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 * Use is subject to license terms. 30 */ 31 32 #include <sys/param.h> 33 #include <sys/types.h> 34 #include <sys/systm.h> 35 #include <sys/cred.h> 36 #include <sys/proc.h> 37 #include <sys/user.h> 38 #include <sys/buf.h> 39 #include <sys/vfs.h> 40 #include <sys/vnode.h> 41 #include <sys/pathname.h> 42 #include <sys/uio.h> 43 #include <sys/file.h> 44 #include <sys/stat.h> 45 #include <sys/errno.h> 46 #include <sys/socket.h> 47 #include <sys/sysmacros.h> 48 #include <sys/siginfo.h> 49 #include <sys/tiuser.h> 50 #include <sys/statvfs.h> 51 #include <sys/stream.h> 52 #include <sys/strsubr.h> 53 #include <sys/stropts.h> 54 #include <sys/timod.h> 55 #include <sys/t_kuser.h> 56 #include <sys/kmem.h> 57 #include <sys/kstat.h> 58 #include <sys/dirent.h> 59 #include <sys/cmn_err.h> 60 #include <sys/debug.h> 61 #include <sys/unistd.h> 62 #include <sys/vtrace.h> 63 #include <sys/mode.h> 64 #include <sys/acl.h> 65 #include <sys/sdt.h> 66 67 #include <rpc/types.h> 68 #include <rpc/auth.h> 69 #include <rpc/auth_unix.h> 70 #include <rpc/auth_des.h> 71 #include <rpc/svc.h> 72 #include <rpc/xdr.h> 73 74 #include <nfs/nfs.h> 75 #include <nfs/export.h> 76 #include <nfs/nfssys.h> 77 #include <nfs/nfs_clnt.h> 78 #include <nfs/nfs_acl.h> 79 #include <nfs/nfs_log.h> 80 #include <nfs/nfs_cmd.h> 81 #include <nfs/lm.h> 82 #include <nfs/nfs_dispatch.h> 83 #include <nfs/nfs4_drc.h> 84 85 #include <sys/modctl.h> 86 #include <sys/cladm.h> 87 #include <sys/clconf.h> 88 89 #define MAXHOST 32 90 const char *kinet_ntop6(uchar_t *, char *, size_t); 91 92 /* 93 * Module linkage information. 94 */ 95 96 static struct modlmisc modlmisc = { 97 &mod_miscops, "NFS server module" 98 }; 99 100 static struct modlinkage modlinkage = { 101 MODREV_1, (void *)&modlmisc, NULL 102 }; 103 104 char _depends_on[] = "misc/klmmod"; 105 106 int 107 _init(void) 108 { 109 int status; 110 111 if ((status = nfs_srvinit()) != 0) { 112 cmn_err(CE_WARN, "_init: nfs_srvinit failed"); 113 return (status); 114 } 115 116 status = mod_install((struct modlinkage *)&modlinkage); 117 if (status != 0) { 118 /* 119 * Could not load module, cleanup previous 120 * initialization work. 121 */ 122 nfs_srvfini(); 123 } 124 125 /* 126 * Initialise some placeholders for nfssys() calls. These have 127 * to be declared by the nfs module, since that handles nfssys() 128 * calls - also used by NFS clients - but are provided by this 129 * nfssrv module. These also then serve as confirmation to the 130 * relevant code in nfs that nfssrv has been loaded, as they're 131 * initially NULL. 132 */ 133 nfs_srv_quiesce_func = nfs_srv_quiesce_all; 134 nfs_srv_dss_func = rfs4_dss_setpaths; 135 136 /* setup DSS paths here; must be done before initial server startup */ 137 rfs4_dss_paths = rfs4_dss_oldpaths = NULL; 138 139 return (status); 140 } 141 142 int 143 _fini() 144 { 145 return (EBUSY); 146 } 147 148 int 149 _info(struct modinfo *modinfop) 150 { 151 return (mod_info(&modlinkage, modinfop)); 152 } 153 154 /* 155 * PUBLICFH_CHECK() checks if the dispatch routine supports 156 * RPC_PUBLICFH_OK, if the filesystem is exported public, and if the 157 * incoming request is using the public filehandle. The check duplicates 158 * the exportmatch() call done in checkexport(), and we should consider 159 * modifying those routines to avoid the duplication. For now, we optimize 160 * by calling exportmatch() only after checking that the dispatch routine 161 * supports RPC_PUBLICFH_OK, and if the filesystem is explicitly exported 162 * public (i.e., not the placeholder). 163 */ 164 #define PUBLICFH_CHECK(disp, exi, fsid, xfid) \ 165 ((disp->dis_flags & RPC_PUBLICFH_OK) && \ 166 ((exi->exi_export.ex_flags & EX_PUBLIC) || \ 167 (exi == exi_public && exportmatch(exi_root, \ 168 fsid, xfid)))) 169 170 static void nfs_srv_shutdown_all(int); 171 static void rfs4_server_start(int); 172 static void nullfree(void); 173 static void rfs_dispatch(struct svc_req *, SVCXPRT *); 174 static void acl_dispatch(struct svc_req *, SVCXPRT *); 175 static void common_dispatch(struct svc_req *, SVCXPRT *, 176 rpcvers_t, rpcvers_t, char *, 177 struct rpc_disptable *); 178 static void hanfsv4_failover(void); 179 static int checkauth(struct exportinfo *, struct svc_req *, cred_t *, int, 180 bool_t); 181 static char *client_name(struct svc_req *req); 182 static char *client_addr(struct svc_req *req, char *buf); 183 extern int sec_svc_getcred(struct svc_req *, cred_t *cr, char **, int *); 184 extern bool_t sec_svc_inrootlist(int, caddr_t, int, caddr_t *); 185 186 #define NFSLOG_COPY_NETBUF(exi, xprt, nb) { \ 187 (nb)->maxlen = (xprt)->xp_rtaddr.maxlen; \ 188 (nb)->len = (xprt)->xp_rtaddr.len; \ 189 (nb)->buf = kmem_alloc((nb)->len, KM_SLEEP); \ 190 bcopy((xprt)->xp_rtaddr.buf, (nb)->buf, (nb)->len); \ 191 } 192 193 /* 194 * Public Filehandle common nfs routines 195 */ 196 static int MCLpath(char **); 197 static void URLparse(char *); 198 199 /* 200 * NFS callout table. 201 * This table is used by svc_getreq() to dispatch a request with 202 * a given prog/vers pair to an appropriate service provider 203 * dispatch routine. 204 * 205 * NOTE: ordering is relied upon below when resetting the version min/max 206 * for NFS_PROGRAM. Careful, if this is ever changed. 207 */ 208 static SVC_CALLOUT __nfs_sc_clts[] = { 209 { NFS_PROGRAM, NFS_VERSMIN, NFS_VERSMAX, rfs_dispatch }, 210 { NFS_ACL_PROGRAM, NFS_ACL_VERSMIN, NFS_ACL_VERSMAX, acl_dispatch } 211 }; 212 213 static SVC_CALLOUT_TABLE nfs_sct_clts = { 214 sizeof (__nfs_sc_clts) / sizeof (__nfs_sc_clts[0]), FALSE, 215 __nfs_sc_clts 216 }; 217 218 static SVC_CALLOUT __nfs_sc_cots[] = { 219 { NFS_PROGRAM, NFS_VERSMIN, NFS_VERSMAX, rfs_dispatch }, 220 { NFS_ACL_PROGRAM, NFS_ACL_VERSMIN, NFS_ACL_VERSMAX, acl_dispatch } 221 }; 222 223 static SVC_CALLOUT_TABLE nfs_sct_cots = { 224 sizeof (__nfs_sc_cots) / sizeof (__nfs_sc_cots[0]), FALSE, __nfs_sc_cots 225 }; 226 227 static SVC_CALLOUT __nfs_sc_rdma[] = { 228 { NFS_PROGRAM, NFS_VERSMIN, NFS_VERSMAX, rfs_dispatch }, 229 { NFS_ACL_PROGRAM, NFS_ACL_VERSMIN, NFS_ACL_VERSMAX, acl_dispatch } 230 }; 231 232 static SVC_CALLOUT_TABLE nfs_sct_rdma = { 233 sizeof (__nfs_sc_rdma) / sizeof (__nfs_sc_rdma[0]), FALSE, __nfs_sc_rdma 234 }; 235 rpcvers_t nfs_versmin = NFS_VERSMIN_DEFAULT; 236 rpcvers_t nfs_versmax = NFS_VERSMAX_DEFAULT; 237 238 /* 239 * Used to track the state of the server so that initialization 240 * can be done properly. 241 */ 242 typedef enum { 243 NFS_SERVER_STOPPED, /* server state destroyed */ 244 NFS_SERVER_STOPPING, /* server state being destroyed */ 245 NFS_SERVER_RUNNING, 246 NFS_SERVER_QUIESCED, /* server state preserved */ 247 NFS_SERVER_OFFLINE /* server pool offline */ 248 } nfs_server_running_t; 249 250 static nfs_server_running_t nfs_server_upordown; 251 static kmutex_t nfs_server_upordown_lock; 252 static kcondvar_t nfs_server_upordown_cv; 253 254 /* 255 * DSS: distributed stable storage 256 * lists of all DSS paths: current, and before last warmstart 257 */ 258 nvlist_t *rfs4_dss_paths, *rfs4_dss_oldpaths; 259 260 int rfs4_dispatch(struct rpcdisp *, struct svc_req *, SVCXPRT *, char *); 261 bool_t rfs4_minorvers_mismatch(struct svc_req *, SVCXPRT *, void *); 262 263 /* 264 * RDMA wait variables. 265 */ 266 static kcondvar_t rdma_wait_cv; 267 static kmutex_t rdma_wait_mutex; 268 269 /* 270 * Will be called at the point the server pool is being unregistered 271 * from the pool list. From that point onwards, the pool is waiting 272 * to be drained and as such the server state is stale and pertains 273 * to the old instantiation of the NFS server pool. 274 */ 275 void 276 nfs_srv_offline(void) 277 { 278 mutex_enter(&nfs_server_upordown_lock); 279 if (nfs_server_upordown == NFS_SERVER_RUNNING) { 280 nfs_server_upordown = NFS_SERVER_OFFLINE; 281 } 282 mutex_exit(&nfs_server_upordown_lock); 283 } 284 285 /* 286 * Will be called at the point the server pool is being destroyed so 287 * all transports have been closed and no service threads are in 288 * existence. 289 * 290 * If we quiesce the server, we're shutting it down without destroying the 291 * server state. This allows it to warm start subsequently. 292 */ 293 void 294 nfs_srv_stop_all(void) 295 { 296 int quiesce = 0; 297 nfs_srv_shutdown_all(quiesce); 298 } 299 300 /* 301 * This alternative shutdown routine can be requested via nfssys() 302 */ 303 void 304 nfs_srv_quiesce_all(void) 305 { 306 int quiesce = 1; 307 nfs_srv_shutdown_all(quiesce); 308 } 309 310 static void 311 nfs_srv_shutdown_all(int quiesce) { 312 mutex_enter(&nfs_server_upordown_lock); 313 if (quiesce) { 314 if (nfs_server_upordown == NFS_SERVER_RUNNING || 315 nfs_server_upordown == NFS_SERVER_OFFLINE) { 316 nfs_server_upordown = NFS_SERVER_QUIESCED; 317 cv_signal(&nfs_server_upordown_cv); 318 319 /* reset DSS state, for subsequent warm restart */ 320 rfs4_dss_numnewpaths = 0; 321 rfs4_dss_newpaths = NULL; 322 323 cmn_err(CE_NOTE, "nfs_server: server is now quiesced; " 324 "NFSv4 state has been preserved"); 325 } 326 } else { 327 if (nfs_server_upordown == NFS_SERVER_OFFLINE) { 328 nfs_server_upordown = NFS_SERVER_STOPPING; 329 mutex_exit(&nfs_server_upordown_lock); 330 rfs4_state_fini(); 331 rfs4_fini_drc(nfs4_drc); 332 mutex_enter(&nfs_server_upordown_lock); 333 nfs_server_upordown = NFS_SERVER_STOPPED; 334 cv_signal(&nfs_server_upordown_cv); 335 } 336 } 337 mutex_exit(&nfs_server_upordown_lock); 338 } 339 340 static int 341 nfs_srv_set_sc_versions(struct file *fp, SVC_CALLOUT_TABLE **sctpp, 342 rpcvers_t versmin, rpcvers_t versmax) 343 { 344 struct strioctl strioc; 345 struct T_info_ack tinfo; 346 int error, retval; 347 348 /* 349 * Find out what type of transport this is. 350 */ 351 strioc.ic_cmd = TI_GETINFO; 352 strioc.ic_timout = -1; 353 strioc.ic_len = sizeof (tinfo); 354 strioc.ic_dp = (char *)&tinfo; 355 tinfo.PRIM_type = T_INFO_REQ; 356 357 error = strioctl(fp->f_vnode, I_STR, (intptr_t)&strioc, 0, K_TO_K, 358 CRED(), &retval); 359 if (error || retval) 360 return (error); 361 362 /* 363 * Based on our query of the transport type... 364 * 365 * Reset the min/max versions based on the caller's request 366 * NOTE: This assumes that NFS_PROGRAM is first in the array!! 367 * And the second entry is the NFS_ACL_PROGRAM. 368 */ 369 switch (tinfo.SERV_type) { 370 case T_CLTS: 371 if (versmax == NFS_V4) 372 return (EINVAL); 373 __nfs_sc_clts[0].sc_versmin = versmin; 374 __nfs_sc_clts[0].sc_versmax = versmax; 375 __nfs_sc_clts[1].sc_versmin = versmin; 376 __nfs_sc_clts[1].sc_versmax = versmax; 377 *sctpp = &nfs_sct_clts; 378 break; 379 case T_COTS: 380 case T_COTS_ORD: 381 __nfs_sc_cots[0].sc_versmin = versmin; 382 __nfs_sc_cots[0].sc_versmax = versmax; 383 /* For the NFS_ACL program, check the max version */ 384 if (versmax > NFS_ACL_VERSMAX) 385 versmax = NFS_ACL_VERSMAX; 386 __nfs_sc_cots[1].sc_versmin = versmin; 387 __nfs_sc_cots[1].sc_versmax = versmax; 388 *sctpp = &nfs_sct_cots; 389 break; 390 default: 391 error = EINVAL; 392 } 393 394 return (error); 395 } 396 397 /* 398 * NFS Server system call. 399 * Does all of the work of running a NFS server. 400 * uap->fd is the fd of an open transport provider 401 */ 402 int 403 nfs_svc(struct nfs_svc_args *arg, model_t model) 404 { 405 file_t *fp; 406 SVCMASTERXPRT *xprt; 407 int error; 408 int readsize; 409 char buf[KNC_STRSIZE]; 410 size_t len; 411 STRUCT_HANDLE(nfs_svc_args, uap); 412 struct netbuf addrmask; 413 SVC_CALLOUT_TABLE *sctp = NULL; 414 415 #ifdef lint 416 model = model; /* STRUCT macros don't always refer to it */ 417 #endif 418 419 STRUCT_SET_HANDLE(uap, model, arg); 420 421 /* Check privileges in nfssys() */ 422 423 if ((fp = getf(STRUCT_FGET(uap, fd))) == NULL) 424 return (EBADF); 425 426 /* 427 * Set read buffer size to rsize 428 * and add room for RPC headers. 429 */ 430 readsize = nfs3tsize() + (RPC_MAXDATASIZE - NFS_MAXDATA); 431 if (readsize < RPC_MAXDATASIZE) 432 readsize = RPC_MAXDATASIZE; 433 434 error = copyinstr((const char *)STRUCT_FGETP(uap, netid), buf, 435 KNC_STRSIZE, &len); 436 if (error) { 437 releasef(STRUCT_FGET(uap, fd)); 438 return (error); 439 } 440 441 addrmask.len = STRUCT_FGET(uap, addrmask.len); 442 addrmask.maxlen = STRUCT_FGET(uap, addrmask.maxlen); 443 addrmask.buf = kmem_alloc(addrmask.maxlen, KM_SLEEP); 444 error = copyin(STRUCT_FGETP(uap, addrmask.buf), addrmask.buf, 445 addrmask.len); 446 if (error) { 447 releasef(STRUCT_FGET(uap, fd)); 448 kmem_free(addrmask.buf, addrmask.maxlen); 449 return (error); 450 } 451 452 nfs_versmin = STRUCT_FGET(uap, versmin); 453 nfs_versmax = STRUCT_FGET(uap, versmax); 454 455 /* Double check the vers min/max ranges */ 456 if ((nfs_versmin > nfs_versmax) || 457 (nfs_versmin < NFS_VERSMIN) || 458 (nfs_versmax > NFS_VERSMAX)) { 459 nfs_versmin = NFS_VERSMIN_DEFAULT; 460 nfs_versmax = NFS_VERSMAX_DEFAULT; 461 } 462 463 if (error = 464 nfs_srv_set_sc_versions(fp, &sctp, nfs_versmin, nfs_versmax)) { 465 releasef(STRUCT_FGET(uap, fd)); 466 kmem_free(addrmask.buf, addrmask.maxlen); 467 return (error); 468 } 469 470 /* Initialize nfsv4 server */ 471 if (nfs_versmax == (rpcvers_t)NFS_V4) 472 rfs4_server_start(STRUCT_FGET(uap, delegation)); 473 474 /* Create a transport handle. */ 475 error = svc_tli_kcreate(fp, readsize, buf, &addrmask, &xprt, 476 sctp, NULL, NFS_SVCPOOL_ID, TRUE); 477 478 if (error) 479 kmem_free(addrmask.buf, addrmask.maxlen); 480 481 releasef(STRUCT_FGET(uap, fd)); 482 483 /* HA-NFSv4: save the cluster nodeid */ 484 if (cluster_bootflags & CLUSTER_BOOTED) 485 lm_global_nlmid = clconf_get_nodeid(); 486 487 return (error); 488 } 489 490 static void 491 rfs4_server_start(int nfs4_srv_delegation) 492 { 493 /* 494 * Determine if the server has previously been "started" and 495 * if not, do the per instance initialization 496 */ 497 mutex_enter(&nfs_server_upordown_lock); 498 499 if (nfs_server_upordown != NFS_SERVER_RUNNING) { 500 /* Do we need to stop and wait on the previous server? */ 501 while (nfs_server_upordown == NFS_SERVER_STOPPING || 502 nfs_server_upordown == NFS_SERVER_OFFLINE) 503 cv_wait(&nfs_server_upordown_cv, 504 &nfs_server_upordown_lock); 505 506 if (nfs_server_upordown != NFS_SERVER_RUNNING) { 507 (void) svc_pool_control(NFS_SVCPOOL_ID, 508 SVCPSET_UNREGISTER_PROC, (void *)&nfs_srv_offline); 509 (void) svc_pool_control(NFS_SVCPOOL_ID, 510 SVCPSET_SHUTDOWN_PROC, (void *)&nfs_srv_stop_all); 511 512 /* is this an nfsd warm start? */ 513 if (nfs_server_upordown == NFS_SERVER_QUIESCED) { 514 cmn_err(CE_NOTE, "nfs_server: " 515 "server was previously quiesced; " 516 "existing NFSv4 state will be re-used"); 517 518 /* 519 * HA-NFSv4: this is also the signal 520 * that a Resource Group failover has 521 * occurred. 522 */ 523 if (cluster_bootflags & CLUSTER_BOOTED) 524 hanfsv4_failover(); 525 } else { 526 /* cold start */ 527 rfs4_state_init(); 528 nfs4_drc = rfs4_init_drc(nfs4_drc_max, 529 nfs4_drc_hash); 530 } 531 532 /* 533 * Check to see if delegation is to be 534 * enabled at the server 535 */ 536 if (nfs4_srv_delegation != FALSE) 537 rfs4_set_deleg_policy(SRV_NORMAL_DELEGATE); 538 539 nfs_server_upordown = NFS_SERVER_RUNNING; 540 } 541 cv_signal(&nfs_server_upordown_cv); 542 } 543 mutex_exit(&nfs_server_upordown_lock); 544 } 545 546 /* 547 * If RDMA device available, 548 * start RDMA listener. 549 */ 550 int 551 rdma_start(struct rdma_svc_args *rsa) 552 { 553 int error; 554 rdma_xprt_group_t started_rdma_xprts; 555 556 /* Double check the vers min/max ranges */ 557 if ((rsa->nfs_versmin > rsa->nfs_versmax) || 558 (rsa->nfs_versmin < NFS_VERSMIN) || 559 (rsa->nfs_versmax > NFS_VERSMAX)) { 560 rsa->nfs_versmin = NFS_VERSMIN_DEFAULT; 561 rsa->nfs_versmax = NFS_VERSMAX_DEFAULT; 562 } 563 nfs_versmin = rsa->nfs_versmin; 564 nfs_versmax = rsa->nfs_versmax; 565 566 /* Set the versions in the callout table */ 567 __nfs_sc_rdma[0].sc_versmin = rsa->nfs_versmin; 568 __nfs_sc_rdma[0].sc_versmax = rsa->nfs_versmax; 569 /* For the NFS_ACL program, check the max version */ 570 __nfs_sc_rdma[1].sc_versmin = rsa->nfs_versmin; 571 if (rsa->nfs_versmax > NFS_ACL_VERSMAX) 572 __nfs_sc_rdma[1].sc_versmax = NFS_ACL_VERSMAX; 573 else 574 __nfs_sc_rdma[1].sc_versmax = rsa->nfs_versmax; 575 576 /* Initialize nfsv4 server */ 577 if (rsa->nfs_versmax == (rpcvers_t)NFS_V4) 578 rfs4_server_start(rsa->delegation); 579 580 started_rdma_xprts.rtg_count = 0; 581 started_rdma_xprts.rtg_listhead = NULL; 582 started_rdma_xprts.rtg_poolid = rsa->poolid; 583 error = svc_rdma_kcreate(rsa->netid, &nfs_sct_rdma, rsa->poolid, 584 &started_rdma_xprts); 585 586 if (error == 0) { 587 mutex_enter(&rdma_wait_mutex); 588 if (!cv_wait_sig(&rdma_wait_cv, &rdma_wait_mutex)) { 589 rdma_stop(started_rdma_xprts); 590 } 591 mutex_exit(&rdma_wait_mutex); 592 } 593 594 return (error); 595 } 596 597 /* ARGSUSED */ 598 void 599 rpc_null(caddr_t *argp, caddr_t *resp) 600 { 601 } 602 603 /* ARGSUSED */ 604 void 605 rpc_null_v3(caddr_t *argp, caddr_t *resp, struct exportinfo *exi, 606 struct svc_req *req, cred_t *cr) 607 { 608 DTRACE_NFSV3_3(op__null__start, struct svc_req *, req, 609 cred_t *, cr, vnode_t *, NULL); 610 DTRACE_NFSV3_3(op__null__done, struct svc_req *, req, 611 cred_t *, cr, vnode_t *, NULL); 612 } 613 614 /* ARGSUSED */ 615 static void 616 rfs_error(caddr_t *argp, caddr_t *resp) 617 { 618 /* return (EOPNOTSUPP); */ 619 } 620 621 static void 622 nullfree(void) 623 { 624 } 625 626 static char *rfscallnames_v2[] = { 627 "RFS2_NULL", 628 "RFS2_GETATTR", 629 "RFS2_SETATTR", 630 "RFS2_ROOT", 631 "RFS2_LOOKUP", 632 "RFS2_READLINK", 633 "RFS2_READ", 634 "RFS2_WRITECACHE", 635 "RFS2_WRITE", 636 "RFS2_CREATE", 637 "RFS2_REMOVE", 638 "RFS2_RENAME", 639 "RFS2_LINK", 640 "RFS2_SYMLINK", 641 "RFS2_MKDIR", 642 "RFS2_RMDIR", 643 "RFS2_READDIR", 644 "RFS2_STATFS" 645 }; 646 647 static struct rpcdisp rfsdisptab_v2[] = { 648 /* 649 * NFS VERSION 2 650 */ 651 652 /* RFS_NULL = 0 */ 653 {rpc_null, 654 xdr_void, NULL_xdrproc_t, 0, 655 xdr_void, NULL_xdrproc_t, 0, 656 nullfree, RPC_IDEMPOTENT, 657 0}, 658 659 /* RFS_GETATTR = 1 */ 660 {rfs_getattr, 661 xdr_fhandle, xdr_fastfhandle, sizeof (fhandle_t), 662 xdr_attrstat, xdr_fastattrstat, sizeof (struct nfsattrstat), 663 nullfree, RPC_IDEMPOTENT|RPC_ALLOWANON|RPC_MAPRESP, 664 rfs_getattr_getfh}, 665 666 /* RFS_SETATTR = 2 */ 667 {rfs_setattr, 668 xdr_saargs, NULL_xdrproc_t, sizeof (struct nfssaargs), 669 xdr_attrstat, xdr_fastattrstat, sizeof (struct nfsattrstat), 670 nullfree, RPC_MAPRESP, 671 rfs_setattr_getfh}, 672 673 /* RFS_ROOT = 3 *** NO LONGER SUPPORTED *** */ 674 {rfs_error, 675 xdr_void, NULL_xdrproc_t, 0, 676 xdr_void, NULL_xdrproc_t, 0, 677 nullfree, RPC_IDEMPOTENT, 678 0}, 679 680 /* RFS_LOOKUP = 4 */ 681 {rfs_lookup, 682 xdr_diropargs, NULL_xdrproc_t, sizeof (struct nfsdiropargs), 683 xdr_diropres, xdr_fastdiropres, sizeof (struct nfsdiropres), 684 nullfree, RPC_IDEMPOTENT|RPC_MAPRESP|RPC_PUBLICFH_OK, 685 rfs_lookup_getfh}, 686 687 /* RFS_READLINK = 5 */ 688 {rfs_readlink, 689 xdr_fhandle, xdr_fastfhandle, sizeof (fhandle_t), 690 xdr_rdlnres, NULL_xdrproc_t, sizeof (struct nfsrdlnres), 691 rfs_rlfree, RPC_IDEMPOTENT, 692 rfs_readlink_getfh}, 693 694 /* RFS_READ = 6 */ 695 {rfs_read, 696 xdr_readargs, NULL_xdrproc_t, sizeof (struct nfsreadargs), 697 xdr_rdresult, NULL_xdrproc_t, sizeof (struct nfsrdresult), 698 rfs_rdfree, RPC_IDEMPOTENT, 699 rfs_read_getfh}, 700 701 /* RFS_WRITECACHE = 7 *** NO LONGER SUPPORTED *** */ 702 {rfs_error, 703 xdr_void, NULL_xdrproc_t, 0, 704 xdr_void, NULL_xdrproc_t, 0, 705 nullfree, RPC_IDEMPOTENT, 706 0}, 707 708 /* RFS_WRITE = 8 */ 709 {rfs_write, 710 xdr_writeargs, NULL_xdrproc_t, sizeof (struct nfswriteargs), 711 xdr_attrstat, xdr_fastattrstat, sizeof (struct nfsattrstat), 712 nullfree, RPC_MAPRESP, 713 rfs_write_getfh}, 714 715 /* RFS_CREATE = 9 */ 716 {rfs_create, 717 xdr_creatargs, NULL_xdrproc_t, sizeof (struct nfscreatargs), 718 xdr_diropres, xdr_fastdiropres, sizeof (struct nfsdiropres), 719 nullfree, RPC_MAPRESP, 720 rfs_create_getfh}, 721 722 /* RFS_REMOVE = 10 */ 723 {rfs_remove, 724 xdr_diropargs, NULL_xdrproc_t, sizeof (struct nfsdiropargs), 725 #ifdef _LITTLE_ENDIAN 726 xdr_enum, xdr_fastenum, sizeof (enum nfsstat), 727 #else 728 xdr_enum, NULL_xdrproc_t, sizeof (enum nfsstat), 729 #endif 730 nullfree, RPC_MAPRESP, 731 rfs_remove_getfh}, 732 733 /* RFS_RENAME = 11 */ 734 {rfs_rename, 735 xdr_rnmargs, NULL_xdrproc_t, sizeof (struct nfsrnmargs), 736 #ifdef _LITTLE_ENDIAN 737 xdr_enum, xdr_fastenum, sizeof (enum nfsstat), 738 #else 739 xdr_enum, NULL_xdrproc_t, sizeof (enum nfsstat), 740 #endif 741 nullfree, RPC_MAPRESP, 742 rfs_rename_getfh}, 743 744 /* RFS_LINK = 12 */ 745 {rfs_link, 746 xdr_linkargs, NULL_xdrproc_t, sizeof (struct nfslinkargs), 747 #ifdef _LITTLE_ENDIAN 748 xdr_enum, xdr_fastenum, sizeof (enum nfsstat), 749 #else 750 xdr_enum, NULL_xdrproc_t, sizeof (enum nfsstat), 751 #endif 752 nullfree, RPC_MAPRESP, 753 rfs_link_getfh}, 754 755 /* RFS_SYMLINK = 13 */ 756 {rfs_symlink, 757 xdr_slargs, NULL_xdrproc_t, sizeof (struct nfsslargs), 758 #ifdef _LITTLE_ENDIAN 759 xdr_enum, xdr_fastenum, sizeof (enum nfsstat), 760 #else 761 xdr_enum, NULL_xdrproc_t, sizeof (enum nfsstat), 762 #endif 763 nullfree, RPC_MAPRESP, 764 rfs_symlink_getfh}, 765 766 /* RFS_MKDIR = 14 */ 767 {rfs_mkdir, 768 xdr_creatargs, NULL_xdrproc_t, sizeof (struct nfscreatargs), 769 xdr_diropres, xdr_fastdiropres, sizeof (struct nfsdiropres), 770 nullfree, RPC_MAPRESP, 771 rfs_mkdir_getfh}, 772 773 /* RFS_RMDIR = 15 */ 774 {rfs_rmdir, 775 xdr_diropargs, NULL_xdrproc_t, sizeof (struct nfsdiropargs), 776 #ifdef _LITTLE_ENDIAN 777 xdr_enum, xdr_fastenum, sizeof (enum nfsstat), 778 #else 779 xdr_enum, NULL_xdrproc_t, sizeof (enum nfsstat), 780 #endif 781 nullfree, RPC_MAPRESP, 782 rfs_rmdir_getfh}, 783 784 /* RFS_READDIR = 16 */ 785 {rfs_readdir, 786 xdr_rddirargs, NULL_xdrproc_t, sizeof (struct nfsrddirargs), 787 xdr_putrddirres, NULL_xdrproc_t, sizeof (struct nfsrddirres), 788 rfs_rddirfree, RPC_IDEMPOTENT, 789 rfs_readdir_getfh}, 790 791 /* RFS_STATFS = 17 */ 792 {rfs_statfs, 793 xdr_fhandle, xdr_fastfhandle, sizeof (fhandle_t), 794 xdr_statfs, xdr_faststatfs, sizeof (struct nfsstatfs), 795 nullfree, RPC_IDEMPOTENT|RPC_ALLOWANON|RPC_MAPRESP, 796 rfs_statfs_getfh}, 797 }; 798 799 static char *rfscallnames_v3[] = { 800 "RFS3_NULL", 801 "RFS3_GETATTR", 802 "RFS3_SETATTR", 803 "RFS3_LOOKUP", 804 "RFS3_ACCESS", 805 "RFS3_READLINK", 806 "RFS3_READ", 807 "RFS3_WRITE", 808 "RFS3_CREATE", 809 "RFS3_MKDIR", 810 "RFS3_SYMLINK", 811 "RFS3_MKNOD", 812 "RFS3_REMOVE", 813 "RFS3_RMDIR", 814 "RFS3_RENAME", 815 "RFS3_LINK", 816 "RFS3_READDIR", 817 "RFS3_READDIRPLUS", 818 "RFS3_FSSTAT", 819 "RFS3_FSINFO", 820 "RFS3_PATHCONF", 821 "RFS3_COMMIT" 822 }; 823 824 static struct rpcdisp rfsdisptab_v3[] = { 825 /* 826 * NFS VERSION 3 827 */ 828 829 /* RFS_NULL = 0 */ 830 {rpc_null_v3, 831 xdr_void, NULL_xdrproc_t, 0, 832 xdr_void, NULL_xdrproc_t, 0, 833 nullfree, RPC_IDEMPOTENT, 834 0}, 835 836 /* RFS3_GETATTR = 1 */ 837 {rfs3_getattr, 838 xdr_nfs_fh3_server, NULL_xdrproc_t, sizeof (GETATTR3args), 839 xdr_GETATTR3res, NULL_xdrproc_t, sizeof (GETATTR3res), 840 nullfree, (RPC_IDEMPOTENT | RPC_ALLOWANON), 841 rfs3_getattr_getfh}, 842 843 /* RFS3_SETATTR = 2 */ 844 {rfs3_setattr, 845 xdr_SETATTR3args, NULL_xdrproc_t, sizeof (SETATTR3args), 846 xdr_SETATTR3res, NULL_xdrproc_t, sizeof (SETATTR3res), 847 nullfree, 0, 848 rfs3_setattr_getfh}, 849 850 /* RFS3_LOOKUP = 3 */ 851 {rfs3_lookup, 852 xdr_diropargs3, NULL_xdrproc_t, sizeof (LOOKUP3args), 853 xdr_LOOKUP3res, NULL_xdrproc_t, sizeof (LOOKUP3res), 854 nullfree, (RPC_IDEMPOTENT | RPC_PUBLICFH_OK), 855 rfs3_lookup_getfh}, 856 857 /* RFS3_ACCESS = 4 */ 858 {rfs3_access, 859 xdr_ACCESS3args, NULL_xdrproc_t, sizeof (ACCESS3args), 860 xdr_ACCESS3res, NULL_xdrproc_t, sizeof (ACCESS3res), 861 nullfree, RPC_IDEMPOTENT, 862 rfs3_access_getfh}, 863 864 /* RFS3_READLINK = 5 */ 865 {rfs3_readlink, 866 xdr_nfs_fh3_server, NULL_xdrproc_t, sizeof (READLINK3args), 867 xdr_READLINK3res, NULL_xdrproc_t, sizeof (READLINK3res), 868 rfs3_readlink_free, RPC_IDEMPOTENT, 869 rfs3_readlink_getfh}, 870 871 /* RFS3_READ = 6 */ 872 {rfs3_read, 873 xdr_READ3args, NULL_xdrproc_t, sizeof (READ3args), 874 xdr_READ3res, NULL_xdrproc_t, sizeof (READ3res), 875 rfs3_read_free, RPC_IDEMPOTENT, 876 rfs3_read_getfh}, 877 878 /* RFS3_WRITE = 7 */ 879 {rfs3_write, 880 xdr_WRITE3args, NULL_xdrproc_t, sizeof (WRITE3args), 881 xdr_WRITE3res, NULL_xdrproc_t, sizeof (WRITE3res), 882 nullfree, 0, 883 rfs3_write_getfh}, 884 885 /* RFS3_CREATE = 8 */ 886 {rfs3_create, 887 xdr_CREATE3args, NULL_xdrproc_t, sizeof (CREATE3args), 888 xdr_CREATE3res, NULL_xdrproc_t, sizeof (CREATE3res), 889 nullfree, 0, 890 rfs3_create_getfh}, 891 892 /* RFS3_MKDIR = 9 */ 893 {rfs3_mkdir, 894 xdr_MKDIR3args, NULL_xdrproc_t, sizeof (MKDIR3args), 895 xdr_MKDIR3res, NULL_xdrproc_t, sizeof (MKDIR3res), 896 nullfree, 0, 897 rfs3_mkdir_getfh}, 898 899 /* RFS3_SYMLINK = 10 */ 900 {rfs3_symlink, 901 xdr_SYMLINK3args, NULL_xdrproc_t, sizeof (SYMLINK3args), 902 xdr_SYMLINK3res, NULL_xdrproc_t, sizeof (SYMLINK3res), 903 nullfree, 0, 904 rfs3_symlink_getfh}, 905 906 /* RFS3_MKNOD = 11 */ 907 {rfs3_mknod, 908 xdr_MKNOD3args, NULL_xdrproc_t, sizeof (MKNOD3args), 909 xdr_MKNOD3res, NULL_xdrproc_t, sizeof (MKNOD3res), 910 nullfree, 0, 911 rfs3_mknod_getfh}, 912 913 /* RFS3_REMOVE = 12 */ 914 {rfs3_remove, 915 xdr_diropargs3, NULL_xdrproc_t, sizeof (REMOVE3args), 916 xdr_REMOVE3res, NULL_xdrproc_t, sizeof (REMOVE3res), 917 nullfree, 0, 918 rfs3_remove_getfh}, 919 920 /* RFS3_RMDIR = 13 */ 921 {rfs3_rmdir, 922 xdr_diropargs3, NULL_xdrproc_t, sizeof (RMDIR3args), 923 xdr_RMDIR3res, NULL_xdrproc_t, sizeof (RMDIR3res), 924 nullfree, 0, 925 rfs3_rmdir_getfh}, 926 927 /* RFS3_RENAME = 14 */ 928 {rfs3_rename, 929 xdr_RENAME3args, NULL_xdrproc_t, sizeof (RENAME3args), 930 xdr_RENAME3res, NULL_xdrproc_t, sizeof (RENAME3res), 931 nullfree, 0, 932 rfs3_rename_getfh}, 933 934 /* RFS3_LINK = 15 */ 935 {rfs3_link, 936 xdr_LINK3args, NULL_xdrproc_t, sizeof (LINK3args), 937 xdr_LINK3res, NULL_xdrproc_t, sizeof (LINK3res), 938 nullfree, 0, 939 rfs3_link_getfh}, 940 941 /* RFS3_READDIR = 16 */ 942 {rfs3_readdir, 943 xdr_READDIR3args, NULL_xdrproc_t, sizeof (READDIR3args), 944 xdr_READDIR3res, NULL_xdrproc_t, sizeof (READDIR3res), 945 rfs3_readdir_free, RPC_IDEMPOTENT, 946 rfs3_readdir_getfh}, 947 948 /* RFS3_READDIRPLUS = 17 */ 949 {rfs3_readdirplus, 950 xdr_READDIRPLUS3args, NULL_xdrproc_t, sizeof (READDIRPLUS3args), 951 xdr_READDIRPLUS3res, NULL_xdrproc_t, sizeof (READDIRPLUS3res), 952 rfs3_readdirplus_free, RPC_AVOIDWORK, 953 rfs3_readdirplus_getfh}, 954 955 /* RFS3_FSSTAT = 18 */ 956 {rfs3_fsstat, 957 xdr_nfs_fh3_server, NULL_xdrproc_t, sizeof (FSSTAT3args), 958 xdr_FSSTAT3res, NULL_xdrproc_t, sizeof (FSSTAT3res), 959 nullfree, RPC_IDEMPOTENT, 960 rfs3_fsstat_getfh}, 961 962 /* RFS3_FSINFO = 19 */ 963 {rfs3_fsinfo, 964 xdr_nfs_fh3_server, NULL_xdrproc_t, sizeof (FSINFO3args), 965 xdr_FSINFO3res, NULL_xdrproc_t, sizeof (FSINFO3res), 966 nullfree, RPC_IDEMPOTENT|RPC_ALLOWANON, 967 rfs3_fsinfo_getfh}, 968 969 /* RFS3_PATHCONF = 20 */ 970 {rfs3_pathconf, 971 xdr_nfs_fh3_server, NULL_xdrproc_t, sizeof (PATHCONF3args), 972 xdr_PATHCONF3res, NULL_xdrproc_t, sizeof (PATHCONF3res), 973 nullfree, RPC_IDEMPOTENT, 974 rfs3_pathconf_getfh}, 975 976 /* RFS3_COMMIT = 21 */ 977 {rfs3_commit, 978 xdr_COMMIT3args, NULL_xdrproc_t, sizeof (COMMIT3args), 979 xdr_COMMIT3res, NULL_xdrproc_t, sizeof (COMMIT3res), 980 nullfree, RPC_IDEMPOTENT, 981 rfs3_commit_getfh}, 982 }; 983 984 static char *rfscallnames_v4[] = { 985 "RFS4_NULL", 986 "RFS4_COMPOUND", 987 "RFS4_NULL", 988 "RFS4_NULL", 989 "RFS4_NULL", 990 "RFS4_NULL", 991 "RFS4_NULL", 992 "RFS4_NULL", 993 "RFS4_CREATE" 994 }; 995 996 static struct rpcdisp rfsdisptab_v4[] = { 997 /* 998 * NFS VERSION 4 999 */ 1000 1001 /* RFS_NULL = 0 */ 1002 {rpc_null, 1003 xdr_void, NULL_xdrproc_t, 0, 1004 xdr_void, NULL_xdrproc_t, 0, 1005 nullfree, RPC_IDEMPOTENT, 0}, 1006 1007 /* RFS4_compound = 1 */ 1008 {rfs4_compound, 1009 xdr_COMPOUND4args_srv, NULL_xdrproc_t, sizeof (COMPOUND4args), 1010 xdr_COMPOUND4res_srv, NULL_xdrproc_t, sizeof (COMPOUND4res), 1011 rfs4_compound_free, 0, 0}, 1012 }; 1013 1014 union rfs_args { 1015 /* 1016 * NFS VERSION 2 1017 */ 1018 1019 /* RFS_NULL = 0 */ 1020 1021 /* RFS_GETATTR = 1 */ 1022 fhandle_t nfs2_getattr_args; 1023 1024 /* RFS_SETATTR = 2 */ 1025 struct nfssaargs nfs2_setattr_args; 1026 1027 /* RFS_ROOT = 3 *** NO LONGER SUPPORTED *** */ 1028 1029 /* RFS_LOOKUP = 4 */ 1030 struct nfsdiropargs nfs2_lookup_args; 1031 1032 /* RFS_READLINK = 5 */ 1033 fhandle_t nfs2_readlink_args; 1034 1035 /* RFS_READ = 6 */ 1036 struct nfsreadargs nfs2_read_args; 1037 1038 /* RFS_WRITECACHE = 7 *** NO LONGER SUPPORTED *** */ 1039 1040 /* RFS_WRITE = 8 */ 1041 struct nfswriteargs nfs2_write_args; 1042 1043 /* RFS_CREATE = 9 */ 1044 struct nfscreatargs nfs2_create_args; 1045 1046 /* RFS_REMOVE = 10 */ 1047 struct nfsdiropargs nfs2_remove_args; 1048 1049 /* RFS_RENAME = 11 */ 1050 struct nfsrnmargs nfs2_rename_args; 1051 1052 /* RFS_LINK = 12 */ 1053 struct nfslinkargs nfs2_link_args; 1054 1055 /* RFS_SYMLINK = 13 */ 1056 struct nfsslargs nfs2_symlink_args; 1057 1058 /* RFS_MKDIR = 14 */ 1059 struct nfscreatargs nfs2_mkdir_args; 1060 1061 /* RFS_RMDIR = 15 */ 1062 struct nfsdiropargs nfs2_rmdir_args; 1063 1064 /* RFS_READDIR = 16 */ 1065 struct nfsrddirargs nfs2_readdir_args; 1066 1067 /* RFS_STATFS = 17 */ 1068 fhandle_t nfs2_statfs_args; 1069 1070 /* 1071 * NFS VERSION 3 1072 */ 1073 1074 /* RFS_NULL = 0 */ 1075 1076 /* RFS3_GETATTR = 1 */ 1077 GETATTR3args nfs3_getattr_args; 1078 1079 /* RFS3_SETATTR = 2 */ 1080 SETATTR3args nfs3_setattr_args; 1081 1082 /* RFS3_LOOKUP = 3 */ 1083 LOOKUP3args nfs3_lookup_args; 1084 1085 /* RFS3_ACCESS = 4 */ 1086 ACCESS3args nfs3_access_args; 1087 1088 /* RFS3_READLINK = 5 */ 1089 READLINK3args nfs3_readlink_args; 1090 1091 /* RFS3_READ = 6 */ 1092 READ3args nfs3_read_args; 1093 1094 /* RFS3_WRITE = 7 */ 1095 WRITE3args nfs3_write_args; 1096 1097 /* RFS3_CREATE = 8 */ 1098 CREATE3args nfs3_create_args; 1099 1100 /* RFS3_MKDIR = 9 */ 1101 MKDIR3args nfs3_mkdir_args; 1102 1103 /* RFS3_SYMLINK = 10 */ 1104 SYMLINK3args nfs3_symlink_args; 1105 1106 /* RFS3_MKNOD = 11 */ 1107 MKNOD3args nfs3_mknod_args; 1108 1109 /* RFS3_REMOVE = 12 */ 1110 REMOVE3args nfs3_remove_args; 1111 1112 /* RFS3_RMDIR = 13 */ 1113 RMDIR3args nfs3_rmdir_args; 1114 1115 /* RFS3_RENAME = 14 */ 1116 RENAME3args nfs3_rename_args; 1117 1118 /* RFS3_LINK = 15 */ 1119 LINK3args nfs3_link_args; 1120 1121 /* RFS3_READDIR = 16 */ 1122 READDIR3args nfs3_readdir_args; 1123 1124 /* RFS3_READDIRPLUS = 17 */ 1125 READDIRPLUS3args nfs3_readdirplus_args; 1126 1127 /* RFS3_FSSTAT = 18 */ 1128 FSSTAT3args nfs3_fsstat_args; 1129 1130 /* RFS3_FSINFO = 19 */ 1131 FSINFO3args nfs3_fsinfo_args; 1132 1133 /* RFS3_PATHCONF = 20 */ 1134 PATHCONF3args nfs3_pathconf_args; 1135 1136 /* RFS3_COMMIT = 21 */ 1137 COMMIT3args nfs3_commit_args; 1138 1139 /* 1140 * NFS VERSION 4 1141 */ 1142 1143 /* RFS_NULL = 0 */ 1144 1145 /* COMPUND = 1 */ 1146 COMPOUND4args nfs4_compound_args; 1147 }; 1148 1149 union rfs_res { 1150 /* 1151 * NFS VERSION 2 1152 */ 1153 1154 /* RFS_NULL = 0 */ 1155 1156 /* RFS_GETATTR = 1 */ 1157 struct nfsattrstat nfs2_getattr_res; 1158 1159 /* RFS_SETATTR = 2 */ 1160 struct nfsattrstat nfs2_setattr_res; 1161 1162 /* RFS_ROOT = 3 *** NO LONGER SUPPORTED *** */ 1163 1164 /* RFS_LOOKUP = 4 */ 1165 struct nfsdiropres nfs2_lookup_res; 1166 1167 /* RFS_READLINK = 5 */ 1168 struct nfsrdlnres nfs2_readlink_res; 1169 1170 /* RFS_READ = 6 */ 1171 struct nfsrdresult nfs2_read_res; 1172 1173 /* RFS_WRITECACHE = 7 *** NO LONGER SUPPORTED *** */ 1174 1175 /* RFS_WRITE = 8 */ 1176 struct nfsattrstat nfs2_write_res; 1177 1178 /* RFS_CREATE = 9 */ 1179 struct nfsdiropres nfs2_create_res; 1180 1181 /* RFS_REMOVE = 10 */ 1182 enum nfsstat nfs2_remove_res; 1183 1184 /* RFS_RENAME = 11 */ 1185 enum nfsstat nfs2_rename_res; 1186 1187 /* RFS_LINK = 12 */ 1188 enum nfsstat nfs2_link_res; 1189 1190 /* RFS_SYMLINK = 13 */ 1191 enum nfsstat nfs2_symlink_res; 1192 1193 /* RFS_MKDIR = 14 */ 1194 struct nfsdiropres nfs2_mkdir_res; 1195 1196 /* RFS_RMDIR = 15 */ 1197 enum nfsstat nfs2_rmdir_res; 1198 1199 /* RFS_READDIR = 16 */ 1200 struct nfsrddirres nfs2_readdir_res; 1201 1202 /* RFS_STATFS = 17 */ 1203 struct nfsstatfs nfs2_statfs_res; 1204 1205 /* 1206 * NFS VERSION 3 1207 */ 1208 1209 /* RFS_NULL = 0 */ 1210 1211 /* RFS3_GETATTR = 1 */ 1212 GETATTR3res nfs3_getattr_res; 1213 1214 /* RFS3_SETATTR = 2 */ 1215 SETATTR3res nfs3_setattr_res; 1216 1217 /* RFS3_LOOKUP = 3 */ 1218 LOOKUP3res nfs3_lookup_res; 1219 1220 /* RFS3_ACCESS = 4 */ 1221 ACCESS3res nfs3_access_res; 1222 1223 /* RFS3_READLINK = 5 */ 1224 READLINK3res nfs3_readlink_res; 1225 1226 /* RFS3_READ = 6 */ 1227 READ3res nfs3_read_res; 1228 1229 /* RFS3_WRITE = 7 */ 1230 WRITE3res nfs3_write_res; 1231 1232 /* RFS3_CREATE = 8 */ 1233 CREATE3res nfs3_create_res; 1234 1235 /* RFS3_MKDIR = 9 */ 1236 MKDIR3res nfs3_mkdir_res; 1237 1238 /* RFS3_SYMLINK = 10 */ 1239 SYMLINK3res nfs3_symlink_res; 1240 1241 /* RFS3_MKNOD = 11 */ 1242 MKNOD3res nfs3_mknod_res; 1243 1244 /* RFS3_REMOVE = 12 */ 1245 REMOVE3res nfs3_remove_res; 1246 1247 /* RFS3_RMDIR = 13 */ 1248 RMDIR3res nfs3_rmdir_res; 1249 1250 /* RFS3_RENAME = 14 */ 1251 RENAME3res nfs3_rename_res; 1252 1253 /* RFS3_LINK = 15 */ 1254 LINK3res nfs3_link_res; 1255 1256 /* RFS3_READDIR = 16 */ 1257 READDIR3res nfs3_readdir_res; 1258 1259 /* RFS3_READDIRPLUS = 17 */ 1260 READDIRPLUS3res nfs3_readdirplus_res; 1261 1262 /* RFS3_FSSTAT = 18 */ 1263 FSSTAT3res nfs3_fsstat_res; 1264 1265 /* RFS3_FSINFO = 19 */ 1266 FSINFO3res nfs3_fsinfo_res; 1267 1268 /* RFS3_PATHCONF = 20 */ 1269 PATHCONF3res nfs3_pathconf_res; 1270 1271 /* RFS3_COMMIT = 21 */ 1272 COMMIT3res nfs3_commit_res; 1273 1274 /* 1275 * NFS VERSION 4 1276 */ 1277 1278 /* RFS_NULL = 0 */ 1279 1280 /* RFS4_COMPOUND = 1 */ 1281 COMPOUND4res nfs4_compound_res; 1282 1283 }; 1284 1285 static struct rpc_disptable rfs_disptable[] = { 1286 {sizeof (rfsdisptab_v2) / sizeof (rfsdisptab_v2[0]), 1287 rfscallnames_v2, 1288 &rfsproccnt_v2_ptr, rfsdisptab_v2}, 1289 {sizeof (rfsdisptab_v3) / sizeof (rfsdisptab_v3[0]), 1290 rfscallnames_v3, 1291 &rfsproccnt_v3_ptr, rfsdisptab_v3}, 1292 {sizeof (rfsdisptab_v4) / sizeof (rfsdisptab_v4[0]), 1293 rfscallnames_v4, 1294 &rfsproccnt_v4_ptr, rfsdisptab_v4}, 1295 }; 1296 1297 /* 1298 * If nfs_portmon is set, then clients are required to use privileged 1299 * ports (ports < IPPORT_RESERVED) in order to get NFS services. 1300 * 1301 * N.B.: this attempt to carry forward the already ill-conceived notion 1302 * of privileged ports for TCP/UDP is really quite ineffectual. Not only 1303 * is it transport-dependent, it's laughably easy to spoof. If you're 1304 * really interested in security, you must start with secure RPC instead. 1305 */ 1306 static int nfs_portmon = 0; 1307 1308 #ifdef DEBUG 1309 static int cred_hits = 0; 1310 static int cred_misses = 0; 1311 #endif 1312 1313 1314 #ifdef DEBUG 1315 /* 1316 * Debug code to allow disabling of rfs_dispatch() use of 1317 * fastxdrargs() and fastxdrres() calls for testing purposes. 1318 */ 1319 static int rfs_no_fast_xdrargs = 0; 1320 static int rfs_no_fast_xdrres = 0; 1321 #endif 1322 1323 union acl_args { 1324 /* 1325 * ACL VERSION 2 1326 */ 1327 1328 /* ACL2_NULL = 0 */ 1329 1330 /* ACL2_GETACL = 1 */ 1331 GETACL2args acl2_getacl_args; 1332 1333 /* ACL2_SETACL = 2 */ 1334 SETACL2args acl2_setacl_args; 1335 1336 /* ACL2_GETATTR = 3 */ 1337 GETATTR2args acl2_getattr_args; 1338 1339 /* ACL2_ACCESS = 4 */ 1340 ACCESS2args acl2_access_args; 1341 1342 /* ACL2_GETXATTRDIR = 5 */ 1343 GETXATTRDIR2args acl2_getxattrdir_args; 1344 1345 /* 1346 * ACL VERSION 3 1347 */ 1348 1349 /* ACL3_NULL = 0 */ 1350 1351 /* ACL3_GETACL = 1 */ 1352 GETACL3args acl3_getacl_args; 1353 1354 /* ACL3_SETACL = 2 */ 1355 SETACL3args acl3_setacl; 1356 1357 /* ACL3_GETXATTRDIR = 3 */ 1358 GETXATTRDIR3args acl3_getxattrdir_args; 1359 1360 }; 1361 1362 union acl_res { 1363 /* 1364 * ACL VERSION 2 1365 */ 1366 1367 /* ACL2_NULL = 0 */ 1368 1369 /* ACL2_GETACL = 1 */ 1370 GETACL2res acl2_getacl_res; 1371 1372 /* ACL2_SETACL = 2 */ 1373 SETACL2res acl2_setacl_res; 1374 1375 /* ACL2_GETATTR = 3 */ 1376 GETATTR2res acl2_getattr_res; 1377 1378 /* ACL2_ACCESS = 4 */ 1379 ACCESS2res acl2_access_res; 1380 1381 /* ACL2_GETXATTRDIR = 5 */ 1382 GETXATTRDIR2args acl2_getxattrdir_res; 1383 1384 /* 1385 * ACL VERSION 3 1386 */ 1387 1388 /* ACL3_NULL = 0 */ 1389 1390 /* ACL3_GETACL = 1 */ 1391 GETACL3res acl3_getacl_res; 1392 1393 /* ACL3_SETACL = 2 */ 1394 SETACL3res acl3_setacl_res; 1395 1396 /* ACL3_GETXATTRDIR = 3 */ 1397 GETXATTRDIR3res acl3_getxattrdir_res; 1398 1399 }; 1400 1401 static bool_t 1402 auth_tooweak(struct svc_req *req, char *res) 1403 { 1404 1405 if (req->rq_vers == NFS_VERSION && req->rq_proc == RFS_LOOKUP) { 1406 struct nfsdiropres *dr = (struct nfsdiropres *)res; 1407 if (dr->dr_status == WNFSERR_CLNT_FLAVOR) 1408 return (TRUE); 1409 } else if (req->rq_vers == NFS_V3 && req->rq_proc == NFSPROC3_LOOKUP) { 1410 LOOKUP3res *resp = (LOOKUP3res *)res; 1411 if (resp->status == WNFSERR_CLNT_FLAVOR) 1412 return (TRUE); 1413 } 1414 return (FALSE); 1415 } 1416 1417 1418 static void 1419 common_dispatch(struct svc_req *req, SVCXPRT *xprt, rpcvers_t min_vers, 1420 rpcvers_t max_vers, char *pgmname, 1421 struct rpc_disptable *disptable) 1422 { 1423 int which; 1424 rpcvers_t vers; 1425 char *args; 1426 union { 1427 union rfs_args ra; 1428 union acl_args aa; 1429 } args_buf; 1430 char *res; 1431 union { 1432 union rfs_res rr; 1433 union acl_res ar; 1434 } res_buf; 1435 struct rpcdisp *disp = NULL; 1436 int dis_flags = 0; 1437 cred_t *cr; 1438 int error = 0; 1439 int anon_ok; 1440 struct exportinfo *exi = NULL; 1441 unsigned int nfslog_rec_id; 1442 int dupstat; 1443 struct dupreq *dr; 1444 int authres; 1445 bool_t publicfh_ok = FALSE; 1446 enum_t auth_flavor; 1447 bool_t dupcached = FALSE; 1448 struct netbuf nb; 1449 bool_t logging_enabled = FALSE; 1450 struct exportinfo *nfslog_exi = NULL; 1451 char **procnames; 1452 char cbuf[INET6_ADDRSTRLEN]; /* to hold both IPv4 and IPv6 addr */ 1453 1454 vers = req->rq_vers; 1455 1456 if (vers < min_vers || vers > max_vers) { 1457 svcerr_progvers(req->rq_xprt, min_vers, max_vers); 1458 error++; 1459 cmn_err(CE_NOTE, "%s: bad version number %u", pgmname, vers); 1460 goto done; 1461 } 1462 vers -= min_vers; 1463 1464 which = req->rq_proc; 1465 if (which < 0 || which >= disptable[(int)vers].dis_nprocs) { 1466 svcerr_noproc(req->rq_xprt); 1467 error++; 1468 goto done; 1469 } 1470 1471 (*(disptable[(int)vers].dis_proccntp))[which].value.ui64++; 1472 1473 disp = &disptable[(int)vers].dis_table[which]; 1474 procnames = disptable[(int)vers].dis_procnames; 1475 1476 auth_flavor = req->rq_cred.oa_flavor; 1477 1478 /* 1479 * Deserialize into the args struct. 1480 */ 1481 args = (char *)&args_buf; 1482 1483 #ifdef DEBUG 1484 if (rfs_no_fast_xdrargs || (auth_flavor == RPCSEC_GSS) || 1485 disp->dis_fastxdrargs == NULL_xdrproc_t || 1486 !SVC_GETARGS(xprt, disp->dis_fastxdrargs, (char *)&args)) 1487 #else 1488 if ((auth_flavor == RPCSEC_GSS) || 1489 disp->dis_fastxdrargs == NULL_xdrproc_t || 1490 !SVC_GETARGS(xprt, disp->dis_fastxdrargs, (char *)&args)) 1491 #endif 1492 { 1493 bzero(args, disp->dis_argsz); 1494 if (!SVC_GETARGS(xprt, disp->dis_xdrargs, args)) { 1495 error++; 1496 /* 1497 * Check if we are outside our capabilities. 1498 */ 1499 if (rfs4_minorvers_mismatch(req, xprt, (void *)args)) 1500 goto done; 1501 1502 svcerr_decode(xprt); 1503 cmn_err(CE_NOTE, 1504 "Failed to decode arguments for %s version %u " 1505 "procedure %s client %s%s", 1506 pgmname, vers + min_vers, procnames[which], 1507 client_name(req), client_addr(req, cbuf)); 1508 goto done; 1509 } 1510 } 1511 1512 /* 1513 * If Version 4 use that specific dispatch function. 1514 */ 1515 if (req->rq_vers == 4) { 1516 error += rfs4_dispatch(disp, req, xprt, args); 1517 goto done; 1518 } 1519 1520 dis_flags = disp->dis_flags; 1521 1522 /* 1523 * Find export information and check authentication, 1524 * setting the credential if everything is ok. 1525 */ 1526 if (disp->dis_getfh != NULL) { 1527 void *fh; 1528 fsid_t *fsid; 1529 fid_t *fid, *xfid; 1530 fhandle_t *fh2; 1531 nfs_fh3 *fh3; 1532 1533 fh = (*disp->dis_getfh)(args); 1534 switch (req->rq_vers) { 1535 case NFS_VERSION: 1536 fh2 = (fhandle_t *)fh; 1537 fsid = &fh2->fh_fsid; 1538 fid = (fid_t *)&fh2->fh_len; 1539 xfid = (fid_t *)&fh2->fh_xlen; 1540 break; 1541 case NFS_V3: 1542 fh3 = (nfs_fh3 *)fh; 1543 fsid = &fh3->fh3_fsid; 1544 fid = FH3TOFIDP(fh3); 1545 xfid = FH3TOXFIDP(fh3); 1546 break; 1547 } 1548 1549 /* 1550 * Fix for bug 1038302 - corbin 1551 * There is a problem here if anonymous access is 1552 * disallowed. If the current request is part of the 1553 * client's mount process for the requested filesystem, 1554 * then it will carry root (uid 0) credentials on it, and 1555 * will be denied by checkauth if that client does not 1556 * have explicit root=0 permission. This will cause the 1557 * client's mount operation to fail. As a work-around, 1558 * we check here to see if the request is a getattr or 1559 * statfs operation on the exported vnode itself, and 1560 * pass a flag to checkauth with the result of this test. 1561 * 1562 * The filehandle refers to the mountpoint itself if 1563 * the fh_data and fh_xdata portions of the filehandle 1564 * are equal. 1565 * 1566 * Added anon_ok argument to checkauth(). 1567 */ 1568 1569 if ((dis_flags & RPC_ALLOWANON) && EQFID(fid, xfid)) 1570 anon_ok = 1; 1571 else 1572 anon_ok = 0; 1573 1574 cr = xprt->xp_cred; 1575 ASSERT(cr != NULL); 1576 #ifdef DEBUG 1577 if (crgetref(cr) != 1) { 1578 crfree(cr); 1579 cr = crget(); 1580 xprt->xp_cred = cr; 1581 cred_misses++; 1582 } else 1583 cred_hits++; 1584 #else 1585 if (crgetref(cr) != 1) { 1586 crfree(cr); 1587 cr = crget(); 1588 xprt->xp_cred = cr; 1589 } 1590 #endif 1591 1592 exi = checkexport(fsid, xfid); 1593 1594 if (exi != NULL) { 1595 publicfh_ok = PUBLICFH_CHECK(disp, exi, fsid, xfid); 1596 1597 /* 1598 * Don't allow non-V4 clients access 1599 * to pseudo exports 1600 */ 1601 if (PSEUDO(exi)) { 1602 svcerr_weakauth(xprt); 1603 error++; 1604 goto done; 1605 } 1606 1607 authres = checkauth(exi, req, cr, anon_ok, publicfh_ok); 1608 /* 1609 * authres > 0: authentication OK - proceed 1610 * authres == 0: authentication weak - return error 1611 * authres < 0: authentication timeout - drop 1612 */ 1613 if (authres <= 0) { 1614 if (authres == 0) { 1615 svcerr_weakauth(xprt); 1616 error++; 1617 } 1618 goto done; 1619 } 1620 1621 /* check to see if we might need charmap */ 1622 if (exi->exi_export.ex_flags & EX_CHARMAP) { 1623 struct sockaddr *ca; 1624 ca = (struct sockaddr *) 1625 svc_getrpccaller(req->rq_xprt)->buf; 1626 (void) nfscmd_charmap(exi, ca); 1627 } 1628 } 1629 } else 1630 cr = NULL; 1631 1632 if ((dis_flags & RPC_MAPRESP) && (auth_flavor != RPCSEC_GSS)) { 1633 res = (char *)SVC_GETRES(xprt, disp->dis_ressz); 1634 if (res == NULL) 1635 res = (char *)&res_buf; 1636 } else 1637 res = (char *)&res_buf; 1638 1639 if (!(dis_flags & RPC_IDEMPOTENT)) { 1640 dupstat = SVC_DUP_EXT(xprt, req, res, disp->dis_ressz, &dr, 1641 &dupcached); 1642 1643 switch (dupstat) { 1644 case DUP_ERROR: 1645 svcerr_systemerr(xprt); 1646 error++; 1647 goto done; 1648 /* NOTREACHED */ 1649 case DUP_INPROGRESS: 1650 if (res != (char *)&res_buf) 1651 SVC_FREERES(xprt); 1652 error++; 1653 goto done; 1654 /* NOTREACHED */ 1655 case DUP_NEW: 1656 case DUP_DROP: 1657 curthread->t_flag |= T_DONTPEND; 1658 1659 (*disp->dis_proc)(args, res, exi, req, cr); 1660 1661 curthread->t_flag &= ~T_DONTPEND; 1662 if (curthread->t_flag & T_WOULDBLOCK) { 1663 curthread->t_flag &= ~T_WOULDBLOCK; 1664 SVC_DUPDONE_EXT(xprt, dr, res, NULL, 1665 disp->dis_ressz, DUP_DROP); 1666 if (res != (char *)&res_buf) 1667 SVC_FREERES(xprt); 1668 error++; 1669 goto done; 1670 } 1671 if (dis_flags & RPC_AVOIDWORK) { 1672 SVC_DUPDONE_EXT(xprt, dr, res, NULL, 1673 disp->dis_ressz, DUP_DROP); 1674 } else { 1675 SVC_DUPDONE_EXT(xprt, dr, res, 1676 disp->dis_resfree == nullfree ? NULL : 1677 disp->dis_resfree, 1678 disp->dis_ressz, DUP_DONE); 1679 dupcached = TRUE; 1680 } 1681 break; 1682 case DUP_DONE: 1683 break; 1684 } 1685 1686 } else { 1687 curthread->t_flag |= T_DONTPEND; 1688 1689 (*disp->dis_proc)(args, res, exi, req, cr); 1690 1691 curthread->t_flag &= ~T_DONTPEND; 1692 if (curthread->t_flag & T_WOULDBLOCK) { 1693 curthread->t_flag &= ~T_WOULDBLOCK; 1694 if (res != (char *)&res_buf) 1695 SVC_FREERES(xprt); 1696 error++; 1697 goto done; 1698 } 1699 } 1700 1701 if (auth_tooweak(req, res)) { 1702 svcerr_weakauth(xprt); 1703 error++; 1704 goto done; 1705 } 1706 1707 /* 1708 * Check to see if logging has been enabled on the server. 1709 * If so, then obtain the export info struct to be used for 1710 * the later writing of the log record. This is done for 1711 * the case that a lookup is done across a non-logged public 1712 * file system. 1713 */ 1714 if (nfslog_buffer_list != NULL) { 1715 nfslog_exi = nfslog_get_exi(exi, req, res, &nfslog_rec_id); 1716 /* 1717 * Is logging enabled? 1718 */ 1719 logging_enabled = (nfslog_exi != NULL); 1720 1721 /* 1722 * Copy the netbuf for logging purposes, before it is 1723 * freed by svc_sendreply(). 1724 */ 1725 if (logging_enabled) { 1726 NFSLOG_COPY_NETBUF(nfslog_exi, xprt, &nb); 1727 /* 1728 * If RPC_MAPRESP flag set (i.e. in V2 ops) the 1729 * res gets copied directly into the mbuf and 1730 * may be freed soon after the sendreply. So we 1731 * must copy it here to a safe place... 1732 */ 1733 if (res != (char *)&res_buf) { 1734 bcopy(res, (char *)&res_buf, disp->dis_ressz); 1735 } 1736 } 1737 } 1738 1739 /* 1740 * Serialize and send results struct 1741 */ 1742 #ifdef DEBUG 1743 if (rfs_no_fast_xdrres == 0 && res != (char *)&res_buf) 1744 #else 1745 if (res != (char *)&res_buf) 1746 #endif 1747 { 1748 if (!svc_sendreply(xprt, disp->dis_fastxdrres, res)) { 1749 cmn_err(CE_NOTE, "%s: bad sendreply", pgmname); 1750 error++; 1751 } 1752 } else { 1753 if (!svc_sendreply(xprt, disp->dis_xdrres, res)) { 1754 cmn_err(CE_NOTE, "%s: bad sendreply", pgmname); 1755 error++; 1756 } 1757 } 1758 1759 /* 1760 * Log if needed 1761 */ 1762 if (logging_enabled) { 1763 nfslog_write_record(nfslog_exi, req, args, (char *)&res_buf, 1764 cr, &nb, nfslog_rec_id, NFSLOG_ONE_BUFFER); 1765 exi_rele(nfslog_exi); 1766 kmem_free((&nb)->buf, (&nb)->len); 1767 } 1768 1769 /* 1770 * Free results struct. With the addition of NFS V4 we can 1771 * have non-idempotent procedures with functions. 1772 */ 1773 if (disp->dis_resfree != nullfree && dupcached == FALSE) { 1774 (*disp->dis_resfree)(res); 1775 } 1776 1777 done: 1778 /* 1779 * Free arguments struct 1780 */ 1781 if (disp) { 1782 if (!SVC_FREEARGS(xprt, disp->dis_xdrargs, args)) { 1783 cmn_err(CE_NOTE, "%s: bad freeargs", pgmname); 1784 error++; 1785 } 1786 } else { 1787 if (!SVC_FREEARGS(xprt, (xdrproc_t)0, (caddr_t)0)) { 1788 cmn_err(CE_NOTE, "%s: bad freeargs", pgmname); 1789 error++; 1790 } 1791 } 1792 1793 if (exi != NULL) 1794 exi_rele(exi); 1795 1796 global_svstat_ptr[req->rq_vers][NFS_BADCALLS].value.ui64 += error; 1797 1798 global_svstat_ptr[req->rq_vers][NFS_CALLS].value.ui64++; 1799 } 1800 1801 static void 1802 rfs_dispatch(struct svc_req *req, SVCXPRT *xprt) 1803 { 1804 common_dispatch(req, xprt, NFS_VERSMIN, NFS_VERSMAX, 1805 "NFS", rfs_disptable); 1806 } 1807 1808 static char *aclcallnames_v2[] = { 1809 "ACL2_NULL", 1810 "ACL2_GETACL", 1811 "ACL2_SETACL", 1812 "ACL2_GETATTR", 1813 "ACL2_ACCESS", 1814 "ACL2_GETXATTRDIR" 1815 }; 1816 1817 static struct rpcdisp acldisptab_v2[] = { 1818 /* 1819 * ACL VERSION 2 1820 */ 1821 1822 /* ACL2_NULL = 0 */ 1823 {rpc_null, 1824 xdr_void, NULL_xdrproc_t, 0, 1825 xdr_void, NULL_xdrproc_t, 0, 1826 nullfree, RPC_IDEMPOTENT, 1827 0}, 1828 1829 /* ACL2_GETACL = 1 */ 1830 {acl2_getacl, 1831 xdr_GETACL2args, xdr_fastGETACL2args, sizeof (GETACL2args), 1832 xdr_GETACL2res, NULL_xdrproc_t, sizeof (GETACL2res), 1833 acl2_getacl_free, RPC_IDEMPOTENT, 1834 acl2_getacl_getfh}, 1835 1836 /* ACL2_SETACL = 2 */ 1837 {acl2_setacl, 1838 xdr_SETACL2args, NULL_xdrproc_t, sizeof (SETACL2args), 1839 #ifdef _LITTLE_ENDIAN 1840 xdr_SETACL2res, xdr_fastSETACL2res, sizeof (SETACL2res), 1841 #else 1842 xdr_SETACL2res, NULL_xdrproc_t, sizeof (SETACL2res), 1843 #endif 1844 nullfree, RPC_MAPRESP, 1845 acl2_setacl_getfh}, 1846 1847 /* ACL2_GETATTR = 3 */ 1848 {acl2_getattr, 1849 xdr_GETATTR2args, xdr_fastGETATTR2args, sizeof (GETATTR2args), 1850 #ifdef _LITTLE_ENDIAN 1851 xdr_GETATTR2res, xdr_fastGETATTR2res, sizeof (GETATTR2res), 1852 #else 1853 xdr_GETATTR2res, NULL_xdrproc_t, sizeof (GETATTR2res), 1854 #endif 1855 nullfree, RPC_IDEMPOTENT|RPC_ALLOWANON|RPC_MAPRESP, 1856 acl2_getattr_getfh}, 1857 1858 /* ACL2_ACCESS = 4 */ 1859 {acl2_access, 1860 xdr_ACCESS2args, xdr_fastACCESS2args, sizeof (ACCESS2args), 1861 #ifdef _LITTLE_ENDIAN 1862 xdr_ACCESS2res, xdr_fastACCESS2res, sizeof (ACCESS2res), 1863 #else 1864 xdr_ACCESS2res, NULL_xdrproc_t, sizeof (ACCESS2res), 1865 #endif 1866 nullfree, RPC_IDEMPOTENT|RPC_MAPRESP, 1867 acl2_access_getfh}, 1868 1869 /* ACL2_GETXATTRDIR = 5 */ 1870 {acl2_getxattrdir, 1871 xdr_GETXATTRDIR2args, NULL_xdrproc_t, sizeof (GETXATTRDIR2args), 1872 xdr_GETXATTRDIR2res, NULL_xdrproc_t, sizeof (GETXATTRDIR2res), 1873 nullfree, RPC_IDEMPOTENT, 1874 acl2_getxattrdir_getfh}, 1875 }; 1876 1877 static char *aclcallnames_v3[] = { 1878 "ACL3_NULL", 1879 "ACL3_GETACL", 1880 "ACL3_SETACL", 1881 "ACL3_GETXATTRDIR" 1882 }; 1883 1884 static struct rpcdisp acldisptab_v3[] = { 1885 /* 1886 * ACL VERSION 3 1887 */ 1888 1889 /* ACL3_NULL = 0 */ 1890 {rpc_null, 1891 xdr_void, NULL_xdrproc_t, 0, 1892 xdr_void, NULL_xdrproc_t, 0, 1893 nullfree, RPC_IDEMPOTENT, 1894 0}, 1895 1896 /* ACL3_GETACL = 1 */ 1897 {acl3_getacl, 1898 xdr_GETACL3args, NULL_xdrproc_t, sizeof (GETACL3args), 1899 xdr_GETACL3res, NULL_xdrproc_t, sizeof (GETACL3res), 1900 acl3_getacl_free, RPC_IDEMPOTENT, 1901 acl3_getacl_getfh}, 1902 1903 /* ACL3_SETACL = 2 */ 1904 {acl3_setacl, 1905 xdr_SETACL3args, NULL_xdrproc_t, sizeof (SETACL3args), 1906 xdr_SETACL3res, NULL_xdrproc_t, sizeof (SETACL3res), 1907 nullfree, 0, 1908 acl3_setacl_getfh}, 1909 1910 /* ACL3_GETXATTRDIR = 3 */ 1911 {acl3_getxattrdir, 1912 xdr_GETXATTRDIR3args, NULL_xdrproc_t, sizeof (GETXATTRDIR3args), 1913 xdr_GETXATTRDIR3res, NULL_xdrproc_t, sizeof (GETXATTRDIR3res), 1914 nullfree, RPC_IDEMPOTENT, 1915 acl3_getxattrdir_getfh}, 1916 }; 1917 1918 static struct rpc_disptable acl_disptable[] = { 1919 {sizeof (acldisptab_v2) / sizeof (acldisptab_v2[0]), 1920 aclcallnames_v2, 1921 &aclproccnt_v2_ptr, acldisptab_v2}, 1922 {sizeof (acldisptab_v3) / sizeof (acldisptab_v3[0]), 1923 aclcallnames_v3, 1924 &aclproccnt_v3_ptr, acldisptab_v3}, 1925 }; 1926 1927 static void 1928 acl_dispatch(struct svc_req *req, SVCXPRT *xprt) 1929 { 1930 common_dispatch(req, xprt, NFS_ACL_VERSMIN, NFS_ACL_VERSMAX, 1931 "ACL", acl_disptable); 1932 } 1933 1934 int 1935 checkwin(int flavor, int window, struct svc_req *req) 1936 { 1937 struct authdes_cred *adc; 1938 1939 switch (flavor) { 1940 case AUTH_DES: 1941 adc = (struct authdes_cred *)req->rq_clntcred; 1942 if (adc->adc_fullname.window > window) 1943 return (0); 1944 break; 1945 1946 default: 1947 break; 1948 } 1949 return (1); 1950 } 1951 1952 1953 /* 1954 * checkauth() will check the access permission against the export 1955 * information. Then map root uid/gid to appropriate uid/gid. 1956 * 1957 * This routine is used by NFS V3 and V2 code. 1958 */ 1959 static int 1960 checkauth(struct exportinfo *exi, struct svc_req *req, cred_t *cr, int anon_ok, 1961 bool_t publicfh_ok) 1962 { 1963 int i, nfsflavor, rpcflavor, stat, access; 1964 struct secinfo *secp; 1965 caddr_t principal; 1966 char buf[INET6_ADDRSTRLEN]; /* to hold both IPv4 and IPv6 addr */ 1967 int anon_res = 0; 1968 1969 /* 1970 * Check for privileged port number 1971 * N.B.: this assumes that we know the format of a netbuf. 1972 */ 1973 if (nfs_portmon) { 1974 struct sockaddr *ca; 1975 ca = (struct sockaddr *)svc_getrpccaller(req->rq_xprt)->buf; 1976 1977 if (ca == NULL) 1978 return (0); 1979 1980 if ((ca->sa_family == AF_INET && 1981 ntohs(((struct sockaddr_in *)ca)->sin_port) >= 1982 IPPORT_RESERVED) || 1983 (ca->sa_family == AF_INET6 && 1984 ntohs(((struct sockaddr_in6 *)ca)->sin6_port) >= 1985 IPPORT_RESERVED)) { 1986 cmn_err(CE_NOTE, 1987 "nfs_server: client %s%ssent NFS request from " 1988 "unprivileged port", 1989 client_name(req), client_addr(req, buf)); 1990 return (0); 1991 } 1992 } 1993 1994 /* 1995 * return 1 on success or 0 on failure 1996 */ 1997 stat = sec_svc_getcred(req, cr, &principal, &nfsflavor); 1998 1999 /* 2000 * A failed AUTH_UNIX svc_get_cred() implies we couldn't set 2001 * the credentials; below we map that to anonymous. 2002 */ 2003 if (!stat && nfsflavor != AUTH_UNIX) { 2004 cmn_err(CE_NOTE, 2005 "nfs_server: couldn't get unix cred for %s", 2006 client_name(req)); 2007 return (0); 2008 } 2009 2010 /* 2011 * Short circuit checkauth() on operations that support the 2012 * public filehandle, and if the request for that operation 2013 * is using the public filehandle. Note that we must call 2014 * sec_svc_getcred() first so that xp_cookie is set to the 2015 * right value. Normally xp_cookie is just the RPC flavor 2016 * of the the request, but in the case of RPCSEC_GSS it 2017 * could be a pseudo flavor. 2018 */ 2019 if (publicfh_ok) 2020 return (1); 2021 2022 rpcflavor = req->rq_cred.oa_flavor; 2023 /* 2024 * Check if the auth flavor is valid for this export 2025 */ 2026 access = nfsauth_access(exi, req); 2027 if (access & NFSAUTH_DROP) 2028 return (-1); /* drop the request */ 2029 2030 if (access & NFSAUTH_DENIED) { 2031 /* 2032 * If anon_ok == 1 and we got NFSAUTH_DENIED, it was 2033 * probably due to the flavor not matching during the 2034 * the mount attempt. So map the flavor to AUTH_NONE 2035 * so that the credentials get mapped to the anonymous 2036 * user. 2037 */ 2038 if (anon_ok == 1) 2039 rpcflavor = AUTH_NONE; 2040 else 2041 return (0); /* deny access */ 2042 2043 } else if (access & NFSAUTH_MAPNONE) { 2044 /* 2045 * Access was granted even though the flavor mismatched 2046 * because AUTH_NONE was one of the exported flavors. 2047 */ 2048 rpcflavor = AUTH_NONE; 2049 2050 } else if (access & NFSAUTH_WRONGSEC) { 2051 /* 2052 * NFSAUTH_WRONGSEC is used for NFSv4. If we get here, 2053 * it means a client ignored the list of allowed flavors 2054 * returned via the MOUNT protocol. So we just disallow it! 2055 */ 2056 return (0); 2057 } 2058 2059 switch (rpcflavor) { 2060 case AUTH_NONE: 2061 anon_res = crsetugid(cr, exi->exi_export.ex_anon, 2062 exi->exi_export.ex_anon); 2063 (void) crsetgroups(cr, 0, NULL); 2064 break; 2065 2066 case AUTH_UNIX: 2067 if (!stat || crgetuid(cr) == 0 && !(access & NFSAUTH_ROOT)) { 2068 anon_res = crsetugid(cr, exi->exi_export.ex_anon, 2069 exi->exi_export.ex_anon); 2070 (void) crsetgroups(cr, 0, NULL); 2071 } else if (!stat || crgetuid(cr) == 0 && 2072 access & NFSAUTH_ROOT) { 2073 /* 2074 * It is root, so apply rootid to get real UID 2075 * Find the secinfo structure. We should be able 2076 * to find it by the time we reach here. 2077 * nfsauth_access() has done the checking. 2078 */ 2079 secp = NULL; 2080 for (i = 0; i < exi->exi_export.ex_seccnt; i++) { 2081 struct secinfo *sptr; 2082 sptr = &exi->exi_export.ex_secinfo[i]; 2083 if (sptr->s_secinfo.sc_nfsnum == nfsflavor) { 2084 secp = sptr; 2085 break; 2086 } 2087 } 2088 if (secp != NULL) { 2089 (void) crsetugid(cr, secp->s_rootid, 2090 secp->s_rootid); 2091 (void) crsetgroups(cr, 0, NULL); 2092 } 2093 } 2094 break; 2095 2096 case AUTH_DES: 2097 case RPCSEC_GSS: 2098 /* 2099 * Find the secinfo structure. We should be able 2100 * to find it by the time we reach here. 2101 * nfsauth_access() has done the checking. 2102 */ 2103 secp = NULL; 2104 for (i = 0; i < exi->exi_export.ex_seccnt; i++) { 2105 if (exi->exi_export.ex_secinfo[i].s_secinfo.sc_nfsnum == 2106 nfsflavor) { 2107 secp = &exi->exi_export.ex_secinfo[i]; 2108 break; 2109 } 2110 } 2111 2112 if (!secp) { 2113 cmn_err(CE_NOTE, "nfs_server: client %s%shad " 2114 "no secinfo data for flavor %d", 2115 client_name(req), client_addr(req, buf), 2116 nfsflavor); 2117 return (0); 2118 } 2119 2120 if (!checkwin(rpcflavor, secp->s_window, req)) { 2121 cmn_err(CE_NOTE, 2122 "nfs_server: client %s%sused invalid " 2123 "auth window value", 2124 client_name(req), client_addr(req, buf)); 2125 return (0); 2126 } 2127 2128 /* 2129 * Map root principals listed in the share's root= list to root, 2130 * and map any others principals that were mapped to root by RPC 2131 * to anon. 2132 */ 2133 if (principal && sec_svc_inrootlist(rpcflavor, principal, 2134 secp->s_rootcnt, secp->s_rootnames)) { 2135 if (crgetuid(cr) == 0 && secp->s_rootid == 0) 2136 return (1); 2137 2138 2139 (void) crsetugid(cr, secp->s_rootid, secp->s_rootid); 2140 2141 /* 2142 * NOTE: If and when kernel-land privilege tracing is 2143 * added this may have to be replaced with code that 2144 * retrieves root's supplementary groups (e.g., using 2145 * kgss_get_group_info(). In the meantime principals 2146 * mapped to uid 0 get all privileges, so setting cr's 2147 * supplementary groups for them does nothing. 2148 */ 2149 (void) crsetgroups(cr, 0, NULL); 2150 2151 return (1); 2152 } 2153 2154 /* 2155 * Not a root princ, or not in root list, map UID 0/nobody to 2156 * the anon ID for the share. (RPC sets cr's UIDs and GIDs to 2157 * UID_NOBODY and GID_NOBODY, respectively.) 2158 */ 2159 if (crgetuid(cr) != 0 && 2160 (crgetuid(cr) != UID_NOBODY || crgetgid(cr) != GID_NOBODY)) 2161 return (1); 2162 2163 anon_res = crsetugid(cr, exi->exi_export.ex_anon, 2164 exi->exi_export.ex_anon); 2165 (void) crsetgroups(cr, 0, NULL); 2166 break; 2167 default: 2168 return (0); 2169 } /* switch on rpcflavor */ 2170 2171 /* 2172 * Even if anon access is disallowed via ex_anon == -1, we allow 2173 * this access if anon_ok is set. So set creds to the default 2174 * "nobody" id. 2175 */ 2176 if (anon_res != 0) { 2177 if (anon_ok == 0) { 2178 cmn_err(CE_NOTE, 2179 "nfs_server: client %s%ssent wrong " 2180 "authentication for %s", 2181 client_name(req), client_addr(req, buf), 2182 exi->exi_export.ex_path ? 2183 exi->exi_export.ex_path : "?"); 2184 return (0); 2185 } 2186 2187 if (crsetugid(cr, UID_NOBODY, GID_NOBODY) != 0) 2188 return (0); 2189 } 2190 2191 return (1); 2192 } 2193 2194 /* 2195 * returns 0 on failure, -1 on a drop, -2 on wrong security flavor, 2196 * and 1 on success 2197 */ 2198 int 2199 checkauth4(struct compound_state *cs, struct svc_req *req) 2200 { 2201 int i, rpcflavor, access; 2202 struct secinfo *secp; 2203 char buf[MAXHOST + 1]; 2204 int anon_res = 0, nfsflavor; 2205 struct exportinfo *exi; 2206 cred_t *cr; 2207 caddr_t principal; 2208 2209 exi = cs->exi; 2210 cr = cs->cr; 2211 principal = cs->principal; 2212 nfsflavor = cs->nfsflavor; 2213 2214 ASSERT(cr != NULL); 2215 2216 rpcflavor = req->rq_cred.oa_flavor; 2217 cs->access &= ~CS_ACCESS_LIMITED; 2218 2219 /* 2220 * Check the access right per auth flavor on the vnode of 2221 * this export for the given request. 2222 */ 2223 access = nfsauth4_access(cs->exi, cs->vp, req); 2224 2225 if (access & NFSAUTH_WRONGSEC) 2226 return (-2); /* no access for this security flavor */ 2227 2228 if (access & NFSAUTH_DROP) 2229 return (-1); /* drop the request */ 2230 2231 if (access & NFSAUTH_DENIED) { 2232 2233 if (exi->exi_export.ex_seccnt > 0) 2234 return (0); /* deny access */ 2235 2236 } else if (access & NFSAUTH_LIMITED) { 2237 2238 cs->access |= CS_ACCESS_LIMITED; 2239 2240 } else if (access & NFSAUTH_MAPNONE) { 2241 /* 2242 * Access was granted even though the flavor mismatched 2243 * because AUTH_NONE was one of the exported flavors. 2244 */ 2245 rpcflavor = AUTH_NONE; 2246 } 2247 2248 /* 2249 * XXX probably need to redo some of it for nfsv4? 2250 * return 1 on success or 0 on failure 2251 */ 2252 2253 switch (rpcflavor) { 2254 case AUTH_NONE: 2255 anon_res = crsetugid(cr, exi->exi_export.ex_anon, 2256 exi->exi_export.ex_anon); 2257 (void) crsetgroups(cr, 0, NULL); 2258 break; 2259 2260 case AUTH_UNIX: 2261 if (crgetuid(cr) == 0 && !(access & NFSAUTH_ROOT)) { 2262 anon_res = crsetugid(cr, exi->exi_export.ex_anon, 2263 exi->exi_export.ex_anon); 2264 (void) crsetgroups(cr, 0, NULL); 2265 } else if (crgetuid(cr) == 0 && access & NFSAUTH_ROOT) { 2266 /* 2267 * It is root, so apply rootid to get real UID 2268 * Find the secinfo structure. We should be able 2269 * to find it by the time we reach here. 2270 * nfsauth_access() has done the checking. 2271 */ 2272 secp = NULL; 2273 for (i = 0; i < exi->exi_export.ex_seccnt; i++) { 2274 struct secinfo *sptr; 2275 sptr = &exi->exi_export.ex_secinfo[i]; 2276 if (sptr->s_secinfo.sc_nfsnum == nfsflavor) { 2277 secp = &exi->exi_export.ex_secinfo[i]; 2278 break; 2279 } 2280 } 2281 if (secp != NULL) { 2282 (void) crsetugid(cr, secp->s_rootid, 2283 secp->s_rootid); 2284 (void) crsetgroups(cr, 0, NULL); 2285 } 2286 } 2287 break; 2288 2289 default: 2290 /* 2291 * Find the secinfo structure. We should be able 2292 * to find it by the time we reach here. 2293 * nfsauth_access() has done the checking. 2294 */ 2295 secp = NULL; 2296 for (i = 0; i < exi->exi_export.ex_seccnt; i++) { 2297 if (exi->exi_export.ex_secinfo[i].s_secinfo.sc_nfsnum == 2298 nfsflavor) { 2299 secp = &exi->exi_export.ex_secinfo[i]; 2300 break; 2301 } 2302 } 2303 2304 if (!secp) { 2305 cmn_err(CE_NOTE, "nfs_server: client %s%shad " 2306 "no secinfo data for flavor %d", 2307 client_name(req), client_addr(req, buf), 2308 nfsflavor); 2309 return (0); 2310 } 2311 2312 if (!checkwin(rpcflavor, secp->s_window, req)) { 2313 cmn_err(CE_NOTE, 2314 "nfs_server: client %s%sused invalid " 2315 "auth window value", 2316 client_name(req), client_addr(req, buf)); 2317 return (0); 2318 } 2319 2320 /* 2321 * Map root principals listed in the share's root= list to root, 2322 * and map any others principals that were mapped to root by RPC 2323 * to anon. If not going to anon, set to rootid (root_mapping). 2324 */ 2325 if (principal && sec_svc_inrootlist(rpcflavor, principal, 2326 secp->s_rootcnt, secp->s_rootnames)) { 2327 if (crgetuid(cr) == 0 && secp->s_rootid == 0) 2328 return (1); 2329 2330 (void) crsetugid(cr, secp->s_rootid, secp->s_rootid); 2331 2332 /* 2333 * NOTE: If and when kernel-land privilege tracing is 2334 * added this may have to be replaced with code that 2335 * retrieves root's supplementary groups (e.g., using 2336 * kgss_get_group_info(). In the meantime principals 2337 * mapped to uid 0 get all privileges, so setting cr's 2338 * supplementary groups for them does nothing. 2339 */ 2340 (void) crsetgroups(cr, 0, NULL); 2341 2342 return (1); 2343 } 2344 2345 /* 2346 * Not a root princ, or not in root list, map UID 0/nobody to 2347 * the anon ID for the share. (RPC sets cr's UIDs and GIDs to 2348 * UID_NOBODY and GID_NOBODY, respectively.) 2349 */ 2350 if (crgetuid(cr) != 0 && 2351 (crgetuid(cr) != UID_NOBODY || crgetgid(cr) != GID_NOBODY)) 2352 return (1); 2353 2354 anon_res = crsetugid(cr, exi->exi_export.ex_anon, 2355 exi->exi_export.ex_anon); 2356 (void) crsetgroups(cr, 0, NULL); 2357 break; 2358 } /* switch on rpcflavor */ 2359 2360 /* 2361 * Even if anon access is disallowed via ex_anon == -1, we allow 2362 * this access if anon_ok is set. So set creds to the default 2363 * "nobody" id. 2364 */ 2365 2366 if (anon_res != 0) { 2367 cmn_err(CE_NOTE, 2368 "nfs_server: client %s%ssent wrong " 2369 "authentication for %s", 2370 client_name(req), client_addr(req, buf), 2371 exi->exi_export.ex_path ? 2372 exi->exi_export.ex_path : "?"); 2373 return (0); 2374 } 2375 2376 return (1); 2377 } 2378 2379 2380 static char * 2381 client_name(struct svc_req *req) 2382 { 2383 char *hostname = NULL; 2384 2385 /* 2386 * If it's a Unix cred then use the 2387 * hostname from the credential. 2388 */ 2389 if (req->rq_cred.oa_flavor == AUTH_UNIX) { 2390 hostname = ((struct authunix_parms *) 2391 req->rq_clntcred)->aup_machname; 2392 } 2393 if (hostname == NULL) 2394 hostname = ""; 2395 2396 return (hostname); 2397 } 2398 2399 static char * 2400 client_addr(struct svc_req *req, char *buf) 2401 { 2402 struct sockaddr *ca; 2403 uchar_t *b; 2404 char *frontspace = ""; 2405 2406 /* 2407 * We assume we are called in tandem with client_name and the 2408 * format string looks like "...client %s%sblah blah..." 2409 * 2410 * If it's a Unix cred then client_name returned 2411 * a host name, so we need insert a space between host name 2412 * and IP address. 2413 */ 2414 if (req->rq_cred.oa_flavor == AUTH_UNIX) 2415 frontspace = " "; 2416 2417 /* 2418 * Convert the caller's IP address to a dotted string 2419 */ 2420 ca = (struct sockaddr *)svc_getrpccaller(req->rq_xprt)->buf; 2421 2422 if (ca->sa_family == AF_INET) { 2423 b = (uchar_t *)&((struct sockaddr_in *)ca)->sin_addr; 2424 (void) sprintf(buf, "%s(%d.%d.%d.%d) ", frontspace, 2425 b[0] & 0xFF, b[1] & 0xFF, b[2] & 0xFF, b[3] & 0xFF); 2426 } else if (ca->sa_family == AF_INET6) { 2427 struct sockaddr_in6 *sin6; 2428 sin6 = (struct sockaddr_in6 *)ca; 2429 (void) kinet_ntop6((uchar_t *)&sin6->sin6_addr, 2430 buf, INET6_ADDRSTRLEN); 2431 2432 } else { 2433 2434 /* 2435 * No IP address to print. If there was a host name 2436 * printed, then we print a space. 2437 */ 2438 (void) sprintf(buf, frontspace); 2439 } 2440 2441 return (buf); 2442 } 2443 2444 /* 2445 * NFS Server initialization routine. This routine should only be called 2446 * once. It performs the following tasks: 2447 * - Call sub-initialization routines (localize access to variables) 2448 * - Initialize all locks 2449 * - initialize the version 3 write verifier 2450 */ 2451 int 2452 nfs_srvinit(void) 2453 { 2454 int error; 2455 2456 error = nfs_exportinit(); 2457 if (error != 0) 2458 return (error); 2459 error = rfs4_srvrinit(); 2460 if (error != 0) { 2461 nfs_exportfini(); 2462 return (error); 2463 } 2464 rfs_srvrinit(); 2465 rfs3_srvrinit(); 2466 nfsauth_init(); 2467 2468 /* Init the stuff to control start/stop */ 2469 nfs_server_upordown = NFS_SERVER_STOPPED; 2470 mutex_init(&nfs_server_upordown_lock, NULL, MUTEX_DEFAULT, NULL); 2471 cv_init(&nfs_server_upordown_cv, NULL, CV_DEFAULT, NULL); 2472 mutex_init(&rdma_wait_mutex, NULL, MUTEX_DEFAULT, NULL); 2473 cv_init(&rdma_wait_cv, NULL, CV_DEFAULT, NULL); 2474 2475 return (0); 2476 } 2477 2478 /* 2479 * NFS Server finalization routine. This routine is called to cleanup the 2480 * initialization work previously performed if the NFS server module could 2481 * not be loaded correctly. 2482 */ 2483 void 2484 nfs_srvfini(void) 2485 { 2486 nfsauth_fini(); 2487 rfs3_srvrfini(); 2488 rfs_srvrfini(); 2489 nfs_exportfini(); 2490 2491 mutex_destroy(&nfs_server_upordown_lock); 2492 cv_destroy(&nfs_server_upordown_cv); 2493 mutex_destroy(&rdma_wait_mutex); 2494 cv_destroy(&rdma_wait_cv); 2495 } 2496 2497 /* 2498 * Set up an iovec array of up to cnt pointers. 2499 */ 2500 2501 void 2502 mblk_to_iov(mblk_t *m, int cnt, struct iovec *iovp) 2503 { 2504 while (m != NULL && cnt-- > 0) { 2505 iovp->iov_base = (caddr_t)m->b_rptr; 2506 iovp->iov_len = (m->b_wptr - m->b_rptr); 2507 iovp++; 2508 m = m->b_cont; 2509 } 2510 } 2511 2512 /* 2513 * Common code between NFS Version 2 and NFS Version 3 for the public 2514 * filehandle multicomponent lookups. 2515 */ 2516 2517 /* 2518 * Public filehandle evaluation of a multi-component lookup, following 2519 * symbolic links, if necessary. This may result in a vnode in another 2520 * filesystem, which is OK as long as the other filesystem is exported. 2521 * 2522 * Note that the exi will be set either to NULL or a new reference to the 2523 * exportinfo struct that corresponds to the vnode of the multi-component path. 2524 * It is the callers responsibility to release this reference. 2525 */ 2526 int 2527 rfs_publicfh_mclookup(char *p, vnode_t *dvp, cred_t *cr, vnode_t **vpp, 2528 struct exportinfo **exi, struct sec_ol *sec) 2529 { 2530 int pathflag; 2531 vnode_t *mc_dvp = NULL; 2532 vnode_t *realvp; 2533 int error; 2534 2535 *exi = NULL; 2536 2537 /* 2538 * check if the given path is a url or native path. Since p is 2539 * modified by MCLpath(), it may be empty after returning from 2540 * there, and should be checked. 2541 */ 2542 if ((pathflag = MCLpath(&p)) == -1) 2543 return (EIO); 2544 2545 /* 2546 * If pathflag is SECURITY_QUERY, turn the SEC_QUERY bit 2547 * on in sec->sec_flags. This bit will later serve as an 2548 * indication in makefh_ol() or makefh3_ol() to overload the 2549 * filehandle to contain the sec modes used by the server for 2550 * the path. 2551 */ 2552 if (pathflag == SECURITY_QUERY) { 2553 if ((sec->sec_index = (uint_t)(*p)) > 0) { 2554 sec->sec_flags |= SEC_QUERY; 2555 p++; 2556 if ((pathflag = MCLpath(&p)) == -1) 2557 return (EIO); 2558 } else { 2559 cmn_err(CE_NOTE, 2560 "nfs_server: invalid security index %d, " 2561 "violating WebNFS SNEGO protocol.", sec->sec_index); 2562 return (EIO); 2563 } 2564 } 2565 2566 if (p[0] == '\0') { 2567 error = ENOENT; 2568 goto publicfh_done; 2569 } 2570 2571 error = rfs_pathname(p, &mc_dvp, vpp, dvp, cr, pathflag); 2572 2573 /* 2574 * If name resolves to "/" we get EINVAL since we asked for 2575 * the vnode of the directory that the file is in. Try again 2576 * with NULL directory vnode. 2577 */ 2578 if (error == EINVAL) { 2579 error = rfs_pathname(p, NULL, vpp, dvp, cr, pathflag); 2580 if (!error) { 2581 ASSERT(*vpp != NULL); 2582 if ((*vpp)->v_type == VDIR) { 2583 VN_HOLD(*vpp); 2584 mc_dvp = *vpp; 2585 } else { 2586 /* 2587 * This should not happen, the filesystem is 2588 * in an inconsistent state. Fail the lookup 2589 * at this point. 2590 */ 2591 VN_RELE(*vpp); 2592 error = EINVAL; 2593 } 2594 } 2595 } 2596 2597 if (error) 2598 goto publicfh_done; 2599 2600 if (*vpp == NULL) { 2601 error = ENOENT; 2602 goto publicfh_done; 2603 } 2604 2605 ASSERT(mc_dvp != NULL); 2606 ASSERT(*vpp != NULL); 2607 2608 if ((*vpp)->v_type == VDIR) { 2609 do { 2610 /* 2611 * *vpp may be an AutoFS node, so we perform 2612 * a VOP_ACCESS() to trigger the mount of the intended 2613 * filesystem, so we can perform the lookup in the 2614 * intended filesystem. 2615 */ 2616 (void) VOP_ACCESS(*vpp, 0, 0, cr, NULL); 2617 2618 /* 2619 * If vnode is covered, get the 2620 * the topmost vnode. 2621 */ 2622 if (vn_mountedvfs(*vpp) != NULL) { 2623 error = traverse(vpp); 2624 if (error) { 2625 VN_RELE(*vpp); 2626 goto publicfh_done; 2627 } 2628 } 2629 2630 if (VOP_REALVP(*vpp, &realvp, NULL) == 0 && 2631 realvp != *vpp) { 2632 /* 2633 * If realvp is different from *vpp 2634 * then release our reference on *vpp, so that 2635 * the export access check be performed on the 2636 * real filesystem instead. 2637 */ 2638 VN_HOLD(realvp); 2639 VN_RELE(*vpp); 2640 *vpp = realvp; 2641 } else { 2642 break; 2643 } 2644 /* LINTED */ 2645 } while (TRUE); 2646 2647 /* 2648 * Let nfs_vptexi() figure what the real parent is. 2649 */ 2650 VN_RELE(mc_dvp); 2651 mc_dvp = NULL; 2652 2653 } else { 2654 /* 2655 * If vnode is covered, get the 2656 * the topmost vnode. 2657 */ 2658 if (vn_mountedvfs(mc_dvp) != NULL) { 2659 error = traverse(&mc_dvp); 2660 if (error) { 2661 VN_RELE(*vpp); 2662 goto publicfh_done; 2663 } 2664 } 2665 2666 if (VOP_REALVP(mc_dvp, &realvp, NULL) == 0 && 2667 realvp != mc_dvp) { 2668 /* 2669 * *vpp is a file, obtain realvp of the parent 2670 * directory vnode. 2671 */ 2672 VN_HOLD(realvp); 2673 VN_RELE(mc_dvp); 2674 mc_dvp = realvp; 2675 } 2676 } 2677 2678 /* 2679 * The pathname may take us from the public filesystem to another. 2680 * If that's the case then just set the exportinfo to the new export 2681 * and build filehandle for it. Thanks to per-access checking there's 2682 * no security issues with doing this. If the client is not allowed 2683 * access to this new export then it will get an access error when it 2684 * tries to use the filehandle 2685 */ 2686 if (error = nfs_check_vpexi(mc_dvp, *vpp, kcred, exi)) { 2687 VN_RELE(*vpp); 2688 goto publicfh_done; 2689 } 2690 2691 /* 2692 * Not allowed access to pseudo exports. 2693 */ 2694 if (PSEUDO(*exi)) { 2695 error = ENOENT; 2696 VN_RELE(*vpp); 2697 goto publicfh_done; 2698 } 2699 2700 /* 2701 * Do a lookup for the index file. We know the index option doesn't 2702 * allow paths through handling in the share command, so mc_dvp will 2703 * be the parent for the index file vnode, if its present. Use 2704 * temporary pointers to preserve and reuse the vnode pointers of the 2705 * original directory in case there's no index file. Note that the 2706 * index file is a native path, and should not be interpreted by 2707 * the URL parser in rfs_pathname() 2708 */ 2709 if (((*exi)->exi_export.ex_flags & EX_INDEX) && 2710 ((*vpp)->v_type == VDIR) && (pathflag == URLPATH)) { 2711 vnode_t *tvp, *tmc_dvp; /* temporary vnode pointers */ 2712 2713 tmc_dvp = mc_dvp; 2714 mc_dvp = tvp = *vpp; 2715 2716 error = rfs_pathname((*exi)->exi_export.ex_index, NULL, vpp, 2717 mc_dvp, cr, NATIVEPATH); 2718 2719 if (error == ENOENT) { 2720 *vpp = tvp; 2721 mc_dvp = tmc_dvp; 2722 error = 0; 2723 } else { /* ok or error other than ENOENT */ 2724 if (tmc_dvp) 2725 VN_RELE(tmc_dvp); 2726 if (error) 2727 goto publicfh_done; 2728 2729 /* 2730 * Found a valid vp for index "filename". Sanity check 2731 * for odd case where a directory is provided as index 2732 * option argument and leads us to another filesystem 2733 */ 2734 2735 /* Release the reference on the old exi value */ 2736 ASSERT(*exi != NULL); 2737 exi_rele(*exi); 2738 2739 if (error = nfs_check_vpexi(mc_dvp, *vpp, kcred, exi)) { 2740 VN_RELE(*vpp); 2741 goto publicfh_done; 2742 } 2743 } 2744 } 2745 2746 publicfh_done: 2747 if (mc_dvp) 2748 VN_RELE(mc_dvp); 2749 2750 return (error); 2751 } 2752 2753 /* 2754 * Evaluate a multi-component path 2755 */ 2756 int 2757 rfs_pathname( 2758 char *path, /* pathname to evaluate */ 2759 vnode_t **dirvpp, /* ret for ptr to parent dir vnode */ 2760 vnode_t **compvpp, /* ret for ptr to component vnode */ 2761 vnode_t *startdvp, /* starting vnode */ 2762 cred_t *cr, /* user's credential */ 2763 int pathflag) /* flag to identify path, e.g. URL */ 2764 { 2765 char namebuf[TYPICALMAXPATHLEN]; 2766 struct pathname pn; 2767 int error; 2768 2769 /* 2770 * If pathname starts with '/', then set startdvp to root. 2771 */ 2772 if (*path == '/') { 2773 while (*path == '/') 2774 path++; 2775 2776 startdvp = rootdir; 2777 } 2778 2779 error = pn_get_buf(path, UIO_SYSSPACE, &pn, namebuf, sizeof (namebuf)); 2780 if (error == 0) { 2781 /* 2782 * Call the URL parser for URL paths to modify the original 2783 * string to handle any '%' encoded characters that exist. 2784 * Done here to avoid an extra bcopy in the lookup. 2785 * We need to be careful about pathlen's. We know that 2786 * rfs_pathname() is called with a non-empty path. However, 2787 * it could be emptied due to the path simply being all /'s, 2788 * which is valid to proceed with the lookup, or due to the 2789 * URL parser finding an encoded null character at the 2790 * beginning of path which should not proceed with the lookup. 2791 */ 2792 if (pn.pn_pathlen != 0 && pathflag == URLPATH) { 2793 URLparse(pn.pn_path); 2794 if ((pn.pn_pathlen = strlen(pn.pn_path)) == 0) 2795 return (ENOENT); 2796 } 2797 VN_HOLD(startdvp); 2798 error = lookuppnvp(&pn, NULL, NO_FOLLOW, dirvpp, compvpp, 2799 rootdir, startdvp, cr); 2800 } 2801 if (error == ENAMETOOLONG) { 2802 /* 2803 * This thread used a pathname > TYPICALMAXPATHLEN bytes long. 2804 */ 2805 if (error = pn_get(path, UIO_SYSSPACE, &pn)) 2806 return (error); 2807 if (pn.pn_pathlen != 0 && pathflag == URLPATH) { 2808 URLparse(pn.pn_path); 2809 if ((pn.pn_pathlen = strlen(pn.pn_path)) == 0) { 2810 pn_free(&pn); 2811 return (ENOENT); 2812 } 2813 } 2814 VN_HOLD(startdvp); 2815 error = lookuppnvp(&pn, NULL, NO_FOLLOW, dirvpp, compvpp, 2816 rootdir, startdvp, cr); 2817 pn_free(&pn); 2818 } 2819 2820 return (error); 2821 } 2822 2823 /* 2824 * Adapt the multicomponent lookup path depending on the pathtype 2825 */ 2826 static int 2827 MCLpath(char **path) 2828 { 2829 unsigned char c = (unsigned char)**path; 2830 2831 /* 2832 * If the MCL path is between 0x20 and 0x7E (graphic printable 2833 * character of the US-ASCII coded character set), its a URL path, 2834 * per RFC 1738. 2835 */ 2836 if (c >= 0x20 && c <= 0x7E) 2837 return (URLPATH); 2838 2839 /* 2840 * If the first octet of the MCL path is not an ASCII character 2841 * then it must be interpreted as a tag value that describes the 2842 * format of the remaining octets of the MCL path. 2843 * 2844 * If the first octet of the MCL path is 0x81 it is a query 2845 * for the security info. 2846 */ 2847 switch (c) { 2848 case 0x80: /* native path, i.e. MCL via mount protocol */ 2849 (*path)++; 2850 return (NATIVEPATH); 2851 case 0x81: /* security query */ 2852 (*path)++; 2853 return (SECURITY_QUERY); 2854 default: 2855 return (-1); 2856 } 2857 } 2858 2859 #define fromhex(c) ((c >= '0' && c <= '9') ? (c - '0') : \ 2860 ((c >= 'A' && c <= 'F') ? (c - 'A' + 10) :\ 2861 ((c >= 'a' && c <= 'f') ? (c - 'a' + 10) : 0))) 2862 2863 /* 2864 * The implementation of URLparse guarantees that the final string will 2865 * fit in the original one. Replaces '%' occurrences followed by 2 characters 2866 * with its corresponding hexadecimal character. 2867 */ 2868 static void 2869 URLparse(char *str) 2870 { 2871 char *p, *q; 2872 2873 p = q = str; 2874 while (*p) { 2875 *q = *p; 2876 if (*p++ == '%') { 2877 if (*p) { 2878 *q = fromhex(*p) * 16; 2879 p++; 2880 if (*p) { 2881 *q += fromhex(*p); 2882 p++; 2883 } 2884 } 2885 } 2886 q++; 2887 } 2888 *q = '\0'; 2889 } 2890 2891 2892 /* 2893 * Get the export information for the lookup vnode, and verify its 2894 * useable. 2895 */ 2896 int 2897 nfs_check_vpexi(vnode_t *mc_dvp, vnode_t *vp, cred_t *cr, 2898 struct exportinfo **exi) 2899 { 2900 int walk; 2901 int error = 0; 2902 2903 *exi = nfs_vptoexi(mc_dvp, vp, cr, &walk, NULL, FALSE); 2904 if (*exi == NULL) 2905 error = EACCES; 2906 else { 2907 /* 2908 * If nosub is set for this export then 2909 * a lookup relative to the public fh 2910 * must not terminate below the 2911 * exported directory. 2912 */ 2913 if ((*exi)->exi_export.ex_flags & EX_NOSUB && walk > 0) 2914 error = EACCES; 2915 } 2916 2917 return (error); 2918 } 2919 2920 /* 2921 * Do the main work of handling HA-NFSv4 Resource Group failover on 2922 * Sun Cluster. 2923 * We need to detect whether any RG admin paths have been added or removed, 2924 * and adjust resources accordingly. 2925 * Currently we're using a very inefficient algorithm, ~ 2 * O(n**2). In 2926 * order to scale, the list and array of paths need to be held in more 2927 * suitable data structures. 2928 */ 2929 static void 2930 hanfsv4_failover(void) 2931 { 2932 int i, start_grace, numadded_paths = 0; 2933 char **added_paths = NULL; 2934 rfs4_dss_path_t *dss_path; 2935 2936 /* 2937 * Note: currently, rfs4_dss_pathlist cannot be NULL, since 2938 * it will always include an entry for NFS4_DSS_VAR_DIR. If we 2939 * make the latter dynamically specified too, the following will 2940 * need to be adjusted. 2941 */ 2942 2943 /* 2944 * First, look for removed paths: RGs that have been failed-over 2945 * away from this node. 2946 * Walk the "currently-serving" rfs4_dss_pathlist and, for each 2947 * path, check if it is on the "passed-in" rfs4_dss_newpaths array 2948 * from nfsd. If not, that RG path has been removed. 2949 * 2950 * Note that nfsd has sorted rfs4_dss_newpaths for us, and removed 2951 * any duplicates. 2952 */ 2953 dss_path = rfs4_dss_pathlist; 2954 do { 2955 int found = 0; 2956 char *path = dss_path->path; 2957 2958 /* used only for non-HA so may not be removed */ 2959 if (strcmp(path, NFS4_DSS_VAR_DIR) == 0) { 2960 dss_path = dss_path->next; 2961 continue; 2962 } 2963 2964 for (i = 0; i < rfs4_dss_numnewpaths; i++) { 2965 int cmpret; 2966 char *newpath = rfs4_dss_newpaths[i]; 2967 2968 /* 2969 * Since nfsd has sorted rfs4_dss_newpaths for us, 2970 * once the return from strcmp is negative we know 2971 * we've passed the point where "path" should be, 2972 * and can stop searching: "path" has been removed. 2973 */ 2974 cmpret = strcmp(path, newpath); 2975 if (cmpret < 0) 2976 break; 2977 if (cmpret == 0) { 2978 found = 1; 2979 break; 2980 } 2981 } 2982 2983 if (found == 0) { 2984 unsigned index = dss_path->index; 2985 rfs4_servinst_t *sip = dss_path->sip; 2986 rfs4_dss_path_t *path_next = dss_path->next; 2987 2988 /* 2989 * This path has been removed. 2990 * We must clear out the servinst reference to 2991 * it, since it's now owned by another 2992 * node: we should not attempt to touch it. 2993 */ 2994 ASSERT(dss_path == sip->dss_paths[index]); 2995 sip->dss_paths[index] = NULL; 2996 2997 /* remove from "currently-serving" list, and destroy */ 2998 remque(dss_path); 2999 /* allow for NUL */ 3000 kmem_free(dss_path->path, strlen(dss_path->path) + 1); 3001 kmem_free(dss_path, sizeof (rfs4_dss_path_t)); 3002 3003 dss_path = path_next; 3004 } else { 3005 /* path was found; not removed */ 3006 dss_path = dss_path->next; 3007 } 3008 } while (dss_path != rfs4_dss_pathlist); 3009 3010 /* 3011 * Now, look for added paths: RGs that have been failed-over 3012 * to this node. 3013 * Walk the "passed-in" rfs4_dss_newpaths array from nfsd and, 3014 * for each path, check if it is on the "currently-serving" 3015 * rfs4_dss_pathlist. If not, that RG path has been added. 3016 * 3017 * Note: we don't do duplicate detection here; nfsd does that for us. 3018 * 3019 * Note: numadded_paths <= rfs4_dss_numnewpaths, which gives us 3020 * an upper bound for the size needed for added_paths[numadded_paths]. 3021 */ 3022 3023 /* probably more space than we need, but guaranteed to be enough */ 3024 if (rfs4_dss_numnewpaths > 0) { 3025 size_t sz = rfs4_dss_numnewpaths * sizeof (char *); 3026 added_paths = kmem_zalloc(sz, KM_SLEEP); 3027 } 3028 3029 /* walk the "passed-in" rfs4_dss_newpaths array from nfsd */ 3030 for (i = 0; i < rfs4_dss_numnewpaths; i++) { 3031 int found = 0; 3032 char *newpath = rfs4_dss_newpaths[i]; 3033 3034 dss_path = rfs4_dss_pathlist; 3035 do { 3036 char *path = dss_path->path; 3037 3038 /* used only for non-HA */ 3039 if (strcmp(path, NFS4_DSS_VAR_DIR) == 0) { 3040 dss_path = dss_path->next; 3041 continue; 3042 } 3043 3044 if (strncmp(path, newpath, strlen(path)) == 0) { 3045 found = 1; 3046 break; 3047 } 3048 3049 dss_path = dss_path->next; 3050 } while (dss_path != rfs4_dss_pathlist); 3051 3052 if (found == 0) { 3053 added_paths[numadded_paths] = newpath; 3054 numadded_paths++; 3055 } 3056 } 3057 3058 /* did we find any added paths? */ 3059 if (numadded_paths > 0) { 3060 /* create a new server instance, and start its grace period */ 3061 start_grace = 1; 3062 rfs4_servinst_create(start_grace, numadded_paths, added_paths); 3063 3064 /* read in the stable storage state from these paths */ 3065 rfs4_dss_readstate(numadded_paths, added_paths); 3066 3067 /* 3068 * Multiple failovers during a grace period will cause 3069 * clients of the same resource group to be partitioned 3070 * into different server instances, with different 3071 * grace periods. Since clients of the same resource 3072 * group must be subject to the same grace period, 3073 * we need to reset all currently active grace periods. 3074 */ 3075 rfs4_grace_reset_all(); 3076 } 3077 3078 if (rfs4_dss_numnewpaths > 0) 3079 kmem_free(added_paths, rfs4_dss_numnewpaths * sizeof (char *)); 3080 } 3081