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