1 /* 2 * CDDL HEADER START 3 * 4 * The contents of this file are subject to the terms of the 5 * Common Development and Distribution License (the "License"). 6 * You may not use this file except in compliance with the License. 7 * 8 * You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE 9 * or http://www.opensolaris.org/os/licensing. 10 * See the License for the specific language governing permissions 11 * and limitations under the License. 12 * 13 * When distributing Covered Code, include this CDDL HEADER in each 14 * file and include the License file at usr/src/OPENSOLARIS.LICENSE. 15 * If applicable, add the following below this CDDL HEADER, with the 16 * fields enclosed by brackets "[]" replaced with your own identifying 17 * information: Portions Copyright [yyyy] [name of copyright owner] 18 * 19 * CDDL HEADER END 20 */ 21 /* 22 * Copyright (c) 2003, 2010, Oracle and/or its affiliates. All rights reserved. 23 * Copyright 2012 Nexenta Systems, Inc. All rights reserved. 24 */ 25 26 27 /* 28 * Copyright (c) 1983,1984,1985,1986,1987,1988,1989 AT&T. 29 * All Rights Reserved 30 */ 31 32 #include <sys/param.h> 33 #include <sys/types.h> 34 #include <sys/systm.h> 35 #include <sys/cred.h> 36 #include <sys/buf.h> 37 #include <sys/vfs.h> 38 #include <sys/vfs_opreg.h> 39 #include <sys/vnode.h> 40 #include <sys/uio.h> 41 #include <sys/errno.h> 42 #include <sys/sysmacros.h> 43 #include <sys/statvfs.h> 44 #include <sys/kmem.h> 45 #include <sys/dirent.h> 46 #include <sys/cmn_err.h> 47 #include <sys/debug.h> 48 #include <sys/systeminfo.h> 49 #include <sys/flock.h> 50 #include <sys/pathname.h> 51 #include <sys/nbmlock.h> 52 #include <sys/share.h> 53 #include <sys/atomic.h> 54 #include <sys/policy.h> 55 #include <sys/fem.h> 56 #include <sys/sdt.h> 57 #include <sys/ddi.h> 58 #include <sys/zone.h> 59 60 #include <fs/fs_reparse.h> 61 62 #include <rpc/types.h> 63 #include <rpc/auth.h> 64 #include <rpc/rpcsec_gss.h> 65 #include <rpc/svc.h> 66 67 #include <nfs/nfs.h> 68 #include <nfs/export.h> 69 #include <nfs/nfs_cmd.h> 70 #include <nfs/lm.h> 71 #include <nfs/nfs4.h> 72 73 #include <sys/strsubr.h> 74 #include <sys/strsun.h> 75 76 #include <inet/common.h> 77 #include <inet/ip.h> 78 #include <inet/ip6.h> 79 80 #include <sys/tsol/label.h> 81 #include <sys/tsol/tndb.h> 82 83 #define RFS4_MAXLOCK_TRIES 4 /* Try to get the lock this many times */ 84 static int rfs4_maxlock_tries = RFS4_MAXLOCK_TRIES; 85 #define RFS4_LOCK_DELAY 10 /* Milliseconds */ 86 static clock_t rfs4_lock_delay = RFS4_LOCK_DELAY; 87 extern struct svc_ops rdma_svc_ops; 88 extern int nfs_loaned_buffers; 89 /* End of Tunables */ 90 91 static int rdma_setup_read_data4(READ4args *, READ4res *); 92 93 /* 94 * Used to bump the stateid4.seqid value and show changes in the stateid 95 */ 96 #define next_stateid(sp) (++(sp)->bits.chgseq) 97 98 /* 99 * RFS4_MINLEN_ENTRY4: XDR-encoded size of smallest possible dirent. 100 * This is used to return NFS4ERR_TOOSMALL when clients specify 101 * maxcount that isn't large enough to hold the smallest possible 102 * XDR encoded dirent. 103 * 104 * sizeof cookie (8 bytes) + 105 * sizeof name_len (4 bytes) + 106 * sizeof smallest (padded) name (4 bytes) + 107 * sizeof bitmap4_len (12 bytes) + NOTE: we always encode len=2 bm4 108 * sizeof attrlist4_len (4 bytes) + 109 * sizeof next boolean (4 bytes) 110 * 111 * RFS4_MINLEN_RDDIR4: XDR-encoded size of READDIR op reply containing 112 * the smallest possible entry4 (assumes no attrs requested). 113 * sizeof nfsstat4 (4 bytes) + 114 * sizeof verifier4 (8 bytes) + 115 * sizeof entry4list bool (4 bytes) + 116 * sizeof entry4 (36 bytes) + 117 * sizeof eof bool (4 bytes) 118 * 119 * RFS4_MINLEN_RDDIR_BUF: minimum length of buffer server will provide to 120 * VOP_READDIR. Its value is the size of the maximum possible dirent 121 * for solaris. The DIRENT64_RECLEN macro returns the size of dirent 122 * required for a given name length. MAXNAMELEN is the maximum 123 * filename length allowed in Solaris. The first two DIRENT64_RECLEN() 124 * macros are to allow for . and .. entries -- just a minor tweak to try 125 * and guarantee that buffer we give to VOP_READDIR will be large enough 126 * to hold ., .., and the largest possible solaris dirent64. 127 */ 128 #define RFS4_MINLEN_ENTRY4 36 129 #define RFS4_MINLEN_RDDIR4 (4 + NFS4_VERIFIER_SIZE + 4 + RFS4_MINLEN_ENTRY4 + 4) 130 #define RFS4_MINLEN_RDDIR_BUF \ 131 (DIRENT64_RECLEN(1) + DIRENT64_RECLEN(2) + DIRENT64_RECLEN(MAXNAMELEN)) 132 133 /* 134 * It would be better to pad to 4 bytes since that's what XDR would do, 135 * but the dirents UFS gives us are already padded to 8, so just take 136 * what we're given. Dircount is only a hint anyway. Currently the 137 * solaris kernel is ASCII only, so there's no point in calling the 138 * UTF8 functions. 139 * 140 * dirent64: named padded to provide 8 byte struct alignment 141 * d_ino(8) + d_off(8) + d_reclen(2) + d_name(namelen + null(1) + pad) 142 * 143 * cookie: uint64_t + utf8namelen: uint_t + utf8name padded to 8 bytes 144 * 145 */ 146 #define DIRENT64_TO_DIRCOUNT(dp) \ 147 (3 * BYTES_PER_XDR_UNIT + DIRENT64_NAMELEN((dp)->d_reclen)) 148 149 time_t rfs4_start_time; /* Initialized in rfs4_srvrinit */ 150 151 static sysid_t lockt_sysid; /* dummy sysid for all LOCKT calls */ 152 153 u_longlong_t nfs4_srv_caller_id; 154 uint_t nfs4_srv_vkey = 0; 155 156 verifier4 Write4verf; 157 verifier4 Readdir4verf; 158 159 void rfs4_init_compound_state(struct compound_state *); 160 161 static void nullfree(caddr_t); 162 static void rfs4_op_inval(nfs_argop4 *, nfs_resop4 *, struct svc_req *, 163 struct compound_state *); 164 static void rfs4_op_access(nfs_argop4 *, nfs_resop4 *, struct svc_req *, 165 struct compound_state *); 166 static void rfs4_op_close(nfs_argop4 *, nfs_resop4 *, struct svc_req *, 167 struct compound_state *); 168 static void rfs4_op_commit(nfs_argop4 *, nfs_resop4 *, struct svc_req *, 169 struct compound_state *); 170 static void rfs4_op_create(nfs_argop4 *, nfs_resop4 *, struct svc_req *, 171 struct compound_state *); 172 static void rfs4_op_create_free(nfs_resop4 *resop); 173 static void rfs4_op_delegreturn(nfs_argop4 *, nfs_resop4 *, 174 struct svc_req *, struct compound_state *); 175 static void rfs4_op_delegpurge(nfs_argop4 *, nfs_resop4 *, 176 struct svc_req *, struct compound_state *); 177 static void rfs4_op_getattr(nfs_argop4 *, nfs_resop4 *, struct svc_req *, 178 struct compound_state *); 179 static void rfs4_op_getattr_free(nfs_resop4 *); 180 static void rfs4_op_getfh(nfs_argop4 *, nfs_resop4 *, struct svc_req *, 181 struct compound_state *); 182 static void rfs4_op_getfh_free(nfs_resop4 *); 183 static void rfs4_op_illegal(nfs_argop4 *, nfs_resop4 *, struct svc_req *, 184 struct compound_state *); 185 static void rfs4_op_link(nfs_argop4 *, nfs_resop4 *, struct svc_req *, 186 struct compound_state *); 187 static void rfs4_op_lock(nfs_argop4 *, nfs_resop4 *, struct svc_req *, 188 struct compound_state *); 189 static void lock_denied_free(nfs_resop4 *); 190 static void rfs4_op_locku(nfs_argop4 *, nfs_resop4 *, struct svc_req *, 191 struct compound_state *); 192 static void rfs4_op_lockt(nfs_argop4 *, nfs_resop4 *, struct svc_req *, 193 struct compound_state *); 194 static void rfs4_op_lookup(nfs_argop4 *, nfs_resop4 *, struct svc_req *, 195 struct compound_state *); 196 static void rfs4_op_lookupp(nfs_argop4 *, nfs_resop4 *, struct svc_req *, 197 struct compound_state *); 198 static void rfs4_op_openattr(nfs_argop4 *argop, nfs_resop4 *resop, 199 struct svc_req *req, struct compound_state *cs); 200 static void rfs4_op_nverify(nfs_argop4 *, nfs_resop4 *, struct svc_req *, 201 struct compound_state *); 202 static void rfs4_op_open(nfs_argop4 *, nfs_resop4 *, struct svc_req *, 203 struct compound_state *); 204 static void rfs4_op_open_confirm(nfs_argop4 *, nfs_resop4 *, 205 struct svc_req *, struct compound_state *); 206 static void rfs4_op_open_downgrade(nfs_argop4 *, nfs_resop4 *, 207 struct svc_req *, struct compound_state *); 208 static void rfs4_op_putfh(nfs_argop4 *, nfs_resop4 *, struct svc_req *, 209 struct compound_state *); 210 static void rfs4_op_putpubfh(nfs_argop4 *, nfs_resop4 *, struct svc_req *, 211 struct compound_state *); 212 static void rfs4_op_putrootfh(nfs_argop4 *, nfs_resop4 *, struct svc_req *, 213 struct compound_state *); 214 static void rfs4_op_read(nfs_argop4 *, nfs_resop4 *, struct svc_req *, 215 struct compound_state *); 216 static void rfs4_op_read_free(nfs_resop4 *); 217 static void rfs4_op_readdir_free(nfs_resop4 *resop); 218 static void rfs4_op_readlink(nfs_argop4 *, nfs_resop4 *, struct svc_req *, 219 struct compound_state *); 220 static void rfs4_op_readlink_free(nfs_resop4 *); 221 static void rfs4_op_release_lockowner(nfs_argop4 *, nfs_resop4 *, 222 struct svc_req *, struct compound_state *); 223 static void rfs4_op_remove(nfs_argop4 *, nfs_resop4 *, struct svc_req *, 224 struct compound_state *); 225 static void rfs4_op_rename(nfs_argop4 *, nfs_resop4 *, struct svc_req *, 226 struct compound_state *); 227 static void rfs4_op_renew(nfs_argop4 *, nfs_resop4 *, struct svc_req *, 228 struct compound_state *); 229 static void rfs4_op_restorefh(nfs_argop4 *, nfs_resop4 *, struct svc_req *, 230 struct compound_state *); 231 static void rfs4_op_savefh(nfs_argop4 *, nfs_resop4 *, struct svc_req *, 232 struct compound_state *); 233 static void rfs4_op_setattr(nfs_argop4 *, nfs_resop4 *, struct svc_req *, 234 struct compound_state *); 235 static void rfs4_op_verify(nfs_argop4 *, nfs_resop4 *, struct svc_req *, 236 struct compound_state *); 237 static void rfs4_op_write(nfs_argop4 *, nfs_resop4 *, struct svc_req *, 238 struct compound_state *); 239 static void rfs4_op_setclientid(nfs_argop4 *, nfs_resop4 *, 240 struct svc_req *, struct compound_state *); 241 static void rfs4_op_setclientid_confirm(nfs_argop4 *, nfs_resop4 *, 242 struct svc_req *req, struct compound_state *); 243 static void rfs4_op_secinfo(nfs_argop4 *, nfs_resop4 *, struct svc_req *, 244 struct compound_state *); 245 static void rfs4_op_secinfo_free(nfs_resop4 *); 246 247 static nfsstat4 check_open_access(uint32_t, 248 struct compound_state *, struct svc_req *); 249 nfsstat4 rfs4_client_sysid(rfs4_client_t *, sysid_t *); 250 void rfs4_ss_clid(rfs4_client_t *); 251 252 /* 253 * translation table for attrs 254 */ 255 struct nfs4_ntov_table { 256 union nfs4_attr_u *na; 257 uint8_t amap[NFS4_MAXNUM_ATTRS]; 258 int attrcnt; 259 bool_t vfsstat; 260 }; 261 262 static void nfs4_ntov_table_init(struct nfs4_ntov_table *ntovp); 263 static void nfs4_ntov_table_free(struct nfs4_ntov_table *ntovp, 264 struct nfs4_svgetit_arg *sargp); 265 266 static nfsstat4 do_rfs4_set_attrs(bitmap4 *resp, fattr4 *fattrp, 267 struct compound_state *cs, struct nfs4_svgetit_arg *sargp, 268 struct nfs4_ntov_table *ntovp, nfs4_attr_cmd_t cmd); 269 270 fem_t *deleg_rdops; 271 fem_t *deleg_wrops; 272 273 rfs4_servinst_t *rfs4_cur_servinst = NULL; /* current server instance */ 274 kmutex_t rfs4_servinst_lock; /* protects linked list */ 275 int rfs4_seen_first_compound; /* set first time we see one */ 276 277 /* 278 * NFS4 op dispatch table 279 */ 280 281 struct rfsv4disp { 282 void (*dis_proc)(); /* proc to call */ 283 void (*dis_resfree)(); /* frees space allocated by proc */ 284 int dis_flags; /* RPC_IDEMPOTENT, etc... */ 285 }; 286 287 static struct rfsv4disp rfsv4disptab[] = { 288 /* 289 * NFS VERSION 4 290 */ 291 292 /* RFS_NULL = 0 */ 293 {rfs4_op_illegal, nullfree, 0}, 294 295 /* UNUSED = 1 */ 296 {rfs4_op_illegal, nullfree, 0}, 297 298 /* UNUSED = 2 */ 299 {rfs4_op_illegal, nullfree, 0}, 300 301 /* OP_ACCESS = 3 */ 302 {rfs4_op_access, nullfree, RPC_IDEMPOTENT}, 303 304 /* OP_CLOSE = 4 */ 305 {rfs4_op_close, nullfree, 0}, 306 307 /* OP_COMMIT = 5 */ 308 {rfs4_op_commit, nullfree, RPC_IDEMPOTENT}, 309 310 /* OP_CREATE = 6 */ 311 {rfs4_op_create, nullfree, 0}, 312 313 /* OP_DELEGPURGE = 7 */ 314 {rfs4_op_delegpurge, nullfree, 0}, 315 316 /* OP_DELEGRETURN = 8 */ 317 {rfs4_op_delegreturn, nullfree, 0}, 318 319 /* OP_GETATTR = 9 */ 320 {rfs4_op_getattr, rfs4_op_getattr_free, RPC_IDEMPOTENT}, 321 322 /* OP_GETFH = 10 */ 323 {rfs4_op_getfh, rfs4_op_getfh_free, RPC_ALL}, 324 325 /* OP_LINK = 11 */ 326 {rfs4_op_link, nullfree, 0}, 327 328 /* OP_LOCK = 12 */ 329 {rfs4_op_lock, lock_denied_free, 0}, 330 331 /* OP_LOCKT = 13 */ 332 {rfs4_op_lockt, lock_denied_free, 0}, 333 334 /* OP_LOCKU = 14 */ 335 {rfs4_op_locku, nullfree, 0}, 336 337 /* OP_LOOKUP = 15 */ 338 {rfs4_op_lookup, nullfree, (RPC_IDEMPOTENT | RPC_PUBLICFH_OK)}, 339 340 /* OP_LOOKUPP = 16 */ 341 {rfs4_op_lookupp, nullfree, (RPC_IDEMPOTENT | RPC_PUBLICFH_OK)}, 342 343 /* OP_NVERIFY = 17 */ 344 {rfs4_op_nverify, nullfree, RPC_IDEMPOTENT}, 345 346 /* OP_OPEN = 18 */ 347 {rfs4_op_open, rfs4_free_reply, 0}, 348 349 /* OP_OPENATTR = 19 */ 350 {rfs4_op_openattr, nullfree, 0}, 351 352 /* OP_OPEN_CONFIRM = 20 */ 353 {rfs4_op_open_confirm, nullfree, 0}, 354 355 /* OP_OPEN_DOWNGRADE = 21 */ 356 {rfs4_op_open_downgrade, nullfree, 0}, 357 358 /* OP_OPEN_PUTFH = 22 */ 359 {rfs4_op_putfh, nullfree, RPC_ALL}, 360 361 /* OP_PUTPUBFH = 23 */ 362 {rfs4_op_putpubfh, nullfree, RPC_ALL}, 363 364 /* OP_PUTROOTFH = 24 */ 365 {rfs4_op_putrootfh, nullfree, RPC_ALL}, 366 367 /* OP_READ = 25 */ 368 {rfs4_op_read, rfs4_op_read_free, RPC_IDEMPOTENT}, 369 370 /* OP_READDIR = 26 */ 371 {rfs4_op_readdir, rfs4_op_readdir_free, RPC_IDEMPOTENT}, 372 373 /* OP_READLINK = 27 */ 374 {rfs4_op_readlink, rfs4_op_readlink_free, RPC_IDEMPOTENT}, 375 376 /* OP_REMOVE = 28 */ 377 {rfs4_op_remove, nullfree, 0}, 378 379 /* OP_RENAME = 29 */ 380 {rfs4_op_rename, nullfree, 0}, 381 382 /* OP_RENEW = 30 */ 383 {rfs4_op_renew, nullfree, 0}, 384 385 /* OP_RESTOREFH = 31 */ 386 {rfs4_op_restorefh, nullfree, RPC_ALL}, 387 388 /* OP_SAVEFH = 32 */ 389 {rfs4_op_savefh, nullfree, RPC_ALL}, 390 391 /* OP_SECINFO = 33 */ 392 {rfs4_op_secinfo, rfs4_op_secinfo_free, 0}, 393 394 /* OP_SETATTR = 34 */ 395 {rfs4_op_setattr, nullfree, 0}, 396 397 /* OP_SETCLIENTID = 35 */ 398 {rfs4_op_setclientid, nullfree, 0}, 399 400 /* OP_SETCLIENTID_CONFIRM = 36 */ 401 {rfs4_op_setclientid_confirm, nullfree, 0}, 402 403 /* OP_VERIFY = 37 */ 404 {rfs4_op_verify, nullfree, RPC_IDEMPOTENT}, 405 406 /* OP_WRITE = 38 */ 407 {rfs4_op_write, nullfree, 0}, 408 409 /* OP_RELEASE_LOCKOWNER = 39 */ 410 {rfs4_op_release_lockowner, nullfree, 0}, 411 }; 412 413 static uint_t rfsv4disp_cnt = sizeof (rfsv4disptab) / sizeof (rfsv4disptab[0]); 414 415 #define OP_ILLEGAL_IDX (rfsv4disp_cnt) 416 417 #ifdef DEBUG 418 419 int rfs4_fillone_debug = 0; 420 int rfs4_no_stub_access = 1; 421 int rfs4_rddir_debug = 0; 422 423 static char *rfs4_op_string[] = { 424 "rfs4_op_null", 425 "rfs4_op_1 unused", 426 "rfs4_op_2 unused", 427 "rfs4_op_access", 428 "rfs4_op_close", 429 "rfs4_op_commit", 430 "rfs4_op_create", 431 "rfs4_op_delegpurge", 432 "rfs4_op_delegreturn", 433 "rfs4_op_getattr", 434 "rfs4_op_getfh", 435 "rfs4_op_link", 436 "rfs4_op_lock", 437 "rfs4_op_lockt", 438 "rfs4_op_locku", 439 "rfs4_op_lookup", 440 "rfs4_op_lookupp", 441 "rfs4_op_nverify", 442 "rfs4_op_open", 443 "rfs4_op_openattr", 444 "rfs4_op_open_confirm", 445 "rfs4_op_open_downgrade", 446 "rfs4_op_putfh", 447 "rfs4_op_putpubfh", 448 "rfs4_op_putrootfh", 449 "rfs4_op_read", 450 "rfs4_op_readdir", 451 "rfs4_op_readlink", 452 "rfs4_op_remove", 453 "rfs4_op_rename", 454 "rfs4_op_renew", 455 "rfs4_op_restorefh", 456 "rfs4_op_savefh", 457 "rfs4_op_secinfo", 458 "rfs4_op_setattr", 459 "rfs4_op_setclientid", 460 "rfs4_op_setclient_confirm", 461 "rfs4_op_verify", 462 "rfs4_op_write", 463 "rfs4_op_release_lockowner", 464 "rfs4_op_illegal" 465 }; 466 #endif 467 468 void rfs4_ss_chkclid(rfs4_client_t *); 469 470 extern size_t strlcpy(char *dst, const char *src, size_t dstsize); 471 472 extern void rfs4_free_fs_locations4(fs_locations4 *); 473 474 #ifdef nextdp 475 #undef nextdp 476 #endif 477 #define nextdp(dp) ((struct dirent64 *)((char *)(dp) + (dp)->d_reclen)) 478 479 static const fs_operation_def_t nfs4_rd_deleg_tmpl[] = { 480 VOPNAME_OPEN, { .femop_open = deleg_rd_open }, 481 VOPNAME_WRITE, { .femop_write = deleg_rd_write }, 482 VOPNAME_SETATTR, { .femop_setattr = deleg_rd_setattr }, 483 VOPNAME_RWLOCK, { .femop_rwlock = deleg_rd_rwlock }, 484 VOPNAME_SPACE, { .femop_space = deleg_rd_space }, 485 VOPNAME_SETSECATTR, { .femop_setsecattr = deleg_rd_setsecattr }, 486 VOPNAME_VNEVENT, { .femop_vnevent = deleg_rd_vnevent }, 487 NULL, NULL 488 }; 489 static const fs_operation_def_t nfs4_wr_deleg_tmpl[] = { 490 VOPNAME_OPEN, { .femop_open = deleg_wr_open }, 491 VOPNAME_READ, { .femop_read = deleg_wr_read }, 492 VOPNAME_WRITE, { .femop_write = deleg_wr_write }, 493 VOPNAME_SETATTR, { .femop_setattr = deleg_wr_setattr }, 494 VOPNAME_RWLOCK, { .femop_rwlock = deleg_wr_rwlock }, 495 VOPNAME_SPACE, { .femop_space = deleg_wr_space }, 496 VOPNAME_SETSECATTR, { .femop_setsecattr = deleg_wr_setsecattr }, 497 VOPNAME_VNEVENT, { .femop_vnevent = deleg_wr_vnevent }, 498 NULL, NULL 499 }; 500 501 int 502 rfs4_srvrinit(void) 503 { 504 timespec32_t verf; 505 int error; 506 extern void rfs4_attr_init(); 507 extern krwlock_t rfs4_deleg_policy_lock; 508 509 /* 510 * The following algorithm attempts to find a unique verifier 511 * to be used as the write verifier returned from the server 512 * to the client. It is important that this verifier change 513 * whenever the server reboots. Of secondary importance, it 514 * is important for the verifier to be unique between two 515 * different servers. 516 * 517 * Thus, an attempt is made to use the system hostid and the 518 * current time in seconds when the nfssrv kernel module is 519 * loaded. It is assumed that an NFS server will not be able 520 * to boot and then to reboot in less than a second. If the 521 * hostid has not been set, then the current high resolution 522 * time is used. This will ensure different verifiers each 523 * time the server reboots and minimize the chances that two 524 * different servers will have the same verifier. 525 * XXX - this is broken on LP64 kernels. 526 */ 527 verf.tv_sec = (time_t)zone_get_hostid(NULL); 528 if (verf.tv_sec != 0) { 529 verf.tv_nsec = gethrestime_sec(); 530 } else { 531 timespec_t tverf; 532 533 gethrestime(&tverf); 534 verf.tv_sec = (time_t)tverf.tv_sec; 535 verf.tv_nsec = tverf.tv_nsec; 536 } 537 538 Write4verf = *(uint64_t *)&verf; 539 540 rfs4_attr_init(); 541 mutex_init(&rfs4_deleg_lock, NULL, MUTEX_DEFAULT, NULL); 542 543 /* Used to manage create/destroy of server state */ 544 mutex_init(&rfs4_state_lock, NULL, MUTEX_DEFAULT, NULL); 545 546 /* Used to manage access to server instance linked list */ 547 mutex_init(&rfs4_servinst_lock, NULL, MUTEX_DEFAULT, NULL); 548 549 /* Used to manage access to rfs4_deleg_policy */ 550 rw_init(&rfs4_deleg_policy_lock, NULL, RW_DEFAULT, NULL); 551 552 error = fem_create("deleg_rdops", nfs4_rd_deleg_tmpl, &deleg_rdops); 553 if (error != 0) { 554 rfs4_disable_delegation(); 555 } else { 556 error = fem_create("deleg_wrops", nfs4_wr_deleg_tmpl, 557 &deleg_wrops); 558 if (error != 0) { 559 rfs4_disable_delegation(); 560 fem_free(deleg_rdops); 561 } 562 } 563 564 nfs4_srv_caller_id = fs_new_caller_id(); 565 566 lockt_sysid = lm_alloc_sysidt(); 567 568 vsd_create(&nfs4_srv_vkey, NULL); 569 570 return (0); 571 } 572 573 void 574 rfs4_srvrfini(void) 575 { 576 extern krwlock_t rfs4_deleg_policy_lock; 577 578 if (lockt_sysid != LM_NOSYSID) { 579 lm_free_sysidt(lockt_sysid); 580 lockt_sysid = LM_NOSYSID; 581 } 582 583 mutex_destroy(&rfs4_deleg_lock); 584 mutex_destroy(&rfs4_state_lock); 585 rw_destroy(&rfs4_deleg_policy_lock); 586 587 fem_free(deleg_rdops); 588 fem_free(deleg_wrops); 589 } 590 591 void 592 rfs4_init_compound_state(struct compound_state *cs) 593 { 594 bzero(cs, sizeof (*cs)); 595 cs->cont = TRUE; 596 cs->access = CS_ACCESS_DENIED; 597 cs->deleg = FALSE; 598 cs->mandlock = FALSE; 599 cs->fh.nfs_fh4_val = cs->fhbuf; 600 } 601 602 void 603 rfs4_grace_start(rfs4_servinst_t *sip) 604 { 605 rw_enter(&sip->rwlock, RW_WRITER); 606 sip->start_time = (time_t)TICK_TO_SEC(ddi_get_lbolt()); 607 sip->grace_period = rfs4_grace_period; 608 rw_exit(&sip->rwlock); 609 } 610 611 /* 612 * returns true if the instance's grace period has never been started 613 */ 614 int 615 rfs4_servinst_grace_new(rfs4_servinst_t *sip) 616 { 617 time_t start_time; 618 619 rw_enter(&sip->rwlock, RW_READER); 620 start_time = sip->start_time; 621 rw_exit(&sip->rwlock); 622 623 return (start_time == 0); 624 } 625 626 /* 627 * Indicates if server instance is within the 628 * grace period. 629 */ 630 int 631 rfs4_servinst_in_grace(rfs4_servinst_t *sip) 632 { 633 time_t grace_expiry; 634 635 rw_enter(&sip->rwlock, RW_READER); 636 grace_expiry = sip->start_time + sip->grace_period; 637 rw_exit(&sip->rwlock); 638 639 return (((time_t)TICK_TO_SEC(ddi_get_lbolt())) < grace_expiry); 640 } 641 642 int 643 rfs4_clnt_in_grace(rfs4_client_t *cp) 644 { 645 ASSERT(rfs4_dbe_refcnt(cp->rc_dbe) > 0); 646 647 return (rfs4_servinst_in_grace(cp->rc_server_instance)); 648 } 649 650 /* 651 * reset all currently active grace periods 652 */ 653 void 654 rfs4_grace_reset_all(void) 655 { 656 rfs4_servinst_t *sip; 657 658 mutex_enter(&rfs4_servinst_lock); 659 for (sip = rfs4_cur_servinst; sip != NULL; sip = sip->prev) 660 if (rfs4_servinst_in_grace(sip)) 661 rfs4_grace_start(sip); 662 mutex_exit(&rfs4_servinst_lock); 663 } 664 665 /* 666 * start any new instances' grace periods 667 */ 668 void 669 rfs4_grace_start_new(void) 670 { 671 rfs4_servinst_t *sip; 672 673 mutex_enter(&rfs4_servinst_lock); 674 for (sip = rfs4_cur_servinst; sip != NULL; sip = sip->prev) 675 if (rfs4_servinst_grace_new(sip)) 676 rfs4_grace_start(sip); 677 mutex_exit(&rfs4_servinst_lock); 678 } 679 680 static rfs4_dss_path_t * 681 rfs4_dss_newpath(rfs4_servinst_t *sip, char *path, unsigned index) 682 { 683 size_t len; 684 rfs4_dss_path_t *dss_path; 685 686 dss_path = kmem_alloc(sizeof (rfs4_dss_path_t), KM_SLEEP); 687 688 /* 689 * Take a copy of the string, since the original may be overwritten. 690 * Sadly, no strdup() in the kernel. 691 */ 692 /* allow for NUL */ 693 len = strlen(path) + 1; 694 dss_path->path = kmem_alloc(len, KM_SLEEP); 695 (void) strlcpy(dss_path->path, path, len); 696 697 /* associate with servinst */ 698 dss_path->sip = sip; 699 dss_path->index = index; 700 701 /* 702 * Add to list of served paths. 703 * No locking required, as we're only ever called at startup. 704 */ 705 if (rfs4_dss_pathlist == NULL) { 706 /* this is the first dss_path_t */ 707 708 /* needed for insque/remque */ 709 dss_path->next = dss_path->prev = dss_path; 710 711 rfs4_dss_pathlist = dss_path; 712 } else { 713 insque(dss_path, rfs4_dss_pathlist); 714 } 715 716 return (dss_path); 717 } 718 719 /* 720 * Create a new server instance, and make it the currently active instance. 721 * Note that starting the grace period too early will reduce the clients' 722 * recovery window. 723 */ 724 void 725 rfs4_servinst_create(int start_grace, int dss_npaths, char **dss_paths) 726 { 727 unsigned i; 728 rfs4_servinst_t *sip; 729 rfs4_oldstate_t *oldstate; 730 731 sip = kmem_alloc(sizeof (rfs4_servinst_t), KM_SLEEP); 732 rw_init(&sip->rwlock, NULL, RW_DEFAULT, NULL); 733 734 sip->start_time = (time_t)0; 735 sip->grace_period = (time_t)0; 736 sip->next = NULL; 737 sip->prev = NULL; 738 739 rw_init(&sip->oldstate_lock, NULL, RW_DEFAULT, NULL); 740 /* 741 * This initial dummy entry is required to setup for insque/remque. 742 * It must be skipped over whenever the list is traversed. 743 */ 744 oldstate = kmem_alloc(sizeof (rfs4_oldstate_t), KM_SLEEP); 745 /* insque/remque require initial list entry to be self-terminated */ 746 oldstate->next = oldstate; 747 oldstate->prev = oldstate; 748 sip->oldstate = oldstate; 749 750 751 sip->dss_npaths = dss_npaths; 752 sip->dss_paths = kmem_alloc(dss_npaths * 753 sizeof (rfs4_dss_path_t *), KM_SLEEP); 754 755 for (i = 0; i < dss_npaths; i++) { 756 sip->dss_paths[i] = rfs4_dss_newpath(sip, dss_paths[i], i); 757 } 758 759 mutex_enter(&rfs4_servinst_lock); 760 if (rfs4_cur_servinst != NULL) { 761 /* add to linked list */ 762 sip->prev = rfs4_cur_servinst; 763 rfs4_cur_servinst->next = sip; 764 } 765 if (start_grace) 766 rfs4_grace_start(sip); 767 /* make the new instance "current" */ 768 rfs4_cur_servinst = sip; 769 770 mutex_exit(&rfs4_servinst_lock); 771 } 772 773 /* 774 * In future, we might add a rfs4_servinst_destroy(sip) but, for now, destroy 775 * all instances directly. 776 */ 777 void 778 rfs4_servinst_destroy_all(void) 779 { 780 rfs4_servinst_t *sip, *prev, *current; 781 #ifdef DEBUG 782 int n = 0; 783 #endif 784 785 mutex_enter(&rfs4_servinst_lock); 786 ASSERT(rfs4_cur_servinst != NULL); 787 current = rfs4_cur_servinst; 788 rfs4_cur_servinst = NULL; 789 for (sip = current; sip != NULL; sip = prev) { 790 prev = sip->prev; 791 rw_destroy(&sip->rwlock); 792 if (sip->oldstate) 793 kmem_free(sip->oldstate, sizeof (rfs4_oldstate_t)); 794 if (sip->dss_paths) 795 kmem_free(sip->dss_paths, 796 sip->dss_npaths * sizeof (rfs4_dss_path_t *)); 797 kmem_free(sip, sizeof (rfs4_servinst_t)); 798 #ifdef DEBUG 799 n++; 800 #endif 801 } 802 mutex_exit(&rfs4_servinst_lock); 803 } 804 805 /* 806 * Assign the current server instance to a client_t. 807 * Should be called with cp->rc_dbe held. 808 */ 809 void 810 rfs4_servinst_assign(rfs4_client_t *cp, rfs4_servinst_t *sip) 811 { 812 ASSERT(rfs4_dbe_refcnt(cp->rc_dbe) > 0); 813 814 /* 815 * The lock ensures that if the current instance is in the process 816 * of changing, we will see the new one. 817 */ 818 mutex_enter(&rfs4_servinst_lock); 819 cp->rc_server_instance = sip; 820 mutex_exit(&rfs4_servinst_lock); 821 } 822 823 rfs4_servinst_t * 824 rfs4_servinst(rfs4_client_t *cp) 825 { 826 ASSERT(rfs4_dbe_refcnt(cp->rc_dbe) > 0); 827 828 return (cp->rc_server_instance); 829 } 830 831 /* ARGSUSED */ 832 static void 833 nullfree(caddr_t resop) 834 { 835 } 836 837 /* 838 * This is a fall-through for invalid or not implemented (yet) ops 839 */ 840 /* ARGSUSED */ 841 static void 842 rfs4_op_inval(nfs_argop4 *argop, nfs_resop4 *resop, struct svc_req *req, 843 struct compound_state *cs) 844 { 845 *cs->statusp = *((nfsstat4 *)&(resop)->nfs_resop4_u) = NFS4ERR_INVAL; 846 } 847 848 /* 849 * Check if the security flavor, nfsnum, is in the flavor_list. 850 */ 851 bool_t 852 in_flavor_list(int nfsnum, int *flavor_list, int count) 853 { 854 int i; 855 856 for (i = 0; i < count; i++) { 857 if (nfsnum == flavor_list[i]) 858 return (TRUE); 859 } 860 return (FALSE); 861 } 862 863 /* 864 * Used by rfs4_op_secinfo to get the security information from the 865 * export structure associated with the component. 866 */ 867 /* ARGSUSED */ 868 static nfsstat4 869 do_rfs4_op_secinfo(struct compound_state *cs, char *nm, SECINFO4res *resp) 870 { 871 int error, different_export = 0; 872 vnode_t *dvp, *vp, *tvp; 873 struct exportinfo *exi = NULL; 874 fid_t fid; 875 uint_t count, i; 876 secinfo4 *resok_val; 877 struct secinfo *secp; 878 seconfig_t *si; 879 bool_t did_traverse = FALSE; 880 int dotdot, walk; 881 882 dvp = cs->vp; 883 dotdot = (nm[0] == '.' && nm[1] == '.' && nm[2] == '\0'); 884 885 /* 886 * If dotdotting, then need to check whether it's above the 887 * root of a filesystem, or above an export point. 888 */ 889 if (dotdot) { 890 891 /* 892 * If dotdotting at the root of a filesystem, then 893 * need to traverse back to the mounted-on filesystem 894 * and do the dotdot lookup there. 895 */ 896 if (cs->vp->v_flag & VROOT) { 897 898 /* 899 * If at the system root, then can 900 * go up no further. 901 */ 902 if (VN_CMP(dvp, rootdir)) 903 return (puterrno4(ENOENT)); 904 905 /* 906 * Traverse back to the mounted-on filesystem 907 */ 908 dvp = untraverse(cs->vp); 909 910 /* 911 * Set the different_export flag so we remember 912 * to pick up a new exportinfo entry for 913 * this new filesystem. 914 */ 915 different_export = 1; 916 } else { 917 918 /* 919 * If dotdotting above an export point then set 920 * the different_export to get new export info. 921 */ 922 different_export = nfs_exported(cs->exi, cs->vp); 923 } 924 } 925 926 /* 927 * Get the vnode for the component "nm". 928 */ 929 error = VOP_LOOKUP(dvp, nm, &vp, NULL, 0, NULL, cs->cr, 930 NULL, NULL, NULL); 931 if (error) 932 return (puterrno4(error)); 933 934 /* 935 * If the vnode is in a pseudo filesystem, or if the security flavor 936 * used in the request is valid but not an explicitly shared flavor, 937 * or the access bit indicates that this is a limited access, 938 * check whether this vnode is visible. 939 */ 940 if (!different_export && 941 (PSEUDO(cs->exi) || ! is_exported_sec(cs->nfsflavor, cs->exi) || 942 cs->access & CS_ACCESS_LIMITED)) { 943 if (! nfs_visible(cs->exi, vp, &different_export)) { 944 VN_RELE(vp); 945 return (puterrno4(ENOENT)); 946 } 947 } 948 949 /* 950 * If it's a mountpoint, then traverse it. 951 */ 952 if (vn_ismntpt(vp)) { 953 tvp = vp; 954 if ((error = traverse(&tvp)) != 0) { 955 VN_RELE(vp); 956 return (puterrno4(error)); 957 } 958 /* remember that we had to traverse mountpoint */ 959 did_traverse = TRUE; 960 vp = tvp; 961 different_export = 1; 962 } else if (vp->v_vfsp != dvp->v_vfsp) { 963 /* 964 * If vp isn't a mountpoint and the vfs ptrs aren't the same, 965 * then vp is probably an LOFS object. We don't need the 966 * realvp, we just need to know that we might have crossed 967 * a server fs boundary and need to call checkexport4. 968 * (LOFS lookup hides server fs mountpoints, and actually calls 969 * traverse) 970 */ 971 different_export = 1; 972 } 973 974 /* 975 * Get the export information for it. 976 */ 977 if (different_export) { 978 979 bzero(&fid, sizeof (fid)); 980 fid.fid_len = MAXFIDSZ; 981 error = vop_fid_pseudo(vp, &fid); 982 if (error) { 983 VN_RELE(vp); 984 return (puterrno4(error)); 985 } 986 987 if (dotdot) 988 exi = nfs_vptoexi(NULL, vp, cs->cr, &walk, NULL, TRUE); 989 else 990 exi = checkexport4(&vp->v_vfsp->vfs_fsid, &fid, vp); 991 992 if (exi == NULL) { 993 if (did_traverse == TRUE) { 994 /* 995 * If this vnode is a mounted-on vnode, 996 * but the mounted-on file system is not 997 * exported, send back the secinfo for 998 * the exported node that the mounted-on 999 * vnode lives in. 1000 */ 1001 exi = cs->exi; 1002 } else { 1003 VN_RELE(vp); 1004 return (puterrno4(EACCES)); 1005 } 1006 } 1007 } else { 1008 exi = cs->exi; 1009 } 1010 ASSERT(exi != NULL); 1011 1012 1013 /* 1014 * Create the secinfo result based on the security information 1015 * from the exportinfo structure (exi). 1016 * 1017 * Return all flavors for a pseudo node. 1018 * For a real export node, return the flavor that the client 1019 * has access with. 1020 */ 1021 ASSERT(RW_LOCK_HELD(&exported_lock)); 1022 if (PSEUDO(exi)) { 1023 count = exi->exi_export.ex_seccnt; /* total sec count */ 1024 resok_val = kmem_alloc(count * sizeof (secinfo4), KM_SLEEP); 1025 secp = exi->exi_export.ex_secinfo; 1026 1027 for (i = 0; i < count; i++) { 1028 si = &secp[i].s_secinfo; 1029 resok_val[i].flavor = si->sc_rpcnum; 1030 if (resok_val[i].flavor == RPCSEC_GSS) { 1031 rpcsec_gss_info *info; 1032 1033 info = &resok_val[i].flavor_info; 1034 info->qop = si->sc_qop; 1035 info->service = (rpc_gss_svc_t)si->sc_service; 1036 1037 /* get oid opaque data */ 1038 info->oid.sec_oid4_len = 1039 si->sc_gss_mech_type->length; 1040 info->oid.sec_oid4_val = kmem_alloc( 1041 si->sc_gss_mech_type->length, KM_SLEEP); 1042 bcopy( 1043 si->sc_gss_mech_type->elements, 1044 info->oid.sec_oid4_val, 1045 info->oid.sec_oid4_len); 1046 } 1047 } 1048 resp->SECINFO4resok_len = count; 1049 resp->SECINFO4resok_val = resok_val; 1050 } else { 1051 int ret_cnt = 0, k = 0; 1052 int *flavor_list; 1053 1054 count = exi->exi_export.ex_seccnt; /* total sec count */ 1055 secp = exi->exi_export.ex_secinfo; 1056 1057 flavor_list = kmem_alloc(count * sizeof (int), KM_SLEEP); 1058 /* find out which flavors to return */ 1059 for (i = 0; i < count; i ++) { 1060 int access, flavor, perm; 1061 1062 flavor = secp[i].s_secinfo.sc_nfsnum; 1063 perm = secp[i].s_flags; 1064 1065 access = nfsauth4_secinfo_access(exi, cs->req, 1066 flavor, perm); 1067 1068 if (! (access & NFSAUTH_DENIED) && 1069 ! (access & NFSAUTH_WRONGSEC)) { 1070 flavor_list[ret_cnt] = flavor; 1071 ret_cnt++; 1072 } 1073 } 1074 1075 /* Create the returning SECINFO value */ 1076 resok_val = kmem_alloc(ret_cnt * sizeof (secinfo4), KM_SLEEP); 1077 1078 for (i = 0; i < count; i++) { 1079 /* 1080 * If the flavor is in the flavor list, 1081 * fill in resok_val. 1082 */ 1083 si = &secp[i].s_secinfo; 1084 if (in_flavor_list(si->sc_nfsnum, 1085 flavor_list, ret_cnt)) { 1086 resok_val[k].flavor = si->sc_rpcnum; 1087 if (resok_val[k].flavor == RPCSEC_GSS) { 1088 rpcsec_gss_info *info; 1089 1090 info = &resok_val[k].flavor_info; 1091 info->qop = si->sc_qop; 1092 info->service = (rpc_gss_svc_t) 1093 si->sc_service; 1094 1095 /* get oid opaque data */ 1096 info->oid.sec_oid4_len = 1097 si->sc_gss_mech_type->length; 1098 info->oid.sec_oid4_val = kmem_alloc( 1099 si->sc_gss_mech_type->length, 1100 KM_SLEEP); 1101 bcopy(si->sc_gss_mech_type->elements, 1102 info->oid.sec_oid4_val, 1103 info->oid.sec_oid4_len); 1104 } 1105 k++; 1106 } 1107 if (k >= ret_cnt) 1108 break; 1109 } 1110 resp->SECINFO4resok_len = ret_cnt; 1111 resp->SECINFO4resok_val = resok_val; 1112 kmem_free(flavor_list, count * sizeof (int)); 1113 } 1114 1115 VN_RELE(vp); 1116 return (NFS4_OK); 1117 } 1118 1119 /* 1120 * SECINFO (Operation 33): Obtain required security information on 1121 * the component name in the format of (security-mechanism-oid, qop, service) 1122 * triplets. 1123 */ 1124 /* ARGSUSED */ 1125 static void 1126 rfs4_op_secinfo(nfs_argop4 *argop, nfs_resop4 *resop, struct svc_req *req, 1127 struct compound_state *cs) 1128 { 1129 SECINFO4args *args = &argop->nfs_argop4_u.opsecinfo; 1130 SECINFO4res *resp = &resop->nfs_resop4_u.opsecinfo; 1131 utf8string *utfnm = &args->name; 1132 uint_t len; 1133 char *nm; 1134 struct sockaddr *ca; 1135 char *name = NULL; 1136 nfsstat4 status = NFS4_OK; 1137 1138 DTRACE_NFSV4_2(op__secinfo__start, struct compound_state *, cs, 1139 SECINFO4args *, args); 1140 1141 /* 1142 * Current file handle (cfh) should have been set before getting 1143 * into this function. If not, return error. 1144 */ 1145 if (cs->vp == NULL) { 1146 *cs->statusp = resp->status = NFS4ERR_NOFILEHANDLE; 1147 goto out; 1148 } 1149 1150 if (cs->vp->v_type != VDIR) { 1151 *cs->statusp = resp->status = NFS4ERR_NOTDIR; 1152 goto out; 1153 } 1154 1155 /* 1156 * Verify the component name. If failed, error out, but 1157 * do not error out if the component name is a "..". 1158 * SECINFO will return its parents secinfo data for SECINFO "..". 1159 */ 1160 status = utf8_dir_verify(utfnm); 1161 if (status != NFS4_OK) { 1162 if (utfnm->utf8string_len != 2 || 1163 utfnm->utf8string_val[0] != '.' || 1164 utfnm->utf8string_val[1] != '.') { 1165 *cs->statusp = resp->status = status; 1166 goto out; 1167 } 1168 } 1169 1170 nm = utf8_to_str(utfnm, &len, NULL); 1171 if (nm == NULL) { 1172 *cs->statusp = resp->status = NFS4ERR_INVAL; 1173 goto out; 1174 } 1175 1176 if (len > MAXNAMELEN) { 1177 *cs->statusp = resp->status = NFS4ERR_NAMETOOLONG; 1178 kmem_free(nm, len); 1179 goto out; 1180 } 1181 1182 ca = (struct sockaddr *)svc_getrpccaller(req->rq_xprt)->buf; 1183 name = nfscmd_convname(ca, cs->exi, nm, NFSCMD_CONV_INBOUND, 1184 MAXPATHLEN + 1); 1185 1186 if (name == NULL) { 1187 *cs->statusp = resp->status = NFS4ERR_INVAL; 1188 kmem_free(nm, len); 1189 goto out; 1190 } 1191 1192 1193 *cs->statusp = resp->status = do_rfs4_op_secinfo(cs, name, resp); 1194 1195 if (name != nm) 1196 kmem_free(name, MAXPATHLEN + 1); 1197 kmem_free(nm, len); 1198 1199 out: 1200 DTRACE_NFSV4_2(op__secinfo__done, struct compound_state *, cs, 1201 SECINFO4res *, resp); 1202 } 1203 1204 /* 1205 * Free SECINFO result. 1206 */ 1207 /* ARGSUSED */ 1208 static void 1209 rfs4_op_secinfo_free(nfs_resop4 *resop) 1210 { 1211 SECINFO4res *resp = &resop->nfs_resop4_u.opsecinfo; 1212 int count, i; 1213 secinfo4 *resok_val; 1214 1215 /* If this is not an Ok result, nothing to free. */ 1216 if (resp->status != NFS4_OK) { 1217 return; 1218 } 1219 1220 count = resp->SECINFO4resok_len; 1221 resok_val = resp->SECINFO4resok_val; 1222 1223 for (i = 0; i < count; i++) { 1224 if (resok_val[i].flavor == RPCSEC_GSS) { 1225 rpcsec_gss_info *info; 1226 1227 info = &resok_val[i].flavor_info; 1228 kmem_free(info->oid.sec_oid4_val, 1229 info->oid.sec_oid4_len); 1230 } 1231 } 1232 kmem_free(resok_val, count * sizeof (secinfo4)); 1233 resp->SECINFO4resok_len = 0; 1234 resp->SECINFO4resok_val = NULL; 1235 } 1236 1237 /* ARGSUSED */ 1238 static void 1239 rfs4_op_access(nfs_argop4 *argop, nfs_resop4 *resop, struct svc_req *req, 1240 struct compound_state *cs) 1241 { 1242 ACCESS4args *args = &argop->nfs_argop4_u.opaccess; 1243 ACCESS4res *resp = &resop->nfs_resop4_u.opaccess; 1244 int error; 1245 vnode_t *vp; 1246 struct vattr va; 1247 int checkwriteperm; 1248 cred_t *cr = cs->cr; 1249 bslabel_t *clabel, *slabel; 1250 ts_label_t *tslabel; 1251 boolean_t admin_low_client; 1252 1253 DTRACE_NFSV4_2(op__access__start, struct compound_state *, cs, 1254 ACCESS4args *, args); 1255 1256 #if 0 /* XXX allow access even if !cs->access. Eventually only pseudo fs */ 1257 if (cs->access == CS_ACCESS_DENIED) { 1258 *cs->statusp = resp->status = NFS4ERR_ACCESS; 1259 goto out; 1260 } 1261 #endif 1262 if (cs->vp == NULL) { 1263 *cs->statusp = resp->status = NFS4ERR_NOFILEHANDLE; 1264 goto out; 1265 } 1266 1267 ASSERT(cr != NULL); 1268 1269 vp = cs->vp; 1270 1271 /* 1272 * If the file system is exported read only, it is not appropriate 1273 * to check write permissions for regular files and directories. 1274 * Special files are interpreted by the client, so the underlying 1275 * permissions are sent back to the client for interpretation. 1276 */ 1277 if (rdonly4(cs->exi, cs->vp, req) && 1278 (vp->v_type == VREG || vp->v_type == VDIR)) 1279 checkwriteperm = 0; 1280 else 1281 checkwriteperm = 1; 1282 1283 /* 1284 * XXX 1285 * We need the mode so that we can correctly determine access 1286 * permissions relative to a mandatory lock file. Access to 1287 * mandatory lock files is denied on the server, so it might 1288 * as well be reflected to the server during the open. 1289 */ 1290 va.va_mask = AT_MODE; 1291 error = VOP_GETATTR(vp, &va, 0, cr, NULL); 1292 if (error) { 1293 *cs->statusp = resp->status = puterrno4(error); 1294 goto out; 1295 } 1296 resp->access = 0; 1297 resp->supported = 0; 1298 1299 if (is_system_labeled()) { 1300 ASSERT(req->rq_label != NULL); 1301 clabel = req->rq_label; 1302 DTRACE_PROBE2(tx__rfs4__log__info__opaccess__clabel, char *, 1303 "got client label from request(1)", 1304 struct svc_req *, req); 1305 if (!blequal(&l_admin_low->tsl_label, clabel)) { 1306 if ((tslabel = nfs_getflabel(vp, cs->exi)) == NULL) { 1307 *cs->statusp = resp->status = puterrno4(EACCES); 1308 goto out; 1309 } 1310 slabel = label2bslabel(tslabel); 1311 DTRACE_PROBE3(tx__rfs4__log__info__opaccess__slabel, 1312 char *, "got server label(1) for vp(2)", 1313 bslabel_t *, slabel, vnode_t *, vp); 1314 1315 admin_low_client = B_FALSE; 1316 } else 1317 admin_low_client = B_TRUE; 1318 } 1319 1320 if (args->access & ACCESS4_READ) { 1321 error = VOP_ACCESS(vp, VREAD, 0, cr, NULL); 1322 if (!error && !MANDLOCK(vp, va.va_mode) && 1323 (!is_system_labeled() || admin_low_client || 1324 bldominates(clabel, slabel))) 1325 resp->access |= ACCESS4_READ; 1326 resp->supported |= ACCESS4_READ; 1327 } 1328 if ((args->access & ACCESS4_LOOKUP) && vp->v_type == VDIR) { 1329 error = VOP_ACCESS(vp, VEXEC, 0, cr, NULL); 1330 if (!error && (!is_system_labeled() || admin_low_client || 1331 bldominates(clabel, slabel))) 1332 resp->access |= ACCESS4_LOOKUP; 1333 resp->supported |= ACCESS4_LOOKUP; 1334 } 1335 if (checkwriteperm && 1336 (args->access & (ACCESS4_MODIFY|ACCESS4_EXTEND))) { 1337 error = VOP_ACCESS(vp, VWRITE, 0, cr, NULL); 1338 if (!error && !MANDLOCK(vp, va.va_mode) && 1339 (!is_system_labeled() || admin_low_client || 1340 blequal(clabel, slabel))) 1341 resp->access |= 1342 (args->access & (ACCESS4_MODIFY | ACCESS4_EXTEND)); 1343 resp->supported |= 1344 resp->access & (ACCESS4_MODIFY | ACCESS4_EXTEND); 1345 } 1346 1347 if (checkwriteperm && 1348 (args->access & ACCESS4_DELETE) && vp->v_type == VDIR) { 1349 error = VOP_ACCESS(vp, VWRITE, 0, cr, NULL); 1350 if (!error && (!is_system_labeled() || admin_low_client || 1351 blequal(clabel, slabel))) 1352 resp->access |= ACCESS4_DELETE; 1353 resp->supported |= ACCESS4_DELETE; 1354 } 1355 if (args->access & ACCESS4_EXECUTE && vp->v_type != VDIR) { 1356 error = VOP_ACCESS(vp, VEXEC, 0, cr, NULL); 1357 if (!error && !MANDLOCK(vp, va.va_mode) && 1358 (!is_system_labeled() || admin_low_client || 1359 bldominates(clabel, slabel))) 1360 resp->access |= ACCESS4_EXECUTE; 1361 resp->supported |= ACCESS4_EXECUTE; 1362 } 1363 1364 if (is_system_labeled() && !admin_low_client) 1365 label_rele(tslabel); 1366 1367 *cs->statusp = resp->status = NFS4_OK; 1368 out: 1369 DTRACE_NFSV4_2(op__access__done, struct compound_state *, cs, 1370 ACCESS4res *, resp); 1371 } 1372 1373 /* ARGSUSED */ 1374 static void 1375 rfs4_op_commit(nfs_argop4 *argop, nfs_resop4 *resop, struct svc_req *req, 1376 struct compound_state *cs) 1377 { 1378 COMMIT4args *args = &argop->nfs_argop4_u.opcommit; 1379 COMMIT4res *resp = &resop->nfs_resop4_u.opcommit; 1380 int error; 1381 vnode_t *vp = cs->vp; 1382 cred_t *cr = cs->cr; 1383 vattr_t va; 1384 1385 DTRACE_NFSV4_2(op__commit__start, struct compound_state *, cs, 1386 COMMIT4args *, args); 1387 1388 if (vp == NULL) { 1389 *cs->statusp = resp->status = NFS4ERR_NOFILEHANDLE; 1390 goto out; 1391 } 1392 if (cs->access == CS_ACCESS_DENIED) { 1393 *cs->statusp = resp->status = NFS4ERR_ACCESS; 1394 goto out; 1395 } 1396 1397 if (args->offset + args->count < args->offset) { 1398 *cs->statusp = resp->status = NFS4ERR_INVAL; 1399 goto out; 1400 } 1401 1402 va.va_mask = AT_UID; 1403 error = VOP_GETATTR(vp, &va, 0, cr, NULL); 1404 1405 /* 1406 * If we can't get the attributes, then we can't do the 1407 * right access checking. So, we'll fail the request. 1408 */ 1409 if (error) { 1410 *cs->statusp = resp->status = puterrno4(error); 1411 goto out; 1412 } 1413 if (rdonly4(cs->exi, cs->vp, req)) { 1414 *cs->statusp = resp->status = NFS4ERR_ROFS; 1415 goto out; 1416 } 1417 1418 if (vp->v_type != VREG) { 1419 if (vp->v_type == VDIR) 1420 resp->status = NFS4ERR_ISDIR; 1421 else 1422 resp->status = NFS4ERR_INVAL; 1423 *cs->statusp = resp->status; 1424 goto out; 1425 } 1426 1427 if (crgetuid(cr) != va.va_uid && 1428 (error = VOP_ACCESS(vp, VWRITE, 0, cs->cr, NULL))) { 1429 *cs->statusp = resp->status = puterrno4(error); 1430 goto out; 1431 } 1432 1433 error = VOP_FSYNC(vp, FSYNC, cr, NULL); 1434 1435 if (error) { 1436 *cs->statusp = resp->status = puterrno4(error); 1437 goto out; 1438 } 1439 1440 *cs->statusp = resp->status = NFS4_OK; 1441 resp->writeverf = Write4verf; 1442 out: 1443 DTRACE_NFSV4_2(op__commit__done, struct compound_state *, cs, 1444 COMMIT4res *, resp); 1445 } 1446 1447 /* 1448 * rfs4_op_mknod is called from rfs4_op_create after all initial verification 1449 * was completed. It does the nfsv4 create for special files. 1450 */ 1451 /* ARGSUSED */ 1452 static vnode_t * 1453 do_rfs4_op_mknod(CREATE4args *args, CREATE4res *resp, struct svc_req *req, 1454 struct compound_state *cs, vattr_t *vap, char *nm) 1455 { 1456 int error; 1457 cred_t *cr = cs->cr; 1458 vnode_t *dvp = cs->vp; 1459 vnode_t *vp = NULL; 1460 int mode; 1461 enum vcexcl excl; 1462 1463 switch (args->type) { 1464 case NF4CHR: 1465 case NF4BLK: 1466 if (secpolicy_sys_devices(cr) != 0) { 1467 *cs->statusp = resp->status = NFS4ERR_PERM; 1468 return (NULL); 1469 } 1470 if (args->type == NF4CHR) 1471 vap->va_type = VCHR; 1472 else 1473 vap->va_type = VBLK; 1474 vap->va_rdev = makedevice(args->ftype4_u.devdata.specdata1, 1475 args->ftype4_u.devdata.specdata2); 1476 vap->va_mask |= AT_RDEV; 1477 break; 1478 case NF4SOCK: 1479 vap->va_type = VSOCK; 1480 break; 1481 case NF4FIFO: 1482 vap->va_type = VFIFO; 1483 break; 1484 default: 1485 *cs->statusp = resp->status = NFS4ERR_BADTYPE; 1486 return (NULL); 1487 } 1488 1489 /* 1490 * Must specify the mode. 1491 */ 1492 if (!(vap->va_mask & AT_MODE)) { 1493 *cs->statusp = resp->status = NFS4ERR_INVAL; 1494 return (NULL); 1495 } 1496 1497 excl = EXCL; 1498 1499 mode = 0; 1500 1501 error = VOP_CREATE(dvp, nm, vap, excl, mode, &vp, cr, 0, NULL, NULL); 1502 if (error) { 1503 *cs->statusp = resp->status = puterrno4(error); 1504 return (NULL); 1505 } 1506 return (vp); 1507 } 1508 1509 /* 1510 * nfsv4 create is used to create non-regular files. For regular files, 1511 * use nfsv4 open. 1512 */ 1513 /* ARGSUSED */ 1514 static void 1515 rfs4_op_create(nfs_argop4 *argop, nfs_resop4 *resop, struct svc_req *req, 1516 struct compound_state *cs) 1517 { 1518 CREATE4args *args = &argop->nfs_argop4_u.opcreate; 1519 CREATE4res *resp = &resop->nfs_resop4_u.opcreate; 1520 int error; 1521 struct vattr bva, iva, iva2, ava, *vap; 1522 cred_t *cr = cs->cr; 1523 vnode_t *dvp = cs->vp; 1524 vnode_t *vp = NULL; 1525 vnode_t *realvp; 1526 char *nm, *lnm; 1527 uint_t len, llen; 1528 int syncval = 0; 1529 struct nfs4_svgetit_arg sarg; 1530 struct nfs4_ntov_table ntov; 1531 struct statvfs64 sb; 1532 nfsstat4 status; 1533 struct sockaddr *ca; 1534 char *name = NULL; 1535 char *lname = NULL; 1536 1537 DTRACE_NFSV4_2(op__create__start, struct compound_state *, cs, 1538 CREATE4args *, args); 1539 1540 resp->attrset = 0; 1541 1542 if (dvp == NULL) { 1543 *cs->statusp = resp->status = NFS4ERR_NOFILEHANDLE; 1544 goto out; 1545 } 1546 1547 /* 1548 * If there is an unshared filesystem mounted on this vnode, 1549 * do not allow to create an object in this directory. 1550 */ 1551 if (vn_ismntpt(dvp)) { 1552 *cs->statusp = resp->status = NFS4ERR_ACCESS; 1553 goto out; 1554 } 1555 1556 /* Verify that type is correct */ 1557 switch (args->type) { 1558 case NF4LNK: 1559 case NF4BLK: 1560 case NF4CHR: 1561 case NF4SOCK: 1562 case NF4FIFO: 1563 case NF4DIR: 1564 break; 1565 default: 1566 *cs->statusp = resp->status = NFS4ERR_BADTYPE; 1567 goto out; 1568 }; 1569 1570 if (cs->access == CS_ACCESS_DENIED) { 1571 *cs->statusp = resp->status = NFS4ERR_ACCESS; 1572 goto out; 1573 } 1574 if (dvp->v_type != VDIR) { 1575 *cs->statusp = resp->status = NFS4ERR_NOTDIR; 1576 goto out; 1577 } 1578 status = utf8_dir_verify(&args->objname); 1579 if (status != NFS4_OK) { 1580 *cs->statusp = resp->status = status; 1581 goto out; 1582 } 1583 1584 if (rdonly4(cs->exi, cs->vp, req)) { 1585 *cs->statusp = resp->status = NFS4ERR_ROFS; 1586 goto out; 1587 } 1588 1589 /* 1590 * Name of newly created object 1591 */ 1592 nm = utf8_to_fn(&args->objname, &len, NULL); 1593 if (nm == NULL) { 1594 *cs->statusp = resp->status = NFS4ERR_INVAL; 1595 goto out; 1596 } 1597 1598 if (len > MAXNAMELEN) { 1599 *cs->statusp = resp->status = NFS4ERR_NAMETOOLONG; 1600 kmem_free(nm, len); 1601 goto out; 1602 } 1603 1604 ca = (struct sockaddr *)svc_getrpccaller(req->rq_xprt)->buf; 1605 name = nfscmd_convname(ca, cs->exi, nm, NFSCMD_CONV_INBOUND, 1606 MAXPATHLEN + 1); 1607 1608 if (name == NULL) { 1609 *cs->statusp = resp->status = NFS4ERR_INVAL; 1610 kmem_free(nm, len); 1611 goto out; 1612 } 1613 1614 resp->attrset = 0; 1615 1616 sarg.sbp = &sb; 1617 sarg.is_referral = B_FALSE; 1618 nfs4_ntov_table_init(&ntov); 1619 1620 status = do_rfs4_set_attrs(&resp->attrset, 1621 &args->createattrs, cs, &sarg, &ntov, NFS4ATTR_SETIT); 1622 1623 if (sarg.vap->va_mask == 0 && status == NFS4_OK) 1624 status = NFS4ERR_INVAL; 1625 1626 if (status != NFS4_OK) { 1627 *cs->statusp = resp->status = status; 1628 if (name != nm) 1629 kmem_free(name, MAXPATHLEN + 1); 1630 kmem_free(nm, len); 1631 nfs4_ntov_table_free(&ntov, &sarg); 1632 resp->attrset = 0; 1633 goto out; 1634 } 1635 1636 /* Get "before" change value */ 1637 bva.va_mask = AT_CTIME|AT_SEQ|AT_MODE; 1638 error = VOP_GETATTR(dvp, &bva, 0, cr, NULL); 1639 if (error) { 1640 *cs->statusp = resp->status = puterrno4(error); 1641 if (name != nm) 1642 kmem_free(name, MAXPATHLEN + 1); 1643 kmem_free(nm, len); 1644 nfs4_ntov_table_free(&ntov, &sarg); 1645 resp->attrset = 0; 1646 goto out; 1647 } 1648 NFS4_SET_FATTR4_CHANGE(resp->cinfo.before, bva.va_ctime) 1649 1650 vap = sarg.vap; 1651 1652 /* 1653 * Set the default initial values for attributes when the parent 1654 * directory does not have the VSUID/VSGID bit set and they have 1655 * not been specified in createattrs. 1656 */ 1657 if (!(bva.va_mode & VSUID) && (vap->va_mask & AT_UID) == 0) { 1658 vap->va_uid = crgetuid(cr); 1659 vap->va_mask |= AT_UID; 1660 } 1661 if (!(bva.va_mode & VSGID) && (vap->va_mask & AT_GID) == 0) { 1662 vap->va_gid = crgetgid(cr); 1663 vap->va_mask |= AT_GID; 1664 } 1665 1666 vap->va_mask |= AT_TYPE; 1667 switch (args->type) { 1668 case NF4DIR: 1669 vap->va_type = VDIR; 1670 if ((vap->va_mask & AT_MODE) == 0) { 1671 vap->va_mode = 0700; /* default: owner rwx only */ 1672 vap->va_mask |= AT_MODE; 1673 } 1674 error = VOP_MKDIR(dvp, name, vap, &vp, cr, NULL, 0, NULL); 1675 if (error) 1676 break; 1677 1678 /* 1679 * Get the initial "after" sequence number, if it fails, 1680 * set to zero 1681 */ 1682 iva.va_mask = AT_SEQ; 1683 if (VOP_GETATTR(dvp, &iva, 0, cs->cr, NULL)) 1684 iva.va_seq = 0; 1685 break; 1686 case NF4LNK: 1687 vap->va_type = VLNK; 1688 if ((vap->va_mask & AT_MODE) == 0) { 1689 vap->va_mode = 0700; /* default: owner rwx only */ 1690 vap->va_mask |= AT_MODE; 1691 } 1692 1693 /* 1694 * symlink names must be treated as data 1695 */ 1696 lnm = utf8_to_str(&args->ftype4_u.linkdata, &llen, NULL); 1697 1698 if (lnm == NULL) { 1699 *cs->statusp = resp->status = NFS4ERR_INVAL; 1700 if (name != nm) 1701 kmem_free(name, MAXPATHLEN + 1); 1702 kmem_free(nm, len); 1703 nfs4_ntov_table_free(&ntov, &sarg); 1704 resp->attrset = 0; 1705 goto out; 1706 } 1707 1708 if (llen > MAXPATHLEN) { 1709 *cs->statusp = resp->status = NFS4ERR_NAMETOOLONG; 1710 if (name != nm) 1711 kmem_free(name, MAXPATHLEN + 1); 1712 kmem_free(nm, len); 1713 kmem_free(lnm, llen); 1714 nfs4_ntov_table_free(&ntov, &sarg); 1715 resp->attrset = 0; 1716 goto out; 1717 } 1718 1719 lname = nfscmd_convname(ca, cs->exi, lnm, 1720 NFSCMD_CONV_INBOUND, MAXPATHLEN + 1); 1721 1722 if (lname == NULL) { 1723 *cs->statusp = resp->status = NFS4ERR_SERVERFAULT; 1724 if (name != nm) 1725 kmem_free(name, MAXPATHLEN + 1); 1726 kmem_free(nm, len); 1727 kmem_free(lnm, llen); 1728 nfs4_ntov_table_free(&ntov, &sarg); 1729 resp->attrset = 0; 1730 goto out; 1731 } 1732 1733 error = VOP_SYMLINK(dvp, name, vap, lname, cr, NULL, 0); 1734 if (lname != lnm) 1735 kmem_free(lname, MAXPATHLEN + 1); 1736 kmem_free(lnm, llen); 1737 if (error) 1738 break; 1739 1740 /* 1741 * Get the initial "after" sequence number, if it fails, 1742 * set to zero 1743 */ 1744 iva.va_mask = AT_SEQ; 1745 if (VOP_GETATTR(dvp, &iva, 0, cs->cr, NULL)) 1746 iva.va_seq = 0; 1747 1748 error = VOP_LOOKUP(dvp, name, &vp, NULL, 0, NULL, cr, 1749 NULL, NULL, NULL); 1750 if (error) 1751 break; 1752 1753 /* 1754 * va_seq is not safe over VOP calls, check it again 1755 * if it has changed zero out iva to force atomic = FALSE. 1756 */ 1757 iva2.va_mask = AT_SEQ; 1758 if (VOP_GETATTR(dvp, &iva2, 0, cs->cr, NULL) || 1759 iva2.va_seq != iva.va_seq) 1760 iva.va_seq = 0; 1761 break; 1762 default: 1763 /* 1764 * probably a special file. 1765 */ 1766 if ((vap->va_mask & AT_MODE) == 0) { 1767 vap->va_mode = 0600; /* default: owner rw only */ 1768 vap->va_mask |= AT_MODE; 1769 } 1770 syncval = FNODSYNC; 1771 /* 1772 * We know this will only generate one VOP call 1773 */ 1774 vp = do_rfs4_op_mknod(args, resp, req, cs, vap, name); 1775 1776 if (vp == NULL) { 1777 if (name != nm) 1778 kmem_free(name, MAXPATHLEN + 1); 1779 kmem_free(nm, len); 1780 nfs4_ntov_table_free(&ntov, &sarg); 1781 resp->attrset = 0; 1782 goto out; 1783 } 1784 1785 /* 1786 * Get the initial "after" sequence number, if it fails, 1787 * set to zero 1788 */ 1789 iva.va_mask = AT_SEQ; 1790 if (VOP_GETATTR(dvp, &iva, 0, cs->cr, NULL)) 1791 iva.va_seq = 0; 1792 1793 break; 1794 } 1795 if (name != nm) 1796 kmem_free(name, MAXPATHLEN + 1); 1797 kmem_free(nm, len); 1798 1799 if (error) { 1800 *cs->statusp = resp->status = puterrno4(error); 1801 } 1802 1803 /* 1804 * Force modified data and metadata out to stable storage. 1805 */ 1806 (void) VOP_FSYNC(dvp, 0, cr, NULL); 1807 1808 if (resp->status != NFS4_OK) { 1809 if (vp != NULL) 1810 VN_RELE(vp); 1811 nfs4_ntov_table_free(&ntov, &sarg); 1812 resp->attrset = 0; 1813 goto out; 1814 } 1815 1816 /* 1817 * Finish setup of cinfo response, "before" value already set. 1818 * Get "after" change value, if it fails, simply return the 1819 * before value. 1820 */ 1821 ava.va_mask = AT_CTIME|AT_SEQ; 1822 if (VOP_GETATTR(dvp, &ava, 0, cr, NULL)) { 1823 ava.va_ctime = bva.va_ctime; 1824 ava.va_seq = 0; 1825 } 1826 NFS4_SET_FATTR4_CHANGE(resp->cinfo.after, ava.va_ctime); 1827 1828 /* 1829 * True verification that object was created with correct 1830 * attrs is impossible. The attrs could have been changed 1831 * immediately after object creation. If attributes did 1832 * not verify, the only recourse for the server is to 1833 * destroy the object. Maybe if some attrs (like gid) 1834 * are set incorrectly, the object should be destroyed; 1835 * however, seems bad as a default policy. Do we really 1836 * want to destroy an object over one of the times not 1837 * verifying correctly? For these reasons, the server 1838 * currently sets bits in attrset for createattrs 1839 * that were set; however, no verification is done. 1840 * 1841 * vmask_to_nmask accounts for vattr bits set on create 1842 * [do_rfs4_set_attrs() only sets resp bits for 1843 * non-vattr/vfs bits.] 1844 * Mask off any bits set by default so as not to return 1845 * more attrset bits than were requested in createattrs 1846 */ 1847 nfs4_vmask_to_nmask(sarg.vap->va_mask, &resp->attrset); 1848 resp->attrset &= args->createattrs.attrmask; 1849 nfs4_ntov_table_free(&ntov, &sarg); 1850 1851 error = makefh4(&cs->fh, vp, cs->exi); 1852 if (error) { 1853 *cs->statusp = resp->status = puterrno4(error); 1854 } 1855 1856 /* 1857 * The cinfo.atomic = TRUE only if we got no errors, we have 1858 * non-zero va_seq's, and it has incremented by exactly one 1859 * during the creation and it didn't change during the VOP_LOOKUP 1860 * or VOP_FSYNC. 1861 */ 1862 if (!error && bva.va_seq && iva.va_seq && ava.va_seq && 1863 iva.va_seq == (bva.va_seq + 1) && iva.va_seq == ava.va_seq) 1864 resp->cinfo.atomic = TRUE; 1865 else 1866 resp->cinfo.atomic = FALSE; 1867 1868 /* 1869 * Force modified metadata out to stable storage. 1870 * 1871 * if a underlying vp exists, pass it to VOP_FSYNC 1872 */ 1873 if (VOP_REALVP(vp, &realvp, NULL) == 0) 1874 (void) VOP_FSYNC(realvp, syncval, cr, NULL); 1875 else 1876 (void) VOP_FSYNC(vp, syncval, cr, NULL); 1877 1878 if (resp->status != NFS4_OK) { 1879 VN_RELE(vp); 1880 goto out; 1881 } 1882 if (cs->vp) 1883 VN_RELE(cs->vp); 1884 1885 cs->vp = vp; 1886 *cs->statusp = resp->status = NFS4_OK; 1887 out: 1888 DTRACE_NFSV4_2(op__create__done, struct compound_state *, cs, 1889 CREATE4res *, resp); 1890 } 1891 1892 /*ARGSUSED*/ 1893 static void 1894 rfs4_op_delegpurge(nfs_argop4 *argop, nfs_resop4 *resop, struct svc_req *req, 1895 struct compound_state *cs) 1896 { 1897 DTRACE_NFSV4_2(op__delegpurge__start, struct compound_state *, cs, 1898 DELEGPURGE4args *, &argop->nfs_argop4_u.opdelegpurge); 1899 1900 rfs4_op_inval(argop, resop, req, cs); 1901 1902 DTRACE_NFSV4_2(op__delegpurge__done, struct compound_state *, cs, 1903 DELEGPURGE4res *, &resop->nfs_resop4_u.opdelegpurge); 1904 } 1905 1906 /*ARGSUSED*/ 1907 static void 1908 rfs4_op_delegreturn(nfs_argop4 *argop, nfs_resop4 *resop, struct svc_req *req, 1909 struct compound_state *cs) 1910 { 1911 DELEGRETURN4args *args = &argop->nfs_argop4_u.opdelegreturn; 1912 DELEGRETURN4res *resp = &resop->nfs_resop4_u.opdelegreturn; 1913 rfs4_deleg_state_t *dsp; 1914 nfsstat4 status; 1915 1916 DTRACE_NFSV4_2(op__delegreturn__start, struct compound_state *, cs, 1917 DELEGRETURN4args *, args); 1918 1919 status = rfs4_get_deleg_state(&args->deleg_stateid, &dsp); 1920 resp->status = *cs->statusp = status; 1921 if (status != NFS4_OK) 1922 goto out; 1923 1924 /* Ensure specified filehandle matches */ 1925 if (cs->vp != dsp->rds_finfo->rf_vp) { 1926 resp->status = *cs->statusp = NFS4ERR_BAD_STATEID; 1927 } else 1928 rfs4_return_deleg(dsp, FALSE); 1929 1930 rfs4_update_lease(dsp->rds_client); 1931 1932 rfs4_deleg_state_rele(dsp); 1933 out: 1934 DTRACE_NFSV4_2(op__delegreturn__done, struct compound_state *, cs, 1935 DELEGRETURN4res *, resp); 1936 } 1937 1938 /* 1939 * Check to see if a given "flavor" is an explicitly shared flavor. 1940 * The assumption of this routine is the "flavor" is already a valid 1941 * flavor in the secinfo list of "exi". 1942 * 1943 * e.g. 1944 * # share -o sec=flavor1 /export 1945 * # share -o sec=flavor2 /export/home 1946 * 1947 * flavor2 is not an explicitly shared flavor for /export, 1948 * however it is in the secinfo list for /export thru the 1949 * server namespace setup. 1950 */ 1951 int 1952 is_exported_sec(int flavor, struct exportinfo *exi) 1953 { 1954 int i; 1955 struct secinfo *sp; 1956 1957 sp = exi->exi_export.ex_secinfo; 1958 for (i = 0; i < exi->exi_export.ex_seccnt; i++) { 1959 if (flavor == sp[i].s_secinfo.sc_nfsnum || 1960 sp[i].s_secinfo.sc_nfsnum == AUTH_NONE) { 1961 return (SEC_REF_EXPORTED(&sp[i])); 1962 } 1963 } 1964 1965 /* Should not reach this point based on the assumption */ 1966 return (0); 1967 } 1968 1969 /* 1970 * Check if the security flavor used in the request matches what is 1971 * required at the export point or at the root pseudo node (exi_root). 1972 * 1973 * returns 1 if there's a match or if exported with AUTH_NONE; 0 otherwise. 1974 * 1975 */ 1976 static int 1977 secinfo_match_or_authnone(struct compound_state *cs) 1978 { 1979 int i; 1980 struct secinfo *sp; 1981 1982 /* 1983 * Check cs->nfsflavor (from the request) against 1984 * the current export data in cs->exi. 1985 */ 1986 sp = cs->exi->exi_export.ex_secinfo; 1987 for (i = 0; i < cs->exi->exi_export.ex_seccnt; i++) { 1988 if (cs->nfsflavor == sp[i].s_secinfo.sc_nfsnum || 1989 sp[i].s_secinfo.sc_nfsnum == AUTH_NONE) 1990 return (1); 1991 } 1992 1993 return (0); 1994 } 1995 1996 /* 1997 * Check the access authority for the client and return the correct error. 1998 */ 1999 nfsstat4 2000 call_checkauth4(struct compound_state *cs, struct svc_req *req) 2001 { 2002 int authres; 2003 2004 /* 2005 * First, check if the security flavor used in the request 2006 * are among the flavors set in the server namespace. 2007 */ 2008 if (!secinfo_match_or_authnone(cs)) { 2009 *cs->statusp = NFS4ERR_WRONGSEC; 2010 return (*cs->statusp); 2011 } 2012 2013 authres = checkauth4(cs, req); 2014 2015 if (authres > 0) { 2016 *cs->statusp = NFS4_OK; 2017 if (! (cs->access & CS_ACCESS_LIMITED)) 2018 cs->access = CS_ACCESS_OK; 2019 } else if (authres == 0) { 2020 *cs->statusp = NFS4ERR_ACCESS; 2021 } else if (authres == -2) { 2022 *cs->statusp = NFS4ERR_WRONGSEC; 2023 } else { 2024 *cs->statusp = NFS4ERR_DELAY; 2025 } 2026 return (*cs->statusp); 2027 } 2028 2029 /* 2030 * bitmap4_to_attrmask is called by getattr and readdir. 2031 * It sets up the vattr mask and determines whether vfsstat call is needed 2032 * based on the input bitmap. 2033 * Returns nfsv4 status. 2034 */ 2035 static nfsstat4 2036 bitmap4_to_attrmask(bitmap4 breq, struct nfs4_svgetit_arg *sargp) 2037 { 2038 int i; 2039 uint_t va_mask; 2040 struct statvfs64 *sbp = sargp->sbp; 2041 2042 sargp->sbp = NULL; 2043 sargp->flag = 0; 2044 sargp->rdattr_error = NFS4_OK; 2045 sargp->mntdfid_set = FALSE; 2046 if (sargp->cs->vp) 2047 sargp->xattr = get_fh4_flag(&sargp->cs->fh, 2048 FH4_ATTRDIR | FH4_NAMEDATTR); 2049 else 2050 sargp->xattr = 0; 2051 2052 /* 2053 * Set rdattr_error_req to true if return error per 2054 * failed entry rather than fail the readdir. 2055 */ 2056 if (breq & FATTR4_RDATTR_ERROR_MASK) 2057 sargp->rdattr_error_req = 1; 2058 else 2059 sargp->rdattr_error_req = 0; 2060 2061 /* 2062 * generate the va_mask 2063 * Handle the easy cases first 2064 */ 2065 switch (breq) { 2066 case NFS4_NTOV_ATTR_MASK: 2067 sargp->vap->va_mask = NFS4_NTOV_ATTR_AT_MASK; 2068 return (NFS4_OK); 2069 2070 case NFS4_FS_ATTR_MASK: 2071 sargp->vap->va_mask = NFS4_FS_ATTR_AT_MASK; 2072 sargp->sbp = sbp; 2073 return (NFS4_OK); 2074 2075 case NFS4_NTOV_ATTR_CACHE_MASK: 2076 sargp->vap->va_mask = NFS4_NTOV_ATTR_CACHE_AT_MASK; 2077 return (NFS4_OK); 2078 2079 case FATTR4_LEASE_TIME_MASK: 2080 sargp->vap->va_mask = 0; 2081 return (NFS4_OK); 2082 2083 default: 2084 va_mask = 0; 2085 for (i = 0; i < nfs4_ntov_map_size; i++) { 2086 if ((breq & nfs4_ntov_map[i].fbit) && 2087 nfs4_ntov_map[i].vbit) 2088 va_mask |= nfs4_ntov_map[i].vbit; 2089 } 2090 2091 /* 2092 * Check is vfsstat is needed 2093 */ 2094 if (breq & NFS4_FS_ATTR_MASK) 2095 sargp->sbp = sbp; 2096 2097 sargp->vap->va_mask = va_mask; 2098 return (NFS4_OK); 2099 } 2100 /* NOTREACHED */ 2101 } 2102 2103 /* 2104 * bitmap4_get_sysattrs is called by getattr and readdir. 2105 * It calls both VOP_GETATTR and VFS_STATVFS calls to get the attrs. 2106 * Returns nfsv4 status. 2107 */ 2108 static nfsstat4 2109 bitmap4_get_sysattrs(struct nfs4_svgetit_arg *sargp) 2110 { 2111 int error; 2112 struct compound_state *cs = sargp->cs; 2113 vnode_t *vp = cs->vp; 2114 2115 if (sargp->sbp != NULL) { 2116 if (error = VFS_STATVFS(vp->v_vfsp, sargp->sbp)) { 2117 sargp->sbp = NULL; /* to identify error */ 2118 return (puterrno4(error)); 2119 } 2120 } 2121 2122 return (rfs4_vop_getattr(vp, sargp->vap, 0, cs->cr)); 2123 } 2124 2125 static void 2126 nfs4_ntov_table_init(struct nfs4_ntov_table *ntovp) 2127 { 2128 ntovp->na = kmem_zalloc(sizeof (union nfs4_attr_u) * nfs4_ntov_map_size, 2129 KM_SLEEP); 2130 ntovp->attrcnt = 0; 2131 ntovp->vfsstat = FALSE; 2132 } 2133 2134 static void 2135 nfs4_ntov_table_free(struct nfs4_ntov_table *ntovp, 2136 struct nfs4_svgetit_arg *sargp) 2137 { 2138 int i; 2139 union nfs4_attr_u *na; 2140 uint8_t *amap; 2141 2142 /* 2143 * XXX Should do the same checks for whether the bit is set 2144 */ 2145 for (i = 0, na = ntovp->na, amap = ntovp->amap; 2146 i < ntovp->attrcnt; i++, na++, amap++) { 2147 (void) (*nfs4_ntov_map[*amap].sv_getit)( 2148 NFS4ATTR_FREEIT, sargp, na); 2149 } 2150 if ((sargp->op == NFS4ATTR_SETIT) || (sargp->op == NFS4ATTR_VERIT)) { 2151 /* 2152 * xdr_free for getattr will be done later 2153 */ 2154 for (i = 0, na = ntovp->na, amap = ntovp->amap; 2155 i < ntovp->attrcnt; i++, na++, amap++) { 2156 xdr_free(nfs4_ntov_map[*amap].xfunc, (caddr_t)na); 2157 } 2158 } 2159 kmem_free(ntovp->na, sizeof (union nfs4_attr_u) * nfs4_ntov_map_size); 2160 } 2161 2162 /* 2163 * do_rfs4_op_getattr gets the system attrs and converts into fattr4. 2164 */ 2165 static nfsstat4 2166 do_rfs4_op_getattr(bitmap4 breq, fattr4 *fattrp, 2167 struct nfs4_svgetit_arg *sargp) 2168 { 2169 int error = 0; 2170 int i, k; 2171 struct nfs4_ntov_table ntov; 2172 XDR xdr; 2173 ulong_t xdr_size; 2174 char *xdr_attrs; 2175 nfsstat4 status = NFS4_OK; 2176 nfsstat4 prev_rdattr_error = sargp->rdattr_error; 2177 union nfs4_attr_u *na; 2178 uint8_t *amap; 2179 2180 sargp->op = NFS4ATTR_GETIT; 2181 sargp->flag = 0; 2182 2183 fattrp->attrmask = 0; 2184 /* if no bits requested, then return empty fattr4 */ 2185 if (breq == 0) { 2186 fattrp->attrlist4_len = 0; 2187 fattrp->attrlist4 = NULL; 2188 return (NFS4_OK); 2189 } 2190 2191 /* 2192 * return NFS4ERR_INVAL when client requests write-only attrs 2193 */ 2194 if (breq & (FATTR4_TIME_ACCESS_SET_MASK | FATTR4_TIME_MODIFY_SET_MASK)) 2195 return (NFS4ERR_INVAL); 2196 2197 nfs4_ntov_table_init(&ntov); 2198 na = ntov.na; 2199 amap = ntov.amap; 2200 2201 /* 2202 * Now loop to get or verify the attrs 2203 */ 2204 for (i = 0; i < nfs4_ntov_map_size; i++) { 2205 if (breq & nfs4_ntov_map[i].fbit) { 2206 if ((*nfs4_ntov_map[i].sv_getit)( 2207 NFS4ATTR_SUPPORTED, sargp, NULL) == 0) { 2208 2209 error = (*nfs4_ntov_map[i].sv_getit)( 2210 NFS4ATTR_GETIT, sargp, na); 2211 2212 /* 2213 * Possible error values: 2214 * >0 if sv_getit failed to 2215 * get the attr; 0 if succeeded; 2216 * <0 if rdattr_error and the 2217 * attribute cannot be returned. 2218 */ 2219 if (error && !(sargp->rdattr_error_req)) 2220 goto done; 2221 /* 2222 * If error then just for entry 2223 */ 2224 if (error == 0) { 2225 fattrp->attrmask |= 2226 nfs4_ntov_map[i].fbit; 2227 *amap++ = 2228 (uint8_t)nfs4_ntov_map[i].nval; 2229 na++; 2230 (ntov.attrcnt)++; 2231 } else if ((error > 0) && 2232 (sargp->rdattr_error == NFS4_OK)) { 2233 sargp->rdattr_error = puterrno4(error); 2234 } 2235 error = 0; 2236 } 2237 } 2238 } 2239 2240 /* 2241 * If rdattr_error was set after the return value for it was assigned, 2242 * update it. 2243 */ 2244 if (prev_rdattr_error != sargp->rdattr_error) { 2245 na = ntov.na; 2246 amap = ntov.amap; 2247 for (i = 0; i < ntov.attrcnt; i++, na++, amap++) { 2248 k = *amap; 2249 if (k < FATTR4_RDATTR_ERROR) { 2250 continue; 2251 } 2252 if ((k == FATTR4_RDATTR_ERROR) && 2253 ((*nfs4_ntov_map[k].sv_getit)( 2254 NFS4ATTR_SUPPORTED, sargp, NULL) == 0)) { 2255 2256 (void) (*nfs4_ntov_map[k].sv_getit)( 2257 NFS4ATTR_GETIT, sargp, na); 2258 } 2259 break; 2260 } 2261 } 2262 2263 xdr_size = 0; 2264 na = ntov.na; 2265 amap = ntov.amap; 2266 for (i = 0; i < ntov.attrcnt; i++, na++, amap++) { 2267 xdr_size += xdr_sizeof(nfs4_ntov_map[*amap].xfunc, na); 2268 } 2269 2270 fattrp->attrlist4_len = xdr_size; 2271 if (xdr_size) { 2272 /* freed by rfs4_op_getattr_free() */ 2273 fattrp->attrlist4 = xdr_attrs = kmem_zalloc(xdr_size, KM_SLEEP); 2274 2275 xdrmem_create(&xdr, xdr_attrs, xdr_size, XDR_ENCODE); 2276 2277 na = ntov.na; 2278 amap = ntov.amap; 2279 for (i = 0; i < ntov.attrcnt; i++, na++, amap++) { 2280 if (!(*nfs4_ntov_map[*amap].xfunc)(&xdr, na)) { 2281 DTRACE_PROBE1(nfss__e__getattr4_encfail, 2282 int, *amap); 2283 status = NFS4ERR_SERVERFAULT; 2284 break; 2285 } 2286 } 2287 /* xdrmem_destroy(&xdrs); */ /* NO-OP */ 2288 } else { 2289 fattrp->attrlist4 = NULL; 2290 } 2291 done: 2292 2293 nfs4_ntov_table_free(&ntov, sargp); 2294 2295 if (error != 0) 2296 status = puterrno4(error); 2297 2298 return (status); 2299 } 2300 2301 /* ARGSUSED */ 2302 static void 2303 rfs4_op_getattr(nfs_argop4 *argop, nfs_resop4 *resop, struct svc_req *req, 2304 struct compound_state *cs) 2305 { 2306 GETATTR4args *args = &argop->nfs_argop4_u.opgetattr; 2307 GETATTR4res *resp = &resop->nfs_resop4_u.opgetattr; 2308 struct nfs4_svgetit_arg sarg; 2309 struct statvfs64 sb; 2310 nfsstat4 status; 2311 2312 DTRACE_NFSV4_2(op__getattr__start, struct compound_state *, cs, 2313 GETATTR4args *, args); 2314 2315 if (cs->vp == NULL) { 2316 *cs->statusp = resp->status = NFS4ERR_NOFILEHANDLE; 2317 goto out; 2318 } 2319 2320 if (cs->access == CS_ACCESS_DENIED) { 2321 *cs->statusp = resp->status = NFS4ERR_ACCESS; 2322 goto out; 2323 } 2324 2325 sarg.sbp = &sb; 2326 sarg.cs = cs; 2327 sarg.is_referral = B_FALSE; 2328 2329 status = bitmap4_to_attrmask(args->attr_request, &sarg); 2330 if (status == NFS4_OK) { 2331 2332 status = bitmap4_get_sysattrs(&sarg); 2333 if (status == NFS4_OK) { 2334 2335 /* Is this a referral? */ 2336 if (vn_is_nfs_reparse(cs->vp, cs->cr)) { 2337 /* Older V4 Solaris client sees a link */ 2338 if (client_is_downrev(req)) 2339 sarg.vap->va_type = VLNK; 2340 else 2341 sarg.is_referral = B_TRUE; 2342 } 2343 2344 status = do_rfs4_op_getattr(args->attr_request, 2345 &resp->obj_attributes, &sarg); 2346 } 2347 } 2348 *cs->statusp = resp->status = status; 2349 out: 2350 DTRACE_NFSV4_2(op__getattr__done, struct compound_state *, cs, 2351 GETATTR4res *, resp); 2352 } 2353 2354 static void 2355 rfs4_op_getattr_free(nfs_resop4 *resop) 2356 { 2357 GETATTR4res *resp = &resop->nfs_resop4_u.opgetattr; 2358 2359 nfs4_fattr4_free(&resp->obj_attributes); 2360 } 2361 2362 /* ARGSUSED */ 2363 static void 2364 rfs4_op_getfh(nfs_argop4 *argop, nfs_resop4 *resop, struct svc_req *req, 2365 struct compound_state *cs) 2366 { 2367 GETFH4res *resp = &resop->nfs_resop4_u.opgetfh; 2368 2369 DTRACE_NFSV4_1(op__getfh__start, struct compound_state *, cs); 2370 2371 if (cs->vp == NULL) { 2372 *cs->statusp = resp->status = NFS4ERR_NOFILEHANDLE; 2373 goto out; 2374 } 2375 if (cs->access == CS_ACCESS_DENIED) { 2376 *cs->statusp = resp->status = NFS4ERR_ACCESS; 2377 goto out; 2378 } 2379 2380 /* check for reparse point at the share point */ 2381 if (cs->exi->exi_moved || vn_is_nfs_reparse(cs->exi->exi_vp, cs->cr)) { 2382 /* it's all bad */ 2383 cs->exi->exi_moved = 1; 2384 *cs->statusp = resp->status = NFS4ERR_MOVED; 2385 DTRACE_PROBE2(nfs4serv__func__referral__shared__moved, 2386 vnode_t *, cs->vp, char *, "rfs4_op_getfh"); 2387 return; 2388 } 2389 2390 /* check for reparse point at vp */ 2391 if (vn_is_nfs_reparse(cs->vp, cs->cr) && !client_is_downrev(req)) { 2392 /* it's not all bad */ 2393 *cs->statusp = resp->status = NFS4ERR_MOVED; 2394 DTRACE_PROBE2(nfs4serv__func__referral__moved, 2395 vnode_t *, cs->vp, char *, "rfs4_op_getfh"); 2396 return; 2397 } 2398 2399 resp->object.nfs_fh4_val = 2400 kmem_alloc(cs->fh.nfs_fh4_len, KM_SLEEP); 2401 nfs_fh4_copy(&cs->fh, &resp->object); 2402 *cs->statusp = resp->status = NFS4_OK; 2403 out: 2404 DTRACE_NFSV4_2(op__getfh__done, struct compound_state *, cs, 2405 GETFH4res *, resp); 2406 } 2407 2408 static void 2409 rfs4_op_getfh_free(nfs_resop4 *resop) 2410 { 2411 GETFH4res *resp = &resop->nfs_resop4_u.opgetfh; 2412 2413 if (resp->status == NFS4_OK && 2414 resp->object.nfs_fh4_val != NULL) { 2415 kmem_free(resp->object.nfs_fh4_val, resp->object.nfs_fh4_len); 2416 resp->object.nfs_fh4_val = NULL; 2417 resp->object.nfs_fh4_len = 0; 2418 } 2419 } 2420 2421 /* 2422 * illegal: args: void 2423 * res : status (NFS4ERR_OP_ILLEGAL) 2424 */ 2425 /* ARGSUSED */ 2426 static void 2427 rfs4_op_illegal(nfs_argop4 *argop, nfs_resop4 *resop, 2428 struct svc_req *req, struct compound_state *cs) 2429 { 2430 ILLEGAL4res *resp = &resop->nfs_resop4_u.opillegal; 2431 2432 resop->resop = OP_ILLEGAL; 2433 *cs->statusp = resp->status = NFS4ERR_OP_ILLEGAL; 2434 } 2435 2436 /* 2437 * link: args: SAVED_FH: file, CURRENT_FH: target directory 2438 * res: status. If success - CURRENT_FH unchanged, return change_info 2439 */ 2440 /* ARGSUSED */ 2441 static void 2442 rfs4_op_link(nfs_argop4 *argop, nfs_resop4 *resop, struct svc_req *req, 2443 struct compound_state *cs) 2444 { 2445 LINK4args *args = &argop->nfs_argop4_u.oplink; 2446 LINK4res *resp = &resop->nfs_resop4_u.oplink; 2447 int error; 2448 vnode_t *vp; 2449 vnode_t *dvp; 2450 struct vattr bdva, idva, adva; 2451 char *nm; 2452 uint_t len; 2453 struct sockaddr *ca; 2454 char *name = NULL; 2455 nfsstat4 status; 2456 2457 DTRACE_NFSV4_2(op__link__start, struct compound_state *, cs, 2458 LINK4args *, args); 2459 2460 /* SAVED_FH: source object */ 2461 vp = cs->saved_vp; 2462 if (vp == NULL) { 2463 *cs->statusp = resp->status = NFS4ERR_NOFILEHANDLE; 2464 goto out; 2465 } 2466 2467 /* CURRENT_FH: target directory */ 2468 dvp = cs->vp; 2469 if (dvp == NULL) { 2470 *cs->statusp = resp->status = NFS4ERR_NOFILEHANDLE; 2471 goto out; 2472 } 2473 2474 /* 2475 * If there is a non-shared filesystem mounted on this vnode, 2476 * do not allow to link any file in this directory. 2477 */ 2478 if (vn_ismntpt(dvp)) { 2479 *cs->statusp = resp->status = NFS4ERR_ACCESS; 2480 goto out; 2481 } 2482 2483 if (cs->access == CS_ACCESS_DENIED) { 2484 *cs->statusp = resp->status = NFS4ERR_ACCESS; 2485 goto out; 2486 } 2487 2488 /* Check source object's type validity */ 2489 if (vp->v_type == VDIR) { 2490 *cs->statusp = resp->status = NFS4ERR_ISDIR; 2491 goto out; 2492 } 2493 2494 /* Check target directory's type */ 2495 if (dvp->v_type != VDIR) { 2496 *cs->statusp = resp->status = NFS4ERR_NOTDIR; 2497 goto out; 2498 } 2499 2500 if (cs->saved_exi != cs->exi) { 2501 *cs->statusp = resp->status = NFS4ERR_XDEV; 2502 goto out; 2503 } 2504 2505 status = utf8_dir_verify(&args->newname); 2506 if (status != NFS4_OK) { 2507 *cs->statusp = resp->status = status; 2508 goto out; 2509 } 2510 2511 nm = utf8_to_fn(&args->newname, &len, NULL); 2512 if (nm == NULL) { 2513 *cs->statusp = resp->status = NFS4ERR_INVAL; 2514 goto out; 2515 } 2516 2517 if (len > MAXNAMELEN) { 2518 *cs->statusp = resp->status = NFS4ERR_NAMETOOLONG; 2519 kmem_free(nm, len); 2520 goto out; 2521 } 2522 2523 if (rdonly4(cs->exi, cs->vp, req)) { 2524 *cs->statusp = resp->status = NFS4ERR_ROFS; 2525 kmem_free(nm, len); 2526 goto out; 2527 } 2528 2529 /* Get "before" change value */ 2530 bdva.va_mask = AT_CTIME|AT_SEQ; 2531 error = VOP_GETATTR(dvp, &bdva, 0, cs->cr, NULL); 2532 if (error) { 2533 *cs->statusp = resp->status = puterrno4(error); 2534 kmem_free(nm, len); 2535 goto out; 2536 } 2537 2538 ca = (struct sockaddr *)svc_getrpccaller(req->rq_xprt)->buf; 2539 name = nfscmd_convname(ca, cs->exi, nm, NFSCMD_CONV_INBOUND, 2540 MAXPATHLEN + 1); 2541 2542 if (name == NULL) { 2543 *cs->statusp = resp->status = NFS4ERR_INVAL; 2544 kmem_free(nm, len); 2545 goto out; 2546 } 2547 2548 NFS4_SET_FATTR4_CHANGE(resp->cinfo.before, bdva.va_ctime) 2549 2550 error = VOP_LINK(dvp, vp, name, cs->cr, NULL, 0); 2551 2552 if (nm != name) 2553 kmem_free(name, MAXPATHLEN + 1); 2554 kmem_free(nm, len); 2555 2556 /* 2557 * Get the initial "after" sequence number, if it fails, set to zero 2558 */ 2559 idva.va_mask = AT_SEQ; 2560 if (VOP_GETATTR(dvp, &idva, 0, cs->cr, NULL)) 2561 idva.va_seq = 0; 2562 2563 /* 2564 * Force modified data and metadata out to stable storage. 2565 */ 2566 (void) VOP_FSYNC(vp, FNODSYNC, cs->cr, NULL); 2567 (void) VOP_FSYNC(dvp, 0, cs->cr, NULL); 2568 2569 if (error) { 2570 *cs->statusp = resp->status = puterrno4(error); 2571 goto out; 2572 } 2573 2574 /* 2575 * Get "after" change value, if it fails, simply return the 2576 * before value. 2577 */ 2578 adva.va_mask = AT_CTIME|AT_SEQ; 2579 if (VOP_GETATTR(dvp, &adva, 0, cs->cr, NULL)) { 2580 adva.va_ctime = bdva.va_ctime; 2581 adva.va_seq = 0; 2582 } 2583 2584 NFS4_SET_FATTR4_CHANGE(resp->cinfo.after, adva.va_ctime) 2585 2586 /* 2587 * The cinfo.atomic = TRUE only if we have 2588 * non-zero va_seq's, and it has incremented by exactly one 2589 * during the VOP_LINK and it didn't change during the VOP_FSYNC. 2590 */ 2591 if (bdva.va_seq && idva.va_seq && adva.va_seq && 2592 idva.va_seq == (bdva.va_seq + 1) && idva.va_seq == adva.va_seq) 2593 resp->cinfo.atomic = TRUE; 2594 else 2595 resp->cinfo.atomic = FALSE; 2596 2597 *cs->statusp = resp->status = NFS4_OK; 2598 out: 2599 DTRACE_NFSV4_2(op__link__done, struct compound_state *, cs, 2600 LINK4res *, resp); 2601 } 2602 2603 /* 2604 * Used by rfs4_op_lookup and rfs4_op_lookupp to do the actual work. 2605 */ 2606 2607 /* ARGSUSED */ 2608 static nfsstat4 2609 do_rfs4_op_lookup(char *nm, struct svc_req *req, struct compound_state *cs) 2610 { 2611 int error; 2612 int different_export = 0; 2613 vnode_t *vp, *tvp, *pre_tvp = NULL, *oldvp = NULL; 2614 struct exportinfo *exi = NULL, *pre_exi = NULL; 2615 nfsstat4 stat; 2616 fid_t fid; 2617 int attrdir, dotdot, walk; 2618 bool_t is_newvp = FALSE; 2619 2620 if (cs->vp->v_flag & V_XATTRDIR) { 2621 attrdir = 1; 2622 ASSERT(get_fh4_flag(&cs->fh, FH4_ATTRDIR)); 2623 } else { 2624 attrdir = 0; 2625 ASSERT(! get_fh4_flag(&cs->fh, FH4_ATTRDIR)); 2626 } 2627 2628 dotdot = (nm[0] == '.' && nm[1] == '.' && nm[2] == '\0'); 2629 2630 /* 2631 * If dotdotting, then need to check whether it's 2632 * above the root of a filesystem, or above an 2633 * export point. 2634 */ 2635 if (dotdot) { 2636 2637 /* 2638 * If dotdotting at the root of a filesystem, then 2639 * need to traverse back to the mounted-on filesystem 2640 * and do the dotdot lookup there. 2641 */ 2642 if (cs->vp->v_flag & VROOT) { 2643 2644 /* 2645 * If at the system root, then can 2646 * go up no further. 2647 */ 2648 if (VN_CMP(cs->vp, rootdir)) 2649 return (puterrno4(ENOENT)); 2650 2651 /* 2652 * Traverse back to the mounted-on filesystem 2653 */ 2654 cs->vp = untraverse(cs->vp); 2655 2656 /* 2657 * Set the different_export flag so we remember 2658 * to pick up a new exportinfo entry for 2659 * this new filesystem. 2660 */ 2661 different_export = 1; 2662 } else { 2663 2664 /* 2665 * If dotdotting above an export point then set 2666 * the different_export to get new export info. 2667 */ 2668 different_export = nfs_exported(cs->exi, cs->vp); 2669 } 2670 } 2671 2672 error = VOP_LOOKUP(cs->vp, nm, &vp, NULL, 0, NULL, cs->cr, 2673 NULL, NULL, NULL); 2674 if (error) 2675 return (puterrno4(error)); 2676 2677 /* 2678 * If the vnode is in a pseudo filesystem, check whether it is visible. 2679 * 2680 * XXX if the vnode is a symlink and it is not visible in 2681 * a pseudo filesystem, return ENOENT (not following symlink). 2682 * V4 client can not mount such symlink. This is a regression 2683 * from V2/V3. 2684 * 2685 * In the same exported filesystem, if the security flavor used 2686 * is not an explicitly shared flavor, limit the view to the visible 2687 * list entries only. This is not a WRONGSEC case because it's already 2688 * checked via PUTROOTFH/PUTPUBFH or PUTFH. 2689 */ 2690 if (!different_export && 2691 (PSEUDO(cs->exi) || ! is_exported_sec(cs->nfsflavor, cs->exi) || 2692 cs->access & CS_ACCESS_LIMITED)) { 2693 if (! nfs_visible(cs->exi, vp, &different_export)) { 2694 VN_RELE(vp); 2695 return (puterrno4(ENOENT)); 2696 } 2697 } 2698 2699 /* 2700 * If it's a mountpoint, then traverse it. 2701 */ 2702 if (vn_ismntpt(vp)) { 2703 pre_exi = cs->exi; /* save pre-traversed exportinfo */ 2704 pre_tvp = vp; /* save pre-traversed vnode */ 2705 2706 /* 2707 * hold pre_tvp to counteract rele by traverse. We will 2708 * need pre_tvp below if checkexport4 fails 2709 */ 2710 VN_HOLD(pre_tvp); 2711 tvp = vp; 2712 if ((error = traverse(&tvp)) != 0) { 2713 VN_RELE(vp); 2714 VN_RELE(pre_tvp); 2715 return (puterrno4(error)); 2716 } 2717 vp = tvp; 2718 different_export = 1; 2719 } else if (vp->v_vfsp != cs->vp->v_vfsp) { 2720 /* 2721 * The vfsp comparison is to handle the case where 2722 * a LOFS mount is shared. lo_lookup traverses mount points, 2723 * and NFS is unaware of local fs transistions because 2724 * v_vfsmountedhere isn't set. For this special LOFS case, 2725 * the dir and the obj returned by lookup will have different 2726 * vfs ptrs. 2727 */ 2728 different_export = 1; 2729 } 2730 2731 if (different_export) { 2732 2733 bzero(&fid, sizeof (fid)); 2734 fid.fid_len = MAXFIDSZ; 2735 error = vop_fid_pseudo(vp, &fid); 2736 if (error) { 2737 VN_RELE(vp); 2738 if (pre_tvp) 2739 VN_RELE(pre_tvp); 2740 return (puterrno4(error)); 2741 } 2742 2743 if (dotdot) 2744 exi = nfs_vptoexi(NULL, vp, cs->cr, &walk, NULL, TRUE); 2745 else 2746 exi = checkexport4(&vp->v_vfsp->vfs_fsid, &fid, vp); 2747 2748 if (exi == NULL) { 2749 if (pre_tvp) { 2750 /* 2751 * If this vnode is a mounted-on vnode, 2752 * but the mounted-on file system is not 2753 * exported, send back the filehandle for 2754 * the mounted-on vnode, not the root of 2755 * the mounted-on file system. 2756 */ 2757 VN_RELE(vp); 2758 vp = pre_tvp; 2759 exi = pre_exi; 2760 } else { 2761 VN_RELE(vp); 2762 return (puterrno4(EACCES)); 2763 } 2764 } else if (pre_tvp) { 2765 /* we're done with pre_tvp now. release extra hold */ 2766 VN_RELE(pre_tvp); 2767 } 2768 2769 cs->exi = exi; 2770 2771 /* 2772 * Now we do a checkauth4. The reason is that 2773 * this client/user may not have access to the new 2774 * exported file system, and if he does, 2775 * the client/user may be mapped to a different uid. 2776 * 2777 * We start with a new cr, because the checkauth4 done 2778 * in the PUT*FH operation over wrote the cred's uid, 2779 * gid, etc, and we want the real thing before calling 2780 * checkauth4() 2781 */ 2782 crfree(cs->cr); 2783 cs->cr = crdup(cs->basecr); 2784 2785 oldvp = cs->vp; 2786 cs->vp = vp; 2787 is_newvp = TRUE; 2788 2789 stat = call_checkauth4(cs, req); 2790 if (stat != NFS4_OK) { 2791 VN_RELE(cs->vp); 2792 cs->vp = oldvp; 2793 return (stat); 2794 } 2795 } 2796 2797 /* 2798 * After various NFS checks, do a label check on the path 2799 * component. The label on this path should either be the 2800 * global zone's label or a zone's label. We are only 2801 * interested in the zone's label because exported files 2802 * in global zone is accessible (though read-only) to 2803 * clients. The exportability/visibility check is already 2804 * done before reaching this code. 2805 */ 2806 if (is_system_labeled()) { 2807 bslabel_t *clabel; 2808 2809 ASSERT(req->rq_label != NULL); 2810 clabel = req->rq_label; 2811 DTRACE_PROBE2(tx__rfs4__log__info__oplookup__clabel, char *, 2812 "got client label from request(1)", struct svc_req *, req); 2813 2814 if (!blequal(&l_admin_low->tsl_label, clabel)) { 2815 if (!do_rfs_label_check(clabel, vp, DOMINANCE_CHECK, 2816 cs->exi)) { 2817 error = EACCES; 2818 goto err_out; 2819 } 2820 } else { 2821 /* 2822 * We grant access to admin_low label clients 2823 * only if the client is trusted, i.e. also 2824 * running Solaris Trusted Extension. 2825 */ 2826 struct sockaddr *ca; 2827 int addr_type; 2828 void *ipaddr; 2829 tsol_tpc_t *tp; 2830 2831 ca = (struct sockaddr *)svc_getrpccaller( 2832 req->rq_xprt)->buf; 2833 if (ca->sa_family == AF_INET) { 2834 addr_type = IPV4_VERSION; 2835 ipaddr = &((struct sockaddr_in *)ca)->sin_addr; 2836 } else if (ca->sa_family == AF_INET6) { 2837 addr_type = IPV6_VERSION; 2838 ipaddr = &((struct sockaddr_in6 *) 2839 ca)->sin6_addr; 2840 } 2841 tp = find_tpc(ipaddr, addr_type, B_FALSE); 2842 if (tp == NULL || tp->tpc_tp.tp_doi != 2843 l_admin_low->tsl_doi || tp->tpc_tp.host_type != 2844 SUN_CIPSO) { 2845 if (tp != NULL) 2846 TPC_RELE(tp); 2847 error = EACCES; 2848 goto err_out; 2849 } 2850 TPC_RELE(tp); 2851 } 2852 } 2853 2854 error = makefh4(&cs->fh, vp, cs->exi); 2855 2856 err_out: 2857 if (error) { 2858 if (is_newvp) { 2859 VN_RELE(cs->vp); 2860 cs->vp = oldvp; 2861 } else 2862 VN_RELE(vp); 2863 return (puterrno4(error)); 2864 } 2865 2866 if (!is_newvp) { 2867 if (cs->vp) 2868 VN_RELE(cs->vp); 2869 cs->vp = vp; 2870 } else if (oldvp) 2871 VN_RELE(oldvp); 2872 2873 /* 2874 * if did lookup on attrdir and didn't lookup .., set named 2875 * attr fh flag 2876 */ 2877 if (attrdir && ! dotdot) 2878 set_fh4_flag(&cs->fh, FH4_NAMEDATTR); 2879 2880 /* Assume false for now, open proc will set this */ 2881 cs->mandlock = FALSE; 2882 2883 return (NFS4_OK); 2884 } 2885 2886 /* ARGSUSED */ 2887 static void 2888 rfs4_op_lookup(nfs_argop4 *argop, nfs_resop4 *resop, struct svc_req *req, 2889 struct compound_state *cs) 2890 { 2891 LOOKUP4args *args = &argop->nfs_argop4_u.oplookup; 2892 LOOKUP4res *resp = &resop->nfs_resop4_u.oplookup; 2893 char *nm; 2894 uint_t len; 2895 struct sockaddr *ca; 2896 char *name = NULL; 2897 nfsstat4 status; 2898 2899 DTRACE_NFSV4_2(op__lookup__start, struct compound_state *, cs, 2900 LOOKUP4args *, args); 2901 2902 if (cs->vp == NULL) { 2903 *cs->statusp = resp->status = NFS4ERR_NOFILEHANDLE; 2904 goto out; 2905 } 2906 2907 if (cs->vp->v_type == VLNK) { 2908 *cs->statusp = resp->status = NFS4ERR_SYMLINK; 2909 goto out; 2910 } 2911 2912 if (cs->vp->v_type != VDIR) { 2913 *cs->statusp = resp->status = NFS4ERR_NOTDIR; 2914 goto out; 2915 } 2916 2917 status = utf8_dir_verify(&args->objname); 2918 if (status != NFS4_OK) { 2919 *cs->statusp = resp->status = status; 2920 goto out; 2921 } 2922 2923 nm = utf8_to_str(&args->objname, &len, NULL); 2924 if (nm == NULL) { 2925 *cs->statusp = resp->status = NFS4ERR_INVAL; 2926 goto out; 2927 } 2928 2929 if (len > MAXNAMELEN) { 2930 *cs->statusp = resp->status = NFS4ERR_NAMETOOLONG; 2931 kmem_free(nm, len); 2932 goto out; 2933 } 2934 2935 ca = (struct sockaddr *)svc_getrpccaller(req->rq_xprt)->buf; 2936 name = nfscmd_convname(ca, cs->exi, nm, NFSCMD_CONV_INBOUND, 2937 MAXPATHLEN + 1); 2938 2939 if (name == NULL) { 2940 *cs->statusp = resp->status = NFS4ERR_INVAL; 2941 kmem_free(nm, len); 2942 goto out; 2943 } 2944 2945 *cs->statusp = resp->status = do_rfs4_op_lookup(name, req, cs); 2946 2947 if (name != nm) 2948 kmem_free(name, MAXPATHLEN + 1); 2949 kmem_free(nm, len); 2950 2951 out: 2952 DTRACE_NFSV4_2(op__lookup__done, struct compound_state *, cs, 2953 LOOKUP4res *, resp); 2954 } 2955 2956 /* ARGSUSED */ 2957 static void 2958 rfs4_op_lookupp(nfs_argop4 *args, nfs_resop4 *resop, struct svc_req *req, 2959 struct compound_state *cs) 2960 { 2961 LOOKUPP4res *resp = &resop->nfs_resop4_u.oplookupp; 2962 2963 DTRACE_NFSV4_1(op__lookupp__start, struct compound_state *, cs); 2964 2965 if (cs->vp == NULL) { 2966 *cs->statusp = resp->status = NFS4ERR_NOFILEHANDLE; 2967 goto out; 2968 } 2969 2970 if (cs->vp->v_type != VDIR) { 2971 *cs->statusp = resp->status = NFS4ERR_NOTDIR; 2972 goto out; 2973 } 2974 2975 *cs->statusp = resp->status = do_rfs4_op_lookup("..", req, cs); 2976 2977 /* 2978 * From NFSV4 Specification, LOOKUPP should not check for 2979 * NFS4ERR_WRONGSEC. Retrun NFS4_OK instead. 2980 */ 2981 if (resp->status == NFS4ERR_WRONGSEC) { 2982 *cs->statusp = resp->status = NFS4_OK; 2983 } 2984 2985 out: 2986 DTRACE_NFSV4_2(op__lookupp__done, struct compound_state *, cs, 2987 LOOKUPP4res *, resp); 2988 } 2989 2990 2991 /*ARGSUSED2*/ 2992 static void 2993 rfs4_op_openattr(nfs_argop4 *argop, nfs_resop4 *resop, struct svc_req *req, 2994 struct compound_state *cs) 2995 { 2996 OPENATTR4args *args = &argop->nfs_argop4_u.opopenattr; 2997 OPENATTR4res *resp = &resop->nfs_resop4_u.opopenattr; 2998 vnode_t *avp = NULL; 2999 int lookup_flags = LOOKUP_XATTR, error; 3000 int exp_ro = 0; 3001 3002 DTRACE_NFSV4_2(op__openattr__start, struct compound_state *, cs, 3003 OPENATTR4args *, args); 3004 3005 if (cs->vp == NULL) { 3006 *cs->statusp = resp->status = NFS4ERR_NOFILEHANDLE; 3007 goto out; 3008 } 3009 3010 if ((cs->vp->v_vfsp->vfs_flag & VFS_XATTR) == 0 && 3011 !vfs_has_feature(cs->vp->v_vfsp, VFSFT_SYSATTR_VIEWS)) { 3012 *cs->statusp = resp->status = puterrno4(ENOTSUP); 3013 goto out; 3014 } 3015 3016 /* 3017 * If file system supports passing ACE mask to VOP_ACCESS then 3018 * check for ACE_READ_NAMED_ATTRS, otherwise do legacy checks 3019 */ 3020 3021 if (vfs_has_feature(cs->vp->v_vfsp, VFSFT_ACEMASKONACCESS)) 3022 error = VOP_ACCESS(cs->vp, ACE_READ_NAMED_ATTRS, 3023 V_ACE_MASK, cs->cr, NULL); 3024 else 3025 error = ((VOP_ACCESS(cs->vp, VREAD, 0, cs->cr, NULL) != 0) && 3026 (VOP_ACCESS(cs->vp, VWRITE, 0, cs->cr, NULL) != 0) && 3027 (VOP_ACCESS(cs->vp, VEXEC, 0, cs->cr, NULL) != 0)); 3028 3029 if (error) { 3030 *cs->statusp = resp->status = puterrno4(EACCES); 3031 goto out; 3032 } 3033 3034 /* 3035 * The CREATE_XATTR_DIR VOP flag cannot be specified if 3036 * the file system is exported read-only -- regardless of 3037 * createdir flag. Otherwise the attrdir would be created 3038 * (assuming server fs isn't mounted readonly locally). If 3039 * VOP_LOOKUP returns ENOENT in this case, the error will 3040 * be translated into EROFS. ENOSYS is mapped to ENOTSUP 3041 * because specfs has no VOP_LOOKUP op, so the macro would 3042 * return ENOSYS. EINVAL is returned by all (current) 3043 * Solaris file system implementations when any of their 3044 * restrictions are violated (xattr(dir) can't have xattrdir). 3045 * Returning NOTSUPP is more appropriate in this case 3046 * because the object will never be able to have an attrdir. 3047 */ 3048 if (args->createdir && ! (exp_ro = rdonly4(cs->exi, cs->vp, req))) 3049 lookup_flags |= CREATE_XATTR_DIR; 3050 3051 error = VOP_LOOKUP(cs->vp, "", &avp, NULL, lookup_flags, NULL, cs->cr, 3052 NULL, NULL, NULL); 3053 3054 if (error) { 3055 if (error == ENOENT && args->createdir && exp_ro) 3056 *cs->statusp = resp->status = puterrno4(EROFS); 3057 else if (error == EINVAL || error == ENOSYS) 3058 *cs->statusp = resp->status = puterrno4(ENOTSUP); 3059 else 3060 *cs->statusp = resp->status = puterrno4(error); 3061 goto out; 3062 } 3063 3064 ASSERT(avp->v_flag & V_XATTRDIR); 3065 3066 error = makefh4(&cs->fh, avp, cs->exi); 3067 3068 if (error) { 3069 VN_RELE(avp); 3070 *cs->statusp = resp->status = puterrno4(error); 3071 goto out; 3072 } 3073 3074 VN_RELE(cs->vp); 3075 cs->vp = avp; 3076 3077 /* 3078 * There is no requirement for an attrdir fh flag 3079 * because the attrdir has a vnode flag to distinguish 3080 * it from regular (non-xattr) directories. The 3081 * FH4_ATTRDIR flag is set for future sanity checks. 3082 */ 3083 set_fh4_flag(&cs->fh, FH4_ATTRDIR); 3084 *cs->statusp = resp->status = NFS4_OK; 3085 3086 out: 3087 DTRACE_NFSV4_2(op__openattr__done, struct compound_state *, cs, 3088 OPENATTR4res *, resp); 3089 } 3090 3091 static int 3092 do_io(int direction, vnode_t *vp, struct uio *uio, int ioflag, cred_t *cred, 3093 caller_context_t *ct) 3094 { 3095 int error; 3096 int i; 3097 clock_t delaytime; 3098 3099 delaytime = MSEC_TO_TICK_ROUNDUP(rfs4_lock_delay); 3100 3101 /* 3102 * Don't block on mandatory locks. If this routine returns 3103 * EAGAIN, the caller should return NFS4ERR_LOCKED. 3104 */ 3105 uio->uio_fmode = FNONBLOCK; 3106 3107 for (i = 0; i < rfs4_maxlock_tries; i++) { 3108 3109 3110 if (direction == FREAD) { 3111 (void) VOP_RWLOCK(vp, V_WRITELOCK_FALSE, ct); 3112 error = VOP_READ(vp, uio, ioflag, cred, ct); 3113 VOP_RWUNLOCK(vp, V_WRITELOCK_FALSE, ct); 3114 } else { 3115 (void) VOP_RWLOCK(vp, V_WRITELOCK_TRUE, ct); 3116 error = VOP_WRITE(vp, uio, ioflag, cred, ct); 3117 VOP_RWUNLOCK(vp, V_WRITELOCK_TRUE, ct); 3118 } 3119 3120 if (error != EAGAIN) 3121 break; 3122 3123 if (i < rfs4_maxlock_tries - 1) { 3124 delay(delaytime); 3125 delaytime *= 2; 3126 } 3127 } 3128 3129 return (error); 3130 } 3131 3132 /* ARGSUSED */ 3133 static void 3134 rfs4_op_read(nfs_argop4 *argop, nfs_resop4 *resop, struct svc_req *req, 3135 struct compound_state *cs) 3136 { 3137 READ4args *args = &argop->nfs_argop4_u.opread; 3138 READ4res *resp = &resop->nfs_resop4_u.opread; 3139 int error; 3140 int verror; 3141 vnode_t *vp; 3142 struct vattr va; 3143 struct iovec iov; 3144 struct uio uio; 3145 u_offset_t offset; 3146 bool_t *deleg = &cs->deleg; 3147 nfsstat4 stat; 3148 int in_crit = 0; 3149 mblk_t *mp = NULL; 3150 int alloc_err = 0; 3151 int rdma_used = 0; 3152 int loaned_buffers; 3153 caller_context_t ct; 3154 struct uio *uiop; 3155 3156 DTRACE_NFSV4_2(op__read__start, struct compound_state *, cs, 3157 READ4args, args); 3158 3159 vp = cs->vp; 3160 if (vp == NULL) { 3161 *cs->statusp = resp->status = NFS4ERR_NOFILEHANDLE; 3162 goto out; 3163 } 3164 if (cs->access == CS_ACCESS_DENIED) { 3165 *cs->statusp = resp->status = NFS4ERR_ACCESS; 3166 goto out; 3167 } 3168 3169 if ((stat = rfs4_check_stateid(FREAD, vp, &args->stateid, FALSE, 3170 deleg, TRUE, &ct)) != NFS4_OK) { 3171 *cs->statusp = resp->status = stat; 3172 goto out; 3173 } 3174 3175 /* 3176 * Enter the critical region before calling VOP_RWLOCK 3177 * to avoid a deadlock with write requests. 3178 */ 3179 if (nbl_need_check(vp)) { 3180 nbl_start_crit(vp, RW_READER); 3181 in_crit = 1; 3182 if (nbl_conflict(vp, NBL_READ, args->offset, args->count, 0, 3183 &ct)) { 3184 *cs->statusp = resp->status = NFS4ERR_LOCKED; 3185 goto out; 3186 } 3187 } 3188 3189 if ((stat = rfs4_check_stateid(FREAD, vp, &args->stateid, FALSE, 3190 deleg, TRUE, &ct)) != NFS4_OK) { 3191 *cs->statusp = resp->status = stat; 3192 goto out; 3193 } 3194 3195 if (args->wlist) { 3196 if (args->count > clist_len(args->wlist)) { 3197 *cs->statusp = resp->status = NFS4ERR_INVAL; 3198 goto out; 3199 } 3200 rdma_used = 1; 3201 } 3202 3203 /* use loaned buffers for TCP */ 3204 loaned_buffers = (nfs_loaned_buffers && !rdma_used) ? 1 : 0; 3205 3206 va.va_mask = AT_MODE|AT_SIZE|AT_UID; 3207 verror = VOP_GETATTR(vp, &va, 0, cs->cr, &ct); 3208 3209 /* 3210 * If we can't get the attributes, then we can't do the 3211 * right access checking. So, we'll fail the request. 3212 */ 3213 if (verror) { 3214 *cs->statusp = resp->status = puterrno4(verror); 3215 goto out; 3216 } 3217 3218 if (vp->v_type != VREG) { 3219 *cs->statusp = resp->status = 3220 ((vp->v_type == VDIR) ? NFS4ERR_ISDIR : NFS4ERR_INVAL); 3221 goto out; 3222 } 3223 3224 if (crgetuid(cs->cr) != va.va_uid && 3225 (error = VOP_ACCESS(vp, VREAD, 0, cs->cr, &ct)) && 3226 (error = VOP_ACCESS(vp, VEXEC, 0, cs->cr, &ct))) { 3227 *cs->statusp = resp->status = puterrno4(error); 3228 goto out; 3229 } 3230 3231 if (MANDLOCK(vp, va.va_mode)) { /* XXX - V4 supports mand locking */ 3232 *cs->statusp = resp->status = NFS4ERR_ACCESS; 3233 goto out; 3234 } 3235 3236 offset = args->offset; 3237 if (offset >= va.va_size) { 3238 *cs->statusp = resp->status = NFS4_OK; 3239 resp->eof = TRUE; 3240 resp->data_len = 0; 3241 resp->data_val = NULL; 3242 resp->mblk = NULL; 3243 /* RDMA */ 3244 resp->wlist = args->wlist; 3245 resp->wlist_len = resp->data_len; 3246 *cs->statusp = resp->status = NFS4_OK; 3247 if (resp->wlist) 3248 clist_zero_len(resp->wlist); 3249 goto out; 3250 } 3251 3252 if (args->count == 0) { 3253 *cs->statusp = resp->status = NFS4_OK; 3254 resp->eof = FALSE; 3255 resp->data_len = 0; 3256 resp->data_val = NULL; 3257 resp->mblk = NULL; 3258 /* RDMA */ 3259 resp->wlist = args->wlist; 3260 resp->wlist_len = resp->data_len; 3261 if (resp->wlist) 3262 clist_zero_len(resp->wlist); 3263 goto out; 3264 } 3265 3266 /* 3267 * Do not allocate memory more than maximum allowed 3268 * transfer size 3269 */ 3270 if (args->count > rfs4_tsize(req)) 3271 args->count = rfs4_tsize(req); 3272 3273 if (loaned_buffers) { 3274 uiop = (uio_t *)rfs_setup_xuio(vp); 3275 ASSERT(uiop != NULL); 3276 uiop->uio_segflg = UIO_SYSSPACE; 3277 uiop->uio_loffset = args->offset; 3278 uiop->uio_resid = args->count; 3279 3280 /* Jump to do the read if successful */ 3281 if (!VOP_REQZCBUF(vp, UIO_READ, (xuio_t *)uiop, cs->cr, &ct)) { 3282 /* 3283 * Need to hold the vnode until after VOP_RETZCBUF() 3284 * is called. 3285 */ 3286 VN_HOLD(vp); 3287 goto doio_read; 3288 } 3289 3290 DTRACE_PROBE2(nfss__i__reqzcbuf_failed, int, 3291 uiop->uio_loffset, int, uiop->uio_resid); 3292 3293 uiop->uio_extflg = 0; 3294 3295 /* failure to setup for zero copy */ 3296 rfs_free_xuio((void *)uiop); 3297 loaned_buffers = 0; 3298 } 3299 3300 /* 3301 * If returning data via RDMA Write, then grab the chunk list. If we 3302 * aren't returning READ data w/RDMA_WRITE, then grab a mblk. 3303 */ 3304 if (rdma_used) { 3305 mp = NULL; 3306 (void) rdma_get_wchunk(req, &iov, args->wlist); 3307 } else { 3308 /* 3309 * mp will contain the data to be sent out in the read reply. 3310 * It will be freed after the reply has been sent. Let's 3311 * roundup the data to a BYTES_PER_XDR_UNIT multiple, so that 3312 * the call to xdrmblk_putmblk() never fails. If the first 3313 * alloc of the requested size fails, then decrease the size to 3314 * something more reasonable and wait for the allocation to 3315 * occur. 3316 */ 3317 mp = allocb(RNDUP(args->count), BPRI_MED); 3318 if (mp == NULL) { 3319 if (args->count > MAXBSIZE) 3320 args->count = MAXBSIZE; 3321 mp = allocb_wait(RNDUP(args->count), BPRI_MED, 3322 STR_NOSIG, &alloc_err); 3323 } 3324 ASSERT(mp != NULL); 3325 ASSERT(alloc_err == 0); 3326 3327 iov.iov_base = (caddr_t)mp->b_datap->db_base; 3328 iov.iov_len = args->count; 3329 } 3330 3331 uio.uio_iov = &iov; 3332 uio.uio_iovcnt = 1; 3333 uio.uio_segflg = UIO_SYSSPACE; 3334 uio.uio_extflg = UIO_COPY_CACHED; 3335 uio.uio_loffset = args->offset; 3336 uio.uio_resid = args->count; 3337 uiop = &uio; 3338 3339 doio_read: 3340 error = do_io(FREAD, vp, uiop, 0, cs->cr, &ct); 3341 3342 va.va_mask = AT_SIZE; 3343 verror = VOP_GETATTR(vp, &va, 0, cs->cr, &ct); 3344 3345 if (error) { 3346 if (mp) 3347 freemsg(mp); 3348 *cs->statusp = resp->status = puterrno4(error); 3349 goto out; 3350 } 3351 3352 /* make mblk using zc buffers */ 3353 if (loaned_buffers) { 3354 mp = uio_to_mblk(uiop); 3355 ASSERT(mp != NULL); 3356 } 3357 3358 *cs->statusp = resp->status = NFS4_OK; 3359 3360 ASSERT(uiop->uio_resid >= 0); 3361 resp->data_len = args->count - uiop->uio_resid; 3362 if (mp) { 3363 resp->data_val = (char *)mp->b_datap->db_base; 3364 rfs_rndup_mblks(mp, resp->data_len, loaned_buffers); 3365 } else { 3366 resp->data_val = (caddr_t)iov.iov_base; 3367 } 3368 3369 resp->mblk = mp; 3370 3371 if (!verror && offset + resp->data_len == va.va_size) 3372 resp->eof = TRUE; 3373 else 3374 resp->eof = FALSE; 3375 3376 if (rdma_used) { 3377 if (!rdma_setup_read_data4(args, resp)) { 3378 *cs->statusp = resp->status = NFS4ERR_INVAL; 3379 } 3380 } else { 3381 resp->wlist = NULL; 3382 } 3383 3384 out: 3385 if (in_crit) 3386 nbl_end_crit(vp); 3387 3388 DTRACE_NFSV4_2(op__read__done, struct compound_state *, cs, 3389 READ4res *, resp); 3390 } 3391 3392 static void 3393 rfs4_op_read_free(nfs_resop4 *resop) 3394 { 3395 READ4res *resp = &resop->nfs_resop4_u.opread; 3396 3397 if (resp->status == NFS4_OK && resp->mblk != NULL) { 3398 freemsg(resp->mblk); 3399 resp->mblk = NULL; 3400 resp->data_val = NULL; 3401 resp->data_len = 0; 3402 } 3403 } 3404 3405 static void 3406 rfs4_op_readdir_free(nfs_resop4 * resop) 3407 { 3408 READDIR4res *resp = &resop->nfs_resop4_u.opreaddir; 3409 3410 if (resp->status == NFS4_OK && resp->mblk != NULL) { 3411 freeb(resp->mblk); 3412 resp->mblk = NULL; 3413 resp->data_len = 0; 3414 } 3415 } 3416 3417 3418 /* ARGSUSED */ 3419 static void 3420 rfs4_op_putpubfh(nfs_argop4 *args, nfs_resop4 *resop, struct svc_req *req, 3421 struct compound_state *cs) 3422 { 3423 PUTPUBFH4res *resp = &resop->nfs_resop4_u.opputpubfh; 3424 int error; 3425 vnode_t *vp; 3426 struct exportinfo *exi, *sav_exi; 3427 nfs_fh4_fmt_t *fh_fmtp; 3428 3429 DTRACE_NFSV4_1(op__putpubfh__start, struct compound_state *, cs); 3430 3431 if (cs->vp) { 3432 VN_RELE(cs->vp); 3433 cs->vp = NULL; 3434 } 3435 3436 if (cs->cr) 3437 crfree(cs->cr); 3438 3439 cs->cr = crdup(cs->basecr); 3440 3441 vp = exi_public->exi_vp; 3442 if (vp == NULL) { 3443 *cs->statusp = resp->status = NFS4ERR_SERVERFAULT; 3444 goto out; 3445 } 3446 3447 error = makefh4(&cs->fh, vp, exi_public); 3448 if (error != 0) { 3449 *cs->statusp = resp->status = puterrno4(error); 3450 goto out; 3451 } 3452 sav_exi = cs->exi; 3453 if (exi_public == exi_root) { 3454 /* 3455 * No filesystem is actually shared public, so we default 3456 * to exi_root. In this case, we must check whether root 3457 * is exported. 3458 */ 3459 fh_fmtp = (nfs_fh4_fmt_t *)cs->fh.nfs_fh4_val; 3460 3461 /* 3462 * if root filesystem is exported, the exportinfo struct that we 3463 * should use is what checkexport4 returns, because root_exi is 3464 * actually a mostly empty struct. 3465 */ 3466 exi = checkexport4(&fh_fmtp->fh4_fsid, 3467 (fid_t *)&fh_fmtp->fh4_xlen, NULL); 3468 cs->exi = ((exi != NULL) ? exi : exi_public); 3469 } else { 3470 /* 3471 * it's a properly shared filesystem 3472 */ 3473 cs->exi = exi_public; 3474 } 3475 3476 if (is_system_labeled()) { 3477 bslabel_t *clabel; 3478 3479 ASSERT(req->rq_label != NULL); 3480 clabel = req->rq_label; 3481 DTRACE_PROBE2(tx__rfs4__log__info__opputpubfh__clabel, char *, 3482 "got client label from request(1)", 3483 struct svc_req *, req); 3484 if (!blequal(&l_admin_low->tsl_label, clabel)) { 3485 if (!do_rfs_label_check(clabel, vp, DOMINANCE_CHECK, 3486 cs->exi)) { 3487 *cs->statusp = resp->status = 3488 NFS4ERR_SERVERFAULT; 3489 goto out; 3490 } 3491 } 3492 } 3493 3494 VN_HOLD(vp); 3495 cs->vp = vp; 3496 3497 if ((resp->status = call_checkauth4(cs, req)) != NFS4_OK) { 3498 VN_RELE(cs->vp); 3499 cs->vp = NULL; 3500 cs->exi = sav_exi; 3501 goto out; 3502 } 3503 3504 *cs->statusp = resp->status = NFS4_OK; 3505 out: 3506 DTRACE_NFSV4_2(op__putpubfh__done, struct compound_state *, cs, 3507 PUTPUBFH4res *, resp); 3508 } 3509 3510 /* 3511 * XXX - issue with put*fh operations. Suppose /export/home is exported. 3512 * Suppose an NFS client goes to mount /export/home/joe. If /export, home, 3513 * or joe have restrictive search permissions, then we shouldn't let 3514 * the client get a file handle. This is easy to enforce. However, we 3515 * don't know what security flavor should be used until we resolve the 3516 * path name. Another complication is uid mapping. If root is 3517 * the user, then it will be mapped to the anonymous user by default, 3518 * but we won't know that till we've resolved the path name. And we won't 3519 * know what the anonymous user is. 3520 * Luckily, SECINFO is specified to take a full filename. 3521 * So what we will have to in rfs4_op_lookup is check that flavor of 3522 * the target object matches that of the request, and if root was the 3523 * caller, check for the root= and anon= options, and if necessary, 3524 * repeat the lookup using the right cred_t. But that's not done yet. 3525 */ 3526 /* ARGSUSED */ 3527 static void 3528 rfs4_op_putfh(nfs_argop4 *argop, nfs_resop4 *resop, struct svc_req *req, 3529 struct compound_state *cs) 3530 { 3531 PUTFH4args *args = &argop->nfs_argop4_u.opputfh; 3532 PUTFH4res *resp = &resop->nfs_resop4_u.opputfh; 3533 nfs_fh4_fmt_t *fh_fmtp; 3534 3535 DTRACE_NFSV4_2(op__putfh__start, struct compound_state *, cs, 3536 PUTFH4args *, args); 3537 3538 if (cs->vp) { 3539 VN_RELE(cs->vp); 3540 cs->vp = NULL; 3541 } 3542 3543 if (cs->cr) { 3544 crfree(cs->cr); 3545 cs->cr = NULL; 3546 } 3547 3548 3549 if (args->object.nfs_fh4_len < NFS_FH4_LEN) { 3550 *cs->statusp = resp->status = NFS4ERR_BADHANDLE; 3551 goto out; 3552 } 3553 3554 fh_fmtp = (nfs_fh4_fmt_t *)args->object.nfs_fh4_val; 3555 cs->exi = checkexport4(&fh_fmtp->fh4_fsid, (fid_t *)&fh_fmtp->fh4_xlen, 3556 NULL); 3557 3558 if (cs->exi == NULL) { 3559 *cs->statusp = resp->status = NFS4ERR_STALE; 3560 goto out; 3561 } 3562 3563 cs->cr = crdup(cs->basecr); 3564 3565 ASSERT(cs->cr != NULL); 3566 3567 if (! (cs->vp = nfs4_fhtovp(&args->object, cs->exi, &resp->status))) { 3568 *cs->statusp = resp->status; 3569 goto out; 3570 } 3571 3572 if ((resp->status = call_checkauth4(cs, req)) != NFS4_OK) { 3573 VN_RELE(cs->vp); 3574 cs->vp = NULL; 3575 goto out; 3576 } 3577 3578 nfs_fh4_copy(&args->object, &cs->fh); 3579 *cs->statusp = resp->status = NFS4_OK; 3580 cs->deleg = FALSE; 3581 3582 out: 3583 DTRACE_NFSV4_2(op__putfh__done, struct compound_state *, cs, 3584 PUTFH4res *, resp); 3585 } 3586 3587 /* ARGSUSED */ 3588 static void 3589 rfs4_op_putrootfh(nfs_argop4 *argop, nfs_resop4 *resop, struct svc_req *req, 3590 struct compound_state *cs) 3591 { 3592 PUTROOTFH4res *resp = &resop->nfs_resop4_u.opputrootfh; 3593 int error; 3594 fid_t fid; 3595 struct exportinfo *exi, *sav_exi; 3596 3597 DTRACE_NFSV4_1(op__putrootfh__start, struct compound_state *, cs); 3598 3599 if (cs->vp) { 3600 VN_RELE(cs->vp); 3601 cs->vp = NULL; 3602 } 3603 3604 if (cs->cr) 3605 crfree(cs->cr); 3606 3607 cs->cr = crdup(cs->basecr); 3608 3609 /* 3610 * Using rootdir, the system root vnode, 3611 * get its fid. 3612 */ 3613 bzero(&fid, sizeof (fid)); 3614 fid.fid_len = MAXFIDSZ; 3615 error = vop_fid_pseudo(rootdir, &fid); 3616 if (error != 0) { 3617 *cs->statusp = resp->status = puterrno4(error); 3618 goto out; 3619 } 3620 3621 /* 3622 * Then use the root fsid & fid it to find out if it's exported 3623 * 3624 * If the server root isn't exported directly, then 3625 * it should at least be a pseudo export based on 3626 * one or more exports further down in the server's 3627 * file tree. 3628 */ 3629 exi = checkexport4(&rootdir->v_vfsp->vfs_fsid, &fid, NULL); 3630 if (exi == NULL || exi->exi_export.ex_flags & EX_PUBLIC) { 3631 NFS4_DEBUG(rfs4_debug, 3632 (CE_WARN, "rfs4_op_putrootfh: export check failure")); 3633 *cs->statusp = resp->status = NFS4ERR_SERVERFAULT; 3634 goto out; 3635 } 3636 3637 /* 3638 * Now make a filehandle based on the root 3639 * export and root vnode. 3640 */ 3641 error = makefh4(&cs->fh, rootdir, exi); 3642 if (error != 0) { 3643 *cs->statusp = resp->status = puterrno4(error); 3644 goto out; 3645 } 3646 3647 sav_exi = cs->exi; 3648 cs->exi = exi; 3649 3650 VN_HOLD(rootdir); 3651 cs->vp = rootdir; 3652 3653 if ((resp->status = call_checkauth4(cs, req)) != NFS4_OK) { 3654 VN_RELE(rootdir); 3655 cs->vp = NULL; 3656 cs->exi = sav_exi; 3657 goto out; 3658 } 3659 3660 *cs->statusp = resp->status = NFS4_OK; 3661 cs->deleg = FALSE; 3662 out: 3663 DTRACE_NFSV4_2(op__putrootfh__done, struct compound_state *, cs, 3664 PUTROOTFH4res *, resp); 3665 } 3666 3667 /* 3668 * A directory entry is a valid nfsv4 entry if 3669 * - it has a non-zero ino 3670 * - it is not a dot or dotdot name 3671 * - it is visible in a pseudo export or in a real export that can 3672 * only have a limited view. 3673 */ 3674 static bool_t 3675 valid_nfs4_entry(struct exportinfo *exi, struct dirent64 *dp, 3676 int *expseudo, int check_visible) 3677 { 3678 if (dp->d_ino == 0 || NFS_IS_DOTNAME(dp->d_name)) { 3679 *expseudo = 0; 3680 return (FALSE); 3681 } 3682 3683 if (! check_visible) { 3684 *expseudo = 0; 3685 return (TRUE); 3686 } 3687 3688 return (nfs_visible_inode(exi, dp->d_ino, expseudo)); 3689 } 3690 3691 /* 3692 * set_rdattr_params sets up the variables used to manage what information 3693 * to get for each directory entry. 3694 */ 3695 static nfsstat4 3696 set_rdattr_params(struct nfs4_svgetit_arg *sargp, 3697 bitmap4 attrs, bool_t *need_to_lookup) 3698 { 3699 uint_t va_mask; 3700 nfsstat4 status; 3701 bitmap4 objbits; 3702 3703 status = bitmap4_to_attrmask(attrs, sargp); 3704 if (status != NFS4_OK) { 3705 /* 3706 * could not even figure attr mask 3707 */ 3708 return (status); 3709 } 3710 va_mask = sargp->vap->va_mask; 3711 3712 /* 3713 * dirent's d_ino is always correct value for mounted_on_fileid. 3714 * mntdfid_set is set once here, but mounted_on_fileid is 3715 * set in main dirent processing loop for each dirent. 3716 * The mntdfid_set is a simple optimization that lets the 3717 * server attr code avoid work when caller is readdir. 3718 */ 3719 sargp->mntdfid_set = TRUE; 3720 3721 /* 3722 * Lookup entry only if client asked for any of the following: 3723 * a) vattr attrs 3724 * b) vfs attrs 3725 * c) attrs w/per-object scope requested (change, filehandle, etc) 3726 * other than mounted_on_fileid (which we can take from dirent) 3727 */ 3728 objbits = attrs ? attrs & NFS4_VP_ATTR_MASK : 0; 3729 3730 if (va_mask || sargp->sbp || (objbits & ~FATTR4_MOUNTED_ON_FILEID_MASK)) 3731 *need_to_lookup = TRUE; 3732 else 3733 *need_to_lookup = FALSE; 3734 3735 if (sargp->sbp == NULL) 3736 return (NFS4_OK); 3737 3738 /* 3739 * If filesystem attrs are requested, get them now from the 3740 * directory vp, as most entries will have same filesystem. The only 3741 * exception are mounted over entries but we handle 3742 * those as we go (XXX mounted over detection not yet implemented). 3743 */ 3744 sargp->vap->va_mask = 0; /* to avoid VOP_GETATTR */ 3745 status = bitmap4_get_sysattrs(sargp); 3746 sargp->vap->va_mask = va_mask; 3747 3748 if ((status != NFS4_OK) && sargp->rdattr_error_req) { 3749 /* 3750 * Failed to get filesystem attributes. 3751 * Return a rdattr_error for each entry, but don't fail. 3752 * However, don't get any obj-dependent attrs. 3753 */ 3754 sargp->rdattr_error = status; /* for rdattr_error */ 3755 *need_to_lookup = FALSE; 3756 /* 3757 * At least get fileid for regular readdir output 3758 */ 3759 sargp->vap->va_mask &= AT_NODEID; 3760 status = NFS4_OK; 3761 } 3762 3763 return (status); 3764 } 3765 3766 /* 3767 * readlink: args: CURRENT_FH. 3768 * res: status. If success - CURRENT_FH unchanged, return linktext. 3769 */ 3770 3771 /* ARGSUSED */ 3772 static void 3773 rfs4_op_readlink(nfs_argop4 *argop, nfs_resop4 *resop, struct svc_req *req, 3774 struct compound_state *cs) 3775 { 3776 READLINK4res *resp = &resop->nfs_resop4_u.opreadlink; 3777 int error; 3778 vnode_t *vp; 3779 struct iovec iov; 3780 struct vattr va; 3781 struct uio uio; 3782 char *data; 3783 struct sockaddr *ca; 3784 char *name = NULL; 3785 int is_referral; 3786 3787 DTRACE_NFSV4_1(op__readlink__start, struct compound_state *, cs); 3788 3789 /* CURRENT_FH: directory */ 3790 vp = cs->vp; 3791 if (vp == NULL) { 3792 *cs->statusp = resp->status = NFS4ERR_NOFILEHANDLE; 3793 goto out; 3794 } 3795 3796 if (cs->access == CS_ACCESS_DENIED) { 3797 *cs->statusp = resp->status = NFS4ERR_ACCESS; 3798 goto out; 3799 } 3800 3801 /* Is it a referral? */ 3802 if (vn_is_nfs_reparse(vp, cs->cr) && client_is_downrev(req)) { 3803 3804 is_referral = 1; 3805 3806 } else { 3807 3808 is_referral = 0; 3809 3810 if (vp->v_type == VDIR) { 3811 *cs->statusp = resp->status = NFS4ERR_ISDIR; 3812 goto out; 3813 } 3814 3815 if (vp->v_type != VLNK) { 3816 *cs->statusp = resp->status = NFS4ERR_INVAL; 3817 goto out; 3818 } 3819 3820 } 3821 3822 va.va_mask = AT_MODE; 3823 error = VOP_GETATTR(vp, &va, 0, cs->cr, NULL); 3824 if (error) { 3825 *cs->statusp = resp->status = puterrno4(error); 3826 goto out; 3827 } 3828 3829 if (MANDLOCK(vp, va.va_mode)) { 3830 *cs->statusp = resp->status = NFS4ERR_ACCESS; 3831 goto out; 3832 } 3833 3834 data = kmem_alloc(MAXPATHLEN + 1, KM_SLEEP); 3835 3836 if (is_referral) { 3837 char *s; 3838 size_t strsz; 3839 3840 /* Get an artificial symlink based on a referral */ 3841 s = build_symlink(vp, cs->cr, &strsz); 3842 global_svstat_ptr[4][NFS_REFERLINKS].value.ui64++; 3843 DTRACE_PROBE2(nfs4serv__func__referral__reflink, 3844 vnode_t *, vp, char *, s); 3845 if (s == NULL) 3846 error = EINVAL; 3847 else { 3848 error = 0; 3849 (void) strlcpy(data, s, MAXPATHLEN + 1); 3850 kmem_free(s, strsz); 3851 } 3852 3853 } else { 3854 3855 iov.iov_base = data; 3856 iov.iov_len = MAXPATHLEN; 3857 uio.uio_iov = &iov; 3858 uio.uio_iovcnt = 1; 3859 uio.uio_segflg = UIO_SYSSPACE; 3860 uio.uio_extflg = UIO_COPY_CACHED; 3861 uio.uio_loffset = 0; 3862 uio.uio_resid = MAXPATHLEN; 3863 3864 error = VOP_READLINK(vp, &uio, cs->cr, NULL); 3865 3866 if (!error) 3867 *(data + MAXPATHLEN - uio.uio_resid) = '\0'; 3868 } 3869 3870 if (error) { 3871 kmem_free((caddr_t)data, (uint_t)MAXPATHLEN + 1); 3872 *cs->statusp = resp->status = puterrno4(error); 3873 goto out; 3874 } 3875 3876 ca = (struct sockaddr *)svc_getrpccaller(req->rq_xprt)->buf; 3877 name = nfscmd_convname(ca, cs->exi, data, NFSCMD_CONV_OUTBOUND, 3878 MAXPATHLEN + 1); 3879 3880 if (name == NULL) { 3881 /* 3882 * Even though the conversion failed, we return 3883 * something. We just don't translate it. 3884 */ 3885 name = data; 3886 } 3887 3888 /* 3889 * treat link name as data 3890 */ 3891 (void) str_to_utf8(name, &resp->link); 3892 3893 if (name != data) 3894 kmem_free(name, MAXPATHLEN + 1); 3895 kmem_free((caddr_t)data, (uint_t)MAXPATHLEN + 1); 3896 *cs->statusp = resp->status = NFS4_OK; 3897 3898 out: 3899 DTRACE_NFSV4_2(op__readlink__done, struct compound_state *, cs, 3900 READLINK4res *, resp); 3901 } 3902 3903 static void 3904 rfs4_op_readlink_free(nfs_resop4 *resop) 3905 { 3906 READLINK4res *resp = &resop->nfs_resop4_u.opreadlink; 3907 utf8string *symlink = &resp->link; 3908 3909 if (symlink->utf8string_val) { 3910 UTF8STRING_FREE(*symlink) 3911 } 3912 } 3913 3914 /* 3915 * release_lockowner: 3916 * Release any state associated with the supplied 3917 * lockowner. Note if any lo_state is holding locks we will not 3918 * rele that lo_state and thus the lockowner will not be destroyed. 3919 * A client using lock after the lock owner stateid has been released 3920 * will suffer the consequence of NFS4ERR_BAD_STATEID and would have 3921 * to reissue the lock with new_lock_owner set to TRUE. 3922 * args: lock_owner 3923 * res: status 3924 */ 3925 /* ARGSUSED */ 3926 static void 3927 rfs4_op_release_lockowner(nfs_argop4 *argop, nfs_resop4 *resop, 3928 struct svc_req *req, struct compound_state *cs) 3929 { 3930 RELEASE_LOCKOWNER4args *ap = &argop->nfs_argop4_u.oprelease_lockowner; 3931 RELEASE_LOCKOWNER4res *resp = &resop->nfs_resop4_u.oprelease_lockowner; 3932 rfs4_lockowner_t *lo; 3933 rfs4_openowner_t *oo; 3934 rfs4_state_t *sp; 3935 rfs4_lo_state_t *lsp; 3936 rfs4_client_t *cp; 3937 bool_t create = FALSE; 3938 locklist_t *llist; 3939 sysid_t sysid; 3940 3941 DTRACE_NFSV4_2(op__release__lockowner__start, struct compound_state *, 3942 cs, RELEASE_LOCKOWNER4args *, ap); 3943 3944 /* Make sure there is a clientid around for this request */ 3945 cp = rfs4_findclient_by_id(ap->lock_owner.clientid, FALSE); 3946 3947 if (cp == NULL) { 3948 *cs->statusp = resp->status = 3949 rfs4_check_clientid(&ap->lock_owner.clientid, 0); 3950 goto out; 3951 } 3952 rfs4_client_rele(cp); 3953 3954 lo = rfs4_findlockowner(&ap->lock_owner, &create); 3955 if (lo == NULL) { 3956 *cs->statusp = resp->status = NFS4_OK; 3957 goto out; 3958 } 3959 ASSERT(lo->rl_client != NULL); 3960 3961 /* 3962 * Check for EXPIRED client. If so will reap state with in a lease 3963 * period or on next set_clientid_confirm step 3964 */ 3965 if (rfs4_lease_expired(lo->rl_client)) { 3966 rfs4_lockowner_rele(lo); 3967 *cs->statusp = resp->status = NFS4ERR_EXPIRED; 3968 goto out; 3969 } 3970 3971 /* 3972 * If no sysid has been assigned, then no locks exist; just return. 3973 */ 3974 rfs4_dbe_lock(lo->rl_client->rc_dbe); 3975 if (lo->rl_client->rc_sysidt == LM_NOSYSID) { 3976 rfs4_lockowner_rele(lo); 3977 rfs4_dbe_unlock(lo->rl_client->rc_dbe); 3978 goto out; 3979 } 3980 3981 sysid = lo->rl_client->rc_sysidt; 3982 rfs4_dbe_unlock(lo->rl_client->rc_dbe); 3983 3984 /* 3985 * Mark the lockowner invalid. 3986 */ 3987 rfs4_dbe_hide(lo->rl_dbe); 3988 3989 /* 3990 * sysid-pid pair should now not be used since the lockowner is 3991 * invalid. If the client were to instantiate the lockowner again 3992 * it would be assigned a new pid. Thus we can get the list of 3993 * current locks. 3994 */ 3995 3996 llist = flk_get_active_locks(sysid, lo->rl_pid); 3997 /* If we are still holding locks fail */ 3998 if (llist != NULL) { 3999 4000 *cs->statusp = resp->status = NFS4ERR_LOCKS_HELD; 4001 4002 flk_free_locklist(llist); 4003 /* 4004 * We need to unhide the lockowner so the client can 4005 * try it again. The bad thing here is if the client 4006 * has a logic error that took it here in the first place 4007 * he probably has lost accounting of the locks that it 4008 * is holding. So we may have dangling state until the 4009 * open owner state is reaped via close. One scenario 4010 * that could possibly occur is that the client has 4011 * sent the unlock request(s) in separate threads 4012 * and has not waited for the replies before sending the 4013 * RELEASE_LOCKOWNER request. Presumably, it would expect 4014 * and deal appropriately with NFS4ERR_LOCKS_HELD, by 4015 * reissuing the request. 4016 */ 4017 rfs4_dbe_unhide(lo->rl_dbe); 4018 rfs4_lockowner_rele(lo); 4019 goto out; 4020 } 4021 4022 /* 4023 * For the corresponding client we need to check each open 4024 * owner for any opens that have lockowner state associated 4025 * with this lockowner. 4026 */ 4027 4028 rfs4_dbe_lock(lo->rl_client->rc_dbe); 4029 for (oo = list_head(&lo->rl_client->rc_openownerlist); oo != NULL; 4030 oo = list_next(&lo->rl_client->rc_openownerlist, oo)) { 4031 4032 rfs4_dbe_lock(oo->ro_dbe); 4033 for (sp = list_head(&oo->ro_statelist); sp != NULL; 4034 sp = list_next(&oo->ro_statelist, sp)) { 4035 4036 rfs4_dbe_lock(sp->rs_dbe); 4037 for (lsp = list_head(&sp->rs_lostatelist); 4038 lsp != NULL; 4039 lsp = list_next(&sp->rs_lostatelist, lsp)) { 4040 if (lsp->rls_locker == lo) { 4041 rfs4_dbe_lock(lsp->rls_dbe); 4042 rfs4_dbe_invalidate(lsp->rls_dbe); 4043 rfs4_dbe_unlock(lsp->rls_dbe); 4044 } 4045 } 4046 rfs4_dbe_unlock(sp->rs_dbe); 4047 } 4048 rfs4_dbe_unlock(oo->ro_dbe); 4049 } 4050 rfs4_dbe_unlock(lo->rl_client->rc_dbe); 4051 4052 rfs4_lockowner_rele(lo); 4053 4054 *cs->statusp = resp->status = NFS4_OK; 4055 4056 out: 4057 DTRACE_NFSV4_2(op__release__lockowner__done, struct compound_state *, 4058 cs, RELEASE_LOCKOWNER4res *, resp); 4059 } 4060 4061 /* 4062 * short utility function to lookup a file and recall the delegation 4063 */ 4064 static rfs4_file_t * 4065 rfs4_lookup_and_findfile(vnode_t *dvp, char *nm, vnode_t **vpp, 4066 int *lkup_error, cred_t *cr) 4067 { 4068 vnode_t *vp; 4069 rfs4_file_t *fp = NULL; 4070 bool_t fcreate = FALSE; 4071 int error; 4072 4073 if (vpp) 4074 *vpp = NULL; 4075 4076 if ((error = VOP_LOOKUP(dvp, nm, &vp, NULL, 0, NULL, cr, NULL, NULL, 4077 NULL)) == 0) { 4078 if (vp->v_type == VREG) 4079 fp = rfs4_findfile(vp, NULL, &fcreate); 4080 if (vpp) 4081 *vpp = vp; 4082 else 4083 VN_RELE(vp); 4084 } 4085 4086 if (lkup_error) 4087 *lkup_error = error; 4088 4089 return (fp); 4090 } 4091 4092 /* 4093 * remove: args: CURRENT_FH: directory; name. 4094 * res: status. If success - CURRENT_FH unchanged, return change_info 4095 * for directory. 4096 */ 4097 /* ARGSUSED */ 4098 static void 4099 rfs4_op_remove(nfs_argop4 *argop, nfs_resop4 *resop, struct svc_req *req, 4100 struct compound_state *cs) 4101 { 4102 REMOVE4args *args = &argop->nfs_argop4_u.opremove; 4103 REMOVE4res *resp = &resop->nfs_resop4_u.opremove; 4104 int error; 4105 vnode_t *dvp, *vp; 4106 struct vattr bdva, idva, adva; 4107 char *nm; 4108 uint_t len; 4109 rfs4_file_t *fp; 4110 int in_crit = 0; 4111 bslabel_t *clabel; 4112 struct sockaddr *ca; 4113 char *name = NULL; 4114 nfsstat4 status; 4115 4116 DTRACE_NFSV4_2(op__remove__start, struct compound_state *, cs, 4117 REMOVE4args *, args); 4118 4119 /* CURRENT_FH: directory */ 4120 dvp = cs->vp; 4121 if (dvp == NULL) { 4122 *cs->statusp = resp->status = NFS4ERR_NOFILEHANDLE; 4123 goto out; 4124 } 4125 4126 if (cs->access == CS_ACCESS_DENIED) { 4127 *cs->statusp = resp->status = NFS4ERR_ACCESS; 4128 goto out; 4129 } 4130 4131 /* 4132 * If there is an unshared filesystem mounted on this vnode, 4133 * Do not allow to remove anything in this directory. 4134 */ 4135 if (vn_ismntpt(dvp)) { 4136 *cs->statusp = resp->status = NFS4ERR_ACCESS; 4137 goto out; 4138 } 4139 4140 if (dvp->v_type != VDIR) { 4141 *cs->statusp = resp->status = NFS4ERR_NOTDIR; 4142 goto out; 4143 } 4144 4145 status = utf8_dir_verify(&args->target); 4146 if (status != NFS4_OK) { 4147 *cs->statusp = resp->status = status; 4148 goto out; 4149 } 4150 4151 /* 4152 * Lookup the file so that we can check if it's a directory 4153 */ 4154 nm = utf8_to_fn(&args->target, &len, NULL); 4155 if (nm == NULL) { 4156 *cs->statusp = resp->status = NFS4ERR_INVAL; 4157 goto out; 4158 } 4159 4160 if (len > MAXNAMELEN) { 4161 *cs->statusp = resp->status = NFS4ERR_NAMETOOLONG; 4162 kmem_free(nm, len); 4163 goto out; 4164 } 4165 4166 if (rdonly4(cs->exi, cs->vp, req)) { 4167 *cs->statusp = resp->status = NFS4ERR_ROFS; 4168 kmem_free(nm, len); 4169 goto out; 4170 } 4171 4172 ca = (struct sockaddr *)svc_getrpccaller(req->rq_xprt)->buf; 4173 name = nfscmd_convname(ca, cs->exi, nm, NFSCMD_CONV_INBOUND, 4174 MAXPATHLEN + 1); 4175 4176 if (name == NULL) { 4177 *cs->statusp = resp->status = NFS4ERR_INVAL; 4178 kmem_free(nm, len); 4179 goto out; 4180 } 4181 4182 /* 4183 * Lookup the file to determine type and while we are see if 4184 * there is a file struct around and check for delegation. 4185 * We don't need to acquire va_seq before this lookup, if 4186 * it causes an update, cinfo.before will not match, which will 4187 * trigger a cache flush even if atomic is TRUE. 4188 */ 4189 if (fp = rfs4_lookup_and_findfile(dvp, name, &vp, &error, cs->cr)) { 4190 if (rfs4_check_delegated_byfp(FWRITE, fp, TRUE, TRUE, TRUE, 4191 NULL)) { 4192 VN_RELE(vp); 4193 rfs4_file_rele(fp); 4194 *cs->statusp = resp->status = NFS4ERR_DELAY; 4195 if (nm != name) 4196 kmem_free(name, MAXPATHLEN + 1); 4197 kmem_free(nm, len); 4198 goto out; 4199 } 4200 } 4201 4202 /* Didn't find anything to remove */ 4203 if (vp == NULL) { 4204 *cs->statusp = resp->status = error; 4205 if (nm != name) 4206 kmem_free(name, MAXPATHLEN + 1); 4207 kmem_free(nm, len); 4208 goto out; 4209 } 4210 4211 if (nbl_need_check(vp)) { 4212 nbl_start_crit(vp, RW_READER); 4213 in_crit = 1; 4214 if (nbl_conflict(vp, NBL_REMOVE, 0, 0, 0, NULL)) { 4215 *cs->statusp = resp->status = NFS4ERR_FILE_OPEN; 4216 if (nm != name) 4217 kmem_free(name, MAXPATHLEN + 1); 4218 kmem_free(nm, len); 4219 nbl_end_crit(vp); 4220 VN_RELE(vp); 4221 if (fp) { 4222 rfs4_clear_dont_grant(fp); 4223 rfs4_file_rele(fp); 4224 } 4225 goto out; 4226 } 4227 } 4228 4229 /* check label before allowing removal */ 4230 if (is_system_labeled()) { 4231 ASSERT(req->rq_label != NULL); 4232 clabel = req->rq_label; 4233 DTRACE_PROBE2(tx__rfs4__log__info__opremove__clabel, char *, 4234 "got client label from request(1)", 4235 struct svc_req *, req); 4236 if (!blequal(&l_admin_low->tsl_label, clabel)) { 4237 if (!do_rfs_label_check(clabel, vp, EQUALITY_CHECK, 4238 cs->exi)) { 4239 *cs->statusp = resp->status = NFS4ERR_ACCESS; 4240 if (name != nm) 4241 kmem_free(name, MAXPATHLEN + 1); 4242 kmem_free(nm, len); 4243 if (in_crit) 4244 nbl_end_crit(vp); 4245 VN_RELE(vp); 4246 if (fp) { 4247 rfs4_clear_dont_grant(fp); 4248 rfs4_file_rele(fp); 4249 } 4250 goto out; 4251 } 4252 } 4253 } 4254 4255 /* Get dir "before" change value */ 4256 bdva.va_mask = AT_CTIME|AT_SEQ; 4257 error = VOP_GETATTR(dvp, &bdva, 0, cs->cr, NULL); 4258 if (error) { 4259 *cs->statusp = resp->status = puterrno4(error); 4260 if (nm != name) 4261 kmem_free(name, MAXPATHLEN + 1); 4262 kmem_free(nm, len); 4263 if (in_crit) 4264 nbl_end_crit(vp); 4265 VN_RELE(vp); 4266 if (fp) { 4267 rfs4_clear_dont_grant(fp); 4268 rfs4_file_rele(fp); 4269 } 4270 goto out; 4271 } 4272 NFS4_SET_FATTR4_CHANGE(resp->cinfo.before, bdva.va_ctime) 4273 4274 /* Actually do the REMOVE operation */ 4275 if (vp->v_type == VDIR) { 4276 /* 4277 * Can't remove a directory that has a mounted-on filesystem. 4278 */ 4279 if (vn_ismntpt(vp)) { 4280 error = EACCES; 4281 } else { 4282 /* 4283 * System V defines rmdir to return EEXIST, 4284 * not ENOTEMPTY, if the directory is not 4285 * empty. A System V NFS server needs to map 4286 * NFS4ERR_EXIST to NFS4ERR_NOTEMPTY to 4287 * transmit over the wire. 4288 */ 4289 if ((error = VOP_RMDIR(dvp, name, rootdir, cs->cr, 4290 NULL, 0)) == EEXIST) 4291 error = ENOTEMPTY; 4292 } 4293 } else { 4294 if ((error = VOP_REMOVE(dvp, name, cs->cr, NULL, 0)) == 0 && 4295 fp != NULL) { 4296 struct vattr va; 4297 vnode_t *tvp; 4298 4299 rfs4_dbe_lock(fp->rf_dbe); 4300 tvp = fp->rf_vp; 4301 if (tvp) 4302 VN_HOLD(tvp); 4303 rfs4_dbe_unlock(fp->rf_dbe); 4304 4305 if (tvp) { 4306 /* 4307 * This is va_seq safe because we are not 4308 * manipulating dvp. 4309 */ 4310 va.va_mask = AT_NLINK; 4311 if (!VOP_GETATTR(tvp, &va, 0, cs->cr, NULL) && 4312 va.va_nlink == 0) { 4313 /* Remove state on file remove */ 4314 if (in_crit) { 4315 nbl_end_crit(vp); 4316 in_crit = 0; 4317 } 4318 rfs4_close_all_state(fp); 4319 } 4320 VN_RELE(tvp); 4321 } 4322 } 4323 } 4324 4325 if (in_crit) 4326 nbl_end_crit(vp); 4327 VN_RELE(vp); 4328 4329 if (fp) { 4330 rfs4_clear_dont_grant(fp); 4331 rfs4_file_rele(fp); 4332 } 4333 if (nm != name) 4334 kmem_free(name, MAXPATHLEN + 1); 4335 kmem_free(nm, len); 4336 4337 if (error) { 4338 *cs->statusp = resp->status = puterrno4(error); 4339 goto out; 4340 } 4341 4342 /* 4343 * Get the initial "after" sequence number, if it fails, set to zero 4344 */ 4345 idva.va_mask = AT_SEQ; 4346 if (VOP_GETATTR(dvp, &idva, 0, cs->cr, NULL)) 4347 idva.va_seq = 0; 4348 4349 /* 4350 * Force modified data and metadata out to stable storage. 4351 */ 4352 (void) VOP_FSYNC(dvp, 0, cs->cr, NULL); 4353 4354 /* 4355 * Get "after" change value, if it fails, simply return the 4356 * before value. 4357 */ 4358 adva.va_mask = AT_CTIME|AT_SEQ; 4359 if (VOP_GETATTR(dvp, &adva, 0, cs->cr, NULL)) { 4360 adva.va_ctime = bdva.va_ctime; 4361 adva.va_seq = 0; 4362 } 4363 4364 NFS4_SET_FATTR4_CHANGE(resp->cinfo.after, adva.va_ctime) 4365 4366 /* 4367 * The cinfo.atomic = TRUE only if we have 4368 * non-zero va_seq's, and it has incremented by exactly one 4369 * during the VOP_REMOVE/RMDIR and it didn't change during 4370 * the VOP_FSYNC. 4371 */ 4372 if (bdva.va_seq && idva.va_seq && adva.va_seq && 4373 idva.va_seq == (bdva.va_seq + 1) && idva.va_seq == adva.va_seq) 4374 resp->cinfo.atomic = TRUE; 4375 else 4376 resp->cinfo.atomic = FALSE; 4377 4378 *cs->statusp = resp->status = NFS4_OK; 4379 4380 out: 4381 DTRACE_NFSV4_2(op__remove__done, struct compound_state *, cs, 4382 REMOVE4res *, resp); 4383 } 4384 4385 /* 4386 * rename: args: SAVED_FH: from directory, CURRENT_FH: target directory, 4387 * oldname and newname. 4388 * res: status. If success - CURRENT_FH unchanged, return change_info 4389 * for both from and target directories. 4390 */ 4391 /* ARGSUSED */ 4392 static void 4393 rfs4_op_rename(nfs_argop4 *argop, nfs_resop4 *resop, struct svc_req *req, 4394 struct compound_state *cs) 4395 { 4396 RENAME4args *args = &argop->nfs_argop4_u.oprename; 4397 RENAME4res *resp = &resop->nfs_resop4_u.oprename; 4398 int error; 4399 vnode_t *odvp; 4400 vnode_t *ndvp; 4401 vnode_t *srcvp, *targvp; 4402 struct vattr obdva, oidva, oadva; 4403 struct vattr nbdva, nidva, nadva; 4404 char *onm, *nnm; 4405 uint_t olen, nlen; 4406 rfs4_file_t *fp, *sfp; 4407 int in_crit_src, in_crit_targ; 4408 int fp_rele_grant_hold, sfp_rele_grant_hold; 4409 bslabel_t *clabel; 4410 struct sockaddr *ca; 4411 char *converted_onm = NULL; 4412 char *converted_nnm = NULL; 4413 nfsstat4 status; 4414 4415 DTRACE_NFSV4_2(op__rename__start, struct compound_state *, cs, 4416 RENAME4args *, args); 4417 4418 fp = sfp = NULL; 4419 srcvp = targvp = NULL; 4420 in_crit_src = in_crit_targ = 0; 4421 fp_rele_grant_hold = sfp_rele_grant_hold = 0; 4422 4423 /* CURRENT_FH: target directory */ 4424 ndvp = cs->vp; 4425 if (ndvp == NULL) { 4426 *cs->statusp = resp->status = NFS4ERR_NOFILEHANDLE; 4427 goto out; 4428 } 4429 4430 /* SAVED_FH: from directory */ 4431 odvp = cs->saved_vp; 4432 if (odvp == NULL) { 4433 *cs->statusp = resp->status = NFS4ERR_NOFILEHANDLE; 4434 goto out; 4435 } 4436 4437 if (cs->access == CS_ACCESS_DENIED) { 4438 *cs->statusp = resp->status = NFS4ERR_ACCESS; 4439 goto out; 4440 } 4441 4442 /* 4443 * If there is an unshared filesystem mounted on this vnode, 4444 * do not allow to rename objects in this directory. 4445 */ 4446 if (vn_ismntpt(odvp)) { 4447 *cs->statusp = resp->status = NFS4ERR_ACCESS; 4448 goto out; 4449 } 4450 4451 /* 4452 * If there is an unshared filesystem mounted on this vnode, 4453 * do not allow to rename to this directory. 4454 */ 4455 if (vn_ismntpt(ndvp)) { 4456 *cs->statusp = resp->status = NFS4ERR_ACCESS; 4457 goto out; 4458 } 4459 4460 if (odvp->v_type != VDIR || ndvp->v_type != VDIR) { 4461 *cs->statusp = resp->status = NFS4ERR_NOTDIR; 4462 goto out; 4463 } 4464 4465 if (cs->saved_exi != cs->exi) { 4466 *cs->statusp = resp->status = NFS4ERR_XDEV; 4467 goto out; 4468 } 4469 4470 status = utf8_dir_verify(&args->oldname); 4471 if (status != NFS4_OK) { 4472 *cs->statusp = resp->status = status; 4473 goto out; 4474 } 4475 4476 status = utf8_dir_verify(&args->newname); 4477 if (status != NFS4_OK) { 4478 *cs->statusp = resp->status = status; 4479 goto out; 4480 } 4481 4482 onm = utf8_to_fn(&args->oldname, &olen, NULL); 4483 if (onm == NULL) { 4484 *cs->statusp = resp->status = NFS4ERR_INVAL; 4485 goto out; 4486 } 4487 ca = (struct sockaddr *)svc_getrpccaller(req->rq_xprt)->buf; 4488 nlen = MAXPATHLEN + 1; 4489 converted_onm = nfscmd_convname(ca, cs->exi, onm, NFSCMD_CONV_INBOUND, 4490 nlen); 4491 4492 if (converted_onm == NULL) { 4493 *cs->statusp = resp->status = NFS4ERR_INVAL; 4494 kmem_free(onm, olen); 4495 goto out; 4496 } 4497 4498 nnm = utf8_to_fn(&args->newname, &nlen, NULL); 4499 if (nnm == NULL) { 4500 *cs->statusp = resp->status = NFS4ERR_INVAL; 4501 if (onm != converted_onm) 4502 kmem_free(converted_onm, MAXPATHLEN + 1); 4503 kmem_free(onm, olen); 4504 goto out; 4505 } 4506 converted_nnm = nfscmd_convname(ca, cs->exi, nnm, NFSCMD_CONV_INBOUND, 4507 MAXPATHLEN + 1); 4508 4509 if (converted_nnm == NULL) { 4510 *cs->statusp = resp->status = NFS4ERR_INVAL; 4511 kmem_free(nnm, nlen); 4512 nnm = NULL; 4513 if (onm != converted_onm) 4514 kmem_free(converted_onm, MAXPATHLEN + 1); 4515 kmem_free(onm, olen); 4516 goto out; 4517 } 4518 4519 4520 if (olen > MAXNAMELEN || nlen > MAXNAMELEN) { 4521 *cs->statusp = resp->status = NFS4ERR_NAMETOOLONG; 4522 kmem_free(onm, olen); 4523 kmem_free(nnm, nlen); 4524 goto out; 4525 } 4526 4527 4528 if (rdonly4(cs->exi, cs->vp, req)) { 4529 *cs->statusp = resp->status = NFS4ERR_ROFS; 4530 if (onm != converted_onm) 4531 kmem_free(converted_onm, MAXPATHLEN + 1); 4532 kmem_free(onm, olen); 4533 if (nnm != converted_nnm) 4534 kmem_free(converted_nnm, MAXPATHLEN + 1); 4535 kmem_free(nnm, nlen); 4536 goto out; 4537 } 4538 4539 /* check label of the target dir */ 4540 if (is_system_labeled()) { 4541 ASSERT(req->rq_label != NULL); 4542 clabel = req->rq_label; 4543 DTRACE_PROBE2(tx__rfs4__log__info__oprename__clabel, char *, 4544 "got client label from request(1)", 4545 struct svc_req *, req); 4546 if (!blequal(&l_admin_low->tsl_label, clabel)) { 4547 if (!do_rfs_label_check(clabel, ndvp, 4548 EQUALITY_CHECK, cs->exi)) { 4549 *cs->statusp = resp->status = NFS4ERR_ACCESS; 4550 goto err_out; 4551 } 4552 } 4553 } 4554 4555 /* 4556 * Is the source a file and have a delegation? 4557 * We don't need to acquire va_seq before these lookups, if 4558 * it causes an update, cinfo.before will not match, which will 4559 * trigger a cache flush even if atomic is TRUE. 4560 */ 4561 if (sfp = rfs4_lookup_and_findfile(odvp, converted_onm, &srcvp, 4562 &error, cs->cr)) { 4563 if (rfs4_check_delegated_byfp(FWRITE, sfp, TRUE, TRUE, TRUE, 4564 NULL)) { 4565 *cs->statusp = resp->status = NFS4ERR_DELAY; 4566 goto err_out; 4567 } 4568 } 4569 4570 if (srcvp == NULL) { 4571 *cs->statusp = resp->status = puterrno4(error); 4572 if (onm != converted_onm) 4573 kmem_free(converted_onm, MAXPATHLEN + 1); 4574 kmem_free(onm, olen); 4575 if (nnm != converted_nnm) 4576 kmem_free(converted_nnm, MAXPATHLEN + 1); 4577 kmem_free(nnm, nlen); 4578 goto out; 4579 } 4580 4581 sfp_rele_grant_hold = 1; 4582 4583 /* Does the destination exist and a file and have a delegation? */ 4584 if (fp = rfs4_lookup_and_findfile(ndvp, converted_nnm, &targvp, 4585 NULL, cs->cr)) { 4586 if (rfs4_check_delegated_byfp(FWRITE, fp, TRUE, TRUE, TRUE, 4587 NULL)) { 4588 *cs->statusp = resp->status = NFS4ERR_DELAY; 4589 goto err_out; 4590 } 4591 } 4592 fp_rele_grant_hold = 1; 4593 4594 4595 /* Check for NBMAND lock on both source and target */ 4596 if (nbl_need_check(srcvp)) { 4597 nbl_start_crit(srcvp, RW_READER); 4598 in_crit_src = 1; 4599 if (nbl_conflict(srcvp, NBL_RENAME, 0, 0, 0, NULL)) { 4600 *cs->statusp = resp->status = NFS4ERR_FILE_OPEN; 4601 goto err_out; 4602 } 4603 } 4604 4605 if (targvp && nbl_need_check(targvp)) { 4606 nbl_start_crit(targvp, RW_READER); 4607 in_crit_targ = 1; 4608 if (nbl_conflict(targvp, NBL_REMOVE, 0, 0, 0, NULL)) { 4609 *cs->statusp = resp->status = NFS4ERR_FILE_OPEN; 4610 goto err_out; 4611 } 4612 } 4613 4614 /* Get source "before" change value */ 4615 obdva.va_mask = AT_CTIME|AT_SEQ; 4616 error = VOP_GETATTR(odvp, &obdva, 0, cs->cr, NULL); 4617 if (!error) { 4618 nbdva.va_mask = AT_CTIME|AT_SEQ; 4619 error = VOP_GETATTR(ndvp, &nbdva, 0, cs->cr, NULL); 4620 } 4621 if (error) { 4622 *cs->statusp = resp->status = puterrno4(error); 4623 goto err_out; 4624 } 4625 4626 NFS4_SET_FATTR4_CHANGE(resp->source_cinfo.before, obdva.va_ctime) 4627 NFS4_SET_FATTR4_CHANGE(resp->target_cinfo.before, nbdva.va_ctime) 4628 4629 if ((error = VOP_RENAME(odvp, converted_onm, ndvp, converted_nnm, 4630 cs->cr, NULL, 0)) == 0 && fp != NULL) { 4631 struct vattr va; 4632 vnode_t *tvp; 4633 4634 rfs4_dbe_lock(fp->rf_dbe); 4635 tvp = fp->rf_vp; 4636 if (tvp) 4637 VN_HOLD(tvp); 4638 rfs4_dbe_unlock(fp->rf_dbe); 4639 4640 if (tvp) { 4641 va.va_mask = AT_NLINK; 4642 if (!VOP_GETATTR(tvp, &va, 0, cs->cr, NULL) && 4643 va.va_nlink == 0) { 4644 /* The file is gone and so should the state */ 4645 if (in_crit_targ) { 4646 nbl_end_crit(targvp); 4647 in_crit_targ = 0; 4648 } 4649 rfs4_close_all_state(fp); 4650 } 4651 VN_RELE(tvp); 4652 } 4653 } 4654 if (error == 0) 4655 vn_renamepath(ndvp, srcvp, nnm, nlen - 1); 4656 4657 if (in_crit_src) 4658 nbl_end_crit(srcvp); 4659 if (srcvp) 4660 VN_RELE(srcvp); 4661 if (in_crit_targ) 4662 nbl_end_crit(targvp); 4663 if (targvp) 4664 VN_RELE(targvp); 4665 4666 if (sfp) { 4667 rfs4_clear_dont_grant(sfp); 4668 rfs4_file_rele(sfp); 4669 } 4670 if (fp) { 4671 rfs4_clear_dont_grant(fp); 4672 rfs4_file_rele(fp); 4673 } 4674 4675 if (converted_onm != onm) 4676 kmem_free(converted_onm, MAXPATHLEN + 1); 4677 kmem_free(onm, olen); 4678 if (converted_nnm != nnm) 4679 kmem_free(converted_nnm, MAXPATHLEN + 1); 4680 kmem_free(nnm, nlen); 4681 4682 /* 4683 * Get the initial "after" sequence number, if it fails, set to zero 4684 */ 4685 oidva.va_mask = AT_SEQ; 4686 if (VOP_GETATTR(odvp, &oidva, 0, cs->cr, NULL)) 4687 oidva.va_seq = 0; 4688 4689 nidva.va_mask = AT_SEQ; 4690 if (VOP_GETATTR(ndvp, &nidva, 0, cs->cr, NULL)) 4691 nidva.va_seq = 0; 4692 4693 /* 4694 * Force modified data and metadata out to stable storage. 4695 */ 4696 (void) VOP_FSYNC(odvp, 0, cs->cr, NULL); 4697 (void) VOP_FSYNC(ndvp, 0, cs->cr, NULL); 4698 4699 if (error) { 4700 *cs->statusp = resp->status = puterrno4(error); 4701 goto out; 4702 } 4703 4704 /* 4705 * Get "after" change values, if it fails, simply return the 4706 * before value. 4707 */ 4708 oadva.va_mask = AT_CTIME|AT_SEQ; 4709 if (VOP_GETATTR(odvp, &oadva, 0, cs->cr, NULL)) { 4710 oadva.va_ctime = obdva.va_ctime; 4711 oadva.va_seq = 0; 4712 } 4713 4714 nadva.va_mask = AT_CTIME|AT_SEQ; 4715 if (VOP_GETATTR(odvp, &nadva, 0, cs->cr, NULL)) { 4716 nadva.va_ctime = nbdva.va_ctime; 4717 nadva.va_seq = 0; 4718 } 4719 4720 NFS4_SET_FATTR4_CHANGE(resp->source_cinfo.after, oadva.va_ctime) 4721 NFS4_SET_FATTR4_CHANGE(resp->target_cinfo.after, nadva.va_ctime) 4722 4723 /* 4724 * The cinfo.atomic = TRUE only if we have 4725 * non-zero va_seq's, and it has incremented by exactly one 4726 * during the VOP_RENAME and it didn't change during the VOP_FSYNC. 4727 */ 4728 if (obdva.va_seq && oidva.va_seq && oadva.va_seq && 4729 oidva.va_seq == (obdva.va_seq + 1) && oidva.va_seq == oadva.va_seq) 4730 resp->source_cinfo.atomic = TRUE; 4731 else 4732 resp->source_cinfo.atomic = FALSE; 4733 4734 if (nbdva.va_seq && nidva.va_seq && nadva.va_seq && 4735 nidva.va_seq == (nbdva.va_seq + 1) && nidva.va_seq == nadva.va_seq) 4736 resp->target_cinfo.atomic = TRUE; 4737 else 4738 resp->target_cinfo.atomic = FALSE; 4739 4740 #ifdef VOLATILE_FH_TEST 4741 { 4742 extern void add_volrnm_fh(struct exportinfo *, vnode_t *); 4743 4744 /* 4745 * Add the renamed file handle to the volatile rename list 4746 */ 4747 if (cs->exi->exi_export.ex_flags & EX_VOLRNM) { 4748 /* file handles may expire on rename */ 4749 vnode_t *vp; 4750 4751 nnm = utf8_to_fn(&args->newname, &nlen, NULL); 4752 /* 4753 * Already know that nnm will be a valid string 4754 */ 4755 error = VOP_LOOKUP(ndvp, nnm, &vp, NULL, 0, NULL, cs->cr, 4756 NULL, NULL, NULL); 4757 kmem_free(nnm, nlen); 4758 if (!error) { 4759 add_volrnm_fh(cs->exi, vp); 4760 VN_RELE(vp); 4761 } 4762 } 4763 } 4764 #endif /* VOLATILE_FH_TEST */ 4765 4766 *cs->statusp = resp->status = NFS4_OK; 4767 out: 4768 DTRACE_NFSV4_2(op__rename__done, struct compound_state *, cs, 4769 RENAME4res *, resp); 4770 return; 4771 4772 err_out: 4773 if (onm != converted_onm) 4774 kmem_free(converted_onm, MAXPATHLEN + 1); 4775 if (onm != NULL) 4776 kmem_free(onm, olen); 4777 if (nnm != converted_nnm) 4778 kmem_free(converted_nnm, MAXPATHLEN + 1); 4779 if (nnm != NULL) 4780 kmem_free(nnm, nlen); 4781 4782 if (in_crit_src) nbl_end_crit(srcvp); 4783 if (in_crit_targ) nbl_end_crit(targvp); 4784 if (targvp) VN_RELE(targvp); 4785 if (srcvp) VN_RELE(srcvp); 4786 if (sfp) { 4787 if (sfp_rele_grant_hold) rfs4_clear_dont_grant(sfp); 4788 rfs4_file_rele(sfp); 4789 } 4790 if (fp) { 4791 if (fp_rele_grant_hold) rfs4_clear_dont_grant(fp); 4792 rfs4_file_rele(fp); 4793 } 4794 4795 DTRACE_NFSV4_2(op__rename__done, struct compound_state *, cs, 4796 RENAME4res *, resp); 4797 } 4798 4799 /* ARGSUSED */ 4800 static void 4801 rfs4_op_renew(nfs_argop4 *argop, nfs_resop4 *resop, struct svc_req *req, 4802 struct compound_state *cs) 4803 { 4804 RENEW4args *args = &argop->nfs_argop4_u.oprenew; 4805 RENEW4res *resp = &resop->nfs_resop4_u.oprenew; 4806 rfs4_client_t *cp; 4807 4808 DTRACE_NFSV4_2(op__renew__start, struct compound_state *, cs, 4809 RENEW4args *, args); 4810 4811 if ((cp = rfs4_findclient_by_id(args->clientid, FALSE)) == NULL) { 4812 *cs->statusp = resp->status = 4813 rfs4_check_clientid(&args->clientid, 0); 4814 goto out; 4815 } 4816 4817 if (rfs4_lease_expired(cp)) { 4818 rfs4_client_rele(cp); 4819 *cs->statusp = resp->status = NFS4ERR_EXPIRED; 4820 goto out; 4821 } 4822 4823 rfs4_update_lease(cp); 4824 4825 mutex_enter(cp->rc_cbinfo.cb_lock); 4826 if (cp->rc_cbinfo.cb_notified_of_cb_path_down == FALSE) { 4827 cp->rc_cbinfo.cb_notified_of_cb_path_down = TRUE; 4828 *cs->statusp = resp->status = NFS4ERR_CB_PATH_DOWN; 4829 } else { 4830 *cs->statusp = resp->status = NFS4_OK; 4831 } 4832 mutex_exit(cp->rc_cbinfo.cb_lock); 4833 4834 rfs4_client_rele(cp); 4835 4836 out: 4837 DTRACE_NFSV4_2(op__renew__done, struct compound_state *, cs, 4838 RENEW4res *, resp); 4839 } 4840 4841 /* ARGSUSED */ 4842 static void 4843 rfs4_op_restorefh(nfs_argop4 *args, nfs_resop4 *resop, struct svc_req *req, 4844 struct compound_state *cs) 4845 { 4846 RESTOREFH4res *resp = &resop->nfs_resop4_u.oprestorefh; 4847 4848 DTRACE_NFSV4_1(op__restorefh__start, struct compound_state *, cs); 4849 4850 /* No need to check cs->access - we are not accessing any object */ 4851 if ((cs->saved_vp == NULL) || (cs->saved_fh.nfs_fh4_val == NULL)) { 4852 *cs->statusp = resp->status = NFS4ERR_RESTOREFH; 4853 goto out; 4854 } 4855 if (cs->vp != NULL) { 4856 VN_RELE(cs->vp); 4857 } 4858 cs->vp = cs->saved_vp; 4859 cs->saved_vp = NULL; 4860 cs->exi = cs->saved_exi; 4861 nfs_fh4_copy(&cs->saved_fh, &cs->fh); 4862 *cs->statusp = resp->status = NFS4_OK; 4863 cs->deleg = FALSE; 4864 4865 out: 4866 DTRACE_NFSV4_2(op__restorefh__done, struct compound_state *, cs, 4867 RESTOREFH4res *, resp); 4868 } 4869 4870 /* ARGSUSED */ 4871 static void 4872 rfs4_op_savefh(nfs_argop4 *argop, nfs_resop4 *resop, struct svc_req *req, 4873 struct compound_state *cs) 4874 { 4875 SAVEFH4res *resp = &resop->nfs_resop4_u.opsavefh; 4876 4877 DTRACE_NFSV4_1(op__savefh__start, struct compound_state *, cs); 4878 4879 /* No need to check cs->access - we are not accessing any object */ 4880 if (cs->vp == NULL) { 4881 *cs->statusp = resp->status = NFS4ERR_NOFILEHANDLE; 4882 goto out; 4883 } 4884 if (cs->saved_vp != NULL) { 4885 VN_RELE(cs->saved_vp); 4886 } 4887 cs->saved_vp = cs->vp; 4888 VN_HOLD(cs->saved_vp); 4889 cs->saved_exi = cs->exi; 4890 /* 4891 * since SAVEFH is fairly rare, don't alloc space for its fh 4892 * unless necessary. 4893 */ 4894 if (cs->saved_fh.nfs_fh4_val == NULL) { 4895 cs->saved_fh.nfs_fh4_val = kmem_alloc(NFS4_FHSIZE, KM_SLEEP); 4896 } 4897 nfs_fh4_copy(&cs->fh, &cs->saved_fh); 4898 *cs->statusp = resp->status = NFS4_OK; 4899 4900 out: 4901 DTRACE_NFSV4_2(op__savefh__done, struct compound_state *, cs, 4902 SAVEFH4res *, resp); 4903 } 4904 4905 /* 4906 * rfs4_verify_attr is called when nfsv4 Setattr failed, but we wish to 4907 * return the bitmap of attrs that were set successfully. It is also 4908 * called by Verify/Nverify to test the vattr/vfsstat attrs. It should 4909 * always be called only after rfs4_do_set_attrs(). 4910 * 4911 * Verify that the attributes are same as the expected ones. sargp->vap 4912 * and sargp->sbp contain the input attributes as translated from fattr4. 4913 * 4914 * This function verifies only the attrs that correspond to a vattr or 4915 * vfsstat struct. That is because of the extra step needed to get the 4916 * corresponding system structs. Other attributes have already been set or 4917 * verified by do_rfs4_set_attrs. 4918 * 4919 * Return 0 if all attrs match, -1 if some don't, error if error processing. 4920 */ 4921 static int 4922 rfs4_verify_attr(struct nfs4_svgetit_arg *sargp, 4923 bitmap4 *resp, struct nfs4_ntov_table *ntovp) 4924 { 4925 int error, ret_error = 0; 4926 int i, k; 4927 uint_t sva_mask = sargp->vap->va_mask; 4928 uint_t vbit; 4929 union nfs4_attr_u *na; 4930 uint8_t *amap; 4931 bool_t getsb = ntovp->vfsstat; 4932 4933 if (sva_mask != 0) { 4934 /* 4935 * Okay to overwrite sargp->vap because we verify based 4936 * on the incoming values. 4937 */ 4938 ret_error = VOP_GETATTR(sargp->cs->vp, sargp->vap, 0, 4939 sargp->cs->cr, NULL); 4940 if (ret_error) { 4941 if (resp == NULL) 4942 return (ret_error); 4943 /* 4944 * Must return bitmap of successful attrs 4945 */ 4946 sva_mask = 0; /* to prevent checking vap later */ 4947 } else { 4948 /* 4949 * Some file systems clobber va_mask. it is probably 4950 * wrong of them to do so, nonethless we practice 4951 * defensive coding. 4952 * See bug id 4276830. 4953 */ 4954 sargp->vap->va_mask = sva_mask; 4955 } 4956 } 4957 4958 if (getsb) { 4959 /* 4960 * Now get the superblock and loop on the bitmap, as there is 4961 * no simple way of translating from superblock to bitmap4. 4962 */ 4963 ret_error = VFS_STATVFS(sargp->cs->vp->v_vfsp, sargp->sbp); 4964 if (ret_error) { 4965 if (resp == NULL) 4966 goto errout; 4967 getsb = FALSE; 4968 } 4969 } 4970 4971 /* 4972 * Now loop and verify each attribute which getattr returned 4973 * whether it's the same as the input. 4974 */ 4975 if (resp == NULL && !getsb && (sva_mask == 0)) 4976 goto errout; 4977 4978 na = ntovp->na; 4979 amap = ntovp->amap; 4980 k = 0; 4981 for (i = 0; i < ntovp->attrcnt; i++, na++, amap++) { 4982 k = *amap; 4983 ASSERT(nfs4_ntov_map[k].nval == k); 4984 vbit = nfs4_ntov_map[k].vbit; 4985 4986 /* 4987 * If vattr attribute but VOP_GETATTR failed, or it's 4988 * superblock attribute but VFS_STATVFS failed, skip 4989 */ 4990 if (vbit) { 4991 if ((vbit & sva_mask) == 0) 4992 continue; 4993 } else if (!(getsb && nfs4_ntov_map[k].vfsstat)) { 4994 continue; 4995 } 4996 error = (*nfs4_ntov_map[k].sv_getit)(NFS4ATTR_VERIT, sargp, na); 4997 if (resp != NULL) { 4998 if (error) 4999 ret_error = -1; /* not all match */ 5000 else /* update response bitmap */ 5001 *resp |= nfs4_ntov_map[k].fbit; 5002 continue; 5003 } 5004 if (error) { 5005 ret_error = -1; /* not all match */ 5006 break; 5007 } 5008 } 5009 errout: 5010 return (ret_error); 5011 } 5012 5013 /* 5014 * Decode the attribute to be set/verified. If the attr requires a sys op 5015 * (VOP_GETATTR, VFS_VFSSTAT), and the request is to verify, then don't 5016 * call the sv_getit function for it, because the sys op hasn't yet been done. 5017 * Return 0 for success, error code if failed. 5018 * 5019 * Note: the decoded arg is not freed here but in nfs4_ntov_table_free. 5020 */ 5021 static int 5022 decode_fattr4_attr(nfs4_attr_cmd_t cmd, struct nfs4_svgetit_arg *sargp, 5023 int k, XDR *xdrp, bitmap4 *resp_bval, union nfs4_attr_u *nap) 5024 { 5025 int error = 0; 5026 bool_t set_later; 5027 5028 sargp->vap->va_mask |= nfs4_ntov_map[k].vbit; 5029 5030 if ((*nfs4_ntov_map[k].xfunc)(xdrp, nap)) { 5031 set_later = nfs4_ntov_map[k].vbit || nfs4_ntov_map[k].vfsstat; 5032 /* 5033 * don't verify yet if a vattr or sb dependent attr, 5034 * because we don't have their sys values yet. 5035 * Will be done later. 5036 */ 5037 if (! (set_later && (cmd == NFS4ATTR_VERIT))) { 5038 /* 5039 * ACLs are a special case, since setting the MODE 5040 * conflicts with setting the ACL. We delay setting 5041 * the ACL until all other attributes have been set. 5042 * The ACL gets set in do_rfs4_op_setattr(). 5043 */ 5044 if (nfs4_ntov_map[k].fbit != FATTR4_ACL_MASK) { 5045 error = (*nfs4_ntov_map[k].sv_getit)(cmd, 5046 sargp, nap); 5047 if (error) { 5048 xdr_free(nfs4_ntov_map[k].xfunc, 5049 (caddr_t)nap); 5050 } 5051 } 5052 } 5053 } else { 5054 #ifdef DEBUG 5055 cmn_err(CE_NOTE, "decode_fattr4_attr: error " 5056 "decoding attribute %d\n", k); 5057 #endif 5058 error = EINVAL; 5059 } 5060 if (!error && resp_bval && !set_later) { 5061 *resp_bval |= nfs4_ntov_map[k].fbit; 5062 } 5063 5064 return (error); 5065 } 5066 5067 /* 5068 * Set vattr based on incoming fattr4 attrs - used by setattr. 5069 * Set response mask. Ignore any values that are not writable vattr attrs. 5070 */ 5071 static nfsstat4 5072 do_rfs4_set_attrs(bitmap4 *resp, fattr4 *fattrp, struct compound_state *cs, 5073 struct nfs4_svgetit_arg *sargp, struct nfs4_ntov_table *ntovp, 5074 nfs4_attr_cmd_t cmd) 5075 { 5076 int error = 0; 5077 int i; 5078 char *attrs = fattrp->attrlist4; 5079 uint32_t attrslen = fattrp->attrlist4_len; 5080 XDR xdr; 5081 nfsstat4 status = NFS4_OK; 5082 vnode_t *vp = cs->vp; 5083 union nfs4_attr_u *na; 5084 uint8_t *amap; 5085 5086 #ifndef lint 5087 /* 5088 * Make sure that maximum attribute number can be expressed as an 5089 * 8 bit quantity. 5090 */ 5091 ASSERT(NFS4_MAXNUM_ATTRS <= (UINT8_MAX + 1)); 5092 #endif 5093 5094 if (vp == NULL) { 5095 if (resp) 5096 *resp = 0; 5097 return (NFS4ERR_NOFILEHANDLE); 5098 } 5099 if (cs->access == CS_ACCESS_DENIED) { 5100 if (resp) 5101 *resp = 0; 5102 return (NFS4ERR_ACCESS); 5103 } 5104 5105 sargp->op = cmd; 5106 sargp->cs = cs; 5107 sargp->flag = 0; /* may be set later */ 5108 sargp->vap->va_mask = 0; 5109 sargp->rdattr_error = NFS4_OK; 5110 sargp->rdattr_error_req = FALSE; 5111 /* sargp->sbp is set by the caller */ 5112 5113 xdrmem_create(&xdr, attrs, attrslen, XDR_DECODE); 5114 5115 na = ntovp->na; 5116 amap = ntovp->amap; 5117 5118 /* 5119 * The following loop iterates on the nfs4_ntov_map checking 5120 * if the fbit is set in the requested bitmap. 5121 * If set then we process the arguments using the 5122 * rfs4_fattr4 conversion functions to populate the setattr 5123 * vattr and va_mask. Any settable attrs that are not using vattr 5124 * will be set in this loop. 5125 */ 5126 for (i = 0; i < nfs4_ntov_map_size; i++) { 5127 if (!(fattrp->attrmask & nfs4_ntov_map[i].fbit)) { 5128 continue; 5129 } 5130 /* 5131 * If setattr, must be a writable attr. 5132 * If verify/nverify, must be a readable attr. 5133 */ 5134 if ((error = (*nfs4_ntov_map[i].sv_getit)( 5135 NFS4ATTR_SUPPORTED, sargp, NULL)) != 0) { 5136 /* 5137 * Client tries to set/verify an 5138 * unsupported attribute, tries to set 5139 * a read only attr or verify a write 5140 * only one - error! 5141 */ 5142 break; 5143 } 5144 /* 5145 * Decode the attribute to set/verify 5146 */ 5147 error = decode_fattr4_attr(cmd, sargp, nfs4_ntov_map[i].nval, 5148 &xdr, resp ? resp : NULL, na); 5149 if (error) 5150 break; 5151 *amap++ = (uint8_t)nfs4_ntov_map[i].nval; 5152 na++; 5153 (ntovp->attrcnt)++; 5154 if (nfs4_ntov_map[i].vfsstat) 5155 ntovp->vfsstat = TRUE; 5156 } 5157 5158 if (error != 0) 5159 status = (error == ENOTSUP ? NFS4ERR_ATTRNOTSUPP : 5160 puterrno4(error)); 5161 /* xdrmem_destroy(&xdrs); */ /* NO-OP */ 5162 return (status); 5163 } 5164 5165 static nfsstat4 5166 do_rfs4_op_setattr(bitmap4 *resp, fattr4 *fattrp, struct compound_state *cs, 5167 stateid4 *stateid) 5168 { 5169 int error = 0; 5170 struct nfs4_svgetit_arg sarg; 5171 bool_t trunc; 5172 5173 nfsstat4 status = NFS4_OK; 5174 cred_t *cr = cs->cr; 5175 vnode_t *vp = cs->vp; 5176 struct nfs4_ntov_table ntov; 5177 struct statvfs64 sb; 5178 struct vattr bva; 5179 struct flock64 bf; 5180 int in_crit = 0; 5181 uint_t saved_mask = 0; 5182 caller_context_t ct; 5183 5184 *resp = 0; 5185 sarg.sbp = &sb; 5186 sarg.is_referral = B_FALSE; 5187 nfs4_ntov_table_init(&ntov); 5188 status = do_rfs4_set_attrs(resp, fattrp, cs, &sarg, &ntov, 5189 NFS4ATTR_SETIT); 5190 if (status != NFS4_OK) { 5191 /* 5192 * failed set attrs 5193 */ 5194 goto done; 5195 } 5196 if ((sarg.vap->va_mask == 0) && 5197 (! (fattrp->attrmask & FATTR4_ACL_MASK))) { 5198 /* 5199 * no further work to be done 5200 */ 5201 goto done; 5202 } 5203 5204 /* 5205 * If we got a request to set the ACL and the MODE, only 5206 * allow changing VSUID, VSGID, and VSVTX. Attempting 5207 * to change any other bits, along with setting an ACL, 5208 * gives NFS4ERR_INVAL. 5209 */ 5210 if ((fattrp->attrmask & FATTR4_ACL_MASK) && 5211 (fattrp->attrmask & FATTR4_MODE_MASK)) { 5212 vattr_t va; 5213 5214 va.va_mask = AT_MODE; 5215 error = VOP_GETATTR(vp, &va, 0, cs->cr, NULL); 5216 if (error) { 5217 status = puterrno4(error); 5218 goto done; 5219 } 5220 if ((sarg.vap->va_mode ^ va.va_mode) & 5221 ~(VSUID | VSGID | VSVTX)) { 5222 status = NFS4ERR_INVAL; 5223 goto done; 5224 } 5225 } 5226 5227 /* Check stateid only if size has been set */ 5228 if (sarg.vap->va_mask & AT_SIZE) { 5229 trunc = (sarg.vap->va_size == 0); 5230 status = rfs4_check_stateid(FWRITE, cs->vp, stateid, 5231 trunc, &cs->deleg, sarg.vap->va_mask & AT_SIZE, &ct); 5232 if (status != NFS4_OK) 5233 goto done; 5234 } else { 5235 ct.cc_sysid = 0; 5236 ct.cc_pid = 0; 5237 ct.cc_caller_id = nfs4_srv_caller_id; 5238 ct.cc_flags = CC_DONTBLOCK; 5239 } 5240 5241 /* XXX start of possible race with delegations */ 5242 5243 /* 5244 * We need to specially handle size changes because it is 5245 * possible for the client to create a file with read-only 5246 * modes, but with the file opened for writing. If the client 5247 * then tries to set the file size, e.g. ftruncate(3C), 5248 * fcntl(F_FREESP), the normal access checking done in 5249 * VOP_SETATTR would prevent the client from doing it even though 5250 * it should be allowed to do so. To get around this, we do the 5251 * access checking for ourselves and use VOP_SPACE which doesn't 5252 * do the access checking. 5253 * Also the client should not be allowed to change the file 5254 * size if there is a conflicting non-blocking mandatory lock in 5255 * the region of the change. 5256 */ 5257 if (vp->v_type == VREG && (sarg.vap->va_mask & AT_SIZE)) { 5258 u_offset_t offset; 5259 ssize_t length; 5260 5261 /* 5262 * ufs_setattr clears AT_SIZE from vap->va_mask, but 5263 * before returning, sarg.vap->va_mask is used to 5264 * generate the setattr reply bitmap. We also clear 5265 * AT_SIZE below before calling VOP_SPACE. For both 5266 * of these cases, the va_mask needs to be saved here 5267 * and restored after calling VOP_SETATTR. 5268 */ 5269 saved_mask = sarg.vap->va_mask; 5270 5271 /* 5272 * Check any possible conflict due to NBMAND locks. 5273 * Get into critical region before VOP_GETATTR, so the 5274 * size attribute is valid when checking conflicts. 5275 */ 5276 if (nbl_need_check(vp)) { 5277 nbl_start_crit(vp, RW_READER); 5278 in_crit = 1; 5279 } 5280 5281 bva.va_mask = AT_UID|AT_SIZE; 5282 if (error = VOP_GETATTR(vp, &bva, 0, cr, &ct)) { 5283 status = puterrno4(error); 5284 goto done; 5285 } 5286 5287 if (in_crit) { 5288 if (sarg.vap->va_size < bva.va_size) { 5289 offset = sarg.vap->va_size; 5290 length = bva.va_size - sarg.vap->va_size; 5291 } else { 5292 offset = bva.va_size; 5293 length = sarg.vap->va_size - bva.va_size; 5294 } 5295 if (nbl_conflict(vp, NBL_WRITE, offset, length, 0, 5296 &ct)) { 5297 status = NFS4ERR_LOCKED; 5298 goto done; 5299 } 5300 } 5301 5302 if (crgetuid(cr) == bva.va_uid) { 5303 sarg.vap->va_mask &= ~AT_SIZE; 5304 bf.l_type = F_WRLCK; 5305 bf.l_whence = 0; 5306 bf.l_start = (off64_t)sarg.vap->va_size; 5307 bf.l_len = 0; 5308 bf.l_sysid = 0; 5309 bf.l_pid = 0; 5310 error = VOP_SPACE(vp, F_FREESP, &bf, FWRITE, 5311 (offset_t)sarg.vap->va_size, cr, &ct); 5312 } 5313 } 5314 5315 if (!error && sarg.vap->va_mask != 0) 5316 error = VOP_SETATTR(vp, sarg.vap, sarg.flag, cr, &ct); 5317 5318 /* restore va_mask -- ufs_setattr clears AT_SIZE */ 5319 if (saved_mask & AT_SIZE) 5320 sarg.vap->va_mask |= AT_SIZE; 5321 5322 /* 5323 * If an ACL was being set, it has been delayed until now, 5324 * in order to set the mode (via the VOP_SETATTR() above) first. 5325 */ 5326 if ((! error) && (fattrp->attrmask & FATTR4_ACL_MASK)) { 5327 int i; 5328 5329 for (i = 0; i < NFS4_MAXNUM_ATTRS; i++) 5330 if (ntov.amap[i] == FATTR4_ACL) 5331 break; 5332 if (i < NFS4_MAXNUM_ATTRS) { 5333 error = (*nfs4_ntov_map[FATTR4_ACL].sv_getit)( 5334 NFS4ATTR_SETIT, &sarg, &ntov.na[i]); 5335 if (error == 0) { 5336 *resp |= FATTR4_ACL_MASK; 5337 } else if (error == ENOTSUP) { 5338 (void) rfs4_verify_attr(&sarg, resp, &ntov); 5339 status = NFS4ERR_ATTRNOTSUPP; 5340 goto done; 5341 } 5342 } else { 5343 NFS4_DEBUG(rfs4_debug, 5344 (CE_NOTE, "do_rfs4_op_setattr: " 5345 "unable to find ACL in fattr4")); 5346 error = EINVAL; 5347 } 5348 } 5349 5350 if (error) { 5351 /* check if a monitor detected a delegation conflict */ 5352 if (error == EAGAIN && (ct.cc_flags & CC_WOULDBLOCK)) 5353 status = NFS4ERR_DELAY; 5354 else 5355 status = puterrno4(error); 5356 5357 /* 5358 * Set the response bitmap when setattr failed. 5359 * If VOP_SETATTR partially succeeded, test by doing a 5360 * VOP_GETATTR on the object and comparing the data 5361 * to the setattr arguments. 5362 */ 5363 (void) rfs4_verify_attr(&sarg, resp, &ntov); 5364 } else { 5365 /* 5366 * Force modified metadata out to stable storage. 5367 */ 5368 (void) VOP_FSYNC(vp, FNODSYNC, cr, &ct); 5369 /* 5370 * Set response bitmap 5371 */ 5372 nfs4_vmask_to_nmask_set(sarg.vap->va_mask, resp); 5373 } 5374 5375 /* Return early and already have a NFSv4 error */ 5376 done: 5377 /* 5378 * Except for nfs4_vmask_to_nmask_set(), vattr --> fattr 5379 * conversion sets both readable and writeable NFS4 attrs 5380 * for AT_MTIME and AT_ATIME. The line below masks out 5381 * unrequested attrs from the setattr result bitmap. This 5382 * is placed after the done: label to catch the ATTRNOTSUP 5383 * case. 5384 */ 5385 *resp &= fattrp->attrmask; 5386 5387 if (in_crit) 5388 nbl_end_crit(vp); 5389 5390 nfs4_ntov_table_free(&ntov, &sarg); 5391 5392 return (status); 5393 } 5394 5395 /* ARGSUSED */ 5396 static void 5397 rfs4_op_setattr(nfs_argop4 *argop, nfs_resop4 *resop, struct svc_req *req, 5398 struct compound_state *cs) 5399 { 5400 SETATTR4args *args = &argop->nfs_argop4_u.opsetattr; 5401 SETATTR4res *resp = &resop->nfs_resop4_u.opsetattr; 5402 bslabel_t *clabel; 5403 5404 DTRACE_NFSV4_2(op__setattr__start, struct compound_state *, cs, 5405 SETATTR4args *, args); 5406 5407 if (cs->vp == NULL) { 5408 *cs->statusp = resp->status = NFS4ERR_NOFILEHANDLE; 5409 goto out; 5410 } 5411 5412 /* 5413 * If there is an unshared filesystem mounted on this vnode, 5414 * do not allow to setattr on this vnode. 5415 */ 5416 if (vn_ismntpt(cs->vp)) { 5417 *cs->statusp = resp->status = NFS4ERR_ACCESS; 5418 goto out; 5419 } 5420 5421 resp->attrsset = 0; 5422 5423 if (rdonly4(cs->exi, cs->vp, req)) { 5424 *cs->statusp = resp->status = NFS4ERR_ROFS; 5425 goto out; 5426 } 5427 5428 /* check label before setting attributes */ 5429 if (is_system_labeled()) { 5430 ASSERT(req->rq_label != NULL); 5431 clabel = req->rq_label; 5432 DTRACE_PROBE2(tx__rfs4__log__info__opsetattr__clabel, char *, 5433 "got client label from request(1)", 5434 struct svc_req *, req); 5435 if (!blequal(&l_admin_low->tsl_label, clabel)) { 5436 if (!do_rfs_label_check(clabel, cs->vp, 5437 EQUALITY_CHECK, cs->exi)) { 5438 *cs->statusp = resp->status = NFS4ERR_ACCESS; 5439 goto out; 5440 } 5441 } 5442 } 5443 5444 *cs->statusp = resp->status = 5445 do_rfs4_op_setattr(&resp->attrsset, &args->obj_attributes, cs, 5446 &args->stateid); 5447 5448 out: 5449 DTRACE_NFSV4_2(op__setattr__done, struct compound_state *, cs, 5450 SETATTR4res *, resp); 5451 } 5452 5453 /* ARGSUSED */ 5454 static void 5455 rfs4_op_verify(nfs_argop4 *argop, nfs_resop4 *resop, struct svc_req *req, 5456 struct compound_state *cs) 5457 { 5458 /* 5459 * verify and nverify are exactly the same, except that nverify 5460 * succeeds when some argument changed, and verify succeeds when 5461 * when none changed. 5462 */ 5463 5464 VERIFY4args *args = &argop->nfs_argop4_u.opverify; 5465 VERIFY4res *resp = &resop->nfs_resop4_u.opverify; 5466 5467 int error; 5468 struct nfs4_svgetit_arg sarg; 5469 struct statvfs64 sb; 5470 struct nfs4_ntov_table ntov; 5471 5472 DTRACE_NFSV4_2(op__verify__start, struct compound_state *, cs, 5473 VERIFY4args *, args); 5474 5475 if (cs->vp == NULL) { 5476 *cs->statusp = resp->status = NFS4ERR_NOFILEHANDLE; 5477 goto out; 5478 } 5479 5480 sarg.sbp = &sb; 5481 sarg.is_referral = B_FALSE; 5482 nfs4_ntov_table_init(&ntov); 5483 resp->status = do_rfs4_set_attrs(NULL, &args->obj_attributes, cs, 5484 &sarg, &ntov, NFS4ATTR_VERIT); 5485 if (resp->status != NFS4_OK) { 5486 /* 5487 * do_rfs4_set_attrs will try to verify systemwide attrs, 5488 * so could return -1 for "no match". 5489 */ 5490 if (resp->status == -1) 5491 resp->status = NFS4ERR_NOT_SAME; 5492 goto done; 5493 } 5494 error = rfs4_verify_attr(&sarg, NULL, &ntov); 5495 switch (error) { 5496 case 0: 5497 resp->status = NFS4_OK; 5498 break; 5499 case -1: 5500 resp->status = NFS4ERR_NOT_SAME; 5501 break; 5502 default: 5503 resp->status = puterrno4(error); 5504 break; 5505 } 5506 done: 5507 *cs->statusp = resp->status; 5508 nfs4_ntov_table_free(&ntov, &sarg); 5509 out: 5510 DTRACE_NFSV4_2(op__verify__done, struct compound_state *, cs, 5511 VERIFY4res *, resp); 5512 } 5513 5514 /* ARGSUSED */ 5515 static void 5516 rfs4_op_nverify(nfs_argop4 *argop, nfs_resop4 *resop, struct svc_req *req, 5517 struct compound_state *cs) 5518 { 5519 /* 5520 * verify and nverify are exactly the same, except that nverify 5521 * succeeds when some argument changed, and verify succeeds when 5522 * when none changed. 5523 */ 5524 5525 NVERIFY4args *args = &argop->nfs_argop4_u.opnverify; 5526 NVERIFY4res *resp = &resop->nfs_resop4_u.opnverify; 5527 5528 int error; 5529 struct nfs4_svgetit_arg sarg; 5530 struct statvfs64 sb; 5531 struct nfs4_ntov_table ntov; 5532 5533 DTRACE_NFSV4_2(op__nverify__start, struct compound_state *, cs, 5534 NVERIFY4args *, args); 5535 5536 if (cs->vp == NULL) { 5537 *cs->statusp = resp->status = NFS4ERR_NOFILEHANDLE; 5538 DTRACE_NFSV4_2(op__nverify__done, struct compound_state *, cs, 5539 NVERIFY4res *, resp); 5540 return; 5541 } 5542 sarg.sbp = &sb; 5543 sarg.is_referral = B_FALSE; 5544 nfs4_ntov_table_init(&ntov); 5545 resp->status = do_rfs4_set_attrs(NULL, &args->obj_attributes, cs, 5546 &sarg, &ntov, NFS4ATTR_VERIT); 5547 if (resp->status != NFS4_OK) { 5548 /* 5549 * do_rfs4_set_attrs will try to verify systemwide attrs, 5550 * so could return -1 for "no match". 5551 */ 5552 if (resp->status == -1) 5553 resp->status = NFS4_OK; 5554 goto done; 5555 } 5556 error = rfs4_verify_attr(&sarg, NULL, &ntov); 5557 switch (error) { 5558 case 0: 5559 resp->status = NFS4ERR_SAME; 5560 break; 5561 case -1: 5562 resp->status = NFS4_OK; 5563 break; 5564 default: 5565 resp->status = puterrno4(error); 5566 break; 5567 } 5568 done: 5569 *cs->statusp = resp->status; 5570 nfs4_ntov_table_free(&ntov, &sarg); 5571 5572 DTRACE_NFSV4_2(op__nverify__done, struct compound_state *, cs, 5573 NVERIFY4res *, resp); 5574 } 5575 5576 /* 5577 * XXX - This should live in an NFS header file. 5578 */ 5579 #define MAX_IOVECS 12 5580 5581 /* ARGSUSED */ 5582 static void 5583 rfs4_op_write(nfs_argop4 *argop, nfs_resop4 *resop, struct svc_req *req, 5584 struct compound_state *cs) 5585 { 5586 WRITE4args *args = &argop->nfs_argop4_u.opwrite; 5587 WRITE4res *resp = &resop->nfs_resop4_u.opwrite; 5588 int error; 5589 vnode_t *vp; 5590 struct vattr bva; 5591 u_offset_t rlimit; 5592 struct uio uio; 5593 struct iovec iov[MAX_IOVECS]; 5594 struct iovec *iovp; 5595 int iovcnt; 5596 int ioflag; 5597 cred_t *savecred, *cr; 5598 bool_t *deleg = &cs->deleg; 5599 nfsstat4 stat; 5600 int in_crit = 0; 5601 caller_context_t ct; 5602 5603 DTRACE_NFSV4_2(op__write__start, struct compound_state *, cs, 5604 WRITE4args *, args); 5605 5606 vp = cs->vp; 5607 if (vp == NULL) { 5608 *cs->statusp = resp->status = NFS4ERR_NOFILEHANDLE; 5609 goto out; 5610 } 5611 if (cs->access == CS_ACCESS_DENIED) { 5612 *cs->statusp = resp->status = NFS4ERR_ACCESS; 5613 goto out; 5614 } 5615 5616 cr = cs->cr; 5617 5618 if ((stat = rfs4_check_stateid(FWRITE, vp, &args->stateid, FALSE, 5619 deleg, TRUE, &ct)) != NFS4_OK) { 5620 *cs->statusp = resp->status = stat; 5621 goto out; 5622 } 5623 5624 /* 5625 * We have to enter the critical region before calling VOP_RWLOCK 5626 * to avoid a deadlock with ufs. 5627 */ 5628 if (nbl_need_check(vp)) { 5629 nbl_start_crit(vp, RW_READER); 5630 in_crit = 1; 5631 if (nbl_conflict(vp, NBL_WRITE, 5632 args->offset, args->data_len, 0, &ct)) { 5633 *cs->statusp = resp->status = NFS4ERR_LOCKED; 5634 goto out; 5635 } 5636 } 5637 5638 bva.va_mask = AT_MODE | AT_UID; 5639 error = VOP_GETATTR(vp, &bva, 0, cr, &ct); 5640 5641 /* 5642 * If we can't get the attributes, then we can't do the 5643 * right access checking. So, we'll fail the request. 5644 */ 5645 if (error) { 5646 *cs->statusp = resp->status = puterrno4(error); 5647 goto out; 5648 } 5649 5650 if (rdonly4(cs->exi, cs->vp, req)) { 5651 *cs->statusp = resp->status = NFS4ERR_ROFS; 5652 goto out; 5653 } 5654 5655 if (vp->v_type != VREG) { 5656 *cs->statusp = resp->status = 5657 ((vp->v_type == VDIR) ? NFS4ERR_ISDIR : NFS4ERR_INVAL); 5658 goto out; 5659 } 5660 5661 if (crgetuid(cr) != bva.va_uid && 5662 (error = VOP_ACCESS(vp, VWRITE, 0, cr, &ct))) { 5663 *cs->statusp = resp->status = puterrno4(error); 5664 goto out; 5665 } 5666 5667 if (MANDLOCK(vp, bva.va_mode)) { 5668 *cs->statusp = resp->status = NFS4ERR_ACCESS; 5669 goto out; 5670 } 5671 5672 if (args->data_len == 0) { 5673 *cs->statusp = resp->status = NFS4_OK; 5674 resp->count = 0; 5675 resp->committed = args->stable; 5676 resp->writeverf = Write4verf; 5677 goto out; 5678 } 5679 5680 if (args->mblk != NULL) { 5681 mblk_t *m; 5682 uint_t bytes, round_len; 5683 5684 iovcnt = 0; 5685 bytes = 0; 5686 round_len = roundup(args->data_len, BYTES_PER_XDR_UNIT); 5687 for (m = args->mblk; 5688 m != NULL && bytes < round_len; 5689 m = m->b_cont) { 5690 iovcnt++; 5691 bytes += MBLKL(m); 5692 } 5693 #ifdef DEBUG 5694 /* should have ended on an mblk boundary */ 5695 if (bytes != round_len) { 5696 printf("bytes=0x%x, round_len=0x%x, req len=0x%x\n", 5697 bytes, round_len, args->data_len); 5698 printf("args=%p, args->mblk=%p, m=%p", (void *)args, 5699 (void *)args->mblk, (void *)m); 5700 ASSERT(bytes == round_len); 5701 } 5702 #endif 5703 if (iovcnt <= MAX_IOVECS) { 5704 iovp = iov; 5705 } else { 5706 iovp = kmem_alloc(sizeof (*iovp) * iovcnt, KM_SLEEP); 5707 } 5708 mblk_to_iov(args->mblk, iovcnt, iovp); 5709 } else if (args->rlist != NULL) { 5710 iovcnt = 1; 5711 iovp = iov; 5712 iovp->iov_base = (char *)((args->rlist)->u.c_daddr3); 5713 iovp->iov_len = args->data_len; 5714 } else { 5715 iovcnt = 1; 5716 iovp = iov; 5717 iovp->iov_base = args->data_val; 5718 iovp->iov_len = args->data_len; 5719 } 5720 5721 uio.uio_iov = iovp; 5722 uio.uio_iovcnt = iovcnt; 5723 5724 uio.uio_segflg = UIO_SYSSPACE; 5725 uio.uio_extflg = UIO_COPY_DEFAULT; 5726 uio.uio_loffset = args->offset; 5727 uio.uio_resid = args->data_len; 5728 uio.uio_llimit = curproc->p_fsz_ctl; 5729 rlimit = uio.uio_llimit - args->offset; 5730 if (rlimit < (u_offset_t)uio.uio_resid) 5731 uio.uio_resid = (int)rlimit; 5732 5733 if (args->stable == UNSTABLE4) 5734 ioflag = 0; 5735 else if (args->stable == FILE_SYNC4) 5736 ioflag = FSYNC; 5737 else if (args->stable == DATA_SYNC4) 5738 ioflag = FDSYNC; 5739 else { 5740 if (iovp != iov) 5741 kmem_free(iovp, sizeof (*iovp) * iovcnt); 5742 *cs->statusp = resp->status = NFS4ERR_INVAL; 5743 goto out; 5744 } 5745 5746 /* 5747 * We're changing creds because VM may fault and we need 5748 * the cred of the current thread to be used if quota 5749 * checking is enabled. 5750 */ 5751 savecred = curthread->t_cred; 5752 curthread->t_cred = cr; 5753 error = do_io(FWRITE, vp, &uio, ioflag, cr, &ct); 5754 curthread->t_cred = savecred; 5755 5756 if (iovp != iov) 5757 kmem_free(iovp, sizeof (*iovp) * iovcnt); 5758 5759 if (error) { 5760 *cs->statusp = resp->status = puterrno4(error); 5761 goto out; 5762 } 5763 5764 *cs->statusp = resp->status = NFS4_OK; 5765 resp->count = args->data_len - uio.uio_resid; 5766 5767 if (ioflag == 0) 5768 resp->committed = UNSTABLE4; 5769 else 5770 resp->committed = FILE_SYNC4; 5771 5772 resp->writeverf = Write4verf; 5773 5774 out: 5775 if (in_crit) 5776 nbl_end_crit(vp); 5777 5778 DTRACE_NFSV4_2(op__write__done, struct compound_state *, cs, 5779 WRITE4res *, resp); 5780 } 5781 5782 5783 /* XXX put in a header file */ 5784 extern int sec_svc_getcred(struct svc_req *, cred_t *, caddr_t *, int *); 5785 5786 void 5787 rfs4_compound(COMPOUND4args *args, COMPOUND4res *resp, struct exportinfo *exi, 5788 struct svc_req *req, cred_t *cr, int *rv) 5789 { 5790 uint_t i; 5791 struct compound_state cs; 5792 5793 if (rv != NULL) 5794 *rv = 0; 5795 rfs4_init_compound_state(&cs); 5796 /* 5797 * Form a reply tag by copying over the reqeuest tag. 5798 */ 5799 resp->tag.utf8string_val = 5800 kmem_alloc(args->tag.utf8string_len, KM_SLEEP); 5801 resp->tag.utf8string_len = args->tag.utf8string_len; 5802 bcopy(args->tag.utf8string_val, resp->tag.utf8string_val, 5803 resp->tag.utf8string_len); 5804 5805 cs.statusp = &resp->status; 5806 cs.req = req; 5807 5808 resp->status = utf8_name_verify(&(resp->tag)); 5809 if (resp->status != NFS4_OK) { 5810 resp->array_len = 0; 5811 resp->array = NULL; 5812 return; 5813 } 5814 5815 /* 5816 * XXX for now, minorversion should be zero 5817 */ 5818 if (args->minorversion != NFS4_MINORVERSION) { 5819 DTRACE_NFSV4_2(compound__start, struct compound_state *, 5820 &cs, COMPOUND4args *, args); 5821 resp->array_len = 0; 5822 resp->array = NULL; 5823 resp->status = NFS4ERR_MINOR_VERS_MISMATCH; 5824 DTRACE_NFSV4_2(compound__done, struct compound_state *, 5825 &cs, COMPOUND4res *, resp); 5826 return; 5827 } 5828 5829 ASSERT(exi == NULL); 5830 ASSERT(cr == NULL); 5831 5832 cr = crget(); 5833 ASSERT(cr != NULL); 5834 5835 if (sec_svc_getcred(req, cr, &cs.principal, &cs.nfsflavor) == 0) { 5836 DTRACE_NFSV4_2(compound__start, struct compound_state *, 5837 &cs, COMPOUND4args *, args); 5838 crfree(cr); 5839 DTRACE_NFSV4_2(compound__done, struct compound_state *, 5840 &cs, COMPOUND4res *, resp); 5841 svcerr_badcred(req->rq_xprt); 5842 if (rv != NULL) 5843 *rv = 1; 5844 return; 5845 } 5846 resp->array_len = args->array_len; 5847 resp->array = kmem_zalloc(args->array_len * sizeof (nfs_resop4), 5848 KM_SLEEP); 5849 5850 cs.basecr = cr; 5851 5852 DTRACE_NFSV4_2(compound__start, struct compound_state *, &cs, 5853 COMPOUND4args *, args); 5854 5855 /* 5856 * For now, NFS4 compound processing must be protected by 5857 * exported_lock because it can access more than one exportinfo 5858 * per compound and share/unshare can now change multiple 5859 * exinfo structs. The NFS2/3 code only refs 1 exportinfo 5860 * per proc (excluding public exinfo), and exi_count design 5861 * is sufficient to protect concurrent execution of NFS2/3 5862 * ops along with unexport. This lock will be removed as 5863 * part of the NFSv4 phase 2 namespace redesign work. 5864 */ 5865 rw_enter(&exported_lock, RW_READER); 5866 5867 /* 5868 * If this is the first compound we've seen, we need to start all 5869 * new instances' grace periods. 5870 */ 5871 if (rfs4_seen_first_compound == 0) { 5872 rfs4_grace_start_new(); 5873 /* 5874 * This must be set after rfs4_grace_start_new(), otherwise 5875 * another thread could proceed past here before the former 5876 * is finished. 5877 */ 5878 rfs4_seen_first_compound = 1; 5879 } 5880 5881 for (i = 0; i < args->array_len && cs.cont; i++) { 5882 nfs_argop4 *argop; 5883 nfs_resop4 *resop; 5884 uint_t op; 5885 5886 argop = &args->array[i]; 5887 resop = &resp->array[i]; 5888 resop->resop = argop->argop; 5889 op = (uint_t)resop->resop; 5890 5891 if (op < rfsv4disp_cnt) { 5892 /* 5893 * Count the individual ops here; NULL and COMPOUND 5894 * are counted in common_dispatch() 5895 */ 5896 rfsproccnt_v4_ptr[op].value.ui64++; 5897 5898 NFS4_DEBUG(rfs4_debug > 1, 5899 (CE_NOTE, "Executing %s", rfs4_op_string[op])); 5900 (*rfsv4disptab[op].dis_proc)(argop, resop, req, &cs); 5901 NFS4_DEBUG(rfs4_debug > 1, (CE_NOTE, "%s returned %d", 5902 rfs4_op_string[op], *cs.statusp)); 5903 if (*cs.statusp != NFS4_OK) 5904 cs.cont = FALSE; 5905 } else { 5906 /* 5907 * This is effectively dead code since XDR code 5908 * will have already returned BADXDR if op doesn't 5909 * decode to legal value. This only done for a 5910 * day when XDR code doesn't verify v4 opcodes. 5911 */ 5912 op = OP_ILLEGAL; 5913 rfsproccnt_v4_ptr[OP_ILLEGAL_IDX].value.ui64++; 5914 5915 rfs4_op_illegal(argop, resop, req, &cs); 5916 cs.cont = FALSE; 5917 } 5918 5919 /* 5920 * If not at last op, and if we are to stop, then 5921 * compact the results array. 5922 */ 5923 if ((i + 1) < args->array_len && !cs.cont) { 5924 nfs_resop4 *new_res = kmem_alloc( 5925 (i+1) * sizeof (nfs_resop4), KM_SLEEP); 5926 bcopy(resp->array, 5927 new_res, (i+1) * sizeof (nfs_resop4)); 5928 kmem_free(resp->array, 5929 args->array_len * sizeof (nfs_resop4)); 5930 5931 resp->array_len = i + 1; 5932 resp->array = new_res; 5933 } 5934 } 5935 5936 rw_exit(&exported_lock); 5937 5938 DTRACE_NFSV4_2(compound__done, struct compound_state *, &cs, 5939 COMPOUND4res *, resp); 5940 5941 if (cs.vp) 5942 VN_RELE(cs.vp); 5943 if (cs.saved_vp) 5944 VN_RELE(cs.saved_vp); 5945 if (cs.saved_fh.nfs_fh4_val) 5946 kmem_free(cs.saved_fh.nfs_fh4_val, NFS4_FHSIZE); 5947 5948 if (cs.basecr) 5949 crfree(cs.basecr); 5950 if (cs.cr) 5951 crfree(cs.cr); 5952 /* 5953 * done with this compound request, free the label 5954 */ 5955 5956 if (req->rq_label != NULL) { 5957 kmem_free(req->rq_label, sizeof (bslabel_t)); 5958 req->rq_label = NULL; 5959 } 5960 } 5961 5962 /* 5963 * XXX because of what appears to be duplicate calls to rfs4_compound_free 5964 * XXX zero out the tag and array values. Need to investigate why the 5965 * XXX calls occur, but at least prevent the panic for now. 5966 */ 5967 void 5968 rfs4_compound_free(COMPOUND4res *resp) 5969 { 5970 uint_t i; 5971 5972 if (resp->tag.utf8string_val) { 5973 UTF8STRING_FREE(resp->tag) 5974 } 5975 5976 for (i = 0; i < resp->array_len; i++) { 5977 nfs_resop4 *resop; 5978 uint_t op; 5979 5980 resop = &resp->array[i]; 5981 op = (uint_t)resop->resop; 5982 if (op < rfsv4disp_cnt) { 5983 (*rfsv4disptab[op].dis_resfree)(resop); 5984 } 5985 } 5986 if (resp->array != NULL) { 5987 kmem_free(resp->array, resp->array_len * sizeof (nfs_resop4)); 5988 } 5989 } 5990 5991 /* 5992 * Process the value of the compound request rpc flags, as a bit-AND 5993 * of the individual per-op flags (idempotent, allowork, publicfh_ok) 5994 */ 5995 void 5996 rfs4_compound_flagproc(COMPOUND4args *args, int *flagp) 5997 { 5998 int i; 5999 int flag = RPC_ALL; 6000 6001 for (i = 0; flag && i < args->array_len; i++) { 6002 uint_t op; 6003 6004 op = (uint_t)args->array[i].argop; 6005 6006 if (op < rfsv4disp_cnt) 6007 flag &= rfsv4disptab[op].dis_flags; 6008 else 6009 flag = 0; 6010 } 6011 *flagp = flag; 6012 } 6013 6014 nfsstat4 6015 rfs4_client_sysid(rfs4_client_t *cp, sysid_t *sp) 6016 { 6017 nfsstat4 e; 6018 6019 rfs4_dbe_lock(cp->rc_dbe); 6020 6021 if (cp->rc_sysidt != LM_NOSYSID) { 6022 *sp = cp->rc_sysidt; 6023 e = NFS4_OK; 6024 6025 } else if ((cp->rc_sysidt = lm_alloc_sysidt()) != LM_NOSYSID) { 6026 *sp = cp->rc_sysidt; 6027 e = NFS4_OK; 6028 6029 NFS4_DEBUG(rfs4_debug, (CE_NOTE, 6030 "rfs4_client_sysid: allocated 0x%x\n", *sp)); 6031 } else 6032 e = NFS4ERR_DELAY; 6033 6034 rfs4_dbe_unlock(cp->rc_dbe); 6035 return (e); 6036 } 6037 6038 #if defined(DEBUG) && ! defined(lint) 6039 static void lock_print(char *str, int operation, struct flock64 *flk) 6040 { 6041 char *op, *type; 6042 6043 switch (operation) { 6044 case F_GETLK: op = "F_GETLK"; 6045 break; 6046 case F_SETLK: op = "F_SETLK"; 6047 break; 6048 case F_SETLK_NBMAND: op = "F_SETLK_NBMAND"; 6049 break; 6050 default: op = "F_UNKNOWN"; 6051 break; 6052 } 6053 switch (flk->l_type) { 6054 case F_UNLCK: type = "F_UNLCK"; 6055 break; 6056 case F_RDLCK: type = "F_RDLCK"; 6057 break; 6058 case F_WRLCK: type = "F_WRLCK"; 6059 break; 6060 default: type = "F_UNKNOWN"; 6061 break; 6062 } 6063 6064 ASSERT(flk->l_whence == 0); 6065 cmn_err(CE_NOTE, "%s: %s, type = %s, off = %llx len = %llx pid = %d", 6066 str, op, type, (longlong_t)flk->l_start, 6067 flk->l_len ? (longlong_t)flk->l_len : ~0LL, flk->l_pid); 6068 } 6069 6070 #define LOCK_PRINT(d, s, t, f) if (d) lock_print(s, t, f) 6071 #else 6072 #define LOCK_PRINT(d, s, t, f) 6073 #endif 6074 6075 /*ARGSUSED*/ 6076 static bool_t 6077 creds_ok(cred_set_t cr_set, struct svc_req *req, struct compound_state *cs) 6078 { 6079 return (TRUE); 6080 } 6081 6082 /* 6083 * Look up the pathname using the vp in cs as the directory vnode. 6084 * cs->vp will be the vnode for the file on success 6085 */ 6086 6087 static nfsstat4 6088 rfs4_lookup(component4 *component, struct svc_req *req, 6089 struct compound_state *cs) 6090 { 6091 char *nm; 6092 uint32_t len; 6093 nfsstat4 status; 6094 struct sockaddr *ca; 6095 char *name; 6096 6097 if (cs->vp == NULL) { 6098 return (NFS4ERR_NOFILEHANDLE); 6099 } 6100 if (cs->vp->v_type != VDIR) { 6101 return (NFS4ERR_NOTDIR); 6102 } 6103 6104 status = utf8_dir_verify(component); 6105 if (status != NFS4_OK) 6106 return (status); 6107 6108 nm = utf8_to_fn(component, &len, NULL); 6109 if (nm == NULL) { 6110 return (NFS4ERR_INVAL); 6111 } 6112 6113 if (len > MAXNAMELEN) { 6114 kmem_free(nm, len); 6115 return (NFS4ERR_NAMETOOLONG); 6116 } 6117 6118 ca = (struct sockaddr *)svc_getrpccaller(req->rq_xprt)->buf; 6119 name = nfscmd_convname(ca, cs->exi, nm, NFSCMD_CONV_INBOUND, 6120 MAXPATHLEN + 1); 6121 6122 if (name == NULL) { 6123 kmem_free(nm, len); 6124 return (NFS4ERR_INVAL); 6125 } 6126 6127 status = do_rfs4_op_lookup(name, req, cs); 6128 6129 if (name != nm) 6130 kmem_free(name, MAXPATHLEN + 1); 6131 6132 kmem_free(nm, len); 6133 6134 return (status); 6135 } 6136 6137 static nfsstat4 6138 rfs4_lookupfile(component4 *component, struct svc_req *req, 6139 struct compound_state *cs, uint32_t access, change_info4 *cinfo) 6140 { 6141 nfsstat4 status; 6142 vnode_t *dvp = cs->vp; 6143 vattr_t bva, ava, fva; 6144 int error; 6145 6146 /* Get "before" change value */ 6147 bva.va_mask = AT_CTIME|AT_SEQ; 6148 error = VOP_GETATTR(dvp, &bva, 0, cs->cr, NULL); 6149 if (error) 6150 return (puterrno4(error)); 6151 6152 /* rfs4_lookup may VN_RELE directory */ 6153 VN_HOLD(dvp); 6154 6155 status = rfs4_lookup(component, req, cs); 6156 if (status != NFS4_OK) { 6157 VN_RELE(dvp); 6158 return (status); 6159 } 6160 6161 /* 6162 * Get "after" change value, if it fails, simply return the 6163 * before value. 6164 */ 6165 ava.va_mask = AT_CTIME|AT_SEQ; 6166 if (VOP_GETATTR(dvp, &ava, 0, cs->cr, NULL)) { 6167 ava.va_ctime = bva.va_ctime; 6168 ava.va_seq = 0; 6169 } 6170 VN_RELE(dvp); 6171 6172 /* 6173 * Validate the file is a file 6174 */ 6175 fva.va_mask = AT_TYPE|AT_MODE; 6176 error = VOP_GETATTR(cs->vp, &fva, 0, cs->cr, NULL); 6177 if (error) 6178 return (puterrno4(error)); 6179 6180 if (fva.va_type != VREG) { 6181 if (fva.va_type == VDIR) 6182 return (NFS4ERR_ISDIR); 6183 if (fva.va_type == VLNK) 6184 return (NFS4ERR_SYMLINK); 6185 return (NFS4ERR_INVAL); 6186 } 6187 6188 NFS4_SET_FATTR4_CHANGE(cinfo->before, bva.va_ctime); 6189 NFS4_SET_FATTR4_CHANGE(cinfo->after, ava.va_ctime); 6190 6191 /* 6192 * It is undefined if VOP_LOOKUP will change va_seq, so 6193 * cinfo.atomic = TRUE only if we have 6194 * non-zero va_seq's, and they have not changed. 6195 */ 6196 if (bva.va_seq && ava.va_seq && ava.va_seq == bva.va_seq) 6197 cinfo->atomic = TRUE; 6198 else 6199 cinfo->atomic = FALSE; 6200 6201 /* Check for mandatory locking */ 6202 cs->mandlock = MANDLOCK(cs->vp, fva.va_mode); 6203 return (check_open_access(access, cs, req)); 6204 } 6205 6206 static nfsstat4 6207 create_vnode(vnode_t *dvp, char *nm, vattr_t *vap, createmode4 mode, 6208 timespec32_t *mtime, cred_t *cr, vnode_t **vpp, bool_t *created) 6209 { 6210 int error; 6211 nfsstat4 status = NFS4_OK; 6212 vattr_t va; 6213 6214 tryagain: 6215 6216 /* 6217 * The file open mode used is VWRITE. If the client needs 6218 * some other semantic, then it should do the access checking 6219 * itself. It would have been nice to have the file open mode 6220 * passed as part of the arguments. 6221 */ 6222 6223 *created = TRUE; 6224 error = VOP_CREATE(dvp, nm, vap, EXCL, VWRITE, vpp, cr, 0, NULL, NULL); 6225 6226 if (error) { 6227 *created = FALSE; 6228 6229 /* 6230 * If we got something other than file already exists 6231 * then just return this error. Otherwise, we got 6232 * EEXIST. If we were doing a GUARDED create, then 6233 * just return this error. Otherwise, we need to 6234 * make sure that this wasn't a duplicate of an 6235 * exclusive create request. 6236 * 6237 * The assumption is made that a non-exclusive create 6238 * request will never return EEXIST. 6239 */ 6240 6241 if (error != EEXIST || mode == GUARDED4) { 6242 status = puterrno4(error); 6243 return (status); 6244 } 6245 error = VOP_LOOKUP(dvp, nm, vpp, NULL, 0, NULL, cr, 6246 NULL, NULL, NULL); 6247 6248 if (error) { 6249 /* 6250 * We couldn't find the file that we thought that 6251 * we just created. So, we'll just try creating 6252 * it again. 6253 */ 6254 if (error == ENOENT) 6255 goto tryagain; 6256 6257 status = puterrno4(error); 6258 return (status); 6259 } 6260 6261 if (mode == UNCHECKED4) { 6262 /* existing object must be regular file */ 6263 if ((*vpp)->v_type != VREG) { 6264 if ((*vpp)->v_type == VDIR) 6265 status = NFS4ERR_ISDIR; 6266 else if ((*vpp)->v_type == VLNK) 6267 status = NFS4ERR_SYMLINK; 6268 else 6269 status = NFS4ERR_INVAL; 6270 VN_RELE(*vpp); 6271 return (status); 6272 } 6273 6274 return (NFS4_OK); 6275 } 6276 6277 /* Check for duplicate request */ 6278 ASSERT(mtime != 0); 6279 va.va_mask = AT_MTIME; 6280 error = VOP_GETATTR(*vpp, &va, 0, cr, NULL); 6281 if (!error) { 6282 /* We found the file */ 6283 if (va.va_mtime.tv_sec != mtime->tv_sec || 6284 va.va_mtime.tv_nsec != mtime->tv_nsec) { 6285 /* but its not our creation */ 6286 VN_RELE(*vpp); 6287 return (NFS4ERR_EXIST); 6288 } 6289 *created = TRUE; /* retrans of create == created */ 6290 return (NFS4_OK); 6291 } 6292 VN_RELE(*vpp); 6293 return (NFS4ERR_EXIST); 6294 } 6295 6296 return (NFS4_OK); 6297 } 6298 6299 static nfsstat4 6300 check_open_access(uint32_t access, struct compound_state *cs, 6301 struct svc_req *req) 6302 { 6303 int error; 6304 vnode_t *vp; 6305 bool_t readonly; 6306 cred_t *cr = cs->cr; 6307 6308 /* For now we don't allow mandatory locking as per V2/V3 */ 6309 if (cs->access == CS_ACCESS_DENIED || cs->mandlock) { 6310 return (NFS4ERR_ACCESS); 6311 } 6312 6313 vp = cs->vp; 6314 ASSERT(cr != NULL && vp->v_type == VREG); 6315 6316 /* 6317 * If the file system is exported read only and we are trying 6318 * to open for write, then return NFS4ERR_ROFS 6319 */ 6320 6321 readonly = rdonly4(cs->exi, cs->vp, req); 6322 6323 if ((access & OPEN4_SHARE_ACCESS_WRITE) && readonly) 6324 return (NFS4ERR_ROFS); 6325 6326 if (access & OPEN4_SHARE_ACCESS_READ) { 6327 if ((VOP_ACCESS(vp, VREAD, 0, cr, NULL) != 0) && 6328 (VOP_ACCESS(vp, VEXEC, 0, cr, NULL) != 0)) { 6329 return (NFS4ERR_ACCESS); 6330 } 6331 } 6332 6333 if (access & OPEN4_SHARE_ACCESS_WRITE) { 6334 error = VOP_ACCESS(vp, VWRITE, 0, cr, NULL); 6335 if (error) 6336 return (NFS4ERR_ACCESS); 6337 } 6338 6339 return (NFS4_OK); 6340 } 6341 6342 static nfsstat4 6343 rfs4_createfile(OPEN4args *args, struct svc_req *req, struct compound_state *cs, 6344 change_info4 *cinfo, bitmap4 *attrset, clientid4 clientid) 6345 { 6346 struct nfs4_svgetit_arg sarg; 6347 struct nfs4_ntov_table ntov; 6348 6349 bool_t ntov_table_init = FALSE; 6350 struct statvfs64 sb; 6351 nfsstat4 status; 6352 vnode_t *vp; 6353 vattr_t bva, ava, iva, cva, *vap; 6354 vnode_t *dvp; 6355 timespec32_t *mtime; 6356 char *nm = NULL; 6357 uint_t buflen; 6358 bool_t created; 6359 bool_t setsize = FALSE; 6360 len_t reqsize; 6361 int error; 6362 bool_t trunc; 6363 caller_context_t ct; 6364 component4 *component; 6365 bslabel_t *clabel; 6366 struct sockaddr *ca; 6367 char *name = NULL; 6368 6369 sarg.sbp = &sb; 6370 sarg.is_referral = B_FALSE; 6371 6372 dvp = cs->vp; 6373 6374 /* Check if the file system is read only */ 6375 if (rdonly4(cs->exi, dvp, req)) 6376 return (NFS4ERR_ROFS); 6377 6378 /* check the label of including directory */ 6379 if (is_system_labeled()) { 6380 ASSERT(req->rq_label != NULL); 6381 clabel = req->rq_label; 6382 DTRACE_PROBE2(tx__rfs4__log__info__opremove__clabel, char *, 6383 "got client label from request(1)", 6384 struct svc_req *, req); 6385 if (!blequal(&l_admin_low->tsl_label, clabel)) { 6386 if (!do_rfs_label_check(clabel, dvp, EQUALITY_CHECK, 6387 cs->exi)) { 6388 return (NFS4ERR_ACCESS); 6389 } 6390 } 6391 } 6392 6393 /* 6394 * Get the last component of path name in nm. cs will reference 6395 * the including directory on success. 6396 */ 6397 component = &args->open_claim4_u.file; 6398 status = utf8_dir_verify(component); 6399 if (status != NFS4_OK) 6400 return (status); 6401 6402 nm = utf8_to_fn(component, &buflen, NULL); 6403 6404 if (nm == NULL) 6405 return (NFS4ERR_RESOURCE); 6406 6407 if (buflen > MAXNAMELEN) { 6408 kmem_free(nm, buflen); 6409 return (NFS4ERR_NAMETOOLONG); 6410 } 6411 6412 bva.va_mask = AT_TYPE|AT_CTIME|AT_SEQ; 6413 error = VOP_GETATTR(dvp, &bva, 0, cs->cr, NULL); 6414 if (error) { 6415 kmem_free(nm, buflen); 6416 return (puterrno4(error)); 6417 } 6418 6419 if (bva.va_type != VDIR) { 6420 kmem_free(nm, buflen); 6421 return (NFS4ERR_NOTDIR); 6422 } 6423 6424 NFS4_SET_FATTR4_CHANGE(cinfo->before, bva.va_ctime) 6425 6426 switch (args->mode) { 6427 case GUARDED4: 6428 /*FALLTHROUGH*/ 6429 case UNCHECKED4: 6430 nfs4_ntov_table_init(&ntov); 6431 ntov_table_init = TRUE; 6432 6433 *attrset = 0; 6434 status = do_rfs4_set_attrs(attrset, 6435 &args->createhow4_u.createattrs, 6436 cs, &sarg, &ntov, NFS4ATTR_SETIT); 6437 6438 if (status == NFS4_OK && (sarg.vap->va_mask & AT_TYPE) && 6439 sarg.vap->va_type != VREG) { 6440 if (sarg.vap->va_type == VDIR) 6441 status = NFS4ERR_ISDIR; 6442 else if (sarg.vap->va_type == VLNK) 6443 status = NFS4ERR_SYMLINK; 6444 else 6445 status = NFS4ERR_INVAL; 6446 } 6447 6448 if (status != NFS4_OK) { 6449 kmem_free(nm, buflen); 6450 nfs4_ntov_table_free(&ntov, &sarg); 6451 *attrset = 0; 6452 return (status); 6453 } 6454 6455 vap = sarg.vap; 6456 vap->va_type = VREG; 6457 vap->va_mask |= AT_TYPE; 6458 6459 if ((vap->va_mask & AT_MODE) == 0) { 6460 vap->va_mask |= AT_MODE; 6461 vap->va_mode = (mode_t)0600; 6462 } 6463 6464 if (vap->va_mask & AT_SIZE) { 6465 6466 /* Disallow create with a non-zero size */ 6467 6468 if ((reqsize = sarg.vap->va_size) != 0) { 6469 kmem_free(nm, buflen); 6470 nfs4_ntov_table_free(&ntov, &sarg); 6471 *attrset = 0; 6472 return (NFS4ERR_INVAL); 6473 } 6474 setsize = TRUE; 6475 } 6476 break; 6477 6478 case EXCLUSIVE4: 6479 /* prohibit EXCL create of named attributes */ 6480 if (dvp->v_flag & V_XATTRDIR) { 6481 kmem_free(nm, buflen); 6482 *attrset = 0; 6483 return (NFS4ERR_INVAL); 6484 } 6485 6486 cva.va_mask = AT_TYPE | AT_MTIME | AT_MODE; 6487 cva.va_type = VREG; 6488 /* 6489 * Ensure no time overflows. Assumes underlying 6490 * filesystem supports at least 32 bits. 6491 * Truncate nsec to usec resolution to allow valid 6492 * compares even if the underlying filesystem truncates. 6493 */ 6494 mtime = (timespec32_t *)&args->createhow4_u.createverf; 6495 cva.va_mtime.tv_sec = mtime->tv_sec % TIME32_MAX; 6496 cva.va_mtime.tv_nsec = (mtime->tv_nsec / 1000) * 1000; 6497 cva.va_mode = (mode_t)0; 6498 vap = &cva; 6499 6500 /* 6501 * For EXCL create, attrset is set to the server attr 6502 * used to cache the client's verifier. 6503 */ 6504 *attrset = FATTR4_TIME_MODIFY_MASK; 6505 break; 6506 } 6507 6508 ca = (struct sockaddr *)svc_getrpccaller(req->rq_xprt)->buf; 6509 name = nfscmd_convname(ca, cs->exi, nm, NFSCMD_CONV_INBOUND, 6510 MAXPATHLEN + 1); 6511 6512 if (name == NULL) { 6513 kmem_free(nm, buflen); 6514 return (NFS4ERR_SERVERFAULT); 6515 } 6516 6517 status = create_vnode(dvp, name, vap, args->mode, mtime, 6518 cs->cr, &vp, &created); 6519 if (nm != name) 6520 kmem_free(name, MAXPATHLEN + 1); 6521 kmem_free(nm, buflen); 6522 6523 if (status != NFS4_OK) { 6524 if (ntov_table_init) 6525 nfs4_ntov_table_free(&ntov, &sarg); 6526 *attrset = 0; 6527 return (status); 6528 } 6529 6530 trunc = (setsize && !created); 6531 6532 if (args->mode != EXCLUSIVE4) { 6533 bitmap4 createmask = args->createhow4_u.createattrs.attrmask; 6534 6535 /* 6536 * True verification that object was created with correct 6537 * attrs is impossible. The attrs could have been changed 6538 * immediately after object creation. If attributes did 6539 * not verify, the only recourse for the server is to 6540 * destroy the object. Maybe if some attrs (like gid) 6541 * are set incorrectly, the object should be destroyed; 6542 * however, seems bad as a default policy. Do we really 6543 * want to destroy an object over one of the times not 6544 * verifying correctly? For these reasons, the server 6545 * currently sets bits in attrset for createattrs 6546 * that were set; however, no verification is done. 6547 * 6548 * vmask_to_nmask accounts for vattr bits set on create 6549 * [do_rfs4_set_attrs() only sets resp bits for 6550 * non-vattr/vfs bits.] 6551 * Mask off any bits we set by default so as not to return 6552 * more attrset bits than were requested in createattrs 6553 */ 6554 if (created) { 6555 nfs4_vmask_to_nmask(sarg.vap->va_mask, attrset); 6556 *attrset &= createmask; 6557 } else { 6558 /* 6559 * We did not create the vnode (we tried but it 6560 * already existed). In this case, the only createattr 6561 * that the spec allows the server to set is size, 6562 * and even then, it can only be set if it is 0. 6563 */ 6564 *attrset = 0; 6565 if (trunc) 6566 *attrset = FATTR4_SIZE_MASK; 6567 } 6568 } 6569 if (ntov_table_init) 6570 nfs4_ntov_table_free(&ntov, &sarg); 6571 6572 /* 6573 * Get the initial "after" sequence number, if it fails, 6574 * set to zero, time to before. 6575 */ 6576 iva.va_mask = AT_CTIME|AT_SEQ; 6577 if (VOP_GETATTR(dvp, &iva, 0, cs->cr, NULL)) { 6578 iva.va_seq = 0; 6579 iva.va_ctime = bva.va_ctime; 6580 } 6581 6582 /* 6583 * create_vnode attempts to create the file exclusive, 6584 * if it already exists the VOP_CREATE will fail and 6585 * may not increase va_seq. It is atomic if 6586 * we haven't changed the directory, but if it has changed 6587 * we don't know what changed it. 6588 */ 6589 if (!created) { 6590 if (bva.va_seq && iva.va_seq && 6591 bva.va_seq == iva.va_seq) 6592 cinfo->atomic = TRUE; 6593 else 6594 cinfo->atomic = FALSE; 6595 NFS4_SET_FATTR4_CHANGE(cinfo->after, iva.va_ctime); 6596 } else { 6597 /* 6598 * The entry was created, we need to sync the 6599 * directory metadata. 6600 */ 6601 (void) VOP_FSYNC(dvp, 0, cs->cr, NULL); 6602 6603 /* 6604 * Get "after" change value, if it fails, simply return the 6605 * before value. 6606 */ 6607 ava.va_mask = AT_CTIME|AT_SEQ; 6608 if (VOP_GETATTR(dvp, &ava, 0, cs->cr, NULL)) { 6609 ava.va_ctime = bva.va_ctime; 6610 ava.va_seq = 0; 6611 } 6612 6613 NFS4_SET_FATTR4_CHANGE(cinfo->after, ava.va_ctime); 6614 6615 /* 6616 * The cinfo->atomic = TRUE only if we have 6617 * non-zero va_seq's, and it has incremented by exactly one 6618 * during the create_vnode and it didn't 6619 * change during the VOP_FSYNC. 6620 */ 6621 if (bva.va_seq && iva.va_seq && ava.va_seq && 6622 iva.va_seq == (bva.va_seq + 1) && iva.va_seq == ava.va_seq) 6623 cinfo->atomic = TRUE; 6624 else 6625 cinfo->atomic = FALSE; 6626 } 6627 6628 /* Check for mandatory locking and that the size gets set. */ 6629 cva.va_mask = AT_MODE; 6630 if (setsize) 6631 cva.va_mask |= AT_SIZE; 6632 6633 /* Assume the worst */ 6634 cs->mandlock = TRUE; 6635 6636 if (VOP_GETATTR(vp, &cva, 0, cs->cr, NULL) == 0) { 6637 cs->mandlock = MANDLOCK(cs->vp, cva.va_mode); 6638 6639 /* 6640 * Truncate the file if necessary; this would be 6641 * the case for create over an existing file. 6642 */ 6643 6644 if (trunc) { 6645 int in_crit = 0; 6646 rfs4_file_t *fp; 6647 bool_t create = FALSE; 6648 6649 /* 6650 * We are writing over an existing file. 6651 * Check to see if we need to recall a delegation. 6652 */ 6653 rfs4_hold_deleg_policy(); 6654 if ((fp = rfs4_findfile(vp, NULL, &create)) != NULL) { 6655 if (rfs4_check_delegated_byfp(FWRITE, fp, 6656 (reqsize == 0), FALSE, FALSE, &clientid)) { 6657 rfs4_file_rele(fp); 6658 rfs4_rele_deleg_policy(); 6659 VN_RELE(vp); 6660 *attrset = 0; 6661 return (NFS4ERR_DELAY); 6662 } 6663 rfs4_file_rele(fp); 6664 } 6665 rfs4_rele_deleg_policy(); 6666 6667 if (nbl_need_check(vp)) { 6668 in_crit = 1; 6669 6670 ASSERT(reqsize == 0); 6671 6672 nbl_start_crit(vp, RW_READER); 6673 if (nbl_conflict(vp, NBL_WRITE, 0, 6674 cva.va_size, 0, NULL)) { 6675 in_crit = 0; 6676 nbl_end_crit(vp); 6677 VN_RELE(vp); 6678 *attrset = 0; 6679 return (NFS4ERR_ACCESS); 6680 } 6681 } 6682 ct.cc_sysid = 0; 6683 ct.cc_pid = 0; 6684 ct.cc_caller_id = nfs4_srv_caller_id; 6685 ct.cc_flags = CC_DONTBLOCK; 6686 6687 cva.va_mask = AT_SIZE; 6688 cva.va_size = reqsize; 6689 (void) VOP_SETATTR(vp, &cva, 0, cs->cr, &ct); 6690 if (in_crit) 6691 nbl_end_crit(vp); 6692 } 6693 } 6694 6695 error = makefh4(&cs->fh, vp, cs->exi); 6696 6697 /* 6698 * Force modified data and metadata out to stable storage. 6699 */ 6700 (void) VOP_FSYNC(vp, FNODSYNC, cs->cr, NULL); 6701 6702 if (error) { 6703 VN_RELE(vp); 6704 *attrset = 0; 6705 return (puterrno4(error)); 6706 } 6707 6708 /* if parent dir is attrdir, set namedattr fh flag */ 6709 if (dvp->v_flag & V_XATTRDIR) 6710 set_fh4_flag(&cs->fh, FH4_NAMEDATTR); 6711 6712 if (cs->vp) 6713 VN_RELE(cs->vp); 6714 6715 cs->vp = vp; 6716 6717 /* 6718 * if we did not create the file, we will need to check 6719 * the access bits on the file 6720 */ 6721 6722 if (!created) { 6723 if (setsize) 6724 args->share_access |= OPEN4_SHARE_ACCESS_WRITE; 6725 status = check_open_access(args->share_access, cs, req); 6726 if (status != NFS4_OK) 6727 *attrset = 0; 6728 } 6729 return (status); 6730 } 6731 6732 /*ARGSUSED*/ 6733 static void 6734 rfs4_do_open(struct compound_state *cs, struct svc_req *req, 6735 rfs4_openowner_t *oo, delegreq_t deleg, 6736 uint32_t access, uint32_t deny, 6737 OPEN4res *resp, int deleg_cur) 6738 { 6739 /* XXX Currently not using req */ 6740 rfs4_state_t *sp; 6741 rfs4_file_t *fp; 6742 bool_t screate = TRUE; 6743 bool_t fcreate = TRUE; 6744 uint32_t open_a, share_a; 6745 uint32_t open_d, share_d; 6746 rfs4_deleg_state_t *dsp; 6747 sysid_t sysid; 6748 nfsstat4 status; 6749 caller_context_t ct; 6750 int fflags = 0; 6751 int recall = 0; 6752 int err; 6753 int first_open; 6754 6755 /* get the file struct and hold a lock on it during initial open */ 6756 fp = rfs4_findfile_withlock(cs->vp, &cs->fh, &fcreate); 6757 if (fp == NULL) { 6758 resp->status = NFS4ERR_RESOURCE; 6759 DTRACE_PROBE1(nfss__e__do__open1, nfsstat4, resp->status); 6760 return; 6761 } 6762 6763 sp = rfs4_findstate_by_owner_file(oo, fp, &screate); 6764 if (sp == NULL) { 6765 resp->status = NFS4ERR_RESOURCE; 6766 DTRACE_PROBE1(nfss__e__do__open2, nfsstat4, resp->status); 6767 /* No need to keep any reference */ 6768 rw_exit(&fp->rf_file_rwlock); 6769 rfs4_file_rele(fp); 6770 return; 6771 } 6772 6773 /* try to get the sysid before continuing */ 6774 if ((status = rfs4_client_sysid(oo->ro_client, &sysid)) != NFS4_OK) { 6775 resp->status = status; 6776 rfs4_file_rele(fp); 6777 /* Not a fully formed open; "close" it */ 6778 if (screate == TRUE) 6779 rfs4_state_close(sp, FALSE, FALSE, cs->cr); 6780 rfs4_state_rele(sp); 6781 return; 6782 } 6783 6784 /* Calculate the fflags for this OPEN. */ 6785 if (access & OPEN4_SHARE_ACCESS_READ) 6786 fflags |= FREAD; 6787 if (access & OPEN4_SHARE_ACCESS_WRITE) 6788 fflags |= FWRITE; 6789 6790 rfs4_dbe_lock(sp->rs_dbe); 6791 6792 /* 6793 * Calculate the new deny and access mode that this open is adding to 6794 * the file for this open owner; 6795 */ 6796 open_d = (deny & ~sp->rs_open_deny); 6797 open_a = (access & ~sp->rs_open_access); 6798 6799 /* 6800 * Calculate the new share access and share deny modes that this open 6801 * is adding to the file for this open owner; 6802 */ 6803 share_a = (access & ~sp->rs_share_access); 6804 share_d = (deny & ~sp->rs_share_deny); 6805 6806 first_open = (sp->rs_open_access & OPEN4_SHARE_ACCESS_BOTH) == 0; 6807 6808 /* 6809 * Check to see the client has already sent an open for this 6810 * open owner on this file with the same share/deny modes. 6811 * If so, we don't need to check for a conflict and we don't 6812 * need to add another shrlock. If not, then we need to 6813 * check for conflicts in deny and access before checking for 6814 * conflicts in delegation. We don't want to recall a 6815 * delegation based on an open that will eventually fail based 6816 * on shares modes. 6817 */ 6818 6819 if (share_a || share_d) { 6820 if ((err = rfs4_share(sp, access, deny)) != 0) { 6821 rfs4_dbe_unlock(sp->rs_dbe); 6822 resp->status = err; 6823 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 return; 6830 } 6831 } 6832 6833 rfs4_dbe_lock(fp->rf_dbe); 6834 6835 /* 6836 * Check to see if this file is delegated and if so, if a 6837 * recall needs to be done. 6838 */ 6839 if (rfs4_check_recall(sp, access)) { 6840 rfs4_dbe_unlock(fp->rf_dbe); 6841 rfs4_dbe_unlock(sp->rs_dbe); 6842 rfs4_recall_deleg(fp, FALSE, sp->rs_owner->ro_client); 6843 delay(NFS4_DELEGATION_CONFLICT_DELAY); 6844 rfs4_dbe_lock(sp->rs_dbe); 6845 6846 /* if state closed while lock was dropped */ 6847 if (sp->rs_closed) { 6848 if (share_a || share_d) 6849 (void) rfs4_unshare(sp); 6850 rfs4_dbe_unlock(sp->rs_dbe); 6851 rfs4_file_rele(fp); 6852 /* Not a fully formed open; "close" it */ 6853 if (screate == TRUE) 6854 rfs4_state_close(sp, FALSE, FALSE, cs->cr); 6855 rfs4_state_rele(sp); 6856 resp->status = NFS4ERR_OLD_STATEID; 6857 return; 6858 } 6859 6860 rfs4_dbe_lock(fp->rf_dbe); 6861 /* Let's see if the delegation was returned */ 6862 if (rfs4_check_recall(sp, access)) { 6863 rfs4_dbe_unlock(fp->rf_dbe); 6864 if (share_a || share_d) 6865 (void) rfs4_unshare(sp); 6866 rfs4_dbe_unlock(sp->rs_dbe); 6867 rfs4_file_rele(fp); 6868 rfs4_update_lease(sp->rs_owner->ro_client); 6869 6870 /* Not a fully formed open; "close" it */ 6871 if (screate == TRUE) 6872 rfs4_state_close(sp, FALSE, FALSE, cs->cr); 6873 rfs4_state_rele(sp); 6874 resp->status = NFS4ERR_DELAY; 6875 return; 6876 } 6877 } 6878 /* 6879 * the share check passed and any delegation conflict has been 6880 * taken care of, now call vop_open. 6881 * if this is the first open then call vop_open with fflags. 6882 * if not, call vn_open_upgrade with just the upgrade flags. 6883 * 6884 * if the file has been opened already, it will have the current 6885 * access mode in the state struct. if it has no share access, then 6886 * this is a new open. 6887 * 6888 * However, if this is open with CLAIM_DLEGATE_CUR, then don't 6889 * call VOP_OPEN(), just do the open upgrade. 6890 */ 6891 if (first_open && !deleg_cur) { 6892 ct.cc_sysid = sysid; 6893 ct.cc_pid = rfs4_dbe_getid(sp->rs_owner->ro_dbe); 6894 ct.cc_caller_id = nfs4_srv_caller_id; 6895 ct.cc_flags = CC_DONTBLOCK; 6896 err = VOP_OPEN(&cs->vp, fflags, cs->cr, &ct); 6897 if (err) { 6898 rfs4_dbe_unlock(fp->rf_dbe); 6899 if (share_a || share_d) 6900 (void) rfs4_unshare(sp); 6901 rfs4_dbe_unlock(sp->rs_dbe); 6902 rfs4_file_rele(fp); 6903 6904 /* Not a fully formed open; "close" it */ 6905 if (screate == TRUE) 6906 rfs4_state_close(sp, FALSE, FALSE, cs->cr); 6907 rfs4_state_rele(sp); 6908 /* check if a monitor detected a delegation conflict */ 6909 if (err == EAGAIN && (ct.cc_flags & CC_WOULDBLOCK)) 6910 resp->status = NFS4ERR_DELAY; 6911 else 6912 resp->status = NFS4ERR_SERVERFAULT; 6913 return; 6914 } 6915 } else { /* open upgrade */ 6916 /* 6917 * calculate the fflags for the new mode that is being added 6918 * by this upgrade. 6919 */ 6920 fflags = 0; 6921 if (open_a & OPEN4_SHARE_ACCESS_READ) 6922 fflags |= FREAD; 6923 if (open_a & OPEN4_SHARE_ACCESS_WRITE) 6924 fflags |= FWRITE; 6925 vn_open_upgrade(cs->vp, fflags); 6926 } 6927 sp->rs_open_access |= access; 6928 sp->rs_open_deny |= deny; 6929 6930 if (open_d & OPEN4_SHARE_DENY_READ) 6931 fp->rf_deny_read++; 6932 if (open_d & OPEN4_SHARE_DENY_WRITE) 6933 fp->rf_deny_write++; 6934 fp->rf_share_deny |= deny; 6935 6936 if (open_a & OPEN4_SHARE_ACCESS_READ) 6937 fp->rf_access_read++; 6938 if (open_a & OPEN4_SHARE_ACCESS_WRITE) 6939 fp->rf_access_write++; 6940 fp->rf_share_access |= access; 6941 6942 /* 6943 * Check for delegation here. if the deleg argument is not 6944 * DELEG_ANY, then this is a reclaim from a client and 6945 * we must honor the delegation requested. If necessary we can 6946 * set the recall flag. 6947 */ 6948 6949 dsp = rfs4_grant_delegation(deleg, sp, &recall); 6950 6951 cs->deleg = (fp->rf_dinfo.rd_dtype == OPEN_DELEGATE_WRITE); 6952 6953 next_stateid(&sp->rs_stateid); 6954 6955 resp->stateid = sp->rs_stateid.stateid; 6956 6957 rfs4_dbe_unlock(fp->rf_dbe); 6958 rfs4_dbe_unlock(sp->rs_dbe); 6959 6960 if (dsp) { 6961 rfs4_set_deleg_response(dsp, &resp->delegation, NULL, recall); 6962 rfs4_deleg_state_rele(dsp); 6963 } 6964 6965 rfs4_file_rele(fp); 6966 rfs4_state_rele(sp); 6967 6968 resp->status = NFS4_OK; 6969 } 6970 6971 /*ARGSUSED*/ 6972 static void 6973 rfs4_do_opennull(struct compound_state *cs, struct svc_req *req, 6974 OPEN4args *args, rfs4_openowner_t *oo, OPEN4res *resp) 6975 { 6976 change_info4 *cinfo = &resp->cinfo; 6977 bitmap4 *attrset = &resp->attrset; 6978 6979 if (args->opentype == OPEN4_NOCREATE) 6980 resp->status = rfs4_lookupfile(&args->open_claim4_u.file, 6981 req, cs, args->share_access, cinfo); 6982 else { 6983 /* inhibit delegation grants during exclusive create */ 6984 6985 if (args->mode == EXCLUSIVE4) 6986 rfs4_disable_delegation(); 6987 6988 resp->status = rfs4_createfile(args, req, cs, cinfo, attrset, 6989 oo->ro_client->rc_clientid); 6990 } 6991 6992 if (resp->status == NFS4_OK) { 6993 6994 /* cs->vp cs->fh now reference the desired file */ 6995 6996 rfs4_do_open(cs, req, oo, 6997 oo->ro_need_confirm ? DELEG_NONE : DELEG_ANY, 6998 args->share_access, args->share_deny, resp, 0); 6999 7000 /* 7001 * If rfs4_createfile set attrset, we must 7002 * clear this attrset before the response is copied. 7003 */ 7004 if (resp->status != NFS4_OK && resp->attrset) { 7005 resp->attrset = 0; 7006 } 7007 } 7008 else 7009 *cs->statusp = resp->status; 7010 7011 if (args->mode == EXCLUSIVE4) 7012 rfs4_enable_delegation(); 7013 } 7014 7015 /*ARGSUSED*/ 7016 static void 7017 rfs4_do_openprev(struct compound_state *cs, struct svc_req *req, 7018 OPEN4args *args, rfs4_openowner_t *oo, OPEN4res *resp) 7019 { 7020 change_info4 *cinfo = &resp->cinfo; 7021 vattr_t va; 7022 vtype_t v_type = cs->vp->v_type; 7023 int error = 0; 7024 7025 /* Verify that we have a regular file */ 7026 if (v_type != VREG) { 7027 if (v_type == VDIR) 7028 resp->status = NFS4ERR_ISDIR; 7029 else if (v_type == VLNK) 7030 resp->status = NFS4ERR_SYMLINK; 7031 else 7032 resp->status = NFS4ERR_INVAL; 7033 return; 7034 } 7035 7036 va.va_mask = AT_MODE|AT_UID; 7037 error = VOP_GETATTR(cs->vp, &va, 0, cs->cr, NULL); 7038 if (error) { 7039 resp->status = puterrno4(error); 7040 return; 7041 } 7042 7043 cs->mandlock = MANDLOCK(cs->vp, va.va_mode); 7044 7045 /* 7046 * Check if we have access to the file, Note the the file 7047 * could have originally been open UNCHECKED or GUARDED 7048 * with mode bits that will now fail, but there is nothing 7049 * we can really do about that except in the case that the 7050 * owner of the file is the one requesting the open. 7051 */ 7052 if (crgetuid(cs->cr) != va.va_uid) { 7053 resp->status = check_open_access(args->share_access, cs, req); 7054 if (resp->status != NFS4_OK) { 7055 return; 7056 } 7057 } 7058 7059 /* 7060 * cinfo on a CLAIM_PREVIOUS is undefined, initialize to zero 7061 */ 7062 cinfo->before = 0; 7063 cinfo->after = 0; 7064 cinfo->atomic = FALSE; 7065 7066 rfs4_do_open(cs, req, oo, 7067 NFS4_DELEG4TYPE2REQTYPE(args->open_claim4_u.delegate_type), 7068 args->share_access, args->share_deny, resp, 0); 7069 } 7070 7071 static void 7072 rfs4_do_opendelcur(struct compound_state *cs, struct svc_req *req, 7073 OPEN4args *args, rfs4_openowner_t *oo, OPEN4res *resp) 7074 { 7075 int error; 7076 nfsstat4 status; 7077 stateid4 stateid = 7078 args->open_claim4_u.delegate_cur_info.delegate_stateid; 7079 rfs4_deleg_state_t *dsp; 7080 7081 /* 7082 * Find the state info from the stateid and confirm that the 7083 * file is delegated. If the state openowner is the same as 7084 * the supplied openowner we're done. If not, get the file 7085 * info from the found state info. Use that file info to 7086 * create the state for this lock owner. Note solaris doen't 7087 * really need the pathname to find the file. We may want to 7088 * lookup the pathname and make sure that the vp exist and 7089 * matches the vp in the file structure. However it is 7090 * possible that the pathname nolonger exists (local process 7091 * unlinks the file), so this may not be that useful. 7092 */ 7093 7094 status = rfs4_get_deleg_state(&stateid, &dsp); 7095 if (status != NFS4_OK) { 7096 resp->status = status; 7097 return; 7098 } 7099 7100 ASSERT(dsp->rds_finfo->rf_dinfo.rd_dtype != OPEN_DELEGATE_NONE); 7101 7102 /* 7103 * New lock owner, create state. Since this was probably called 7104 * in response to a CB_RECALL we set deleg to DELEG_NONE 7105 */ 7106 7107 ASSERT(cs->vp != NULL); 7108 VN_RELE(cs->vp); 7109 VN_HOLD(dsp->rds_finfo->rf_vp); 7110 cs->vp = dsp->rds_finfo->rf_vp; 7111 7112 if (error = makefh4(&cs->fh, cs->vp, cs->exi)) { 7113 rfs4_deleg_state_rele(dsp); 7114 *cs->statusp = resp->status = puterrno4(error); 7115 return; 7116 } 7117 7118 /* Mark progress for delegation returns */ 7119 dsp->rds_finfo->rf_dinfo.rd_time_lastwrite = gethrestime_sec(); 7120 rfs4_deleg_state_rele(dsp); 7121 rfs4_do_open(cs, req, oo, DELEG_NONE, 7122 args->share_access, args->share_deny, resp, 1); 7123 } 7124 7125 /*ARGSUSED*/ 7126 static void 7127 rfs4_do_opendelprev(struct compound_state *cs, struct svc_req *req, 7128 OPEN4args *args, rfs4_openowner_t *oo, OPEN4res *resp) 7129 { 7130 /* 7131 * Lookup the pathname, it must already exist since this file 7132 * was delegated. 7133 * 7134 * Find the file and state info for this vp and open owner pair. 7135 * check that they are in fact delegated. 7136 * check that the state access and deny modes are the same. 7137 * 7138 * Return the delgation possibly seting the recall flag. 7139 */ 7140 rfs4_file_t *fp; 7141 rfs4_state_t *sp; 7142 bool_t create = FALSE; 7143 bool_t dcreate = FALSE; 7144 rfs4_deleg_state_t *dsp; 7145 nfsace4 *ace; 7146 7147 /* Note we ignore oflags */ 7148 resp->status = rfs4_lookupfile(&args->open_claim4_u.file_delegate_prev, 7149 req, cs, args->share_access, &resp->cinfo); 7150 7151 if (resp->status != NFS4_OK) { 7152 return; 7153 } 7154 7155 /* get the file struct and hold a lock on it during initial open */ 7156 fp = rfs4_findfile_withlock(cs->vp, NULL, &create); 7157 if (fp == NULL) { 7158 resp->status = NFS4ERR_RESOURCE; 7159 DTRACE_PROBE1(nfss__e__do_opendelprev1, nfsstat4, resp->status); 7160 return; 7161 } 7162 7163 sp = rfs4_findstate_by_owner_file(oo, fp, &create); 7164 if (sp == NULL) { 7165 resp->status = NFS4ERR_SERVERFAULT; 7166 DTRACE_PROBE1(nfss__e__do_opendelprev2, nfsstat4, resp->status); 7167 rw_exit(&fp->rf_file_rwlock); 7168 rfs4_file_rele(fp); 7169 return; 7170 } 7171 7172 rfs4_dbe_lock(sp->rs_dbe); 7173 rfs4_dbe_lock(fp->rf_dbe); 7174 if (args->share_access != sp->rs_share_access || 7175 args->share_deny != sp->rs_share_deny || 7176 sp->rs_finfo->rf_dinfo.rd_dtype == OPEN_DELEGATE_NONE) { 7177 NFS4_DEBUG(rfs4_debug, 7178 (CE_NOTE, "rfs4_do_opendelprev: state mixup")); 7179 rfs4_dbe_unlock(fp->rf_dbe); 7180 rfs4_dbe_unlock(sp->rs_dbe); 7181 rfs4_file_rele(fp); 7182 rfs4_state_rele(sp); 7183 resp->status = NFS4ERR_SERVERFAULT; 7184 return; 7185 } 7186 rfs4_dbe_unlock(fp->rf_dbe); 7187 rfs4_dbe_unlock(sp->rs_dbe); 7188 7189 dsp = rfs4_finddeleg(sp, &dcreate); 7190 if (dsp == NULL) { 7191 rfs4_state_rele(sp); 7192 rfs4_file_rele(fp); 7193 resp->status = NFS4ERR_SERVERFAULT; 7194 return; 7195 } 7196 7197 next_stateid(&sp->rs_stateid); 7198 7199 resp->stateid = sp->rs_stateid.stateid; 7200 7201 resp->delegation.delegation_type = dsp->rds_dtype; 7202 7203 if (dsp->rds_dtype == OPEN_DELEGATE_READ) { 7204 open_read_delegation4 *rv = 7205 &resp->delegation.open_delegation4_u.read; 7206 7207 rv->stateid = dsp->rds_delegid.stateid; 7208 rv->recall = FALSE; /* no policy in place to set to TRUE */ 7209 ace = &rv->permissions; 7210 } else { 7211 open_write_delegation4 *rv = 7212 &resp->delegation.open_delegation4_u.write; 7213 7214 rv->stateid = dsp->rds_delegid.stateid; 7215 rv->recall = FALSE; /* no policy in place to set to TRUE */ 7216 ace = &rv->permissions; 7217 rv->space_limit.limitby = NFS_LIMIT_SIZE; 7218 rv->space_limit.nfs_space_limit4_u.filesize = UINT64_MAX; 7219 } 7220 7221 /* XXX For now */ 7222 ace->type = ACE4_ACCESS_ALLOWED_ACE_TYPE; 7223 ace->flag = 0; 7224 ace->access_mask = 0; 7225 ace->who.utf8string_len = 0; 7226 ace->who.utf8string_val = 0; 7227 7228 rfs4_deleg_state_rele(dsp); 7229 rfs4_state_rele(sp); 7230 rfs4_file_rele(fp); 7231 } 7232 7233 typedef enum { 7234 NFS4_CHKSEQ_OKAY = 0, 7235 NFS4_CHKSEQ_REPLAY = 1, 7236 NFS4_CHKSEQ_BAD = 2 7237 } rfs4_chkseq_t; 7238 7239 /* 7240 * Generic function for sequence number checks. 7241 */ 7242 static rfs4_chkseq_t 7243 rfs4_check_seqid(seqid4 seqid, nfs_resop4 *lastop, 7244 seqid4 rqst_seq, nfs_resop4 *resop, bool_t copyres) 7245 { 7246 /* Same sequence ids and matching operations? */ 7247 if (seqid == rqst_seq && resop->resop == lastop->resop) { 7248 if (copyres == TRUE) { 7249 rfs4_free_reply(resop); 7250 rfs4_copy_reply(resop, lastop); 7251 } 7252 NFS4_DEBUG(rfs4_debug, (CE_NOTE, 7253 "Replayed SEQID %d\n", seqid)); 7254 return (NFS4_CHKSEQ_REPLAY); 7255 } 7256 7257 /* If the incoming sequence is not the next expected then it is bad */ 7258 if (rqst_seq != seqid + 1) { 7259 if (rqst_seq == seqid) { 7260 NFS4_DEBUG(rfs4_debug, 7261 (CE_NOTE, "BAD SEQID: Replayed sequence id " 7262 "but last op was %d current op is %d\n", 7263 lastop->resop, resop->resop)); 7264 return (NFS4_CHKSEQ_BAD); 7265 } 7266 NFS4_DEBUG(rfs4_debug, 7267 (CE_NOTE, "BAD SEQID: got %u expecting %u\n", 7268 rqst_seq, seqid)); 7269 return (NFS4_CHKSEQ_BAD); 7270 } 7271 7272 /* Everything okay -- next expected */ 7273 return (NFS4_CHKSEQ_OKAY); 7274 } 7275 7276 7277 static rfs4_chkseq_t 7278 rfs4_check_open_seqid(seqid4 seqid, rfs4_openowner_t *op, nfs_resop4 *resop) 7279 { 7280 rfs4_chkseq_t rc; 7281 7282 rfs4_dbe_lock(op->ro_dbe); 7283 rc = rfs4_check_seqid(op->ro_open_seqid, &op->ro_reply, seqid, resop, 7284 TRUE); 7285 rfs4_dbe_unlock(op->ro_dbe); 7286 7287 if (rc == NFS4_CHKSEQ_OKAY) 7288 rfs4_update_lease(op->ro_client); 7289 7290 return (rc); 7291 } 7292 7293 static rfs4_chkseq_t 7294 rfs4_check_olo_seqid(seqid4 olo_seqid, rfs4_openowner_t *op, nfs_resop4 *resop) 7295 { 7296 rfs4_chkseq_t rc; 7297 7298 rfs4_dbe_lock(op->ro_dbe); 7299 rc = rfs4_check_seqid(op->ro_open_seqid, &op->ro_reply, 7300 olo_seqid, resop, FALSE); 7301 rfs4_dbe_unlock(op->ro_dbe); 7302 7303 return (rc); 7304 } 7305 7306 static rfs4_chkseq_t 7307 rfs4_check_lock_seqid(seqid4 seqid, rfs4_lo_state_t *lsp, nfs_resop4 *resop) 7308 { 7309 rfs4_chkseq_t rc = NFS4_CHKSEQ_OKAY; 7310 7311 rfs4_dbe_lock(lsp->rls_dbe); 7312 if (!lsp->rls_skip_seqid_check) 7313 rc = rfs4_check_seqid(lsp->rls_seqid, &lsp->rls_reply, seqid, 7314 resop, TRUE); 7315 rfs4_dbe_unlock(lsp->rls_dbe); 7316 7317 return (rc); 7318 } 7319 7320 static void 7321 rfs4_op_open(nfs_argop4 *argop, nfs_resop4 *resop, 7322 struct svc_req *req, struct compound_state *cs) 7323 { 7324 OPEN4args *args = &argop->nfs_argop4_u.opopen; 7325 OPEN4res *resp = &resop->nfs_resop4_u.opopen; 7326 open_owner4 *owner = &args->owner; 7327 open_claim_type4 claim = args->claim; 7328 rfs4_client_t *cp; 7329 rfs4_openowner_t *oo; 7330 bool_t create; 7331 bool_t replay = FALSE; 7332 int can_reclaim; 7333 7334 DTRACE_NFSV4_2(op__open__start, struct compound_state *, cs, 7335 OPEN4args *, args); 7336 7337 if (cs->vp == NULL) { 7338 *cs->statusp = resp->status = NFS4ERR_NOFILEHANDLE; 7339 goto end; 7340 } 7341 7342 /* 7343 * Need to check clientid and lease expiration first based on 7344 * error ordering and incrementing sequence id. 7345 */ 7346 cp = rfs4_findclient_by_id(owner->clientid, FALSE); 7347 if (cp == NULL) { 7348 *cs->statusp = resp->status = 7349 rfs4_check_clientid(&owner->clientid, 0); 7350 goto end; 7351 } 7352 7353 if (rfs4_lease_expired(cp)) { 7354 rfs4_client_close(cp); 7355 *cs->statusp = resp->status = NFS4ERR_EXPIRED; 7356 goto end; 7357 } 7358 can_reclaim = cp->rc_can_reclaim; 7359 7360 /* 7361 * Find the open_owner for use from this point forward. Take 7362 * care in updating the sequence id based on the type of error 7363 * being returned. 7364 */ 7365 retry: 7366 create = TRUE; 7367 oo = rfs4_findopenowner(owner, &create, args->seqid); 7368 if (oo == NULL) { 7369 *cs->statusp = resp->status = NFS4ERR_STALE_CLIENTID; 7370 rfs4_client_rele(cp); 7371 goto end; 7372 } 7373 7374 /* Hold off access to the sequence space while the open is done */ 7375 rfs4_sw_enter(&oo->ro_sw); 7376 7377 /* 7378 * If the open_owner existed before at the server, then check 7379 * the sequence id. 7380 */ 7381 if (!create && !oo->ro_postpone_confirm) { 7382 switch (rfs4_check_open_seqid(args->seqid, oo, resop)) { 7383 case NFS4_CHKSEQ_BAD: 7384 if ((args->seqid > oo->ro_open_seqid) && 7385 oo->ro_need_confirm) { 7386 rfs4_free_opens(oo, TRUE, FALSE); 7387 rfs4_sw_exit(&oo->ro_sw); 7388 rfs4_openowner_rele(oo); 7389 goto retry; 7390 } 7391 resp->status = NFS4ERR_BAD_SEQID; 7392 goto out; 7393 case NFS4_CHKSEQ_REPLAY: /* replay of previous request */ 7394 replay = TRUE; 7395 goto out; 7396 default: 7397 break; 7398 } 7399 7400 /* 7401 * Sequence was ok and open owner exists 7402 * check to see if we have yet to see an 7403 * open_confirm. 7404 */ 7405 if (oo->ro_need_confirm) { 7406 rfs4_free_opens(oo, TRUE, FALSE); 7407 rfs4_sw_exit(&oo->ro_sw); 7408 rfs4_openowner_rele(oo); 7409 goto retry; 7410 } 7411 } 7412 /* Grace only applies to regular-type OPENs */ 7413 if (rfs4_clnt_in_grace(cp) && 7414 (claim == CLAIM_NULL || claim == CLAIM_DELEGATE_CUR)) { 7415 *cs->statusp = resp->status = NFS4ERR_GRACE; 7416 goto out; 7417 } 7418 7419 /* 7420 * If previous state at the server existed then can_reclaim 7421 * will be set. If not reply NFS4ERR_NO_GRACE to the 7422 * client. 7423 */ 7424 if (rfs4_clnt_in_grace(cp) && claim == CLAIM_PREVIOUS && !can_reclaim) { 7425 *cs->statusp = resp->status = NFS4ERR_NO_GRACE; 7426 goto out; 7427 } 7428 7429 7430 /* 7431 * Reject the open if the client has missed the grace period 7432 */ 7433 if (!rfs4_clnt_in_grace(cp) && claim == CLAIM_PREVIOUS) { 7434 *cs->statusp = resp->status = NFS4ERR_NO_GRACE; 7435 goto out; 7436 } 7437 7438 /* Couple of up-front bookkeeping items */ 7439 if (oo->ro_need_confirm) { 7440 /* 7441 * If this is a reclaim OPEN then we should not ask 7442 * for a confirmation of the open_owner per the 7443 * protocol specification. 7444 */ 7445 if (claim == CLAIM_PREVIOUS) 7446 oo->ro_need_confirm = FALSE; 7447 else 7448 resp->rflags |= OPEN4_RESULT_CONFIRM; 7449 } 7450 resp->rflags |= OPEN4_RESULT_LOCKTYPE_POSIX; 7451 7452 /* 7453 * If there is an unshared filesystem mounted on this vnode, 7454 * do not allow to open/create in this directory. 7455 */ 7456 if (vn_ismntpt(cs->vp)) { 7457 *cs->statusp = resp->status = NFS4ERR_ACCESS; 7458 goto out; 7459 } 7460 7461 /* 7462 * access must READ, WRITE, or BOTH. No access is invalid. 7463 * deny can be READ, WRITE, BOTH, or NONE. 7464 * bits not defined for access/deny are invalid. 7465 */ 7466 if (! (args->share_access & OPEN4_SHARE_ACCESS_BOTH) || 7467 (args->share_access & ~OPEN4_SHARE_ACCESS_BOTH) || 7468 (args->share_deny & ~OPEN4_SHARE_DENY_BOTH)) { 7469 *cs->statusp = resp->status = NFS4ERR_INVAL; 7470 goto out; 7471 } 7472 7473 7474 /* 7475 * make sure attrset is zero before response is built. 7476 */ 7477 resp->attrset = 0; 7478 7479 switch (claim) { 7480 case CLAIM_NULL: 7481 rfs4_do_opennull(cs, req, args, oo, resp); 7482 break; 7483 case CLAIM_PREVIOUS: 7484 rfs4_do_openprev(cs, req, args, oo, resp); 7485 break; 7486 case CLAIM_DELEGATE_CUR: 7487 rfs4_do_opendelcur(cs, req, args, oo, resp); 7488 break; 7489 case CLAIM_DELEGATE_PREV: 7490 rfs4_do_opendelprev(cs, req, args, oo, resp); 7491 break; 7492 default: 7493 resp->status = NFS4ERR_INVAL; 7494 break; 7495 } 7496 7497 out: 7498 rfs4_client_rele(cp); 7499 7500 /* Catch sequence id handling here to make it a little easier */ 7501 switch (resp->status) { 7502 case NFS4ERR_BADXDR: 7503 case NFS4ERR_BAD_SEQID: 7504 case NFS4ERR_BAD_STATEID: 7505 case NFS4ERR_NOFILEHANDLE: 7506 case NFS4ERR_RESOURCE: 7507 case NFS4ERR_STALE_CLIENTID: 7508 case NFS4ERR_STALE_STATEID: 7509 /* 7510 * The protocol states that if any of these errors are 7511 * being returned, the sequence id should not be 7512 * incremented. Any other return requires an 7513 * increment. 7514 */ 7515 break; 7516 default: 7517 /* Always update the lease in this case */ 7518 rfs4_update_lease(oo->ro_client); 7519 7520 /* Regular response - copy the result */ 7521 if (!replay) 7522 rfs4_update_open_resp(oo, resop, &cs->fh); 7523 7524 /* 7525 * REPLAY case: Only if the previous response was OK 7526 * do we copy the filehandle. If not OK, no 7527 * filehandle to copy. 7528 */ 7529 if (replay == TRUE && 7530 resp->status == NFS4_OK && 7531 oo->ro_reply_fh.nfs_fh4_val) { 7532 /* 7533 * If this is a replay, we must restore the 7534 * current filehandle/vp to that of what was 7535 * returned originally. Try our best to do 7536 * it. 7537 */ 7538 nfs_fh4_fmt_t *fh_fmtp = 7539 (nfs_fh4_fmt_t *)oo->ro_reply_fh.nfs_fh4_val; 7540 7541 cs->exi = checkexport4(&fh_fmtp->fh4_fsid, 7542 (fid_t *)&fh_fmtp->fh4_xlen, NULL); 7543 7544 if (cs->exi == NULL) { 7545 resp->status = NFS4ERR_STALE; 7546 goto finish; 7547 } 7548 7549 VN_RELE(cs->vp); 7550 7551 cs->vp = nfs4_fhtovp(&oo->ro_reply_fh, cs->exi, 7552 &resp->status); 7553 7554 if (cs->vp == NULL) 7555 goto finish; 7556 7557 nfs_fh4_copy(&oo->ro_reply_fh, &cs->fh); 7558 } 7559 7560 /* 7561 * If this was a replay, no need to update the 7562 * sequence id. If the open_owner was not created on 7563 * this pass, then update. The first use of an 7564 * open_owner will not bump the sequence id. 7565 */ 7566 if (replay == FALSE && !create) 7567 rfs4_update_open_sequence(oo); 7568 /* 7569 * If the client is receiving an error and the 7570 * open_owner needs to be confirmed, there is no way 7571 * to notify the client of this fact ignoring the fact 7572 * that the server has no method of returning a 7573 * stateid to confirm. Therefore, the server needs to 7574 * mark this open_owner in a way as to avoid the 7575 * sequence id checking the next time the client uses 7576 * this open_owner. 7577 */ 7578 if (resp->status != NFS4_OK && oo->ro_need_confirm) 7579 oo->ro_postpone_confirm = TRUE; 7580 /* 7581 * If OK response then clear the postpone flag and 7582 * reset the sequence id to keep in sync with the 7583 * client. 7584 */ 7585 if (resp->status == NFS4_OK && oo->ro_postpone_confirm) { 7586 oo->ro_postpone_confirm = FALSE; 7587 oo->ro_open_seqid = args->seqid; 7588 } 7589 break; 7590 } 7591 7592 finish: 7593 *cs->statusp = resp->status; 7594 7595 rfs4_sw_exit(&oo->ro_sw); 7596 rfs4_openowner_rele(oo); 7597 7598 end: 7599 DTRACE_NFSV4_2(op__open__done, struct compound_state *, cs, 7600 OPEN4res *, resp); 7601 } 7602 7603 /*ARGSUSED*/ 7604 void 7605 rfs4_op_open_confirm(nfs_argop4 *argop, nfs_resop4 *resop, 7606 struct svc_req *req, struct compound_state *cs) 7607 { 7608 OPEN_CONFIRM4args *args = &argop->nfs_argop4_u.opopen_confirm; 7609 OPEN_CONFIRM4res *resp = &resop->nfs_resop4_u.opopen_confirm; 7610 rfs4_state_t *sp; 7611 nfsstat4 status; 7612 7613 DTRACE_NFSV4_2(op__open__confirm__start, struct compound_state *, cs, 7614 OPEN_CONFIRM4args *, args); 7615 7616 if (cs->vp == NULL) { 7617 *cs->statusp = resp->status = NFS4ERR_NOFILEHANDLE; 7618 goto out; 7619 } 7620 7621 status = rfs4_get_state(&args->open_stateid, &sp, RFS4_DBS_VALID); 7622 if (status != NFS4_OK) { 7623 *cs->statusp = resp->status = status; 7624 goto out; 7625 } 7626 7627 /* Ensure specified filehandle matches */ 7628 if (cs->vp != sp->rs_finfo->rf_vp) { 7629 rfs4_state_rele(sp); 7630 *cs->statusp = resp->status = NFS4ERR_BAD_STATEID; 7631 goto out; 7632 } 7633 7634 /* hold off other access to open_owner while we tinker */ 7635 rfs4_sw_enter(&sp->rs_owner->ro_sw); 7636 7637 switch (rfs4_check_stateid_seqid(sp, &args->open_stateid)) { 7638 case NFS4_CHECK_STATEID_OKAY: 7639 if (rfs4_check_open_seqid(args->seqid, sp->rs_owner, 7640 resop) != 0) { 7641 *cs->statusp = resp->status = NFS4ERR_BAD_SEQID; 7642 break; 7643 } 7644 /* 7645 * If it is the appropriate stateid and determined to 7646 * be "OKAY" then this means that the stateid does not 7647 * need to be confirmed and the client is in error for 7648 * sending an OPEN_CONFIRM. 7649 */ 7650 *cs->statusp = resp->status = NFS4ERR_BAD_STATEID; 7651 break; 7652 case NFS4_CHECK_STATEID_OLD: 7653 *cs->statusp = resp->status = NFS4ERR_OLD_STATEID; 7654 break; 7655 case NFS4_CHECK_STATEID_BAD: 7656 *cs->statusp = resp->status = NFS4ERR_BAD_STATEID; 7657 break; 7658 case NFS4_CHECK_STATEID_EXPIRED: 7659 *cs->statusp = resp->status = NFS4ERR_EXPIRED; 7660 break; 7661 case NFS4_CHECK_STATEID_CLOSED: 7662 *cs->statusp = resp->status = NFS4ERR_OLD_STATEID; 7663 break; 7664 case NFS4_CHECK_STATEID_REPLAY: 7665 switch (rfs4_check_open_seqid(args->seqid, sp->rs_owner, 7666 resop)) { 7667 case NFS4_CHKSEQ_OKAY: 7668 /* 7669 * This is replayed stateid; if seqid matches 7670 * next expected, then client is using wrong seqid. 7671 */ 7672 /* fall through */ 7673 case NFS4_CHKSEQ_BAD: 7674 *cs->statusp = resp->status = NFS4ERR_BAD_SEQID; 7675 break; 7676 case NFS4_CHKSEQ_REPLAY: 7677 /* 7678 * Note this case is the duplicate case so 7679 * resp->status is already set. 7680 */ 7681 *cs->statusp = resp->status; 7682 rfs4_update_lease(sp->rs_owner->ro_client); 7683 break; 7684 } 7685 break; 7686 case NFS4_CHECK_STATEID_UNCONFIRMED: 7687 if (rfs4_check_open_seqid(args->seqid, sp->rs_owner, 7688 resop) != NFS4_CHKSEQ_OKAY) { 7689 *cs->statusp = resp->status = NFS4ERR_BAD_SEQID; 7690 break; 7691 } 7692 *cs->statusp = resp->status = NFS4_OK; 7693 7694 next_stateid(&sp->rs_stateid); 7695 resp->open_stateid = sp->rs_stateid.stateid; 7696 sp->rs_owner->ro_need_confirm = FALSE; 7697 rfs4_update_lease(sp->rs_owner->ro_client); 7698 rfs4_update_open_sequence(sp->rs_owner); 7699 rfs4_update_open_resp(sp->rs_owner, resop, NULL); 7700 break; 7701 default: 7702 ASSERT(FALSE); 7703 *cs->statusp = resp->status = NFS4ERR_SERVERFAULT; 7704 break; 7705 } 7706 rfs4_sw_exit(&sp->rs_owner->ro_sw); 7707 rfs4_state_rele(sp); 7708 7709 out: 7710 DTRACE_NFSV4_2(op__open__confirm__done, struct compound_state *, cs, 7711 OPEN_CONFIRM4res *, resp); 7712 } 7713 7714 /*ARGSUSED*/ 7715 void 7716 rfs4_op_open_downgrade(nfs_argop4 *argop, nfs_resop4 *resop, 7717 struct svc_req *req, struct compound_state *cs) 7718 { 7719 OPEN_DOWNGRADE4args *args = &argop->nfs_argop4_u.opopen_downgrade; 7720 OPEN_DOWNGRADE4res *resp = &resop->nfs_resop4_u.opopen_downgrade; 7721 uint32_t access = args->share_access; 7722 uint32_t deny = args->share_deny; 7723 nfsstat4 status; 7724 rfs4_state_t *sp; 7725 rfs4_file_t *fp; 7726 int fflags = 0; 7727 7728 DTRACE_NFSV4_2(op__open__downgrade__start, struct compound_state *, cs, 7729 OPEN_DOWNGRADE4args *, args); 7730 7731 if (cs->vp == NULL) { 7732 *cs->statusp = resp->status = NFS4ERR_NOFILEHANDLE; 7733 goto out; 7734 } 7735 7736 status = rfs4_get_state(&args->open_stateid, &sp, RFS4_DBS_VALID); 7737 if (status != NFS4_OK) { 7738 *cs->statusp = resp->status = status; 7739 goto out; 7740 } 7741 7742 /* Ensure specified filehandle matches */ 7743 if (cs->vp != sp->rs_finfo->rf_vp) { 7744 rfs4_state_rele(sp); 7745 *cs->statusp = resp->status = NFS4ERR_BAD_STATEID; 7746 goto out; 7747 } 7748 7749 /* hold off other access to open_owner while we tinker */ 7750 rfs4_sw_enter(&sp->rs_owner->ro_sw); 7751 7752 switch (rfs4_check_stateid_seqid(sp, &args->open_stateid)) { 7753 case NFS4_CHECK_STATEID_OKAY: 7754 if (rfs4_check_open_seqid(args->seqid, sp->rs_owner, 7755 resop) != NFS4_CHKSEQ_OKAY) { 7756 *cs->statusp = resp->status = NFS4ERR_BAD_SEQID; 7757 goto end; 7758 } 7759 break; 7760 case NFS4_CHECK_STATEID_OLD: 7761 *cs->statusp = resp->status = NFS4ERR_OLD_STATEID; 7762 goto end; 7763 case NFS4_CHECK_STATEID_BAD: 7764 *cs->statusp = resp->status = NFS4ERR_BAD_STATEID; 7765 goto end; 7766 case NFS4_CHECK_STATEID_EXPIRED: 7767 *cs->statusp = resp->status = NFS4ERR_EXPIRED; 7768 goto end; 7769 case NFS4_CHECK_STATEID_CLOSED: 7770 *cs->statusp = resp->status = NFS4ERR_OLD_STATEID; 7771 goto end; 7772 case NFS4_CHECK_STATEID_UNCONFIRMED: 7773 *cs->statusp = resp->status = NFS4ERR_BAD_STATEID; 7774 goto end; 7775 case NFS4_CHECK_STATEID_REPLAY: 7776 /* Check the sequence id for the open owner */ 7777 switch (rfs4_check_open_seqid(args->seqid, sp->rs_owner, 7778 resop)) { 7779 case NFS4_CHKSEQ_OKAY: 7780 /* 7781 * This is replayed stateid; if seqid matches 7782 * next expected, then client is using wrong seqid. 7783 */ 7784 /* fall through */ 7785 case NFS4_CHKSEQ_BAD: 7786 *cs->statusp = resp->status = NFS4ERR_BAD_SEQID; 7787 goto end; 7788 case NFS4_CHKSEQ_REPLAY: 7789 /* 7790 * Note this case is the duplicate case so 7791 * resp->status is already set. 7792 */ 7793 *cs->statusp = resp->status; 7794 rfs4_update_lease(sp->rs_owner->ro_client); 7795 goto end; 7796 } 7797 break; 7798 default: 7799 ASSERT(FALSE); 7800 break; 7801 } 7802 7803 rfs4_dbe_lock(sp->rs_dbe); 7804 /* 7805 * Check that the new access modes and deny modes are valid. 7806 * Check that no invalid bits are set. 7807 */ 7808 if ((access & ~(OPEN4_SHARE_ACCESS_READ | OPEN4_SHARE_ACCESS_WRITE)) || 7809 (deny & ~(OPEN4_SHARE_DENY_READ | OPEN4_SHARE_DENY_WRITE))) { 7810 *cs->statusp = resp->status = NFS4ERR_INVAL; 7811 rfs4_update_open_sequence(sp->rs_owner); 7812 rfs4_dbe_unlock(sp->rs_dbe); 7813 goto end; 7814 } 7815 7816 /* 7817 * The new modes must be a subset of the current modes and 7818 * the access must specify at least one mode. To test that 7819 * the new mode is a subset of the current modes we bitwise 7820 * AND them together and check that the result equals the new 7821 * mode. For example: 7822 * New mode, access == R and current mode, sp->rs_open_access == RW 7823 * access & sp->rs_open_access == R == access, so the new access mode 7824 * is valid. Consider access == RW, sp->rs_open_access = R 7825 * access & sp->rs_open_access == R != access, so the new access mode 7826 * is invalid. 7827 */ 7828 if ((access & sp->rs_open_access) != access || 7829 (deny & sp->rs_open_deny) != deny || 7830 (access & 7831 (OPEN4_SHARE_ACCESS_READ | OPEN4_SHARE_ACCESS_WRITE)) == 0) { 7832 *cs->statusp = resp->status = NFS4ERR_INVAL; 7833 rfs4_update_open_sequence(sp->rs_owner); 7834 rfs4_dbe_unlock(sp->rs_dbe); 7835 goto end; 7836 } 7837 7838 /* 7839 * Release any share locks associated with this stateID. 7840 * Strictly speaking, this violates the spec because the 7841 * spec effectively requires that open downgrade be atomic. 7842 * At present, fs_shrlock does not have this capability. 7843 */ 7844 (void) rfs4_unshare(sp); 7845 7846 status = rfs4_share(sp, access, deny); 7847 if (status != NFS4_OK) { 7848 *cs->statusp = resp->status = NFS4ERR_SERVERFAULT; 7849 rfs4_update_open_sequence(sp->rs_owner); 7850 rfs4_dbe_unlock(sp->rs_dbe); 7851 goto end; 7852 } 7853 7854 fp = sp->rs_finfo; 7855 rfs4_dbe_lock(fp->rf_dbe); 7856 7857 /* 7858 * If the current mode has deny read and the new mode 7859 * does not, decrement the number of deny read mode bits 7860 * and if it goes to zero turn off the deny read bit 7861 * on the file. 7862 */ 7863 if ((sp->rs_open_deny & OPEN4_SHARE_DENY_READ) && 7864 (deny & OPEN4_SHARE_DENY_READ) == 0) { 7865 fp->rf_deny_read--; 7866 if (fp->rf_deny_read == 0) 7867 fp->rf_share_deny &= ~OPEN4_SHARE_DENY_READ; 7868 } 7869 7870 /* 7871 * If the current mode has deny write and the new mode 7872 * does not, decrement the number of deny write mode bits 7873 * and if it goes to zero turn off the deny write bit 7874 * on the file. 7875 */ 7876 if ((sp->rs_open_deny & OPEN4_SHARE_DENY_WRITE) && 7877 (deny & OPEN4_SHARE_DENY_WRITE) == 0) { 7878 fp->rf_deny_write--; 7879 if (fp->rf_deny_write == 0) 7880 fp->rf_share_deny &= ~OPEN4_SHARE_DENY_WRITE; 7881 } 7882 7883 /* 7884 * If the current mode has access read and the new mode 7885 * does not, decrement the number of access read mode bits 7886 * and if it goes to zero turn off the access read bit 7887 * on the file. set fflags to FREAD for the call to 7888 * vn_open_downgrade(). 7889 */ 7890 if ((sp->rs_open_access & OPEN4_SHARE_ACCESS_READ) && 7891 (access & OPEN4_SHARE_ACCESS_READ) == 0) { 7892 fp->rf_access_read--; 7893 if (fp->rf_access_read == 0) 7894 fp->rf_share_access &= ~OPEN4_SHARE_ACCESS_READ; 7895 fflags |= FREAD; 7896 } 7897 7898 /* 7899 * If the current mode has access write and the new mode 7900 * does not, decrement the number of access write mode bits 7901 * and if it goes to zero turn off the access write bit 7902 * on the file. set fflags to FWRITE for the call to 7903 * vn_open_downgrade(). 7904 */ 7905 if ((sp->rs_open_access & OPEN4_SHARE_ACCESS_WRITE) && 7906 (access & OPEN4_SHARE_ACCESS_WRITE) == 0) { 7907 fp->rf_access_write--; 7908 if (fp->rf_access_write == 0) 7909 fp->rf_share_deny &= ~OPEN4_SHARE_ACCESS_WRITE; 7910 fflags |= FWRITE; 7911 } 7912 7913 /* Check that the file is still accessible */ 7914 ASSERT(fp->rf_share_access); 7915 7916 rfs4_dbe_unlock(fp->rf_dbe); 7917 7918 /* now set the new open access and deny modes */ 7919 sp->rs_open_access = access; 7920 sp->rs_open_deny = deny; 7921 7922 /* 7923 * we successfully downgraded the share lock, now we need to downgrade 7924 * the open. it is possible that the downgrade was only for a deny 7925 * mode and we have nothing else to do. 7926 */ 7927 if ((fflags & (FREAD|FWRITE)) != 0) 7928 vn_open_downgrade(cs->vp, fflags); 7929 7930 /* Update the stateid */ 7931 next_stateid(&sp->rs_stateid); 7932 resp->open_stateid = sp->rs_stateid.stateid; 7933 7934 rfs4_dbe_unlock(sp->rs_dbe); 7935 7936 *cs->statusp = resp->status = NFS4_OK; 7937 /* Update the lease */ 7938 rfs4_update_lease(sp->rs_owner->ro_client); 7939 /* And the sequence */ 7940 rfs4_update_open_sequence(sp->rs_owner); 7941 rfs4_update_open_resp(sp->rs_owner, resop, NULL); 7942 7943 end: 7944 rfs4_sw_exit(&sp->rs_owner->ro_sw); 7945 rfs4_state_rele(sp); 7946 out: 7947 DTRACE_NFSV4_2(op__open__downgrade__done, struct compound_state *, cs, 7948 OPEN_DOWNGRADE4res *, resp); 7949 } 7950 7951 /* 7952 * The logic behind this function is detailed in the NFSv4 RFC in the 7953 * SETCLIENTID operation description under IMPLEMENTATION. Refer to 7954 * that section for explicit guidance to server behavior for 7955 * SETCLIENTID. 7956 */ 7957 void 7958 rfs4_op_setclientid(nfs_argop4 *argop, nfs_resop4 *resop, 7959 struct svc_req *req, struct compound_state *cs) 7960 { 7961 SETCLIENTID4args *args = &argop->nfs_argop4_u.opsetclientid; 7962 SETCLIENTID4res *res = &resop->nfs_resop4_u.opsetclientid; 7963 rfs4_client_t *cp, *newcp, *cp_confirmed, *cp_unconfirmed; 7964 rfs4_clntip_t *ci; 7965 bool_t create; 7966 char *addr, *netid; 7967 int len; 7968 7969 DTRACE_NFSV4_2(op__setclientid__start, struct compound_state *, cs, 7970 SETCLIENTID4args *, args); 7971 retry: 7972 newcp = cp_confirmed = cp_unconfirmed = NULL; 7973 7974 /* 7975 * Save the caller's IP address 7976 */ 7977 args->client.cl_addr = 7978 (struct sockaddr *)svc_getrpccaller(req->rq_xprt)->buf; 7979 7980 /* 7981 * Record if it is a Solaris client that cannot handle referrals. 7982 */ 7983 if (strstr(args->client.id_val, "Solaris") && 7984 !strstr(args->client.id_val, "+referrals")) { 7985 /* Add a "yes, it's downrev" record */ 7986 create = TRUE; 7987 ci = rfs4_find_clntip(args->client.cl_addr, &create); 7988 ASSERT(ci != NULL); 7989 rfs4_dbe_rele(ci->ri_dbe); 7990 } else { 7991 /* Remove any previous record */ 7992 rfs4_invalidate_clntip(args->client.cl_addr); 7993 } 7994 7995 /* 7996 * In search of an EXISTING client matching the incoming 7997 * request to establish a new client identifier at the server 7998 */ 7999 create = TRUE; 8000 cp = rfs4_findclient(&args->client, &create, NULL); 8001 8002 /* Should never happen */ 8003 ASSERT(cp != NULL); 8004 8005 if (cp == NULL) { 8006 *cs->statusp = res->status = NFS4ERR_SERVERFAULT; 8007 goto out; 8008 } 8009 8010 /* 8011 * Easiest case. Client identifier is newly created and is 8012 * unconfirmed. Also note that for this case, no other 8013 * entries exist for the client identifier. Nothing else to 8014 * check. Just setup the response and respond. 8015 */ 8016 if (create) { 8017 *cs->statusp = res->status = NFS4_OK; 8018 res->SETCLIENTID4res_u.resok4.clientid = cp->rc_clientid; 8019 res->SETCLIENTID4res_u.resok4.setclientid_confirm = 8020 cp->rc_confirm_verf; 8021 /* Setup callback information; CB_NULL confirmation later */ 8022 rfs4_client_setcb(cp, &args->callback, args->callback_ident); 8023 8024 rfs4_client_rele(cp); 8025 goto out; 8026 } 8027 8028 /* 8029 * An existing, confirmed client may exist but it may not have 8030 * been active for at least one lease period. If so, then 8031 * "close" the client and create a new client identifier 8032 */ 8033 if (rfs4_lease_expired(cp)) { 8034 rfs4_client_close(cp); 8035 goto retry; 8036 } 8037 8038 if (cp->rc_need_confirm == TRUE) 8039 cp_unconfirmed = cp; 8040 else 8041 cp_confirmed = cp; 8042 8043 cp = NULL; 8044 8045 /* 8046 * We have a confirmed client, now check for an 8047 * unconfimred entry 8048 */ 8049 if (cp_confirmed) { 8050 /* If creds don't match then client identifier is inuse */ 8051 if (!creds_ok(cp_confirmed->rc_cr_set, req, cs)) { 8052 rfs4_cbinfo_t *cbp; 8053 /* 8054 * Some one else has established this client 8055 * id. Try and say * who they are. We will use 8056 * the call back address supplied by * the 8057 * first client. 8058 */ 8059 *cs->statusp = res->status = NFS4ERR_CLID_INUSE; 8060 8061 addr = netid = NULL; 8062 8063 cbp = &cp_confirmed->rc_cbinfo; 8064 if (cbp->cb_callback.cb_location.r_addr && 8065 cbp->cb_callback.cb_location.r_netid) { 8066 cb_client4 *cbcp = &cbp->cb_callback; 8067 8068 len = strlen(cbcp->cb_location.r_addr)+1; 8069 addr = kmem_alloc(len, KM_SLEEP); 8070 bcopy(cbcp->cb_location.r_addr, addr, len); 8071 len = strlen(cbcp->cb_location.r_netid)+1; 8072 netid = kmem_alloc(len, KM_SLEEP); 8073 bcopy(cbcp->cb_location.r_netid, netid, len); 8074 } 8075 8076 res->SETCLIENTID4res_u.client_using.r_addr = addr; 8077 res->SETCLIENTID4res_u.client_using.r_netid = netid; 8078 8079 rfs4_client_rele(cp_confirmed); 8080 } 8081 8082 /* 8083 * Confirmed, creds match, and verifier matches; must 8084 * be an update of the callback info 8085 */ 8086 if (cp_confirmed->rc_nfs_client.verifier == 8087 args->client.verifier) { 8088 /* Setup callback information */ 8089 rfs4_client_setcb(cp_confirmed, &args->callback, 8090 args->callback_ident); 8091 8092 /* everything okay -- move ahead */ 8093 *cs->statusp = res->status = NFS4_OK; 8094 res->SETCLIENTID4res_u.resok4.clientid = 8095 cp_confirmed->rc_clientid; 8096 8097 /* update the confirm_verifier and return it */ 8098 rfs4_client_scv_next(cp_confirmed); 8099 res->SETCLIENTID4res_u.resok4.setclientid_confirm = 8100 cp_confirmed->rc_confirm_verf; 8101 8102 rfs4_client_rele(cp_confirmed); 8103 goto out; 8104 } 8105 8106 /* 8107 * Creds match but the verifier doesn't. Must search 8108 * for an unconfirmed client that would be replaced by 8109 * this request. 8110 */ 8111 create = FALSE; 8112 cp_unconfirmed = rfs4_findclient(&args->client, &create, 8113 cp_confirmed); 8114 } 8115 8116 /* 8117 * At this point, we have taken care of the brand new client 8118 * struct, INUSE case, update of an existing, and confirmed 8119 * client struct. 8120 */ 8121 8122 /* 8123 * check to see if things have changed while we originally 8124 * picked up the client struct. If they have, then return and 8125 * retry the processing of this SETCLIENTID request. 8126 */ 8127 if (cp_unconfirmed) { 8128 rfs4_dbe_lock(cp_unconfirmed->rc_dbe); 8129 if (!cp_unconfirmed->rc_need_confirm) { 8130 rfs4_dbe_unlock(cp_unconfirmed->rc_dbe); 8131 rfs4_client_rele(cp_unconfirmed); 8132 if (cp_confirmed) 8133 rfs4_client_rele(cp_confirmed); 8134 goto retry; 8135 } 8136 /* do away with the old unconfirmed one */ 8137 rfs4_dbe_invalidate(cp_unconfirmed->rc_dbe); 8138 rfs4_dbe_unlock(cp_unconfirmed->rc_dbe); 8139 rfs4_client_rele(cp_unconfirmed); 8140 cp_unconfirmed = NULL; 8141 } 8142 8143 /* 8144 * This search will temporarily hide the confirmed client 8145 * struct while a new client struct is created as the 8146 * unconfirmed one. 8147 */ 8148 create = TRUE; 8149 newcp = rfs4_findclient(&args->client, &create, cp_confirmed); 8150 8151 ASSERT(newcp != NULL); 8152 8153 if (newcp == NULL) { 8154 *cs->statusp = res->status = NFS4ERR_SERVERFAULT; 8155 rfs4_client_rele(cp_confirmed); 8156 goto out; 8157 } 8158 8159 /* 8160 * If one was not created, then a similar request must be in 8161 * process so release and start over with this one 8162 */ 8163 if (create != TRUE) { 8164 rfs4_client_rele(newcp); 8165 if (cp_confirmed) 8166 rfs4_client_rele(cp_confirmed); 8167 goto retry; 8168 } 8169 8170 *cs->statusp = res->status = NFS4_OK; 8171 res->SETCLIENTID4res_u.resok4.clientid = newcp->rc_clientid; 8172 res->SETCLIENTID4res_u.resok4.setclientid_confirm = 8173 newcp->rc_confirm_verf; 8174 /* Setup callback information; CB_NULL confirmation later */ 8175 rfs4_client_setcb(newcp, &args->callback, args->callback_ident); 8176 8177 newcp->rc_cp_confirmed = cp_confirmed; 8178 8179 rfs4_client_rele(newcp); 8180 8181 out: 8182 DTRACE_NFSV4_2(op__setclientid__done, struct compound_state *, cs, 8183 SETCLIENTID4res *, res); 8184 } 8185 8186 /*ARGSUSED*/ 8187 void 8188 rfs4_op_setclientid_confirm(nfs_argop4 *argop, nfs_resop4 *resop, 8189 struct svc_req *req, struct compound_state *cs) 8190 { 8191 SETCLIENTID_CONFIRM4args *args = 8192 &argop->nfs_argop4_u.opsetclientid_confirm; 8193 SETCLIENTID_CONFIRM4res *res = 8194 &resop->nfs_resop4_u.opsetclientid_confirm; 8195 rfs4_client_t *cp, *cptoclose = NULL; 8196 8197 DTRACE_NFSV4_2(op__setclientid__confirm__start, 8198 struct compound_state *, cs, 8199 SETCLIENTID_CONFIRM4args *, args); 8200 8201 *cs->statusp = res->status = NFS4_OK; 8202 8203 cp = rfs4_findclient_by_id(args->clientid, TRUE); 8204 8205 if (cp == NULL) { 8206 *cs->statusp = res->status = 8207 rfs4_check_clientid(&args->clientid, 1); 8208 goto out; 8209 } 8210 8211 if (!creds_ok(cp, req, cs)) { 8212 *cs->statusp = res->status = NFS4ERR_CLID_INUSE; 8213 rfs4_client_rele(cp); 8214 goto out; 8215 } 8216 8217 /* If the verifier doesn't match, the record doesn't match */ 8218 if (cp->rc_confirm_verf != args->setclientid_confirm) { 8219 *cs->statusp = res->status = NFS4ERR_STALE_CLIENTID; 8220 rfs4_client_rele(cp); 8221 goto out; 8222 } 8223 8224 rfs4_dbe_lock(cp->rc_dbe); 8225 cp->rc_need_confirm = FALSE; 8226 if (cp->rc_cp_confirmed) { 8227 cptoclose = cp->rc_cp_confirmed; 8228 cptoclose->rc_ss_remove = 1; 8229 cp->rc_cp_confirmed = NULL; 8230 } 8231 8232 /* 8233 * Update the client's associated server instance, if it's changed 8234 * since the client was created. 8235 */ 8236 if (rfs4_servinst(cp) != rfs4_cur_servinst) 8237 rfs4_servinst_assign(cp, rfs4_cur_servinst); 8238 8239 /* 8240 * Record clientid in stable storage. 8241 * Must be done after server instance has been assigned. 8242 */ 8243 rfs4_ss_clid(cp); 8244 8245 rfs4_dbe_unlock(cp->rc_dbe); 8246 8247 if (cptoclose) 8248 /* don't need to rele, client_close does it */ 8249 rfs4_client_close(cptoclose); 8250 8251 /* If needed, initiate CB_NULL call for callback path */ 8252 rfs4_deleg_cb_check(cp); 8253 rfs4_update_lease(cp); 8254 8255 /* 8256 * Check to see if client can perform reclaims 8257 */ 8258 rfs4_ss_chkclid(cp); 8259 8260 rfs4_client_rele(cp); 8261 8262 out: 8263 DTRACE_NFSV4_2(op__setclientid__confirm__done, 8264 struct compound_state *, cs, 8265 SETCLIENTID_CONFIRM4 *, res); 8266 } 8267 8268 8269 /*ARGSUSED*/ 8270 void 8271 rfs4_op_close(nfs_argop4 *argop, nfs_resop4 *resop, 8272 struct svc_req *req, struct compound_state *cs) 8273 { 8274 CLOSE4args *args = &argop->nfs_argop4_u.opclose; 8275 CLOSE4res *resp = &resop->nfs_resop4_u.opclose; 8276 rfs4_state_t *sp; 8277 nfsstat4 status; 8278 8279 DTRACE_NFSV4_2(op__close__start, struct compound_state *, cs, 8280 CLOSE4args *, args); 8281 8282 if (cs->vp == NULL) { 8283 *cs->statusp = resp->status = NFS4ERR_NOFILEHANDLE; 8284 goto out; 8285 } 8286 8287 status = rfs4_get_state(&args->open_stateid, &sp, RFS4_DBS_INVALID); 8288 if (status != NFS4_OK) { 8289 *cs->statusp = resp->status = status; 8290 goto out; 8291 } 8292 8293 /* Ensure specified filehandle matches */ 8294 if (cs->vp != sp->rs_finfo->rf_vp) { 8295 rfs4_state_rele(sp); 8296 *cs->statusp = resp->status = NFS4ERR_BAD_STATEID; 8297 goto out; 8298 } 8299 8300 /* hold off other access to open_owner while we tinker */ 8301 rfs4_sw_enter(&sp->rs_owner->ro_sw); 8302 8303 switch (rfs4_check_stateid_seqid(sp, &args->open_stateid)) { 8304 case NFS4_CHECK_STATEID_OKAY: 8305 if (rfs4_check_open_seqid(args->seqid, sp->rs_owner, 8306 resop) != NFS4_CHKSEQ_OKAY) { 8307 *cs->statusp = resp->status = NFS4ERR_BAD_SEQID; 8308 goto end; 8309 } 8310 break; 8311 case NFS4_CHECK_STATEID_OLD: 8312 *cs->statusp = resp->status = NFS4ERR_OLD_STATEID; 8313 goto end; 8314 case NFS4_CHECK_STATEID_BAD: 8315 *cs->statusp = resp->status = NFS4ERR_BAD_STATEID; 8316 goto end; 8317 case NFS4_CHECK_STATEID_EXPIRED: 8318 *cs->statusp = resp->status = NFS4ERR_EXPIRED; 8319 goto end; 8320 case NFS4_CHECK_STATEID_CLOSED: 8321 *cs->statusp = resp->status = NFS4ERR_OLD_STATEID; 8322 goto end; 8323 case NFS4_CHECK_STATEID_UNCONFIRMED: 8324 *cs->statusp = resp->status = NFS4ERR_BAD_STATEID; 8325 goto end; 8326 case NFS4_CHECK_STATEID_REPLAY: 8327 /* Check the sequence id for the open owner */ 8328 switch (rfs4_check_open_seqid(args->seqid, sp->rs_owner, 8329 resop)) { 8330 case NFS4_CHKSEQ_OKAY: 8331 /* 8332 * This is replayed stateid; if seqid matches 8333 * next expected, then client is using wrong seqid. 8334 */ 8335 /* FALL THROUGH */ 8336 case NFS4_CHKSEQ_BAD: 8337 *cs->statusp = resp->status = NFS4ERR_BAD_SEQID; 8338 goto end; 8339 case NFS4_CHKSEQ_REPLAY: 8340 /* 8341 * Note this case is the duplicate case so 8342 * resp->status is already set. 8343 */ 8344 *cs->statusp = resp->status; 8345 rfs4_update_lease(sp->rs_owner->ro_client); 8346 goto end; 8347 } 8348 break; 8349 default: 8350 ASSERT(FALSE); 8351 break; 8352 } 8353 8354 rfs4_dbe_lock(sp->rs_dbe); 8355 8356 /* Update the stateid. */ 8357 next_stateid(&sp->rs_stateid); 8358 resp->open_stateid = sp->rs_stateid.stateid; 8359 8360 rfs4_dbe_unlock(sp->rs_dbe); 8361 8362 rfs4_update_lease(sp->rs_owner->ro_client); 8363 rfs4_update_open_sequence(sp->rs_owner); 8364 rfs4_update_open_resp(sp->rs_owner, resop, NULL); 8365 8366 rfs4_state_close(sp, FALSE, FALSE, cs->cr); 8367 8368 *cs->statusp = resp->status = status; 8369 8370 end: 8371 rfs4_sw_exit(&sp->rs_owner->ro_sw); 8372 rfs4_state_rele(sp); 8373 out: 8374 DTRACE_NFSV4_2(op__close__done, struct compound_state *, cs, 8375 CLOSE4res *, resp); 8376 } 8377 8378 /* 8379 * Manage the counts on the file struct and close all file locks 8380 */ 8381 /*ARGSUSED*/ 8382 void 8383 rfs4_release_share_lock_state(rfs4_state_t *sp, cred_t *cr, 8384 bool_t close_of_client) 8385 { 8386 rfs4_file_t *fp = sp->rs_finfo; 8387 rfs4_lo_state_t *lsp; 8388 int fflags = 0; 8389 8390 /* 8391 * If this call is part of the larger closing down of client 8392 * state then it is just easier to release all locks 8393 * associated with this client instead of going through each 8394 * individual file and cleaning locks there. 8395 */ 8396 if (close_of_client) { 8397 if (sp->rs_owner->ro_client->rc_unlksys_completed == FALSE && 8398 !list_is_empty(&sp->rs_lostatelist) && 8399 sp->rs_owner->ro_client->rc_sysidt != LM_NOSYSID) { 8400 /* Is the PxFS kernel module loaded? */ 8401 if (lm_remove_file_locks != NULL) { 8402 int new_sysid; 8403 8404 /* Encode the cluster nodeid in new sysid */ 8405 new_sysid = sp->rs_owner->ro_client->rc_sysidt; 8406 lm_set_nlmid_flk(&new_sysid); 8407 8408 /* 8409 * This PxFS routine removes file locks for a 8410 * client over all nodes of a cluster. 8411 */ 8412 NFS4_DEBUG(rfs4_debug, (CE_NOTE, 8413 "lm_remove_file_locks(sysid=0x%x)\n", 8414 new_sysid)); 8415 (*lm_remove_file_locks)(new_sysid); 8416 } else { 8417 struct flock64 flk; 8418 8419 /* Release all locks for this client */ 8420 flk.l_type = F_UNLKSYS; 8421 flk.l_whence = 0; 8422 flk.l_start = 0; 8423 flk.l_len = 0; 8424 flk.l_sysid = 8425 sp->rs_owner->ro_client->rc_sysidt; 8426 flk.l_pid = 0; 8427 (void) VOP_FRLOCK(sp->rs_finfo->rf_vp, F_SETLK, 8428 &flk, F_REMOTELOCK | FREAD | FWRITE, 8429 (u_offset_t)0, NULL, CRED(), NULL); 8430 } 8431 8432 sp->rs_owner->ro_client->rc_unlksys_completed = TRUE; 8433 } 8434 } 8435 8436 /* 8437 * Release all locks on this file by this lock owner or at 8438 * least mark the locks as having been released 8439 */ 8440 for (lsp = list_head(&sp->rs_lostatelist); lsp != NULL; 8441 lsp = list_next(&sp->rs_lostatelist, lsp)) { 8442 lsp->rls_locks_cleaned = TRUE; 8443 8444 /* Was this already taken care of above? */ 8445 if (!close_of_client && 8446 sp->rs_owner->ro_client->rc_sysidt != LM_NOSYSID) 8447 (void) cleanlocks(sp->rs_finfo->rf_vp, 8448 lsp->rls_locker->rl_pid, 8449 lsp->rls_locker->rl_client->rc_sysidt); 8450 } 8451 8452 /* 8453 * Release any shrlocks associated with this open state ID. 8454 * This must be done before the rfs4_state gets marked closed. 8455 */ 8456 if (sp->rs_owner->ro_client->rc_sysidt != LM_NOSYSID) 8457 (void) rfs4_unshare(sp); 8458 8459 if (sp->rs_open_access) { 8460 rfs4_dbe_lock(fp->rf_dbe); 8461 8462 /* 8463 * Decrement the count for each access and deny bit that this 8464 * state has contributed to the file. 8465 * If the file counts go to zero 8466 * clear the appropriate bit in the appropriate mask. 8467 */ 8468 if (sp->rs_open_access & OPEN4_SHARE_ACCESS_READ) { 8469 fp->rf_access_read--; 8470 fflags |= FREAD; 8471 if (fp->rf_access_read == 0) 8472 fp->rf_share_access &= ~OPEN4_SHARE_ACCESS_READ; 8473 } 8474 if (sp->rs_open_access & OPEN4_SHARE_ACCESS_WRITE) { 8475 fp->rf_access_write--; 8476 fflags |= FWRITE; 8477 if (fp->rf_access_write == 0) 8478 fp->rf_share_access &= 8479 ~OPEN4_SHARE_ACCESS_WRITE; 8480 } 8481 if (sp->rs_open_deny & OPEN4_SHARE_DENY_READ) { 8482 fp->rf_deny_read--; 8483 if (fp->rf_deny_read == 0) 8484 fp->rf_share_deny &= ~OPEN4_SHARE_DENY_READ; 8485 } 8486 if (sp->rs_open_deny & OPEN4_SHARE_DENY_WRITE) { 8487 fp->rf_deny_write--; 8488 if (fp->rf_deny_write == 0) 8489 fp->rf_share_deny &= ~OPEN4_SHARE_DENY_WRITE; 8490 } 8491 8492 (void) VOP_CLOSE(fp->rf_vp, fflags, 1, (offset_t)0, cr, NULL); 8493 8494 rfs4_dbe_unlock(fp->rf_dbe); 8495 8496 sp->rs_open_access = 0; 8497 sp->rs_open_deny = 0; 8498 } 8499 } 8500 8501 /* 8502 * lock_denied: Fill in a LOCK4deneid structure given an flock64 structure. 8503 */ 8504 static nfsstat4 8505 lock_denied(LOCK4denied *dp, struct flock64 *flk) 8506 { 8507 rfs4_lockowner_t *lo; 8508 rfs4_client_t *cp; 8509 uint32_t len; 8510 8511 lo = rfs4_findlockowner_by_pid(flk->l_pid); 8512 if (lo != NULL) { 8513 cp = lo->rl_client; 8514 if (rfs4_lease_expired(cp)) { 8515 rfs4_lockowner_rele(lo); 8516 rfs4_dbe_hold(cp->rc_dbe); 8517 rfs4_client_close(cp); 8518 return (NFS4ERR_EXPIRED); 8519 } 8520 dp->owner.clientid = lo->rl_owner.clientid; 8521 len = lo->rl_owner.owner_len; 8522 dp->owner.owner_val = kmem_alloc(len, KM_SLEEP); 8523 bcopy(lo->rl_owner.owner_val, dp->owner.owner_val, len); 8524 dp->owner.owner_len = len; 8525 rfs4_lockowner_rele(lo); 8526 goto finish; 8527 } 8528 8529 /* 8530 * Its not a NFS4 lock. We take advantage that the upper 32 bits 8531 * of the client id contain the boot time for a NFS4 lock. So we 8532 * fabricate and identity by setting clientid to the sysid, and 8533 * the lock owner to the pid. 8534 */ 8535 dp->owner.clientid = flk->l_sysid; 8536 len = sizeof (pid_t); 8537 dp->owner.owner_len = len; 8538 dp->owner.owner_val = kmem_alloc(len, KM_SLEEP); 8539 bcopy(&flk->l_pid, dp->owner.owner_val, len); 8540 finish: 8541 dp->offset = flk->l_start; 8542 dp->length = flk->l_len; 8543 8544 if (flk->l_type == F_RDLCK) 8545 dp->locktype = READ_LT; 8546 else if (flk->l_type == F_WRLCK) 8547 dp->locktype = WRITE_LT; 8548 else 8549 return (NFS4ERR_INVAL); /* no mapping from POSIX ltype to v4 */ 8550 8551 return (NFS4_OK); 8552 } 8553 8554 static int 8555 setlock(vnode_t *vp, struct flock64 *flock, int flag, cred_t *cred) 8556 { 8557 int error; 8558 struct flock64 flk; 8559 int i; 8560 clock_t delaytime; 8561 int cmd; 8562 8563 cmd = nbl_need_check(vp) ? F_SETLK_NBMAND : F_SETLK; 8564 retry: 8565 delaytime = MSEC_TO_TICK_ROUNDUP(rfs4_lock_delay); 8566 8567 for (i = 0; i < rfs4_maxlock_tries; i++) { 8568 LOCK_PRINT(rfs4_debug, "setlock", cmd, flock); 8569 error = VOP_FRLOCK(vp, cmd, 8570 flock, flag, (u_offset_t)0, NULL, cred, NULL); 8571 8572 if (error != EAGAIN && error != EACCES) 8573 break; 8574 8575 if (i < rfs4_maxlock_tries - 1) { 8576 delay(delaytime); 8577 delaytime *= 2; 8578 } 8579 } 8580 8581 if (error == EAGAIN || error == EACCES) { 8582 /* Get the owner of the lock */ 8583 flk = *flock; 8584 LOCK_PRINT(rfs4_debug, "setlock", F_GETLK, &flk); 8585 if (VOP_FRLOCK(vp, F_GETLK, &flk, flag, 8586 (u_offset_t)0, NULL, cred, NULL) == 0) { 8587 if (flk.l_type == F_UNLCK) { 8588 /* No longer locked, retry */ 8589 goto retry; 8590 } 8591 *flock = flk; 8592 LOCK_PRINT(rfs4_debug, "setlock(blocking lock)", 8593 F_GETLK, &flk); 8594 } 8595 } 8596 8597 return (error); 8598 } 8599 8600 /*ARGSUSED*/ 8601 static nfsstat4 8602 rfs4_do_lock(rfs4_lo_state_t *lsp, nfs_lock_type4 locktype, 8603 offset4 offset, length4 length, cred_t *cred, nfs_resop4 *resop) 8604 { 8605 nfsstat4 status; 8606 rfs4_lockowner_t *lo = lsp->rls_locker; 8607 rfs4_state_t *sp = lsp->rls_state; 8608 struct flock64 flock; 8609 int16_t ltype; 8610 int flag; 8611 int error; 8612 sysid_t sysid; 8613 LOCK4res *lres; 8614 8615 if (rfs4_lease_expired(lo->rl_client)) { 8616 return (NFS4ERR_EXPIRED); 8617 } 8618 8619 if ((status = rfs4_client_sysid(lo->rl_client, &sysid)) != NFS4_OK) 8620 return (status); 8621 8622 /* Check for zero length. To lock to end of file use all ones for V4 */ 8623 if (length == 0) 8624 return (NFS4ERR_INVAL); 8625 else if (length == (length4)(~0)) 8626 length = 0; /* Posix to end of file */ 8627 8628 retry: 8629 rfs4_dbe_lock(sp->rs_dbe); 8630 if (sp->rs_closed) { 8631 rfs4_dbe_unlock(sp->rs_dbe); 8632 return (NFS4ERR_OLD_STATEID); 8633 } 8634 8635 if (resop->resop != OP_LOCKU) { 8636 switch (locktype) { 8637 case READ_LT: 8638 case READW_LT: 8639 if ((sp->rs_share_access 8640 & OPEN4_SHARE_ACCESS_READ) == 0) { 8641 rfs4_dbe_unlock(sp->rs_dbe); 8642 8643 return (NFS4ERR_OPENMODE); 8644 } 8645 ltype = F_RDLCK; 8646 break; 8647 case WRITE_LT: 8648 case WRITEW_LT: 8649 if ((sp->rs_share_access 8650 & OPEN4_SHARE_ACCESS_WRITE) == 0) { 8651 rfs4_dbe_unlock(sp->rs_dbe); 8652 8653 return (NFS4ERR_OPENMODE); 8654 } 8655 ltype = F_WRLCK; 8656 break; 8657 } 8658 } else 8659 ltype = F_UNLCK; 8660 8661 flock.l_type = ltype; 8662 flock.l_whence = 0; /* SEEK_SET */ 8663 flock.l_start = offset; 8664 flock.l_len = length; 8665 flock.l_sysid = sysid; 8666 flock.l_pid = lsp->rls_locker->rl_pid; 8667 8668 /* Note that length4 is uint64_t but l_len and l_start are off64_t */ 8669 if (flock.l_len < 0 || flock.l_start < 0) { 8670 rfs4_dbe_unlock(sp->rs_dbe); 8671 return (NFS4ERR_INVAL); 8672 } 8673 8674 /* 8675 * N.B. FREAD has the same value as OPEN4_SHARE_ACCESS_READ and 8676 * FWRITE has the same value as OPEN4_SHARE_ACCESS_WRITE. 8677 */ 8678 flag = (int)sp->rs_share_access | F_REMOTELOCK; 8679 8680 error = setlock(sp->rs_finfo->rf_vp, &flock, flag, cred); 8681 if (error == 0) { 8682 rfs4_dbe_lock(lsp->rls_dbe); 8683 next_stateid(&lsp->rls_lockid); 8684 rfs4_dbe_unlock(lsp->rls_dbe); 8685 } 8686 8687 rfs4_dbe_unlock(sp->rs_dbe); 8688 8689 /* 8690 * N.B. We map error values to nfsv4 errors. This is differrent 8691 * than puterrno4 routine. 8692 */ 8693 switch (error) { 8694 case 0: 8695 status = NFS4_OK; 8696 break; 8697 case EAGAIN: 8698 case EACCES: /* Old value */ 8699 /* Can only get here if op is OP_LOCK */ 8700 ASSERT(resop->resop == OP_LOCK); 8701 lres = &resop->nfs_resop4_u.oplock; 8702 status = NFS4ERR_DENIED; 8703 if (lock_denied(&lres->LOCK4res_u.denied, &flock) 8704 == NFS4ERR_EXPIRED) 8705 goto retry; 8706 break; 8707 case ENOLCK: 8708 status = NFS4ERR_DELAY; 8709 break; 8710 case EOVERFLOW: 8711 status = NFS4ERR_INVAL; 8712 break; 8713 case EINVAL: 8714 status = NFS4ERR_NOTSUPP; 8715 break; 8716 default: 8717 status = NFS4ERR_SERVERFAULT; 8718 break; 8719 } 8720 8721 return (status); 8722 } 8723 8724 /*ARGSUSED*/ 8725 void 8726 rfs4_op_lock(nfs_argop4 *argop, nfs_resop4 *resop, 8727 struct svc_req *req, struct compound_state *cs) 8728 { 8729 LOCK4args *args = &argop->nfs_argop4_u.oplock; 8730 LOCK4res *resp = &resop->nfs_resop4_u.oplock; 8731 nfsstat4 status; 8732 stateid4 *stateid; 8733 rfs4_lockowner_t *lo; 8734 rfs4_client_t *cp; 8735 rfs4_state_t *sp = NULL; 8736 rfs4_lo_state_t *lsp = NULL; 8737 bool_t ls_sw_held = FALSE; 8738 bool_t create = TRUE; 8739 bool_t lcreate = TRUE; 8740 bool_t dup_lock = FALSE; 8741 int rc; 8742 8743 DTRACE_NFSV4_2(op__lock__start, struct compound_state *, cs, 8744 LOCK4args *, args); 8745 8746 if (cs->vp == NULL) { 8747 *cs->statusp = resp->status = NFS4ERR_NOFILEHANDLE; 8748 DTRACE_NFSV4_2(op__lock__done, struct compound_state *, 8749 cs, LOCK4res *, resp); 8750 return; 8751 } 8752 8753 if (args->locker.new_lock_owner) { 8754 /* Create a new lockowner for this instance */ 8755 open_to_lock_owner4 *olo = &args->locker.locker4_u.open_owner; 8756 8757 NFS4_DEBUG(rfs4_debug, (CE_NOTE, "Creating new lock owner")); 8758 8759 stateid = &olo->open_stateid; 8760 status = rfs4_get_state(stateid, &sp, RFS4_DBS_VALID); 8761 if (status != NFS4_OK) { 8762 NFS4_DEBUG(rfs4_debug, 8763 (CE_NOTE, "Get state failed in lock %d", status)); 8764 *cs->statusp = resp->status = status; 8765 DTRACE_NFSV4_2(op__lock__done, struct compound_state *, 8766 cs, LOCK4res *, resp); 8767 return; 8768 } 8769 8770 /* Ensure specified filehandle matches */ 8771 if (cs->vp != sp->rs_finfo->rf_vp) { 8772 rfs4_state_rele(sp); 8773 *cs->statusp = resp->status = NFS4ERR_BAD_STATEID; 8774 DTRACE_NFSV4_2(op__lock__done, struct compound_state *, 8775 cs, LOCK4res *, resp); 8776 return; 8777 } 8778 8779 /* hold off other access to open_owner while we tinker */ 8780 rfs4_sw_enter(&sp->rs_owner->ro_sw); 8781 8782 switch (rc = rfs4_check_stateid_seqid(sp, stateid)) { 8783 case NFS4_CHECK_STATEID_OLD: 8784 *cs->statusp = resp->status = NFS4ERR_OLD_STATEID; 8785 goto end; 8786 case NFS4_CHECK_STATEID_BAD: 8787 *cs->statusp = resp->status = NFS4ERR_BAD_STATEID; 8788 goto end; 8789 case NFS4_CHECK_STATEID_EXPIRED: 8790 *cs->statusp = resp->status = NFS4ERR_EXPIRED; 8791 goto end; 8792 case NFS4_CHECK_STATEID_UNCONFIRMED: 8793 *cs->statusp = resp->status = NFS4ERR_BAD_STATEID; 8794 goto end; 8795 case NFS4_CHECK_STATEID_CLOSED: 8796 *cs->statusp = resp->status = NFS4ERR_OLD_STATEID; 8797 goto end; 8798 case NFS4_CHECK_STATEID_OKAY: 8799 case NFS4_CHECK_STATEID_REPLAY: 8800 switch (rfs4_check_olo_seqid(olo->open_seqid, 8801 sp->rs_owner, resop)) { 8802 case NFS4_CHKSEQ_OKAY: 8803 if (rc == NFS4_CHECK_STATEID_OKAY) 8804 break; 8805 /* 8806 * This is replayed stateid; if seqid 8807 * matches next expected, then client 8808 * is using wrong seqid. 8809 */ 8810 /* FALLTHROUGH */ 8811 case NFS4_CHKSEQ_BAD: 8812 *cs->statusp = resp->status = NFS4ERR_BAD_SEQID; 8813 goto end; 8814 case NFS4_CHKSEQ_REPLAY: 8815 /* This is a duplicate LOCK request */ 8816 dup_lock = TRUE; 8817 8818 /* 8819 * For a duplicate we do not want to 8820 * create a new lockowner as it should 8821 * already exist. 8822 * Turn off the lockowner create flag. 8823 */ 8824 lcreate = FALSE; 8825 } 8826 break; 8827 } 8828 8829 lo = rfs4_findlockowner(&olo->lock_owner, &lcreate); 8830 if (lo == NULL) { 8831 NFS4_DEBUG(rfs4_debug, 8832 (CE_NOTE, "rfs4_op_lock: no lock owner")); 8833 *cs->statusp = resp->status = NFS4ERR_RESOURCE; 8834 goto end; 8835 } 8836 8837 lsp = rfs4_findlo_state_by_owner(lo, sp, &create); 8838 if (lsp == NULL) { 8839 rfs4_update_lease(sp->rs_owner->ro_client); 8840 /* 8841 * Only update theh open_seqid if this is not 8842 * a duplicate request 8843 */ 8844 if (dup_lock == FALSE) { 8845 rfs4_update_open_sequence(sp->rs_owner); 8846 } 8847 8848 NFS4_DEBUG(rfs4_debug, 8849 (CE_NOTE, "rfs4_op_lock: no state")); 8850 *cs->statusp = resp->status = NFS4ERR_SERVERFAULT; 8851 rfs4_update_open_resp(sp->rs_owner, resop, NULL); 8852 rfs4_lockowner_rele(lo); 8853 goto end; 8854 } 8855 8856 /* 8857 * This is the new_lock_owner branch and the client is 8858 * supposed to be associating a new lock_owner with 8859 * the open file at this point. If we find that a 8860 * lock_owner/state association already exists and a 8861 * successful LOCK request was returned to the client, 8862 * an error is returned to the client since this is 8863 * not appropriate. The client should be using the 8864 * existing lock_owner branch. 8865 */ 8866 if (dup_lock == FALSE && create == FALSE) { 8867 if (lsp->rls_lock_completed == TRUE) { 8868 *cs->statusp = 8869 resp->status = NFS4ERR_BAD_SEQID; 8870 rfs4_lockowner_rele(lo); 8871 goto end; 8872 } 8873 } 8874 8875 rfs4_update_lease(sp->rs_owner->ro_client); 8876 8877 /* 8878 * Only update theh open_seqid if this is not 8879 * a duplicate request 8880 */ 8881 if (dup_lock == FALSE) { 8882 rfs4_update_open_sequence(sp->rs_owner); 8883 } 8884 8885 /* 8886 * If this is a duplicate lock request, just copy the 8887 * previously saved reply and return. 8888 */ 8889 if (dup_lock == TRUE) { 8890 /* verify that lock_seqid's match */ 8891 if (lsp->rls_seqid != olo->lock_seqid) { 8892 NFS4_DEBUG(rfs4_debug, 8893 (CE_NOTE, "rfs4_op_lock: Dup-Lock seqid bad" 8894 "lsp->seqid=%d old->seqid=%d", 8895 lsp->rls_seqid, olo->lock_seqid)); 8896 *cs->statusp = resp->status = NFS4ERR_BAD_SEQID; 8897 } else { 8898 rfs4_copy_reply(resop, &lsp->rls_reply); 8899 /* 8900 * Make sure to copy the just 8901 * retrieved reply status into the 8902 * overall compound status 8903 */ 8904 *cs->statusp = resp->status; 8905 } 8906 rfs4_lockowner_rele(lo); 8907 goto end; 8908 } 8909 8910 rfs4_dbe_lock(lsp->rls_dbe); 8911 8912 /* Make sure to update the lock sequence id */ 8913 lsp->rls_seqid = olo->lock_seqid; 8914 8915 NFS4_DEBUG(rfs4_debug, 8916 (CE_NOTE, "Lock seqid established as %d", lsp->rls_seqid)); 8917 8918 /* 8919 * This is used to signify the newly created lockowner 8920 * stateid and its sequence number. The checks for 8921 * sequence number and increment don't occur on the 8922 * very first lock request for a lockowner. 8923 */ 8924 lsp->rls_skip_seqid_check = TRUE; 8925 8926 /* hold off other access to lsp while we tinker */ 8927 rfs4_sw_enter(&lsp->rls_sw); 8928 ls_sw_held = TRUE; 8929 8930 rfs4_dbe_unlock(lsp->rls_dbe); 8931 8932 rfs4_lockowner_rele(lo); 8933 } else { 8934 stateid = &args->locker.locker4_u.lock_owner.lock_stateid; 8935 /* get lsp and hold the lock on the underlying file struct */ 8936 if ((status = rfs4_get_lo_state(stateid, &lsp, TRUE)) 8937 != NFS4_OK) { 8938 *cs->statusp = resp->status = status; 8939 DTRACE_NFSV4_2(op__lock__done, struct compound_state *, 8940 cs, LOCK4res *, resp); 8941 return; 8942 } 8943 create = FALSE; /* We didn't create lsp */ 8944 8945 /* Ensure specified filehandle matches */ 8946 if (cs->vp != lsp->rls_state->rs_finfo->rf_vp) { 8947 rfs4_lo_state_rele(lsp, TRUE); 8948 *cs->statusp = resp->status = NFS4ERR_BAD_STATEID; 8949 DTRACE_NFSV4_2(op__lock__done, struct compound_state *, 8950 cs, LOCK4res *, resp); 8951 return; 8952 } 8953 8954 /* hold off other access to lsp while we tinker */ 8955 rfs4_sw_enter(&lsp->rls_sw); 8956 ls_sw_held = TRUE; 8957 8958 switch (rfs4_check_lo_stateid_seqid(lsp, stateid)) { 8959 /* 8960 * The stateid looks like it was okay (expected to be 8961 * the next one) 8962 */ 8963 case NFS4_CHECK_STATEID_OKAY: 8964 /* 8965 * The sequence id is now checked. Determine 8966 * if this is a replay or if it is in the 8967 * expected (next) sequence. In the case of a 8968 * replay, there are two replay conditions 8969 * that may occur. The first is the normal 8970 * condition where a LOCK is done with a 8971 * NFS4_OK response and the stateid is 8972 * updated. That case is handled below when 8973 * the stateid is identified as a REPLAY. The 8974 * second is the case where an error is 8975 * returned, like NFS4ERR_DENIED, and the 8976 * sequence number is updated but the stateid 8977 * is not updated. This second case is dealt 8978 * with here. So it may seem odd that the 8979 * stateid is okay but the sequence id is a 8980 * replay but it is okay. 8981 */ 8982 switch (rfs4_check_lock_seqid( 8983 args->locker.locker4_u.lock_owner.lock_seqid, 8984 lsp, resop)) { 8985 case NFS4_CHKSEQ_REPLAY: 8986 if (resp->status != NFS4_OK) { 8987 /* 8988 * Here is our replay and need 8989 * to verify that the last 8990 * response was an error. 8991 */ 8992 *cs->statusp = resp->status; 8993 goto end; 8994 } 8995 /* 8996 * This is done since the sequence id 8997 * looked like a replay but it didn't 8998 * pass our check so a BAD_SEQID is 8999 * returned as a result. 9000 */ 9001 /*FALLTHROUGH*/ 9002 case NFS4_CHKSEQ_BAD: 9003 *cs->statusp = resp->status = NFS4ERR_BAD_SEQID; 9004 goto end; 9005 case NFS4_CHKSEQ_OKAY: 9006 /* Everything looks okay move ahead */ 9007 break; 9008 } 9009 break; 9010 case NFS4_CHECK_STATEID_OLD: 9011 *cs->statusp = resp->status = NFS4ERR_OLD_STATEID; 9012 goto end; 9013 case NFS4_CHECK_STATEID_BAD: 9014 *cs->statusp = resp->status = NFS4ERR_BAD_STATEID; 9015 goto end; 9016 case NFS4_CHECK_STATEID_EXPIRED: 9017 *cs->statusp = resp->status = NFS4ERR_EXPIRED; 9018 goto end; 9019 case NFS4_CHECK_STATEID_CLOSED: 9020 *cs->statusp = resp->status = NFS4ERR_OLD_STATEID; 9021 goto end; 9022 case NFS4_CHECK_STATEID_REPLAY: 9023 switch (rfs4_check_lock_seqid( 9024 args->locker.locker4_u.lock_owner.lock_seqid, 9025 lsp, resop)) { 9026 case NFS4_CHKSEQ_OKAY: 9027 /* 9028 * This is a replayed stateid; if 9029 * seqid matches the next expected, 9030 * then client is using wrong seqid. 9031 */ 9032 case NFS4_CHKSEQ_BAD: 9033 *cs->statusp = resp->status = NFS4ERR_BAD_SEQID; 9034 goto end; 9035 case NFS4_CHKSEQ_REPLAY: 9036 rfs4_update_lease(lsp->rls_locker->rl_client); 9037 *cs->statusp = status = resp->status; 9038 goto end; 9039 } 9040 break; 9041 default: 9042 ASSERT(FALSE); 9043 break; 9044 } 9045 9046 rfs4_update_lock_sequence(lsp); 9047 rfs4_update_lease(lsp->rls_locker->rl_client); 9048 } 9049 9050 /* 9051 * NFS4 only allows locking on regular files, so 9052 * verify type of object. 9053 */ 9054 if (cs->vp->v_type != VREG) { 9055 if (cs->vp->v_type == VDIR) 9056 status = NFS4ERR_ISDIR; 9057 else 9058 status = NFS4ERR_INVAL; 9059 goto out; 9060 } 9061 9062 cp = lsp->rls_state->rs_owner->ro_client; 9063 9064 if (rfs4_clnt_in_grace(cp) && !args->reclaim) { 9065 status = NFS4ERR_GRACE; 9066 goto out; 9067 } 9068 9069 if (rfs4_clnt_in_grace(cp) && args->reclaim && !cp->rc_can_reclaim) { 9070 status = NFS4ERR_NO_GRACE; 9071 goto out; 9072 } 9073 9074 if (!rfs4_clnt_in_grace(cp) && args->reclaim) { 9075 status = NFS4ERR_NO_GRACE; 9076 goto out; 9077 } 9078 9079 if (lsp->rls_state->rs_finfo->rf_dinfo.rd_dtype == OPEN_DELEGATE_WRITE) 9080 cs->deleg = TRUE; 9081 9082 status = rfs4_do_lock(lsp, args->locktype, 9083 args->offset, args->length, cs->cr, resop); 9084 9085 out: 9086 lsp->rls_skip_seqid_check = FALSE; 9087 9088 *cs->statusp = resp->status = status; 9089 9090 if (status == NFS4_OK) { 9091 resp->LOCK4res_u.lock_stateid = lsp->rls_lockid.stateid; 9092 lsp->rls_lock_completed = TRUE; 9093 } 9094 /* 9095 * Only update the "OPEN" response here if this was a new 9096 * lock_owner 9097 */ 9098 if (sp) 9099 rfs4_update_open_resp(sp->rs_owner, resop, NULL); 9100 9101 rfs4_update_lock_resp(lsp, resop); 9102 9103 end: 9104 if (lsp) { 9105 if (ls_sw_held) 9106 rfs4_sw_exit(&lsp->rls_sw); 9107 /* 9108 * If an sp obtained, then the lsp does not represent 9109 * a lock on the file struct. 9110 */ 9111 if (sp != NULL) 9112 rfs4_lo_state_rele(lsp, FALSE); 9113 else 9114 rfs4_lo_state_rele(lsp, TRUE); 9115 } 9116 if (sp) { 9117 rfs4_sw_exit(&sp->rs_owner->ro_sw); 9118 rfs4_state_rele(sp); 9119 } 9120 9121 DTRACE_NFSV4_2(op__lock__done, struct compound_state *, cs, 9122 LOCK4res *, resp); 9123 } 9124 9125 /* free function for LOCK/LOCKT */ 9126 static void 9127 lock_denied_free(nfs_resop4 *resop) 9128 { 9129 LOCK4denied *dp = NULL; 9130 9131 switch (resop->resop) { 9132 case OP_LOCK: 9133 if (resop->nfs_resop4_u.oplock.status == NFS4ERR_DENIED) 9134 dp = &resop->nfs_resop4_u.oplock.LOCK4res_u.denied; 9135 break; 9136 case OP_LOCKT: 9137 if (resop->nfs_resop4_u.oplockt.status == NFS4ERR_DENIED) 9138 dp = &resop->nfs_resop4_u.oplockt.denied; 9139 break; 9140 default: 9141 break; 9142 } 9143 9144 if (dp) 9145 kmem_free(dp->owner.owner_val, dp->owner.owner_len); 9146 } 9147 9148 /*ARGSUSED*/ 9149 void 9150 rfs4_op_locku(nfs_argop4 *argop, nfs_resop4 *resop, 9151 struct svc_req *req, struct compound_state *cs) 9152 { 9153 LOCKU4args *args = &argop->nfs_argop4_u.oplocku; 9154 LOCKU4res *resp = &resop->nfs_resop4_u.oplocku; 9155 nfsstat4 status; 9156 stateid4 *stateid = &args->lock_stateid; 9157 rfs4_lo_state_t *lsp; 9158 9159 DTRACE_NFSV4_2(op__locku__start, struct compound_state *, cs, 9160 LOCKU4args *, args); 9161 9162 if (cs->vp == NULL) { 9163 *cs->statusp = resp->status = NFS4ERR_NOFILEHANDLE; 9164 DTRACE_NFSV4_2(op__locku__done, struct compound_state *, cs, 9165 LOCKU4res *, resp); 9166 return; 9167 } 9168 9169 if ((status = rfs4_get_lo_state(stateid, &lsp, TRUE)) != NFS4_OK) { 9170 *cs->statusp = resp->status = status; 9171 DTRACE_NFSV4_2(op__locku__done, struct compound_state *, cs, 9172 LOCKU4res *, resp); 9173 return; 9174 } 9175 9176 /* Ensure specified filehandle matches */ 9177 if (cs->vp != lsp->rls_state->rs_finfo->rf_vp) { 9178 rfs4_lo_state_rele(lsp, TRUE); 9179 *cs->statusp = resp->status = NFS4ERR_BAD_STATEID; 9180 DTRACE_NFSV4_2(op__locku__done, struct compound_state *, cs, 9181 LOCKU4res *, resp); 9182 return; 9183 } 9184 9185 /* hold off other access to lsp while we tinker */ 9186 rfs4_sw_enter(&lsp->rls_sw); 9187 9188 switch (rfs4_check_lo_stateid_seqid(lsp, stateid)) { 9189 case NFS4_CHECK_STATEID_OKAY: 9190 if (rfs4_check_lock_seqid(args->seqid, lsp, resop) 9191 != NFS4_CHKSEQ_OKAY) { 9192 *cs->statusp = resp->status = NFS4ERR_BAD_SEQID; 9193 goto end; 9194 } 9195 break; 9196 case NFS4_CHECK_STATEID_OLD: 9197 *cs->statusp = resp->status = NFS4ERR_OLD_STATEID; 9198 goto end; 9199 case NFS4_CHECK_STATEID_BAD: 9200 *cs->statusp = resp->status = NFS4ERR_BAD_STATEID; 9201 goto end; 9202 case NFS4_CHECK_STATEID_EXPIRED: 9203 *cs->statusp = resp->status = NFS4ERR_EXPIRED; 9204 goto end; 9205 case NFS4_CHECK_STATEID_CLOSED: 9206 *cs->statusp = resp->status = NFS4ERR_OLD_STATEID; 9207 goto end; 9208 case NFS4_CHECK_STATEID_REPLAY: 9209 switch (rfs4_check_lock_seqid(args->seqid, lsp, resop)) { 9210 case NFS4_CHKSEQ_OKAY: 9211 /* 9212 * This is a replayed stateid; if 9213 * seqid matches the next expected, 9214 * then client is using wrong seqid. 9215 */ 9216 case NFS4_CHKSEQ_BAD: 9217 *cs->statusp = resp->status = NFS4ERR_BAD_SEQID; 9218 goto end; 9219 case NFS4_CHKSEQ_REPLAY: 9220 rfs4_update_lease(lsp->rls_locker->rl_client); 9221 *cs->statusp = status = resp->status; 9222 goto end; 9223 } 9224 break; 9225 default: 9226 ASSERT(FALSE); 9227 break; 9228 } 9229 9230 rfs4_update_lock_sequence(lsp); 9231 rfs4_update_lease(lsp->rls_locker->rl_client); 9232 9233 /* 9234 * NFS4 only allows locking on regular files, so 9235 * verify type of object. 9236 */ 9237 if (cs->vp->v_type != VREG) { 9238 if (cs->vp->v_type == VDIR) 9239 status = NFS4ERR_ISDIR; 9240 else 9241 status = NFS4ERR_INVAL; 9242 goto out; 9243 } 9244 9245 if (rfs4_clnt_in_grace(lsp->rls_state->rs_owner->ro_client)) { 9246 status = NFS4ERR_GRACE; 9247 goto out; 9248 } 9249 9250 status = rfs4_do_lock(lsp, args->locktype, 9251 args->offset, args->length, cs->cr, resop); 9252 9253 out: 9254 *cs->statusp = resp->status = status; 9255 9256 if (status == NFS4_OK) 9257 resp->lock_stateid = lsp->rls_lockid.stateid; 9258 9259 rfs4_update_lock_resp(lsp, resop); 9260 9261 end: 9262 rfs4_sw_exit(&lsp->rls_sw); 9263 rfs4_lo_state_rele(lsp, TRUE); 9264 9265 DTRACE_NFSV4_2(op__locku__done, struct compound_state *, cs, 9266 LOCKU4res *, resp); 9267 } 9268 9269 /* 9270 * LOCKT is a best effort routine, the client can not be guaranteed that 9271 * the status return is still in effect by the time the reply is received. 9272 * They are numerous race conditions in this routine, but we are not required 9273 * and can not be accurate. 9274 */ 9275 /*ARGSUSED*/ 9276 void 9277 rfs4_op_lockt(nfs_argop4 *argop, nfs_resop4 *resop, 9278 struct svc_req *req, struct compound_state *cs) 9279 { 9280 LOCKT4args *args = &argop->nfs_argop4_u.oplockt; 9281 LOCKT4res *resp = &resop->nfs_resop4_u.oplockt; 9282 rfs4_lockowner_t *lo; 9283 rfs4_client_t *cp; 9284 bool_t create = FALSE; 9285 struct flock64 flk; 9286 int error; 9287 int flag = FREAD | FWRITE; 9288 int ltype; 9289 length4 posix_length; 9290 sysid_t sysid; 9291 pid_t pid; 9292 9293 DTRACE_NFSV4_2(op__lockt__start, struct compound_state *, cs, 9294 LOCKT4args *, args); 9295 9296 if (cs->vp == NULL) { 9297 *cs->statusp = resp->status = NFS4ERR_NOFILEHANDLE; 9298 goto out; 9299 } 9300 9301 /* 9302 * NFS4 only allows locking on regular files, so 9303 * verify type of object. 9304 */ 9305 if (cs->vp->v_type != VREG) { 9306 if (cs->vp->v_type == VDIR) 9307 *cs->statusp = resp->status = NFS4ERR_ISDIR; 9308 else 9309 *cs->statusp = resp->status = NFS4ERR_INVAL; 9310 goto out; 9311 } 9312 9313 /* 9314 * Check out the clientid to ensure the server knows about it 9315 * so that we correctly inform the client of a server reboot. 9316 */ 9317 if ((cp = rfs4_findclient_by_id(args->owner.clientid, FALSE)) 9318 == NULL) { 9319 *cs->statusp = resp->status = 9320 rfs4_check_clientid(&args->owner.clientid, 0); 9321 goto out; 9322 } 9323 if (rfs4_lease_expired(cp)) { 9324 rfs4_client_close(cp); 9325 /* 9326 * Protocol doesn't allow returning NFS4ERR_STALE as 9327 * other operations do on this check so STALE_CLIENTID 9328 * is returned instead 9329 */ 9330 *cs->statusp = resp->status = NFS4ERR_STALE_CLIENTID; 9331 goto out; 9332 } 9333 9334 if (rfs4_clnt_in_grace(cp) && !(cp->rc_can_reclaim)) { 9335 *cs->statusp = resp->status = NFS4ERR_GRACE; 9336 rfs4_client_rele(cp); 9337 goto out; 9338 } 9339 rfs4_client_rele(cp); 9340 9341 resp->status = NFS4_OK; 9342 9343 switch (args->locktype) { 9344 case READ_LT: 9345 case READW_LT: 9346 ltype = F_RDLCK; 9347 break; 9348 case WRITE_LT: 9349 case WRITEW_LT: 9350 ltype = F_WRLCK; 9351 break; 9352 } 9353 9354 posix_length = args->length; 9355 /* Check for zero length. To lock to end of file use all ones for V4 */ 9356 if (posix_length == 0) { 9357 *cs->statusp = resp->status = NFS4ERR_INVAL; 9358 goto out; 9359 } else if (posix_length == (length4)(~0)) { 9360 posix_length = 0; /* Posix to end of file */ 9361 } 9362 9363 /* Find or create a lockowner */ 9364 lo = rfs4_findlockowner(&args->owner, &create); 9365 9366 if (lo) { 9367 pid = lo->rl_pid; 9368 if ((resp->status = 9369 rfs4_client_sysid(lo->rl_client, &sysid)) != NFS4_OK) 9370 goto err; 9371 } else { 9372 pid = 0; 9373 sysid = lockt_sysid; 9374 } 9375 retry: 9376 flk.l_type = ltype; 9377 flk.l_whence = 0; /* SEEK_SET */ 9378 flk.l_start = args->offset; 9379 flk.l_len = posix_length; 9380 flk.l_sysid = sysid; 9381 flk.l_pid = pid; 9382 flag |= F_REMOTELOCK; 9383 9384 LOCK_PRINT(rfs4_debug, "rfs4_op_lockt", F_GETLK, &flk); 9385 9386 /* Note that length4 is uint64_t but l_len and l_start are off64_t */ 9387 if (flk.l_len < 0 || flk.l_start < 0) { 9388 resp->status = NFS4ERR_INVAL; 9389 goto err; 9390 } 9391 error = VOP_FRLOCK(cs->vp, F_GETLK, &flk, flag, (u_offset_t)0, 9392 NULL, cs->cr, NULL); 9393 9394 /* 9395 * N.B. We map error values to nfsv4 errors. This is differrent 9396 * than puterrno4 routine. 9397 */ 9398 switch (error) { 9399 case 0: 9400 if (flk.l_type == F_UNLCK) 9401 resp->status = NFS4_OK; 9402 else { 9403 if (lock_denied(&resp->denied, &flk) == NFS4ERR_EXPIRED) 9404 goto retry; 9405 resp->status = NFS4ERR_DENIED; 9406 } 9407 break; 9408 case EOVERFLOW: 9409 resp->status = NFS4ERR_INVAL; 9410 break; 9411 case EINVAL: 9412 resp->status = NFS4ERR_NOTSUPP; 9413 break; 9414 default: 9415 cmn_err(CE_WARN, "rfs4_op_lockt: unexpected errno (%d)", 9416 error); 9417 resp->status = NFS4ERR_SERVERFAULT; 9418 break; 9419 } 9420 9421 err: 9422 if (lo) 9423 rfs4_lockowner_rele(lo); 9424 *cs->statusp = resp->status; 9425 out: 9426 DTRACE_NFSV4_2(op__lockt__done, struct compound_state *, cs, 9427 LOCKT4res *, resp); 9428 } 9429 9430 int 9431 rfs4_share(rfs4_state_t *sp, uint32_t access, uint32_t deny) 9432 { 9433 int err; 9434 int cmd; 9435 vnode_t *vp; 9436 struct shrlock shr; 9437 struct shr_locowner shr_loco; 9438 int fflags = 0; 9439 9440 ASSERT(rfs4_dbe_islocked(sp->rs_dbe)); 9441 ASSERT(sp->rs_owner->ro_client->rc_sysidt != LM_NOSYSID); 9442 9443 if (sp->rs_closed) 9444 return (NFS4ERR_OLD_STATEID); 9445 9446 vp = sp->rs_finfo->rf_vp; 9447 ASSERT(vp); 9448 9449 shr.s_access = shr.s_deny = 0; 9450 9451 if (access & OPEN4_SHARE_ACCESS_READ) { 9452 fflags |= FREAD; 9453 shr.s_access |= F_RDACC; 9454 } 9455 if (access & OPEN4_SHARE_ACCESS_WRITE) { 9456 fflags |= FWRITE; 9457 shr.s_access |= F_WRACC; 9458 } 9459 ASSERT(shr.s_access); 9460 9461 if (deny & OPEN4_SHARE_DENY_READ) 9462 shr.s_deny |= F_RDDNY; 9463 if (deny & OPEN4_SHARE_DENY_WRITE) 9464 shr.s_deny |= F_WRDNY; 9465 9466 shr.s_pid = rfs4_dbe_getid(sp->rs_owner->ro_dbe); 9467 shr.s_sysid = sp->rs_owner->ro_client->rc_sysidt; 9468 shr_loco.sl_pid = shr.s_pid; 9469 shr_loco.sl_id = shr.s_sysid; 9470 shr.s_owner = (caddr_t)&shr_loco; 9471 shr.s_own_len = sizeof (shr_loco); 9472 9473 cmd = nbl_need_check(vp) ? F_SHARE_NBMAND : F_SHARE; 9474 9475 err = VOP_SHRLOCK(vp, cmd, &shr, fflags, CRED(), NULL); 9476 if (err != 0) { 9477 if (err == EAGAIN) 9478 err = NFS4ERR_SHARE_DENIED; 9479 else 9480 err = puterrno4(err); 9481 return (err); 9482 } 9483 9484 sp->rs_share_access |= access; 9485 sp->rs_share_deny |= deny; 9486 9487 return (0); 9488 } 9489 9490 int 9491 rfs4_unshare(rfs4_state_t *sp) 9492 { 9493 int err; 9494 struct shrlock shr; 9495 struct shr_locowner shr_loco; 9496 9497 ASSERT(rfs4_dbe_islocked(sp->rs_dbe)); 9498 9499 if (sp->rs_closed || sp->rs_share_access == 0) 9500 return (0); 9501 9502 ASSERT(sp->rs_owner->ro_client->rc_sysidt != LM_NOSYSID); 9503 ASSERT(sp->rs_finfo->rf_vp); 9504 9505 shr.s_access = shr.s_deny = 0; 9506 shr.s_pid = rfs4_dbe_getid(sp->rs_owner->ro_dbe); 9507 shr.s_sysid = sp->rs_owner->ro_client->rc_sysidt; 9508 shr_loco.sl_pid = shr.s_pid; 9509 shr_loco.sl_id = shr.s_sysid; 9510 shr.s_owner = (caddr_t)&shr_loco; 9511 shr.s_own_len = sizeof (shr_loco); 9512 9513 err = VOP_SHRLOCK(sp->rs_finfo->rf_vp, F_UNSHARE, &shr, 0, CRED(), 9514 NULL); 9515 if (err != 0) { 9516 err = puterrno4(err); 9517 return (err); 9518 } 9519 9520 sp->rs_share_access = 0; 9521 sp->rs_share_deny = 0; 9522 9523 return (0); 9524 9525 } 9526 9527 static int 9528 rdma_setup_read_data4(READ4args *args, READ4res *rok) 9529 { 9530 struct clist *wcl; 9531 count4 count = rok->data_len; 9532 int wlist_len; 9533 9534 wcl = args->wlist; 9535 if (rdma_setup_read_chunks(wcl, count, &wlist_len) == FALSE) { 9536 return (FALSE); 9537 } 9538 wcl = args->wlist; 9539 rok->wlist_len = wlist_len; 9540 rok->wlist = wcl; 9541 return (TRUE); 9542 } 9543 9544 /* tunable to disable server referrals */ 9545 int rfs4_no_referrals = 0; 9546 9547 /* 9548 * Find an NFS record in reparse point data. 9549 * Returns 0 for success and <0 or an errno value on failure. 9550 */ 9551 int 9552 vn_find_nfs_record(vnode_t *vp, nvlist_t **nvlp, char **svcp, char **datap) 9553 { 9554 int err; 9555 char *stype, *val; 9556 nvlist_t *nvl; 9557 nvpair_t *curr; 9558 9559 if ((nvl = reparse_init()) == NULL) 9560 return (-1); 9561 9562 if ((err = reparse_vnode_parse(vp, nvl)) != 0) { 9563 reparse_free(nvl); 9564 return (err); 9565 } 9566 9567 curr = NULL; 9568 while ((curr = nvlist_next_nvpair(nvl, curr)) != NULL) { 9569 if ((stype = nvpair_name(curr)) == NULL) { 9570 reparse_free(nvl); 9571 return (-2); 9572 } 9573 if (strncasecmp(stype, "NFS", 3) == 0) 9574 break; 9575 } 9576 9577 if ((curr == NULL) || 9578 (nvpair_value_string(curr, &val))) { 9579 reparse_free(nvl); 9580 return (-3); 9581 } 9582 *nvlp = nvl; 9583 *svcp = stype; 9584 *datap = val; 9585 return (0); 9586 } 9587 9588 int 9589 vn_is_nfs_reparse(vnode_t *vp, cred_t *cr) 9590 { 9591 nvlist_t *nvl; 9592 char *s, *d; 9593 9594 if (rfs4_no_referrals != 0) 9595 return (B_FALSE); 9596 9597 if (vn_is_reparse(vp, cr, NULL) == B_FALSE) 9598 return (B_FALSE); 9599 9600 if (vn_find_nfs_record(vp, &nvl, &s, &d) != 0) 9601 return (B_FALSE); 9602 9603 reparse_free(nvl); 9604 9605 return (B_TRUE); 9606 } 9607 9608 /* 9609 * There is a user-level copy of this routine in ref_subr.c. 9610 * Changes should be kept in sync. 9611 */ 9612 static int 9613 nfs4_create_components(char *path, component4 *comp4) 9614 { 9615 int slen, plen, ncomp; 9616 char *ori_path, *nxtc, buf[MAXNAMELEN]; 9617 9618 if (path == NULL) 9619 return (0); 9620 9621 plen = strlen(path) + 1; /* include the terminator */ 9622 ori_path = path; 9623 ncomp = 0; 9624 9625 /* count number of components in the path */ 9626 for (nxtc = path; nxtc < ori_path + plen; nxtc++) { 9627 if (*nxtc == '/' || *nxtc == '\0' || *nxtc == '\n') { 9628 if ((slen = nxtc - path) == 0) { 9629 path = nxtc + 1; 9630 continue; 9631 } 9632 9633 if (comp4 != NULL) { 9634 bcopy(path, buf, slen); 9635 buf[slen] = '\0'; 9636 (void) str_to_utf8(buf, &comp4[ncomp]); 9637 } 9638 9639 ncomp++; /* 1 valid component */ 9640 path = nxtc + 1; 9641 } 9642 if (*nxtc == '\0' || *nxtc == '\n') 9643 break; 9644 } 9645 9646 return (ncomp); 9647 } 9648 9649 /* 9650 * There is a user-level copy of this routine in ref_subr.c. 9651 * Changes should be kept in sync. 9652 */ 9653 static int 9654 make_pathname4(char *path, pathname4 *pathname) 9655 { 9656 int ncomp; 9657 component4 *comp4; 9658 9659 if (pathname == NULL) 9660 return (0); 9661 9662 if (path == NULL) { 9663 pathname->pathname4_val = NULL; 9664 pathname->pathname4_len = 0; 9665 return (0); 9666 } 9667 9668 /* count number of components to alloc buffer */ 9669 if ((ncomp = nfs4_create_components(path, NULL)) == 0) { 9670 pathname->pathname4_val = NULL; 9671 pathname->pathname4_len = 0; 9672 return (0); 9673 } 9674 comp4 = kmem_zalloc(ncomp * sizeof (component4), KM_SLEEP); 9675 9676 /* copy components into allocated buffer */ 9677 ncomp = nfs4_create_components(path, comp4); 9678 9679 pathname->pathname4_val = comp4; 9680 pathname->pathname4_len = ncomp; 9681 9682 return (ncomp); 9683 } 9684 9685 #define xdr_fs_locations4 xdr_fattr4_fs_locations 9686 9687 fs_locations4 * 9688 fetch_referral(vnode_t *vp, cred_t *cr) 9689 { 9690 nvlist_t *nvl; 9691 char *stype, *sdata; 9692 fs_locations4 *result; 9693 char buf[1024]; 9694 size_t bufsize; 9695 XDR xdr; 9696 int err; 9697 9698 /* 9699 * Check attrs to ensure it's a reparse point 9700 */ 9701 if (vn_is_reparse(vp, cr, NULL) == B_FALSE) 9702 return (NULL); 9703 9704 /* 9705 * Look for an NFS record and get the type and data 9706 */ 9707 if (vn_find_nfs_record(vp, &nvl, &stype, &sdata) != 0) 9708 return (NULL); 9709 9710 /* 9711 * With the type and data, upcall to get the referral 9712 */ 9713 bufsize = sizeof (buf); 9714 bzero(buf, sizeof (buf)); 9715 err = reparse_kderef((const char *)stype, (const char *)sdata, 9716 buf, &bufsize); 9717 reparse_free(nvl); 9718 9719 DTRACE_PROBE4(nfs4serv__func__referral__upcall, 9720 char *, stype, char *, sdata, char *, buf, int, err); 9721 if (err) { 9722 cmn_err(CE_NOTE, 9723 "reparsed daemon not running: unable to get referral (%d)", 9724 err); 9725 return (NULL); 9726 } 9727 9728 /* 9729 * We get an XDR'ed record back from the kderef call 9730 */ 9731 xdrmem_create(&xdr, buf, bufsize, XDR_DECODE); 9732 result = kmem_alloc(sizeof (fs_locations4), KM_SLEEP); 9733 err = xdr_fs_locations4(&xdr, result); 9734 XDR_DESTROY(&xdr); 9735 if (err != TRUE) { 9736 DTRACE_PROBE1(nfs4serv__func__referral__upcall__xdrfail, 9737 int, err); 9738 return (NULL); 9739 } 9740 9741 /* 9742 * Look at path to recover fs_root, ignoring the leading '/' 9743 */ 9744 (void) make_pathname4(vp->v_path, &result->fs_root); 9745 9746 return (result); 9747 } 9748 9749 char * 9750 build_symlink(vnode_t *vp, cred_t *cr, size_t *strsz) 9751 { 9752 fs_locations4 *fsl; 9753 fs_location4 *fs; 9754 char *server, *path, *symbuf; 9755 static char *prefix = "/net/"; 9756 int i, size, npaths; 9757 uint_t len; 9758 9759 /* Get the referral */ 9760 if ((fsl = fetch_referral(vp, cr)) == NULL) 9761 return (NULL); 9762 9763 /* Deal with only the first location and first server */ 9764 fs = &fsl->locations_val[0]; 9765 server = utf8_to_str(&fs->server_val[0], &len, NULL); 9766 if (server == NULL) { 9767 rfs4_free_fs_locations4(fsl); 9768 kmem_free(fsl, sizeof (fs_locations4)); 9769 return (NULL); 9770 } 9771 9772 /* Figure out size for "/net/" + host + /path/path/path + NULL */ 9773 size = strlen(prefix) + len; 9774 for (i = 0; i < fs->rootpath.pathname4_len; i++) 9775 size += fs->rootpath.pathname4_val[i].utf8string_len + 1; 9776 9777 /* Allocate the symlink buffer and fill it */ 9778 symbuf = kmem_zalloc(size, KM_SLEEP); 9779 (void) strcat(symbuf, prefix); 9780 (void) strcat(symbuf, server); 9781 kmem_free(server, len); 9782 9783 npaths = 0; 9784 for (i = 0; i < fs->rootpath.pathname4_len; i++) { 9785 path = utf8_to_str(&fs->rootpath.pathname4_val[i], &len, NULL); 9786 if (path == NULL) 9787 continue; 9788 (void) strcat(symbuf, "/"); 9789 (void) strcat(symbuf, path); 9790 npaths++; 9791 kmem_free(path, len); 9792 } 9793 9794 rfs4_free_fs_locations4(fsl); 9795 kmem_free(fsl, sizeof (fs_locations4)); 9796 9797 if (strsz != NULL) 9798 *strsz = size; 9799 return (symbuf); 9800 } 9801 9802 /* 9803 * Check to see if we have a downrev Solaris client, so that we 9804 * can send it a symlink instead of a referral. 9805 */ 9806 int 9807 client_is_downrev(struct svc_req *req) 9808 { 9809 struct sockaddr *ca; 9810 rfs4_clntip_t *ci; 9811 bool_t create = FALSE; 9812 int is_downrev; 9813 9814 ca = (struct sockaddr *)svc_getrpccaller(req->rq_xprt)->buf; 9815 ASSERT(ca); 9816 ci = rfs4_find_clntip(ca, &create); 9817 if (ci == NULL) 9818 return (0); 9819 is_downrev = ci->ri_no_referrals; 9820 rfs4_dbe_rele(ci->ri_dbe); 9821 return (is_downrev); 9822 } 9823