97 98 /* 99 * RFS4_MINLEN_ENTRY4: XDR-encoded size of smallest possible dirent. 100 * This is used to return NFS4ERR_TOOSMALL when clients specify 101 * maxcount that isn't large enough to hold the smallest possible 102 * XDR encoded dirent. 103 * 104 * sizeof cookie (8 bytes) + 105 * sizeof name_len (4 bytes) + 106 * sizeof smallest (padded) name (4 bytes) + 107 * sizeof bitmap4_len (12 bytes) + NOTE: we always encode len=2 bm4 108 * sizeof attrlist4_len (4 bytes) + 109 * sizeof next boolean (4 bytes) 110 * 111 * RFS4_MINLEN_RDDIR4: XDR-encoded size of READDIR op reply containing 112 * the smallest possible entry4 (assumes no attrs requested). 113 * sizeof nfsstat4 (4 bytes) + 114 * sizeof verifier4 (8 bytes) + 115 * sizeof entry4list bool (4 bytes) + 116 * sizeof entry4 (36 bytes) + 117 * sizeof eof bool (4 bytes) 118 * 119 * RFS4_MINLEN_RDDIR_BUF: minimum length of buffer server will provide to 120 * VOP_READDIR. Its value is the size of the maximum possible dirent 121 * for solaris. The DIRENT64_RECLEN macro returns the size of dirent 122 * required for a given name length. MAXNAMELEN is the maximum 123 * filename length allowed in Solaris. The first two DIRENT64_RECLEN() 124 * macros are to allow for . and .. entries -- just a minor tweak to try 125 * and guarantee that buffer we give to VOP_READDIR will be large enough 126 * to hold ., .., and the largest possible solaris dirent64. 127 */ 128 #define RFS4_MINLEN_ENTRY4 36 129 #define RFS4_MINLEN_RDDIR4 (4 + NFS4_VERIFIER_SIZE + 4 + RFS4_MINLEN_ENTRY4 + 4) 130 #define RFS4_MINLEN_RDDIR_BUF \ 131 (DIRENT64_RECLEN(1) + DIRENT64_RECLEN(2) + DIRENT64_RECLEN(MAXNAMELEN)) 132 133 /* 134 * It would be better to pad to 4 bytes since that's what XDR would do, 135 * but the dirents UFS gives us are already padded to 8, so just take 136 * what we're given. Dircount is only a hint anyway. Currently the 137 * solaris kernel is ASCII only, so there's no point in calling the 138 * UTF8 functions. 139 * 140 * dirent64: named padded to provide 8 byte struct alignment 141 * d_ino(8) + d_off(8) + d_reclen(2) + d_name(namelen + null(1) + pad) 142 * 143 * cookie: uint64_t + utf8namelen: uint_t + utf8name padded to 8 bytes 144 * 145 */ 146 #define DIRENT64_TO_DIRCOUNT(dp) \ 147 (3 * BYTES_PER_XDR_UNIT + DIRENT64_NAMELEN((dp)->d_reclen)) 148 149 time_t rfs4_start_time; /* Initialized in rfs4_srvrinit */ 150 151 static sysid_t lockt_sysid; /* dummy sysid for all LOCKT calls */ 152 153 u_longlong_t nfs4_srv_caller_id; 154 uint_t nfs4_srv_vkey = 0; 155 156 verifier4 Write4verf; 157 verifier4 Readdir4verf; 158 159 void rfs4_init_compound_state(struct compound_state *); 160 161 static void nullfree(caddr_t); 162 static void rfs4_op_inval(nfs_argop4 *, nfs_resop4 *, struct svc_req *, 163 struct compound_state *); 164 static void rfs4_op_access(nfs_argop4 *, nfs_resop4 *, struct svc_req *, 165 struct compound_state *); 166 static void rfs4_op_close(nfs_argop4 *, nfs_resop4 *, struct svc_req *, 167 struct compound_state *); 168 static void rfs4_op_commit(nfs_argop4 *, nfs_resop4 *, struct svc_req *, 169 struct compound_state *); 170 static void rfs4_op_create(nfs_argop4 *, nfs_resop4 *, struct svc_req *, 171 struct compound_state *); 172 static void rfs4_op_create_free(nfs_resop4 *resop); 173 static void rfs4_op_delegreturn(nfs_argop4 *, nfs_resop4 *, 174 struct svc_req *, struct compound_state *); 175 static void rfs4_op_delegpurge(nfs_argop4 *, nfs_resop4 *, 176 struct svc_req *, struct compound_state *); 177 static void rfs4_op_getattr(nfs_argop4 *, nfs_resop4 *, struct svc_req *, 178 struct compound_state *); 179 static void rfs4_op_getattr_free(nfs_resop4 *); 180 static void rfs4_op_getfh(nfs_argop4 *, nfs_resop4 *, struct svc_req *, 181 struct compound_state *); 182 static void rfs4_op_getfh_free(nfs_resop4 *); 183 static void rfs4_op_illegal(nfs_argop4 *, nfs_resop4 *, struct svc_req *, 184 struct compound_state *); 185 static void rfs4_op_link(nfs_argop4 *, nfs_resop4 *, struct svc_req *, 186 struct compound_state *); 187 static void rfs4_op_lock(nfs_argop4 *, nfs_resop4 *, struct svc_req *, 188 struct compound_state *); 189 static void lock_denied_free(nfs_resop4 *); 190 static void rfs4_op_locku(nfs_argop4 *, nfs_resop4 *, struct svc_req *, 191 struct compound_state *); 192 static void rfs4_op_lockt(nfs_argop4 *, nfs_resop4 *, struct svc_req *, 193 struct compound_state *); 194 static void rfs4_op_lookup(nfs_argop4 *, nfs_resop4 *, struct svc_req *, 195 struct compound_state *); 196 static void rfs4_op_lookupp(nfs_argop4 *, nfs_resop4 *, struct svc_req *, 197 struct compound_state *); 198 static void rfs4_op_openattr(nfs_argop4 *argop, nfs_resop4 *resop, 199 struct svc_req *req, struct compound_state *cs); 200 static void rfs4_op_nverify(nfs_argop4 *, nfs_resop4 *, struct svc_req *, 201 struct compound_state *); 202 static void rfs4_op_open(nfs_argop4 *, nfs_resop4 *, struct svc_req *, 203 struct compound_state *); 204 static void rfs4_op_open_confirm(nfs_argop4 *, nfs_resop4 *, 205 struct svc_req *, struct compound_state *); 206 static void rfs4_op_open_downgrade(nfs_argop4 *, nfs_resop4 *, 207 struct svc_req *, struct compound_state *); 208 static void rfs4_op_putfh(nfs_argop4 *, nfs_resop4 *, struct svc_req *, 209 struct compound_state *); 210 static void rfs4_op_putpubfh(nfs_argop4 *, nfs_resop4 *, struct svc_req *, 211 struct compound_state *); 212 static void rfs4_op_putrootfh(nfs_argop4 *, nfs_resop4 *, struct svc_req *, 213 struct compound_state *); 214 static void rfs4_op_read(nfs_argop4 *, nfs_resop4 *, struct svc_req *, 215 struct compound_state *); 216 static void rfs4_op_read_free(nfs_resop4 *); 217 static void rfs4_op_readdir_free(nfs_resop4 *resop); 218 static void rfs4_op_readlink(nfs_argop4 *, nfs_resop4 *, struct svc_req *, 219 struct compound_state *); 220 static void rfs4_op_readlink_free(nfs_resop4 *); 221 static void rfs4_op_release_lockowner(nfs_argop4 *, nfs_resop4 *, 222 struct svc_req *, struct compound_state *); 223 static void rfs4_op_remove(nfs_argop4 *, nfs_resop4 *, struct svc_req *, 224 struct compound_state *); 225 static void rfs4_op_rename(nfs_argop4 *, nfs_resop4 *, struct svc_req *, 226 struct compound_state *); 227 static void rfs4_op_renew(nfs_argop4 *, nfs_resop4 *, struct svc_req *, 228 struct compound_state *); 229 static void rfs4_op_restorefh(nfs_argop4 *, nfs_resop4 *, struct svc_req *, 230 struct compound_state *); 231 static void rfs4_op_savefh(nfs_argop4 *, nfs_resop4 *, struct svc_req *, 232 struct compound_state *); 233 static void rfs4_op_setattr(nfs_argop4 *, nfs_resop4 *, struct svc_req *, 234 struct compound_state *); 235 static void rfs4_op_verify(nfs_argop4 *, nfs_resop4 *, struct svc_req *, 236 struct compound_state *); 237 static void rfs4_op_write(nfs_argop4 *, nfs_resop4 *, struct svc_req *, 238 struct compound_state *); 239 static void rfs4_op_setclientid(nfs_argop4 *, nfs_resop4 *, 240 struct svc_req *, struct compound_state *); 241 static void rfs4_op_setclientid_confirm(nfs_argop4 *, nfs_resop4 *, 242 struct svc_req *req, struct compound_state *); 243 static void rfs4_op_secinfo(nfs_argop4 *, nfs_resop4 *, struct svc_req *, 244 struct compound_state *); 245 static void rfs4_op_secinfo_free(nfs_resop4 *); 246 247 static nfsstat4 check_open_access(uint32_t, 248 struct compound_state *, struct svc_req *); 249 nfsstat4 rfs4_client_sysid(rfs4_client_t *, sysid_t *); 250 void rfs4_ss_clid(rfs4_client_t *); 251 252 /* 253 * translation table for attrs 254 */ 255 struct nfs4_ntov_table { 256 union nfs4_attr_u *na; 257 uint8_t amap[NFS4_MAXNUM_ATTRS]; 258 int attrcnt; 259 bool_t vfsstat; 260 }; 261 262 static void nfs4_ntov_table_init(struct nfs4_ntov_table *ntovp); 263 static void nfs4_ntov_table_free(struct nfs4_ntov_table *ntovp, 264 struct nfs4_svgetit_arg *sargp); 265 266 static nfsstat4 do_rfs4_set_attrs(bitmap4 *resp, fattr4 *fattrp, 267 struct compound_state *cs, struct nfs4_svgetit_arg *sargp, 268 struct nfs4_ntov_table *ntovp, nfs4_attr_cmd_t cmd); 269 270 fem_t *deleg_rdops; 271 fem_t *deleg_wrops; 272 273 rfs4_servinst_t *rfs4_cur_servinst = NULL; /* current server instance */ 274 kmutex_t rfs4_servinst_lock; /* protects linked list */ 275 int rfs4_seen_first_compound; /* set first time we see one */ 276 277 /* 278 * NFS4 op dispatch table 279 */ 280 281 struct rfsv4disp { 282 void (*dis_proc)(); /* proc to call */ 283 void (*dis_resfree)(); /* frees space allocated by proc */ 284 int dis_flags; /* RPC_IDEMPOTENT, etc... */ 285 }; 286 287 static struct rfsv4disp rfsv4disptab[] = { 288 /* 289 * NFS VERSION 4 290 */ 291 292 /* RFS_NULL = 0 */ 293 {rfs4_op_illegal, nullfree, 0}, 294 295 /* UNUSED = 1 */ 296 {rfs4_op_illegal, nullfree, 0}, 297 298 /* UNUSED = 2 */ 299 {rfs4_op_illegal, nullfree, 0}, 300 301 /* OP_ACCESS = 3 */ 302 {rfs4_op_access, nullfree, RPC_IDEMPOTENT}, 303 304 /* OP_CLOSE = 4 */ 305 {rfs4_op_close, nullfree, 0}, 306 307 /* OP_COMMIT = 5 */ 308 {rfs4_op_commit, nullfree, RPC_IDEMPOTENT}, 309 310 /* OP_CREATE = 6 */ 311 {rfs4_op_create, nullfree, 0}, 312 313 /* OP_DELEGPURGE = 7 */ 314 {rfs4_op_delegpurge, nullfree, 0}, 315 316 /* OP_DELEGRETURN = 8 */ 317 {rfs4_op_delegreturn, nullfree, 0}, 318 319 /* OP_GETATTR = 9 */ 320 {rfs4_op_getattr, rfs4_op_getattr_free, RPC_IDEMPOTENT}, 321 322 /* OP_GETFH = 10 */ 323 {rfs4_op_getfh, rfs4_op_getfh_free, RPC_ALL}, 324 325 /* OP_LINK = 11 */ 326 {rfs4_op_link, nullfree, 0}, 327 328 /* OP_LOCK = 12 */ 329 {rfs4_op_lock, lock_denied_free, 0}, 330 331 /* OP_LOCKT = 13 */ 332 {rfs4_op_lockt, lock_denied_free, 0}, 333 334 /* OP_LOCKU = 14 */ 335 {rfs4_op_locku, nullfree, 0}, 336 337 /* OP_LOOKUP = 15 */ 338 {rfs4_op_lookup, nullfree, (RPC_IDEMPOTENT | RPC_PUBLICFH_OK)}, 339 340 /* OP_LOOKUPP = 16 */ 341 {rfs4_op_lookupp, nullfree, (RPC_IDEMPOTENT | RPC_PUBLICFH_OK)}, 342 343 /* OP_NVERIFY = 17 */ 344 {rfs4_op_nverify, nullfree, RPC_IDEMPOTENT}, 345 346 /* OP_OPEN = 18 */ 347 {rfs4_op_open, rfs4_free_reply, 0}, 348 349 /* OP_OPENATTR = 19 */ 350 {rfs4_op_openattr, nullfree, 0}, 351 352 /* OP_OPEN_CONFIRM = 20 */ 353 {rfs4_op_open_confirm, nullfree, 0}, 354 355 /* OP_OPEN_DOWNGRADE = 21 */ 356 {rfs4_op_open_downgrade, nullfree, 0}, 357 358 /* OP_OPEN_PUTFH = 22 */ 359 {rfs4_op_putfh, nullfree, RPC_ALL}, 360 361 /* OP_PUTPUBFH = 23 */ 362 {rfs4_op_putpubfh, nullfree, RPC_ALL}, 363 364 /* OP_PUTROOTFH = 24 */ 365 {rfs4_op_putrootfh, nullfree, RPC_ALL}, 366 367 /* OP_READ = 25 */ 368 {rfs4_op_read, rfs4_op_read_free, RPC_IDEMPOTENT}, 369 370 /* OP_READDIR = 26 */ 371 {rfs4_op_readdir, rfs4_op_readdir_free, RPC_IDEMPOTENT}, 372 373 /* OP_READLINK = 27 */ 374 {rfs4_op_readlink, rfs4_op_readlink_free, RPC_IDEMPOTENT}, 375 376 /* OP_REMOVE = 28 */ 377 {rfs4_op_remove, nullfree, 0}, 378 379 /* OP_RENAME = 29 */ 380 {rfs4_op_rename, nullfree, 0}, 381 382 /* OP_RENEW = 30 */ 383 {rfs4_op_renew, nullfree, 0}, 384 385 /* OP_RESTOREFH = 31 */ 386 {rfs4_op_restorefh, nullfree, RPC_ALL}, 387 388 /* OP_SAVEFH = 32 */ 389 {rfs4_op_savefh, nullfree, RPC_ALL}, 390 391 /* OP_SECINFO = 33 */ 392 {rfs4_op_secinfo, rfs4_op_secinfo_free, 0}, 393 394 /* OP_SETATTR = 34 */ 395 {rfs4_op_setattr, nullfree, 0}, 396 397 /* OP_SETCLIENTID = 35 */ 398 {rfs4_op_setclientid, nullfree, 0}, 399 400 /* OP_SETCLIENTID_CONFIRM = 36 */ 401 {rfs4_op_setclientid_confirm, nullfree, 0}, 402 403 /* OP_VERIFY = 37 */ 404 {rfs4_op_verify, nullfree, RPC_IDEMPOTENT}, 405 406 /* OP_WRITE = 38 */ 407 {rfs4_op_write, nullfree, 0}, 408 409 /* OP_RELEASE_LOCKOWNER = 39 */ 410 {rfs4_op_release_lockowner, nullfree, 0}, 411 }; 412 413 static uint_t rfsv4disp_cnt = sizeof (rfsv4disptab) / sizeof (rfsv4disptab[0]); 414 415 #define OP_ILLEGAL_IDX (rfsv4disp_cnt) 416 417 #ifdef DEBUG 418 419 int rfs4_fillone_debug = 0; 420 int rfs4_no_stub_access = 1; 421 int rfs4_rddir_debug = 0; 422 423 static char *rfs4_op_string[] = { 424 "rfs4_op_null", 425 "rfs4_op_1 unused", 426 "rfs4_op_2 unused", 427 "rfs4_op_access", 428 "rfs4_op_close", 429 "rfs4_op_commit", 430 "rfs4_op_create", 431 "rfs4_op_delegpurge", 432 "rfs4_op_delegreturn", 433 "rfs4_op_getattr", 434 "rfs4_op_getfh", 435 "rfs4_op_link", 436 "rfs4_op_lock", 437 "rfs4_op_lockt", 438 "rfs4_op_locku", 439 "rfs4_op_lookup", 440 "rfs4_op_lookupp", 441 "rfs4_op_nverify", 442 "rfs4_op_open", 443 "rfs4_op_openattr", 444 "rfs4_op_open_confirm", 445 "rfs4_op_open_downgrade", 446 "rfs4_op_putfh", 447 "rfs4_op_putpubfh", 448 "rfs4_op_putrootfh", 449 "rfs4_op_read", 450 "rfs4_op_readdir", 451 "rfs4_op_readlink", 452 "rfs4_op_remove", 453 "rfs4_op_rename", 454 "rfs4_op_renew", 455 "rfs4_op_restorefh", 456 "rfs4_op_savefh", 457 "rfs4_op_secinfo", 458 "rfs4_op_setattr", 459 "rfs4_op_setclientid", 460 "rfs4_op_setclient_confirm", 461 "rfs4_op_verify", 462 "rfs4_op_write", 463 "rfs4_op_release_lockowner", 464 "rfs4_op_illegal" 465 }; 466 #endif 467 468 void rfs4_ss_chkclid(rfs4_client_t *); 469 470 extern size_t strlcpy(char *dst, const char *src, size_t dstsize); 471 472 extern void rfs4_free_fs_locations4(fs_locations4 *); 473 474 #ifdef nextdp 475 #undef nextdp 476 #endif 477 #define nextdp(dp) ((struct dirent64 *)((char *)(dp) + (dp)->d_reclen)) 478 479 static const fs_operation_def_t nfs4_rd_deleg_tmpl[] = { 480 VOPNAME_OPEN, { .femop_open = deleg_rd_open }, 481 VOPNAME_WRITE, { .femop_write = deleg_rd_write }, 482 VOPNAME_SETATTR, { .femop_setattr = deleg_rd_setattr }, 483 VOPNAME_RWLOCK, { .femop_rwlock = deleg_rd_rwlock }, 484 VOPNAME_SPACE, { .femop_space = deleg_rd_space }, 485 VOPNAME_SETSECATTR, { .femop_setsecattr = deleg_rd_setsecattr }, 486 VOPNAME_VNEVENT, { .femop_vnevent = deleg_rd_vnevent }, 487 NULL, NULL 488 }; 489 static const fs_operation_def_t nfs4_wr_deleg_tmpl[] = { 490 VOPNAME_OPEN, { .femop_open = deleg_wr_open }, 491 VOPNAME_READ, { .femop_read = deleg_wr_read }, 492 VOPNAME_WRITE, { .femop_write = deleg_wr_write }, 493 VOPNAME_SETATTR, { .femop_setattr = deleg_wr_setattr }, 494 VOPNAME_RWLOCK, { .femop_rwlock = deleg_wr_rwlock }, 495 VOPNAME_SPACE, { .femop_space = deleg_wr_space }, 496 VOPNAME_SETSECATTR, { .femop_setsecattr = deleg_wr_setsecattr }, 497 VOPNAME_VNEVENT, { .femop_vnevent = deleg_wr_vnevent }, 498 NULL, NULL 499 }; 500 501 int 502 rfs4_srvrinit(void) 503 { 504 timespec32_t verf; 505 int error; 506 extern void rfs4_attr_init(); 507 extern krwlock_t rfs4_deleg_policy_lock; 508 509 /* 510 * The following algorithm attempts to find a unique verifier 511 * to be used as the write verifier returned from the server 512 * to the client. It is important that this verifier change 513 * whenever the server reboots. Of secondary importance, it 514 * is important for the verifier to be unique between two 515 * different servers. 516 * 517 * Thus, an attempt is made to use the system hostid and the 518 * current time in seconds when the nfssrv kernel module is 519 * loaded. It is assumed that an NFS server will not be able 520 * to boot and then to reboot in less than a second. If the 521 * hostid has not been set, then the current high resolution 522 * time is used. This will ensure different verifiers each 523 * time the server reboots and minimize the chances that two 524 * different servers will have the same verifier. 525 * XXX - this is broken on LP64 kernels. 526 */ 527 verf.tv_sec = (time_t)zone_get_hostid(NULL); 528 if (verf.tv_sec != 0) { 529 verf.tv_nsec = gethrestime_sec(); 530 } else { 531 timespec_t tverf; 532 533 gethrestime(&tverf); 534 verf.tv_sec = (time_t)tverf.tv_sec; 535 verf.tv_nsec = tverf.tv_nsec; 536 } 537 538 Write4verf = *(uint64_t *)&verf; 539 540 rfs4_attr_init(); 541 mutex_init(&rfs4_deleg_lock, NULL, MUTEX_DEFAULT, NULL); 542 543 /* Used to manage create/destroy of server state */ 544 mutex_init(&rfs4_state_lock, NULL, MUTEX_DEFAULT, NULL); 545 546 /* Used to manage access to server instance linked list */ 547 mutex_init(&rfs4_servinst_lock, NULL, MUTEX_DEFAULT, NULL); 548 549 /* Used to manage access to rfs4_deleg_policy */ 550 rw_init(&rfs4_deleg_policy_lock, NULL, RW_DEFAULT, NULL); 551 552 error = fem_create("deleg_rdops", nfs4_rd_deleg_tmpl, &deleg_rdops); 553 if (error != 0) { 554 rfs4_disable_delegation(); 555 } else { 556 error = fem_create("deleg_wrops", nfs4_wr_deleg_tmpl, 557 &deleg_wrops); 558 if (error != 0) { 559 rfs4_disable_delegation(); 560 fem_free(deleg_rdops); 561 } 562 } 563 564 nfs4_srv_caller_id = fs_new_caller_id(); 565 566 lockt_sysid = lm_alloc_sysidt(); 567 568 vsd_create(&nfs4_srv_vkey, NULL); 569 570 return (0); 571 } 572 573 void 574 rfs4_srvrfini(void) 575 { 576 extern krwlock_t rfs4_deleg_policy_lock; 577 578 if (lockt_sysid != LM_NOSYSID) { 579 lm_free_sysidt(lockt_sysid); 580 lockt_sysid = LM_NOSYSID; 581 } 582 583 mutex_destroy(&rfs4_deleg_lock); 584 mutex_destroy(&rfs4_state_lock); 585 rw_destroy(&rfs4_deleg_policy_lock); 586 587 fem_free(deleg_rdops); 588 fem_free(deleg_wrops); 589 } 590 591 void 592 rfs4_init_compound_state(struct compound_state *cs) 593 { 594 bzero(cs, sizeof (*cs)); 595 cs->cont = TRUE; 596 cs->access = CS_ACCESS_DENIED; 597 cs->deleg = FALSE; 598 cs->mandlock = FALSE; 599 cs->fh.nfs_fh4_val = cs->fhbuf; 600 } 601 602 void 603 rfs4_grace_start(rfs4_servinst_t *sip) 604 { 605 rw_enter(&sip->rwlock, RW_WRITER); 606 sip->start_time = (time_t)TICK_TO_SEC(ddi_get_lbolt()); 607 sip->grace_period = rfs4_grace_period; 608 rw_exit(&sip->rwlock); 609 } 610 611 /* 612 * returns true if the instance's grace period has never been started 613 */ 614 int 615 rfs4_servinst_grace_new(rfs4_servinst_t *sip) 616 { 617 time_t start_time; 618 619 rw_enter(&sip->rwlock, RW_READER); 620 start_time = sip->start_time; 621 rw_exit(&sip->rwlock); 622 623 return (start_time == 0); 624 } 625 626 /* 627 * Indicates if server instance is within the 628 * grace period. 629 */ 630 int 631 rfs4_servinst_in_grace(rfs4_servinst_t *sip) 632 { 633 time_t grace_expiry; 634 635 rw_enter(&sip->rwlock, RW_READER); 636 grace_expiry = sip->start_time + sip->grace_period; 637 rw_exit(&sip->rwlock); 638 639 return (((time_t)TICK_TO_SEC(ddi_get_lbolt())) < grace_expiry); 640 } 641 642 int 643 rfs4_clnt_in_grace(rfs4_client_t *cp) 644 { 645 ASSERT(rfs4_dbe_refcnt(cp->rc_dbe) > 0); 646 647 return (rfs4_servinst_in_grace(cp->rc_server_instance)); 648 } 649 650 /* 651 * reset all currently active grace periods 652 */ 653 void 654 rfs4_grace_reset_all(void) 655 { 656 rfs4_servinst_t *sip; 657 658 mutex_enter(&rfs4_servinst_lock); 659 for (sip = rfs4_cur_servinst; sip != NULL; sip = sip->prev) 660 if (rfs4_servinst_in_grace(sip)) 661 rfs4_grace_start(sip); 662 mutex_exit(&rfs4_servinst_lock); 663 } 664 665 /* 666 * start any new instances' grace periods 667 */ 668 void 669 rfs4_grace_start_new(void) 670 { 671 rfs4_servinst_t *sip; 672 673 mutex_enter(&rfs4_servinst_lock); 674 for (sip = rfs4_cur_servinst; sip != NULL; sip = sip->prev) 675 if (rfs4_servinst_grace_new(sip)) 676 rfs4_grace_start(sip); 677 mutex_exit(&rfs4_servinst_lock); 678 } 679 680 static rfs4_dss_path_t * 681 rfs4_dss_newpath(rfs4_servinst_t *sip, char *path, unsigned index) 682 { 683 size_t len; 684 rfs4_dss_path_t *dss_path; 685 686 dss_path = kmem_alloc(sizeof (rfs4_dss_path_t), KM_SLEEP); 687 688 /* 689 * Take a copy of the string, since the original may be overwritten. 690 * Sadly, no strdup() in the kernel. 691 */ 692 /* allow for NUL */ 693 len = strlen(path) + 1; 694 dss_path->path = kmem_alloc(len, KM_SLEEP); 695 (void) strlcpy(dss_path->path, path, len); 696 697 /* associate with servinst */ 698 dss_path->sip = sip; 699 dss_path->index = index; 700 701 /* 702 * Add to list of served paths. 703 * No locking required, as we're only ever called at startup. 704 */ 705 if (rfs4_dss_pathlist == NULL) { 706 /* this is the first dss_path_t */ 707 708 /* needed for insque/remque */ 709 dss_path->next = dss_path->prev = dss_path; 710 711 rfs4_dss_pathlist = dss_path; 712 } else { 713 insque(dss_path, rfs4_dss_pathlist); 714 } 715 716 return (dss_path); 717 } 718 719 /* 720 * Create a new server instance, and make it the currently active instance. 721 * Note that starting the grace period too early will reduce the clients' 722 * recovery window. 723 */ 724 void 725 rfs4_servinst_create(int start_grace, int dss_npaths, char **dss_paths) 726 { 727 unsigned i; 728 rfs4_servinst_t *sip; 729 rfs4_oldstate_t *oldstate; 730 731 sip = kmem_alloc(sizeof (rfs4_servinst_t), KM_SLEEP); 732 rw_init(&sip->rwlock, NULL, RW_DEFAULT, NULL); 733 734 sip->start_time = (time_t)0; 735 sip->grace_period = (time_t)0; 736 sip->next = NULL; 737 sip->prev = NULL; 738 739 rw_init(&sip->oldstate_lock, NULL, RW_DEFAULT, NULL); 740 /* 741 * This initial dummy entry is required to setup for insque/remque. 742 * It must be skipped over whenever the list is traversed. 743 */ 744 oldstate = kmem_alloc(sizeof (rfs4_oldstate_t), KM_SLEEP); 745 /* insque/remque require initial list entry to be self-terminated */ 746 oldstate->next = oldstate; 747 oldstate->prev = oldstate; 748 sip->oldstate = oldstate; 749 750 751 sip->dss_npaths = dss_npaths; 752 sip->dss_paths = kmem_alloc(dss_npaths * 753 sizeof (rfs4_dss_path_t *), KM_SLEEP); 754 755 for (i = 0; i < dss_npaths; i++) { 756 sip->dss_paths[i] = rfs4_dss_newpath(sip, dss_paths[i], i); 757 } 758 759 mutex_enter(&rfs4_servinst_lock); 760 if (rfs4_cur_servinst != NULL) { 761 /* add to linked list */ 762 sip->prev = rfs4_cur_servinst; 763 rfs4_cur_servinst->next = sip; 764 } 765 if (start_grace) 766 rfs4_grace_start(sip); 767 /* make the new instance "current" */ 768 rfs4_cur_servinst = sip; 769 770 mutex_exit(&rfs4_servinst_lock); 771 } 772 773 /* 774 * In future, we might add a rfs4_servinst_destroy(sip) but, for now, destroy 775 * all instances directly. 776 */ 777 void 778 rfs4_servinst_destroy_all(void) 779 { 780 rfs4_servinst_t *sip, *prev, *current; 781 #ifdef DEBUG 782 int n = 0; 783 #endif 784 785 mutex_enter(&rfs4_servinst_lock); 786 ASSERT(rfs4_cur_servinst != NULL); 787 current = rfs4_cur_servinst; 788 rfs4_cur_servinst = NULL; 789 for (sip = current; sip != NULL; sip = prev) { 790 prev = sip->prev; 791 rw_destroy(&sip->rwlock); 792 if (sip->oldstate) 793 kmem_free(sip->oldstate, sizeof (rfs4_oldstate_t)); 794 if (sip->dss_paths) 795 kmem_free(sip->dss_paths, 796 sip->dss_npaths * sizeof (rfs4_dss_path_t *)); 797 kmem_free(sip, sizeof (rfs4_servinst_t)); 798 #ifdef DEBUG 799 n++; 800 #endif 801 } 802 mutex_exit(&rfs4_servinst_lock); 803 } 804 805 /* 806 * Assign the current server instance to a client_t. 807 * Should be called with cp->rc_dbe held. 808 */ 809 void 810 rfs4_servinst_assign(rfs4_client_t *cp, rfs4_servinst_t *sip) 811 { 812 ASSERT(rfs4_dbe_refcnt(cp->rc_dbe) > 0); 813 814 /* 815 * The lock ensures that if the current instance is in the process 816 * of changing, we will see the new one. 817 */ 818 mutex_enter(&rfs4_servinst_lock); 819 cp->rc_server_instance = sip; 820 mutex_exit(&rfs4_servinst_lock); 821 } 822 823 rfs4_servinst_t * 824 rfs4_servinst(rfs4_client_t *cp) 825 { 826 ASSERT(rfs4_dbe_refcnt(cp->rc_dbe) > 0); 827 828 return (cp->rc_server_instance); 829 } 830 831 /* ARGSUSED */ 832 static void 833 nullfree(caddr_t resop) 834 { 835 } 836 837 /* 838 * This is a fall-through for invalid or not implemented (yet) ops 839 */ 840 /* ARGSUSED */ 841 static void 842 rfs4_op_inval(nfs_argop4 *argop, nfs_resop4 *resop, struct svc_req *req, 843 struct compound_state *cs) 844 { 845 *cs->statusp = *((nfsstat4 *)&(resop)->nfs_resop4_u) = NFS4ERR_INVAL; 846 } 847 848 /* 849 * Check if the security flavor, nfsnum, is in the flavor_list. 850 */ 851 bool_t 852 in_flavor_list(int nfsnum, int *flavor_list, int count) 853 { 854 int i; 855 856 for (i = 0; i < count; i++) { 857 if (nfsnum == flavor_list[i]) 858 return (TRUE); 859 } 860 return (FALSE); 861 } 862 863 /* 864 * Used by rfs4_op_secinfo to get the security information from the 865 * export structure associated with the component. 866 */ 867 /* ARGSUSED */ 868 static nfsstat4 869 do_rfs4_op_secinfo(struct compound_state *cs, char *nm, SECINFO4res *resp) 870 { 871 int error, different_export = 0; 872 vnode_t *dvp, *vp, *tvp; 873 struct exportinfo *exi = NULL; 874 fid_t fid; 875 uint_t count, i; 876 secinfo4 *resok_val; 877 struct secinfo *secp; 878 seconfig_t *si; 879 bool_t did_traverse = FALSE; 880 int dotdot, walk; 881 882 dvp = cs->vp; 883 dotdot = (nm[0] == '.' && nm[1] == '.' && nm[2] == '\0'); 884 885 /* 886 * If dotdotting, then need to check whether it's above the 887 * root of a filesystem, or above an export point. 888 */ 889 if (dotdot) { 890 891 /* 892 * If dotdotting at the root of a filesystem, then 893 * need to traverse back to the mounted-on filesystem 894 * and do the dotdot lookup there. 895 */ 896 if (cs->vp->v_flag & VROOT) { 897 898 /* 899 * If at the system root, then can 900 * go up no further. 901 */ 902 if (VN_CMP(dvp, rootdir)) 903 return (puterrno4(ENOENT)); 904 905 /* 906 * Traverse back to the mounted-on filesystem 907 */ 908 dvp = untraverse(cs->vp); 909 910 /* 911 * Set the different_export flag so we remember 912 * to pick up a new exportinfo entry for 913 * this new filesystem. 914 */ 915 different_export = 1; 916 } else { 917 918 /* 919 * If dotdotting above an export point then set 920 * the different_export to get new export info. 921 */ 922 different_export = nfs_exported(cs->exi, cs->vp); 923 } 924 } 925 926 /* 927 * Get the vnode for the component "nm". 928 */ 929 error = VOP_LOOKUP(dvp, nm, &vp, NULL, 0, NULL, cs->cr, 930 NULL, NULL, NULL); 931 if (error) 932 return (puterrno4(error)); 933 934 /* 935 * If the vnode is in a pseudo filesystem, or if the security flavor 936 * used in the request is valid but not an explicitly shared flavor, 937 * or the access bit indicates that this is a limited access, 938 * check whether this vnode is visible. 939 */ 940 if (!different_export && 941 (PSEUDO(cs->exi) || ! is_exported_sec(cs->nfsflavor, cs->exi) || 942 cs->access & CS_ACCESS_LIMITED)) { 943 if (! nfs_visible(cs->exi, vp, &different_export)) { 944 VN_RELE(vp); 945 return (puterrno4(ENOENT)); 946 } 947 } 948 949 /* 950 * If it's a mountpoint, then traverse it. 951 */ 952 if (vn_ismntpt(vp)) { 953 tvp = vp; 954 if ((error = traverse(&tvp)) != 0) { 955 VN_RELE(vp); 956 return (puterrno4(error)); 957 } 958 /* remember that we had to traverse mountpoint */ 959 did_traverse = TRUE; 960 vp = tvp; 961 different_export = 1; 962 } else if (vp->v_vfsp != dvp->v_vfsp) { 963 /* 964 * If vp isn't a mountpoint and the vfs ptrs aren't the same, 965 * then vp is probably an LOFS object. We don't need the 966 * realvp, we just need to know that we might have crossed 967 * a server fs boundary and need to call checkexport4. 968 * (LOFS lookup hides server fs mountpoints, and actually calls 969 * traverse) 970 */ 971 different_export = 1; 972 } 973 974 /* 975 * Get the export information for it. 976 */ 977 if (different_export) { 978 979 bzero(&fid, sizeof (fid)); 980 fid.fid_len = MAXFIDSZ; 981 error = vop_fid_pseudo(vp, &fid); 982 if (error) { 983 VN_RELE(vp); 984 return (puterrno4(error)); 985 } 986 987 if (dotdot) 988 exi = nfs_vptoexi(NULL, vp, cs->cr, &walk, NULL, TRUE); 989 else 990 exi = checkexport4(&vp->v_vfsp->vfs_fsid, &fid, vp); 991 992 if (exi == NULL) { 993 if (did_traverse == TRUE) { 994 /* 995 * If this vnode is a mounted-on vnode, 996 * but the mounted-on file system is not 997 * exported, send back the secinfo for 998 * the exported node that the mounted-on 999 * vnode lives in. 1000 */ 1001 exi = cs->exi; 1002 } else { 1003 VN_RELE(vp); 1004 return (puterrno4(EACCES)); 1005 } 1006 } 1007 } else { 1008 exi = cs->exi; 1009 } 1010 ASSERT(exi != NULL); 1011 1012 1013 /* 1014 * Create the secinfo result based on the security information 1015 * from the exportinfo structure (exi). 1016 * 1017 * Return all flavors for a pseudo node. 1018 * For a real export node, return the flavor that the client 1019 * has access with. 1020 */ 1021 ASSERT(RW_LOCK_HELD(&exported_lock)); 1022 if (PSEUDO(exi)) { 1023 count = exi->exi_export.ex_seccnt; /* total sec count */ 1024 resok_val = kmem_alloc(count * sizeof (secinfo4), KM_SLEEP); 1025 secp = exi->exi_export.ex_secinfo; 1026 1027 for (i = 0; i < count; i++) { 1028 si = &secp[i].s_secinfo; 1029 resok_val[i].flavor = si->sc_rpcnum; 1030 if (resok_val[i].flavor == RPCSEC_GSS) { 1031 rpcsec_gss_info *info; 1032 1033 info = &resok_val[i].flavor_info; 1034 info->qop = si->sc_qop; 1035 info->service = (rpc_gss_svc_t)si->sc_service; 1036 1037 /* get oid opaque data */ 1038 info->oid.sec_oid4_len = 1039 si->sc_gss_mech_type->length; 1040 info->oid.sec_oid4_val = kmem_alloc( 1041 si->sc_gss_mech_type->length, KM_SLEEP); 1042 bcopy( 1043 si->sc_gss_mech_type->elements, 1044 info->oid.sec_oid4_val, 1045 info->oid.sec_oid4_len); 1046 } 1047 } 1048 resp->SECINFO4resok_len = count; 1049 resp->SECINFO4resok_val = resok_val; 1050 } else { 1051 int ret_cnt = 0, k = 0; 1052 int *flavor_list; 1053 1054 count = exi->exi_export.ex_seccnt; /* total sec count */ 1055 secp = exi->exi_export.ex_secinfo; 1056 1057 flavor_list = kmem_alloc(count * sizeof (int), KM_SLEEP); 1058 /* find out which flavors to return */ 1059 for (i = 0; i < count; i ++) { 1060 int access, flavor, perm; 1061 1062 flavor = secp[i].s_secinfo.sc_nfsnum; 1063 perm = secp[i].s_flags; 1064 1065 access = nfsauth4_secinfo_access(exi, cs->req, 1066 flavor, perm); 1067 1068 if (! (access & NFSAUTH_DENIED) && 1069 ! (access & NFSAUTH_WRONGSEC)) { 1070 flavor_list[ret_cnt] = flavor; 1071 ret_cnt++; 1072 } 1073 } 1074 1075 /* Create the returning SECINFO value */ 1076 resok_val = kmem_alloc(ret_cnt * sizeof (secinfo4), KM_SLEEP); 1077 1078 for (i = 0; i < count; i++) { 1079 /* 1080 * If the flavor is in the flavor list, 1081 * fill in resok_val. 1082 */ 1083 si = &secp[i].s_secinfo; 1084 if (in_flavor_list(si->sc_nfsnum, 1085 flavor_list, ret_cnt)) { 1086 resok_val[k].flavor = si->sc_rpcnum; 1087 if (resok_val[k].flavor == RPCSEC_GSS) { 1088 rpcsec_gss_info *info; 1089 1090 info = &resok_val[k].flavor_info; 1091 info->qop = si->sc_qop; 1092 info->service = (rpc_gss_svc_t) 1093 si->sc_service; 1094 1095 /* get oid opaque data */ 1096 info->oid.sec_oid4_len = 1097 si->sc_gss_mech_type->length; 1098 info->oid.sec_oid4_val = kmem_alloc( 1099 si->sc_gss_mech_type->length, 1100 KM_SLEEP); 1101 bcopy(si->sc_gss_mech_type->elements, 1102 info->oid.sec_oid4_val, 1103 info->oid.sec_oid4_len); 1104 } 1105 k++; 1106 } 1107 if (k >= ret_cnt) 1108 break; 1109 } 1110 resp->SECINFO4resok_len = ret_cnt; 1111 resp->SECINFO4resok_val = resok_val; 1112 kmem_free(flavor_list, count * sizeof (int)); 1113 } 1114 1115 VN_RELE(vp); 1116 return (NFS4_OK); 1117 } 1118 1119 /* 1120 * SECINFO (Operation 33): Obtain required security information on 1121 * the component name in the format of (security-mechanism-oid, qop, service) 1122 * triplets. 1123 */ 1124 /* ARGSUSED */ 1125 static void 1126 rfs4_op_secinfo(nfs_argop4 *argop, nfs_resop4 *resop, struct svc_req *req, 1127 struct compound_state *cs) 1128 { 1129 SECINFO4args *args = &argop->nfs_argop4_u.opsecinfo; 1130 SECINFO4res *resp = &resop->nfs_resop4_u.opsecinfo; 1131 utf8string *utfnm = &args->name; 1132 uint_t len; 1133 char *nm; 1134 struct sockaddr *ca; 1135 char *name = NULL; 1136 nfsstat4 status = NFS4_OK; 1137 1138 DTRACE_NFSV4_2(op__secinfo__start, struct compound_state *, cs, 1139 SECINFO4args *, args); 1140 1141 /* 1142 * Current file handle (cfh) should have been set before getting 1143 * into this function. If not, return error. 1144 */ 1145 if (cs->vp == NULL) { 1146 *cs->statusp = resp->status = NFS4ERR_NOFILEHANDLE; 1147 goto out; 1148 } 1149 1150 if (cs->vp->v_type != VDIR) { 1151 *cs->statusp = resp->status = NFS4ERR_NOTDIR; 1152 goto out; 1153 } 1154 1155 /* 1156 * Verify the component name. If failed, error out, but 1157 * do not error out if the component name is a "..". 1158 * SECINFO will return its parents secinfo data for SECINFO "..". 1159 */ 1160 status = utf8_dir_verify(utfnm); 1161 if (status != NFS4_OK) { 1162 if (utfnm->utf8string_len != 2 || 1163 utfnm->utf8string_val[0] != '.' || 1164 utfnm->utf8string_val[1] != '.') { 1165 *cs->statusp = resp->status = status; 1166 goto out; 1167 } 1168 } 1169 1170 nm = utf8_to_str(utfnm, &len, NULL); 1171 if (nm == NULL) { 1172 *cs->statusp = resp->status = NFS4ERR_INVAL; 1173 goto out; 1174 } 1175 1176 if (len > MAXNAMELEN) { 1177 *cs->statusp = resp->status = NFS4ERR_NAMETOOLONG; 1178 kmem_free(nm, len); 1179 goto out; 1180 } 1181 1182 ca = (struct sockaddr *)svc_getrpccaller(req->rq_xprt)->buf; 1183 name = nfscmd_convname(ca, cs->exi, nm, NFSCMD_CONV_INBOUND, 1184 MAXPATHLEN + 1); 1185 1186 if (name == NULL) { 1187 *cs->statusp = resp->status = NFS4ERR_INVAL; 1188 kmem_free(nm, len); 1189 goto out; 1190 } 1191 1192 1193 *cs->statusp = resp->status = do_rfs4_op_secinfo(cs, name, resp); 1194 1195 if (name != nm) 1196 kmem_free(name, MAXPATHLEN + 1); 1197 kmem_free(nm, len); 1198 1199 out: 1200 DTRACE_NFSV4_2(op__secinfo__done, struct compound_state *, cs, 1201 SECINFO4res *, resp); 1202 } 1203 1204 /* 1205 * Free SECINFO result. 1206 */ 1207 /* ARGSUSED */ 1208 static void 1209 rfs4_op_secinfo_free(nfs_resop4 *resop) 1210 { 1211 SECINFO4res *resp = &resop->nfs_resop4_u.opsecinfo; 1212 int count, i; 1213 secinfo4 *resok_val; 1214 1215 /* If this is not an Ok result, nothing to free. */ 1216 if (resp->status != NFS4_OK) { 1217 return; 1218 } 1219 1220 count = resp->SECINFO4resok_len; 1221 resok_val = resp->SECINFO4resok_val; 1222 1223 for (i = 0; i < count; i++) { 1224 if (resok_val[i].flavor == RPCSEC_GSS) { 1225 rpcsec_gss_info *info; 1226 1227 info = &resok_val[i].flavor_info; 1228 kmem_free(info->oid.sec_oid4_val, 1229 info->oid.sec_oid4_len); 1230 } 1231 } 1232 kmem_free(resok_val, count * sizeof (secinfo4)); 1233 resp->SECINFO4resok_len = 0; 1234 resp->SECINFO4resok_val = NULL; 1235 } 1236 1237 /* ARGSUSED */ 1238 static void 1239 rfs4_op_access(nfs_argop4 *argop, nfs_resop4 *resop, struct svc_req *req, 1240 struct compound_state *cs) 1241 { 1242 ACCESS4args *args = &argop->nfs_argop4_u.opaccess; 1243 ACCESS4res *resp = &resop->nfs_resop4_u.opaccess; 1244 int error; 1245 vnode_t *vp; 1246 struct vattr va; 1247 int checkwriteperm; 1248 cred_t *cr = cs->cr; 1249 bslabel_t *clabel, *slabel; 1250 ts_label_t *tslabel; 1251 boolean_t admin_low_client; 1252 1253 DTRACE_NFSV4_2(op__access__start, struct compound_state *, cs, 1254 ACCESS4args *, args); 1255 1256 #if 0 /* XXX allow access even if !cs->access. Eventually only pseudo fs */ 1257 if (cs->access == CS_ACCESS_DENIED) { 1258 *cs->statusp = resp->status = NFS4ERR_ACCESS; 1259 goto out; 1260 } 1261 #endif 1262 if (cs->vp == NULL) { 1263 *cs->statusp = resp->status = NFS4ERR_NOFILEHANDLE; 1264 goto out; 1265 } 1266 1267 ASSERT(cr != NULL); 1268 1269 vp = cs->vp; 1270 1271 /* 1272 * If the file system is exported read only, it is not appropriate 1273 * to check write permissions for regular files and directories. 1274 * Special files are interpreted by the client, so the underlying 1275 * permissions are sent back to the client for interpretation. 1276 */ 1277 if (rdonly4(cs->exi, cs->vp, req) && 1278 (vp->v_type == VREG || vp->v_type == VDIR)) 1279 checkwriteperm = 0; 1280 else 1281 checkwriteperm = 1; 1282 1283 /* 1284 * XXX 1285 * We need the mode so that we can correctly determine access 1286 * permissions relative to a mandatory lock file. Access to 1287 * mandatory lock files is denied on the server, so it might 1288 * as well be reflected to the server during the open. 1289 */ 1290 va.va_mask = AT_MODE; 1291 error = VOP_GETATTR(vp, &va, 0, cr, NULL); 1292 if (error) { 1293 *cs->statusp = resp->status = puterrno4(error); 1294 goto out; 1295 } 1296 resp->access = 0; 1297 resp->supported = 0; 1298 1299 if (is_system_labeled()) { 1300 ASSERT(req->rq_label != NULL); 1301 clabel = req->rq_label; 1302 DTRACE_PROBE2(tx__rfs4__log__info__opaccess__clabel, char *, 1303 "got client label from request(1)", 1304 struct svc_req *, req); 1305 if (!blequal(&l_admin_low->tsl_label, clabel)) { 1306 if ((tslabel = nfs_getflabel(vp, cs->exi)) == NULL) { 1307 *cs->statusp = resp->status = puterrno4(EACCES); 1308 goto out; 1309 } 1310 slabel = label2bslabel(tslabel); 1311 DTRACE_PROBE3(tx__rfs4__log__info__opaccess__slabel, 1312 char *, "got server label(1) for vp(2)", 1313 bslabel_t *, slabel, vnode_t *, vp); 1314 1315 admin_low_client = B_FALSE; 1316 } else 1317 admin_low_client = B_TRUE; 1318 } 1319 1320 if (args->access & ACCESS4_READ) { 1321 error = VOP_ACCESS(vp, VREAD, 0, cr, NULL); 1322 if (!error && !MANDLOCK(vp, va.va_mode) && 1323 (!is_system_labeled() || admin_low_client || 1324 bldominates(clabel, slabel))) 1325 resp->access |= ACCESS4_READ; 1326 resp->supported |= ACCESS4_READ; 1327 } 1328 if ((args->access & ACCESS4_LOOKUP) && vp->v_type == VDIR) { 1329 error = VOP_ACCESS(vp, VEXEC, 0, cr, NULL); 1330 if (!error && (!is_system_labeled() || admin_low_client || 1331 bldominates(clabel, slabel))) 1332 resp->access |= ACCESS4_LOOKUP; 1333 resp->supported |= ACCESS4_LOOKUP; 1334 } 1335 if (checkwriteperm && 1336 (args->access & (ACCESS4_MODIFY|ACCESS4_EXTEND))) { 1337 error = VOP_ACCESS(vp, VWRITE, 0, cr, NULL); 1338 if (!error && !MANDLOCK(vp, va.va_mode) && 1339 (!is_system_labeled() || admin_low_client || 1340 blequal(clabel, slabel))) 1341 resp->access |= 1342 (args->access & (ACCESS4_MODIFY | ACCESS4_EXTEND)); 1343 resp->supported |= (ACCESS4_MODIFY | ACCESS4_EXTEND); 1344 } 1345 1346 if (checkwriteperm && 1347 (args->access & ACCESS4_DELETE) && vp->v_type == VDIR) { 1348 error = VOP_ACCESS(vp, VWRITE, 0, cr, NULL); 1349 if (!error && (!is_system_labeled() || admin_low_client || 1350 blequal(clabel, slabel))) 1351 resp->access |= ACCESS4_DELETE; 1352 resp->supported |= ACCESS4_DELETE; 1353 } 1354 if (args->access & ACCESS4_EXECUTE && vp->v_type != VDIR) { 1355 error = VOP_ACCESS(vp, VEXEC, 0, cr, NULL); 1356 if (!error && !MANDLOCK(vp, va.va_mode) && 1357 (!is_system_labeled() || admin_low_client || 1358 bldominates(clabel, slabel))) 1359 resp->access |= ACCESS4_EXECUTE; 1360 resp->supported |= ACCESS4_EXECUTE; 1361 } 1362 1363 if (is_system_labeled() && !admin_low_client) 1364 label_rele(tslabel); 1365 1366 *cs->statusp = resp->status = NFS4_OK; 1367 out: 1368 DTRACE_NFSV4_2(op__access__done, struct compound_state *, cs, 1369 ACCESS4res *, resp); 1370 } 1371 1372 /* ARGSUSED */ 1373 static void 1374 rfs4_op_commit(nfs_argop4 *argop, nfs_resop4 *resop, struct svc_req *req, 1375 struct compound_state *cs) 1376 { 1377 COMMIT4args *args = &argop->nfs_argop4_u.opcommit; 1378 COMMIT4res *resp = &resop->nfs_resop4_u.opcommit; 1379 int error; 1380 vnode_t *vp = cs->vp; 1381 cred_t *cr = cs->cr; 1382 vattr_t va; 1383 1384 DTRACE_NFSV4_2(op__commit__start, struct compound_state *, cs, 1385 COMMIT4args *, args); 1386 1387 if (vp == NULL) { 1388 *cs->statusp = resp->status = NFS4ERR_NOFILEHANDLE; 1389 goto out; 1390 } 1391 if (cs->access == CS_ACCESS_DENIED) { 1392 *cs->statusp = resp->status = NFS4ERR_ACCESS; 1393 goto out; 1394 } 1395 1396 if (args->offset + args->count < args->offset) { 1397 *cs->statusp = resp->status = NFS4ERR_INVAL; 1398 goto out; 1399 } 1400 1401 va.va_mask = AT_UID; 1402 error = VOP_GETATTR(vp, &va, 0, cr, NULL); 1403 1404 /* 1405 * If we can't get the attributes, then we can't do the 1406 * right access checking. So, we'll fail the request. 1407 */ 1408 if (error) { 1409 *cs->statusp = resp->status = puterrno4(error); 1410 goto out; 1411 } 1412 if (rdonly4(cs->exi, cs->vp, req)) { 1413 *cs->statusp = resp->status = NFS4ERR_ROFS; 1414 goto out; 1415 } 1416 1417 if (vp->v_type != VREG) { 1418 if (vp->v_type == VDIR) 1419 resp->status = NFS4ERR_ISDIR; 1420 else 1421 resp->status = NFS4ERR_INVAL; 1422 *cs->statusp = resp->status; 1423 goto out; 1424 } 1425 1426 if (crgetuid(cr) != va.va_uid && 1427 (error = VOP_ACCESS(vp, VWRITE, 0, cs->cr, NULL))) { 1428 *cs->statusp = resp->status = puterrno4(error); 1429 goto out; 1430 } 1431 1432 error = VOP_FSYNC(vp, FSYNC, cr, NULL); 1433 1434 if (error) { 1435 *cs->statusp = resp->status = puterrno4(error); 1436 goto out; 1437 } 1438 1439 *cs->statusp = resp->status = NFS4_OK; 1440 resp->writeverf = Write4verf; 1441 out: 1442 DTRACE_NFSV4_2(op__commit__done, struct compound_state *, cs, 1443 COMMIT4res *, resp); 1444 } 1445 1446 /* 1447 * rfs4_op_mknod is called from rfs4_op_create after all initial verification 1448 * was completed. It does the nfsv4 create for special files. 1449 */ 1450 /* ARGSUSED */ 1451 static vnode_t * 1452 do_rfs4_op_mknod(CREATE4args *args, CREATE4res *resp, struct svc_req *req, 1453 struct compound_state *cs, vattr_t *vap, char *nm) 1454 { 1455 int error; 1456 cred_t *cr = cs->cr; 1457 vnode_t *dvp = cs->vp; 1458 vnode_t *vp = NULL; 1459 int mode; 1460 enum vcexcl excl; 1461 1462 switch (args->type) { 1463 case NF4CHR: 1464 case NF4BLK: 1465 if (secpolicy_sys_devices(cr) != 0) { 1466 *cs->statusp = resp->status = NFS4ERR_PERM; 1467 return (NULL); 1468 } 1469 if (args->type == NF4CHR) 1470 vap->va_type = VCHR; 1471 else 1472 vap->va_type = VBLK; 1473 vap->va_rdev = makedevice(args->ftype4_u.devdata.specdata1, 1474 args->ftype4_u.devdata.specdata2); 1475 vap->va_mask |= AT_RDEV; 1476 break; 1477 case NF4SOCK: 1478 vap->va_type = VSOCK; 1479 break; 1480 case NF4FIFO: 1481 vap->va_type = VFIFO; 1482 break; 1483 default: 1484 *cs->statusp = resp->status = NFS4ERR_BADTYPE; 1485 return (NULL); 1486 } 1487 1488 /* 1489 * Must specify the mode. 1490 */ 1491 if (!(vap->va_mask & AT_MODE)) { 1492 *cs->statusp = resp->status = NFS4ERR_INVAL; 1493 return (NULL); 1494 } 1495 1496 excl = EXCL; 1497 1498 mode = 0; 1499 1500 error = VOP_CREATE(dvp, nm, vap, excl, mode, &vp, cr, 0, NULL, NULL); 1501 if (error) { 1502 *cs->statusp = resp->status = puterrno4(error); 1503 return (NULL); 1504 } 1505 return (vp); 1506 } 1507 1508 /* 1509 * nfsv4 create is used to create non-regular files. For regular files, 1510 * use nfsv4 open. 1511 */ 1512 /* ARGSUSED */ 1513 static void 1514 rfs4_op_create(nfs_argop4 *argop, nfs_resop4 *resop, struct svc_req *req, 1515 struct compound_state *cs) 1516 { 1517 CREATE4args *args = &argop->nfs_argop4_u.opcreate; 1518 CREATE4res *resp = &resop->nfs_resop4_u.opcreate; 1519 int error; 1520 struct vattr bva, iva, iva2, ava, *vap; 1521 cred_t *cr = cs->cr; 1522 vnode_t *dvp = cs->vp; 1523 vnode_t *vp = NULL; 1524 vnode_t *realvp; 1525 char *nm, *lnm; 1526 uint_t len, llen; 1527 int syncval = 0; 1528 struct nfs4_svgetit_arg sarg; 1529 struct nfs4_ntov_table ntov; 1530 struct statvfs64 sb; 1531 nfsstat4 status; 1532 struct sockaddr *ca; 1533 char *name = NULL; 1534 char *lname = NULL; 1535 1536 DTRACE_NFSV4_2(op__create__start, struct compound_state *, cs, 1537 CREATE4args *, args); 1538 1539 resp->attrset = 0; 1540 1541 if (dvp == NULL) { 1542 *cs->statusp = resp->status = NFS4ERR_NOFILEHANDLE; 1543 goto out; 1544 } 1545 1546 /* 1547 * If there is an unshared filesystem mounted on this vnode, 1548 * do not allow to create an object in this directory. 1549 */ 1550 if (vn_ismntpt(dvp)) { 1551 *cs->statusp = resp->status = NFS4ERR_ACCESS; 1552 goto out; 1553 } 1554 1555 /* Verify that type is correct */ 1556 switch (args->type) { 1557 case NF4LNK: 1558 case NF4BLK: 1559 case NF4CHR: 1560 case NF4SOCK: 1561 case NF4FIFO: 1562 case NF4DIR: 1563 break; 1564 default: 1565 *cs->statusp = resp->status = NFS4ERR_BADTYPE; 1566 goto out; 1567 }; 1568 1569 if (cs->access == CS_ACCESS_DENIED) { 1570 *cs->statusp = resp->status = NFS4ERR_ACCESS; 1571 goto out; 1572 } 1573 if (dvp->v_type != VDIR) { 1574 *cs->statusp = resp->status = NFS4ERR_NOTDIR; 1575 goto out; 1576 } 1577 status = utf8_dir_verify(&args->objname); 1578 if (status != NFS4_OK) { 1579 *cs->statusp = resp->status = status; 1580 goto out; 1581 } 1582 1583 if (rdonly4(cs->exi, cs->vp, req)) { 1584 *cs->statusp = resp->status = NFS4ERR_ROFS; 1585 goto out; 1586 } 1587 1588 /* 1589 * Name of newly created object 1590 */ 1591 nm = utf8_to_fn(&args->objname, &len, NULL); 1592 if (nm == NULL) { 1593 *cs->statusp = resp->status = NFS4ERR_INVAL; 1594 goto out; 1595 } 1596 1597 if (len > MAXNAMELEN) { 1598 *cs->statusp = resp->status = NFS4ERR_NAMETOOLONG; 1599 kmem_free(nm, len); 1600 goto out; 1601 } 1602 1603 ca = (struct sockaddr *)svc_getrpccaller(req->rq_xprt)->buf; 1604 name = nfscmd_convname(ca, cs->exi, nm, NFSCMD_CONV_INBOUND, 1605 MAXPATHLEN + 1); 1606 1607 if (name == NULL) { 1608 *cs->statusp = resp->status = NFS4ERR_INVAL; 1609 kmem_free(nm, len); 1610 goto out; 1611 } 1612 1613 resp->attrset = 0; 1614 1615 sarg.sbp = &sb; 1616 sarg.is_referral = B_FALSE; 1617 nfs4_ntov_table_init(&ntov); 1618 1619 status = do_rfs4_set_attrs(&resp->attrset, 1620 &args->createattrs, cs, &sarg, &ntov, NFS4ATTR_SETIT); 1621 1622 if (sarg.vap->va_mask == 0 && status == NFS4_OK) 1623 status = NFS4ERR_INVAL; 1624 1625 if (status != NFS4_OK) { 1626 *cs->statusp = resp->status = status; 1627 if (name != nm) 1628 kmem_free(name, MAXPATHLEN + 1); 1629 kmem_free(nm, len); 1630 nfs4_ntov_table_free(&ntov, &sarg); 1631 resp->attrset = 0; 1632 goto out; 1633 } 1634 1635 /* Get "before" change value */ 1636 bva.va_mask = AT_CTIME|AT_SEQ|AT_MODE; 1637 error = VOP_GETATTR(dvp, &bva, 0, cr, NULL); 1638 if (error) { 1639 *cs->statusp = resp->status = puterrno4(error); 1640 if (name != nm) 1641 kmem_free(name, MAXPATHLEN + 1); 1642 kmem_free(nm, len); 1643 nfs4_ntov_table_free(&ntov, &sarg); 1644 resp->attrset = 0; 1645 goto out; 1646 } 1647 NFS4_SET_FATTR4_CHANGE(resp->cinfo.before, bva.va_ctime) 1648 1649 vap = sarg.vap; 1650 1651 /* 1652 * Set the default initial values for attributes when the parent 1653 * directory does not have the VSUID/VSGID bit set and they have 1654 * not been specified in createattrs. 1655 */ 1656 if (!(bva.va_mode & VSUID) && (vap->va_mask & AT_UID) == 0) { 1657 vap->va_uid = crgetuid(cr); 1658 vap->va_mask |= AT_UID; 1659 } 1660 if (!(bva.va_mode & VSGID) && (vap->va_mask & AT_GID) == 0) { 1661 vap->va_gid = crgetgid(cr); 1662 vap->va_mask |= AT_GID; 1663 } 1664 1665 vap->va_mask |= AT_TYPE; 1666 switch (args->type) { 1667 case NF4DIR: 1668 vap->va_type = VDIR; 1669 if ((vap->va_mask & AT_MODE) == 0) { 1670 vap->va_mode = 0700; /* default: owner rwx only */ 1671 vap->va_mask |= AT_MODE; 1672 } 1673 error = VOP_MKDIR(dvp, name, vap, &vp, cr, NULL, 0, NULL); 1674 if (error) 1675 break; 1676 1677 /* 1678 * Get the initial "after" sequence number, if it fails, 1679 * set to zero 1680 */ 1681 iva.va_mask = AT_SEQ; 1682 if (VOP_GETATTR(dvp, &iva, 0, cs->cr, NULL)) 1683 iva.va_seq = 0; 1684 break; 1685 case NF4LNK: 1686 vap->va_type = VLNK; 1687 if ((vap->va_mask & AT_MODE) == 0) { 1688 vap->va_mode = 0700; /* default: owner rwx only */ 1689 vap->va_mask |= AT_MODE; 1690 } 1691 1692 /* 1693 * symlink names must be treated as data 1694 */ 1695 lnm = utf8_to_str(&args->ftype4_u.linkdata, &llen, NULL); 1696 1697 if (lnm == NULL) { 1698 *cs->statusp = resp->status = NFS4ERR_INVAL; 1699 if (name != nm) 1700 kmem_free(name, MAXPATHLEN + 1); 1701 kmem_free(nm, len); 1702 nfs4_ntov_table_free(&ntov, &sarg); 1703 resp->attrset = 0; 1704 goto out; 1705 } 1706 1707 if (llen > MAXPATHLEN) { 1708 *cs->statusp = resp->status = NFS4ERR_NAMETOOLONG; 1709 if (name != nm) 1710 kmem_free(name, MAXPATHLEN + 1); 1711 kmem_free(nm, len); 1712 kmem_free(lnm, llen); 1713 nfs4_ntov_table_free(&ntov, &sarg); 1714 resp->attrset = 0; 1715 goto out; 1716 } 1717 1718 lname = nfscmd_convname(ca, cs->exi, lnm, 1719 NFSCMD_CONV_INBOUND, MAXPATHLEN + 1); 1720 1721 if (lname == NULL) { 1722 *cs->statusp = resp->status = NFS4ERR_SERVERFAULT; 1723 if (name != nm) 1724 kmem_free(name, MAXPATHLEN + 1); 1725 kmem_free(nm, len); 1726 kmem_free(lnm, llen); 1727 nfs4_ntov_table_free(&ntov, &sarg); 1728 resp->attrset = 0; 1729 goto out; 1730 } 1731 1732 error = VOP_SYMLINK(dvp, name, vap, lname, cr, NULL, 0); 1733 if (lname != lnm) 1734 kmem_free(lname, MAXPATHLEN + 1); 1735 kmem_free(lnm, llen); 1736 if (error) 1737 break; 1738 1739 /* 1740 * Get the initial "after" sequence number, if it fails, 1741 * set to zero 1742 */ 1743 iva.va_mask = AT_SEQ; 1744 if (VOP_GETATTR(dvp, &iva, 0, cs->cr, NULL)) 1745 iva.va_seq = 0; 1746 1747 error = VOP_LOOKUP(dvp, name, &vp, NULL, 0, NULL, cr, 1748 NULL, NULL, NULL); 1749 if (error) 1750 break; 1751 1752 /* 1753 * va_seq is not safe over VOP calls, check it again 1754 * if it has changed zero out iva to force atomic = FALSE. 1755 */ 1756 iva2.va_mask = AT_SEQ; 1757 if (VOP_GETATTR(dvp, &iva2, 0, cs->cr, NULL) || 1758 iva2.va_seq != iva.va_seq) 1759 iva.va_seq = 0; 1760 break; 1761 default: 1762 /* 1763 * probably a special file. 1764 */ 1765 if ((vap->va_mask & AT_MODE) == 0) { 1766 vap->va_mode = 0600; /* default: owner rw only */ 1767 vap->va_mask |= AT_MODE; 1768 } 1769 syncval = FNODSYNC; 1770 /* 1771 * We know this will only generate one VOP call 1772 */ 1773 vp = do_rfs4_op_mknod(args, resp, req, cs, vap, name); 1774 1775 if (vp == NULL) { 1776 if (name != nm) 1777 kmem_free(name, MAXPATHLEN + 1); 1778 kmem_free(nm, len); 1779 nfs4_ntov_table_free(&ntov, &sarg); 1780 resp->attrset = 0; 1781 goto out; 1782 } 1783 1784 /* 1785 * Get the initial "after" sequence number, if it fails, 1786 * set to zero 1787 */ 1788 iva.va_mask = AT_SEQ; 1789 if (VOP_GETATTR(dvp, &iva, 0, cs->cr, NULL)) 1790 iva.va_seq = 0; 1791 1792 break; 1793 } 1794 if (name != nm) 1795 kmem_free(name, MAXPATHLEN + 1); 1796 kmem_free(nm, len); 1797 1798 if (error) { 1799 *cs->statusp = resp->status = puterrno4(error); 1800 } 1801 1802 /* 1803 * Force modified data and metadata out to stable storage. 1804 */ 1805 (void) VOP_FSYNC(dvp, 0, cr, NULL); 1806 1807 if (resp->status != NFS4_OK) { 1808 if (vp != NULL) 1809 VN_RELE(vp); 1810 nfs4_ntov_table_free(&ntov, &sarg); 1811 resp->attrset = 0; 1812 goto out; 1813 } 1814 1815 /* 1816 * Finish setup of cinfo response, "before" value already set. 1817 * Get "after" change value, if it fails, simply return the 1818 * before value. 1819 */ 1820 ava.va_mask = AT_CTIME|AT_SEQ; 1821 if (VOP_GETATTR(dvp, &ava, 0, cr, NULL)) { 1822 ava.va_ctime = bva.va_ctime; 1823 ava.va_seq = 0; 1824 } 1825 NFS4_SET_FATTR4_CHANGE(resp->cinfo.after, ava.va_ctime); 1826 1827 /* 1828 * True verification that object was created with correct 1829 * attrs is impossible. The attrs could have been changed 1830 * immediately after object creation. If attributes did 1831 * not verify, the only recourse for the server is to 1832 * destroy the object. Maybe if some attrs (like gid) 1833 * are set incorrectly, the object should be destroyed; 1834 * however, seems bad as a default policy. Do we really 1835 * want to destroy an object over one of the times not 1836 * verifying correctly? For these reasons, the server 1837 * currently sets bits in attrset for createattrs 1838 * that were set; however, no verification is done. 1839 * 1840 * vmask_to_nmask accounts for vattr bits set on create 1841 * [do_rfs4_set_attrs() only sets resp bits for 1842 * non-vattr/vfs bits.] 1843 * Mask off any bits set by default so as not to return 1844 * more attrset bits than were requested in createattrs 1845 */ 1846 nfs4_vmask_to_nmask(sarg.vap->va_mask, &resp->attrset); 1847 resp->attrset &= args->createattrs.attrmask; 1848 nfs4_ntov_table_free(&ntov, &sarg); 1849 1850 error = makefh4(&cs->fh, vp, cs->exi); 1851 if (error) { 1852 *cs->statusp = resp->status = puterrno4(error); 1853 } 1854 1855 /* 1856 * The cinfo.atomic = TRUE only if we got no errors, we have 1857 * non-zero va_seq's, and it has incremented by exactly one 1858 * during the creation and it didn't change during the VOP_LOOKUP 1859 * or VOP_FSYNC. 1860 */ 1861 if (!error && bva.va_seq && iva.va_seq && ava.va_seq && 1862 iva.va_seq == (bva.va_seq + 1) && iva.va_seq == ava.va_seq) 1863 resp->cinfo.atomic = TRUE; 1864 else 1865 resp->cinfo.atomic = FALSE; 1866 1867 /* 1868 * Force modified metadata out to stable storage. 1869 * 1870 * if a underlying vp exists, pass it to VOP_FSYNC 1871 */ 1872 if (VOP_REALVP(vp, &realvp, NULL) == 0) 1873 (void) VOP_FSYNC(realvp, syncval, cr, NULL); 1874 else 1875 (void) VOP_FSYNC(vp, syncval, cr, NULL); 1876 1877 if (resp->status != NFS4_OK) { 1878 VN_RELE(vp); 1879 goto out; 1880 } 1881 if (cs->vp) 1882 VN_RELE(cs->vp); 1883 1884 cs->vp = vp; 1885 *cs->statusp = resp->status = NFS4_OK; 1886 out: 1887 DTRACE_NFSV4_2(op__create__done, struct compound_state *, cs, 1888 CREATE4res *, resp); 1889 } 1890 1891 /*ARGSUSED*/ 1892 static void 1893 rfs4_op_delegpurge(nfs_argop4 *argop, nfs_resop4 *resop, struct svc_req *req, 1894 struct compound_state *cs) 1895 { 1896 DTRACE_NFSV4_2(op__delegpurge__start, struct compound_state *, cs, 1897 DELEGPURGE4args *, &argop->nfs_argop4_u.opdelegpurge); 1898 1899 rfs4_op_inval(argop, resop, req, cs); 1900 1901 DTRACE_NFSV4_2(op__delegpurge__done, struct compound_state *, cs, 1902 DELEGPURGE4res *, &resop->nfs_resop4_u.opdelegpurge); 1903 } 1904 1905 /*ARGSUSED*/ 1906 static void 1907 rfs4_op_delegreturn(nfs_argop4 *argop, nfs_resop4 *resop, struct svc_req *req, 1908 struct compound_state *cs) 1909 { 1910 DELEGRETURN4args *args = &argop->nfs_argop4_u.opdelegreturn; 1911 DELEGRETURN4res *resp = &resop->nfs_resop4_u.opdelegreturn; 1912 rfs4_deleg_state_t *dsp; 1913 nfsstat4 status; 1914 1915 DTRACE_NFSV4_2(op__delegreturn__start, struct compound_state *, cs, 1916 DELEGRETURN4args *, args); 1917 1918 status = rfs4_get_deleg_state(&args->deleg_stateid, &dsp); 1919 resp->status = *cs->statusp = status; 1920 if (status != NFS4_OK) 1921 goto out; 1922 1923 /* Ensure specified filehandle matches */ 1924 if (cs->vp != dsp->rds_finfo->rf_vp) { 1925 resp->status = *cs->statusp = NFS4ERR_BAD_STATEID; 1926 } else 1927 rfs4_return_deleg(dsp, FALSE); 1928 1929 rfs4_update_lease(dsp->rds_client); 1930 1931 rfs4_deleg_state_rele(dsp); 1932 out: 1933 DTRACE_NFSV4_2(op__delegreturn__done, struct compound_state *, cs, 1934 DELEGRETURN4res *, resp); 1935 } 1936 1937 /* 1938 * Check to see if a given "flavor" is an explicitly shared flavor. 1939 * The assumption of this routine is the "flavor" is already a valid 1940 * flavor in the secinfo list of "exi". 1941 * 1942 * e.g. 1943 * # share -o sec=flavor1 /export 1944 * # share -o sec=flavor2 /export/home 1945 * 1946 * flavor2 is not an explicitly shared flavor for /export, 1947 * however it is in the secinfo list for /export thru the 1948 * server namespace setup. 1949 */ 1950 int 1951 is_exported_sec(int flavor, struct exportinfo *exi) 1952 { 1953 int i; 1954 struct secinfo *sp; 1955 1956 sp = exi->exi_export.ex_secinfo; 1957 for (i = 0; i < exi->exi_export.ex_seccnt; i++) { 1958 if (flavor == sp[i].s_secinfo.sc_nfsnum || 1959 sp[i].s_secinfo.sc_nfsnum == AUTH_NONE) { 1960 return (SEC_REF_EXPORTED(&sp[i])); 1961 } 1962 } 1963 1964 /* Should not reach this point based on the assumption */ 1965 return (0); 1966 } 1967 1968 /* 1969 * Check if the security flavor used in the request matches what is 1970 * required at the export point or at the root pseudo node (exi_root). 1971 * 1972 * returns 1 if there's a match or if exported with AUTH_NONE; 0 otherwise. 1973 * 1974 */ 1975 static int 1976 secinfo_match_or_authnone(struct compound_state *cs) 1977 { 1978 int i; 1979 struct secinfo *sp; 1980 1981 /* 1982 * Check cs->nfsflavor (from the request) against 1983 * the current export data in cs->exi. 1984 */ 1985 sp = cs->exi->exi_export.ex_secinfo; 1986 for (i = 0; i < cs->exi->exi_export.ex_seccnt; i++) { 1987 if (cs->nfsflavor == sp[i].s_secinfo.sc_nfsnum || 1988 sp[i].s_secinfo.sc_nfsnum == AUTH_NONE) 1989 return (1); 1990 } 1991 1992 return (0); 1993 } 1994 1995 /* 1996 * Check the access authority for the client and return the correct error. 1997 */ 1998 nfsstat4 1999 call_checkauth4(struct compound_state *cs, struct svc_req *req) 2000 { 2001 int authres; 2002 2003 /* 2004 * First, check if the security flavor used in the request 2005 * are among the flavors set in the server namespace. 2006 */ 2007 if (!secinfo_match_or_authnone(cs)) { 2008 *cs->statusp = NFS4ERR_WRONGSEC; 2009 return (*cs->statusp); 2010 } 2011 2012 authres = checkauth4(cs, req); 2013 2014 if (authres > 0) { 2015 *cs->statusp = NFS4_OK; 2016 if (! (cs->access & CS_ACCESS_LIMITED)) 2017 cs->access = CS_ACCESS_OK; 2018 } else if (authres == 0) { 2019 *cs->statusp = NFS4ERR_ACCESS; 2020 } else if (authres == -2) { 2021 *cs->statusp = NFS4ERR_WRONGSEC; 2022 } else { 2023 *cs->statusp = NFS4ERR_DELAY; 2024 } 2025 return (*cs->statusp); 2026 } 2027 2028 /* 2029 * bitmap4_to_attrmask is called by getattr and readdir. 2030 * It sets up the vattr mask and determines whether vfsstat call is needed 2031 * based on the input bitmap. 2032 * Returns nfsv4 status. 2033 */ 2034 static nfsstat4 2035 bitmap4_to_attrmask(bitmap4 breq, struct nfs4_svgetit_arg *sargp) 2036 { 2037 int i; 2038 uint_t va_mask; 2039 struct statvfs64 *sbp = sargp->sbp; 2040 2041 sargp->sbp = NULL; 2042 sargp->flag = 0; 2043 sargp->rdattr_error = NFS4_OK; 2044 sargp->mntdfid_set = FALSE; 2045 if (sargp->cs->vp) 2046 sargp->xattr = get_fh4_flag(&sargp->cs->fh, 2047 FH4_ATTRDIR | FH4_NAMEDATTR); 2048 else 2049 sargp->xattr = 0; 2050 2051 /* 2052 * Set rdattr_error_req to true if return error per 2053 * failed entry rather than fail the readdir. 2054 */ 2055 if (breq & FATTR4_RDATTR_ERROR_MASK) 2056 sargp->rdattr_error_req = 1; 2057 else 2058 sargp->rdattr_error_req = 0; 2059 2060 /* 2061 * generate the va_mask 2062 * Handle the easy cases first 2063 */ 2064 switch (breq) { 2065 case NFS4_NTOV_ATTR_MASK: 2066 sargp->vap->va_mask = NFS4_NTOV_ATTR_AT_MASK; 2067 return (NFS4_OK); 2068 2069 case NFS4_FS_ATTR_MASK: 2070 sargp->vap->va_mask = NFS4_FS_ATTR_AT_MASK; 2071 sargp->sbp = sbp; 2072 return (NFS4_OK); 2073 2074 case NFS4_NTOV_ATTR_CACHE_MASK: 2075 sargp->vap->va_mask = NFS4_NTOV_ATTR_CACHE_AT_MASK; 2076 return (NFS4_OK); 2077 2078 case FATTR4_LEASE_TIME_MASK: 2079 sargp->vap->va_mask = 0; 2080 return (NFS4_OK); 2081 2082 default: 2083 va_mask = 0; 2084 for (i = 0; i < nfs4_ntov_map_size; i++) { 2085 if ((breq & nfs4_ntov_map[i].fbit) && 2086 nfs4_ntov_map[i].vbit) 2087 va_mask |= nfs4_ntov_map[i].vbit; 2088 } 2089 2090 /* 2091 * Check is vfsstat is needed 2092 */ 2093 if (breq & NFS4_FS_ATTR_MASK) 2094 sargp->sbp = sbp; 2095 2096 sargp->vap->va_mask = va_mask; 2097 return (NFS4_OK); 2098 } 2099 /* NOTREACHED */ 2100 } 2101 2102 /* 2103 * bitmap4_get_sysattrs is called by getattr and readdir. 2104 * It calls both VOP_GETATTR and VFS_STATVFS calls to get the attrs. 2105 * Returns nfsv4 status. 2106 */ 2107 static nfsstat4 2108 bitmap4_get_sysattrs(struct nfs4_svgetit_arg *sargp) 2109 { 2110 int error; 2111 struct compound_state *cs = sargp->cs; 2112 vnode_t *vp = cs->vp; 2113 2114 if (sargp->sbp != NULL) { 2115 if (error = VFS_STATVFS(vp->v_vfsp, sargp->sbp)) { 2116 sargp->sbp = NULL; /* to identify error */ 2117 return (puterrno4(error)); 2118 } 2119 } 2120 2121 return (rfs4_vop_getattr(vp, sargp->vap, 0, cs->cr)); 2122 } 2123 2124 static void 2125 nfs4_ntov_table_init(struct nfs4_ntov_table *ntovp) 2126 { 2127 ntovp->na = kmem_zalloc(sizeof (union nfs4_attr_u) * nfs4_ntov_map_size, 2128 KM_SLEEP); 2129 ntovp->attrcnt = 0; 2130 ntovp->vfsstat = FALSE; 2131 } 2132 2133 static void 2134 nfs4_ntov_table_free(struct nfs4_ntov_table *ntovp, 2135 struct nfs4_svgetit_arg *sargp) 2136 { 2137 int i; 2138 union nfs4_attr_u *na; 2139 uint8_t *amap; 2140 2141 /* 2142 * XXX Should do the same checks for whether the bit is set 2143 */ 2144 for (i = 0, na = ntovp->na, amap = ntovp->amap; 2145 i < ntovp->attrcnt; i++, na++, amap++) { 2146 (void) (*nfs4_ntov_map[*amap].sv_getit)( 2147 NFS4ATTR_FREEIT, sargp, na); 2148 } 2149 if ((sargp->op == NFS4ATTR_SETIT) || (sargp->op == NFS4ATTR_VERIT)) { 2150 /* 2151 * xdr_free for getattr will be done later 2152 */ 2153 for (i = 0, na = ntovp->na, amap = ntovp->amap; 2154 i < ntovp->attrcnt; i++, na++, amap++) { 2155 xdr_free(nfs4_ntov_map[*amap].xfunc, (caddr_t)na); 2156 } 2157 } 2158 kmem_free(ntovp->na, sizeof (union nfs4_attr_u) * nfs4_ntov_map_size); 2159 } 2160 2161 /* 2162 * do_rfs4_op_getattr gets the system attrs and converts into fattr4. 2163 */ 2164 static nfsstat4 2165 do_rfs4_op_getattr(bitmap4 breq, fattr4 *fattrp, 2166 struct nfs4_svgetit_arg *sargp) 2167 { 2168 int error = 0; 2169 int i, k; 2170 struct nfs4_ntov_table ntov; 2171 XDR xdr; 2172 ulong_t xdr_size; 2173 char *xdr_attrs; 2174 nfsstat4 status = NFS4_OK; 2175 nfsstat4 prev_rdattr_error = sargp->rdattr_error; 2176 union nfs4_attr_u *na; 2177 uint8_t *amap; 2178 2179 sargp->op = NFS4ATTR_GETIT; 2180 sargp->flag = 0; 2181 2182 fattrp->attrmask = 0; 2183 /* if no bits requested, then return empty fattr4 */ 2184 if (breq == 0) { 2185 fattrp->attrlist4_len = 0; 2186 fattrp->attrlist4 = NULL; 2187 return (NFS4_OK); 2188 } 2189 2190 /* 2191 * return NFS4ERR_INVAL when client requests write-only attrs 2192 */ 2193 if (breq & (FATTR4_TIME_ACCESS_SET_MASK | FATTR4_TIME_MODIFY_SET_MASK)) 2194 return (NFS4ERR_INVAL); 2195 2196 nfs4_ntov_table_init(&ntov); 2197 na = ntov.na; 2198 amap = ntov.amap; 2199 2200 /* 2201 * Now loop to get or verify the attrs 2202 */ 2203 for (i = 0; i < nfs4_ntov_map_size; i++) { 2204 if (breq & nfs4_ntov_map[i].fbit) { 2205 if ((*nfs4_ntov_map[i].sv_getit)( 2206 NFS4ATTR_SUPPORTED, sargp, NULL) == 0) { 2207 2208 error = (*nfs4_ntov_map[i].sv_getit)( 2209 NFS4ATTR_GETIT, sargp, na); 2210 2211 /* 2212 * Possible error values: 2213 * >0 if sv_getit failed to 2214 * get the attr; 0 if succeeded; 2215 * <0 if rdattr_error and the 2216 * attribute cannot be returned. 2217 */ 2218 if (error && !(sargp->rdattr_error_req)) 2219 goto done; 2220 /* 2221 * If error then just for entry 2222 */ 2223 if (error == 0) { 2224 fattrp->attrmask |= 2225 nfs4_ntov_map[i].fbit; 2226 *amap++ = 2227 (uint8_t)nfs4_ntov_map[i].nval; 2228 na++; 2229 (ntov.attrcnt)++; 2230 } else if ((error > 0) && 2231 (sargp->rdattr_error == NFS4_OK)) { 2232 sargp->rdattr_error = puterrno4(error); 2233 } 2234 error = 0; 2235 } 2236 } 2237 } 2238 2239 /* 2240 * If rdattr_error was set after the return value for it was assigned, 2241 * update it. 2242 */ 2243 if (prev_rdattr_error != sargp->rdattr_error) { 2244 na = ntov.na; 2245 amap = ntov.amap; 2246 for (i = 0; i < ntov.attrcnt; i++, na++, amap++) { 2247 k = *amap; 2248 if (k < FATTR4_RDATTR_ERROR) { 2249 continue; 2250 } 2251 if ((k == FATTR4_RDATTR_ERROR) && 2252 ((*nfs4_ntov_map[k].sv_getit)( 2253 NFS4ATTR_SUPPORTED, sargp, NULL) == 0)) { 2254 2255 (void) (*nfs4_ntov_map[k].sv_getit)( 2256 NFS4ATTR_GETIT, sargp, na); 2257 } 2258 break; 2259 } 2260 } 2261 2262 xdr_size = 0; 2263 na = ntov.na; 2264 amap = ntov.amap; 2265 for (i = 0; i < ntov.attrcnt; i++, na++, amap++) { 2266 xdr_size += xdr_sizeof(nfs4_ntov_map[*amap].xfunc, na); 2267 } 2268 2269 fattrp->attrlist4_len = xdr_size; 2270 if (xdr_size) { 2271 /* freed by rfs4_op_getattr_free() */ 2272 fattrp->attrlist4 = xdr_attrs = kmem_zalloc(xdr_size, KM_SLEEP); 2273 2274 xdrmem_create(&xdr, xdr_attrs, xdr_size, XDR_ENCODE); 2275 2276 na = ntov.na; 2277 amap = ntov.amap; 2278 for (i = 0; i < ntov.attrcnt; i++, na++, amap++) { 2279 if (!(*nfs4_ntov_map[*amap].xfunc)(&xdr, na)) { 2280 DTRACE_PROBE1(nfss__e__getattr4_encfail, 2281 int, *amap); 2282 status = NFS4ERR_SERVERFAULT; 2283 break; 2284 } 2285 } 2286 /* xdrmem_destroy(&xdrs); */ /* NO-OP */ 2287 } else { 2288 fattrp->attrlist4 = NULL; 2289 } 2290 done: 2291 2292 nfs4_ntov_table_free(&ntov, sargp); 2293 2294 if (error != 0) 2295 status = puterrno4(error); 2296 2297 return (status); 2298 } 2299 2300 /* ARGSUSED */ 2301 static void 2302 rfs4_op_getattr(nfs_argop4 *argop, nfs_resop4 *resop, struct svc_req *req, 2303 struct compound_state *cs) 2304 { 2305 GETATTR4args *args = &argop->nfs_argop4_u.opgetattr; 2306 GETATTR4res *resp = &resop->nfs_resop4_u.opgetattr; 2307 struct nfs4_svgetit_arg sarg; 2308 struct statvfs64 sb; 2309 nfsstat4 status; 2310 2311 DTRACE_NFSV4_2(op__getattr__start, struct compound_state *, cs, 2312 GETATTR4args *, args); 2313 2314 if (cs->vp == NULL) { 2315 *cs->statusp = resp->status = NFS4ERR_NOFILEHANDLE; 2316 goto out; 2317 } 2318 2319 if (cs->access == CS_ACCESS_DENIED) { 2320 *cs->statusp = resp->status = NFS4ERR_ACCESS; 2321 goto out; 2322 } 2323 2324 sarg.sbp = &sb; 2325 sarg.cs = cs; 2326 sarg.is_referral = B_FALSE; 2327 2328 status = bitmap4_to_attrmask(args->attr_request, &sarg); 2329 if (status == NFS4_OK) { 2330 2331 status = bitmap4_get_sysattrs(&sarg); 2332 if (status == NFS4_OK) { 2333 2334 /* Is this a referral? */ 2335 if (vn_is_nfs_reparse(cs->vp, cs->cr)) { 2336 /* Older V4 Solaris client sees a link */ 2337 if (client_is_downrev(req)) 2338 sarg.vap->va_type = VLNK; 2339 else 2340 sarg.is_referral = B_TRUE; 2341 } 2342 2343 status = do_rfs4_op_getattr(args->attr_request, 2344 &resp->obj_attributes, &sarg); 2345 } 2346 } 2347 *cs->statusp = resp->status = status; 2348 out: 2349 DTRACE_NFSV4_2(op__getattr__done, struct compound_state *, cs, 2350 GETATTR4res *, resp); 2351 } 2352 2353 static void 2354 rfs4_op_getattr_free(nfs_resop4 *resop) 2355 { 2356 GETATTR4res *resp = &resop->nfs_resop4_u.opgetattr; 2357 2358 nfs4_fattr4_free(&resp->obj_attributes); 2359 } 2360 2361 /* ARGSUSED */ 2362 static void 2363 rfs4_op_getfh(nfs_argop4 *argop, nfs_resop4 *resop, struct svc_req *req, 2364 struct compound_state *cs) 2365 { 2366 GETFH4res *resp = &resop->nfs_resop4_u.opgetfh; 2367 2368 DTRACE_NFSV4_1(op__getfh__start, struct compound_state *, cs); 2369 2370 if (cs->vp == NULL) { 2371 *cs->statusp = resp->status = NFS4ERR_NOFILEHANDLE; 2372 goto out; 2373 } 2374 if (cs->access == CS_ACCESS_DENIED) { 2375 *cs->statusp = resp->status = NFS4ERR_ACCESS; 2376 goto out; 2377 } 2378 2379 /* check for reparse point at the share point */ 2380 if (cs->exi->exi_moved || vn_is_nfs_reparse(cs->exi->exi_vp, cs->cr)) { 2381 /* it's all bad */ 2382 cs->exi->exi_moved = 1; 2383 *cs->statusp = resp->status = NFS4ERR_MOVED; 2384 DTRACE_PROBE2(nfs4serv__func__referral__shared__moved, 2385 vnode_t *, cs->vp, char *, "rfs4_op_getfh"); 2386 return; 2387 } 2388 2389 /* check for reparse point at vp */ 2390 if (vn_is_nfs_reparse(cs->vp, cs->cr) && !client_is_downrev(req)) { 2391 /* it's not all bad */ 2392 *cs->statusp = resp->status = NFS4ERR_MOVED; 2393 DTRACE_PROBE2(nfs4serv__func__referral__moved, 2394 vnode_t *, cs->vp, char *, "rfs4_op_getfh"); 2395 return; 2396 } 2397 2398 resp->object.nfs_fh4_val = 2399 kmem_alloc(cs->fh.nfs_fh4_len, KM_SLEEP); 2400 nfs_fh4_copy(&cs->fh, &resp->object); 2401 *cs->statusp = resp->status = NFS4_OK; 2402 out: 2403 DTRACE_NFSV4_2(op__getfh__done, struct compound_state *, cs, 2404 GETFH4res *, resp); 2405 } 2406 2407 static void 2408 rfs4_op_getfh_free(nfs_resop4 *resop) 2409 { 2410 GETFH4res *resp = &resop->nfs_resop4_u.opgetfh; 2411 2412 if (resp->status == NFS4_OK && 2413 resp->object.nfs_fh4_val != NULL) { 2414 kmem_free(resp->object.nfs_fh4_val, resp->object.nfs_fh4_len); 2415 resp->object.nfs_fh4_val = NULL; 2416 resp->object.nfs_fh4_len = 0; 2417 } 2418 } 2419 2420 /* 2421 * illegal: args: void 2422 * res : status (NFS4ERR_OP_ILLEGAL) 2423 */ 2424 /* ARGSUSED */ 2425 static void 2426 rfs4_op_illegal(nfs_argop4 *argop, nfs_resop4 *resop, 2427 struct svc_req *req, struct compound_state *cs) 2428 { 2429 ILLEGAL4res *resp = &resop->nfs_resop4_u.opillegal; 2430 2431 resop->resop = OP_ILLEGAL; 2432 *cs->statusp = resp->status = NFS4ERR_OP_ILLEGAL; 2433 } 2434 2435 /* 2436 * link: args: SAVED_FH: file, CURRENT_FH: target directory 2437 * res: status. If success - CURRENT_FH unchanged, return change_info 2438 */ 2439 /* ARGSUSED */ 2440 static void 2441 rfs4_op_link(nfs_argop4 *argop, nfs_resop4 *resop, struct svc_req *req, 2442 struct compound_state *cs) 2443 { 2444 LINK4args *args = &argop->nfs_argop4_u.oplink; 2445 LINK4res *resp = &resop->nfs_resop4_u.oplink; 2446 int error; 2447 vnode_t *vp; 2448 vnode_t *dvp; 2449 struct vattr bdva, idva, adva; 2450 char *nm; 2451 uint_t len; 2452 struct sockaddr *ca; 2453 char *name = NULL; 2454 nfsstat4 status; 2455 2456 DTRACE_NFSV4_2(op__link__start, struct compound_state *, cs, 2457 LINK4args *, args); 2458 2459 /* SAVED_FH: source object */ 2460 vp = cs->saved_vp; 2461 if (vp == NULL) { 2462 *cs->statusp = resp->status = NFS4ERR_NOFILEHANDLE; 2463 goto out; 2464 } 2465 2466 /* CURRENT_FH: target directory */ 2467 dvp = cs->vp; 2468 if (dvp == NULL) { 2469 *cs->statusp = resp->status = NFS4ERR_NOFILEHANDLE; 2470 goto out; 2471 } 2472 2473 /* 2474 * If there is a non-shared filesystem mounted on this vnode, 2475 * do not allow to link any file in this directory. 2476 */ 2477 if (vn_ismntpt(dvp)) { 2478 *cs->statusp = resp->status = NFS4ERR_ACCESS; 2479 goto out; 2480 } 2481 2482 if (cs->access == CS_ACCESS_DENIED) { 2483 *cs->statusp = resp->status = NFS4ERR_ACCESS; 2484 goto out; 2485 } 2486 2487 /* Check source object's type validity */ 2488 if (vp->v_type == VDIR) { 2489 *cs->statusp = resp->status = NFS4ERR_ISDIR; 2490 goto out; 2491 } 2492 2493 /* Check target directory's type */ 2494 if (dvp->v_type != VDIR) { 2495 *cs->statusp = resp->status = NFS4ERR_NOTDIR; 2496 goto out; 2497 } 2498 2499 if (cs->saved_exi != cs->exi) { 2500 *cs->statusp = resp->status = NFS4ERR_XDEV; 2501 goto out; 2502 } 2503 2504 status = utf8_dir_verify(&args->newname); 2505 if (status != NFS4_OK) { 2506 *cs->statusp = resp->status = status; 2507 goto out; 2508 } 2509 2510 nm = utf8_to_fn(&args->newname, &len, NULL); 2511 if (nm == NULL) { 2512 *cs->statusp = resp->status = NFS4ERR_INVAL; 2513 goto out; 2514 } 2515 2516 if (len > MAXNAMELEN) { 2517 *cs->statusp = resp->status = NFS4ERR_NAMETOOLONG; 2518 kmem_free(nm, len); 2519 goto out; 2520 } 2521 2522 if (rdonly4(cs->exi, cs->vp, req)) { 2523 *cs->statusp = resp->status = NFS4ERR_ROFS; 2524 kmem_free(nm, len); 2525 goto out; 2526 } 2527 2528 /* Get "before" change value */ 2529 bdva.va_mask = AT_CTIME|AT_SEQ; 2530 error = VOP_GETATTR(dvp, &bdva, 0, cs->cr, NULL); 2531 if (error) { 2532 *cs->statusp = resp->status = puterrno4(error); 2533 kmem_free(nm, len); 2534 goto out; 2535 } 2536 2537 ca = (struct sockaddr *)svc_getrpccaller(req->rq_xprt)->buf; 2538 name = nfscmd_convname(ca, cs->exi, nm, NFSCMD_CONV_INBOUND, 2539 MAXPATHLEN + 1); 2540 2541 if (name == NULL) { 2542 *cs->statusp = resp->status = NFS4ERR_INVAL; 2543 kmem_free(nm, len); 2544 goto out; 2545 } 2546 2547 NFS4_SET_FATTR4_CHANGE(resp->cinfo.before, bdva.va_ctime) 2548 2549 error = VOP_LINK(dvp, vp, name, cs->cr, NULL, 0); 2550 2551 if (nm != name) 2552 kmem_free(name, MAXPATHLEN + 1); 2553 kmem_free(nm, len); 2554 2555 /* 2556 * Get the initial "after" sequence number, if it fails, set to zero 2557 */ 2558 idva.va_mask = AT_SEQ; 2559 if (VOP_GETATTR(dvp, &idva, 0, cs->cr, NULL)) 2560 idva.va_seq = 0; 2561 2562 /* 2563 * Force modified data and metadata out to stable storage. 2564 */ 2565 (void) VOP_FSYNC(vp, FNODSYNC, cs->cr, NULL); 2566 (void) VOP_FSYNC(dvp, 0, cs->cr, NULL); 2567 2568 if (error) { 2569 *cs->statusp = resp->status = puterrno4(error); 2570 goto out; 2571 } 2572 2573 /* 2574 * Get "after" change value, if it fails, simply return the 2575 * before value. 2576 */ 2577 adva.va_mask = AT_CTIME|AT_SEQ; 2578 if (VOP_GETATTR(dvp, &adva, 0, cs->cr, NULL)) { 2579 adva.va_ctime = bdva.va_ctime; 2580 adva.va_seq = 0; 2581 } 2582 2583 NFS4_SET_FATTR4_CHANGE(resp->cinfo.after, adva.va_ctime) 2584 2585 /* 2586 * The cinfo.atomic = TRUE only if we have 2587 * non-zero va_seq's, and it has incremented by exactly one 2588 * during the VOP_LINK and it didn't change during the VOP_FSYNC. 2589 */ 2590 if (bdva.va_seq && idva.va_seq && adva.va_seq && 2591 idva.va_seq == (bdva.va_seq + 1) && idva.va_seq == adva.va_seq) 2592 resp->cinfo.atomic = TRUE; 2593 else 2594 resp->cinfo.atomic = FALSE; 2595 2596 *cs->statusp = resp->status = NFS4_OK; 2597 out: 2598 DTRACE_NFSV4_2(op__link__done, struct compound_state *, cs, 2599 LINK4res *, resp); 2600 } 2601 2602 /* 2603 * Used by rfs4_op_lookup and rfs4_op_lookupp to do the actual work. 2604 */ 2605 2606 /* ARGSUSED */ 2607 static nfsstat4 2608 do_rfs4_op_lookup(char *nm, struct svc_req *req, struct compound_state *cs) 2609 { 2610 int error; 2611 int different_export = 0; 2612 vnode_t *vp, *tvp, *pre_tvp = NULL, *oldvp = NULL; 2613 struct exportinfo *exi = NULL, *pre_exi = NULL; 2614 nfsstat4 stat; 2615 fid_t fid; 2616 int attrdir, dotdot, walk; 2617 bool_t is_newvp = FALSE; 2618 2619 if (cs->vp->v_flag & V_XATTRDIR) { 2620 attrdir = 1; 2621 ASSERT(get_fh4_flag(&cs->fh, FH4_ATTRDIR)); 2622 } else { 2623 attrdir = 0; 2624 ASSERT(! get_fh4_flag(&cs->fh, FH4_ATTRDIR)); 2625 } 2626 2627 dotdot = (nm[0] == '.' && nm[1] == '.' && nm[2] == '\0'); 2628 2629 /* 2630 * If dotdotting, then need to check whether it's 2631 * above the root of a filesystem, or above an 2632 * export point. 2633 */ 2634 if (dotdot) { 2635 2636 /* 2637 * If dotdotting at the root of a filesystem, then 2638 * need to traverse back to the mounted-on filesystem 2639 * and do the dotdot lookup there. 2640 */ 2641 if (cs->vp->v_flag & VROOT) { 2642 2643 /* 2644 * If at the system root, then can 2645 * go up no further. 2646 */ 2647 if (VN_CMP(cs->vp, rootdir)) 2648 return (puterrno4(ENOENT)); 2649 2650 /* 2651 * Traverse back to the mounted-on filesystem 2652 */ 2653 cs->vp = untraverse(cs->vp); 2654 2655 /* 2656 * Set the different_export flag so we remember 2657 * to pick up a new exportinfo entry for 2658 * this new filesystem. 2659 */ 2660 different_export = 1; 2661 } else { 2662 2663 /* 2664 * If dotdotting above an export point then set 2665 * the different_export to get new export info. 2666 */ 2667 different_export = nfs_exported(cs->exi, cs->vp); 2668 } 2669 } 2670 2671 error = VOP_LOOKUP(cs->vp, nm, &vp, NULL, 0, NULL, cs->cr, 2672 NULL, NULL, NULL); 2673 if (error) 2674 return (puterrno4(error)); 2675 2676 /* 2677 * If the vnode is in a pseudo filesystem, check whether it is visible. 2678 * 2679 * XXX if the vnode is a symlink and it is not visible in 2680 * a pseudo filesystem, return ENOENT (not following symlink). 2681 * V4 client can not mount such symlink. This is a regression 2682 * from V2/V3. 2683 * 2684 * In the same exported filesystem, if the security flavor used 2685 * is not an explicitly shared flavor, limit the view to the visible 2686 * list entries only. This is not a WRONGSEC case because it's already 2687 * checked via PUTROOTFH/PUTPUBFH or PUTFH. 2688 */ 2689 if (!different_export && 2690 (PSEUDO(cs->exi) || ! is_exported_sec(cs->nfsflavor, cs->exi) || 2691 cs->access & CS_ACCESS_LIMITED)) { 2692 if (! nfs_visible(cs->exi, vp, &different_export)) { 2693 VN_RELE(vp); 2694 return (puterrno4(ENOENT)); 2695 } 2696 } 2697 2698 /* 2699 * If it's a mountpoint, then traverse it. 2700 */ 2701 if (vn_ismntpt(vp)) { 2702 pre_exi = cs->exi; /* save pre-traversed exportinfo */ 2703 pre_tvp = vp; /* save pre-traversed vnode */ 2704 2705 /* 2706 * hold pre_tvp to counteract rele by traverse. We will 2707 * need pre_tvp below if checkexport4 fails 2708 */ 2709 VN_HOLD(pre_tvp); 2710 tvp = vp; 2711 if ((error = traverse(&tvp)) != 0) { 2712 VN_RELE(vp); 2713 VN_RELE(pre_tvp); 2714 return (puterrno4(error)); 2715 } 2716 vp = tvp; 2717 different_export = 1; 2718 } else if (vp->v_vfsp != cs->vp->v_vfsp) { 2719 /* 2720 * The vfsp comparison is to handle the case where 2721 * a LOFS mount is shared. lo_lookup traverses mount points, 2722 * and NFS is unaware of local fs transistions because 2723 * v_vfsmountedhere isn't set. For this special LOFS case, 2724 * the dir and the obj returned by lookup will have different 2725 * vfs ptrs. 2726 */ 2727 different_export = 1; 2728 } 2729 2730 if (different_export) { 2731 2732 bzero(&fid, sizeof (fid)); 2733 fid.fid_len = MAXFIDSZ; 2734 error = vop_fid_pseudo(vp, &fid); 2735 if (error) { 2736 VN_RELE(vp); 2737 if (pre_tvp) 2738 VN_RELE(pre_tvp); 2739 return (puterrno4(error)); 2740 } 2741 2742 if (dotdot) 2743 exi = nfs_vptoexi(NULL, vp, cs->cr, &walk, NULL, TRUE); 2744 else 2745 exi = checkexport4(&vp->v_vfsp->vfs_fsid, &fid, vp); 2746 2747 if (exi == NULL) { 2748 if (pre_tvp) { 2749 /* 2750 * If this vnode is a mounted-on vnode, 2751 * but the mounted-on file system is not 2752 * exported, send back the filehandle for 2753 * the mounted-on vnode, not the root of 2754 * the mounted-on file system. 2755 */ 2756 VN_RELE(vp); 2757 vp = pre_tvp; 2758 exi = pre_exi; 2759 } else { 2760 VN_RELE(vp); 2761 return (puterrno4(EACCES)); 2762 } 2763 } else if (pre_tvp) { 2764 /* we're done with pre_tvp now. release extra hold */ 2765 VN_RELE(pre_tvp); 2766 } 2767 2768 cs->exi = exi; 2769 2770 /* 2771 * Now we do a checkauth4. The reason is that 2772 * this client/user may not have access to the new 2773 * exported file system, and if he does, 2774 * the client/user may be mapped to a different uid. 2775 * 2776 * We start with a new cr, because the checkauth4 done 2777 * in the PUT*FH operation over wrote the cred's uid, 2778 * gid, etc, and we want the real thing before calling 2779 * checkauth4() 2780 */ 2781 crfree(cs->cr); 2782 cs->cr = crdup(cs->basecr); 2783 2784 oldvp = cs->vp; 2785 cs->vp = vp; 2786 is_newvp = TRUE; 2787 2788 stat = call_checkauth4(cs, req); 2789 if (stat != NFS4_OK) { 2790 VN_RELE(cs->vp); 2791 cs->vp = oldvp; 2792 return (stat); 2793 } 2794 } 2795 2796 /* 2797 * After various NFS checks, do a label check on the path 2798 * component. The label on this path should either be the 2799 * global zone's label or a zone's label. We are only 2800 * interested in the zone's label because exported files 2801 * in global zone is accessible (though read-only) to 2802 * clients. The exportability/visibility check is already 2803 * done before reaching this code. 2804 */ 2805 if (is_system_labeled()) { 2806 bslabel_t *clabel; 2807 2808 ASSERT(req->rq_label != NULL); 2809 clabel = req->rq_label; 2810 DTRACE_PROBE2(tx__rfs4__log__info__oplookup__clabel, char *, 2811 "got client label from request(1)", struct svc_req *, req); 2812 2813 if (!blequal(&l_admin_low->tsl_label, clabel)) { 2814 if (!do_rfs_label_check(clabel, vp, DOMINANCE_CHECK, 2815 cs->exi)) { 2816 error = EACCES; 2817 goto err_out; 2818 } 2819 } else { 2820 /* 2821 * We grant access to admin_low label clients 2822 * only if the client is trusted, i.e. also 2823 * running Solaris Trusted Extension. 2824 */ 2825 struct sockaddr *ca; 2826 int addr_type; 2827 void *ipaddr; 2828 tsol_tpc_t *tp; 2829 2830 ca = (struct sockaddr *)svc_getrpccaller( 2831 req->rq_xprt)->buf; 2832 if (ca->sa_family == AF_INET) { 2833 addr_type = IPV4_VERSION; 2834 ipaddr = &((struct sockaddr_in *)ca)->sin_addr; 2835 } else if (ca->sa_family == AF_INET6) { 2836 addr_type = IPV6_VERSION; 2837 ipaddr = &((struct sockaddr_in6 *) 2838 ca)->sin6_addr; 2839 } 2840 tp = find_tpc(ipaddr, addr_type, B_FALSE); 2841 if (tp == NULL || tp->tpc_tp.tp_doi != 2842 l_admin_low->tsl_doi || tp->tpc_tp.host_type != 2843 SUN_CIPSO) { 2844 if (tp != NULL) 2845 TPC_RELE(tp); 2846 error = EACCES; 2847 goto err_out; 2848 } 2849 TPC_RELE(tp); 2850 } 2851 } 2852 2853 error = makefh4(&cs->fh, vp, cs->exi); 2854 2855 err_out: 2856 if (error) { 2857 if (is_newvp) { 2858 VN_RELE(cs->vp); 2859 cs->vp = oldvp; 2860 } else 2861 VN_RELE(vp); 2862 return (puterrno4(error)); 2863 } 2864 2865 if (!is_newvp) { 2866 if (cs->vp) 2867 VN_RELE(cs->vp); 2868 cs->vp = vp; 2869 } else if (oldvp) 2870 VN_RELE(oldvp); 2871 2872 /* 2873 * if did lookup on attrdir and didn't lookup .., set named 2874 * attr fh flag 2875 */ 2876 if (attrdir && ! dotdot) 2877 set_fh4_flag(&cs->fh, FH4_NAMEDATTR); 2878 2879 /* Assume false for now, open proc will set this */ 2880 cs->mandlock = FALSE; 2881 2882 return (NFS4_OK); 2883 } 2884 2885 /* ARGSUSED */ 2886 static void 2887 rfs4_op_lookup(nfs_argop4 *argop, nfs_resop4 *resop, struct svc_req *req, 2888 struct compound_state *cs) 2889 { 2890 LOOKUP4args *args = &argop->nfs_argop4_u.oplookup; 2891 LOOKUP4res *resp = &resop->nfs_resop4_u.oplookup; 2892 char *nm; 2893 uint_t len; 2894 struct sockaddr *ca; 2895 char *name = NULL; 2896 nfsstat4 status; 2897 2898 DTRACE_NFSV4_2(op__lookup__start, struct compound_state *, cs, 2899 LOOKUP4args *, args); 2900 2901 if (cs->vp == NULL) { 2902 *cs->statusp = resp->status = NFS4ERR_NOFILEHANDLE; 2903 goto out; 2904 } 2905 2906 if (cs->vp->v_type == VLNK) { 2907 *cs->statusp = resp->status = NFS4ERR_SYMLINK; 2908 goto out; 2909 } 2910 2911 if (cs->vp->v_type != VDIR) { 2912 *cs->statusp = resp->status = NFS4ERR_NOTDIR; 2913 goto out; 2914 } 2915 2916 status = utf8_dir_verify(&args->objname); 2917 if (status != NFS4_OK) { 2918 *cs->statusp = resp->status = status; 2919 goto out; 2920 } 2921 2922 nm = utf8_to_str(&args->objname, &len, NULL); 2923 if (nm == NULL) { 2924 *cs->statusp = resp->status = NFS4ERR_INVAL; 2925 goto out; 2926 } 2927 2928 if (len > MAXNAMELEN) { 2929 *cs->statusp = resp->status = NFS4ERR_NAMETOOLONG; 2930 kmem_free(nm, len); 2931 goto out; 2932 } 2933 2934 ca = (struct sockaddr *)svc_getrpccaller(req->rq_xprt)->buf; 2935 name = nfscmd_convname(ca, cs->exi, nm, NFSCMD_CONV_INBOUND, 2936 MAXPATHLEN + 1); 2937 2938 if (name == NULL) { 2939 *cs->statusp = resp->status = NFS4ERR_INVAL; 2940 kmem_free(nm, len); 2941 goto out; 2942 } 2943 2944 *cs->statusp = resp->status = do_rfs4_op_lookup(name, req, cs); 2945 2946 if (name != nm) 2947 kmem_free(name, MAXPATHLEN + 1); 2948 kmem_free(nm, len); 2949 2950 out: 2951 DTRACE_NFSV4_2(op__lookup__done, struct compound_state *, cs, 2952 LOOKUP4res *, resp); 2953 } 2954 2955 /* ARGSUSED */ 2956 static void 2957 rfs4_op_lookupp(nfs_argop4 *args, nfs_resop4 *resop, struct svc_req *req, 2958 struct compound_state *cs) 2959 { 2960 LOOKUPP4res *resp = &resop->nfs_resop4_u.oplookupp; 2961 2962 DTRACE_NFSV4_1(op__lookupp__start, struct compound_state *, cs); 2963 2964 if (cs->vp == NULL) { 2965 *cs->statusp = resp->status = NFS4ERR_NOFILEHANDLE; 2966 goto out; 2967 } 2968 2969 if (cs->vp->v_type != VDIR) { 2970 *cs->statusp = resp->status = NFS4ERR_NOTDIR; 2971 goto out; 2972 } 2973 2974 *cs->statusp = resp->status = do_rfs4_op_lookup("..", req, cs); 2975 2976 /* 2977 * From NFSV4 Specification, LOOKUPP should not check for 2978 * NFS4ERR_WRONGSEC. Retrun NFS4_OK instead. 2979 */ 2980 if (resp->status == NFS4ERR_WRONGSEC) { 2981 *cs->statusp = resp->status = NFS4_OK; 2982 } 2983 2984 out: 2985 DTRACE_NFSV4_2(op__lookupp__done, struct compound_state *, cs, 2986 LOOKUPP4res *, resp); 2987 } 2988 2989 2990 /*ARGSUSED2*/ 2991 static void 2992 rfs4_op_openattr(nfs_argop4 *argop, nfs_resop4 *resop, struct svc_req *req, 2993 struct compound_state *cs) 2994 { 2995 OPENATTR4args *args = &argop->nfs_argop4_u.opopenattr; 2996 OPENATTR4res *resp = &resop->nfs_resop4_u.opopenattr; 2997 vnode_t *avp = NULL; 2998 int lookup_flags = LOOKUP_XATTR, error; 2999 int exp_ro = 0; 3000 3001 DTRACE_NFSV4_2(op__openattr__start, struct compound_state *, cs, 3002 OPENATTR4args *, args); 3003 3004 if (cs->vp == NULL) { 3005 *cs->statusp = resp->status = NFS4ERR_NOFILEHANDLE; 3006 goto out; 3007 } 3008 3009 if ((cs->vp->v_vfsp->vfs_flag & VFS_XATTR) == 0 && 3010 !vfs_has_feature(cs->vp->v_vfsp, VFSFT_SYSATTR_VIEWS)) { 3011 *cs->statusp = resp->status = puterrno4(ENOTSUP); 3012 goto out; 3013 } 3014 3015 /* 3016 * If file system supports passing ACE mask to VOP_ACCESS then 3017 * check for ACE_READ_NAMED_ATTRS, otherwise do legacy checks 3018 */ 3019 3020 if (vfs_has_feature(cs->vp->v_vfsp, VFSFT_ACEMASKONACCESS)) 3021 error = VOP_ACCESS(cs->vp, ACE_READ_NAMED_ATTRS, 3022 V_ACE_MASK, cs->cr, NULL); 3023 else 3024 error = ((VOP_ACCESS(cs->vp, VREAD, 0, cs->cr, NULL) != 0) && 3025 (VOP_ACCESS(cs->vp, VWRITE, 0, cs->cr, NULL) != 0) && 3026 (VOP_ACCESS(cs->vp, VEXEC, 0, cs->cr, NULL) != 0)); 3027 3028 if (error) { 3029 *cs->statusp = resp->status = puterrno4(EACCES); 3030 goto out; 3031 } 3032 3033 /* 3034 * The CREATE_XATTR_DIR VOP flag cannot be specified if 3035 * the file system is exported read-only -- regardless of 3036 * createdir flag. Otherwise the attrdir would be created 3037 * (assuming server fs isn't mounted readonly locally). If 3038 * VOP_LOOKUP returns ENOENT in this case, the error will 3039 * be translated into EROFS. ENOSYS is mapped to ENOTSUP 3040 * because specfs has no VOP_LOOKUP op, so the macro would 3041 * return ENOSYS. EINVAL is returned by all (current) 3042 * Solaris file system implementations when any of their 3043 * restrictions are violated (xattr(dir) can't have xattrdir). 3044 * Returning NOTSUPP is more appropriate in this case 3045 * because the object will never be able to have an attrdir. 3046 */ 3047 if (args->createdir && ! (exp_ro = rdonly4(cs->exi, cs->vp, req))) 3048 lookup_flags |= CREATE_XATTR_DIR; 3049 3050 error = VOP_LOOKUP(cs->vp, "", &avp, NULL, lookup_flags, NULL, cs->cr, 3051 NULL, NULL, NULL); 3052 3053 if (error) { 3054 if (error == ENOENT && args->createdir && exp_ro) 3055 *cs->statusp = resp->status = puterrno4(EROFS); 3056 else if (error == EINVAL || error == ENOSYS) 3057 *cs->statusp = resp->status = puterrno4(ENOTSUP); 3058 else 3059 *cs->statusp = resp->status = puterrno4(error); 3060 goto out; 3061 } 3062 3063 ASSERT(avp->v_flag & V_XATTRDIR); 3064 3065 error = makefh4(&cs->fh, avp, cs->exi); 3066 3067 if (error) { 3068 VN_RELE(avp); 3069 *cs->statusp = resp->status = puterrno4(error); 3070 goto out; 3071 } 3072 3073 VN_RELE(cs->vp); 3074 cs->vp = avp; 3075 3076 /* 3077 * There is no requirement for an attrdir fh flag 3078 * because the attrdir has a vnode flag to distinguish 3079 * it from regular (non-xattr) directories. The 3080 * FH4_ATTRDIR flag is set for future sanity checks. 3081 */ 3082 set_fh4_flag(&cs->fh, FH4_ATTRDIR); 3083 *cs->statusp = resp->status = NFS4_OK; 3084 3085 out: 3086 DTRACE_NFSV4_2(op__openattr__done, struct compound_state *, cs, 3087 OPENATTR4res *, resp); 3088 } 3089 3090 static int 3091 do_io(int direction, vnode_t *vp, struct uio *uio, int ioflag, cred_t *cred, 3092 caller_context_t *ct) 3093 { 3094 int error; 3095 int i; 3096 clock_t delaytime; 3097 3098 delaytime = MSEC_TO_TICK_ROUNDUP(rfs4_lock_delay); 3099 3100 /* 3101 * Don't block on mandatory locks. If this routine returns 3102 * EAGAIN, the caller should return NFS4ERR_LOCKED. 3103 */ 3104 uio->uio_fmode = FNONBLOCK; 3105 3106 for (i = 0; i < rfs4_maxlock_tries; i++) { 3107 3108 3109 if (direction == FREAD) { 3110 (void) VOP_RWLOCK(vp, V_WRITELOCK_FALSE, ct); 3111 error = VOP_READ(vp, uio, ioflag, cred, ct); 3112 VOP_RWUNLOCK(vp, V_WRITELOCK_FALSE, ct); 3113 } else { 3114 (void) VOP_RWLOCK(vp, V_WRITELOCK_TRUE, ct); 3115 error = VOP_WRITE(vp, uio, ioflag, cred, ct); 3116 VOP_RWUNLOCK(vp, V_WRITELOCK_TRUE, ct); 3117 } 3118 3119 if (error != EAGAIN) 3120 break; 3121 3122 if (i < rfs4_maxlock_tries - 1) { 3123 delay(delaytime); 3124 delaytime *= 2; 3125 } 3126 } 3127 3128 return (error); 3129 } 3130 3131 /* ARGSUSED */ 3132 static void 3133 rfs4_op_read(nfs_argop4 *argop, nfs_resop4 *resop, struct svc_req *req, 3134 struct compound_state *cs) 3135 { 3136 READ4args *args = &argop->nfs_argop4_u.opread; 3137 READ4res *resp = &resop->nfs_resop4_u.opread; 3138 int error; 3139 int verror; 3140 vnode_t *vp; 3141 struct vattr va; 3142 struct iovec iov; 3143 struct uio uio; 3144 u_offset_t offset; 3145 bool_t *deleg = &cs->deleg; 3146 nfsstat4 stat; 3147 int in_crit = 0; 3148 mblk_t *mp = NULL; 3149 int alloc_err = 0; 3150 int rdma_used = 0; 3151 int loaned_buffers; 3152 caller_context_t ct; 3153 struct uio *uiop; 3154 3155 DTRACE_NFSV4_2(op__read__start, struct compound_state *, cs, 3156 READ4args, args); 3157 3158 vp = cs->vp; 3159 if (vp == NULL) { 3160 *cs->statusp = resp->status = NFS4ERR_NOFILEHANDLE; 3161 goto out; 3162 } 3163 if (cs->access == CS_ACCESS_DENIED) { 3164 *cs->statusp = resp->status = NFS4ERR_ACCESS; 3165 goto out; 3166 } 3167 3168 if ((stat = rfs4_check_stateid(FREAD, vp, &args->stateid, FALSE, 3169 deleg, TRUE, &ct)) != NFS4_OK) { 3170 *cs->statusp = resp->status = stat; 3171 goto out; 3172 } 3173 3174 /* 3175 * Enter the critical region before calling VOP_RWLOCK 3176 * to avoid a deadlock with write requests. 3177 */ 3178 if (nbl_need_check(vp)) { 3179 nbl_start_crit(vp, RW_READER); 3180 in_crit = 1; 3181 if (nbl_conflict(vp, NBL_READ, args->offset, args->count, 0, 3182 &ct)) { 3183 *cs->statusp = resp->status = NFS4ERR_LOCKED; 3184 goto out; 3185 } 3186 } 3187 3188 if ((stat = rfs4_check_stateid(FREAD, vp, &args->stateid, FALSE, 3189 deleg, TRUE, &ct)) != NFS4_OK) { 3190 *cs->statusp = resp->status = stat; 3191 goto out; 3192 } 3193 3194 if (args->wlist) { 3195 if (args->count > clist_len(args->wlist)) { 3196 *cs->statusp = resp->status = NFS4ERR_INVAL; 3197 goto out; 3198 } 3199 rdma_used = 1; 3200 } 3201 3202 /* use loaned buffers for TCP */ 3203 loaned_buffers = (nfs_loaned_buffers && !rdma_used) ? 1 : 0; 3204 3205 va.va_mask = AT_MODE|AT_SIZE|AT_UID; 3206 verror = VOP_GETATTR(vp, &va, 0, cs->cr, &ct); 3207 3208 /* 3209 * If we can't get the attributes, then we can't do the 3210 * right access checking. So, we'll fail the request. 3211 */ 3212 if (verror) { 3213 *cs->statusp = resp->status = puterrno4(verror); 3214 goto out; 3215 } 3216 3217 if (vp->v_type != VREG) { 3218 *cs->statusp = resp->status = 3219 ((vp->v_type == VDIR) ? NFS4ERR_ISDIR : NFS4ERR_INVAL); 3220 goto out; 3221 } 3222 3223 if (crgetuid(cs->cr) != va.va_uid && 3224 (error = VOP_ACCESS(vp, VREAD, 0, cs->cr, &ct)) && 3225 (error = VOP_ACCESS(vp, VEXEC, 0, cs->cr, &ct))) { 3226 *cs->statusp = resp->status = puterrno4(error); 3227 goto out; 3228 } 3229 3230 if (MANDLOCK(vp, va.va_mode)) { /* XXX - V4 supports mand locking */ 3231 *cs->statusp = resp->status = NFS4ERR_ACCESS; 3232 goto out; 3233 } 3234 3235 offset = args->offset; 3236 if (offset >= va.va_size) { 3237 *cs->statusp = resp->status = NFS4_OK; 3238 resp->eof = TRUE; 3239 resp->data_len = 0; 3240 resp->data_val = NULL; 3241 resp->mblk = NULL; 3242 /* RDMA */ 3243 resp->wlist = args->wlist; 3244 resp->wlist_len = resp->data_len; 3245 *cs->statusp = resp->status = NFS4_OK; 3246 if (resp->wlist) 3247 clist_zero_len(resp->wlist); 3248 goto out; 3249 } 3250 3251 if (args->count == 0) { 3252 *cs->statusp = resp->status = NFS4_OK; 3253 resp->eof = FALSE; 3254 resp->data_len = 0; 3255 resp->data_val = NULL; 3256 resp->mblk = NULL; 3257 /* RDMA */ 3258 resp->wlist = args->wlist; 3259 resp->wlist_len = resp->data_len; 3260 if (resp->wlist) 3261 clist_zero_len(resp->wlist); 3262 goto out; 3263 } 3264 3265 /* 3266 * Do not allocate memory more than maximum allowed 3267 * transfer size 3268 */ 3269 if (args->count > rfs4_tsize(req)) 3270 args->count = rfs4_tsize(req); 3271 3272 if (loaned_buffers) { 3273 uiop = (uio_t *)rfs_setup_xuio(vp); 3274 ASSERT(uiop != NULL); 3275 uiop->uio_segflg = UIO_SYSSPACE; 3276 uiop->uio_loffset = args->offset; 3277 uiop->uio_resid = args->count; 3278 3279 /* Jump to do the read if successful */ 3280 if (!VOP_REQZCBUF(vp, UIO_READ, (xuio_t *)uiop, cs->cr, &ct)) { 3281 /* 3282 * Need to hold the vnode until after VOP_RETZCBUF() 3283 * is called. 3284 */ 3285 VN_HOLD(vp); 3286 goto doio_read; 3287 } 3288 3289 DTRACE_PROBE2(nfss__i__reqzcbuf_failed, int, 3290 uiop->uio_loffset, int, uiop->uio_resid); 3291 3292 uiop->uio_extflg = 0; 3293 3294 /* failure to setup for zero copy */ 3295 rfs_free_xuio((void *)uiop); 3296 loaned_buffers = 0; 3297 } 3298 3299 /* 3300 * If returning data via RDMA Write, then grab the chunk list. If we 3301 * aren't returning READ data w/RDMA_WRITE, then grab a mblk. 3302 */ 3303 if (rdma_used) { 3304 mp = NULL; 3305 (void) rdma_get_wchunk(req, &iov, args->wlist); 3306 } else { 3307 /* 3308 * mp will contain the data to be sent out in the read reply. 3309 * It will be freed after the reply has been sent. Let's 3310 * roundup the data to a BYTES_PER_XDR_UNIT multiple, so that 3311 * the call to xdrmblk_putmblk() never fails. If the first 3312 * alloc of the requested size fails, then decrease the size to 3313 * something more reasonable and wait for the allocation to 3314 * occur. 3315 */ 3316 mp = allocb(RNDUP(args->count), BPRI_MED); 3317 if (mp == NULL) { 3318 if (args->count > MAXBSIZE) 3319 args->count = MAXBSIZE; 3320 mp = allocb_wait(RNDUP(args->count), BPRI_MED, 3321 STR_NOSIG, &alloc_err); 3322 } 3323 ASSERT(mp != NULL); 3324 ASSERT(alloc_err == 0); 3325 3326 iov.iov_base = (caddr_t)mp->b_datap->db_base; 3327 iov.iov_len = args->count; 3328 } 3329 3330 uio.uio_iov = &iov; 3331 uio.uio_iovcnt = 1; 3332 uio.uio_segflg = UIO_SYSSPACE; 3333 uio.uio_extflg = UIO_COPY_CACHED; 3334 uio.uio_loffset = args->offset; 3335 uio.uio_resid = args->count; 3336 uiop = &uio; 3337 3338 doio_read: 3339 error = do_io(FREAD, vp, uiop, 0, cs->cr, &ct); 3340 3341 va.va_mask = AT_SIZE; 3342 verror = VOP_GETATTR(vp, &va, 0, cs->cr, &ct); 3343 3344 if (error) { 3345 if (mp) 3346 freemsg(mp); 3347 *cs->statusp = resp->status = puterrno4(error); 3348 goto out; 3349 } 3350 3351 /* make mblk using zc buffers */ 3352 if (loaned_buffers) { 3353 mp = uio_to_mblk(uiop); 3354 ASSERT(mp != NULL); 3355 } 3356 3357 *cs->statusp = resp->status = NFS4_OK; 3358 3359 ASSERT(uiop->uio_resid >= 0); 3360 resp->data_len = args->count - uiop->uio_resid; 3361 if (mp) { 3362 resp->data_val = (char *)mp->b_datap->db_base; 3363 rfs_rndup_mblks(mp, resp->data_len, loaned_buffers); 3364 } else { 3365 resp->data_val = (caddr_t)iov.iov_base; 3366 } 3367 3368 resp->mblk = mp; 3369 3370 if (!verror && offset + resp->data_len == va.va_size) 3371 resp->eof = TRUE; 3372 else 3373 resp->eof = FALSE; 3374 3375 if (rdma_used) { 3376 if (!rdma_setup_read_data4(args, resp)) { 3377 *cs->statusp = resp->status = NFS4ERR_INVAL; 3378 } 3379 } else { 3380 resp->wlist = NULL; 3381 } 3382 3383 out: 3384 if (in_crit) 3385 nbl_end_crit(vp); 3386 3387 DTRACE_NFSV4_2(op__read__done, struct compound_state *, cs, 3388 READ4res *, resp); 3389 } 3390 3391 static void 3392 rfs4_op_read_free(nfs_resop4 *resop) 3393 { 3394 READ4res *resp = &resop->nfs_resop4_u.opread; 3395 3396 if (resp->status == NFS4_OK && resp->mblk != NULL) { 3397 freemsg(resp->mblk); 3398 resp->mblk = NULL; 3399 resp->data_val = NULL; 3400 resp->data_len = 0; 3401 } 3402 } 3403 3404 static void 3405 rfs4_op_readdir_free(nfs_resop4 * resop) 3406 { 3407 READDIR4res *resp = &resop->nfs_resop4_u.opreaddir; 3408 3409 if (resp->status == NFS4_OK && resp->mblk != NULL) { 3410 freeb(resp->mblk); 3411 resp->mblk = NULL; 3412 resp->data_len = 0; 3413 } 3414 } 3415 3416 3417 /* ARGSUSED */ 3418 static void 3419 rfs4_op_putpubfh(nfs_argop4 *args, nfs_resop4 *resop, struct svc_req *req, 3420 struct compound_state *cs) 3421 { 3422 PUTPUBFH4res *resp = &resop->nfs_resop4_u.opputpubfh; 3423 int error; 3424 vnode_t *vp; 3425 struct exportinfo *exi, *sav_exi; 3426 nfs_fh4_fmt_t *fh_fmtp; 3427 3428 DTRACE_NFSV4_1(op__putpubfh__start, struct compound_state *, cs); 3429 3430 if (cs->vp) { 3431 VN_RELE(cs->vp); 3432 cs->vp = NULL; 3433 } 3434 3435 if (cs->cr) 3436 crfree(cs->cr); 3437 3438 cs->cr = crdup(cs->basecr); 3439 3440 vp = exi_public->exi_vp; 3441 if (vp == NULL) { 3442 *cs->statusp = resp->status = NFS4ERR_SERVERFAULT; 3443 goto out; 3444 } 3445 3446 error = makefh4(&cs->fh, vp, exi_public); 3447 if (error != 0) { 3448 *cs->statusp = resp->status = puterrno4(error); 3449 goto out; 3450 } 3451 sav_exi = cs->exi; 3452 if (exi_public == exi_root) { 3453 /* 3454 * No filesystem is actually shared public, so we default 3455 * to exi_root. In this case, we must check whether root 3456 * is exported. 3457 */ 3458 fh_fmtp = (nfs_fh4_fmt_t *)cs->fh.nfs_fh4_val; 3459 3460 /* 3461 * if root filesystem is exported, the exportinfo struct that we 3462 * should use is what checkexport4 returns, because root_exi is 3463 * actually a mostly empty struct. 3464 */ 3465 exi = checkexport4(&fh_fmtp->fh4_fsid, 3466 (fid_t *)&fh_fmtp->fh4_xlen, NULL); 3467 cs->exi = ((exi != NULL) ? exi : exi_public); 3468 } else { 3469 /* 3470 * it's a properly shared filesystem 3471 */ 3472 cs->exi = exi_public; 3473 } 3474 3475 if (is_system_labeled()) { 3476 bslabel_t *clabel; 3477 3478 ASSERT(req->rq_label != NULL); 3479 clabel = req->rq_label; 3480 DTRACE_PROBE2(tx__rfs4__log__info__opputpubfh__clabel, char *, 3481 "got client label from request(1)", 3482 struct svc_req *, req); 3483 if (!blequal(&l_admin_low->tsl_label, clabel)) { 3484 if (!do_rfs_label_check(clabel, vp, DOMINANCE_CHECK, 3485 cs->exi)) { 3486 *cs->statusp = resp->status = 3487 NFS4ERR_SERVERFAULT; 3488 goto out; 3489 } 3490 } 3491 } 3492 3493 VN_HOLD(vp); 3494 cs->vp = vp; 3495 3496 if ((resp->status = call_checkauth4(cs, req)) != NFS4_OK) { 3497 VN_RELE(cs->vp); 3498 cs->vp = NULL; 3499 cs->exi = sav_exi; 3500 goto out; 3501 } 3502 3503 *cs->statusp = resp->status = NFS4_OK; 3504 out: 3505 DTRACE_NFSV4_2(op__putpubfh__done, struct compound_state *, cs, 3506 PUTPUBFH4res *, resp); 3507 } 3508 3509 /* 3510 * XXX - issue with put*fh operations. Suppose /export/home is exported. 3511 * Suppose an NFS client goes to mount /export/home/joe. If /export, home, 3512 * or joe have restrictive search permissions, then we shouldn't let 3513 * the client get a file handle. This is easy to enforce. However, we 3514 * don't know what security flavor should be used until we resolve the 3515 * path name. Another complication is uid mapping. If root is 3516 * the user, then it will be mapped to the anonymous user by default, 3517 * but we won't know that till we've resolved the path name. And we won't 3518 * know what the anonymous user is. 3519 * Luckily, SECINFO is specified to take a full filename. 3520 * So what we will have to in rfs4_op_lookup is check that flavor of 3521 * the target object matches that of the request, and if root was the 3522 * caller, check for the root= and anon= options, and if necessary, 3523 * repeat the lookup using the right cred_t. But that's not done yet. 3524 */ 3525 /* ARGSUSED */ 3526 static void 3527 rfs4_op_putfh(nfs_argop4 *argop, nfs_resop4 *resop, struct svc_req *req, 3528 struct compound_state *cs) 3529 { 3530 PUTFH4args *args = &argop->nfs_argop4_u.opputfh; 3531 PUTFH4res *resp = &resop->nfs_resop4_u.opputfh; 3532 nfs_fh4_fmt_t *fh_fmtp; 3533 3534 DTRACE_NFSV4_2(op__putfh__start, struct compound_state *, cs, 3535 PUTFH4args *, args); 3536 3537 if (cs->vp) { 3538 VN_RELE(cs->vp); 3539 cs->vp = NULL; 3540 } 3541 3542 if (cs->cr) { 3543 crfree(cs->cr); 3544 cs->cr = NULL; 3545 } 3546 3547 3548 if (args->object.nfs_fh4_len < NFS_FH4_LEN) { 3549 *cs->statusp = resp->status = NFS4ERR_BADHANDLE; 3550 goto out; 3551 } 3552 3553 fh_fmtp = (nfs_fh4_fmt_t *)args->object.nfs_fh4_val; 3554 cs->exi = checkexport4(&fh_fmtp->fh4_fsid, (fid_t *)&fh_fmtp->fh4_xlen, 3555 NULL); 3556 3557 if (cs->exi == NULL) { 3558 *cs->statusp = resp->status = NFS4ERR_STALE; 3559 goto out; 3560 } 3561 3562 cs->cr = crdup(cs->basecr); 3563 3564 ASSERT(cs->cr != NULL); 3565 3566 if (! (cs->vp = nfs4_fhtovp(&args->object, cs->exi, &resp->status))) { 3567 *cs->statusp = resp->status; 3568 goto out; 3569 } 3570 3571 if ((resp->status = call_checkauth4(cs, req)) != NFS4_OK) { 3572 VN_RELE(cs->vp); 3573 cs->vp = NULL; 3574 goto out; 3575 } 3576 3577 nfs_fh4_copy(&args->object, &cs->fh); 3578 *cs->statusp = resp->status = NFS4_OK; 3579 cs->deleg = FALSE; 3580 3581 out: 3582 DTRACE_NFSV4_2(op__putfh__done, struct compound_state *, cs, 3583 PUTFH4res *, resp); 3584 } 3585 3586 /* ARGSUSED */ 3587 static void 3588 rfs4_op_putrootfh(nfs_argop4 *argop, nfs_resop4 *resop, struct svc_req *req, 3589 struct compound_state *cs) 3590 { 3591 PUTROOTFH4res *resp = &resop->nfs_resop4_u.opputrootfh; 3592 int error; 3593 fid_t fid; 3594 struct exportinfo *exi, *sav_exi; 3595 3596 DTRACE_NFSV4_1(op__putrootfh__start, struct compound_state *, cs); 3597 3598 if (cs->vp) { 3599 VN_RELE(cs->vp); 3600 cs->vp = NULL; 3601 } 3602 3603 if (cs->cr) 3604 crfree(cs->cr); 3605 3606 cs->cr = crdup(cs->basecr); 3607 3608 /* 3609 * Using rootdir, the system root vnode, 3610 * get its fid. 3611 */ 3612 bzero(&fid, sizeof (fid)); 3613 fid.fid_len = MAXFIDSZ; 3614 error = vop_fid_pseudo(rootdir, &fid); 3615 if (error != 0) { 3616 *cs->statusp = resp->status = puterrno4(error); 3617 goto out; 3618 } 3619 3620 /* 3621 * Then use the root fsid & fid it to find out if it's exported 3622 * 3623 * If the server root isn't exported directly, then 3624 * it should at least be a pseudo export based on 3625 * one or more exports further down in the server's 3626 * file tree. 3627 */ 3628 exi = checkexport4(&rootdir->v_vfsp->vfs_fsid, &fid, NULL); 3629 if (exi == NULL || exi->exi_export.ex_flags & EX_PUBLIC) { 3630 NFS4_DEBUG(rfs4_debug, 3631 (CE_WARN, "rfs4_op_putrootfh: export check failure")); 3632 *cs->statusp = resp->status = NFS4ERR_SERVERFAULT; 3633 goto out; 3634 } 3635 3636 /* 3637 * Now make a filehandle based on the root 3638 * export and root vnode. 3639 */ 3640 error = makefh4(&cs->fh, rootdir, exi); 3641 if (error != 0) { 3642 *cs->statusp = resp->status = puterrno4(error); 3643 goto out; 3644 } 3645 3646 sav_exi = cs->exi; 3647 cs->exi = exi; 3648 3649 VN_HOLD(rootdir); 3650 cs->vp = rootdir; 3651 3652 if ((resp->status = call_checkauth4(cs, req)) != NFS4_OK) { 3653 VN_RELE(rootdir); 3654 cs->vp = NULL; 3655 cs->exi = sav_exi; 3656 goto out; 3657 } 3658 3659 *cs->statusp = resp->status = NFS4_OK; 3660 cs->deleg = FALSE; 3661 out: 3662 DTRACE_NFSV4_2(op__putrootfh__done, struct compound_state *, cs, 3663 PUTROOTFH4res *, resp); 3664 } 3665 3666 /* 3667 * A directory entry is a valid nfsv4 entry if 3668 * - it has a non-zero ino 3669 * - it is not a dot or dotdot name 3670 * - it is visible in a pseudo export or in a real export that can 3671 * only have a limited view. 3672 */ 3673 static bool_t 3674 valid_nfs4_entry(struct exportinfo *exi, struct dirent64 *dp, 3675 int *expseudo, int check_visible) 3676 { 3677 if (dp->d_ino == 0 || NFS_IS_DOTNAME(dp->d_name)) { 3678 *expseudo = 0; 3679 return (FALSE); 3680 } 3681 3682 if (! check_visible) { 3683 *expseudo = 0; 3684 return (TRUE); 3685 } 3686 3687 return (nfs_visible_inode(exi, dp->d_ino, expseudo)); 3688 } 3689 3690 /* 3691 * set_rdattr_params sets up the variables used to manage what information 3692 * to get for each directory entry. 3693 */ 3694 static nfsstat4 3695 set_rdattr_params(struct nfs4_svgetit_arg *sargp, 3696 bitmap4 attrs, bool_t *need_to_lookup) 3697 { 3698 uint_t va_mask; 3699 nfsstat4 status; 3700 bitmap4 objbits; 3701 3702 status = bitmap4_to_attrmask(attrs, sargp); 3703 if (status != NFS4_OK) { 3704 /* 3705 * could not even figure attr mask 3706 */ 3707 return (status); 3708 } 3709 va_mask = sargp->vap->va_mask; 3710 3711 /* 3712 * dirent's d_ino is always correct value for mounted_on_fileid. 3713 * mntdfid_set is set once here, but mounted_on_fileid is 3714 * set in main dirent processing loop for each dirent. 3715 * The mntdfid_set is a simple optimization that lets the 3716 * server attr code avoid work when caller is readdir. 3717 */ 3718 sargp->mntdfid_set = TRUE; 3719 3720 /* 3721 * Lookup entry only if client asked for any of the following: 3722 * a) vattr attrs 3723 * b) vfs attrs 3724 * c) attrs w/per-object scope requested (change, filehandle, etc) 3725 * other than mounted_on_fileid (which we can take from dirent) 3726 */ 3727 objbits = attrs ? attrs & NFS4_VP_ATTR_MASK : 0; 3728 3729 if (va_mask || sargp->sbp || (objbits & ~FATTR4_MOUNTED_ON_FILEID_MASK)) 3730 *need_to_lookup = TRUE; 3731 else 3732 *need_to_lookup = FALSE; 3733 3734 if (sargp->sbp == NULL) 3735 return (NFS4_OK); 3736 3737 /* 3738 * If filesystem attrs are requested, get them now from the 3739 * directory vp, as most entries will have same filesystem. The only 3740 * exception are mounted over entries but we handle 3741 * those as we go (XXX mounted over detection not yet implemented). 3742 */ 3743 sargp->vap->va_mask = 0; /* to avoid VOP_GETATTR */ 3744 status = bitmap4_get_sysattrs(sargp); 3745 sargp->vap->va_mask = va_mask; 3746 3747 if ((status != NFS4_OK) && sargp->rdattr_error_req) { 3748 /* 3749 * Failed to get filesystem attributes. 3750 * Return a rdattr_error for each entry, but don't fail. 3751 * However, don't get any obj-dependent attrs. 3752 */ 3753 sargp->rdattr_error = status; /* for rdattr_error */ 3754 *need_to_lookup = FALSE; 3755 /* 3756 * At least get fileid for regular readdir output 3757 */ 3758 sargp->vap->va_mask &= AT_NODEID; 3759 status = NFS4_OK; 3760 } 3761 3762 return (status); 3763 } 3764 3765 /* 3766 * readlink: args: CURRENT_FH. 3767 * res: status. If success - CURRENT_FH unchanged, return linktext. 3768 */ 3769 3770 /* ARGSUSED */ 3771 static void 3772 rfs4_op_readlink(nfs_argop4 *argop, nfs_resop4 *resop, struct svc_req *req, 3773 struct compound_state *cs) 3774 { 3775 READLINK4res *resp = &resop->nfs_resop4_u.opreadlink; 3776 int error; 3777 vnode_t *vp; 3778 struct iovec iov; 3779 struct vattr va; 3780 struct uio uio; 3781 char *data; 3782 struct sockaddr *ca; 3783 char *name = NULL; 3784 int is_referral; 3785 3786 DTRACE_NFSV4_1(op__readlink__start, struct compound_state *, cs); 3787 3788 /* CURRENT_FH: directory */ 3789 vp = cs->vp; 3790 if (vp == NULL) { 3791 *cs->statusp = resp->status = NFS4ERR_NOFILEHANDLE; 3792 goto out; 3793 } 3794 3795 if (cs->access == CS_ACCESS_DENIED) { 3796 *cs->statusp = resp->status = NFS4ERR_ACCESS; 3797 goto out; 3798 } 3799 3800 /* Is it a referral? */ 3801 if (vn_is_nfs_reparse(vp, cs->cr) && client_is_downrev(req)) { 3802 3803 is_referral = 1; 3804 3805 } else { 3806 3807 is_referral = 0; 3808 3809 if (vp->v_type == VDIR) { 3810 *cs->statusp = resp->status = NFS4ERR_ISDIR; 3811 goto out; 3812 } 3813 3814 if (vp->v_type != VLNK) { 3815 *cs->statusp = resp->status = NFS4ERR_INVAL; 3816 goto out; 3817 } 3818 3819 } 3820 3821 va.va_mask = AT_MODE; 3822 error = VOP_GETATTR(vp, &va, 0, cs->cr, NULL); 3823 if (error) { 3824 *cs->statusp = resp->status = puterrno4(error); 3825 goto out; 3826 } 3827 3828 if (MANDLOCK(vp, va.va_mode)) { 3829 *cs->statusp = resp->status = NFS4ERR_ACCESS; 3830 goto out; 3831 } 3832 3833 data = kmem_alloc(MAXPATHLEN + 1, KM_SLEEP); 3834 3835 if (is_referral) { 3836 char *s; 3837 size_t strsz; 3838 3839 /* Get an artificial symlink based on a referral */ 3840 s = build_symlink(vp, cs->cr, &strsz); 3841 global_svstat_ptr[4][NFS_REFERLINKS].value.ui64++; 3842 DTRACE_PROBE2(nfs4serv__func__referral__reflink, 3843 vnode_t *, vp, char *, s); 3844 if (s == NULL) 3845 error = EINVAL; 3846 else { 3847 error = 0; 3848 (void) strlcpy(data, s, MAXPATHLEN + 1); 3849 kmem_free(s, strsz); 3850 } 3851 3852 } else { 3853 3854 iov.iov_base = data; 3855 iov.iov_len = MAXPATHLEN; 3856 uio.uio_iov = &iov; 3857 uio.uio_iovcnt = 1; 3858 uio.uio_segflg = UIO_SYSSPACE; 3859 uio.uio_extflg = UIO_COPY_CACHED; 3860 uio.uio_loffset = 0; 3861 uio.uio_resid = MAXPATHLEN; 3862 3863 error = VOP_READLINK(vp, &uio, cs->cr, NULL); 3864 3865 if (!error) 3866 *(data + MAXPATHLEN - uio.uio_resid) = '\0'; 3867 } 3868 3869 if (error) { 3870 kmem_free((caddr_t)data, (uint_t)MAXPATHLEN + 1); 3871 *cs->statusp = resp->status = puterrno4(error); 3872 goto out; 3873 } 3874 3875 ca = (struct sockaddr *)svc_getrpccaller(req->rq_xprt)->buf; 3876 name = nfscmd_convname(ca, cs->exi, data, NFSCMD_CONV_OUTBOUND, 3877 MAXPATHLEN + 1); 3878 3879 if (name == NULL) { 3880 /* 3881 * Even though the conversion failed, we return 3882 * something. We just don't translate it. 3883 */ 3884 name = data; 3885 } 3886 3887 /* 3888 * treat link name as data 3889 */ 3890 (void) str_to_utf8(name, &resp->link); 3891 3892 if (name != data) 3893 kmem_free(name, MAXPATHLEN + 1); 3894 kmem_free((caddr_t)data, (uint_t)MAXPATHLEN + 1); 3895 *cs->statusp = resp->status = NFS4_OK; 3896 3897 out: 3898 DTRACE_NFSV4_2(op__readlink__done, struct compound_state *, cs, 3899 READLINK4res *, resp); 3900 } 3901 3902 static void 3903 rfs4_op_readlink_free(nfs_resop4 *resop) 3904 { 3905 READLINK4res *resp = &resop->nfs_resop4_u.opreadlink; 3906 utf8string *symlink = &resp->link; 3907 3908 if (symlink->utf8string_val) { 3909 UTF8STRING_FREE(*symlink) 3910 } 3911 } 3912 3913 /* 3914 * release_lockowner: 3915 * Release any state associated with the supplied 3916 * lockowner. Note if any lo_state is holding locks we will not 3917 * rele that lo_state and thus the lockowner will not be destroyed. 3918 * A client using lock after the lock owner stateid has been released 3919 * will suffer the consequence of NFS4ERR_BAD_STATEID and would have 3920 * to reissue the lock with new_lock_owner set to TRUE. 3921 * args: lock_owner 3922 * res: status 3923 */ 3924 /* ARGSUSED */ 3925 static void 3926 rfs4_op_release_lockowner(nfs_argop4 *argop, nfs_resop4 *resop, 3927 struct svc_req *req, struct compound_state *cs) 3928 { 3929 RELEASE_LOCKOWNER4args *ap = &argop->nfs_argop4_u.oprelease_lockowner; 3930 RELEASE_LOCKOWNER4res *resp = &resop->nfs_resop4_u.oprelease_lockowner; 3931 rfs4_lockowner_t *lo; 3932 rfs4_openowner_t *oo; 3933 rfs4_state_t *sp; 3934 rfs4_lo_state_t *lsp; 3935 rfs4_client_t *cp; 3936 bool_t create = FALSE; 3937 locklist_t *llist; 3938 sysid_t sysid; 3939 3940 DTRACE_NFSV4_2(op__release__lockowner__start, struct compound_state *, 3941 cs, RELEASE_LOCKOWNER4args *, ap); 3942 3943 /* Make sure there is a clientid around for this request */ 3944 cp = rfs4_findclient_by_id(ap->lock_owner.clientid, FALSE); 3945 3946 if (cp == NULL) { 3947 *cs->statusp = resp->status = 3948 rfs4_check_clientid(&ap->lock_owner.clientid, 0); 3949 goto out; 3950 } 3951 rfs4_client_rele(cp); 3952 3953 lo = rfs4_findlockowner(&ap->lock_owner, &create); 3954 if (lo == NULL) { 3955 *cs->statusp = resp->status = NFS4_OK; 3956 goto out; 3957 } 3958 ASSERT(lo->rl_client != NULL); 3959 3960 /* 3961 * Check for EXPIRED client. If so will reap state with in a lease 3962 * period or on next set_clientid_confirm step 3963 */ 3964 if (rfs4_lease_expired(lo->rl_client)) { 3965 rfs4_lockowner_rele(lo); 3966 *cs->statusp = resp->status = NFS4ERR_EXPIRED; 3967 goto out; 3968 } 3969 3970 /* 3971 * If no sysid has been assigned, then no locks exist; just return. 3972 */ 3973 rfs4_dbe_lock(lo->rl_client->rc_dbe); 3974 if (lo->rl_client->rc_sysidt == LM_NOSYSID) { 3975 rfs4_lockowner_rele(lo); 3976 rfs4_dbe_unlock(lo->rl_client->rc_dbe); 3977 goto out; 3978 } 3979 3980 sysid = lo->rl_client->rc_sysidt; 3981 rfs4_dbe_unlock(lo->rl_client->rc_dbe); 3982 3983 /* 3984 * Mark the lockowner invalid. 3985 */ 3986 rfs4_dbe_hide(lo->rl_dbe); 3987 3988 /* 3989 * sysid-pid pair should now not be used since the lockowner is 3990 * invalid. If the client were to instantiate the lockowner again 3991 * it would be assigned a new pid. Thus we can get the list of 3992 * current locks. 3993 */ 3994 3995 llist = flk_get_active_locks(sysid, lo->rl_pid); 3996 /* If we are still holding locks fail */ 3997 if (llist != NULL) { 3998 3999 *cs->statusp = resp->status = NFS4ERR_LOCKS_HELD; 4000 4001 flk_free_locklist(llist); 4002 /* 4003 * We need to unhide the lockowner so the client can 4004 * try it again. The bad thing here is if the client 4005 * has a logic error that took it here in the first place 4006 * he probably has lost accounting of the locks that it 4007 * is holding. So we may have dangling state until the 4008 * open owner state is reaped via close. One scenario 4009 * that could possibly occur is that the client has 4010 * sent the unlock request(s) in separate threads 4011 * and has not waited for the replies before sending the 4012 * RELEASE_LOCKOWNER request. Presumably, it would expect 4013 * and deal appropriately with NFS4ERR_LOCKS_HELD, by 4014 * reissuing the request. 4015 */ 4016 rfs4_dbe_unhide(lo->rl_dbe); 4017 rfs4_lockowner_rele(lo); 4018 goto out; 4019 } 4020 4021 /* 4022 * For the corresponding client we need to check each open 4023 * owner for any opens that have lockowner state associated 4024 * with this lockowner. 4025 */ 4026 4027 rfs4_dbe_lock(lo->rl_client->rc_dbe); 4028 for (oo = list_head(&lo->rl_client->rc_openownerlist); oo != NULL; 4029 oo = list_next(&lo->rl_client->rc_openownerlist, oo)) { 4030 4031 rfs4_dbe_lock(oo->ro_dbe); 4032 for (sp = list_head(&oo->ro_statelist); sp != NULL; 4033 sp = list_next(&oo->ro_statelist, sp)) { 4034 4035 rfs4_dbe_lock(sp->rs_dbe); 4036 for (lsp = list_head(&sp->rs_lostatelist); 4037 lsp != NULL; 4038 lsp = list_next(&sp->rs_lostatelist, lsp)) { 4039 if (lsp->rls_locker == lo) { 4040 rfs4_dbe_lock(lsp->rls_dbe); 4041 rfs4_dbe_invalidate(lsp->rls_dbe); 4042 rfs4_dbe_unlock(lsp->rls_dbe); 4043 } 4044 } 4045 rfs4_dbe_unlock(sp->rs_dbe); 4046 } 4047 rfs4_dbe_unlock(oo->ro_dbe); 4048 } 4049 rfs4_dbe_unlock(lo->rl_client->rc_dbe); 4050 4051 rfs4_lockowner_rele(lo); 4052 4053 *cs->statusp = resp->status = NFS4_OK; 4054 4055 out: 4056 DTRACE_NFSV4_2(op__release__lockowner__done, struct compound_state *, 4057 cs, RELEASE_LOCKOWNER4res *, resp); 4058 } 4059 4060 /* 4061 * short utility function to lookup a file and recall the delegation 4062 */ 4063 static rfs4_file_t * 4064 rfs4_lookup_and_findfile(vnode_t *dvp, char *nm, vnode_t **vpp, 4065 int *lkup_error, cred_t *cr) 4066 { 4067 vnode_t *vp; 4068 rfs4_file_t *fp = NULL; 4069 bool_t fcreate = FALSE; 4070 int error; 4071 4072 if (vpp) 4073 *vpp = NULL; 4074 4075 if ((error = VOP_LOOKUP(dvp, nm, &vp, NULL, 0, NULL, cr, NULL, NULL, 4076 NULL)) == 0) { 4077 if (vp->v_type == VREG) 4078 fp = rfs4_findfile(vp, NULL, &fcreate); 4079 if (vpp) 4080 *vpp = vp; 4081 else 4082 VN_RELE(vp); 4083 } 4084 4085 if (lkup_error) 4086 *lkup_error = error; 4087 4088 return (fp); 4089 } 4090 4091 /* 4092 * remove: args: CURRENT_FH: directory; name. 4093 * res: status. If success - CURRENT_FH unchanged, return change_info 4094 * for directory. 4095 */ 4096 /* ARGSUSED */ 4097 static void 4098 rfs4_op_remove(nfs_argop4 *argop, nfs_resop4 *resop, struct svc_req *req, 4099 struct compound_state *cs) 4100 { 4101 REMOVE4args *args = &argop->nfs_argop4_u.opremove; 4102 REMOVE4res *resp = &resop->nfs_resop4_u.opremove; 4103 int error; 4104 vnode_t *dvp, *vp; 4105 struct vattr bdva, idva, adva; 4106 char *nm; 4107 uint_t len; 4108 rfs4_file_t *fp; 4109 int in_crit = 0; 4110 bslabel_t *clabel; 4111 struct sockaddr *ca; 4112 char *name = NULL; 4113 nfsstat4 status; 4114 4115 DTRACE_NFSV4_2(op__remove__start, struct compound_state *, cs, 4116 REMOVE4args *, args); 4117 4118 /* CURRENT_FH: directory */ 4119 dvp = cs->vp; 4120 if (dvp == NULL) { 4121 *cs->statusp = resp->status = NFS4ERR_NOFILEHANDLE; 4122 goto out; 4123 } 4124 4125 if (cs->access == CS_ACCESS_DENIED) { 4126 *cs->statusp = resp->status = NFS4ERR_ACCESS; 4127 goto out; 4128 } 4129 4130 /* 4131 * If there is an unshared filesystem mounted on this vnode, 4132 * Do not allow to remove anything in this directory. 4133 */ 4134 if (vn_ismntpt(dvp)) { 4135 *cs->statusp = resp->status = NFS4ERR_ACCESS; 4136 goto out; 4137 } 4138 4139 if (dvp->v_type != VDIR) { 4140 *cs->statusp = resp->status = NFS4ERR_NOTDIR; 4141 goto out; 4142 } 4143 4144 status = utf8_dir_verify(&args->target); 4145 if (status != NFS4_OK) { 4146 *cs->statusp = resp->status = status; 4147 goto out; 4148 } 4149 4150 /* 4151 * Lookup the file so that we can check if it's a directory 4152 */ 4153 nm = utf8_to_fn(&args->target, &len, NULL); 4154 if (nm == NULL) { 4155 *cs->statusp = resp->status = NFS4ERR_INVAL; 4156 goto out; 4157 } 4158 4159 if (len > MAXNAMELEN) { 4160 *cs->statusp = resp->status = NFS4ERR_NAMETOOLONG; 4161 kmem_free(nm, len); 4162 goto out; 4163 } 4164 4165 if (rdonly4(cs->exi, cs->vp, req)) { 4166 *cs->statusp = resp->status = NFS4ERR_ROFS; 4167 kmem_free(nm, len); 4168 goto out; 4169 } 4170 4171 ca = (struct sockaddr *)svc_getrpccaller(req->rq_xprt)->buf; 4172 name = nfscmd_convname(ca, cs->exi, nm, NFSCMD_CONV_INBOUND, 4173 MAXPATHLEN + 1); 4174 4175 if (name == NULL) { 4176 *cs->statusp = resp->status = NFS4ERR_INVAL; 4177 kmem_free(nm, len); 4178 goto out; 4179 } 4180 4181 /* 4182 * Lookup the file to determine type and while we are see if 4183 * there is a file struct around and check for delegation. 4184 * We don't need to acquire va_seq before this lookup, if 4185 * it causes an update, cinfo.before will not match, which will 4186 * trigger a cache flush even if atomic is TRUE. 4187 */ 4188 if (fp = rfs4_lookup_and_findfile(dvp, name, &vp, &error, cs->cr)) { 4189 if (rfs4_check_delegated_byfp(FWRITE, fp, TRUE, TRUE, TRUE, 4190 NULL)) { 4191 VN_RELE(vp); 4192 rfs4_file_rele(fp); 4193 *cs->statusp = resp->status = NFS4ERR_DELAY; 4194 if (nm != name) 4195 kmem_free(name, MAXPATHLEN + 1); 4196 kmem_free(nm, len); 4197 goto out; 4198 } 4199 } 4200 4201 /* Didn't find anything to remove */ 4202 if (vp == NULL) { 4203 *cs->statusp = resp->status = error; 4204 if (nm != name) 4205 kmem_free(name, MAXPATHLEN + 1); 4206 kmem_free(nm, len); 4207 goto out; 4208 } 4209 4210 if (nbl_need_check(vp)) { 4211 nbl_start_crit(vp, RW_READER); 4212 in_crit = 1; 4213 if (nbl_conflict(vp, NBL_REMOVE, 0, 0, 0, NULL)) { 4214 *cs->statusp = resp->status = NFS4ERR_FILE_OPEN; 4215 if (nm != name) 4216 kmem_free(name, MAXPATHLEN + 1); 4217 kmem_free(nm, len); 4218 nbl_end_crit(vp); 4219 VN_RELE(vp); 4220 if (fp) { 4221 rfs4_clear_dont_grant(fp); 4222 rfs4_file_rele(fp); 4223 } 4224 goto out; 4225 } 4226 } 4227 4228 /* check label before allowing removal */ 4229 if (is_system_labeled()) { 4230 ASSERT(req->rq_label != NULL); 4231 clabel = req->rq_label; 4232 DTRACE_PROBE2(tx__rfs4__log__info__opremove__clabel, char *, 4233 "got client label from request(1)", 4234 struct svc_req *, req); 4235 if (!blequal(&l_admin_low->tsl_label, clabel)) { 4236 if (!do_rfs_label_check(clabel, vp, EQUALITY_CHECK, 4237 cs->exi)) { 4238 *cs->statusp = resp->status = NFS4ERR_ACCESS; 4239 if (name != nm) 4240 kmem_free(name, MAXPATHLEN + 1); 4241 kmem_free(nm, len); 4242 if (in_crit) 4243 nbl_end_crit(vp); 4244 VN_RELE(vp); 4245 if (fp) { 4246 rfs4_clear_dont_grant(fp); 4247 rfs4_file_rele(fp); 4248 } 4249 goto out; 4250 } 4251 } 4252 } 4253 4254 /* Get dir "before" change value */ 4255 bdva.va_mask = AT_CTIME|AT_SEQ; 4256 error = VOP_GETATTR(dvp, &bdva, 0, cs->cr, NULL); 4257 if (error) { 4258 *cs->statusp = resp->status = puterrno4(error); 4259 if (nm != name) 4260 kmem_free(name, MAXPATHLEN + 1); 4261 kmem_free(nm, len); 4262 if (in_crit) 4263 nbl_end_crit(vp); 4264 VN_RELE(vp); 4265 if (fp) { 4266 rfs4_clear_dont_grant(fp); 4267 rfs4_file_rele(fp); 4268 } 4269 goto out; 4270 } 4271 NFS4_SET_FATTR4_CHANGE(resp->cinfo.before, bdva.va_ctime) 4272 4273 /* Actually do the REMOVE operation */ 4274 if (vp->v_type == VDIR) { 4275 /* 4276 * Can't remove a directory that has a mounted-on filesystem. 4277 */ 4278 if (vn_ismntpt(vp)) { 4279 error = EACCES; 4280 } else { 4281 /* 4282 * System V defines rmdir to return EEXIST, 4283 * not ENOTEMPTY, if the directory is not 4284 * empty. A System V NFS server needs to map 4285 * NFS4ERR_EXIST to NFS4ERR_NOTEMPTY to 4286 * transmit over the wire. 4287 */ 4288 if ((error = VOP_RMDIR(dvp, name, rootdir, cs->cr, 4289 NULL, 0)) == EEXIST) 4290 error = ENOTEMPTY; 4291 } 4292 } else { 4293 if ((error = VOP_REMOVE(dvp, name, cs->cr, NULL, 0)) == 0 && 4294 fp != NULL) { 4295 struct vattr va; 4296 vnode_t *tvp; 4297 4298 rfs4_dbe_lock(fp->rf_dbe); 4299 tvp = fp->rf_vp; 4300 if (tvp) 4301 VN_HOLD(tvp); 4302 rfs4_dbe_unlock(fp->rf_dbe); 4303 4304 if (tvp) { 4305 /* 4306 * This is va_seq safe because we are not 4307 * manipulating dvp. 4308 */ 4309 va.va_mask = AT_NLINK; 4310 if (!VOP_GETATTR(tvp, &va, 0, cs->cr, NULL) && 4311 va.va_nlink == 0) { 4312 /* Remove state on file remove */ 4313 if (in_crit) { 4314 nbl_end_crit(vp); 4315 in_crit = 0; 4316 } 4317 rfs4_close_all_state(fp); 4318 } 4319 VN_RELE(tvp); 4320 } 4321 } 4322 } 4323 4324 if (in_crit) 4325 nbl_end_crit(vp); 4326 VN_RELE(vp); 4327 4328 if (fp) { 4329 rfs4_clear_dont_grant(fp); 4330 rfs4_file_rele(fp); 4331 } 4332 if (nm != name) 4333 kmem_free(name, MAXPATHLEN + 1); 4334 kmem_free(nm, len); 4335 4336 if (error) { 4337 *cs->statusp = resp->status = puterrno4(error); 4338 goto out; 4339 } 4340 4341 /* 4342 * Get the initial "after" sequence number, if it fails, set to zero 4343 */ 4344 idva.va_mask = AT_SEQ; 4345 if (VOP_GETATTR(dvp, &idva, 0, cs->cr, NULL)) 4346 idva.va_seq = 0; 4347 4348 /* 4349 * Force modified data and metadata out to stable storage. 4350 */ 4351 (void) VOP_FSYNC(dvp, 0, cs->cr, NULL); 4352 4353 /* 4354 * Get "after" change value, if it fails, simply return the 4355 * before value. 4356 */ 4357 adva.va_mask = AT_CTIME|AT_SEQ; 4358 if (VOP_GETATTR(dvp, &adva, 0, cs->cr, NULL)) { 4359 adva.va_ctime = bdva.va_ctime; 4360 adva.va_seq = 0; 4361 } 4362 4363 NFS4_SET_FATTR4_CHANGE(resp->cinfo.after, adva.va_ctime) 4364 4365 /* 4366 * The cinfo.atomic = TRUE only if we have 4367 * non-zero va_seq's, and it has incremented by exactly one 4368 * during the VOP_REMOVE/RMDIR and it didn't change during 4369 * the VOP_FSYNC. 4370 */ 4371 if (bdva.va_seq && idva.va_seq && adva.va_seq && 4372 idva.va_seq == (bdva.va_seq + 1) && idva.va_seq == adva.va_seq) 4373 resp->cinfo.atomic = TRUE; 4374 else 4375 resp->cinfo.atomic = FALSE; 4376 4377 *cs->statusp = resp->status = NFS4_OK; 4378 4379 out: 4380 DTRACE_NFSV4_2(op__remove__done, struct compound_state *, cs, 4381 REMOVE4res *, resp); 4382 } 4383 4384 /* 4385 * rename: args: SAVED_FH: from directory, CURRENT_FH: target directory, 4386 * oldname and newname. 4387 * res: status. If success - CURRENT_FH unchanged, return change_info 4388 * for both from and target directories. 4389 */ 4390 /* ARGSUSED */ 4391 static void 4392 rfs4_op_rename(nfs_argop4 *argop, nfs_resop4 *resop, struct svc_req *req, 4393 struct compound_state *cs) 4394 { 4395 RENAME4args *args = &argop->nfs_argop4_u.oprename; 4396 RENAME4res *resp = &resop->nfs_resop4_u.oprename; 4397 int error; 4398 vnode_t *odvp; 4399 vnode_t *ndvp; 4400 vnode_t *srcvp, *targvp; 4401 struct vattr obdva, oidva, oadva; 4402 struct vattr nbdva, nidva, nadva; 4403 char *onm, *nnm; 4404 uint_t olen, nlen; 4405 rfs4_file_t *fp, *sfp; 4406 int in_crit_src, in_crit_targ; 4407 int fp_rele_grant_hold, sfp_rele_grant_hold; 4408 bslabel_t *clabel; 4409 struct sockaddr *ca; 4410 char *converted_onm = NULL; 4411 char *converted_nnm = NULL; 4412 nfsstat4 status; 4413 4414 DTRACE_NFSV4_2(op__rename__start, struct compound_state *, cs, 4415 RENAME4args *, args); 4416 4417 fp = sfp = NULL; 4418 srcvp = targvp = NULL; 4419 in_crit_src = in_crit_targ = 0; 4420 fp_rele_grant_hold = sfp_rele_grant_hold = 0; 4421 4422 /* CURRENT_FH: target directory */ 4423 ndvp = cs->vp; 4424 if (ndvp == NULL) { 4425 *cs->statusp = resp->status = NFS4ERR_NOFILEHANDLE; 4426 goto out; 4427 } 4428 4429 /* SAVED_FH: from directory */ 4430 odvp = cs->saved_vp; 4431 if (odvp == NULL) { 4432 *cs->statusp = resp->status = NFS4ERR_NOFILEHANDLE; 4433 goto out; 4434 } 4435 4436 if (cs->access == CS_ACCESS_DENIED) { 4437 *cs->statusp = resp->status = NFS4ERR_ACCESS; 4438 goto out; 4439 } 4440 4441 /* 4442 * If there is an unshared filesystem mounted on this vnode, 4443 * do not allow to rename objects in this directory. 4444 */ 4445 if (vn_ismntpt(odvp)) { 4446 *cs->statusp = resp->status = NFS4ERR_ACCESS; 4447 goto out; 4448 } 4449 4450 /* 4451 * If there is an unshared filesystem mounted on this vnode, 4452 * do not allow to rename to this directory. 4453 */ 4454 if (vn_ismntpt(ndvp)) { 4455 *cs->statusp = resp->status = NFS4ERR_ACCESS; 4456 goto out; 4457 } 4458 4459 if (odvp->v_type != VDIR || ndvp->v_type != VDIR) { 4460 *cs->statusp = resp->status = NFS4ERR_NOTDIR; 4461 goto out; 4462 } 4463 4464 if (cs->saved_exi != cs->exi) { 4465 *cs->statusp = resp->status = NFS4ERR_XDEV; 4466 goto out; 4467 } 4468 4469 status = utf8_dir_verify(&args->oldname); 4470 if (status != NFS4_OK) { 4471 *cs->statusp = resp->status = status; 4472 goto out; 4473 } 4474 4475 status = utf8_dir_verify(&args->newname); 4476 if (status != NFS4_OK) { 4477 *cs->statusp = resp->status = status; 4478 goto out; 4479 } 4480 4481 onm = utf8_to_fn(&args->oldname, &olen, NULL); 4482 if (onm == NULL) { 4483 *cs->statusp = resp->status = NFS4ERR_INVAL; 4484 goto out; 4485 } 4486 ca = (struct sockaddr *)svc_getrpccaller(req->rq_xprt)->buf; 4487 nlen = MAXPATHLEN + 1; 4488 converted_onm = nfscmd_convname(ca, cs->exi, onm, NFSCMD_CONV_INBOUND, 4489 nlen); 4490 4491 if (converted_onm == NULL) { 4492 *cs->statusp = resp->status = NFS4ERR_INVAL; 4493 kmem_free(onm, olen); 4494 goto out; 4495 } 4496 4497 nnm = utf8_to_fn(&args->newname, &nlen, NULL); 4498 if (nnm == NULL) { 4499 *cs->statusp = resp->status = NFS4ERR_INVAL; 4500 if (onm != converted_onm) 4501 kmem_free(converted_onm, MAXPATHLEN + 1); 4502 kmem_free(onm, olen); 4503 goto out; 4504 } 4505 converted_nnm = nfscmd_convname(ca, cs->exi, nnm, NFSCMD_CONV_INBOUND, 4506 MAXPATHLEN + 1); 4507 4508 if (converted_nnm == NULL) { 4509 *cs->statusp = resp->status = NFS4ERR_INVAL; 4510 kmem_free(nnm, nlen); 4511 nnm = NULL; 4512 if (onm != converted_onm) 4513 kmem_free(converted_onm, MAXPATHLEN + 1); 4514 kmem_free(onm, olen); 4515 goto out; 4516 } 4517 4518 4519 if (olen > MAXNAMELEN || nlen > MAXNAMELEN) { 4520 *cs->statusp = resp->status = NFS4ERR_NAMETOOLONG; 4521 kmem_free(onm, olen); 4522 kmem_free(nnm, nlen); 4523 goto out; 4524 } 4525 4526 4527 if (rdonly4(cs->exi, cs->vp, req)) { 4528 *cs->statusp = resp->status = NFS4ERR_ROFS; 4529 if (onm != converted_onm) 4530 kmem_free(converted_onm, MAXPATHLEN + 1); 4531 kmem_free(onm, olen); 4532 if (nnm != converted_nnm) 4533 kmem_free(converted_nnm, MAXPATHLEN + 1); 4534 kmem_free(nnm, nlen); 4535 goto out; 4536 } 4537 4538 /* check label of the target dir */ 4539 if (is_system_labeled()) { 4540 ASSERT(req->rq_label != NULL); 4541 clabel = req->rq_label; 4542 DTRACE_PROBE2(tx__rfs4__log__info__oprename__clabel, char *, 4543 "got client label from request(1)", 4544 struct svc_req *, req); 4545 if (!blequal(&l_admin_low->tsl_label, clabel)) { 4546 if (!do_rfs_label_check(clabel, ndvp, 4547 EQUALITY_CHECK, cs->exi)) { 4548 *cs->statusp = resp->status = NFS4ERR_ACCESS; 4549 goto err_out; 4550 } 4551 } 4552 } 4553 4554 /* 4555 * Is the source a file and have a delegation? 4556 * We don't need to acquire va_seq before these lookups, if 4557 * it causes an update, cinfo.before will not match, which will 4558 * trigger a cache flush even if atomic is TRUE. 4559 */ 4560 if (sfp = rfs4_lookup_and_findfile(odvp, converted_onm, &srcvp, 4561 &error, cs->cr)) { 4562 if (rfs4_check_delegated_byfp(FWRITE, sfp, TRUE, TRUE, TRUE, 4563 NULL)) { 4564 *cs->statusp = resp->status = NFS4ERR_DELAY; 4565 goto err_out; 4566 } 4567 } 4568 4569 if (srcvp == NULL) { 4570 *cs->statusp = resp->status = puterrno4(error); 4571 if (onm != converted_onm) 4572 kmem_free(converted_onm, MAXPATHLEN + 1); 4573 kmem_free(onm, olen); 4574 if (nnm != converted_nnm) 4575 kmem_free(converted_nnm, MAXPATHLEN + 1); 4576 kmem_free(nnm, nlen); 4577 goto out; 4578 } 4579 4580 sfp_rele_grant_hold = 1; 4581 4582 /* Does the destination exist and a file and have a delegation? */ 4583 if (fp = rfs4_lookup_and_findfile(ndvp, converted_nnm, &targvp, 4584 NULL, cs->cr)) { 4585 if (rfs4_check_delegated_byfp(FWRITE, fp, TRUE, TRUE, TRUE, 4586 NULL)) { 4587 *cs->statusp = resp->status = NFS4ERR_DELAY; 4588 goto err_out; 4589 } 4590 } 4591 fp_rele_grant_hold = 1; 4592 4593 4594 /* Check for NBMAND lock on both source and target */ 4595 if (nbl_need_check(srcvp)) { 4596 nbl_start_crit(srcvp, RW_READER); 4597 in_crit_src = 1; 4598 if (nbl_conflict(srcvp, NBL_RENAME, 0, 0, 0, NULL)) { 4599 *cs->statusp = resp->status = NFS4ERR_FILE_OPEN; 4600 goto err_out; 4601 } 4602 } 4603 4604 if (targvp && nbl_need_check(targvp)) { 4605 nbl_start_crit(targvp, RW_READER); 4606 in_crit_targ = 1; 4607 if (nbl_conflict(targvp, NBL_REMOVE, 0, 0, 0, NULL)) { 4608 *cs->statusp = resp->status = NFS4ERR_FILE_OPEN; 4609 goto err_out; 4610 } 4611 } 4612 4613 /* Get source "before" change value */ 4614 obdva.va_mask = AT_CTIME|AT_SEQ; 4615 error = VOP_GETATTR(odvp, &obdva, 0, cs->cr, NULL); 4616 if (!error) { 4617 nbdva.va_mask = AT_CTIME|AT_SEQ; 4618 error = VOP_GETATTR(ndvp, &nbdva, 0, cs->cr, NULL); 4619 } 4620 if (error) { 4621 *cs->statusp = resp->status = puterrno4(error); 4622 goto err_out; 4623 } 4624 4625 NFS4_SET_FATTR4_CHANGE(resp->source_cinfo.before, obdva.va_ctime) 4626 NFS4_SET_FATTR4_CHANGE(resp->target_cinfo.before, nbdva.va_ctime) 4627 4628 if ((error = VOP_RENAME(odvp, converted_onm, ndvp, converted_nnm, 4629 cs->cr, NULL, 0)) == 0 && fp != NULL) { 4630 struct vattr va; 4631 vnode_t *tvp; 4632 4633 rfs4_dbe_lock(fp->rf_dbe); 4634 tvp = fp->rf_vp; 4635 if (tvp) 4636 VN_HOLD(tvp); 4637 rfs4_dbe_unlock(fp->rf_dbe); 4638 4639 if (tvp) { 4640 va.va_mask = AT_NLINK; 4641 if (!VOP_GETATTR(tvp, &va, 0, cs->cr, NULL) && 4642 va.va_nlink == 0) { 4643 /* The file is gone and so should the state */ 4644 if (in_crit_targ) { 4645 nbl_end_crit(targvp); 4646 in_crit_targ = 0; 4647 } 4648 rfs4_close_all_state(fp); 4649 } 4650 VN_RELE(tvp); 4651 } 4652 } 4653 if (error == 0) 4654 vn_renamepath(ndvp, srcvp, nnm, nlen - 1); 4655 4656 if (in_crit_src) 4657 nbl_end_crit(srcvp); 4658 if (srcvp) 4659 VN_RELE(srcvp); 4660 if (in_crit_targ) 4661 nbl_end_crit(targvp); 4662 if (targvp) 4663 VN_RELE(targvp); 4664 4665 if (sfp) { 4666 rfs4_clear_dont_grant(sfp); 4667 rfs4_file_rele(sfp); 4668 } 4669 if (fp) { 4670 rfs4_clear_dont_grant(fp); 4671 rfs4_file_rele(fp); 4672 } 4673 4674 if (converted_onm != onm) 4675 kmem_free(converted_onm, MAXPATHLEN + 1); 4676 kmem_free(onm, olen); 4677 if (converted_nnm != nnm) 4678 kmem_free(converted_nnm, MAXPATHLEN + 1); 4679 kmem_free(nnm, nlen); 4680 4681 /* 4682 * Get the initial "after" sequence number, if it fails, set to zero 4683 */ 4684 oidva.va_mask = AT_SEQ; 4685 if (VOP_GETATTR(odvp, &oidva, 0, cs->cr, NULL)) 4686 oidva.va_seq = 0; 4687 4688 nidva.va_mask = AT_SEQ; 4689 if (VOP_GETATTR(ndvp, &nidva, 0, cs->cr, NULL)) 4690 nidva.va_seq = 0; 4691 4692 /* 4693 * Force modified data and metadata out to stable storage. 4694 */ 4695 (void) VOP_FSYNC(odvp, 0, cs->cr, NULL); 4696 (void) VOP_FSYNC(ndvp, 0, cs->cr, NULL); 4697 4698 if (error) { 4699 *cs->statusp = resp->status = puterrno4(error); 4700 goto out; 4701 } 4702 4703 /* 4704 * Get "after" change values, if it fails, simply return the 4705 * before value. 4706 */ 4707 oadva.va_mask = AT_CTIME|AT_SEQ; 4708 if (VOP_GETATTR(odvp, &oadva, 0, cs->cr, NULL)) { 4709 oadva.va_ctime = obdva.va_ctime; 4710 oadva.va_seq = 0; 4711 } 4712 4713 nadva.va_mask = AT_CTIME|AT_SEQ; 4714 if (VOP_GETATTR(odvp, &nadva, 0, cs->cr, NULL)) { 4715 nadva.va_ctime = nbdva.va_ctime; 4716 nadva.va_seq = 0; 4717 } 4718 4719 NFS4_SET_FATTR4_CHANGE(resp->source_cinfo.after, oadva.va_ctime) 4720 NFS4_SET_FATTR4_CHANGE(resp->target_cinfo.after, nadva.va_ctime) 4721 4722 /* 4723 * The cinfo.atomic = TRUE only if we have 4724 * non-zero va_seq's, and it has incremented by exactly one 4725 * during the VOP_RENAME and it didn't change during the VOP_FSYNC. 4726 */ 4727 if (obdva.va_seq && oidva.va_seq && oadva.va_seq && 4728 oidva.va_seq == (obdva.va_seq + 1) && oidva.va_seq == oadva.va_seq) 4729 resp->source_cinfo.atomic = TRUE; 4730 else 4731 resp->source_cinfo.atomic = FALSE; 4732 4733 if (nbdva.va_seq && nidva.va_seq && nadva.va_seq && 4734 nidva.va_seq == (nbdva.va_seq + 1) && nidva.va_seq == nadva.va_seq) 4735 resp->target_cinfo.atomic = TRUE; 4736 else 4737 resp->target_cinfo.atomic = FALSE; 4738 4739 #ifdef VOLATILE_FH_TEST 4740 { 4741 extern void add_volrnm_fh(struct exportinfo *, vnode_t *); 4742 4743 /* 4744 * Add the renamed file handle to the volatile rename list 4745 */ 4746 if (cs->exi->exi_export.ex_flags & EX_VOLRNM) { 4747 /* file handles may expire on rename */ 4748 vnode_t *vp; 4749 4750 nnm = utf8_to_fn(&args->newname, &nlen, NULL); 4751 /* 4752 * Already know that nnm will be a valid string 4753 */ 4754 error = VOP_LOOKUP(ndvp, nnm, &vp, NULL, 0, NULL, cs->cr, 4755 NULL, NULL, NULL); 4756 kmem_free(nnm, nlen); 4757 if (!error) { 4758 add_volrnm_fh(cs->exi, vp); 4759 VN_RELE(vp); 4760 } 4761 } 4762 } 4763 #endif /* VOLATILE_FH_TEST */ 4764 4765 *cs->statusp = resp->status = NFS4_OK; 4766 out: 4767 DTRACE_NFSV4_2(op__rename__done, struct compound_state *, cs, 4768 RENAME4res *, resp); 4769 return; 4770 4771 err_out: 4772 if (onm != converted_onm) 4773 kmem_free(converted_onm, MAXPATHLEN + 1); 4774 if (onm != NULL) 4775 kmem_free(onm, olen); 4776 if (nnm != converted_nnm) 4777 kmem_free(converted_nnm, MAXPATHLEN + 1); 4778 if (nnm != NULL) 4779 kmem_free(nnm, nlen); 4780 4781 if (in_crit_src) nbl_end_crit(srcvp); 4782 if (in_crit_targ) nbl_end_crit(targvp); 4783 if (targvp) VN_RELE(targvp); 4784 if (srcvp) VN_RELE(srcvp); 4785 if (sfp) { 4786 if (sfp_rele_grant_hold) rfs4_clear_dont_grant(sfp); 4787 rfs4_file_rele(sfp); 4788 } 4789 if (fp) { 4790 if (fp_rele_grant_hold) rfs4_clear_dont_grant(fp); 4791 rfs4_file_rele(fp); 4792 } 4793 4794 DTRACE_NFSV4_2(op__rename__done, struct compound_state *, cs, 4795 RENAME4res *, resp); 4796 } 4797 4798 /* ARGSUSED */ 4799 static void 4800 rfs4_op_renew(nfs_argop4 *argop, nfs_resop4 *resop, struct svc_req *req, 4801 struct compound_state *cs) 4802 { 4803 RENEW4args *args = &argop->nfs_argop4_u.oprenew; 4804 RENEW4res *resp = &resop->nfs_resop4_u.oprenew; 4805 rfs4_client_t *cp; 4806 4807 DTRACE_NFSV4_2(op__renew__start, struct compound_state *, cs, 4808 RENEW4args *, args); 4809 4810 if ((cp = rfs4_findclient_by_id(args->clientid, FALSE)) == NULL) { 4811 *cs->statusp = resp->status = 4812 rfs4_check_clientid(&args->clientid, 0); 4813 goto out; 4814 } 4815 4816 if (rfs4_lease_expired(cp)) { 4817 rfs4_client_rele(cp); 4818 *cs->statusp = resp->status = NFS4ERR_EXPIRED; 4819 goto out; 4820 } 4821 4822 rfs4_update_lease(cp); 4823 4824 mutex_enter(cp->rc_cbinfo.cb_lock); 4825 if (cp->rc_cbinfo.cb_notified_of_cb_path_down == FALSE) { 4826 cp->rc_cbinfo.cb_notified_of_cb_path_down = TRUE; 4827 *cs->statusp = resp->status = NFS4ERR_CB_PATH_DOWN; 4828 } else { 4829 *cs->statusp = resp->status = NFS4_OK; 4830 } 4831 mutex_exit(cp->rc_cbinfo.cb_lock); 4832 4833 rfs4_client_rele(cp); 4834 4835 out: 4836 DTRACE_NFSV4_2(op__renew__done, struct compound_state *, cs, 4837 RENEW4res *, resp); 4838 } 4839 4840 /* ARGSUSED */ 4841 static void 4842 rfs4_op_restorefh(nfs_argop4 *args, nfs_resop4 *resop, struct svc_req *req, 4843 struct compound_state *cs) 4844 { 4845 RESTOREFH4res *resp = &resop->nfs_resop4_u.oprestorefh; 4846 4847 DTRACE_NFSV4_1(op__restorefh__start, struct compound_state *, cs); 4848 4849 /* No need to check cs->access - we are not accessing any object */ 4850 if ((cs->saved_vp == NULL) || (cs->saved_fh.nfs_fh4_val == NULL)) { 4851 *cs->statusp = resp->status = NFS4ERR_RESTOREFH; 4852 goto out; 4853 } 4854 if (cs->vp != NULL) { 4855 VN_RELE(cs->vp); 4856 } 4857 cs->vp = cs->saved_vp; 4858 cs->saved_vp = NULL; 4859 cs->exi = cs->saved_exi; 4860 nfs_fh4_copy(&cs->saved_fh, &cs->fh); 4861 *cs->statusp = resp->status = NFS4_OK; 4862 cs->deleg = FALSE; 4863 4864 out: 4865 DTRACE_NFSV4_2(op__restorefh__done, struct compound_state *, cs, 4866 RESTOREFH4res *, resp); 4867 } 4868 4869 /* ARGSUSED */ 4870 static void 4871 rfs4_op_savefh(nfs_argop4 *argop, nfs_resop4 *resop, struct svc_req *req, 4872 struct compound_state *cs) 4873 { 4874 SAVEFH4res *resp = &resop->nfs_resop4_u.opsavefh; 4875 4876 DTRACE_NFSV4_1(op__savefh__start, struct compound_state *, cs); 4877 4878 /* No need to check cs->access - we are not accessing any object */ 4879 if (cs->vp == NULL) { 4880 *cs->statusp = resp->status = NFS4ERR_NOFILEHANDLE; 4881 goto out; 4882 } 4883 if (cs->saved_vp != NULL) { 4884 VN_RELE(cs->saved_vp); 4885 } 4886 cs->saved_vp = cs->vp; 4887 VN_HOLD(cs->saved_vp); 4888 cs->saved_exi = cs->exi; 4889 /* 4890 * since SAVEFH is fairly rare, don't alloc space for its fh 4891 * unless necessary. 4892 */ 4893 if (cs->saved_fh.nfs_fh4_val == NULL) { 4894 cs->saved_fh.nfs_fh4_val = kmem_alloc(NFS4_FHSIZE, KM_SLEEP); 4895 } 4896 nfs_fh4_copy(&cs->fh, &cs->saved_fh); 4897 *cs->statusp = resp->status = NFS4_OK; 4898 4899 out: 4900 DTRACE_NFSV4_2(op__savefh__done, struct compound_state *, cs, 4901 SAVEFH4res *, resp); 4902 } 4903 4904 /* 4905 * rfs4_verify_attr is called when nfsv4 Setattr failed, but we wish to 4906 * return the bitmap of attrs that were set successfully. It is also 4907 * called by Verify/Nverify to test the vattr/vfsstat attrs. It should 4908 * always be called only after rfs4_do_set_attrs(). 4909 * 4910 * Verify that the attributes are same as the expected ones. sargp->vap 4911 * and sargp->sbp contain the input attributes as translated from fattr4. 4912 * 4913 * This function verifies only the attrs that correspond to a vattr or 4914 * vfsstat struct. That is because of the extra step needed to get the 4915 * corresponding system structs. Other attributes have already been set or 4916 * verified by do_rfs4_set_attrs. 4917 * 4918 * Return 0 if all attrs match, -1 if some don't, error if error processing. 4919 */ 4920 static int 4921 rfs4_verify_attr(struct nfs4_svgetit_arg *sargp, 4922 bitmap4 *resp, struct nfs4_ntov_table *ntovp) 4923 { 4924 int error, ret_error = 0; 4925 int i, k; 4926 uint_t sva_mask = sargp->vap->va_mask; 4927 uint_t vbit; 4928 union nfs4_attr_u *na; 4929 uint8_t *amap; 4930 bool_t getsb = ntovp->vfsstat; 4931 4932 if (sva_mask != 0) { 4933 /* 4934 * Okay to overwrite sargp->vap because we verify based 4935 * on the incoming values. 4936 */ 4937 ret_error = VOP_GETATTR(sargp->cs->vp, sargp->vap, 0, 4938 sargp->cs->cr, NULL); 4939 if (ret_error) { 4940 if (resp == NULL) 4941 return (ret_error); 4942 /* 4943 * Must return bitmap of successful attrs 4944 */ 4945 sva_mask = 0; /* to prevent checking vap later */ 4946 } else { 4947 /* 4948 * Some file systems clobber va_mask. it is probably 4949 * wrong of them to do so, nonethless we practice 4950 * defensive coding. 4951 * See bug id 4276830. 4952 */ 4953 sargp->vap->va_mask = sva_mask; 4954 } 4955 } 4956 4957 if (getsb) { 4958 /* 4959 * Now get the superblock and loop on the bitmap, as there is 4960 * no simple way of translating from superblock to bitmap4. 4961 */ 4962 ret_error = VFS_STATVFS(sargp->cs->vp->v_vfsp, sargp->sbp); 4963 if (ret_error) { 4964 if (resp == NULL) 4965 goto errout; 4966 getsb = FALSE; 4967 } 4968 } 4969 4970 /* 4971 * Now loop and verify each attribute which getattr returned 4972 * whether it's the same as the input. 4973 */ 4974 if (resp == NULL && !getsb && (sva_mask == 0)) 4975 goto errout; 4976 4977 na = ntovp->na; 4978 amap = ntovp->amap; 4979 k = 0; 4980 for (i = 0; i < ntovp->attrcnt; i++, na++, amap++) { 4981 k = *amap; 4982 ASSERT(nfs4_ntov_map[k].nval == k); 4983 vbit = nfs4_ntov_map[k].vbit; 4984 4985 /* 4986 * If vattr attribute but VOP_GETATTR failed, or it's 4987 * superblock attribute but VFS_STATVFS failed, skip 4988 */ 4989 if (vbit) { 4990 if ((vbit & sva_mask) == 0) 4991 continue; 4992 } else if (!(getsb && nfs4_ntov_map[k].vfsstat)) { 4993 continue; 4994 } 4995 error = (*nfs4_ntov_map[k].sv_getit)(NFS4ATTR_VERIT, sargp, na); 4996 if (resp != NULL) { 4997 if (error) 4998 ret_error = -1; /* not all match */ 4999 else /* update response bitmap */ 5000 *resp |= nfs4_ntov_map[k].fbit; 5001 continue; 5002 } 5003 if (error) { 5004 ret_error = -1; /* not all match */ 5005 break; 5006 } 5007 } 5008 errout: 5009 return (ret_error); 5010 } 5011 5012 /* 5013 * Decode the attribute to be set/verified. If the attr requires a sys op 5014 * (VOP_GETATTR, VFS_VFSSTAT), and the request is to verify, then don't 5015 * call the sv_getit function for it, because the sys op hasn't yet been done. 5016 * Return 0 for success, error code if failed. 5017 * 5018 * Note: the decoded arg is not freed here but in nfs4_ntov_table_free. 5019 */ 5020 static int 5021 decode_fattr4_attr(nfs4_attr_cmd_t cmd, struct nfs4_svgetit_arg *sargp, 5022 int k, XDR *xdrp, bitmap4 *resp_bval, union nfs4_attr_u *nap) 5023 { 5024 int error = 0; 5025 bool_t set_later; 5026 5027 sargp->vap->va_mask |= nfs4_ntov_map[k].vbit; 5028 5029 if ((*nfs4_ntov_map[k].xfunc)(xdrp, nap)) { 5030 set_later = nfs4_ntov_map[k].vbit || nfs4_ntov_map[k].vfsstat; 5031 /* 5032 * don't verify yet if a vattr or sb dependent attr, 5033 * because we don't have their sys values yet. 5034 * Will be done later. 5035 */ 5036 if (! (set_later && (cmd == NFS4ATTR_VERIT))) { 5037 /* 5038 * ACLs are a special case, since setting the MODE 5039 * conflicts with setting the ACL. We delay setting 5040 * the ACL until all other attributes have been set. 5041 * The ACL gets set in do_rfs4_op_setattr(). 5042 */ 5043 if (nfs4_ntov_map[k].fbit != FATTR4_ACL_MASK) { 5044 error = (*nfs4_ntov_map[k].sv_getit)(cmd, 5045 sargp, nap); 5046 if (error) { 5047 xdr_free(nfs4_ntov_map[k].xfunc, 5048 (caddr_t)nap); 5049 } 5050 } 5051 } 5052 } else { 5053 #ifdef DEBUG 5054 cmn_err(CE_NOTE, "decode_fattr4_attr: error " 5055 "decoding attribute %d\n", k); 5056 #endif 5057 error = EINVAL; 5058 } 5059 if (!error && resp_bval && !set_later) { 5060 *resp_bval |= nfs4_ntov_map[k].fbit; 5061 } 5062 5063 return (error); 5064 } 5065 5066 /* 5067 * Set vattr based on incoming fattr4 attrs - used by setattr. 5068 * Set response mask. Ignore any values that are not writable vattr attrs. 5069 */ 5070 static nfsstat4 5071 do_rfs4_set_attrs(bitmap4 *resp, fattr4 *fattrp, struct compound_state *cs, 5072 struct nfs4_svgetit_arg *sargp, struct nfs4_ntov_table *ntovp, 5073 nfs4_attr_cmd_t cmd) 5074 { 5075 int error = 0; 5076 int i; 5077 char *attrs = fattrp->attrlist4; 5078 uint32_t attrslen = fattrp->attrlist4_len; 5079 XDR xdr; 5080 nfsstat4 status = NFS4_OK; 5081 vnode_t *vp = cs->vp; 5082 union nfs4_attr_u *na; 5083 uint8_t *amap; 5084 5085 #ifndef lint 5086 /* 5087 * Make sure that maximum attribute number can be expressed as an 5088 * 8 bit quantity. 5089 */ 5090 ASSERT(NFS4_MAXNUM_ATTRS <= (UINT8_MAX + 1)); 5091 #endif 5092 5093 if (vp == NULL) { 5094 if (resp) 5095 *resp = 0; 5096 return (NFS4ERR_NOFILEHANDLE); 5097 } 5098 if (cs->access == CS_ACCESS_DENIED) { 5099 if (resp) 5100 *resp = 0; 5101 return (NFS4ERR_ACCESS); 5102 } 5103 5104 sargp->op = cmd; 5105 sargp->cs = cs; 5106 sargp->flag = 0; /* may be set later */ 5107 sargp->vap->va_mask = 0; 5108 sargp->rdattr_error = NFS4_OK; 5109 sargp->rdattr_error_req = FALSE; 5110 /* sargp->sbp is set by the caller */ 5111 5112 xdrmem_create(&xdr, attrs, attrslen, XDR_DECODE); 5113 5114 na = ntovp->na; 5115 amap = ntovp->amap; 5116 5117 /* 5118 * The following loop iterates on the nfs4_ntov_map checking 5119 * if the fbit is set in the requested bitmap. 5120 * If set then we process the arguments using the 5121 * rfs4_fattr4 conversion functions to populate the setattr 5122 * vattr and va_mask. Any settable attrs that are not using vattr 5123 * will be set in this loop. 5124 */ 5125 for (i = 0; i < nfs4_ntov_map_size; i++) { 5126 if (!(fattrp->attrmask & nfs4_ntov_map[i].fbit)) { 5127 continue; 5128 } 5129 /* 5130 * If setattr, must be a writable attr. 5131 * If verify/nverify, must be a readable attr. 5132 */ 5133 if ((error = (*nfs4_ntov_map[i].sv_getit)( 5134 NFS4ATTR_SUPPORTED, sargp, NULL)) != 0) { 5135 /* 5136 * Client tries to set/verify an 5137 * unsupported attribute, tries to set 5138 * a read only attr or verify a write 5139 * only one - error! 5140 */ 5141 break; 5142 } 5143 /* 5144 * Decode the attribute to set/verify 5145 */ 5146 error = decode_fattr4_attr(cmd, sargp, nfs4_ntov_map[i].nval, 5147 &xdr, resp ? resp : NULL, na); 5148 if (error) 5149 break; 5150 *amap++ = (uint8_t)nfs4_ntov_map[i].nval; 5151 na++; 5152 (ntovp->attrcnt)++; 5153 if (nfs4_ntov_map[i].vfsstat) 5154 ntovp->vfsstat = TRUE; 5155 } 5156 5157 if (error != 0) 5158 status = (error == ENOTSUP ? NFS4ERR_ATTRNOTSUPP : 5159 puterrno4(error)); 5160 /* xdrmem_destroy(&xdrs); */ /* NO-OP */ 5161 return (status); 5162 } 5163 5164 static nfsstat4 5165 do_rfs4_op_setattr(bitmap4 *resp, fattr4 *fattrp, struct compound_state *cs, 5166 stateid4 *stateid) 5167 { 5168 int error = 0; 5169 struct nfs4_svgetit_arg sarg; 5170 bool_t trunc; 5171 5172 nfsstat4 status = NFS4_OK; 5173 cred_t *cr = cs->cr; 5174 vnode_t *vp = cs->vp; 5175 struct nfs4_ntov_table ntov; 5176 struct statvfs64 sb; 5177 struct vattr bva; 5178 struct flock64 bf; 5179 int in_crit = 0; 5180 uint_t saved_mask = 0; 5181 caller_context_t ct; 5182 5183 *resp = 0; 5184 sarg.sbp = &sb; 5185 sarg.is_referral = B_FALSE; 5186 nfs4_ntov_table_init(&ntov); 5187 status = do_rfs4_set_attrs(resp, fattrp, cs, &sarg, &ntov, 5188 NFS4ATTR_SETIT); 5189 if (status != NFS4_OK) { 5190 /* 5191 * failed set attrs 5192 */ 5193 goto done; 5194 } 5195 if ((sarg.vap->va_mask == 0) && 5196 (! (fattrp->attrmask & FATTR4_ACL_MASK))) { 5197 /* 5198 * no further work to be done 5199 */ 5200 goto done; 5201 } 5202 5203 /* 5204 * If we got a request to set the ACL and the MODE, only 5205 * allow changing VSUID, VSGID, and VSVTX. Attempting 5206 * to change any other bits, along with setting an ACL, 5207 * gives NFS4ERR_INVAL. 5208 */ 5209 if ((fattrp->attrmask & FATTR4_ACL_MASK) && 5210 (fattrp->attrmask & FATTR4_MODE_MASK)) { 5211 vattr_t va; 5212 5213 va.va_mask = AT_MODE; 5214 error = VOP_GETATTR(vp, &va, 0, cs->cr, NULL); 5215 if (error) { 5216 status = puterrno4(error); 5217 goto done; 5218 } 5219 if ((sarg.vap->va_mode ^ va.va_mode) & 5220 ~(VSUID | VSGID | VSVTX)) { 5221 status = NFS4ERR_INVAL; 5222 goto done; 5223 } 5224 } 5225 5226 /* Check stateid only if size has been set */ 5227 if (sarg.vap->va_mask & AT_SIZE) { 5228 trunc = (sarg.vap->va_size == 0); 5229 status = rfs4_check_stateid(FWRITE, cs->vp, stateid, 5230 trunc, &cs->deleg, sarg.vap->va_mask & AT_SIZE, &ct); 5231 if (status != NFS4_OK) 5232 goto done; 5233 } else { 5234 ct.cc_sysid = 0; 5235 ct.cc_pid = 0; 5236 ct.cc_caller_id = nfs4_srv_caller_id; 5237 ct.cc_flags = CC_DONTBLOCK; 5238 } 5239 5240 /* XXX start of possible race with delegations */ 5241 5242 /* 5243 * We need to specially handle size changes because it is 5244 * possible for the client to create a file with read-only 5245 * modes, but with the file opened for writing. If the client 5246 * then tries to set the file size, e.g. ftruncate(3C), 5247 * fcntl(F_FREESP), the normal access checking done in 5248 * VOP_SETATTR would prevent the client from doing it even though 5249 * it should be allowed to do so. To get around this, we do the 5250 * access checking for ourselves and use VOP_SPACE which doesn't 5251 * do the access checking. 5252 * Also the client should not be allowed to change the file 5253 * size if there is a conflicting non-blocking mandatory lock in 5254 * the region of the change. 5255 */ 5256 if (vp->v_type == VREG && (sarg.vap->va_mask & AT_SIZE)) { 5257 u_offset_t offset; 5258 ssize_t length; 5259 5260 /* 5261 * ufs_setattr clears AT_SIZE from vap->va_mask, but 5262 * before returning, sarg.vap->va_mask is used to 5263 * generate the setattr reply bitmap. We also clear 5264 * AT_SIZE below before calling VOP_SPACE. For both 5265 * of these cases, the va_mask needs to be saved here 5266 * and restored after calling VOP_SETATTR. 5267 */ 5268 saved_mask = sarg.vap->va_mask; 5269 5270 /* 5271 * Check any possible conflict due to NBMAND locks. 5272 * Get into critical region before VOP_GETATTR, so the 5273 * size attribute is valid when checking conflicts. 5274 */ 5275 if (nbl_need_check(vp)) { 5276 nbl_start_crit(vp, RW_READER); 5277 in_crit = 1; 5278 } 5279 5280 bva.va_mask = AT_UID|AT_SIZE; 5281 if (error = VOP_GETATTR(vp, &bva, 0, cr, &ct)) { 5282 status = puterrno4(error); 5283 goto done; 5284 } 5285 5286 if (in_crit) { 5287 if (sarg.vap->va_size < bva.va_size) { 5288 offset = sarg.vap->va_size; 5289 length = bva.va_size - sarg.vap->va_size; 5290 } else { 5291 offset = bva.va_size; 5292 length = sarg.vap->va_size - bva.va_size; 5293 } 5294 if (nbl_conflict(vp, NBL_WRITE, offset, length, 0, 5295 &ct)) { 5296 status = NFS4ERR_LOCKED; 5297 goto done; 5298 } 5299 } 5300 5301 if (crgetuid(cr) == bva.va_uid) { 5302 sarg.vap->va_mask &= ~AT_SIZE; 5303 bf.l_type = F_WRLCK; 5304 bf.l_whence = 0; 5305 bf.l_start = (off64_t)sarg.vap->va_size; 5306 bf.l_len = 0; 5307 bf.l_sysid = 0; 5308 bf.l_pid = 0; 5309 error = VOP_SPACE(vp, F_FREESP, &bf, FWRITE, 5310 (offset_t)sarg.vap->va_size, cr, &ct); 5311 } 5312 } 5313 5314 if (!error && sarg.vap->va_mask != 0) 5315 error = VOP_SETATTR(vp, sarg.vap, sarg.flag, cr, &ct); 5316 5317 /* restore va_mask -- ufs_setattr clears AT_SIZE */ 5318 if (saved_mask & AT_SIZE) 5319 sarg.vap->va_mask |= AT_SIZE; 5320 5321 /* 5322 * If an ACL was being set, it has been delayed until now, 5323 * in order to set the mode (via the VOP_SETATTR() above) first. 5324 */ 5325 if ((! error) && (fattrp->attrmask & FATTR4_ACL_MASK)) { 5326 int i; 5327 5328 for (i = 0; i < NFS4_MAXNUM_ATTRS; i++) 5329 if (ntov.amap[i] == FATTR4_ACL) 5330 break; 5331 if (i < NFS4_MAXNUM_ATTRS) { 5332 error = (*nfs4_ntov_map[FATTR4_ACL].sv_getit)( 5333 NFS4ATTR_SETIT, &sarg, &ntov.na[i]); 5334 if (error == 0) { 5335 *resp |= FATTR4_ACL_MASK; 5336 } else if (error == ENOTSUP) { 5337 (void) rfs4_verify_attr(&sarg, resp, &ntov); 5338 status = NFS4ERR_ATTRNOTSUPP; 5339 goto done; 5340 } 5341 } else { 5342 NFS4_DEBUG(rfs4_debug, 5343 (CE_NOTE, "do_rfs4_op_setattr: " 5344 "unable to find ACL in fattr4")); 5345 error = EINVAL; 5346 } 5347 } 5348 5349 if (error) { 5350 /* check if a monitor detected a delegation conflict */ 5351 if (error == EAGAIN && (ct.cc_flags & CC_WOULDBLOCK)) 5352 status = NFS4ERR_DELAY; 5353 else 5354 status = puterrno4(error); 5355 5356 /* 5357 * Set the response bitmap when setattr failed. 5358 * If VOP_SETATTR partially succeeded, test by doing a 5359 * VOP_GETATTR on the object and comparing the data 5360 * to the setattr arguments. 5361 */ 5362 (void) rfs4_verify_attr(&sarg, resp, &ntov); 5363 } else { 5364 /* 5365 * Force modified metadata out to stable storage. 5366 */ 5367 (void) VOP_FSYNC(vp, FNODSYNC, cr, &ct); 5368 /* 5369 * Set response bitmap 5370 */ 5371 nfs4_vmask_to_nmask_set(sarg.vap->va_mask, resp); 5372 } 5373 5374 /* Return early and already have a NFSv4 error */ 5375 done: 5376 /* 5377 * Except for nfs4_vmask_to_nmask_set(), vattr --> fattr 5378 * conversion sets both readable and writeable NFS4 attrs 5379 * for AT_MTIME and AT_ATIME. The line below masks out 5380 * unrequested attrs from the setattr result bitmap. This 5381 * is placed after the done: label to catch the ATTRNOTSUP 5382 * case. 5383 */ 5384 *resp &= fattrp->attrmask; 5385 5386 if (in_crit) 5387 nbl_end_crit(vp); 5388 5389 nfs4_ntov_table_free(&ntov, &sarg); 5390 5391 return (status); 5392 } 5393 5394 /* ARGSUSED */ 5395 static void 5396 rfs4_op_setattr(nfs_argop4 *argop, nfs_resop4 *resop, struct svc_req *req, 5397 struct compound_state *cs) 5398 { 5399 SETATTR4args *args = &argop->nfs_argop4_u.opsetattr; 5400 SETATTR4res *resp = &resop->nfs_resop4_u.opsetattr; 5401 bslabel_t *clabel; 5402 5403 DTRACE_NFSV4_2(op__setattr__start, struct compound_state *, cs, 5404 SETATTR4args *, args); 5405 5406 if (cs->vp == NULL) { 5407 *cs->statusp = resp->status = NFS4ERR_NOFILEHANDLE; 5408 goto out; 5409 } 5410 5411 /* 5412 * If there is an unshared filesystem mounted on this vnode, 5413 * do not allow to setattr on this vnode. 5414 */ 5415 if (vn_ismntpt(cs->vp)) { 5416 *cs->statusp = resp->status = NFS4ERR_ACCESS; 5417 goto out; 5418 } 5419 5420 resp->attrsset = 0; 5421 5422 if (rdonly4(cs->exi, cs->vp, req)) { 5423 *cs->statusp = resp->status = NFS4ERR_ROFS; 5424 goto out; 5425 } 5426 5427 /* check label before setting attributes */ 5428 if (is_system_labeled()) { 5429 ASSERT(req->rq_label != NULL); 5430 clabel = req->rq_label; 5431 DTRACE_PROBE2(tx__rfs4__log__info__opsetattr__clabel, char *, 5432 "got client label from request(1)", 5433 struct svc_req *, req); 5434 if (!blequal(&l_admin_low->tsl_label, clabel)) { 5435 if (!do_rfs_label_check(clabel, cs->vp, 5436 EQUALITY_CHECK, cs->exi)) { 5437 *cs->statusp = resp->status = NFS4ERR_ACCESS; 5438 goto out; 5439 } 5440 } 5441 } 5442 5443 *cs->statusp = resp->status = 5444 do_rfs4_op_setattr(&resp->attrsset, &args->obj_attributes, cs, 5445 &args->stateid); 5446 5447 out: 5448 DTRACE_NFSV4_2(op__setattr__done, struct compound_state *, cs, 5449 SETATTR4res *, resp); 5450 } 5451 5452 /* ARGSUSED */ 5453 static void 5454 rfs4_op_verify(nfs_argop4 *argop, nfs_resop4 *resop, struct svc_req *req, 5455 struct compound_state *cs) 5456 { 5457 /* 5458 * verify and nverify are exactly the same, except that nverify 5459 * succeeds when some argument changed, and verify succeeds when 5460 * when none changed. 5461 */ 5462 5463 VERIFY4args *args = &argop->nfs_argop4_u.opverify; 5464 VERIFY4res *resp = &resop->nfs_resop4_u.opverify; 5465 5466 int error; 5467 struct nfs4_svgetit_arg sarg; 5468 struct statvfs64 sb; 5469 struct nfs4_ntov_table ntov; 5470 5471 DTRACE_NFSV4_2(op__verify__start, struct compound_state *, cs, 5472 VERIFY4args *, args); 5473 5474 if (cs->vp == NULL) { 5475 *cs->statusp = resp->status = NFS4ERR_NOFILEHANDLE; 5476 goto out; 5477 } 5478 5479 sarg.sbp = &sb; 5480 sarg.is_referral = B_FALSE; 5481 nfs4_ntov_table_init(&ntov); 5482 resp->status = do_rfs4_set_attrs(NULL, &args->obj_attributes, cs, 5483 &sarg, &ntov, NFS4ATTR_VERIT); 5484 if (resp->status != NFS4_OK) { 5485 /* 5486 * do_rfs4_set_attrs will try to verify systemwide attrs, 5487 * so could return -1 for "no match". 5488 */ 5489 if (resp->status == -1) 5490 resp->status = NFS4ERR_NOT_SAME; 5491 goto done; 5492 } 5493 error = rfs4_verify_attr(&sarg, NULL, &ntov); 5494 switch (error) { 5495 case 0: 5496 resp->status = NFS4_OK; 5497 break; 5498 case -1: 5499 resp->status = NFS4ERR_NOT_SAME; 5500 break; 5501 default: 5502 resp->status = puterrno4(error); 5503 break; 5504 } 5505 done: 5506 *cs->statusp = resp->status; 5507 nfs4_ntov_table_free(&ntov, &sarg); 5508 out: 5509 DTRACE_NFSV4_2(op__verify__done, struct compound_state *, cs, 5510 VERIFY4res *, resp); 5511 } 5512 5513 /* ARGSUSED */ 5514 static void 5515 rfs4_op_nverify(nfs_argop4 *argop, nfs_resop4 *resop, struct svc_req *req, 5516 struct compound_state *cs) 5517 { 5518 /* 5519 * verify and nverify are exactly the same, except that nverify 5520 * succeeds when some argument changed, and verify succeeds when 5521 * when none changed. 5522 */ 5523 5524 NVERIFY4args *args = &argop->nfs_argop4_u.opnverify; 5525 NVERIFY4res *resp = &resop->nfs_resop4_u.opnverify; 5526 5527 int error; 5528 struct nfs4_svgetit_arg sarg; 5529 struct statvfs64 sb; 5530 struct nfs4_ntov_table ntov; 5531 5532 DTRACE_NFSV4_2(op__nverify__start, struct compound_state *, cs, 5533 NVERIFY4args *, args); 5534 5535 if (cs->vp == NULL) { 5536 *cs->statusp = resp->status = NFS4ERR_NOFILEHANDLE; 5537 DTRACE_NFSV4_2(op__nverify__done, struct compound_state *, cs, 5538 NVERIFY4res *, resp); 5539 return; 5540 } 5541 sarg.sbp = &sb; 5542 sarg.is_referral = B_FALSE; 5543 nfs4_ntov_table_init(&ntov); 5544 resp->status = do_rfs4_set_attrs(NULL, &args->obj_attributes, cs, 5545 &sarg, &ntov, NFS4ATTR_VERIT); 5546 if (resp->status != NFS4_OK) { 5547 /* 5548 * do_rfs4_set_attrs will try to verify systemwide attrs, 5549 * so could return -1 for "no match". 5550 */ 5551 if (resp->status == -1) 5552 resp->status = NFS4_OK; 5553 goto done; 5554 } 5555 error = rfs4_verify_attr(&sarg, NULL, &ntov); 5556 switch (error) { 5557 case 0: 5558 resp->status = NFS4ERR_SAME; 5559 break; 5560 case -1: 5561 resp->status = NFS4_OK; 5562 break; 5563 default: 5564 resp->status = puterrno4(error); 5565 break; 5566 } 5567 done: 5568 *cs->statusp = resp->status; 5569 nfs4_ntov_table_free(&ntov, &sarg); 5570 5571 DTRACE_NFSV4_2(op__nverify__done, struct compound_state *, cs, 5572 NVERIFY4res *, resp); 5573 } 5574 5575 /* 5576 * XXX - This should live in an NFS header file. 5577 */ 5578 #define MAX_IOVECS 12 5579 5580 /* ARGSUSED */ 5581 static void 5582 rfs4_op_write(nfs_argop4 *argop, nfs_resop4 *resop, struct svc_req *req, 5583 struct compound_state *cs) 5584 { 5585 WRITE4args *args = &argop->nfs_argop4_u.opwrite; 5586 WRITE4res *resp = &resop->nfs_resop4_u.opwrite; 5587 int error; 5588 vnode_t *vp; 5589 struct vattr bva; 5590 u_offset_t rlimit; 5591 struct uio uio; 5592 struct iovec iov[MAX_IOVECS]; 5593 struct iovec *iovp; 5594 int iovcnt; 5595 int ioflag; 5596 cred_t *savecred, *cr; 5597 bool_t *deleg = &cs->deleg; 5598 nfsstat4 stat; 5599 int in_crit = 0; 5600 caller_context_t ct; 5601 5602 DTRACE_NFSV4_2(op__write__start, struct compound_state *, cs, 5603 WRITE4args *, args); 5604 5605 vp = cs->vp; 5606 if (vp == NULL) { 5607 *cs->statusp = resp->status = NFS4ERR_NOFILEHANDLE; 5608 goto out; 5609 } 5610 if (cs->access == CS_ACCESS_DENIED) { 5611 *cs->statusp = resp->status = NFS4ERR_ACCESS; 5612 goto out; 5613 } 5614 5615 cr = cs->cr; 5616 5617 if ((stat = rfs4_check_stateid(FWRITE, vp, &args->stateid, FALSE, 5618 deleg, TRUE, &ct)) != NFS4_OK) { 5619 *cs->statusp = resp->status = stat; 5620 goto out; 5621 } 5622 5623 /* 5624 * We have to enter the critical region before calling VOP_RWLOCK 5625 * to avoid a deadlock with ufs. 5626 */ 5627 if (nbl_need_check(vp)) { 5628 nbl_start_crit(vp, RW_READER); 5629 in_crit = 1; 5630 if (nbl_conflict(vp, NBL_WRITE, 5631 args->offset, args->data_len, 0, &ct)) { 5632 *cs->statusp = resp->status = NFS4ERR_LOCKED; 5633 goto out; 5634 } 5635 } 5636 5637 bva.va_mask = AT_MODE | AT_UID; 5638 error = VOP_GETATTR(vp, &bva, 0, cr, &ct); 5639 5640 /* 5641 * If we can't get the attributes, then we can't do the 5642 * right access checking. So, we'll fail the request. 5643 */ 5644 if (error) { 5645 *cs->statusp = resp->status = puterrno4(error); 5646 goto out; 5647 } 5648 5649 if (rdonly4(cs->exi, cs->vp, req)) { 5650 *cs->statusp = resp->status = NFS4ERR_ROFS; 5651 goto out; 5652 } 5653 5654 if (vp->v_type != VREG) { 5655 *cs->statusp = resp->status = 5656 ((vp->v_type == VDIR) ? NFS4ERR_ISDIR : NFS4ERR_INVAL); 5657 goto out; 5658 } 5659 5660 if (crgetuid(cr) != bva.va_uid && 5661 (error = VOP_ACCESS(vp, VWRITE, 0, cr, &ct))) { 5662 *cs->statusp = resp->status = puterrno4(error); 5663 goto out; 5664 } 5665 5666 if (MANDLOCK(vp, bva.va_mode)) { 5667 *cs->statusp = resp->status = NFS4ERR_ACCESS; 5668 goto out; 5669 } 5670 5671 if (args->data_len == 0) { 5672 *cs->statusp = resp->status = NFS4_OK; 5673 resp->count = 0; 5674 resp->committed = args->stable; 5675 resp->writeverf = Write4verf; 5676 goto out; 5677 } 5678 5679 if (args->mblk != NULL) { 5680 mblk_t *m; 5681 uint_t bytes, round_len; 5682 5683 iovcnt = 0; 5684 bytes = 0; 5685 round_len = roundup(args->data_len, BYTES_PER_XDR_UNIT); 5686 for (m = args->mblk; 5687 m != NULL && bytes < round_len; 5688 m = m->b_cont) { 5689 iovcnt++; 5690 bytes += MBLKL(m); 5691 } 5692 #ifdef DEBUG 5693 /* should have ended on an mblk boundary */ 5694 if (bytes != round_len) { 5695 printf("bytes=0x%x, round_len=0x%x, req len=0x%x\n", 5696 bytes, round_len, args->data_len); 5697 printf("args=%p, args->mblk=%p, m=%p", (void *)args, 5698 (void *)args->mblk, (void *)m); 5699 ASSERT(bytes == round_len); 5700 } 5701 #endif 5702 if (iovcnt <= MAX_IOVECS) { 5703 iovp = iov; 5704 } else { 5705 iovp = kmem_alloc(sizeof (*iovp) * iovcnt, KM_SLEEP); 5706 } 5707 mblk_to_iov(args->mblk, iovcnt, iovp); 5708 } else if (args->rlist != NULL) { 5709 iovcnt = 1; 5710 iovp = iov; 5711 iovp->iov_base = (char *)((args->rlist)->u.c_daddr3); 5712 iovp->iov_len = args->data_len; 5713 } else { 5714 iovcnt = 1; 5715 iovp = iov; 5716 iovp->iov_base = args->data_val; 5717 iovp->iov_len = args->data_len; 5718 } 5719 5720 uio.uio_iov = iovp; 5721 uio.uio_iovcnt = iovcnt; 5722 5723 uio.uio_segflg = UIO_SYSSPACE; 5724 uio.uio_extflg = UIO_COPY_DEFAULT; 5725 uio.uio_loffset = args->offset; 5726 uio.uio_resid = args->data_len; 5727 uio.uio_llimit = curproc->p_fsz_ctl; 5728 rlimit = uio.uio_llimit - args->offset; 5729 if (rlimit < (u_offset_t)uio.uio_resid) 5730 uio.uio_resid = (int)rlimit; 5731 5732 if (args->stable == UNSTABLE4) 5733 ioflag = 0; 5734 else if (args->stable == FILE_SYNC4) 5735 ioflag = FSYNC; 5736 else if (args->stable == DATA_SYNC4) 5737 ioflag = FDSYNC; 5738 else { 5739 if (iovp != iov) 5740 kmem_free(iovp, sizeof (*iovp) * iovcnt); 5741 *cs->statusp = resp->status = NFS4ERR_INVAL; 5742 goto out; 5743 } 5744 5745 /* 5746 * We're changing creds because VM may fault and we need 5747 * the cred of the current thread to be used if quota 5748 * checking is enabled. 5749 */ 5750 savecred = curthread->t_cred; 5751 curthread->t_cred = cr; 5752 error = do_io(FWRITE, vp, &uio, ioflag, cr, &ct); 5753 curthread->t_cred = savecred; 5754 5755 if (iovp != iov) 5756 kmem_free(iovp, sizeof (*iovp) * iovcnt); 5757 5758 if (error) { 5759 *cs->statusp = resp->status = puterrno4(error); 5760 goto out; 5761 } 5762 5763 *cs->statusp = resp->status = NFS4_OK; 5764 resp->count = args->data_len - uio.uio_resid; 5765 5766 if (ioflag == 0) 5767 resp->committed = UNSTABLE4; 5768 else 5769 resp->committed = FILE_SYNC4; 5770 5771 resp->writeverf = Write4verf; 5772 5773 out: 5774 if (in_crit) 5775 nbl_end_crit(vp); 5776 5777 DTRACE_NFSV4_2(op__write__done, struct compound_state *, cs, 5778 WRITE4res *, resp); 5779 } 5780 5781 5782 /* XXX put in a header file */ 5783 extern int sec_svc_getcred(struct svc_req *, cred_t *, caddr_t *, int *); 5784 5785 void 5786 rfs4_compound(COMPOUND4args *args, COMPOUND4res *resp, struct exportinfo *exi, 5787 struct svc_req *req, cred_t *cr, int *rv) 5788 { 5789 uint_t i; 5790 struct compound_state cs; 5791 5792 if (rv != NULL) 5793 *rv = 0; 5794 rfs4_init_compound_state(&cs); 5795 /* 5796 * Form a reply tag by copying over the reqeuest tag. 5797 */ 5798 resp->tag.utf8string_val = 5799 kmem_alloc(args->tag.utf8string_len, KM_SLEEP); 5800 resp->tag.utf8string_len = args->tag.utf8string_len; 5801 bcopy(args->tag.utf8string_val, resp->tag.utf8string_val, 5802 resp->tag.utf8string_len); 5803 5804 cs.statusp = &resp->status; 5805 cs.req = req; 5806 5807 resp->status = utf8_name_verify(&(resp->tag)); 5808 if (resp->status != NFS4_OK) { 5809 resp->array_len = 0; 5810 resp->array = NULL; 5811 return; 5812 } 5813 5814 /* 5815 * XXX for now, minorversion should be zero 5816 */ 5817 if (args->minorversion != NFS4_MINORVERSION) { 5818 DTRACE_NFSV4_2(compound__start, struct compound_state *, 5819 &cs, COMPOUND4args *, args); 5820 resp->array_len = 0; 5821 resp->array = NULL; 5822 resp->status = NFS4ERR_MINOR_VERS_MISMATCH; 5823 DTRACE_NFSV4_2(compound__done, struct compound_state *, 5824 &cs, COMPOUND4res *, resp); 5825 return; 5826 } 5827 5828 ASSERT(exi == NULL); 5829 ASSERT(cr == NULL); 5830 5831 cr = crget(); 5832 ASSERT(cr != NULL); 5833 5834 if (sec_svc_getcred(req, cr, &cs.principal, &cs.nfsflavor) == 0) { 5835 DTRACE_NFSV4_2(compound__start, struct compound_state *, 5836 &cs, COMPOUND4args *, args); 5837 crfree(cr); 5838 DTRACE_NFSV4_2(compound__done, struct compound_state *, 5839 &cs, COMPOUND4res *, resp); 5840 svcerr_badcred(req->rq_xprt); 5841 if (rv != NULL) 5842 *rv = 1; 5843 return; 5844 } 5845 resp->array_len = args->array_len; 5846 resp->array = kmem_zalloc(args->array_len * sizeof (nfs_resop4), 5847 KM_SLEEP); 5848 5849 cs.basecr = cr; 5850 5851 DTRACE_NFSV4_2(compound__start, struct compound_state *, &cs, 5852 COMPOUND4args *, args); 5853 5854 /* 5855 * For now, NFS4 compound processing must be protected by 5856 * exported_lock because it can access more than one exportinfo 5857 * per compound and share/unshare can now change multiple 5858 * exinfo structs. The NFS2/3 code only refs 1 exportinfo 5859 * per proc (excluding public exinfo), and exi_count design 5860 * is sufficient to protect concurrent execution of NFS2/3 5861 * ops along with unexport. This lock will be removed as 5862 * part of the NFSv4 phase 2 namespace redesign work. 5863 */ 5864 rw_enter(&exported_lock, RW_READER); 5865 5866 /* 5867 * If this is the first compound we've seen, we need to start all 5868 * new instances' grace periods. 5869 */ 5870 if (rfs4_seen_first_compound == 0) { 5871 rfs4_grace_start_new(); 5872 /* 5873 * This must be set after rfs4_grace_start_new(), otherwise 5874 * another thread could proceed past here before the former 5875 * is finished. 5876 */ 5877 rfs4_seen_first_compound = 1; 5878 } 5879 5880 for (i = 0; i < args->array_len && cs.cont; i++) { 5881 nfs_argop4 *argop; 5882 nfs_resop4 *resop; 5883 uint_t op; 5884 5885 argop = &args->array[i]; 5886 resop = &resp->array[i]; 5887 resop->resop = argop->argop; 5888 op = (uint_t)resop->resop; 5889 5890 if (op < rfsv4disp_cnt) { 5891 /* 5892 * Count the individual ops here; NULL and COMPOUND 5893 * are counted in common_dispatch() 5894 */ 5895 rfsproccnt_v4_ptr[op].value.ui64++; 5896 5897 NFS4_DEBUG(rfs4_debug > 1, 5898 (CE_NOTE, "Executing %s", rfs4_op_string[op])); 5899 (*rfsv4disptab[op].dis_proc)(argop, resop, req, &cs); 5900 NFS4_DEBUG(rfs4_debug > 1, (CE_NOTE, "%s returned %d", 5901 rfs4_op_string[op], *cs.statusp)); 5902 if (*cs.statusp != NFS4_OK) 5903 cs.cont = FALSE; 5904 } else { 5905 /* 5906 * This is effectively dead code since XDR code 5907 * will have already returned BADXDR if op doesn't 5908 * decode to legal value. This only done for a 5909 * day when XDR code doesn't verify v4 opcodes. 5910 */ 5911 op = OP_ILLEGAL; 5912 rfsproccnt_v4_ptr[OP_ILLEGAL_IDX].value.ui64++; 5913 5914 rfs4_op_illegal(argop, resop, req, &cs); 5915 cs.cont = FALSE; 5916 } 5917 5918 /* 5919 * If not at last op, and if we are to stop, then 5920 * compact the results array. 5921 */ 5922 if ((i + 1) < args->array_len && !cs.cont) { 5923 nfs_resop4 *new_res = kmem_alloc( 5924 (i+1) * sizeof (nfs_resop4), KM_SLEEP); 5925 bcopy(resp->array, 5926 new_res, (i+1) * sizeof (nfs_resop4)); 5927 kmem_free(resp->array, 5928 args->array_len * sizeof (nfs_resop4)); 5929 5930 resp->array_len = i + 1; 5931 resp->array = new_res; 5932 } 5933 } 5934 5935 rw_exit(&exported_lock); 5936 5937 DTRACE_NFSV4_2(compound__done, struct compound_state *, &cs, 5938 COMPOUND4res *, resp); 5939 5940 if (cs.vp) 5941 VN_RELE(cs.vp); 5942 if (cs.saved_vp) 5943 VN_RELE(cs.saved_vp); 5944 if (cs.saved_fh.nfs_fh4_val) 5945 kmem_free(cs.saved_fh.nfs_fh4_val, NFS4_FHSIZE); 5946 5947 if (cs.basecr) 5948 crfree(cs.basecr); 5949 if (cs.cr) 5950 crfree(cs.cr); 5951 /* 5952 * done with this compound request, free the label 5953 */ 5954 5955 if (req->rq_label != NULL) { 5956 kmem_free(req->rq_label, sizeof (bslabel_t)); 5957 req->rq_label = NULL; 5958 } 5959 } 5960 5961 /* 5962 * XXX because of what appears to be duplicate calls to rfs4_compound_free 5963 * XXX zero out the tag and array values. Need to investigate why the 5964 * XXX calls occur, but at least prevent the panic for now. 5965 */ 5966 void 5967 rfs4_compound_free(COMPOUND4res *resp) 5968 { 5969 uint_t i; 5970 5971 if (resp->tag.utf8string_val) { 5972 UTF8STRING_FREE(resp->tag) 5973 } 5974 5975 for (i = 0; i < resp->array_len; i++) { 5976 nfs_resop4 *resop; 5977 uint_t op; 5978 5979 resop = &resp->array[i]; 5980 op = (uint_t)resop->resop; 5981 if (op < rfsv4disp_cnt) { 5982 (*rfsv4disptab[op].dis_resfree)(resop); 5983 } 5984 } 5985 if (resp->array != NULL) { 5986 kmem_free(resp->array, resp->array_len * sizeof (nfs_resop4)); 5987 } 5988 } 5989 5990 /* 5991 * Process the value of the compound request rpc flags, as a bit-AND 5992 * of the individual per-op flags (idempotent, allowork, publicfh_ok) 5993 */ 5994 void 5995 rfs4_compound_flagproc(COMPOUND4args *args, int *flagp) 5996 { 5997 int i; 5998 int flag = RPC_ALL; 5999 6000 for (i = 0; flag && i < args->array_len; i++) { 6001 uint_t op; 6002 6003 op = (uint_t)args->array[i].argop; 6004 6005 if (op < rfsv4disp_cnt) 6006 flag &= rfsv4disptab[op].dis_flags; 6007 else 6008 flag = 0; 6009 } 6010 *flagp = flag; 6011 } 6012 6013 nfsstat4 6014 rfs4_client_sysid(rfs4_client_t *cp, sysid_t *sp) 6015 { 6016 nfsstat4 e; 6017 6018 rfs4_dbe_lock(cp->rc_dbe); 6019 6020 if (cp->rc_sysidt != LM_NOSYSID) { 6021 *sp = cp->rc_sysidt; 6022 e = NFS4_OK; 6023 6024 } else if ((cp->rc_sysidt = lm_alloc_sysidt()) != LM_NOSYSID) { 6025 *sp = cp->rc_sysidt; 6026 e = NFS4_OK; 6027 6028 NFS4_DEBUG(rfs4_debug, (CE_NOTE, 6029 "rfs4_client_sysid: allocated 0x%x\n", *sp)); 6030 } else 6031 e = NFS4ERR_DELAY; 6032 6033 rfs4_dbe_unlock(cp->rc_dbe); 6034 return (e); 6035 } 6036 6037 #if defined(DEBUG) && ! defined(lint) 6038 static void lock_print(char *str, int operation, struct flock64 *flk) 6039 { 6040 char *op, *type; 6041 6042 switch (operation) { 6043 case F_GETLK: op = "F_GETLK"; 6044 break; 6045 case F_SETLK: op = "F_SETLK"; 6046 break; 6047 case F_SETLK_NBMAND: op = "F_SETLK_NBMAND"; 6048 break; 6049 default: op = "F_UNKNOWN"; 6050 break; 6051 } 6052 switch (flk->l_type) { 6053 case F_UNLCK: type = "F_UNLCK"; 6054 break; 6055 case F_RDLCK: type = "F_RDLCK"; 6056 break; 6057 case F_WRLCK: type = "F_WRLCK"; 6058 break; 6059 default: type = "F_UNKNOWN"; 6060 break; 6061 } 6062 6063 ASSERT(flk->l_whence == 0); 6064 cmn_err(CE_NOTE, "%s: %s, type = %s, off = %llx len = %llx pid = %d", 6065 str, op, type, (longlong_t)flk->l_start, 6066 flk->l_len ? (longlong_t)flk->l_len : ~0LL, flk->l_pid); 6067 } 6068 6069 #define LOCK_PRINT(d, s, t, f) if (d) lock_print(s, t, f) 6070 #else 6071 #define LOCK_PRINT(d, s, t, f) 6072 #endif 6073 6074 /*ARGSUSED*/ 6075 static bool_t 6076 creds_ok(cred_set_t cr_set, struct svc_req *req, struct compound_state *cs) 6077 { 6078 return (TRUE); 6079 } 6080 6081 /* 6082 * Look up the pathname using the vp in cs as the directory vnode. 6083 * cs->vp will be the vnode for the file on success 6084 */ 6085 6086 static nfsstat4 6087 rfs4_lookup(component4 *component, struct svc_req *req, 6088 struct compound_state *cs) 6089 { 6090 char *nm; 6091 uint32_t len; 6092 nfsstat4 status; 6093 struct sockaddr *ca; 6094 char *name; 6095 6096 if (cs->vp == NULL) { 6097 return (NFS4ERR_NOFILEHANDLE); 6098 } 6099 if (cs->vp->v_type != VDIR) { 6100 return (NFS4ERR_NOTDIR); 6101 } 6102 6103 status = utf8_dir_verify(component); 6104 if (status != NFS4_OK) 6105 return (status); 6106 6107 nm = utf8_to_fn(component, &len, NULL); 6108 if (nm == NULL) { 6109 return (NFS4ERR_INVAL); 6110 } 6111 6112 if (len > MAXNAMELEN) { 6113 kmem_free(nm, len); 6114 return (NFS4ERR_NAMETOOLONG); 6115 } 6116 6117 ca = (struct sockaddr *)svc_getrpccaller(req->rq_xprt)->buf; 6118 name = nfscmd_convname(ca, cs->exi, nm, NFSCMD_CONV_INBOUND, 6119 MAXPATHLEN + 1); 6120 6121 if (name == NULL) { 6122 kmem_free(nm, len); 6123 return (NFS4ERR_INVAL); 6124 } 6125 6126 status = do_rfs4_op_lookup(name, req, cs); 6127 6128 if (name != nm) 6129 kmem_free(name, MAXPATHLEN + 1); 6130 6131 kmem_free(nm, len); 6132 6133 return (status); 6134 } 6135 6136 static nfsstat4 6137 rfs4_lookupfile(component4 *component, struct svc_req *req, 6138 struct compound_state *cs, uint32_t access, change_info4 *cinfo) 6139 { 6140 nfsstat4 status; 6141 vnode_t *dvp = cs->vp; 6142 vattr_t bva, ava, fva; 6143 int error; 6144 6145 /* Get "before" change value */ 6146 bva.va_mask = AT_CTIME|AT_SEQ; 6147 error = VOP_GETATTR(dvp, &bva, 0, cs->cr, NULL); 6148 if (error) 6149 return (puterrno4(error)); 6150 6151 /* rfs4_lookup may VN_RELE directory */ 6152 VN_HOLD(dvp); 6153 6154 status = rfs4_lookup(component, req, cs); 6155 if (status != NFS4_OK) { 6156 VN_RELE(dvp); 6157 return (status); 6158 } 6159 6160 /* 6161 * Get "after" change value, if it fails, simply return the 6162 * before value. 6163 */ 6164 ava.va_mask = AT_CTIME|AT_SEQ; 6165 if (VOP_GETATTR(dvp, &ava, 0, cs->cr, NULL)) { 6166 ava.va_ctime = bva.va_ctime; 6167 ava.va_seq = 0; 6168 } 6169 VN_RELE(dvp); 6170 6171 /* 6172 * Validate the file is a file 6173 */ 6174 fva.va_mask = AT_TYPE|AT_MODE; 6175 error = VOP_GETATTR(cs->vp, &fva, 0, cs->cr, NULL); 6176 if (error) 6177 return (puterrno4(error)); 6178 6179 if (fva.va_type != VREG) { 6180 if (fva.va_type == VDIR) 6181 return (NFS4ERR_ISDIR); 6182 if (fva.va_type == VLNK) 6183 return (NFS4ERR_SYMLINK); 6184 return (NFS4ERR_INVAL); 6185 } 6186 6187 NFS4_SET_FATTR4_CHANGE(cinfo->before, bva.va_ctime); 6188 NFS4_SET_FATTR4_CHANGE(cinfo->after, ava.va_ctime); 6189 6190 /* 6191 * It is undefined if VOP_LOOKUP will change va_seq, so 6192 * cinfo.atomic = TRUE only if we have 6193 * non-zero va_seq's, and they have not changed. 6194 */ 6195 if (bva.va_seq && ava.va_seq && ava.va_seq == bva.va_seq) 6196 cinfo->atomic = TRUE; 6197 else 6198 cinfo->atomic = FALSE; 6199 6200 /* Check for mandatory locking */ 6201 cs->mandlock = MANDLOCK(cs->vp, fva.va_mode); 6202 return (check_open_access(access, cs, req)); 6203 } 6204 6205 static nfsstat4 6206 create_vnode(vnode_t *dvp, char *nm, vattr_t *vap, createmode4 mode, 6207 timespec32_t *mtime, cred_t *cr, vnode_t **vpp, bool_t *created) 6208 { 6209 int error; 6210 nfsstat4 status = NFS4_OK; 6211 vattr_t va; 6212 6213 tryagain: 6214 6215 /* 6216 * The file open mode used is VWRITE. If the client needs 6217 * some other semantic, then it should do the access checking 6218 * itself. It would have been nice to have the file open mode 6219 * passed as part of the arguments. 6220 */ 6221 6222 *created = TRUE; 6223 error = VOP_CREATE(dvp, nm, vap, EXCL, VWRITE, vpp, cr, 0, NULL, NULL); 6224 6225 if (error) { 6226 *created = FALSE; 6227 6228 /* 6229 * If we got something other than file already exists 6230 * then just return this error. Otherwise, we got 6231 * EEXIST. If we were doing a GUARDED create, then 6232 * just return this error. Otherwise, we need to 6233 * make sure that this wasn't a duplicate of an 6234 * exclusive create request. 6235 * 6236 * The assumption is made that a non-exclusive create 6237 * request will never return EEXIST. 6238 */ 6239 6240 if (error != EEXIST || mode == GUARDED4) { 6241 status = puterrno4(error); 6242 return (status); 6243 } 6244 error = VOP_LOOKUP(dvp, nm, vpp, NULL, 0, NULL, cr, 6245 NULL, NULL, NULL); 6246 6247 if (error) { 6248 /* 6249 * We couldn't find the file that we thought that 6250 * we just created. So, we'll just try creating 6251 * it again. 6252 */ 6253 if (error == ENOENT) 6254 goto tryagain; 6255 6256 status = puterrno4(error); 6257 return (status); 6258 } 6259 6260 if (mode == UNCHECKED4) { 6261 /* existing object must be regular file */ 6262 if ((*vpp)->v_type != VREG) { 6263 if ((*vpp)->v_type == VDIR) 6264 status = NFS4ERR_ISDIR; 6265 else if ((*vpp)->v_type == VLNK) 6266 status = NFS4ERR_SYMLINK; 6267 else 6268 status = NFS4ERR_INVAL; 6269 VN_RELE(*vpp); 6270 return (status); 6271 } 6272 6273 return (NFS4_OK); 6274 } 6275 6276 /* Check for duplicate request */ 6277 ASSERT(mtime != 0); 6278 va.va_mask = AT_MTIME; 6279 error = VOP_GETATTR(*vpp, &va, 0, cr, NULL); 6280 if (!error) { 6281 /* We found the file */ 6282 if (va.va_mtime.tv_sec != mtime->tv_sec || 6283 va.va_mtime.tv_nsec != mtime->tv_nsec) { 6284 /* but its not our creation */ 6285 VN_RELE(*vpp); 6286 return (NFS4ERR_EXIST); 6287 } 6288 *created = TRUE; /* retrans of create == created */ 6289 return (NFS4_OK); 6290 } 6291 VN_RELE(*vpp); 6292 return (NFS4ERR_EXIST); 6293 } 6294 6295 return (NFS4_OK); 6296 } 6297 6298 static nfsstat4 6299 check_open_access(uint32_t access, struct compound_state *cs, 6300 struct svc_req *req) 6301 { 6302 int error; 6303 vnode_t *vp; 6304 bool_t readonly; 6305 cred_t *cr = cs->cr; 6306 6307 /* For now we don't allow mandatory locking as per V2/V3 */ 6308 if (cs->access == CS_ACCESS_DENIED || cs->mandlock) { 6309 return (NFS4ERR_ACCESS); 6310 } 6311 6312 vp = cs->vp; 6313 ASSERT(cr != NULL && vp->v_type == VREG); 6314 6315 /* 6316 * If the file system is exported read only and we are trying 6317 * to open for write, then return NFS4ERR_ROFS 6318 */ 6319 6320 readonly = rdonly4(cs->exi, cs->vp, req); 6321 6322 if ((access & OPEN4_SHARE_ACCESS_WRITE) && readonly) 6323 return (NFS4ERR_ROFS); 6324 6325 if (access & OPEN4_SHARE_ACCESS_READ) { 6326 if ((VOP_ACCESS(vp, VREAD, 0, cr, NULL) != 0) && 6327 (VOP_ACCESS(vp, VEXEC, 0, cr, NULL) != 0)) { 6328 return (NFS4ERR_ACCESS); 6329 } 6330 } 6331 6332 if (access & OPEN4_SHARE_ACCESS_WRITE) { 6333 error = VOP_ACCESS(vp, VWRITE, 0, cr, NULL); 6334 if (error) 6335 return (NFS4ERR_ACCESS); 6336 } 6337 6338 return (NFS4_OK); 6339 } 6340 6341 static nfsstat4 6342 rfs4_createfile(OPEN4args *args, struct svc_req *req, struct compound_state *cs, 6343 change_info4 *cinfo, bitmap4 *attrset, clientid4 clientid) 6344 { 6345 struct nfs4_svgetit_arg sarg; 6346 struct nfs4_ntov_table ntov; 6347 6348 bool_t ntov_table_init = FALSE; 6349 struct statvfs64 sb; 6350 nfsstat4 status; 6351 vnode_t *vp; 6352 vattr_t bva, ava, iva, cva, *vap; 6353 vnode_t *dvp; 6354 timespec32_t *mtime; 6355 char *nm = NULL; 6356 uint_t buflen; 6357 bool_t created; 6358 bool_t setsize = FALSE; 6359 len_t reqsize; 6360 int error; 6361 bool_t trunc; 6362 caller_context_t ct; 6363 component4 *component; 6364 bslabel_t *clabel; 6365 struct sockaddr *ca; 6366 char *name = NULL; 6367 6368 sarg.sbp = &sb; 6369 sarg.is_referral = B_FALSE; 6370 6371 dvp = cs->vp; 6372 6373 /* Check if the file system is read only */ 6374 if (rdonly4(cs->exi, dvp, req)) 6375 return (NFS4ERR_ROFS); 6376 6377 /* check the label of including directory */ 6378 if (is_system_labeled()) { 6379 ASSERT(req->rq_label != NULL); 6380 clabel = req->rq_label; 6381 DTRACE_PROBE2(tx__rfs4__log__info__opremove__clabel, char *, 6382 "got client label from request(1)", 6383 struct svc_req *, req); 6384 if (!blequal(&l_admin_low->tsl_label, clabel)) { 6385 if (!do_rfs_label_check(clabel, dvp, EQUALITY_CHECK, 6386 cs->exi)) { 6387 return (NFS4ERR_ACCESS); 6388 } 6389 } 6390 } 6391 6392 /* 6393 * Get the last component of path name in nm. cs will reference 6394 * the including directory on success. 6395 */ 6396 component = &args->open_claim4_u.file; 6397 status = utf8_dir_verify(component); 6398 if (status != NFS4_OK) 6399 return (status); 6400 6401 nm = utf8_to_fn(component, &buflen, NULL); 6402 6403 if (nm == NULL) 6404 return (NFS4ERR_RESOURCE); 6405 6406 if (buflen > MAXNAMELEN) { 6407 kmem_free(nm, buflen); 6408 return (NFS4ERR_NAMETOOLONG); 6409 } 6410 6411 bva.va_mask = AT_TYPE|AT_CTIME|AT_SEQ; 6412 error = VOP_GETATTR(dvp, &bva, 0, cs->cr, NULL); 6413 if (error) { 6414 kmem_free(nm, buflen); 6415 return (puterrno4(error)); 6416 } 6417 6418 if (bva.va_type != VDIR) { 6419 kmem_free(nm, buflen); 6420 return (NFS4ERR_NOTDIR); 6421 } 6422 6423 NFS4_SET_FATTR4_CHANGE(cinfo->before, bva.va_ctime) 6424 6425 switch (args->mode) { 6426 case GUARDED4: 6427 /*FALLTHROUGH*/ 6428 case UNCHECKED4: 6429 nfs4_ntov_table_init(&ntov); 6430 ntov_table_init = TRUE; 6431 6432 *attrset = 0; 6433 status = do_rfs4_set_attrs(attrset, 6434 &args->createhow4_u.createattrs, 6435 cs, &sarg, &ntov, NFS4ATTR_SETIT); 6436 6437 if (status == NFS4_OK && (sarg.vap->va_mask & AT_TYPE) && 6438 sarg.vap->va_type != VREG) { 6439 if (sarg.vap->va_type == VDIR) 6440 status = NFS4ERR_ISDIR; 6441 else if (sarg.vap->va_type == VLNK) 6442 status = NFS4ERR_SYMLINK; 6443 else 6444 status = NFS4ERR_INVAL; 6445 } 6446 6447 if (status != NFS4_OK) { 6448 kmem_free(nm, buflen); 6449 nfs4_ntov_table_free(&ntov, &sarg); 6450 *attrset = 0; 6451 return (status); 6452 } 6453 6454 vap = sarg.vap; 6455 vap->va_type = VREG; 6456 vap->va_mask |= AT_TYPE; 6457 6458 if ((vap->va_mask & AT_MODE) == 0) { 6459 vap->va_mask |= AT_MODE; 6460 vap->va_mode = (mode_t)0600; 6461 } 6462 6463 if (vap->va_mask & AT_SIZE) { 6464 6465 /* Disallow create with a non-zero size */ 6466 6467 if ((reqsize = sarg.vap->va_size) != 0) { 6468 kmem_free(nm, buflen); 6469 nfs4_ntov_table_free(&ntov, &sarg); 6470 *attrset = 0; 6471 return (NFS4ERR_INVAL); 6472 } 6473 setsize = TRUE; 6474 } 6475 break; 6476 6477 case EXCLUSIVE4: 6478 /* prohibit EXCL create of named attributes */ 6479 if (dvp->v_flag & V_XATTRDIR) { 6480 kmem_free(nm, buflen); 6481 *attrset = 0; 6482 return (NFS4ERR_INVAL); 6483 } 6484 6485 cva.va_mask = AT_TYPE | AT_MTIME | AT_MODE; 6486 cva.va_type = VREG; 6487 /* 6488 * Ensure no time overflows. Assumes underlying 6489 * filesystem supports at least 32 bits. 6490 * Truncate nsec to usec resolution to allow valid 6491 * compares even if the underlying filesystem truncates. 6492 */ 6493 mtime = (timespec32_t *)&args->createhow4_u.createverf; 6494 cva.va_mtime.tv_sec = mtime->tv_sec % TIME32_MAX; 6495 cva.va_mtime.tv_nsec = (mtime->tv_nsec / 1000) * 1000; 6496 cva.va_mode = (mode_t)0; 6497 vap = &cva; 6498 6499 /* 6500 * For EXCL create, attrset is set to the server attr 6501 * used to cache the client's verifier. 6502 */ 6503 *attrset = FATTR4_TIME_MODIFY_MASK; 6504 break; 6505 } 6506 6507 ca = (struct sockaddr *)svc_getrpccaller(req->rq_xprt)->buf; 6508 name = nfscmd_convname(ca, cs->exi, nm, NFSCMD_CONV_INBOUND, 6509 MAXPATHLEN + 1); 6510 6511 if (name == NULL) { 6512 kmem_free(nm, buflen); 6513 return (NFS4ERR_SERVERFAULT); 6514 } 6515 6516 status = create_vnode(dvp, name, vap, args->mode, mtime, 6517 cs->cr, &vp, &created); 6518 if (nm != name) 6519 kmem_free(name, MAXPATHLEN + 1); 6520 kmem_free(nm, buflen); 6521 6522 if (status != NFS4_OK) { 6523 if (ntov_table_init) 6524 nfs4_ntov_table_free(&ntov, &sarg); 6525 *attrset = 0; 6526 return (status); 6527 } 6528 6529 trunc = (setsize && !created); 6530 6531 if (args->mode != EXCLUSIVE4) { 6532 bitmap4 createmask = args->createhow4_u.createattrs.attrmask; 6533 6534 /* 6535 * True verification that object was created with correct 6536 * attrs is impossible. The attrs could have been changed 6537 * immediately after object creation. If attributes did 6538 * not verify, the only recourse for the server is to 6539 * destroy the object. Maybe if some attrs (like gid) 6540 * are set incorrectly, the object should be destroyed; 6541 * however, seems bad as a default policy. Do we really 6542 * want to destroy an object over one of the times not 6543 * verifying correctly? For these reasons, the server 6544 * currently sets bits in attrset for createattrs 6545 * that were set; however, no verification is done. 6546 * 6547 * vmask_to_nmask accounts for vattr bits set on create 6548 * [do_rfs4_set_attrs() only sets resp bits for 6549 * non-vattr/vfs bits.] 6550 * Mask off any bits we set by default so as not to return 6551 * more attrset bits than were requested in createattrs 6552 */ 6553 if (created) { 6554 nfs4_vmask_to_nmask(sarg.vap->va_mask, attrset); 6555 *attrset &= createmask; 6556 } else { 6557 /* 6558 * We did not create the vnode (we tried but it 6559 * already existed). In this case, the only createattr 6560 * that the spec allows the server to set is size, 6561 * and even then, it can only be set if it is 0. 6562 */ 6563 *attrset = 0; 6564 if (trunc) 6565 *attrset = FATTR4_SIZE_MASK; 6566 } 6567 } 6568 if (ntov_table_init) 6569 nfs4_ntov_table_free(&ntov, &sarg); 6570 6571 /* 6572 * Get the initial "after" sequence number, if it fails, 6573 * set to zero, time to before. 6574 */ 6575 iva.va_mask = AT_CTIME|AT_SEQ; 6576 if (VOP_GETATTR(dvp, &iva, 0, cs->cr, NULL)) { 6577 iva.va_seq = 0; 6578 iva.va_ctime = bva.va_ctime; 6579 } 6580 6581 /* 6582 * create_vnode attempts to create the file exclusive, 6583 * if it already exists the VOP_CREATE will fail and 6584 * may not increase va_seq. It is atomic if 6585 * we haven't changed the directory, but if it has changed 6586 * we don't know what changed it. 6587 */ 6588 if (!created) { 6589 if (bva.va_seq && iva.va_seq && 6590 bva.va_seq == iva.va_seq) 6591 cinfo->atomic = TRUE; 6592 else 6593 cinfo->atomic = FALSE; 6594 NFS4_SET_FATTR4_CHANGE(cinfo->after, iva.va_ctime); 6595 } else { 6596 /* 6597 * The entry was created, we need to sync the 6598 * directory metadata. 6599 */ 6600 (void) VOP_FSYNC(dvp, 0, cs->cr, NULL); 6601 6602 /* 6603 * Get "after" change value, if it fails, simply return the 6604 * before value. 6605 */ 6606 ava.va_mask = AT_CTIME|AT_SEQ; 6607 if (VOP_GETATTR(dvp, &ava, 0, cs->cr, NULL)) { 6608 ava.va_ctime = bva.va_ctime; 6609 ava.va_seq = 0; 6610 } 6611 6612 NFS4_SET_FATTR4_CHANGE(cinfo->after, ava.va_ctime); 6613 6614 /* 6615 * The cinfo->atomic = TRUE only if we have 6616 * non-zero va_seq's, and it has incremented by exactly one 6617 * during the create_vnode and it didn't 6618 * change during the VOP_FSYNC. 6619 */ 6620 if (bva.va_seq && iva.va_seq && ava.va_seq && 6621 iva.va_seq == (bva.va_seq + 1) && iva.va_seq == ava.va_seq) 6622 cinfo->atomic = TRUE; 6623 else 6624 cinfo->atomic = FALSE; 6625 } 6626 6627 /* Check for mandatory locking and that the size gets set. */ 6628 cva.va_mask = AT_MODE; 6629 if (setsize) 6630 cva.va_mask |= AT_SIZE; 6631 6632 /* Assume the worst */ 6633 cs->mandlock = TRUE; 6634 6635 if (VOP_GETATTR(vp, &cva, 0, cs->cr, NULL) == 0) { 6636 cs->mandlock = MANDLOCK(cs->vp, cva.va_mode); 6637 6638 /* 6639 * Truncate the file if necessary; this would be 6640 * the case for create over an existing file. 6641 */ 6642 6643 if (trunc) { 6644 int in_crit = 0; 6645 rfs4_file_t *fp; 6646 bool_t create = FALSE; 6647 6648 /* 6649 * We are writing over an existing file. 6650 * Check to see if we need to recall a delegation. 6651 */ 6652 rfs4_hold_deleg_policy(); 6653 if ((fp = rfs4_findfile(vp, NULL, &create)) != NULL) { 6654 if (rfs4_check_delegated_byfp(FWRITE, fp, 6655 (reqsize == 0), FALSE, FALSE, &clientid)) { 6656 rfs4_file_rele(fp); 6657 rfs4_rele_deleg_policy(); 6658 VN_RELE(vp); 6659 *attrset = 0; 6660 return (NFS4ERR_DELAY); 6661 } 6662 rfs4_file_rele(fp); 6663 } 6664 rfs4_rele_deleg_policy(); 6665 6666 if (nbl_need_check(vp)) { 6667 in_crit = 1; 6668 6669 ASSERT(reqsize == 0); 6670 6671 nbl_start_crit(vp, RW_READER); 6672 if (nbl_conflict(vp, NBL_WRITE, 0, 6673 cva.va_size, 0, NULL)) { 6674 in_crit = 0; 6675 nbl_end_crit(vp); 6676 VN_RELE(vp); 6677 *attrset = 0; 6678 return (NFS4ERR_ACCESS); 6679 } 6680 } 6681 ct.cc_sysid = 0; 6682 ct.cc_pid = 0; 6683 ct.cc_caller_id = nfs4_srv_caller_id; 6684 ct.cc_flags = CC_DONTBLOCK; 6685 6686 cva.va_mask = AT_SIZE; 6687 cva.va_size = reqsize; 6688 (void) VOP_SETATTR(vp, &cva, 0, cs->cr, &ct); 6689 if (in_crit) 6690 nbl_end_crit(vp); 6691 } 6692 } 6693 6694 error = makefh4(&cs->fh, vp, cs->exi); 6695 6696 /* 6697 * Force modified data and metadata out to stable storage. 6698 */ 6699 (void) VOP_FSYNC(vp, FNODSYNC, cs->cr, NULL); 6700 6701 if (error) { 6702 VN_RELE(vp); 6703 *attrset = 0; 6704 return (puterrno4(error)); 6705 } 6706 6707 /* if parent dir is attrdir, set namedattr fh flag */ 6708 if (dvp->v_flag & V_XATTRDIR) 6709 set_fh4_flag(&cs->fh, FH4_NAMEDATTR); 6710 6711 if (cs->vp) 6712 VN_RELE(cs->vp); 6713 6714 cs->vp = vp; 6715 6716 /* 6717 * if we did not create the file, we will need to check 6718 * the access bits on the file 6719 */ 6720 6721 if (!created) { 6722 if (setsize) 6723 args->share_access |= OPEN4_SHARE_ACCESS_WRITE; 6724 status = check_open_access(args->share_access, cs, req); 6725 if (status != NFS4_OK) 6726 *attrset = 0; 6727 } 6728 return (status); 6729 } 6730 6731 /*ARGSUSED*/ 6732 static void 6733 rfs4_do_open(struct compound_state *cs, struct svc_req *req, 6734 rfs4_openowner_t *oo, delegreq_t deleg, 6735 uint32_t access, uint32_t deny, 6736 OPEN4res *resp, int deleg_cur) 6737 { 6738 /* XXX Currently not using req */ 6739 rfs4_state_t *sp; 6740 rfs4_file_t *fp; 6741 bool_t screate = TRUE; 6742 bool_t fcreate = TRUE; 6743 uint32_t open_a, share_a; 6744 uint32_t open_d, share_d; 6745 rfs4_deleg_state_t *dsp; 6746 sysid_t sysid; 6747 nfsstat4 status; 6748 caller_context_t ct; 6749 int fflags = 0; 6750 int recall = 0; 6751 int err; 6752 int first_open; 6753 6754 /* get the file struct and hold a lock on it during initial open */ 6755 fp = rfs4_findfile_withlock(cs->vp, &cs->fh, &fcreate); 6756 if (fp == NULL) { 6757 resp->status = NFS4ERR_RESOURCE; 6758 DTRACE_PROBE1(nfss__e__do__open1, nfsstat4, resp->status); 6759 return; 6760 } 6761 6762 sp = rfs4_findstate_by_owner_file(oo, fp, &screate); 6763 if (sp == NULL) { 6764 resp->status = NFS4ERR_RESOURCE; 6765 DTRACE_PROBE1(nfss__e__do__open2, nfsstat4, resp->status); 6766 /* No need to keep any reference */ 6767 rw_exit(&fp->rf_file_rwlock); 6768 rfs4_file_rele(fp); 6769 return; 6770 } 6771 6772 /* try to get the sysid before continuing */ 6773 if ((status = rfs4_client_sysid(oo->ro_client, &sysid)) != NFS4_OK) { 6774 resp->status = status; 6775 rfs4_file_rele(fp); 6776 /* Not a fully formed open; "close" it */ 6777 if (screate == TRUE) 6778 rfs4_state_close(sp, FALSE, FALSE, cs->cr); 6779 rfs4_state_rele(sp); 6780 return; 6781 } 6782 6783 /* Calculate the fflags for this OPEN. */ 6784 if (access & OPEN4_SHARE_ACCESS_READ) 6785 fflags |= FREAD; 6786 if (access & OPEN4_SHARE_ACCESS_WRITE) 6787 fflags |= FWRITE; 6788 6789 rfs4_dbe_lock(sp->rs_dbe); 6790 6791 /* 6792 * Calculate the new deny and access mode that this open is adding to 6793 * the file for this open owner; 6794 */ 6795 open_d = (deny & ~sp->rs_open_deny); 6796 open_a = (access & ~sp->rs_open_access); 6797 6798 /* 6799 * Calculate the new share access and share deny modes that this open 6800 * is adding to the file for this open owner; 6801 */ 6802 share_a = (access & ~sp->rs_share_access); 6803 share_d = (deny & ~sp->rs_share_deny); 6804 6805 first_open = (sp->rs_open_access & OPEN4_SHARE_ACCESS_BOTH) == 0; 6806 6807 /* 6808 * Check to see the client has already sent an open for this 6809 * open owner on this file with the same share/deny modes. 6810 * If so, we don't need to check for a conflict and we don't 6811 * need to add another shrlock. If not, then we need to 6812 * check for conflicts in deny and access before checking for 6813 * conflicts in delegation. We don't want to recall a 6814 * delegation based on an open that will eventually fail based 6815 * on shares modes. 6816 */ 6817 6818 if (share_a || share_d) { 6819 if ((err = rfs4_share(sp, access, deny)) != 0) { 6820 rfs4_dbe_unlock(sp->rs_dbe); 6821 resp->status = err; 6822 6823 rfs4_file_rele(fp); 6824 /* Not a fully formed open; "close" it */ 6825 if (screate == TRUE) 6826 rfs4_state_close(sp, FALSE, FALSE, cs->cr); 6827 rfs4_state_rele(sp); 6828 return; 6829 } 6830 } 6831 6832 rfs4_dbe_lock(fp->rf_dbe); 6833 6834 /* 6835 * Check to see if this file is delegated and if so, if a 6836 * recall needs to be done. 6837 */ 6838 if (rfs4_check_recall(sp, access)) { 6839 rfs4_dbe_unlock(fp->rf_dbe); 6840 rfs4_dbe_unlock(sp->rs_dbe); 6841 rfs4_recall_deleg(fp, FALSE, sp->rs_owner->ro_client); 6842 delay(NFS4_DELEGATION_CONFLICT_DELAY); 6843 rfs4_dbe_lock(sp->rs_dbe); 6844 6845 /* if state closed while lock was dropped */ 6846 if (sp->rs_closed) { 6847 if (share_a || share_d) 6848 (void) rfs4_unshare(sp); 6849 rfs4_dbe_unlock(sp->rs_dbe); 6850 rfs4_file_rele(fp); 6851 /* Not a fully formed open; "close" it */ 6852 if (screate == TRUE) 6853 rfs4_state_close(sp, FALSE, FALSE, cs->cr); 6854 rfs4_state_rele(sp); 6855 resp->status = NFS4ERR_OLD_STATEID; 6856 return; 6857 } 6858 6859 rfs4_dbe_lock(fp->rf_dbe); 6860 /* Let's see if the delegation was returned */ 6861 if (rfs4_check_recall(sp, access)) { 6862 rfs4_dbe_unlock(fp->rf_dbe); 6863 if (share_a || share_d) 6864 (void) rfs4_unshare(sp); 6865 rfs4_dbe_unlock(sp->rs_dbe); 6866 rfs4_file_rele(fp); 6867 rfs4_update_lease(sp->rs_owner->ro_client); 6868 6869 /* Not a fully formed open; "close" it */ 6870 if (screate == TRUE) 6871 rfs4_state_close(sp, FALSE, FALSE, cs->cr); 6872 rfs4_state_rele(sp); 6873 resp->status = NFS4ERR_DELAY; 6874 return; 6875 } 6876 } 6877 /* 6878 * the share check passed and any delegation conflict has been 6879 * taken care of, now call vop_open. 6880 * if this is the first open then call vop_open with fflags. 6881 * if not, call vn_open_upgrade with just the upgrade flags. 6882 * 6883 * if the file has been opened already, it will have the current 6884 * access mode in the state struct. if it has no share access, then 6885 * this is a new open. 6886 * 6887 * However, if this is open with CLAIM_DLEGATE_CUR, then don't 6888 * call VOP_OPEN(), just do the open upgrade. 6889 */ 6890 if (first_open && !deleg_cur) { 6891 ct.cc_sysid = sysid; 6892 ct.cc_pid = rfs4_dbe_getid(sp->rs_owner->ro_dbe); 6893 ct.cc_caller_id = nfs4_srv_caller_id; 6894 ct.cc_flags = CC_DONTBLOCK; 6895 err = VOP_OPEN(&cs->vp, fflags, cs->cr, &ct); 6896 if (err) { 6897 rfs4_dbe_unlock(fp->rf_dbe); 6898 if (share_a || share_d) 6899 (void) rfs4_unshare(sp); 6900 rfs4_dbe_unlock(sp->rs_dbe); 6901 rfs4_file_rele(fp); 6902 6903 /* Not a fully formed open; "close" it */ 6904 if (screate == TRUE) 6905 rfs4_state_close(sp, FALSE, FALSE, cs->cr); 6906 rfs4_state_rele(sp); 6907 /* check if a monitor detected a delegation conflict */ 6908 if (err == EAGAIN && (ct.cc_flags & CC_WOULDBLOCK)) 6909 resp->status = NFS4ERR_DELAY; 6910 else 6911 resp->status = NFS4ERR_SERVERFAULT; 6912 return; 6913 } 6914 } else { /* open upgrade */ 6915 /* 6916 * calculate the fflags for the new mode that is being added 6917 * by this upgrade. 6918 */ 6919 fflags = 0; 6920 if (open_a & OPEN4_SHARE_ACCESS_READ) 6921 fflags |= FREAD; 6922 if (open_a & OPEN4_SHARE_ACCESS_WRITE) 6923 fflags |= FWRITE; 6924 vn_open_upgrade(cs->vp, fflags); 6925 } 6926 sp->rs_open_access |= access; 6927 sp->rs_open_deny |= deny; 6928 6929 if (open_d & OPEN4_SHARE_DENY_READ) 6930 fp->rf_deny_read++; 6931 if (open_d & OPEN4_SHARE_DENY_WRITE) 6932 fp->rf_deny_write++; 6933 fp->rf_share_deny |= deny; 6934 6935 if (open_a & OPEN4_SHARE_ACCESS_READ) 6936 fp->rf_access_read++; 6937 if (open_a & OPEN4_SHARE_ACCESS_WRITE) 6938 fp->rf_access_write++; 6939 fp->rf_share_access |= access; 6940 6941 /* 6942 * Check for delegation here. if the deleg argument is not 6943 * DELEG_ANY, then this is a reclaim from a client and 6944 * we must honor the delegation requested. If necessary we can 6945 * set the recall flag. 6946 */ 6947 6948 dsp = rfs4_grant_delegation(deleg, sp, &recall); 6949 6950 cs->deleg = (fp->rf_dinfo.rd_dtype == OPEN_DELEGATE_WRITE); 6951 6952 next_stateid(&sp->rs_stateid); 6953 6954 resp->stateid = sp->rs_stateid.stateid; 6955 6956 rfs4_dbe_unlock(fp->rf_dbe); 6957 rfs4_dbe_unlock(sp->rs_dbe); 6958 6959 if (dsp) { 6960 rfs4_set_deleg_response(dsp, &resp->delegation, NULL, recall); 6961 rfs4_deleg_state_rele(dsp); 6962 } 6963 6964 rfs4_file_rele(fp); 6965 rfs4_state_rele(sp); 6966 6967 resp->status = NFS4_OK; 6968 } 6969 6970 /*ARGSUSED*/ 6971 static void 6972 rfs4_do_opennull(struct compound_state *cs, struct svc_req *req, 6973 OPEN4args *args, rfs4_openowner_t *oo, OPEN4res *resp) 6974 { 6975 change_info4 *cinfo = &resp->cinfo; 6976 bitmap4 *attrset = &resp->attrset; 6977 6978 if (args->opentype == OPEN4_NOCREATE) 6979 resp->status = rfs4_lookupfile(&args->open_claim4_u.file, 6980 req, cs, args->share_access, cinfo); 6981 else { 6982 /* inhibit delegation grants during exclusive create */ 6983 6984 if (args->mode == EXCLUSIVE4) 6985 rfs4_disable_delegation(); 6986 6987 resp->status = rfs4_createfile(args, req, cs, cinfo, attrset, 6988 oo->ro_client->rc_clientid); 6989 } 6990 6991 if (resp->status == NFS4_OK) { 6992 6993 /* cs->vp cs->fh now reference the desired file */ 6994 6995 rfs4_do_open(cs, req, oo, 6996 oo->ro_need_confirm ? DELEG_NONE : DELEG_ANY, 6997 args->share_access, args->share_deny, resp, 0); 6998 6999 /* 7000 * If rfs4_createfile set attrset, we must 7001 * clear this attrset before the response is copied. 7002 */ 7003 if (resp->status != NFS4_OK && resp->attrset) { 7004 resp->attrset = 0; 7005 } 7006 } 7007 else 7008 *cs->statusp = resp->status; 7009 7010 if (args->mode == EXCLUSIVE4) 7011 rfs4_enable_delegation(); 7012 } 7013 7014 /*ARGSUSED*/ 7015 static void 7016 rfs4_do_openprev(struct compound_state *cs, struct svc_req *req, 7017 OPEN4args *args, rfs4_openowner_t *oo, OPEN4res *resp) 7018 { 7019 change_info4 *cinfo = &resp->cinfo; 7020 vattr_t va; 7021 vtype_t v_type = cs->vp->v_type; 7022 int error = 0; 7023 7024 /* Verify that we have a regular file */ 7025 if (v_type != VREG) { 7026 if (v_type == VDIR) 7027 resp->status = NFS4ERR_ISDIR; 7028 else if (v_type == VLNK) 7029 resp->status = NFS4ERR_SYMLINK; 7030 else 7031 resp->status = NFS4ERR_INVAL; 7032 return; 7033 } 7034 7035 va.va_mask = AT_MODE|AT_UID; 7036 error = VOP_GETATTR(cs->vp, &va, 0, cs->cr, NULL); 7037 if (error) { 7038 resp->status = puterrno4(error); 7039 return; 7040 } 7041 7042 cs->mandlock = MANDLOCK(cs->vp, va.va_mode); 7043 7044 /* 7045 * Check if we have access to the file, Note the the file 7046 * could have originally been open UNCHECKED or GUARDED 7047 * with mode bits that will now fail, but there is nothing 7048 * we can really do about that except in the case that the 7049 * owner of the file is the one requesting the open. 7050 */ 7051 if (crgetuid(cs->cr) != va.va_uid) { 7052 resp->status = check_open_access(args->share_access, cs, req); 7053 if (resp->status != NFS4_OK) { 7054 return; 7055 } 7056 } 7057 7058 /* 7059 * cinfo on a CLAIM_PREVIOUS is undefined, initialize to zero 7060 */ 7061 cinfo->before = 0; 7062 cinfo->after = 0; 7063 cinfo->atomic = FALSE; 7064 7065 rfs4_do_open(cs, req, oo, 7066 NFS4_DELEG4TYPE2REQTYPE(args->open_claim4_u.delegate_type), 7067 args->share_access, args->share_deny, resp, 0); 7068 } 7069 7070 static void 7071 rfs4_do_opendelcur(struct compound_state *cs, struct svc_req *req, 7072 OPEN4args *args, rfs4_openowner_t *oo, OPEN4res *resp) 7073 { 7074 int error; 7075 nfsstat4 status; 7076 stateid4 stateid = 7077 args->open_claim4_u.delegate_cur_info.delegate_stateid; 7078 rfs4_deleg_state_t *dsp; 7079 7080 /* 7081 * Find the state info from the stateid and confirm that the 7082 * file is delegated. If the state openowner is the same as 7083 * the supplied openowner we're done. If not, get the file 7084 * info from the found state info. Use that file info to 7085 * create the state for this lock owner. Note solaris doen't 7086 * really need the pathname to find the file. We may want to 7087 * lookup the pathname and make sure that the vp exist and 7088 * matches the vp in the file structure. However it is 7089 * possible that the pathname nolonger exists (local process 7090 * unlinks the file), so this may not be that useful. 7091 */ 7092 7093 status = rfs4_get_deleg_state(&stateid, &dsp); 7094 if (status != NFS4_OK) { 7095 resp->status = status; 7096 return; 7097 } 7098 7099 ASSERT(dsp->rds_finfo->rf_dinfo.rd_dtype != OPEN_DELEGATE_NONE); 7100 7101 /* 7102 * New lock owner, create state. Since this was probably called 7103 * in response to a CB_RECALL we set deleg to DELEG_NONE 7104 */ 7105 7106 ASSERT(cs->vp != NULL); 7107 VN_RELE(cs->vp); 7108 VN_HOLD(dsp->rds_finfo->rf_vp); 7109 cs->vp = dsp->rds_finfo->rf_vp; 7110 7111 if (error = makefh4(&cs->fh, cs->vp, cs->exi)) { 7112 rfs4_deleg_state_rele(dsp); 7113 *cs->statusp = resp->status = puterrno4(error); 7114 return; 7115 } 7116 7117 /* Mark progress for delegation returns */ 7118 dsp->rds_finfo->rf_dinfo.rd_time_lastwrite = gethrestime_sec(); 7119 rfs4_deleg_state_rele(dsp); 7120 rfs4_do_open(cs, req, oo, DELEG_NONE, 7121 args->share_access, args->share_deny, resp, 1); 7122 } 7123 7124 /*ARGSUSED*/ 7125 static void 7126 rfs4_do_opendelprev(struct compound_state *cs, struct svc_req *req, 7127 OPEN4args *args, rfs4_openowner_t *oo, OPEN4res *resp) 7128 { 7129 /* 7130 * Lookup the pathname, it must already exist since this file 7131 * was delegated. 7132 * 7133 * Find the file and state info for this vp and open owner pair. 7134 * check that they are in fact delegated. 7135 * check that the state access and deny modes are the same. 7136 * 7137 * Return the delgation possibly seting the recall flag. 7138 */ 7139 rfs4_file_t *fp; 7140 rfs4_state_t *sp; 7141 bool_t create = FALSE; 7142 bool_t dcreate = FALSE; 7143 rfs4_deleg_state_t *dsp; 7144 nfsace4 *ace; 7145 7146 /* Note we ignore oflags */ 7147 resp->status = rfs4_lookupfile(&args->open_claim4_u.file_delegate_prev, 7148 req, cs, args->share_access, &resp->cinfo); 7149 7150 if (resp->status != NFS4_OK) { 7151 return; 7152 } 7153 7154 /* get the file struct and hold a lock on it during initial open */ 7155 fp = rfs4_findfile_withlock(cs->vp, NULL, &create); 7156 if (fp == NULL) { 7157 resp->status = NFS4ERR_RESOURCE; 7158 DTRACE_PROBE1(nfss__e__do_opendelprev1, nfsstat4, resp->status); 7159 return; 7160 } 7161 7162 sp = rfs4_findstate_by_owner_file(oo, fp, &create); 7163 if (sp == NULL) { 7164 resp->status = NFS4ERR_SERVERFAULT; 7165 DTRACE_PROBE1(nfss__e__do_opendelprev2, nfsstat4, resp->status); 7166 rw_exit(&fp->rf_file_rwlock); 7167 rfs4_file_rele(fp); 7168 return; 7169 } 7170 7171 rfs4_dbe_lock(sp->rs_dbe); 7172 rfs4_dbe_lock(fp->rf_dbe); 7173 if (args->share_access != sp->rs_share_access || 7174 args->share_deny != sp->rs_share_deny || 7175 sp->rs_finfo->rf_dinfo.rd_dtype == OPEN_DELEGATE_NONE) { 7176 NFS4_DEBUG(rfs4_debug, 7177 (CE_NOTE, "rfs4_do_opendelprev: state mixup")); 7178 rfs4_dbe_unlock(fp->rf_dbe); 7179 rfs4_dbe_unlock(sp->rs_dbe); 7180 rfs4_file_rele(fp); 7181 rfs4_state_rele(sp); 7182 resp->status = NFS4ERR_SERVERFAULT; 7183 return; 7184 } 7185 rfs4_dbe_unlock(fp->rf_dbe); 7186 rfs4_dbe_unlock(sp->rs_dbe); 7187 7188 dsp = rfs4_finddeleg(sp, &dcreate); 7189 if (dsp == NULL) { 7190 rfs4_state_rele(sp); 7191 rfs4_file_rele(fp); 7192 resp->status = NFS4ERR_SERVERFAULT; 7193 return; 7194 } 7195 7196 next_stateid(&sp->rs_stateid); 7197 7198 resp->stateid = sp->rs_stateid.stateid; 7199 7200 resp->delegation.delegation_type = dsp->rds_dtype; 7201 7202 if (dsp->rds_dtype == OPEN_DELEGATE_READ) { 7203 open_read_delegation4 *rv = 7204 &resp->delegation.open_delegation4_u.read; 7205 7206 rv->stateid = dsp->rds_delegid.stateid; 7207 rv->recall = FALSE; /* no policy in place to set to TRUE */ 7208 ace = &rv->permissions; 7209 } else { 7210 open_write_delegation4 *rv = 7211 &resp->delegation.open_delegation4_u.write; 7212 7213 rv->stateid = dsp->rds_delegid.stateid; 7214 rv->recall = FALSE; /* no policy in place to set to TRUE */ 7215 ace = &rv->permissions; 7216 rv->space_limit.limitby = NFS_LIMIT_SIZE; 7217 rv->space_limit.nfs_space_limit4_u.filesize = UINT64_MAX; 7218 } 7219 7220 /* XXX For now */ 7221 ace->type = ACE4_ACCESS_ALLOWED_ACE_TYPE; 7222 ace->flag = 0; 7223 ace->access_mask = 0; 7224 ace->who.utf8string_len = 0; 7225 ace->who.utf8string_val = 0; 7226 7227 rfs4_deleg_state_rele(dsp); 7228 rfs4_state_rele(sp); 7229 rfs4_file_rele(fp); 7230 } 7231 7232 typedef enum { 7233 NFS4_CHKSEQ_OKAY = 0, 7234 NFS4_CHKSEQ_REPLAY = 1, 7235 NFS4_CHKSEQ_BAD = 2 7236 } rfs4_chkseq_t; 7237 7238 /* 7239 * Generic function for sequence number checks. 7240 */ 7241 static rfs4_chkseq_t 7242 rfs4_check_seqid(seqid4 seqid, nfs_resop4 *lastop, 7243 seqid4 rqst_seq, nfs_resop4 *resop, bool_t copyres) 7244 { 7245 /* Same sequence ids and matching operations? */ 7246 if (seqid == rqst_seq && resop->resop == lastop->resop) { 7247 if (copyres == TRUE) { 7248 rfs4_free_reply(resop); 7249 rfs4_copy_reply(resop, lastop); 7250 } 7251 NFS4_DEBUG(rfs4_debug, (CE_NOTE, 7252 "Replayed SEQID %d\n", seqid)); 7253 return (NFS4_CHKSEQ_REPLAY); 7254 } 7255 7256 /* If the incoming sequence is not the next expected then it is bad */ 7257 if (rqst_seq != seqid + 1) { 7258 if (rqst_seq == seqid) { 7259 NFS4_DEBUG(rfs4_debug, 7260 (CE_NOTE, "BAD SEQID: Replayed sequence id " 7261 "but last op was %d current op is %d\n", 7262 lastop->resop, resop->resop)); 7263 return (NFS4_CHKSEQ_BAD); 7264 } 7265 NFS4_DEBUG(rfs4_debug, 7266 (CE_NOTE, "BAD SEQID: got %u expecting %u\n", 7267 rqst_seq, seqid)); 7268 return (NFS4_CHKSEQ_BAD); 7269 } 7270 7271 /* Everything okay -- next expected */ 7272 return (NFS4_CHKSEQ_OKAY); 7273 } 7274 7275 7276 static rfs4_chkseq_t 7277 rfs4_check_open_seqid(seqid4 seqid, rfs4_openowner_t *op, nfs_resop4 *resop) 7278 { 7279 rfs4_chkseq_t rc; 7280 7281 rfs4_dbe_lock(op->ro_dbe); 7282 rc = rfs4_check_seqid(op->ro_open_seqid, &op->ro_reply, seqid, resop, 7283 TRUE); 7284 rfs4_dbe_unlock(op->ro_dbe); 7285 7286 if (rc == NFS4_CHKSEQ_OKAY) 7287 rfs4_update_lease(op->ro_client); 7288 7289 return (rc); 7290 } 7291 7292 static rfs4_chkseq_t 7293 rfs4_check_olo_seqid(seqid4 olo_seqid, rfs4_openowner_t *op, nfs_resop4 *resop) 7294 { 7295 rfs4_chkseq_t rc; 7296 7297 rfs4_dbe_lock(op->ro_dbe); 7298 rc = rfs4_check_seqid(op->ro_open_seqid, &op->ro_reply, 7299 olo_seqid, resop, FALSE); 7300 rfs4_dbe_unlock(op->ro_dbe); 7301 7302 return (rc); 7303 } 7304 7305 static rfs4_chkseq_t 7306 rfs4_check_lock_seqid(seqid4 seqid, rfs4_lo_state_t *lsp, nfs_resop4 *resop) 7307 { 7308 rfs4_chkseq_t rc = NFS4_CHKSEQ_OKAY; 7309 7310 rfs4_dbe_lock(lsp->rls_dbe); 7311 if (!lsp->rls_skip_seqid_check) 7312 rc = rfs4_check_seqid(lsp->rls_seqid, &lsp->rls_reply, seqid, 7313 resop, TRUE); 7314 rfs4_dbe_unlock(lsp->rls_dbe); 7315 7316 return (rc); 7317 } 7318 7319 static void 7320 rfs4_op_open(nfs_argop4 *argop, nfs_resop4 *resop, 7321 struct svc_req *req, struct compound_state *cs) 7322 { 7323 OPEN4args *args = &argop->nfs_argop4_u.opopen; 7324 OPEN4res *resp = &resop->nfs_resop4_u.opopen; 7325 open_owner4 *owner = &args->owner; 7326 open_claim_type4 claim = args->claim; 7327 rfs4_client_t *cp; 7328 rfs4_openowner_t *oo; 7329 bool_t create; 7330 bool_t replay = FALSE; 7331 int can_reclaim; 7332 7333 DTRACE_NFSV4_2(op__open__start, struct compound_state *, cs, 7334 OPEN4args *, args); 7335 7336 if (cs->vp == NULL) { 7337 *cs->statusp = resp->status = NFS4ERR_NOFILEHANDLE; 7338 goto end; 7339 } 7340 7341 /* 7342 * Need to check clientid and lease expiration first based on 7343 * error ordering and incrementing sequence id. 7344 */ 7345 cp = rfs4_findclient_by_id(owner->clientid, FALSE); 7346 if (cp == NULL) { 7347 *cs->statusp = resp->status = 7348 rfs4_check_clientid(&owner->clientid, 0); 7349 goto end; 7350 } 7351 7352 if (rfs4_lease_expired(cp)) { 7353 rfs4_client_close(cp); 7354 *cs->statusp = resp->status = NFS4ERR_EXPIRED; 7355 goto end; 7356 } 7357 can_reclaim = cp->rc_can_reclaim; 7358 7359 /* 7360 * Find the open_owner for use from this point forward. Take 7361 * care in updating the sequence id based on the type of error 7362 * being returned. 7363 */ 7364 retry: 7365 create = TRUE; 7366 oo = rfs4_findopenowner(owner, &create, args->seqid); 7367 if (oo == NULL) { 7368 *cs->statusp = resp->status = NFS4ERR_STALE_CLIENTID; 7369 rfs4_client_rele(cp); 7370 goto end; 7371 } 7372 7373 /* Hold off access to the sequence space while the open is done */ 7374 rfs4_sw_enter(&oo->ro_sw); 7375 7376 /* 7377 * If the open_owner existed before at the server, then check 7378 * the sequence id. 7379 */ 7380 if (!create && !oo->ro_postpone_confirm) { 7381 switch (rfs4_check_open_seqid(args->seqid, oo, resop)) { 7382 case NFS4_CHKSEQ_BAD: 7383 if ((args->seqid > oo->ro_open_seqid) && 7384 oo->ro_need_confirm) { 7385 rfs4_free_opens(oo, TRUE, FALSE); 7386 rfs4_sw_exit(&oo->ro_sw); 7387 rfs4_openowner_rele(oo); 7388 goto retry; 7389 } 7390 resp->status = NFS4ERR_BAD_SEQID; 7391 goto out; 7392 case NFS4_CHKSEQ_REPLAY: /* replay of previous request */ 7393 replay = TRUE; 7394 goto out; 7395 default: 7396 break; 7397 } 7398 7399 /* 7400 * Sequence was ok and open owner exists 7401 * check to see if we have yet to see an 7402 * open_confirm. 7403 */ 7404 if (oo->ro_need_confirm) { 7405 rfs4_free_opens(oo, TRUE, FALSE); 7406 rfs4_sw_exit(&oo->ro_sw); 7407 rfs4_openowner_rele(oo); 7408 goto retry; 7409 } 7410 } 7411 /* Grace only applies to regular-type OPENs */ 7412 if (rfs4_clnt_in_grace(cp) && 7413 (claim == CLAIM_NULL || claim == CLAIM_DELEGATE_CUR)) { 7414 *cs->statusp = resp->status = NFS4ERR_GRACE; 7415 goto out; 7416 } 7417 7418 /* 7419 * If previous state at the server existed then can_reclaim 7420 * will be set. If not reply NFS4ERR_NO_GRACE to the 7421 * client. 7422 */ 7423 if (rfs4_clnt_in_grace(cp) && claim == CLAIM_PREVIOUS && !can_reclaim) { 7424 *cs->statusp = resp->status = NFS4ERR_NO_GRACE; 7425 goto out; 7426 } 7427 7428 7429 /* 7430 * Reject the open if the client has missed the grace period 7431 */ 7432 if (!rfs4_clnt_in_grace(cp) && claim == CLAIM_PREVIOUS) { 7433 *cs->statusp = resp->status = NFS4ERR_NO_GRACE; 7434 goto out; 7435 } 7436 7437 /* Couple of up-front bookkeeping items */ 7438 if (oo->ro_need_confirm) { 7439 /* 7440 * If this is a reclaim OPEN then we should not ask 7441 * for a confirmation of the open_owner per the 7442 * protocol specification. 7443 */ 7444 if (claim == CLAIM_PREVIOUS) 7445 oo->ro_need_confirm = FALSE; 7446 else 7447 resp->rflags |= OPEN4_RESULT_CONFIRM; 7448 } 7449 resp->rflags |= OPEN4_RESULT_LOCKTYPE_POSIX; 7450 7451 /* 7452 * If there is an unshared filesystem mounted on this vnode, 7453 * do not allow to open/create in this directory. 7454 */ 7455 if (vn_ismntpt(cs->vp)) { 7456 *cs->statusp = resp->status = NFS4ERR_ACCESS; 7457 goto out; 7458 } 7459 7460 /* 7461 * access must READ, WRITE, or BOTH. No access is invalid. 7462 * deny can be READ, WRITE, BOTH, or NONE. 7463 * bits not defined for access/deny are invalid. 7464 */ 7465 if (! (args->share_access & OPEN4_SHARE_ACCESS_BOTH) || 7466 (args->share_access & ~OPEN4_SHARE_ACCESS_BOTH) || 7467 (args->share_deny & ~OPEN4_SHARE_DENY_BOTH)) { 7468 *cs->statusp = resp->status = NFS4ERR_INVAL; 7469 goto out; 7470 } 7471 7472 7473 /* 7474 * make sure attrset is zero before response is built. 7475 */ 7476 resp->attrset = 0; 7477 7478 switch (claim) { 7479 case CLAIM_NULL: 7480 rfs4_do_opennull(cs, req, args, oo, resp); 7481 break; 7482 case CLAIM_PREVIOUS: 7483 rfs4_do_openprev(cs, req, args, oo, resp); 7484 break; 7485 case CLAIM_DELEGATE_CUR: 7486 rfs4_do_opendelcur(cs, req, args, oo, resp); 7487 break; 7488 case CLAIM_DELEGATE_PREV: 7489 rfs4_do_opendelprev(cs, req, args, oo, resp); 7490 break; 7491 default: 7492 resp->status = NFS4ERR_INVAL; 7493 break; 7494 } 7495 7496 out: 7497 rfs4_client_rele(cp); 7498 7499 /* Catch sequence id handling here to make it a little easier */ 7500 switch (resp->status) { 7501 case NFS4ERR_BADXDR: 7502 case NFS4ERR_BAD_SEQID: 7503 case NFS4ERR_BAD_STATEID: 7504 case NFS4ERR_NOFILEHANDLE: 7505 case NFS4ERR_RESOURCE: 7506 case NFS4ERR_STALE_CLIENTID: 7507 case NFS4ERR_STALE_STATEID: 7508 /* 7509 * The protocol states that if any of these errors are 7510 * being returned, the sequence id should not be 7511 * incremented. Any other return requires an 7512 * increment. 7513 */ 7514 break; 7515 default: 7516 /* Always update the lease in this case */ 7517 rfs4_update_lease(oo->ro_client); 7518 7519 /* Regular response - copy the result */ 7520 if (!replay) 7521 rfs4_update_open_resp(oo, resop, &cs->fh); 7522 7523 /* 7524 * REPLAY case: Only if the previous response was OK 7525 * do we copy the filehandle. If not OK, no 7526 * filehandle to copy. 7527 */ 7528 if (replay == TRUE && 7529 resp->status == NFS4_OK && 7530 oo->ro_reply_fh.nfs_fh4_val) { 7531 /* 7532 * If this is a replay, we must restore the 7533 * current filehandle/vp to that of what was 7534 * returned originally. Try our best to do 7535 * it. 7536 */ 7537 nfs_fh4_fmt_t *fh_fmtp = 7538 (nfs_fh4_fmt_t *)oo->ro_reply_fh.nfs_fh4_val; 7539 7540 cs->exi = checkexport4(&fh_fmtp->fh4_fsid, 7541 (fid_t *)&fh_fmtp->fh4_xlen, NULL); 7542 7543 if (cs->exi == NULL) { 7544 resp->status = NFS4ERR_STALE; 7545 goto finish; 7546 } 7547 7548 VN_RELE(cs->vp); 7549 7550 cs->vp = nfs4_fhtovp(&oo->ro_reply_fh, cs->exi, 7551 &resp->status); 7552 7553 if (cs->vp == NULL) 7554 goto finish; 7555 7556 nfs_fh4_copy(&oo->ro_reply_fh, &cs->fh); 7557 } 7558 7559 /* 7560 * If this was a replay, no need to update the 7561 * sequence id. If the open_owner was not created on 7562 * this pass, then update. The first use of an 7563 * open_owner will not bump the sequence id. 7564 */ 7565 if (replay == FALSE && !create) 7566 rfs4_update_open_sequence(oo); 7567 /* 7568 * If the client is receiving an error and the 7569 * open_owner needs to be confirmed, there is no way 7570 * to notify the client of this fact ignoring the fact 7571 * that the server has no method of returning a 7572 * stateid to confirm. Therefore, the server needs to 7573 * mark this open_owner in a way as to avoid the 7574 * sequence id checking the next time the client uses 7575 * this open_owner. 7576 */ 7577 if (resp->status != NFS4_OK && oo->ro_need_confirm) 7578 oo->ro_postpone_confirm = TRUE; 7579 /* 7580 * If OK response then clear the postpone flag and 7581 * reset the sequence id to keep in sync with the 7582 * client. 7583 */ 7584 if (resp->status == NFS4_OK && oo->ro_postpone_confirm) { 7585 oo->ro_postpone_confirm = FALSE; 7586 oo->ro_open_seqid = args->seqid; 7587 } 7588 break; 7589 } 7590 7591 finish: 7592 *cs->statusp = resp->status; 7593 7594 rfs4_sw_exit(&oo->ro_sw); 7595 rfs4_openowner_rele(oo); 7596 7597 end: 7598 DTRACE_NFSV4_2(op__open__done, struct compound_state *, cs, 7599 OPEN4res *, resp); 7600 } 7601 7602 /*ARGSUSED*/ 7603 void 7604 rfs4_op_open_confirm(nfs_argop4 *argop, nfs_resop4 *resop, 7605 struct svc_req *req, struct compound_state *cs) 7606 { 7607 OPEN_CONFIRM4args *args = &argop->nfs_argop4_u.opopen_confirm; 7608 OPEN_CONFIRM4res *resp = &resop->nfs_resop4_u.opopen_confirm; 7609 rfs4_state_t *sp; 7610 nfsstat4 status; 7611 7612 DTRACE_NFSV4_2(op__open__confirm__start, struct compound_state *, cs, 7613 OPEN_CONFIRM4args *, args); 7614 7615 if (cs->vp == NULL) { 7616 *cs->statusp = resp->status = NFS4ERR_NOFILEHANDLE; 7617 goto out; 7618 } 7619 7620 status = rfs4_get_state(&args->open_stateid, &sp, RFS4_DBS_VALID); 7621 if (status != NFS4_OK) { 7622 *cs->statusp = resp->status = status; 7623 goto out; 7624 } 7625 7626 /* Ensure specified filehandle matches */ 7627 if (cs->vp != sp->rs_finfo->rf_vp) { 7628 rfs4_state_rele(sp); 7629 *cs->statusp = resp->status = NFS4ERR_BAD_STATEID; 7630 goto out; 7631 } 7632 7633 /* hold off other access to open_owner while we tinker */ 7634 rfs4_sw_enter(&sp->rs_owner->ro_sw); 7635 7636 switch (rfs4_check_stateid_seqid(sp, &args->open_stateid)) { 7637 case NFS4_CHECK_STATEID_OKAY: 7638 if (rfs4_check_open_seqid(args->seqid, sp->rs_owner, 7639 resop) != 0) { 7640 *cs->statusp = resp->status = NFS4ERR_BAD_SEQID; 7641 break; 7642 } 7643 /* 7644 * If it is the appropriate stateid and determined to 7645 * be "OKAY" then this means that the stateid does not 7646 * need to be confirmed and the client is in error for 7647 * sending an OPEN_CONFIRM. 7648 */ 7649 *cs->statusp = resp->status = NFS4ERR_BAD_STATEID; 7650 break; 7651 case NFS4_CHECK_STATEID_OLD: 7652 *cs->statusp = resp->status = NFS4ERR_OLD_STATEID; 7653 break; 7654 case NFS4_CHECK_STATEID_BAD: 7655 *cs->statusp = resp->status = NFS4ERR_BAD_STATEID; 7656 break; 7657 case NFS4_CHECK_STATEID_EXPIRED: 7658 *cs->statusp = resp->status = NFS4ERR_EXPIRED; 7659 break; 7660 case NFS4_CHECK_STATEID_CLOSED: 7661 *cs->statusp = resp->status = NFS4ERR_OLD_STATEID; 7662 break; 7663 case NFS4_CHECK_STATEID_REPLAY: 7664 switch (rfs4_check_open_seqid(args->seqid, sp->rs_owner, 7665 resop)) { 7666 case NFS4_CHKSEQ_OKAY: 7667 /* 7668 * This is replayed stateid; if seqid matches 7669 * next expected, then client is using wrong seqid. 7670 */ 7671 /* fall through */ 7672 case NFS4_CHKSEQ_BAD: 7673 *cs->statusp = resp->status = NFS4ERR_BAD_SEQID; 7674 break; 7675 case NFS4_CHKSEQ_REPLAY: 7676 /* 7677 * Note this case is the duplicate case so 7678 * resp->status is already set. 7679 */ 7680 *cs->statusp = resp->status; 7681 rfs4_update_lease(sp->rs_owner->ro_client); 7682 break; 7683 } 7684 break; 7685 case NFS4_CHECK_STATEID_UNCONFIRMED: 7686 if (rfs4_check_open_seqid(args->seqid, sp->rs_owner, 7687 resop) != NFS4_CHKSEQ_OKAY) { 7688 *cs->statusp = resp->status = NFS4ERR_BAD_SEQID; 7689 break; 7690 } 7691 *cs->statusp = resp->status = NFS4_OK; 7692 7693 next_stateid(&sp->rs_stateid); 7694 resp->open_stateid = sp->rs_stateid.stateid; 7695 sp->rs_owner->ro_need_confirm = FALSE; 7696 rfs4_update_lease(sp->rs_owner->ro_client); 7697 rfs4_update_open_sequence(sp->rs_owner); 7698 rfs4_update_open_resp(sp->rs_owner, resop, NULL); 7699 break; 7700 default: 7701 ASSERT(FALSE); 7702 *cs->statusp = resp->status = NFS4ERR_SERVERFAULT; 7703 break; 7704 } 7705 rfs4_sw_exit(&sp->rs_owner->ro_sw); 7706 rfs4_state_rele(sp); 7707 7708 out: 7709 DTRACE_NFSV4_2(op__open__confirm__done, struct compound_state *, cs, 7710 OPEN_CONFIRM4res *, resp); 7711 } 7712 7713 /*ARGSUSED*/ 7714 void 7715 rfs4_op_open_downgrade(nfs_argop4 *argop, nfs_resop4 *resop, 7716 struct svc_req *req, struct compound_state *cs) 7717 { 7718 OPEN_DOWNGRADE4args *args = &argop->nfs_argop4_u.opopen_downgrade; 7719 OPEN_DOWNGRADE4res *resp = &resop->nfs_resop4_u.opopen_downgrade; 7720 uint32_t access = args->share_access; 7721 uint32_t deny = args->share_deny; 7722 nfsstat4 status; 7723 rfs4_state_t *sp; 7724 rfs4_file_t *fp; 7725 int fflags = 0; 7726 7727 DTRACE_NFSV4_2(op__open__downgrade__start, struct compound_state *, cs, 7728 OPEN_DOWNGRADE4args *, args); 7729 7730 if (cs->vp == NULL) { 7731 *cs->statusp = resp->status = NFS4ERR_NOFILEHANDLE; 7732 goto out; 7733 } 7734 7735 status = rfs4_get_state(&args->open_stateid, &sp, RFS4_DBS_VALID); 7736 if (status != NFS4_OK) { 7737 *cs->statusp = resp->status = status; 7738 goto out; 7739 } 7740 7741 /* Ensure specified filehandle matches */ 7742 if (cs->vp != sp->rs_finfo->rf_vp) { 7743 rfs4_state_rele(sp); 7744 *cs->statusp = resp->status = NFS4ERR_BAD_STATEID; 7745 goto out; 7746 } 7747 7748 /* hold off other access to open_owner while we tinker */ 7749 rfs4_sw_enter(&sp->rs_owner->ro_sw); 7750 7751 switch (rfs4_check_stateid_seqid(sp, &args->open_stateid)) { 7752 case NFS4_CHECK_STATEID_OKAY: 7753 if (rfs4_check_open_seqid(args->seqid, sp->rs_owner, 7754 resop) != NFS4_CHKSEQ_OKAY) { 7755 *cs->statusp = resp->status = NFS4ERR_BAD_SEQID; 7756 goto end; 7757 } 7758 break; 7759 case NFS4_CHECK_STATEID_OLD: 7760 *cs->statusp = resp->status = NFS4ERR_OLD_STATEID; 7761 goto end; 7762 case NFS4_CHECK_STATEID_BAD: 7763 *cs->statusp = resp->status = NFS4ERR_BAD_STATEID; 7764 goto end; 7765 case NFS4_CHECK_STATEID_EXPIRED: 7766 *cs->statusp = resp->status = NFS4ERR_EXPIRED; 7767 goto end; 7768 case NFS4_CHECK_STATEID_CLOSED: 7769 *cs->statusp = resp->status = NFS4ERR_OLD_STATEID; 7770 goto end; 7771 case NFS4_CHECK_STATEID_UNCONFIRMED: 7772 *cs->statusp = resp->status = NFS4ERR_BAD_STATEID; 7773 goto end; 7774 case NFS4_CHECK_STATEID_REPLAY: 7775 /* Check the sequence id for the open owner */ 7776 switch (rfs4_check_open_seqid(args->seqid, sp->rs_owner, 7777 resop)) { 7778 case NFS4_CHKSEQ_OKAY: 7779 /* 7780 * This is replayed stateid; if seqid matches 7781 * next expected, then client is using wrong seqid. 7782 */ 7783 /* fall through */ 7784 case NFS4_CHKSEQ_BAD: 7785 *cs->statusp = resp->status = NFS4ERR_BAD_SEQID; 7786 goto end; 7787 case NFS4_CHKSEQ_REPLAY: 7788 /* 7789 * Note this case is the duplicate case so 7790 * resp->status is already set. 7791 */ 7792 *cs->statusp = resp->status; 7793 rfs4_update_lease(sp->rs_owner->ro_client); 7794 goto end; 7795 } 7796 break; 7797 default: 7798 ASSERT(FALSE); 7799 break; 7800 } 7801 7802 rfs4_dbe_lock(sp->rs_dbe); 7803 /* 7804 * Check that the new access modes and deny modes are valid. 7805 * Check that no invalid bits are set. 7806 */ 7807 if ((access & ~(OPEN4_SHARE_ACCESS_READ | OPEN4_SHARE_ACCESS_WRITE)) || 7808 (deny & ~(OPEN4_SHARE_DENY_READ | OPEN4_SHARE_DENY_WRITE))) { 7809 *cs->statusp = resp->status = NFS4ERR_INVAL; 7810 rfs4_update_open_sequence(sp->rs_owner); 7811 rfs4_dbe_unlock(sp->rs_dbe); 7812 goto end; 7813 } 7814 7815 /* 7816 * The new modes must be a subset of the current modes and 7817 * the access must specify at least one mode. To test that 7818 * the new mode is a subset of the current modes we bitwise 7819 * AND them together and check that the result equals the new 7820 * mode. For example: 7821 * New mode, access == R and current mode, sp->rs_open_access == RW 7822 * access & sp->rs_open_access == R == access, so the new access mode 7823 * is valid. Consider access == RW, sp->rs_open_access = R 7824 * access & sp->rs_open_access == R != access, so the new access mode 7825 * is invalid. 7826 */ 7827 if ((access & sp->rs_open_access) != access || 7828 (deny & sp->rs_open_deny) != deny || 7829 (access & 7830 (OPEN4_SHARE_ACCESS_READ | OPEN4_SHARE_ACCESS_WRITE)) == 0) { 7831 *cs->statusp = resp->status = NFS4ERR_INVAL; 7832 rfs4_update_open_sequence(sp->rs_owner); 7833 rfs4_dbe_unlock(sp->rs_dbe); 7834 goto end; 7835 } 7836 7837 /* 7838 * Release any share locks associated with this stateID. 7839 * Strictly speaking, this violates the spec because the 7840 * spec effectively requires that open downgrade be atomic. 7841 * At present, fs_shrlock does not have this capability. 7842 */ 7843 (void) rfs4_unshare(sp); 7844 7845 status = rfs4_share(sp, access, deny); 7846 if (status != NFS4_OK) { 7847 *cs->statusp = resp->status = NFS4ERR_SERVERFAULT; 7848 rfs4_update_open_sequence(sp->rs_owner); 7849 rfs4_dbe_unlock(sp->rs_dbe); 7850 goto end; 7851 } 7852 7853 fp = sp->rs_finfo; 7854 rfs4_dbe_lock(fp->rf_dbe); 7855 7856 /* 7857 * If the current mode has deny read and the new mode 7858 * does not, decrement the number of deny read mode bits 7859 * and if it goes to zero turn off the deny read bit 7860 * on the file. 7861 */ 7862 if ((sp->rs_open_deny & OPEN4_SHARE_DENY_READ) && 7863 (deny & OPEN4_SHARE_DENY_READ) == 0) { 7864 fp->rf_deny_read--; 7865 if (fp->rf_deny_read == 0) 7866 fp->rf_share_deny &= ~OPEN4_SHARE_DENY_READ; 7867 } 7868 7869 /* 7870 * If the current mode has deny write and the new mode 7871 * does not, decrement the number of deny write mode bits 7872 * and if it goes to zero turn off the deny write bit 7873 * on the file. 7874 */ 7875 if ((sp->rs_open_deny & OPEN4_SHARE_DENY_WRITE) && 7876 (deny & OPEN4_SHARE_DENY_WRITE) == 0) { 7877 fp->rf_deny_write--; 7878 if (fp->rf_deny_write == 0) 7879 fp->rf_share_deny &= ~OPEN4_SHARE_DENY_WRITE; 7880 } 7881 7882 /* 7883 * If the current mode has access read and the new mode 7884 * does not, decrement the number of access read mode bits 7885 * and if it goes to zero turn off the access read bit 7886 * on the file. set fflags to FREAD for the call to 7887 * vn_open_downgrade(). 7888 */ 7889 if ((sp->rs_open_access & OPEN4_SHARE_ACCESS_READ) && 7890 (access & OPEN4_SHARE_ACCESS_READ) == 0) { 7891 fp->rf_access_read--; 7892 if (fp->rf_access_read == 0) 7893 fp->rf_share_access &= ~OPEN4_SHARE_ACCESS_READ; 7894 fflags |= FREAD; 7895 } 7896 7897 /* 7898 * If the current mode has access write and the new mode 7899 * does not, decrement the number of access write mode bits 7900 * and if it goes to zero turn off the access write bit 7901 * on the file. set fflags to FWRITE for the call to 7902 * vn_open_downgrade(). 7903 */ 7904 if ((sp->rs_open_access & OPEN4_SHARE_ACCESS_WRITE) && 7905 (access & OPEN4_SHARE_ACCESS_WRITE) == 0) { 7906 fp->rf_access_write--; 7907 if (fp->rf_access_write == 0) 7908 fp->rf_share_deny &= ~OPEN4_SHARE_ACCESS_WRITE; 7909 fflags |= FWRITE; 7910 } 7911 7912 /* Check that the file is still accessible */ 7913 ASSERT(fp->rf_share_access); 7914 7915 rfs4_dbe_unlock(fp->rf_dbe); 7916 7917 /* now set the new open access and deny modes */ 7918 sp->rs_open_access = access; 7919 sp->rs_open_deny = deny; 7920 7921 /* 7922 * we successfully downgraded the share lock, now we need to downgrade 7923 * the open. it is possible that the downgrade was only for a deny 7924 * mode and we have nothing else to do. 7925 */ 7926 if ((fflags & (FREAD|FWRITE)) != 0) 7927 vn_open_downgrade(cs->vp, fflags); 7928 7929 /* Update the stateid */ 7930 next_stateid(&sp->rs_stateid); 7931 resp->open_stateid = sp->rs_stateid.stateid; 7932 7933 rfs4_dbe_unlock(sp->rs_dbe); 7934 7935 *cs->statusp = resp->status = NFS4_OK; 7936 /* Update the lease */ 7937 rfs4_update_lease(sp->rs_owner->ro_client); 7938 /* And the sequence */ 7939 rfs4_update_open_sequence(sp->rs_owner); 7940 rfs4_update_open_resp(sp->rs_owner, resop, NULL); 7941 7942 end: 7943 rfs4_sw_exit(&sp->rs_owner->ro_sw); 7944 rfs4_state_rele(sp); 7945 out: 7946 DTRACE_NFSV4_2(op__open__downgrade__done, struct compound_state *, cs, 7947 OPEN_DOWNGRADE4res *, resp); 7948 } 7949 7950 /* 7951 * The logic behind this function is detailed in the NFSv4 RFC in the 7952 * SETCLIENTID operation description under IMPLEMENTATION. Refer to 7953 * that section for explicit guidance to server behavior for 7954 * SETCLIENTID. 7955 */ 7956 void 7957 rfs4_op_setclientid(nfs_argop4 *argop, nfs_resop4 *resop, 7958 struct svc_req *req, struct compound_state *cs) 7959 { 7960 SETCLIENTID4args *args = &argop->nfs_argop4_u.opsetclientid; 7961 SETCLIENTID4res *res = &resop->nfs_resop4_u.opsetclientid; 7962 rfs4_client_t *cp, *newcp, *cp_confirmed, *cp_unconfirmed; 7963 rfs4_clntip_t *ci; 7964 bool_t create; 7965 char *addr, *netid; 7966 int len; 7967 7968 DTRACE_NFSV4_2(op__setclientid__start, struct compound_state *, cs, 7969 SETCLIENTID4args *, args); 7970 retry: 7971 newcp = cp_confirmed = cp_unconfirmed = NULL; 7972 7973 /* 7974 * Save the caller's IP address 7975 */ 7976 args->client.cl_addr = 7977 (struct sockaddr *)svc_getrpccaller(req->rq_xprt)->buf; 7978 7979 /* 7980 * Record if it is a Solaris client that cannot handle referrals. 7981 */ 7982 if (strstr(args->client.id_val, "Solaris") && 7983 !strstr(args->client.id_val, "+referrals")) { 7984 /* Add a "yes, it's downrev" record */ 7985 create = TRUE; 7986 ci = rfs4_find_clntip(args->client.cl_addr, &create); 7987 ASSERT(ci != NULL); 7988 rfs4_dbe_rele(ci->ri_dbe); 7989 } else { 7990 /* Remove any previous record */ 7991 rfs4_invalidate_clntip(args->client.cl_addr); 7992 } 7993 7994 /* 7995 * In search of an EXISTING client matching the incoming 7996 * request to establish a new client identifier at the server 7997 */ 7998 create = TRUE; 7999 cp = rfs4_findclient(&args->client, &create, NULL); 8000 8001 /* Should never happen */ 8002 ASSERT(cp != NULL); 8003 8004 if (cp == NULL) { 8005 *cs->statusp = res->status = NFS4ERR_SERVERFAULT; 8006 goto out; 8007 } 8008 8009 /* 8010 * Easiest case. Client identifier is newly created and is 8011 * unconfirmed. Also note that for this case, no other 8012 * entries exist for the client identifier. Nothing else to 8013 * check. Just setup the response and respond. 8014 */ 8015 if (create) { 8016 *cs->statusp = res->status = NFS4_OK; 8017 res->SETCLIENTID4res_u.resok4.clientid = cp->rc_clientid; 8018 res->SETCLIENTID4res_u.resok4.setclientid_confirm = 8019 cp->rc_confirm_verf; 8020 /* Setup callback information; CB_NULL confirmation later */ 8021 rfs4_client_setcb(cp, &args->callback, args->callback_ident); 8022 8023 rfs4_client_rele(cp); 8024 goto out; 8025 } 8026 8027 /* 8028 * An existing, confirmed client may exist but it may not have 8029 * been active for at least one lease period. If so, then 8030 * "close" the client and create a new client identifier 8031 */ 8032 if (rfs4_lease_expired(cp)) { 8033 rfs4_client_close(cp); 8034 goto retry; 8035 } 8036 8037 if (cp->rc_need_confirm == TRUE) 8038 cp_unconfirmed = cp; 8039 else 8040 cp_confirmed = cp; 8041 8042 cp = NULL; 8043 8044 /* 8045 * We have a confirmed client, now check for an 8046 * unconfimred entry 8047 */ 8048 if (cp_confirmed) { 8049 /* If creds don't match then client identifier is inuse */ 8050 if (!creds_ok(cp_confirmed->rc_cr_set, req, cs)) { 8051 rfs4_cbinfo_t *cbp; 8052 /* 8053 * Some one else has established this client 8054 * id. Try and say * who they are. We will use 8055 * the call back address supplied by * the 8056 * first client. 8057 */ 8058 *cs->statusp = res->status = NFS4ERR_CLID_INUSE; 8059 8060 addr = netid = NULL; 8061 8062 cbp = &cp_confirmed->rc_cbinfo; 8063 if (cbp->cb_callback.cb_location.r_addr && 8064 cbp->cb_callback.cb_location.r_netid) { 8065 cb_client4 *cbcp = &cbp->cb_callback; 8066 8067 len = strlen(cbcp->cb_location.r_addr)+1; 8068 addr = kmem_alloc(len, KM_SLEEP); 8069 bcopy(cbcp->cb_location.r_addr, addr, len); 8070 len = strlen(cbcp->cb_location.r_netid)+1; 8071 netid = kmem_alloc(len, KM_SLEEP); 8072 bcopy(cbcp->cb_location.r_netid, netid, len); 8073 } 8074 8075 res->SETCLIENTID4res_u.client_using.r_addr = addr; 8076 res->SETCLIENTID4res_u.client_using.r_netid = netid; 8077 8078 rfs4_client_rele(cp_confirmed); 8079 } 8080 8081 /* 8082 * Confirmed, creds match, and verifier matches; must 8083 * be an update of the callback info 8084 */ 8085 if (cp_confirmed->rc_nfs_client.verifier == 8086 args->client.verifier) { 8087 /* Setup callback information */ 8088 rfs4_client_setcb(cp_confirmed, &args->callback, 8089 args->callback_ident); 8090 8091 /* everything okay -- move ahead */ 8092 *cs->statusp = res->status = NFS4_OK; 8093 res->SETCLIENTID4res_u.resok4.clientid = 8094 cp_confirmed->rc_clientid; 8095 8096 /* update the confirm_verifier and return it */ 8097 rfs4_client_scv_next(cp_confirmed); 8098 res->SETCLIENTID4res_u.resok4.setclientid_confirm = 8099 cp_confirmed->rc_confirm_verf; 8100 8101 rfs4_client_rele(cp_confirmed); 8102 goto out; 8103 } 8104 8105 /* 8106 * Creds match but the verifier doesn't. Must search 8107 * for an unconfirmed client that would be replaced by 8108 * this request. 8109 */ 8110 create = FALSE; 8111 cp_unconfirmed = rfs4_findclient(&args->client, &create, 8112 cp_confirmed); 8113 } 8114 8115 /* 8116 * At this point, we have taken care of the brand new client 8117 * struct, INUSE case, update of an existing, and confirmed 8118 * client struct. 8119 */ 8120 8121 /* 8122 * check to see if things have changed while we originally 8123 * picked up the client struct. If they have, then return and 8124 * retry the processing of this SETCLIENTID request. 8125 */ 8126 if (cp_unconfirmed) { 8127 rfs4_dbe_lock(cp_unconfirmed->rc_dbe); 8128 if (!cp_unconfirmed->rc_need_confirm) { 8129 rfs4_dbe_unlock(cp_unconfirmed->rc_dbe); 8130 rfs4_client_rele(cp_unconfirmed); 8131 if (cp_confirmed) 8132 rfs4_client_rele(cp_confirmed); 8133 goto retry; 8134 } 8135 /* do away with the old unconfirmed one */ 8136 rfs4_dbe_invalidate(cp_unconfirmed->rc_dbe); 8137 rfs4_dbe_unlock(cp_unconfirmed->rc_dbe); 8138 rfs4_client_rele(cp_unconfirmed); 8139 cp_unconfirmed = NULL; 8140 } 8141 8142 /* 8143 * This search will temporarily hide the confirmed client 8144 * struct while a new client struct is created as the 8145 * unconfirmed one. 8146 */ 8147 create = TRUE; 8148 newcp = rfs4_findclient(&args->client, &create, cp_confirmed); 8149 8150 ASSERT(newcp != NULL); 8151 8152 if (newcp == NULL) { 8153 *cs->statusp = res->status = NFS4ERR_SERVERFAULT; 8154 rfs4_client_rele(cp_confirmed); 8155 goto out; 8156 } 8157 8158 /* 8159 * If one was not created, then a similar request must be in 8160 * process so release and start over with this one 8161 */ 8162 if (create != TRUE) { 8163 rfs4_client_rele(newcp); 8164 if (cp_confirmed) 8165 rfs4_client_rele(cp_confirmed); 8166 goto retry; 8167 } 8168 8169 *cs->statusp = res->status = NFS4_OK; 8170 res->SETCLIENTID4res_u.resok4.clientid = newcp->rc_clientid; 8171 res->SETCLIENTID4res_u.resok4.setclientid_confirm = 8172 newcp->rc_confirm_verf; 8173 /* Setup callback information; CB_NULL confirmation later */ 8174 rfs4_client_setcb(newcp, &args->callback, args->callback_ident); 8175 8176 newcp->rc_cp_confirmed = cp_confirmed; 8177 8178 rfs4_client_rele(newcp); 8179 8180 out: 8181 DTRACE_NFSV4_2(op__setclientid__done, struct compound_state *, cs, 8182 SETCLIENTID4res *, res); 8183 } 8184 8185 /*ARGSUSED*/ 8186 void 8187 rfs4_op_setclientid_confirm(nfs_argop4 *argop, nfs_resop4 *resop, 8188 struct svc_req *req, struct compound_state *cs) 8189 { 8190 SETCLIENTID_CONFIRM4args *args = 8191 &argop->nfs_argop4_u.opsetclientid_confirm; 8192 SETCLIENTID_CONFIRM4res *res = 8193 &resop->nfs_resop4_u.opsetclientid_confirm; 8194 rfs4_client_t *cp, *cptoclose = NULL; 8195 8196 DTRACE_NFSV4_2(op__setclientid__confirm__start, 8197 struct compound_state *, cs, 8198 SETCLIENTID_CONFIRM4args *, args); 8199 8200 *cs->statusp = res->status = NFS4_OK; 8201 8202 cp = rfs4_findclient_by_id(args->clientid, TRUE); 8203 8204 if (cp == NULL) { 8205 *cs->statusp = res->status = 8206 rfs4_check_clientid(&args->clientid, 1); 8207 goto out; 8208 } 8209 8210 if (!creds_ok(cp, req, cs)) { 8211 *cs->statusp = res->status = NFS4ERR_CLID_INUSE; 8212 rfs4_client_rele(cp); 8213 goto out; 8214 } 8215 8216 /* If the verifier doesn't match, the record doesn't match */ 8217 if (cp->rc_confirm_verf != args->setclientid_confirm) { 8218 *cs->statusp = res->status = NFS4ERR_STALE_CLIENTID; 8219 rfs4_client_rele(cp); 8220 goto out; 8221 } 8222 8223 rfs4_dbe_lock(cp->rc_dbe); 8224 cp->rc_need_confirm = FALSE; 8225 if (cp->rc_cp_confirmed) { 8226 cptoclose = cp->rc_cp_confirmed; 8227 cptoclose->rc_ss_remove = 1; 8228 cp->rc_cp_confirmed = NULL; 8229 } 8230 8231 /* 8232 * Update the client's associated server instance, if it's changed 8233 * since the client was created. 8234 */ 8235 if (rfs4_servinst(cp) != rfs4_cur_servinst) 8236 rfs4_servinst_assign(cp, rfs4_cur_servinst); 8237 8238 /* 8239 * Record clientid in stable storage. 8240 * Must be done after server instance has been assigned. 8241 */ 8242 rfs4_ss_clid(cp); 8243 8244 rfs4_dbe_unlock(cp->rc_dbe); 8245 8246 if (cptoclose) 8247 /* don't need to rele, client_close does it */ 8248 rfs4_client_close(cptoclose); 8249 8250 /* If needed, initiate CB_NULL call for callback path */ 8251 rfs4_deleg_cb_check(cp); 8252 rfs4_update_lease(cp); 8253 8254 /* 8255 * Check to see if client can perform reclaims 8256 */ 8257 rfs4_ss_chkclid(cp); 8258 8259 rfs4_client_rele(cp); 8260 8261 out: 8262 DTRACE_NFSV4_2(op__setclientid__confirm__done, 8263 struct compound_state *, cs, 8264 SETCLIENTID_CONFIRM4 *, res); 8265 } 8266 8267 8268 /*ARGSUSED*/ 8269 void 8270 rfs4_op_close(nfs_argop4 *argop, nfs_resop4 *resop, 8271 struct svc_req *req, struct compound_state *cs) 8272 { 8273 CLOSE4args *args = &argop->nfs_argop4_u.opclose; 8274 CLOSE4res *resp = &resop->nfs_resop4_u.opclose; 8275 rfs4_state_t *sp; 8276 nfsstat4 status; 8277 8278 DTRACE_NFSV4_2(op__close__start, struct compound_state *, cs, 8279 CLOSE4args *, args); 8280 8281 if (cs->vp == NULL) { 8282 *cs->statusp = resp->status = NFS4ERR_NOFILEHANDLE; 8283 goto out; 8284 } 8285 8286 status = rfs4_get_state(&args->open_stateid, &sp, RFS4_DBS_INVALID); 8287 if (status != NFS4_OK) { 8288 *cs->statusp = resp->status = status; 8289 goto out; 8290 } 8291 8292 /* Ensure specified filehandle matches */ 8293 if (cs->vp != sp->rs_finfo->rf_vp) { 8294 rfs4_state_rele(sp); 8295 *cs->statusp = resp->status = NFS4ERR_BAD_STATEID; 8296 goto out; 8297 } 8298 8299 /* hold off other access to open_owner while we tinker */ 8300 rfs4_sw_enter(&sp->rs_owner->ro_sw); 8301 8302 switch (rfs4_check_stateid_seqid(sp, &args->open_stateid)) { 8303 case NFS4_CHECK_STATEID_OKAY: 8304 if (rfs4_check_open_seqid(args->seqid, sp->rs_owner, 8305 resop) != NFS4_CHKSEQ_OKAY) { 8306 *cs->statusp = resp->status = NFS4ERR_BAD_SEQID; 8307 goto end; 8308 } 8309 break; 8310 case NFS4_CHECK_STATEID_OLD: 8311 *cs->statusp = resp->status = NFS4ERR_OLD_STATEID; 8312 goto end; 8313 case NFS4_CHECK_STATEID_BAD: 8314 *cs->statusp = resp->status = NFS4ERR_BAD_STATEID; 8315 goto end; 8316 case NFS4_CHECK_STATEID_EXPIRED: 8317 *cs->statusp = resp->status = NFS4ERR_EXPIRED; 8318 goto end; 8319 case NFS4_CHECK_STATEID_CLOSED: 8320 *cs->statusp = resp->status = NFS4ERR_OLD_STATEID; 8321 goto end; 8322 case NFS4_CHECK_STATEID_UNCONFIRMED: 8323 *cs->statusp = resp->status = NFS4ERR_BAD_STATEID; 8324 goto end; 8325 case NFS4_CHECK_STATEID_REPLAY: 8326 /* Check the sequence id for the open owner */ 8327 switch (rfs4_check_open_seqid(args->seqid, sp->rs_owner, 8328 resop)) { 8329 case NFS4_CHKSEQ_OKAY: 8330 /* 8331 * This is replayed stateid; if seqid matches 8332 * next expected, then client is using wrong seqid. 8333 */ 8334 /* FALL THROUGH */ 8335 case NFS4_CHKSEQ_BAD: 8336 *cs->statusp = resp->status = NFS4ERR_BAD_SEQID; 8337 goto end; 8338 case NFS4_CHKSEQ_REPLAY: 8339 /* 8340 * Note this case is the duplicate case so 8341 * resp->status is already set. 8342 */ 8343 *cs->statusp = resp->status; 8344 rfs4_update_lease(sp->rs_owner->ro_client); 8345 goto end; 8346 } 8347 break; 8348 default: 8349 ASSERT(FALSE); 8350 break; 8351 } 8352 8353 rfs4_dbe_lock(sp->rs_dbe); 8354 8355 /* Update the stateid. */ 8356 next_stateid(&sp->rs_stateid); 8357 resp->open_stateid = sp->rs_stateid.stateid; 8358 8359 rfs4_dbe_unlock(sp->rs_dbe); 8360 8361 rfs4_update_lease(sp->rs_owner->ro_client); 8362 rfs4_update_open_sequence(sp->rs_owner); 8363 rfs4_update_open_resp(sp->rs_owner, resop, NULL); 8364 8365 rfs4_state_close(sp, FALSE, FALSE, cs->cr); 8366 8367 *cs->statusp = resp->status = status; 8368 8369 end: 8370 rfs4_sw_exit(&sp->rs_owner->ro_sw); 8371 rfs4_state_rele(sp); 8372 out: 8373 DTRACE_NFSV4_2(op__close__done, struct compound_state *, cs, 8374 CLOSE4res *, resp); 8375 } 8376 8377 /* 8378 * Manage the counts on the file struct and close all file locks 8379 */ 8380 /*ARGSUSED*/ 8381 void 8382 rfs4_release_share_lock_state(rfs4_state_t *sp, cred_t *cr, 8383 bool_t close_of_client) 8384 { 8385 rfs4_file_t *fp = sp->rs_finfo; 8386 rfs4_lo_state_t *lsp; 8387 int fflags = 0; 8388 8389 /* 8390 * If this call is part of the larger closing down of client 8391 * state then it is just easier to release all locks 8392 * associated with this client instead of going through each 8393 * individual file and cleaning locks there. 8394 */ 8395 if (close_of_client) { 8396 if (sp->rs_owner->ro_client->rc_unlksys_completed == FALSE && 8397 !list_is_empty(&sp->rs_lostatelist) && 8398 sp->rs_owner->ro_client->rc_sysidt != LM_NOSYSID) { 8399 /* Is the PxFS kernel module loaded? */ 8400 if (lm_remove_file_locks != NULL) { 8401 int new_sysid; 8402 8403 /* Encode the cluster nodeid in new sysid */ 8404 new_sysid = sp->rs_owner->ro_client->rc_sysidt; 8405 lm_set_nlmid_flk(&new_sysid); 8406 8407 /* 8408 * This PxFS routine removes file locks for a 8409 * client over all nodes of a cluster. 8410 */ 8411 NFS4_DEBUG(rfs4_debug, (CE_NOTE, 8412 "lm_remove_file_locks(sysid=0x%x)\n", 8413 new_sysid)); 8414 (*lm_remove_file_locks)(new_sysid); 8415 } else { 8416 struct flock64 flk; 8417 8418 /* Release all locks for this client */ 8419 flk.l_type = F_UNLKSYS; 8420 flk.l_whence = 0; 8421 flk.l_start = 0; 8422 flk.l_len = 0; 8423 flk.l_sysid = 8424 sp->rs_owner->ro_client->rc_sysidt; 8425 flk.l_pid = 0; 8426 (void) VOP_FRLOCK(sp->rs_finfo->rf_vp, F_SETLK, 8427 &flk, F_REMOTELOCK | FREAD | FWRITE, 8428 (u_offset_t)0, NULL, CRED(), NULL); 8429 } 8430 8431 sp->rs_owner->ro_client->rc_unlksys_completed = TRUE; 8432 } 8433 } 8434 8435 /* 8436 * Release all locks on this file by this lock owner or at 8437 * least mark the locks as having been released 8438 */ 8439 for (lsp = list_head(&sp->rs_lostatelist); lsp != NULL; 8440 lsp = list_next(&sp->rs_lostatelist, lsp)) { 8441 lsp->rls_locks_cleaned = TRUE; 8442 8443 /* Was this already taken care of above? */ 8444 if (!close_of_client && 8445 sp->rs_owner->ro_client->rc_sysidt != LM_NOSYSID) 8446 (void) cleanlocks(sp->rs_finfo->rf_vp, 8447 lsp->rls_locker->rl_pid, 8448 lsp->rls_locker->rl_client->rc_sysidt); 8449 } 8450 8451 /* 8452 * Release any shrlocks associated with this open state ID. 8453 * This must be done before the rfs4_state gets marked closed. 8454 */ 8455 if (sp->rs_owner->ro_client->rc_sysidt != LM_NOSYSID) 8456 (void) rfs4_unshare(sp); 8457 8458 if (sp->rs_open_access) { 8459 rfs4_dbe_lock(fp->rf_dbe); 8460 8461 /* 8462 * Decrement the count for each access and deny bit that this 8463 * state has contributed to the file. 8464 * If the file counts go to zero 8465 * clear the appropriate bit in the appropriate mask. 8466 */ 8467 if (sp->rs_open_access & OPEN4_SHARE_ACCESS_READ) { 8468 fp->rf_access_read--; 8469 fflags |= FREAD; 8470 if (fp->rf_access_read == 0) 8471 fp->rf_share_access &= ~OPEN4_SHARE_ACCESS_READ; 8472 } 8473 if (sp->rs_open_access & OPEN4_SHARE_ACCESS_WRITE) { 8474 fp->rf_access_write--; 8475 fflags |= FWRITE; 8476 if (fp->rf_access_write == 0) 8477 fp->rf_share_access &= 8478 ~OPEN4_SHARE_ACCESS_WRITE; 8479 } 8480 if (sp->rs_open_deny & OPEN4_SHARE_DENY_READ) { 8481 fp->rf_deny_read--; 8482 if (fp->rf_deny_read == 0) 8483 fp->rf_share_deny &= ~OPEN4_SHARE_DENY_READ; 8484 } 8485 if (sp->rs_open_deny & OPEN4_SHARE_DENY_WRITE) { 8486 fp->rf_deny_write--; 8487 if (fp->rf_deny_write == 0) 8488 fp->rf_share_deny &= ~OPEN4_SHARE_DENY_WRITE; 8489 } 8490 8491 (void) VOP_CLOSE(fp->rf_vp, fflags, 1, (offset_t)0, cr, NULL); 8492 8493 rfs4_dbe_unlock(fp->rf_dbe); 8494 8495 sp->rs_open_access = 0; 8496 sp->rs_open_deny = 0; 8497 } 8498 } 8499 8500 /* 8501 * lock_denied: Fill in a LOCK4deneid structure given an flock64 structure. 8502 */ 8503 static nfsstat4 8504 lock_denied(LOCK4denied *dp, struct flock64 *flk) 8505 { 8506 rfs4_lockowner_t *lo; 8507 rfs4_client_t *cp; 8508 uint32_t len; 8509 8510 lo = rfs4_findlockowner_by_pid(flk->l_pid); 8511 if (lo != NULL) { 8512 cp = lo->rl_client; 8513 if (rfs4_lease_expired(cp)) { 8514 rfs4_lockowner_rele(lo); 8515 rfs4_dbe_hold(cp->rc_dbe); 8516 rfs4_client_close(cp); 8517 return (NFS4ERR_EXPIRED); 8518 } 8519 dp->owner.clientid = lo->rl_owner.clientid; 8520 len = lo->rl_owner.owner_len; 8521 dp->owner.owner_val = kmem_alloc(len, KM_SLEEP); 8522 bcopy(lo->rl_owner.owner_val, dp->owner.owner_val, len); 8523 dp->owner.owner_len = len; 8524 rfs4_lockowner_rele(lo); 8525 goto finish; 8526 } 8527 8528 /* 8529 * Its not a NFS4 lock. We take advantage that the upper 32 bits 8530 * of the client id contain the boot time for a NFS4 lock. So we 8531 * fabricate and identity by setting clientid to the sysid, and 8532 * the lock owner to the pid. 8533 */ 8534 dp->owner.clientid = flk->l_sysid; 8535 len = sizeof (pid_t); 8536 dp->owner.owner_len = len; 8537 dp->owner.owner_val = kmem_alloc(len, KM_SLEEP); 8538 bcopy(&flk->l_pid, dp->owner.owner_val, len); 8539 finish: 8540 dp->offset = flk->l_start; 8541 dp->length = flk->l_len; 8542 8543 if (flk->l_type == F_RDLCK) 8544 dp->locktype = READ_LT; 8545 else if (flk->l_type == F_WRLCK) 8546 dp->locktype = WRITE_LT; 8547 else 8548 return (NFS4ERR_INVAL); /* no mapping from POSIX ltype to v4 */ 8549 8550 return (NFS4_OK); 8551 } 8552 8553 static int 8554 setlock(vnode_t *vp, struct flock64 *flock, int flag, cred_t *cred) 8555 { 8556 int error; 8557 struct flock64 flk; 8558 int i; 8559 clock_t delaytime; 8560 int cmd; 8561 8562 cmd = nbl_need_check(vp) ? F_SETLK_NBMAND : F_SETLK; 8563 retry: 8564 delaytime = MSEC_TO_TICK_ROUNDUP(rfs4_lock_delay); 8565 8566 for (i = 0; i < rfs4_maxlock_tries; i++) { 8567 LOCK_PRINT(rfs4_debug, "setlock", cmd, flock); 8568 error = VOP_FRLOCK(vp, cmd, 8569 flock, flag, (u_offset_t)0, NULL, cred, NULL); 8570 8571 if (error != EAGAIN && error != EACCES) 8572 break; 8573 8574 if (i < rfs4_maxlock_tries - 1) { 8575 delay(delaytime); 8576 delaytime *= 2; 8577 } 8578 } 8579 8580 if (error == EAGAIN || error == EACCES) { 8581 /* Get the owner of the lock */ 8582 flk = *flock; 8583 LOCK_PRINT(rfs4_debug, "setlock", F_GETLK, &flk); 8584 if (VOP_FRLOCK(vp, F_GETLK, &flk, flag, 8585 (u_offset_t)0, NULL, cred, NULL) == 0) { 8586 if (flk.l_type == F_UNLCK) { 8587 /* No longer locked, retry */ 8588 goto retry; 8589 } 8590 *flock = flk; 8591 LOCK_PRINT(rfs4_debug, "setlock(blocking lock)", 8592 F_GETLK, &flk); 8593 } 8594 } 8595 8596 return (error); 8597 } 8598 8599 /*ARGSUSED*/ 8600 static nfsstat4 8601 rfs4_do_lock(rfs4_lo_state_t *lsp, nfs_lock_type4 locktype, 8602 offset4 offset, length4 length, cred_t *cred, nfs_resop4 *resop) 8603 { 8604 nfsstat4 status; 8605 rfs4_lockowner_t *lo = lsp->rls_locker; 8606 rfs4_state_t *sp = lsp->rls_state; 8607 struct flock64 flock; 8608 int16_t ltype; 8609 int flag; 8610 int error; 8611 sysid_t sysid; 8612 LOCK4res *lres; 8613 8614 if (rfs4_lease_expired(lo->rl_client)) { 8615 return (NFS4ERR_EXPIRED); 8616 } 8617 8618 if ((status = rfs4_client_sysid(lo->rl_client, &sysid)) != NFS4_OK) 8619 return (status); 8620 8621 /* Check for zero length. To lock to end of file use all ones for V4 */ 8622 if (length == 0) 8623 return (NFS4ERR_INVAL); 8624 else if (length == (length4)(~0)) 8625 length = 0; /* Posix to end of file */ 8626 8627 retry: 8628 rfs4_dbe_lock(sp->rs_dbe); 8629 if (sp->rs_closed) { 8630 rfs4_dbe_unlock(sp->rs_dbe); 8631 return (NFS4ERR_OLD_STATEID); 8632 } 8633 8634 if (resop->resop != OP_LOCKU) { 8635 switch (locktype) { 8636 case READ_LT: 8637 case READW_LT: 8638 if ((sp->rs_share_access 8639 & OPEN4_SHARE_ACCESS_READ) == 0) { 8640 rfs4_dbe_unlock(sp->rs_dbe); 8641 8642 return (NFS4ERR_OPENMODE); 8643 } 8644 ltype = F_RDLCK; 8645 break; 8646 case WRITE_LT: 8647 case WRITEW_LT: 8648 if ((sp->rs_share_access 8649 & OPEN4_SHARE_ACCESS_WRITE) == 0) { 8650 rfs4_dbe_unlock(sp->rs_dbe); 8651 8652 return (NFS4ERR_OPENMODE); 8653 } 8654 ltype = F_WRLCK; 8655 break; 8656 } 8657 } else 8658 ltype = F_UNLCK; 8659 8660 flock.l_type = ltype; 8661 flock.l_whence = 0; /* SEEK_SET */ 8662 flock.l_start = offset; 8663 flock.l_len = length; 8664 flock.l_sysid = sysid; 8665 flock.l_pid = lsp->rls_locker->rl_pid; 8666 8667 /* Note that length4 is uint64_t but l_len and l_start are off64_t */ 8668 if (flock.l_len < 0 || flock.l_start < 0) { 8669 rfs4_dbe_unlock(sp->rs_dbe); 8670 return (NFS4ERR_INVAL); 8671 } 8672 8673 /* 8674 * N.B. FREAD has the same value as OPEN4_SHARE_ACCESS_READ and 8675 * FWRITE has the same value as OPEN4_SHARE_ACCESS_WRITE. 8676 */ 8677 flag = (int)sp->rs_share_access | F_REMOTELOCK; 8678 8679 error = setlock(sp->rs_finfo->rf_vp, &flock, flag, cred); 8680 if (error == 0) { 8681 rfs4_dbe_lock(lsp->rls_dbe); 8682 next_stateid(&lsp->rls_lockid); 8683 rfs4_dbe_unlock(lsp->rls_dbe); 8684 } 8685 8686 rfs4_dbe_unlock(sp->rs_dbe); 8687 8688 /* 8689 * N.B. We map error values to nfsv4 errors. This is differrent 8690 * than puterrno4 routine. 8691 */ 8692 switch (error) { 8693 case 0: 8694 status = NFS4_OK; 8695 break; 8696 case EAGAIN: 8697 case EACCES: /* Old value */ 8698 /* Can only get here if op is OP_LOCK */ 8699 ASSERT(resop->resop == OP_LOCK); 8700 lres = &resop->nfs_resop4_u.oplock; 8701 status = NFS4ERR_DENIED; 8702 if (lock_denied(&lres->LOCK4res_u.denied, &flock) 8703 == NFS4ERR_EXPIRED) 8704 goto retry; 8705 break; 8706 case ENOLCK: 8707 status = NFS4ERR_DELAY; 8708 break; 8709 case EOVERFLOW: 8710 status = NFS4ERR_INVAL; 8711 break; 8712 case EINVAL: 8713 status = NFS4ERR_NOTSUPP; 8714 break; 8715 default: 8716 status = NFS4ERR_SERVERFAULT; 8717 break; 8718 } 8719 8720 return (status); 8721 } 8722 8723 /*ARGSUSED*/ 8724 void 8725 rfs4_op_lock(nfs_argop4 *argop, nfs_resop4 *resop, 8726 struct svc_req *req, struct compound_state *cs) 8727 { 8728 LOCK4args *args = &argop->nfs_argop4_u.oplock; 8729 LOCK4res *resp = &resop->nfs_resop4_u.oplock; 8730 nfsstat4 status; 8731 stateid4 *stateid; 8732 rfs4_lockowner_t *lo; 8733 rfs4_client_t *cp; 8734 rfs4_state_t *sp = NULL; 8735 rfs4_lo_state_t *lsp = NULL; 8736 bool_t ls_sw_held = FALSE; 8737 bool_t create = TRUE; 8738 bool_t lcreate = TRUE; 8739 bool_t dup_lock = FALSE; 8740 int rc; 8741 8742 DTRACE_NFSV4_2(op__lock__start, struct compound_state *, cs, 8743 LOCK4args *, args); 8744 8745 if (cs->vp == NULL) { 8746 *cs->statusp = resp->status = NFS4ERR_NOFILEHANDLE; 8747 DTRACE_NFSV4_2(op__lock__done, struct compound_state *, 8748 cs, LOCK4res *, resp); 8749 return; 8750 } 8751 8752 if (args->locker.new_lock_owner) { 8753 /* Create a new lockowner for this instance */ 8754 open_to_lock_owner4 *olo = &args->locker.locker4_u.open_owner; 8755 8756 NFS4_DEBUG(rfs4_debug, (CE_NOTE, "Creating new lock owner")); 8757 8758 stateid = &olo->open_stateid; 8759 status = rfs4_get_state(stateid, &sp, RFS4_DBS_VALID); 8760 if (status != NFS4_OK) { 8761 NFS4_DEBUG(rfs4_debug, 8762 (CE_NOTE, "Get state failed in lock %d", status)); 8763 *cs->statusp = resp->status = status; 8764 DTRACE_NFSV4_2(op__lock__done, struct compound_state *, 8765 cs, LOCK4res *, resp); 8766 return; 8767 } 8768 8769 /* Ensure specified filehandle matches */ 8770 if (cs->vp != sp->rs_finfo->rf_vp) { 8771 rfs4_state_rele(sp); 8772 *cs->statusp = resp->status = NFS4ERR_BAD_STATEID; 8773 DTRACE_NFSV4_2(op__lock__done, struct compound_state *, 8774 cs, LOCK4res *, resp); 8775 return; 8776 } 8777 8778 /* hold off other access to open_owner while we tinker */ 8779 rfs4_sw_enter(&sp->rs_owner->ro_sw); 8780 8781 switch (rc = rfs4_check_stateid_seqid(sp, stateid)) { 8782 case NFS4_CHECK_STATEID_OLD: 8783 *cs->statusp = resp->status = NFS4ERR_OLD_STATEID; 8784 goto end; 8785 case NFS4_CHECK_STATEID_BAD: 8786 *cs->statusp = resp->status = NFS4ERR_BAD_STATEID; 8787 goto end; 8788 case NFS4_CHECK_STATEID_EXPIRED: 8789 *cs->statusp = resp->status = NFS4ERR_EXPIRED; 8790 goto end; 8791 case NFS4_CHECK_STATEID_UNCONFIRMED: 8792 *cs->statusp = resp->status = NFS4ERR_BAD_STATEID; 8793 goto end; 8794 case NFS4_CHECK_STATEID_CLOSED: 8795 *cs->statusp = resp->status = NFS4ERR_OLD_STATEID; 8796 goto end; 8797 case NFS4_CHECK_STATEID_OKAY: 8798 case NFS4_CHECK_STATEID_REPLAY: 8799 switch (rfs4_check_olo_seqid(olo->open_seqid, 8800 sp->rs_owner, resop)) { 8801 case NFS4_CHKSEQ_OKAY: 8802 if (rc == NFS4_CHECK_STATEID_OKAY) 8803 break; 8804 /* 8805 * This is replayed stateid; if seqid 8806 * matches next expected, then client 8807 * is using wrong seqid. 8808 */ 8809 /* FALLTHROUGH */ 8810 case NFS4_CHKSEQ_BAD: 8811 *cs->statusp = resp->status = NFS4ERR_BAD_SEQID; 8812 goto end; 8813 case NFS4_CHKSEQ_REPLAY: 8814 /* This is a duplicate LOCK request */ 8815 dup_lock = TRUE; 8816 8817 /* 8818 * For a duplicate we do not want to 8819 * create a new lockowner as it should 8820 * already exist. 8821 * Turn off the lockowner create flag. 8822 */ 8823 lcreate = FALSE; 8824 } 8825 break; 8826 } 8827 8828 lo = rfs4_findlockowner(&olo->lock_owner, &lcreate); 8829 if (lo == NULL) { 8830 NFS4_DEBUG(rfs4_debug, 8831 (CE_NOTE, "rfs4_op_lock: no lock owner")); 8832 *cs->statusp = resp->status = NFS4ERR_RESOURCE; 8833 goto end; 8834 } 8835 8836 lsp = rfs4_findlo_state_by_owner(lo, sp, &create); 8837 if (lsp == NULL) { 8838 rfs4_update_lease(sp->rs_owner->ro_client); 8839 /* 8840 * Only update theh open_seqid if this is not 8841 * a duplicate request 8842 */ 8843 if (dup_lock == FALSE) { 8844 rfs4_update_open_sequence(sp->rs_owner); 8845 } 8846 8847 NFS4_DEBUG(rfs4_debug, 8848 (CE_NOTE, "rfs4_op_lock: no state")); 8849 *cs->statusp = resp->status = NFS4ERR_SERVERFAULT; 8850 rfs4_update_open_resp(sp->rs_owner, resop, NULL); 8851 rfs4_lockowner_rele(lo); 8852 goto end; 8853 } 8854 8855 /* 8856 * This is the new_lock_owner branch and the client is 8857 * supposed to be associating a new lock_owner with 8858 * the open file at this point. If we find that a 8859 * lock_owner/state association already exists and a 8860 * successful LOCK request was returned to the client, 8861 * an error is returned to the client since this is 8862 * not appropriate. The client should be using the 8863 * existing lock_owner branch. 8864 */ 8865 if (dup_lock == FALSE && create == FALSE) { 8866 if (lsp->rls_lock_completed == TRUE) { 8867 *cs->statusp = 8868 resp->status = NFS4ERR_BAD_SEQID; 8869 rfs4_lockowner_rele(lo); 8870 goto end; 8871 } 8872 } 8873 8874 rfs4_update_lease(sp->rs_owner->ro_client); 8875 8876 /* 8877 * Only update theh open_seqid if this is not 8878 * a duplicate request 8879 */ 8880 if (dup_lock == FALSE) { 8881 rfs4_update_open_sequence(sp->rs_owner); 8882 } 8883 8884 /* 8885 * If this is a duplicate lock request, just copy the 8886 * previously saved reply and return. 8887 */ 8888 if (dup_lock == TRUE) { 8889 /* verify that lock_seqid's match */ 8890 if (lsp->rls_seqid != olo->lock_seqid) { 8891 NFS4_DEBUG(rfs4_debug, 8892 (CE_NOTE, "rfs4_op_lock: Dup-Lock seqid bad" 8893 "lsp->seqid=%d old->seqid=%d", 8894 lsp->rls_seqid, olo->lock_seqid)); 8895 *cs->statusp = resp->status = NFS4ERR_BAD_SEQID; 8896 } else { 8897 rfs4_copy_reply(resop, &lsp->rls_reply); 8898 /* 8899 * Make sure to copy the just 8900 * retrieved reply status into the 8901 * overall compound status 8902 */ 8903 *cs->statusp = resp->status; 8904 } 8905 rfs4_lockowner_rele(lo); 8906 goto end; 8907 } 8908 8909 rfs4_dbe_lock(lsp->rls_dbe); 8910 8911 /* Make sure to update the lock sequence id */ 8912 lsp->rls_seqid = olo->lock_seqid; 8913 8914 NFS4_DEBUG(rfs4_debug, 8915 (CE_NOTE, "Lock seqid established as %d", lsp->rls_seqid)); 8916 8917 /* 8918 * This is used to signify the newly created lockowner 8919 * stateid and its sequence number. The checks for 8920 * sequence number and increment don't occur on the 8921 * very first lock request for a lockowner. 8922 */ 8923 lsp->rls_skip_seqid_check = TRUE; 8924 8925 /* hold off other access to lsp while we tinker */ 8926 rfs4_sw_enter(&lsp->rls_sw); 8927 ls_sw_held = TRUE; 8928 8929 rfs4_dbe_unlock(lsp->rls_dbe); 8930 8931 rfs4_lockowner_rele(lo); 8932 } else { 8933 stateid = &args->locker.locker4_u.lock_owner.lock_stateid; 8934 /* get lsp and hold the lock on the underlying file struct */ 8935 if ((status = rfs4_get_lo_state(stateid, &lsp, TRUE)) 8936 != NFS4_OK) { 8937 *cs->statusp = resp->status = status; 8938 DTRACE_NFSV4_2(op__lock__done, struct compound_state *, 8939 cs, LOCK4res *, resp); 8940 return; 8941 } 8942 create = FALSE; /* We didn't create lsp */ 8943 8944 /* Ensure specified filehandle matches */ 8945 if (cs->vp != lsp->rls_state->rs_finfo->rf_vp) { 8946 rfs4_lo_state_rele(lsp, TRUE); 8947 *cs->statusp = resp->status = NFS4ERR_BAD_STATEID; 8948 DTRACE_NFSV4_2(op__lock__done, struct compound_state *, 8949 cs, LOCK4res *, resp); 8950 return; 8951 } 8952 8953 /* hold off other access to lsp while we tinker */ 8954 rfs4_sw_enter(&lsp->rls_sw); 8955 ls_sw_held = TRUE; 8956 8957 switch (rfs4_check_lo_stateid_seqid(lsp, stateid)) { 8958 /* 8959 * The stateid looks like it was okay (expected to be 8960 * the next one) 8961 */ 8962 case NFS4_CHECK_STATEID_OKAY: 8963 /* 8964 * The sequence id is now checked. Determine 8965 * if this is a replay or if it is in the 8966 * expected (next) sequence. In the case of a 8967 * replay, there are two replay conditions 8968 * that may occur. The first is the normal 8969 * condition where a LOCK is done with a 8970 * NFS4_OK response and the stateid is 8971 * updated. That case is handled below when 8972 * the stateid is identified as a REPLAY. The 8973 * second is the case where an error is 8974 * returned, like NFS4ERR_DENIED, and the 8975 * sequence number is updated but the stateid 8976 * is not updated. This second case is dealt 8977 * with here. So it may seem odd that the 8978 * stateid is okay but the sequence id is a 8979 * replay but it is okay. 8980 */ 8981 switch (rfs4_check_lock_seqid( 8982 args->locker.locker4_u.lock_owner.lock_seqid, 8983 lsp, resop)) { 8984 case NFS4_CHKSEQ_REPLAY: 8985 if (resp->status != NFS4_OK) { 8986 /* 8987 * Here is our replay and need 8988 * to verify that the last 8989 * response was an error. 8990 */ 8991 *cs->statusp = resp->status; 8992 goto end; 8993 } 8994 /* 8995 * This is done since the sequence id 8996 * looked like a replay but it didn't 8997 * pass our check so a BAD_SEQID is 8998 * returned as a result. 8999 */ 9000 /*FALLTHROUGH*/ 9001 case NFS4_CHKSEQ_BAD: 9002 *cs->statusp = resp->status = NFS4ERR_BAD_SEQID; 9003 goto end; 9004 case NFS4_CHKSEQ_OKAY: 9005 /* Everything looks okay move ahead */ 9006 break; 9007 } 9008 break; 9009 case NFS4_CHECK_STATEID_OLD: 9010 *cs->statusp = resp->status = NFS4ERR_OLD_STATEID; 9011 goto end; 9012 case NFS4_CHECK_STATEID_BAD: 9013 *cs->statusp = resp->status = NFS4ERR_BAD_STATEID; 9014 goto end; 9015 case NFS4_CHECK_STATEID_EXPIRED: 9016 *cs->statusp = resp->status = NFS4ERR_EXPIRED; 9017 goto end; 9018 case NFS4_CHECK_STATEID_CLOSED: 9019 *cs->statusp = resp->status = NFS4ERR_OLD_STATEID; 9020 goto end; 9021 case NFS4_CHECK_STATEID_REPLAY: 9022 switch (rfs4_check_lock_seqid( 9023 args->locker.locker4_u.lock_owner.lock_seqid, 9024 lsp, resop)) { 9025 case NFS4_CHKSEQ_OKAY: 9026 /* 9027 * This is a replayed stateid; if 9028 * seqid matches the next expected, 9029 * then client is using wrong seqid. 9030 */ 9031 case NFS4_CHKSEQ_BAD: 9032 *cs->statusp = resp->status = NFS4ERR_BAD_SEQID; 9033 goto end; 9034 case NFS4_CHKSEQ_REPLAY: 9035 rfs4_update_lease(lsp->rls_locker->rl_client); 9036 *cs->statusp = status = resp->status; 9037 goto end; 9038 } 9039 break; 9040 default: 9041 ASSERT(FALSE); 9042 break; 9043 } 9044 9045 rfs4_update_lock_sequence(lsp); 9046 rfs4_update_lease(lsp->rls_locker->rl_client); 9047 } 9048 9049 /* 9050 * NFS4 only allows locking on regular files, so 9051 * verify type of object. 9052 */ 9053 if (cs->vp->v_type != VREG) { 9054 if (cs->vp->v_type == VDIR) 9055 status = NFS4ERR_ISDIR; 9056 else 9057 status = NFS4ERR_INVAL; 9058 goto out; 9059 } 9060 9061 cp = lsp->rls_state->rs_owner->ro_client; 9062 9063 if (rfs4_clnt_in_grace(cp) && !args->reclaim) { 9064 status = NFS4ERR_GRACE; 9065 goto out; 9066 } 9067 9068 if (rfs4_clnt_in_grace(cp) && args->reclaim && !cp->rc_can_reclaim) { 9069 status = NFS4ERR_NO_GRACE; 9070 goto out; 9071 } 9072 9073 if (!rfs4_clnt_in_grace(cp) && args->reclaim) { 9074 status = NFS4ERR_NO_GRACE; 9075 goto out; 9076 } 9077 9078 if (lsp->rls_state->rs_finfo->rf_dinfo.rd_dtype == OPEN_DELEGATE_WRITE) 9079 cs->deleg = TRUE; 9080 9081 status = rfs4_do_lock(lsp, args->locktype, 9082 args->offset, args->length, cs->cr, resop); 9083 9084 out: 9085 lsp->rls_skip_seqid_check = FALSE; 9086 9087 *cs->statusp = resp->status = status; 9088 9089 if (status == NFS4_OK) { 9090 resp->LOCK4res_u.lock_stateid = lsp->rls_lockid.stateid; 9091 lsp->rls_lock_completed = TRUE; 9092 } 9093 /* 9094 * Only update the "OPEN" response here if this was a new 9095 * lock_owner 9096 */ 9097 if (sp) 9098 rfs4_update_open_resp(sp->rs_owner, resop, NULL); 9099 9100 rfs4_update_lock_resp(lsp, resop); 9101 9102 end: 9103 if (lsp) { 9104 if (ls_sw_held) 9105 rfs4_sw_exit(&lsp->rls_sw); 9106 /* 9107 * If an sp obtained, then the lsp does not represent 9108 * a lock on the file struct. 9109 */ 9110 if (sp != NULL) 9111 rfs4_lo_state_rele(lsp, FALSE); 9112 else 9113 rfs4_lo_state_rele(lsp, TRUE); 9114 } 9115 if (sp) { 9116 rfs4_sw_exit(&sp->rs_owner->ro_sw); 9117 rfs4_state_rele(sp); 9118 } 9119 9120 DTRACE_NFSV4_2(op__lock__done, struct compound_state *, cs, 9121 LOCK4res *, resp); 9122 } 9123 9124 /* free function for LOCK/LOCKT */ 9125 static void 9126 lock_denied_free(nfs_resop4 *resop) 9127 { 9128 LOCK4denied *dp = NULL; 9129 9130 switch (resop->resop) { 9131 case OP_LOCK: 9132 if (resop->nfs_resop4_u.oplock.status == NFS4ERR_DENIED) 9133 dp = &resop->nfs_resop4_u.oplock.LOCK4res_u.denied; 9134 break; 9135 case OP_LOCKT: 9136 if (resop->nfs_resop4_u.oplockt.status == NFS4ERR_DENIED) 9137 dp = &resop->nfs_resop4_u.oplockt.denied; 9138 break; 9139 default: 9140 break; 9141 } 9142 9143 if (dp) 9144 kmem_free(dp->owner.owner_val, dp->owner.owner_len); 9145 } 9146 9147 /*ARGSUSED*/ 9148 void 9149 rfs4_op_locku(nfs_argop4 *argop, nfs_resop4 *resop, 9150 struct svc_req *req, struct compound_state *cs) 9151 { 9152 LOCKU4args *args = &argop->nfs_argop4_u.oplocku; 9153 LOCKU4res *resp = &resop->nfs_resop4_u.oplocku; 9154 nfsstat4 status; 9155 stateid4 *stateid = &args->lock_stateid; 9156 rfs4_lo_state_t *lsp; 9157 9158 DTRACE_NFSV4_2(op__locku__start, struct compound_state *, cs, 9159 LOCKU4args *, args); 9160 9161 if (cs->vp == NULL) { 9162 *cs->statusp = resp->status = NFS4ERR_NOFILEHANDLE; 9163 DTRACE_NFSV4_2(op__locku__done, struct compound_state *, cs, 9164 LOCKU4res *, resp); 9165 return; 9166 } 9167 9168 if ((status = rfs4_get_lo_state(stateid, &lsp, TRUE)) != NFS4_OK) { 9169 *cs->statusp = resp->status = status; 9170 DTRACE_NFSV4_2(op__locku__done, struct compound_state *, cs, 9171 LOCKU4res *, resp); 9172 return; 9173 } 9174 9175 /* Ensure specified filehandle matches */ 9176 if (cs->vp != lsp->rls_state->rs_finfo->rf_vp) { 9177 rfs4_lo_state_rele(lsp, TRUE); 9178 *cs->statusp = resp->status = NFS4ERR_BAD_STATEID; 9179 DTRACE_NFSV4_2(op__locku__done, struct compound_state *, cs, 9180 LOCKU4res *, resp); 9181 return; 9182 } 9183 9184 /* hold off other access to lsp while we tinker */ 9185 rfs4_sw_enter(&lsp->rls_sw); 9186 9187 switch (rfs4_check_lo_stateid_seqid(lsp, stateid)) { 9188 case NFS4_CHECK_STATEID_OKAY: 9189 if (rfs4_check_lock_seqid(args->seqid, lsp, resop) 9190 != NFS4_CHKSEQ_OKAY) { 9191 *cs->statusp = resp->status = NFS4ERR_BAD_SEQID; 9192 goto end; 9193 } 9194 break; 9195 case NFS4_CHECK_STATEID_OLD: 9196 *cs->statusp = resp->status = NFS4ERR_OLD_STATEID; 9197 goto end; 9198 case NFS4_CHECK_STATEID_BAD: 9199 *cs->statusp = resp->status = NFS4ERR_BAD_STATEID; 9200 goto end; 9201 case NFS4_CHECK_STATEID_EXPIRED: 9202 *cs->statusp = resp->status = NFS4ERR_EXPIRED; 9203 goto end; 9204 case NFS4_CHECK_STATEID_CLOSED: 9205 *cs->statusp = resp->status = NFS4ERR_OLD_STATEID; 9206 goto end; 9207 case NFS4_CHECK_STATEID_REPLAY: 9208 switch (rfs4_check_lock_seqid(args->seqid, lsp, resop)) { 9209 case NFS4_CHKSEQ_OKAY: 9210 /* 9211 * This is a replayed stateid; if 9212 * seqid matches the next expected, 9213 * then client is using wrong seqid. 9214 */ 9215 case NFS4_CHKSEQ_BAD: 9216 *cs->statusp = resp->status = NFS4ERR_BAD_SEQID; 9217 goto end; 9218 case NFS4_CHKSEQ_REPLAY: 9219 rfs4_update_lease(lsp->rls_locker->rl_client); 9220 *cs->statusp = status = resp->status; 9221 goto end; 9222 } 9223 break; 9224 default: 9225 ASSERT(FALSE); 9226 break; 9227 } 9228 9229 rfs4_update_lock_sequence(lsp); 9230 rfs4_update_lease(lsp->rls_locker->rl_client); 9231 9232 /* 9233 * NFS4 only allows locking on regular files, so 9234 * verify type of object. 9235 */ 9236 if (cs->vp->v_type != VREG) { 9237 if (cs->vp->v_type == VDIR) 9238 status = NFS4ERR_ISDIR; 9239 else 9240 status = NFS4ERR_INVAL; 9241 goto out; 9242 } 9243 9244 if (rfs4_clnt_in_grace(lsp->rls_state->rs_owner->ro_client)) { 9245 status = NFS4ERR_GRACE; 9246 goto out; 9247 } 9248 9249 status = rfs4_do_lock(lsp, args->locktype, 9250 args->offset, args->length, cs->cr, resop); 9251 9252 out: 9253 *cs->statusp = resp->status = status; 9254 9255 if (status == NFS4_OK) 9256 resp->lock_stateid = lsp->rls_lockid.stateid; 9257 9258 rfs4_update_lock_resp(lsp, resop); 9259 9260 end: 9261 rfs4_sw_exit(&lsp->rls_sw); 9262 rfs4_lo_state_rele(lsp, TRUE); 9263 9264 DTRACE_NFSV4_2(op__locku__done, struct compound_state *, cs, 9265 LOCKU4res *, resp); 9266 } 9267 9268 /* 9269 * LOCKT is a best effort routine, the client can not be guaranteed that 9270 * the status return is still in effect by the time the reply is received. 9271 * They are numerous race conditions in this routine, but we are not required 9272 * and can not be accurate. 9273 */ 9274 /*ARGSUSED*/ 9275 void 9276 rfs4_op_lockt(nfs_argop4 *argop, nfs_resop4 *resop, 9277 struct svc_req *req, struct compound_state *cs) 9278 { 9279 LOCKT4args *args = &argop->nfs_argop4_u.oplockt; 9280 LOCKT4res *resp = &resop->nfs_resop4_u.oplockt; 9281 rfs4_lockowner_t *lo; 9282 rfs4_client_t *cp; 9283 bool_t create = FALSE; 9284 struct flock64 flk; 9285 int error; 9286 int flag = FREAD | FWRITE; 9287 int ltype; 9288 length4 posix_length; 9289 sysid_t sysid; 9290 pid_t pid; 9291 9292 DTRACE_NFSV4_2(op__lockt__start, struct compound_state *, cs, 9293 LOCKT4args *, args); 9294 9295 if (cs->vp == NULL) { 9296 *cs->statusp = resp->status = NFS4ERR_NOFILEHANDLE; 9297 goto out; 9298 } 9299 9300 /* 9301 * NFS4 only allows locking on regular files, so 9302 * verify type of object. 9303 */ 9304 if (cs->vp->v_type != VREG) { 9305 if (cs->vp->v_type == VDIR) 9306 *cs->statusp = resp->status = NFS4ERR_ISDIR; 9307 else 9308 *cs->statusp = resp->status = NFS4ERR_INVAL; 9309 goto out; 9310 } 9311 9312 /* 9313 * Check out the clientid to ensure the server knows about it 9314 * so that we correctly inform the client of a server reboot. 9315 */ 9316 if ((cp = rfs4_findclient_by_id(args->owner.clientid, FALSE)) 9317 == NULL) { 9318 *cs->statusp = resp->status = 9319 rfs4_check_clientid(&args->owner.clientid, 0); 9320 goto out; 9321 } 9322 if (rfs4_lease_expired(cp)) { 9323 rfs4_client_close(cp); 9324 /* 9325 * Protocol doesn't allow returning NFS4ERR_STALE as 9326 * other operations do on this check so STALE_CLIENTID 9327 * is returned instead 9328 */ 9329 *cs->statusp = resp->status = NFS4ERR_STALE_CLIENTID; 9330 goto out; 9331 } 9332 9333 if (rfs4_clnt_in_grace(cp) && !(cp->rc_can_reclaim)) { 9334 *cs->statusp = resp->status = NFS4ERR_GRACE; 9335 rfs4_client_rele(cp); 9336 goto out; 9337 } 9338 rfs4_client_rele(cp); 9339 9340 resp->status = NFS4_OK; 9341 9342 switch (args->locktype) { 9343 case READ_LT: 9344 case READW_LT: 9345 ltype = F_RDLCK; 9346 break; 9347 case WRITE_LT: 9348 case WRITEW_LT: 9349 ltype = F_WRLCK; 9350 break; 9351 } 9352 9353 posix_length = args->length; 9354 /* Check for zero length. To lock to end of file use all ones for V4 */ 9355 if (posix_length == 0) { 9356 *cs->statusp = resp->status = NFS4ERR_INVAL; 9357 goto out; 9358 } else if (posix_length == (length4)(~0)) { 9359 posix_length = 0; /* Posix to end of file */ 9360 } 9361 9362 /* Find or create a lockowner */ 9363 lo = rfs4_findlockowner(&args->owner, &create); 9364 9365 if (lo) { 9366 pid = lo->rl_pid; 9367 if ((resp->status = 9368 rfs4_client_sysid(lo->rl_client, &sysid)) != NFS4_OK) 9369 goto err; 9370 } else { 9371 pid = 0; 9372 sysid = lockt_sysid; 9373 } 9374 retry: 9375 flk.l_type = ltype; 9376 flk.l_whence = 0; /* SEEK_SET */ 9377 flk.l_start = args->offset; 9378 flk.l_len = posix_length; 9379 flk.l_sysid = sysid; 9380 flk.l_pid = pid; 9381 flag |= F_REMOTELOCK; 9382 9383 LOCK_PRINT(rfs4_debug, "rfs4_op_lockt", F_GETLK, &flk); 9384 9385 /* Note that length4 is uint64_t but l_len and l_start are off64_t */ 9386 if (flk.l_len < 0 || flk.l_start < 0) { 9387 resp->status = NFS4ERR_INVAL; 9388 goto err; 9389 } 9390 error = VOP_FRLOCK(cs->vp, F_GETLK, &flk, flag, (u_offset_t)0, 9391 NULL, cs->cr, NULL); 9392 9393 /* 9394 * N.B. We map error values to nfsv4 errors. This is differrent 9395 * than puterrno4 routine. 9396 */ 9397 switch (error) { 9398 case 0: 9399 if (flk.l_type == F_UNLCK) 9400 resp->status = NFS4_OK; 9401 else { 9402 if (lock_denied(&resp->denied, &flk) == NFS4ERR_EXPIRED) 9403 goto retry; 9404 resp->status = NFS4ERR_DENIED; 9405 } 9406 break; 9407 case EOVERFLOW: 9408 resp->status = NFS4ERR_INVAL; 9409 break; 9410 case EINVAL: 9411 resp->status = NFS4ERR_NOTSUPP; 9412 break; 9413 default: 9414 cmn_err(CE_WARN, "rfs4_op_lockt: unexpected errno (%d)", 9415 error); 9416 resp->status = NFS4ERR_SERVERFAULT; 9417 break; 9418 } 9419 9420 err: 9421 if (lo) 9422 rfs4_lockowner_rele(lo); 9423 *cs->statusp = resp->status; 9424 out: 9425 DTRACE_NFSV4_2(op__lockt__done, struct compound_state *, cs, 9426 LOCKT4res *, resp); 9427 } 9428 9429 int 9430 rfs4_share(rfs4_state_t *sp, uint32_t access, uint32_t deny) 9431 { 9432 int err; 9433 int cmd; 9434 vnode_t *vp; 9435 struct shrlock shr; 9436 struct shr_locowner shr_loco; 9437 int fflags = 0; 9438 9439 ASSERT(rfs4_dbe_islocked(sp->rs_dbe)); 9440 ASSERT(sp->rs_owner->ro_client->rc_sysidt != LM_NOSYSID); 9441 9442 if (sp->rs_closed) 9443 return (NFS4ERR_OLD_STATEID); 9444 9445 vp = sp->rs_finfo->rf_vp; 9446 ASSERT(vp); 9447 9448 shr.s_access = shr.s_deny = 0; 9449 9450 if (access & OPEN4_SHARE_ACCESS_READ) { 9451 fflags |= FREAD; 9452 shr.s_access |= F_RDACC; 9453 } 9454 if (access & OPEN4_SHARE_ACCESS_WRITE) { 9455 fflags |= FWRITE; 9456 shr.s_access |= F_WRACC; 9457 } 9458 ASSERT(shr.s_access); 9459 9460 if (deny & OPEN4_SHARE_DENY_READ) 9461 shr.s_deny |= F_RDDNY; 9462 if (deny & OPEN4_SHARE_DENY_WRITE) 9463 shr.s_deny |= F_WRDNY; 9464 9465 shr.s_pid = rfs4_dbe_getid(sp->rs_owner->ro_dbe); 9466 shr.s_sysid = sp->rs_owner->ro_client->rc_sysidt; 9467 shr_loco.sl_pid = shr.s_pid; 9468 shr_loco.sl_id = shr.s_sysid; 9469 shr.s_owner = (caddr_t)&shr_loco; 9470 shr.s_own_len = sizeof (shr_loco); 9471 9472 cmd = nbl_need_check(vp) ? F_SHARE_NBMAND : F_SHARE; 9473 9474 err = VOP_SHRLOCK(vp, cmd, &shr, fflags, CRED(), NULL); 9475 if (err != 0) { 9476 if (err == EAGAIN) 9477 err = NFS4ERR_SHARE_DENIED; 9478 else 9479 err = puterrno4(err); 9480 return (err); 9481 } 9482 9483 sp->rs_share_access |= access; 9484 sp->rs_share_deny |= deny; 9485 9486 return (0); 9487 } 9488 9489 int 9490 rfs4_unshare(rfs4_state_t *sp) 9491 { 9492 int err; 9493 struct shrlock shr; 9494 struct shr_locowner shr_loco; 9495 9496 ASSERT(rfs4_dbe_islocked(sp->rs_dbe)); 9497 9498 if (sp->rs_closed || sp->rs_share_access == 0) 9499 return (0); 9500 9501 ASSERT(sp->rs_owner->ro_client->rc_sysidt != LM_NOSYSID); 9502 ASSERT(sp->rs_finfo->rf_vp); 9503 9504 shr.s_access = shr.s_deny = 0; 9505 shr.s_pid = rfs4_dbe_getid(sp->rs_owner->ro_dbe); 9506 shr.s_sysid = sp->rs_owner->ro_client->rc_sysidt; 9507 shr_loco.sl_pid = shr.s_pid; 9508 shr_loco.sl_id = shr.s_sysid; 9509 shr.s_owner = (caddr_t)&shr_loco; 9510 shr.s_own_len = sizeof (shr_loco); 9511 9512 err = VOP_SHRLOCK(sp->rs_finfo->rf_vp, F_UNSHARE, &shr, 0, CRED(), 9513 NULL); 9514 if (err != 0) { 9515 err = puterrno4(err); 9516 return (err); 9517 } 9518 9519 sp->rs_share_access = 0; 9520 sp->rs_share_deny = 0; 9521 9522 return (0); 9523 9524 } 9525 9526 static int 9527 rdma_setup_read_data4(READ4args *args, READ4res *rok) 9528 { 9529 struct clist *wcl; 9530 count4 count = rok->data_len; 9531 int wlist_len; 9532 9533 wcl = args->wlist; 9534 if (rdma_setup_read_chunks(wcl, count, &wlist_len) == FALSE) { 9535 return (FALSE); 9536 } 9537 wcl = args->wlist; 9538 rok->wlist_len = wlist_len; 9539 rok->wlist = wcl; 9540 return (TRUE); 9541 } 9542 9543 /* tunable to disable server referrals */ 9544 int rfs4_no_referrals = 0; 9545 9546 /* 9547 * Find an NFS record in reparse point data. 9548 * Returns 0 for success and <0 or an errno value on failure. 9549 */ 9550 int 9551 vn_find_nfs_record(vnode_t *vp, nvlist_t **nvlp, char **svcp, char **datap) 9552 { 9553 int err; 9554 char *stype, *val; 9555 nvlist_t *nvl; 9556 nvpair_t *curr; 9557 9558 if ((nvl = reparse_init()) == NULL) 9559 return (-1); 9560 9561 if ((err = reparse_vnode_parse(vp, nvl)) != 0) { 9562 reparse_free(nvl); 9563 return (err); 9564 } 9565 9566 curr = NULL; 9567 while ((curr = nvlist_next_nvpair(nvl, curr)) != NULL) { 9568 if ((stype = nvpair_name(curr)) == NULL) { 9569 reparse_free(nvl); 9570 return (-2); 9571 } 9572 if (strncasecmp(stype, "NFS", 3) == 0) 9573 break; 9574 } 9575 9576 if ((curr == NULL) || 9577 (nvpair_value_string(curr, &val))) { 9578 reparse_free(nvl); 9579 return (-3); 9580 } 9581 *nvlp = nvl; 9582 *svcp = stype; 9583 *datap = val; 9584 return (0); 9585 } 9586 9587 int 9588 vn_is_nfs_reparse(vnode_t *vp, cred_t *cr) 9589 { 9590 nvlist_t *nvl; 9591 char *s, *d; 9592 9593 if (rfs4_no_referrals != 0) 9594 return (B_FALSE); 9595 9596 if (vn_is_reparse(vp, cr, NULL) == B_FALSE) 9597 return (B_FALSE); 9598 9599 if (vn_find_nfs_record(vp, &nvl, &s, &d) != 0) 9600 return (B_FALSE); 9601 9602 reparse_free(nvl); 9603 9604 return (B_TRUE); 9605 } 9606 9607 /* 9608 * There is a user-level copy of this routine in ref_subr.c. 9609 * Changes should be kept in sync. 9610 */ 9611 static int 9612 nfs4_create_components(char *path, component4 *comp4) 9613 { 9614 int slen, plen, ncomp; 9615 char *ori_path, *nxtc, buf[MAXNAMELEN]; 9616 9617 if (path == NULL) 9618 return (0); 9619 9620 plen = strlen(path) + 1; /* include the terminator */ 9621 ori_path = path; 9622 ncomp = 0; 9623 9624 /* count number of components in the path */ 9625 for (nxtc = path; nxtc < ori_path + plen; nxtc++) { 9626 if (*nxtc == '/' || *nxtc == '\0' || *nxtc == '\n') { 9627 if ((slen = nxtc - path) == 0) { 9628 path = nxtc + 1; 9629 continue; 9630 } 9631 9632 if (comp4 != NULL) { 9633 bcopy(path, buf, slen); 9634 buf[slen] = '\0'; 9635 (void) str_to_utf8(buf, &comp4[ncomp]); 9636 } 9637 9638 ncomp++; /* 1 valid component */ 9639 path = nxtc + 1; 9640 } 9641 if (*nxtc == '\0' || *nxtc == '\n') 9642 break; 9643 } 9644 9645 return (ncomp); 9646 } 9647 9648 /* 9649 * There is a user-level copy of this routine in ref_subr.c. 9650 * Changes should be kept in sync. 9651 */ 9652 static int 9653 make_pathname4(char *path, pathname4 *pathname) 9654 { 9655 int ncomp; 9656 component4 *comp4; 9657 9658 if (pathname == NULL) 9659 return (0); 9660 9661 if (path == NULL) { 9662 pathname->pathname4_val = NULL; 9663 pathname->pathname4_len = 0; 9664 return (0); 9665 } 9666 9667 /* count number of components to alloc buffer */ 9668 if ((ncomp = nfs4_create_components(path, NULL)) == 0) { 9669 pathname->pathname4_val = NULL; 9670 pathname->pathname4_len = 0; 9671 return (0); 9672 } 9673 comp4 = kmem_zalloc(ncomp * sizeof (component4), KM_SLEEP); 9674 9675 /* copy components into allocated buffer */ 9676 ncomp = nfs4_create_components(path, comp4); 9677 9678 pathname->pathname4_val = comp4; 9679 pathname->pathname4_len = ncomp; 9680 9681 return (ncomp); 9682 } 9683 9684 #define xdr_fs_locations4 xdr_fattr4_fs_locations 9685 9686 fs_locations4 * 9687 fetch_referral(vnode_t *vp, cred_t *cr) 9688 { 9689 nvlist_t *nvl; 9690 char *stype, *sdata; 9691 fs_locations4 *result; 9692 char buf[1024]; 9693 size_t bufsize; 9694 XDR xdr; 9695 int err; 9696 9697 /* 9698 * Check attrs to ensure it's a reparse point 9699 */ 9700 if (vn_is_reparse(vp, cr, NULL) == B_FALSE) 9701 return (NULL); 9702 9703 /* 9704 * Look for an NFS record and get the type and data 9705 */ 9706 if (vn_find_nfs_record(vp, &nvl, &stype, &sdata) != 0) 9707 return (NULL); 9708 9709 /* 9710 * With the type and data, upcall to get the referral 9711 */ 9712 bufsize = sizeof (buf); 9713 bzero(buf, sizeof (buf)); 9714 err = reparse_kderef((const char *)stype, (const char *)sdata, 9715 buf, &bufsize); 9716 reparse_free(nvl); 9717 9718 DTRACE_PROBE4(nfs4serv__func__referral__upcall, 9719 char *, stype, char *, sdata, char *, buf, int, err); 9720 if (err) { 9721 cmn_err(CE_NOTE, 9722 "reparsed daemon not running: unable to get referral (%d)", 9723 err); 9724 return (NULL); 9725 } 9726 9727 /* 9728 * We get an XDR'ed record back from the kderef call 9729 */ 9730 xdrmem_create(&xdr, buf, bufsize, XDR_DECODE); 9731 result = kmem_alloc(sizeof (fs_locations4), KM_SLEEP); 9732 err = xdr_fs_locations4(&xdr, result); 9733 XDR_DESTROY(&xdr); 9734 if (err != TRUE) { 9735 DTRACE_PROBE1(nfs4serv__func__referral__upcall__xdrfail, 9736 int, err); 9737 return (NULL); 9738 } 9739 9740 /* 9741 * Look at path to recover fs_root, ignoring the leading '/' 9742 */ 9743 (void) make_pathname4(vp->v_path, &result->fs_root); 9744 9745 return (result); 9746 } 9747 9748 char * 9749 build_symlink(vnode_t *vp, cred_t *cr, size_t *strsz) 9750 { 9751 fs_locations4 *fsl; 9752 fs_location4 *fs; 9753 char *server, *path, *symbuf; 9754 static char *prefix = "/net/"; 9755 int i, size, npaths; 9756 uint_t len; 9757 9758 /* Get the referral */ 9759 if ((fsl = fetch_referral(vp, cr)) == NULL) 9760 return (NULL); 9761 9762 /* Deal with only the first location and first server */ 9763 fs = &fsl->locations_val[0]; 9764 server = utf8_to_str(&fs->server_val[0], &len, NULL); 9765 if (server == NULL) { 9766 rfs4_free_fs_locations4(fsl); 9767 kmem_free(fsl, sizeof (fs_locations4)); 9768 return (NULL); 9769 } 9770 9771 /* Figure out size for "/net/" + host + /path/path/path + NULL */ 9772 size = strlen(prefix) + len; 9773 for (i = 0; i < fs->rootpath.pathname4_len; i++) 9774 size += fs->rootpath.pathname4_val[i].utf8string_len + 1; 9775 9776 /* Allocate the symlink buffer and fill it */ 9777 symbuf = kmem_zalloc(size, KM_SLEEP); 9778 (void) strcat(symbuf, prefix); 9779 (void) strcat(symbuf, server); 9780 kmem_free(server, len); 9781 9782 npaths = 0; 9783 for (i = 0; i < fs->rootpath.pathname4_len; i++) { 9784 path = utf8_to_str(&fs->rootpath.pathname4_val[i], &len, NULL); 9785 if (path == NULL) 9786 continue; 9787 (void) strcat(symbuf, "/"); 9788 (void) strcat(symbuf, path); 9789 npaths++; 9790 kmem_free(path, len); 9791 } 9792 9793 rfs4_free_fs_locations4(fsl); 9794 kmem_free(fsl, sizeof (fs_locations4)); 9795 9796 if (strsz != NULL) 9797 *strsz = size; 9798 return (symbuf); 9799 } 9800 9801 /* 9802 * Check to see if we have a downrev Solaris client, so that we 9803 * can send it a symlink instead of a referral. 9804 */ 9805 int 9806 client_is_downrev(struct svc_req *req) 9807 { 9808 struct sockaddr *ca; 9809 rfs4_clntip_t *ci; 9810 bool_t create = FALSE; 9811 int is_downrev; 9812 9813 ca = (struct sockaddr *)svc_getrpccaller(req->rq_xprt)->buf; 9814 ASSERT(ca); 9815 ci = rfs4_find_clntip(ca, &create); 9816 if (ci == NULL) 9817 return (0); 9818 is_downrev = ci->ri_no_referrals; 9819 rfs4_dbe_rele(ci->ri_dbe); 9820 return (is_downrev); 9821 } 9822