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 2008 Sun Microsystems, Inc. All rights reserved. 23 * Use is subject to license terms. 24 */ 25 26 /* 27 * Copyright (c) 1983,1984,1985,1986,1987,1988,1989 AT&T. 28 * All Rights Reserved 29 */ 30 31 #include <sys/param.h> 32 #include <sys/types.h> 33 #include <sys/systm.h> 34 #include <sys/cred.h> 35 #include <sys/buf.h> 36 #include <sys/vfs.h> 37 #include <sys/vfs_opreg.h> 38 #include <sys/vnode.h> 39 #include <sys/uio.h> 40 #include <sys/errno.h> 41 #include <sys/sysmacros.h> 42 #include <sys/statvfs.h> 43 #include <sys/kmem.h> 44 #include <sys/dirent.h> 45 #include <sys/cmn_err.h> 46 #include <sys/debug.h> 47 #include <sys/systeminfo.h> 48 #include <sys/flock.h> 49 #include <sys/pathname.h> 50 #include <sys/nbmlock.h> 51 #include <sys/share.h> 52 #include <sys/atomic.h> 53 #include <sys/policy.h> 54 #include <sys/fem.h> 55 #include <sys/sdt.h> 56 #include <sys/ddi.h> 57 58 #include <rpc/types.h> 59 #include <rpc/auth.h> 60 #include <rpc/rpcsec_gss.h> 61 #include <rpc/svc.h> 62 63 #include <nfs/nfs.h> 64 #include <nfs/export.h> 65 #include <nfs/lm.h> 66 #include <nfs/nfs4.h> 67 68 #include <sys/strsubr.h> 69 #include <sys/strsun.h> 70 71 #include <inet/common.h> 72 #include <inet/ip.h> 73 #include <inet/ip6.h> 74 75 #include <sys/tsol/label.h> 76 #include <sys/tsol/tndb.h> 77 78 #define RFS4_MAXLOCK_TRIES 4 /* Try to get the lock this many times */ 79 static int rfs4_maxlock_tries = RFS4_MAXLOCK_TRIES; 80 #define RFS4_LOCK_DELAY 10 /* Milliseconds */ 81 static clock_t rfs4_lock_delay = RFS4_LOCK_DELAY; 82 extern struct svc_ops rdma_svc_ops; 83 /* End of Tunables */ 84 85 static int rdma_setup_read_data4(READ4args *, READ4res *); 86 87 /* 88 * Used to bump the stateid4.seqid value and show changes in the stateid 89 */ 90 #define next_stateid(sp) (++(sp)->bits.chgseq) 91 92 /* 93 * RFS4_MINLEN_ENTRY4: XDR-encoded size of smallest possible dirent. 94 * This is used to return NFS4ERR_TOOSMALL when clients specify 95 * maxcount that isn't large enough to hold the smallest possible 96 * XDR encoded dirent. 97 * 98 * sizeof cookie (8 bytes) + 99 * sizeof name_len (4 bytes) + 100 * sizeof smallest (padded) name (4 bytes) + 101 * sizeof bitmap4_len (12 bytes) + NOTE: we always encode len=2 bm4 102 * sizeof attrlist4_len (4 bytes) + 103 * sizeof next boolean (4 bytes) 104 * 105 * RFS4_MINLEN_RDDIR4: XDR-encoded size of READDIR op reply containing 106 * the smallest possible entry4 (assumes no attrs requested). 107 * sizeof nfsstat4 (4 bytes) + 108 * sizeof verifier4 (8 bytes) + 109 * sizeof entry4list bool (4 bytes) + 110 * sizeof entry4 (36 bytes) + 111 * sizeof eof bool (4 bytes) 112 * 113 * RFS4_MINLEN_RDDIR_BUF: minimum length of buffer server will provide to 114 * VOP_READDIR. Its value is the size of the maximum possible dirent 115 * for solaris. The DIRENT64_RECLEN macro returns the size of dirent 116 * required for a given name length. MAXNAMELEN is the maximum 117 * filename length allowed in Solaris. The first two DIRENT64_RECLEN() 118 * macros are to allow for . and .. entries -- just a minor tweak to try 119 * and guarantee that buffer we give to VOP_READDIR will be large enough 120 * to hold ., .., and the largest possible solaris dirent64. 121 */ 122 #define RFS4_MINLEN_ENTRY4 36 123 #define RFS4_MINLEN_RDDIR4 (4 + NFS4_VERIFIER_SIZE + 4 + RFS4_MINLEN_ENTRY4 + 4) 124 #define RFS4_MINLEN_RDDIR_BUF \ 125 (DIRENT64_RECLEN(1) + DIRENT64_RECLEN(2) + DIRENT64_RECLEN(MAXNAMELEN)) 126 127 /* 128 * It would be better to pad to 4 bytes since that's what XDR would do, 129 * but the dirents UFS gives us are already padded to 8, so just take 130 * what we're given. Dircount is only a hint anyway. Currently the 131 * solaris kernel is ASCII only, so there's no point in calling the 132 * UTF8 functions. 133 * 134 * dirent64: named padded to provide 8 byte struct alignment 135 * d_ino(8) + d_off(8) + d_reclen(2) + d_name(namelen + null(1) + pad) 136 * 137 * cookie: uint64_t + utf8namelen: uint_t + utf8name padded to 8 bytes 138 * 139 */ 140 #define DIRENT64_TO_DIRCOUNT(dp) \ 141 (3 * BYTES_PER_XDR_UNIT + DIRENT64_NAMELEN((dp)->d_reclen)) 142 143 time_t rfs4_start_time; /* Initialized in rfs4_srvrinit */ 144 145 static sysid_t lockt_sysid; /* dummy sysid for all LOCKT calls */ 146 147 u_longlong_t nfs4_srv_caller_id; 148 uint_t nfs4_srv_vkey = 0; 149 150 verifier4 Write4verf; 151 verifier4 Readdir4verf; 152 153 void rfs4_init_compound_state(struct compound_state *); 154 155 static void nullfree(caddr_t); 156 static void rfs4_op_inval(nfs_argop4 *, nfs_resop4 *, struct svc_req *, 157 struct compound_state *); 158 static void rfs4_op_access(nfs_argop4 *, nfs_resop4 *, struct svc_req *, 159 struct compound_state *); 160 static void rfs4_op_close(nfs_argop4 *, nfs_resop4 *, struct svc_req *, 161 struct compound_state *); 162 static void rfs4_op_commit(nfs_argop4 *, nfs_resop4 *, struct svc_req *, 163 struct compound_state *); 164 static void rfs4_op_create(nfs_argop4 *, nfs_resop4 *, struct svc_req *, 165 struct compound_state *); 166 static void rfs4_op_create_free(nfs_resop4 *resop); 167 static void rfs4_op_delegreturn(nfs_argop4 *, nfs_resop4 *, 168 struct svc_req *, struct compound_state *); 169 static void rfs4_op_delegpurge(nfs_argop4 *, nfs_resop4 *, 170 struct svc_req *, struct compound_state *); 171 static void rfs4_op_getattr(nfs_argop4 *, nfs_resop4 *, struct svc_req *, 172 struct compound_state *); 173 static void rfs4_op_getattr_free(nfs_resop4 *); 174 static void rfs4_op_getfh(nfs_argop4 *, nfs_resop4 *, struct svc_req *, 175 struct compound_state *); 176 static void rfs4_op_getfh_free(nfs_resop4 *); 177 static void rfs4_op_illegal(nfs_argop4 *, nfs_resop4 *, struct svc_req *, 178 struct compound_state *); 179 static void rfs4_op_link(nfs_argop4 *, nfs_resop4 *, struct svc_req *, 180 struct compound_state *); 181 static void rfs4_op_lock(nfs_argop4 *, nfs_resop4 *, struct svc_req *, 182 struct compound_state *); 183 static void lock_denied_free(nfs_resop4 *); 184 static void rfs4_op_locku(nfs_argop4 *, nfs_resop4 *, struct svc_req *, 185 struct compound_state *); 186 static void rfs4_op_lockt(nfs_argop4 *, nfs_resop4 *, struct svc_req *, 187 struct compound_state *); 188 static void rfs4_op_lookup(nfs_argop4 *, nfs_resop4 *, struct svc_req *, 189 struct compound_state *); 190 static void rfs4_op_lookupp(nfs_argop4 *, nfs_resop4 *, struct svc_req *, 191 struct compound_state *); 192 static void rfs4_op_openattr(nfs_argop4 *argop, nfs_resop4 *resop, 193 struct svc_req *req, struct compound_state *cs); 194 static void rfs4_op_nverify(nfs_argop4 *, nfs_resop4 *, struct svc_req *, 195 struct compound_state *); 196 static void rfs4_op_open(nfs_argop4 *, nfs_resop4 *, struct svc_req *, 197 struct compound_state *); 198 static void rfs4_op_open_confirm(nfs_argop4 *, nfs_resop4 *, 199 struct svc_req *, struct compound_state *); 200 static void rfs4_op_open_downgrade(nfs_argop4 *, nfs_resop4 *, 201 struct svc_req *, struct compound_state *); 202 static void rfs4_op_putfh(nfs_argop4 *, nfs_resop4 *, struct svc_req *, 203 struct compound_state *); 204 static void rfs4_op_putpubfh(nfs_argop4 *, nfs_resop4 *, struct svc_req *, 205 struct compound_state *); 206 static void rfs4_op_putrootfh(nfs_argop4 *, nfs_resop4 *, struct svc_req *, 207 struct compound_state *); 208 static void rfs4_op_read(nfs_argop4 *, nfs_resop4 *, struct svc_req *, 209 struct compound_state *); 210 static void rfs4_op_read_free(nfs_resop4 *); 211 static void rfs4_op_readdir_free(nfs_resop4 *resop); 212 static void rfs4_op_readlink(nfs_argop4 *, nfs_resop4 *, struct svc_req *, 213 struct compound_state *); 214 static void rfs4_op_readlink_free(nfs_resop4 *); 215 static void rfs4_op_release_lockowner(nfs_argop4 *, nfs_resop4 *, 216 struct svc_req *, struct compound_state *); 217 static void rfs4_op_remove(nfs_argop4 *, nfs_resop4 *, struct svc_req *, 218 struct compound_state *); 219 static void rfs4_op_rename(nfs_argop4 *, nfs_resop4 *, struct svc_req *, 220 struct compound_state *); 221 static void rfs4_op_renew(nfs_argop4 *, nfs_resop4 *, struct svc_req *, 222 struct compound_state *); 223 static void rfs4_op_restorefh(nfs_argop4 *, nfs_resop4 *, struct svc_req *, 224 struct compound_state *); 225 static void rfs4_op_savefh(nfs_argop4 *, nfs_resop4 *, struct svc_req *, 226 struct compound_state *); 227 static void rfs4_op_setattr(nfs_argop4 *, nfs_resop4 *, struct svc_req *, 228 struct compound_state *); 229 static void rfs4_op_verify(nfs_argop4 *, nfs_resop4 *, struct svc_req *, 230 struct compound_state *); 231 static void rfs4_op_write(nfs_argop4 *, nfs_resop4 *, struct svc_req *, 232 struct compound_state *); 233 static void rfs4_op_setclientid(nfs_argop4 *, nfs_resop4 *, 234 struct svc_req *, struct compound_state *); 235 static void rfs4_op_setclientid_confirm(nfs_argop4 *, nfs_resop4 *, 236 struct svc_req *req, struct compound_state *); 237 static void rfs4_op_secinfo(nfs_argop4 *, nfs_resop4 *, struct svc_req *, 238 struct compound_state *); 239 static void rfs4_op_secinfo_free(nfs_resop4 *); 240 241 static nfsstat4 check_open_access(uint32_t, 242 struct compound_state *, struct svc_req *); 243 nfsstat4 rfs4_client_sysid(rfs4_client_t *, sysid_t *); 244 static int vop_shrlock(vnode_t *, int, struct shrlock *, int); 245 static int rfs4_shrlock(rfs4_state_t *, int); 246 static int rfs4_share(rfs4_state_t *); 247 void rfs4_ss_clid(rfs4_client_t *, struct svc_req *); 248 249 /* 250 * translation table for attrs 251 */ 252 struct nfs4_ntov_table { 253 union nfs4_attr_u *na; 254 uint8_t amap[NFS4_MAXNUM_ATTRS]; 255 int attrcnt; 256 bool_t vfsstat; 257 }; 258 259 static void nfs4_ntov_table_init(struct nfs4_ntov_table *ntovp); 260 static void nfs4_ntov_table_free(struct nfs4_ntov_table *ntovp, 261 struct nfs4_svgetit_arg *sargp); 262 263 static nfsstat4 do_rfs4_set_attrs(bitmap4 *resp, fattr4 *fattrp, 264 struct compound_state *cs, struct nfs4_svgetit_arg *sargp, 265 struct nfs4_ntov_table *ntovp, nfs4_attr_cmd_t cmd); 266 267 fem_t *deleg_rdops; 268 fem_t *deleg_wrops; 269 270 rfs4_servinst_t *rfs4_cur_servinst = NULL; /* current server instance */ 271 kmutex_t rfs4_servinst_lock; /* protects linked list */ 272 int rfs4_seen_first_compound; /* set first time we see one */ 273 274 /* 275 * NFS4 op dispatch table 276 */ 277 278 struct rfsv4disp { 279 void (*dis_proc)(); /* proc to call */ 280 void (*dis_resfree)(); /* frees space allocated by proc */ 281 int dis_flags; /* RPC_IDEMPOTENT, etc... */ 282 }; 283 284 static struct rfsv4disp rfsv4disptab[] = { 285 /* 286 * NFS VERSION 4 287 */ 288 289 /* RFS_NULL = 0 */ 290 {rfs4_op_illegal, nullfree, 0}, 291 292 /* UNUSED = 1 */ 293 {rfs4_op_illegal, nullfree, 0}, 294 295 /* UNUSED = 2 */ 296 {rfs4_op_illegal, nullfree, 0}, 297 298 /* OP_ACCESS = 3 */ 299 {rfs4_op_access, nullfree, RPC_IDEMPOTENT}, 300 301 /* OP_CLOSE = 4 */ 302 {rfs4_op_close, nullfree, 0}, 303 304 /* OP_COMMIT = 5 */ 305 {rfs4_op_commit, nullfree, RPC_IDEMPOTENT}, 306 307 /* OP_CREATE = 6 */ 308 {rfs4_op_create, nullfree, 0}, 309 310 /* OP_DELEGPURGE = 7 */ 311 {rfs4_op_delegpurge, nullfree, 0}, 312 313 /* OP_DELEGRETURN = 8 */ 314 {rfs4_op_delegreturn, nullfree, 0}, 315 316 /* OP_GETATTR = 9 */ 317 {rfs4_op_getattr, rfs4_op_getattr_free, RPC_IDEMPOTENT}, 318 319 /* OP_GETFH = 10 */ 320 {rfs4_op_getfh, rfs4_op_getfh_free, RPC_ALL}, 321 322 /* OP_LINK = 11 */ 323 {rfs4_op_link, nullfree, 0}, 324 325 /* OP_LOCK = 12 */ 326 {rfs4_op_lock, lock_denied_free, 0}, 327 328 /* OP_LOCKT = 13 */ 329 {rfs4_op_lockt, lock_denied_free, 0}, 330 331 /* OP_LOCKU = 14 */ 332 {rfs4_op_locku, nullfree, 0}, 333 334 /* OP_LOOKUP = 15 */ 335 {rfs4_op_lookup, nullfree, (RPC_IDEMPOTENT | RPC_PUBLICFH_OK)}, 336 337 /* OP_LOOKUPP = 16 */ 338 {rfs4_op_lookupp, nullfree, (RPC_IDEMPOTENT | RPC_PUBLICFH_OK)}, 339 340 /* OP_NVERIFY = 17 */ 341 {rfs4_op_nverify, nullfree, RPC_IDEMPOTENT}, 342 343 /* OP_OPEN = 18 */ 344 {rfs4_op_open, rfs4_free_reply, 0}, 345 346 /* OP_OPENATTR = 19 */ 347 {rfs4_op_openattr, nullfree, 0}, 348 349 /* OP_OPEN_CONFIRM = 20 */ 350 {rfs4_op_open_confirm, nullfree, 0}, 351 352 /* OP_OPEN_DOWNGRADE = 21 */ 353 {rfs4_op_open_downgrade, nullfree, 0}, 354 355 /* OP_OPEN_PUTFH = 22 */ 356 {rfs4_op_putfh, nullfree, RPC_ALL}, 357 358 /* OP_PUTPUBFH = 23 */ 359 {rfs4_op_putpubfh, nullfree, RPC_ALL}, 360 361 /* OP_PUTROOTFH = 24 */ 362 {rfs4_op_putrootfh, nullfree, RPC_ALL}, 363 364 /* OP_READ = 25 */ 365 {rfs4_op_read, rfs4_op_read_free, RPC_IDEMPOTENT}, 366 367 /* OP_READDIR = 26 */ 368 {rfs4_op_readdir, rfs4_op_readdir_free, RPC_IDEMPOTENT}, 369 370 /* OP_READLINK = 27 */ 371 {rfs4_op_readlink, rfs4_op_readlink_free, RPC_IDEMPOTENT}, 372 373 /* OP_REMOVE = 28 */ 374 {rfs4_op_remove, nullfree, 0}, 375 376 /* OP_RENAME = 29 */ 377 {rfs4_op_rename, nullfree, 0}, 378 379 /* OP_RENEW = 30 */ 380 {rfs4_op_renew, nullfree, 0}, 381 382 /* OP_RESTOREFH = 31 */ 383 {rfs4_op_restorefh, nullfree, RPC_ALL}, 384 385 /* OP_SAVEFH = 32 */ 386 {rfs4_op_savefh, nullfree, RPC_ALL}, 387 388 /* OP_SECINFO = 33 */ 389 {rfs4_op_secinfo, rfs4_op_secinfo_free, 0}, 390 391 /* OP_SETATTR = 34 */ 392 {rfs4_op_setattr, nullfree, 0}, 393 394 /* OP_SETCLIENTID = 35 */ 395 {rfs4_op_setclientid, nullfree, 0}, 396 397 /* OP_SETCLIENTID_CONFIRM = 36 */ 398 {rfs4_op_setclientid_confirm, nullfree, 0}, 399 400 /* OP_VERIFY = 37 */ 401 {rfs4_op_verify, nullfree, RPC_IDEMPOTENT}, 402 403 /* OP_WRITE = 38 */ 404 {rfs4_op_write, nullfree, 0}, 405 406 /* OP_RELEASE_LOCKOWNER = 39 */ 407 {rfs4_op_release_lockowner, nullfree, 0}, 408 }; 409 410 static uint_t rfsv4disp_cnt = sizeof (rfsv4disptab) / sizeof (rfsv4disptab[0]); 411 412 #define OP_ILLEGAL_IDX (rfsv4disp_cnt) 413 414 #ifdef DEBUG 415 416 int rfs4_fillone_debug = 0; 417 int rfs4_shrlock_debug = 0; 418 int rfs4_no_stub_access = 1; 419 int rfs4_rddir_debug = 0; 420 421 static char *rfs4_op_string[] = { 422 "rfs4_op_null", 423 "rfs4_op_1 unused", 424 "rfs4_op_2 unused", 425 "rfs4_op_access", 426 "rfs4_op_close", 427 "rfs4_op_commit", 428 "rfs4_op_create", 429 "rfs4_op_delegpurge", 430 "rfs4_op_delegreturn", 431 "rfs4_op_getattr", 432 "rfs4_op_getfh", 433 "rfs4_op_link", 434 "rfs4_op_lock", 435 "rfs4_op_lockt", 436 "rfs4_op_locku", 437 "rfs4_op_lookup", 438 "rfs4_op_lookupp", 439 "rfs4_op_nverify", 440 "rfs4_op_open", 441 "rfs4_op_openattr", 442 "rfs4_op_open_confirm", 443 "rfs4_op_open_downgrade", 444 "rfs4_op_putfh", 445 "rfs4_op_putpubfh", 446 "rfs4_op_putrootfh", 447 "rfs4_op_read", 448 "rfs4_op_readdir", 449 "rfs4_op_readlink", 450 "rfs4_op_remove", 451 "rfs4_op_rename", 452 "rfs4_op_renew", 453 "rfs4_op_restorefh", 454 "rfs4_op_savefh", 455 "rfs4_op_secinfo", 456 "rfs4_op_setattr", 457 "rfs4_op_setclientid", 458 "rfs4_op_setclient_confirm", 459 "rfs4_op_verify", 460 "rfs4_op_write", 461 "rfs4_op_release_lockowner", 462 "rfs4_op_illegal" 463 }; 464 #endif 465 466 void rfs4_ss_chkclid(rfs4_client_t *); 467 468 extern size_t strlcpy(char *dst, const char *src, size_t dstsize); 469 470 #ifdef nextdp 471 #undef nextdp 472 #endif 473 #define nextdp(dp) ((struct dirent64 *)((char *)(dp) + (dp)->d_reclen)) 474 475 static const fs_operation_def_t nfs4_rd_deleg_tmpl[] = { 476 VOPNAME_OPEN, { .femop_open = deleg_rd_open }, 477 VOPNAME_WRITE, { .femop_write = deleg_rd_write }, 478 VOPNAME_SETATTR, { .femop_setattr = deleg_rd_setattr }, 479 VOPNAME_RWLOCK, { .femop_rwlock = deleg_rd_rwlock }, 480 VOPNAME_SPACE, { .femop_space = deleg_rd_space }, 481 VOPNAME_SETSECATTR, { .femop_setsecattr = deleg_rd_setsecattr }, 482 VOPNAME_VNEVENT, { .femop_vnevent = deleg_rd_vnevent }, 483 NULL, NULL 484 }; 485 static const fs_operation_def_t nfs4_wr_deleg_tmpl[] = { 486 VOPNAME_OPEN, { .femop_open = deleg_wr_open }, 487 VOPNAME_READ, { .femop_read = deleg_wr_read }, 488 VOPNAME_WRITE, { .femop_write = deleg_wr_write }, 489 VOPNAME_SETATTR, { .femop_setattr = deleg_wr_setattr }, 490 VOPNAME_RWLOCK, { .femop_rwlock = deleg_wr_rwlock }, 491 VOPNAME_SPACE, { .femop_space = deleg_wr_space }, 492 VOPNAME_SETSECATTR, { .femop_setsecattr = deleg_wr_setsecattr }, 493 VOPNAME_VNEVENT, { .femop_vnevent = deleg_wr_vnevent }, 494 NULL, NULL 495 }; 496 497 int 498 rfs4_srvrinit(void) 499 { 500 timespec32_t verf; 501 int error; 502 extern void rfs4_attr_init(); 503 extern krwlock_t rfs4_deleg_policy_lock; 504 505 /* 506 * The following algorithm attempts to find a unique verifier 507 * to be used as the write verifier returned from the server 508 * to the client. It is important that this verifier change 509 * whenever the server reboots. Of secondary importance, it 510 * is important for the verifier to be unique between two 511 * different servers. 512 * 513 * Thus, an attempt is made to use the system hostid and the 514 * current time in seconds when the nfssrv kernel module is 515 * loaded. It is assumed that an NFS server will not be able 516 * to boot and then to reboot in less than a second. If the 517 * hostid has not been set, then the current high resolution 518 * time is used. This will ensure different verifiers each 519 * time the server reboots and minimize the chances that two 520 * different servers will have the same verifier. 521 * XXX - this is broken on LP64 kernels. 522 */ 523 verf.tv_sec = (time_t)nfs_atoi(hw_serial); 524 if (verf.tv_sec != 0) { 525 verf.tv_nsec = gethrestime_sec(); 526 } else { 527 timespec_t tverf; 528 529 gethrestime(&tverf); 530 verf.tv_sec = (time_t)tverf.tv_sec; 531 verf.tv_nsec = tverf.tv_nsec; 532 } 533 534 Write4verf = *(uint64_t *)&verf; 535 536 rfs4_attr_init(); 537 mutex_init(&rfs4_deleg_lock, NULL, MUTEX_DEFAULT, NULL); 538 539 /* Used to manage create/destroy of server state */ 540 mutex_init(&rfs4_state_lock, NULL, MUTEX_DEFAULT, NULL); 541 542 /* Used to manage access to server instance linked list */ 543 mutex_init(&rfs4_servinst_lock, NULL, MUTEX_DEFAULT, NULL); 544 545 /* Used to manage access to rfs4_deleg_policy */ 546 rw_init(&rfs4_deleg_policy_lock, NULL, RW_DEFAULT, NULL); 547 548 error = fem_create("deleg_rdops", nfs4_rd_deleg_tmpl, &deleg_rdops); 549 if (error != 0) { 550 rfs4_disable_delegation(); 551 } else { 552 error = fem_create("deleg_wrops", nfs4_wr_deleg_tmpl, 553 &deleg_wrops); 554 if (error != 0) { 555 rfs4_disable_delegation(); 556 fem_free(deleg_rdops); 557 } 558 } 559 560 nfs4_srv_caller_id = fs_new_caller_id(); 561 562 lockt_sysid = lm_alloc_sysidt(); 563 564 vsd_create(&nfs4_srv_vkey, NULL); 565 566 return (0); 567 } 568 569 void 570 rfs4_srvrfini(void) 571 { 572 extern krwlock_t rfs4_deleg_policy_lock; 573 574 if (lockt_sysid != LM_NOSYSID) { 575 lm_free_sysidt(lockt_sysid); 576 lockt_sysid = LM_NOSYSID; 577 } 578 579 mutex_destroy(&rfs4_deleg_lock); 580 mutex_destroy(&rfs4_state_lock); 581 rw_destroy(&rfs4_deleg_policy_lock); 582 583 fem_free(deleg_rdops); 584 fem_free(deleg_wrops); 585 } 586 587 void 588 rfs4_init_compound_state(struct compound_state *cs) 589 { 590 bzero(cs, sizeof (*cs)); 591 cs->cont = TRUE; 592 cs->access = CS_ACCESS_DENIED; 593 cs->deleg = FALSE; 594 cs->mandlock = FALSE; 595 cs->fh.nfs_fh4_val = cs->fhbuf; 596 } 597 598 void 599 rfs4_grace_start(rfs4_servinst_t *sip) 600 { 601 time_t now = gethrestime_sec(); 602 603 rw_enter(&sip->rwlock, RW_WRITER); 604 sip->start_time = now; 605 sip->grace_period = rfs4_grace_period; 606 rw_exit(&sip->rwlock); 607 } 608 609 /* 610 * returns true if the instance's grace period has never been started 611 */ 612 int 613 rfs4_servinst_grace_new(rfs4_servinst_t *sip) 614 { 615 time_t start_time; 616 617 rw_enter(&sip->rwlock, RW_READER); 618 start_time = sip->start_time; 619 rw_exit(&sip->rwlock); 620 621 return (start_time == 0); 622 } 623 624 /* 625 * Indicates if server instance is within the 626 * grace period. 627 */ 628 int 629 rfs4_servinst_in_grace(rfs4_servinst_t *sip) 630 { 631 time_t grace_expiry; 632 633 rw_enter(&sip->rwlock, RW_READER); 634 grace_expiry = sip->start_time + sip->grace_period; 635 rw_exit(&sip->rwlock); 636 637 return (gethrestime_sec() < grace_expiry); 638 } 639 640 int 641 rfs4_clnt_in_grace(rfs4_client_t *cp) 642 { 643 ASSERT(rfs4_dbe_refcnt(cp->dbe) > 0); 644 645 return (rfs4_servinst_in_grace(cp->server_instance)); 646 } 647 648 /* 649 * reset all currently active grace periods 650 */ 651 void 652 rfs4_grace_reset_all(void) 653 { 654 rfs4_servinst_t *sip; 655 656 mutex_enter(&rfs4_servinst_lock); 657 for (sip = rfs4_cur_servinst; sip != NULL; sip = sip->prev) 658 if (rfs4_servinst_in_grace(sip)) 659 rfs4_grace_start(sip); 660 mutex_exit(&rfs4_servinst_lock); 661 } 662 663 /* 664 * start any new instances' grace periods 665 */ 666 void 667 rfs4_grace_start_new(void) 668 { 669 rfs4_servinst_t *sip; 670 671 mutex_enter(&rfs4_servinst_lock); 672 for (sip = rfs4_cur_servinst; sip != NULL; sip = sip->prev) 673 if (rfs4_servinst_grace_new(sip)) 674 rfs4_grace_start(sip); 675 mutex_exit(&rfs4_servinst_lock); 676 } 677 678 static rfs4_dss_path_t * 679 rfs4_dss_newpath(rfs4_servinst_t *sip, char *path, unsigned index) 680 { 681 size_t len; 682 rfs4_dss_path_t *dss_path; 683 684 dss_path = kmem_alloc(sizeof (rfs4_dss_path_t), KM_SLEEP); 685 686 /* 687 * Take a copy of the string, since the original may be overwritten. 688 * Sadly, no strdup() in the kernel. 689 */ 690 /* allow for NUL */ 691 len = strlen(path) + 1; 692 dss_path->path = kmem_alloc(len, KM_SLEEP); 693 (void) strlcpy(dss_path->path, path, len); 694 695 /* associate with servinst */ 696 dss_path->sip = sip; 697 dss_path->index = index; 698 699 /* 700 * Add to list of served paths. 701 * No locking required, as we're only ever called at startup. 702 */ 703 if (rfs4_dss_pathlist == NULL) { 704 /* this is the first dss_path_t */ 705 706 /* needed for insque/remque */ 707 dss_path->next = dss_path->prev = dss_path; 708 709 rfs4_dss_pathlist = dss_path; 710 } else { 711 insque(dss_path, rfs4_dss_pathlist); 712 } 713 714 return (dss_path); 715 } 716 717 /* 718 * Create a new server instance, and make it the currently active instance. 719 * Note that starting the grace period too early will reduce the clients' 720 * recovery window. 721 */ 722 void 723 rfs4_servinst_create(int start_grace, int dss_npaths, char **dss_paths) 724 { 725 unsigned i; 726 rfs4_servinst_t *sip; 727 rfs4_oldstate_t *oldstate; 728 729 sip = kmem_alloc(sizeof (rfs4_servinst_t), KM_SLEEP); 730 rw_init(&sip->rwlock, NULL, RW_DEFAULT, NULL); 731 732 sip->start_time = (time_t)0; 733 sip->grace_period = (time_t)0; 734 sip->next = NULL; 735 sip->prev = NULL; 736 737 rw_init(&sip->oldstate_lock, NULL, RW_DEFAULT, NULL); 738 /* 739 * This initial dummy entry is required to setup for insque/remque. 740 * It must be skipped over whenever the list is traversed. 741 */ 742 oldstate = kmem_alloc(sizeof (rfs4_oldstate_t), KM_SLEEP); 743 /* insque/remque require initial list entry to be self-terminated */ 744 oldstate->next = oldstate; 745 oldstate->prev = oldstate; 746 sip->oldstate = oldstate; 747 748 749 sip->dss_npaths = dss_npaths; 750 sip->dss_paths = kmem_alloc(dss_npaths * 751 sizeof (rfs4_dss_path_t *), KM_SLEEP); 752 753 for (i = 0; i < dss_npaths; i++) { 754 sip->dss_paths[i] = rfs4_dss_newpath(sip, dss_paths[i], i); 755 } 756 757 mutex_enter(&rfs4_servinst_lock); 758 if (rfs4_cur_servinst != NULL) { 759 /* add to linked list */ 760 sip->prev = rfs4_cur_servinst; 761 rfs4_cur_servinst->next = sip; 762 } 763 if (start_grace) 764 rfs4_grace_start(sip); 765 /* make the new instance "current" */ 766 rfs4_cur_servinst = sip; 767 768 mutex_exit(&rfs4_servinst_lock); 769 } 770 771 /* 772 * In future, we might add a rfs4_servinst_destroy(sip) but, for now, destroy 773 * all instances directly. 774 */ 775 void 776 rfs4_servinst_destroy_all(void) 777 { 778 rfs4_servinst_t *sip, *prev, *current; 779 #ifdef DEBUG 780 int n = 0; 781 #endif 782 783 mutex_enter(&rfs4_servinst_lock); 784 ASSERT(rfs4_cur_servinst != NULL); 785 current = rfs4_cur_servinst; 786 rfs4_cur_servinst = NULL; 787 for (sip = current; sip != NULL; sip = prev) { 788 prev = sip->prev; 789 rw_destroy(&sip->rwlock); 790 if (sip->oldstate) 791 kmem_free(sip->oldstate, sizeof (rfs4_oldstate_t)); 792 if (sip->dss_paths) 793 kmem_free(sip->dss_paths, 794 sip->dss_npaths * sizeof (rfs4_dss_path_t *)); 795 kmem_free(sip, sizeof (rfs4_servinst_t)); 796 #ifdef DEBUG 797 n++; 798 #endif 799 } 800 mutex_exit(&rfs4_servinst_lock); 801 } 802 803 /* 804 * Assign the current server instance to a client_t. 805 * Should be called with cp->dbe held. 806 */ 807 void 808 rfs4_servinst_assign(rfs4_client_t *cp, rfs4_servinst_t *sip) 809 { 810 ASSERT(rfs4_dbe_refcnt(cp->dbe) > 0); 811 812 /* 813 * The lock ensures that if the current instance is in the process 814 * of changing, we will see the new one. 815 */ 816 mutex_enter(&rfs4_servinst_lock); 817 cp->server_instance = sip; 818 mutex_exit(&rfs4_servinst_lock); 819 } 820 821 rfs4_servinst_t * 822 rfs4_servinst(rfs4_client_t *cp) 823 { 824 ASSERT(rfs4_dbe_refcnt(cp->dbe) > 0); 825 826 return (cp->server_instance); 827 } 828 829 /* ARGSUSED */ 830 static void 831 nullfree(caddr_t resop) 832 { 833 } 834 835 /* 836 * This is a fall-through for invalid or not implemented (yet) ops 837 */ 838 /* ARGSUSED */ 839 static void 840 rfs4_op_inval(nfs_argop4 *argop, nfs_resop4 *resop, struct svc_req *req, 841 struct compound_state *cs) 842 { 843 *cs->statusp = *((nfsstat4 *)&(resop)->nfs_resop4_u) = NFS4ERR_INVAL; 844 } 845 846 /* 847 * Check if the security flavor, nfsnum, is in the flavor_list. 848 */ 849 bool_t 850 in_flavor_list(int nfsnum, int *flavor_list, int count) 851 { 852 int i; 853 854 for (i = 0; i < count; i++) { 855 if (nfsnum == flavor_list[i]) 856 return (TRUE); 857 } 858 return (FALSE); 859 } 860 861 /* 862 * Used by rfs4_op_secinfo to get the security information from the 863 * export structure associated with the component. 864 */ 865 /* ARGSUSED */ 866 static nfsstat4 867 do_rfs4_op_secinfo(struct compound_state *cs, char *nm, SECINFO4res *resp) 868 { 869 int error, different_export = 0; 870 vnode_t *dvp, *vp, *tvp; 871 struct exportinfo *exi = NULL; 872 fid_t fid; 873 uint_t count, i; 874 secinfo4 *resok_val; 875 struct secinfo *secp; 876 seconfig_t *si; 877 bool_t did_traverse; 878 int dotdot, walk; 879 880 dvp = cs->vp; 881 dotdot = (nm[0] == '.' && nm[1] == '.' && nm[2] == '\0'); 882 883 /* 884 * If dotdotting, then need to check whether it's above the 885 * root of a filesystem, or above an export point. 886 */ 887 if (dotdot) { 888 889 /* 890 * If dotdotting at the root of a filesystem, then 891 * need to traverse back to the mounted-on filesystem 892 * and do the dotdot lookup there. 893 */ 894 if (cs->vp->v_flag & VROOT) { 895 896 /* 897 * If at the system root, then can 898 * go up no further. 899 */ 900 if (VN_CMP(dvp, rootdir)) 901 return (puterrno4(ENOENT)); 902 903 /* 904 * Traverse back to the mounted-on filesystem 905 */ 906 dvp = untraverse(cs->vp); 907 908 /* 909 * Set the different_export flag so we remember 910 * to pick up a new exportinfo entry for 911 * this new filesystem. 912 */ 913 different_export = 1; 914 } else { 915 916 /* 917 * If dotdotting above an export point then set 918 * the different_export to get new export info. 919 */ 920 different_export = nfs_exported(cs->exi, cs->vp); 921 } 922 } 923 924 /* 925 * Get the vnode for the component "nm". 926 */ 927 error = VOP_LOOKUP(dvp, nm, &vp, NULL, 0, NULL, cs->cr, 928 NULL, NULL, NULL); 929 if (error) 930 return (puterrno4(error)); 931 932 /* 933 * If the vnode is in a pseudo filesystem, or if the security flavor 934 * used in the request is valid but not an explicitly shared flavor, 935 * or the access bit indicates that this is a limited access, 936 * check whether this vnode is visible. 937 */ 938 if (!different_export && 939 (PSEUDO(cs->exi) || ! is_exported_sec(cs->nfsflavor, cs->exi) || 940 cs->access & CS_ACCESS_LIMITED)) { 941 if (! nfs_visible(cs->exi, vp, &different_export)) { 942 VN_RELE(vp); 943 return (puterrno4(ENOENT)); 944 } 945 } 946 947 /* 948 * If it's a mountpoint, then traverse it. 949 */ 950 if (vn_ismntpt(vp)) { 951 tvp = vp; 952 if ((error = traverse(&tvp)) != 0) { 953 VN_RELE(vp); 954 return (puterrno4(error)); 955 } 956 /* remember that we had to traverse mountpoint */ 957 did_traverse = TRUE; 958 vp = tvp; 959 different_export = 1; 960 } else if (vp->v_vfsp != dvp->v_vfsp) { 961 /* 962 * If vp isn't a mountpoint and the vfs ptrs aren't the same, 963 * then vp is probably an LOFS object. We don't need the 964 * realvp, we just need to know that we might have crossed 965 * a server fs boundary and need to call checkexport4. 966 * (LOFS lookup hides server fs mountpoints, and actually calls 967 * traverse) 968 */ 969 different_export = 1; 970 did_traverse = FALSE; 971 } 972 973 /* 974 * Get the export information for it. 975 */ 976 if (different_export) { 977 978 bzero(&fid, sizeof (fid)); 979 fid.fid_len = MAXFIDSZ; 980 error = vop_fid_pseudo(vp, &fid); 981 if (error) { 982 VN_RELE(vp); 983 return (puterrno4(error)); 984 } 985 986 if (dotdot) 987 exi = nfs_vptoexi(NULL, vp, cs->cr, &walk, NULL, TRUE); 988 else 989 exi = checkexport4(&vp->v_vfsp->vfs_fsid, &fid, vp); 990 991 if (exi == NULL) { 992 if (did_traverse == TRUE) { 993 /* 994 * If this vnode is a mounted-on vnode, 995 * but the mounted-on file system is not 996 * exported, send back the secinfo for 997 * the exported node that the mounted-on 998 * vnode lives in. 999 */ 1000 exi = cs->exi; 1001 } else { 1002 VN_RELE(vp); 1003 return (puterrno4(EACCES)); 1004 } 1005 } 1006 } else { 1007 exi = cs->exi; 1008 } 1009 ASSERT(exi != NULL); 1010 1011 1012 /* 1013 * Create the secinfo result based on the security information 1014 * from the exportinfo structure (exi). 1015 * 1016 * Return all flavors for a pseudo node. 1017 * For a real export node, return the flavor that the client 1018 * has access with. 1019 */ 1020 ASSERT(RW_LOCK_HELD(&exported_lock)); 1021 if (PSEUDO(exi)) { 1022 count = exi->exi_export.ex_seccnt; /* total sec count */ 1023 resok_val = kmem_alloc(count * sizeof (secinfo4), KM_SLEEP); 1024 secp = exi->exi_export.ex_secinfo; 1025 1026 for (i = 0; i < count; i++) { 1027 si = &secp[i].s_secinfo; 1028 resok_val[i].flavor = si->sc_rpcnum; 1029 if (resok_val[i].flavor == RPCSEC_GSS) { 1030 rpcsec_gss_info *info; 1031 1032 info = &resok_val[i].flavor_info; 1033 info->qop = si->sc_qop; 1034 info->service = (rpc_gss_svc_t)si->sc_service; 1035 1036 /* get oid opaque data */ 1037 info->oid.sec_oid4_len = 1038 si->sc_gss_mech_type->length; 1039 info->oid.sec_oid4_val = kmem_alloc( 1040 si->sc_gss_mech_type->length, KM_SLEEP); 1041 bcopy( 1042 si->sc_gss_mech_type->elements, 1043 info->oid.sec_oid4_val, 1044 info->oid.sec_oid4_len); 1045 } 1046 } 1047 resp->SECINFO4resok_len = count; 1048 resp->SECINFO4resok_val = resok_val; 1049 } else { 1050 int ret_cnt = 0, k = 0; 1051 int *flavor_list; 1052 1053 count = exi->exi_export.ex_seccnt; /* total sec count */ 1054 secp = exi->exi_export.ex_secinfo; 1055 1056 flavor_list = kmem_alloc(count * sizeof (int), KM_SLEEP); 1057 /* find out which flavors to return */ 1058 for (i = 0; i < count; i ++) { 1059 int access, flavor, perm; 1060 1061 flavor = secp[i].s_secinfo.sc_nfsnum; 1062 perm = secp[i].s_flags; 1063 1064 access = nfsauth4_secinfo_access(exi, cs->req, 1065 flavor, perm); 1066 1067 if (! (access & NFSAUTH_DENIED) && 1068 ! (access & NFSAUTH_WRONGSEC)) { 1069 flavor_list[ret_cnt] = flavor; 1070 ret_cnt++; 1071 } 1072 } 1073 1074 /* Create the returning SECINFO value */ 1075 resok_val = kmem_alloc(ret_cnt * sizeof (secinfo4), KM_SLEEP); 1076 1077 for (i = 0; i < count; i++) { 1078 /* 1079 * If the flavor is in the flavor list, 1080 * fill in resok_val. 1081 */ 1082 si = &secp[i].s_secinfo; 1083 if (in_flavor_list(si->sc_nfsnum, 1084 flavor_list, ret_cnt)) { 1085 resok_val[k].flavor = si->sc_rpcnum; 1086 if (resok_val[k].flavor == RPCSEC_GSS) { 1087 rpcsec_gss_info *info; 1088 1089 info = &resok_val[k].flavor_info; 1090 info->qop = si->sc_qop; 1091 info->service = (rpc_gss_svc_t) 1092 si->sc_service; 1093 1094 /* get oid opaque data */ 1095 info->oid.sec_oid4_len = 1096 si->sc_gss_mech_type->length; 1097 info->oid.sec_oid4_val = kmem_alloc( 1098 si->sc_gss_mech_type->length, 1099 KM_SLEEP); 1100 bcopy(si->sc_gss_mech_type->elements, 1101 info->oid.sec_oid4_val, 1102 info->oid.sec_oid4_len); 1103 } 1104 k++; 1105 } 1106 if (k >= ret_cnt) 1107 break; 1108 } 1109 resp->SECINFO4resok_len = ret_cnt; 1110 resp->SECINFO4resok_val = resok_val; 1111 kmem_free(flavor_list, count * sizeof (int)); 1112 } 1113 1114 VN_RELE(vp); 1115 return (NFS4_OK); 1116 } 1117 1118 /* 1119 * SECINFO (Operation 33): Obtain required security information on 1120 * the component name in the format of (security-mechanism-oid, qop, service) 1121 * triplets. 1122 */ 1123 /* ARGSUSED */ 1124 static void 1125 rfs4_op_secinfo(nfs_argop4 *argop, nfs_resop4 *resop, struct svc_req *req, 1126 struct compound_state *cs) 1127 { 1128 SECINFO4args *args = &argop->nfs_argop4_u.opsecinfo; 1129 SECINFO4res *resp = &resop->nfs_resop4_u.opsecinfo; 1130 utf8string *utfnm = &args->name; 1131 uint_t len; 1132 char *nm; 1133 1134 DTRACE_NFSV4_2(op__secinfo__start, struct compound_state *, cs, 1135 SECINFO4args *, args); 1136 1137 /* 1138 * Current file handle (cfh) should have been set before getting 1139 * into this function. If not, return error. 1140 */ 1141 if (cs->vp == NULL) { 1142 *cs->statusp = resp->status = NFS4ERR_NOFILEHANDLE; 1143 goto out; 1144 } 1145 1146 if (cs->vp->v_type != VDIR) { 1147 *cs->statusp = resp->status = NFS4ERR_NOTDIR; 1148 goto out; 1149 } 1150 1151 /* 1152 * Verify the component name. If failed, error out, but 1153 * do not error out if the component name is a "..". 1154 * SECINFO will return its parents secinfo data for SECINFO "..". 1155 */ 1156 if (!utf8_dir_verify(utfnm)) { 1157 if (utfnm->utf8string_len != 2 || 1158 utfnm->utf8string_val[0] != '.' || 1159 utfnm->utf8string_val[1] != '.') { 1160 *cs->statusp = resp->status = NFS4ERR_INVAL; 1161 goto out; 1162 } 1163 } 1164 1165 nm = utf8_to_str(utfnm, &len, NULL); 1166 if (nm == NULL) { 1167 *cs->statusp = resp->status = NFS4ERR_INVAL; 1168 goto out; 1169 } 1170 1171 if (len > MAXNAMELEN) { 1172 *cs->statusp = resp->status = NFS4ERR_NAMETOOLONG; 1173 kmem_free(nm, len); 1174 goto out; 1175 } 1176 1177 *cs->statusp = resp->status = do_rfs4_op_secinfo(cs, nm, resp); 1178 1179 kmem_free(nm, len); 1180 1181 out: 1182 DTRACE_NFSV4_2(op__secinfo__done, struct compound_state *, cs, 1183 SECINFO4res *, resp); 1184 } 1185 1186 /* 1187 * Free SECINFO result. 1188 */ 1189 /* ARGSUSED */ 1190 static void 1191 rfs4_op_secinfo_free(nfs_resop4 *resop) 1192 { 1193 SECINFO4res *resp = &resop->nfs_resop4_u.opsecinfo; 1194 int count, i; 1195 secinfo4 *resok_val; 1196 1197 /* If this is not an Ok result, nothing to free. */ 1198 if (resp->status != NFS4_OK) { 1199 return; 1200 } 1201 1202 count = resp->SECINFO4resok_len; 1203 resok_val = resp->SECINFO4resok_val; 1204 1205 for (i = 0; i < count; i++) { 1206 if (resok_val[i].flavor == RPCSEC_GSS) { 1207 rpcsec_gss_info *info; 1208 1209 info = &resok_val[i].flavor_info; 1210 kmem_free(info->oid.sec_oid4_val, 1211 info->oid.sec_oid4_len); 1212 } 1213 } 1214 kmem_free(resok_val, count * sizeof (secinfo4)); 1215 resp->SECINFO4resok_len = 0; 1216 resp->SECINFO4resok_val = NULL; 1217 } 1218 1219 /* ARGSUSED */ 1220 static void 1221 rfs4_op_access(nfs_argop4 *argop, nfs_resop4 *resop, struct svc_req *req, 1222 struct compound_state *cs) 1223 { 1224 ACCESS4args *args = &argop->nfs_argop4_u.opaccess; 1225 ACCESS4res *resp = &resop->nfs_resop4_u.opaccess; 1226 int error; 1227 vnode_t *vp; 1228 struct vattr va; 1229 int checkwriteperm; 1230 cred_t *cr = cs->cr; 1231 bslabel_t *clabel, *slabel; 1232 ts_label_t *tslabel; 1233 boolean_t admin_low_client; 1234 1235 DTRACE_NFSV4_2(op__access__start, struct compound_state *, cs, 1236 ACCESS4args *, args); 1237 1238 #if 0 /* XXX allow access even if !cs->access. Eventually only pseudo fs */ 1239 if (cs->access == CS_ACCESS_DENIED) { 1240 *cs->statusp = resp->status = NFS4ERR_ACCESS; 1241 goto out; 1242 } 1243 #endif 1244 if (cs->vp == NULL) { 1245 *cs->statusp = resp->status = NFS4ERR_NOFILEHANDLE; 1246 goto out; 1247 } 1248 1249 ASSERT(cr != NULL); 1250 1251 vp = cs->vp; 1252 1253 /* 1254 * If the file system is exported read only, it is not appropriate 1255 * to check write permissions for regular files and directories. 1256 * Special files are interpreted by the client, so the underlying 1257 * permissions are sent back to the client for interpretation. 1258 */ 1259 if (rdonly4(cs->exi, cs->vp, req) && 1260 (vp->v_type == VREG || vp->v_type == VDIR)) 1261 checkwriteperm = 0; 1262 else 1263 checkwriteperm = 1; 1264 1265 /* 1266 * XXX 1267 * We need the mode so that we can correctly determine access 1268 * permissions relative to a mandatory lock file. Access to 1269 * mandatory lock files is denied on the server, so it might 1270 * as well be reflected to the server during the open. 1271 */ 1272 va.va_mask = AT_MODE; 1273 error = VOP_GETATTR(vp, &va, 0, cr, NULL); 1274 if (error) { 1275 *cs->statusp = resp->status = puterrno4(error); 1276 goto out; 1277 } 1278 resp->access = 0; 1279 resp->supported = 0; 1280 1281 if (is_system_labeled()) { 1282 ASSERT(req->rq_label != NULL); 1283 clabel = req->rq_label; 1284 DTRACE_PROBE2(tx__rfs4__log__info__opaccess__clabel, char *, 1285 "got client label from request(1)", 1286 struct svc_req *, req); 1287 if (!blequal(&l_admin_low->tsl_label, clabel)) { 1288 if ((tslabel = nfs_getflabel(vp)) == NULL) { 1289 *cs->statusp = resp->status = puterrno4(EACCES); 1290 goto out; 1291 } 1292 slabel = label2bslabel(tslabel); 1293 DTRACE_PROBE3(tx__rfs4__log__info__opaccess__slabel, 1294 char *, "got server label(1) for vp(2)", 1295 bslabel_t *, slabel, vnode_t *, vp); 1296 1297 admin_low_client = B_FALSE; 1298 } else 1299 admin_low_client = B_TRUE; 1300 } 1301 1302 if (args->access & ACCESS4_READ) { 1303 error = VOP_ACCESS(vp, VREAD, 0, cr, NULL); 1304 if (!error && !MANDLOCK(vp, va.va_mode) && 1305 (!is_system_labeled() || admin_low_client || 1306 bldominates(clabel, slabel))) 1307 resp->access |= ACCESS4_READ; 1308 resp->supported |= ACCESS4_READ; 1309 } 1310 if ((args->access & ACCESS4_LOOKUP) && vp->v_type == VDIR) { 1311 error = VOP_ACCESS(vp, VEXEC, 0, cr, NULL); 1312 if (!error && (!is_system_labeled() || admin_low_client || 1313 bldominates(clabel, slabel))) 1314 resp->access |= ACCESS4_LOOKUP; 1315 resp->supported |= ACCESS4_LOOKUP; 1316 } 1317 if (checkwriteperm && 1318 (args->access & (ACCESS4_MODIFY|ACCESS4_EXTEND))) { 1319 error = VOP_ACCESS(vp, VWRITE, 0, cr, NULL); 1320 if (!error && !MANDLOCK(vp, va.va_mode) && 1321 (!is_system_labeled() || admin_low_client || 1322 blequal(clabel, slabel))) 1323 resp->access |= 1324 (args->access & (ACCESS4_MODIFY | ACCESS4_EXTEND)); 1325 resp->supported |= (ACCESS4_MODIFY | ACCESS4_EXTEND); 1326 } 1327 1328 if (checkwriteperm && 1329 (args->access & ACCESS4_DELETE) && vp->v_type == VDIR) { 1330 error = VOP_ACCESS(vp, VWRITE, 0, cr, NULL); 1331 if (!error && (!is_system_labeled() || admin_low_client || 1332 blequal(clabel, slabel))) 1333 resp->access |= ACCESS4_DELETE; 1334 resp->supported |= ACCESS4_DELETE; 1335 } 1336 if (args->access & ACCESS4_EXECUTE && vp->v_type != VDIR) { 1337 error = VOP_ACCESS(vp, VEXEC, 0, cr, NULL); 1338 if (!error && !MANDLOCK(vp, va.va_mode) && 1339 (!is_system_labeled() || admin_low_client || 1340 bldominates(clabel, slabel))) 1341 resp->access |= ACCESS4_EXECUTE; 1342 resp->supported |= ACCESS4_EXECUTE; 1343 } 1344 1345 if (is_system_labeled() && !admin_low_client) 1346 label_rele(tslabel); 1347 1348 *cs->statusp = resp->status = NFS4_OK; 1349 out: 1350 DTRACE_NFSV4_2(op__access__done, struct compound_state *, cs, 1351 ACCESS4res *, resp); 1352 } 1353 1354 /* ARGSUSED */ 1355 static void 1356 rfs4_op_commit(nfs_argop4 *argop, nfs_resop4 *resop, struct svc_req *req, 1357 struct compound_state *cs) 1358 { 1359 COMMIT4args *args = &argop->nfs_argop4_u.opcommit; 1360 COMMIT4res *resp = &resop->nfs_resop4_u.opcommit; 1361 int error; 1362 vnode_t *vp = cs->vp; 1363 cred_t *cr = cs->cr; 1364 vattr_t va; 1365 1366 DTRACE_NFSV4_2(op__commit__start, struct compound_state *, cs, 1367 COMMIT4args *, args); 1368 1369 if (vp == NULL) { 1370 *cs->statusp = resp->status = NFS4ERR_NOFILEHANDLE; 1371 goto out; 1372 } 1373 if (cs->access == CS_ACCESS_DENIED) { 1374 *cs->statusp = resp->status = NFS4ERR_ACCESS; 1375 goto out; 1376 } 1377 1378 if (args->offset + args->count < args->offset) { 1379 *cs->statusp = resp->status = NFS4ERR_INVAL; 1380 goto out; 1381 } 1382 1383 va.va_mask = AT_UID; 1384 error = VOP_GETATTR(vp, &va, 0, cr, NULL); 1385 1386 /* 1387 * If we can't get the attributes, then we can't do the 1388 * right access checking. So, we'll fail the request. 1389 */ 1390 if (error) { 1391 *cs->statusp = resp->status = puterrno4(error); 1392 goto out; 1393 } 1394 if (rdonly4(cs->exi, cs->vp, req)) { 1395 *cs->statusp = resp->status = NFS4ERR_ROFS; 1396 goto out; 1397 } 1398 1399 if (vp->v_type != VREG) { 1400 if (vp->v_type == VDIR) 1401 resp->status = NFS4ERR_ISDIR; 1402 else 1403 resp->status = NFS4ERR_INVAL; 1404 *cs->statusp = resp->status; 1405 goto out; 1406 } 1407 1408 if (crgetuid(cr) != va.va_uid && 1409 (error = VOP_ACCESS(vp, VWRITE, 0, cs->cr, NULL))) { 1410 *cs->statusp = resp->status = puterrno4(error); 1411 goto out; 1412 } 1413 1414 error = VOP_PUTPAGE(vp, args->offset, args->count, 0, cr, NULL); 1415 if (!error) 1416 error = VOP_FSYNC(vp, FNODSYNC, cr, NULL); 1417 1418 if (error) { 1419 *cs->statusp = resp->status = puterrno4(error); 1420 goto out; 1421 } 1422 1423 *cs->statusp = resp->status = NFS4_OK; 1424 resp->writeverf = Write4verf; 1425 out: 1426 DTRACE_NFSV4_2(op__commit__done, struct compound_state *, cs, 1427 COMMIT4res *, resp); 1428 } 1429 1430 /* 1431 * rfs4_op_mknod is called from rfs4_op_create after all initial verification 1432 * was completed. It does the nfsv4 create for special files. 1433 */ 1434 /* ARGSUSED */ 1435 static vnode_t * 1436 do_rfs4_op_mknod(CREATE4args *args, CREATE4res *resp, struct svc_req *req, 1437 struct compound_state *cs, vattr_t *vap, char *nm) 1438 { 1439 int error; 1440 cred_t *cr = cs->cr; 1441 vnode_t *dvp = cs->vp; 1442 vnode_t *vp = NULL; 1443 int mode; 1444 enum vcexcl excl; 1445 1446 switch (args->type) { 1447 case NF4CHR: 1448 case NF4BLK: 1449 if (secpolicy_sys_devices(cr) != 0) { 1450 *cs->statusp = resp->status = NFS4ERR_PERM; 1451 return (NULL); 1452 } 1453 if (args->type == NF4CHR) 1454 vap->va_type = VCHR; 1455 else 1456 vap->va_type = VBLK; 1457 vap->va_rdev = makedevice(args->ftype4_u.devdata.specdata1, 1458 args->ftype4_u.devdata.specdata2); 1459 vap->va_mask |= AT_RDEV; 1460 break; 1461 case NF4SOCK: 1462 vap->va_type = VSOCK; 1463 break; 1464 case NF4FIFO: 1465 vap->va_type = VFIFO; 1466 break; 1467 default: 1468 *cs->statusp = resp->status = NFS4ERR_BADTYPE; 1469 return (NULL); 1470 } 1471 1472 /* 1473 * Must specify the mode. 1474 */ 1475 if (!(vap->va_mask & AT_MODE)) { 1476 *cs->statusp = resp->status = NFS4ERR_INVAL; 1477 return (NULL); 1478 } 1479 1480 excl = EXCL; 1481 1482 mode = 0; 1483 1484 error = VOP_CREATE(dvp, nm, vap, excl, mode, &vp, cr, 0, NULL, NULL); 1485 if (error) { 1486 *cs->statusp = resp->status = puterrno4(error); 1487 return (NULL); 1488 } 1489 return (vp); 1490 } 1491 1492 /* 1493 * nfsv4 create is used to create non-regular files. For regular files, 1494 * use nfsv4 open. 1495 */ 1496 /* ARGSUSED */ 1497 static void 1498 rfs4_op_create(nfs_argop4 *argop, nfs_resop4 *resop, struct svc_req *req, 1499 struct compound_state *cs) 1500 { 1501 CREATE4args *args = &argop->nfs_argop4_u.opcreate; 1502 CREATE4res *resp = &resop->nfs_resop4_u.opcreate; 1503 int error; 1504 struct vattr bva, iva, iva2, ava, *vap; 1505 cred_t *cr = cs->cr; 1506 vnode_t *dvp = cs->vp; 1507 vnode_t *vp = NULL; 1508 vnode_t *realvp; 1509 char *nm, *lnm; 1510 uint_t len, llen; 1511 int syncval = 0; 1512 struct nfs4_svgetit_arg sarg; 1513 struct nfs4_ntov_table ntov; 1514 struct statvfs64 sb; 1515 nfsstat4 status; 1516 1517 DTRACE_NFSV4_2(op__create__start, struct compound_state *, cs, 1518 CREATE4args *, args); 1519 1520 resp->attrset = 0; 1521 1522 if (dvp == NULL) { 1523 *cs->statusp = resp->status = NFS4ERR_NOFILEHANDLE; 1524 goto out; 1525 } 1526 1527 /* 1528 * If there is an unshared filesystem mounted on this vnode, 1529 * do not allow to create an object in this directory. 1530 */ 1531 if (vn_ismntpt(dvp)) { 1532 *cs->statusp = resp->status = NFS4ERR_ACCESS; 1533 goto out; 1534 } 1535 1536 /* Verify that type is correct */ 1537 switch (args->type) { 1538 case NF4LNK: 1539 case NF4BLK: 1540 case NF4CHR: 1541 case NF4SOCK: 1542 case NF4FIFO: 1543 case NF4DIR: 1544 break; 1545 default: 1546 *cs->statusp = resp->status = NFS4ERR_BADTYPE; 1547 goto out; 1548 }; 1549 1550 if (cs->access == CS_ACCESS_DENIED) { 1551 *cs->statusp = resp->status = NFS4ERR_ACCESS; 1552 goto out; 1553 } 1554 if (dvp->v_type != VDIR) { 1555 *cs->statusp = resp->status = NFS4ERR_NOTDIR; 1556 goto out; 1557 } 1558 if (!utf8_dir_verify(&args->objname)) { 1559 *cs->statusp = resp->status = NFS4ERR_INVAL; 1560 goto out; 1561 } 1562 1563 if (rdonly4(cs->exi, cs->vp, req)) { 1564 *cs->statusp = resp->status = NFS4ERR_ROFS; 1565 goto out; 1566 } 1567 1568 /* 1569 * Name of newly created object 1570 */ 1571 nm = utf8_to_fn(&args->objname, &len, NULL); 1572 if (nm == NULL) { 1573 *cs->statusp = resp->status = NFS4ERR_INVAL; 1574 goto out; 1575 } 1576 1577 if (len > MAXNAMELEN) { 1578 *cs->statusp = resp->status = NFS4ERR_NAMETOOLONG; 1579 kmem_free(nm, len); 1580 goto out; 1581 } 1582 1583 resp->attrset = 0; 1584 1585 sarg.sbp = &sb; 1586 nfs4_ntov_table_init(&ntov); 1587 1588 status = do_rfs4_set_attrs(&resp->attrset, 1589 &args->createattrs, cs, &sarg, &ntov, NFS4ATTR_SETIT); 1590 1591 if (sarg.vap->va_mask == 0 && status == NFS4_OK) 1592 status = NFS4ERR_INVAL; 1593 1594 if (status != NFS4_OK) { 1595 *cs->statusp = resp->status = status; 1596 kmem_free(nm, len); 1597 nfs4_ntov_table_free(&ntov, &sarg); 1598 resp->attrset = 0; 1599 goto out; 1600 } 1601 1602 /* Get "before" change value */ 1603 bva.va_mask = AT_CTIME|AT_SEQ; 1604 error = VOP_GETATTR(dvp, &bva, 0, cr, NULL); 1605 if (error) { 1606 *cs->statusp = resp->status = puterrno4(error); 1607 kmem_free(nm, len); 1608 nfs4_ntov_table_free(&ntov, &sarg); 1609 resp->attrset = 0; 1610 goto out; 1611 } 1612 NFS4_SET_FATTR4_CHANGE(resp->cinfo.before, bva.va_ctime) 1613 1614 vap = sarg.vap; 1615 1616 /* 1617 * Set default initial values for attributes when not specified 1618 * in createattrs. 1619 */ 1620 if ((vap->va_mask & AT_UID) == 0) { 1621 vap->va_uid = crgetuid(cr); 1622 vap->va_mask |= AT_UID; 1623 } 1624 if ((vap->va_mask & AT_GID) == 0) { 1625 vap->va_gid = crgetgid(cr); 1626 vap->va_mask |= AT_GID; 1627 } 1628 1629 vap->va_mask |= AT_TYPE; 1630 switch (args->type) { 1631 case NF4DIR: 1632 vap->va_type = VDIR; 1633 if ((vap->va_mask & AT_MODE) == 0) { 1634 vap->va_mode = 0700; /* default: owner rwx only */ 1635 vap->va_mask |= AT_MODE; 1636 } 1637 error = VOP_MKDIR(dvp, nm, vap, &vp, cr, NULL, 0, NULL); 1638 if (error) 1639 break; 1640 1641 /* 1642 * Get the initial "after" sequence number, if it fails, 1643 * set to zero 1644 */ 1645 iva.va_mask = AT_SEQ; 1646 if (VOP_GETATTR(dvp, &iva, 0, cs->cr, NULL)) 1647 iva.va_seq = 0; 1648 break; 1649 case NF4LNK: 1650 vap->va_type = VLNK; 1651 if ((vap->va_mask & AT_MODE) == 0) { 1652 vap->va_mode = 0700; /* default: owner rwx only */ 1653 vap->va_mask |= AT_MODE; 1654 } 1655 1656 /* 1657 * symlink names must be treated as data 1658 */ 1659 lnm = utf8_to_str(&args->ftype4_u.linkdata, &llen, NULL); 1660 1661 if (lnm == NULL) { 1662 *cs->statusp = resp->status = NFS4ERR_INVAL; 1663 kmem_free(nm, len); 1664 nfs4_ntov_table_free(&ntov, &sarg); 1665 resp->attrset = 0; 1666 goto out; 1667 } 1668 1669 if (llen > MAXPATHLEN) { 1670 *cs->statusp = resp->status = NFS4ERR_NAMETOOLONG; 1671 kmem_free(nm, len); 1672 kmem_free(lnm, llen); 1673 nfs4_ntov_table_free(&ntov, &sarg); 1674 resp->attrset = 0; 1675 goto out; 1676 } 1677 1678 error = VOP_SYMLINK(dvp, nm, vap, lnm, cr, NULL, 0); 1679 if (lnm != NULL) 1680 kmem_free(lnm, llen); 1681 if (error) 1682 break; 1683 1684 /* 1685 * Get the initial "after" sequence number, if it fails, 1686 * set to zero 1687 */ 1688 iva.va_mask = AT_SEQ; 1689 if (VOP_GETATTR(dvp, &iva, 0, cs->cr, NULL)) 1690 iva.va_seq = 0; 1691 1692 error = VOP_LOOKUP(dvp, nm, &vp, NULL, 0, NULL, cr, 1693 NULL, NULL, NULL); 1694 if (error) 1695 break; 1696 1697 /* 1698 * va_seq is not safe over VOP calls, check it again 1699 * if it has changed zero out iva to force atomic = FALSE. 1700 */ 1701 iva2.va_mask = AT_SEQ; 1702 if (VOP_GETATTR(dvp, &iva2, 0, cs->cr, NULL) || 1703 iva2.va_seq != iva.va_seq) 1704 iva.va_seq = 0; 1705 break; 1706 default: 1707 /* 1708 * probably a special file. 1709 */ 1710 if ((vap->va_mask & AT_MODE) == 0) { 1711 vap->va_mode = 0600; /* default: owner rw only */ 1712 vap->va_mask |= AT_MODE; 1713 } 1714 syncval = FNODSYNC; 1715 /* 1716 * We know this will only generate one VOP call 1717 */ 1718 vp = do_rfs4_op_mknod(args, resp, req, cs, vap, nm); 1719 1720 if (vp == NULL) { 1721 kmem_free(nm, len); 1722 nfs4_ntov_table_free(&ntov, &sarg); 1723 resp->attrset = 0; 1724 goto out; 1725 } 1726 1727 /* 1728 * Get the initial "after" sequence number, if it fails, 1729 * set to zero 1730 */ 1731 iva.va_mask = AT_SEQ; 1732 if (VOP_GETATTR(dvp, &iva, 0, cs->cr, NULL)) 1733 iva.va_seq = 0; 1734 1735 break; 1736 } 1737 kmem_free(nm, len); 1738 1739 if (error) { 1740 *cs->statusp = resp->status = puterrno4(error); 1741 } 1742 1743 /* 1744 * Force modified data and metadata out to stable storage. 1745 */ 1746 (void) VOP_FSYNC(dvp, 0, cr, NULL); 1747 1748 if (resp->status != NFS4_OK) { 1749 if (vp != NULL) 1750 VN_RELE(vp); 1751 nfs4_ntov_table_free(&ntov, &sarg); 1752 resp->attrset = 0; 1753 goto out; 1754 } 1755 1756 /* 1757 * Finish setup of cinfo response, "before" value already set. 1758 * Get "after" change value, if it fails, simply return the 1759 * before value. 1760 */ 1761 ava.va_mask = AT_CTIME|AT_SEQ; 1762 if (VOP_GETATTR(dvp, &ava, 0, cr, NULL)) { 1763 ava.va_ctime = bva.va_ctime; 1764 ava.va_seq = 0; 1765 } 1766 NFS4_SET_FATTR4_CHANGE(resp->cinfo.after, ava.va_ctime); 1767 1768 /* 1769 * True verification that object was created with correct 1770 * attrs is impossible. The attrs could have been changed 1771 * immediately after object creation. If attributes did 1772 * not verify, the only recourse for the server is to 1773 * destroy the object. Maybe if some attrs (like gid) 1774 * are set incorrectly, the object should be destroyed; 1775 * however, seems bad as a default policy. Do we really 1776 * want to destroy an object over one of the times not 1777 * verifying correctly? For these reasons, the server 1778 * currently sets bits in attrset for createattrs 1779 * that were set; however, no verification is done. 1780 * 1781 * vmask_to_nmask accounts for vattr bits set on create 1782 * [do_rfs4_set_attrs() only sets resp bits for 1783 * non-vattr/vfs bits.] 1784 * Mask off any bits set by default so as not to return 1785 * more attrset bits than were requested in createattrs 1786 */ 1787 nfs4_vmask_to_nmask(sarg.vap->va_mask, &resp->attrset); 1788 resp->attrset &= args->createattrs.attrmask; 1789 nfs4_ntov_table_free(&ntov, &sarg); 1790 1791 error = makefh4(&cs->fh, vp, cs->exi); 1792 if (error) { 1793 *cs->statusp = resp->status = puterrno4(error); 1794 } 1795 1796 /* 1797 * The cinfo.atomic = TRUE only if we got no errors, we have 1798 * non-zero va_seq's, and it has incremented by exactly one 1799 * during the creation and it didn't change during the VOP_LOOKUP 1800 * or VOP_FSYNC. 1801 */ 1802 if (!error && bva.va_seq && iva.va_seq && ava.va_seq && 1803 iva.va_seq == (bva.va_seq + 1) && iva.va_seq == ava.va_seq) 1804 resp->cinfo.atomic = TRUE; 1805 else 1806 resp->cinfo.atomic = FALSE; 1807 1808 /* 1809 * Force modified metadata out to stable storage. 1810 * 1811 * if a underlying vp exists, pass it to VOP_FSYNC 1812 */ 1813 if (VOP_REALVP(vp, &realvp, NULL) == 0) 1814 (void) VOP_FSYNC(realvp, syncval, cr, NULL); 1815 else 1816 (void) VOP_FSYNC(vp, syncval, cr, NULL); 1817 1818 if (resp->status != NFS4_OK) { 1819 VN_RELE(vp); 1820 goto out; 1821 } 1822 if (cs->vp) 1823 VN_RELE(cs->vp); 1824 1825 cs->vp = vp; 1826 *cs->statusp = resp->status = NFS4_OK; 1827 out: 1828 DTRACE_NFSV4_2(op__create__done, struct compound_state *, cs, 1829 CREATE4res *, resp); 1830 } 1831 1832 /*ARGSUSED*/ 1833 static void 1834 rfs4_op_delegpurge(nfs_argop4 *argop, nfs_resop4 *resop, struct svc_req *req, 1835 struct compound_state *cs) 1836 { 1837 DTRACE_NFSV4_2(op__delegpurge__start, struct compound_state *, cs, 1838 DELEGPURGE4args *, &argop->nfs_argop4_u.opdelegpurge); 1839 1840 rfs4_op_inval(argop, resop, req, cs); 1841 1842 DTRACE_NFSV4_2(op__delegpurge__done, struct compound_state *, cs, 1843 DELEGPURGE4res *, &resop->nfs_resop4_u.opdelegpurge); 1844 } 1845 1846 /*ARGSUSED*/ 1847 static void 1848 rfs4_op_delegreturn(nfs_argop4 *argop, nfs_resop4 *resop, struct svc_req *req, 1849 struct compound_state *cs) 1850 { 1851 DELEGRETURN4args *args = &argop->nfs_argop4_u.opdelegreturn; 1852 DELEGRETURN4res *resp = &resop->nfs_resop4_u.opdelegreturn; 1853 rfs4_deleg_state_t *dsp; 1854 nfsstat4 status; 1855 1856 DTRACE_NFSV4_2(op__delegreturn__start, struct compound_state *, cs, 1857 DELEGRETURN4args *, args); 1858 1859 status = rfs4_get_deleg_state(&args->deleg_stateid, &dsp); 1860 resp->status = *cs->statusp = status; 1861 if (status != NFS4_OK) 1862 goto out; 1863 1864 /* Ensure specified filehandle matches */ 1865 if (cs->vp != dsp->finfo->vp) { 1866 resp->status = *cs->statusp = NFS4ERR_BAD_STATEID; 1867 } else 1868 rfs4_return_deleg(dsp, FALSE); 1869 1870 rfs4_update_lease(dsp->client); 1871 1872 rfs4_deleg_state_rele(dsp); 1873 out: 1874 DTRACE_NFSV4_2(op__delegreturn__done, struct compound_state *, cs, 1875 DELEGRETURN4res *, resp); 1876 } 1877 1878 /* 1879 * Check to see if a given "flavor" is an explicitly shared flavor. 1880 * The assumption of this routine is the "flavor" is already a valid 1881 * flavor in the secinfo list of "exi". 1882 * 1883 * e.g. 1884 * # share -o sec=flavor1 /export 1885 * # share -o sec=flavor2 /export/home 1886 * 1887 * flavor2 is not an explicitly shared flavor for /export, 1888 * however it is in the secinfo list for /export thru the 1889 * server namespace setup. 1890 */ 1891 int 1892 is_exported_sec(int flavor, struct exportinfo *exi) 1893 { 1894 int i; 1895 struct secinfo *sp; 1896 1897 sp = exi->exi_export.ex_secinfo; 1898 for (i = 0; i < exi->exi_export.ex_seccnt; i++) { 1899 if (flavor == sp[i].s_secinfo.sc_nfsnum || 1900 sp[i].s_secinfo.sc_nfsnum == AUTH_NONE) { 1901 return (SEC_REF_EXPORTED(&sp[i])); 1902 } 1903 } 1904 1905 /* Should not reach this point based on the assumption */ 1906 return (0); 1907 } 1908 1909 /* 1910 * Check if the security flavor used in the request matches what is 1911 * required at the export point or at the root pseudo node (exi_root). 1912 * 1913 * returns 1 if there's a match or if exported with AUTH_NONE; 0 otherwise. 1914 * 1915 */ 1916 static int 1917 secinfo_match_or_authnone(struct compound_state *cs) 1918 { 1919 int i; 1920 struct secinfo *sp; 1921 1922 /* 1923 * Check cs->nfsflavor (from the request) against 1924 * the current export data in cs->exi. 1925 */ 1926 sp = cs->exi->exi_export.ex_secinfo; 1927 for (i = 0; i < cs->exi->exi_export.ex_seccnt; i++) { 1928 if (cs->nfsflavor == sp[i].s_secinfo.sc_nfsnum || 1929 sp[i].s_secinfo.sc_nfsnum == AUTH_NONE) 1930 return (1); 1931 } 1932 1933 return (0); 1934 } 1935 1936 /* 1937 * Check the access authority for the client and return the correct error. 1938 */ 1939 nfsstat4 1940 call_checkauth4(struct compound_state *cs, struct svc_req *req) 1941 { 1942 int authres; 1943 1944 /* 1945 * First, check if the security flavor used in the request 1946 * are among the flavors set in the server namespace. 1947 */ 1948 if (!secinfo_match_or_authnone(cs)) { 1949 *cs->statusp = NFS4ERR_WRONGSEC; 1950 return (*cs->statusp); 1951 } 1952 1953 authres = checkauth4(cs, req); 1954 1955 if (authres > 0) { 1956 *cs->statusp = NFS4_OK; 1957 if (! (cs->access & CS_ACCESS_LIMITED)) 1958 cs->access = CS_ACCESS_OK; 1959 } else if (authres == 0) { 1960 *cs->statusp = NFS4ERR_ACCESS; 1961 } else if (authres == -2) { 1962 *cs->statusp = NFS4ERR_WRONGSEC; 1963 } else { 1964 *cs->statusp = NFS4ERR_DELAY; 1965 } 1966 return (*cs->statusp); 1967 } 1968 1969 /* 1970 * bitmap4_to_attrmask is called by getattr and readdir. 1971 * It sets up the vattr mask and determines whether vfsstat call is needed 1972 * based on the input bitmap. 1973 * Returns nfsv4 status. 1974 */ 1975 static nfsstat4 1976 bitmap4_to_attrmask(bitmap4 breq, struct nfs4_svgetit_arg *sargp) 1977 { 1978 int i; 1979 uint_t va_mask; 1980 struct statvfs64 *sbp = sargp->sbp; 1981 1982 sargp->sbp = NULL; 1983 sargp->flag = 0; 1984 sargp->rdattr_error = NFS4_OK; 1985 sargp->mntdfid_set = FALSE; 1986 if (sargp->cs->vp) 1987 sargp->xattr = get_fh4_flag(&sargp->cs->fh, 1988 FH4_ATTRDIR | FH4_NAMEDATTR); 1989 else 1990 sargp->xattr = 0; 1991 1992 /* 1993 * Set rdattr_error_req to true if return error per 1994 * failed entry rather than fail the readdir. 1995 */ 1996 if (breq & FATTR4_RDATTR_ERROR_MASK) 1997 sargp->rdattr_error_req = 1; 1998 else 1999 sargp->rdattr_error_req = 0; 2000 2001 /* 2002 * generate the va_mask 2003 * Handle the easy cases first 2004 */ 2005 switch (breq) { 2006 case NFS4_NTOV_ATTR_MASK: 2007 sargp->vap->va_mask = NFS4_NTOV_ATTR_AT_MASK; 2008 return (NFS4_OK); 2009 2010 case NFS4_FS_ATTR_MASK: 2011 sargp->vap->va_mask = NFS4_FS_ATTR_AT_MASK; 2012 sargp->sbp = sbp; 2013 return (NFS4_OK); 2014 2015 case NFS4_NTOV_ATTR_CACHE_MASK: 2016 sargp->vap->va_mask = NFS4_NTOV_ATTR_CACHE_AT_MASK; 2017 return (NFS4_OK); 2018 2019 case FATTR4_LEASE_TIME_MASK: 2020 sargp->vap->va_mask = 0; 2021 return (NFS4_OK); 2022 2023 default: 2024 va_mask = 0; 2025 for (i = 0; i < nfs4_ntov_map_size; i++) { 2026 if ((breq & nfs4_ntov_map[i].fbit) && 2027 nfs4_ntov_map[i].vbit) 2028 va_mask |= nfs4_ntov_map[i].vbit; 2029 } 2030 2031 /* 2032 * Check is vfsstat is needed 2033 */ 2034 if (breq & NFS4_FS_ATTR_MASK) 2035 sargp->sbp = sbp; 2036 2037 sargp->vap->va_mask = va_mask; 2038 return (NFS4_OK); 2039 } 2040 /* NOTREACHED */ 2041 } 2042 2043 /* 2044 * bitmap4_get_sysattrs is called by getattr and readdir. 2045 * It calls both VOP_GETATTR and VFS_STATVFS calls to get the attrs. 2046 * Returns nfsv4 status. 2047 */ 2048 static nfsstat4 2049 bitmap4_get_sysattrs(struct nfs4_svgetit_arg *sargp) 2050 { 2051 int error; 2052 struct compound_state *cs = sargp->cs; 2053 vnode_t *vp = cs->vp; 2054 2055 if (sargp->sbp != NULL) { 2056 if (error = VFS_STATVFS(vp->v_vfsp, sargp->sbp)) { 2057 sargp->sbp = NULL; /* to identify error */ 2058 return (puterrno4(error)); 2059 } 2060 } 2061 2062 return (rfs4_vop_getattr(vp, sargp->vap, 0, cs->cr)); 2063 } 2064 2065 static void 2066 nfs4_ntov_table_init(struct nfs4_ntov_table *ntovp) 2067 { 2068 ntovp->na = kmem_zalloc(sizeof (union nfs4_attr_u) * nfs4_ntov_map_size, 2069 KM_SLEEP); 2070 ntovp->attrcnt = 0; 2071 ntovp->vfsstat = FALSE; 2072 } 2073 2074 static void 2075 nfs4_ntov_table_free(struct nfs4_ntov_table *ntovp, 2076 struct nfs4_svgetit_arg *sargp) 2077 { 2078 int i; 2079 union nfs4_attr_u *na; 2080 uint8_t *amap; 2081 2082 /* 2083 * XXX Should do the same checks for whether the bit is set 2084 */ 2085 for (i = 0, na = ntovp->na, amap = ntovp->amap; 2086 i < ntovp->attrcnt; i++, na++, amap++) { 2087 (void) (*nfs4_ntov_map[*amap].sv_getit)( 2088 NFS4ATTR_FREEIT, sargp, na); 2089 } 2090 if ((sargp->op == NFS4ATTR_SETIT) || (sargp->op == NFS4ATTR_VERIT)) { 2091 /* 2092 * xdr_free for getattr will be done later 2093 */ 2094 for (i = 0, na = ntovp->na, amap = ntovp->amap; 2095 i < ntovp->attrcnt; i++, na++, amap++) { 2096 xdr_free(nfs4_ntov_map[*amap].xfunc, (caddr_t)na); 2097 } 2098 } 2099 kmem_free(ntovp->na, sizeof (union nfs4_attr_u) * nfs4_ntov_map_size); 2100 } 2101 2102 /* 2103 * do_rfs4_op_getattr gets the system attrs and converts into fattr4. 2104 */ 2105 static nfsstat4 2106 do_rfs4_op_getattr(bitmap4 breq, fattr4 *fattrp, 2107 struct nfs4_svgetit_arg *sargp) 2108 { 2109 int error = 0; 2110 int i, k; 2111 struct nfs4_ntov_table ntov; 2112 XDR xdr; 2113 ulong_t xdr_size; 2114 char *xdr_attrs; 2115 nfsstat4 status = NFS4_OK; 2116 nfsstat4 prev_rdattr_error = sargp->rdattr_error; 2117 union nfs4_attr_u *na; 2118 uint8_t *amap; 2119 2120 sargp->op = NFS4ATTR_GETIT; 2121 sargp->flag = 0; 2122 2123 fattrp->attrmask = 0; 2124 /* if no bits requested, then return empty fattr4 */ 2125 if (breq == 0) { 2126 fattrp->attrlist4_len = 0; 2127 fattrp->attrlist4 = NULL; 2128 return (NFS4_OK); 2129 } 2130 2131 /* 2132 * return NFS4ERR_INVAL when client requests write-only attrs 2133 */ 2134 if (breq & (FATTR4_TIME_ACCESS_SET_MASK | FATTR4_TIME_MODIFY_SET_MASK)) 2135 return (NFS4ERR_INVAL); 2136 2137 nfs4_ntov_table_init(&ntov); 2138 na = ntov.na; 2139 amap = ntov.amap; 2140 2141 /* 2142 * Now loop to get or verify the attrs 2143 */ 2144 for (i = 0; i < nfs4_ntov_map_size; i++) { 2145 if (breq & nfs4_ntov_map[i].fbit) { 2146 if ((*nfs4_ntov_map[i].sv_getit)( 2147 NFS4ATTR_SUPPORTED, sargp, NULL) == 0) { 2148 2149 error = (*nfs4_ntov_map[i].sv_getit)( 2150 NFS4ATTR_GETIT, sargp, na); 2151 2152 /* 2153 * Possible error values: 2154 * >0 if sv_getit failed to 2155 * get the attr; 0 if succeeded; 2156 * <0 if rdattr_error and the 2157 * attribute cannot be returned. 2158 */ 2159 if (error && !(sargp->rdattr_error_req)) 2160 goto done; 2161 /* 2162 * If error then just for entry 2163 */ 2164 if (error == 0) { 2165 fattrp->attrmask |= 2166 nfs4_ntov_map[i].fbit; 2167 *amap++ = 2168 (uint8_t)nfs4_ntov_map[i].nval; 2169 na++; 2170 (ntov.attrcnt)++; 2171 } else if ((error > 0) && 2172 (sargp->rdattr_error == NFS4_OK)) { 2173 sargp->rdattr_error = puterrno4(error); 2174 } 2175 error = 0; 2176 } 2177 } 2178 } 2179 2180 /* 2181 * If rdattr_error was set after the return value for it was assigned, 2182 * update it. 2183 */ 2184 if (prev_rdattr_error != sargp->rdattr_error) { 2185 na = ntov.na; 2186 amap = ntov.amap; 2187 for (i = 0; i < ntov.attrcnt; i++, na++, amap++) { 2188 k = *amap; 2189 if (k < FATTR4_RDATTR_ERROR) { 2190 continue; 2191 } 2192 if ((k == FATTR4_RDATTR_ERROR) && 2193 ((*nfs4_ntov_map[k].sv_getit)( 2194 NFS4ATTR_SUPPORTED, sargp, NULL) == 0)) { 2195 2196 (void) (*nfs4_ntov_map[k].sv_getit)( 2197 NFS4ATTR_GETIT, sargp, na); 2198 } 2199 break; 2200 } 2201 } 2202 2203 xdr_size = 0; 2204 na = ntov.na; 2205 amap = ntov.amap; 2206 for (i = 0; i < ntov.attrcnt; i++, na++, amap++) { 2207 xdr_size += xdr_sizeof(nfs4_ntov_map[*amap].xfunc, na); 2208 } 2209 2210 fattrp->attrlist4_len = xdr_size; 2211 if (xdr_size) { 2212 /* freed by rfs4_op_getattr_free() */ 2213 fattrp->attrlist4 = xdr_attrs = kmem_zalloc(xdr_size, KM_SLEEP); 2214 2215 xdrmem_create(&xdr, xdr_attrs, xdr_size, XDR_ENCODE); 2216 2217 na = ntov.na; 2218 amap = ntov.amap; 2219 for (i = 0; i < ntov.attrcnt; i++, na++, amap++) { 2220 if (!(*nfs4_ntov_map[*amap].xfunc)(&xdr, na)) { 2221 DTRACE_PROBE1(nfss__e__getattr4_encfail, 2222 int, *amap); 2223 status = NFS4ERR_SERVERFAULT; 2224 break; 2225 } 2226 } 2227 /* xdrmem_destroy(&xdrs); */ /* NO-OP */ 2228 } else { 2229 fattrp->attrlist4 = NULL; 2230 } 2231 done: 2232 2233 nfs4_ntov_table_free(&ntov, sargp); 2234 2235 if (error != 0) 2236 status = puterrno4(error); 2237 2238 return (status); 2239 } 2240 2241 /* ARGSUSED */ 2242 static void 2243 rfs4_op_getattr(nfs_argop4 *argop, nfs_resop4 *resop, struct svc_req *req, 2244 struct compound_state *cs) 2245 { 2246 GETATTR4args *args = &argop->nfs_argop4_u.opgetattr; 2247 GETATTR4res *resp = &resop->nfs_resop4_u.opgetattr; 2248 struct nfs4_svgetit_arg sarg; 2249 struct statvfs64 sb; 2250 nfsstat4 status; 2251 2252 DTRACE_NFSV4_2(op__getattr__start, struct compound_state *, cs, 2253 GETATTR4args *, args); 2254 2255 if (cs->vp == NULL) { 2256 *cs->statusp = resp->status = NFS4ERR_NOFILEHANDLE; 2257 goto out; 2258 } 2259 2260 if (cs->access == CS_ACCESS_DENIED) { 2261 *cs->statusp = resp->status = NFS4ERR_ACCESS; 2262 goto out; 2263 } 2264 2265 sarg.sbp = &sb; 2266 sarg.cs = cs; 2267 2268 status = bitmap4_to_attrmask(args->attr_request, &sarg); 2269 if (status == NFS4_OK) { 2270 status = bitmap4_get_sysattrs(&sarg); 2271 if (status == NFS4_OK) 2272 status = do_rfs4_op_getattr(args->attr_request, 2273 &resp->obj_attributes, &sarg); 2274 } 2275 *cs->statusp = resp->status = status; 2276 out: 2277 DTRACE_NFSV4_2(op__getattr__done, struct compound_state *, cs, 2278 GETATTR4res *, resp); 2279 } 2280 2281 static void 2282 rfs4_op_getattr_free(nfs_resop4 *resop) 2283 { 2284 GETATTR4res *resp = &resop->nfs_resop4_u.opgetattr; 2285 2286 nfs4_fattr4_free(&resp->obj_attributes); 2287 } 2288 2289 /* ARGSUSED */ 2290 static void 2291 rfs4_op_getfh(nfs_argop4 *argop, nfs_resop4 *resop, struct svc_req *req, 2292 struct compound_state *cs) 2293 { 2294 GETFH4res *resp = &resop->nfs_resop4_u.opgetfh; 2295 2296 DTRACE_NFSV4_1(op__getfh__start, struct compound_state *, cs); 2297 2298 if (cs->vp == NULL) { 2299 *cs->statusp = resp->status = NFS4ERR_NOFILEHANDLE; 2300 goto out; 2301 } 2302 if (cs->access == CS_ACCESS_DENIED) { 2303 *cs->statusp = resp->status = NFS4ERR_ACCESS; 2304 goto out; 2305 } 2306 2307 resp->object.nfs_fh4_val = 2308 kmem_alloc(cs->fh.nfs_fh4_len, KM_SLEEP); 2309 nfs_fh4_copy(&cs->fh, &resp->object); 2310 *cs->statusp = resp->status = NFS4_OK; 2311 out: 2312 DTRACE_NFSV4_2(op__getfh__done, struct compound_state *, cs, 2313 GETFH4res *, resp); 2314 } 2315 2316 static void 2317 rfs4_op_getfh_free(nfs_resop4 *resop) 2318 { 2319 GETFH4res *resp = &resop->nfs_resop4_u.opgetfh; 2320 2321 if (resp->status == NFS4_OK && 2322 resp->object.nfs_fh4_val != NULL) { 2323 kmem_free(resp->object.nfs_fh4_val, resp->object.nfs_fh4_len); 2324 resp->object.nfs_fh4_val = NULL; 2325 resp->object.nfs_fh4_len = 0; 2326 } 2327 } 2328 2329 /* 2330 * illegal: args: void 2331 * res : status (NFS4ERR_OP_ILLEGAL) 2332 */ 2333 /* ARGSUSED */ 2334 static void 2335 rfs4_op_illegal(nfs_argop4 *argop, nfs_resop4 *resop, 2336 struct svc_req *req, struct compound_state *cs) 2337 { 2338 ILLEGAL4res *resp = &resop->nfs_resop4_u.opillegal; 2339 2340 resop->resop = OP_ILLEGAL; 2341 *cs->statusp = resp->status = NFS4ERR_OP_ILLEGAL; 2342 } 2343 2344 /* 2345 * link: args: SAVED_FH: file, CURRENT_FH: target directory 2346 * res: status. If success - CURRENT_FH unchanged, return change_info 2347 */ 2348 /* ARGSUSED */ 2349 static void 2350 rfs4_op_link(nfs_argop4 *argop, nfs_resop4 *resop, struct svc_req *req, 2351 struct compound_state *cs) 2352 { 2353 LINK4args *args = &argop->nfs_argop4_u.oplink; 2354 LINK4res *resp = &resop->nfs_resop4_u.oplink; 2355 int error; 2356 vnode_t *vp; 2357 vnode_t *dvp; 2358 struct vattr bdva, idva, adva; 2359 char *nm; 2360 uint_t len; 2361 2362 DTRACE_NFSV4_2(op__link__start, struct compound_state *, cs, 2363 LINK4args *, args); 2364 2365 /* SAVED_FH: source object */ 2366 vp = cs->saved_vp; 2367 if (vp == NULL) { 2368 *cs->statusp = resp->status = NFS4ERR_NOFILEHANDLE; 2369 goto out; 2370 } 2371 2372 /* CURRENT_FH: target directory */ 2373 dvp = cs->vp; 2374 if (dvp == NULL) { 2375 *cs->statusp = resp->status = NFS4ERR_NOFILEHANDLE; 2376 goto out; 2377 } 2378 2379 /* 2380 * If there is a non-shared filesystem mounted on this vnode, 2381 * do not allow to link any file in this directory. 2382 */ 2383 if (vn_ismntpt(dvp)) { 2384 *cs->statusp = resp->status = NFS4ERR_ACCESS; 2385 goto out; 2386 } 2387 2388 if (cs->access == CS_ACCESS_DENIED) { 2389 *cs->statusp = resp->status = NFS4ERR_ACCESS; 2390 goto out; 2391 } 2392 2393 /* Check source object's type validity */ 2394 if (vp->v_type == VDIR) { 2395 *cs->statusp = resp->status = NFS4ERR_ISDIR; 2396 goto out; 2397 } 2398 2399 /* Check target directory's type */ 2400 if (dvp->v_type != VDIR) { 2401 *cs->statusp = resp->status = NFS4ERR_NOTDIR; 2402 goto out; 2403 } 2404 2405 if (cs->saved_exi != cs->exi) { 2406 *cs->statusp = resp->status = NFS4ERR_XDEV; 2407 goto out; 2408 } 2409 2410 if (!utf8_dir_verify(&args->newname)) { 2411 *cs->statusp = resp->status = NFS4ERR_INVAL; 2412 goto out; 2413 } 2414 2415 nm = utf8_to_fn(&args->newname, &len, NULL); 2416 if (nm == NULL) { 2417 *cs->statusp = resp->status = NFS4ERR_INVAL; 2418 goto out; 2419 } 2420 2421 if (len > MAXNAMELEN) { 2422 *cs->statusp = resp->status = NFS4ERR_NAMETOOLONG; 2423 kmem_free(nm, len); 2424 goto out; 2425 } 2426 2427 if (rdonly4(cs->exi, cs->vp, req)) { 2428 *cs->statusp = resp->status = NFS4ERR_ROFS; 2429 kmem_free(nm, len); 2430 goto out; 2431 } 2432 2433 /* Get "before" change value */ 2434 bdva.va_mask = AT_CTIME|AT_SEQ; 2435 error = VOP_GETATTR(dvp, &bdva, 0, cs->cr, NULL); 2436 if (error) { 2437 *cs->statusp = resp->status = puterrno4(error); 2438 kmem_free(nm, len); 2439 goto out; 2440 } 2441 2442 NFS4_SET_FATTR4_CHANGE(resp->cinfo.before, bdva.va_ctime) 2443 2444 error = VOP_LINK(dvp, vp, nm, cs->cr, NULL, 0); 2445 2446 kmem_free(nm, len); 2447 2448 /* 2449 * Get the initial "after" sequence number, if it fails, set to zero 2450 */ 2451 idva.va_mask = AT_SEQ; 2452 if (VOP_GETATTR(dvp, &idva, 0, cs->cr, NULL)) 2453 idva.va_seq = 0; 2454 2455 /* 2456 * Force modified data and metadata out to stable storage. 2457 */ 2458 (void) VOP_FSYNC(vp, FNODSYNC, cs->cr, NULL); 2459 (void) VOP_FSYNC(dvp, 0, cs->cr, NULL); 2460 2461 if (error) { 2462 *cs->statusp = resp->status = puterrno4(error); 2463 goto out; 2464 } 2465 2466 /* 2467 * Get "after" change value, if it fails, simply return the 2468 * before value. 2469 */ 2470 adva.va_mask = AT_CTIME|AT_SEQ; 2471 if (VOP_GETATTR(dvp, &adva, 0, cs->cr, NULL)) { 2472 adva.va_ctime = bdva.va_ctime; 2473 adva.va_seq = 0; 2474 } 2475 2476 NFS4_SET_FATTR4_CHANGE(resp->cinfo.after, adva.va_ctime) 2477 2478 /* 2479 * The cinfo.atomic = TRUE only if we have 2480 * non-zero va_seq's, and it has incremented by exactly one 2481 * during the VOP_LINK and it didn't change during the VOP_FSYNC. 2482 */ 2483 if (bdva.va_seq && idva.va_seq && adva.va_seq && 2484 idva.va_seq == (bdva.va_seq + 1) && idva.va_seq == adva.va_seq) 2485 resp->cinfo.atomic = TRUE; 2486 else 2487 resp->cinfo.atomic = FALSE; 2488 2489 *cs->statusp = resp->status = NFS4_OK; 2490 out: 2491 DTRACE_NFSV4_2(op__link__done, struct compound_state *, cs, 2492 LINK4res *, resp); 2493 } 2494 2495 /* 2496 * Used by rfs4_op_lookup and rfs4_op_lookupp to do the actual work. 2497 */ 2498 2499 /* ARGSUSED */ 2500 static nfsstat4 2501 do_rfs4_op_lookup(char *nm, uint_t buflen, struct svc_req *req, 2502 struct compound_state *cs) 2503 { 2504 int error; 2505 int different_export = 0; 2506 vnode_t *vp, *tvp, *pre_tvp = NULL, *oldvp = NULL; 2507 struct exportinfo *exi = NULL, *pre_exi = NULL; 2508 nfsstat4 stat; 2509 fid_t fid; 2510 int attrdir, dotdot, walk; 2511 bool_t is_newvp = FALSE; 2512 2513 if (cs->vp->v_flag & V_XATTRDIR) { 2514 attrdir = 1; 2515 ASSERT(get_fh4_flag(&cs->fh, FH4_ATTRDIR)); 2516 } else { 2517 attrdir = 0; 2518 ASSERT(! get_fh4_flag(&cs->fh, FH4_ATTRDIR)); 2519 } 2520 2521 dotdot = (nm[0] == '.' && nm[1] == '.' && nm[2] == '\0'); 2522 2523 /* 2524 * If dotdotting, then need to check whether it's 2525 * above the root of a filesystem, or above an 2526 * export point. 2527 */ 2528 if (dotdot) { 2529 2530 /* 2531 * If dotdotting at the root of a filesystem, then 2532 * need to traverse back to the mounted-on filesystem 2533 * and do the dotdot lookup there. 2534 */ 2535 if (cs->vp->v_flag & VROOT) { 2536 2537 /* 2538 * If at the system root, then can 2539 * go up no further. 2540 */ 2541 if (VN_CMP(cs->vp, rootdir)) 2542 return (puterrno4(ENOENT)); 2543 2544 /* 2545 * Traverse back to the mounted-on filesystem 2546 */ 2547 cs->vp = untraverse(cs->vp); 2548 2549 /* 2550 * Set the different_export flag so we remember 2551 * to pick up a new exportinfo entry for 2552 * this new filesystem. 2553 */ 2554 different_export = 1; 2555 } else { 2556 2557 /* 2558 * If dotdotting above an export point then set 2559 * the different_export to get new export info. 2560 */ 2561 different_export = nfs_exported(cs->exi, cs->vp); 2562 } 2563 } 2564 2565 error = VOP_LOOKUP(cs->vp, nm, &vp, NULL, 0, NULL, cs->cr, 2566 NULL, NULL, NULL); 2567 if (error) 2568 return (puterrno4(error)); 2569 2570 /* 2571 * If the vnode is in a pseudo filesystem, check whether it is visible. 2572 * 2573 * XXX if the vnode is a symlink and it is not visible in 2574 * a pseudo filesystem, return ENOENT (not following symlink). 2575 * V4 client can not mount such symlink. This is a regression 2576 * from V2/V3. 2577 * 2578 * In the same exported filesystem, if the security flavor used 2579 * is not an explicitly shared flavor, limit the view to the visible 2580 * list entries only. This is not a WRONGSEC case because it's already 2581 * checked via PUTROOTFH/PUTPUBFH or PUTFH. 2582 */ 2583 if (!different_export && 2584 (PSEUDO(cs->exi) || ! is_exported_sec(cs->nfsflavor, cs->exi) || 2585 cs->access & CS_ACCESS_LIMITED)) { 2586 if (! nfs_visible(cs->exi, vp, &different_export)) { 2587 VN_RELE(vp); 2588 return (puterrno4(ENOENT)); 2589 } 2590 } 2591 2592 /* 2593 * If it's a mountpoint, then traverse it. 2594 */ 2595 if (vn_ismntpt(vp)) { 2596 pre_exi = cs->exi; /* save pre-traversed exportinfo */ 2597 pre_tvp = vp; /* save pre-traversed vnode */ 2598 2599 /* 2600 * hold pre_tvp to counteract rele by traverse. We will 2601 * need pre_tvp below if checkexport4 fails 2602 */ 2603 VN_HOLD(pre_tvp); 2604 tvp = vp; 2605 if ((error = traverse(&tvp)) != 0) { 2606 VN_RELE(vp); 2607 VN_RELE(pre_tvp); 2608 return (puterrno4(error)); 2609 } 2610 vp = tvp; 2611 different_export = 1; 2612 } else if (vp->v_vfsp != cs->vp->v_vfsp) { 2613 /* 2614 * The vfsp comparison is to handle the case where 2615 * a LOFS mount is shared. lo_lookup traverses mount points, 2616 * and NFS is unaware of local fs transistions because 2617 * v_vfsmountedhere isn't set. For this special LOFS case, 2618 * the dir and the obj returned by lookup will have different 2619 * vfs ptrs. 2620 */ 2621 different_export = 1; 2622 } 2623 2624 if (different_export) { 2625 2626 bzero(&fid, sizeof (fid)); 2627 fid.fid_len = MAXFIDSZ; 2628 error = vop_fid_pseudo(vp, &fid); 2629 if (error) { 2630 VN_RELE(vp); 2631 if (pre_tvp) 2632 VN_RELE(pre_tvp); 2633 return (puterrno4(error)); 2634 } 2635 2636 if (dotdot) 2637 exi = nfs_vptoexi(NULL, vp, cs->cr, &walk, NULL, TRUE); 2638 else 2639 exi = checkexport4(&vp->v_vfsp->vfs_fsid, &fid, vp); 2640 2641 if (exi == NULL) { 2642 if (pre_tvp) { 2643 /* 2644 * If this vnode is a mounted-on vnode, 2645 * but the mounted-on file system is not 2646 * exported, send back the filehandle for 2647 * the mounted-on vnode, not the root of 2648 * the mounted-on file system. 2649 */ 2650 VN_RELE(vp); 2651 vp = pre_tvp; 2652 exi = pre_exi; 2653 } else { 2654 VN_RELE(vp); 2655 return (puterrno4(EACCES)); 2656 } 2657 } else if (pre_tvp) { 2658 /* we're done with pre_tvp now. release extra hold */ 2659 VN_RELE(pre_tvp); 2660 } 2661 2662 cs->exi = exi; 2663 2664 /* 2665 * Now we do a checkauth4. The reason is that 2666 * this client/user may not have access to the new 2667 * exported file system, and if he does, 2668 * the client/user may be mapped to a different uid. 2669 * 2670 * We start with a new cr, because the checkauth4 done 2671 * in the PUT*FH operation over wrote the cred's uid, 2672 * gid, etc, and we want the real thing before calling 2673 * checkauth4() 2674 */ 2675 crfree(cs->cr); 2676 cs->cr = crdup(cs->basecr); 2677 2678 if (cs->vp) 2679 oldvp = cs->vp; 2680 cs->vp = vp; 2681 is_newvp = TRUE; 2682 2683 stat = call_checkauth4(cs, req); 2684 if (stat != NFS4_OK) { 2685 VN_RELE(cs->vp); 2686 cs->vp = oldvp; 2687 return (stat); 2688 } 2689 } 2690 2691 /* 2692 * After various NFS checks, do a label check on the path 2693 * component. The label on this path should either be the 2694 * global zone's label or a zone's label. We are only 2695 * interested in the zone's label because exported files 2696 * in global zone is accessible (though read-only) to 2697 * clients. The exportability/visibility check is already 2698 * done before reaching this code. 2699 */ 2700 if (is_system_labeled()) { 2701 bslabel_t *clabel; 2702 2703 ASSERT(req->rq_label != NULL); 2704 clabel = req->rq_label; 2705 DTRACE_PROBE2(tx__rfs4__log__info__oplookup__clabel, char *, 2706 "got client label from request(1)", struct svc_req *, req); 2707 2708 if (!blequal(&l_admin_low->tsl_label, clabel)) { 2709 if (!do_rfs_label_check(clabel, vp, DOMINANCE_CHECK)) { 2710 error = EACCES; 2711 goto err_out; 2712 } 2713 } else { 2714 /* 2715 * We grant access to admin_low label clients 2716 * only if the client is trusted, i.e. also 2717 * running Solaris Trusted Extension. 2718 */ 2719 struct sockaddr *ca; 2720 int addr_type; 2721 void *ipaddr; 2722 tsol_tpc_t *tp; 2723 2724 ca = (struct sockaddr *)svc_getrpccaller( 2725 req->rq_xprt)->buf; 2726 if (ca->sa_family == AF_INET) { 2727 addr_type = IPV4_VERSION; 2728 ipaddr = &((struct sockaddr_in *)ca)->sin_addr; 2729 } else if (ca->sa_family == AF_INET6) { 2730 addr_type = IPV6_VERSION; 2731 ipaddr = &((struct sockaddr_in6 *) 2732 ca)->sin6_addr; 2733 } 2734 tp = find_tpc(ipaddr, addr_type, B_FALSE); 2735 if (tp == NULL || tp->tpc_tp.tp_doi != 2736 l_admin_low->tsl_doi || tp->tpc_tp.host_type != 2737 SUN_CIPSO) { 2738 if (tp != NULL) 2739 TPC_RELE(tp); 2740 error = EACCES; 2741 goto err_out; 2742 } 2743 TPC_RELE(tp); 2744 } 2745 } 2746 2747 error = makefh4(&cs->fh, vp, cs->exi); 2748 2749 err_out: 2750 if (error) { 2751 if (is_newvp) { 2752 VN_RELE(cs->vp); 2753 cs->vp = oldvp; 2754 } else 2755 VN_RELE(vp); 2756 return (puterrno4(error)); 2757 } 2758 2759 if (!is_newvp) { 2760 if (cs->vp) 2761 VN_RELE(cs->vp); 2762 cs->vp = vp; 2763 } else if (oldvp) 2764 VN_RELE(oldvp); 2765 2766 /* 2767 * if did lookup on attrdir and didn't lookup .., set named 2768 * attr fh flag 2769 */ 2770 if (attrdir && ! dotdot) 2771 set_fh4_flag(&cs->fh, FH4_NAMEDATTR); 2772 2773 /* Assume false for now, open proc will set this */ 2774 cs->mandlock = FALSE; 2775 2776 return (NFS4_OK); 2777 } 2778 2779 /* ARGSUSED */ 2780 static void 2781 rfs4_op_lookup(nfs_argop4 *argop, nfs_resop4 *resop, struct svc_req *req, 2782 struct compound_state *cs) 2783 { 2784 LOOKUP4args *args = &argop->nfs_argop4_u.oplookup; 2785 LOOKUP4res *resp = &resop->nfs_resop4_u.oplookup; 2786 char *nm; 2787 uint_t len; 2788 2789 DTRACE_NFSV4_2(op__lookup__start, struct compound_state *, cs, 2790 LOOKUP4args *, args); 2791 2792 if (cs->vp == NULL) { 2793 *cs->statusp = resp->status = NFS4ERR_NOFILEHANDLE; 2794 goto out; 2795 } 2796 2797 if (cs->vp->v_type == VLNK) { 2798 *cs->statusp = resp->status = NFS4ERR_SYMLINK; 2799 goto out; 2800 } 2801 2802 if (cs->vp->v_type != VDIR) { 2803 *cs->statusp = resp->status = NFS4ERR_NOTDIR; 2804 goto out; 2805 } 2806 2807 if (!utf8_dir_verify(&args->objname)) { 2808 *cs->statusp = resp->status = NFS4ERR_INVAL; 2809 goto out; 2810 } 2811 2812 nm = utf8_to_str(&args->objname, &len, NULL); 2813 if (nm == NULL) { 2814 *cs->statusp = resp->status = NFS4ERR_INVAL; 2815 goto out; 2816 } 2817 2818 if (len > MAXNAMELEN) { 2819 *cs->statusp = resp->status = NFS4ERR_NAMETOOLONG; 2820 kmem_free(nm, len); 2821 goto out; 2822 } 2823 2824 *cs->statusp = resp->status = do_rfs4_op_lookup(nm, len, req, cs); 2825 2826 kmem_free(nm, len); 2827 2828 out: 2829 DTRACE_NFSV4_2(op__lookup__done, struct compound_state *, cs, 2830 LOOKUP4res *, resp); 2831 } 2832 2833 /* ARGSUSED */ 2834 static void 2835 rfs4_op_lookupp(nfs_argop4 *args, nfs_resop4 *resop, struct svc_req *req, 2836 struct compound_state *cs) 2837 { 2838 LOOKUPP4res *resp = &resop->nfs_resop4_u.oplookupp; 2839 2840 DTRACE_NFSV4_1(op__lookupp__start, struct compound_state *, cs); 2841 2842 if (cs->vp == NULL) { 2843 *cs->statusp = resp->status = NFS4ERR_NOFILEHANDLE; 2844 goto out; 2845 } 2846 2847 if (cs->vp->v_type != VDIR) { 2848 *cs->statusp = resp->status = NFS4ERR_NOTDIR; 2849 goto out; 2850 } 2851 2852 *cs->statusp = resp->status = do_rfs4_op_lookup("..", 3, req, cs); 2853 2854 /* 2855 * From NFSV4 Specification, LOOKUPP should not check for 2856 * NFS4ERR_WRONGSEC. Retrun NFS4_OK instead. 2857 */ 2858 if (resp->status == NFS4ERR_WRONGSEC) { 2859 *cs->statusp = resp->status = NFS4_OK; 2860 } 2861 2862 out: 2863 DTRACE_NFSV4_2(op__lookupp__done, struct compound_state *, cs, 2864 LOOKUPP4res *, resp); 2865 } 2866 2867 2868 /*ARGSUSED2*/ 2869 static void 2870 rfs4_op_openattr(nfs_argop4 *argop, nfs_resop4 *resop, struct svc_req *req, 2871 struct compound_state *cs) 2872 { 2873 OPENATTR4args *args = &argop->nfs_argop4_u.opopenattr; 2874 OPENATTR4res *resp = &resop->nfs_resop4_u.opopenattr; 2875 vnode_t *avp = NULL; 2876 int lookup_flags = LOOKUP_XATTR, error; 2877 int exp_ro = 0; 2878 2879 DTRACE_NFSV4_2(op__openattr__start, struct compound_state *, cs, 2880 OPENATTR4args *, args); 2881 2882 if (cs->vp == NULL) { 2883 *cs->statusp = resp->status = NFS4ERR_NOFILEHANDLE; 2884 goto out; 2885 } 2886 2887 if ((cs->vp->v_vfsp->vfs_flag & VFS_XATTR) == 0 && 2888 !vfs_has_feature(cs->vp->v_vfsp, VFSFT_SYSATTR_VIEWS)) { 2889 *cs->statusp = resp->status = puterrno4(ENOTSUP); 2890 goto out; 2891 } 2892 2893 /* 2894 * If file system supports passing ACE mask to VOP_ACCESS then 2895 * check for ACE_READ_NAMED_ATTRS, otherwise do legacy checks 2896 */ 2897 2898 if (vfs_has_feature(cs->vp->v_vfsp, VFSFT_ACEMASKONACCESS)) 2899 error = VOP_ACCESS(cs->vp, ACE_READ_NAMED_ATTRS, 2900 V_ACE_MASK, cs->cr, NULL); 2901 else 2902 error = ((VOP_ACCESS(cs->vp, VREAD, 0, cs->cr, NULL) != 0) && 2903 (VOP_ACCESS(cs->vp, VWRITE, 0, cs->cr, NULL) != 0) && 2904 (VOP_ACCESS(cs->vp, VEXEC, 0, cs->cr, NULL) != 0)); 2905 2906 if (error) { 2907 *cs->statusp = resp->status = puterrno4(EACCES); 2908 goto out; 2909 } 2910 2911 /* 2912 * The CREATE_XATTR_DIR VOP flag cannot be specified if 2913 * the file system is exported read-only -- regardless of 2914 * createdir flag. Otherwise the attrdir would be created 2915 * (assuming server fs isn't mounted readonly locally). If 2916 * VOP_LOOKUP returns ENOENT in this case, the error will 2917 * be translated into EROFS. ENOSYS is mapped to ENOTSUP 2918 * because specfs has no VOP_LOOKUP op, so the macro would 2919 * return ENOSYS. EINVAL is returned by all (current) 2920 * Solaris file system implementations when any of their 2921 * restrictions are violated (xattr(dir) can't have xattrdir). 2922 * Returning NOTSUPP is more appropriate in this case 2923 * because the object will never be able to have an attrdir. 2924 */ 2925 if (args->createdir && ! (exp_ro = rdonly4(cs->exi, cs->vp, req))) 2926 lookup_flags |= CREATE_XATTR_DIR; 2927 2928 error = VOP_LOOKUP(cs->vp, "", &avp, NULL, lookup_flags, NULL, cs->cr, 2929 NULL, NULL, NULL); 2930 2931 if (error) { 2932 if (error == ENOENT && args->createdir && exp_ro) 2933 *cs->statusp = resp->status = puterrno4(EROFS); 2934 else if (error == EINVAL || error == ENOSYS) 2935 *cs->statusp = resp->status = puterrno4(ENOTSUP); 2936 else 2937 *cs->statusp = resp->status = puterrno4(error); 2938 goto out; 2939 } 2940 2941 ASSERT(avp->v_flag & V_XATTRDIR); 2942 2943 error = makefh4(&cs->fh, avp, cs->exi); 2944 2945 if (error) { 2946 VN_RELE(avp); 2947 *cs->statusp = resp->status = puterrno4(error); 2948 goto out; 2949 } 2950 2951 VN_RELE(cs->vp); 2952 cs->vp = avp; 2953 2954 /* 2955 * There is no requirement for an attrdir fh flag 2956 * because the attrdir has a vnode flag to distinguish 2957 * it from regular (non-xattr) directories. The 2958 * FH4_ATTRDIR flag is set for future sanity checks. 2959 */ 2960 set_fh4_flag(&cs->fh, FH4_ATTRDIR); 2961 *cs->statusp = resp->status = NFS4_OK; 2962 2963 out: 2964 DTRACE_NFSV4_2(op__openattr__done, struct compound_state *, cs, 2965 OPENATTR4res *, resp); 2966 } 2967 2968 static int 2969 do_io(int direction, vnode_t *vp, struct uio *uio, int ioflag, cred_t *cred, 2970 caller_context_t *ct) 2971 { 2972 int error; 2973 int i; 2974 clock_t delaytime; 2975 2976 delaytime = MSEC_TO_TICK_ROUNDUP(rfs4_lock_delay); 2977 2978 /* 2979 * Don't block on mandatory locks. If this routine returns 2980 * EAGAIN, the caller should return NFS4ERR_LOCKED. 2981 */ 2982 uio->uio_fmode = FNONBLOCK; 2983 2984 for (i = 0; i < rfs4_maxlock_tries; i++) { 2985 2986 2987 if (direction == FREAD) { 2988 (void) VOP_RWLOCK(vp, V_WRITELOCK_FALSE, ct); 2989 error = VOP_READ(vp, uio, ioflag, cred, ct); 2990 VOP_RWUNLOCK(vp, V_WRITELOCK_FALSE, ct); 2991 } else { 2992 (void) VOP_RWLOCK(vp, V_WRITELOCK_TRUE, ct); 2993 error = VOP_WRITE(vp, uio, ioflag, cred, ct); 2994 VOP_RWUNLOCK(vp, V_WRITELOCK_TRUE, ct); 2995 } 2996 2997 if (error != EAGAIN) 2998 break; 2999 3000 if (i < rfs4_maxlock_tries - 1) { 3001 delay(delaytime); 3002 delaytime *= 2; 3003 } 3004 } 3005 3006 return (error); 3007 } 3008 3009 /* ARGSUSED */ 3010 static void 3011 rfs4_op_read(nfs_argop4 *argop, nfs_resop4 *resop, struct svc_req *req, 3012 struct compound_state *cs) 3013 { 3014 READ4args *args = &argop->nfs_argop4_u.opread; 3015 READ4res *resp = &resop->nfs_resop4_u.opread; 3016 int error; 3017 int verror; 3018 vnode_t *vp; 3019 struct vattr va; 3020 struct iovec iov; 3021 struct uio uio; 3022 u_offset_t offset; 3023 bool_t *deleg = &cs->deleg; 3024 nfsstat4 stat; 3025 int in_crit = 0; 3026 mblk_t *mp; 3027 int alloc_err = 0; 3028 caller_context_t ct; 3029 3030 DTRACE_NFSV4_2(op__read__start, struct compound_state *, cs, 3031 READ4args, args); 3032 3033 vp = cs->vp; 3034 if (vp == NULL) { 3035 *cs->statusp = resp->status = NFS4ERR_NOFILEHANDLE; 3036 goto out; 3037 } 3038 if (cs->access == CS_ACCESS_DENIED) { 3039 *cs->statusp = resp->status = NFS4ERR_ACCESS; 3040 goto out; 3041 } 3042 3043 if ((stat = rfs4_check_stateid(FREAD, vp, &args->stateid, FALSE, 3044 deleg, TRUE, &ct)) != NFS4_OK) { 3045 *cs->statusp = resp->status = stat; 3046 goto out; 3047 } 3048 3049 /* 3050 * Enter the critical region before calling VOP_RWLOCK 3051 * to avoid a deadlock with write requests. 3052 */ 3053 if (nbl_need_check(vp)) { 3054 nbl_start_crit(vp, RW_READER); 3055 in_crit = 1; 3056 if (nbl_conflict(vp, NBL_READ, args->offset, args->count, 0, 3057 &ct)) { 3058 *cs->statusp = resp->status = NFS4ERR_LOCKED; 3059 goto out; 3060 } 3061 } 3062 3063 if ((stat = rfs4_check_stateid(FREAD, vp, &args->stateid, FALSE, 3064 deleg, TRUE, &ct)) != NFS4_OK) { 3065 *cs->statusp = resp->status = stat; 3066 goto out; 3067 } 3068 3069 va.va_mask = AT_MODE|AT_SIZE|AT_UID; 3070 verror = VOP_GETATTR(vp, &va, 0, cs->cr, &ct); 3071 3072 /* 3073 * If we can't get the attributes, then we can't do the 3074 * right access checking. So, we'll fail the request. 3075 */ 3076 if (verror) { 3077 *cs->statusp = resp->status = puterrno4(verror); 3078 goto out; 3079 } 3080 3081 if (vp->v_type != VREG) { 3082 *cs->statusp = resp->status = 3083 ((vp->v_type == VDIR) ? NFS4ERR_ISDIR : NFS4ERR_INVAL); 3084 goto out; 3085 } 3086 3087 if (crgetuid(cs->cr) != va.va_uid && 3088 (error = VOP_ACCESS(vp, VREAD, 0, cs->cr, &ct)) && 3089 (error = VOP_ACCESS(vp, VEXEC, 0, cs->cr, &ct))) { 3090 *cs->statusp = resp->status = puterrno4(error); 3091 goto out; 3092 } 3093 3094 if (MANDLOCK(vp, va.va_mode)) { /* XXX - V4 supports mand locking */ 3095 *cs->statusp = resp->status = NFS4ERR_ACCESS; 3096 goto out; 3097 } 3098 3099 offset = args->offset; 3100 if (offset >= va.va_size) { 3101 *cs->statusp = resp->status = NFS4_OK; 3102 resp->eof = TRUE; 3103 resp->data_len = 0; 3104 resp->data_val = NULL; 3105 resp->mblk = NULL; 3106 /* RDMA */ 3107 resp->wlist = args->wlist; 3108 resp->wlist_len = resp->data_len; 3109 *cs->statusp = resp->status = NFS4_OK; 3110 goto out; 3111 } 3112 3113 if (args->count == 0) { 3114 *cs->statusp = resp->status = NFS4_OK; 3115 resp->eof = FALSE; 3116 resp->data_len = 0; 3117 resp->data_val = NULL; 3118 resp->mblk = NULL; 3119 /* RDMA */ 3120 resp->wlist = args->wlist; 3121 resp->wlist_len = resp->data_len; 3122 goto out; 3123 } 3124 3125 /* 3126 * Do not allocate memory more than maximum allowed 3127 * transfer size 3128 */ 3129 if (args->count > rfs4_tsize(req)) 3130 args->count = rfs4_tsize(req); 3131 3132 /* 3133 * If returning data via RDMA Write, then grab the chunk list. If we 3134 * aren't returning READ data w/RDMA_WRITE, then grab a mblk. 3135 */ 3136 if (args->wlist) { 3137 mp = NULL; 3138 (void) rdma_get_wchunk(req, &iov, args->wlist); 3139 } else { 3140 /* 3141 * mp will contain the data to be sent out in the read reply. 3142 * It will be freed after the reply has been sent. Let's 3143 * roundup the data to a BYTES_PER_XDR_UNIT multiple, so that 3144 * the call to xdrmblk_putmblk() never fails. If the first 3145 * alloc of the requested size fails, then decrease the size to 3146 * something more reasonable and wait for the allocation to 3147 * occur. 3148 */ 3149 mp = allocb(RNDUP(args->count), BPRI_MED); 3150 if (mp == NULL) { 3151 if (args->count > MAXBSIZE) 3152 args->count = MAXBSIZE; 3153 mp = allocb_wait(RNDUP(args->count), BPRI_MED, 3154 STR_NOSIG, &alloc_err); 3155 } 3156 ASSERT(mp != NULL); 3157 ASSERT(alloc_err == 0); 3158 3159 iov.iov_base = (caddr_t)mp->b_datap->db_base; 3160 iov.iov_len = args->count; 3161 } 3162 3163 uio.uio_iov = &iov; 3164 uio.uio_iovcnt = 1; 3165 uio.uio_segflg = UIO_SYSSPACE; 3166 uio.uio_extflg = UIO_COPY_CACHED; 3167 uio.uio_loffset = args->offset; 3168 uio.uio_resid = args->count; 3169 3170 error = do_io(FREAD, vp, &uio, 0, cs->cr, &ct); 3171 3172 va.va_mask = AT_SIZE; 3173 verror = VOP_GETATTR(vp, &va, 0, cs->cr, &ct); 3174 3175 if (error) { 3176 freeb(mp); 3177 *cs->statusp = resp->status = puterrno4(error); 3178 goto out; 3179 } 3180 3181 *cs->statusp = resp->status = NFS4_OK; 3182 3183 ASSERT(uio.uio_resid >= 0); 3184 resp->data_len = args->count - uio.uio_resid; 3185 if (mp) { 3186 resp->data_val = (char *)mp->b_datap->db_base; 3187 } else { 3188 resp->data_val = (caddr_t)iov.iov_base; 3189 } 3190 resp->mblk = mp; 3191 3192 if (!verror && offset + resp->data_len == va.va_size) 3193 resp->eof = TRUE; 3194 else 3195 resp->eof = FALSE; 3196 3197 if (args->wlist) { 3198 if (!rdma_setup_read_data4(args, resp)) { 3199 *cs->statusp = resp->status = NFS4ERR_INVAL; 3200 } 3201 } else { 3202 resp->wlist = NULL; 3203 } 3204 3205 out: 3206 if (in_crit) 3207 nbl_end_crit(vp); 3208 3209 DTRACE_NFSV4_2(op__read__done, struct compound_state *, cs, 3210 READ4res *, resp); 3211 } 3212 3213 static void 3214 rfs4_op_read_free(nfs_resop4 *resop) 3215 { 3216 READ4res *resp = &resop->nfs_resop4_u.opread; 3217 3218 if (resp->status == NFS4_OK && resp->mblk != NULL) { 3219 freeb(resp->mblk); 3220 resp->mblk = NULL; 3221 resp->data_val = NULL; 3222 resp->data_len = 0; 3223 } 3224 } 3225 3226 static void 3227 rfs4_op_readdir_free(nfs_resop4 * resop) 3228 { 3229 READDIR4res *resp = &resop->nfs_resop4_u.opreaddir; 3230 3231 if (resp->status == NFS4_OK && resp->mblk != NULL) { 3232 freeb(resp->mblk); 3233 resp->mblk = NULL; 3234 resp->data_len = 0; 3235 } 3236 } 3237 3238 3239 /* ARGSUSED */ 3240 static void 3241 rfs4_op_putpubfh(nfs_argop4 *args, nfs_resop4 *resop, struct svc_req *req, 3242 struct compound_state *cs) 3243 { 3244 PUTPUBFH4res *resp = &resop->nfs_resop4_u.opputpubfh; 3245 int error; 3246 vnode_t *vp; 3247 struct exportinfo *exi, *sav_exi; 3248 nfs_fh4_fmt_t *fh_fmtp; 3249 3250 DTRACE_NFSV4_1(op__putpubfh__start, struct compound_state *, cs); 3251 3252 if (cs->vp) { 3253 VN_RELE(cs->vp); 3254 cs->vp = NULL; 3255 } 3256 3257 if (cs->cr) 3258 crfree(cs->cr); 3259 3260 cs->cr = crdup(cs->basecr); 3261 3262 vp = exi_public->exi_vp; 3263 if (vp == NULL) { 3264 *cs->statusp = resp->status = NFS4ERR_SERVERFAULT; 3265 goto out; 3266 } 3267 3268 if (is_system_labeled()) { 3269 bslabel_t *clabel; 3270 3271 ASSERT(req->rq_label != NULL); 3272 clabel = req->rq_label; 3273 DTRACE_PROBE2(tx__rfs4__log__info__opputpubfh__clabel, char *, 3274 "got client label from request(1)", 3275 struct svc_req *, req); 3276 if (!blequal(&l_admin_low->tsl_label, clabel)) { 3277 if (!do_rfs_label_check(clabel, vp, DOMINANCE_CHECK)) { 3278 *cs->statusp = resp->status = 3279 NFS4ERR_SERVERFAULT; 3280 return; 3281 } 3282 } 3283 } 3284 3285 error = makefh4(&cs->fh, vp, exi_public); 3286 if (error != 0) { 3287 *cs->statusp = resp->status = puterrno4(error); 3288 goto out; 3289 } 3290 sav_exi = cs->exi; 3291 if (exi_public == exi_root) { 3292 /* 3293 * No filesystem is actually shared public, so we default 3294 * to exi_root. In this case, we must check whether root 3295 * is exported. 3296 */ 3297 fh_fmtp = (nfs_fh4_fmt_t *)cs->fh.nfs_fh4_val; 3298 3299 /* 3300 * if root filesystem is exported, the exportinfo struct that we 3301 * should use is what checkexport4 returns, because root_exi is 3302 * actually a mostly empty struct. 3303 */ 3304 exi = checkexport4(&fh_fmtp->fh4_fsid, 3305 (fid_t *)&fh_fmtp->fh4_xlen, NULL); 3306 cs->exi = ((exi != NULL) ? exi : exi_public); 3307 } else { 3308 /* 3309 * it's a properly shared filesystem 3310 */ 3311 cs->exi = exi_public; 3312 } 3313 3314 VN_HOLD(vp); 3315 cs->vp = vp; 3316 3317 if ((resp->status = call_checkauth4(cs, req)) != NFS4_OK) { 3318 VN_RELE(cs->vp); 3319 cs->vp = NULL; 3320 cs->exi = sav_exi; 3321 goto out; 3322 } 3323 3324 *cs->statusp = resp->status = NFS4_OK; 3325 out: 3326 DTRACE_NFSV4_2(op__putpubfh__done, struct compound_state *, cs, 3327 PUTPUBFH4res *, resp); 3328 } 3329 3330 /* 3331 * XXX - issue with put*fh operations. Suppose /export/home is exported. 3332 * Suppose an NFS client goes to mount /export/home/joe. If /export, home, 3333 * or joe have restrictive search permissions, then we shouldn't let 3334 * the client get a file handle. This is easy to enforce. However, we 3335 * don't know what security flavor should be used until we resolve the 3336 * path name. Another complication is uid mapping. If root is 3337 * the user, then it will be mapped to the anonymous user by default, 3338 * but we won't know that till we've resolved the path name. And we won't 3339 * know what the anonymous user is. 3340 * Luckily, SECINFO is specified to take a full filename. 3341 * So what we will have to in rfs4_op_lookup is check that flavor of 3342 * the target object matches that of the request, and if root was the 3343 * caller, check for the root= and anon= options, and if necessary, 3344 * repeat the lookup using the right cred_t. But that's not done yet. 3345 */ 3346 /* ARGSUSED */ 3347 static void 3348 rfs4_op_putfh(nfs_argop4 *argop, nfs_resop4 *resop, struct svc_req *req, 3349 struct compound_state *cs) 3350 { 3351 PUTFH4args *args = &argop->nfs_argop4_u.opputfh; 3352 PUTFH4res *resp = &resop->nfs_resop4_u.opputfh; 3353 nfs_fh4_fmt_t *fh_fmtp; 3354 3355 DTRACE_NFSV4_2(op__putfh__start, struct compound_state *, cs, 3356 PUTFH4args *, args); 3357 3358 if (cs->vp) { 3359 VN_RELE(cs->vp); 3360 cs->vp = NULL; 3361 } 3362 3363 if (cs->cr) { 3364 crfree(cs->cr); 3365 cs->cr = NULL; 3366 } 3367 3368 3369 if (args->object.nfs_fh4_len < NFS_FH4_LEN) { 3370 *cs->statusp = resp->status = NFS4ERR_BADHANDLE; 3371 goto out; 3372 } 3373 3374 fh_fmtp = (nfs_fh4_fmt_t *)args->object.nfs_fh4_val; 3375 cs->exi = checkexport4(&fh_fmtp->fh4_fsid, (fid_t *)&fh_fmtp->fh4_xlen, 3376 NULL); 3377 3378 if (cs->exi == NULL) { 3379 *cs->statusp = resp->status = NFS4ERR_STALE; 3380 goto out; 3381 } 3382 3383 cs->cr = crdup(cs->basecr); 3384 3385 ASSERT(cs->cr != NULL); 3386 3387 if (! (cs->vp = nfs4_fhtovp(&args->object, cs->exi, &resp->status))) { 3388 *cs->statusp = resp->status; 3389 goto out; 3390 } 3391 3392 if ((resp->status = call_checkauth4(cs, req)) != NFS4_OK) { 3393 VN_RELE(cs->vp); 3394 cs->vp = NULL; 3395 goto out; 3396 } 3397 3398 nfs_fh4_copy(&args->object, &cs->fh); 3399 *cs->statusp = resp->status = NFS4_OK; 3400 cs->deleg = FALSE; 3401 3402 out: 3403 DTRACE_NFSV4_2(op__putfh__done, struct compound_state *, cs, 3404 PUTFH4res *, resp); 3405 } 3406 3407 /* ARGSUSED */ 3408 static void 3409 rfs4_op_putrootfh(nfs_argop4 *argop, nfs_resop4 *resop, struct svc_req *req, 3410 struct compound_state *cs) 3411 3412 { 3413 PUTROOTFH4res *resp = &resop->nfs_resop4_u.opputrootfh; 3414 int error; 3415 fid_t fid; 3416 struct exportinfo *exi, *sav_exi; 3417 3418 DTRACE_NFSV4_1(op__putrootfh__start, struct compound_state *, cs); 3419 3420 if (cs->vp) { 3421 VN_RELE(cs->vp); 3422 cs->vp = NULL; 3423 } 3424 3425 if (cs->cr) 3426 crfree(cs->cr); 3427 3428 cs->cr = crdup(cs->basecr); 3429 3430 /* 3431 * Using rootdir, the system root vnode, 3432 * get its fid. 3433 */ 3434 bzero(&fid, sizeof (fid)); 3435 fid.fid_len = MAXFIDSZ; 3436 error = vop_fid_pseudo(rootdir, &fid); 3437 if (error != 0) { 3438 *cs->statusp = resp->status = puterrno4(error); 3439 goto out; 3440 } 3441 3442 /* 3443 * Then use the root fsid & fid it to find out if it's exported 3444 * 3445 * If the server root isn't exported directly, then 3446 * it should at least be a pseudo export based on 3447 * one or more exports further down in the server's 3448 * file tree. 3449 */ 3450 exi = checkexport4(&rootdir->v_vfsp->vfs_fsid, &fid, NULL); 3451 if (exi == NULL || exi->exi_export.ex_flags & EX_PUBLIC) { 3452 NFS4_DEBUG(rfs4_debug, 3453 (CE_WARN, "rfs4_op_putrootfh: export check failure")); 3454 *cs->statusp = resp->status = NFS4ERR_SERVERFAULT; 3455 goto out; 3456 } 3457 3458 /* 3459 * Now make a filehandle based on the root 3460 * export and root vnode. 3461 */ 3462 error = makefh4(&cs->fh, rootdir, exi); 3463 if (error != 0) { 3464 *cs->statusp = resp->status = puterrno4(error); 3465 goto out; 3466 } 3467 3468 sav_exi = cs->exi; 3469 cs->exi = exi; 3470 3471 VN_HOLD(rootdir); 3472 cs->vp = rootdir; 3473 3474 if ((resp->status = call_checkauth4(cs, req)) != NFS4_OK) { 3475 VN_RELE(rootdir); 3476 cs->vp = NULL; 3477 cs->exi = sav_exi; 3478 goto out; 3479 } 3480 3481 *cs->statusp = resp->status = NFS4_OK; 3482 cs->deleg = FALSE; 3483 out: 3484 DTRACE_NFSV4_2(op__putrootfh__done, struct compound_state *, cs, 3485 PUTROOTFH4res *, resp); 3486 } 3487 3488 /* 3489 * A directory entry is a valid nfsv4 entry if 3490 * - it has a non-zero ino 3491 * - it is not a dot or dotdot name 3492 * - it is visible in a pseudo export or in a real export that can 3493 * only have a limited view. 3494 */ 3495 static bool_t 3496 valid_nfs4_entry(struct exportinfo *exi, struct dirent64 *dp, 3497 int *expseudo, int check_visible) 3498 { 3499 if (dp->d_ino == 0 || NFS_IS_DOTNAME(dp->d_name)) { 3500 *expseudo = 0; 3501 return (FALSE); 3502 } 3503 3504 if (! check_visible) { 3505 *expseudo = 0; 3506 return (TRUE); 3507 } 3508 3509 return (nfs_visible_inode(exi, dp->d_ino, expseudo)); 3510 } 3511 3512 /* 3513 * set_rdattr_params sets up the variables used to manage what information 3514 * to get for each directory entry. 3515 */ 3516 static nfsstat4 3517 set_rdattr_params(struct nfs4_svgetit_arg *sargp, 3518 bitmap4 attrs, bool_t *need_to_lookup) 3519 { 3520 uint_t va_mask; 3521 nfsstat4 status; 3522 bitmap4 objbits; 3523 3524 status = bitmap4_to_attrmask(attrs, sargp); 3525 if (status != NFS4_OK) { 3526 /* 3527 * could not even figure attr mask 3528 */ 3529 return (status); 3530 } 3531 va_mask = sargp->vap->va_mask; 3532 3533 /* 3534 * dirent's d_ino is always correct value for mounted_on_fileid. 3535 * mntdfid_set is set once here, but mounted_on_fileid is 3536 * set in main dirent processing loop for each dirent. 3537 * The mntdfid_set is a simple optimization that lets the 3538 * server attr code avoid work when caller is readdir. 3539 */ 3540 sargp->mntdfid_set = TRUE; 3541 3542 /* 3543 * Lookup entry only if client asked for any of the following: 3544 * a) vattr attrs 3545 * b) vfs attrs 3546 * c) attrs w/per-object scope requested (change, filehandle, etc) 3547 * other than mounted_on_fileid (which we can take from dirent) 3548 */ 3549 objbits = attrs ? attrs & NFS4_VP_ATTR_MASK : 0; 3550 3551 if (va_mask || sargp->sbp || (objbits & ~FATTR4_MOUNTED_ON_FILEID_MASK)) 3552 *need_to_lookup = TRUE; 3553 else 3554 *need_to_lookup = FALSE; 3555 3556 if (sargp->sbp == NULL) 3557 return (NFS4_OK); 3558 3559 /* 3560 * If filesystem attrs are requested, get them now from the 3561 * directory vp, as most entries will have same filesystem. The only 3562 * exception are mounted over entries but we handle 3563 * those as we go (XXX mounted over detection not yet implemented). 3564 */ 3565 sargp->vap->va_mask = 0; /* to avoid VOP_GETATTR */ 3566 status = bitmap4_get_sysattrs(sargp); 3567 sargp->vap->va_mask = va_mask; 3568 3569 if ((status != NFS4_OK) && sargp->rdattr_error_req) { 3570 /* 3571 * Failed to get filesystem attributes. 3572 * Return a rdattr_error for each entry, but don't fail. 3573 * However, don't get any obj-dependent attrs. 3574 */ 3575 sargp->rdattr_error = status; /* for rdattr_error */ 3576 *need_to_lookup = FALSE; 3577 /* 3578 * At least get fileid for regular readdir output 3579 */ 3580 sargp->vap->va_mask &= AT_NODEID; 3581 status = NFS4_OK; 3582 } 3583 3584 return (status); 3585 } 3586 3587 /* 3588 * readlink: args: CURRENT_FH. 3589 * res: status. If success - CURRENT_FH unchanged, return linktext. 3590 */ 3591 3592 /* ARGSUSED */ 3593 static void 3594 rfs4_op_readlink(nfs_argop4 *argop, nfs_resop4 *resop, struct svc_req *req, 3595 struct compound_state *cs) 3596 { 3597 READLINK4res *resp = &resop->nfs_resop4_u.opreadlink; 3598 int error; 3599 vnode_t *vp; 3600 struct iovec iov; 3601 struct vattr va; 3602 struct uio uio; 3603 char *data; 3604 3605 DTRACE_NFSV4_1(op__readlink__start, struct compound_state *, cs); 3606 3607 /* CURRENT_FH: directory */ 3608 vp = cs->vp; 3609 if (vp == NULL) { 3610 *cs->statusp = resp->status = NFS4ERR_NOFILEHANDLE; 3611 goto out; 3612 } 3613 3614 if (cs->access == CS_ACCESS_DENIED) { 3615 *cs->statusp = resp->status = NFS4ERR_ACCESS; 3616 goto out; 3617 } 3618 3619 if (vp->v_type == VDIR) { 3620 *cs->statusp = resp->status = NFS4ERR_ISDIR; 3621 goto out; 3622 } 3623 3624 if (vp->v_type != VLNK) { 3625 *cs->statusp = resp->status = NFS4ERR_INVAL; 3626 goto out; 3627 } 3628 3629 va.va_mask = AT_MODE; 3630 error = VOP_GETATTR(vp, &va, 0, cs->cr, NULL); 3631 if (error) { 3632 *cs->statusp = resp->status = puterrno4(error); 3633 goto out; 3634 } 3635 3636 if (MANDLOCK(vp, va.va_mode)) { 3637 *cs->statusp = resp->status = NFS4ERR_ACCESS; 3638 goto out; 3639 } 3640 3641 data = kmem_alloc(MAXPATHLEN + 1, KM_SLEEP); 3642 3643 iov.iov_base = data; 3644 iov.iov_len = MAXPATHLEN; 3645 uio.uio_iov = &iov; 3646 uio.uio_iovcnt = 1; 3647 uio.uio_segflg = UIO_SYSSPACE; 3648 uio.uio_extflg = UIO_COPY_CACHED; 3649 uio.uio_loffset = 0; 3650 uio.uio_resid = MAXPATHLEN; 3651 3652 error = VOP_READLINK(vp, &uio, cs->cr, NULL); 3653 3654 if (error) { 3655 kmem_free((caddr_t)data, (uint_t)MAXPATHLEN + 1); 3656 *cs->statusp = resp->status = puterrno4(error); 3657 goto out; 3658 } 3659 3660 *(data + MAXPATHLEN - uio.uio_resid) = '\0'; 3661 3662 /* 3663 * treat link name as data 3664 */ 3665 (void) str_to_utf8(data, &resp->link); 3666 3667 kmem_free((caddr_t)data, (uint_t)MAXPATHLEN + 1); 3668 *cs->statusp = resp->status = NFS4_OK; 3669 3670 out: 3671 DTRACE_NFSV4_2(op__readlink__done, struct compound_state *, cs, 3672 READLINK4res *, resp); 3673 } 3674 3675 static void 3676 rfs4_op_readlink_free(nfs_resop4 *resop) 3677 { 3678 READLINK4res *resp = &resop->nfs_resop4_u.opreadlink; 3679 utf8string *symlink = &resp->link; 3680 3681 if (symlink->utf8string_val) { 3682 UTF8STRING_FREE(*symlink) 3683 } 3684 } 3685 3686 /* 3687 * release_lockowner: 3688 * Release any state associated with the supplied 3689 * lockowner. Note if any lo_state is holding locks we will not 3690 * rele that lo_state and thus the lockowner will not be destroyed. 3691 * A client using lock after the lock owner stateid has been released 3692 * will suffer the consequence of NFS4ERR_BAD_STATEID and would have 3693 * to reissue the lock with new_lock_owner set to TRUE. 3694 * args: lock_owner 3695 * res: status 3696 */ 3697 /* ARGSUSED */ 3698 static void 3699 rfs4_op_release_lockowner(nfs_argop4 *argop, nfs_resop4 *resop, 3700 struct svc_req *req, struct compound_state *cs) 3701 { 3702 RELEASE_LOCKOWNER4args *ap = &argop->nfs_argop4_u.oprelease_lockowner; 3703 RELEASE_LOCKOWNER4res *resp = &resop->nfs_resop4_u.oprelease_lockowner; 3704 rfs4_lockowner_t *lo; 3705 rfs4_openowner_t *oop; 3706 rfs4_state_t *sp; 3707 rfs4_lo_state_t *lsp; 3708 rfs4_client_t *cp; 3709 bool_t create = FALSE; 3710 locklist_t *llist; 3711 sysid_t sysid; 3712 3713 DTRACE_NFSV4_2(op__release__lockowner__start, struct compound_state *, 3714 cs, RELEASE_LOCKOWNER4args *, ap); 3715 3716 /* Make sure there is a clientid around for this request */ 3717 cp = rfs4_findclient_by_id(ap->lock_owner.clientid, FALSE); 3718 3719 if (cp == NULL) { 3720 *cs->statusp = resp->status = 3721 rfs4_check_clientid(&ap->lock_owner.clientid, 0); 3722 goto out; 3723 } 3724 rfs4_client_rele(cp); 3725 3726 lo = rfs4_findlockowner(&ap->lock_owner, &create); 3727 if (lo == NULL) { 3728 *cs->statusp = resp->status = NFS4_OK; 3729 goto out; 3730 } 3731 ASSERT(lo->client != NULL); 3732 3733 /* 3734 * Check for EXPIRED client. If so will reap state with in a lease 3735 * period or on next set_clientid_confirm step 3736 */ 3737 if (rfs4_lease_expired(lo->client)) { 3738 rfs4_lockowner_rele(lo); 3739 *cs->statusp = resp->status = NFS4ERR_EXPIRED; 3740 goto out; 3741 } 3742 3743 /* 3744 * If no sysid has been assigned, then no locks exist; just return. 3745 */ 3746 rfs4_dbe_lock(lo->client->dbe); 3747 if (lo->client->sysidt == LM_NOSYSID) { 3748 rfs4_lockowner_rele(lo); 3749 rfs4_dbe_unlock(lo->client->dbe); 3750 goto out; 3751 } 3752 3753 sysid = lo->client->sysidt; 3754 rfs4_dbe_unlock(lo->client->dbe); 3755 3756 /* 3757 * Mark the lockowner invalid. 3758 */ 3759 rfs4_dbe_hide(lo->dbe); 3760 3761 /* 3762 * sysid-pid pair should now not be used since the lockowner is 3763 * invalid. If the client were to instantiate the lockowner again 3764 * it would be assigned a new pid. Thus we can get the list of 3765 * current locks. 3766 */ 3767 3768 llist = flk_get_active_locks(sysid, lo->pid); 3769 /* If we are still holding locks fail */ 3770 if (llist != NULL) { 3771 3772 *cs->statusp = resp->status = NFS4ERR_LOCKS_HELD; 3773 3774 flk_free_locklist(llist); 3775 /* 3776 * We need to unhide the lockowner so the client can 3777 * try it again. The bad thing here is if the client 3778 * has a logic error that took it here in the first place 3779 * he probably has lost accounting of the locks that it 3780 * is holding. So we may have dangling state until the 3781 * open owner state is reaped via close. One scenario 3782 * that could possibly occur is that the client has 3783 * sent the unlock request(s) in separate threads 3784 * and has not waited for the replies before sending the 3785 * RELEASE_LOCKOWNER request. Presumably, it would expect 3786 * and deal appropriately with NFS4ERR_LOCKS_HELD, by 3787 * reissuing the request. 3788 */ 3789 rfs4_dbe_unhide(lo->dbe); 3790 rfs4_lockowner_rele(lo); 3791 goto out; 3792 } 3793 3794 /* 3795 * For the corresponding client we need to check each open 3796 * owner for any opens that have lockowner state associated 3797 * with this lockowner. 3798 */ 3799 3800 rfs4_dbe_lock(lo->client->dbe); 3801 for (oop = lo->client->openownerlist.next->oop; oop != NULL; 3802 oop = oop->openownerlist.next->oop) { 3803 3804 rfs4_dbe_lock(oop->dbe); 3805 for (sp = oop->ownerstateids.next->sp; sp != NULL; 3806 sp = sp->ownerstateids.next->sp) { 3807 3808 rfs4_dbe_lock(sp->dbe); 3809 for (lsp = sp->lockownerlist.next->lsp; 3810 lsp != NULL; lsp = lsp->lockownerlist.next->lsp) { 3811 if (lsp->locker == lo) { 3812 rfs4_dbe_lock(lsp->dbe); 3813 rfs4_dbe_invalidate(lsp->dbe); 3814 rfs4_dbe_unlock(lsp->dbe); 3815 } 3816 } 3817 rfs4_dbe_unlock(sp->dbe); 3818 } 3819 rfs4_dbe_unlock(oop->dbe); 3820 } 3821 rfs4_dbe_unlock(lo->client->dbe); 3822 3823 rfs4_lockowner_rele(lo); 3824 3825 *cs->statusp = resp->status = NFS4_OK; 3826 3827 out: 3828 DTRACE_NFSV4_2(op__release__lockowner__done, struct compound_state *, 3829 cs, RELEASE_LOCKOWNER4res *, resp); 3830 } 3831 3832 /* 3833 * short utility function to lookup a file and recall the delegation 3834 */ 3835 static rfs4_file_t * 3836 rfs4_lookup_and_findfile(vnode_t *dvp, char *nm, vnode_t **vpp, 3837 int *lkup_error, cred_t *cr) 3838 { 3839 vnode_t *vp; 3840 rfs4_file_t *fp = NULL; 3841 bool_t fcreate = FALSE; 3842 int error; 3843 3844 if (vpp) 3845 *vpp = NULL; 3846 3847 if ((error = VOP_LOOKUP(dvp, nm, &vp, NULL, 0, NULL, cr, NULL, NULL, 3848 NULL)) == 0) { 3849 if (vp->v_type == VREG) 3850 fp = rfs4_findfile(vp, NULL, &fcreate); 3851 if (vpp) 3852 *vpp = vp; 3853 else 3854 VN_RELE(vp); 3855 } 3856 3857 if (lkup_error) 3858 *lkup_error = error; 3859 3860 return (fp); 3861 } 3862 3863 /* 3864 * remove: args: CURRENT_FH: directory; name. 3865 * res: status. If success - CURRENT_FH unchanged, return change_info 3866 * for directory. 3867 */ 3868 /* ARGSUSED */ 3869 static void 3870 rfs4_op_remove(nfs_argop4 *argop, nfs_resop4 *resop, struct svc_req *req, 3871 struct compound_state *cs) 3872 { 3873 REMOVE4args *args = &argop->nfs_argop4_u.opremove; 3874 REMOVE4res *resp = &resop->nfs_resop4_u.opremove; 3875 int error; 3876 vnode_t *dvp, *vp; 3877 struct vattr bdva, idva, adva; 3878 char *nm; 3879 uint_t len; 3880 rfs4_file_t *fp; 3881 int in_crit = 0; 3882 bslabel_t *clabel; 3883 3884 DTRACE_NFSV4_2(op__remove__start, struct compound_state *, cs, 3885 REMOVE4args *, args); 3886 3887 /* CURRENT_FH: directory */ 3888 dvp = cs->vp; 3889 if (dvp == NULL) { 3890 *cs->statusp = resp->status = NFS4ERR_NOFILEHANDLE; 3891 goto out; 3892 } 3893 3894 if (cs->access == CS_ACCESS_DENIED) { 3895 *cs->statusp = resp->status = NFS4ERR_ACCESS; 3896 goto out; 3897 } 3898 3899 /* 3900 * If there is an unshared filesystem mounted on this vnode, 3901 * Do not allow to remove anything in this directory. 3902 */ 3903 if (vn_ismntpt(dvp)) { 3904 *cs->statusp = resp->status = NFS4ERR_ACCESS; 3905 goto out; 3906 } 3907 3908 if (dvp->v_type != VDIR) { 3909 *cs->statusp = resp->status = NFS4ERR_NOTDIR; 3910 goto out; 3911 } 3912 3913 if (!utf8_dir_verify(&args->target)) { 3914 *cs->statusp = resp->status = NFS4ERR_INVAL; 3915 goto out; 3916 } 3917 3918 /* 3919 * Lookup the file so that we can check if it's a directory 3920 */ 3921 nm = utf8_to_fn(&args->target, &len, NULL); 3922 if (nm == NULL) { 3923 *cs->statusp = resp->status = NFS4ERR_INVAL; 3924 goto out; 3925 } 3926 3927 if (len > MAXNAMELEN) { 3928 *cs->statusp = resp->status = NFS4ERR_NAMETOOLONG; 3929 kmem_free(nm, len); 3930 goto out; 3931 } 3932 3933 if (rdonly4(cs->exi, cs->vp, req)) { 3934 *cs->statusp = resp->status = NFS4ERR_ROFS; 3935 kmem_free(nm, len); 3936 goto out; 3937 } 3938 3939 /* 3940 * Lookup the file to determine type and while we are see if 3941 * there is a file struct around and check for delegation. 3942 * We don't need to acquire va_seq before this lookup, if 3943 * it causes an update, cinfo.before will not match, which will 3944 * trigger a cache flush even if atomic is TRUE. 3945 */ 3946 if (fp = rfs4_lookup_and_findfile(dvp, nm, &vp, &error, cs->cr)) { 3947 if (rfs4_check_delegated_byfp(FWRITE, fp, TRUE, TRUE, TRUE, 3948 NULL)) { 3949 VN_RELE(vp); 3950 rfs4_file_rele(fp); 3951 *cs->statusp = resp->status = NFS4ERR_DELAY; 3952 kmem_free(nm, len); 3953 goto out; 3954 } 3955 } 3956 3957 /* Didn't find anything to remove */ 3958 if (vp == NULL) { 3959 *cs->statusp = resp->status = error; 3960 kmem_free(nm, len); 3961 goto out; 3962 } 3963 3964 if (nbl_need_check(vp)) { 3965 nbl_start_crit(vp, RW_READER); 3966 in_crit = 1; 3967 if (nbl_conflict(vp, NBL_REMOVE, 0, 0, 0, NULL)) { 3968 *cs->statusp = resp->status = NFS4ERR_FILE_OPEN; 3969 kmem_free(nm, len); 3970 nbl_end_crit(vp); 3971 VN_RELE(vp); 3972 if (fp) { 3973 rfs4_clear_dont_grant(fp); 3974 rfs4_file_rele(fp); 3975 } 3976 goto out; 3977 } 3978 } 3979 3980 /* check label before allowing removal */ 3981 if (is_system_labeled()) { 3982 ASSERT(req->rq_label != NULL); 3983 clabel = req->rq_label; 3984 DTRACE_PROBE2(tx__rfs4__log__info__opremove__clabel, char *, 3985 "got client label from request(1)", 3986 struct svc_req *, req); 3987 if (!blequal(&l_admin_low->tsl_label, clabel)) { 3988 if (!do_rfs_label_check(clabel, vp, EQUALITY_CHECK)) { 3989 *cs->statusp = resp->status = NFS4ERR_ACCESS; 3990 kmem_free(nm, len); 3991 if (in_crit) 3992 nbl_end_crit(vp); 3993 VN_RELE(vp); 3994 if (fp) { 3995 rfs4_clear_dont_grant(fp); 3996 rfs4_file_rele(fp); 3997 } 3998 goto out; 3999 } 4000 } 4001 } 4002 4003 /* Get dir "before" change value */ 4004 bdva.va_mask = AT_CTIME|AT_SEQ; 4005 error = VOP_GETATTR(dvp, &bdva, 0, cs->cr, NULL); 4006 if (error) { 4007 *cs->statusp = resp->status = puterrno4(error); 4008 kmem_free(nm, len); 4009 goto out; 4010 } 4011 NFS4_SET_FATTR4_CHANGE(resp->cinfo.before, bdva.va_ctime) 4012 4013 /* Actually do the REMOVE operation */ 4014 if (vp->v_type == VDIR) { 4015 /* 4016 * Can't remove a directory that has a mounted-on filesystem. 4017 */ 4018 if (vn_ismntpt(vp)) { 4019 error = EACCES; 4020 } else { 4021 /* 4022 * System V defines rmdir to return EEXIST, 4023 * not * ENOTEMPTY, if the directory is not 4024 * empty. A System V NFS server needs to map 4025 * NFS4ERR_EXIST to NFS4ERR_NOTEMPTY to 4026 * transmit over the wire. 4027 */ 4028 if ((error = VOP_RMDIR(dvp, nm, rootdir, cs->cr, 4029 NULL, 0)) == EEXIST) 4030 error = ENOTEMPTY; 4031 } 4032 } else { 4033 if ((error = VOP_REMOVE(dvp, nm, cs->cr, NULL, 0)) == 0 && 4034 fp != NULL) { 4035 struct vattr va; 4036 vnode_t *tvp; 4037 4038 rfs4_dbe_lock(fp->dbe); 4039 tvp = fp->vp; 4040 if (tvp) 4041 VN_HOLD(tvp); 4042 rfs4_dbe_unlock(fp->dbe); 4043 4044 if (tvp) { 4045 /* 4046 * This is va_seq safe because we are not 4047 * manipulating dvp. 4048 */ 4049 va.va_mask = AT_NLINK; 4050 if (!VOP_GETATTR(tvp, &va, 0, cs->cr, NULL) && 4051 va.va_nlink == 0) { 4052 /* Remove state on file remove */ 4053 if (in_crit) { 4054 nbl_end_crit(vp); 4055 in_crit = 0; 4056 } 4057 rfs4_close_all_state(fp); 4058 } 4059 VN_RELE(tvp); 4060 } 4061 } 4062 } 4063 4064 if (in_crit) 4065 nbl_end_crit(vp); 4066 VN_RELE(vp); 4067 4068 if (fp) { 4069 rfs4_clear_dont_grant(fp); 4070 rfs4_file_rele(fp); 4071 } 4072 kmem_free(nm, len); 4073 4074 if (error) { 4075 *cs->statusp = resp->status = puterrno4(error); 4076 goto out; 4077 } 4078 4079 /* 4080 * Get the initial "after" sequence number, if it fails, set to zero 4081 */ 4082 idva.va_mask = AT_SEQ; 4083 if (VOP_GETATTR(dvp, &idva, 0, cs->cr, NULL)) 4084 idva.va_seq = 0; 4085 4086 /* 4087 * Force modified data and metadata out to stable storage. 4088 */ 4089 (void) VOP_FSYNC(dvp, 0, cs->cr, NULL); 4090 4091 /* 4092 * Get "after" change value, if it fails, simply return the 4093 * before value. 4094 */ 4095 adva.va_mask = AT_CTIME|AT_SEQ; 4096 if (VOP_GETATTR(dvp, &adva, 0, cs->cr, NULL)) { 4097 adva.va_ctime = bdva.va_ctime; 4098 adva.va_seq = 0; 4099 } 4100 4101 NFS4_SET_FATTR4_CHANGE(resp->cinfo.after, adva.va_ctime) 4102 4103 /* 4104 * The cinfo.atomic = TRUE only if we have 4105 * non-zero va_seq's, and it has incremented by exactly one 4106 * during the VOP_REMOVE/RMDIR and it didn't change during 4107 * the VOP_FSYNC. 4108 */ 4109 if (bdva.va_seq && idva.va_seq && adva.va_seq && 4110 idva.va_seq == (bdva.va_seq + 1) && idva.va_seq == adva.va_seq) 4111 resp->cinfo.atomic = TRUE; 4112 else 4113 resp->cinfo.atomic = FALSE; 4114 4115 *cs->statusp = resp->status = NFS4_OK; 4116 4117 out: 4118 DTRACE_NFSV4_2(op__remove__done, struct compound_state *, cs, 4119 REMOVE4res *, resp); 4120 } 4121 4122 /* 4123 * rename: args: SAVED_FH: from directory, CURRENT_FH: target directory, 4124 * oldname and newname. 4125 * res: status. If success - CURRENT_FH unchanged, return change_info 4126 * for both from and target directories. 4127 */ 4128 /* ARGSUSED */ 4129 static void 4130 rfs4_op_rename(nfs_argop4 *argop, nfs_resop4 *resop, struct svc_req *req, 4131 struct compound_state *cs) 4132 { 4133 RENAME4args *args = &argop->nfs_argop4_u.oprename; 4134 RENAME4res *resp = &resop->nfs_resop4_u.oprename; 4135 int error; 4136 vnode_t *odvp; 4137 vnode_t *ndvp; 4138 vnode_t *srcvp, *targvp; 4139 struct vattr obdva, oidva, oadva; 4140 struct vattr nbdva, nidva, nadva; 4141 char *onm, *nnm; 4142 uint_t olen, nlen; 4143 rfs4_file_t *fp, *sfp; 4144 int in_crit_src, in_crit_targ; 4145 int fp_rele_grant_hold, sfp_rele_grant_hold; 4146 bslabel_t *clabel; 4147 4148 DTRACE_NFSV4_2(op__rename__start, struct compound_state *, cs, 4149 RENAME4args *, args); 4150 4151 fp = sfp = NULL; 4152 srcvp = targvp = NULL; 4153 in_crit_src = in_crit_targ = 0; 4154 fp_rele_grant_hold = sfp_rele_grant_hold = 0; 4155 4156 /* CURRENT_FH: target directory */ 4157 ndvp = cs->vp; 4158 if (ndvp == NULL) { 4159 *cs->statusp = resp->status = NFS4ERR_NOFILEHANDLE; 4160 goto out; 4161 } 4162 4163 /* SAVED_FH: from directory */ 4164 odvp = cs->saved_vp; 4165 if (odvp == NULL) { 4166 *cs->statusp = resp->status = NFS4ERR_NOFILEHANDLE; 4167 goto out; 4168 } 4169 4170 if (cs->access == CS_ACCESS_DENIED) { 4171 *cs->statusp = resp->status = NFS4ERR_ACCESS; 4172 goto out; 4173 } 4174 4175 /* 4176 * If there is an unshared filesystem mounted on this vnode, 4177 * do not allow to rename objects in this directory. 4178 */ 4179 if (vn_ismntpt(odvp)) { 4180 *cs->statusp = resp->status = NFS4ERR_ACCESS; 4181 goto out; 4182 } 4183 4184 /* 4185 * If there is an unshared filesystem mounted on this vnode, 4186 * do not allow to rename to this directory. 4187 */ 4188 if (vn_ismntpt(ndvp)) { 4189 *cs->statusp = resp->status = NFS4ERR_ACCESS; 4190 goto out; 4191 } 4192 4193 if (odvp->v_type != VDIR || ndvp->v_type != VDIR) { 4194 *cs->statusp = resp->status = NFS4ERR_NOTDIR; 4195 goto out; 4196 } 4197 4198 if (cs->saved_exi != cs->exi) { 4199 *cs->statusp = resp->status = NFS4ERR_XDEV; 4200 goto out; 4201 } 4202 4203 if (!utf8_dir_verify(&args->oldname)) { 4204 *cs->statusp = resp->status = NFS4ERR_INVAL; 4205 goto out; 4206 } 4207 4208 if (!utf8_dir_verify(&args->newname)) { 4209 *cs->statusp = resp->status = NFS4ERR_INVAL; 4210 goto out; 4211 } 4212 4213 onm = utf8_to_fn(&args->oldname, &olen, NULL); 4214 if (onm == NULL) { 4215 *cs->statusp = resp->status = NFS4ERR_INVAL; 4216 goto out; 4217 } 4218 4219 nnm = utf8_to_fn(&args->newname, &nlen, NULL); 4220 if (nnm == NULL) { 4221 *cs->statusp = resp->status = NFS4ERR_INVAL; 4222 kmem_free(onm, olen); 4223 goto out; 4224 } 4225 4226 if (olen > MAXNAMELEN || nlen > MAXNAMELEN) { 4227 *cs->statusp = resp->status = NFS4ERR_NAMETOOLONG; 4228 kmem_free(onm, olen); 4229 kmem_free(nnm, nlen); 4230 goto out; 4231 } 4232 4233 4234 if (rdonly4(cs->exi, cs->vp, req)) { 4235 *cs->statusp = resp->status = NFS4ERR_ROFS; 4236 kmem_free(onm, olen); 4237 kmem_free(nnm, nlen); 4238 goto out; 4239 } 4240 4241 /* check label of the target dir */ 4242 if (is_system_labeled()) { 4243 ASSERT(req->rq_label != NULL); 4244 clabel = req->rq_label; 4245 DTRACE_PROBE2(tx__rfs4__log__info__oprename__clabel, char *, 4246 "got client label from request(1)", 4247 struct svc_req *, req); 4248 if (!blequal(&l_admin_low->tsl_label, clabel)) { 4249 if (!do_rfs_label_check(clabel, ndvp, 4250 EQUALITY_CHECK)) { 4251 *cs->statusp = resp->status = NFS4ERR_ACCESS; 4252 goto out; 4253 } 4254 } 4255 } 4256 4257 /* 4258 * Is the source a file and have a delegation? 4259 * We don't need to acquire va_seq before these lookups, if 4260 * it causes an update, cinfo.before will not match, which will 4261 * trigger a cache flush even if atomic is TRUE. 4262 */ 4263 if (sfp = rfs4_lookup_and_findfile(odvp, onm, &srcvp, &error, cs->cr)) { 4264 if (rfs4_check_delegated_byfp(FWRITE, sfp, TRUE, TRUE, TRUE, 4265 NULL)) { 4266 *cs->statusp = resp->status = NFS4ERR_DELAY; 4267 goto err_out; 4268 } 4269 } 4270 4271 if (srcvp == NULL) { 4272 *cs->statusp = resp->status = puterrno4(error); 4273 kmem_free(onm, olen); 4274 kmem_free(nnm, nlen); 4275 goto out; 4276 } 4277 4278 sfp_rele_grant_hold = 1; 4279 4280 /* Does the destination exist and a file and have a delegation? */ 4281 if (fp = rfs4_lookup_and_findfile(ndvp, nnm, &targvp, NULL, cs->cr)) { 4282 if (rfs4_check_delegated_byfp(FWRITE, fp, TRUE, TRUE, TRUE, 4283 NULL)) { 4284 *cs->statusp = resp->status = NFS4ERR_DELAY; 4285 goto err_out; 4286 } 4287 } 4288 fp_rele_grant_hold = 1; 4289 4290 4291 /* Check for NBMAND lock on both source and target */ 4292 if (nbl_need_check(srcvp)) { 4293 nbl_start_crit(srcvp, RW_READER); 4294 in_crit_src = 1; 4295 if (nbl_conflict(srcvp, NBL_RENAME, 0, 0, 0, NULL)) { 4296 *cs->statusp = resp->status = NFS4ERR_FILE_OPEN; 4297 goto err_out; 4298 } 4299 } 4300 4301 if (targvp && nbl_need_check(targvp)) { 4302 nbl_start_crit(targvp, RW_READER); 4303 in_crit_targ = 1; 4304 if (nbl_conflict(targvp, NBL_REMOVE, 0, 0, 0, NULL)) { 4305 *cs->statusp = resp->status = NFS4ERR_FILE_OPEN; 4306 goto err_out; 4307 } 4308 } 4309 4310 /* Get source "before" change value */ 4311 obdva.va_mask = AT_CTIME|AT_SEQ; 4312 error = VOP_GETATTR(odvp, &obdva, 0, cs->cr, NULL); 4313 if (!error) { 4314 nbdva.va_mask = AT_CTIME|AT_SEQ; 4315 error = VOP_GETATTR(ndvp, &nbdva, 0, cs->cr, NULL); 4316 } 4317 if (error) { 4318 *cs->statusp = resp->status = puterrno4(error); 4319 goto err_out; 4320 } 4321 4322 NFS4_SET_FATTR4_CHANGE(resp->source_cinfo.before, obdva.va_ctime) 4323 NFS4_SET_FATTR4_CHANGE(resp->target_cinfo.before, nbdva.va_ctime) 4324 4325 if ((error = VOP_RENAME(odvp, onm, ndvp, nnm, cs->cr, NULL, 0)) == 0 && 4326 fp != NULL) { 4327 struct vattr va; 4328 vnode_t *tvp; 4329 4330 rfs4_dbe_lock(fp->dbe); 4331 tvp = fp->vp; 4332 if (tvp) 4333 VN_HOLD(tvp); 4334 rfs4_dbe_unlock(fp->dbe); 4335 4336 if (tvp) { 4337 va.va_mask = AT_NLINK; 4338 if (!VOP_GETATTR(tvp, &va, 0, cs->cr, NULL) && 4339 va.va_nlink == 0) { 4340 /* The file is gone and so should the state */ 4341 if (in_crit_targ) { 4342 nbl_end_crit(targvp); 4343 in_crit_targ = 0; 4344 } 4345 rfs4_close_all_state(fp); 4346 } 4347 VN_RELE(tvp); 4348 } 4349 } 4350 if (error == 0) 4351 vn_renamepath(ndvp, srcvp, nnm, nlen - 1); 4352 4353 if (in_crit_src) 4354 nbl_end_crit(srcvp); 4355 if (srcvp) 4356 VN_RELE(srcvp); 4357 if (in_crit_targ) 4358 nbl_end_crit(targvp); 4359 if (targvp) 4360 VN_RELE(targvp); 4361 4362 if (sfp) { 4363 rfs4_clear_dont_grant(sfp); 4364 rfs4_file_rele(sfp); 4365 } 4366 if (fp) { 4367 rfs4_clear_dont_grant(fp); 4368 rfs4_file_rele(fp); 4369 } 4370 4371 kmem_free(onm, olen); 4372 kmem_free(nnm, nlen); 4373 4374 /* 4375 * Get the initial "after" sequence number, if it fails, set to zero 4376 */ 4377 oidva.va_mask = AT_SEQ; 4378 if (VOP_GETATTR(odvp, &oidva, 0, cs->cr, NULL)) 4379 oidva.va_seq = 0; 4380 4381 nidva.va_mask = AT_SEQ; 4382 if (VOP_GETATTR(ndvp, &nidva, 0, cs->cr, NULL)) 4383 nidva.va_seq = 0; 4384 4385 /* 4386 * Force modified data and metadata out to stable storage. 4387 */ 4388 (void) VOP_FSYNC(odvp, 0, cs->cr, NULL); 4389 (void) VOP_FSYNC(ndvp, 0, cs->cr, NULL); 4390 4391 if (error) { 4392 *cs->statusp = resp->status = puterrno4(error); 4393 goto out; 4394 } 4395 4396 /* 4397 * Get "after" change values, if it fails, simply return the 4398 * before value. 4399 */ 4400 oadva.va_mask = AT_CTIME|AT_SEQ; 4401 if (VOP_GETATTR(odvp, &oadva, 0, cs->cr, NULL)) { 4402 oadva.va_ctime = obdva.va_ctime; 4403 oadva.va_seq = 0; 4404 } 4405 4406 nadva.va_mask = AT_CTIME|AT_SEQ; 4407 if (VOP_GETATTR(odvp, &nadva, 0, cs->cr, NULL)) { 4408 nadva.va_ctime = nbdva.va_ctime; 4409 nadva.va_seq = 0; 4410 } 4411 4412 NFS4_SET_FATTR4_CHANGE(resp->source_cinfo.after, oadva.va_ctime) 4413 NFS4_SET_FATTR4_CHANGE(resp->target_cinfo.after, nadva.va_ctime) 4414 4415 /* 4416 * The cinfo.atomic = TRUE only if we have 4417 * non-zero va_seq's, and it has incremented by exactly one 4418 * during the VOP_RENAME and it didn't change during the VOP_FSYNC. 4419 */ 4420 if (obdva.va_seq && oidva.va_seq && oadva.va_seq && 4421 oidva.va_seq == (obdva.va_seq + 1) && oidva.va_seq == oadva.va_seq) 4422 resp->source_cinfo.atomic = TRUE; 4423 else 4424 resp->source_cinfo.atomic = FALSE; 4425 4426 if (nbdva.va_seq && nidva.va_seq && nadva.va_seq && 4427 nidva.va_seq == (nbdva.va_seq + 1) && nidva.va_seq == nadva.va_seq) 4428 resp->target_cinfo.atomic = TRUE; 4429 else 4430 resp->target_cinfo.atomic = FALSE; 4431 4432 #ifdef VOLATILE_FH_TEST 4433 { 4434 extern void add_volrnm_fh(struct exportinfo *, vnode_t *); 4435 4436 /* 4437 * Add the renamed file handle to the volatile rename list 4438 */ 4439 if (cs->exi->exi_export.ex_flags & EX_VOLRNM) { 4440 /* file handles may expire on rename */ 4441 vnode_t *vp; 4442 4443 nnm = utf8_to_fn(&args->newname, &nlen, NULL); 4444 /* 4445 * Already know that nnm will be a valid string 4446 */ 4447 error = VOP_LOOKUP(ndvp, nnm, &vp, NULL, 0, NULL, cs->cr, 4448 NULL, NULL, NULL); 4449 kmem_free(nnm, nlen); 4450 if (!error) { 4451 add_volrnm_fh(cs->exi, vp); 4452 VN_RELE(vp); 4453 } 4454 } 4455 } 4456 #endif /* VOLATILE_FH_TEST */ 4457 4458 *cs->statusp = resp->status = NFS4_OK; 4459 out: 4460 DTRACE_NFSV4_2(op__rename__done, struct compound_state *, cs, 4461 RENAME4res *, resp); 4462 return; 4463 4464 err_out: 4465 kmem_free(onm, olen); 4466 kmem_free(nnm, nlen); 4467 4468 if (in_crit_src) nbl_end_crit(srcvp); 4469 if (in_crit_targ) nbl_end_crit(targvp); 4470 if (targvp) VN_RELE(targvp); 4471 if (srcvp) VN_RELE(srcvp); 4472 if (sfp) { 4473 if (sfp_rele_grant_hold) rfs4_clear_dont_grant(sfp); 4474 rfs4_file_rele(sfp); 4475 } 4476 if (fp) { 4477 if (fp_rele_grant_hold) rfs4_clear_dont_grant(fp); 4478 rfs4_file_rele(fp); 4479 } 4480 4481 DTRACE_NFSV4_2(op__rename__done, struct compound_state *, cs, 4482 RENAME4res *, resp); 4483 } 4484 4485 /* ARGSUSED */ 4486 static void 4487 rfs4_op_renew(nfs_argop4 *argop, nfs_resop4 *resop, struct svc_req *req, 4488 struct compound_state *cs) 4489 { 4490 RENEW4args *args = &argop->nfs_argop4_u.oprenew; 4491 RENEW4res *resp = &resop->nfs_resop4_u.oprenew; 4492 rfs4_client_t *cp; 4493 4494 DTRACE_NFSV4_2(op__renew__start, struct compound_state *, cs, 4495 RENEW4args *, args); 4496 4497 if ((cp = rfs4_findclient_by_id(args->clientid, FALSE)) == NULL) { 4498 *cs->statusp = resp->status = 4499 rfs4_check_clientid(&args->clientid, 0); 4500 goto out; 4501 } 4502 4503 if (rfs4_lease_expired(cp)) { 4504 rfs4_client_rele(cp); 4505 *cs->statusp = resp->status = NFS4ERR_EXPIRED; 4506 goto out; 4507 } 4508 4509 rfs4_update_lease(cp); 4510 4511 mutex_enter(cp->cbinfo.cb_lock); 4512 if (cp->cbinfo.cb_notified_of_cb_path_down == FALSE) { 4513 cp->cbinfo.cb_notified_of_cb_path_down = TRUE; 4514 *cs->statusp = resp->status = NFS4ERR_CB_PATH_DOWN; 4515 } else { 4516 *cs->statusp = resp->status = NFS4_OK; 4517 } 4518 mutex_exit(cp->cbinfo.cb_lock); 4519 4520 rfs4_client_rele(cp); 4521 4522 out: 4523 DTRACE_NFSV4_2(op__renew__done, struct compound_state *, cs, 4524 RENEW4res *, resp); 4525 } 4526 4527 /* ARGSUSED */ 4528 static void 4529 rfs4_op_restorefh(nfs_argop4 *args, nfs_resop4 *resop, struct svc_req *req, 4530 struct compound_state *cs) 4531 { 4532 RESTOREFH4res *resp = &resop->nfs_resop4_u.oprestorefh; 4533 4534 DTRACE_NFSV4_1(op__restorefh__start, struct compound_state *, cs); 4535 4536 /* No need to check cs->access - we are not accessing any object */ 4537 if ((cs->saved_vp == NULL) || (cs->saved_fh.nfs_fh4_val == NULL)) { 4538 *cs->statusp = resp->status = NFS4ERR_RESTOREFH; 4539 goto out; 4540 } 4541 if (cs->vp != NULL) { 4542 VN_RELE(cs->vp); 4543 } 4544 cs->vp = cs->saved_vp; 4545 cs->saved_vp = NULL; 4546 cs->exi = cs->saved_exi; 4547 nfs_fh4_copy(&cs->saved_fh, &cs->fh); 4548 *cs->statusp = resp->status = NFS4_OK; 4549 cs->deleg = FALSE; 4550 4551 out: 4552 DTRACE_NFSV4_2(op__restorefh__done, struct compound_state *, cs, 4553 RESTOREFH4res *, resp); 4554 } 4555 4556 /* ARGSUSED */ 4557 static void 4558 rfs4_op_savefh(nfs_argop4 *argop, nfs_resop4 *resop, struct svc_req *req, 4559 struct compound_state *cs) 4560 { 4561 SAVEFH4res *resp = &resop->nfs_resop4_u.opsavefh; 4562 4563 DTRACE_NFSV4_1(op__savefh__start, struct compound_state *, cs); 4564 4565 /* No need to check cs->access - we are not accessing any object */ 4566 if (cs->vp == NULL) { 4567 *cs->statusp = resp->status = NFS4ERR_NOFILEHANDLE; 4568 goto out; 4569 } 4570 if (cs->saved_vp != NULL) { 4571 VN_RELE(cs->saved_vp); 4572 } 4573 cs->saved_vp = cs->vp; 4574 VN_HOLD(cs->saved_vp); 4575 cs->saved_exi = cs->exi; 4576 /* 4577 * since SAVEFH is fairly rare, don't alloc space for its fh 4578 * unless necessary. 4579 */ 4580 if (cs->saved_fh.nfs_fh4_val == NULL) { 4581 cs->saved_fh.nfs_fh4_val = kmem_alloc(NFS4_FHSIZE, KM_SLEEP); 4582 } 4583 nfs_fh4_copy(&cs->fh, &cs->saved_fh); 4584 *cs->statusp = resp->status = NFS4_OK; 4585 4586 out: 4587 DTRACE_NFSV4_2(op__savefh__done, struct compound_state *, cs, 4588 SAVEFH4res *, resp); 4589 } 4590 4591 /* 4592 * rfs4_verify_attr is called when nfsv4 Setattr failed, but we wish to 4593 * return the bitmap of attrs that were set successfully. It is also 4594 * called by Verify/Nverify to test the vattr/vfsstat attrs. It should 4595 * always be called only after rfs4_do_set_attrs(). 4596 * 4597 * Verify that the attributes are same as the expected ones. sargp->vap 4598 * and sargp->sbp contain the input attributes as translated from fattr4. 4599 * 4600 * This function verifies only the attrs that correspond to a vattr or 4601 * vfsstat struct. That is because of the extra step needed to get the 4602 * corresponding system structs. Other attributes have already been set or 4603 * verified by do_rfs4_set_attrs. 4604 * 4605 * Return 0 if all attrs match, -1 if some don't, error if error processing. 4606 */ 4607 static int 4608 rfs4_verify_attr(struct nfs4_svgetit_arg *sargp, 4609 bitmap4 *resp, struct nfs4_ntov_table *ntovp) 4610 { 4611 int error, ret_error = 0; 4612 int i, k; 4613 uint_t sva_mask = sargp->vap->va_mask; 4614 uint_t vbit; 4615 union nfs4_attr_u *na; 4616 uint8_t *amap; 4617 bool_t getsb = ntovp->vfsstat; 4618 4619 if (sva_mask != 0) { 4620 /* 4621 * Okay to overwrite sargp->vap because we verify based 4622 * on the incoming values. 4623 */ 4624 ret_error = VOP_GETATTR(sargp->cs->vp, sargp->vap, 0, 4625 sargp->cs->cr, NULL); 4626 if (ret_error) { 4627 if (resp == NULL) 4628 return (ret_error); 4629 /* 4630 * Must return bitmap of successful attrs 4631 */ 4632 sva_mask = 0; /* to prevent checking vap later */ 4633 } else { 4634 /* 4635 * Some file systems clobber va_mask. it is probably 4636 * wrong of them to do so, nonethless we practice 4637 * defensive coding. 4638 * See bug id 4276830. 4639 */ 4640 sargp->vap->va_mask = sva_mask; 4641 } 4642 } 4643 4644 if (getsb) { 4645 /* 4646 * Now get the superblock and loop on the bitmap, as there is 4647 * no simple way of translating from superblock to bitmap4. 4648 */ 4649 ret_error = VFS_STATVFS(sargp->cs->vp->v_vfsp, sargp->sbp); 4650 if (ret_error) { 4651 if (resp == NULL) 4652 goto errout; 4653 getsb = FALSE; 4654 } 4655 } 4656 4657 /* 4658 * Now loop and verify each attribute which getattr returned 4659 * whether it's the same as the input. 4660 */ 4661 if (resp == NULL && !getsb && (sva_mask == 0)) 4662 goto errout; 4663 4664 na = ntovp->na; 4665 amap = ntovp->amap; 4666 k = 0; 4667 for (i = 0; i < ntovp->attrcnt; i++, na++, amap++) { 4668 k = *amap; 4669 ASSERT(nfs4_ntov_map[k].nval == k); 4670 vbit = nfs4_ntov_map[k].vbit; 4671 4672 /* 4673 * If vattr attribute but VOP_GETATTR failed, or it's 4674 * superblock attribute but VFS_STATVFS failed, skip 4675 */ 4676 if (vbit) { 4677 if ((vbit & sva_mask) == 0) 4678 continue; 4679 } else if (!(getsb && nfs4_ntov_map[k].vfsstat)) { 4680 continue; 4681 } 4682 error = (*nfs4_ntov_map[k].sv_getit)(NFS4ATTR_VERIT, sargp, na); 4683 if (resp != NULL) { 4684 if (error) 4685 ret_error = -1; /* not all match */ 4686 else /* update response bitmap */ 4687 *resp |= nfs4_ntov_map[k].fbit; 4688 continue; 4689 } 4690 if (error) { 4691 ret_error = -1; /* not all match */ 4692 break; 4693 } 4694 } 4695 errout: 4696 return (ret_error); 4697 } 4698 4699 /* 4700 * Decode the attribute to be set/verified. If the attr requires a sys op 4701 * (VOP_GETATTR, VFS_VFSSTAT), and the request is to verify, then don't 4702 * call the sv_getit function for it, because the sys op hasn't yet been done. 4703 * Return 0 for success, error code if failed. 4704 * 4705 * Note: the decoded arg is not freed here but in nfs4_ntov_table_free. 4706 */ 4707 static int 4708 decode_fattr4_attr(nfs4_attr_cmd_t cmd, struct nfs4_svgetit_arg *sargp, 4709 int k, XDR *xdrp, bitmap4 *resp_bval, union nfs4_attr_u *nap) 4710 { 4711 int error = 0; 4712 bool_t set_later; 4713 4714 sargp->vap->va_mask |= nfs4_ntov_map[k].vbit; 4715 4716 if ((*nfs4_ntov_map[k].xfunc)(xdrp, nap)) { 4717 set_later = nfs4_ntov_map[k].vbit || nfs4_ntov_map[k].vfsstat; 4718 /* 4719 * don't verify yet if a vattr or sb dependent attr, 4720 * because we don't have their sys values yet. 4721 * Will be done later. 4722 */ 4723 if (! (set_later && (cmd == NFS4ATTR_VERIT))) { 4724 /* 4725 * ACLs are a special case, since setting the MODE 4726 * conflicts with setting the ACL. We delay setting 4727 * the ACL until all other attributes have been set. 4728 * The ACL gets set in do_rfs4_op_setattr(). 4729 */ 4730 if (nfs4_ntov_map[k].fbit != FATTR4_ACL_MASK) { 4731 error = (*nfs4_ntov_map[k].sv_getit)(cmd, 4732 sargp, nap); 4733 if (error) { 4734 xdr_free(nfs4_ntov_map[k].xfunc, 4735 (caddr_t)nap); 4736 } 4737 } 4738 } 4739 } else { 4740 #ifdef DEBUG 4741 cmn_err(CE_NOTE, "decode_fattr4_attr: error " 4742 "decoding attribute %d\n", k); 4743 #endif 4744 error = EINVAL; 4745 } 4746 if (!error && resp_bval && !set_later) { 4747 *resp_bval |= nfs4_ntov_map[k].fbit; 4748 } 4749 4750 return (error); 4751 } 4752 4753 /* 4754 * Set vattr based on incoming fattr4 attrs - used by setattr. 4755 * Set response mask. Ignore any values that are not writable vattr attrs. 4756 */ 4757 static nfsstat4 4758 do_rfs4_set_attrs(bitmap4 *resp, fattr4 *fattrp, struct compound_state *cs, 4759 struct nfs4_svgetit_arg *sargp, struct nfs4_ntov_table *ntovp, 4760 nfs4_attr_cmd_t cmd) 4761 { 4762 int error = 0; 4763 int i; 4764 char *attrs = fattrp->attrlist4; 4765 uint32_t attrslen = fattrp->attrlist4_len; 4766 XDR xdr; 4767 nfsstat4 status = NFS4_OK; 4768 vnode_t *vp = cs->vp; 4769 union nfs4_attr_u *na; 4770 uint8_t *amap; 4771 4772 #ifndef lint 4773 /* 4774 * Make sure that maximum attribute number can be expressed as an 4775 * 8 bit quantity. 4776 */ 4777 ASSERT(NFS4_MAXNUM_ATTRS <= (UINT8_MAX + 1)); 4778 #endif 4779 4780 if (vp == NULL) { 4781 if (resp) 4782 *resp = 0; 4783 return (NFS4ERR_NOFILEHANDLE); 4784 } 4785 if (cs->access == CS_ACCESS_DENIED) { 4786 if (resp) 4787 *resp = 0; 4788 return (NFS4ERR_ACCESS); 4789 } 4790 4791 sargp->op = cmd; 4792 sargp->cs = cs; 4793 sargp->flag = 0; /* may be set later */ 4794 sargp->vap->va_mask = 0; 4795 sargp->rdattr_error = NFS4_OK; 4796 sargp->rdattr_error_req = FALSE; 4797 /* sargp->sbp is set by the caller */ 4798 4799 xdrmem_create(&xdr, attrs, attrslen, XDR_DECODE); 4800 4801 na = ntovp->na; 4802 amap = ntovp->amap; 4803 4804 /* 4805 * The following loop iterates on the nfs4_ntov_map checking 4806 * if the fbit is set in the requested bitmap. 4807 * If set then we process the arguments using the 4808 * rfs4_fattr4 conversion functions to populate the setattr 4809 * vattr and va_mask. Any settable attrs that are not using vattr 4810 * will be set in this loop. 4811 */ 4812 for (i = 0; i < nfs4_ntov_map_size; i++) { 4813 if (!(fattrp->attrmask & nfs4_ntov_map[i].fbit)) { 4814 continue; 4815 } 4816 /* 4817 * If setattr, must be a writable attr. 4818 * If verify/nverify, must be a readable attr. 4819 */ 4820 if ((error = (*nfs4_ntov_map[i].sv_getit)( 4821 NFS4ATTR_SUPPORTED, sargp, NULL)) != 0) { 4822 /* 4823 * Client tries to set/verify an 4824 * unsupported attribute, tries to set 4825 * a read only attr or verify a write 4826 * only one - error! 4827 */ 4828 break; 4829 } 4830 /* 4831 * Decode the attribute to set/verify 4832 */ 4833 error = decode_fattr4_attr(cmd, sargp, nfs4_ntov_map[i].nval, 4834 &xdr, resp ? resp : NULL, na); 4835 if (error) 4836 break; 4837 *amap++ = (uint8_t)nfs4_ntov_map[i].nval; 4838 na++; 4839 (ntovp->attrcnt)++; 4840 if (nfs4_ntov_map[i].vfsstat) 4841 ntovp->vfsstat = TRUE; 4842 } 4843 4844 if (error != 0) 4845 status = (error == ENOTSUP ? NFS4ERR_ATTRNOTSUPP : 4846 puterrno4(error)); 4847 /* xdrmem_destroy(&xdrs); */ /* NO-OP */ 4848 return (status); 4849 } 4850 4851 static nfsstat4 4852 do_rfs4_op_setattr(bitmap4 *resp, fattr4 *fattrp, struct compound_state *cs, 4853 stateid4 *stateid) 4854 { 4855 int error = 0; 4856 struct nfs4_svgetit_arg sarg; 4857 bool_t trunc; 4858 4859 nfsstat4 status = NFS4_OK; 4860 cred_t *cr = cs->cr; 4861 vnode_t *vp = cs->vp; 4862 struct nfs4_ntov_table ntov; 4863 struct statvfs64 sb; 4864 struct vattr bva; 4865 struct flock64 bf; 4866 int in_crit = 0; 4867 uint_t saved_mask = 0; 4868 caller_context_t ct; 4869 4870 *resp = 0; 4871 sarg.sbp = &sb; 4872 nfs4_ntov_table_init(&ntov); 4873 status = do_rfs4_set_attrs(resp, fattrp, cs, &sarg, &ntov, 4874 NFS4ATTR_SETIT); 4875 if (status != NFS4_OK) { 4876 /* 4877 * failed set attrs 4878 */ 4879 goto done; 4880 } 4881 if ((sarg.vap->va_mask == 0) && 4882 (! (fattrp->attrmask & FATTR4_ACL_MASK))) { 4883 /* 4884 * no further work to be done 4885 */ 4886 goto done; 4887 } 4888 4889 /* 4890 * If we got a request to set the ACL and the MODE, only 4891 * allow changing VSUID, VSGID, and VSVTX. Attempting 4892 * to change any other bits, along with setting an ACL, 4893 * gives NFS4ERR_INVAL. 4894 */ 4895 if ((fattrp->attrmask & FATTR4_ACL_MASK) && 4896 (fattrp->attrmask & FATTR4_MODE_MASK)) { 4897 vattr_t va; 4898 4899 va.va_mask = AT_MODE; 4900 error = VOP_GETATTR(vp, &va, 0, cs->cr, NULL); 4901 if (error) { 4902 status = puterrno4(error); 4903 goto done; 4904 } 4905 if ((sarg.vap->va_mode ^ va.va_mode) & 4906 ~(VSUID | VSGID | VSVTX)) { 4907 status = NFS4ERR_INVAL; 4908 goto done; 4909 } 4910 } 4911 4912 /* Check stateid only if size has been set */ 4913 if (sarg.vap->va_mask & AT_SIZE) { 4914 trunc = (sarg.vap->va_size == 0); 4915 status = rfs4_check_stateid(FWRITE, cs->vp, stateid, 4916 trunc, &cs->deleg, sarg.vap->va_mask & AT_SIZE, &ct); 4917 if (status != NFS4_OK) 4918 goto done; 4919 } else { 4920 ct.cc_sysid = 0; 4921 ct.cc_pid = 0; 4922 ct.cc_caller_id = nfs4_srv_caller_id; 4923 ct.cc_flags = CC_DONTBLOCK; 4924 } 4925 4926 /* XXX start of possible race with delegations */ 4927 4928 /* 4929 * We need to specially handle size changes because it is 4930 * possible for the client to create a file with read-only 4931 * modes, but with the file opened for writing. If the client 4932 * then tries to set the file size, e.g. ftruncate(3C), 4933 * fcntl(F_FREESP), the normal access checking done in 4934 * VOP_SETATTR would prevent the client from doing it even though 4935 * it should be allowed to do so. To get around this, we do the 4936 * access checking for ourselves and use VOP_SPACE which doesn't 4937 * do the access checking. 4938 * Also the client should not be allowed to change the file 4939 * size if there is a conflicting non-blocking mandatory lock in 4940 * the region of the change. 4941 */ 4942 if (vp->v_type == VREG && (sarg.vap->va_mask & AT_SIZE)) { 4943 u_offset_t offset; 4944 ssize_t length; 4945 4946 /* 4947 * ufs_setattr clears AT_SIZE from vap->va_mask, but 4948 * before returning, sarg.vap->va_mask is used to 4949 * generate the setattr reply bitmap. We also clear 4950 * AT_SIZE below before calling VOP_SPACE. For both 4951 * of these cases, the va_mask needs to be saved here 4952 * and restored after calling VOP_SETATTR. 4953 */ 4954 saved_mask = sarg.vap->va_mask; 4955 4956 /* 4957 * Check any possible conflict due to NBMAND locks. 4958 * Get into critical region before VOP_GETATTR, so the 4959 * size attribute is valid when checking conflicts. 4960 */ 4961 if (nbl_need_check(vp)) { 4962 nbl_start_crit(vp, RW_READER); 4963 in_crit = 1; 4964 } 4965 4966 bva.va_mask = AT_UID|AT_SIZE; 4967 if (error = VOP_GETATTR(vp, &bva, 0, cr, &ct)) { 4968 status = puterrno4(error); 4969 goto done; 4970 } 4971 4972 if (in_crit) { 4973 if (sarg.vap->va_size < bva.va_size) { 4974 offset = sarg.vap->va_size; 4975 length = bva.va_size - sarg.vap->va_size; 4976 } else { 4977 offset = bva.va_size; 4978 length = sarg.vap->va_size - bva.va_size; 4979 } 4980 if (nbl_conflict(vp, NBL_WRITE, offset, length, 0, 4981 &ct)) { 4982 status = NFS4ERR_LOCKED; 4983 goto done; 4984 } 4985 } 4986 4987 if (crgetuid(cr) == bva.va_uid) { 4988 sarg.vap->va_mask &= ~AT_SIZE; 4989 bf.l_type = F_WRLCK; 4990 bf.l_whence = 0; 4991 bf.l_start = (off64_t)sarg.vap->va_size; 4992 bf.l_len = 0; 4993 bf.l_sysid = 0; 4994 bf.l_pid = 0; 4995 error = VOP_SPACE(vp, F_FREESP, &bf, FWRITE, 4996 (offset_t)sarg.vap->va_size, cr, &ct); 4997 } 4998 } 4999 5000 if (!error && sarg.vap->va_mask != 0) 5001 error = VOP_SETATTR(vp, sarg.vap, sarg.flag, cr, &ct); 5002 5003 /* restore va_mask -- ufs_setattr clears AT_SIZE */ 5004 if (saved_mask & AT_SIZE) 5005 sarg.vap->va_mask |= AT_SIZE; 5006 5007 /* 5008 * If an ACL was being set, it has been delayed until now, 5009 * in order to set the mode (via the VOP_SETATTR() above) first. 5010 */ 5011 if ((! error) && (fattrp->attrmask & FATTR4_ACL_MASK)) { 5012 int i; 5013 5014 for (i = 0; i < NFS4_MAXNUM_ATTRS; i++) 5015 if (ntov.amap[i] == FATTR4_ACL) 5016 break; 5017 if (i < NFS4_MAXNUM_ATTRS) { 5018 error = (*nfs4_ntov_map[FATTR4_ACL].sv_getit)( 5019 NFS4ATTR_SETIT, &sarg, &ntov.na[i]); 5020 if (error == 0) { 5021 *resp |= FATTR4_ACL_MASK; 5022 } else if (error == ENOTSUP) { 5023 (void) rfs4_verify_attr(&sarg, resp, &ntov); 5024 status = NFS4ERR_ATTRNOTSUPP; 5025 goto done; 5026 } 5027 } else { 5028 NFS4_DEBUG(rfs4_debug, 5029 (CE_NOTE, "do_rfs4_op_setattr: " 5030 "unable to find ACL in fattr4")); 5031 error = EINVAL; 5032 } 5033 } 5034 5035 if (error) { 5036 /* check if a monitor detected a delegation conflict */ 5037 if (error == EAGAIN && (ct.cc_flags & CC_WOULDBLOCK)) 5038 status = NFS4ERR_DELAY; 5039 else 5040 status = puterrno4(error); 5041 5042 /* 5043 * Set the response bitmap when setattr failed. 5044 * If VOP_SETATTR partially succeeded, test by doing a 5045 * VOP_GETATTR on the object and comparing the data 5046 * to the setattr arguments. 5047 */ 5048 (void) rfs4_verify_attr(&sarg, resp, &ntov); 5049 } else { 5050 /* 5051 * Force modified metadata out to stable storage. 5052 */ 5053 (void) VOP_FSYNC(vp, FNODSYNC, cr, &ct); 5054 /* 5055 * Set response bitmap 5056 */ 5057 nfs4_vmask_to_nmask_set(sarg.vap->va_mask, resp); 5058 } 5059 5060 /* Return early and already have a NFSv4 error */ 5061 done: 5062 /* 5063 * Except for nfs4_vmask_to_nmask_set(), vattr --> fattr 5064 * conversion sets both readable and writeable NFS4 attrs 5065 * for AT_MTIME and AT_ATIME. The line below masks out 5066 * unrequested attrs from the setattr result bitmap. This 5067 * is placed after the done: label to catch the ATTRNOTSUP 5068 * case. 5069 */ 5070 *resp &= fattrp->attrmask; 5071 5072 if (in_crit) 5073 nbl_end_crit(vp); 5074 5075 nfs4_ntov_table_free(&ntov, &sarg); 5076 5077 return (status); 5078 } 5079 5080 /* ARGSUSED */ 5081 static void 5082 rfs4_op_setattr(nfs_argop4 *argop, nfs_resop4 *resop, struct svc_req *req, 5083 struct compound_state *cs) 5084 { 5085 SETATTR4args *args = &argop->nfs_argop4_u.opsetattr; 5086 SETATTR4res *resp = &resop->nfs_resop4_u.opsetattr; 5087 bslabel_t *clabel; 5088 5089 DTRACE_NFSV4_2(op__setattr__start, struct compound_state *, cs, 5090 SETATTR4args *, args); 5091 5092 if (cs->vp == NULL) { 5093 *cs->statusp = resp->status = NFS4ERR_NOFILEHANDLE; 5094 goto out; 5095 } 5096 5097 /* 5098 * If there is an unshared filesystem mounted on this vnode, 5099 * do not allow to setattr on this vnode. 5100 */ 5101 if (vn_ismntpt(cs->vp)) { 5102 *cs->statusp = resp->status = NFS4ERR_ACCESS; 5103 goto out; 5104 } 5105 5106 resp->attrsset = 0; 5107 5108 if (rdonly4(cs->exi, cs->vp, req)) { 5109 *cs->statusp = resp->status = NFS4ERR_ROFS; 5110 goto out; 5111 } 5112 5113 /* check label before setting attributes */ 5114 if (is_system_labeled()) { 5115 ASSERT(req->rq_label != NULL); 5116 clabel = req->rq_label; 5117 DTRACE_PROBE2(tx__rfs4__log__info__opsetattr__clabel, char *, 5118 "got client label from request(1)", 5119 struct svc_req *, req); 5120 if (!blequal(&l_admin_low->tsl_label, clabel)) { 5121 if (!do_rfs_label_check(clabel, cs->vp, 5122 EQUALITY_CHECK)) { 5123 *cs->statusp = resp->status = NFS4ERR_ACCESS; 5124 goto out; 5125 } 5126 } 5127 } 5128 5129 *cs->statusp = resp->status = 5130 do_rfs4_op_setattr(&resp->attrsset, &args->obj_attributes, cs, 5131 &args->stateid); 5132 5133 out: 5134 DTRACE_NFSV4_2(op__setattr__done, struct compound_state *, cs, 5135 SETATTR4res *, resp); 5136 } 5137 5138 /* ARGSUSED */ 5139 static void 5140 rfs4_op_verify(nfs_argop4 *argop, nfs_resop4 *resop, struct svc_req *req, 5141 struct compound_state *cs) 5142 { 5143 /* 5144 * verify and nverify are exactly the same, except that nverify 5145 * succeeds when some argument changed, and verify succeeds when 5146 * when none changed. 5147 */ 5148 5149 VERIFY4args *args = &argop->nfs_argop4_u.opverify; 5150 VERIFY4res *resp = &resop->nfs_resop4_u.opverify; 5151 5152 int error; 5153 struct nfs4_svgetit_arg sarg; 5154 struct statvfs64 sb; 5155 struct nfs4_ntov_table ntov; 5156 5157 DTRACE_NFSV4_2(op__verify__start, struct compound_state *, cs, 5158 VERIFY4args *, args); 5159 5160 if (cs->vp == NULL) { 5161 *cs->statusp = resp->status = NFS4ERR_NOFILEHANDLE; 5162 goto out; 5163 } 5164 5165 sarg.sbp = &sb; 5166 nfs4_ntov_table_init(&ntov); 5167 resp->status = do_rfs4_set_attrs(NULL, &args->obj_attributes, cs, 5168 &sarg, &ntov, NFS4ATTR_VERIT); 5169 if (resp->status != NFS4_OK) { 5170 /* 5171 * do_rfs4_set_attrs will try to verify systemwide attrs, 5172 * so could return -1 for "no match". 5173 */ 5174 if (resp->status == -1) 5175 resp->status = NFS4ERR_NOT_SAME; 5176 goto done; 5177 } 5178 error = rfs4_verify_attr(&sarg, NULL, &ntov); 5179 switch (error) { 5180 case 0: 5181 resp->status = NFS4_OK; 5182 break; 5183 case -1: 5184 resp->status = NFS4ERR_NOT_SAME; 5185 break; 5186 default: 5187 resp->status = puterrno4(error); 5188 break; 5189 } 5190 done: 5191 *cs->statusp = resp->status; 5192 nfs4_ntov_table_free(&ntov, &sarg); 5193 out: 5194 DTRACE_NFSV4_2(op__verify__done, struct compound_state *, cs, 5195 VERIFY4res *, resp); 5196 } 5197 5198 /* ARGSUSED */ 5199 static void 5200 rfs4_op_nverify(nfs_argop4 *argop, nfs_resop4 *resop, struct svc_req *req, 5201 struct compound_state *cs) 5202 { 5203 /* 5204 * verify and nverify are exactly the same, except that nverify 5205 * succeeds when some argument changed, and verify succeeds when 5206 * when none changed. 5207 */ 5208 5209 NVERIFY4args *args = &argop->nfs_argop4_u.opnverify; 5210 NVERIFY4res *resp = &resop->nfs_resop4_u.opnverify; 5211 5212 int error; 5213 struct nfs4_svgetit_arg sarg; 5214 struct statvfs64 sb; 5215 struct nfs4_ntov_table ntov; 5216 5217 DTRACE_NFSV4_2(op__nverify__start, struct compound_state *, cs, 5218 NVERIFY4args *, args); 5219 5220 if (cs->vp == NULL) { 5221 *cs->statusp = resp->status = NFS4ERR_NOFILEHANDLE; 5222 DTRACE_NFSV4_2(op__nverify__done, struct compound_state *, cs, 5223 NVERIFY4res *, resp); 5224 return; 5225 } 5226 sarg.sbp = &sb; 5227 nfs4_ntov_table_init(&ntov); 5228 resp->status = do_rfs4_set_attrs(NULL, &args->obj_attributes, cs, 5229 &sarg, &ntov, NFS4ATTR_VERIT); 5230 if (resp->status != NFS4_OK) { 5231 /* 5232 * do_rfs4_set_attrs will try to verify systemwide attrs, 5233 * so could return -1 for "no match". 5234 */ 5235 if (resp->status == -1) 5236 resp->status = NFS4_OK; 5237 goto done; 5238 } 5239 error = rfs4_verify_attr(&sarg, NULL, &ntov); 5240 switch (error) { 5241 case 0: 5242 resp->status = NFS4ERR_SAME; 5243 break; 5244 case -1: 5245 resp->status = NFS4_OK; 5246 break; 5247 default: 5248 resp->status = puterrno4(error); 5249 break; 5250 } 5251 done: 5252 *cs->statusp = resp->status; 5253 nfs4_ntov_table_free(&ntov, &sarg); 5254 5255 DTRACE_NFSV4_2(op__nverify__done, struct compound_state *, cs, 5256 NVERIFY4res *, resp); 5257 } 5258 5259 /* 5260 * XXX - This should live in an NFS header file. 5261 */ 5262 #define MAX_IOVECS 12 5263 5264 /* ARGSUSED */ 5265 static void 5266 rfs4_op_write(nfs_argop4 *argop, nfs_resop4 *resop, struct svc_req *req, 5267 struct compound_state *cs) 5268 { 5269 WRITE4args *args = &argop->nfs_argop4_u.opwrite; 5270 WRITE4res *resp = &resop->nfs_resop4_u.opwrite; 5271 int error; 5272 vnode_t *vp; 5273 struct vattr bva; 5274 u_offset_t rlimit; 5275 struct uio uio; 5276 struct iovec iov[MAX_IOVECS]; 5277 struct iovec *iovp; 5278 int iovcnt; 5279 int ioflag; 5280 cred_t *savecred, *cr; 5281 bool_t *deleg = &cs->deleg; 5282 nfsstat4 stat; 5283 int in_crit = 0; 5284 caller_context_t ct; 5285 5286 DTRACE_NFSV4_2(op__write__start, struct compound_state *, cs, 5287 WRITE4args *, args); 5288 5289 vp = cs->vp; 5290 if (vp == NULL) { 5291 *cs->statusp = resp->status = NFS4ERR_NOFILEHANDLE; 5292 goto out; 5293 } 5294 if (cs->access == CS_ACCESS_DENIED) { 5295 *cs->statusp = resp->status = NFS4ERR_ACCESS; 5296 goto out; 5297 } 5298 5299 cr = cs->cr; 5300 5301 if ((stat = rfs4_check_stateid(FWRITE, vp, &args->stateid, FALSE, 5302 deleg, TRUE, &ct)) != NFS4_OK) { 5303 *cs->statusp = resp->status = stat; 5304 goto out; 5305 } 5306 5307 /* 5308 * We have to enter the critical region before calling VOP_RWLOCK 5309 * to avoid a deadlock with ufs. 5310 */ 5311 if (nbl_need_check(vp)) { 5312 nbl_start_crit(vp, RW_READER); 5313 in_crit = 1; 5314 if (nbl_conflict(vp, NBL_WRITE, 5315 args->offset, args->data_len, 0, &ct)) { 5316 *cs->statusp = resp->status = NFS4ERR_LOCKED; 5317 goto out; 5318 } 5319 } 5320 5321 bva.va_mask = AT_MODE | AT_UID; 5322 error = VOP_GETATTR(vp, &bva, 0, cr, &ct); 5323 5324 /* 5325 * If we can't get the attributes, then we can't do the 5326 * right access checking. So, we'll fail the request. 5327 */ 5328 if (error) { 5329 *cs->statusp = resp->status = puterrno4(error); 5330 goto out; 5331 } 5332 5333 if (rdonly4(cs->exi, cs->vp, req)) { 5334 *cs->statusp = resp->status = NFS4ERR_ROFS; 5335 goto out; 5336 } 5337 5338 if (vp->v_type != VREG) { 5339 *cs->statusp = resp->status = 5340 ((vp->v_type == VDIR) ? NFS4ERR_ISDIR : NFS4ERR_INVAL); 5341 goto out; 5342 } 5343 5344 if (crgetuid(cr) != bva.va_uid && 5345 (error = VOP_ACCESS(vp, VWRITE, 0, cr, &ct))) { 5346 *cs->statusp = resp->status = puterrno4(error); 5347 goto out; 5348 } 5349 5350 if (MANDLOCK(vp, bva.va_mode)) { 5351 *cs->statusp = resp->status = NFS4ERR_ACCESS; 5352 goto out; 5353 } 5354 5355 if (args->data_len == 0) { 5356 *cs->statusp = resp->status = NFS4_OK; 5357 resp->count = 0; 5358 resp->committed = args->stable; 5359 resp->writeverf = Write4verf; 5360 goto out; 5361 } 5362 5363 if (args->mblk != NULL) { 5364 mblk_t *m; 5365 uint_t bytes, round_len; 5366 5367 iovcnt = 0; 5368 bytes = 0; 5369 round_len = roundup(args->data_len, BYTES_PER_XDR_UNIT); 5370 for (m = args->mblk; 5371 m != NULL && bytes < round_len; 5372 m = m->b_cont) { 5373 iovcnt++; 5374 bytes += MBLKL(m); 5375 } 5376 #ifdef DEBUG 5377 /* should have ended on an mblk boundary */ 5378 if (bytes != round_len) { 5379 printf("bytes=0x%x, round_len=0x%x, req len=0x%x\n", 5380 bytes, round_len, args->data_len); 5381 printf("args=%p, args->mblk=%p, m=%p", (void *)args, 5382 (void *)args->mblk, (void *)m); 5383 ASSERT(bytes == round_len); 5384 } 5385 #endif 5386 if (iovcnt <= MAX_IOVECS) { 5387 iovp = iov; 5388 } else { 5389 iovp = kmem_alloc(sizeof (*iovp) * iovcnt, KM_SLEEP); 5390 } 5391 mblk_to_iov(args->mblk, iovcnt, iovp); 5392 } else if (args->rlist != NULL) { 5393 iovcnt = 1; 5394 iovp = iov; 5395 iovp->iov_base = (char *)((args->rlist)->u.c_daddr3); 5396 iovp->iov_len = args->data_len; 5397 } else { 5398 iovcnt = 1; 5399 iovp = iov; 5400 iovp->iov_base = args->data_val; 5401 iovp->iov_len = args->data_len; 5402 } 5403 5404 uio.uio_iov = iovp; 5405 uio.uio_iovcnt = iovcnt; 5406 5407 uio.uio_segflg = UIO_SYSSPACE; 5408 uio.uio_extflg = UIO_COPY_DEFAULT; 5409 uio.uio_loffset = args->offset; 5410 uio.uio_resid = args->data_len; 5411 uio.uio_llimit = curproc->p_fsz_ctl; 5412 rlimit = uio.uio_llimit - args->offset; 5413 if (rlimit < (u_offset_t)uio.uio_resid) 5414 uio.uio_resid = (int)rlimit; 5415 5416 if (args->stable == UNSTABLE4) 5417 ioflag = 0; 5418 else if (args->stable == FILE_SYNC4) 5419 ioflag = FSYNC; 5420 else if (args->stable == DATA_SYNC4) 5421 ioflag = FDSYNC; 5422 else { 5423 if (iovp != iov) 5424 kmem_free(iovp, sizeof (*iovp) * iovcnt); 5425 *cs->statusp = resp->status = NFS4ERR_INVAL; 5426 goto out; 5427 } 5428 5429 /* 5430 * We're changing creds because VM may fault and we need 5431 * the cred of the current thread to be used if quota 5432 * checking is enabled. 5433 */ 5434 savecred = curthread->t_cred; 5435 curthread->t_cred = cr; 5436 error = do_io(FWRITE, vp, &uio, ioflag, cr, &ct); 5437 curthread->t_cred = savecred; 5438 5439 if (iovp != iov) 5440 kmem_free(iovp, sizeof (*iovp) * iovcnt); 5441 5442 if (error) { 5443 *cs->statusp = resp->status = puterrno4(error); 5444 goto out; 5445 } 5446 5447 *cs->statusp = resp->status = NFS4_OK; 5448 resp->count = args->data_len - uio.uio_resid; 5449 5450 if (ioflag == 0) 5451 resp->committed = UNSTABLE4; 5452 else 5453 resp->committed = FILE_SYNC4; 5454 5455 resp->writeverf = Write4verf; 5456 5457 out: 5458 if (in_crit) 5459 nbl_end_crit(vp); 5460 5461 DTRACE_NFSV4_2(op__write__done, struct compound_state *, cs, 5462 WRITE4res *, resp); 5463 } 5464 5465 5466 /* XXX put in a header file */ 5467 extern int sec_svc_getcred(struct svc_req *, cred_t *, caddr_t *, int *); 5468 5469 void 5470 rfs4_compound(COMPOUND4args *args, COMPOUND4res *resp, struct exportinfo *exi, 5471 struct svc_req *req, cred_t *cr, int *rv) 5472 { 5473 uint_t i; 5474 struct compound_state cs; 5475 5476 if (rv != NULL) 5477 *rv = 0; 5478 rfs4_init_compound_state(&cs); 5479 /* 5480 * Form a reply tag by copying over the reqeuest tag. 5481 */ 5482 resp->tag.utf8string_val = 5483 kmem_alloc(args->tag.utf8string_len, KM_SLEEP); 5484 resp->tag.utf8string_len = args->tag.utf8string_len; 5485 bcopy(args->tag.utf8string_val, resp->tag.utf8string_val, 5486 resp->tag.utf8string_len); 5487 5488 cs.statusp = &resp->status; 5489 cs.req = req; 5490 5491 /* 5492 * XXX for now, minorversion should be zero 5493 */ 5494 if (args->minorversion != NFS4_MINORVERSION) { 5495 DTRACE_NFSV4_2(compound__start, struct compound_state *, 5496 &cs, COMPOUND4args *, args); 5497 resp->array_len = 0; 5498 resp->array = NULL; 5499 resp->status = NFS4ERR_MINOR_VERS_MISMATCH; 5500 DTRACE_NFSV4_2(compound__done, struct compound_state *, 5501 &cs, COMPOUND4res *, resp); 5502 return; 5503 } 5504 5505 ASSERT(exi == NULL); 5506 ASSERT(cr == NULL); 5507 5508 cr = crget(); 5509 ASSERT(cr != NULL); 5510 5511 if (sec_svc_getcred(req, cr, &cs.principal, &cs.nfsflavor) == 0) { 5512 DTRACE_NFSV4_2(compound__start, struct compound_state *, 5513 &cs, COMPOUND4args *, args); 5514 crfree(cr); 5515 DTRACE_NFSV4_2(compound__done, struct compound_state *, 5516 &cs, COMPOUND4res *, resp); 5517 svcerr_badcred(req->rq_xprt); 5518 if (rv != NULL) 5519 *rv = 1; 5520 return; 5521 } 5522 resp->array_len = args->array_len; 5523 resp->array = kmem_zalloc(args->array_len * sizeof (nfs_resop4), 5524 KM_SLEEP); 5525 5526 cs.basecr = cr; 5527 5528 DTRACE_NFSV4_2(compound__start, struct compound_state *, &cs, 5529 COMPOUND4args *, args); 5530 5531 /* 5532 * For now, NFS4 compound processing must be protected by 5533 * exported_lock because it can access more than one exportinfo 5534 * per compound and share/unshare can now change multiple 5535 * exinfo structs. The NFS2/3 code only refs 1 exportinfo 5536 * per proc (excluding public exinfo), and exi_count design 5537 * is sufficient to protect concurrent execution of NFS2/3 5538 * ops along with unexport. This lock will be removed as 5539 * part of the NFSv4 phase 2 namespace redesign work. 5540 */ 5541 rw_enter(&exported_lock, RW_READER); 5542 5543 /* 5544 * If this is the first compound we've seen, we need to start all 5545 * new instances' grace periods. 5546 */ 5547 if (rfs4_seen_first_compound == 0) { 5548 rfs4_grace_start_new(); 5549 /* 5550 * This must be set after rfs4_grace_start_new(), otherwise 5551 * another thread could proceed past here before the former 5552 * is finished. 5553 */ 5554 rfs4_seen_first_compound = 1; 5555 } 5556 5557 for (i = 0; i < args->array_len && cs.cont; i++) { 5558 nfs_argop4 *argop; 5559 nfs_resop4 *resop; 5560 uint_t op; 5561 5562 argop = &args->array[i]; 5563 resop = &resp->array[i]; 5564 resop->resop = argop->argop; 5565 op = (uint_t)resop->resop; 5566 5567 if (op < rfsv4disp_cnt) { 5568 /* 5569 * Count the individual ops here; NULL and COMPOUND 5570 * are counted in common_dispatch() 5571 */ 5572 rfsproccnt_v4_ptr[op].value.ui64++; 5573 5574 NFS4_DEBUG(rfs4_debug > 1, 5575 (CE_NOTE, "Executing %s", rfs4_op_string[op])); 5576 (*rfsv4disptab[op].dis_proc)(argop, resop, req, &cs); 5577 NFS4_DEBUG(rfs4_debug > 1, (CE_NOTE, "%s returned %d", 5578 rfs4_op_string[op], *cs.statusp)); 5579 if (*cs.statusp != NFS4_OK) 5580 cs.cont = FALSE; 5581 } else { 5582 /* 5583 * This is effectively dead code since XDR code 5584 * will have already returned BADXDR if op doesn't 5585 * decode to legal value. This only done for a 5586 * day when XDR code doesn't verify v4 opcodes. 5587 */ 5588 op = OP_ILLEGAL; 5589 rfsproccnt_v4_ptr[OP_ILLEGAL_IDX].value.ui64++; 5590 5591 rfs4_op_illegal(argop, resop, req, &cs); 5592 cs.cont = FALSE; 5593 } 5594 5595 /* 5596 * If not at last op, and if we are to stop, then 5597 * compact the results array. 5598 */ 5599 if ((i + 1) < args->array_len && !cs.cont) { 5600 nfs_resop4 *new_res = kmem_alloc( 5601 (i+1) * sizeof (nfs_resop4), KM_SLEEP); 5602 bcopy(resp->array, 5603 new_res, (i+1) * sizeof (nfs_resop4)); 5604 kmem_free(resp->array, 5605 args->array_len * sizeof (nfs_resop4)); 5606 5607 resp->array_len = i + 1; 5608 resp->array = new_res; 5609 } 5610 } 5611 5612 rw_exit(&exported_lock); 5613 5614 DTRACE_NFSV4_2(compound__done, struct compound_state *, &cs, 5615 COMPOUND4res *, resp); 5616 5617 if (cs.vp) 5618 VN_RELE(cs.vp); 5619 if (cs.saved_vp) 5620 VN_RELE(cs.saved_vp); 5621 if (cs.saved_fh.nfs_fh4_val) 5622 kmem_free(cs.saved_fh.nfs_fh4_val, NFS4_FHSIZE); 5623 5624 if (cs.basecr) 5625 crfree(cs.basecr); 5626 if (cs.cr) 5627 crfree(cs.cr); 5628 /* 5629 * done with this compound request, free the label 5630 */ 5631 5632 if (req->rq_label != NULL) { 5633 kmem_free(req->rq_label, sizeof (bslabel_t)); 5634 req->rq_label = NULL; 5635 } 5636 } 5637 5638 /* 5639 * XXX because of what appears to be duplicate calls to rfs4_compound_free 5640 * XXX zero out the tag and array values. Need to investigate why the 5641 * XXX calls occur, but at least prevent the panic for now. 5642 */ 5643 void 5644 rfs4_compound_free(COMPOUND4res *resp) 5645 { 5646 uint_t i; 5647 5648 if (resp->tag.utf8string_val) { 5649 UTF8STRING_FREE(resp->tag) 5650 } 5651 5652 for (i = 0; i < resp->array_len; i++) { 5653 nfs_resop4 *resop; 5654 uint_t op; 5655 5656 resop = &resp->array[i]; 5657 op = (uint_t)resop->resop; 5658 if (op < rfsv4disp_cnt) { 5659 (*rfsv4disptab[op].dis_resfree)(resop); 5660 } 5661 } 5662 if (resp->array != NULL) { 5663 kmem_free(resp->array, resp->array_len * sizeof (nfs_resop4)); 5664 } 5665 } 5666 5667 /* 5668 * Process the value of the compound request rpc flags, as a bit-AND 5669 * of the individual per-op flags (idempotent, allowork, publicfh_ok) 5670 */ 5671 void 5672 rfs4_compound_flagproc(COMPOUND4args *args, int *flagp) 5673 { 5674 int i; 5675 int flag = RPC_ALL; 5676 5677 for (i = 0; flag && i < args->array_len; i++) { 5678 uint_t op; 5679 5680 op = (uint_t)args->array[i].argop; 5681 5682 if (op < rfsv4disp_cnt) 5683 flag &= rfsv4disptab[op].dis_flags; 5684 else 5685 flag = 0; 5686 } 5687 *flagp = flag; 5688 } 5689 5690 nfsstat4 5691 rfs4_client_sysid(rfs4_client_t *cp, sysid_t *sp) 5692 { 5693 nfsstat4 e; 5694 5695 rfs4_dbe_lock(cp->dbe); 5696 5697 if (cp->sysidt != LM_NOSYSID) { 5698 *sp = cp->sysidt; 5699 e = NFS4_OK; 5700 5701 } else if ((cp->sysidt = lm_alloc_sysidt()) != LM_NOSYSID) { 5702 *sp = cp->sysidt; 5703 e = NFS4_OK; 5704 5705 NFS4_DEBUG(rfs4_debug, (CE_NOTE, 5706 "rfs4_client_sysid: allocated 0x%x\n", *sp)); 5707 } else 5708 e = NFS4ERR_DELAY; 5709 5710 rfs4_dbe_unlock(cp->dbe); 5711 return (e); 5712 } 5713 5714 #if defined(DEBUG) && ! defined(lint) 5715 static void lock_print(char *str, int operation, struct flock64 *flk) 5716 { 5717 char *op, *type; 5718 5719 switch (operation) { 5720 case F_GETLK: op = "F_GETLK"; 5721 break; 5722 case F_SETLK: op = "F_SETLK"; 5723 break; 5724 case F_SETLK_NBMAND: op = "F_SETLK_NBMAND"; 5725 break; 5726 default: op = "F_UNKNOWN"; 5727 break; 5728 } 5729 switch (flk->l_type) { 5730 case F_UNLCK: type = "F_UNLCK"; 5731 break; 5732 case F_RDLCK: type = "F_RDLCK"; 5733 break; 5734 case F_WRLCK: type = "F_WRLCK"; 5735 break; 5736 default: type = "F_UNKNOWN"; 5737 break; 5738 } 5739 5740 ASSERT(flk->l_whence == 0); 5741 cmn_err(CE_NOTE, "%s: %s, type = %s, off = %llx len = %llx pid = %d", 5742 str, op, type, (longlong_t)flk->l_start, 5743 flk->l_len ? (longlong_t)flk->l_len : ~0LL, flk->l_pid); 5744 } 5745 5746 #define LOCK_PRINT(d, s, t, f) if (d) lock_print(s, t, f) 5747 #else 5748 #define LOCK_PRINT(d, s, t, f) 5749 #endif 5750 5751 /*ARGSUSED*/ 5752 static bool_t 5753 creds_ok(cred_set_t cr_set, struct svc_req *req, struct compound_state *cs) 5754 { 5755 return (TRUE); 5756 } 5757 5758 /* 5759 * Look up the pathname using the vp in cs as the directory vnode. 5760 * cs->vp will be the vnode for the file on success 5761 */ 5762 5763 static nfsstat4 5764 rfs4_lookup(component4 *component, struct svc_req *req, 5765 struct compound_state *cs) 5766 { 5767 char *nm; 5768 uint32_t len; 5769 nfsstat4 status; 5770 5771 if (cs->vp == NULL) { 5772 return (NFS4ERR_NOFILEHANDLE); 5773 } 5774 if (cs->vp->v_type != VDIR) { 5775 return (NFS4ERR_NOTDIR); 5776 } 5777 5778 if (!utf8_dir_verify(component)) 5779 return (NFS4ERR_INVAL); 5780 5781 nm = utf8_to_fn(component, &len, NULL); 5782 if (nm == NULL) { 5783 return (NFS4ERR_INVAL); 5784 } 5785 5786 if (len > MAXNAMELEN) { 5787 kmem_free(nm, len); 5788 return (NFS4ERR_NAMETOOLONG); 5789 } 5790 5791 status = do_rfs4_op_lookup(nm, len, req, cs); 5792 5793 kmem_free(nm, len); 5794 5795 return (status); 5796 } 5797 5798 static nfsstat4 5799 rfs4_lookupfile(component4 *component, struct svc_req *req, 5800 struct compound_state *cs, uint32_t access, 5801 change_info4 *cinfo) 5802 { 5803 nfsstat4 status; 5804 vnode_t *dvp = cs->vp; 5805 vattr_t bva, ava, fva; 5806 int error; 5807 5808 /* Get "before" change value */ 5809 bva.va_mask = AT_CTIME|AT_SEQ; 5810 error = VOP_GETATTR(dvp, &bva, 0, cs->cr, NULL); 5811 if (error) 5812 return (puterrno4(error)); 5813 5814 /* rfs4_lookup may VN_RELE directory */ 5815 VN_HOLD(dvp); 5816 5817 status = rfs4_lookup(component, req, cs); 5818 if (status != NFS4_OK) { 5819 VN_RELE(dvp); 5820 return (status); 5821 } 5822 5823 /* 5824 * Get "after" change value, if it fails, simply return the 5825 * before value. 5826 */ 5827 ava.va_mask = AT_CTIME|AT_SEQ; 5828 if (VOP_GETATTR(dvp, &ava, 0, cs->cr, NULL)) { 5829 ava.va_ctime = bva.va_ctime; 5830 ava.va_seq = 0; 5831 } 5832 VN_RELE(dvp); 5833 5834 /* 5835 * Validate the file is a file 5836 */ 5837 fva.va_mask = AT_TYPE|AT_MODE; 5838 error = VOP_GETATTR(cs->vp, &fva, 0, cs->cr, NULL); 5839 if (error) 5840 return (puterrno4(error)); 5841 5842 if (fva.va_type != VREG) { 5843 if (fva.va_type == VDIR) 5844 return (NFS4ERR_ISDIR); 5845 if (fva.va_type == VLNK) 5846 return (NFS4ERR_SYMLINK); 5847 return (NFS4ERR_INVAL); 5848 } 5849 5850 NFS4_SET_FATTR4_CHANGE(cinfo->before, bva.va_ctime); 5851 NFS4_SET_FATTR4_CHANGE(cinfo->after, ava.va_ctime); 5852 5853 /* 5854 * It is undefined if VOP_LOOKUP will change va_seq, so 5855 * cinfo.atomic = TRUE only if we have 5856 * non-zero va_seq's, and they have not changed. 5857 */ 5858 if (bva.va_seq && ava.va_seq && ava.va_seq == bva.va_seq) 5859 cinfo->atomic = TRUE; 5860 else 5861 cinfo->atomic = FALSE; 5862 5863 /* Check for mandatory locking */ 5864 cs->mandlock = MANDLOCK(cs->vp, fva.va_mode); 5865 return (check_open_access(access, cs, req)); 5866 } 5867 5868 static nfsstat4 5869 create_vnode(vnode_t *dvp, char *nm, vattr_t *vap, createmode4 mode, 5870 timespec32_t *mtime, cred_t *cr, vnode_t **vpp, bool_t *created) 5871 { 5872 int error; 5873 nfsstat4 status = NFS4_OK; 5874 vattr_t va; 5875 5876 tryagain: 5877 5878 /* 5879 * The file open mode used is VWRITE. If the client needs 5880 * some other semantic, then it should do the access checking 5881 * itself. It would have been nice to have the file open mode 5882 * passed as part of the arguments. 5883 */ 5884 5885 *created = TRUE; 5886 error = VOP_CREATE(dvp, nm, vap, EXCL, VWRITE, vpp, cr, 0, NULL, NULL); 5887 5888 if (error) { 5889 *created = FALSE; 5890 5891 /* 5892 * If we got something other than file already exists 5893 * then just return this error. Otherwise, we got 5894 * EEXIST. If we were doing a GUARDED create, then 5895 * just return this error. Otherwise, we need to 5896 * make sure that this wasn't a duplicate of an 5897 * exclusive create request. 5898 * 5899 * The assumption is made that a non-exclusive create 5900 * request will never return EEXIST. 5901 */ 5902 5903 if (error != EEXIST || mode == GUARDED4) { 5904 status = puterrno4(error); 5905 return (status); 5906 } 5907 error = VOP_LOOKUP(dvp, nm, vpp, NULL, 0, NULL, cr, 5908 NULL, NULL, NULL); 5909 5910 if (error) { 5911 /* 5912 * We couldn't find the file that we thought that 5913 * we just created. So, we'll just try creating 5914 * it again. 5915 */ 5916 if (error == ENOENT) 5917 goto tryagain; 5918 5919 status = puterrno4(error); 5920 return (status); 5921 } 5922 5923 if (mode == UNCHECKED4) { 5924 /* existing object must be regular file */ 5925 if ((*vpp)->v_type != VREG) { 5926 if ((*vpp)->v_type == VDIR) 5927 status = NFS4ERR_ISDIR; 5928 else if ((*vpp)->v_type == VLNK) 5929 status = NFS4ERR_SYMLINK; 5930 else 5931 status = NFS4ERR_INVAL; 5932 VN_RELE(*vpp); 5933 return (status); 5934 } 5935 5936 return (NFS4_OK); 5937 } 5938 5939 /* Check for duplicate request */ 5940 ASSERT(mtime != 0); 5941 va.va_mask = AT_MTIME; 5942 error = VOP_GETATTR(*vpp, &va, 0, cr, NULL); 5943 if (!error) { 5944 /* We found the file */ 5945 if (va.va_mtime.tv_sec != mtime->tv_sec || 5946 va.va_mtime.tv_nsec != mtime->tv_nsec) { 5947 /* but its not our creation */ 5948 VN_RELE(*vpp); 5949 return (NFS4ERR_EXIST); 5950 } 5951 *created = TRUE; /* retrans of create == created */ 5952 return (NFS4_OK); 5953 } 5954 VN_RELE(*vpp); 5955 return (NFS4ERR_EXIST); 5956 } 5957 5958 return (NFS4_OK); 5959 } 5960 5961 static nfsstat4 5962 check_open_access(uint32_t access, 5963 struct compound_state *cs, struct svc_req *req) 5964 { 5965 int error; 5966 vnode_t *vp; 5967 bool_t readonly; 5968 cred_t *cr = cs->cr; 5969 5970 /* For now we don't allow mandatory locking as per V2/V3 */ 5971 if (cs->access == CS_ACCESS_DENIED || cs->mandlock) { 5972 return (NFS4ERR_ACCESS); 5973 } 5974 5975 vp = cs->vp; 5976 ASSERT(cr != NULL && vp->v_type == VREG); 5977 5978 /* 5979 * If the file system is exported read only and we are trying 5980 * to open for write, then return NFS4ERR_ROFS 5981 */ 5982 5983 readonly = rdonly4(cs->exi, cs->vp, req); 5984 5985 if ((access & OPEN4_SHARE_ACCESS_WRITE) && readonly) 5986 return (NFS4ERR_ROFS); 5987 5988 if (access & OPEN4_SHARE_ACCESS_READ) { 5989 if ((VOP_ACCESS(vp, VREAD, 0, cr, NULL) != 0) && 5990 (VOP_ACCESS(vp, VEXEC, 0, cr, NULL) != 0)) { 5991 return (NFS4ERR_ACCESS); 5992 } 5993 } 5994 5995 if (access & OPEN4_SHARE_ACCESS_WRITE) { 5996 error = VOP_ACCESS(vp, VWRITE, 0, cr, NULL); 5997 if (error) 5998 return (NFS4ERR_ACCESS); 5999 } 6000 6001 return (NFS4_OK); 6002 } 6003 6004 static nfsstat4 6005 rfs4_createfile(OPEN4args *args, struct svc_req *req, struct compound_state *cs, 6006 change_info4 *cinfo, bitmap4 *attrset, clientid4 clientid) 6007 { 6008 struct nfs4_svgetit_arg sarg; 6009 struct nfs4_ntov_table ntov; 6010 6011 bool_t ntov_table_init = FALSE; 6012 struct statvfs64 sb; 6013 nfsstat4 status; 6014 vnode_t *vp; 6015 vattr_t bva, ava, iva, cva, *vap; 6016 vnode_t *dvp; 6017 timespec32_t *mtime; 6018 char *nm = NULL; 6019 uint_t buflen; 6020 bool_t created; 6021 bool_t setsize = FALSE; 6022 len_t reqsize; 6023 int error; 6024 bool_t trunc; 6025 caller_context_t ct; 6026 component4 *component; 6027 bslabel_t *clabel; 6028 6029 sarg.sbp = &sb; 6030 6031 dvp = cs->vp; 6032 6033 /* Check if the file system is read only */ 6034 if (rdonly4(cs->exi, dvp, req)) 6035 return (NFS4ERR_ROFS); 6036 6037 /* check the label of including directory */ 6038 if (is_system_labeled()) { 6039 ASSERT(req->rq_label != NULL); 6040 clabel = req->rq_label; 6041 DTRACE_PROBE2(tx__rfs4__log__info__opremove__clabel, char *, 6042 "got client label from request(1)", 6043 struct svc_req *, req); 6044 if (!blequal(&l_admin_low->tsl_label, clabel)) { 6045 if (!do_rfs_label_check(clabel, dvp, EQUALITY_CHECK)) { 6046 return (NFS4ERR_ACCESS); 6047 } 6048 } 6049 } 6050 6051 /* 6052 * Get the last component of path name in nm. cs will reference 6053 * the including directory on success. 6054 */ 6055 component = &args->open_claim4_u.file; 6056 if (!utf8_dir_verify(component)) 6057 return (NFS4ERR_INVAL); 6058 6059 nm = utf8_to_fn(component, &buflen, NULL); 6060 6061 if (nm == NULL) 6062 return (NFS4ERR_RESOURCE); 6063 6064 if (buflen > MAXNAMELEN) { 6065 kmem_free(nm, buflen); 6066 return (NFS4ERR_NAMETOOLONG); 6067 } 6068 6069 bva.va_mask = AT_TYPE|AT_CTIME|AT_SEQ; 6070 error = VOP_GETATTR(dvp, &bva, 0, cs->cr, NULL); 6071 if (error) { 6072 kmem_free(nm, buflen); 6073 return (puterrno4(error)); 6074 } 6075 6076 if (bva.va_type != VDIR) { 6077 kmem_free(nm, buflen); 6078 return (NFS4ERR_NOTDIR); 6079 } 6080 6081 NFS4_SET_FATTR4_CHANGE(cinfo->before, bva.va_ctime) 6082 6083 switch (args->mode) { 6084 case GUARDED4: 6085 /*FALLTHROUGH*/ 6086 case UNCHECKED4: 6087 nfs4_ntov_table_init(&ntov); 6088 ntov_table_init = TRUE; 6089 6090 *attrset = 0; 6091 status = do_rfs4_set_attrs(attrset, 6092 &args->createhow4_u.createattrs, 6093 cs, &sarg, &ntov, NFS4ATTR_SETIT); 6094 6095 if (status == NFS4_OK && (sarg.vap->va_mask & AT_TYPE) && 6096 sarg.vap->va_type != VREG) { 6097 if (sarg.vap->va_type == VDIR) 6098 status = NFS4ERR_ISDIR; 6099 else if (sarg.vap->va_type == VLNK) 6100 status = NFS4ERR_SYMLINK; 6101 else 6102 status = NFS4ERR_INVAL; 6103 } 6104 6105 if (status != NFS4_OK) { 6106 kmem_free(nm, buflen); 6107 nfs4_ntov_table_free(&ntov, &sarg); 6108 *attrset = 0; 6109 return (status); 6110 } 6111 6112 vap = sarg.vap; 6113 vap->va_type = VREG; 6114 vap->va_mask |= AT_TYPE; 6115 6116 if ((vap->va_mask & AT_MODE) == 0) { 6117 vap->va_mask |= AT_MODE; 6118 vap->va_mode = (mode_t)0600; 6119 } 6120 6121 if (vap->va_mask & AT_SIZE) { 6122 6123 /* Disallow create with a non-zero size */ 6124 6125 if ((reqsize = sarg.vap->va_size) != 0) { 6126 kmem_free(nm, buflen); 6127 nfs4_ntov_table_free(&ntov, &sarg); 6128 *attrset = 0; 6129 return (NFS4ERR_INVAL); 6130 } 6131 setsize = TRUE; 6132 } 6133 break; 6134 6135 case EXCLUSIVE4: 6136 /* prohibit EXCL create of named attributes */ 6137 if (dvp->v_flag & V_XATTRDIR) { 6138 kmem_free(nm, buflen); 6139 *attrset = 0; 6140 return (NFS4ERR_INVAL); 6141 } 6142 6143 cva.va_mask = AT_TYPE | AT_MTIME | AT_MODE; 6144 cva.va_type = VREG; 6145 /* 6146 * Ensure no time overflows. Assumes underlying 6147 * filesystem supports at least 32 bits. 6148 * Truncate nsec to usec resolution to allow valid 6149 * compares even if the underlying filesystem truncates. 6150 */ 6151 mtime = (timespec32_t *)&args->createhow4_u.createverf; 6152 cva.va_mtime.tv_sec = mtime->tv_sec % TIME32_MAX; 6153 cva.va_mtime.tv_nsec = (mtime->tv_nsec / 1000) * 1000; 6154 cva.va_mode = (mode_t)0; 6155 vap = &cva; 6156 break; 6157 } 6158 6159 status = create_vnode(dvp, nm, vap, args->mode, mtime, 6160 cs->cr, &vp, &created); 6161 kmem_free(nm, buflen); 6162 6163 if (status != NFS4_OK) { 6164 if (ntov_table_init) 6165 nfs4_ntov_table_free(&ntov, &sarg); 6166 *attrset = 0; 6167 return (status); 6168 } 6169 6170 trunc = (setsize && !created); 6171 6172 if (args->mode != EXCLUSIVE4) { 6173 bitmap4 createmask = args->createhow4_u.createattrs.attrmask; 6174 6175 /* 6176 * True verification that object was created with correct 6177 * attrs is impossible. The attrs could have been changed 6178 * immediately after object creation. If attributes did 6179 * not verify, the only recourse for the server is to 6180 * destroy the object. Maybe if some attrs (like gid) 6181 * are set incorrectly, the object should be destroyed; 6182 * however, seems bad as a default policy. Do we really 6183 * want to destroy an object over one of the times not 6184 * verifying correctly? For these reasons, the server 6185 * currently sets bits in attrset for createattrs 6186 * that were set; however, no verification is done. 6187 * 6188 * vmask_to_nmask accounts for vattr bits set on create 6189 * [do_rfs4_set_attrs() only sets resp bits for 6190 * non-vattr/vfs bits.] 6191 * Mask off any bits we set by default so as not to return 6192 * more attrset bits than were requested in createattrs 6193 */ 6194 if (created) { 6195 nfs4_vmask_to_nmask(sarg.vap->va_mask, attrset); 6196 *attrset &= createmask; 6197 } else { 6198 /* 6199 * We did not create the vnode (we tried but it 6200 * already existed). In this case, the only createattr 6201 * that the spec allows the server to set is size, 6202 * and even then, it can only be set if it is 0. 6203 */ 6204 *attrset = 0; 6205 if (trunc) 6206 *attrset = FATTR4_SIZE_MASK; 6207 } 6208 } 6209 if (ntov_table_init) 6210 nfs4_ntov_table_free(&ntov, &sarg); 6211 6212 /* 6213 * Get the initial "after" sequence number, if it fails, 6214 * set to zero, time to before. 6215 */ 6216 iva.va_mask = AT_CTIME|AT_SEQ; 6217 if (VOP_GETATTR(dvp, &iva, 0, cs->cr, NULL)) { 6218 iva.va_seq = 0; 6219 iva.va_ctime = bva.va_ctime; 6220 } 6221 6222 /* 6223 * create_vnode attempts to create the file exclusive, 6224 * if it already exists the VOP_CREATE will fail and 6225 * may not increase va_seq. It is atomic if 6226 * we haven't changed the directory, but if it has changed 6227 * we don't know what changed it. 6228 */ 6229 if (!created) { 6230 if (bva.va_seq && iva.va_seq && 6231 bva.va_seq == iva.va_seq) 6232 cinfo->atomic = TRUE; 6233 else 6234 cinfo->atomic = FALSE; 6235 NFS4_SET_FATTR4_CHANGE(cinfo->after, iva.va_ctime); 6236 } else { 6237 /* 6238 * The entry was created, we need to sync the 6239 * directory metadata. 6240 */ 6241 (void) VOP_FSYNC(dvp, 0, cs->cr, NULL); 6242 6243 /* 6244 * Get "after" change value, if it fails, simply return the 6245 * before value. 6246 */ 6247 ava.va_mask = AT_CTIME|AT_SEQ; 6248 if (VOP_GETATTR(dvp, &ava, 0, cs->cr, NULL)) { 6249 ava.va_ctime = bva.va_ctime; 6250 ava.va_seq = 0; 6251 } 6252 6253 NFS4_SET_FATTR4_CHANGE(cinfo->after, ava.va_ctime); 6254 6255 /* 6256 * The cinfo->atomic = TRUE only if we have 6257 * non-zero va_seq's, and it has incremented by exactly one 6258 * during the create_vnode and it didn't 6259 * change during the VOP_FSYNC. 6260 */ 6261 if (bva.va_seq && iva.va_seq && ava.va_seq && 6262 iva.va_seq == (bva.va_seq + 1) && iva.va_seq == ava.va_seq) 6263 cinfo->atomic = TRUE; 6264 else 6265 cinfo->atomic = FALSE; 6266 } 6267 6268 /* Check for mandatory locking and that the size gets set. */ 6269 cva.va_mask = AT_MODE; 6270 if (setsize) 6271 cva.va_mask |= AT_SIZE; 6272 6273 /* Assume the worst */ 6274 cs->mandlock = TRUE; 6275 6276 if (VOP_GETATTR(vp, &cva, 0, cs->cr, NULL) == 0) { 6277 cs->mandlock = MANDLOCK(cs->vp, cva.va_mode); 6278 6279 /* 6280 * Truncate the file if necessary; this would be 6281 * the case for create over an existing file. 6282 */ 6283 6284 if (trunc) { 6285 int in_crit = 0; 6286 rfs4_file_t *fp; 6287 bool_t create = FALSE; 6288 6289 /* 6290 * We are writing over an existing file. 6291 * Check to see if we need to recall a delegation. 6292 */ 6293 rfs4_hold_deleg_policy(); 6294 if ((fp = rfs4_findfile(vp, NULL, &create)) != NULL) { 6295 if (rfs4_check_delegated_byfp(FWRITE, fp, 6296 (reqsize == 0), FALSE, FALSE, &clientid)) { 6297 rfs4_file_rele(fp); 6298 rfs4_rele_deleg_policy(); 6299 VN_RELE(vp); 6300 *attrset = 0; 6301 return (NFS4ERR_DELAY); 6302 } 6303 rfs4_file_rele(fp); 6304 } 6305 rfs4_rele_deleg_policy(); 6306 6307 if (nbl_need_check(vp)) { 6308 in_crit = 1; 6309 6310 ASSERT(reqsize == 0); 6311 6312 nbl_start_crit(vp, RW_READER); 6313 if (nbl_conflict(vp, NBL_WRITE, 0, 6314 cva.va_size, 0, NULL)) { 6315 in_crit = 0; 6316 nbl_end_crit(vp); 6317 VN_RELE(vp); 6318 *attrset = 0; 6319 return (NFS4ERR_ACCESS); 6320 } 6321 } 6322 ct.cc_sysid = 0; 6323 ct.cc_pid = 0; 6324 ct.cc_caller_id = nfs4_srv_caller_id; 6325 ct.cc_flags = CC_DONTBLOCK; 6326 6327 cva.va_mask = AT_SIZE; 6328 cva.va_size = reqsize; 6329 (void) VOP_SETATTR(vp, &cva, 0, cs->cr, &ct); 6330 if (in_crit) 6331 nbl_end_crit(vp); 6332 } 6333 } 6334 6335 error = makefh4(&cs->fh, vp, cs->exi); 6336 6337 /* 6338 * Force modified data and metadata out to stable storage. 6339 */ 6340 (void) VOP_FSYNC(vp, FNODSYNC, cs->cr, NULL); 6341 6342 if (error) { 6343 VN_RELE(vp); 6344 *attrset = 0; 6345 return (puterrno4(error)); 6346 } 6347 6348 /* if parent dir is attrdir, set namedattr fh flag */ 6349 if (dvp->v_flag & V_XATTRDIR) 6350 set_fh4_flag(&cs->fh, FH4_NAMEDATTR); 6351 6352 if (cs->vp) 6353 VN_RELE(cs->vp); 6354 6355 cs->vp = vp; 6356 6357 /* 6358 * if we did not create the file, we will need to check 6359 * the access bits on the file 6360 */ 6361 6362 if (!created) { 6363 if (setsize) 6364 args->share_access |= OPEN4_SHARE_ACCESS_WRITE; 6365 status = check_open_access(args->share_access, cs, req); 6366 if (status != NFS4_OK) 6367 *attrset = 0; 6368 } 6369 return (status); 6370 } 6371 6372 /*ARGSUSED*/ 6373 static void 6374 rfs4_do_open(struct compound_state *cs, struct svc_req *req, 6375 rfs4_openowner_t *oo, delegreq_t deleg, 6376 uint32_t access, uint32_t deny, 6377 OPEN4res *resp, int deleg_cur) 6378 { 6379 /* XXX Currently not using req */ 6380 rfs4_state_t *state; 6381 rfs4_file_t *file; 6382 bool_t screate = TRUE; 6383 bool_t fcreate = TRUE; 6384 uint32_t amodes; 6385 uint32_t dmodes; 6386 rfs4_deleg_state_t *dsp; 6387 struct shrlock shr; 6388 struct shr_locowner shr_loco; 6389 sysid_t sysid; 6390 nfsstat4 status; 6391 caller_context_t ct; 6392 int fflags = 0; 6393 int recall = 0; 6394 int err; 6395 int cmd; 6396 6397 /* get the file struct and hold a lock on it during initial open */ 6398 file = rfs4_findfile_withlock(cs->vp, &cs->fh, &fcreate); 6399 if (file == NULL) { 6400 NFS4_DEBUG(rfs4_debug, 6401 (CE_NOTE, "rfs4_do_open: can't find file")); 6402 resp->status = NFS4ERR_SERVERFAULT; 6403 return; 6404 } 6405 6406 state = rfs4_findstate_by_owner_file(oo, file, &screate); 6407 if (state == NULL) { 6408 NFS4_DEBUG(rfs4_debug, 6409 (CE_NOTE, "rfs4_do_open: can't find state")); 6410 resp->status = NFS4ERR_RESOURCE; 6411 /* No need to keep any reference */ 6412 rfs4_file_rele_withunlock(file); 6413 return; 6414 } 6415 6416 /* try to get the sysid before continuing */ 6417 if ((status = rfs4_client_sysid(oo->client, &sysid)) != NFS4_OK) { 6418 resp->status = status; 6419 rfs4_file_rele(file); 6420 /* Not a fully formed open; "close" it */ 6421 if (screate == TRUE) 6422 rfs4_state_close(state, FALSE, FALSE, cs->cr); 6423 rfs4_state_rele(state); 6424 return; 6425 } 6426 6427 /* Calculate the fflags for this OPEN. */ 6428 if (access & OPEN4_SHARE_ACCESS_READ) 6429 fflags |= FREAD; 6430 if (access & OPEN4_SHARE_ACCESS_WRITE) 6431 fflags |= FWRITE; 6432 6433 /* 6434 * Calculate the new deny and access mode that this open is adding to 6435 * the file for this open owner; 6436 */ 6437 dmodes = (deny & ~state->share_deny); 6438 amodes = (access & ~state->share_access); 6439 6440 /* 6441 * Check to see the client has already sent an open for this 6442 * open owner on this file with the same share/deny modes. 6443 * If so, we don't need to check for a conflict and we don't 6444 * need to add another shrlock. If not, then we need to 6445 * check for conflicts in deny and access before checking for 6446 * conflicts in delegation. We don't want to recall a 6447 * delegation based on an open that will eventually fail based 6448 * on shares modes. 6449 */ 6450 6451 if (dmodes || amodes) { 6452 shr.s_access = (short)access; 6453 shr.s_deny = (short)deny; 6454 shr.s_pid = rfs4_dbe_getid(oo->dbe); 6455 shr.s_sysid = sysid; 6456 shr_loco.sl_pid = shr.s_pid; 6457 shr_loco.sl_id = shr.s_sysid; 6458 shr.s_owner = (caddr_t)&shr_loco; 6459 shr.s_own_len = sizeof (shr_loco); 6460 6461 cmd = nbl_need_check(cs->vp) ? F_SHARE_NBMAND : F_SHARE; 6462 if ((err = vop_shrlock(cs->vp, cmd, &shr, fflags)) != 0) { 6463 6464 resp->status = err == EAGAIN ? 6465 NFS4ERR_SHARE_DENIED : puterrno4(err); 6466 6467 rfs4_file_rele(file); 6468 /* Not a fully formed open; "close" it */ 6469 if (screate == TRUE) 6470 rfs4_state_close(state, FALSE, FALSE, cs->cr); 6471 rfs4_state_rele(state); 6472 return; 6473 } 6474 } 6475 6476 rfs4_dbe_lock(state->dbe); 6477 rfs4_dbe_lock(file->dbe); 6478 6479 /* 6480 * Check to see if this file is delegated and if so, if a 6481 * recall needs to be done. 6482 */ 6483 if (rfs4_check_recall(state, access)) { 6484 rfs4_dbe_unlock(file->dbe); 6485 rfs4_dbe_unlock(state->dbe); 6486 rfs4_recall_deleg(file, FALSE, state->owner->client); 6487 delay(NFS4_DELEGATION_CONFLICT_DELAY); 6488 rfs4_dbe_lock(state->dbe); 6489 rfs4_dbe_lock(file->dbe); 6490 /* Let's see if the delegation was returned */ 6491 if (rfs4_check_recall(state, access)) { 6492 rfs4_dbe_unlock(file->dbe); 6493 rfs4_dbe_unlock(state->dbe); 6494 rfs4_file_rele(file); 6495 rfs4_update_lease(state->owner->client); 6496 (void) vop_shrlock(cs->vp, F_UNSHARE, &shr, fflags); 6497 /* Not a fully formed open; "close" it */ 6498 if (screate == TRUE) 6499 rfs4_state_close(state, FALSE, FALSE, cs->cr); 6500 rfs4_state_rele(state); 6501 resp->status = NFS4ERR_DELAY; 6502 return; 6503 } 6504 } 6505 /* 6506 * the share check passed and any delegation conflict has been 6507 * taken care of, now call vop_open. 6508 * if this is the first open then call vop_open with fflags. 6509 * if not, call vn_open_upgrade with just the upgrade flags. 6510 * 6511 * if the file has been opened already, it will have the current 6512 * access mode in the state struct. if it has no share access, then 6513 * this is a new open. 6514 * 6515 * However, if this is open with CLAIM_DLEGATE_CUR, then don't 6516 * call VOP_OPEN(), just do the open upgrade. 6517 */ 6518 if (((state->share_access & OPEN4_SHARE_ACCESS_BOTH) == 0) && 6519 !deleg_cur) { 6520 ct.cc_sysid = sysid; 6521 ct.cc_pid = shr.s_pid; 6522 ct.cc_caller_id = nfs4_srv_caller_id; 6523 ct.cc_flags = CC_DONTBLOCK; 6524 err = VOP_OPEN(&cs->vp, fflags, cs->cr, &ct); 6525 if (err) { 6526 rfs4_dbe_unlock(file->dbe); 6527 rfs4_dbe_unlock(state->dbe); 6528 rfs4_file_rele(file); 6529 (void) vop_shrlock(cs->vp, F_UNSHARE, &shr, fflags); 6530 /* Not a fully formed open; "close" it */ 6531 if (screate == TRUE) 6532 rfs4_state_close(state, FALSE, FALSE, cs->cr); 6533 rfs4_state_rele(state); 6534 /* check if a monitor detected a delegation conflict */ 6535 if (err == EAGAIN && (ct.cc_flags & CC_WOULDBLOCK)) 6536 resp->status = NFS4ERR_DELAY; 6537 else 6538 resp->status = NFS4ERR_SERVERFAULT; 6539 return; 6540 } 6541 } else { /* open upgrade */ 6542 /* 6543 * calculate the fflags for the new mode that is being added 6544 * by this upgrade. 6545 */ 6546 fflags = 0; 6547 if (amodes & OPEN4_SHARE_ACCESS_READ) 6548 fflags |= FREAD; 6549 if (amodes & OPEN4_SHARE_ACCESS_WRITE) 6550 fflags |= FWRITE; 6551 vn_open_upgrade(cs->vp, fflags); 6552 } 6553 6554 if (dmodes & OPEN4_SHARE_DENY_READ) 6555 file->deny_read++; 6556 if (dmodes & OPEN4_SHARE_DENY_WRITE) 6557 file->deny_write++; 6558 file->share_deny |= deny; 6559 state->share_deny |= deny; 6560 6561 if (amodes & OPEN4_SHARE_ACCESS_READ) 6562 file->access_read++; 6563 if (amodes & OPEN4_SHARE_ACCESS_WRITE) 6564 file->access_write++; 6565 file->share_access |= access; 6566 state->share_access |= access; 6567 6568 /* 6569 * Check for delegation here. if the deleg argument is not 6570 * DELEG_ANY, then this is a reclaim from a client and 6571 * we must honor the delegation requested. If necessary we can 6572 * set the recall flag. 6573 */ 6574 6575 dsp = rfs4_grant_delegation(deleg, state, &recall); 6576 6577 cs->deleg = (file->dinfo->dtype == OPEN_DELEGATE_WRITE); 6578 6579 next_stateid(&state->stateid); 6580 6581 resp->stateid = state->stateid.stateid; 6582 6583 rfs4_dbe_unlock(file->dbe); 6584 rfs4_dbe_unlock(state->dbe); 6585 6586 if (dsp) { 6587 rfs4_set_deleg_response(dsp, &resp->delegation, NULL, recall); 6588 rfs4_deleg_state_rele(dsp); 6589 } 6590 6591 rfs4_file_rele(file); 6592 rfs4_state_rele(state); 6593 6594 resp->status = NFS4_OK; 6595 } 6596 6597 /*ARGSUSED*/ 6598 static void 6599 rfs4_do_opennull(struct compound_state *cs, struct svc_req *req, 6600 OPEN4args *args, rfs4_openowner_t *oo, OPEN4res *resp) 6601 { 6602 change_info4 *cinfo = &resp->cinfo; 6603 bitmap4 *attrset = &resp->attrset; 6604 6605 if (args->opentype == OPEN4_NOCREATE) 6606 resp->status = rfs4_lookupfile(&args->open_claim4_u.file, 6607 req, cs, args->share_access, cinfo); 6608 else { 6609 /* inhibit delegation grants during exclusive create */ 6610 6611 if (args->mode == EXCLUSIVE4) 6612 rfs4_disable_delegation(); 6613 6614 resp->status = rfs4_createfile(args, req, cs, cinfo, attrset, 6615 oo->client->clientid); 6616 } 6617 6618 if (resp->status == NFS4_OK) { 6619 6620 /* cs->vp cs->fh now reference the desired file */ 6621 6622 rfs4_do_open(cs, req, oo, DELEG_ANY, args->share_access, 6623 args->share_deny, resp, 0); 6624 6625 /* 6626 * If rfs4_createfile set attrset, we must 6627 * clear this attrset before the response is copied. 6628 */ 6629 if (resp->status != NFS4_OK && resp->attrset) { 6630 resp->attrset = 0; 6631 } 6632 } 6633 else 6634 *cs->statusp = resp->status; 6635 6636 if (args->mode == EXCLUSIVE4) 6637 rfs4_enable_delegation(); 6638 } 6639 6640 /*ARGSUSED*/ 6641 static void 6642 rfs4_do_openprev(struct compound_state *cs, struct svc_req *req, 6643 OPEN4args *args, rfs4_openowner_t *oo, OPEN4res *resp) 6644 { 6645 change_info4 *cinfo = &resp->cinfo; 6646 vattr_t va; 6647 vtype_t v_type = cs->vp->v_type; 6648 int error = 0; 6649 6650 /* Verify that we have a regular file */ 6651 if (v_type != VREG) { 6652 if (v_type == VDIR) 6653 resp->status = NFS4ERR_ISDIR; 6654 else if (v_type == VLNK) 6655 resp->status = NFS4ERR_SYMLINK; 6656 else 6657 resp->status = NFS4ERR_INVAL; 6658 return; 6659 } 6660 6661 va.va_mask = AT_MODE|AT_UID; 6662 error = VOP_GETATTR(cs->vp, &va, 0, cs->cr, NULL); 6663 if (error) { 6664 resp->status = puterrno4(error); 6665 return; 6666 } 6667 6668 cs->mandlock = MANDLOCK(cs->vp, va.va_mode); 6669 6670 /* 6671 * Check if we have access to the file, Note the the file 6672 * could have originally been open UNCHECKED or GUARDED 6673 * with mode bits that will now fail, but there is nothing 6674 * we can really do about that except in the case that the 6675 * owner of the file is the one requesting the open. 6676 */ 6677 if (crgetuid(cs->cr) != va.va_uid) { 6678 resp->status = check_open_access(args->share_access, cs, req); 6679 if (resp->status != NFS4_OK) { 6680 return; 6681 } 6682 } 6683 6684 /* 6685 * cinfo on a CLAIM_PREVIOUS is undefined, initialize to zero 6686 */ 6687 cinfo->before = 0; 6688 cinfo->after = 0; 6689 cinfo->atomic = FALSE; 6690 6691 rfs4_do_open(cs, req, oo, 6692 NFS4_DELEG4TYPE2REQTYPE(args->open_claim4_u.delegate_type), 6693 args->share_access, args->share_deny, resp, 0); 6694 } 6695 6696 static void 6697 rfs4_do_opendelcur(struct compound_state *cs, struct svc_req *req, 6698 OPEN4args *args, rfs4_openowner_t *oo, OPEN4res *resp) 6699 { 6700 int error; 6701 nfsstat4 status; 6702 stateid4 stateid = 6703 args->open_claim4_u.delegate_cur_info.delegate_stateid; 6704 rfs4_deleg_state_t *dsp; 6705 6706 /* 6707 * Find the state info from the stateid and confirm that the 6708 * file is delegated. If the state openowner is the same as 6709 * the supplied openowner we're done. If not, get the file 6710 * info from the found state info. Use that file info to 6711 * create the state for this lock owner. Note solaris doen't 6712 * really need the pathname to find the file. We may want to 6713 * lookup the pathname and make sure that the vp exist and 6714 * matches the vp in the file structure. However it is 6715 * possible that the pathname nolonger exists (local process 6716 * unlinks the file), so this may not be that useful. 6717 */ 6718 6719 status = rfs4_get_deleg_state(&stateid, &dsp); 6720 if (status != NFS4_OK) { 6721 resp->status = status; 6722 return; 6723 } 6724 6725 ASSERT(dsp->finfo->dinfo->dtype != OPEN_DELEGATE_NONE); 6726 6727 /* 6728 * New lock owner, create state. Since this was probably called 6729 * in response to a CB_RECALL we set deleg to DELEG_NONE 6730 */ 6731 6732 ASSERT(cs->vp != NULL); 6733 VN_RELE(cs->vp); 6734 VN_HOLD(dsp->finfo->vp); 6735 cs->vp = dsp->finfo->vp; 6736 6737 if (error = makefh4(&cs->fh, cs->vp, cs->exi)) { 6738 rfs4_deleg_state_rele(dsp); 6739 *cs->statusp = resp->status = puterrno4(error); 6740 return; 6741 } 6742 6743 /* Mark progress for delegation returns */ 6744 dsp->finfo->dinfo->time_lastwrite = gethrestime_sec(); 6745 rfs4_deleg_state_rele(dsp); 6746 rfs4_do_open(cs, req, oo, DELEG_NONE, 6747 args->share_access, args->share_deny, resp, 1); 6748 } 6749 6750 /*ARGSUSED*/ 6751 static void 6752 rfs4_do_opendelprev(struct compound_state *cs, struct svc_req *req, 6753 OPEN4args *args, rfs4_openowner_t *oo, OPEN4res *resp) 6754 { 6755 /* 6756 * Lookup the pathname, it must already exist since this file 6757 * was delegated. 6758 * 6759 * Find the file and state info for this vp and open owner pair. 6760 * check that they are in fact delegated. 6761 * check that the state access and deny modes are the same. 6762 * 6763 * Return the delgation possibly seting the recall flag. 6764 */ 6765 rfs4_file_t *file; 6766 rfs4_state_t *state; 6767 bool_t create = FALSE; 6768 bool_t dcreate = FALSE; 6769 rfs4_deleg_state_t *dsp; 6770 nfsace4 *ace; 6771 6772 6773 /* Note we ignore oflags */ 6774 resp->status = rfs4_lookupfile(&args->open_claim4_u.file_delegate_prev, 6775 req, cs, args->share_access, &resp->cinfo); 6776 6777 if (resp->status != NFS4_OK) { 6778 return; 6779 } 6780 6781 /* get the file struct and hold a lock on it during initial open */ 6782 file = rfs4_findfile_withlock(cs->vp, NULL, &create); 6783 if (file == NULL) { 6784 NFS4_DEBUG(rfs4_debug, 6785 (CE_NOTE, "rfs4_do_opendelprev: can't find file")); 6786 resp->status = NFS4ERR_SERVERFAULT; 6787 return; 6788 } 6789 6790 state = rfs4_findstate_by_owner_file(oo, file, &create); 6791 if (state == NULL) { 6792 NFS4_DEBUG(rfs4_debug, 6793 (CE_NOTE, "rfs4_do_opendelprev: can't find state")); 6794 resp->status = NFS4ERR_SERVERFAULT; 6795 rfs4_file_rele_withunlock(file); 6796 return; 6797 } 6798 6799 rfs4_dbe_lock(state->dbe); 6800 rfs4_dbe_lock(file->dbe); 6801 if (args->share_access != state->share_access || 6802 args->share_deny != state->share_deny || 6803 state->finfo->dinfo->dtype == OPEN_DELEGATE_NONE) { 6804 NFS4_DEBUG(rfs4_debug, 6805 (CE_NOTE, "rfs4_do_opendelprev: state mixup")); 6806 rfs4_dbe_unlock(file->dbe); 6807 rfs4_dbe_unlock(state->dbe); 6808 rfs4_file_rele(file); 6809 rfs4_state_rele(state); 6810 resp->status = NFS4ERR_SERVERFAULT; 6811 return; 6812 } 6813 rfs4_dbe_unlock(file->dbe); 6814 rfs4_dbe_unlock(state->dbe); 6815 6816 dsp = rfs4_finddeleg(state, &dcreate); 6817 if (dsp == NULL) { 6818 rfs4_state_rele(state); 6819 rfs4_file_rele(file); 6820 resp->status = NFS4ERR_SERVERFAULT; 6821 return; 6822 } 6823 6824 next_stateid(&state->stateid); 6825 6826 resp->stateid = state->stateid.stateid; 6827 6828 resp->delegation.delegation_type = dsp->dtype; 6829 6830 if (dsp->dtype == OPEN_DELEGATE_READ) { 6831 open_read_delegation4 *rv = 6832 &resp->delegation.open_delegation4_u.read; 6833 6834 rv->stateid = dsp->delegid.stateid; 6835 rv->recall = FALSE; /* no policy in place to set to TRUE */ 6836 ace = &rv->permissions; 6837 } else { 6838 open_write_delegation4 *rv = 6839 &resp->delegation.open_delegation4_u.write; 6840 6841 rv->stateid = dsp->delegid.stateid; 6842 rv->recall = FALSE; /* no policy in place to set to TRUE */ 6843 ace = &rv->permissions; 6844 rv->space_limit.limitby = NFS_LIMIT_SIZE; 6845 rv->space_limit.nfs_space_limit4_u.filesize = UINT64_MAX; 6846 } 6847 6848 /* XXX For now */ 6849 ace->type = ACE4_ACCESS_ALLOWED_ACE_TYPE; 6850 ace->flag = 0; 6851 ace->access_mask = 0; 6852 ace->who.utf8string_len = 0; 6853 ace->who.utf8string_val = 0; 6854 6855 rfs4_deleg_state_rele(dsp); 6856 rfs4_state_rele(state); 6857 rfs4_file_rele(file); 6858 } 6859 6860 typedef enum { 6861 NFS4_CHKSEQ_OKAY = 0, 6862 NFS4_CHKSEQ_REPLAY = 1, 6863 NFS4_CHKSEQ_BAD = 2 6864 } rfs4_chkseq_t; 6865 6866 /* 6867 * Generic function for sequence number checks. 6868 */ 6869 static rfs4_chkseq_t 6870 rfs4_check_seqid(seqid4 seqid, nfs_resop4 *lastop, 6871 seqid4 rqst_seq, nfs_resop4 *resop, bool_t copyres) 6872 { 6873 /* Same sequence ids and matching operations? */ 6874 if (seqid == rqst_seq && resop->resop == lastop->resop) { 6875 if (copyres == TRUE) { 6876 rfs4_free_reply(resop); 6877 rfs4_copy_reply(resop, lastop); 6878 } 6879 NFS4_DEBUG(rfs4_debug, (CE_NOTE, 6880 "Replayed SEQID %d\n", seqid)); 6881 return (NFS4_CHKSEQ_REPLAY); 6882 } 6883 6884 /* If the incoming sequence is not the next expected then it is bad */ 6885 if (rqst_seq != seqid + 1) { 6886 if (rqst_seq == seqid) { 6887 NFS4_DEBUG(rfs4_debug, 6888 (CE_NOTE, "BAD SEQID: Replayed sequence id " 6889 "but last op was %d current op is %d\n", 6890 lastop->resop, resop->resop)); 6891 return (NFS4_CHKSEQ_BAD); 6892 } 6893 NFS4_DEBUG(rfs4_debug, 6894 (CE_NOTE, "BAD SEQID: got %u expecting %u\n", 6895 rqst_seq, seqid)); 6896 return (NFS4_CHKSEQ_BAD); 6897 } 6898 6899 /* Everything okay -- next expected */ 6900 return (NFS4_CHKSEQ_OKAY); 6901 } 6902 6903 6904 static rfs4_chkseq_t 6905 rfs4_check_open_seqid(seqid4 seqid, rfs4_openowner_t *op, nfs_resop4 *resop) 6906 { 6907 rfs4_chkseq_t rc; 6908 6909 rfs4_dbe_lock(op->dbe); 6910 rc = rfs4_check_seqid(op->open_seqid, op->reply, seqid, resop, TRUE); 6911 rfs4_dbe_unlock(op->dbe); 6912 6913 if (rc == NFS4_CHKSEQ_OKAY) 6914 rfs4_update_lease(op->client); 6915 6916 return (rc); 6917 } 6918 6919 static rfs4_chkseq_t 6920 rfs4_check_olo_seqid(seqid4 olo_seqid, rfs4_openowner_t *op, 6921 nfs_resop4 *resop) 6922 { 6923 rfs4_chkseq_t rc; 6924 6925 rfs4_dbe_lock(op->dbe); 6926 rc = rfs4_check_seqid(op->open_seqid, op->reply, 6927 olo_seqid, resop, FALSE); 6928 rfs4_dbe_unlock(op->dbe); 6929 6930 return (rc); 6931 } 6932 6933 static rfs4_chkseq_t 6934 rfs4_check_lock_seqid(seqid4 seqid, rfs4_lo_state_t *lp, nfs_resop4 *resop) 6935 { 6936 rfs4_chkseq_t rc = NFS4_CHKSEQ_OKAY; 6937 6938 rfs4_dbe_lock(lp->dbe); 6939 if (!lp->skip_seqid_check) 6940 rc = rfs4_check_seqid(lp->seqid, lp->reply, seqid, resop, TRUE); 6941 rfs4_dbe_unlock(lp->dbe); 6942 6943 return (rc); 6944 } 6945 6946 static void 6947 rfs4_op_open(nfs_argop4 *argop, nfs_resop4 *resop, 6948 struct svc_req *req, struct compound_state *cs) 6949 { 6950 OPEN4args *args = &argop->nfs_argop4_u.opopen; 6951 OPEN4res *resp = &resop->nfs_resop4_u.opopen; 6952 open_owner4 *owner = &args->owner; 6953 open_claim_type4 claim = args->claim; 6954 rfs4_client_t *cp; 6955 rfs4_openowner_t *oo; 6956 bool_t create; 6957 bool_t replay = FALSE; 6958 int can_reclaim; 6959 6960 DTRACE_NFSV4_2(op__open__start, struct compound_state *, cs, 6961 OPEN4args *, args); 6962 6963 if (cs->vp == NULL) { 6964 *cs->statusp = resp->status = NFS4ERR_NOFILEHANDLE; 6965 goto end; 6966 } 6967 6968 /* 6969 * Need to check clientid and lease expiration first based on 6970 * error ordering and incrementing sequence id. 6971 */ 6972 cp = rfs4_findclient_by_id(owner->clientid, FALSE); 6973 if (cp == NULL) { 6974 *cs->statusp = resp->status = 6975 rfs4_check_clientid(&owner->clientid, 0); 6976 goto end; 6977 } 6978 6979 if (rfs4_lease_expired(cp)) { 6980 rfs4_client_close(cp); 6981 *cs->statusp = resp->status = NFS4ERR_EXPIRED; 6982 goto end; 6983 } 6984 can_reclaim = cp->can_reclaim; 6985 6986 /* 6987 * Find the open_owner for use from this point forward. Take 6988 * care in updating the sequence id based on the type of error 6989 * being returned. 6990 */ 6991 retry: 6992 create = TRUE; 6993 oo = rfs4_findopenowner(owner, &create, args->seqid); 6994 if (oo == NULL) { 6995 *cs->statusp = resp->status = NFS4ERR_STALE_CLIENTID; 6996 rfs4_client_rele(cp); 6997 goto end; 6998 } 6999 7000 /* Hold off access to the sequence space while the open is done */ 7001 rfs4_sw_enter(&oo->oo_sw); 7002 7003 /* 7004 * If the open_owner existed before at the server, then check 7005 * the sequence id. 7006 */ 7007 if (!create && !oo->postpone_confirm) { 7008 switch (rfs4_check_open_seqid(args->seqid, oo, resop)) { 7009 case NFS4_CHKSEQ_BAD: 7010 if ((args->seqid > oo->open_seqid) && 7011 oo->need_confirm) { 7012 rfs4_free_opens(oo, TRUE, FALSE); 7013 rfs4_sw_exit(&oo->oo_sw); 7014 rfs4_openowner_rele(oo); 7015 goto retry; 7016 } 7017 resp->status = NFS4ERR_BAD_SEQID; 7018 goto out; 7019 case NFS4_CHKSEQ_REPLAY: /* replay of previous request */ 7020 replay = TRUE; 7021 goto out; 7022 default: 7023 break; 7024 } 7025 7026 /* 7027 * Sequence was ok and open owner exists 7028 * check to see if we have yet to see an 7029 * open_confirm. 7030 */ 7031 if (oo->need_confirm) { 7032 rfs4_free_opens(oo, TRUE, FALSE); 7033 rfs4_sw_exit(&oo->oo_sw); 7034 rfs4_openowner_rele(oo); 7035 goto retry; 7036 } 7037 } 7038 /* Grace only applies to regular-type OPENs */ 7039 if (rfs4_clnt_in_grace(cp) && 7040 (claim == CLAIM_NULL || claim == CLAIM_DELEGATE_CUR)) { 7041 *cs->statusp = resp->status = NFS4ERR_GRACE; 7042 goto out; 7043 } 7044 7045 /* 7046 * If previous state at the server existed then can_reclaim 7047 * will be set. If not reply NFS4ERR_NO_GRACE to the 7048 * client. 7049 */ 7050 if (rfs4_clnt_in_grace(cp) && claim == CLAIM_PREVIOUS && !can_reclaim) { 7051 *cs->statusp = resp->status = NFS4ERR_NO_GRACE; 7052 goto out; 7053 } 7054 7055 7056 /* 7057 * Reject the open if the client has missed the grace period 7058 */ 7059 if (!rfs4_clnt_in_grace(cp) && claim == CLAIM_PREVIOUS) { 7060 *cs->statusp = resp->status = NFS4ERR_NO_GRACE; 7061 goto out; 7062 } 7063 7064 /* Couple of up-front bookkeeping items */ 7065 if (oo->need_confirm) { 7066 /* 7067 * If this is a reclaim OPEN then we should not ask 7068 * for a confirmation of the open_owner per the 7069 * protocol specification. 7070 */ 7071 if (claim == CLAIM_PREVIOUS) 7072 oo->need_confirm = FALSE; 7073 else 7074 resp->rflags |= OPEN4_RESULT_CONFIRM; 7075 } 7076 resp->rflags |= OPEN4_RESULT_LOCKTYPE_POSIX; 7077 7078 /* 7079 * If there is an unshared filesystem mounted on this vnode, 7080 * do not allow to open/create in this directory. 7081 */ 7082 if (vn_ismntpt(cs->vp)) { 7083 *cs->statusp = resp->status = NFS4ERR_ACCESS; 7084 goto out; 7085 } 7086 7087 /* 7088 * access must READ, WRITE, or BOTH. No access is invalid. 7089 * deny can be READ, WRITE, BOTH, or NONE. 7090 * bits not defined for access/deny are invalid. 7091 */ 7092 if (! (args->share_access & OPEN4_SHARE_ACCESS_BOTH) || 7093 (args->share_access & ~OPEN4_SHARE_ACCESS_BOTH) || 7094 (args->share_deny & ~OPEN4_SHARE_DENY_BOTH)) { 7095 *cs->statusp = resp->status = NFS4ERR_INVAL; 7096 goto out; 7097 } 7098 7099 7100 /* 7101 * make sure attrset is zero before response is built. 7102 */ 7103 resp->attrset = 0; 7104 7105 switch (claim) { 7106 case CLAIM_NULL: 7107 rfs4_do_opennull(cs, req, args, oo, resp); 7108 break; 7109 case CLAIM_PREVIOUS: 7110 rfs4_do_openprev(cs, req, args, oo, resp); 7111 break; 7112 case CLAIM_DELEGATE_CUR: 7113 rfs4_do_opendelcur(cs, req, args, oo, resp); 7114 break; 7115 case CLAIM_DELEGATE_PREV: 7116 rfs4_do_opendelprev(cs, req, args, oo, resp); 7117 break; 7118 default: 7119 resp->status = NFS4ERR_INVAL; 7120 break; 7121 } 7122 7123 out: 7124 rfs4_client_rele(cp); 7125 7126 /* Catch sequence id handling here to make it a little easier */ 7127 switch (resp->status) { 7128 case NFS4ERR_BADXDR: 7129 case NFS4ERR_BAD_SEQID: 7130 case NFS4ERR_BAD_STATEID: 7131 case NFS4ERR_NOFILEHANDLE: 7132 case NFS4ERR_RESOURCE: 7133 case NFS4ERR_STALE_CLIENTID: 7134 case NFS4ERR_STALE_STATEID: 7135 /* 7136 * The protocol states that if any of these errors are 7137 * being returned, the sequence id should not be 7138 * incremented. Any other return requires an 7139 * increment. 7140 */ 7141 break; 7142 default: 7143 /* Always update the lease in this case */ 7144 rfs4_update_lease(oo->client); 7145 7146 /* Regular response - copy the result */ 7147 if (!replay) 7148 rfs4_update_open_resp(oo, resop, &cs->fh); 7149 7150 /* 7151 * REPLAY case: Only if the previous response was OK 7152 * do we copy the filehandle. If not OK, no 7153 * filehandle to copy. 7154 */ 7155 if (replay == TRUE && 7156 resp->status == NFS4_OK && 7157 oo->reply_fh.nfs_fh4_val) { 7158 /* 7159 * If this is a replay, we must restore the 7160 * current filehandle/vp to that of what was 7161 * returned originally. Try our best to do 7162 * it. 7163 */ 7164 nfs_fh4_fmt_t *fh_fmtp = 7165 (nfs_fh4_fmt_t *)oo->reply_fh.nfs_fh4_val; 7166 7167 cs->exi = checkexport4(&fh_fmtp->fh4_fsid, 7168 (fid_t *)&fh_fmtp->fh4_xlen, NULL); 7169 7170 if (cs->exi == NULL) { 7171 resp->status = NFS4ERR_STALE; 7172 goto finish; 7173 } 7174 7175 VN_RELE(cs->vp); 7176 7177 cs->vp = nfs4_fhtovp(&oo->reply_fh, cs->exi, 7178 &resp->status); 7179 7180 if (cs->vp == NULL) 7181 goto finish; 7182 7183 nfs_fh4_copy(&oo->reply_fh, &cs->fh); 7184 } 7185 7186 /* 7187 * If this was a replay, no need to update the 7188 * sequence id. If the open_owner was not created on 7189 * this pass, then update. The first use of an 7190 * open_owner will not bump the sequence id. 7191 */ 7192 if (replay == FALSE && !create) 7193 rfs4_update_open_sequence(oo); 7194 /* 7195 * If the client is receiving an error and the 7196 * open_owner needs to be confirmed, there is no way 7197 * to notify the client of this fact ignoring the fact 7198 * that the server has no method of returning a 7199 * stateid to confirm. Therefore, the server needs to 7200 * mark this open_owner in a way as to avoid the 7201 * sequence id checking the next time the client uses 7202 * this open_owner. 7203 */ 7204 if (resp->status != NFS4_OK && oo->need_confirm) 7205 oo->postpone_confirm = TRUE; 7206 /* 7207 * If OK response then clear the postpone flag and 7208 * reset the sequence id to keep in sync with the 7209 * client. 7210 */ 7211 if (resp->status == NFS4_OK && oo->postpone_confirm) { 7212 oo->postpone_confirm = FALSE; 7213 oo->open_seqid = args->seqid; 7214 } 7215 break; 7216 } 7217 7218 finish: 7219 *cs->statusp = resp->status; 7220 7221 rfs4_sw_exit(&oo->oo_sw); 7222 rfs4_openowner_rele(oo); 7223 7224 end: 7225 DTRACE_NFSV4_2(op__open__done, struct compound_state *, cs, 7226 OPEN4res *, resp); 7227 } 7228 7229 /*ARGSUSED*/ 7230 void 7231 rfs4_op_open_confirm(nfs_argop4 *argop, nfs_resop4 *resop, 7232 struct svc_req *req, struct compound_state *cs) 7233 { 7234 OPEN_CONFIRM4args *args = &argop->nfs_argop4_u.opopen_confirm; 7235 OPEN_CONFIRM4res *resp = &resop->nfs_resop4_u.opopen_confirm; 7236 rfs4_state_t *sp; 7237 nfsstat4 status; 7238 7239 DTRACE_NFSV4_2(op__open__confirm__start, struct compound_state *, cs, 7240 OPEN_CONFIRM4args *, args); 7241 7242 if (cs->vp == NULL) { 7243 *cs->statusp = resp->status = NFS4ERR_NOFILEHANDLE; 7244 goto out; 7245 } 7246 7247 status = rfs4_get_state(&args->open_stateid, &sp, RFS4_DBS_VALID); 7248 if (status != NFS4_OK) { 7249 *cs->statusp = resp->status = status; 7250 goto out; 7251 } 7252 7253 /* Ensure specified filehandle matches */ 7254 if (cs->vp != sp->finfo->vp) { 7255 rfs4_state_rele(sp); 7256 *cs->statusp = resp->status = NFS4ERR_BAD_STATEID; 7257 goto out; 7258 } 7259 7260 /* hold off other access to open_owner while we tinker */ 7261 rfs4_sw_enter(&sp->owner->oo_sw); 7262 7263 switch (rfs4_check_stateid_seqid(sp, &args->open_stateid)) { 7264 case NFS4_CHECK_STATEID_OKAY: 7265 if (rfs4_check_open_seqid(args->seqid, sp->owner, 7266 resop) != 0) { 7267 *cs->statusp = resp->status = NFS4ERR_BAD_SEQID; 7268 break; 7269 } 7270 /* 7271 * If it is the appropriate stateid and determined to 7272 * be "OKAY" then this means that the stateid does not 7273 * need to be confirmed and the client is in error for 7274 * sending an OPEN_CONFIRM. 7275 */ 7276 *cs->statusp = resp->status = NFS4ERR_BAD_STATEID; 7277 break; 7278 case NFS4_CHECK_STATEID_OLD: 7279 *cs->statusp = resp->status = NFS4ERR_OLD_STATEID; 7280 break; 7281 case NFS4_CHECK_STATEID_BAD: 7282 *cs->statusp = resp->status = NFS4ERR_BAD_STATEID; 7283 break; 7284 case NFS4_CHECK_STATEID_EXPIRED: 7285 *cs->statusp = resp->status = NFS4ERR_EXPIRED; 7286 break; 7287 case NFS4_CHECK_STATEID_CLOSED: 7288 *cs->statusp = resp->status = NFS4ERR_OLD_STATEID; 7289 break; 7290 case NFS4_CHECK_STATEID_REPLAY: 7291 switch (rfs4_check_open_seqid(args->seqid, sp->owner, resop)) { 7292 case NFS4_CHKSEQ_OKAY: 7293 /* 7294 * This is replayed stateid; if seqid matches 7295 * next expected, then client is using wrong seqid. 7296 */ 7297 /* fall through */ 7298 case NFS4_CHKSEQ_BAD: 7299 *cs->statusp = resp->status = NFS4ERR_BAD_SEQID; 7300 break; 7301 case NFS4_CHKSEQ_REPLAY: 7302 /* 7303 * Note this case is the duplicate case so 7304 * resp->status is already set. 7305 */ 7306 *cs->statusp = resp->status; 7307 rfs4_update_lease(sp->owner->client); 7308 break; 7309 } 7310 break; 7311 case NFS4_CHECK_STATEID_UNCONFIRMED: 7312 if (rfs4_check_open_seqid(args->seqid, sp->owner, 7313 resop) != NFS4_CHKSEQ_OKAY) { 7314 *cs->statusp = resp->status = NFS4ERR_BAD_SEQID; 7315 break; 7316 } 7317 *cs->statusp = resp->status = NFS4_OK; 7318 7319 next_stateid(&sp->stateid); 7320 resp->open_stateid = sp->stateid.stateid; 7321 sp->owner->need_confirm = FALSE; 7322 rfs4_update_lease(sp->owner->client); 7323 rfs4_update_open_sequence(sp->owner); 7324 rfs4_update_open_resp(sp->owner, resop, NULL); 7325 break; 7326 default: 7327 ASSERT(FALSE); 7328 *cs->statusp = resp->status = NFS4ERR_SERVERFAULT; 7329 break; 7330 } 7331 rfs4_sw_exit(&sp->owner->oo_sw); 7332 rfs4_state_rele(sp); 7333 7334 out: 7335 DTRACE_NFSV4_2(op__open__confirm__done, struct compound_state *, cs, 7336 OPEN_CONFIRM4res *, resp); 7337 } 7338 7339 /*ARGSUSED*/ 7340 void 7341 rfs4_op_open_downgrade(nfs_argop4 *argop, nfs_resop4 *resop, 7342 struct svc_req *req, struct compound_state *cs) 7343 { 7344 OPEN_DOWNGRADE4args *args = &argop->nfs_argop4_u.opopen_downgrade; 7345 OPEN_DOWNGRADE4res *resp = &resop->nfs_resop4_u.opopen_downgrade; 7346 uint32_t access = args->share_access; 7347 uint32_t deny = args->share_deny; 7348 nfsstat4 status; 7349 rfs4_state_t *sp; 7350 rfs4_file_t *fp; 7351 int fflags = 0; 7352 7353 DTRACE_NFSV4_2(op__open__downgrade__start, struct compound_state *, cs, 7354 OPEN_DOWNGRADE4args *, args); 7355 7356 if (cs->vp == NULL) { 7357 *cs->statusp = resp->status = NFS4ERR_NOFILEHANDLE; 7358 goto out; 7359 } 7360 7361 status = rfs4_get_state(&args->open_stateid, &sp, RFS4_DBS_VALID); 7362 if (status != NFS4_OK) { 7363 *cs->statusp = resp->status = status; 7364 goto out; 7365 } 7366 7367 /* Ensure specified filehandle matches */ 7368 if (cs->vp != sp->finfo->vp) { 7369 rfs4_state_rele(sp); 7370 *cs->statusp = resp->status = NFS4ERR_BAD_STATEID; 7371 goto out; 7372 } 7373 7374 /* hold off other access to open_owner while we tinker */ 7375 rfs4_sw_enter(&sp->owner->oo_sw); 7376 7377 switch (rfs4_check_stateid_seqid(sp, &args->open_stateid)) { 7378 case NFS4_CHECK_STATEID_OKAY: 7379 if (rfs4_check_open_seqid(args->seqid, sp->owner, 7380 resop) != NFS4_CHKSEQ_OKAY) { 7381 *cs->statusp = resp->status = NFS4ERR_BAD_SEQID; 7382 goto end; 7383 } 7384 break; 7385 case NFS4_CHECK_STATEID_OLD: 7386 *cs->statusp = resp->status = NFS4ERR_OLD_STATEID; 7387 goto end; 7388 case NFS4_CHECK_STATEID_BAD: 7389 *cs->statusp = resp->status = NFS4ERR_BAD_STATEID; 7390 goto end; 7391 case NFS4_CHECK_STATEID_EXPIRED: 7392 *cs->statusp = resp->status = NFS4ERR_EXPIRED; 7393 goto end; 7394 case NFS4_CHECK_STATEID_CLOSED: 7395 *cs->statusp = resp->status = NFS4ERR_OLD_STATEID; 7396 goto end; 7397 case NFS4_CHECK_STATEID_UNCONFIRMED: 7398 *cs->statusp = resp->status = NFS4ERR_BAD_STATEID; 7399 goto end; 7400 case NFS4_CHECK_STATEID_REPLAY: 7401 /* Check the sequence id for the open owner */ 7402 switch (rfs4_check_open_seqid(args->seqid, sp->owner, resop)) { 7403 case NFS4_CHKSEQ_OKAY: 7404 /* 7405 * This is replayed stateid; if seqid matches 7406 * next expected, then client is using wrong seqid. 7407 */ 7408 /* fall through */ 7409 case NFS4_CHKSEQ_BAD: 7410 *cs->statusp = resp->status = NFS4ERR_BAD_SEQID; 7411 goto end; 7412 case NFS4_CHKSEQ_REPLAY: 7413 /* 7414 * Note this case is the duplicate case so 7415 * resp->status is already set. 7416 */ 7417 *cs->statusp = resp->status; 7418 rfs4_update_lease(sp->owner->client); 7419 goto end; 7420 } 7421 break; 7422 default: 7423 ASSERT(FALSE); 7424 break; 7425 } 7426 7427 rfs4_dbe_lock(sp->dbe); 7428 /* 7429 * Check that the new access modes and deny modes are valid. 7430 * Check that no invalid bits are set. 7431 */ 7432 if ((access & ~(OPEN4_SHARE_ACCESS_READ | OPEN4_SHARE_ACCESS_WRITE)) || 7433 (deny & ~(OPEN4_SHARE_DENY_READ | OPEN4_SHARE_DENY_WRITE))) { 7434 *cs->statusp = resp->status = NFS4ERR_INVAL; 7435 rfs4_update_open_sequence(sp->owner); 7436 rfs4_dbe_unlock(sp->dbe); 7437 goto end; 7438 } 7439 7440 /* 7441 * The new modes must be a subset of the current modes and 7442 * the access must specify at least one mode. To test that 7443 * the new mode is a subset of the current modes we bitwise 7444 * AND them together and check that the result equals the new 7445 * mode. For example: 7446 * New mode, access == R and current mode, sp->share_access == RW 7447 * access & sp->share_access == R == access, so the new access mode 7448 * is valid. Consider access == RW, sp->share_access = R 7449 * access & sp->share_access == R != access, so the new access mode 7450 * is invalid. 7451 */ 7452 if ((access & sp->share_access) != access || 7453 (deny & sp->share_deny) != deny || 7454 (access & 7455 (OPEN4_SHARE_ACCESS_READ | OPEN4_SHARE_ACCESS_WRITE)) == 0) { 7456 *cs->statusp = resp->status = NFS4ERR_INVAL; 7457 rfs4_update_open_sequence(sp->owner); 7458 rfs4_dbe_unlock(sp->dbe); 7459 goto end; 7460 } 7461 7462 /* 7463 * Release any share locks associated with this stateID. 7464 * Strictly speaking, this violates the spec because the 7465 * spec effectively requires that open downgrade be atomic. 7466 * At present, fs_shrlock does not have this capability. 7467 */ 7468 rfs4_dbe_unlock(sp->dbe); 7469 rfs4_unshare(sp); 7470 rfs4_dbe_lock(sp->dbe); 7471 7472 fp = sp->finfo; 7473 rfs4_dbe_lock(fp->dbe); 7474 7475 /* 7476 * If the current mode has deny read and the new mode 7477 * does not, decrement the number of deny read mode bits 7478 * and if it goes to zero turn off the deny read bit 7479 * on the file. 7480 */ 7481 if ((sp->share_deny & OPEN4_SHARE_DENY_READ) && 7482 (deny & OPEN4_SHARE_DENY_READ) == 0) { 7483 fp->deny_read--; 7484 if (fp->deny_read == 0) 7485 fp->share_deny &= ~OPEN4_SHARE_DENY_READ; 7486 } 7487 7488 /* 7489 * If the current mode has deny write and the new mode 7490 * does not, decrement the number of deny write mode bits 7491 * and if it goes to zero turn off the deny write bit 7492 * on the file. 7493 */ 7494 if ((sp->share_deny & OPEN4_SHARE_DENY_WRITE) && 7495 (deny & OPEN4_SHARE_DENY_WRITE) == 0) { 7496 fp->deny_write--; 7497 if (fp->deny_write == 0) 7498 fp->share_deny &= ~OPEN4_SHARE_DENY_WRITE; 7499 } 7500 7501 /* 7502 * If the current mode has access read and the new mode 7503 * does not, decrement the number of access read mode bits 7504 * and if it goes to zero turn off the access read bit 7505 * on the file. set fflags to FREAD for the call to 7506 * vn_open_downgrade(). 7507 */ 7508 if ((sp->share_access & OPEN4_SHARE_ACCESS_READ) && 7509 (access & OPEN4_SHARE_ACCESS_READ) == 0) { 7510 fp->access_read--; 7511 if (fp->access_read == 0) 7512 fp->share_access &= ~OPEN4_SHARE_ACCESS_READ; 7513 fflags |= FREAD; 7514 } 7515 7516 /* 7517 * If the current mode has access write and the new mode 7518 * does not, decrement the number of access write mode bits 7519 * and if it goes to zero turn off the access write bit 7520 * on the file. set fflags to FWRITE for the call to 7521 * vn_open_downgrade(). 7522 */ 7523 if ((sp->share_access & OPEN4_SHARE_ACCESS_WRITE) && 7524 (access & OPEN4_SHARE_ACCESS_WRITE) == 0) { 7525 fp->access_write--; 7526 if (fp->access_write == 0) 7527 fp->share_deny &= ~OPEN4_SHARE_ACCESS_WRITE; 7528 fflags |= FWRITE; 7529 } 7530 7531 /* Set the new access and deny modes */ 7532 sp->share_access = access; 7533 sp->share_deny = deny; 7534 /* Check that the file is still accessible */ 7535 ASSERT(fp->share_access); 7536 7537 rfs4_dbe_unlock(fp->dbe); 7538 7539 rfs4_dbe_unlock(sp->dbe); 7540 7541 if ((status = rfs4_share(sp)) != NFS4_OK) { 7542 *cs->statusp = resp->status = NFS4ERR_SERVERFAULT; 7543 rfs4_update_open_sequence(sp->owner); 7544 goto end; 7545 } 7546 7547 /* 7548 * we successfully downgraded the share lock, now we need to downgrade 7549 * the open. it is possible that the downgrade was only for a deny 7550 * mode and we have nothing else to do. 7551 */ 7552 if ((fflags & (FREAD|FWRITE)) != 0) 7553 vn_open_downgrade(cs->vp, fflags); 7554 7555 rfs4_dbe_lock(sp->dbe); 7556 7557 /* Update the stateid */ 7558 next_stateid(&sp->stateid); 7559 resp->open_stateid = sp->stateid.stateid; 7560 7561 rfs4_dbe_unlock(sp->dbe); 7562 7563 *cs->statusp = resp->status = NFS4_OK; 7564 /* Update the lease */ 7565 rfs4_update_lease(sp->owner->client); 7566 /* And the sequence */ 7567 rfs4_update_open_sequence(sp->owner); 7568 rfs4_update_open_resp(sp->owner, resop, NULL); 7569 7570 end: 7571 rfs4_sw_exit(&sp->owner->oo_sw); 7572 rfs4_state_rele(sp); 7573 out: 7574 DTRACE_NFSV4_2(op__open__downgrade__done, struct compound_state *, cs, 7575 OPEN_DOWNGRADE4res *, resp); 7576 } 7577 7578 /* 7579 * The logic behind this function is detailed in the NFSv4 RFC in the 7580 * SETCLIENTID operation description under IMPLEMENTATION. Refer to 7581 * that section for explicit guidance to server behavior for 7582 * SETCLIENTID. 7583 */ 7584 void 7585 rfs4_op_setclientid(nfs_argop4 *argop, nfs_resop4 *resop, 7586 struct svc_req *req, struct compound_state *cs) 7587 { 7588 SETCLIENTID4args *args = &argop->nfs_argop4_u.opsetclientid; 7589 SETCLIENTID4res *res = &resop->nfs_resop4_u.opsetclientid; 7590 rfs4_client_t *cp, *newcp, *cp_confirmed, *cp_unconfirmed; 7591 bool_t create = TRUE; 7592 char *addr, *netid; 7593 int len; 7594 7595 DTRACE_NFSV4_2(op__setclientid__start, struct compound_state *, cs, 7596 SETCLIENTID4args *, args); 7597 retry: 7598 newcp = cp_confirmed = cp_unconfirmed = NULL; 7599 7600 /* 7601 * In search of an EXISTING client matching the incoming 7602 * request to establish a new client identifier at the server 7603 */ 7604 create = TRUE; 7605 cp = rfs4_findclient(&args->client, &create, NULL); 7606 7607 /* Should never happen */ 7608 ASSERT(cp != NULL); 7609 7610 if (cp == NULL) { 7611 *cs->statusp = res->status = NFS4ERR_SERVERFAULT; 7612 goto out; 7613 } 7614 7615 /* 7616 * Easiest case. Client identifier is newly created and is 7617 * unconfirmed. Also note that for this case, no other 7618 * entries exist for the client identifier. Nothing else to 7619 * check. Just setup the response and respond. 7620 */ 7621 if (create) { 7622 *cs->statusp = res->status = NFS4_OK; 7623 res->SETCLIENTID4res_u.resok4.clientid = cp->clientid; 7624 res->SETCLIENTID4res_u.resok4.setclientid_confirm = 7625 cp->confirm_verf; 7626 /* Setup callback information; CB_NULL confirmation later */ 7627 rfs4_client_setcb(cp, &args->callback, args->callback_ident); 7628 7629 rfs4_client_rele(cp); 7630 goto out; 7631 } 7632 7633 /* 7634 * An existing, confirmed client may exist but it may not have 7635 * been active for at least one lease period. If so, then 7636 * "close" the client and create a new client identifier 7637 */ 7638 if (rfs4_lease_expired(cp)) { 7639 rfs4_client_close(cp); 7640 goto retry; 7641 } 7642 7643 if (cp->need_confirm == TRUE) 7644 cp_unconfirmed = cp; 7645 else 7646 cp_confirmed = cp; 7647 7648 cp = NULL; 7649 7650 /* 7651 * We have a confirmed client, now check for an 7652 * unconfimred entry 7653 */ 7654 if (cp_confirmed) { 7655 /* If creds don't match then client identifier is inuse */ 7656 if (!creds_ok(cp_confirmed->cr_set, req, cs)) { 7657 rfs4_cbinfo_t *cbp; 7658 /* 7659 * Some one else has established this client 7660 * id. Try and say * who they are. We will use 7661 * the call back address supplied by * the 7662 * first client. 7663 */ 7664 *cs->statusp = res->status = NFS4ERR_CLID_INUSE; 7665 7666 addr = netid = NULL; 7667 7668 cbp = &cp_confirmed->cbinfo; 7669 if (cbp->cb_callback.cb_location.r_addr && 7670 cbp->cb_callback.cb_location.r_netid) { 7671 cb_client4 *cbcp = &cbp->cb_callback; 7672 7673 len = strlen(cbcp->cb_location.r_addr)+1; 7674 addr = kmem_alloc(len, KM_SLEEP); 7675 bcopy(cbcp->cb_location.r_addr, addr, len); 7676 len = strlen(cbcp->cb_location.r_netid)+1; 7677 netid = kmem_alloc(len, KM_SLEEP); 7678 bcopy(cbcp->cb_location.r_netid, netid, len); 7679 } 7680 7681 res->SETCLIENTID4res_u.client_using.r_addr = addr; 7682 res->SETCLIENTID4res_u.client_using.r_netid = netid; 7683 7684 rfs4_client_rele(cp_confirmed); 7685 } 7686 7687 /* 7688 * Confirmed, creds match, and verifier matches; must 7689 * be an update of the callback info 7690 */ 7691 if (cp_confirmed->nfs_client.verifier == 7692 args->client.verifier) { 7693 /* Setup callback information */ 7694 rfs4_client_setcb(cp_confirmed, &args->callback, 7695 args->callback_ident); 7696 7697 /* everything okay -- move ahead */ 7698 *cs->statusp = res->status = NFS4_OK; 7699 res->SETCLIENTID4res_u.resok4.clientid = 7700 cp_confirmed->clientid; 7701 7702 /* update the confirm_verifier and return it */ 7703 rfs4_client_scv_next(cp_confirmed); 7704 res->SETCLIENTID4res_u.resok4.setclientid_confirm = 7705 cp_confirmed->confirm_verf; 7706 7707 rfs4_client_rele(cp_confirmed); 7708 goto out; 7709 } 7710 7711 /* 7712 * Creds match but the verifier doesn't. Must search 7713 * for an unconfirmed client that would be replaced by 7714 * this request. 7715 */ 7716 create = FALSE; 7717 cp_unconfirmed = rfs4_findclient(&args->client, &create, 7718 cp_confirmed); 7719 } 7720 7721 /* 7722 * At this point, we have taken care of the brand new client 7723 * struct, INUSE case, update of an existing, and confirmed 7724 * client struct. 7725 */ 7726 7727 /* 7728 * check to see if things have changed while we originally 7729 * picked up the client struct. If they have, then return and 7730 * retry the processing of this SETCLIENTID request. 7731 */ 7732 if (cp_unconfirmed) { 7733 rfs4_dbe_lock(cp_unconfirmed->dbe); 7734 if (!cp_unconfirmed->need_confirm) { 7735 rfs4_dbe_unlock(cp_unconfirmed->dbe); 7736 rfs4_client_rele(cp_unconfirmed); 7737 if (cp_confirmed) 7738 rfs4_client_rele(cp_confirmed); 7739 goto retry; 7740 } 7741 /* do away with the old unconfirmed one */ 7742 rfs4_dbe_invalidate(cp_unconfirmed->dbe); 7743 rfs4_dbe_unlock(cp_unconfirmed->dbe); 7744 rfs4_client_rele(cp_unconfirmed); 7745 cp_unconfirmed = NULL; 7746 } 7747 7748 /* 7749 * This search will temporarily hide the confirmed client 7750 * struct while a new client struct is created as the 7751 * unconfirmed one. 7752 */ 7753 create = TRUE; 7754 newcp = rfs4_findclient(&args->client, &create, cp_confirmed); 7755 7756 ASSERT(newcp != NULL); 7757 7758 if (newcp == NULL) { 7759 *cs->statusp = res->status = NFS4ERR_SERVERFAULT; 7760 rfs4_client_rele(cp_confirmed); 7761 goto out; 7762 } 7763 7764 /* 7765 * If one was not created, then a similar request must be in 7766 * process so release and start over with this one 7767 */ 7768 if (create != TRUE) { 7769 rfs4_client_rele(newcp); 7770 if (cp_confirmed) 7771 rfs4_client_rele(cp_confirmed); 7772 goto retry; 7773 } 7774 7775 *cs->statusp = res->status = NFS4_OK; 7776 res->SETCLIENTID4res_u.resok4.clientid = newcp->clientid; 7777 res->SETCLIENTID4res_u.resok4.setclientid_confirm = newcp->confirm_verf; 7778 /* Setup callback information; CB_NULL confirmation later */ 7779 rfs4_client_setcb(newcp, &args->callback, args->callback_ident); 7780 7781 newcp->cp_confirmed = cp_confirmed; 7782 7783 rfs4_client_rele(newcp); 7784 7785 out: 7786 DTRACE_NFSV4_2(op__setclientid__done, struct compound_state *, cs, 7787 SETCLIENTID4res *, res); 7788 } 7789 7790 /*ARGSUSED*/ 7791 void 7792 rfs4_op_setclientid_confirm(nfs_argop4 *argop, nfs_resop4 *resop, 7793 struct svc_req *req, struct compound_state *cs) 7794 { 7795 SETCLIENTID_CONFIRM4args *args = 7796 &argop->nfs_argop4_u.opsetclientid_confirm; 7797 SETCLIENTID_CONFIRM4res *res = 7798 &resop->nfs_resop4_u.opsetclientid_confirm; 7799 rfs4_client_t *cp, *cptoclose = NULL; 7800 7801 DTRACE_NFSV4_2(op__setclientid__confirm__start, 7802 struct compound_state *, cs, 7803 SETCLIENTID_CONFIRM4args *, args); 7804 7805 *cs->statusp = res->status = NFS4_OK; 7806 7807 cp = rfs4_findclient_by_id(args->clientid, TRUE); 7808 7809 if (cp == NULL) { 7810 *cs->statusp = res->status = 7811 rfs4_check_clientid(&args->clientid, 1); 7812 goto out; 7813 } 7814 7815 if (!creds_ok(cp, req, cs)) { 7816 *cs->statusp = res->status = NFS4ERR_CLID_INUSE; 7817 rfs4_client_rele(cp); 7818 goto out; 7819 } 7820 7821 /* If the verifier doesn't match, the record doesn't match */ 7822 if (cp->confirm_verf != args->setclientid_confirm) { 7823 *cs->statusp = res->status = NFS4ERR_STALE_CLIENTID; 7824 rfs4_client_rele(cp); 7825 goto out; 7826 } 7827 7828 rfs4_dbe_lock(cp->dbe); 7829 cp->need_confirm = FALSE; 7830 if (cp->cp_confirmed) { 7831 cptoclose = cp->cp_confirmed; 7832 cptoclose->ss_remove = 1; 7833 cp->cp_confirmed = NULL; 7834 } 7835 7836 /* 7837 * Update the client's associated server instance, if it's changed 7838 * since the client was created. 7839 */ 7840 if (rfs4_servinst(cp) != rfs4_cur_servinst) 7841 rfs4_servinst_assign(cp, rfs4_cur_servinst); 7842 7843 /* 7844 * Record clientid in stable storage. 7845 * Must be done after server instance has been assigned. 7846 */ 7847 rfs4_ss_clid(cp, req); 7848 7849 rfs4_dbe_unlock(cp->dbe); 7850 7851 if (cptoclose) 7852 /* don't need to rele, client_close does it */ 7853 rfs4_client_close(cptoclose); 7854 7855 /* If needed, initiate CB_NULL call for callback path */ 7856 rfs4_deleg_cb_check(cp); 7857 rfs4_update_lease(cp); 7858 7859 /* 7860 * Check to see if client can perform reclaims 7861 */ 7862 rfs4_ss_chkclid(cp); 7863 7864 rfs4_client_rele(cp); 7865 7866 out: 7867 DTRACE_NFSV4_2(op__setclientid__confirm__done, 7868 struct compound_state *, cs, 7869 SETCLIENTID_CONFIRM4 *, res); 7870 } 7871 7872 7873 /*ARGSUSED*/ 7874 void 7875 rfs4_op_close(nfs_argop4 *argop, nfs_resop4 *resop, 7876 struct svc_req *req, struct compound_state *cs) 7877 { 7878 CLOSE4args *args = &argop->nfs_argop4_u.opclose; 7879 CLOSE4res *resp = &resop->nfs_resop4_u.opclose; 7880 rfs4_state_t *sp; 7881 nfsstat4 status; 7882 7883 DTRACE_NFSV4_2(op__close__start, struct compound_state *, cs, 7884 CLOSE4args *, args); 7885 7886 if (cs->vp == NULL) { 7887 *cs->statusp = resp->status = NFS4ERR_NOFILEHANDLE; 7888 goto out; 7889 } 7890 7891 status = rfs4_get_state(&args->open_stateid, &sp, RFS4_DBS_INVALID); 7892 if (status != NFS4_OK) { 7893 *cs->statusp = resp->status = status; 7894 goto out; 7895 } 7896 7897 /* Ensure specified filehandle matches */ 7898 if (cs->vp != sp->finfo->vp) { 7899 rfs4_state_rele(sp); 7900 *cs->statusp = resp->status = NFS4ERR_BAD_STATEID; 7901 goto out; 7902 } 7903 7904 /* hold off other access to open_owner while we tinker */ 7905 rfs4_sw_enter(&sp->owner->oo_sw); 7906 7907 switch (rfs4_check_stateid_seqid(sp, &args->open_stateid)) { 7908 case NFS4_CHECK_STATEID_OKAY: 7909 if (rfs4_check_open_seqid(args->seqid, sp->owner, 7910 resop) != NFS4_CHKSEQ_OKAY) { 7911 *cs->statusp = resp->status = NFS4ERR_BAD_SEQID; 7912 goto end; 7913 } 7914 break; 7915 case NFS4_CHECK_STATEID_OLD: 7916 *cs->statusp = resp->status = NFS4ERR_OLD_STATEID; 7917 goto end; 7918 case NFS4_CHECK_STATEID_BAD: 7919 *cs->statusp = resp->status = NFS4ERR_BAD_STATEID; 7920 goto end; 7921 case NFS4_CHECK_STATEID_EXPIRED: 7922 *cs->statusp = resp->status = NFS4ERR_EXPIRED; 7923 goto end; 7924 case NFS4_CHECK_STATEID_CLOSED: 7925 *cs->statusp = resp->status = NFS4ERR_OLD_STATEID; 7926 goto end; 7927 case NFS4_CHECK_STATEID_UNCONFIRMED: 7928 *cs->statusp = resp->status = NFS4ERR_BAD_STATEID; 7929 goto end; 7930 case NFS4_CHECK_STATEID_REPLAY: 7931 /* Check the sequence id for the open owner */ 7932 switch (rfs4_check_open_seqid(args->seqid, sp->owner, resop)) { 7933 case NFS4_CHKSEQ_OKAY: 7934 /* 7935 * This is replayed stateid; if seqid matches 7936 * next expected, then client is using wrong seqid. 7937 */ 7938 /* FALL THROUGH */ 7939 case NFS4_CHKSEQ_BAD: 7940 *cs->statusp = resp->status = NFS4ERR_BAD_SEQID; 7941 goto end; 7942 case NFS4_CHKSEQ_REPLAY: 7943 /* 7944 * Note this case is the duplicate case so 7945 * resp->status is already set. 7946 */ 7947 *cs->statusp = resp->status; 7948 rfs4_update_lease(sp->owner->client); 7949 goto end; 7950 } 7951 break; 7952 default: 7953 ASSERT(FALSE); 7954 break; 7955 } 7956 7957 rfs4_dbe_lock(sp->dbe); 7958 7959 /* Update the stateid. */ 7960 next_stateid(&sp->stateid); 7961 resp->open_stateid = sp->stateid.stateid; 7962 7963 rfs4_dbe_unlock(sp->dbe); 7964 7965 rfs4_update_lease(sp->owner->client); 7966 rfs4_update_open_sequence(sp->owner); 7967 rfs4_update_open_resp(sp->owner, resop, NULL); 7968 7969 rfs4_state_close(sp, FALSE, FALSE, cs->cr); 7970 7971 *cs->statusp = resp->status = status; 7972 7973 end: 7974 rfs4_sw_exit(&sp->owner->oo_sw); 7975 rfs4_state_rele(sp); 7976 out: 7977 DTRACE_NFSV4_2(op__close__done, struct compound_state *, cs, 7978 CLOSE4res *, resp); 7979 } 7980 7981 /* 7982 * Manage the counts on the file struct and close all file locks 7983 */ 7984 /*ARGSUSED*/ 7985 void 7986 rfs4_release_share_lock_state(rfs4_state_t *sp, cred_t *cr, 7987 bool_t close_of_client) 7988 { 7989 rfs4_file_t *fp = sp->finfo; 7990 rfs4_lo_state_t *lsp; 7991 struct shrlock shr; 7992 struct shr_locowner shr_loco; 7993 int fflags, s_access, s_deny; 7994 7995 fflags = s_access = s_deny = 0; 7996 /* 7997 * Decrement the count for each access and deny bit that this 7998 * state has contributed to the file. If the file counts go to zero 7999 * clear the appropriate bit in the appropriate mask. 8000 */ 8001 8002 if (sp->share_access & OPEN4_SHARE_ACCESS_READ) { 8003 fp->access_read--; 8004 fflags |= FREAD; 8005 s_access |= F_RDACC; 8006 if (fp->access_read == 0) 8007 fp->share_access &= ~OPEN4_SHARE_ACCESS_READ; 8008 } 8009 if (sp->share_access & OPEN4_SHARE_ACCESS_WRITE) { 8010 fp->access_write--; 8011 fflags |= FWRITE; 8012 s_access |= F_WRACC; 8013 if (fp->access_write == 0) 8014 fp->share_access &= ~OPEN4_SHARE_ACCESS_WRITE; 8015 } 8016 if (sp->share_deny & OPEN4_SHARE_DENY_READ) { 8017 fp->deny_read--; 8018 s_deny |= F_RDDNY; 8019 if (fp->deny_read == 0) 8020 fp->share_deny &= ~OPEN4_SHARE_DENY_READ; 8021 } 8022 if (sp->share_deny & OPEN4_SHARE_DENY_WRITE) { 8023 fp->deny_write--; 8024 s_deny |= F_WRDNY; 8025 if (fp->deny_write == 0) 8026 fp->share_deny &= ~OPEN4_SHARE_DENY_WRITE; 8027 } 8028 8029 /* 8030 * If this call is part of the larger closing down of client 8031 * state then it is just easier to release all locks 8032 * associated with this client instead of going through each 8033 * individual file and cleaning locks there. 8034 */ 8035 if (close_of_client) { 8036 if (sp->owner->client->unlksys_completed == FALSE && 8037 sp->lockownerlist.next->lsp != NULL && 8038 sp->owner->client->sysidt != LM_NOSYSID) { 8039 /* Is the PxFS kernel module loaded? */ 8040 if (lm_remove_file_locks != NULL) { 8041 int new_sysid; 8042 8043 /* Encode the cluster nodeid in new sysid */ 8044 new_sysid = sp->owner->client->sysidt; 8045 lm_set_nlmid_flk(&new_sysid); 8046 8047 /* 8048 * This PxFS routine removes file locks for a 8049 * client over all nodes of a cluster. 8050 */ 8051 NFS4_DEBUG(rfs4_debug, (CE_NOTE, 8052 "lm_remove_file_locks(sysid=0x%x)\n", 8053 new_sysid)); 8054 (*lm_remove_file_locks)(new_sysid); 8055 } else { 8056 struct flock64 flk; 8057 8058 /* Release all locks for this client */ 8059 flk.l_type = F_UNLKSYS; 8060 flk.l_whence = 0; 8061 flk.l_start = 0; 8062 flk.l_len = 0; 8063 flk.l_sysid = sp->owner->client->sysidt; 8064 flk.l_pid = 0; 8065 (void) VOP_FRLOCK(sp->finfo->vp, F_SETLK, &flk, 8066 F_REMOTELOCK | FREAD | FWRITE, 8067 (u_offset_t)0, NULL, CRED(), NULL); 8068 } 8069 8070 sp->owner->client->unlksys_completed = TRUE; 8071 } 8072 } 8073 8074 /* 8075 * Release all locks on this file by this lock owner or at 8076 * least mark the locks as having been released 8077 */ 8078 for (lsp = sp->lockownerlist.next->lsp; lsp != NULL; 8079 lsp = lsp->lockownerlist.next->lsp) { 8080 8081 lsp->locks_cleaned = TRUE; 8082 8083 /* Was this already taken care of above? */ 8084 if (!close_of_client && 8085 sp->owner->client->sysidt != LM_NOSYSID) 8086 (void) cleanlocks(sp->finfo->vp, lsp->locker->pid, 8087 lsp->locker->client->sysidt); 8088 } 8089 8090 /* 8091 * Release any shrlocks associated with this open state ID. 8092 * This must be done before the rfs4_state gets marked closed. 8093 */ 8094 if (sp->owner->client->sysidt != LM_NOSYSID) { 8095 shr.s_access = s_access; 8096 shr.s_deny = s_deny; 8097 shr.s_pid = rfs4_dbe_getid(sp->owner->dbe); 8098 shr.s_sysid = sp->owner->client->sysidt; 8099 shr_loco.sl_pid = shr.s_pid; 8100 shr_loco.sl_id = shr.s_sysid; 8101 shr.s_owner = (caddr_t)&shr_loco; 8102 shr.s_own_len = sizeof (shr_loco); 8103 (void) vop_shrlock(sp->finfo->vp, F_UNSHARE, &shr, fflags); 8104 } 8105 8106 (void) VOP_CLOSE(fp->vp, fflags, 1, (offset_t)0, cr, NULL); 8107 } 8108 8109 /* 8110 * lock_denied: Fill in a LOCK4deneid structure given an flock64 structure. 8111 */ 8112 static nfsstat4 8113 lock_denied(LOCK4denied *dp, struct flock64 *flk) 8114 { 8115 rfs4_lockowner_t *lo; 8116 rfs4_client_t *cp; 8117 uint32_t len; 8118 8119 lo = rfs4_findlockowner_by_pid(flk->l_pid); 8120 if (lo != NULL) { 8121 cp = lo->client; 8122 if (rfs4_lease_expired(cp)) { 8123 rfs4_lockowner_rele(lo); 8124 rfs4_dbe_hold(cp->dbe); 8125 rfs4_client_close(cp); 8126 return (NFS4ERR_EXPIRED); 8127 } 8128 dp->owner.clientid = lo->owner.clientid; 8129 len = lo->owner.owner_len; 8130 dp->owner.owner_val = kmem_alloc(len, KM_SLEEP); 8131 bcopy(lo->owner.owner_val, dp->owner.owner_val, len); 8132 dp->owner.owner_len = len; 8133 rfs4_lockowner_rele(lo); 8134 goto finish; 8135 } 8136 8137 /* 8138 * Its not a NFS4 lock. We take advantage that the upper 32 bits 8139 * of the client id contain the boot time for a NFS4 lock. So we 8140 * fabricate and identity by setting clientid to the sysid, and 8141 * the lock owner to the pid. 8142 */ 8143 dp->owner.clientid = flk->l_sysid; 8144 len = sizeof (pid_t); 8145 dp->owner.owner_len = len; 8146 dp->owner.owner_val = kmem_alloc(len, KM_SLEEP); 8147 bcopy(&flk->l_pid, dp->owner.owner_val, len); 8148 finish: 8149 dp->offset = flk->l_start; 8150 dp->length = flk->l_len; 8151 8152 if (flk->l_type == F_RDLCK) 8153 dp->locktype = READ_LT; 8154 else if (flk->l_type == F_WRLCK) 8155 dp->locktype = WRITE_LT; 8156 else 8157 return (NFS4ERR_INVAL); /* no mapping from POSIX ltype to v4 */ 8158 8159 return (NFS4_OK); 8160 } 8161 8162 static int 8163 setlock(vnode_t *vp, struct flock64 *flock, int flag, cred_t *cred) 8164 { 8165 int error; 8166 struct flock64 flk; 8167 int i; 8168 clock_t delaytime; 8169 int cmd; 8170 8171 cmd = nbl_need_check(vp) ? F_SETLK_NBMAND : F_SETLK; 8172 retry: 8173 delaytime = MSEC_TO_TICK_ROUNDUP(rfs4_lock_delay); 8174 8175 for (i = 0; i < rfs4_maxlock_tries; i++) { 8176 LOCK_PRINT(rfs4_debug, "setlock", cmd, flock); 8177 error = VOP_FRLOCK(vp, cmd, 8178 flock, flag, (u_offset_t)0, NULL, cred, NULL); 8179 8180 if (error != EAGAIN && error != EACCES) 8181 break; 8182 8183 if (i < rfs4_maxlock_tries - 1) { 8184 delay(delaytime); 8185 delaytime *= 2; 8186 } 8187 } 8188 8189 if (error == EAGAIN || error == EACCES) { 8190 /* Get the owner of the lock */ 8191 flk = *flock; 8192 LOCK_PRINT(rfs4_debug, "setlock", F_GETLK, &flk); 8193 if (VOP_FRLOCK(vp, F_GETLK, &flk, flag, 8194 (u_offset_t)0, NULL, cred, NULL) == 0) { 8195 if (flk.l_type == F_UNLCK) { 8196 /* No longer locked, retry */ 8197 goto retry; 8198 } 8199 *flock = flk; 8200 LOCK_PRINT(rfs4_debug, "setlock(blocking lock)", 8201 F_GETLK, &flk); 8202 } 8203 } 8204 8205 return (error); 8206 } 8207 8208 /*ARGSUSED*/ 8209 static nfsstat4 8210 rfs4_do_lock(rfs4_lo_state_t *lp, nfs_lock_type4 locktype, 8211 seqid4 seqid, offset4 offset, 8212 length4 length, cred_t *cred, nfs_resop4 *resop) 8213 { 8214 nfsstat4 status; 8215 rfs4_lockowner_t *lo = lp->locker; 8216 rfs4_state_t *sp = lp->state; 8217 struct flock64 flock; 8218 int16_t ltype; 8219 int flag; 8220 int error; 8221 sysid_t sysid; 8222 LOCK4res *lres; 8223 8224 if (rfs4_lease_expired(lo->client)) { 8225 return (NFS4ERR_EXPIRED); 8226 } 8227 8228 if ((status = rfs4_client_sysid(lo->client, &sysid)) != NFS4_OK) 8229 return (status); 8230 8231 /* Check for zero length. To lock to end of file use all ones for V4 */ 8232 if (length == 0) 8233 return (NFS4ERR_INVAL); 8234 else if (length == (length4)(~0)) 8235 length = 0; /* Posix to end of file */ 8236 8237 retry: 8238 rfs4_dbe_lock(sp->dbe); 8239 8240 8241 if (resop->resop != OP_LOCKU) { 8242 switch (locktype) { 8243 case READ_LT: 8244 case READW_LT: 8245 if ((sp->share_access 8246 & OPEN4_SHARE_ACCESS_READ) == 0) { 8247 rfs4_dbe_unlock(sp->dbe); 8248 8249 return (NFS4ERR_OPENMODE); 8250 } 8251 ltype = F_RDLCK; 8252 break; 8253 case WRITE_LT: 8254 case WRITEW_LT: 8255 if ((sp->share_access 8256 & OPEN4_SHARE_ACCESS_WRITE) == 0) { 8257 rfs4_dbe_unlock(sp->dbe); 8258 8259 return (NFS4ERR_OPENMODE); 8260 } 8261 ltype = F_WRLCK; 8262 break; 8263 } 8264 } else 8265 ltype = F_UNLCK; 8266 8267 flock.l_type = ltype; 8268 flock.l_whence = 0; /* SEEK_SET */ 8269 flock.l_start = offset; 8270 flock.l_len = length; 8271 flock.l_sysid = sysid; 8272 flock.l_pid = lp->locker->pid; 8273 8274 /* Note that length4 is uint64_t but l_len and l_start are off64_t */ 8275 if (flock.l_len < 0 || flock.l_start < 0) { 8276 rfs4_dbe_unlock(sp->dbe); 8277 return (NFS4ERR_INVAL); 8278 } 8279 8280 /* 8281 * N.B. FREAD has the same value as OPEN4_SHARE_ACCESS_READ and 8282 * FWRITE has the same value as OPEN4_SHARE_ACCESS_WRITE. 8283 */ 8284 flag = (int)sp->share_access | F_REMOTELOCK; 8285 8286 error = setlock(sp->finfo->vp, &flock, flag, cred); 8287 if (error == 0) { 8288 rfs4_dbe_lock(lp->dbe); 8289 next_stateid(&lp->lockid); 8290 rfs4_dbe_unlock(lp->dbe); 8291 } 8292 8293 rfs4_dbe_unlock(sp->dbe); 8294 8295 /* 8296 * N.B. We map error values to nfsv4 errors. This is differrent 8297 * than puterrno4 routine. 8298 */ 8299 switch (error) { 8300 case 0: 8301 status = NFS4_OK; 8302 break; 8303 case EAGAIN: 8304 case EACCES: /* Old value */ 8305 /* Can only get here if op is OP_LOCK */ 8306 ASSERT(resop->resop == OP_LOCK); 8307 lres = &resop->nfs_resop4_u.oplock; 8308 status = NFS4ERR_DENIED; 8309 if (lock_denied(&lres->LOCK4res_u.denied, &flock) 8310 == NFS4ERR_EXPIRED) 8311 goto retry; 8312 break; 8313 case ENOLCK: 8314 status = NFS4ERR_DELAY; 8315 break; 8316 case EOVERFLOW: 8317 status = NFS4ERR_INVAL; 8318 break; 8319 case EINVAL: 8320 status = NFS4ERR_NOTSUPP; 8321 break; 8322 default: 8323 cmn_err(CE_WARN, "rfs4_do_lock: unexpected errno (%d)", error); 8324 status = NFS4ERR_SERVERFAULT; 8325 break; 8326 } 8327 8328 return (status); 8329 } 8330 8331 /*ARGSUSED*/ 8332 void 8333 rfs4_op_lock(nfs_argop4 *argop, nfs_resop4 *resop, 8334 struct svc_req *req, struct compound_state *cs) 8335 { 8336 LOCK4args *args = &argop->nfs_argop4_u.oplock; 8337 LOCK4res *resp = &resop->nfs_resop4_u.oplock; 8338 nfsstat4 status; 8339 stateid4 *stateid; 8340 rfs4_lockowner_t *lo; 8341 rfs4_client_t *cp; 8342 rfs4_state_t *sp = NULL; 8343 rfs4_lo_state_t *lsp = NULL; 8344 bool_t ls_sw_held = FALSE; 8345 bool_t create = TRUE; 8346 bool_t lcreate = TRUE; 8347 bool_t dup_lock = FALSE; 8348 int rc; 8349 8350 DTRACE_NFSV4_2(op__lock__start, struct compound_state *, cs, 8351 LOCK4args *, args); 8352 8353 if (cs->vp == NULL) { 8354 *cs->statusp = resp->status = NFS4ERR_NOFILEHANDLE; 8355 DTRACE_NFSV4_2(op__lock__done, struct compound_state *, 8356 cs, LOCK4res *, resp); 8357 return; 8358 } 8359 8360 if (args->locker.new_lock_owner) { 8361 /* Create a new lockowner for this instance */ 8362 open_to_lock_owner4 *olo = &args->locker.locker4_u.open_owner; 8363 8364 NFS4_DEBUG(rfs4_debug, (CE_NOTE, "Creating new lock owner")); 8365 8366 stateid = &olo->open_stateid; 8367 status = rfs4_get_state(stateid, &sp, RFS4_DBS_VALID); 8368 if (status != NFS4_OK) { 8369 NFS4_DEBUG(rfs4_debug, 8370 (CE_NOTE, "Get state failed in lock %d", status)); 8371 *cs->statusp = resp->status = status; 8372 DTRACE_NFSV4_2(op__lock__done, struct compound_state *, 8373 cs, LOCK4res *, resp); 8374 return; 8375 } 8376 8377 /* Ensure specified filehandle matches */ 8378 if (cs->vp != sp->finfo->vp) { 8379 rfs4_state_rele(sp); 8380 *cs->statusp = resp->status = NFS4ERR_BAD_STATEID; 8381 DTRACE_NFSV4_2(op__lock__done, struct compound_state *, 8382 cs, LOCK4res *, resp); 8383 return; 8384 } 8385 8386 /* hold off other access to open_owner while we tinker */ 8387 rfs4_sw_enter(&sp->owner->oo_sw); 8388 8389 switch (rc = rfs4_check_stateid_seqid(sp, stateid)) { 8390 case NFS4_CHECK_STATEID_OLD: 8391 *cs->statusp = resp->status = NFS4ERR_OLD_STATEID; 8392 goto end; 8393 case NFS4_CHECK_STATEID_BAD: 8394 *cs->statusp = resp->status = NFS4ERR_BAD_STATEID; 8395 goto end; 8396 case NFS4_CHECK_STATEID_EXPIRED: 8397 *cs->statusp = resp->status = NFS4ERR_EXPIRED; 8398 goto end; 8399 case NFS4_CHECK_STATEID_UNCONFIRMED: 8400 *cs->statusp = resp->status = NFS4ERR_BAD_STATEID; 8401 goto end; 8402 case NFS4_CHECK_STATEID_CLOSED: 8403 *cs->statusp = resp->status = NFS4ERR_OLD_STATEID; 8404 goto end; 8405 case NFS4_CHECK_STATEID_OKAY: 8406 case NFS4_CHECK_STATEID_REPLAY: 8407 switch (rfs4_check_olo_seqid(olo->open_seqid, 8408 sp->owner, resop)) { 8409 case NFS4_CHKSEQ_OKAY: 8410 if (rc == NFS4_CHECK_STATEID_OKAY) 8411 break; 8412 /* 8413 * This is replayed stateid; if seqid 8414 * matches next expected, then client 8415 * is using wrong seqid. 8416 */ 8417 /* FALLTHROUGH */ 8418 case NFS4_CHKSEQ_BAD: 8419 *cs->statusp = resp->status = NFS4ERR_BAD_SEQID; 8420 goto end; 8421 case NFS4_CHKSEQ_REPLAY: 8422 /* This is a duplicate LOCK request */ 8423 dup_lock = TRUE; 8424 8425 /* 8426 * For a duplicate we do not want to 8427 * create a new lockowner as it should 8428 * already exist. 8429 * Turn off the lockowner create flag. 8430 */ 8431 lcreate = FALSE; 8432 } 8433 break; 8434 } 8435 8436 lo = rfs4_findlockowner(&olo->lock_owner, &lcreate); 8437 if (lo == NULL) { 8438 NFS4_DEBUG(rfs4_debug, 8439 (CE_NOTE, "rfs4_op_lock: no lock owner")); 8440 *cs->statusp = resp->status = NFS4ERR_RESOURCE; 8441 goto end; 8442 } 8443 8444 lsp = rfs4_findlo_state_by_owner(lo, sp, &create); 8445 if (lsp == NULL) { 8446 rfs4_update_lease(sp->owner->client); 8447 /* 8448 * Only update theh open_seqid if this is not 8449 * a duplicate request 8450 */ 8451 if (dup_lock == FALSE) { 8452 rfs4_update_open_sequence(sp->owner); 8453 } 8454 8455 NFS4_DEBUG(rfs4_debug, 8456 (CE_NOTE, "rfs4_op_lock: no state")); 8457 *cs->statusp = resp->status = NFS4ERR_SERVERFAULT; 8458 rfs4_update_open_resp(sp->owner, resop, NULL); 8459 rfs4_lockowner_rele(lo); 8460 goto end; 8461 } 8462 8463 /* 8464 * This is the new_lock_owner branch and the client is 8465 * supposed to be associating a new lock_owner with 8466 * the open file at this point. If we find that a 8467 * lock_owner/state association already exists and a 8468 * successful LOCK request was returned to the client, 8469 * an error is returned to the client since this is 8470 * not appropriate. The client should be using the 8471 * existing lock_owner branch. 8472 */ 8473 if (dup_lock == FALSE && create == FALSE) { 8474 if (lsp->lock_completed == TRUE) { 8475 *cs->statusp = 8476 resp->status = NFS4ERR_BAD_SEQID; 8477 rfs4_lockowner_rele(lo); 8478 goto end; 8479 } 8480 } 8481 8482 rfs4_update_lease(sp->owner->client); 8483 8484 /* 8485 * Only update theh open_seqid if this is not 8486 * a duplicate request 8487 */ 8488 if (dup_lock == FALSE) { 8489 rfs4_update_open_sequence(sp->owner); 8490 } 8491 8492 /* 8493 * If this is a duplicate lock request, just copy the 8494 * previously saved reply and return. 8495 */ 8496 if (dup_lock == TRUE) { 8497 /* verify that lock_seqid's match */ 8498 if (lsp->seqid != olo->lock_seqid) { 8499 NFS4_DEBUG(rfs4_debug, 8500 (CE_NOTE, "rfs4_op_lock: Dup-Lock seqid bad" 8501 "lsp->seqid=%d old->seqid=%d", 8502 lsp->seqid, olo->lock_seqid)); 8503 *cs->statusp = resp->status = NFS4ERR_BAD_SEQID; 8504 } else { 8505 rfs4_copy_reply(resop, lsp->reply); 8506 /* 8507 * Make sure to copy the just 8508 * retrieved reply status into the 8509 * overall compound status 8510 */ 8511 *cs->statusp = resp->status; 8512 } 8513 rfs4_lockowner_rele(lo); 8514 goto end; 8515 } 8516 8517 rfs4_dbe_lock(lsp->dbe); 8518 8519 /* Make sure to update the lock sequence id */ 8520 lsp->seqid = olo->lock_seqid; 8521 8522 NFS4_DEBUG(rfs4_debug, 8523 (CE_NOTE, "Lock seqid established as %d", lsp->seqid)); 8524 8525 /* 8526 * This is used to signify the newly created lockowner 8527 * stateid and its sequence number. The checks for 8528 * sequence number and increment don't occur on the 8529 * very first lock request for a lockowner. 8530 */ 8531 lsp->skip_seqid_check = TRUE; 8532 8533 /* hold off other access to lsp while we tinker */ 8534 rfs4_sw_enter(&lsp->ls_sw); 8535 ls_sw_held = TRUE; 8536 8537 rfs4_dbe_unlock(lsp->dbe); 8538 8539 rfs4_lockowner_rele(lo); 8540 } else { 8541 stateid = &args->locker.locker4_u.lock_owner.lock_stateid; 8542 /* get lsp and hold the lock on the underlying file struct */ 8543 if ((status = rfs4_get_lo_state(stateid, &lsp, TRUE)) 8544 != NFS4_OK) { 8545 *cs->statusp = resp->status = status; 8546 DTRACE_NFSV4_2(op__lock__done, struct compound_state *, 8547 cs, LOCK4res *, resp); 8548 return; 8549 } 8550 create = FALSE; /* We didn't create lsp */ 8551 8552 /* Ensure specified filehandle matches */ 8553 if (cs->vp != lsp->state->finfo->vp) { 8554 rfs4_lo_state_rele(lsp, TRUE); 8555 *cs->statusp = resp->status = NFS4ERR_BAD_STATEID; 8556 DTRACE_NFSV4_2(op__lock__done, struct compound_state *, 8557 cs, LOCK4res *, resp); 8558 return; 8559 } 8560 8561 /* hold off other access to lsp while we tinker */ 8562 rfs4_sw_enter(&lsp->ls_sw); 8563 ls_sw_held = TRUE; 8564 8565 switch (rfs4_check_lo_stateid_seqid(lsp, stateid)) { 8566 /* 8567 * The stateid looks like it was okay (expected to be 8568 * the next one) 8569 */ 8570 case NFS4_CHECK_STATEID_OKAY: 8571 /* 8572 * The sequence id is now checked. Determine 8573 * if this is a replay or if it is in the 8574 * expected (next) sequence. In the case of a 8575 * replay, there are two replay conditions 8576 * that may occur. The first is the normal 8577 * condition where a LOCK is done with a 8578 * NFS4_OK response and the stateid is 8579 * updated. That case is handled below when 8580 * the stateid is identified as a REPLAY. The 8581 * second is the case where an error is 8582 * returned, like NFS4ERR_DENIED, and the 8583 * sequence number is updated but the stateid 8584 * is not updated. This second case is dealt 8585 * with here. So it may seem odd that the 8586 * stateid is okay but the sequence id is a 8587 * replay but it is okay. 8588 */ 8589 switch (rfs4_check_lock_seqid( 8590 args->locker.locker4_u.lock_owner.lock_seqid, 8591 lsp, resop)) { 8592 case NFS4_CHKSEQ_REPLAY: 8593 if (resp->status != NFS4_OK) { 8594 /* 8595 * Here is our replay and need 8596 * to verify that the last 8597 * response was an error. 8598 */ 8599 *cs->statusp = resp->status; 8600 goto end; 8601 } 8602 /* 8603 * This is done since the sequence id 8604 * looked like a replay but it didn't 8605 * pass our check so a BAD_SEQID is 8606 * returned as a result. 8607 */ 8608 /*FALLTHROUGH*/ 8609 case NFS4_CHKSEQ_BAD: 8610 *cs->statusp = resp->status = NFS4ERR_BAD_SEQID; 8611 goto end; 8612 case NFS4_CHKSEQ_OKAY: 8613 /* Everything looks okay move ahead */ 8614 break; 8615 } 8616 break; 8617 case NFS4_CHECK_STATEID_OLD: 8618 *cs->statusp = resp->status = NFS4ERR_OLD_STATEID; 8619 goto end; 8620 case NFS4_CHECK_STATEID_BAD: 8621 *cs->statusp = resp->status = NFS4ERR_BAD_STATEID; 8622 goto end; 8623 case NFS4_CHECK_STATEID_EXPIRED: 8624 *cs->statusp = resp->status = NFS4ERR_EXPIRED; 8625 goto end; 8626 case NFS4_CHECK_STATEID_CLOSED: 8627 *cs->statusp = resp->status = NFS4ERR_OLD_STATEID; 8628 goto end; 8629 case NFS4_CHECK_STATEID_REPLAY: 8630 switch (rfs4_check_lock_seqid( 8631 args->locker.locker4_u.lock_owner.lock_seqid, 8632 lsp, resop)) { 8633 case NFS4_CHKSEQ_OKAY: 8634 /* 8635 * This is a replayed stateid; if 8636 * seqid matches the next expected, 8637 * then client is using wrong seqid. 8638 */ 8639 case NFS4_CHKSEQ_BAD: 8640 *cs->statusp = resp->status = NFS4ERR_BAD_SEQID; 8641 goto end; 8642 case NFS4_CHKSEQ_REPLAY: 8643 rfs4_update_lease(lsp->locker->client); 8644 *cs->statusp = status = resp->status; 8645 goto end; 8646 } 8647 break; 8648 default: 8649 ASSERT(FALSE); 8650 break; 8651 } 8652 8653 rfs4_update_lock_sequence(lsp); 8654 rfs4_update_lease(lsp->locker->client); 8655 } 8656 8657 /* 8658 * NFS4 only allows locking on regular files, so 8659 * verify type of object. 8660 */ 8661 if (cs->vp->v_type != VREG) { 8662 if (cs->vp->v_type == VDIR) 8663 status = NFS4ERR_ISDIR; 8664 else 8665 status = NFS4ERR_INVAL; 8666 goto out; 8667 } 8668 8669 cp = lsp->state->owner->client; 8670 8671 if (rfs4_clnt_in_grace(cp) && !args->reclaim) { 8672 status = NFS4ERR_GRACE; 8673 goto out; 8674 } 8675 8676 if (rfs4_clnt_in_grace(cp) && args->reclaim && !cp->can_reclaim) { 8677 status = NFS4ERR_NO_GRACE; 8678 goto out; 8679 } 8680 8681 if (!rfs4_clnt_in_grace(cp) && args->reclaim) { 8682 status = NFS4ERR_NO_GRACE; 8683 goto out; 8684 } 8685 8686 if (lsp->state->finfo->dinfo->dtype == OPEN_DELEGATE_WRITE) 8687 cs->deleg = TRUE; 8688 8689 status = rfs4_do_lock(lsp, args->locktype, 8690 args->locker.locker4_u.lock_owner.lock_seqid, 8691 args->offset, args->length, cs->cr, resop); 8692 8693 out: 8694 lsp->skip_seqid_check = FALSE; 8695 8696 *cs->statusp = resp->status = status; 8697 8698 if (status == NFS4_OK) { 8699 resp->LOCK4res_u.lock_stateid = lsp->lockid.stateid; 8700 lsp->lock_completed = TRUE; 8701 } 8702 /* 8703 * Only update the "OPEN" response here if this was a new 8704 * lock_owner 8705 */ 8706 if (sp) 8707 rfs4_update_open_resp(sp->owner, resop, NULL); 8708 8709 rfs4_update_lock_resp(lsp, resop); 8710 8711 end: 8712 if (lsp) { 8713 if (ls_sw_held) 8714 rfs4_sw_exit(&lsp->ls_sw); 8715 /* 8716 * If an sp obtained, then the lsp does not represent 8717 * a lock on the file struct. 8718 */ 8719 if (sp != NULL) 8720 rfs4_lo_state_rele(lsp, FALSE); 8721 else 8722 rfs4_lo_state_rele(lsp, TRUE); 8723 } 8724 if (sp) { 8725 rfs4_sw_exit(&sp->owner->oo_sw); 8726 rfs4_state_rele(sp); 8727 } 8728 8729 DTRACE_NFSV4_2(op__lock__done, struct compound_state *, cs, 8730 LOCK4res *, resp); 8731 } 8732 8733 /* free function for LOCK/LOCKT */ 8734 static void 8735 lock_denied_free(nfs_resop4 *resop) 8736 { 8737 LOCK4denied *dp = NULL; 8738 8739 switch (resop->resop) { 8740 case OP_LOCK: 8741 if (resop->nfs_resop4_u.oplock.status == NFS4ERR_DENIED) 8742 dp = &resop->nfs_resop4_u.oplock.LOCK4res_u.denied; 8743 break; 8744 case OP_LOCKT: 8745 if (resop->nfs_resop4_u.oplockt.status == NFS4ERR_DENIED) 8746 dp = &resop->nfs_resop4_u.oplockt.denied; 8747 break; 8748 default: 8749 break; 8750 } 8751 8752 if (dp) 8753 kmem_free(dp->owner.owner_val, dp->owner.owner_len); 8754 } 8755 8756 /*ARGSUSED*/ 8757 void 8758 rfs4_op_locku(nfs_argop4 *argop, nfs_resop4 *resop, 8759 struct svc_req *req, struct compound_state *cs) 8760 { 8761 LOCKU4args *args = &argop->nfs_argop4_u.oplocku; 8762 LOCKU4res *resp = &resop->nfs_resop4_u.oplocku; 8763 nfsstat4 status; 8764 stateid4 *stateid = &args->lock_stateid; 8765 rfs4_lo_state_t *lsp; 8766 8767 DTRACE_NFSV4_2(op__locku__start, struct compound_state *, cs, 8768 LOCKU4args *, args); 8769 8770 if (cs->vp == NULL) { 8771 *cs->statusp = resp->status = NFS4ERR_NOFILEHANDLE; 8772 DTRACE_NFSV4_2(op__locku__done, struct compound_state *, cs, 8773 LOCKU4res *, resp); 8774 return; 8775 } 8776 8777 if ((status = rfs4_get_lo_state(stateid, &lsp, TRUE)) != NFS4_OK) { 8778 *cs->statusp = resp->status = status; 8779 DTRACE_NFSV4_2(op__locku__done, struct compound_state *, cs, 8780 LOCKU4res *, resp); 8781 return; 8782 } 8783 8784 /* Ensure specified filehandle matches */ 8785 if (cs->vp != lsp->state->finfo->vp) { 8786 rfs4_lo_state_rele(lsp, TRUE); 8787 *cs->statusp = resp->status = NFS4ERR_BAD_STATEID; 8788 DTRACE_NFSV4_2(op__locku__done, struct compound_state *, cs, 8789 LOCKU4res *, resp); 8790 return; 8791 } 8792 8793 /* hold off other access to lsp while we tinker */ 8794 rfs4_sw_enter(&lsp->ls_sw); 8795 8796 switch (rfs4_check_lo_stateid_seqid(lsp, stateid)) { 8797 case NFS4_CHECK_STATEID_OKAY: 8798 if (rfs4_check_lock_seqid(args->seqid, lsp, resop) 8799 != NFS4_CHKSEQ_OKAY) { 8800 *cs->statusp = resp->status = NFS4ERR_BAD_SEQID; 8801 goto end; 8802 } 8803 break; 8804 case NFS4_CHECK_STATEID_OLD: 8805 *cs->statusp = resp->status = NFS4ERR_OLD_STATEID; 8806 goto end; 8807 case NFS4_CHECK_STATEID_BAD: 8808 *cs->statusp = resp->status = NFS4ERR_BAD_STATEID; 8809 goto end; 8810 case NFS4_CHECK_STATEID_EXPIRED: 8811 *cs->statusp = resp->status = NFS4ERR_EXPIRED; 8812 goto end; 8813 case NFS4_CHECK_STATEID_CLOSED: 8814 *cs->statusp = resp->status = NFS4ERR_OLD_STATEID; 8815 goto end; 8816 case NFS4_CHECK_STATEID_REPLAY: 8817 switch (rfs4_check_lock_seqid(args->seqid, lsp, resop)) { 8818 case NFS4_CHKSEQ_OKAY: 8819 /* 8820 * This is a replayed stateid; if 8821 * seqid matches the next expected, 8822 * then client is using wrong seqid. 8823 */ 8824 case NFS4_CHKSEQ_BAD: 8825 *cs->statusp = resp->status = NFS4ERR_BAD_SEQID; 8826 goto end; 8827 case NFS4_CHKSEQ_REPLAY: 8828 rfs4_update_lease(lsp->locker->client); 8829 *cs->statusp = status = resp->status; 8830 goto end; 8831 } 8832 break; 8833 default: 8834 ASSERT(FALSE); 8835 break; 8836 } 8837 8838 rfs4_update_lock_sequence(lsp); 8839 rfs4_update_lease(lsp->locker->client); 8840 8841 /* 8842 * NFS4 only allows locking on regular files, so 8843 * verify type of object. 8844 */ 8845 if (cs->vp->v_type != VREG) { 8846 if (cs->vp->v_type == VDIR) 8847 status = NFS4ERR_ISDIR; 8848 else 8849 status = NFS4ERR_INVAL; 8850 goto out; 8851 } 8852 8853 if (rfs4_clnt_in_grace(lsp->state->owner->client)) { 8854 status = NFS4ERR_GRACE; 8855 goto out; 8856 } 8857 8858 status = rfs4_do_lock(lsp, args->locktype, 8859 args->seqid, args->offset, args->length, cs->cr, resop); 8860 8861 out: 8862 *cs->statusp = resp->status = status; 8863 8864 if (status == NFS4_OK) 8865 resp->lock_stateid = lsp->lockid.stateid; 8866 8867 rfs4_update_lock_resp(lsp, resop); 8868 8869 end: 8870 rfs4_sw_exit(&lsp->ls_sw); 8871 rfs4_lo_state_rele(lsp, TRUE); 8872 8873 DTRACE_NFSV4_2(op__locku__done, struct compound_state *, cs, 8874 LOCKU4res *, resp); 8875 } 8876 8877 /* 8878 * LOCKT is a best effort routine, the client can not be guaranteed that 8879 * the status return is still in effect by the time the reply is received. 8880 * They are numerous race conditions in this routine, but we are not required 8881 * and can not be accurate. 8882 */ 8883 /*ARGSUSED*/ 8884 void 8885 rfs4_op_lockt(nfs_argop4 *argop, nfs_resop4 *resop, 8886 struct svc_req *req, struct compound_state *cs) 8887 { 8888 LOCKT4args *args = &argop->nfs_argop4_u.oplockt; 8889 LOCKT4res *resp = &resop->nfs_resop4_u.oplockt; 8890 rfs4_lockowner_t *lo; 8891 rfs4_client_t *cp; 8892 bool_t create = FALSE; 8893 struct flock64 flk; 8894 int error; 8895 int flag = FREAD | FWRITE; 8896 int ltype; 8897 length4 posix_length; 8898 sysid_t sysid; 8899 pid_t pid; 8900 8901 DTRACE_NFSV4_2(op__lockt__start, struct compound_state *, cs, 8902 LOCKT4args *, args); 8903 8904 if (cs->vp == NULL) { 8905 *cs->statusp = resp->status = NFS4ERR_NOFILEHANDLE; 8906 goto out; 8907 } 8908 8909 /* 8910 * NFS4 only allows locking on regular files, so 8911 * verify type of object. 8912 */ 8913 if (cs->vp->v_type != VREG) { 8914 if (cs->vp->v_type == VDIR) 8915 *cs->statusp = resp->status = NFS4ERR_ISDIR; 8916 else 8917 *cs->statusp = resp->status = NFS4ERR_INVAL; 8918 goto out; 8919 } 8920 8921 /* 8922 * Check out the clientid to ensure the server knows about it 8923 * so that we correctly inform the client of a server reboot. 8924 */ 8925 if ((cp = rfs4_findclient_by_id(args->owner.clientid, FALSE)) 8926 == NULL) { 8927 *cs->statusp = resp->status = 8928 rfs4_check_clientid(&args->owner.clientid, 0); 8929 goto out; 8930 } 8931 if (rfs4_lease_expired(cp)) { 8932 rfs4_client_close(cp); 8933 /* 8934 * Protocol doesn't allow returning NFS4ERR_STALE as 8935 * other operations do on this check so STALE_CLIENTID 8936 * is returned instead 8937 */ 8938 *cs->statusp = resp->status = NFS4ERR_STALE_CLIENTID; 8939 goto out; 8940 } 8941 8942 if (rfs4_clnt_in_grace(cp) && !(cp->can_reclaim)) { 8943 *cs->statusp = resp->status = NFS4ERR_GRACE; 8944 rfs4_client_rele(cp); 8945 goto out; 8946 } 8947 rfs4_client_rele(cp); 8948 8949 resp->status = NFS4_OK; 8950 8951 switch (args->locktype) { 8952 case READ_LT: 8953 case READW_LT: 8954 ltype = F_RDLCK; 8955 break; 8956 case WRITE_LT: 8957 case WRITEW_LT: 8958 ltype = F_WRLCK; 8959 break; 8960 } 8961 8962 posix_length = args->length; 8963 /* Check for zero length. To lock to end of file use all ones for V4 */ 8964 if (posix_length == 0) { 8965 *cs->statusp = resp->status = NFS4ERR_INVAL; 8966 goto out; 8967 } else if (posix_length == (length4)(~0)) { 8968 posix_length = 0; /* Posix to end of file */ 8969 } 8970 8971 /* Find or create a lockowner */ 8972 lo = rfs4_findlockowner(&args->owner, &create); 8973 8974 if (lo) { 8975 pid = lo->pid; 8976 if ((resp->status = 8977 rfs4_client_sysid(lo->client, &sysid)) != NFS4_OK) 8978 goto err; 8979 } else { 8980 pid = 0; 8981 sysid = lockt_sysid; 8982 } 8983 retry: 8984 flk.l_type = ltype; 8985 flk.l_whence = 0; /* SEEK_SET */ 8986 flk.l_start = args->offset; 8987 flk.l_len = posix_length; 8988 flk.l_sysid = sysid; 8989 flk.l_pid = pid; 8990 flag |= F_REMOTELOCK; 8991 8992 LOCK_PRINT(rfs4_debug, "rfs4_op_lockt", F_GETLK, &flk); 8993 8994 /* Note that length4 is uint64_t but l_len and l_start are off64_t */ 8995 if (flk.l_len < 0 || flk.l_start < 0) { 8996 resp->status = NFS4ERR_INVAL; 8997 goto err; 8998 } 8999 error = VOP_FRLOCK(cs->vp, F_GETLK, &flk, flag, (u_offset_t)0, 9000 NULL, cs->cr, NULL); 9001 9002 /* 9003 * N.B. We map error values to nfsv4 errors. This is differrent 9004 * than puterrno4 routine. 9005 */ 9006 switch (error) { 9007 case 0: 9008 if (flk.l_type == F_UNLCK) 9009 resp->status = NFS4_OK; 9010 else { 9011 if (lock_denied(&resp->denied, &flk) == NFS4ERR_EXPIRED) 9012 goto retry; 9013 resp->status = NFS4ERR_DENIED; 9014 } 9015 break; 9016 case EOVERFLOW: 9017 resp->status = NFS4ERR_INVAL; 9018 break; 9019 case EINVAL: 9020 resp->status = NFS4ERR_NOTSUPP; 9021 break; 9022 default: 9023 cmn_err(CE_WARN, "rfs4_op_lockt: unexpected errno (%d)", 9024 error); 9025 resp->status = NFS4ERR_SERVERFAULT; 9026 break; 9027 } 9028 9029 err: 9030 if (lo) 9031 rfs4_lockowner_rele(lo); 9032 *cs->statusp = resp->status; 9033 out: 9034 DTRACE_NFSV4_2(op__lockt__done, struct compound_state *, cs, 9035 LOCKT4res *, resp); 9036 } 9037 9038 static int 9039 vop_shrlock(vnode_t *vp, int cmd, struct shrlock *sp, int fflags) 9040 { 9041 int err; 9042 9043 if (cmd == F_UNSHARE && sp->s_deny == 0 && sp->s_access == 0) 9044 return (0); 9045 9046 err = VOP_SHRLOCK(vp, cmd, sp, fflags, CRED(), NULL); 9047 9048 NFS4_DEBUG(rfs4_shrlock_debug, 9049 (CE_NOTE, "rfs4_shrlock %s vp=%p acc=%d dny=%d sysid=%d " 9050 "pid=%d err=%d\n", cmd == F_UNSHARE ? "UNSHR" : "SHARE", 9051 (void *) vp, sp->s_access, sp->s_deny, sp->s_sysid, sp->s_pid, 9052 err)); 9053 9054 return (err); 9055 } 9056 9057 static int 9058 rfs4_shrlock(rfs4_state_t *sp, int cmd) 9059 { 9060 struct shrlock shr; 9061 struct shr_locowner shr_loco; 9062 int fflags; 9063 9064 fflags = shr.s_access = shr.s_deny = 0; 9065 9066 if (sp->share_access & OPEN4_SHARE_ACCESS_READ) { 9067 fflags |= FREAD; 9068 shr.s_access |= F_RDACC; 9069 } 9070 if (sp->share_access & OPEN4_SHARE_ACCESS_WRITE) { 9071 fflags |= FWRITE; 9072 shr.s_access |= F_WRACC; 9073 } 9074 if (sp->share_deny & OPEN4_SHARE_DENY_READ) 9075 shr.s_deny |= F_RDDNY; 9076 if (sp->share_deny & OPEN4_SHARE_DENY_WRITE) 9077 shr.s_deny |= F_WRDNY; 9078 9079 shr.s_pid = rfs4_dbe_getid(sp->owner->dbe); 9080 shr.s_sysid = sp->owner->client->sysidt; 9081 shr_loco.sl_pid = shr.s_pid; 9082 shr_loco.sl_id = shr.s_sysid; 9083 shr.s_owner = (caddr_t)&shr_loco; 9084 shr.s_own_len = sizeof (shr_loco); 9085 return (vop_shrlock(sp->finfo->vp, cmd, &shr, fflags)); 9086 } 9087 9088 static int 9089 rfs4_share(rfs4_state_t *sp) 9090 { 9091 int cmd; 9092 cmd = nbl_need_check(sp->finfo->vp) ? F_SHARE_NBMAND : F_SHARE; 9093 return (rfs4_shrlock(sp, cmd)); 9094 } 9095 9096 void 9097 rfs4_unshare(rfs4_state_t *sp) 9098 { 9099 (void) rfs4_shrlock(sp, F_UNSHARE); 9100 } 9101 9102 static int 9103 rdma_setup_read_data4(READ4args * args, READ4res * rok) 9104 { 9105 struct clist *wcl; 9106 int data_len, avail_len, num; 9107 count4 count = rok->data_len; 9108 9109 data_len = num = avail_len = 0; 9110 9111 wcl = args->wlist; 9112 while (wcl != NULL) { 9113 if (wcl->c_dmemhandle.mrc_rmr == 0) 9114 break; 9115 9116 avail_len += wcl->c_len; 9117 if (wcl->c_len < count) { 9118 data_len += wcl->c_len; 9119 } else { 9120 /* Can make the rest chunks all 0-len */ 9121 data_len += count; 9122 wcl->c_len = count; 9123 } 9124 count -= wcl->c_len; 9125 num++; 9126 wcl = wcl->c_next; 9127 } 9128 9129 /* 9130 * MUST fail if there are still more data 9131 */ 9132 if (count > 0) { 9133 DTRACE_PROBE2(nfss__e__read4_wlist_fail, 9134 int, data_len, int, count); 9135 return (FALSE); 9136 } 9137 wcl = args->wlist; 9138 rok->data_len = data_len; 9139 rok->wlist_len = data_len; 9140 rok->wlist = wcl; 9141 9142 return (TRUE); 9143 } 9144