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 * set_rdattr_params sets up the variables used to manage what information 3669 * to get for each directory entry. 3670 */ 3671 static nfsstat4 3672 set_rdattr_params(struct nfs4_svgetit_arg *sargp, 3673 bitmap4 attrs, bool_t *need_to_lookup) 3674 { 3675 uint_t va_mask; 3676 nfsstat4 status; 3677 bitmap4 objbits; 3678 3679 status = bitmap4_to_attrmask(attrs, sargp); 3680 if (status != NFS4_OK) { 3681 /* 3682 * could not even figure attr mask 3683 */ 3684 return (status); 3685 } 3686 va_mask = sargp->vap->va_mask; 3687 3688 /* 3689 * dirent's d_ino is always correct value for mounted_on_fileid. 3690 * mntdfid_set is set once here, but mounted_on_fileid is 3691 * set in main dirent processing loop for each dirent. 3692 * The mntdfid_set is a simple optimization that lets the 3693 * server attr code avoid work when caller is readdir. 3694 */ 3695 sargp->mntdfid_set = TRUE; 3696 3697 /* 3698 * Lookup entry only if client asked for any of the following: 3699 * a) vattr attrs 3700 * b) vfs attrs 3701 * c) attrs w/per-object scope requested (change, filehandle, etc) 3702 * other than mounted_on_fileid (which we can take from dirent) 3703 */ 3704 objbits = attrs ? attrs & NFS4_VP_ATTR_MASK : 0; 3705 3706 if (va_mask || sargp->sbp || (objbits & ~FATTR4_MOUNTED_ON_FILEID_MASK)) 3707 *need_to_lookup = TRUE; 3708 else 3709 *need_to_lookup = FALSE; 3710 3711 if (sargp->sbp == NULL) 3712 return (NFS4_OK); 3713 3714 /* 3715 * If filesystem attrs are requested, get them now from the 3716 * directory vp, as most entries will have same filesystem. The only 3717 * exception are mounted over entries but we handle 3718 * those as we go (XXX mounted over detection not yet implemented). 3719 */ 3720 sargp->vap->va_mask = 0; /* to avoid VOP_GETATTR */ 3721 status = bitmap4_get_sysattrs(sargp); 3722 sargp->vap->va_mask = va_mask; 3723 3724 if ((status != NFS4_OK) && sargp->rdattr_error_req) { 3725 /* 3726 * Failed to get filesystem attributes. 3727 * Return a rdattr_error for each entry, but don't fail. 3728 * However, don't get any obj-dependent attrs. 3729 */ 3730 sargp->rdattr_error = status; /* for rdattr_error */ 3731 *need_to_lookup = FALSE; 3732 /* 3733 * At least get fileid for regular readdir output 3734 */ 3735 sargp->vap->va_mask &= AT_NODEID; 3736 status = NFS4_OK; 3737 } 3738 3739 return (status); 3740 } 3741 3742 /* 3743 * readlink: args: CURRENT_FH. 3744 * res: status. If success - CURRENT_FH unchanged, return linktext. 3745 */ 3746 3747 /* ARGSUSED */ 3748 static void 3749 rfs4_op_readlink(nfs_argop4 *argop, nfs_resop4 *resop, struct svc_req *req, 3750 struct compound_state *cs) 3751 { 3752 READLINK4res *resp = &resop->nfs_resop4_u.opreadlink; 3753 int error; 3754 vnode_t *vp; 3755 struct iovec iov; 3756 struct vattr va; 3757 struct uio uio; 3758 char *data; 3759 struct sockaddr *ca; 3760 char *name = NULL; 3761 int is_referral; 3762 3763 DTRACE_NFSV4_1(op__readlink__start, struct compound_state *, cs); 3764 3765 /* CURRENT_FH: directory */ 3766 vp = cs->vp; 3767 if (vp == NULL) { 3768 *cs->statusp = resp->status = NFS4ERR_NOFILEHANDLE; 3769 goto out; 3770 } 3771 3772 if (cs->access == CS_ACCESS_DENIED) { 3773 *cs->statusp = resp->status = NFS4ERR_ACCESS; 3774 goto out; 3775 } 3776 3777 /* Is it a referral? */ 3778 if (vn_is_nfs_reparse(vp, cs->cr) && client_is_downrev(req)) { 3779 3780 is_referral = 1; 3781 3782 } else { 3783 3784 is_referral = 0; 3785 3786 if (vp->v_type == VDIR) { 3787 *cs->statusp = resp->status = NFS4ERR_ISDIR; 3788 goto out; 3789 } 3790 3791 if (vp->v_type != VLNK) { 3792 *cs->statusp = resp->status = NFS4ERR_INVAL; 3793 goto out; 3794 } 3795 3796 } 3797 3798 va.va_mask = AT_MODE; 3799 error = VOP_GETATTR(vp, &va, 0, cs->cr, NULL); 3800 if (error) { 3801 *cs->statusp = resp->status = puterrno4(error); 3802 goto out; 3803 } 3804 3805 if (MANDLOCK(vp, va.va_mode)) { 3806 *cs->statusp = resp->status = NFS4ERR_ACCESS; 3807 goto out; 3808 } 3809 3810 data = kmem_alloc(MAXPATHLEN + 1, KM_SLEEP); 3811 3812 if (is_referral) { 3813 char *s; 3814 size_t strsz; 3815 3816 /* Get an artificial symlink based on a referral */ 3817 s = build_symlink(vp, cs->cr, &strsz); 3818 global_svstat_ptr[4][NFS_REFERLINKS].value.ui64++; 3819 DTRACE_PROBE2(nfs4serv__func__referral__reflink, 3820 vnode_t *, vp, char *, s); 3821 if (s == NULL) 3822 error = EINVAL; 3823 else { 3824 error = 0; 3825 (void) strlcpy(data, s, MAXPATHLEN + 1); 3826 kmem_free(s, strsz); 3827 } 3828 3829 } else { 3830 3831 iov.iov_base = data; 3832 iov.iov_len = MAXPATHLEN; 3833 uio.uio_iov = &iov; 3834 uio.uio_iovcnt = 1; 3835 uio.uio_segflg = UIO_SYSSPACE; 3836 uio.uio_extflg = UIO_COPY_CACHED; 3837 uio.uio_loffset = 0; 3838 uio.uio_resid = MAXPATHLEN; 3839 3840 error = VOP_READLINK(vp, &uio, cs->cr, NULL); 3841 3842 if (!error) 3843 *(data + MAXPATHLEN - uio.uio_resid) = '\0'; 3844 } 3845 3846 if (error) { 3847 kmem_free((caddr_t)data, (uint_t)MAXPATHLEN + 1); 3848 *cs->statusp = resp->status = puterrno4(error); 3849 goto out; 3850 } 3851 3852 ca = (struct sockaddr *)svc_getrpccaller(req->rq_xprt)->buf; 3853 name = nfscmd_convname(ca, cs->exi, data, NFSCMD_CONV_OUTBOUND, 3854 MAXPATHLEN + 1); 3855 3856 if (name == NULL) { 3857 /* 3858 * Even though the conversion failed, we return 3859 * something. We just don't translate it. 3860 */ 3861 name = data; 3862 } 3863 3864 /* 3865 * treat link name as data 3866 */ 3867 (void) str_to_utf8(name, &resp->link); 3868 3869 if (name != data) 3870 kmem_free(name, MAXPATHLEN + 1); 3871 kmem_free((caddr_t)data, (uint_t)MAXPATHLEN + 1); 3872 *cs->statusp = resp->status = NFS4_OK; 3873 3874 out: 3875 DTRACE_NFSV4_2(op__readlink__done, struct compound_state *, cs, 3876 READLINK4res *, resp); 3877 } 3878 3879 static void 3880 rfs4_op_readlink_free(nfs_resop4 *resop) 3881 { 3882 READLINK4res *resp = &resop->nfs_resop4_u.opreadlink; 3883 utf8string *symlink = &resp->link; 3884 3885 if (symlink->utf8string_val) { 3886 UTF8STRING_FREE(*symlink) 3887 } 3888 } 3889 3890 /* 3891 * release_lockowner: 3892 * Release any state associated with the supplied 3893 * lockowner. Note if any lo_state is holding locks we will not 3894 * rele that lo_state and thus the lockowner will not be destroyed. 3895 * A client using lock after the lock owner stateid has been released 3896 * will suffer the consequence of NFS4ERR_BAD_STATEID and would have 3897 * to reissue the lock with new_lock_owner set to TRUE. 3898 * args: lock_owner 3899 * res: status 3900 */ 3901 /* ARGSUSED */ 3902 static void 3903 rfs4_op_release_lockowner(nfs_argop4 *argop, nfs_resop4 *resop, 3904 struct svc_req *req, struct compound_state *cs) 3905 { 3906 RELEASE_LOCKOWNER4args *ap = &argop->nfs_argop4_u.oprelease_lockowner; 3907 RELEASE_LOCKOWNER4res *resp = &resop->nfs_resop4_u.oprelease_lockowner; 3908 rfs4_lockowner_t *lo; 3909 rfs4_openowner_t *oo; 3910 rfs4_state_t *sp; 3911 rfs4_lo_state_t *lsp; 3912 rfs4_client_t *cp; 3913 bool_t create = FALSE; 3914 locklist_t *llist; 3915 sysid_t sysid; 3916 3917 DTRACE_NFSV4_2(op__release__lockowner__start, struct compound_state *, 3918 cs, RELEASE_LOCKOWNER4args *, ap); 3919 3920 /* Make sure there is a clientid around for this request */ 3921 cp = rfs4_findclient_by_id(ap->lock_owner.clientid, FALSE); 3922 3923 if (cp == NULL) { 3924 *cs->statusp = resp->status = 3925 rfs4_check_clientid(&ap->lock_owner.clientid, 0); 3926 goto out; 3927 } 3928 rfs4_client_rele(cp); 3929 3930 lo = rfs4_findlockowner(&ap->lock_owner, &create); 3931 if (lo == NULL) { 3932 *cs->statusp = resp->status = NFS4_OK; 3933 goto out; 3934 } 3935 ASSERT(lo->rl_client != NULL); 3936 3937 /* 3938 * Check for EXPIRED client. If so will reap state with in a lease 3939 * period or on next set_clientid_confirm step 3940 */ 3941 if (rfs4_lease_expired(lo->rl_client)) { 3942 rfs4_lockowner_rele(lo); 3943 *cs->statusp = resp->status = NFS4ERR_EXPIRED; 3944 goto out; 3945 } 3946 3947 /* 3948 * If no sysid has been assigned, then no locks exist; just return. 3949 */ 3950 rfs4_dbe_lock(lo->rl_client->rc_dbe); 3951 if (lo->rl_client->rc_sysidt == LM_NOSYSID) { 3952 rfs4_lockowner_rele(lo); 3953 rfs4_dbe_unlock(lo->rl_client->rc_dbe); 3954 goto out; 3955 } 3956 3957 sysid = lo->rl_client->rc_sysidt; 3958 rfs4_dbe_unlock(lo->rl_client->rc_dbe); 3959 3960 /* 3961 * Mark the lockowner invalid. 3962 */ 3963 rfs4_dbe_hide(lo->rl_dbe); 3964 3965 /* 3966 * sysid-pid pair should now not be used since the lockowner is 3967 * invalid. If the client were to instantiate the lockowner again 3968 * it would be assigned a new pid. Thus we can get the list of 3969 * current locks. 3970 */ 3971 3972 llist = flk_get_active_locks(sysid, lo->rl_pid); 3973 /* If we are still holding locks fail */ 3974 if (llist != NULL) { 3975 3976 *cs->statusp = resp->status = NFS4ERR_LOCKS_HELD; 3977 3978 flk_free_locklist(llist); 3979 /* 3980 * We need to unhide the lockowner so the client can 3981 * try it again. The bad thing here is if the client 3982 * has a logic error that took it here in the first place 3983 * he probably has lost accounting of the locks that it 3984 * is holding. So we may have dangling state until the 3985 * open owner state is reaped via close. One scenario 3986 * that could possibly occur is that the client has 3987 * sent the unlock request(s) in separate threads 3988 * and has not waited for the replies before sending the 3989 * RELEASE_LOCKOWNER request. Presumably, it would expect 3990 * and deal appropriately with NFS4ERR_LOCKS_HELD, by 3991 * reissuing the request. 3992 */ 3993 rfs4_dbe_unhide(lo->rl_dbe); 3994 rfs4_lockowner_rele(lo); 3995 goto out; 3996 } 3997 3998 /* 3999 * For the corresponding client we need to check each open 4000 * owner for any opens that have lockowner state associated 4001 * with this lockowner. 4002 */ 4003 4004 rfs4_dbe_lock(lo->rl_client->rc_dbe); 4005 for (oo = list_head(&lo->rl_client->rc_openownerlist); oo != NULL; 4006 oo = list_next(&lo->rl_client->rc_openownerlist, oo)) { 4007 4008 rfs4_dbe_lock(oo->ro_dbe); 4009 for (sp = list_head(&oo->ro_statelist); sp != NULL; 4010 sp = list_next(&oo->ro_statelist, sp)) { 4011 4012 rfs4_dbe_lock(sp->rs_dbe); 4013 for (lsp = list_head(&sp->rs_lostatelist); 4014 lsp != NULL; 4015 lsp = list_next(&sp->rs_lostatelist, lsp)) { 4016 if (lsp->rls_locker == lo) { 4017 rfs4_dbe_lock(lsp->rls_dbe); 4018 rfs4_dbe_invalidate(lsp->rls_dbe); 4019 rfs4_dbe_unlock(lsp->rls_dbe); 4020 } 4021 } 4022 rfs4_dbe_unlock(sp->rs_dbe); 4023 } 4024 rfs4_dbe_unlock(oo->ro_dbe); 4025 } 4026 rfs4_dbe_unlock(lo->rl_client->rc_dbe); 4027 4028 rfs4_lockowner_rele(lo); 4029 4030 *cs->statusp = resp->status = NFS4_OK; 4031 4032 out: 4033 DTRACE_NFSV4_2(op__release__lockowner__done, struct compound_state *, 4034 cs, RELEASE_LOCKOWNER4res *, resp); 4035 } 4036 4037 /* 4038 * short utility function to lookup a file and recall the delegation 4039 */ 4040 static rfs4_file_t * 4041 rfs4_lookup_and_findfile(vnode_t *dvp, char *nm, vnode_t **vpp, 4042 int *lkup_error, cred_t *cr) 4043 { 4044 vnode_t *vp; 4045 rfs4_file_t *fp = NULL; 4046 bool_t fcreate = FALSE; 4047 int error; 4048 4049 if (vpp) 4050 *vpp = NULL; 4051 4052 if ((error = VOP_LOOKUP(dvp, nm, &vp, NULL, 0, NULL, cr, NULL, NULL, 4053 NULL)) == 0) { 4054 if (vp->v_type == VREG) 4055 fp = rfs4_findfile(vp, NULL, &fcreate); 4056 if (vpp) 4057 *vpp = vp; 4058 else 4059 VN_RELE(vp); 4060 } 4061 4062 if (lkup_error) 4063 *lkup_error = error; 4064 4065 return (fp); 4066 } 4067 4068 /* 4069 * remove: args: CURRENT_FH: directory; name. 4070 * res: status. If success - CURRENT_FH unchanged, return change_info 4071 * for directory. 4072 */ 4073 /* ARGSUSED */ 4074 static void 4075 rfs4_op_remove(nfs_argop4 *argop, nfs_resop4 *resop, struct svc_req *req, 4076 struct compound_state *cs) 4077 { 4078 REMOVE4args *args = &argop->nfs_argop4_u.opremove; 4079 REMOVE4res *resp = &resop->nfs_resop4_u.opremove; 4080 int error; 4081 vnode_t *dvp, *vp; 4082 struct vattr bdva, idva, adva; 4083 char *nm; 4084 uint_t len; 4085 rfs4_file_t *fp; 4086 int in_crit = 0; 4087 bslabel_t *clabel; 4088 struct sockaddr *ca; 4089 char *name = NULL; 4090 nfsstat4 status; 4091 4092 DTRACE_NFSV4_2(op__remove__start, struct compound_state *, cs, 4093 REMOVE4args *, args); 4094 4095 /* CURRENT_FH: directory */ 4096 dvp = cs->vp; 4097 if (dvp == NULL) { 4098 *cs->statusp = resp->status = NFS4ERR_NOFILEHANDLE; 4099 goto out; 4100 } 4101 4102 if (cs->access == CS_ACCESS_DENIED) { 4103 *cs->statusp = resp->status = NFS4ERR_ACCESS; 4104 goto out; 4105 } 4106 4107 /* 4108 * If there is an unshared filesystem mounted on this vnode, 4109 * Do not allow to remove anything in this directory. 4110 */ 4111 if (vn_ismntpt(dvp)) { 4112 *cs->statusp = resp->status = NFS4ERR_ACCESS; 4113 goto out; 4114 } 4115 4116 if (dvp->v_type != VDIR) { 4117 *cs->statusp = resp->status = NFS4ERR_NOTDIR; 4118 goto out; 4119 } 4120 4121 status = utf8_dir_verify(&args->target); 4122 if (status != NFS4_OK) { 4123 *cs->statusp = resp->status = status; 4124 goto out; 4125 } 4126 4127 /* 4128 * Lookup the file so that we can check if it's a directory 4129 */ 4130 nm = utf8_to_fn(&args->target, &len, NULL); 4131 if (nm == NULL) { 4132 *cs->statusp = resp->status = NFS4ERR_INVAL; 4133 goto out; 4134 } 4135 4136 if (len > MAXNAMELEN) { 4137 *cs->statusp = resp->status = NFS4ERR_NAMETOOLONG; 4138 kmem_free(nm, len); 4139 goto out; 4140 } 4141 4142 if (rdonly4(cs->exi, cs->vp, req)) { 4143 *cs->statusp = resp->status = NFS4ERR_ROFS; 4144 kmem_free(nm, len); 4145 goto out; 4146 } 4147 4148 ca = (struct sockaddr *)svc_getrpccaller(req->rq_xprt)->buf; 4149 name = nfscmd_convname(ca, cs->exi, nm, NFSCMD_CONV_INBOUND, 4150 MAXPATHLEN + 1); 4151 4152 if (name == NULL) { 4153 *cs->statusp = resp->status = NFS4ERR_INVAL; 4154 kmem_free(nm, len); 4155 goto out; 4156 } 4157 4158 /* 4159 * Lookup the file to determine type and while we are see if 4160 * there is a file struct around and check for delegation. 4161 * We don't need to acquire va_seq before this lookup, if 4162 * it causes an update, cinfo.before will not match, which will 4163 * trigger a cache flush even if atomic is TRUE. 4164 */ 4165 if (fp = rfs4_lookup_and_findfile(dvp, name, &vp, &error, cs->cr)) { 4166 if (rfs4_check_delegated_byfp(FWRITE, fp, TRUE, TRUE, TRUE, 4167 NULL)) { 4168 VN_RELE(vp); 4169 rfs4_file_rele(fp); 4170 *cs->statusp = resp->status = NFS4ERR_DELAY; 4171 if (nm != name) 4172 kmem_free(name, MAXPATHLEN + 1); 4173 kmem_free(nm, len); 4174 goto out; 4175 } 4176 } 4177 4178 /* Didn't find anything to remove */ 4179 if (vp == NULL) { 4180 *cs->statusp = resp->status = error; 4181 if (nm != name) 4182 kmem_free(name, MAXPATHLEN + 1); 4183 kmem_free(nm, len); 4184 goto out; 4185 } 4186 4187 if (nbl_need_check(vp)) { 4188 nbl_start_crit(vp, RW_READER); 4189 in_crit = 1; 4190 if (nbl_conflict(vp, NBL_REMOVE, 0, 0, 0, NULL)) { 4191 *cs->statusp = resp->status = NFS4ERR_FILE_OPEN; 4192 if (nm != name) 4193 kmem_free(name, MAXPATHLEN + 1); 4194 kmem_free(nm, len); 4195 nbl_end_crit(vp); 4196 VN_RELE(vp); 4197 if (fp) { 4198 rfs4_clear_dont_grant(fp); 4199 rfs4_file_rele(fp); 4200 } 4201 goto out; 4202 } 4203 } 4204 4205 /* check label before allowing removal */ 4206 if (is_system_labeled()) { 4207 ASSERT(req->rq_label != NULL); 4208 clabel = req->rq_label; 4209 DTRACE_PROBE2(tx__rfs4__log__info__opremove__clabel, char *, 4210 "got client label from request(1)", 4211 struct svc_req *, req); 4212 if (!blequal(&l_admin_low->tsl_label, clabel)) { 4213 if (!do_rfs_label_check(clabel, vp, EQUALITY_CHECK, 4214 cs->exi)) { 4215 *cs->statusp = resp->status = NFS4ERR_ACCESS; 4216 if (name != nm) 4217 kmem_free(name, MAXPATHLEN + 1); 4218 kmem_free(nm, len); 4219 if (in_crit) 4220 nbl_end_crit(vp); 4221 VN_RELE(vp); 4222 if (fp) { 4223 rfs4_clear_dont_grant(fp); 4224 rfs4_file_rele(fp); 4225 } 4226 goto out; 4227 } 4228 } 4229 } 4230 4231 /* Get dir "before" change value */ 4232 bdva.va_mask = AT_CTIME|AT_SEQ; 4233 error = VOP_GETATTR(dvp, &bdva, 0, cs->cr, NULL); 4234 if (error) { 4235 *cs->statusp = resp->status = puterrno4(error); 4236 if (nm != name) 4237 kmem_free(name, MAXPATHLEN + 1); 4238 kmem_free(nm, len); 4239 if (in_crit) 4240 nbl_end_crit(vp); 4241 VN_RELE(vp); 4242 if (fp) { 4243 rfs4_clear_dont_grant(fp); 4244 rfs4_file_rele(fp); 4245 } 4246 goto out; 4247 } 4248 NFS4_SET_FATTR4_CHANGE(resp->cinfo.before, bdva.va_ctime) 4249 4250 /* Actually do the REMOVE operation */ 4251 if (vp->v_type == VDIR) { 4252 /* 4253 * Can't remove a directory that has a mounted-on filesystem. 4254 */ 4255 if (vn_ismntpt(vp)) { 4256 error = EACCES; 4257 } else { 4258 /* 4259 * System V defines rmdir to return EEXIST, 4260 * not ENOTEMPTY, if the directory is not 4261 * empty. A System V NFS server needs to map 4262 * NFS4ERR_EXIST to NFS4ERR_NOTEMPTY to 4263 * transmit over the wire. 4264 */ 4265 if ((error = VOP_RMDIR(dvp, name, rootdir, cs->cr, 4266 NULL, 0)) == EEXIST) 4267 error = ENOTEMPTY; 4268 } 4269 } else { 4270 if ((error = VOP_REMOVE(dvp, name, cs->cr, NULL, 0)) == 0 && 4271 fp != NULL) { 4272 struct vattr va; 4273 vnode_t *tvp; 4274 4275 rfs4_dbe_lock(fp->rf_dbe); 4276 tvp = fp->rf_vp; 4277 if (tvp) 4278 VN_HOLD(tvp); 4279 rfs4_dbe_unlock(fp->rf_dbe); 4280 4281 if (tvp) { 4282 /* 4283 * This is va_seq safe because we are not 4284 * manipulating dvp. 4285 */ 4286 va.va_mask = AT_NLINK; 4287 if (!VOP_GETATTR(tvp, &va, 0, cs->cr, NULL) && 4288 va.va_nlink == 0) { 4289 /* Remove state on file remove */ 4290 if (in_crit) { 4291 nbl_end_crit(vp); 4292 in_crit = 0; 4293 } 4294 rfs4_close_all_state(fp); 4295 } 4296 VN_RELE(tvp); 4297 } 4298 } 4299 } 4300 4301 if (in_crit) 4302 nbl_end_crit(vp); 4303 VN_RELE(vp); 4304 4305 if (fp) { 4306 rfs4_clear_dont_grant(fp); 4307 rfs4_file_rele(fp); 4308 } 4309 if (nm != name) 4310 kmem_free(name, MAXPATHLEN + 1); 4311 kmem_free(nm, len); 4312 4313 if (error) { 4314 *cs->statusp = resp->status = puterrno4(error); 4315 goto out; 4316 } 4317 4318 /* 4319 * Get the initial "after" sequence number, if it fails, set to zero 4320 */ 4321 idva.va_mask = AT_SEQ; 4322 if (VOP_GETATTR(dvp, &idva, 0, cs->cr, NULL)) 4323 idva.va_seq = 0; 4324 4325 /* 4326 * Force modified data and metadata out to stable storage. 4327 */ 4328 (void) VOP_FSYNC(dvp, 0, cs->cr, NULL); 4329 4330 /* 4331 * Get "after" change value, if it fails, simply return the 4332 * before value. 4333 */ 4334 adva.va_mask = AT_CTIME|AT_SEQ; 4335 if (VOP_GETATTR(dvp, &adva, 0, cs->cr, NULL)) { 4336 adva.va_ctime = bdva.va_ctime; 4337 adva.va_seq = 0; 4338 } 4339 4340 NFS4_SET_FATTR4_CHANGE(resp->cinfo.after, adva.va_ctime) 4341 4342 /* 4343 * The cinfo.atomic = TRUE only if we have 4344 * non-zero va_seq's, and it has incremented by exactly one 4345 * during the VOP_REMOVE/RMDIR and it didn't change during 4346 * the VOP_FSYNC. 4347 */ 4348 if (bdva.va_seq && idva.va_seq && adva.va_seq && 4349 idva.va_seq == (bdva.va_seq + 1) && idva.va_seq == adva.va_seq) 4350 resp->cinfo.atomic = TRUE; 4351 else 4352 resp->cinfo.atomic = FALSE; 4353 4354 *cs->statusp = resp->status = NFS4_OK; 4355 4356 out: 4357 DTRACE_NFSV4_2(op__remove__done, struct compound_state *, cs, 4358 REMOVE4res *, resp); 4359 } 4360 4361 /* 4362 * rename: args: SAVED_FH: from directory, CURRENT_FH: target directory, 4363 * oldname and newname. 4364 * res: status. If success - CURRENT_FH unchanged, return change_info 4365 * for both from and target directories. 4366 */ 4367 /* ARGSUSED */ 4368 static void 4369 rfs4_op_rename(nfs_argop4 *argop, nfs_resop4 *resop, struct svc_req *req, 4370 struct compound_state *cs) 4371 { 4372 RENAME4args *args = &argop->nfs_argop4_u.oprename; 4373 RENAME4res *resp = &resop->nfs_resop4_u.oprename; 4374 int error; 4375 vnode_t *odvp; 4376 vnode_t *ndvp; 4377 vnode_t *srcvp, *targvp; 4378 struct vattr obdva, oidva, oadva; 4379 struct vattr nbdva, nidva, nadva; 4380 char *onm, *nnm; 4381 uint_t olen, nlen; 4382 rfs4_file_t *fp, *sfp; 4383 int in_crit_src, in_crit_targ; 4384 int fp_rele_grant_hold, sfp_rele_grant_hold; 4385 bslabel_t *clabel; 4386 struct sockaddr *ca; 4387 char *converted_onm = NULL; 4388 char *converted_nnm = NULL; 4389 nfsstat4 status; 4390 4391 DTRACE_NFSV4_2(op__rename__start, struct compound_state *, cs, 4392 RENAME4args *, args); 4393 4394 fp = sfp = NULL; 4395 srcvp = targvp = NULL; 4396 in_crit_src = in_crit_targ = 0; 4397 fp_rele_grant_hold = sfp_rele_grant_hold = 0; 4398 4399 /* CURRENT_FH: target directory */ 4400 ndvp = cs->vp; 4401 if (ndvp == NULL) { 4402 *cs->statusp = resp->status = NFS4ERR_NOFILEHANDLE; 4403 goto out; 4404 } 4405 4406 /* SAVED_FH: from directory */ 4407 odvp = cs->saved_vp; 4408 if (odvp == NULL) { 4409 *cs->statusp = resp->status = NFS4ERR_NOFILEHANDLE; 4410 goto out; 4411 } 4412 4413 if (cs->access == CS_ACCESS_DENIED) { 4414 *cs->statusp = resp->status = NFS4ERR_ACCESS; 4415 goto out; 4416 } 4417 4418 /* 4419 * If there is an unshared filesystem mounted on this vnode, 4420 * do not allow to rename objects in this directory. 4421 */ 4422 if (vn_ismntpt(odvp)) { 4423 *cs->statusp = resp->status = NFS4ERR_ACCESS; 4424 goto out; 4425 } 4426 4427 /* 4428 * If there is an unshared filesystem mounted on this vnode, 4429 * do not allow to rename to this directory. 4430 */ 4431 if (vn_ismntpt(ndvp)) { 4432 *cs->statusp = resp->status = NFS4ERR_ACCESS; 4433 goto out; 4434 } 4435 4436 if (odvp->v_type != VDIR || ndvp->v_type != VDIR) { 4437 *cs->statusp = resp->status = NFS4ERR_NOTDIR; 4438 goto out; 4439 } 4440 4441 if (cs->saved_exi != cs->exi) { 4442 *cs->statusp = resp->status = NFS4ERR_XDEV; 4443 goto out; 4444 } 4445 4446 status = utf8_dir_verify(&args->oldname); 4447 if (status != NFS4_OK) { 4448 *cs->statusp = resp->status = status; 4449 goto out; 4450 } 4451 4452 status = utf8_dir_verify(&args->newname); 4453 if (status != NFS4_OK) { 4454 *cs->statusp = resp->status = status; 4455 goto out; 4456 } 4457 4458 onm = utf8_to_fn(&args->oldname, &olen, NULL); 4459 if (onm == NULL) { 4460 *cs->statusp = resp->status = NFS4ERR_INVAL; 4461 goto out; 4462 } 4463 ca = (struct sockaddr *)svc_getrpccaller(req->rq_xprt)->buf; 4464 nlen = MAXPATHLEN + 1; 4465 converted_onm = nfscmd_convname(ca, cs->exi, onm, NFSCMD_CONV_INBOUND, 4466 nlen); 4467 4468 if (converted_onm == NULL) { 4469 *cs->statusp = resp->status = NFS4ERR_INVAL; 4470 kmem_free(onm, olen); 4471 goto out; 4472 } 4473 4474 nnm = utf8_to_fn(&args->newname, &nlen, NULL); 4475 if (nnm == NULL) { 4476 *cs->statusp = resp->status = NFS4ERR_INVAL; 4477 if (onm != converted_onm) 4478 kmem_free(converted_onm, MAXPATHLEN + 1); 4479 kmem_free(onm, olen); 4480 goto out; 4481 } 4482 converted_nnm = nfscmd_convname(ca, cs->exi, nnm, NFSCMD_CONV_INBOUND, 4483 MAXPATHLEN + 1); 4484 4485 if (converted_nnm == NULL) { 4486 *cs->statusp = resp->status = NFS4ERR_INVAL; 4487 kmem_free(nnm, nlen); 4488 nnm = NULL; 4489 if (onm != converted_onm) 4490 kmem_free(converted_onm, MAXPATHLEN + 1); 4491 kmem_free(onm, olen); 4492 goto out; 4493 } 4494 4495 4496 if (olen > MAXNAMELEN || nlen > MAXNAMELEN) { 4497 *cs->statusp = resp->status = NFS4ERR_NAMETOOLONG; 4498 kmem_free(onm, olen); 4499 kmem_free(nnm, nlen); 4500 goto out; 4501 } 4502 4503 4504 if (rdonly4(cs->exi, cs->vp, req)) { 4505 *cs->statusp = resp->status = NFS4ERR_ROFS; 4506 if (onm != converted_onm) 4507 kmem_free(converted_onm, MAXPATHLEN + 1); 4508 kmem_free(onm, olen); 4509 if (nnm != converted_nnm) 4510 kmem_free(converted_nnm, MAXPATHLEN + 1); 4511 kmem_free(nnm, nlen); 4512 goto out; 4513 } 4514 4515 /* check label of the target dir */ 4516 if (is_system_labeled()) { 4517 ASSERT(req->rq_label != NULL); 4518 clabel = req->rq_label; 4519 DTRACE_PROBE2(tx__rfs4__log__info__oprename__clabel, char *, 4520 "got client label from request(1)", 4521 struct svc_req *, req); 4522 if (!blequal(&l_admin_low->tsl_label, clabel)) { 4523 if (!do_rfs_label_check(clabel, ndvp, 4524 EQUALITY_CHECK, cs->exi)) { 4525 *cs->statusp = resp->status = NFS4ERR_ACCESS; 4526 goto err_out; 4527 } 4528 } 4529 } 4530 4531 /* 4532 * Is the source a file and have a delegation? 4533 * We don't need to acquire va_seq before these lookups, if 4534 * it causes an update, cinfo.before will not match, which will 4535 * trigger a cache flush even if atomic is TRUE. 4536 */ 4537 if (sfp = rfs4_lookup_and_findfile(odvp, converted_onm, &srcvp, 4538 &error, cs->cr)) { 4539 if (rfs4_check_delegated_byfp(FWRITE, sfp, TRUE, TRUE, TRUE, 4540 NULL)) { 4541 *cs->statusp = resp->status = NFS4ERR_DELAY; 4542 goto err_out; 4543 } 4544 } 4545 4546 if (srcvp == NULL) { 4547 *cs->statusp = resp->status = puterrno4(error); 4548 if (onm != converted_onm) 4549 kmem_free(converted_onm, MAXPATHLEN + 1); 4550 kmem_free(onm, olen); 4551 if (nnm != converted_nnm) 4552 kmem_free(converted_nnm, MAXPATHLEN + 1); 4553 kmem_free(nnm, nlen); 4554 goto out; 4555 } 4556 4557 sfp_rele_grant_hold = 1; 4558 4559 /* Does the destination exist and a file and have a delegation? */ 4560 if (fp = rfs4_lookup_and_findfile(ndvp, converted_nnm, &targvp, 4561 NULL, cs->cr)) { 4562 if (rfs4_check_delegated_byfp(FWRITE, fp, TRUE, TRUE, TRUE, 4563 NULL)) { 4564 *cs->statusp = resp->status = NFS4ERR_DELAY; 4565 goto err_out; 4566 } 4567 } 4568 fp_rele_grant_hold = 1; 4569 4570 4571 /* Check for NBMAND lock on both source and target */ 4572 if (nbl_need_check(srcvp)) { 4573 nbl_start_crit(srcvp, RW_READER); 4574 in_crit_src = 1; 4575 if (nbl_conflict(srcvp, NBL_RENAME, 0, 0, 0, NULL)) { 4576 *cs->statusp = resp->status = NFS4ERR_FILE_OPEN; 4577 goto err_out; 4578 } 4579 } 4580 4581 if (targvp && nbl_need_check(targvp)) { 4582 nbl_start_crit(targvp, RW_READER); 4583 in_crit_targ = 1; 4584 if (nbl_conflict(targvp, NBL_REMOVE, 0, 0, 0, NULL)) { 4585 *cs->statusp = resp->status = NFS4ERR_FILE_OPEN; 4586 goto err_out; 4587 } 4588 } 4589 4590 /* Get source "before" change value */ 4591 obdva.va_mask = AT_CTIME|AT_SEQ; 4592 error = VOP_GETATTR(odvp, &obdva, 0, cs->cr, NULL); 4593 if (!error) { 4594 nbdva.va_mask = AT_CTIME|AT_SEQ; 4595 error = VOP_GETATTR(ndvp, &nbdva, 0, cs->cr, NULL); 4596 } 4597 if (error) { 4598 *cs->statusp = resp->status = puterrno4(error); 4599 goto err_out; 4600 } 4601 4602 NFS4_SET_FATTR4_CHANGE(resp->source_cinfo.before, obdva.va_ctime) 4603 NFS4_SET_FATTR4_CHANGE(resp->target_cinfo.before, nbdva.va_ctime) 4604 4605 if ((error = VOP_RENAME(odvp, converted_onm, ndvp, converted_nnm, 4606 cs->cr, NULL, 0)) == 0 && fp != NULL) { 4607 struct vattr va; 4608 vnode_t *tvp; 4609 4610 rfs4_dbe_lock(fp->rf_dbe); 4611 tvp = fp->rf_vp; 4612 if (tvp) 4613 VN_HOLD(tvp); 4614 rfs4_dbe_unlock(fp->rf_dbe); 4615 4616 if (tvp) { 4617 va.va_mask = AT_NLINK; 4618 if (!VOP_GETATTR(tvp, &va, 0, cs->cr, NULL) && 4619 va.va_nlink == 0) { 4620 /* The file is gone and so should the state */ 4621 if (in_crit_targ) { 4622 nbl_end_crit(targvp); 4623 in_crit_targ = 0; 4624 } 4625 rfs4_close_all_state(fp); 4626 } 4627 VN_RELE(tvp); 4628 } 4629 } 4630 if (error == 0) 4631 vn_renamepath(ndvp, srcvp, nnm, nlen - 1); 4632 4633 if (in_crit_src) 4634 nbl_end_crit(srcvp); 4635 if (srcvp) 4636 VN_RELE(srcvp); 4637 if (in_crit_targ) 4638 nbl_end_crit(targvp); 4639 if (targvp) 4640 VN_RELE(targvp); 4641 4642 if (sfp) { 4643 rfs4_clear_dont_grant(sfp); 4644 rfs4_file_rele(sfp); 4645 } 4646 if (fp) { 4647 rfs4_clear_dont_grant(fp); 4648 rfs4_file_rele(fp); 4649 } 4650 4651 if (converted_onm != onm) 4652 kmem_free(converted_onm, MAXPATHLEN + 1); 4653 kmem_free(onm, olen); 4654 if (converted_nnm != nnm) 4655 kmem_free(converted_nnm, MAXPATHLEN + 1); 4656 kmem_free(nnm, nlen); 4657 4658 /* 4659 * Get the initial "after" sequence number, if it fails, set to zero 4660 */ 4661 oidva.va_mask = AT_SEQ; 4662 if (VOP_GETATTR(odvp, &oidva, 0, cs->cr, NULL)) 4663 oidva.va_seq = 0; 4664 4665 nidva.va_mask = AT_SEQ; 4666 if (VOP_GETATTR(ndvp, &nidva, 0, cs->cr, NULL)) 4667 nidva.va_seq = 0; 4668 4669 /* 4670 * Force modified data and metadata out to stable storage. 4671 */ 4672 (void) VOP_FSYNC(odvp, 0, cs->cr, NULL); 4673 (void) VOP_FSYNC(ndvp, 0, cs->cr, NULL); 4674 4675 if (error) { 4676 *cs->statusp = resp->status = puterrno4(error); 4677 goto out; 4678 } 4679 4680 /* 4681 * Get "after" change values, if it fails, simply return the 4682 * before value. 4683 */ 4684 oadva.va_mask = AT_CTIME|AT_SEQ; 4685 if (VOP_GETATTR(odvp, &oadva, 0, cs->cr, NULL)) { 4686 oadva.va_ctime = obdva.va_ctime; 4687 oadva.va_seq = 0; 4688 } 4689 4690 nadva.va_mask = AT_CTIME|AT_SEQ; 4691 if (VOP_GETATTR(odvp, &nadva, 0, cs->cr, NULL)) { 4692 nadva.va_ctime = nbdva.va_ctime; 4693 nadva.va_seq = 0; 4694 } 4695 4696 NFS4_SET_FATTR4_CHANGE(resp->source_cinfo.after, oadva.va_ctime) 4697 NFS4_SET_FATTR4_CHANGE(resp->target_cinfo.after, nadva.va_ctime) 4698 4699 /* 4700 * The cinfo.atomic = TRUE only if we have 4701 * non-zero va_seq's, and it has incremented by exactly one 4702 * during the VOP_RENAME and it didn't change during the VOP_FSYNC. 4703 */ 4704 if (obdva.va_seq && oidva.va_seq && oadva.va_seq && 4705 oidva.va_seq == (obdva.va_seq + 1) && oidva.va_seq == oadva.va_seq) 4706 resp->source_cinfo.atomic = TRUE; 4707 else 4708 resp->source_cinfo.atomic = FALSE; 4709 4710 if (nbdva.va_seq && nidva.va_seq && nadva.va_seq && 4711 nidva.va_seq == (nbdva.va_seq + 1) && nidva.va_seq == nadva.va_seq) 4712 resp->target_cinfo.atomic = TRUE; 4713 else 4714 resp->target_cinfo.atomic = FALSE; 4715 4716 #ifdef VOLATILE_FH_TEST 4717 { 4718 extern void add_volrnm_fh(struct exportinfo *, vnode_t *); 4719 4720 /* 4721 * Add the renamed file handle to the volatile rename list 4722 */ 4723 if (cs->exi->exi_export.ex_flags & EX_VOLRNM) { 4724 /* file handles may expire on rename */ 4725 vnode_t *vp; 4726 4727 nnm = utf8_to_fn(&args->newname, &nlen, NULL); 4728 /* 4729 * Already know that nnm will be a valid string 4730 */ 4731 error = VOP_LOOKUP(ndvp, nnm, &vp, NULL, 0, NULL, cs->cr, 4732 NULL, NULL, NULL); 4733 kmem_free(nnm, nlen); 4734 if (!error) { 4735 add_volrnm_fh(cs->exi, vp); 4736 VN_RELE(vp); 4737 } 4738 } 4739 } 4740 #endif /* VOLATILE_FH_TEST */ 4741 4742 *cs->statusp = resp->status = NFS4_OK; 4743 out: 4744 DTRACE_NFSV4_2(op__rename__done, struct compound_state *, cs, 4745 RENAME4res *, resp); 4746 return; 4747 4748 err_out: 4749 if (onm != converted_onm) 4750 kmem_free(converted_onm, MAXPATHLEN + 1); 4751 if (onm != NULL) 4752 kmem_free(onm, olen); 4753 if (nnm != converted_nnm) 4754 kmem_free(converted_nnm, MAXPATHLEN + 1); 4755 if (nnm != NULL) 4756 kmem_free(nnm, nlen); 4757 4758 if (in_crit_src) nbl_end_crit(srcvp); 4759 if (in_crit_targ) nbl_end_crit(targvp); 4760 if (targvp) VN_RELE(targvp); 4761 if (srcvp) VN_RELE(srcvp); 4762 if (sfp) { 4763 if (sfp_rele_grant_hold) rfs4_clear_dont_grant(sfp); 4764 rfs4_file_rele(sfp); 4765 } 4766 if (fp) { 4767 if (fp_rele_grant_hold) rfs4_clear_dont_grant(fp); 4768 rfs4_file_rele(fp); 4769 } 4770 4771 DTRACE_NFSV4_2(op__rename__done, struct compound_state *, cs, 4772 RENAME4res *, resp); 4773 } 4774 4775 /* ARGSUSED */ 4776 static void 4777 rfs4_op_renew(nfs_argop4 *argop, nfs_resop4 *resop, struct svc_req *req, 4778 struct compound_state *cs) 4779 { 4780 RENEW4args *args = &argop->nfs_argop4_u.oprenew; 4781 RENEW4res *resp = &resop->nfs_resop4_u.oprenew; 4782 rfs4_client_t *cp; 4783 4784 DTRACE_NFSV4_2(op__renew__start, struct compound_state *, cs, 4785 RENEW4args *, args); 4786 4787 if ((cp = rfs4_findclient_by_id(args->clientid, FALSE)) == NULL) { 4788 *cs->statusp = resp->status = 4789 rfs4_check_clientid(&args->clientid, 0); 4790 goto out; 4791 } 4792 4793 if (rfs4_lease_expired(cp)) { 4794 rfs4_client_rele(cp); 4795 *cs->statusp = resp->status = NFS4ERR_EXPIRED; 4796 goto out; 4797 } 4798 4799 rfs4_update_lease(cp); 4800 4801 mutex_enter(cp->rc_cbinfo.cb_lock); 4802 if (cp->rc_cbinfo.cb_notified_of_cb_path_down == FALSE) { 4803 cp->rc_cbinfo.cb_notified_of_cb_path_down = TRUE; 4804 *cs->statusp = resp->status = NFS4ERR_CB_PATH_DOWN; 4805 } else { 4806 *cs->statusp = resp->status = NFS4_OK; 4807 } 4808 mutex_exit(cp->rc_cbinfo.cb_lock); 4809 4810 rfs4_client_rele(cp); 4811 4812 out: 4813 DTRACE_NFSV4_2(op__renew__done, struct compound_state *, cs, 4814 RENEW4res *, resp); 4815 } 4816 4817 /* ARGSUSED */ 4818 static void 4819 rfs4_op_restorefh(nfs_argop4 *args, nfs_resop4 *resop, struct svc_req *req, 4820 struct compound_state *cs) 4821 { 4822 RESTOREFH4res *resp = &resop->nfs_resop4_u.oprestorefh; 4823 4824 DTRACE_NFSV4_1(op__restorefh__start, struct compound_state *, cs); 4825 4826 /* No need to check cs->access - we are not accessing any object */ 4827 if ((cs->saved_vp == NULL) || (cs->saved_fh.nfs_fh4_val == NULL)) { 4828 *cs->statusp = resp->status = NFS4ERR_RESTOREFH; 4829 goto out; 4830 } 4831 if (cs->vp != NULL) { 4832 VN_RELE(cs->vp); 4833 } 4834 cs->vp = cs->saved_vp; 4835 cs->saved_vp = NULL; 4836 cs->exi = cs->saved_exi; 4837 nfs_fh4_copy(&cs->saved_fh, &cs->fh); 4838 *cs->statusp = resp->status = NFS4_OK; 4839 cs->deleg = FALSE; 4840 4841 out: 4842 DTRACE_NFSV4_2(op__restorefh__done, struct compound_state *, cs, 4843 RESTOREFH4res *, resp); 4844 } 4845 4846 /* ARGSUSED */ 4847 static void 4848 rfs4_op_savefh(nfs_argop4 *argop, nfs_resop4 *resop, struct svc_req *req, 4849 struct compound_state *cs) 4850 { 4851 SAVEFH4res *resp = &resop->nfs_resop4_u.opsavefh; 4852 4853 DTRACE_NFSV4_1(op__savefh__start, struct compound_state *, cs); 4854 4855 /* No need to check cs->access - we are not accessing any object */ 4856 if (cs->vp == NULL) { 4857 *cs->statusp = resp->status = NFS4ERR_NOFILEHANDLE; 4858 goto out; 4859 } 4860 if (cs->saved_vp != NULL) { 4861 VN_RELE(cs->saved_vp); 4862 } 4863 cs->saved_vp = cs->vp; 4864 VN_HOLD(cs->saved_vp); 4865 cs->saved_exi = cs->exi; 4866 /* 4867 * since SAVEFH is fairly rare, don't alloc space for its fh 4868 * unless necessary. 4869 */ 4870 if (cs->saved_fh.nfs_fh4_val == NULL) { 4871 cs->saved_fh.nfs_fh4_val = kmem_alloc(NFS4_FHSIZE, KM_SLEEP); 4872 } 4873 nfs_fh4_copy(&cs->fh, &cs->saved_fh); 4874 *cs->statusp = resp->status = NFS4_OK; 4875 4876 out: 4877 DTRACE_NFSV4_2(op__savefh__done, struct compound_state *, cs, 4878 SAVEFH4res *, resp); 4879 } 4880 4881 /* 4882 * rfs4_verify_attr is called when nfsv4 Setattr failed, but we wish to 4883 * return the bitmap of attrs that were set successfully. It is also 4884 * called by Verify/Nverify to test the vattr/vfsstat attrs. It should 4885 * always be called only after rfs4_do_set_attrs(). 4886 * 4887 * Verify that the attributes are same as the expected ones. sargp->vap 4888 * and sargp->sbp contain the input attributes as translated from fattr4. 4889 * 4890 * This function verifies only the attrs that correspond to a vattr or 4891 * vfsstat struct. That is because of the extra step needed to get the 4892 * corresponding system structs. Other attributes have already been set or 4893 * verified by do_rfs4_set_attrs. 4894 * 4895 * Return 0 if all attrs match, -1 if some don't, error if error processing. 4896 */ 4897 static int 4898 rfs4_verify_attr(struct nfs4_svgetit_arg *sargp, 4899 bitmap4 *resp, struct nfs4_ntov_table *ntovp) 4900 { 4901 int error, ret_error = 0; 4902 int i, k; 4903 uint_t sva_mask = sargp->vap->va_mask; 4904 uint_t vbit; 4905 union nfs4_attr_u *na; 4906 uint8_t *amap; 4907 bool_t getsb = ntovp->vfsstat; 4908 4909 if (sva_mask != 0) { 4910 /* 4911 * Okay to overwrite sargp->vap because we verify based 4912 * on the incoming values. 4913 */ 4914 ret_error = VOP_GETATTR(sargp->cs->vp, sargp->vap, 0, 4915 sargp->cs->cr, NULL); 4916 if (ret_error) { 4917 if (resp == NULL) 4918 return (ret_error); 4919 /* 4920 * Must return bitmap of successful attrs 4921 */ 4922 sva_mask = 0; /* to prevent checking vap later */ 4923 } else { 4924 /* 4925 * Some file systems clobber va_mask. it is probably 4926 * wrong of them to do so, nonethless we practice 4927 * defensive coding. 4928 * See bug id 4276830. 4929 */ 4930 sargp->vap->va_mask = sva_mask; 4931 } 4932 } 4933 4934 if (getsb) { 4935 /* 4936 * Now get the superblock and loop on the bitmap, as there is 4937 * no simple way of translating from superblock to bitmap4. 4938 */ 4939 ret_error = VFS_STATVFS(sargp->cs->vp->v_vfsp, sargp->sbp); 4940 if (ret_error) { 4941 if (resp == NULL) 4942 goto errout; 4943 getsb = FALSE; 4944 } 4945 } 4946 4947 /* 4948 * Now loop and verify each attribute which getattr returned 4949 * whether it's the same as the input. 4950 */ 4951 if (resp == NULL && !getsb && (sva_mask == 0)) 4952 goto errout; 4953 4954 na = ntovp->na; 4955 amap = ntovp->amap; 4956 k = 0; 4957 for (i = 0; i < ntovp->attrcnt; i++, na++, amap++) { 4958 k = *amap; 4959 ASSERT(nfs4_ntov_map[k].nval == k); 4960 vbit = nfs4_ntov_map[k].vbit; 4961 4962 /* 4963 * If vattr attribute but VOP_GETATTR failed, or it's 4964 * superblock attribute but VFS_STATVFS failed, skip 4965 */ 4966 if (vbit) { 4967 if ((vbit & sva_mask) == 0) 4968 continue; 4969 } else if (!(getsb && nfs4_ntov_map[k].vfsstat)) { 4970 continue; 4971 } 4972 error = (*nfs4_ntov_map[k].sv_getit)(NFS4ATTR_VERIT, sargp, na); 4973 if (resp != NULL) { 4974 if (error) 4975 ret_error = -1; /* not all match */ 4976 else /* update response bitmap */ 4977 *resp |= nfs4_ntov_map[k].fbit; 4978 continue; 4979 } 4980 if (error) { 4981 ret_error = -1; /* not all match */ 4982 break; 4983 } 4984 } 4985 errout: 4986 return (ret_error); 4987 } 4988 4989 /* 4990 * Decode the attribute to be set/verified. If the attr requires a sys op 4991 * (VOP_GETATTR, VFS_VFSSTAT), and the request is to verify, then don't 4992 * call the sv_getit function for it, because the sys op hasn't yet been done. 4993 * Return 0 for success, error code if failed. 4994 * 4995 * Note: the decoded arg is not freed here but in nfs4_ntov_table_free. 4996 */ 4997 static int 4998 decode_fattr4_attr(nfs4_attr_cmd_t cmd, struct nfs4_svgetit_arg *sargp, 4999 int k, XDR *xdrp, bitmap4 *resp_bval, union nfs4_attr_u *nap) 5000 { 5001 int error = 0; 5002 bool_t set_later; 5003 5004 sargp->vap->va_mask |= nfs4_ntov_map[k].vbit; 5005 5006 if ((*nfs4_ntov_map[k].xfunc)(xdrp, nap)) { 5007 set_later = nfs4_ntov_map[k].vbit || nfs4_ntov_map[k].vfsstat; 5008 /* 5009 * don't verify yet if a vattr or sb dependent attr, 5010 * because we don't have their sys values yet. 5011 * Will be done later. 5012 */ 5013 if (! (set_later && (cmd == NFS4ATTR_VERIT))) { 5014 /* 5015 * ACLs are a special case, since setting the MODE 5016 * conflicts with setting the ACL. We delay setting 5017 * the ACL until all other attributes have been set. 5018 * The ACL gets set in do_rfs4_op_setattr(). 5019 */ 5020 if (nfs4_ntov_map[k].fbit != FATTR4_ACL_MASK) { 5021 error = (*nfs4_ntov_map[k].sv_getit)(cmd, 5022 sargp, nap); 5023 if (error) { 5024 xdr_free(nfs4_ntov_map[k].xfunc, 5025 (caddr_t)nap); 5026 } 5027 } 5028 } 5029 } else { 5030 #ifdef DEBUG 5031 cmn_err(CE_NOTE, "decode_fattr4_attr: error " 5032 "decoding attribute %d\n", k); 5033 #endif 5034 error = EINVAL; 5035 } 5036 if (!error && resp_bval && !set_later) { 5037 *resp_bval |= nfs4_ntov_map[k].fbit; 5038 } 5039 5040 return (error); 5041 } 5042 5043 /* 5044 * Set vattr based on incoming fattr4 attrs - used by setattr. 5045 * Set response mask. Ignore any values that are not writable vattr attrs. 5046 */ 5047 static nfsstat4 5048 do_rfs4_set_attrs(bitmap4 *resp, fattr4 *fattrp, struct compound_state *cs, 5049 struct nfs4_svgetit_arg *sargp, struct nfs4_ntov_table *ntovp, 5050 nfs4_attr_cmd_t cmd) 5051 { 5052 int error = 0; 5053 int i; 5054 char *attrs = fattrp->attrlist4; 5055 uint32_t attrslen = fattrp->attrlist4_len; 5056 XDR xdr; 5057 nfsstat4 status = NFS4_OK; 5058 vnode_t *vp = cs->vp; 5059 union nfs4_attr_u *na; 5060 uint8_t *amap; 5061 5062 #ifndef lint 5063 /* 5064 * Make sure that maximum attribute number can be expressed as an 5065 * 8 bit quantity. 5066 */ 5067 ASSERT(NFS4_MAXNUM_ATTRS <= (UINT8_MAX + 1)); 5068 #endif 5069 5070 if (vp == NULL) { 5071 if (resp) 5072 *resp = 0; 5073 return (NFS4ERR_NOFILEHANDLE); 5074 } 5075 if (cs->access == CS_ACCESS_DENIED) { 5076 if (resp) 5077 *resp = 0; 5078 return (NFS4ERR_ACCESS); 5079 } 5080 5081 sargp->op = cmd; 5082 sargp->cs = cs; 5083 sargp->flag = 0; /* may be set later */ 5084 sargp->vap->va_mask = 0; 5085 sargp->rdattr_error = NFS4_OK; 5086 sargp->rdattr_error_req = FALSE; 5087 /* sargp->sbp is set by the caller */ 5088 5089 xdrmem_create(&xdr, attrs, attrslen, XDR_DECODE); 5090 5091 na = ntovp->na; 5092 amap = ntovp->amap; 5093 5094 /* 5095 * The following loop iterates on the nfs4_ntov_map checking 5096 * if the fbit is set in the requested bitmap. 5097 * If set then we process the arguments using the 5098 * rfs4_fattr4 conversion functions to populate the setattr 5099 * vattr and va_mask. Any settable attrs that are not using vattr 5100 * will be set in this loop. 5101 */ 5102 for (i = 0; i < nfs4_ntov_map_size; i++) { 5103 if (!(fattrp->attrmask & nfs4_ntov_map[i].fbit)) { 5104 continue; 5105 } 5106 /* 5107 * If setattr, must be a writable attr. 5108 * If verify/nverify, must be a readable attr. 5109 */ 5110 if ((error = (*nfs4_ntov_map[i].sv_getit)( 5111 NFS4ATTR_SUPPORTED, sargp, NULL)) != 0) { 5112 /* 5113 * Client tries to set/verify an 5114 * unsupported attribute, tries to set 5115 * a read only attr or verify a write 5116 * only one - error! 5117 */ 5118 break; 5119 } 5120 /* 5121 * Decode the attribute to set/verify 5122 */ 5123 error = decode_fattr4_attr(cmd, sargp, nfs4_ntov_map[i].nval, 5124 &xdr, resp ? resp : NULL, na); 5125 if (error) 5126 break; 5127 *amap++ = (uint8_t)nfs4_ntov_map[i].nval; 5128 na++; 5129 (ntovp->attrcnt)++; 5130 if (nfs4_ntov_map[i].vfsstat) 5131 ntovp->vfsstat = TRUE; 5132 } 5133 5134 if (error != 0) 5135 status = (error == ENOTSUP ? NFS4ERR_ATTRNOTSUPP : 5136 puterrno4(error)); 5137 /* xdrmem_destroy(&xdrs); */ /* NO-OP */ 5138 return (status); 5139 } 5140 5141 static nfsstat4 5142 do_rfs4_op_setattr(bitmap4 *resp, fattr4 *fattrp, struct compound_state *cs, 5143 stateid4 *stateid) 5144 { 5145 int error = 0; 5146 struct nfs4_svgetit_arg sarg; 5147 bool_t trunc; 5148 5149 nfsstat4 status = NFS4_OK; 5150 cred_t *cr = cs->cr; 5151 vnode_t *vp = cs->vp; 5152 struct nfs4_ntov_table ntov; 5153 struct statvfs64 sb; 5154 struct vattr bva; 5155 struct flock64 bf; 5156 int in_crit = 0; 5157 uint_t saved_mask = 0; 5158 caller_context_t ct; 5159 5160 *resp = 0; 5161 sarg.sbp = &sb; 5162 sarg.is_referral = B_FALSE; 5163 nfs4_ntov_table_init(&ntov); 5164 status = do_rfs4_set_attrs(resp, fattrp, cs, &sarg, &ntov, 5165 NFS4ATTR_SETIT); 5166 if (status != NFS4_OK) { 5167 /* 5168 * failed set attrs 5169 */ 5170 goto done; 5171 } 5172 if ((sarg.vap->va_mask == 0) && 5173 (! (fattrp->attrmask & FATTR4_ACL_MASK))) { 5174 /* 5175 * no further work to be done 5176 */ 5177 goto done; 5178 } 5179 5180 /* 5181 * If we got a request to set the ACL and the MODE, only 5182 * allow changing VSUID, VSGID, and VSVTX. Attempting 5183 * to change any other bits, along with setting an ACL, 5184 * gives NFS4ERR_INVAL. 5185 */ 5186 if ((fattrp->attrmask & FATTR4_ACL_MASK) && 5187 (fattrp->attrmask & FATTR4_MODE_MASK)) { 5188 vattr_t va; 5189 5190 va.va_mask = AT_MODE; 5191 error = VOP_GETATTR(vp, &va, 0, cs->cr, NULL); 5192 if (error) { 5193 status = puterrno4(error); 5194 goto done; 5195 } 5196 if ((sarg.vap->va_mode ^ va.va_mode) & 5197 ~(VSUID | VSGID | VSVTX)) { 5198 status = NFS4ERR_INVAL; 5199 goto done; 5200 } 5201 } 5202 5203 /* Check stateid only if size has been set */ 5204 if (sarg.vap->va_mask & AT_SIZE) { 5205 trunc = (sarg.vap->va_size == 0); 5206 status = rfs4_check_stateid(FWRITE, cs->vp, stateid, 5207 trunc, &cs->deleg, sarg.vap->va_mask & AT_SIZE, &ct); 5208 if (status != NFS4_OK) 5209 goto done; 5210 } else { 5211 ct.cc_sysid = 0; 5212 ct.cc_pid = 0; 5213 ct.cc_caller_id = nfs4_srv_caller_id; 5214 ct.cc_flags = CC_DONTBLOCK; 5215 } 5216 5217 /* XXX start of possible race with delegations */ 5218 5219 /* 5220 * We need to specially handle size changes because it is 5221 * possible for the client to create a file with read-only 5222 * modes, but with the file opened for writing. If the client 5223 * then tries to set the file size, e.g. ftruncate(3C), 5224 * fcntl(F_FREESP), the normal access checking done in 5225 * VOP_SETATTR would prevent the client from doing it even though 5226 * it should be allowed to do so. To get around this, we do the 5227 * access checking for ourselves and use VOP_SPACE which doesn't 5228 * do the access checking. 5229 * Also the client should not be allowed to change the file 5230 * size if there is a conflicting non-blocking mandatory lock in 5231 * the region of the change. 5232 */ 5233 if (vp->v_type == VREG && (sarg.vap->va_mask & AT_SIZE)) { 5234 u_offset_t offset; 5235 ssize_t length; 5236 5237 /* 5238 * ufs_setattr clears AT_SIZE from vap->va_mask, but 5239 * before returning, sarg.vap->va_mask is used to 5240 * generate the setattr reply bitmap. We also clear 5241 * AT_SIZE below before calling VOP_SPACE. For both 5242 * of these cases, the va_mask needs to be saved here 5243 * and restored after calling VOP_SETATTR. 5244 */ 5245 saved_mask = sarg.vap->va_mask; 5246 5247 /* 5248 * Check any possible conflict due to NBMAND locks. 5249 * Get into critical region before VOP_GETATTR, so the 5250 * size attribute is valid when checking conflicts. 5251 */ 5252 if (nbl_need_check(vp)) { 5253 nbl_start_crit(vp, RW_READER); 5254 in_crit = 1; 5255 } 5256 5257 bva.va_mask = AT_UID|AT_SIZE; 5258 if (error = VOP_GETATTR(vp, &bva, 0, cr, &ct)) { 5259 status = puterrno4(error); 5260 goto done; 5261 } 5262 5263 if (in_crit) { 5264 if (sarg.vap->va_size < bva.va_size) { 5265 offset = sarg.vap->va_size; 5266 length = bva.va_size - sarg.vap->va_size; 5267 } else { 5268 offset = bva.va_size; 5269 length = sarg.vap->va_size - bva.va_size; 5270 } 5271 if (nbl_conflict(vp, NBL_WRITE, offset, length, 0, 5272 &ct)) { 5273 status = NFS4ERR_LOCKED; 5274 goto done; 5275 } 5276 } 5277 5278 if (crgetuid(cr) == bva.va_uid) { 5279 sarg.vap->va_mask &= ~AT_SIZE; 5280 bf.l_type = F_WRLCK; 5281 bf.l_whence = 0; 5282 bf.l_start = (off64_t)sarg.vap->va_size; 5283 bf.l_len = 0; 5284 bf.l_sysid = 0; 5285 bf.l_pid = 0; 5286 error = VOP_SPACE(vp, F_FREESP, &bf, FWRITE, 5287 (offset_t)sarg.vap->va_size, cr, &ct); 5288 } 5289 } 5290 5291 if (!error && sarg.vap->va_mask != 0) 5292 error = VOP_SETATTR(vp, sarg.vap, sarg.flag, cr, &ct); 5293 5294 /* restore va_mask -- ufs_setattr clears AT_SIZE */ 5295 if (saved_mask & AT_SIZE) 5296 sarg.vap->va_mask |= AT_SIZE; 5297 5298 /* 5299 * If an ACL was being set, it has been delayed until now, 5300 * in order to set the mode (via the VOP_SETATTR() above) first. 5301 */ 5302 if ((! error) && (fattrp->attrmask & FATTR4_ACL_MASK)) { 5303 int i; 5304 5305 for (i = 0; i < NFS4_MAXNUM_ATTRS; i++) 5306 if (ntov.amap[i] == FATTR4_ACL) 5307 break; 5308 if (i < NFS4_MAXNUM_ATTRS) { 5309 error = (*nfs4_ntov_map[FATTR4_ACL].sv_getit)( 5310 NFS4ATTR_SETIT, &sarg, &ntov.na[i]); 5311 if (error == 0) { 5312 *resp |= FATTR4_ACL_MASK; 5313 } else if (error == ENOTSUP) { 5314 (void) rfs4_verify_attr(&sarg, resp, &ntov); 5315 status = NFS4ERR_ATTRNOTSUPP; 5316 goto done; 5317 } 5318 } else { 5319 NFS4_DEBUG(rfs4_debug, 5320 (CE_NOTE, "do_rfs4_op_setattr: " 5321 "unable to find ACL in fattr4")); 5322 error = EINVAL; 5323 } 5324 } 5325 5326 if (error) { 5327 /* check if a monitor detected a delegation conflict */ 5328 if (error == EAGAIN && (ct.cc_flags & CC_WOULDBLOCK)) 5329 status = NFS4ERR_DELAY; 5330 else 5331 status = puterrno4(error); 5332 5333 /* 5334 * Set the response bitmap when setattr failed. 5335 * If VOP_SETATTR partially succeeded, test by doing a 5336 * VOP_GETATTR on the object and comparing the data 5337 * to the setattr arguments. 5338 */ 5339 (void) rfs4_verify_attr(&sarg, resp, &ntov); 5340 } else { 5341 /* 5342 * Force modified metadata out to stable storage. 5343 */ 5344 (void) VOP_FSYNC(vp, FNODSYNC, cr, &ct); 5345 /* 5346 * Set response bitmap 5347 */ 5348 nfs4_vmask_to_nmask_set(sarg.vap->va_mask, resp); 5349 } 5350 5351 /* Return early and already have a NFSv4 error */ 5352 done: 5353 /* 5354 * Except for nfs4_vmask_to_nmask_set(), vattr --> fattr 5355 * conversion sets both readable and writeable NFS4 attrs 5356 * for AT_MTIME and AT_ATIME. The line below masks out 5357 * unrequested attrs from the setattr result bitmap. This 5358 * is placed after the done: label to catch the ATTRNOTSUP 5359 * case. 5360 */ 5361 *resp &= fattrp->attrmask; 5362 5363 if (in_crit) 5364 nbl_end_crit(vp); 5365 5366 nfs4_ntov_table_free(&ntov, &sarg); 5367 5368 return (status); 5369 } 5370 5371 /* ARGSUSED */ 5372 static void 5373 rfs4_op_setattr(nfs_argop4 *argop, nfs_resop4 *resop, struct svc_req *req, 5374 struct compound_state *cs) 5375 { 5376 SETATTR4args *args = &argop->nfs_argop4_u.opsetattr; 5377 SETATTR4res *resp = &resop->nfs_resop4_u.opsetattr; 5378 bslabel_t *clabel; 5379 5380 DTRACE_NFSV4_2(op__setattr__start, struct compound_state *, cs, 5381 SETATTR4args *, args); 5382 5383 if (cs->vp == NULL) { 5384 *cs->statusp = resp->status = NFS4ERR_NOFILEHANDLE; 5385 goto out; 5386 } 5387 5388 /* 5389 * If there is an unshared filesystem mounted on this vnode, 5390 * do not allow to setattr on this vnode. 5391 */ 5392 if (vn_ismntpt(cs->vp)) { 5393 *cs->statusp = resp->status = NFS4ERR_ACCESS; 5394 goto out; 5395 } 5396 5397 resp->attrsset = 0; 5398 5399 if (rdonly4(cs->exi, cs->vp, req)) { 5400 *cs->statusp = resp->status = NFS4ERR_ROFS; 5401 goto out; 5402 } 5403 5404 /* check label before setting attributes */ 5405 if (is_system_labeled()) { 5406 ASSERT(req->rq_label != NULL); 5407 clabel = req->rq_label; 5408 DTRACE_PROBE2(tx__rfs4__log__info__opsetattr__clabel, char *, 5409 "got client label from request(1)", 5410 struct svc_req *, req); 5411 if (!blequal(&l_admin_low->tsl_label, clabel)) { 5412 if (!do_rfs_label_check(clabel, cs->vp, 5413 EQUALITY_CHECK, cs->exi)) { 5414 *cs->statusp = resp->status = NFS4ERR_ACCESS; 5415 goto out; 5416 } 5417 } 5418 } 5419 5420 *cs->statusp = resp->status = 5421 do_rfs4_op_setattr(&resp->attrsset, &args->obj_attributes, cs, 5422 &args->stateid); 5423 5424 out: 5425 DTRACE_NFSV4_2(op__setattr__done, struct compound_state *, cs, 5426 SETATTR4res *, resp); 5427 } 5428 5429 /* ARGSUSED */ 5430 static void 5431 rfs4_op_verify(nfs_argop4 *argop, nfs_resop4 *resop, struct svc_req *req, 5432 struct compound_state *cs) 5433 { 5434 /* 5435 * verify and nverify are exactly the same, except that nverify 5436 * succeeds when some argument changed, and verify succeeds when 5437 * when none changed. 5438 */ 5439 5440 VERIFY4args *args = &argop->nfs_argop4_u.opverify; 5441 VERIFY4res *resp = &resop->nfs_resop4_u.opverify; 5442 5443 int error; 5444 struct nfs4_svgetit_arg sarg; 5445 struct statvfs64 sb; 5446 struct nfs4_ntov_table ntov; 5447 5448 DTRACE_NFSV4_2(op__verify__start, struct compound_state *, cs, 5449 VERIFY4args *, args); 5450 5451 if (cs->vp == NULL) { 5452 *cs->statusp = resp->status = NFS4ERR_NOFILEHANDLE; 5453 goto out; 5454 } 5455 5456 sarg.sbp = &sb; 5457 sarg.is_referral = B_FALSE; 5458 nfs4_ntov_table_init(&ntov); 5459 resp->status = do_rfs4_set_attrs(NULL, &args->obj_attributes, cs, 5460 &sarg, &ntov, NFS4ATTR_VERIT); 5461 if (resp->status != NFS4_OK) { 5462 /* 5463 * do_rfs4_set_attrs will try to verify systemwide attrs, 5464 * so could return -1 for "no match". 5465 */ 5466 if (resp->status == -1) 5467 resp->status = NFS4ERR_NOT_SAME; 5468 goto done; 5469 } 5470 error = rfs4_verify_attr(&sarg, NULL, &ntov); 5471 switch (error) { 5472 case 0: 5473 resp->status = NFS4_OK; 5474 break; 5475 case -1: 5476 resp->status = NFS4ERR_NOT_SAME; 5477 break; 5478 default: 5479 resp->status = puterrno4(error); 5480 break; 5481 } 5482 done: 5483 *cs->statusp = resp->status; 5484 nfs4_ntov_table_free(&ntov, &sarg); 5485 out: 5486 DTRACE_NFSV4_2(op__verify__done, struct compound_state *, cs, 5487 VERIFY4res *, resp); 5488 } 5489 5490 /* ARGSUSED */ 5491 static void 5492 rfs4_op_nverify(nfs_argop4 *argop, nfs_resop4 *resop, struct svc_req *req, 5493 struct compound_state *cs) 5494 { 5495 /* 5496 * verify and nverify are exactly the same, except that nverify 5497 * succeeds when some argument changed, and verify succeeds when 5498 * when none changed. 5499 */ 5500 5501 NVERIFY4args *args = &argop->nfs_argop4_u.opnverify; 5502 NVERIFY4res *resp = &resop->nfs_resop4_u.opnverify; 5503 5504 int error; 5505 struct nfs4_svgetit_arg sarg; 5506 struct statvfs64 sb; 5507 struct nfs4_ntov_table ntov; 5508 5509 DTRACE_NFSV4_2(op__nverify__start, struct compound_state *, cs, 5510 NVERIFY4args *, args); 5511 5512 if (cs->vp == NULL) { 5513 *cs->statusp = resp->status = NFS4ERR_NOFILEHANDLE; 5514 DTRACE_NFSV4_2(op__nverify__done, struct compound_state *, cs, 5515 NVERIFY4res *, resp); 5516 return; 5517 } 5518 sarg.sbp = &sb; 5519 sarg.is_referral = B_FALSE; 5520 nfs4_ntov_table_init(&ntov); 5521 resp->status = do_rfs4_set_attrs(NULL, &args->obj_attributes, cs, 5522 &sarg, &ntov, NFS4ATTR_VERIT); 5523 if (resp->status != NFS4_OK) { 5524 /* 5525 * do_rfs4_set_attrs will try to verify systemwide attrs, 5526 * so could return -1 for "no match". 5527 */ 5528 if (resp->status == -1) 5529 resp->status = NFS4_OK; 5530 goto done; 5531 } 5532 error = rfs4_verify_attr(&sarg, NULL, &ntov); 5533 switch (error) { 5534 case 0: 5535 resp->status = NFS4ERR_SAME; 5536 break; 5537 case -1: 5538 resp->status = NFS4_OK; 5539 break; 5540 default: 5541 resp->status = puterrno4(error); 5542 break; 5543 } 5544 done: 5545 *cs->statusp = resp->status; 5546 nfs4_ntov_table_free(&ntov, &sarg); 5547 5548 DTRACE_NFSV4_2(op__nverify__done, struct compound_state *, cs, 5549 NVERIFY4res *, resp); 5550 } 5551 5552 /* 5553 * XXX - This should live in an NFS header file. 5554 */ 5555 #define MAX_IOVECS 12 5556 5557 /* ARGSUSED */ 5558 static void 5559 rfs4_op_write(nfs_argop4 *argop, nfs_resop4 *resop, struct svc_req *req, 5560 struct compound_state *cs) 5561 { 5562 WRITE4args *args = &argop->nfs_argop4_u.opwrite; 5563 WRITE4res *resp = &resop->nfs_resop4_u.opwrite; 5564 int error; 5565 vnode_t *vp; 5566 struct vattr bva; 5567 u_offset_t rlimit; 5568 struct uio uio; 5569 struct iovec iov[MAX_IOVECS]; 5570 struct iovec *iovp; 5571 int iovcnt; 5572 int ioflag; 5573 cred_t *savecred, *cr; 5574 bool_t *deleg = &cs->deleg; 5575 nfsstat4 stat; 5576 int in_crit = 0; 5577 caller_context_t ct; 5578 5579 DTRACE_NFSV4_2(op__write__start, struct compound_state *, cs, 5580 WRITE4args *, args); 5581 5582 vp = cs->vp; 5583 if (vp == NULL) { 5584 *cs->statusp = resp->status = NFS4ERR_NOFILEHANDLE; 5585 goto out; 5586 } 5587 if (cs->access == CS_ACCESS_DENIED) { 5588 *cs->statusp = resp->status = NFS4ERR_ACCESS; 5589 goto out; 5590 } 5591 5592 cr = cs->cr; 5593 5594 if ((stat = rfs4_check_stateid(FWRITE, vp, &args->stateid, FALSE, 5595 deleg, TRUE, &ct)) != NFS4_OK) { 5596 *cs->statusp = resp->status = stat; 5597 goto out; 5598 } 5599 5600 /* 5601 * We have to enter the critical region before calling VOP_RWLOCK 5602 * to avoid a deadlock with ufs. 5603 */ 5604 if (nbl_need_check(vp)) { 5605 nbl_start_crit(vp, RW_READER); 5606 in_crit = 1; 5607 if (nbl_conflict(vp, NBL_WRITE, 5608 args->offset, args->data_len, 0, &ct)) { 5609 *cs->statusp = resp->status = NFS4ERR_LOCKED; 5610 goto out; 5611 } 5612 } 5613 5614 bva.va_mask = AT_MODE | AT_UID; 5615 error = VOP_GETATTR(vp, &bva, 0, cr, &ct); 5616 5617 /* 5618 * If we can't get the attributes, then we can't do the 5619 * right access checking. So, we'll fail the request. 5620 */ 5621 if (error) { 5622 *cs->statusp = resp->status = puterrno4(error); 5623 goto out; 5624 } 5625 5626 if (rdonly4(cs->exi, cs->vp, req)) { 5627 *cs->statusp = resp->status = NFS4ERR_ROFS; 5628 goto out; 5629 } 5630 5631 if (vp->v_type != VREG) { 5632 *cs->statusp = resp->status = 5633 ((vp->v_type == VDIR) ? NFS4ERR_ISDIR : NFS4ERR_INVAL); 5634 goto out; 5635 } 5636 5637 if (crgetuid(cr) != bva.va_uid && 5638 (error = VOP_ACCESS(vp, VWRITE, 0, cr, &ct))) { 5639 *cs->statusp = resp->status = puterrno4(error); 5640 goto out; 5641 } 5642 5643 if (MANDLOCK(vp, bva.va_mode)) { 5644 *cs->statusp = resp->status = NFS4ERR_ACCESS; 5645 goto out; 5646 } 5647 5648 if (args->data_len == 0) { 5649 *cs->statusp = resp->status = NFS4_OK; 5650 resp->count = 0; 5651 resp->committed = args->stable; 5652 resp->writeverf = Write4verf; 5653 goto out; 5654 } 5655 5656 if (args->mblk != NULL) { 5657 mblk_t *m; 5658 uint_t bytes, round_len; 5659 5660 iovcnt = 0; 5661 bytes = 0; 5662 round_len = roundup(args->data_len, BYTES_PER_XDR_UNIT); 5663 for (m = args->mblk; 5664 m != NULL && bytes < round_len; 5665 m = m->b_cont) { 5666 iovcnt++; 5667 bytes += MBLKL(m); 5668 } 5669 #ifdef DEBUG 5670 /* should have ended on an mblk boundary */ 5671 if (bytes != round_len) { 5672 printf("bytes=0x%x, round_len=0x%x, req len=0x%x\n", 5673 bytes, round_len, args->data_len); 5674 printf("args=%p, args->mblk=%p, m=%p", (void *)args, 5675 (void *)args->mblk, (void *)m); 5676 ASSERT(bytes == round_len); 5677 } 5678 #endif 5679 if (iovcnt <= MAX_IOVECS) { 5680 iovp = iov; 5681 } else { 5682 iovp = kmem_alloc(sizeof (*iovp) * iovcnt, KM_SLEEP); 5683 } 5684 mblk_to_iov(args->mblk, iovcnt, iovp); 5685 } else if (args->rlist != NULL) { 5686 iovcnt = 1; 5687 iovp = iov; 5688 iovp->iov_base = (char *)((args->rlist)->u.c_daddr3); 5689 iovp->iov_len = args->data_len; 5690 } else { 5691 iovcnt = 1; 5692 iovp = iov; 5693 iovp->iov_base = args->data_val; 5694 iovp->iov_len = args->data_len; 5695 } 5696 5697 uio.uio_iov = iovp; 5698 uio.uio_iovcnt = iovcnt; 5699 5700 uio.uio_segflg = UIO_SYSSPACE; 5701 uio.uio_extflg = UIO_COPY_DEFAULT; 5702 uio.uio_loffset = args->offset; 5703 uio.uio_resid = args->data_len; 5704 uio.uio_llimit = curproc->p_fsz_ctl; 5705 rlimit = uio.uio_llimit - args->offset; 5706 if (rlimit < (u_offset_t)uio.uio_resid) 5707 uio.uio_resid = (int)rlimit; 5708 5709 if (args->stable == UNSTABLE4) 5710 ioflag = 0; 5711 else if (args->stable == FILE_SYNC4) 5712 ioflag = FSYNC; 5713 else if (args->stable == DATA_SYNC4) 5714 ioflag = FDSYNC; 5715 else { 5716 if (iovp != iov) 5717 kmem_free(iovp, sizeof (*iovp) * iovcnt); 5718 *cs->statusp = resp->status = NFS4ERR_INVAL; 5719 goto out; 5720 } 5721 5722 /* 5723 * We're changing creds because VM may fault and we need 5724 * the cred of the current thread to be used if quota 5725 * checking is enabled. 5726 */ 5727 savecred = curthread->t_cred; 5728 curthread->t_cred = cr; 5729 error = do_io(FWRITE, vp, &uio, ioflag, cr, &ct); 5730 curthread->t_cred = savecred; 5731 5732 if (iovp != iov) 5733 kmem_free(iovp, sizeof (*iovp) * iovcnt); 5734 5735 if (error) { 5736 *cs->statusp = resp->status = puterrno4(error); 5737 goto out; 5738 } 5739 5740 *cs->statusp = resp->status = NFS4_OK; 5741 resp->count = args->data_len - uio.uio_resid; 5742 5743 if (ioflag == 0) 5744 resp->committed = UNSTABLE4; 5745 else 5746 resp->committed = FILE_SYNC4; 5747 5748 resp->writeverf = Write4verf; 5749 5750 out: 5751 if (in_crit) 5752 nbl_end_crit(vp); 5753 5754 DTRACE_NFSV4_2(op__write__done, struct compound_state *, cs, 5755 WRITE4res *, resp); 5756 } 5757 5758 5759 /* XXX put in a header file */ 5760 extern int sec_svc_getcred(struct svc_req *, cred_t *, caddr_t *, int *); 5761 5762 void 5763 rfs4_compound(COMPOUND4args *args, COMPOUND4res *resp, struct exportinfo *exi, 5764 struct svc_req *req, cred_t *cr, int *rv) 5765 { 5766 uint_t i; 5767 struct compound_state cs; 5768 5769 if (rv != NULL) 5770 *rv = 0; 5771 rfs4_init_compound_state(&cs); 5772 /* 5773 * Form a reply tag by copying over the reqeuest tag. 5774 */ 5775 resp->tag.utf8string_val = 5776 kmem_alloc(args->tag.utf8string_len, KM_SLEEP); 5777 resp->tag.utf8string_len = args->tag.utf8string_len; 5778 bcopy(args->tag.utf8string_val, resp->tag.utf8string_val, 5779 resp->tag.utf8string_len); 5780 5781 cs.statusp = &resp->status; 5782 cs.req = req; 5783 resp->array = NULL; 5784 resp->array_len = 0; 5785 5786 resp->status = utf8_name_verify(&(resp->tag)); 5787 if (resp->status != NFS4_OK) 5788 return; 5789 5790 /* 5791 * XXX for now, minorversion should be zero 5792 */ 5793 if (args->minorversion != NFS4_MINORVERSION) { 5794 DTRACE_NFSV4_2(compound__start, struct compound_state *, 5795 &cs, COMPOUND4args *, args); 5796 resp->status = NFS4ERR_MINOR_VERS_MISMATCH; 5797 DTRACE_NFSV4_2(compound__done, struct compound_state *, 5798 &cs, COMPOUND4res *, resp); 5799 return; 5800 } 5801 5802 if (args->array_len == 0) { 5803 resp->status = NFS4_OK; 5804 return; 5805 } 5806 5807 ASSERT(exi == NULL); 5808 ASSERT(cr == NULL); 5809 5810 cr = crget(); 5811 ASSERT(cr != NULL); 5812 5813 if (sec_svc_getcred(req, cr, &cs.principal, &cs.nfsflavor) == 0) { 5814 DTRACE_NFSV4_2(compound__start, struct compound_state *, 5815 &cs, COMPOUND4args *, args); 5816 crfree(cr); 5817 DTRACE_NFSV4_2(compound__done, struct compound_state *, 5818 &cs, COMPOUND4res *, resp); 5819 svcerr_badcred(req->rq_xprt); 5820 if (rv != NULL) 5821 *rv = 1; 5822 return; 5823 } 5824 resp->array_len = args->array_len; 5825 resp->array = kmem_zalloc(args->array_len * sizeof (nfs_resop4), 5826 KM_SLEEP); 5827 5828 cs.basecr = cr; 5829 5830 DTRACE_NFSV4_2(compound__start, struct compound_state *, &cs, 5831 COMPOUND4args *, args); 5832 5833 /* 5834 * For now, NFS4 compound processing must be protected by 5835 * exported_lock because it can access more than one exportinfo 5836 * per compound and share/unshare can now change multiple 5837 * exinfo structs. The NFS2/3 code only refs 1 exportinfo 5838 * per proc (excluding public exinfo), and exi_count design 5839 * is sufficient to protect concurrent execution of NFS2/3 5840 * ops along with unexport. This lock will be removed as 5841 * part of the NFSv4 phase 2 namespace redesign work. 5842 */ 5843 rw_enter(&exported_lock, RW_READER); 5844 5845 /* 5846 * If this is the first compound we've seen, we need to start all 5847 * new instances' grace periods. 5848 */ 5849 if (rfs4_seen_first_compound == 0) { 5850 rfs4_grace_start_new(); 5851 /* 5852 * This must be set after rfs4_grace_start_new(), otherwise 5853 * another thread could proceed past here before the former 5854 * is finished. 5855 */ 5856 rfs4_seen_first_compound = 1; 5857 } 5858 5859 for (i = 0; i < args->array_len && cs.cont; i++) { 5860 nfs_argop4 *argop; 5861 nfs_resop4 *resop; 5862 uint_t op; 5863 5864 argop = &args->array[i]; 5865 resop = &resp->array[i]; 5866 resop->resop = argop->argop; 5867 op = (uint_t)resop->resop; 5868 5869 if (op < rfsv4disp_cnt) { 5870 /* 5871 * Count the individual ops here; NULL and COMPOUND 5872 * are counted in common_dispatch() 5873 */ 5874 rfsproccnt_v4_ptr[op].value.ui64++; 5875 5876 NFS4_DEBUG(rfs4_debug > 1, 5877 (CE_NOTE, "Executing %s", rfs4_op_string[op])); 5878 (*rfsv4disptab[op].dis_proc)(argop, resop, req, &cs); 5879 NFS4_DEBUG(rfs4_debug > 1, (CE_NOTE, "%s returned %d", 5880 rfs4_op_string[op], *cs.statusp)); 5881 if (*cs.statusp != NFS4_OK) 5882 cs.cont = FALSE; 5883 } else { 5884 /* 5885 * This is effectively dead code since XDR code 5886 * will have already returned BADXDR if op doesn't 5887 * decode to legal value. This only done for a 5888 * day when XDR code doesn't verify v4 opcodes. 5889 */ 5890 op = OP_ILLEGAL; 5891 rfsproccnt_v4_ptr[OP_ILLEGAL_IDX].value.ui64++; 5892 5893 rfs4_op_illegal(argop, resop, req, &cs); 5894 cs.cont = FALSE; 5895 } 5896 5897 /* 5898 * If not at last op, and if we are to stop, then 5899 * compact the results array. 5900 */ 5901 if ((i + 1) < args->array_len && !cs.cont) { 5902 nfs_resop4 *new_res = kmem_alloc( 5903 (i+1) * sizeof (nfs_resop4), KM_SLEEP); 5904 bcopy(resp->array, 5905 new_res, (i+1) * sizeof (nfs_resop4)); 5906 kmem_free(resp->array, 5907 args->array_len * sizeof (nfs_resop4)); 5908 5909 resp->array_len = i + 1; 5910 resp->array = new_res; 5911 } 5912 } 5913 5914 rw_exit(&exported_lock); 5915 5916 DTRACE_NFSV4_2(compound__done, struct compound_state *, &cs, 5917 COMPOUND4res *, resp); 5918 5919 if (cs.vp) 5920 VN_RELE(cs.vp); 5921 if (cs.saved_vp) 5922 VN_RELE(cs.saved_vp); 5923 if (cs.saved_fh.nfs_fh4_val) 5924 kmem_free(cs.saved_fh.nfs_fh4_val, NFS4_FHSIZE); 5925 5926 if (cs.basecr) 5927 crfree(cs.basecr); 5928 if (cs.cr) 5929 crfree(cs.cr); 5930 /* 5931 * done with this compound request, free the label 5932 */ 5933 5934 if (req->rq_label != NULL) { 5935 kmem_free(req->rq_label, sizeof (bslabel_t)); 5936 req->rq_label = NULL; 5937 } 5938 } 5939 5940 /* 5941 * XXX because of what appears to be duplicate calls to rfs4_compound_free 5942 * XXX zero out the tag and array values. Need to investigate why the 5943 * XXX calls occur, but at least prevent the panic for now. 5944 */ 5945 void 5946 rfs4_compound_free(COMPOUND4res *resp) 5947 { 5948 uint_t i; 5949 5950 if (resp->tag.utf8string_val) { 5951 UTF8STRING_FREE(resp->tag) 5952 } 5953 5954 for (i = 0; i < resp->array_len; i++) { 5955 nfs_resop4 *resop; 5956 uint_t op; 5957 5958 resop = &resp->array[i]; 5959 op = (uint_t)resop->resop; 5960 if (op < rfsv4disp_cnt) { 5961 (*rfsv4disptab[op].dis_resfree)(resop); 5962 } 5963 } 5964 if (resp->array != NULL) { 5965 kmem_free(resp->array, resp->array_len * sizeof (nfs_resop4)); 5966 } 5967 } 5968 5969 /* 5970 * Process the value of the compound request rpc flags, as a bit-AND 5971 * of the individual per-op flags (idempotent, allowork, publicfh_ok) 5972 */ 5973 void 5974 rfs4_compound_flagproc(COMPOUND4args *args, int *flagp) 5975 { 5976 int i; 5977 int flag = RPC_ALL; 5978 5979 for (i = 0; flag && i < args->array_len; i++) { 5980 uint_t op; 5981 5982 op = (uint_t)args->array[i].argop; 5983 5984 if (op < rfsv4disp_cnt) 5985 flag &= rfsv4disptab[op].dis_flags; 5986 else 5987 flag = 0; 5988 } 5989 *flagp = flag; 5990 } 5991 5992 nfsstat4 5993 rfs4_client_sysid(rfs4_client_t *cp, sysid_t *sp) 5994 { 5995 nfsstat4 e; 5996 5997 rfs4_dbe_lock(cp->rc_dbe); 5998 5999 if (cp->rc_sysidt != LM_NOSYSID) { 6000 *sp = cp->rc_sysidt; 6001 e = NFS4_OK; 6002 6003 } else if ((cp->rc_sysidt = lm_alloc_sysidt()) != LM_NOSYSID) { 6004 *sp = cp->rc_sysidt; 6005 e = NFS4_OK; 6006 6007 NFS4_DEBUG(rfs4_debug, (CE_NOTE, 6008 "rfs4_client_sysid: allocated 0x%x\n", *sp)); 6009 } else 6010 e = NFS4ERR_DELAY; 6011 6012 rfs4_dbe_unlock(cp->rc_dbe); 6013 return (e); 6014 } 6015 6016 #if defined(DEBUG) && ! defined(lint) 6017 static void lock_print(char *str, int operation, struct flock64 *flk) 6018 { 6019 char *op, *type; 6020 6021 switch (operation) { 6022 case F_GETLK: op = "F_GETLK"; 6023 break; 6024 case F_SETLK: op = "F_SETLK"; 6025 break; 6026 case F_SETLK_NBMAND: op = "F_SETLK_NBMAND"; 6027 break; 6028 default: op = "F_UNKNOWN"; 6029 break; 6030 } 6031 switch (flk->l_type) { 6032 case F_UNLCK: type = "F_UNLCK"; 6033 break; 6034 case F_RDLCK: type = "F_RDLCK"; 6035 break; 6036 case F_WRLCK: type = "F_WRLCK"; 6037 break; 6038 default: type = "F_UNKNOWN"; 6039 break; 6040 } 6041 6042 ASSERT(flk->l_whence == 0); 6043 cmn_err(CE_NOTE, "%s: %s, type = %s, off = %llx len = %llx pid = %d", 6044 str, op, type, (longlong_t)flk->l_start, 6045 flk->l_len ? (longlong_t)flk->l_len : ~0LL, flk->l_pid); 6046 } 6047 6048 #define LOCK_PRINT(d, s, t, f) if (d) lock_print(s, t, f) 6049 #else 6050 #define LOCK_PRINT(d, s, t, f) 6051 #endif 6052 6053 /*ARGSUSED*/ 6054 static bool_t 6055 creds_ok(cred_set_t cr_set, struct svc_req *req, struct compound_state *cs) 6056 { 6057 return (TRUE); 6058 } 6059 6060 /* 6061 * Look up the pathname using the vp in cs as the directory vnode. 6062 * cs->vp will be the vnode for the file on success 6063 */ 6064 6065 static nfsstat4 6066 rfs4_lookup(component4 *component, struct svc_req *req, 6067 struct compound_state *cs) 6068 { 6069 char *nm; 6070 uint32_t len; 6071 nfsstat4 status; 6072 struct sockaddr *ca; 6073 char *name; 6074 6075 if (cs->vp == NULL) { 6076 return (NFS4ERR_NOFILEHANDLE); 6077 } 6078 if (cs->vp->v_type != VDIR) { 6079 return (NFS4ERR_NOTDIR); 6080 } 6081 6082 status = utf8_dir_verify(component); 6083 if (status != NFS4_OK) 6084 return (status); 6085 6086 nm = utf8_to_fn(component, &len, NULL); 6087 if (nm == NULL) { 6088 return (NFS4ERR_INVAL); 6089 } 6090 6091 if (len > MAXNAMELEN) { 6092 kmem_free(nm, len); 6093 return (NFS4ERR_NAMETOOLONG); 6094 } 6095 6096 ca = (struct sockaddr *)svc_getrpccaller(req->rq_xprt)->buf; 6097 name = nfscmd_convname(ca, cs->exi, nm, NFSCMD_CONV_INBOUND, 6098 MAXPATHLEN + 1); 6099 6100 if (name == NULL) { 6101 kmem_free(nm, len); 6102 return (NFS4ERR_INVAL); 6103 } 6104 6105 status = do_rfs4_op_lookup(name, req, cs); 6106 6107 if (name != nm) 6108 kmem_free(name, MAXPATHLEN + 1); 6109 6110 kmem_free(nm, len); 6111 6112 return (status); 6113 } 6114 6115 static nfsstat4 6116 rfs4_lookupfile(component4 *component, struct svc_req *req, 6117 struct compound_state *cs, uint32_t access, change_info4 *cinfo) 6118 { 6119 nfsstat4 status; 6120 vnode_t *dvp = cs->vp; 6121 vattr_t bva, ava, fva; 6122 int error; 6123 6124 /* Get "before" change value */ 6125 bva.va_mask = AT_CTIME|AT_SEQ; 6126 error = VOP_GETATTR(dvp, &bva, 0, cs->cr, NULL); 6127 if (error) 6128 return (puterrno4(error)); 6129 6130 /* rfs4_lookup may VN_RELE directory */ 6131 VN_HOLD(dvp); 6132 6133 status = rfs4_lookup(component, req, cs); 6134 if (status != NFS4_OK) { 6135 VN_RELE(dvp); 6136 return (status); 6137 } 6138 6139 /* 6140 * Get "after" change value, if it fails, simply return the 6141 * before value. 6142 */ 6143 ava.va_mask = AT_CTIME|AT_SEQ; 6144 if (VOP_GETATTR(dvp, &ava, 0, cs->cr, NULL)) { 6145 ava.va_ctime = bva.va_ctime; 6146 ava.va_seq = 0; 6147 } 6148 VN_RELE(dvp); 6149 6150 /* 6151 * Validate the file is a file 6152 */ 6153 fva.va_mask = AT_TYPE|AT_MODE; 6154 error = VOP_GETATTR(cs->vp, &fva, 0, cs->cr, NULL); 6155 if (error) 6156 return (puterrno4(error)); 6157 6158 if (fva.va_type != VREG) { 6159 if (fva.va_type == VDIR) 6160 return (NFS4ERR_ISDIR); 6161 if (fva.va_type == VLNK) 6162 return (NFS4ERR_SYMLINK); 6163 return (NFS4ERR_INVAL); 6164 } 6165 6166 NFS4_SET_FATTR4_CHANGE(cinfo->before, bva.va_ctime); 6167 NFS4_SET_FATTR4_CHANGE(cinfo->after, ava.va_ctime); 6168 6169 /* 6170 * It is undefined if VOP_LOOKUP will change va_seq, so 6171 * cinfo.atomic = TRUE only if we have 6172 * non-zero va_seq's, and they have not changed. 6173 */ 6174 if (bva.va_seq && ava.va_seq && ava.va_seq == bva.va_seq) 6175 cinfo->atomic = TRUE; 6176 else 6177 cinfo->atomic = FALSE; 6178 6179 /* Check for mandatory locking */ 6180 cs->mandlock = MANDLOCK(cs->vp, fva.va_mode); 6181 return (check_open_access(access, cs, req)); 6182 } 6183 6184 static nfsstat4 6185 create_vnode(vnode_t *dvp, char *nm, vattr_t *vap, createmode4 mode, 6186 timespec32_t *mtime, cred_t *cr, vnode_t **vpp, bool_t *created) 6187 { 6188 int error; 6189 nfsstat4 status = NFS4_OK; 6190 vattr_t va; 6191 6192 tryagain: 6193 6194 /* 6195 * The file open mode used is VWRITE. If the client needs 6196 * some other semantic, then it should do the access checking 6197 * itself. It would have been nice to have the file open mode 6198 * passed as part of the arguments. 6199 */ 6200 6201 *created = TRUE; 6202 error = VOP_CREATE(dvp, nm, vap, EXCL, VWRITE, vpp, cr, 0, NULL, NULL); 6203 6204 if (error) { 6205 *created = FALSE; 6206 6207 /* 6208 * If we got something other than file already exists 6209 * then just return this error. Otherwise, we got 6210 * EEXIST. If we were doing a GUARDED create, then 6211 * just return this error. Otherwise, we need to 6212 * make sure that this wasn't a duplicate of an 6213 * exclusive create request. 6214 * 6215 * The assumption is made that a non-exclusive create 6216 * request will never return EEXIST. 6217 */ 6218 6219 if (error != EEXIST || mode == GUARDED4) { 6220 status = puterrno4(error); 6221 return (status); 6222 } 6223 error = VOP_LOOKUP(dvp, nm, vpp, NULL, 0, NULL, cr, 6224 NULL, NULL, NULL); 6225 6226 if (error) { 6227 /* 6228 * We couldn't find the file that we thought that 6229 * we just created. So, we'll just try creating 6230 * it again. 6231 */ 6232 if (error == ENOENT) 6233 goto tryagain; 6234 6235 status = puterrno4(error); 6236 return (status); 6237 } 6238 6239 if (mode == UNCHECKED4) { 6240 /* existing object must be regular file */ 6241 if ((*vpp)->v_type != VREG) { 6242 if ((*vpp)->v_type == VDIR) 6243 status = NFS4ERR_ISDIR; 6244 else if ((*vpp)->v_type == VLNK) 6245 status = NFS4ERR_SYMLINK; 6246 else 6247 status = NFS4ERR_INVAL; 6248 VN_RELE(*vpp); 6249 return (status); 6250 } 6251 6252 return (NFS4_OK); 6253 } 6254 6255 /* Check for duplicate request */ 6256 ASSERT(mtime != 0); 6257 va.va_mask = AT_MTIME; 6258 error = VOP_GETATTR(*vpp, &va, 0, cr, NULL); 6259 if (!error) { 6260 /* We found the file */ 6261 if (va.va_mtime.tv_sec != mtime->tv_sec || 6262 va.va_mtime.tv_nsec != mtime->tv_nsec) { 6263 /* but its not our creation */ 6264 VN_RELE(*vpp); 6265 return (NFS4ERR_EXIST); 6266 } 6267 *created = TRUE; /* retrans of create == created */ 6268 return (NFS4_OK); 6269 } 6270 VN_RELE(*vpp); 6271 return (NFS4ERR_EXIST); 6272 } 6273 6274 return (NFS4_OK); 6275 } 6276 6277 static nfsstat4 6278 check_open_access(uint32_t access, struct compound_state *cs, 6279 struct svc_req *req) 6280 { 6281 int error; 6282 vnode_t *vp; 6283 bool_t readonly; 6284 cred_t *cr = cs->cr; 6285 6286 /* For now we don't allow mandatory locking as per V2/V3 */ 6287 if (cs->access == CS_ACCESS_DENIED || cs->mandlock) { 6288 return (NFS4ERR_ACCESS); 6289 } 6290 6291 vp = cs->vp; 6292 ASSERT(cr != NULL && vp->v_type == VREG); 6293 6294 /* 6295 * If the file system is exported read only and we are trying 6296 * to open for write, then return NFS4ERR_ROFS 6297 */ 6298 6299 readonly = rdonly4(cs->exi, cs->vp, req); 6300 6301 if ((access & OPEN4_SHARE_ACCESS_WRITE) && readonly) 6302 return (NFS4ERR_ROFS); 6303 6304 if (access & OPEN4_SHARE_ACCESS_READ) { 6305 if ((VOP_ACCESS(vp, VREAD, 0, cr, NULL) != 0) && 6306 (VOP_ACCESS(vp, VEXEC, 0, cr, NULL) != 0)) { 6307 return (NFS4ERR_ACCESS); 6308 } 6309 } 6310 6311 if (access & OPEN4_SHARE_ACCESS_WRITE) { 6312 error = VOP_ACCESS(vp, VWRITE, 0, cr, NULL); 6313 if (error) 6314 return (NFS4ERR_ACCESS); 6315 } 6316 6317 return (NFS4_OK); 6318 } 6319 6320 static nfsstat4 6321 rfs4_createfile(OPEN4args *args, struct svc_req *req, struct compound_state *cs, 6322 change_info4 *cinfo, bitmap4 *attrset, clientid4 clientid) 6323 { 6324 struct nfs4_svgetit_arg sarg; 6325 struct nfs4_ntov_table ntov; 6326 6327 bool_t ntov_table_init = FALSE; 6328 struct statvfs64 sb; 6329 nfsstat4 status; 6330 vnode_t *vp; 6331 vattr_t bva, ava, iva, cva, *vap; 6332 vnode_t *dvp; 6333 timespec32_t *mtime; 6334 char *nm = NULL; 6335 uint_t buflen; 6336 bool_t created; 6337 bool_t setsize = FALSE; 6338 len_t reqsize; 6339 int error; 6340 bool_t trunc; 6341 caller_context_t ct; 6342 component4 *component; 6343 bslabel_t *clabel; 6344 struct sockaddr *ca; 6345 char *name = NULL; 6346 6347 sarg.sbp = &sb; 6348 sarg.is_referral = B_FALSE; 6349 6350 dvp = cs->vp; 6351 6352 /* Check if the file system is read only */ 6353 if (rdonly4(cs->exi, dvp, req)) 6354 return (NFS4ERR_ROFS); 6355 6356 /* check the label of including directory */ 6357 if (is_system_labeled()) { 6358 ASSERT(req->rq_label != NULL); 6359 clabel = req->rq_label; 6360 DTRACE_PROBE2(tx__rfs4__log__info__opremove__clabel, char *, 6361 "got client label from request(1)", 6362 struct svc_req *, req); 6363 if (!blequal(&l_admin_low->tsl_label, clabel)) { 6364 if (!do_rfs_label_check(clabel, dvp, EQUALITY_CHECK, 6365 cs->exi)) { 6366 return (NFS4ERR_ACCESS); 6367 } 6368 } 6369 } 6370 6371 /* 6372 * Get the last component of path name in nm. cs will reference 6373 * the including directory on success. 6374 */ 6375 component = &args->open_claim4_u.file; 6376 status = utf8_dir_verify(component); 6377 if (status != NFS4_OK) 6378 return (status); 6379 6380 nm = utf8_to_fn(component, &buflen, NULL); 6381 6382 if (nm == NULL) 6383 return (NFS4ERR_RESOURCE); 6384 6385 if (buflen > MAXNAMELEN) { 6386 kmem_free(nm, buflen); 6387 return (NFS4ERR_NAMETOOLONG); 6388 } 6389 6390 bva.va_mask = AT_TYPE|AT_CTIME|AT_SEQ; 6391 error = VOP_GETATTR(dvp, &bva, 0, cs->cr, NULL); 6392 if (error) { 6393 kmem_free(nm, buflen); 6394 return (puterrno4(error)); 6395 } 6396 6397 if (bva.va_type != VDIR) { 6398 kmem_free(nm, buflen); 6399 return (NFS4ERR_NOTDIR); 6400 } 6401 6402 NFS4_SET_FATTR4_CHANGE(cinfo->before, bva.va_ctime) 6403 6404 switch (args->mode) { 6405 case GUARDED4: 6406 /*FALLTHROUGH*/ 6407 case UNCHECKED4: 6408 nfs4_ntov_table_init(&ntov); 6409 ntov_table_init = TRUE; 6410 6411 *attrset = 0; 6412 status = do_rfs4_set_attrs(attrset, 6413 &args->createhow4_u.createattrs, 6414 cs, &sarg, &ntov, NFS4ATTR_SETIT); 6415 6416 if (status == NFS4_OK && (sarg.vap->va_mask & AT_TYPE) && 6417 sarg.vap->va_type != VREG) { 6418 if (sarg.vap->va_type == VDIR) 6419 status = NFS4ERR_ISDIR; 6420 else if (sarg.vap->va_type == VLNK) 6421 status = NFS4ERR_SYMLINK; 6422 else 6423 status = NFS4ERR_INVAL; 6424 } 6425 6426 if (status != NFS4_OK) { 6427 kmem_free(nm, buflen); 6428 nfs4_ntov_table_free(&ntov, &sarg); 6429 *attrset = 0; 6430 return (status); 6431 } 6432 6433 vap = sarg.vap; 6434 vap->va_type = VREG; 6435 vap->va_mask |= AT_TYPE; 6436 6437 if ((vap->va_mask & AT_MODE) == 0) { 6438 vap->va_mask |= AT_MODE; 6439 vap->va_mode = (mode_t)0600; 6440 } 6441 6442 if (vap->va_mask & AT_SIZE) { 6443 6444 /* Disallow create with a non-zero size */ 6445 6446 if ((reqsize = sarg.vap->va_size) != 0) { 6447 kmem_free(nm, buflen); 6448 nfs4_ntov_table_free(&ntov, &sarg); 6449 *attrset = 0; 6450 return (NFS4ERR_INVAL); 6451 } 6452 setsize = TRUE; 6453 } 6454 break; 6455 6456 case EXCLUSIVE4: 6457 /* prohibit EXCL create of named attributes */ 6458 if (dvp->v_flag & V_XATTRDIR) { 6459 kmem_free(nm, buflen); 6460 *attrset = 0; 6461 return (NFS4ERR_INVAL); 6462 } 6463 6464 cva.va_mask = AT_TYPE | AT_MTIME | AT_MODE; 6465 cva.va_type = VREG; 6466 /* 6467 * Ensure no time overflows. Assumes underlying 6468 * filesystem supports at least 32 bits. 6469 * Truncate nsec to usec resolution to allow valid 6470 * compares even if the underlying filesystem truncates. 6471 */ 6472 mtime = (timespec32_t *)&args->createhow4_u.createverf; 6473 cva.va_mtime.tv_sec = mtime->tv_sec % TIME32_MAX; 6474 cva.va_mtime.tv_nsec = (mtime->tv_nsec / 1000) * 1000; 6475 cva.va_mode = (mode_t)0; 6476 vap = &cva; 6477 6478 /* 6479 * For EXCL create, attrset is set to the server attr 6480 * used to cache the client's verifier. 6481 */ 6482 *attrset = FATTR4_TIME_MODIFY_MASK; 6483 break; 6484 } 6485 6486 ca = (struct sockaddr *)svc_getrpccaller(req->rq_xprt)->buf; 6487 name = nfscmd_convname(ca, cs->exi, nm, NFSCMD_CONV_INBOUND, 6488 MAXPATHLEN + 1); 6489 6490 if (name == NULL) { 6491 kmem_free(nm, buflen); 6492 return (NFS4ERR_SERVERFAULT); 6493 } 6494 6495 status = create_vnode(dvp, name, vap, args->mode, mtime, 6496 cs->cr, &vp, &created); 6497 if (nm != name) 6498 kmem_free(name, MAXPATHLEN + 1); 6499 kmem_free(nm, buflen); 6500 6501 if (status != NFS4_OK) { 6502 if (ntov_table_init) 6503 nfs4_ntov_table_free(&ntov, &sarg); 6504 *attrset = 0; 6505 return (status); 6506 } 6507 6508 trunc = (setsize && !created); 6509 6510 if (args->mode != EXCLUSIVE4) { 6511 bitmap4 createmask = args->createhow4_u.createattrs.attrmask; 6512 6513 /* 6514 * True verification that object was created with correct 6515 * attrs is impossible. The attrs could have been changed 6516 * immediately after object creation. If attributes did 6517 * not verify, the only recourse for the server is to 6518 * destroy the object. Maybe if some attrs (like gid) 6519 * are set incorrectly, the object should be destroyed; 6520 * however, seems bad as a default policy. Do we really 6521 * want to destroy an object over one of the times not 6522 * verifying correctly? For these reasons, the server 6523 * currently sets bits in attrset for createattrs 6524 * that were set; however, no verification is done. 6525 * 6526 * vmask_to_nmask accounts for vattr bits set on create 6527 * [do_rfs4_set_attrs() only sets resp bits for 6528 * non-vattr/vfs bits.] 6529 * Mask off any bits we set by default so as not to return 6530 * more attrset bits than were requested in createattrs 6531 */ 6532 if (created) { 6533 nfs4_vmask_to_nmask(sarg.vap->va_mask, attrset); 6534 *attrset &= createmask; 6535 } else { 6536 /* 6537 * We did not create the vnode (we tried but it 6538 * already existed). In this case, the only createattr 6539 * that the spec allows the server to set is size, 6540 * and even then, it can only be set if it is 0. 6541 */ 6542 *attrset = 0; 6543 if (trunc) 6544 *attrset = FATTR4_SIZE_MASK; 6545 } 6546 } 6547 if (ntov_table_init) 6548 nfs4_ntov_table_free(&ntov, &sarg); 6549 6550 /* 6551 * Get the initial "after" sequence number, if it fails, 6552 * set to zero, time to before. 6553 */ 6554 iva.va_mask = AT_CTIME|AT_SEQ; 6555 if (VOP_GETATTR(dvp, &iva, 0, cs->cr, NULL)) { 6556 iva.va_seq = 0; 6557 iva.va_ctime = bva.va_ctime; 6558 } 6559 6560 /* 6561 * create_vnode attempts to create the file exclusive, 6562 * if it already exists the VOP_CREATE will fail and 6563 * may not increase va_seq. It is atomic if 6564 * we haven't changed the directory, but if it has changed 6565 * we don't know what changed it. 6566 */ 6567 if (!created) { 6568 if (bva.va_seq && iva.va_seq && 6569 bva.va_seq == iva.va_seq) 6570 cinfo->atomic = TRUE; 6571 else 6572 cinfo->atomic = FALSE; 6573 NFS4_SET_FATTR4_CHANGE(cinfo->after, iva.va_ctime); 6574 } else { 6575 /* 6576 * The entry was created, we need to sync the 6577 * directory metadata. 6578 */ 6579 (void) VOP_FSYNC(dvp, 0, cs->cr, NULL); 6580 6581 /* 6582 * Get "after" change value, if it fails, simply return the 6583 * before value. 6584 */ 6585 ava.va_mask = AT_CTIME|AT_SEQ; 6586 if (VOP_GETATTR(dvp, &ava, 0, cs->cr, NULL)) { 6587 ava.va_ctime = bva.va_ctime; 6588 ava.va_seq = 0; 6589 } 6590 6591 NFS4_SET_FATTR4_CHANGE(cinfo->after, ava.va_ctime); 6592 6593 /* 6594 * The cinfo->atomic = TRUE only if we have 6595 * non-zero va_seq's, and it has incremented by exactly one 6596 * during the create_vnode and it didn't 6597 * change during the VOP_FSYNC. 6598 */ 6599 if (bva.va_seq && iva.va_seq && ava.va_seq && 6600 iva.va_seq == (bva.va_seq + 1) && iva.va_seq == ava.va_seq) 6601 cinfo->atomic = TRUE; 6602 else 6603 cinfo->atomic = FALSE; 6604 } 6605 6606 /* Check for mandatory locking and that the size gets set. */ 6607 cva.va_mask = AT_MODE; 6608 if (setsize) 6609 cva.va_mask |= AT_SIZE; 6610 6611 /* Assume the worst */ 6612 cs->mandlock = TRUE; 6613 6614 if (VOP_GETATTR(vp, &cva, 0, cs->cr, NULL) == 0) { 6615 cs->mandlock = MANDLOCK(cs->vp, cva.va_mode); 6616 6617 /* 6618 * Truncate the file if necessary; this would be 6619 * the case for create over an existing file. 6620 */ 6621 6622 if (trunc) { 6623 int in_crit = 0; 6624 rfs4_file_t *fp; 6625 bool_t create = FALSE; 6626 6627 /* 6628 * We are writing over an existing file. 6629 * Check to see if we need to recall a delegation. 6630 */ 6631 rfs4_hold_deleg_policy(); 6632 if ((fp = rfs4_findfile(vp, NULL, &create)) != NULL) { 6633 if (rfs4_check_delegated_byfp(FWRITE, fp, 6634 (reqsize == 0), FALSE, FALSE, &clientid)) { 6635 rfs4_file_rele(fp); 6636 rfs4_rele_deleg_policy(); 6637 VN_RELE(vp); 6638 *attrset = 0; 6639 return (NFS4ERR_DELAY); 6640 } 6641 rfs4_file_rele(fp); 6642 } 6643 rfs4_rele_deleg_policy(); 6644 6645 if (nbl_need_check(vp)) { 6646 in_crit = 1; 6647 6648 ASSERT(reqsize == 0); 6649 6650 nbl_start_crit(vp, RW_READER); 6651 if (nbl_conflict(vp, NBL_WRITE, 0, 6652 cva.va_size, 0, NULL)) { 6653 in_crit = 0; 6654 nbl_end_crit(vp); 6655 VN_RELE(vp); 6656 *attrset = 0; 6657 return (NFS4ERR_ACCESS); 6658 } 6659 } 6660 ct.cc_sysid = 0; 6661 ct.cc_pid = 0; 6662 ct.cc_caller_id = nfs4_srv_caller_id; 6663 ct.cc_flags = CC_DONTBLOCK; 6664 6665 cva.va_mask = AT_SIZE; 6666 cva.va_size = reqsize; 6667 (void) VOP_SETATTR(vp, &cva, 0, cs->cr, &ct); 6668 if (in_crit) 6669 nbl_end_crit(vp); 6670 } 6671 } 6672 6673 error = makefh4(&cs->fh, vp, cs->exi); 6674 6675 /* 6676 * Force modified data and metadata out to stable storage. 6677 */ 6678 (void) VOP_FSYNC(vp, FNODSYNC, cs->cr, NULL); 6679 6680 if (error) { 6681 VN_RELE(vp); 6682 *attrset = 0; 6683 return (puterrno4(error)); 6684 } 6685 6686 /* if parent dir is attrdir, set namedattr fh flag */ 6687 if (dvp->v_flag & V_XATTRDIR) 6688 set_fh4_flag(&cs->fh, FH4_NAMEDATTR); 6689 6690 if (cs->vp) 6691 VN_RELE(cs->vp); 6692 6693 cs->vp = vp; 6694 6695 /* 6696 * if we did not create the file, we will need to check 6697 * the access bits on the file 6698 */ 6699 6700 if (!created) { 6701 if (setsize) 6702 args->share_access |= OPEN4_SHARE_ACCESS_WRITE; 6703 status = check_open_access(args->share_access, cs, req); 6704 if (status != NFS4_OK) 6705 *attrset = 0; 6706 } 6707 return (status); 6708 } 6709 6710 /*ARGSUSED*/ 6711 static void 6712 rfs4_do_open(struct compound_state *cs, struct svc_req *req, 6713 rfs4_openowner_t *oo, delegreq_t deleg, 6714 uint32_t access, uint32_t deny, 6715 OPEN4res *resp, int deleg_cur) 6716 { 6717 /* XXX Currently not using req */ 6718 rfs4_state_t *sp; 6719 rfs4_file_t *fp; 6720 bool_t screate = TRUE; 6721 bool_t fcreate = TRUE; 6722 uint32_t open_a, share_a; 6723 uint32_t open_d, share_d; 6724 rfs4_deleg_state_t *dsp; 6725 sysid_t sysid; 6726 nfsstat4 status; 6727 caller_context_t ct; 6728 int fflags = 0; 6729 int recall = 0; 6730 int err; 6731 int first_open; 6732 6733 /* get the file struct and hold a lock on it during initial open */ 6734 fp = rfs4_findfile_withlock(cs->vp, &cs->fh, &fcreate); 6735 if (fp == NULL) { 6736 resp->status = NFS4ERR_RESOURCE; 6737 DTRACE_PROBE1(nfss__e__do__open1, nfsstat4, resp->status); 6738 return; 6739 } 6740 6741 sp = rfs4_findstate_by_owner_file(oo, fp, &screate); 6742 if (sp == NULL) { 6743 resp->status = NFS4ERR_RESOURCE; 6744 DTRACE_PROBE1(nfss__e__do__open2, nfsstat4, resp->status); 6745 /* No need to keep any reference */ 6746 rw_exit(&fp->rf_file_rwlock); 6747 rfs4_file_rele(fp); 6748 return; 6749 } 6750 6751 /* try to get the sysid before continuing */ 6752 if ((status = rfs4_client_sysid(oo->ro_client, &sysid)) != NFS4_OK) { 6753 resp->status = status; 6754 rfs4_file_rele(fp); 6755 /* Not a fully formed open; "close" it */ 6756 if (screate == TRUE) 6757 rfs4_state_close(sp, FALSE, FALSE, cs->cr); 6758 rfs4_state_rele(sp); 6759 return; 6760 } 6761 6762 /* Calculate the fflags for this OPEN. */ 6763 if (access & OPEN4_SHARE_ACCESS_READ) 6764 fflags |= FREAD; 6765 if (access & OPEN4_SHARE_ACCESS_WRITE) 6766 fflags |= FWRITE; 6767 6768 rfs4_dbe_lock(sp->rs_dbe); 6769 6770 /* 6771 * Calculate the new deny and access mode that this open is adding to 6772 * the file for this open owner; 6773 */ 6774 open_d = (deny & ~sp->rs_open_deny); 6775 open_a = (access & ~sp->rs_open_access); 6776 6777 /* 6778 * Calculate the new share access and share deny modes that this open 6779 * is adding to the file for this open owner; 6780 */ 6781 share_a = (access & ~sp->rs_share_access); 6782 share_d = (deny & ~sp->rs_share_deny); 6783 6784 first_open = (sp->rs_open_access & OPEN4_SHARE_ACCESS_BOTH) == 0; 6785 6786 /* 6787 * Check to see the client has already sent an open for this 6788 * open owner on this file with the same share/deny modes. 6789 * If so, we don't need to check for a conflict and we don't 6790 * need to add another shrlock. If not, then we need to 6791 * check for conflicts in deny and access before checking for 6792 * conflicts in delegation. We don't want to recall a 6793 * delegation based on an open that will eventually fail based 6794 * on shares modes. 6795 */ 6796 6797 if (share_a || share_d) { 6798 if ((err = rfs4_share(sp, access, deny)) != 0) { 6799 rfs4_dbe_unlock(sp->rs_dbe); 6800 resp->status = err; 6801 6802 rfs4_file_rele(fp); 6803 /* Not a fully formed open; "close" it */ 6804 if (screate == TRUE) 6805 rfs4_state_close(sp, FALSE, FALSE, cs->cr); 6806 rfs4_state_rele(sp); 6807 return; 6808 } 6809 } 6810 6811 rfs4_dbe_lock(fp->rf_dbe); 6812 6813 /* 6814 * Check to see if this file is delegated and if so, if a 6815 * recall needs to be done. 6816 */ 6817 if (rfs4_check_recall(sp, access)) { 6818 rfs4_dbe_unlock(fp->rf_dbe); 6819 rfs4_dbe_unlock(sp->rs_dbe); 6820 rfs4_recall_deleg(fp, FALSE, sp->rs_owner->ro_client); 6821 delay(NFS4_DELEGATION_CONFLICT_DELAY); 6822 rfs4_dbe_lock(sp->rs_dbe); 6823 6824 /* if state closed while lock was dropped */ 6825 if (sp->rs_closed) { 6826 if (share_a || share_d) 6827 (void) rfs4_unshare(sp); 6828 rfs4_dbe_unlock(sp->rs_dbe); 6829 rfs4_file_rele(fp); 6830 /* Not a fully formed open; "close" it */ 6831 if (screate == TRUE) 6832 rfs4_state_close(sp, FALSE, FALSE, cs->cr); 6833 rfs4_state_rele(sp); 6834 resp->status = NFS4ERR_OLD_STATEID; 6835 return; 6836 } 6837 6838 rfs4_dbe_lock(fp->rf_dbe); 6839 /* Let's see if the delegation was returned */ 6840 if (rfs4_check_recall(sp, access)) { 6841 rfs4_dbe_unlock(fp->rf_dbe); 6842 if (share_a || share_d) 6843 (void) rfs4_unshare(sp); 6844 rfs4_dbe_unlock(sp->rs_dbe); 6845 rfs4_file_rele(fp); 6846 rfs4_update_lease(sp->rs_owner->ro_client); 6847 6848 /* Not a fully formed open; "close" it */ 6849 if (screate == TRUE) 6850 rfs4_state_close(sp, FALSE, FALSE, cs->cr); 6851 rfs4_state_rele(sp); 6852 resp->status = NFS4ERR_DELAY; 6853 return; 6854 } 6855 } 6856 /* 6857 * the share check passed and any delegation conflict has been 6858 * taken care of, now call vop_open. 6859 * if this is the first open then call vop_open with fflags. 6860 * if not, call vn_open_upgrade with just the upgrade flags. 6861 * 6862 * if the file has been opened already, it will have the current 6863 * access mode in the state struct. if it has no share access, then 6864 * this is a new open. 6865 * 6866 * However, if this is open with CLAIM_DLEGATE_CUR, then don't 6867 * call VOP_OPEN(), just do the open upgrade. 6868 */ 6869 if (first_open && !deleg_cur) { 6870 ct.cc_sysid = sysid; 6871 ct.cc_pid = rfs4_dbe_getid(sp->rs_owner->ro_dbe); 6872 ct.cc_caller_id = nfs4_srv_caller_id; 6873 ct.cc_flags = CC_DONTBLOCK; 6874 err = VOP_OPEN(&cs->vp, fflags, cs->cr, &ct); 6875 if (err) { 6876 rfs4_dbe_unlock(fp->rf_dbe); 6877 if (share_a || share_d) 6878 (void) rfs4_unshare(sp); 6879 rfs4_dbe_unlock(sp->rs_dbe); 6880 rfs4_file_rele(fp); 6881 6882 /* Not a fully formed open; "close" it */ 6883 if (screate == TRUE) 6884 rfs4_state_close(sp, FALSE, FALSE, cs->cr); 6885 rfs4_state_rele(sp); 6886 /* check if a monitor detected a delegation conflict */ 6887 if (err == EAGAIN && (ct.cc_flags & CC_WOULDBLOCK)) 6888 resp->status = NFS4ERR_DELAY; 6889 else 6890 resp->status = NFS4ERR_SERVERFAULT; 6891 return; 6892 } 6893 } else { /* open upgrade */ 6894 /* 6895 * calculate the fflags for the new mode that is being added 6896 * by this upgrade. 6897 */ 6898 fflags = 0; 6899 if (open_a & OPEN4_SHARE_ACCESS_READ) 6900 fflags |= FREAD; 6901 if (open_a & OPEN4_SHARE_ACCESS_WRITE) 6902 fflags |= FWRITE; 6903 vn_open_upgrade(cs->vp, fflags); 6904 } 6905 sp->rs_open_access |= access; 6906 sp->rs_open_deny |= deny; 6907 6908 if (open_d & OPEN4_SHARE_DENY_READ) 6909 fp->rf_deny_read++; 6910 if (open_d & OPEN4_SHARE_DENY_WRITE) 6911 fp->rf_deny_write++; 6912 fp->rf_share_deny |= deny; 6913 6914 if (open_a & OPEN4_SHARE_ACCESS_READ) 6915 fp->rf_access_read++; 6916 if (open_a & OPEN4_SHARE_ACCESS_WRITE) 6917 fp->rf_access_write++; 6918 fp->rf_share_access |= access; 6919 6920 /* 6921 * Check for delegation here. if the deleg argument is not 6922 * DELEG_ANY, then this is a reclaim from a client and 6923 * we must honor the delegation requested. If necessary we can 6924 * set the recall flag. 6925 */ 6926 6927 dsp = rfs4_grant_delegation(deleg, sp, &recall); 6928 6929 cs->deleg = (fp->rf_dinfo.rd_dtype == OPEN_DELEGATE_WRITE); 6930 6931 next_stateid(&sp->rs_stateid); 6932 6933 resp->stateid = sp->rs_stateid.stateid; 6934 6935 rfs4_dbe_unlock(fp->rf_dbe); 6936 rfs4_dbe_unlock(sp->rs_dbe); 6937 6938 if (dsp) { 6939 rfs4_set_deleg_response(dsp, &resp->delegation, NULL, recall); 6940 rfs4_deleg_state_rele(dsp); 6941 } 6942 6943 rfs4_file_rele(fp); 6944 rfs4_state_rele(sp); 6945 6946 resp->status = NFS4_OK; 6947 } 6948 6949 /*ARGSUSED*/ 6950 static void 6951 rfs4_do_opennull(struct compound_state *cs, struct svc_req *req, 6952 OPEN4args *args, rfs4_openowner_t *oo, OPEN4res *resp) 6953 { 6954 change_info4 *cinfo = &resp->cinfo; 6955 bitmap4 *attrset = &resp->attrset; 6956 6957 if (args->opentype == OPEN4_NOCREATE) 6958 resp->status = rfs4_lookupfile(&args->open_claim4_u.file, 6959 req, cs, args->share_access, cinfo); 6960 else { 6961 /* inhibit delegation grants during exclusive create */ 6962 6963 if (args->mode == EXCLUSIVE4) 6964 rfs4_disable_delegation(); 6965 6966 resp->status = rfs4_createfile(args, req, cs, cinfo, attrset, 6967 oo->ro_client->rc_clientid); 6968 } 6969 6970 if (resp->status == NFS4_OK) { 6971 6972 /* cs->vp cs->fh now reference the desired file */ 6973 6974 rfs4_do_open(cs, req, oo, 6975 oo->ro_need_confirm ? DELEG_NONE : DELEG_ANY, 6976 args->share_access, args->share_deny, resp, 0); 6977 6978 /* 6979 * If rfs4_createfile set attrset, we must 6980 * clear this attrset before the response is copied. 6981 */ 6982 if (resp->status != NFS4_OK && resp->attrset) { 6983 resp->attrset = 0; 6984 } 6985 } 6986 else 6987 *cs->statusp = resp->status; 6988 6989 if (args->mode == EXCLUSIVE4) 6990 rfs4_enable_delegation(); 6991 } 6992 6993 /*ARGSUSED*/ 6994 static void 6995 rfs4_do_openprev(struct compound_state *cs, struct svc_req *req, 6996 OPEN4args *args, rfs4_openowner_t *oo, OPEN4res *resp) 6997 { 6998 change_info4 *cinfo = &resp->cinfo; 6999 vattr_t va; 7000 vtype_t v_type = cs->vp->v_type; 7001 int error = 0; 7002 7003 /* Verify that we have a regular file */ 7004 if (v_type != VREG) { 7005 if (v_type == VDIR) 7006 resp->status = NFS4ERR_ISDIR; 7007 else if (v_type == VLNK) 7008 resp->status = NFS4ERR_SYMLINK; 7009 else 7010 resp->status = NFS4ERR_INVAL; 7011 return; 7012 } 7013 7014 va.va_mask = AT_MODE|AT_UID; 7015 error = VOP_GETATTR(cs->vp, &va, 0, cs->cr, NULL); 7016 if (error) { 7017 resp->status = puterrno4(error); 7018 return; 7019 } 7020 7021 cs->mandlock = MANDLOCK(cs->vp, va.va_mode); 7022 7023 /* 7024 * Check if we have access to the file, Note the the file 7025 * could have originally been open UNCHECKED or GUARDED 7026 * with mode bits that will now fail, but there is nothing 7027 * we can really do about that except in the case that the 7028 * owner of the file is the one requesting the open. 7029 */ 7030 if (crgetuid(cs->cr) != va.va_uid) { 7031 resp->status = check_open_access(args->share_access, cs, req); 7032 if (resp->status != NFS4_OK) { 7033 return; 7034 } 7035 } 7036 7037 /* 7038 * cinfo on a CLAIM_PREVIOUS is undefined, initialize to zero 7039 */ 7040 cinfo->before = 0; 7041 cinfo->after = 0; 7042 cinfo->atomic = FALSE; 7043 7044 rfs4_do_open(cs, req, oo, 7045 NFS4_DELEG4TYPE2REQTYPE(args->open_claim4_u.delegate_type), 7046 args->share_access, args->share_deny, resp, 0); 7047 } 7048 7049 static void 7050 rfs4_do_opendelcur(struct compound_state *cs, struct svc_req *req, 7051 OPEN4args *args, rfs4_openowner_t *oo, OPEN4res *resp) 7052 { 7053 int error; 7054 nfsstat4 status; 7055 stateid4 stateid = 7056 args->open_claim4_u.delegate_cur_info.delegate_stateid; 7057 rfs4_deleg_state_t *dsp; 7058 7059 /* 7060 * Find the state info from the stateid and confirm that the 7061 * file is delegated. If the state openowner is the same as 7062 * the supplied openowner we're done. If not, get the file 7063 * info from the found state info. Use that file info to 7064 * create the state for this lock owner. Note solaris doen't 7065 * really need the pathname to find the file. We may want to 7066 * lookup the pathname and make sure that the vp exist and 7067 * matches the vp in the file structure. However it is 7068 * possible that the pathname nolonger exists (local process 7069 * unlinks the file), so this may not be that useful. 7070 */ 7071 7072 status = rfs4_get_deleg_state(&stateid, &dsp); 7073 if (status != NFS4_OK) { 7074 resp->status = status; 7075 return; 7076 } 7077 7078 ASSERT(dsp->rds_finfo->rf_dinfo.rd_dtype != OPEN_DELEGATE_NONE); 7079 7080 /* 7081 * New lock owner, create state. Since this was probably called 7082 * in response to a CB_RECALL we set deleg to DELEG_NONE 7083 */ 7084 7085 ASSERT(cs->vp != NULL); 7086 VN_RELE(cs->vp); 7087 VN_HOLD(dsp->rds_finfo->rf_vp); 7088 cs->vp = dsp->rds_finfo->rf_vp; 7089 7090 if (error = makefh4(&cs->fh, cs->vp, cs->exi)) { 7091 rfs4_deleg_state_rele(dsp); 7092 *cs->statusp = resp->status = puterrno4(error); 7093 return; 7094 } 7095 7096 /* Mark progress for delegation returns */ 7097 dsp->rds_finfo->rf_dinfo.rd_time_lastwrite = gethrestime_sec(); 7098 rfs4_deleg_state_rele(dsp); 7099 rfs4_do_open(cs, req, oo, DELEG_NONE, 7100 args->share_access, args->share_deny, resp, 1); 7101 } 7102 7103 /*ARGSUSED*/ 7104 static void 7105 rfs4_do_opendelprev(struct compound_state *cs, struct svc_req *req, 7106 OPEN4args *args, rfs4_openowner_t *oo, OPEN4res *resp) 7107 { 7108 /* 7109 * Lookup the pathname, it must already exist since this file 7110 * was delegated. 7111 * 7112 * Find the file and state info for this vp and open owner pair. 7113 * check that they are in fact delegated. 7114 * check that the state access and deny modes are the same. 7115 * 7116 * Return the delgation possibly seting the recall flag. 7117 */ 7118 rfs4_file_t *fp; 7119 rfs4_state_t *sp; 7120 bool_t create = FALSE; 7121 bool_t dcreate = FALSE; 7122 rfs4_deleg_state_t *dsp; 7123 nfsace4 *ace; 7124 7125 /* Note we ignore oflags */ 7126 resp->status = rfs4_lookupfile(&args->open_claim4_u.file_delegate_prev, 7127 req, cs, args->share_access, &resp->cinfo); 7128 7129 if (resp->status != NFS4_OK) { 7130 return; 7131 } 7132 7133 /* get the file struct and hold a lock on it during initial open */ 7134 fp = rfs4_findfile_withlock(cs->vp, NULL, &create); 7135 if (fp == NULL) { 7136 resp->status = NFS4ERR_RESOURCE; 7137 DTRACE_PROBE1(nfss__e__do_opendelprev1, nfsstat4, resp->status); 7138 return; 7139 } 7140 7141 sp = rfs4_findstate_by_owner_file(oo, fp, &create); 7142 if (sp == NULL) { 7143 resp->status = NFS4ERR_SERVERFAULT; 7144 DTRACE_PROBE1(nfss__e__do_opendelprev2, nfsstat4, resp->status); 7145 rw_exit(&fp->rf_file_rwlock); 7146 rfs4_file_rele(fp); 7147 return; 7148 } 7149 7150 rfs4_dbe_lock(sp->rs_dbe); 7151 rfs4_dbe_lock(fp->rf_dbe); 7152 if (args->share_access != sp->rs_share_access || 7153 args->share_deny != sp->rs_share_deny || 7154 sp->rs_finfo->rf_dinfo.rd_dtype == OPEN_DELEGATE_NONE) { 7155 NFS4_DEBUG(rfs4_debug, 7156 (CE_NOTE, "rfs4_do_opendelprev: state mixup")); 7157 rfs4_dbe_unlock(fp->rf_dbe); 7158 rfs4_dbe_unlock(sp->rs_dbe); 7159 rfs4_file_rele(fp); 7160 rfs4_state_rele(sp); 7161 resp->status = NFS4ERR_SERVERFAULT; 7162 return; 7163 } 7164 rfs4_dbe_unlock(fp->rf_dbe); 7165 rfs4_dbe_unlock(sp->rs_dbe); 7166 7167 dsp = rfs4_finddeleg(sp, &dcreate); 7168 if (dsp == NULL) { 7169 rfs4_state_rele(sp); 7170 rfs4_file_rele(fp); 7171 resp->status = NFS4ERR_SERVERFAULT; 7172 return; 7173 } 7174 7175 next_stateid(&sp->rs_stateid); 7176 7177 resp->stateid = sp->rs_stateid.stateid; 7178 7179 resp->delegation.delegation_type = dsp->rds_dtype; 7180 7181 if (dsp->rds_dtype == OPEN_DELEGATE_READ) { 7182 open_read_delegation4 *rv = 7183 &resp->delegation.open_delegation4_u.read; 7184 7185 rv->stateid = dsp->rds_delegid.stateid; 7186 rv->recall = FALSE; /* no policy in place to set to TRUE */ 7187 ace = &rv->permissions; 7188 } else { 7189 open_write_delegation4 *rv = 7190 &resp->delegation.open_delegation4_u.write; 7191 7192 rv->stateid = dsp->rds_delegid.stateid; 7193 rv->recall = FALSE; /* no policy in place to set to TRUE */ 7194 ace = &rv->permissions; 7195 rv->space_limit.limitby = NFS_LIMIT_SIZE; 7196 rv->space_limit.nfs_space_limit4_u.filesize = UINT64_MAX; 7197 } 7198 7199 /* XXX For now */ 7200 ace->type = ACE4_ACCESS_ALLOWED_ACE_TYPE; 7201 ace->flag = 0; 7202 ace->access_mask = 0; 7203 ace->who.utf8string_len = 0; 7204 ace->who.utf8string_val = 0; 7205 7206 rfs4_deleg_state_rele(dsp); 7207 rfs4_state_rele(sp); 7208 rfs4_file_rele(fp); 7209 } 7210 7211 typedef enum { 7212 NFS4_CHKSEQ_OKAY = 0, 7213 NFS4_CHKSEQ_REPLAY = 1, 7214 NFS4_CHKSEQ_BAD = 2 7215 } rfs4_chkseq_t; 7216 7217 /* 7218 * Generic function for sequence number checks. 7219 */ 7220 static rfs4_chkseq_t 7221 rfs4_check_seqid(seqid4 seqid, nfs_resop4 *lastop, 7222 seqid4 rqst_seq, nfs_resop4 *resop, bool_t copyres) 7223 { 7224 /* Same sequence ids and matching operations? */ 7225 if (seqid == rqst_seq && resop->resop == lastop->resop) { 7226 if (copyres == TRUE) { 7227 rfs4_free_reply(resop); 7228 rfs4_copy_reply(resop, lastop); 7229 } 7230 NFS4_DEBUG(rfs4_debug, (CE_NOTE, 7231 "Replayed SEQID %d\n", seqid)); 7232 return (NFS4_CHKSEQ_REPLAY); 7233 } 7234 7235 /* If the incoming sequence is not the next expected then it is bad */ 7236 if (rqst_seq != seqid + 1) { 7237 if (rqst_seq == seqid) { 7238 NFS4_DEBUG(rfs4_debug, 7239 (CE_NOTE, "BAD SEQID: Replayed sequence id " 7240 "but last op was %d current op is %d\n", 7241 lastop->resop, resop->resop)); 7242 return (NFS4_CHKSEQ_BAD); 7243 } 7244 NFS4_DEBUG(rfs4_debug, 7245 (CE_NOTE, "BAD SEQID: got %u expecting %u\n", 7246 rqst_seq, seqid)); 7247 return (NFS4_CHKSEQ_BAD); 7248 } 7249 7250 /* Everything okay -- next expected */ 7251 return (NFS4_CHKSEQ_OKAY); 7252 } 7253 7254 7255 static rfs4_chkseq_t 7256 rfs4_check_open_seqid(seqid4 seqid, rfs4_openowner_t *op, nfs_resop4 *resop) 7257 { 7258 rfs4_chkseq_t rc; 7259 7260 rfs4_dbe_lock(op->ro_dbe); 7261 rc = rfs4_check_seqid(op->ro_open_seqid, &op->ro_reply, seqid, resop, 7262 TRUE); 7263 rfs4_dbe_unlock(op->ro_dbe); 7264 7265 if (rc == NFS4_CHKSEQ_OKAY) 7266 rfs4_update_lease(op->ro_client); 7267 7268 return (rc); 7269 } 7270 7271 static rfs4_chkseq_t 7272 rfs4_check_olo_seqid(seqid4 olo_seqid, rfs4_openowner_t *op, nfs_resop4 *resop) 7273 { 7274 rfs4_chkseq_t rc; 7275 7276 rfs4_dbe_lock(op->ro_dbe); 7277 rc = rfs4_check_seqid(op->ro_open_seqid, &op->ro_reply, 7278 olo_seqid, resop, FALSE); 7279 rfs4_dbe_unlock(op->ro_dbe); 7280 7281 return (rc); 7282 } 7283 7284 static rfs4_chkseq_t 7285 rfs4_check_lock_seqid(seqid4 seqid, rfs4_lo_state_t *lsp, nfs_resop4 *resop) 7286 { 7287 rfs4_chkseq_t rc = NFS4_CHKSEQ_OKAY; 7288 7289 rfs4_dbe_lock(lsp->rls_dbe); 7290 if (!lsp->rls_skip_seqid_check) 7291 rc = rfs4_check_seqid(lsp->rls_seqid, &lsp->rls_reply, seqid, 7292 resop, TRUE); 7293 rfs4_dbe_unlock(lsp->rls_dbe); 7294 7295 return (rc); 7296 } 7297 7298 static void 7299 rfs4_op_open(nfs_argop4 *argop, nfs_resop4 *resop, 7300 struct svc_req *req, struct compound_state *cs) 7301 { 7302 OPEN4args *args = &argop->nfs_argop4_u.opopen; 7303 OPEN4res *resp = &resop->nfs_resop4_u.opopen; 7304 open_owner4 *owner = &args->owner; 7305 open_claim_type4 claim = args->claim; 7306 rfs4_client_t *cp; 7307 rfs4_openowner_t *oo; 7308 bool_t create; 7309 bool_t replay = FALSE; 7310 int can_reclaim; 7311 7312 DTRACE_NFSV4_2(op__open__start, struct compound_state *, cs, 7313 OPEN4args *, args); 7314 7315 if (cs->vp == NULL) { 7316 *cs->statusp = resp->status = NFS4ERR_NOFILEHANDLE; 7317 goto end; 7318 } 7319 7320 /* 7321 * Need to check clientid and lease expiration first based on 7322 * error ordering and incrementing sequence id. 7323 */ 7324 cp = rfs4_findclient_by_id(owner->clientid, FALSE); 7325 if (cp == NULL) { 7326 *cs->statusp = resp->status = 7327 rfs4_check_clientid(&owner->clientid, 0); 7328 goto end; 7329 } 7330 7331 if (rfs4_lease_expired(cp)) { 7332 rfs4_client_close(cp); 7333 *cs->statusp = resp->status = NFS4ERR_EXPIRED; 7334 goto end; 7335 } 7336 can_reclaim = cp->rc_can_reclaim; 7337 7338 /* 7339 * Find the open_owner for use from this point forward. Take 7340 * care in updating the sequence id based on the type of error 7341 * being returned. 7342 */ 7343 retry: 7344 create = TRUE; 7345 oo = rfs4_findopenowner(owner, &create, args->seqid); 7346 if (oo == NULL) { 7347 *cs->statusp = resp->status = NFS4ERR_STALE_CLIENTID; 7348 rfs4_client_rele(cp); 7349 goto end; 7350 } 7351 7352 /* Hold off access to the sequence space while the open is done */ 7353 rfs4_sw_enter(&oo->ro_sw); 7354 7355 /* 7356 * If the open_owner existed before at the server, then check 7357 * the sequence id. 7358 */ 7359 if (!create && !oo->ro_postpone_confirm) { 7360 switch (rfs4_check_open_seqid(args->seqid, oo, resop)) { 7361 case NFS4_CHKSEQ_BAD: 7362 if ((args->seqid > oo->ro_open_seqid) && 7363 oo->ro_need_confirm) { 7364 rfs4_free_opens(oo, TRUE, FALSE); 7365 rfs4_sw_exit(&oo->ro_sw); 7366 rfs4_openowner_rele(oo); 7367 goto retry; 7368 } 7369 resp->status = NFS4ERR_BAD_SEQID; 7370 goto out; 7371 case NFS4_CHKSEQ_REPLAY: /* replay of previous request */ 7372 replay = TRUE; 7373 goto out; 7374 default: 7375 break; 7376 } 7377 7378 /* 7379 * Sequence was ok and open owner exists 7380 * check to see if we have yet to see an 7381 * open_confirm. 7382 */ 7383 if (oo->ro_need_confirm) { 7384 rfs4_free_opens(oo, TRUE, FALSE); 7385 rfs4_sw_exit(&oo->ro_sw); 7386 rfs4_openowner_rele(oo); 7387 goto retry; 7388 } 7389 } 7390 /* Grace only applies to regular-type OPENs */ 7391 if (rfs4_clnt_in_grace(cp) && 7392 (claim == CLAIM_NULL || claim == CLAIM_DELEGATE_CUR)) { 7393 *cs->statusp = resp->status = NFS4ERR_GRACE; 7394 goto out; 7395 } 7396 7397 /* 7398 * If previous state at the server existed then can_reclaim 7399 * will be set. If not reply NFS4ERR_NO_GRACE to the 7400 * client. 7401 */ 7402 if (rfs4_clnt_in_grace(cp) && claim == CLAIM_PREVIOUS && !can_reclaim) { 7403 *cs->statusp = resp->status = NFS4ERR_NO_GRACE; 7404 goto out; 7405 } 7406 7407 7408 /* 7409 * Reject the open if the client has missed the grace period 7410 */ 7411 if (!rfs4_clnt_in_grace(cp) && claim == CLAIM_PREVIOUS) { 7412 *cs->statusp = resp->status = NFS4ERR_NO_GRACE; 7413 goto out; 7414 } 7415 7416 /* Couple of up-front bookkeeping items */ 7417 if (oo->ro_need_confirm) { 7418 /* 7419 * If this is a reclaim OPEN then we should not ask 7420 * for a confirmation of the open_owner per the 7421 * protocol specification. 7422 */ 7423 if (claim == CLAIM_PREVIOUS) 7424 oo->ro_need_confirm = FALSE; 7425 else 7426 resp->rflags |= OPEN4_RESULT_CONFIRM; 7427 } 7428 resp->rflags |= OPEN4_RESULT_LOCKTYPE_POSIX; 7429 7430 /* 7431 * If there is an unshared filesystem mounted on this vnode, 7432 * do not allow to open/create in this directory. 7433 */ 7434 if (vn_ismntpt(cs->vp)) { 7435 *cs->statusp = resp->status = NFS4ERR_ACCESS; 7436 goto out; 7437 } 7438 7439 /* 7440 * access must READ, WRITE, or BOTH. No access is invalid. 7441 * deny can be READ, WRITE, BOTH, or NONE. 7442 * bits not defined for access/deny are invalid. 7443 */ 7444 if (! (args->share_access & OPEN4_SHARE_ACCESS_BOTH) || 7445 (args->share_access & ~OPEN4_SHARE_ACCESS_BOTH) || 7446 (args->share_deny & ~OPEN4_SHARE_DENY_BOTH)) { 7447 *cs->statusp = resp->status = NFS4ERR_INVAL; 7448 goto out; 7449 } 7450 7451 7452 /* 7453 * make sure attrset is zero before response is built. 7454 */ 7455 resp->attrset = 0; 7456 7457 switch (claim) { 7458 case CLAIM_NULL: 7459 rfs4_do_opennull(cs, req, args, oo, resp); 7460 break; 7461 case CLAIM_PREVIOUS: 7462 rfs4_do_openprev(cs, req, args, oo, resp); 7463 break; 7464 case CLAIM_DELEGATE_CUR: 7465 rfs4_do_opendelcur(cs, req, args, oo, resp); 7466 break; 7467 case CLAIM_DELEGATE_PREV: 7468 rfs4_do_opendelprev(cs, req, args, oo, resp); 7469 break; 7470 default: 7471 resp->status = NFS4ERR_INVAL; 7472 break; 7473 } 7474 7475 out: 7476 rfs4_client_rele(cp); 7477 7478 /* Catch sequence id handling here to make it a little easier */ 7479 switch (resp->status) { 7480 case NFS4ERR_BADXDR: 7481 case NFS4ERR_BAD_SEQID: 7482 case NFS4ERR_BAD_STATEID: 7483 case NFS4ERR_NOFILEHANDLE: 7484 case NFS4ERR_RESOURCE: 7485 case NFS4ERR_STALE_CLIENTID: 7486 case NFS4ERR_STALE_STATEID: 7487 /* 7488 * The protocol states that if any of these errors are 7489 * being returned, the sequence id should not be 7490 * incremented. Any other return requires an 7491 * increment. 7492 */ 7493 break; 7494 default: 7495 /* Always update the lease in this case */ 7496 rfs4_update_lease(oo->ro_client); 7497 7498 /* Regular response - copy the result */ 7499 if (!replay) 7500 rfs4_update_open_resp(oo, resop, &cs->fh); 7501 7502 /* 7503 * REPLAY case: Only if the previous response was OK 7504 * do we copy the filehandle. If not OK, no 7505 * filehandle to copy. 7506 */ 7507 if (replay == TRUE && 7508 resp->status == NFS4_OK && 7509 oo->ro_reply_fh.nfs_fh4_val) { 7510 /* 7511 * If this is a replay, we must restore the 7512 * current filehandle/vp to that of what was 7513 * returned originally. Try our best to do 7514 * it. 7515 */ 7516 nfs_fh4_fmt_t *fh_fmtp = 7517 (nfs_fh4_fmt_t *)oo->ro_reply_fh.nfs_fh4_val; 7518 7519 cs->exi = checkexport4(&fh_fmtp->fh4_fsid, 7520 (fid_t *)&fh_fmtp->fh4_xlen, NULL); 7521 7522 if (cs->exi == NULL) { 7523 resp->status = NFS4ERR_STALE; 7524 goto finish; 7525 } 7526 7527 VN_RELE(cs->vp); 7528 7529 cs->vp = nfs4_fhtovp(&oo->ro_reply_fh, cs->exi, 7530 &resp->status); 7531 7532 if (cs->vp == NULL) 7533 goto finish; 7534 7535 nfs_fh4_copy(&oo->ro_reply_fh, &cs->fh); 7536 } 7537 7538 /* 7539 * If this was a replay, no need to update the 7540 * sequence id. If the open_owner was not created on 7541 * this pass, then update. The first use of an 7542 * open_owner will not bump the sequence id. 7543 */ 7544 if (replay == FALSE && !create) 7545 rfs4_update_open_sequence(oo); 7546 /* 7547 * If the client is receiving an error and the 7548 * open_owner needs to be confirmed, there is no way 7549 * to notify the client of this fact ignoring the fact 7550 * that the server has no method of returning a 7551 * stateid to confirm. Therefore, the server needs to 7552 * mark this open_owner in a way as to avoid the 7553 * sequence id checking the next time the client uses 7554 * this open_owner. 7555 */ 7556 if (resp->status != NFS4_OK && oo->ro_need_confirm) 7557 oo->ro_postpone_confirm = TRUE; 7558 /* 7559 * If OK response then clear the postpone flag and 7560 * reset the sequence id to keep in sync with the 7561 * client. 7562 */ 7563 if (resp->status == NFS4_OK && oo->ro_postpone_confirm) { 7564 oo->ro_postpone_confirm = FALSE; 7565 oo->ro_open_seqid = args->seqid; 7566 } 7567 break; 7568 } 7569 7570 finish: 7571 *cs->statusp = resp->status; 7572 7573 rfs4_sw_exit(&oo->ro_sw); 7574 rfs4_openowner_rele(oo); 7575 7576 end: 7577 DTRACE_NFSV4_2(op__open__done, struct compound_state *, cs, 7578 OPEN4res *, resp); 7579 } 7580 7581 /*ARGSUSED*/ 7582 void 7583 rfs4_op_open_confirm(nfs_argop4 *argop, nfs_resop4 *resop, 7584 struct svc_req *req, struct compound_state *cs) 7585 { 7586 OPEN_CONFIRM4args *args = &argop->nfs_argop4_u.opopen_confirm; 7587 OPEN_CONFIRM4res *resp = &resop->nfs_resop4_u.opopen_confirm; 7588 rfs4_state_t *sp; 7589 nfsstat4 status; 7590 7591 DTRACE_NFSV4_2(op__open__confirm__start, struct compound_state *, cs, 7592 OPEN_CONFIRM4args *, args); 7593 7594 if (cs->vp == NULL) { 7595 *cs->statusp = resp->status = NFS4ERR_NOFILEHANDLE; 7596 goto out; 7597 } 7598 7599 if (cs->vp->v_type != VREG) { 7600 *cs->statusp = resp->status = 7601 cs->vp->v_type == VDIR ? NFS4ERR_ISDIR : NFS4ERR_INVAL; 7602 return; 7603 } 7604 7605 status = rfs4_get_state(&args->open_stateid, &sp, RFS4_DBS_VALID); 7606 if (status != NFS4_OK) { 7607 *cs->statusp = resp->status = status; 7608 goto out; 7609 } 7610 7611 /* Ensure specified filehandle matches */ 7612 if (cs->vp != sp->rs_finfo->rf_vp) { 7613 rfs4_state_rele(sp); 7614 *cs->statusp = resp->status = NFS4ERR_BAD_STATEID; 7615 goto out; 7616 } 7617 7618 /* hold off other access to open_owner while we tinker */ 7619 rfs4_sw_enter(&sp->rs_owner->ro_sw); 7620 7621 switch (rfs4_check_stateid_seqid(sp, &args->open_stateid)) { 7622 case NFS4_CHECK_STATEID_OKAY: 7623 if (rfs4_check_open_seqid(args->seqid, sp->rs_owner, 7624 resop) != 0) { 7625 *cs->statusp = resp->status = NFS4ERR_BAD_SEQID; 7626 break; 7627 } 7628 /* 7629 * If it is the appropriate stateid and determined to 7630 * be "OKAY" then this means that the stateid does not 7631 * need to be confirmed and the client is in error for 7632 * sending an OPEN_CONFIRM. 7633 */ 7634 *cs->statusp = resp->status = NFS4ERR_BAD_STATEID; 7635 break; 7636 case NFS4_CHECK_STATEID_OLD: 7637 *cs->statusp = resp->status = NFS4ERR_OLD_STATEID; 7638 break; 7639 case NFS4_CHECK_STATEID_BAD: 7640 *cs->statusp = resp->status = NFS4ERR_BAD_STATEID; 7641 break; 7642 case NFS4_CHECK_STATEID_EXPIRED: 7643 *cs->statusp = resp->status = NFS4ERR_EXPIRED; 7644 break; 7645 case NFS4_CHECK_STATEID_CLOSED: 7646 *cs->statusp = resp->status = NFS4ERR_OLD_STATEID; 7647 break; 7648 case NFS4_CHECK_STATEID_REPLAY: 7649 switch (rfs4_check_open_seqid(args->seqid, sp->rs_owner, 7650 resop)) { 7651 case NFS4_CHKSEQ_OKAY: 7652 /* 7653 * This is replayed stateid; if seqid matches 7654 * next expected, then client is using wrong seqid. 7655 */ 7656 /* fall through */ 7657 case NFS4_CHKSEQ_BAD: 7658 *cs->statusp = resp->status = NFS4ERR_BAD_SEQID; 7659 break; 7660 case NFS4_CHKSEQ_REPLAY: 7661 /* 7662 * Note this case is the duplicate case so 7663 * resp->status is already set. 7664 */ 7665 *cs->statusp = resp->status; 7666 rfs4_update_lease(sp->rs_owner->ro_client); 7667 break; 7668 } 7669 break; 7670 case NFS4_CHECK_STATEID_UNCONFIRMED: 7671 if (rfs4_check_open_seqid(args->seqid, sp->rs_owner, 7672 resop) != NFS4_CHKSEQ_OKAY) { 7673 *cs->statusp = resp->status = NFS4ERR_BAD_SEQID; 7674 break; 7675 } 7676 *cs->statusp = resp->status = NFS4_OK; 7677 7678 next_stateid(&sp->rs_stateid); 7679 resp->open_stateid = sp->rs_stateid.stateid; 7680 sp->rs_owner->ro_need_confirm = FALSE; 7681 rfs4_update_lease(sp->rs_owner->ro_client); 7682 rfs4_update_open_sequence(sp->rs_owner); 7683 rfs4_update_open_resp(sp->rs_owner, resop, NULL); 7684 break; 7685 default: 7686 ASSERT(FALSE); 7687 *cs->statusp = resp->status = NFS4ERR_SERVERFAULT; 7688 break; 7689 } 7690 rfs4_sw_exit(&sp->rs_owner->ro_sw); 7691 rfs4_state_rele(sp); 7692 7693 out: 7694 DTRACE_NFSV4_2(op__open__confirm__done, struct compound_state *, cs, 7695 OPEN_CONFIRM4res *, resp); 7696 } 7697 7698 /*ARGSUSED*/ 7699 void 7700 rfs4_op_open_downgrade(nfs_argop4 *argop, nfs_resop4 *resop, 7701 struct svc_req *req, struct compound_state *cs) 7702 { 7703 OPEN_DOWNGRADE4args *args = &argop->nfs_argop4_u.opopen_downgrade; 7704 OPEN_DOWNGRADE4res *resp = &resop->nfs_resop4_u.opopen_downgrade; 7705 uint32_t access = args->share_access; 7706 uint32_t deny = args->share_deny; 7707 nfsstat4 status; 7708 rfs4_state_t *sp; 7709 rfs4_file_t *fp; 7710 int fflags = 0; 7711 7712 DTRACE_NFSV4_2(op__open__downgrade__start, struct compound_state *, cs, 7713 OPEN_DOWNGRADE4args *, args); 7714 7715 if (cs->vp == NULL) { 7716 *cs->statusp = resp->status = NFS4ERR_NOFILEHANDLE; 7717 goto out; 7718 } 7719 7720 if (cs->vp->v_type != VREG) { 7721 *cs->statusp = resp->status = NFS4ERR_INVAL; 7722 return; 7723 } 7724 7725 status = rfs4_get_state(&args->open_stateid, &sp, RFS4_DBS_VALID); 7726 if (status != NFS4_OK) { 7727 *cs->statusp = resp->status = status; 7728 goto out; 7729 } 7730 7731 /* Ensure specified filehandle matches */ 7732 if (cs->vp != sp->rs_finfo->rf_vp) { 7733 rfs4_state_rele(sp); 7734 *cs->statusp = resp->status = NFS4ERR_BAD_STATEID; 7735 goto out; 7736 } 7737 7738 /* hold off other access to open_owner while we tinker */ 7739 rfs4_sw_enter(&sp->rs_owner->ro_sw); 7740 7741 switch (rfs4_check_stateid_seqid(sp, &args->open_stateid)) { 7742 case NFS4_CHECK_STATEID_OKAY: 7743 if (rfs4_check_open_seqid(args->seqid, sp->rs_owner, 7744 resop) != NFS4_CHKSEQ_OKAY) { 7745 *cs->statusp = resp->status = NFS4ERR_BAD_SEQID; 7746 goto end; 7747 } 7748 break; 7749 case NFS4_CHECK_STATEID_OLD: 7750 *cs->statusp = resp->status = NFS4ERR_OLD_STATEID; 7751 goto end; 7752 case NFS4_CHECK_STATEID_BAD: 7753 *cs->statusp = resp->status = NFS4ERR_BAD_STATEID; 7754 goto end; 7755 case NFS4_CHECK_STATEID_EXPIRED: 7756 *cs->statusp = resp->status = NFS4ERR_EXPIRED; 7757 goto end; 7758 case NFS4_CHECK_STATEID_CLOSED: 7759 *cs->statusp = resp->status = NFS4ERR_OLD_STATEID; 7760 goto end; 7761 case NFS4_CHECK_STATEID_UNCONFIRMED: 7762 *cs->statusp = resp->status = NFS4ERR_BAD_STATEID; 7763 goto end; 7764 case NFS4_CHECK_STATEID_REPLAY: 7765 /* Check the sequence id for the open owner */ 7766 switch (rfs4_check_open_seqid(args->seqid, sp->rs_owner, 7767 resop)) { 7768 case NFS4_CHKSEQ_OKAY: 7769 /* 7770 * This is replayed stateid; if seqid matches 7771 * next expected, then client is using wrong seqid. 7772 */ 7773 /* fall through */ 7774 case NFS4_CHKSEQ_BAD: 7775 *cs->statusp = resp->status = NFS4ERR_BAD_SEQID; 7776 goto end; 7777 case NFS4_CHKSEQ_REPLAY: 7778 /* 7779 * Note this case is the duplicate case so 7780 * resp->status is already set. 7781 */ 7782 *cs->statusp = resp->status; 7783 rfs4_update_lease(sp->rs_owner->ro_client); 7784 goto end; 7785 } 7786 break; 7787 default: 7788 ASSERT(FALSE); 7789 break; 7790 } 7791 7792 rfs4_dbe_lock(sp->rs_dbe); 7793 /* 7794 * Check that the new access modes and deny modes are valid. 7795 * Check that no invalid bits are set. 7796 */ 7797 if ((access & ~(OPEN4_SHARE_ACCESS_READ | OPEN4_SHARE_ACCESS_WRITE)) || 7798 (deny & ~(OPEN4_SHARE_DENY_READ | OPEN4_SHARE_DENY_WRITE))) { 7799 *cs->statusp = resp->status = NFS4ERR_INVAL; 7800 rfs4_update_open_sequence(sp->rs_owner); 7801 rfs4_dbe_unlock(sp->rs_dbe); 7802 goto end; 7803 } 7804 7805 /* 7806 * The new modes must be a subset of the current modes and 7807 * the access must specify at least one mode. To test that 7808 * the new mode is a subset of the current modes we bitwise 7809 * AND them together and check that the result equals the new 7810 * mode. For example: 7811 * New mode, access == R and current mode, sp->rs_open_access == RW 7812 * access & sp->rs_open_access == R == access, so the new access mode 7813 * is valid. Consider access == RW, sp->rs_open_access = R 7814 * access & sp->rs_open_access == R != access, so the new access mode 7815 * is invalid. 7816 */ 7817 if ((access & sp->rs_open_access) != access || 7818 (deny & sp->rs_open_deny) != deny || 7819 (access & 7820 (OPEN4_SHARE_ACCESS_READ | OPEN4_SHARE_ACCESS_WRITE)) == 0) { 7821 *cs->statusp = resp->status = NFS4ERR_INVAL; 7822 rfs4_update_open_sequence(sp->rs_owner); 7823 rfs4_dbe_unlock(sp->rs_dbe); 7824 goto end; 7825 } 7826 7827 /* 7828 * Release any share locks associated with this stateID. 7829 * Strictly speaking, this violates the spec because the 7830 * spec effectively requires that open downgrade be atomic. 7831 * At present, fs_shrlock does not have this capability. 7832 */ 7833 (void) rfs4_unshare(sp); 7834 7835 status = rfs4_share(sp, access, deny); 7836 if (status != NFS4_OK) { 7837 *cs->statusp = resp->status = NFS4ERR_SERVERFAULT; 7838 rfs4_update_open_sequence(sp->rs_owner); 7839 rfs4_dbe_unlock(sp->rs_dbe); 7840 goto end; 7841 } 7842 7843 fp = sp->rs_finfo; 7844 rfs4_dbe_lock(fp->rf_dbe); 7845 7846 /* 7847 * If the current mode has deny read and the new mode 7848 * does not, decrement the number of deny read mode bits 7849 * and if it goes to zero turn off the deny read bit 7850 * on the file. 7851 */ 7852 if ((sp->rs_open_deny & OPEN4_SHARE_DENY_READ) && 7853 (deny & OPEN4_SHARE_DENY_READ) == 0) { 7854 fp->rf_deny_read--; 7855 if (fp->rf_deny_read == 0) 7856 fp->rf_share_deny &= ~OPEN4_SHARE_DENY_READ; 7857 } 7858 7859 /* 7860 * If the current mode has deny write and the new mode 7861 * does not, decrement the number of deny write mode bits 7862 * and if it goes to zero turn off the deny write bit 7863 * on the file. 7864 */ 7865 if ((sp->rs_open_deny & OPEN4_SHARE_DENY_WRITE) && 7866 (deny & OPEN4_SHARE_DENY_WRITE) == 0) { 7867 fp->rf_deny_write--; 7868 if (fp->rf_deny_write == 0) 7869 fp->rf_share_deny &= ~OPEN4_SHARE_DENY_WRITE; 7870 } 7871 7872 /* 7873 * If the current mode has access read and the new mode 7874 * does not, decrement the number of access read mode bits 7875 * and if it goes to zero turn off the access read bit 7876 * on the file. set fflags to FREAD for the call to 7877 * vn_open_downgrade(). 7878 */ 7879 if ((sp->rs_open_access & OPEN4_SHARE_ACCESS_READ) && 7880 (access & OPEN4_SHARE_ACCESS_READ) == 0) { 7881 fp->rf_access_read--; 7882 if (fp->rf_access_read == 0) 7883 fp->rf_share_access &= ~OPEN4_SHARE_ACCESS_READ; 7884 fflags |= FREAD; 7885 } 7886 7887 /* 7888 * If the current mode has access write and the new mode 7889 * does not, decrement the number of access write mode bits 7890 * and if it goes to zero turn off the access write bit 7891 * on the file. set fflags to FWRITE for the call to 7892 * vn_open_downgrade(). 7893 */ 7894 if ((sp->rs_open_access & OPEN4_SHARE_ACCESS_WRITE) && 7895 (access & OPEN4_SHARE_ACCESS_WRITE) == 0) { 7896 fp->rf_access_write--; 7897 if (fp->rf_access_write == 0) 7898 fp->rf_share_deny &= ~OPEN4_SHARE_ACCESS_WRITE; 7899 fflags |= FWRITE; 7900 } 7901 7902 /* Check that the file is still accessible */ 7903 ASSERT(fp->rf_share_access); 7904 7905 rfs4_dbe_unlock(fp->rf_dbe); 7906 7907 /* now set the new open access and deny modes */ 7908 sp->rs_open_access = access; 7909 sp->rs_open_deny = deny; 7910 7911 /* 7912 * we successfully downgraded the share lock, now we need to downgrade 7913 * the open. it is possible that the downgrade was only for a deny 7914 * mode and we have nothing else to do. 7915 */ 7916 if ((fflags & (FREAD|FWRITE)) != 0) 7917 vn_open_downgrade(cs->vp, fflags); 7918 7919 /* Update the stateid */ 7920 next_stateid(&sp->rs_stateid); 7921 resp->open_stateid = sp->rs_stateid.stateid; 7922 7923 rfs4_dbe_unlock(sp->rs_dbe); 7924 7925 *cs->statusp = resp->status = NFS4_OK; 7926 /* Update the lease */ 7927 rfs4_update_lease(sp->rs_owner->ro_client); 7928 /* And the sequence */ 7929 rfs4_update_open_sequence(sp->rs_owner); 7930 rfs4_update_open_resp(sp->rs_owner, resop, NULL); 7931 7932 end: 7933 rfs4_sw_exit(&sp->rs_owner->ro_sw); 7934 rfs4_state_rele(sp); 7935 out: 7936 DTRACE_NFSV4_2(op__open__downgrade__done, struct compound_state *, cs, 7937 OPEN_DOWNGRADE4res *, resp); 7938 } 7939 7940 /* 7941 * The logic behind this function is detailed in the NFSv4 RFC in the 7942 * SETCLIENTID operation description under IMPLEMENTATION. Refer to 7943 * that section for explicit guidance to server behavior for 7944 * SETCLIENTID. 7945 */ 7946 void 7947 rfs4_op_setclientid(nfs_argop4 *argop, nfs_resop4 *resop, 7948 struct svc_req *req, struct compound_state *cs) 7949 { 7950 SETCLIENTID4args *args = &argop->nfs_argop4_u.opsetclientid; 7951 SETCLIENTID4res *res = &resop->nfs_resop4_u.opsetclientid; 7952 rfs4_client_t *cp, *newcp, *cp_confirmed, *cp_unconfirmed; 7953 rfs4_clntip_t *ci; 7954 bool_t create; 7955 char *addr, *netid; 7956 int len; 7957 7958 DTRACE_NFSV4_2(op__setclientid__start, struct compound_state *, cs, 7959 SETCLIENTID4args *, args); 7960 retry: 7961 newcp = cp_confirmed = cp_unconfirmed = NULL; 7962 7963 /* 7964 * Save the caller's IP address 7965 */ 7966 args->client.cl_addr = 7967 (struct sockaddr *)svc_getrpccaller(req->rq_xprt)->buf; 7968 7969 /* 7970 * Record if it is a Solaris client that cannot handle referrals. 7971 */ 7972 if (strstr(args->client.id_val, "Solaris") && 7973 !strstr(args->client.id_val, "+referrals")) { 7974 /* Add a "yes, it's downrev" record */ 7975 create = TRUE; 7976 ci = rfs4_find_clntip(args->client.cl_addr, &create); 7977 ASSERT(ci != NULL); 7978 rfs4_dbe_rele(ci->ri_dbe); 7979 } else { 7980 /* Remove any previous record */ 7981 rfs4_invalidate_clntip(args->client.cl_addr); 7982 } 7983 7984 /* 7985 * In search of an EXISTING client matching the incoming 7986 * request to establish a new client identifier at the server 7987 */ 7988 create = TRUE; 7989 cp = rfs4_findclient(&args->client, &create, NULL); 7990 7991 /* Should never happen */ 7992 ASSERT(cp != NULL); 7993 7994 if (cp == NULL) { 7995 *cs->statusp = res->status = NFS4ERR_SERVERFAULT; 7996 goto out; 7997 } 7998 7999 /* 8000 * Easiest case. Client identifier is newly created and is 8001 * unconfirmed. Also note that for this case, no other 8002 * entries exist for the client identifier. Nothing else to 8003 * check. Just setup the response and respond. 8004 */ 8005 if (create) { 8006 *cs->statusp = res->status = NFS4_OK; 8007 res->SETCLIENTID4res_u.resok4.clientid = cp->rc_clientid; 8008 res->SETCLIENTID4res_u.resok4.setclientid_confirm = 8009 cp->rc_confirm_verf; 8010 /* Setup callback information; CB_NULL confirmation later */ 8011 rfs4_client_setcb(cp, &args->callback, args->callback_ident); 8012 8013 rfs4_client_rele(cp); 8014 goto out; 8015 } 8016 8017 /* 8018 * An existing, confirmed client may exist but it may not have 8019 * been active for at least one lease period. If so, then 8020 * "close" the client and create a new client identifier 8021 */ 8022 if (rfs4_lease_expired(cp)) { 8023 rfs4_client_close(cp); 8024 goto retry; 8025 } 8026 8027 if (cp->rc_need_confirm == TRUE) 8028 cp_unconfirmed = cp; 8029 else 8030 cp_confirmed = cp; 8031 8032 cp = NULL; 8033 8034 /* 8035 * We have a confirmed client, now check for an 8036 * unconfimred entry 8037 */ 8038 if (cp_confirmed) { 8039 /* If creds don't match then client identifier is inuse */ 8040 if (!creds_ok(cp_confirmed->rc_cr_set, req, cs)) { 8041 rfs4_cbinfo_t *cbp; 8042 /* 8043 * Some one else has established this client 8044 * id. Try and say * who they are. We will use 8045 * the call back address supplied by * the 8046 * first client. 8047 */ 8048 *cs->statusp = res->status = NFS4ERR_CLID_INUSE; 8049 8050 addr = netid = NULL; 8051 8052 cbp = &cp_confirmed->rc_cbinfo; 8053 if (cbp->cb_callback.cb_location.r_addr && 8054 cbp->cb_callback.cb_location.r_netid) { 8055 cb_client4 *cbcp = &cbp->cb_callback; 8056 8057 len = strlen(cbcp->cb_location.r_addr)+1; 8058 addr = kmem_alloc(len, KM_SLEEP); 8059 bcopy(cbcp->cb_location.r_addr, addr, len); 8060 len = strlen(cbcp->cb_location.r_netid)+1; 8061 netid = kmem_alloc(len, KM_SLEEP); 8062 bcopy(cbcp->cb_location.r_netid, netid, len); 8063 } 8064 8065 res->SETCLIENTID4res_u.client_using.r_addr = addr; 8066 res->SETCLIENTID4res_u.client_using.r_netid = netid; 8067 8068 rfs4_client_rele(cp_confirmed); 8069 } 8070 8071 /* 8072 * Confirmed, creds match, and verifier matches; must 8073 * be an update of the callback info 8074 */ 8075 if (cp_confirmed->rc_nfs_client.verifier == 8076 args->client.verifier) { 8077 /* Setup callback information */ 8078 rfs4_client_setcb(cp_confirmed, &args->callback, 8079 args->callback_ident); 8080 8081 /* everything okay -- move ahead */ 8082 *cs->statusp = res->status = NFS4_OK; 8083 res->SETCLIENTID4res_u.resok4.clientid = 8084 cp_confirmed->rc_clientid; 8085 8086 /* update the confirm_verifier and return it */ 8087 rfs4_client_scv_next(cp_confirmed); 8088 res->SETCLIENTID4res_u.resok4.setclientid_confirm = 8089 cp_confirmed->rc_confirm_verf; 8090 8091 rfs4_client_rele(cp_confirmed); 8092 goto out; 8093 } 8094 8095 /* 8096 * Creds match but the verifier doesn't. Must search 8097 * for an unconfirmed client that would be replaced by 8098 * this request. 8099 */ 8100 create = FALSE; 8101 cp_unconfirmed = rfs4_findclient(&args->client, &create, 8102 cp_confirmed); 8103 } 8104 8105 /* 8106 * At this point, we have taken care of the brand new client 8107 * struct, INUSE case, update of an existing, and confirmed 8108 * client struct. 8109 */ 8110 8111 /* 8112 * check to see if things have changed while we originally 8113 * picked up the client struct. If they have, then return and 8114 * retry the processing of this SETCLIENTID request. 8115 */ 8116 if (cp_unconfirmed) { 8117 rfs4_dbe_lock(cp_unconfirmed->rc_dbe); 8118 if (!cp_unconfirmed->rc_need_confirm) { 8119 rfs4_dbe_unlock(cp_unconfirmed->rc_dbe); 8120 rfs4_client_rele(cp_unconfirmed); 8121 if (cp_confirmed) 8122 rfs4_client_rele(cp_confirmed); 8123 goto retry; 8124 } 8125 /* do away with the old unconfirmed one */ 8126 rfs4_dbe_invalidate(cp_unconfirmed->rc_dbe); 8127 rfs4_dbe_unlock(cp_unconfirmed->rc_dbe); 8128 rfs4_client_rele(cp_unconfirmed); 8129 cp_unconfirmed = NULL; 8130 } 8131 8132 /* 8133 * This search will temporarily hide the confirmed client 8134 * struct while a new client struct is created as the 8135 * unconfirmed one. 8136 */ 8137 create = TRUE; 8138 newcp = rfs4_findclient(&args->client, &create, cp_confirmed); 8139 8140 ASSERT(newcp != NULL); 8141 8142 if (newcp == NULL) { 8143 *cs->statusp = res->status = NFS4ERR_SERVERFAULT; 8144 rfs4_client_rele(cp_confirmed); 8145 goto out; 8146 } 8147 8148 /* 8149 * If one was not created, then a similar request must be in 8150 * process so release and start over with this one 8151 */ 8152 if (create != TRUE) { 8153 rfs4_client_rele(newcp); 8154 if (cp_confirmed) 8155 rfs4_client_rele(cp_confirmed); 8156 goto retry; 8157 } 8158 8159 *cs->statusp = res->status = NFS4_OK; 8160 res->SETCLIENTID4res_u.resok4.clientid = newcp->rc_clientid; 8161 res->SETCLIENTID4res_u.resok4.setclientid_confirm = 8162 newcp->rc_confirm_verf; 8163 /* Setup callback information; CB_NULL confirmation later */ 8164 rfs4_client_setcb(newcp, &args->callback, args->callback_ident); 8165 8166 newcp->rc_cp_confirmed = cp_confirmed; 8167 8168 rfs4_client_rele(newcp); 8169 8170 out: 8171 DTRACE_NFSV4_2(op__setclientid__done, struct compound_state *, cs, 8172 SETCLIENTID4res *, res); 8173 } 8174 8175 /*ARGSUSED*/ 8176 void 8177 rfs4_op_setclientid_confirm(nfs_argop4 *argop, nfs_resop4 *resop, 8178 struct svc_req *req, struct compound_state *cs) 8179 { 8180 SETCLIENTID_CONFIRM4args *args = 8181 &argop->nfs_argop4_u.opsetclientid_confirm; 8182 SETCLIENTID_CONFIRM4res *res = 8183 &resop->nfs_resop4_u.opsetclientid_confirm; 8184 rfs4_client_t *cp, *cptoclose = NULL; 8185 8186 DTRACE_NFSV4_2(op__setclientid__confirm__start, 8187 struct compound_state *, cs, 8188 SETCLIENTID_CONFIRM4args *, args); 8189 8190 *cs->statusp = res->status = NFS4_OK; 8191 8192 cp = rfs4_findclient_by_id(args->clientid, TRUE); 8193 8194 if (cp == NULL) { 8195 *cs->statusp = res->status = 8196 rfs4_check_clientid(&args->clientid, 1); 8197 goto out; 8198 } 8199 8200 if (!creds_ok(cp, req, cs)) { 8201 *cs->statusp = res->status = NFS4ERR_CLID_INUSE; 8202 rfs4_client_rele(cp); 8203 goto out; 8204 } 8205 8206 /* If the verifier doesn't match, the record doesn't match */ 8207 if (cp->rc_confirm_verf != args->setclientid_confirm) { 8208 *cs->statusp = res->status = NFS4ERR_STALE_CLIENTID; 8209 rfs4_client_rele(cp); 8210 goto out; 8211 } 8212 8213 rfs4_dbe_lock(cp->rc_dbe); 8214 cp->rc_need_confirm = FALSE; 8215 if (cp->rc_cp_confirmed) { 8216 cptoclose = cp->rc_cp_confirmed; 8217 cptoclose->rc_ss_remove = 1; 8218 cp->rc_cp_confirmed = NULL; 8219 } 8220 8221 /* 8222 * Update the client's associated server instance, if it's changed 8223 * since the client was created. 8224 */ 8225 if (rfs4_servinst(cp) != rfs4_cur_servinst) 8226 rfs4_servinst_assign(cp, rfs4_cur_servinst); 8227 8228 /* 8229 * Record clientid in stable storage. 8230 * Must be done after server instance has been assigned. 8231 */ 8232 rfs4_ss_clid(cp); 8233 8234 rfs4_dbe_unlock(cp->rc_dbe); 8235 8236 if (cptoclose) 8237 /* don't need to rele, client_close does it */ 8238 rfs4_client_close(cptoclose); 8239 8240 /* If needed, initiate CB_NULL call for callback path */ 8241 rfs4_deleg_cb_check(cp); 8242 rfs4_update_lease(cp); 8243 8244 /* 8245 * Check to see if client can perform reclaims 8246 */ 8247 rfs4_ss_chkclid(cp); 8248 8249 rfs4_client_rele(cp); 8250 8251 out: 8252 DTRACE_NFSV4_2(op__setclientid__confirm__done, 8253 struct compound_state *, cs, 8254 SETCLIENTID_CONFIRM4 *, res); 8255 } 8256 8257 8258 /*ARGSUSED*/ 8259 void 8260 rfs4_op_close(nfs_argop4 *argop, nfs_resop4 *resop, 8261 struct svc_req *req, struct compound_state *cs) 8262 { 8263 CLOSE4args *args = &argop->nfs_argop4_u.opclose; 8264 CLOSE4res *resp = &resop->nfs_resop4_u.opclose; 8265 rfs4_state_t *sp; 8266 nfsstat4 status; 8267 8268 DTRACE_NFSV4_2(op__close__start, struct compound_state *, cs, 8269 CLOSE4args *, args); 8270 8271 if (cs->vp == NULL) { 8272 *cs->statusp = resp->status = NFS4ERR_NOFILEHANDLE; 8273 goto out; 8274 } 8275 8276 status = rfs4_get_state(&args->open_stateid, &sp, RFS4_DBS_INVALID); 8277 if (status != NFS4_OK) { 8278 *cs->statusp = resp->status = status; 8279 goto out; 8280 } 8281 8282 /* Ensure specified filehandle matches */ 8283 if (cs->vp != sp->rs_finfo->rf_vp) { 8284 rfs4_state_rele(sp); 8285 *cs->statusp = resp->status = NFS4ERR_BAD_STATEID; 8286 goto out; 8287 } 8288 8289 /* hold off other access to open_owner while we tinker */ 8290 rfs4_sw_enter(&sp->rs_owner->ro_sw); 8291 8292 switch (rfs4_check_stateid_seqid(sp, &args->open_stateid)) { 8293 case NFS4_CHECK_STATEID_OKAY: 8294 if (rfs4_check_open_seqid(args->seqid, sp->rs_owner, 8295 resop) != NFS4_CHKSEQ_OKAY) { 8296 *cs->statusp = resp->status = NFS4ERR_BAD_SEQID; 8297 goto end; 8298 } 8299 break; 8300 case NFS4_CHECK_STATEID_OLD: 8301 *cs->statusp = resp->status = NFS4ERR_OLD_STATEID; 8302 goto end; 8303 case NFS4_CHECK_STATEID_BAD: 8304 *cs->statusp = resp->status = NFS4ERR_BAD_STATEID; 8305 goto end; 8306 case NFS4_CHECK_STATEID_EXPIRED: 8307 *cs->statusp = resp->status = NFS4ERR_EXPIRED; 8308 goto end; 8309 case NFS4_CHECK_STATEID_CLOSED: 8310 *cs->statusp = resp->status = NFS4ERR_OLD_STATEID; 8311 goto end; 8312 case NFS4_CHECK_STATEID_UNCONFIRMED: 8313 *cs->statusp = resp->status = NFS4ERR_BAD_STATEID; 8314 goto end; 8315 case NFS4_CHECK_STATEID_REPLAY: 8316 /* Check the sequence id for the open owner */ 8317 switch (rfs4_check_open_seqid(args->seqid, sp->rs_owner, 8318 resop)) { 8319 case NFS4_CHKSEQ_OKAY: 8320 /* 8321 * This is replayed stateid; if seqid matches 8322 * next expected, then client is using wrong seqid. 8323 */ 8324 /* FALL THROUGH */ 8325 case NFS4_CHKSEQ_BAD: 8326 *cs->statusp = resp->status = NFS4ERR_BAD_SEQID; 8327 goto end; 8328 case NFS4_CHKSEQ_REPLAY: 8329 /* 8330 * Note this case is the duplicate case so 8331 * resp->status is already set. 8332 */ 8333 *cs->statusp = resp->status; 8334 rfs4_update_lease(sp->rs_owner->ro_client); 8335 goto end; 8336 } 8337 break; 8338 default: 8339 ASSERT(FALSE); 8340 break; 8341 } 8342 8343 rfs4_dbe_lock(sp->rs_dbe); 8344 8345 /* Update the stateid. */ 8346 next_stateid(&sp->rs_stateid); 8347 resp->open_stateid = sp->rs_stateid.stateid; 8348 8349 rfs4_dbe_unlock(sp->rs_dbe); 8350 8351 rfs4_update_lease(sp->rs_owner->ro_client); 8352 rfs4_update_open_sequence(sp->rs_owner); 8353 rfs4_update_open_resp(sp->rs_owner, resop, NULL); 8354 8355 rfs4_state_close(sp, FALSE, FALSE, cs->cr); 8356 8357 *cs->statusp = resp->status = status; 8358 8359 end: 8360 rfs4_sw_exit(&sp->rs_owner->ro_sw); 8361 rfs4_state_rele(sp); 8362 out: 8363 DTRACE_NFSV4_2(op__close__done, struct compound_state *, cs, 8364 CLOSE4res *, resp); 8365 } 8366 8367 /* 8368 * Manage the counts on the file struct and close all file locks 8369 */ 8370 /*ARGSUSED*/ 8371 void 8372 rfs4_release_share_lock_state(rfs4_state_t *sp, cred_t *cr, 8373 bool_t close_of_client) 8374 { 8375 rfs4_file_t *fp = sp->rs_finfo; 8376 rfs4_lo_state_t *lsp; 8377 int fflags = 0; 8378 8379 /* 8380 * If this call is part of the larger closing down of client 8381 * state then it is just easier to release all locks 8382 * associated with this client instead of going through each 8383 * individual file and cleaning locks there. 8384 */ 8385 if (close_of_client) { 8386 if (sp->rs_owner->ro_client->rc_unlksys_completed == FALSE && 8387 !list_is_empty(&sp->rs_lostatelist) && 8388 sp->rs_owner->ro_client->rc_sysidt != LM_NOSYSID) { 8389 /* Is the PxFS kernel module loaded? */ 8390 if (lm_remove_file_locks != NULL) { 8391 int new_sysid; 8392 8393 /* Encode the cluster nodeid in new sysid */ 8394 new_sysid = sp->rs_owner->ro_client->rc_sysidt; 8395 lm_set_nlmid_flk(&new_sysid); 8396 8397 /* 8398 * This PxFS routine removes file locks for a 8399 * client over all nodes of a cluster. 8400 */ 8401 NFS4_DEBUG(rfs4_debug, (CE_NOTE, 8402 "lm_remove_file_locks(sysid=0x%x)\n", 8403 new_sysid)); 8404 (*lm_remove_file_locks)(new_sysid); 8405 } else { 8406 struct flock64 flk; 8407 8408 /* Release all locks for this client */ 8409 flk.l_type = F_UNLKSYS; 8410 flk.l_whence = 0; 8411 flk.l_start = 0; 8412 flk.l_len = 0; 8413 flk.l_sysid = 8414 sp->rs_owner->ro_client->rc_sysidt; 8415 flk.l_pid = 0; 8416 (void) VOP_FRLOCK(sp->rs_finfo->rf_vp, F_SETLK, 8417 &flk, F_REMOTELOCK | FREAD | FWRITE, 8418 (u_offset_t)0, NULL, CRED(), NULL); 8419 } 8420 8421 sp->rs_owner->ro_client->rc_unlksys_completed = TRUE; 8422 } 8423 } 8424 8425 /* 8426 * Release all locks on this file by this lock owner or at 8427 * least mark the locks as having been released 8428 */ 8429 for (lsp = list_head(&sp->rs_lostatelist); lsp != NULL; 8430 lsp = list_next(&sp->rs_lostatelist, lsp)) { 8431 lsp->rls_locks_cleaned = TRUE; 8432 8433 /* Was this already taken care of above? */ 8434 if (!close_of_client && 8435 sp->rs_owner->ro_client->rc_sysidt != LM_NOSYSID) 8436 (void) cleanlocks(sp->rs_finfo->rf_vp, 8437 lsp->rls_locker->rl_pid, 8438 lsp->rls_locker->rl_client->rc_sysidt); 8439 } 8440 8441 /* 8442 * Release any shrlocks associated with this open state ID. 8443 * This must be done before the rfs4_state gets marked closed. 8444 */ 8445 if (sp->rs_owner->ro_client->rc_sysidt != LM_NOSYSID) 8446 (void) rfs4_unshare(sp); 8447 8448 if (sp->rs_open_access) { 8449 rfs4_dbe_lock(fp->rf_dbe); 8450 8451 /* 8452 * Decrement the count for each access and deny bit that this 8453 * state has contributed to the file. 8454 * If the file counts go to zero 8455 * clear the appropriate bit in the appropriate mask. 8456 */ 8457 if (sp->rs_open_access & OPEN4_SHARE_ACCESS_READ) { 8458 fp->rf_access_read--; 8459 fflags |= FREAD; 8460 if (fp->rf_access_read == 0) 8461 fp->rf_share_access &= ~OPEN4_SHARE_ACCESS_READ; 8462 } 8463 if (sp->rs_open_access & OPEN4_SHARE_ACCESS_WRITE) { 8464 fp->rf_access_write--; 8465 fflags |= FWRITE; 8466 if (fp->rf_access_write == 0) 8467 fp->rf_share_access &= 8468 ~OPEN4_SHARE_ACCESS_WRITE; 8469 } 8470 if (sp->rs_open_deny & OPEN4_SHARE_DENY_READ) { 8471 fp->rf_deny_read--; 8472 if (fp->rf_deny_read == 0) 8473 fp->rf_share_deny &= ~OPEN4_SHARE_DENY_READ; 8474 } 8475 if (sp->rs_open_deny & OPEN4_SHARE_DENY_WRITE) { 8476 fp->rf_deny_write--; 8477 if (fp->rf_deny_write == 0) 8478 fp->rf_share_deny &= ~OPEN4_SHARE_DENY_WRITE; 8479 } 8480 8481 (void) VOP_CLOSE(fp->rf_vp, fflags, 1, (offset_t)0, cr, NULL); 8482 8483 rfs4_dbe_unlock(fp->rf_dbe); 8484 8485 sp->rs_open_access = 0; 8486 sp->rs_open_deny = 0; 8487 } 8488 } 8489 8490 /* 8491 * lock_denied: Fill in a LOCK4deneid structure given an flock64 structure. 8492 */ 8493 static nfsstat4 8494 lock_denied(LOCK4denied *dp, struct flock64 *flk) 8495 { 8496 rfs4_lockowner_t *lo; 8497 rfs4_client_t *cp; 8498 uint32_t len; 8499 8500 lo = rfs4_findlockowner_by_pid(flk->l_pid); 8501 if (lo != NULL) { 8502 cp = lo->rl_client; 8503 if (rfs4_lease_expired(cp)) { 8504 rfs4_lockowner_rele(lo); 8505 rfs4_dbe_hold(cp->rc_dbe); 8506 rfs4_client_close(cp); 8507 return (NFS4ERR_EXPIRED); 8508 } 8509 dp->owner.clientid = lo->rl_owner.clientid; 8510 len = lo->rl_owner.owner_len; 8511 dp->owner.owner_val = kmem_alloc(len, KM_SLEEP); 8512 bcopy(lo->rl_owner.owner_val, dp->owner.owner_val, len); 8513 dp->owner.owner_len = len; 8514 rfs4_lockowner_rele(lo); 8515 goto finish; 8516 } 8517 8518 /* 8519 * Its not a NFS4 lock. We take advantage that the upper 32 bits 8520 * of the client id contain the boot time for a NFS4 lock. So we 8521 * fabricate and identity by setting clientid to the sysid, and 8522 * the lock owner to the pid. 8523 */ 8524 dp->owner.clientid = flk->l_sysid; 8525 len = sizeof (pid_t); 8526 dp->owner.owner_len = len; 8527 dp->owner.owner_val = kmem_alloc(len, KM_SLEEP); 8528 bcopy(&flk->l_pid, dp->owner.owner_val, len); 8529 finish: 8530 dp->offset = flk->l_start; 8531 dp->length = flk->l_len; 8532 8533 if (flk->l_type == F_RDLCK) 8534 dp->locktype = READ_LT; 8535 else if (flk->l_type == F_WRLCK) 8536 dp->locktype = WRITE_LT; 8537 else 8538 return (NFS4ERR_INVAL); /* no mapping from POSIX ltype to v4 */ 8539 8540 return (NFS4_OK); 8541 } 8542 8543 static int 8544 setlock(vnode_t *vp, struct flock64 *flock, int flag, cred_t *cred) 8545 { 8546 int error; 8547 struct flock64 flk; 8548 int i; 8549 clock_t delaytime; 8550 int cmd; 8551 8552 cmd = nbl_need_check(vp) ? F_SETLK_NBMAND : F_SETLK; 8553 retry: 8554 delaytime = MSEC_TO_TICK_ROUNDUP(rfs4_lock_delay); 8555 8556 for (i = 0; i < rfs4_maxlock_tries; i++) { 8557 LOCK_PRINT(rfs4_debug, "setlock", cmd, flock); 8558 error = VOP_FRLOCK(vp, cmd, 8559 flock, flag, (u_offset_t)0, NULL, cred, NULL); 8560 8561 if (error != EAGAIN && error != EACCES) 8562 break; 8563 8564 if (i < rfs4_maxlock_tries - 1) { 8565 delay(delaytime); 8566 delaytime *= 2; 8567 } 8568 } 8569 8570 if (error == EAGAIN || error == EACCES) { 8571 /* Get the owner of the lock */ 8572 flk = *flock; 8573 LOCK_PRINT(rfs4_debug, "setlock", F_GETLK, &flk); 8574 if (VOP_FRLOCK(vp, F_GETLK, &flk, flag, 8575 (u_offset_t)0, NULL, cred, NULL) == 0) { 8576 if (flk.l_type == F_UNLCK) { 8577 /* No longer locked, retry */ 8578 goto retry; 8579 } 8580 *flock = flk; 8581 LOCK_PRINT(rfs4_debug, "setlock(blocking lock)", 8582 F_GETLK, &flk); 8583 } 8584 } 8585 8586 return (error); 8587 } 8588 8589 /*ARGSUSED*/ 8590 static nfsstat4 8591 rfs4_do_lock(rfs4_lo_state_t *lsp, nfs_lock_type4 locktype, 8592 offset4 offset, length4 length, cred_t *cred, nfs_resop4 *resop) 8593 { 8594 nfsstat4 status; 8595 rfs4_lockowner_t *lo = lsp->rls_locker; 8596 rfs4_state_t *sp = lsp->rls_state; 8597 struct flock64 flock; 8598 int16_t ltype; 8599 int flag; 8600 int error; 8601 sysid_t sysid; 8602 LOCK4res *lres; 8603 8604 if (rfs4_lease_expired(lo->rl_client)) { 8605 return (NFS4ERR_EXPIRED); 8606 } 8607 8608 if ((status = rfs4_client_sysid(lo->rl_client, &sysid)) != NFS4_OK) 8609 return (status); 8610 8611 /* Check for zero length. To lock to end of file use all ones for V4 */ 8612 if (length == 0) 8613 return (NFS4ERR_INVAL); 8614 else if (length == (length4)(~0)) 8615 length = 0; /* Posix to end of file */ 8616 8617 retry: 8618 rfs4_dbe_lock(sp->rs_dbe); 8619 if (sp->rs_closed) { 8620 rfs4_dbe_unlock(sp->rs_dbe); 8621 return (NFS4ERR_OLD_STATEID); 8622 } 8623 8624 if (resop->resop != OP_LOCKU) { 8625 switch (locktype) { 8626 case READ_LT: 8627 case READW_LT: 8628 if ((sp->rs_share_access 8629 & OPEN4_SHARE_ACCESS_READ) == 0) { 8630 rfs4_dbe_unlock(sp->rs_dbe); 8631 8632 return (NFS4ERR_OPENMODE); 8633 } 8634 ltype = F_RDLCK; 8635 break; 8636 case WRITE_LT: 8637 case WRITEW_LT: 8638 if ((sp->rs_share_access 8639 & OPEN4_SHARE_ACCESS_WRITE) == 0) { 8640 rfs4_dbe_unlock(sp->rs_dbe); 8641 8642 return (NFS4ERR_OPENMODE); 8643 } 8644 ltype = F_WRLCK; 8645 break; 8646 } 8647 } else 8648 ltype = F_UNLCK; 8649 8650 flock.l_type = ltype; 8651 flock.l_whence = 0; /* SEEK_SET */ 8652 flock.l_start = offset; 8653 flock.l_len = length; 8654 flock.l_sysid = sysid; 8655 flock.l_pid = lsp->rls_locker->rl_pid; 8656 8657 /* Note that length4 is uint64_t but l_len and l_start are off64_t */ 8658 if (flock.l_len < 0 || flock.l_start < 0) { 8659 rfs4_dbe_unlock(sp->rs_dbe); 8660 return (NFS4ERR_INVAL); 8661 } 8662 8663 /* 8664 * N.B. FREAD has the same value as OPEN4_SHARE_ACCESS_READ and 8665 * FWRITE has the same value as OPEN4_SHARE_ACCESS_WRITE. 8666 */ 8667 flag = (int)sp->rs_share_access | F_REMOTELOCK; 8668 8669 error = setlock(sp->rs_finfo->rf_vp, &flock, flag, cred); 8670 if (error == 0) { 8671 rfs4_dbe_lock(lsp->rls_dbe); 8672 next_stateid(&lsp->rls_lockid); 8673 rfs4_dbe_unlock(lsp->rls_dbe); 8674 } 8675 8676 rfs4_dbe_unlock(sp->rs_dbe); 8677 8678 /* 8679 * N.B. We map error values to nfsv4 errors. This is differrent 8680 * than puterrno4 routine. 8681 */ 8682 switch (error) { 8683 case 0: 8684 status = NFS4_OK; 8685 break; 8686 case EAGAIN: 8687 case EACCES: /* Old value */ 8688 /* Can only get here if op is OP_LOCK */ 8689 ASSERT(resop->resop == OP_LOCK); 8690 lres = &resop->nfs_resop4_u.oplock; 8691 status = NFS4ERR_DENIED; 8692 if (lock_denied(&lres->LOCK4res_u.denied, &flock) 8693 == NFS4ERR_EXPIRED) 8694 goto retry; 8695 break; 8696 case ENOLCK: 8697 status = NFS4ERR_DELAY; 8698 break; 8699 case EOVERFLOW: 8700 status = NFS4ERR_INVAL; 8701 break; 8702 case EINVAL: 8703 status = NFS4ERR_NOTSUPP; 8704 break; 8705 default: 8706 status = NFS4ERR_SERVERFAULT; 8707 break; 8708 } 8709 8710 return (status); 8711 } 8712 8713 /*ARGSUSED*/ 8714 void 8715 rfs4_op_lock(nfs_argop4 *argop, nfs_resop4 *resop, 8716 struct svc_req *req, struct compound_state *cs) 8717 { 8718 LOCK4args *args = &argop->nfs_argop4_u.oplock; 8719 LOCK4res *resp = &resop->nfs_resop4_u.oplock; 8720 nfsstat4 status; 8721 stateid4 *stateid; 8722 rfs4_lockowner_t *lo; 8723 rfs4_client_t *cp; 8724 rfs4_state_t *sp = NULL; 8725 rfs4_lo_state_t *lsp = NULL; 8726 bool_t ls_sw_held = FALSE; 8727 bool_t create = TRUE; 8728 bool_t lcreate = TRUE; 8729 bool_t dup_lock = FALSE; 8730 int rc; 8731 8732 DTRACE_NFSV4_2(op__lock__start, struct compound_state *, cs, 8733 LOCK4args *, args); 8734 8735 if (cs->vp == NULL) { 8736 *cs->statusp = resp->status = NFS4ERR_NOFILEHANDLE; 8737 DTRACE_NFSV4_2(op__lock__done, struct compound_state *, 8738 cs, LOCK4res *, resp); 8739 return; 8740 } 8741 8742 if (args->locker.new_lock_owner) { 8743 /* Create a new lockowner for this instance */ 8744 open_to_lock_owner4 *olo = &args->locker.locker4_u.open_owner; 8745 8746 NFS4_DEBUG(rfs4_debug, (CE_NOTE, "Creating new lock owner")); 8747 8748 stateid = &olo->open_stateid; 8749 status = rfs4_get_state(stateid, &sp, RFS4_DBS_VALID); 8750 if (status != NFS4_OK) { 8751 NFS4_DEBUG(rfs4_debug, 8752 (CE_NOTE, "Get state failed in lock %d", status)); 8753 *cs->statusp = resp->status = status; 8754 DTRACE_NFSV4_2(op__lock__done, struct compound_state *, 8755 cs, LOCK4res *, resp); 8756 return; 8757 } 8758 8759 /* Ensure specified filehandle matches */ 8760 if (cs->vp != sp->rs_finfo->rf_vp) { 8761 rfs4_state_rele(sp); 8762 *cs->statusp = resp->status = NFS4ERR_BAD_STATEID; 8763 DTRACE_NFSV4_2(op__lock__done, struct compound_state *, 8764 cs, LOCK4res *, resp); 8765 return; 8766 } 8767 8768 /* hold off other access to open_owner while we tinker */ 8769 rfs4_sw_enter(&sp->rs_owner->ro_sw); 8770 8771 switch (rc = rfs4_check_stateid_seqid(sp, stateid)) { 8772 case NFS4_CHECK_STATEID_OLD: 8773 *cs->statusp = resp->status = NFS4ERR_OLD_STATEID; 8774 goto end; 8775 case NFS4_CHECK_STATEID_BAD: 8776 *cs->statusp = resp->status = NFS4ERR_BAD_STATEID; 8777 goto end; 8778 case NFS4_CHECK_STATEID_EXPIRED: 8779 *cs->statusp = resp->status = NFS4ERR_EXPIRED; 8780 goto end; 8781 case NFS4_CHECK_STATEID_UNCONFIRMED: 8782 *cs->statusp = resp->status = NFS4ERR_BAD_STATEID; 8783 goto end; 8784 case NFS4_CHECK_STATEID_CLOSED: 8785 *cs->statusp = resp->status = NFS4ERR_OLD_STATEID; 8786 goto end; 8787 case NFS4_CHECK_STATEID_OKAY: 8788 case NFS4_CHECK_STATEID_REPLAY: 8789 switch (rfs4_check_olo_seqid(olo->open_seqid, 8790 sp->rs_owner, resop)) { 8791 case NFS4_CHKSEQ_OKAY: 8792 if (rc == NFS4_CHECK_STATEID_OKAY) 8793 break; 8794 /* 8795 * This is replayed stateid; if seqid 8796 * matches next expected, then client 8797 * is using wrong seqid. 8798 */ 8799 /* FALLTHROUGH */ 8800 case NFS4_CHKSEQ_BAD: 8801 *cs->statusp = resp->status = NFS4ERR_BAD_SEQID; 8802 goto end; 8803 case NFS4_CHKSEQ_REPLAY: 8804 /* This is a duplicate LOCK request */ 8805 dup_lock = TRUE; 8806 8807 /* 8808 * For a duplicate we do not want to 8809 * create a new lockowner as it should 8810 * already exist. 8811 * Turn off the lockowner create flag. 8812 */ 8813 lcreate = FALSE; 8814 } 8815 break; 8816 } 8817 8818 lo = rfs4_findlockowner(&olo->lock_owner, &lcreate); 8819 if (lo == NULL) { 8820 NFS4_DEBUG(rfs4_debug, 8821 (CE_NOTE, "rfs4_op_lock: no lock owner")); 8822 *cs->statusp = resp->status = NFS4ERR_RESOURCE; 8823 goto end; 8824 } 8825 8826 lsp = rfs4_findlo_state_by_owner(lo, sp, &create); 8827 if (lsp == NULL) { 8828 rfs4_update_lease(sp->rs_owner->ro_client); 8829 /* 8830 * Only update theh open_seqid if this is not 8831 * a duplicate request 8832 */ 8833 if (dup_lock == FALSE) { 8834 rfs4_update_open_sequence(sp->rs_owner); 8835 } 8836 8837 NFS4_DEBUG(rfs4_debug, 8838 (CE_NOTE, "rfs4_op_lock: no state")); 8839 *cs->statusp = resp->status = NFS4ERR_SERVERFAULT; 8840 rfs4_update_open_resp(sp->rs_owner, resop, NULL); 8841 rfs4_lockowner_rele(lo); 8842 goto end; 8843 } 8844 8845 /* 8846 * This is the new_lock_owner branch and the client is 8847 * supposed to be associating a new lock_owner with 8848 * the open file at this point. If we find that a 8849 * lock_owner/state association already exists and a 8850 * successful LOCK request was returned to the client, 8851 * an error is returned to the client since this is 8852 * not appropriate. The client should be using the 8853 * existing lock_owner branch. 8854 */ 8855 if (dup_lock == FALSE && create == FALSE) { 8856 if (lsp->rls_lock_completed == TRUE) { 8857 *cs->statusp = 8858 resp->status = NFS4ERR_BAD_SEQID; 8859 rfs4_lockowner_rele(lo); 8860 goto end; 8861 } 8862 } 8863 8864 rfs4_update_lease(sp->rs_owner->ro_client); 8865 8866 /* 8867 * Only update theh open_seqid if this is not 8868 * a duplicate request 8869 */ 8870 if (dup_lock == FALSE) { 8871 rfs4_update_open_sequence(sp->rs_owner); 8872 } 8873 8874 /* 8875 * If this is a duplicate lock request, just copy the 8876 * previously saved reply and return. 8877 */ 8878 if (dup_lock == TRUE) { 8879 /* verify that lock_seqid's match */ 8880 if (lsp->rls_seqid != olo->lock_seqid) { 8881 NFS4_DEBUG(rfs4_debug, 8882 (CE_NOTE, "rfs4_op_lock: Dup-Lock seqid bad" 8883 "lsp->seqid=%d old->seqid=%d", 8884 lsp->rls_seqid, olo->lock_seqid)); 8885 *cs->statusp = resp->status = NFS4ERR_BAD_SEQID; 8886 } else { 8887 rfs4_copy_reply(resop, &lsp->rls_reply); 8888 /* 8889 * Make sure to copy the just 8890 * retrieved reply status into the 8891 * overall compound status 8892 */ 8893 *cs->statusp = resp->status; 8894 } 8895 rfs4_lockowner_rele(lo); 8896 goto end; 8897 } 8898 8899 rfs4_dbe_lock(lsp->rls_dbe); 8900 8901 /* Make sure to update the lock sequence id */ 8902 lsp->rls_seqid = olo->lock_seqid; 8903 8904 NFS4_DEBUG(rfs4_debug, 8905 (CE_NOTE, "Lock seqid established as %d", lsp->rls_seqid)); 8906 8907 /* 8908 * This is used to signify the newly created lockowner 8909 * stateid and its sequence number. The checks for 8910 * sequence number and increment don't occur on the 8911 * very first lock request for a lockowner. 8912 */ 8913 lsp->rls_skip_seqid_check = TRUE; 8914 8915 /* hold off other access to lsp while we tinker */ 8916 rfs4_sw_enter(&lsp->rls_sw); 8917 ls_sw_held = TRUE; 8918 8919 rfs4_dbe_unlock(lsp->rls_dbe); 8920 8921 rfs4_lockowner_rele(lo); 8922 } else { 8923 stateid = &args->locker.locker4_u.lock_owner.lock_stateid; 8924 /* get lsp and hold the lock on the underlying file struct */ 8925 if ((status = rfs4_get_lo_state(stateid, &lsp, TRUE)) 8926 != NFS4_OK) { 8927 *cs->statusp = resp->status = status; 8928 DTRACE_NFSV4_2(op__lock__done, struct compound_state *, 8929 cs, LOCK4res *, resp); 8930 return; 8931 } 8932 create = FALSE; /* We didn't create lsp */ 8933 8934 /* Ensure specified filehandle matches */ 8935 if (cs->vp != lsp->rls_state->rs_finfo->rf_vp) { 8936 rfs4_lo_state_rele(lsp, TRUE); 8937 *cs->statusp = resp->status = NFS4ERR_BAD_STATEID; 8938 DTRACE_NFSV4_2(op__lock__done, struct compound_state *, 8939 cs, LOCK4res *, resp); 8940 return; 8941 } 8942 8943 /* hold off other access to lsp while we tinker */ 8944 rfs4_sw_enter(&lsp->rls_sw); 8945 ls_sw_held = TRUE; 8946 8947 switch (rfs4_check_lo_stateid_seqid(lsp, stateid)) { 8948 /* 8949 * The stateid looks like it was okay (expected to be 8950 * the next one) 8951 */ 8952 case NFS4_CHECK_STATEID_OKAY: 8953 /* 8954 * The sequence id is now checked. Determine 8955 * if this is a replay or if it is in the 8956 * expected (next) sequence. In the case of a 8957 * replay, there are two replay conditions 8958 * that may occur. The first is the normal 8959 * condition where a LOCK is done with a 8960 * NFS4_OK response and the stateid is 8961 * updated. That case is handled below when 8962 * the stateid is identified as a REPLAY. The 8963 * second is the case where an error is 8964 * returned, like NFS4ERR_DENIED, and the 8965 * sequence number is updated but the stateid 8966 * is not updated. This second case is dealt 8967 * with here. So it may seem odd that the 8968 * stateid is okay but the sequence id is a 8969 * replay but it is okay. 8970 */ 8971 switch (rfs4_check_lock_seqid( 8972 args->locker.locker4_u.lock_owner.lock_seqid, 8973 lsp, resop)) { 8974 case NFS4_CHKSEQ_REPLAY: 8975 if (resp->status != NFS4_OK) { 8976 /* 8977 * Here is our replay and need 8978 * to verify that the last 8979 * response was an error. 8980 */ 8981 *cs->statusp = resp->status; 8982 goto end; 8983 } 8984 /* 8985 * This is done since the sequence id 8986 * looked like a replay but it didn't 8987 * pass our check so a BAD_SEQID is 8988 * returned as a result. 8989 */ 8990 /*FALLTHROUGH*/ 8991 case NFS4_CHKSEQ_BAD: 8992 *cs->statusp = resp->status = NFS4ERR_BAD_SEQID; 8993 goto end; 8994 case NFS4_CHKSEQ_OKAY: 8995 /* Everything looks okay move ahead */ 8996 break; 8997 } 8998 break; 8999 case NFS4_CHECK_STATEID_OLD: 9000 *cs->statusp = resp->status = NFS4ERR_OLD_STATEID; 9001 goto end; 9002 case NFS4_CHECK_STATEID_BAD: 9003 *cs->statusp = resp->status = NFS4ERR_BAD_STATEID; 9004 goto end; 9005 case NFS4_CHECK_STATEID_EXPIRED: 9006 *cs->statusp = resp->status = NFS4ERR_EXPIRED; 9007 goto end; 9008 case NFS4_CHECK_STATEID_CLOSED: 9009 *cs->statusp = resp->status = NFS4ERR_OLD_STATEID; 9010 goto end; 9011 case NFS4_CHECK_STATEID_REPLAY: 9012 switch (rfs4_check_lock_seqid( 9013 args->locker.locker4_u.lock_owner.lock_seqid, 9014 lsp, resop)) { 9015 case NFS4_CHKSEQ_OKAY: 9016 /* 9017 * This is a replayed stateid; if 9018 * seqid matches the next expected, 9019 * then client is using wrong seqid. 9020 */ 9021 case NFS4_CHKSEQ_BAD: 9022 *cs->statusp = resp->status = NFS4ERR_BAD_SEQID; 9023 goto end; 9024 case NFS4_CHKSEQ_REPLAY: 9025 rfs4_update_lease(lsp->rls_locker->rl_client); 9026 *cs->statusp = status = resp->status; 9027 goto end; 9028 } 9029 break; 9030 default: 9031 ASSERT(FALSE); 9032 break; 9033 } 9034 9035 rfs4_update_lock_sequence(lsp); 9036 rfs4_update_lease(lsp->rls_locker->rl_client); 9037 } 9038 9039 /* 9040 * NFS4 only allows locking on regular files, so 9041 * verify type of object. 9042 */ 9043 if (cs->vp->v_type != VREG) { 9044 if (cs->vp->v_type == VDIR) 9045 status = NFS4ERR_ISDIR; 9046 else 9047 status = NFS4ERR_INVAL; 9048 goto out; 9049 } 9050 9051 cp = lsp->rls_state->rs_owner->ro_client; 9052 9053 if (rfs4_clnt_in_grace(cp) && !args->reclaim) { 9054 status = NFS4ERR_GRACE; 9055 goto out; 9056 } 9057 9058 if (rfs4_clnt_in_grace(cp) && args->reclaim && !cp->rc_can_reclaim) { 9059 status = NFS4ERR_NO_GRACE; 9060 goto out; 9061 } 9062 9063 if (!rfs4_clnt_in_grace(cp) && args->reclaim) { 9064 status = NFS4ERR_NO_GRACE; 9065 goto out; 9066 } 9067 9068 if (lsp->rls_state->rs_finfo->rf_dinfo.rd_dtype == OPEN_DELEGATE_WRITE) 9069 cs->deleg = TRUE; 9070 9071 status = rfs4_do_lock(lsp, args->locktype, 9072 args->offset, args->length, cs->cr, resop); 9073 9074 out: 9075 lsp->rls_skip_seqid_check = FALSE; 9076 9077 *cs->statusp = resp->status = status; 9078 9079 if (status == NFS4_OK) { 9080 resp->LOCK4res_u.lock_stateid = lsp->rls_lockid.stateid; 9081 lsp->rls_lock_completed = TRUE; 9082 } 9083 /* 9084 * Only update the "OPEN" response here if this was a new 9085 * lock_owner 9086 */ 9087 if (sp) 9088 rfs4_update_open_resp(sp->rs_owner, resop, NULL); 9089 9090 rfs4_update_lock_resp(lsp, resop); 9091 9092 end: 9093 if (lsp) { 9094 if (ls_sw_held) 9095 rfs4_sw_exit(&lsp->rls_sw); 9096 /* 9097 * If an sp obtained, then the lsp does not represent 9098 * a lock on the file struct. 9099 */ 9100 if (sp != NULL) 9101 rfs4_lo_state_rele(lsp, FALSE); 9102 else 9103 rfs4_lo_state_rele(lsp, TRUE); 9104 } 9105 if (sp) { 9106 rfs4_sw_exit(&sp->rs_owner->ro_sw); 9107 rfs4_state_rele(sp); 9108 } 9109 9110 DTRACE_NFSV4_2(op__lock__done, struct compound_state *, cs, 9111 LOCK4res *, resp); 9112 } 9113 9114 /* free function for LOCK/LOCKT */ 9115 static void 9116 lock_denied_free(nfs_resop4 *resop) 9117 { 9118 LOCK4denied *dp = NULL; 9119 9120 switch (resop->resop) { 9121 case OP_LOCK: 9122 if (resop->nfs_resop4_u.oplock.status == NFS4ERR_DENIED) 9123 dp = &resop->nfs_resop4_u.oplock.LOCK4res_u.denied; 9124 break; 9125 case OP_LOCKT: 9126 if (resop->nfs_resop4_u.oplockt.status == NFS4ERR_DENIED) 9127 dp = &resop->nfs_resop4_u.oplockt.denied; 9128 break; 9129 default: 9130 break; 9131 } 9132 9133 if (dp) 9134 kmem_free(dp->owner.owner_val, dp->owner.owner_len); 9135 } 9136 9137 /*ARGSUSED*/ 9138 void 9139 rfs4_op_locku(nfs_argop4 *argop, nfs_resop4 *resop, 9140 struct svc_req *req, struct compound_state *cs) 9141 { 9142 LOCKU4args *args = &argop->nfs_argop4_u.oplocku; 9143 LOCKU4res *resp = &resop->nfs_resop4_u.oplocku; 9144 nfsstat4 status; 9145 stateid4 *stateid = &args->lock_stateid; 9146 rfs4_lo_state_t *lsp; 9147 9148 DTRACE_NFSV4_2(op__locku__start, struct compound_state *, cs, 9149 LOCKU4args *, args); 9150 9151 if (cs->vp == NULL) { 9152 *cs->statusp = resp->status = NFS4ERR_NOFILEHANDLE; 9153 DTRACE_NFSV4_2(op__locku__done, struct compound_state *, cs, 9154 LOCKU4res *, resp); 9155 return; 9156 } 9157 9158 if ((status = rfs4_get_lo_state(stateid, &lsp, TRUE)) != NFS4_OK) { 9159 *cs->statusp = resp->status = status; 9160 DTRACE_NFSV4_2(op__locku__done, struct compound_state *, cs, 9161 LOCKU4res *, resp); 9162 return; 9163 } 9164 9165 /* Ensure specified filehandle matches */ 9166 if (cs->vp != lsp->rls_state->rs_finfo->rf_vp) { 9167 rfs4_lo_state_rele(lsp, TRUE); 9168 *cs->statusp = resp->status = NFS4ERR_BAD_STATEID; 9169 DTRACE_NFSV4_2(op__locku__done, struct compound_state *, cs, 9170 LOCKU4res *, resp); 9171 return; 9172 } 9173 9174 /* hold off other access to lsp while we tinker */ 9175 rfs4_sw_enter(&lsp->rls_sw); 9176 9177 switch (rfs4_check_lo_stateid_seqid(lsp, stateid)) { 9178 case NFS4_CHECK_STATEID_OKAY: 9179 if (rfs4_check_lock_seqid(args->seqid, lsp, resop) 9180 != NFS4_CHKSEQ_OKAY) { 9181 *cs->statusp = resp->status = NFS4ERR_BAD_SEQID; 9182 goto end; 9183 } 9184 break; 9185 case NFS4_CHECK_STATEID_OLD: 9186 *cs->statusp = resp->status = NFS4ERR_OLD_STATEID; 9187 goto end; 9188 case NFS4_CHECK_STATEID_BAD: 9189 *cs->statusp = resp->status = NFS4ERR_BAD_STATEID; 9190 goto end; 9191 case NFS4_CHECK_STATEID_EXPIRED: 9192 *cs->statusp = resp->status = NFS4ERR_EXPIRED; 9193 goto end; 9194 case NFS4_CHECK_STATEID_CLOSED: 9195 *cs->statusp = resp->status = NFS4ERR_OLD_STATEID; 9196 goto end; 9197 case NFS4_CHECK_STATEID_REPLAY: 9198 switch (rfs4_check_lock_seqid(args->seqid, lsp, resop)) { 9199 case NFS4_CHKSEQ_OKAY: 9200 /* 9201 * This is a replayed stateid; if 9202 * seqid matches the next expected, 9203 * then client is using wrong seqid. 9204 */ 9205 case NFS4_CHKSEQ_BAD: 9206 *cs->statusp = resp->status = NFS4ERR_BAD_SEQID; 9207 goto end; 9208 case NFS4_CHKSEQ_REPLAY: 9209 rfs4_update_lease(lsp->rls_locker->rl_client); 9210 *cs->statusp = status = resp->status; 9211 goto end; 9212 } 9213 break; 9214 default: 9215 ASSERT(FALSE); 9216 break; 9217 } 9218 9219 rfs4_update_lock_sequence(lsp); 9220 rfs4_update_lease(lsp->rls_locker->rl_client); 9221 9222 /* 9223 * NFS4 only allows locking on regular files, so 9224 * verify type of object. 9225 */ 9226 if (cs->vp->v_type != VREG) { 9227 if (cs->vp->v_type == VDIR) 9228 status = NFS4ERR_ISDIR; 9229 else 9230 status = NFS4ERR_INVAL; 9231 goto out; 9232 } 9233 9234 if (rfs4_clnt_in_grace(lsp->rls_state->rs_owner->ro_client)) { 9235 status = NFS4ERR_GRACE; 9236 goto out; 9237 } 9238 9239 status = rfs4_do_lock(lsp, args->locktype, 9240 args->offset, args->length, cs->cr, resop); 9241 9242 out: 9243 *cs->statusp = resp->status = status; 9244 9245 if (status == NFS4_OK) 9246 resp->lock_stateid = lsp->rls_lockid.stateid; 9247 9248 rfs4_update_lock_resp(lsp, resop); 9249 9250 end: 9251 rfs4_sw_exit(&lsp->rls_sw); 9252 rfs4_lo_state_rele(lsp, TRUE); 9253 9254 DTRACE_NFSV4_2(op__locku__done, struct compound_state *, cs, 9255 LOCKU4res *, resp); 9256 } 9257 9258 /* 9259 * LOCKT is a best effort routine, the client can not be guaranteed that 9260 * the status return is still in effect by the time the reply is received. 9261 * They are numerous race conditions in this routine, but we are not required 9262 * and can not be accurate. 9263 */ 9264 /*ARGSUSED*/ 9265 void 9266 rfs4_op_lockt(nfs_argop4 *argop, nfs_resop4 *resop, 9267 struct svc_req *req, struct compound_state *cs) 9268 { 9269 LOCKT4args *args = &argop->nfs_argop4_u.oplockt; 9270 LOCKT4res *resp = &resop->nfs_resop4_u.oplockt; 9271 rfs4_lockowner_t *lo; 9272 rfs4_client_t *cp; 9273 bool_t create = FALSE; 9274 struct flock64 flk; 9275 int error; 9276 int flag = FREAD | FWRITE; 9277 int ltype; 9278 length4 posix_length; 9279 sysid_t sysid; 9280 pid_t pid; 9281 9282 DTRACE_NFSV4_2(op__lockt__start, struct compound_state *, cs, 9283 LOCKT4args *, args); 9284 9285 if (cs->vp == NULL) { 9286 *cs->statusp = resp->status = NFS4ERR_NOFILEHANDLE; 9287 goto out; 9288 } 9289 9290 /* 9291 * NFS4 only allows locking on regular files, so 9292 * verify type of object. 9293 */ 9294 if (cs->vp->v_type != VREG) { 9295 if (cs->vp->v_type == VDIR) 9296 *cs->statusp = resp->status = NFS4ERR_ISDIR; 9297 else 9298 *cs->statusp = resp->status = NFS4ERR_INVAL; 9299 goto out; 9300 } 9301 9302 /* 9303 * Check out the clientid to ensure the server knows about it 9304 * so that we correctly inform the client of a server reboot. 9305 */ 9306 if ((cp = rfs4_findclient_by_id(args->owner.clientid, FALSE)) 9307 == NULL) { 9308 *cs->statusp = resp->status = 9309 rfs4_check_clientid(&args->owner.clientid, 0); 9310 goto out; 9311 } 9312 if (rfs4_lease_expired(cp)) { 9313 rfs4_client_close(cp); 9314 /* 9315 * Protocol doesn't allow returning NFS4ERR_STALE as 9316 * other operations do on this check so STALE_CLIENTID 9317 * is returned instead 9318 */ 9319 *cs->statusp = resp->status = NFS4ERR_STALE_CLIENTID; 9320 goto out; 9321 } 9322 9323 if (rfs4_clnt_in_grace(cp) && !(cp->rc_can_reclaim)) { 9324 *cs->statusp = resp->status = NFS4ERR_GRACE; 9325 rfs4_client_rele(cp); 9326 goto out; 9327 } 9328 rfs4_client_rele(cp); 9329 9330 resp->status = NFS4_OK; 9331 9332 switch (args->locktype) { 9333 case READ_LT: 9334 case READW_LT: 9335 ltype = F_RDLCK; 9336 break; 9337 case WRITE_LT: 9338 case WRITEW_LT: 9339 ltype = F_WRLCK; 9340 break; 9341 } 9342 9343 posix_length = args->length; 9344 /* Check for zero length. To lock to end of file use all ones for V4 */ 9345 if (posix_length == 0) { 9346 *cs->statusp = resp->status = NFS4ERR_INVAL; 9347 goto out; 9348 } else if (posix_length == (length4)(~0)) { 9349 posix_length = 0; /* Posix to end of file */ 9350 } 9351 9352 /* Find or create a lockowner */ 9353 lo = rfs4_findlockowner(&args->owner, &create); 9354 9355 if (lo) { 9356 pid = lo->rl_pid; 9357 if ((resp->status = 9358 rfs4_client_sysid(lo->rl_client, &sysid)) != NFS4_OK) 9359 goto err; 9360 } else { 9361 pid = 0; 9362 sysid = lockt_sysid; 9363 } 9364 retry: 9365 flk.l_type = ltype; 9366 flk.l_whence = 0; /* SEEK_SET */ 9367 flk.l_start = args->offset; 9368 flk.l_len = posix_length; 9369 flk.l_sysid = sysid; 9370 flk.l_pid = pid; 9371 flag |= F_REMOTELOCK; 9372 9373 LOCK_PRINT(rfs4_debug, "rfs4_op_lockt", F_GETLK, &flk); 9374 9375 /* Note that length4 is uint64_t but l_len and l_start are off64_t */ 9376 if (flk.l_len < 0 || flk.l_start < 0) { 9377 resp->status = NFS4ERR_INVAL; 9378 goto err; 9379 } 9380 error = VOP_FRLOCK(cs->vp, F_GETLK, &flk, flag, (u_offset_t)0, 9381 NULL, cs->cr, NULL); 9382 9383 /* 9384 * N.B. We map error values to nfsv4 errors. This is differrent 9385 * than puterrno4 routine. 9386 */ 9387 switch (error) { 9388 case 0: 9389 if (flk.l_type == F_UNLCK) 9390 resp->status = NFS4_OK; 9391 else { 9392 if (lock_denied(&resp->denied, &flk) == NFS4ERR_EXPIRED) 9393 goto retry; 9394 resp->status = NFS4ERR_DENIED; 9395 } 9396 break; 9397 case EOVERFLOW: 9398 resp->status = NFS4ERR_INVAL; 9399 break; 9400 case EINVAL: 9401 resp->status = NFS4ERR_NOTSUPP; 9402 break; 9403 default: 9404 cmn_err(CE_WARN, "rfs4_op_lockt: unexpected errno (%d)", 9405 error); 9406 resp->status = NFS4ERR_SERVERFAULT; 9407 break; 9408 } 9409 9410 err: 9411 if (lo) 9412 rfs4_lockowner_rele(lo); 9413 *cs->statusp = resp->status; 9414 out: 9415 DTRACE_NFSV4_2(op__lockt__done, struct compound_state *, cs, 9416 LOCKT4res *, resp); 9417 } 9418 9419 int 9420 rfs4_share(rfs4_state_t *sp, uint32_t access, uint32_t deny) 9421 { 9422 int err; 9423 int cmd; 9424 vnode_t *vp; 9425 struct shrlock shr; 9426 struct shr_locowner shr_loco; 9427 int fflags = 0; 9428 9429 ASSERT(rfs4_dbe_islocked(sp->rs_dbe)); 9430 ASSERT(sp->rs_owner->ro_client->rc_sysidt != LM_NOSYSID); 9431 9432 if (sp->rs_closed) 9433 return (NFS4ERR_OLD_STATEID); 9434 9435 vp = sp->rs_finfo->rf_vp; 9436 ASSERT(vp); 9437 9438 shr.s_access = shr.s_deny = 0; 9439 9440 if (access & OPEN4_SHARE_ACCESS_READ) { 9441 fflags |= FREAD; 9442 shr.s_access |= F_RDACC; 9443 } 9444 if (access & OPEN4_SHARE_ACCESS_WRITE) { 9445 fflags |= FWRITE; 9446 shr.s_access |= F_WRACC; 9447 } 9448 ASSERT(shr.s_access); 9449 9450 if (deny & OPEN4_SHARE_DENY_READ) 9451 shr.s_deny |= F_RDDNY; 9452 if (deny & OPEN4_SHARE_DENY_WRITE) 9453 shr.s_deny |= F_WRDNY; 9454 9455 shr.s_pid = rfs4_dbe_getid(sp->rs_owner->ro_dbe); 9456 shr.s_sysid = sp->rs_owner->ro_client->rc_sysidt; 9457 shr_loco.sl_pid = shr.s_pid; 9458 shr_loco.sl_id = shr.s_sysid; 9459 shr.s_owner = (caddr_t)&shr_loco; 9460 shr.s_own_len = sizeof (shr_loco); 9461 9462 cmd = nbl_need_check(vp) ? F_SHARE_NBMAND : F_SHARE; 9463 9464 err = VOP_SHRLOCK(vp, cmd, &shr, fflags, CRED(), NULL); 9465 if (err != 0) { 9466 if (err == EAGAIN) 9467 err = NFS4ERR_SHARE_DENIED; 9468 else 9469 err = puterrno4(err); 9470 return (err); 9471 } 9472 9473 sp->rs_share_access |= access; 9474 sp->rs_share_deny |= deny; 9475 9476 return (0); 9477 } 9478 9479 int 9480 rfs4_unshare(rfs4_state_t *sp) 9481 { 9482 int err; 9483 struct shrlock shr; 9484 struct shr_locowner shr_loco; 9485 9486 ASSERT(rfs4_dbe_islocked(sp->rs_dbe)); 9487 9488 if (sp->rs_closed || sp->rs_share_access == 0) 9489 return (0); 9490 9491 ASSERT(sp->rs_owner->ro_client->rc_sysidt != LM_NOSYSID); 9492 ASSERT(sp->rs_finfo->rf_vp); 9493 9494 shr.s_access = shr.s_deny = 0; 9495 shr.s_pid = rfs4_dbe_getid(sp->rs_owner->ro_dbe); 9496 shr.s_sysid = sp->rs_owner->ro_client->rc_sysidt; 9497 shr_loco.sl_pid = shr.s_pid; 9498 shr_loco.sl_id = shr.s_sysid; 9499 shr.s_owner = (caddr_t)&shr_loco; 9500 shr.s_own_len = sizeof (shr_loco); 9501 9502 err = VOP_SHRLOCK(sp->rs_finfo->rf_vp, F_UNSHARE, &shr, 0, CRED(), 9503 NULL); 9504 if (err != 0) { 9505 err = puterrno4(err); 9506 return (err); 9507 } 9508 9509 sp->rs_share_access = 0; 9510 sp->rs_share_deny = 0; 9511 9512 return (0); 9513 9514 } 9515 9516 static int 9517 rdma_setup_read_data4(READ4args *args, READ4res *rok) 9518 { 9519 struct clist *wcl; 9520 count4 count = rok->data_len; 9521 int wlist_len; 9522 9523 wcl = args->wlist; 9524 if (rdma_setup_read_chunks(wcl, count, &wlist_len) == FALSE) { 9525 return (FALSE); 9526 } 9527 wcl = args->wlist; 9528 rok->wlist_len = wlist_len; 9529 rok->wlist = wcl; 9530 return (TRUE); 9531 } 9532 9533 /* tunable to disable server referrals */ 9534 int rfs4_no_referrals = 0; 9535 9536 /* 9537 * Find an NFS record in reparse point data. 9538 * Returns 0 for success and <0 or an errno value on failure. 9539 */ 9540 int 9541 vn_find_nfs_record(vnode_t *vp, nvlist_t **nvlp, char **svcp, char **datap) 9542 { 9543 int err; 9544 char *stype, *val; 9545 nvlist_t *nvl; 9546 nvpair_t *curr; 9547 9548 if ((nvl = reparse_init()) == NULL) 9549 return (-1); 9550 9551 if ((err = reparse_vnode_parse(vp, nvl)) != 0) { 9552 reparse_free(nvl); 9553 return (err); 9554 } 9555 9556 curr = NULL; 9557 while ((curr = nvlist_next_nvpair(nvl, curr)) != NULL) { 9558 if ((stype = nvpair_name(curr)) == NULL) { 9559 reparse_free(nvl); 9560 return (-2); 9561 } 9562 if (strncasecmp(stype, "NFS", 3) == 0) 9563 break; 9564 } 9565 9566 if ((curr == NULL) || 9567 (nvpair_value_string(curr, &val))) { 9568 reparse_free(nvl); 9569 return (-3); 9570 } 9571 *nvlp = nvl; 9572 *svcp = stype; 9573 *datap = val; 9574 return (0); 9575 } 9576 9577 int 9578 vn_is_nfs_reparse(vnode_t *vp, cred_t *cr) 9579 { 9580 nvlist_t *nvl; 9581 char *s, *d; 9582 9583 if (rfs4_no_referrals != 0) 9584 return (B_FALSE); 9585 9586 if (vn_is_reparse(vp, cr, NULL) == B_FALSE) 9587 return (B_FALSE); 9588 9589 if (vn_find_nfs_record(vp, &nvl, &s, &d) != 0) 9590 return (B_FALSE); 9591 9592 reparse_free(nvl); 9593 9594 return (B_TRUE); 9595 } 9596 9597 /* 9598 * There is a user-level copy of this routine in ref_subr.c. 9599 * Changes should be kept in sync. 9600 */ 9601 static int 9602 nfs4_create_components(char *path, component4 *comp4) 9603 { 9604 int slen, plen, ncomp; 9605 char *ori_path, *nxtc, buf[MAXNAMELEN]; 9606 9607 if (path == NULL) 9608 return (0); 9609 9610 plen = strlen(path) + 1; /* include the terminator */ 9611 ori_path = path; 9612 ncomp = 0; 9613 9614 /* count number of components in the path */ 9615 for (nxtc = path; nxtc < ori_path + plen; nxtc++) { 9616 if (*nxtc == '/' || *nxtc == '\0' || *nxtc == '\n') { 9617 if ((slen = nxtc - path) == 0) { 9618 path = nxtc + 1; 9619 continue; 9620 } 9621 9622 if (comp4 != NULL) { 9623 bcopy(path, buf, slen); 9624 buf[slen] = '\0'; 9625 (void) str_to_utf8(buf, &comp4[ncomp]); 9626 } 9627 9628 ncomp++; /* 1 valid component */ 9629 path = nxtc + 1; 9630 } 9631 if (*nxtc == '\0' || *nxtc == '\n') 9632 break; 9633 } 9634 9635 return (ncomp); 9636 } 9637 9638 /* 9639 * There is a user-level copy of this routine in ref_subr.c. 9640 * Changes should be kept in sync. 9641 */ 9642 static int 9643 make_pathname4(char *path, pathname4 *pathname) 9644 { 9645 int ncomp; 9646 component4 *comp4; 9647 9648 if (pathname == NULL) 9649 return (0); 9650 9651 if (path == NULL) { 9652 pathname->pathname4_val = NULL; 9653 pathname->pathname4_len = 0; 9654 return (0); 9655 } 9656 9657 /* count number of components to alloc buffer */ 9658 if ((ncomp = nfs4_create_components(path, NULL)) == 0) { 9659 pathname->pathname4_val = NULL; 9660 pathname->pathname4_len = 0; 9661 return (0); 9662 } 9663 comp4 = kmem_zalloc(ncomp * sizeof (component4), KM_SLEEP); 9664 9665 /* copy components into allocated buffer */ 9666 ncomp = nfs4_create_components(path, comp4); 9667 9668 pathname->pathname4_val = comp4; 9669 pathname->pathname4_len = ncomp; 9670 9671 return (ncomp); 9672 } 9673 9674 #define xdr_fs_locations4 xdr_fattr4_fs_locations 9675 9676 fs_locations4 * 9677 fetch_referral(vnode_t *vp, cred_t *cr) 9678 { 9679 nvlist_t *nvl; 9680 char *stype, *sdata; 9681 fs_locations4 *result; 9682 char buf[1024]; 9683 size_t bufsize; 9684 XDR xdr; 9685 int err; 9686 9687 /* 9688 * Check attrs to ensure it's a reparse point 9689 */ 9690 if (vn_is_reparse(vp, cr, NULL) == B_FALSE) 9691 return (NULL); 9692 9693 /* 9694 * Look for an NFS record and get the type and data 9695 */ 9696 if (vn_find_nfs_record(vp, &nvl, &stype, &sdata) != 0) 9697 return (NULL); 9698 9699 /* 9700 * With the type and data, upcall to get the referral 9701 */ 9702 bufsize = sizeof (buf); 9703 bzero(buf, sizeof (buf)); 9704 err = reparse_kderef((const char *)stype, (const char *)sdata, 9705 buf, &bufsize); 9706 reparse_free(nvl); 9707 9708 DTRACE_PROBE4(nfs4serv__func__referral__upcall, 9709 char *, stype, char *, sdata, char *, buf, int, err); 9710 if (err) { 9711 cmn_err(CE_NOTE, 9712 "reparsed daemon not running: unable to get referral (%d)", 9713 err); 9714 return (NULL); 9715 } 9716 9717 /* 9718 * We get an XDR'ed record back from the kderef call 9719 */ 9720 xdrmem_create(&xdr, buf, bufsize, XDR_DECODE); 9721 result = kmem_alloc(sizeof (fs_locations4), KM_SLEEP); 9722 err = xdr_fs_locations4(&xdr, result); 9723 XDR_DESTROY(&xdr); 9724 if (err != TRUE) { 9725 DTRACE_PROBE1(nfs4serv__func__referral__upcall__xdrfail, 9726 int, err); 9727 return (NULL); 9728 } 9729 9730 /* 9731 * Look at path to recover fs_root, ignoring the leading '/' 9732 */ 9733 (void) make_pathname4(vp->v_path, &result->fs_root); 9734 9735 return (result); 9736 } 9737 9738 char * 9739 build_symlink(vnode_t *vp, cred_t *cr, size_t *strsz) 9740 { 9741 fs_locations4 *fsl; 9742 fs_location4 *fs; 9743 char *server, *path, *symbuf; 9744 static char *prefix = "/net/"; 9745 int i, size, npaths; 9746 uint_t len; 9747 9748 /* Get the referral */ 9749 if ((fsl = fetch_referral(vp, cr)) == NULL) 9750 return (NULL); 9751 9752 /* Deal with only the first location and first server */ 9753 fs = &fsl->locations_val[0]; 9754 server = utf8_to_str(&fs->server_val[0], &len, NULL); 9755 if (server == NULL) { 9756 rfs4_free_fs_locations4(fsl); 9757 kmem_free(fsl, sizeof (fs_locations4)); 9758 return (NULL); 9759 } 9760 9761 /* Figure out size for "/net/" + host + /path/path/path + NULL */ 9762 size = strlen(prefix) + len; 9763 for (i = 0; i < fs->rootpath.pathname4_len; i++) 9764 size += fs->rootpath.pathname4_val[i].utf8string_len + 1; 9765 9766 /* Allocate the symlink buffer and fill it */ 9767 symbuf = kmem_zalloc(size, KM_SLEEP); 9768 (void) strcat(symbuf, prefix); 9769 (void) strcat(symbuf, server); 9770 kmem_free(server, len); 9771 9772 npaths = 0; 9773 for (i = 0; i < fs->rootpath.pathname4_len; i++) { 9774 path = utf8_to_str(&fs->rootpath.pathname4_val[i], &len, NULL); 9775 if (path == NULL) 9776 continue; 9777 (void) strcat(symbuf, "/"); 9778 (void) strcat(symbuf, path); 9779 npaths++; 9780 kmem_free(path, len); 9781 } 9782 9783 rfs4_free_fs_locations4(fsl); 9784 kmem_free(fsl, sizeof (fs_locations4)); 9785 9786 if (strsz != NULL) 9787 *strsz = size; 9788 return (symbuf); 9789 } 9790 9791 /* 9792 * Check to see if we have a downrev Solaris client, so that we 9793 * can send it a symlink instead of a referral. 9794 */ 9795 int 9796 client_is_downrev(struct svc_req *req) 9797 { 9798 struct sockaddr *ca; 9799 rfs4_clntip_t *ci; 9800 bool_t create = FALSE; 9801 int is_downrev; 9802 9803 ca = (struct sockaddr *)svc_getrpccaller(req->rq_xprt)->buf; 9804 ASSERT(ca); 9805 ci = rfs4_find_clntip(ca, &create); 9806 if (ci == NULL) 9807 return (0); 9808 is_downrev = ci->ri_no_referrals; 9809 rfs4_dbe_rele(ci->ri_dbe); 9810 return (is_downrev); 9811 } 9812