1 /* 2 * CDDL HEADER START 3 * 4 * The contents of this file are subject to the terms of the 5 * Common Development and Distribution License (the "License"). 6 * You may not use this file except in compliance with the License. 7 * 8 * You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE 9 * or http://www.opensolaris.org/os/licensing. 10 * See the License for the specific language governing permissions 11 * and limitations under the License. 12 * 13 * When distributing Covered Code, include this CDDL HEADER in each 14 * file and include the License file at usr/src/OPENSOLARIS.LICENSE. 15 * If applicable, add the following below this CDDL HEADER, with the 16 * fields enclosed by brackets "[]" replaced with your own identifying 17 * information: Portions Copyright [yyyy] [name of copyright owner] 18 * 19 * CDDL HEADER END 20 */ 21 /* 22 * Copyright (c) 1990, 2010, Oracle and/or its affiliates. All rights reserved. 23 * Copyright (c) 2011 Bayard G. Bell. All rights reserved. 24 * Copyright (c) 2013 by Delphix. All rights reserved. 25 * Copyright (c) 2017 Joyent Inc 26 * Copyright 2019 Nexenta by DDN, Inc. 27 * Copyright 2021 Racktop Systems, Inc. 28 */ 29 30 /* 31 * Copyright (c) 1983,1984,1985,1986,1987,1988,1989 AT&T. 32 * All rights reserved. 33 * Use is subject to license terms. 34 */ 35 36 #include <sys/param.h> 37 #include <sys/types.h> 38 #include <sys/systm.h> 39 #include <sys/cred.h> 40 #include <sys/proc.h> 41 #include <sys/user.h> 42 #include <sys/buf.h> 43 #include <sys/vfs.h> 44 #include <sys/vnode.h> 45 #include <sys/pathname.h> 46 #include <sys/uio.h> 47 #include <sys/file.h> 48 #include <sys/stat.h> 49 #include <sys/errno.h> 50 #include <sys/socket.h> 51 #include <sys/sysmacros.h> 52 #include <sys/siginfo.h> 53 #include <sys/tiuser.h> 54 #include <sys/statvfs.h> 55 #include <sys/stream.h> 56 #include <sys/strsun.h> 57 #include <sys/strsubr.h> 58 #include <sys/stropts.h> 59 #include <sys/timod.h> 60 #include <sys/t_kuser.h> 61 #include <sys/kmem.h> 62 #include <sys/kstat.h> 63 #include <sys/dirent.h> 64 #include <sys/cmn_err.h> 65 #include <sys/debug.h> 66 #include <sys/unistd.h> 67 #include <sys/vtrace.h> 68 #include <sys/mode.h> 69 #include <sys/acl.h> 70 #include <sys/sdt.h> 71 #include <sys/debug.h> 72 73 #include <rpc/types.h> 74 #include <rpc/auth.h> 75 #include <rpc/auth_unix.h> 76 #include <rpc/auth_des.h> 77 #include <rpc/svc.h> 78 #include <rpc/xdr.h> 79 #include <rpc/rpc_rdma.h> 80 81 #include <nfs/nfs.h> 82 #include <nfs/export.h> 83 #include <nfs/nfssys.h> 84 #include <nfs/nfs_clnt.h> 85 #include <nfs/nfs_acl.h> 86 #include <nfs/nfs_log.h> 87 #include <nfs/lm.h> 88 #include <nfs/nfs_dispatch.h> 89 #include <nfs/nfs4_drc.h> 90 91 #include <sys/modctl.h> 92 #include <sys/cladm.h> 93 #include <sys/clconf.h> 94 95 #include <sys/tsol/label.h> 96 97 #define MAXHOST 32 98 const char *kinet_ntop6(uchar_t *, char *, size_t); 99 100 /* 101 * Module linkage information. 102 */ 103 104 static struct modlmisc modlmisc = { 105 &mod_miscops, "NFS server module" 106 }; 107 108 static struct modlinkage modlinkage = { 109 MODREV_1, (void *)&modlmisc, NULL 110 }; 111 112 zone_key_t nfssrv_zone_key; 113 list_t nfssrv_globals_list; 114 krwlock_t nfssrv_globals_rwl; 115 116 kmem_cache_t *nfs_xuio_cache; 117 int nfs_loaned_buffers = 0; 118 119 /* array of paths passed-in from nfsd command-line; stored in nvlist */ 120 char **rfs4_dss_newpaths; 121 uint_t rfs4_dss_numnewpaths; 122 123 /* nvlists of all DSS paths: current, and before last warmstart */ 124 nvlist_t *rfs4_dss_paths, *rfs4_dss_oldpaths; 125 126 int 127 _init(void) 128 { 129 int status; 130 131 nfs_srvinit(); 132 133 status = mod_install((struct modlinkage *)&modlinkage); 134 if (status != 0) { 135 /* 136 * Could not load module, cleanup previous 137 * initialization work. 138 */ 139 nfs_srvfini(); 140 141 return (status); 142 } 143 144 /* 145 * Initialise some placeholders for nfssys() calls. These have 146 * to be declared by the nfs module, since that handles nfssys() 147 * calls - also used by NFS clients - but are provided by this 148 * nfssrv module. These also then serve as confirmation to the 149 * relevant code in nfs that nfssrv has been loaded, as they're 150 * initially NULL. 151 */ 152 nfs_srv_quiesce_func = nfs_srv_quiesce_all; 153 nfs_srv_dss_func = rfs4_dss_setpaths; 154 155 /* setup DSS paths here; must be done before initial server startup */ 156 rfs4_dss_paths = rfs4_dss_oldpaths = NULL; 157 158 /* initialize the copy reduction caches */ 159 160 nfs_xuio_cache = kmem_cache_create("nfs_xuio_cache", 161 sizeof (nfs_xuio_t), 0, NULL, NULL, NULL, NULL, NULL, 0); 162 163 return (status); 164 } 165 166 int 167 _fini() 168 { 169 return (EBUSY); 170 } 171 172 int 173 _info(struct modinfo *modinfop) 174 { 175 return (mod_info(&modlinkage, modinfop)); 176 } 177 178 /* 179 * PUBLICFH_CHECK() checks if the dispatch routine supports 180 * RPC_PUBLICFH_OK, if the filesystem is exported public, and if the 181 * incoming request is using the public filehandle. The check duplicates 182 * the exportmatch() call done in checkexport(), and we should consider 183 * modifying those routines to avoid the duplication. For now, we optimize 184 * by calling exportmatch() only after checking that the dispatch routine 185 * supports RPC_PUBLICFH_OK, and if the filesystem is explicitly exported 186 * public (i.e., not the placeholder). 187 */ 188 #define PUBLICFH_CHECK(ne, disp, exi, fsid, xfid) \ 189 ((disp->dis_flags & RPC_PUBLICFH_OK) && \ 190 ((exi->exi_export.ex_flags & EX_PUBLIC) || \ 191 (exi == ne->exi_public && exportmatch(ne->exi_root, \ 192 fsid, xfid)))) 193 194 static void nfs_srv_shutdown_all(int); 195 static void rfs4_server_start(nfs_globals_t *, int); 196 static void nullfree(void); 197 static void rfs_dispatch(struct svc_req *, SVCXPRT *); 198 static void acl_dispatch(struct svc_req *, SVCXPRT *); 199 static int checkauth(struct exportinfo *, struct svc_req *, cred_t *, int, 200 bool_t, bool_t *); 201 static char *client_name(struct svc_req *req); 202 static char *client_addr(struct svc_req *req, char *buf); 203 extern bool_t sec_svc_inrootlist(int, caddr_t, int, caddr_t *); 204 static void *nfs_server_zone_init(zoneid_t); 205 static void nfs_server_zone_fini(zoneid_t, void *); 206 static void nfs_server_zone_shutdown(zoneid_t, void *); 207 208 #define NFSLOG_COPY_NETBUF(exi, xprt, nb) { \ 209 (nb)->maxlen = (xprt)->xp_rtaddr.maxlen; \ 210 (nb)->len = (xprt)->xp_rtaddr.len; \ 211 (nb)->buf = kmem_alloc((nb)->len, KM_SLEEP); \ 212 bcopy((xprt)->xp_rtaddr.buf, (nb)->buf, (nb)->len); \ 213 } 214 215 /* 216 * Public Filehandle common nfs routines 217 */ 218 static int MCLpath(char **); 219 static void URLparse(char *); 220 221 /* 222 * NFS callout table. 223 * This table is used by svc_getreq() to dispatch a request with 224 * a given prog/vers pair to an appropriate service provider 225 * dispatch routine. 226 * 227 * NOTE: ordering is relied upon below when resetting the version min/max 228 * for NFS_PROGRAM. Careful, if this is ever changed. 229 */ 230 static SVC_CALLOUT __nfs_sc_clts[] = { 231 { NFS_PROGRAM, NFS_VERSMIN, NFS_VERSMAX, rfs_dispatch }, 232 { NFS_ACL_PROGRAM, NFS_ACL_VERSMIN, NFS_ACL_VERSMAX, acl_dispatch } 233 }; 234 235 static SVC_CALLOUT_TABLE nfs_sct_clts = { 236 sizeof (__nfs_sc_clts) / sizeof (__nfs_sc_clts[0]), FALSE, 237 __nfs_sc_clts 238 }; 239 240 static SVC_CALLOUT __nfs_sc_cots[] = { 241 { NFS_PROGRAM, NFS_VERSMIN, NFS_VERSMAX, rfs_dispatch }, 242 { NFS_ACL_PROGRAM, NFS_ACL_VERSMIN, NFS_ACL_VERSMAX, acl_dispatch } 243 }; 244 245 static SVC_CALLOUT_TABLE nfs_sct_cots = { 246 sizeof (__nfs_sc_cots) / sizeof (__nfs_sc_cots[0]), FALSE, __nfs_sc_cots 247 }; 248 249 static SVC_CALLOUT __nfs_sc_rdma[] = { 250 { NFS_PROGRAM, NFS_VERSMIN, NFS_VERSMAX, rfs_dispatch }, 251 { NFS_ACL_PROGRAM, NFS_ACL_VERSMIN, NFS_ACL_VERSMAX, acl_dispatch } 252 }; 253 254 static SVC_CALLOUT_TABLE nfs_sct_rdma = { 255 sizeof (__nfs_sc_rdma) / sizeof (__nfs_sc_rdma[0]), FALSE, __nfs_sc_rdma 256 }; 257 258 /* 259 * DSS: distributed stable storage 260 * lists of all DSS paths: current, and before last warmstart 261 */ 262 nvlist_t *rfs4_dss_paths, *rfs4_dss_oldpaths; 263 264 int rfs4_dispatch(struct rpcdisp *, struct svc_req *, SVCXPRT *, char *); 265 bool_t rfs4_minorvers_mismatch(struct svc_req *, SVCXPRT *, void *); 266 267 /* 268 * Stash NFS zone globals in TSD to avoid some lock contention 269 * from frequent zone_getspecific calls. 270 */ 271 static uint_t nfs_server_tsd_key; 272 273 nfs_globals_t * 274 nfs_srv_getzg(void) 275 { 276 nfs_globals_t *ng; 277 278 ng = tsd_get(nfs_server_tsd_key); 279 if (ng == NULL) { 280 ng = zone_getspecific(nfssrv_zone_key, curzone); 281 (void) tsd_set(nfs_server_tsd_key, ng); 282 } 283 284 return (ng); 285 } 286 287 /* 288 * Will be called at the point the server pool is being unregistered 289 * from the pool list. From that point onwards, the pool is waiting 290 * to be drained and as such the server state is stale and pertains 291 * to the old instantiation of the NFS server pool. 292 */ 293 void 294 nfs_srv_offline(void) 295 { 296 nfs_globals_t *ng; 297 298 ng = nfs_srv_getzg(); 299 300 mutex_enter(&ng->nfs_server_upordown_lock); 301 if (ng->nfs_server_upordown == NFS_SERVER_RUNNING) { 302 ng->nfs_server_upordown = NFS_SERVER_OFFLINE; 303 } 304 mutex_exit(&ng->nfs_server_upordown_lock); 305 } 306 307 /* 308 * Will be called at the point the server pool is being destroyed so 309 * all transports have been closed and no service threads are in 310 * existence. 311 * 312 * If we quiesce the server, we're shutting it down without destroying the 313 * server state. This allows it to warm start subsequently. 314 */ 315 void 316 nfs_srv_stop_all(void) 317 { 318 int quiesce = 0; 319 nfs_srv_shutdown_all(quiesce); 320 } 321 322 /* 323 * This alternative shutdown routine can be requested via nfssys() 324 */ 325 void 326 nfs_srv_quiesce_all(void) 327 { 328 int quiesce = 1; 329 nfs_srv_shutdown_all(quiesce); 330 } 331 332 static void 333 nfs_srv_shutdown_all(int quiesce) 334 { 335 nfs_globals_t *ng = nfs_srv_getzg(); 336 337 mutex_enter(&ng->nfs_server_upordown_lock); 338 if (quiesce) { 339 if (ng->nfs_server_upordown == NFS_SERVER_RUNNING || 340 ng->nfs_server_upordown == NFS_SERVER_OFFLINE) { 341 ng->nfs_server_upordown = NFS_SERVER_QUIESCED; 342 cv_signal(&ng->nfs_server_upordown_cv); 343 344 /* reset DSS state */ 345 rfs4_dss_numnewpaths = 0; 346 rfs4_dss_newpaths = NULL; 347 348 cmn_err(CE_NOTE, "nfs_server: server is now quiesced; " 349 "NFSv4 state has been preserved"); 350 } 351 } else { 352 if (ng->nfs_server_upordown == NFS_SERVER_OFFLINE) { 353 ng->nfs_server_upordown = NFS_SERVER_STOPPING; 354 mutex_exit(&ng->nfs_server_upordown_lock); 355 rfs4_state_zone_fini(); 356 rfs4_fini_drc(); 357 mutex_enter(&ng->nfs_server_upordown_lock); 358 ng->nfs_server_upordown = NFS_SERVER_STOPPED; 359 360 /* reset DSS state */ 361 rfs4_dss_numnewpaths = 0; 362 rfs4_dss_newpaths = NULL; 363 364 cv_signal(&ng->nfs_server_upordown_cv); 365 } 366 } 367 mutex_exit(&ng->nfs_server_upordown_lock); 368 } 369 370 static int 371 nfs_srv_set_sc_versions(struct file *fp, SVC_CALLOUT_TABLE **sctpp, 372 rpcvers_t versmin, rpcvers_t versmax) 373 { 374 struct strioctl strioc; 375 struct T_info_ack tinfo; 376 int error, retval; 377 378 /* 379 * Find out what type of transport this is. 380 */ 381 strioc.ic_cmd = TI_GETINFO; 382 strioc.ic_timout = -1; 383 strioc.ic_len = sizeof (tinfo); 384 strioc.ic_dp = (char *)&tinfo; 385 tinfo.PRIM_type = T_INFO_REQ; 386 387 error = strioctl(fp->f_vnode, I_STR, (intptr_t)&strioc, 0, K_TO_K, 388 CRED(), &retval); 389 if (error || retval) 390 return (error); 391 392 /* 393 * Based on our query of the transport type... 394 * 395 * Reset the min/max versions based on the caller's request 396 * NOTE: This assumes that NFS_PROGRAM is first in the array!! 397 * And the second entry is the NFS_ACL_PROGRAM. 398 */ 399 switch (tinfo.SERV_type) { 400 case T_CLTS: 401 if (versmax == NFS_V4) 402 return (EINVAL); 403 __nfs_sc_clts[0].sc_versmin = versmin; 404 __nfs_sc_clts[0].sc_versmax = versmax; 405 __nfs_sc_clts[1].sc_versmin = versmin; 406 __nfs_sc_clts[1].sc_versmax = versmax; 407 *sctpp = &nfs_sct_clts; 408 break; 409 case T_COTS: 410 case T_COTS_ORD: 411 __nfs_sc_cots[0].sc_versmin = versmin; 412 __nfs_sc_cots[0].sc_versmax = versmax; 413 /* For the NFS_ACL program, check the max version */ 414 if (versmax > NFS_ACL_VERSMAX) 415 versmax = NFS_ACL_VERSMAX; 416 __nfs_sc_cots[1].sc_versmin = versmin; 417 __nfs_sc_cots[1].sc_versmax = versmax; 418 *sctpp = &nfs_sct_cots; 419 break; 420 default: 421 error = EINVAL; 422 } 423 424 return (error); 425 } 426 427 /* 428 * NFS Server system call. 429 * Does all of the work of running a NFS server. 430 * uap->fd is the fd of an open transport provider 431 */ 432 int 433 nfs_svc(struct nfs_svc_args *arg, model_t model) 434 { 435 nfs_globals_t *ng; 436 file_t *fp; 437 SVCMASTERXPRT *xprt; 438 int error; 439 int readsize; 440 char buf[KNC_STRSIZE]; 441 size_t len; 442 STRUCT_HANDLE(nfs_svc_args, uap); 443 struct netbuf addrmask; 444 SVC_CALLOUT_TABLE *sctp = NULL; 445 446 #ifdef lint 447 model = model; /* STRUCT macros don't always refer to it */ 448 #endif 449 450 ng = nfs_srv_getzg(); 451 STRUCT_SET_HANDLE(uap, model, arg); 452 453 /* Check privileges in nfssys() */ 454 455 if ((fp = getf(STRUCT_FGET(uap, fd))) == NULL) 456 return (EBADF); 457 458 /* Setup global file handle in nfs_export */ 459 if ((error = nfs_export_get_rootfh(ng)) != 0) 460 return (error); 461 462 /* 463 * Set read buffer size to rsize 464 * and add room for RPC headers. 465 */ 466 readsize = nfs3tsize() + (RPC_MAXDATASIZE - NFS_MAXDATA); 467 if (readsize < RPC_MAXDATASIZE) 468 readsize = RPC_MAXDATASIZE; 469 470 error = copyinstr((const char *)STRUCT_FGETP(uap, netid), buf, 471 KNC_STRSIZE, &len); 472 if (error) { 473 releasef(STRUCT_FGET(uap, fd)); 474 return (error); 475 } 476 477 addrmask.len = STRUCT_FGET(uap, addrmask.len); 478 addrmask.maxlen = STRUCT_FGET(uap, addrmask.maxlen); 479 addrmask.buf = kmem_alloc(addrmask.maxlen, KM_SLEEP); 480 error = copyin(STRUCT_FGETP(uap, addrmask.buf), addrmask.buf, 481 addrmask.len); 482 if (error) { 483 releasef(STRUCT_FGET(uap, fd)); 484 kmem_free(addrmask.buf, addrmask.maxlen); 485 return (error); 486 } 487 488 ng->nfs_versmin = STRUCT_FGET(uap, versmin); 489 ng->nfs_versmax = STRUCT_FGET(uap, versmax); 490 491 /* Double check the vers min/max ranges */ 492 if ((ng->nfs_versmin > ng->nfs_versmax) || 493 (ng->nfs_versmin < NFS_VERSMIN) || 494 (ng->nfs_versmax > NFS_VERSMAX)) { 495 ng->nfs_versmin = NFS_VERSMIN_DEFAULT; 496 ng->nfs_versmax = NFS_VERSMAX_DEFAULT; 497 } 498 499 if (error = nfs_srv_set_sc_versions(fp, &sctp, ng->nfs_versmin, 500 ng->nfs_versmax)) { 501 releasef(STRUCT_FGET(uap, fd)); 502 kmem_free(addrmask.buf, addrmask.maxlen); 503 return (error); 504 } 505 506 /* Initialize nfsv4 server */ 507 if (ng->nfs_versmax == (rpcvers_t)NFS_V4) 508 rfs4_server_start(ng, STRUCT_FGET(uap, delegation)); 509 510 /* Create a transport handle. */ 511 error = svc_tli_kcreate(fp, readsize, buf, &addrmask, &xprt, 512 sctp, NULL, NFS_SVCPOOL_ID, TRUE); 513 514 if (error) 515 kmem_free(addrmask.buf, addrmask.maxlen); 516 517 releasef(STRUCT_FGET(uap, fd)); 518 519 /* HA-NFSv4: save the cluster nodeid */ 520 if (cluster_bootflags & CLUSTER_BOOTED) 521 lm_global_nlmid = clconf_get_nodeid(); 522 523 return (error); 524 } 525 526 static void 527 rfs4_server_start(nfs_globals_t *ng, int nfs4_srv_delegation) 528 { 529 /* 530 * Determine if the server has previously been "started" and 531 * if not, do the per instance initialization 532 */ 533 mutex_enter(&ng->nfs_server_upordown_lock); 534 535 if (ng->nfs_server_upordown != NFS_SERVER_RUNNING) { 536 /* Do we need to stop and wait on the previous server? */ 537 while (ng->nfs_server_upordown == NFS_SERVER_STOPPING || 538 ng->nfs_server_upordown == NFS_SERVER_OFFLINE) 539 cv_wait(&ng->nfs_server_upordown_cv, 540 &ng->nfs_server_upordown_lock); 541 542 if (ng->nfs_server_upordown != NFS_SERVER_RUNNING) { 543 (void) svc_pool_control(NFS_SVCPOOL_ID, 544 SVCPSET_UNREGISTER_PROC, (void *)&nfs_srv_offline); 545 (void) svc_pool_control(NFS_SVCPOOL_ID, 546 SVCPSET_SHUTDOWN_PROC, (void *)&nfs_srv_stop_all); 547 548 rfs4_do_server_start(ng->nfs_server_upordown, 549 nfs4_srv_delegation, 550 cluster_bootflags & CLUSTER_BOOTED); 551 552 ng->nfs_server_upordown = NFS_SERVER_RUNNING; 553 } 554 cv_signal(&ng->nfs_server_upordown_cv); 555 } 556 mutex_exit(&ng->nfs_server_upordown_lock); 557 } 558 559 /* 560 * If RDMA device available, 561 * start RDMA listener. 562 */ 563 int 564 rdma_start(struct rdma_svc_args *rsa) 565 { 566 nfs_globals_t *ng; 567 int error; 568 rdma_xprt_group_t started_rdma_xprts; 569 rdma_stat stat; 570 int svc_state = 0; 571 572 /* Double check the vers min/max ranges */ 573 if ((rsa->nfs_versmin > rsa->nfs_versmax) || 574 (rsa->nfs_versmin < NFS_VERSMIN) || 575 (rsa->nfs_versmax > NFS_VERSMAX)) { 576 rsa->nfs_versmin = NFS_VERSMIN_DEFAULT; 577 rsa->nfs_versmax = NFS_VERSMAX_DEFAULT; 578 } 579 580 ng = nfs_srv_getzg(); 581 ng->nfs_versmin = rsa->nfs_versmin; 582 ng->nfs_versmax = rsa->nfs_versmax; 583 584 /* Set the versions in the callout table */ 585 __nfs_sc_rdma[0].sc_versmin = rsa->nfs_versmin; 586 __nfs_sc_rdma[0].sc_versmax = rsa->nfs_versmax; 587 /* For the NFS_ACL program, check the max version */ 588 __nfs_sc_rdma[1].sc_versmin = rsa->nfs_versmin; 589 if (rsa->nfs_versmax > NFS_ACL_VERSMAX) 590 __nfs_sc_rdma[1].sc_versmax = NFS_ACL_VERSMAX; 591 else 592 __nfs_sc_rdma[1].sc_versmax = rsa->nfs_versmax; 593 594 /* Initialize nfsv4 server */ 595 if (rsa->nfs_versmax == (rpcvers_t)NFS_V4) 596 rfs4_server_start(ng, rsa->delegation); 597 598 started_rdma_xprts.rtg_count = 0; 599 started_rdma_xprts.rtg_listhead = NULL; 600 started_rdma_xprts.rtg_poolid = rsa->poolid; 601 602 restart: 603 error = svc_rdma_kcreate(rsa->netid, &nfs_sct_rdma, rsa->poolid, 604 &started_rdma_xprts); 605 606 svc_state = !error; 607 608 while (!error) { 609 610 /* 611 * wait till either interrupted by a signal on 612 * nfs service stop/restart or signalled by a 613 * rdma attach/detatch. 614 */ 615 616 stat = rdma_kwait(); 617 618 /* 619 * stop services if running -- either on a HCA detach event 620 * or if the nfs service is stopped/restarted. 621 */ 622 623 if ((stat == RDMA_HCA_DETACH || stat == RDMA_INTR) && 624 svc_state) { 625 rdma_stop(&started_rdma_xprts); 626 svc_state = 0; 627 } 628 629 /* 630 * nfs service stop/restart, break out of the 631 * wait loop and return; 632 */ 633 if (stat == RDMA_INTR) 634 return (0); 635 636 /* 637 * restart stopped services on a HCA attach event 638 * (if not already running) 639 */ 640 641 if ((stat == RDMA_HCA_ATTACH) && (svc_state == 0)) 642 goto restart; 643 644 /* 645 * loop until a nfs service stop/restart 646 */ 647 } 648 649 return (error); 650 } 651 652 /* ARGSUSED */ 653 void 654 rpc_null(caddr_t *argp, caddr_t *resp, struct exportinfo *exi, 655 struct svc_req *req, cred_t *cr, bool_t ro) 656 { 657 } 658 659 /* ARGSUSED */ 660 void 661 rpc_null_v3(caddr_t *argp, caddr_t *resp, struct exportinfo *exi, 662 struct svc_req *req, cred_t *cr, bool_t ro) 663 { 664 DTRACE_NFSV3_4(op__null__start, struct svc_req *, req, 665 cred_t *, cr, vnode_t *, NULL, struct exportinfo *, exi); 666 DTRACE_NFSV3_4(op__null__done, struct svc_req *, req, 667 cred_t *, cr, vnode_t *, NULL, struct exportinfo *, exi); 668 } 669 670 /* ARGSUSED */ 671 static void 672 rfs_error(caddr_t *argp, caddr_t *resp, struct exportinfo *exi, 673 struct svc_req *req, cred_t *cr, bool_t ro) 674 { 675 /* return (EOPNOTSUPP); */ 676 } 677 678 static void 679 nullfree(void) 680 { 681 } 682 683 static char *rfscallnames_v2[] = { 684 "RFS2_NULL", 685 "RFS2_GETATTR", 686 "RFS2_SETATTR", 687 "RFS2_ROOT", 688 "RFS2_LOOKUP", 689 "RFS2_READLINK", 690 "RFS2_READ", 691 "RFS2_WRITECACHE", 692 "RFS2_WRITE", 693 "RFS2_CREATE", 694 "RFS2_REMOVE", 695 "RFS2_RENAME", 696 "RFS2_LINK", 697 "RFS2_SYMLINK", 698 "RFS2_MKDIR", 699 "RFS2_RMDIR", 700 "RFS2_READDIR", 701 "RFS2_STATFS" 702 }; 703 704 static struct rpcdisp rfsdisptab_v2[] = { 705 /* 706 * NFS VERSION 2 707 */ 708 709 /* RFS_NULL = 0 */ 710 {rpc_null, 711 xdr_void, NULL_xdrproc_t, 0, 712 xdr_void, NULL_xdrproc_t, 0, 713 nullfree, RPC_IDEMPOTENT, 714 0}, 715 716 /* RFS_GETATTR = 1 */ 717 {rfs_getattr, 718 xdr_fhandle, xdr_fastfhandle, sizeof (fhandle_t), 719 xdr_attrstat, xdr_fastattrstat, sizeof (struct nfsattrstat), 720 nullfree, RPC_IDEMPOTENT|RPC_ALLOWANON|RPC_MAPRESP, 721 rfs_getattr_getfh}, 722 723 /* RFS_SETATTR = 2 */ 724 {rfs_setattr, 725 xdr_saargs, NULL_xdrproc_t, sizeof (struct nfssaargs), 726 xdr_attrstat, xdr_fastattrstat, sizeof (struct nfsattrstat), 727 nullfree, RPC_MAPRESP, 728 rfs_setattr_getfh}, 729 730 /* RFS_ROOT = 3 *** NO LONGER SUPPORTED *** */ 731 {rfs_error, 732 xdr_void, NULL_xdrproc_t, 0, 733 xdr_void, NULL_xdrproc_t, 0, 734 nullfree, RPC_IDEMPOTENT, 735 0}, 736 737 /* RFS_LOOKUP = 4 */ 738 {rfs_lookup, 739 xdr_diropargs, NULL_xdrproc_t, sizeof (struct nfsdiropargs), 740 xdr_diropres, xdr_fastdiropres, sizeof (struct nfsdiropres), 741 nullfree, RPC_IDEMPOTENT|RPC_MAPRESP|RPC_PUBLICFH_OK, 742 rfs_lookup_getfh}, 743 744 /* RFS_READLINK = 5 */ 745 {rfs_readlink, 746 xdr_fhandle, xdr_fastfhandle, sizeof (fhandle_t), 747 xdr_rdlnres, NULL_xdrproc_t, sizeof (struct nfsrdlnres), 748 rfs_rlfree, RPC_IDEMPOTENT, 749 rfs_readlink_getfh}, 750 751 /* RFS_READ = 6 */ 752 {rfs_read, 753 xdr_readargs, NULL_xdrproc_t, sizeof (struct nfsreadargs), 754 xdr_rdresult, NULL_xdrproc_t, sizeof (struct nfsrdresult), 755 rfs_rdfree, RPC_IDEMPOTENT, 756 rfs_read_getfh}, 757 758 /* RFS_WRITECACHE = 7 *** NO LONGER SUPPORTED *** */ 759 {rfs_error, 760 xdr_void, NULL_xdrproc_t, 0, 761 xdr_void, NULL_xdrproc_t, 0, 762 nullfree, RPC_IDEMPOTENT, 763 0}, 764 765 /* RFS_WRITE = 8 */ 766 {rfs_write, 767 xdr_writeargs, NULL_xdrproc_t, sizeof (struct nfswriteargs), 768 xdr_attrstat, xdr_fastattrstat, sizeof (struct nfsattrstat), 769 nullfree, RPC_MAPRESP, 770 rfs_write_getfh}, 771 772 /* RFS_CREATE = 9 */ 773 {rfs_create, 774 xdr_creatargs, NULL_xdrproc_t, sizeof (struct nfscreatargs), 775 xdr_diropres, xdr_fastdiropres, sizeof (struct nfsdiropres), 776 nullfree, RPC_MAPRESP, 777 rfs_create_getfh}, 778 779 /* RFS_REMOVE = 10 */ 780 {rfs_remove, 781 xdr_diropargs, NULL_xdrproc_t, sizeof (struct nfsdiropargs), 782 #ifdef _LITTLE_ENDIAN 783 xdr_enum, xdr_fastenum, sizeof (enum nfsstat), 784 #else 785 xdr_enum, NULL_xdrproc_t, sizeof (enum nfsstat), 786 #endif 787 nullfree, RPC_MAPRESP, 788 rfs_remove_getfh}, 789 790 /* RFS_RENAME = 11 */ 791 {rfs_rename, 792 xdr_rnmargs, NULL_xdrproc_t, sizeof (struct nfsrnmargs), 793 #ifdef _LITTLE_ENDIAN 794 xdr_enum, xdr_fastenum, sizeof (enum nfsstat), 795 #else 796 xdr_enum, NULL_xdrproc_t, sizeof (enum nfsstat), 797 #endif 798 nullfree, RPC_MAPRESP, 799 rfs_rename_getfh}, 800 801 /* RFS_LINK = 12 */ 802 {rfs_link, 803 xdr_linkargs, NULL_xdrproc_t, sizeof (struct nfslinkargs), 804 #ifdef _LITTLE_ENDIAN 805 xdr_enum, xdr_fastenum, sizeof (enum nfsstat), 806 #else 807 xdr_enum, NULL_xdrproc_t, sizeof (enum nfsstat), 808 #endif 809 nullfree, RPC_MAPRESP, 810 rfs_link_getfh}, 811 812 /* RFS_SYMLINK = 13 */ 813 {rfs_symlink, 814 xdr_slargs, NULL_xdrproc_t, sizeof (struct nfsslargs), 815 #ifdef _LITTLE_ENDIAN 816 xdr_enum, xdr_fastenum, sizeof (enum nfsstat), 817 #else 818 xdr_enum, NULL_xdrproc_t, sizeof (enum nfsstat), 819 #endif 820 nullfree, RPC_MAPRESP, 821 rfs_symlink_getfh}, 822 823 /* RFS_MKDIR = 14 */ 824 {rfs_mkdir, 825 xdr_creatargs, NULL_xdrproc_t, sizeof (struct nfscreatargs), 826 xdr_diropres, xdr_fastdiropres, sizeof (struct nfsdiropres), 827 nullfree, RPC_MAPRESP, 828 rfs_mkdir_getfh}, 829 830 /* RFS_RMDIR = 15 */ 831 {rfs_rmdir, 832 xdr_diropargs, NULL_xdrproc_t, sizeof (struct nfsdiropargs), 833 #ifdef _LITTLE_ENDIAN 834 xdr_enum, xdr_fastenum, sizeof (enum nfsstat), 835 #else 836 xdr_enum, NULL_xdrproc_t, sizeof (enum nfsstat), 837 #endif 838 nullfree, RPC_MAPRESP, 839 rfs_rmdir_getfh}, 840 841 /* RFS_READDIR = 16 */ 842 {rfs_readdir, 843 xdr_rddirargs, NULL_xdrproc_t, sizeof (struct nfsrddirargs), 844 xdr_putrddirres, NULL_xdrproc_t, sizeof (struct nfsrddirres), 845 rfs_rddirfree, RPC_IDEMPOTENT, 846 rfs_readdir_getfh}, 847 848 /* RFS_STATFS = 17 */ 849 {rfs_statfs, 850 xdr_fhandle, xdr_fastfhandle, sizeof (fhandle_t), 851 xdr_statfs, xdr_faststatfs, sizeof (struct nfsstatfs), 852 nullfree, RPC_IDEMPOTENT|RPC_ALLOWANON|RPC_MAPRESP, 853 rfs_statfs_getfh}, 854 }; 855 856 static char *rfscallnames_v3[] = { 857 "RFS3_NULL", 858 "RFS3_GETATTR", 859 "RFS3_SETATTR", 860 "RFS3_LOOKUP", 861 "RFS3_ACCESS", 862 "RFS3_READLINK", 863 "RFS3_READ", 864 "RFS3_WRITE", 865 "RFS3_CREATE", 866 "RFS3_MKDIR", 867 "RFS3_SYMLINK", 868 "RFS3_MKNOD", 869 "RFS3_REMOVE", 870 "RFS3_RMDIR", 871 "RFS3_RENAME", 872 "RFS3_LINK", 873 "RFS3_READDIR", 874 "RFS3_READDIRPLUS", 875 "RFS3_FSSTAT", 876 "RFS3_FSINFO", 877 "RFS3_PATHCONF", 878 "RFS3_COMMIT" 879 }; 880 881 static struct rpcdisp rfsdisptab_v3[] = { 882 /* 883 * NFS VERSION 3 884 */ 885 886 /* RFS_NULL = 0 */ 887 {rpc_null_v3, 888 xdr_void, NULL_xdrproc_t, 0, 889 xdr_void, NULL_xdrproc_t, 0, 890 nullfree, RPC_IDEMPOTENT, 891 0}, 892 893 /* RFS3_GETATTR = 1 */ 894 {rfs3_getattr, 895 xdr_nfs_fh3_server, NULL_xdrproc_t, sizeof (GETATTR3args), 896 xdr_GETATTR3res, NULL_xdrproc_t, sizeof (GETATTR3res), 897 nullfree, (RPC_IDEMPOTENT | RPC_ALLOWANON), 898 rfs3_getattr_getfh}, 899 900 /* RFS3_SETATTR = 2 */ 901 {rfs3_setattr, 902 xdr_SETATTR3args, NULL_xdrproc_t, sizeof (SETATTR3args), 903 xdr_SETATTR3res, NULL_xdrproc_t, sizeof (SETATTR3res), 904 nullfree, 0, 905 rfs3_setattr_getfh}, 906 907 /* RFS3_LOOKUP = 3 */ 908 {rfs3_lookup, 909 xdr_diropargs3, NULL_xdrproc_t, sizeof (LOOKUP3args), 910 xdr_LOOKUP3res, NULL_xdrproc_t, sizeof (LOOKUP3res), 911 nullfree, (RPC_IDEMPOTENT | RPC_PUBLICFH_OK), 912 rfs3_lookup_getfh}, 913 914 /* RFS3_ACCESS = 4 */ 915 {rfs3_access, 916 xdr_ACCESS3args, NULL_xdrproc_t, sizeof (ACCESS3args), 917 xdr_ACCESS3res, NULL_xdrproc_t, sizeof (ACCESS3res), 918 nullfree, RPC_IDEMPOTENT, 919 rfs3_access_getfh}, 920 921 /* RFS3_READLINK = 5 */ 922 {rfs3_readlink, 923 xdr_nfs_fh3_server, NULL_xdrproc_t, sizeof (READLINK3args), 924 xdr_READLINK3res, NULL_xdrproc_t, sizeof (READLINK3res), 925 rfs3_readlink_free, RPC_IDEMPOTENT, 926 rfs3_readlink_getfh}, 927 928 /* RFS3_READ = 6 */ 929 {rfs3_read, 930 xdr_READ3args, NULL_xdrproc_t, sizeof (READ3args), 931 xdr_READ3res, NULL_xdrproc_t, sizeof (READ3res), 932 rfs3_read_free, RPC_IDEMPOTENT, 933 rfs3_read_getfh}, 934 935 /* RFS3_WRITE = 7 */ 936 {rfs3_write, 937 xdr_WRITE3args, NULL_xdrproc_t, sizeof (WRITE3args), 938 xdr_WRITE3res, NULL_xdrproc_t, sizeof (WRITE3res), 939 nullfree, 0, 940 rfs3_write_getfh}, 941 942 /* RFS3_CREATE = 8 */ 943 {rfs3_create, 944 xdr_CREATE3args, NULL_xdrproc_t, sizeof (CREATE3args), 945 xdr_CREATE3res, NULL_xdrproc_t, sizeof (CREATE3res), 946 nullfree, 0, 947 rfs3_create_getfh}, 948 949 /* RFS3_MKDIR = 9 */ 950 {rfs3_mkdir, 951 xdr_MKDIR3args, NULL_xdrproc_t, sizeof (MKDIR3args), 952 xdr_MKDIR3res, NULL_xdrproc_t, sizeof (MKDIR3res), 953 nullfree, 0, 954 rfs3_mkdir_getfh}, 955 956 /* RFS3_SYMLINK = 10 */ 957 {rfs3_symlink, 958 xdr_SYMLINK3args, NULL_xdrproc_t, sizeof (SYMLINK3args), 959 xdr_SYMLINK3res, NULL_xdrproc_t, sizeof (SYMLINK3res), 960 nullfree, 0, 961 rfs3_symlink_getfh}, 962 963 /* RFS3_MKNOD = 11 */ 964 {rfs3_mknod, 965 xdr_MKNOD3args, NULL_xdrproc_t, sizeof (MKNOD3args), 966 xdr_MKNOD3res, NULL_xdrproc_t, sizeof (MKNOD3res), 967 nullfree, 0, 968 rfs3_mknod_getfh}, 969 970 /* RFS3_REMOVE = 12 */ 971 {rfs3_remove, 972 xdr_diropargs3, NULL_xdrproc_t, sizeof (REMOVE3args), 973 xdr_REMOVE3res, NULL_xdrproc_t, sizeof (REMOVE3res), 974 nullfree, 0, 975 rfs3_remove_getfh}, 976 977 /* RFS3_RMDIR = 13 */ 978 {rfs3_rmdir, 979 xdr_diropargs3, NULL_xdrproc_t, sizeof (RMDIR3args), 980 xdr_RMDIR3res, NULL_xdrproc_t, sizeof (RMDIR3res), 981 nullfree, 0, 982 rfs3_rmdir_getfh}, 983 984 /* RFS3_RENAME = 14 */ 985 {rfs3_rename, 986 xdr_RENAME3args, NULL_xdrproc_t, sizeof (RENAME3args), 987 xdr_RENAME3res, NULL_xdrproc_t, sizeof (RENAME3res), 988 nullfree, 0, 989 rfs3_rename_getfh}, 990 991 /* RFS3_LINK = 15 */ 992 {rfs3_link, 993 xdr_LINK3args, NULL_xdrproc_t, sizeof (LINK3args), 994 xdr_LINK3res, NULL_xdrproc_t, sizeof (LINK3res), 995 nullfree, 0, 996 rfs3_link_getfh}, 997 998 /* RFS3_READDIR = 16 */ 999 {rfs3_readdir, 1000 xdr_READDIR3args, NULL_xdrproc_t, sizeof (READDIR3args), 1001 xdr_READDIR3res, NULL_xdrproc_t, sizeof (READDIR3res), 1002 rfs3_readdir_free, RPC_IDEMPOTENT, 1003 rfs3_readdir_getfh}, 1004 1005 /* RFS3_READDIRPLUS = 17 */ 1006 {rfs3_readdirplus, 1007 xdr_READDIRPLUS3args, NULL_xdrproc_t, sizeof (READDIRPLUS3args), 1008 xdr_READDIRPLUS3res, NULL_xdrproc_t, sizeof (READDIRPLUS3res), 1009 rfs3_readdirplus_free, RPC_AVOIDWORK, 1010 rfs3_readdirplus_getfh}, 1011 1012 /* RFS3_FSSTAT = 18 */ 1013 {rfs3_fsstat, 1014 xdr_nfs_fh3_server, NULL_xdrproc_t, sizeof (FSSTAT3args), 1015 xdr_FSSTAT3res, NULL_xdrproc_t, sizeof (FSSTAT3res), 1016 nullfree, RPC_IDEMPOTENT, 1017 rfs3_fsstat_getfh}, 1018 1019 /* RFS3_FSINFO = 19 */ 1020 {rfs3_fsinfo, 1021 xdr_nfs_fh3_server, NULL_xdrproc_t, sizeof (FSINFO3args), 1022 xdr_FSINFO3res, NULL_xdrproc_t, sizeof (FSINFO3res), 1023 nullfree, RPC_IDEMPOTENT|RPC_ALLOWANON, 1024 rfs3_fsinfo_getfh}, 1025 1026 /* RFS3_PATHCONF = 20 */ 1027 {rfs3_pathconf, 1028 xdr_nfs_fh3_server, NULL_xdrproc_t, sizeof (PATHCONF3args), 1029 xdr_PATHCONF3res, NULL_xdrproc_t, sizeof (PATHCONF3res), 1030 nullfree, RPC_IDEMPOTENT, 1031 rfs3_pathconf_getfh}, 1032 1033 /* RFS3_COMMIT = 21 */ 1034 {rfs3_commit, 1035 xdr_COMMIT3args, NULL_xdrproc_t, sizeof (COMMIT3args), 1036 xdr_COMMIT3res, NULL_xdrproc_t, sizeof (COMMIT3res), 1037 nullfree, RPC_IDEMPOTENT, 1038 rfs3_commit_getfh}, 1039 }; 1040 1041 static char *rfscallnames_v4[] = { 1042 "RFS4_NULL", 1043 "RFS4_COMPOUND", 1044 "RFS4_NULL", 1045 "RFS4_NULL", 1046 "RFS4_NULL", 1047 "RFS4_NULL", 1048 "RFS4_NULL", 1049 "RFS4_NULL", 1050 "RFS4_CREATE" 1051 }; 1052 1053 static struct rpcdisp rfsdisptab_v4[] = { 1054 /* 1055 * NFS VERSION 4 1056 */ 1057 1058 /* RFS_NULL = 0 */ 1059 {rpc_null, 1060 xdr_void, NULL_xdrproc_t, 0, 1061 xdr_void, NULL_xdrproc_t, 0, 1062 nullfree, RPC_IDEMPOTENT, 0}, 1063 1064 /* RFS4_compound = 1 */ 1065 {rfs4_compound, 1066 xdr_COMPOUND4args_srv, NULL_xdrproc_t, sizeof (COMPOUND4args), 1067 xdr_COMPOUND4res_srv, NULL_xdrproc_t, sizeof (COMPOUND4res), 1068 rfs4_compound_free, 0, 0}, 1069 }; 1070 1071 union rfs_args { 1072 /* 1073 * NFS VERSION 2 1074 */ 1075 1076 /* RFS_NULL = 0 */ 1077 1078 /* RFS_GETATTR = 1 */ 1079 fhandle_t nfs2_getattr_args; 1080 1081 /* RFS_SETATTR = 2 */ 1082 struct nfssaargs nfs2_setattr_args; 1083 1084 /* RFS_ROOT = 3 *** NO LONGER SUPPORTED *** */ 1085 1086 /* RFS_LOOKUP = 4 */ 1087 struct nfsdiropargs nfs2_lookup_args; 1088 1089 /* RFS_READLINK = 5 */ 1090 fhandle_t nfs2_readlink_args; 1091 1092 /* RFS_READ = 6 */ 1093 struct nfsreadargs nfs2_read_args; 1094 1095 /* RFS_WRITECACHE = 7 *** NO LONGER SUPPORTED *** */ 1096 1097 /* RFS_WRITE = 8 */ 1098 struct nfswriteargs nfs2_write_args; 1099 1100 /* RFS_CREATE = 9 */ 1101 struct nfscreatargs nfs2_create_args; 1102 1103 /* RFS_REMOVE = 10 */ 1104 struct nfsdiropargs nfs2_remove_args; 1105 1106 /* RFS_RENAME = 11 */ 1107 struct nfsrnmargs nfs2_rename_args; 1108 1109 /* RFS_LINK = 12 */ 1110 struct nfslinkargs nfs2_link_args; 1111 1112 /* RFS_SYMLINK = 13 */ 1113 struct nfsslargs nfs2_symlink_args; 1114 1115 /* RFS_MKDIR = 14 */ 1116 struct nfscreatargs nfs2_mkdir_args; 1117 1118 /* RFS_RMDIR = 15 */ 1119 struct nfsdiropargs nfs2_rmdir_args; 1120 1121 /* RFS_READDIR = 16 */ 1122 struct nfsrddirargs nfs2_readdir_args; 1123 1124 /* RFS_STATFS = 17 */ 1125 fhandle_t nfs2_statfs_args; 1126 1127 /* 1128 * NFS VERSION 3 1129 */ 1130 1131 /* RFS_NULL = 0 */ 1132 1133 /* RFS3_GETATTR = 1 */ 1134 GETATTR3args nfs3_getattr_args; 1135 1136 /* RFS3_SETATTR = 2 */ 1137 SETATTR3args nfs3_setattr_args; 1138 1139 /* RFS3_LOOKUP = 3 */ 1140 LOOKUP3args nfs3_lookup_args; 1141 1142 /* RFS3_ACCESS = 4 */ 1143 ACCESS3args nfs3_access_args; 1144 1145 /* RFS3_READLINK = 5 */ 1146 READLINK3args nfs3_readlink_args; 1147 1148 /* RFS3_READ = 6 */ 1149 READ3args nfs3_read_args; 1150 1151 /* RFS3_WRITE = 7 */ 1152 WRITE3args nfs3_write_args; 1153 1154 /* RFS3_CREATE = 8 */ 1155 CREATE3args nfs3_create_args; 1156 1157 /* RFS3_MKDIR = 9 */ 1158 MKDIR3args nfs3_mkdir_args; 1159 1160 /* RFS3_SYMLINK = 10 */ 1161 SYMLINK3args nfs3_symlink_args; 1162 1163 /* RFS3_MKNOD = 11 */ 1164 MKNOD3args nfs3_mknod_args; 1165 1166 /* RFS3_REMOVE = 12 */ 1167 REMOVE3args nfs3_remove_args; 1168 1169 /* RFS3_RMDIR = 13 */ 1170 RMDIR3args nfs3_rmdir_args; 1171 1172 /* RFS3_RENAME = 14 */ 1173 RENAME3args nfs3_rename_args; 1174 1175 /* RFS3_LINK = 15 */ 1176 LINK3args nfs3_link_args; 1177 1178 /* RFS3_READDIR = 16 */ 1179 READDIR3args nfs3_readdir_args; 1180 1181 /* RFS3_READDIRPLUS = 17 */ 1182 READDIRPLUS3args nfs3_readdirplus_args; 1183 1184 /* RFS3_FSSTAT = 18 */ 1185 FSSTAT3args nfs3_fsstat_args; 1186 1187 /* RFS3_FSINFO = 19 */ 1188 FSINFO3args nfs3_fsinfo_args; 1189 1190 /* RFS3_PATHCONF = 20 */ 1191 PATHCONF3args nfs3_pathconf_args; 1192 1193 /* RFS3_COMMIT = 21 */ 1194 COMMIT3args nfs3_commit_args; 1195 1196 /* 1197 * NFS VERSION 4 1198 */ 1199 1200 /* RFS_NULL = 0 */ 1201 1202 /* COMPUND = 1 */ 1203 COMPOUND4args nfs4_compound_args; 1204 }; 1205 1206 union rfs_res { 1207 /* 1208 * NFS VERSION 2 1209 */ 1210 1211 /* RFS_NULL = 0 */ 1212 1213 /* RFS_GETATTR = 1 */ 1214 struct nfsattrstat nfs2_getattr_res; 1215 1216 /* RFS_SETATTR = 2 */ 1217 struct nfsattrstat nfs2_setattr_res; 1218 1219 /* RFS_ROOT = 3 *** NO LONGER SUPPORTED *** */ 1220 1221 /* RFS_LOOKUP = 4 */ 1222 struct nfsdiropres nfs2_lookup_res; 1223 1224 /* RFS_READLINK = 5 */ 1225 struct nfsrdlnres nfs2_readlink_res; 1226 1227 /* RFS_READ = 6 */ 1228 struct nfsrdresult nfs2_read_res; 1229 1230 /* RFS_WRITECACHE = 7 *** NO LONGER SUPPORTED *** */ 1231 1232 /* RFS_WRITE = 8 */ 1233 struct nfsattrstat nfs2_write_res; 1234 1235 /* RFS_CREATE = 9 */ 1236 struct nfsdiropres nfs2_create_res; 1237 1238 /* RFS_REMOVE = 10 */ 1239 enum nfsstat nfs2_remove_res; 1240 1241 /* RFS_RENAME = 11 */ 1242 enum nfsstat nfs2_rename_res; 1243 1244 /* RFS_LINK = 12 */ 1245 enum nfsstat nfs2_link_res; 1246 1247 /* RFS_SYMLINK = 13 */ 1248 enum nfsstat nfs2_symlink_res; 1249 1250 /* RFS_MKDIR = 14 */ 1251 struct nfsdiropres nfs2_mkdir_res; 1252 1253 /* RFS_RMDIR = 15 */ 1254 enum nfsstat nfs2_rmdir_res; 1255 1256 /* RFS_READDIR = 16 */ 1257 struct nfsrddirres nfs2_readdir_res; 1258 1259 /* RFS_STATFS = 17 */ 1260 struct nfsstatfs nfs2_statfs_res; 1261 1262 /* 1263 * NFS VERSION 3 1264 */ 1265 1266 /* RFS_NULL = 0 */ 1267 1268 /* RFS3_GETATTR = 1 */ 1269 GETATTR3res nfs3_getattr_res; 1270 1271 /* RFS3_SETATTR = 2 */ 1272 SETATTR3res nfs3_setattr_res; 1273 1274 /* RFS3_LOOKUP = 3 */ 1275 LOOKUP3res nfs3_lookup_res; 1276 1277 /* RFS3_ACCESS = 4 */ 1278 ACCESS3res nfs3_access_res; 1279 1280 /* RFS3_READLINK = 5 */ 1281 READLINK3res nfs3_readlink_res; 1282 1283 /* RFS3_READ = 6 */ 1284 READ3res nfs3_read_res; 1285 1286 /* RFS3_WRITE = 7 */ 1287 WRITE3res nfs3_write_res; 1288 1289 /* RFS3_CREATE = 8 */ 1290 CREATE3res nfs3_create_res; 1291 1292 /* RFS3_MKDIR = 9 */ 1293 MKDIR3res nfs3_mkdir_res; 1294 1295 /* RFS3_SYMLINK = 10 */ 1296 SYMLINK3res nfs3_symlink_res; 1297 1298 /* RFS3_MKNOD = 11 */ 1299 MKNOD3res nfs3_mknod_res; 1300 1301 /* RFS3_REMOVE = 12 */ 1302 REMOVE3res nfs3_remove_res; 1303 1304 /* RFS3_RMDIR = 13 */ 1305 RMDIR3res nfs3_rmdir_res; 1306 1307 /* RFS3_RENAME = 14 */ 1308 RENAME3res nfs3_rename_res; 1309 1310 /* RFS3_LINK = 15 */ 1311 LINK3res nfs3_link_res; 1312 1313 /* RFS3_READDIR = 16 */ 1314 READDIR3res nfs3_readdir_res; 1315 1316 /* RFS3_READDIRPLUS = 17 */ 1317 READDIRPLUS3res nfs3_readdirplus_res; 1318 1319 /* RFS3_FSSTAT = 18 */ 1320 FSSTAT3res nfs3_fsstat_res; 1321 1322 /* RFS3_FSINFO = 19 */ 1323 FSINFO3res nfs3_fsinfo_res; 1324 1325 /* RFS3_PATHCONF = 20 */ 1326 PATHCONF3res nfs3_pathconf_res; 1327 1328 /* RFS3_COMMIT = 21 */ 1329 COMMIT3res nfs3_commit_res; 1330 1331 /* 1332 * NFS VERSION 4 1333 */ 1334 1335 /* RFS_NULL = 0 */ 1336 1337 /* RFS4_COMPOUND = 1 */ 1338 COMPOUND4res nfs4_compound_res; 1339 1340 }; 1341 1342 static struct rpc_disptable rfs_disptable[] = { 1343 {sizeof (rfsdisptab_v2) / sizeof (rfsdisptab_v2[0]), 1344 rfscallnames_v2, 1345 rfsdisptab_v2}, 1346 {sizeof (rfsdisptab_v3) / sizeof (rfsdisptab_v3[0]), 1347 rfscallnames_v3, 1348 rfsdisptab_v3}, 1349 {sizeof (rfsdisptab_v4) / sizeof (rfsdisptab_v4[0]), 1350 rfscallnames_v4, 1351 rfsdisptab_v4}, 1352 }; 1353 1354 /* 1355 * If nfs_portmon is set, then clients are required to use privileged 1356 * ports (ports < IPPORT_RESERVED) in order to get NFS services. 1357 * 1358 * N.B.: this attempt to carry forward the already ill-conceived notion 1359 * of privileged ports for TCP/UDP is really quite ineffectual. Not only 1360 * is it transport-dependent, it's laughably easy to spoof. If you're 1361 * really interested in security, you must start with secure RPC instead. 1362 */ 1363 static int nfs_portmon = 0; 1364 1365 #ifdef DEBUG 1366 /* 1367 * Debug code to allow disabling of rfs_dispatch() use of 1368 * fastxdrargs() and fastxdrres() calls for testing purposes. 1369 */ 1370 static int rfs_no_fast_xdrargs = 0; 1371 static int rfs_no_fast_xdrres = 0; 1372 #endif 1373 1374 union acl_args { 1375 /* 1376 * ACL VERSION 2 1377 */ 1378 1379 /* ACL2_NULL = 0 */ 1380 1381 /* ACL2_GETACL = 1 */ 1382 GETACL2args acl2_getacl_args; 1383 1384 /* ACL2_SETACL = 2 */ 1385 SETACL2args acl2_setacl_args; 1386 1387 /* ACL2_GETATTR = 3 */ 1388 GETATTR2args acl2_getattr_args; 1389 1390 /* ACL2_ACCESS = 4 */ 1391 ACCESS2args acl2_access_args; 1392 1393 /* ACL2_GETXATTRDIR = 5 */ 1394 GETXATTRDIR2args acl2_getxattrdir_args; 1395 1396 /* 1397 * ACL VERSION 3 1398 */ 1399 1400 /* ACL3_NULL = 0 */ 1401 1402 /* ACL3_GETACL = 1 */ 1403 GETACL3args acl3_getacl_args; 1404 1405 /* ACL3_SETACL = 2 */ 1406 SETACL3args acl3_setacl; 1407 1408 /* ACL3_GETXATTRDIR = 3 */ 1409 GETXATTRDIR3args acl3_getxattrdir_args; 1410 1411 }; 1412 1413 union acl_res { 1414 /* 1415 * ACL VERSION 2 1416 */ 1417 1418 /* ACL2_NULL = 0 */ 1419 1420 /* ACL2_GETACL = 1 */ 1421 GETACL2res acl2_getacl_res; 1422 1423 /* ACL2_SETACL = 2 */ 1424 SETACL2res acl2_setacl_res; 1425 1426 /* ACL2_GETATTR = 3 */ 1427 GETATTR2res acl2_getattr_res; 1428 1429 /* ACL2_ACCESS = 4 */ 1430 ACCESS2res acl2_access_res; 1431 1432 /* ACL2_GETXATTRDIR = 5 */ 1433 GETXATTRDIR2args acl2_getxattrdir_res; 1434 1435 /* 1436 * ACL VERSION 3 1437 */ 1438 1439 /* ACL3_NULL = 0 */ 1440 1441 /* ACL3_GETACL = 1 */ 1442 GETACL3res acl3_getacl_res; 1443 1444 /* ACL3_SETACL = 2 */ 1445 SETACL3res acl3_setacl_res; 1446 1447 /* ACL3_GETXATTRDIR = 3 */ 1448 GETXATTRDIR3res acl3_getxattrdir_res; 1449 1450 }; 1451 1452 static bool_t 1453 auth_tooweak(struct svc_req *req, char *res) 1454 { 1455 1456 if (req->rq_vers == NFS_VERSION && req->rq_proc == RFS_LOOKUP) { 1457 struct nfsdiropres *dr = (struct nfsdiropres *)res; 1458 if ((enum wnfsstat)dr->dr_status == WNFSERR_CLNT_FLAVOR) 1459 return (TRUE); 1460 } else if (req->rq_vers == NFS_V3 && req->rq_proc == NFSPROC3_LOOKUP) { 1461 LOOKUP3res *resp = (LOOKUP3res *)res; 1462 if ((enum wnfsstat)resp->status == WNFSERR_CLNT_FLAVOR) 1463 return (TRUE); 1464 } 1465 return (FALSE); 1466 } 1467 1468 static void 1469 common_dispatch(struct svc_req *req, SVCXPRT *xprt, rpcvers_t min_vers, 1470 rpcvers_t max_vers, char *pgmname, struct rpc_disptable *disptable) 1471 { 1472 int which; 1473 rpcvers_t vers; 1474 char *args; 1475 union { 1476 union rfs_args ra; 1477 union acl_args aa; 1478 } args_buf; 1479 char *res; 1480 union { 1481 union rfs_res rr; 1482 union acl_res ar; 1483 } res_buf; 1484 struct rpcdisp *disp = NULL; 1485 int dis_flags = 0; 1486 cred_t *cr; 1487 int error = 0; 1488 int anon_ok; 1489 struct exportinfo *exi = NULL; 1490 unsigned int nfslog_rec_id; 1491 int dupstat; 1492 struct dupreq *dr; 1493 int authres; 1494 bool_t publicfh_ok = FALSE; 1495 enum_t auth_flavor; 1496 bool_t dupcached = FALSE; 1497 struct netbuf nb; 1498 bool_t logging_enabled = FALSE; 1499 struct exportinfo *nfslog_exi = NULL; 1500 char **procnames; 1501 char cbuf[INET6_ADDRSTRLEN]; /* to hold both IPv4 and IPv6 addr */ 1502 bool_t ro = FALSE; 1503 nfs_globals_t *ng = nfs_srv_getzg(); 1504 nfs_export_t *ne = ng->nfs_export; 1505 kstat_named_t *svstat, *procstat; 1506 1507 ASSERT(req->rq_prog == NFS_PROGRAM || req->rq_prog == NFS_ACL_PROGRAM); 1508 1509 vers = req->rq_vers; 1510 1511 svstat = ng->svstat[req->rq_vers]; 1512 procstat = (req->rq_prog == NFS_PROGRAM) ? 1513 ng->rfsproccnt[vers] : ng->aclproccnt[vers]; 1514 1515 if (vers < min_vers || vers > max_vers) { 1516 svcerr_progvers(req->rq_xprt, min_vers, max_vers); 1517 error++; 1518 cmn_err(CE_NOTE, "%s: bad version number %u", pgmname, vers); 1519 goto done; 1520 } 1521 vers -= min_vers; 1522 1523 which = req->rq_proc; 1524 if (which < 0 || which >= disptable[(int)vers].dis_nprocs) { 1525 svcerr_noproc(req->rq_xprt); 1526 error++; 1527 goto done; 1528 } 1529 1530 procstat[which].value.ui64++; 1531 1532 disp = &disptable[(int)vers].dis_table[which]; 1533 procnames = disptable[(int)vers].dis_procnames; 1534 1535 auth_flavor = req->rq_cred.oa_flavor; 1536 1537 /* 1538 * Deserialize into the args struct. 1539 */ 1540 args = (char *)&args_buf; 1541 1542 #ifdef DEBUG 1543 if (rfs_no_fast_xdrargs || (auth_flavor == RPCSEC_GSS) || 1544 disp->dis_fastxdrargs == NULL_xdrproc_t || 1545 !SVC_GETARGS(xprt, disp->dis_fastxdrargs, (char *)&args)) 1546 #else 1547 if ((auth_flavor == RPCSEC_GSS) || 1548 disp->dis_fastxdrargs == NULL_xdrproc_t || 1549 !SVC_GETARGS(xprt, disp->dis_fastxdrargs, (char *)&args)) 1550 #endif 1551 { 1552 bzero(args, disp->dis_argsz); 1553 if (!SVC_GETARGS(xprt, disp->dis_xdrargs, args)) { 1554 error++; 1555 /* 1556 * Check if we are outside our capabilities. 1557 */ 1558 if (rfs4_minorvers_mismatch(req, xprt, (void *)args)) 1559 goto done; 1560 1561 svcerr_decode(xprt); 1562 cmn_err(CE_NOTE, 1563 "Failed to decode arguments for %s version %u " 1564 "procedure %s client %s%s", 1565 pgmname, vers + min_vers, procnames[which], 1566 client_name(req), client_addr(req, cbuf)); 1567 goto done; 1568 } 1569 } 1570 1571 /* 1572 * If Version 4 use that specific dispatch function. 1573 */ 1574 if (req->rq_vers == 4) { 1575 error += rfs4_dispatch(disp, req, xprt, args); 1576 goto done; 1577 } 1578 1579 dis_flags = disp->dis_flags; 1580 1581 /* 1582 * Find export information and check authentication, 1583 * setting the credential if everything is ok. 1584 */ 1585 if (disp->dis_getfh != NULL) { 1586 void *fh; 1587 fsid_t *fsid; 1588 fid_t *fid, *xfid; 1589 fhandle_t *fh2; 1590 nfs_fh3 *fh3; 1591 1592 fh = (*disp->dis_getfh)(args); 1593 switch (req->rq_vers) { 1594 case NFS_VERSION: 1595 fh2 = (fhandle_t *)fh; 1596 fsid = &fh2->fh_fsid; 1597 fid = (fid_t *)&fh2->fh_len; 1598 xfid = (fid_t *)&fh2->fh_xlen; 1599 break; 1600 case NFS_V3: 1601 fh3 = (nfs_fh3 *)fh; 1602 fsid = &fh3->fh3_fsid; 1603 fid = FH3TOFIDP(fh3); 1604 xfid = FH3TOXFIDP(fh3); 1605 break; 1606 } 1607 1608 /* 1609 * Fix for bug 1038302 - corbin 1610 * There is a problem here if anonymous access is 1611 * disallowed. If the current request is part of the 1612 * client's mount process for the requested filesystem, 1613 * then it will carry root (uid 0) credentials on it, and 1614 * will be denied by checkauth if that client does not 1615 * have explicit root=0 permission. This will cause the 1616 * client's mount operation to fail. As a work-around, 1617 * we check here to see if the request is a getattr or 1618 * statfs operation on the exported vnode itself, and 1619 * pass a flag to checkauth with the result of this test. 1620 * 1621 * The filehandle refers to the mountpoint itself if 1622 * the fh_data and fh_xdata portions of the filehandle 1623 * are equal. 1624 * 1625 * Added anon_ok argument to checkauth(). 1626 */ 1627 1628 if ((dis_flags & RPC_ALLOWANON) && EQFID(fid, xfid)) 1629 anon_ok = 1; 1630 else 1631 anon_ok = 0; 1632 1633 cr = svc_xprt_cred(xprt); 1634 1635 exi = checkexport(fsid, xfid); 1636 1637 if (exi != NULL) { 1638 publicfh_ok = PUBLICFH_CHECK(ne, disp, exi, fsid, xfid); 1639 1640 /* 1641 * Don't allow non-V4 clients access 1642 * to pseudo exports 1643 */ 1644 if (PSEUDO(exi)) { 1645 svcerr_weakauth(xprt); 1646 error++; 1647 goto done; 1648 } 1649 1650 authres = checkauth(exi, req, cr, anon_ok, publicfh_ok, 1651 &ro); 1652 /* 1653 * authres > 0: authentication OK - proceed 1654 * authres == 0: authentication weak - return error 1655 * authres < 0: authentication timeout - drop 1656 */ 1657 if (authres <= 0) { 1658 if (authres == 0) { 1659 svcerr_weakauth(xprt); 1660 error++; 1661 } 1662 goto done; 1663 } 1664 } 1665 } else 1666 cr = NULL; 1667 1668 if ((dis_flags & RPC_MAPRESP) && (auth_flavor != RPCSEC_GSS)) { 1669 res = (char *)SVC_GETRES(xprt, disp->dis_ressz); 1670 if (res == NULL) 1671 res = (char *)&res_buf; 1672 } else 1673 res = (char *)&res_buf; 1674 1675 if (!(dis_flags & RPC_IDEMPOTENT)) { 1676 dupstat = SVC_DUP_EXT(xprt, req, res, disp->dis_ressz, &dr, 1677 &dupcached); 1678 1679 switch (dupstat) { 1680 case DUP_ERROR: 1681 svcerr_systemerr(xprt); 1682 error++; 1683 goto done; 1684 /* NOTREACHED */ 1685 case DUP_INPROGRESS: 1686 if (res != (char *)&res_buf) 1687 SVC_FREERES(xprt); 1688 error++; 1689 goto done; 1690 /* NOTREACHED */ 1691 case DUP_NEW: 1692 case DUP_DROP: 1693 curthread->t_flag |= T_DONTPEND; 1694 1695 (*disp->dis_proc)(args, res, exi, req, cr, ro); 1696 1697 curthread->t_flag &= ~T_DONTPEND; 1698 if (curthread->t_flag & T_WOULDBLOCK) { 1699 curthread->t_flag &= ~T_WOULDBLOCK; 1700 SVC_DUPDONE_EXT(xprt, dr, res, NULL, 1701 disp->dis_ressz, DUP_DROP); 1702 if (res != (char *)&res_buf) 1703 SVC_FREERES(xprt); 1704 error++; 1705 goto done; 1706 } 1707 if (dis_flags & RPC_AVOIDWORK) { 1708 SVC_DUPDONE_EXT(xprt, dr, res, NULL, 1709 disp->dis_ressz, DUP_DROP); 1710 } else { 1711 SVC_DUPDONE_EXT(xprt, dr, res, 1712 disp->dis_resfree == nullfree ? NULL : 1713 disp->dis_resfree, 1714 disp->dis_ressz, DUP_DONE); 1715 dupcached = TRUE; 1716 } 1717 break; 1718 case DUP_DONE: 1719 break; 1720 } 1721 1722 } else { 1723 curthread->t_flag |= T_DONTPEND; 1724 1725 (*disp->dis_proc)(args, res, exi, req, cr, ro); 1726 1727 curthread->t_flag &= ~T_DONTPEND; 1728 if (curthread->t_flag & T_WOULDBLOCK) { 1729 curthread->t_flag &= ~T_WOULDBLOCK; 1730 if (res != (char *)&res_buf) 1731 SVC_FREERES(xprt); 1732 error++; 1733 goto done; 1734 } 1735 } 1736 1737 if (auth_tooweak(req, res)) { 1738 svcerr_weakauth(xprt); 1739 error++; 1740 goto done; 1741 } 1742 1743 /* 1744 * Check to see if logging has been enabled on the server. 1745 * If so, then obtain the export info struct to be used for 1746 * the later writing of the log record. This is done for 1747 * the case that a lookup is done across a non-logged public 1748 * file system. 1749 */ 1750 if (nfslog_buffer_list != NULL) { 1751 nfslog_exi = nfslog_get_exi(ne, exi, req, res, &nfslog_rec_id); 1752 /* 1753 * Is logging enabled? 1754 */ 1755 logging_enabled = (nfslog_exi != NULL); 1756 1757 /* 1758 * Copy the netbuf for logging purposes, before it is 1759 * freed by svc_sendreply(). 1760 */ 1761 if (logging_enabled) { 1762 NFSLOG_COPY_NETBUF(nfslog_exi, xprt, &nb); 1763 /* 1764 * If RPC_MAPRESP flag set (i.e. in V2 ops) the 1765 * res gets copied directly into the mbuf and 1766 * may be freed soon after the sendreply. So we 1767 * must copy it here to a safe place... 1768 */ 1769 if (res != (char *)&res_buf) { 1770 bcopy(res, (char *)&res_buf, disp->dis_ressz); 1771 } 1772 } 1773 } 1774 1775 /* 1776 * Serialize and send results struct 1777 */ 1778 #ifdef DEBUG 1779 if (rfs_no_fast_xdrres == 0 && res != (char *)&res_buf) 1780 #else 1781 if (res != (char *)&res_buf) 1782 #endif 1783 { 1784 if (!svc_sendreply(xprt, disp->dis_fastxdrres, res)) { 1785 cmn_err(CE_NOTE, "%s: bad sendreply", pgmname); 1786 svcerr_systemerr(xprt); 1787 error++; 1788 } 1789 } else { 1790 if (!svc_sendreply(xprt, disp->dis_xdrres, res)) { 1791 cmn_err(CE_NOTE, "%s: bad sendreply", pgmname); 1792 svcerr_systemerr(xprt); 1793 error++; 1794 } 1795 } 1796 1797 /* 1798 * Log if needed 1799 */ 1800 if (logging_enabled) { 1801 nfslog_write_record(nfslog_exi, req, args, (char *)&res_buf, 1802 cr, &nb, nfslog_rec_id, NFSLOG_ONE_BUFFER); 1803 exi_rele(nfslog_exi); 1804 kmem_free((&nb)->buf, (&nb)->len); 1805 } 1806 1807 /* 1808 * Free results struct. With the addition of NFS V4 we can 1809 * have non-idempotent procedures with functions. 1810 */ 1811 if (disp->dis_resfree != nullfree && dupcached == FALSE) { 1812 (*disp->dis_resfree)(res); 1813 } 1814 1815 done: 1816 /* 1817 * Free arguments struct 1818 */ 1819 if (disp) { 1820 if (!SVC_FREEARGS(xprt, disp->dis_xdrargs, args)) { 1821 cmn_err(CE_NOTE, "%s: bad freeargs", pgmname); 1822 error++; 1823 } 1824 } else { 1825 if (!SVC_FREEARGS(xprt, (xdrproc_t)0, (caddr_t)0)) { 1826 cmn_err(CE_NOTE, "%s: bad freeargs", pgmname); 1827 error++; 1828 } 1829 } 1830 1831 if (exi != NULL) 1832 exi_rele(exi); 1833 1834 svstat[NFS_BADCALLS].value.ui64 += error; 1835 svstat[NFS_CALLS].value.ui64++; 1836 } 1837 1838 static void 1839 rfs_dispatch(struct svc_req *req, SVCXPRT *xprt) 1840 { 1841 common_dispatch(req, xprt, NFS_VERSMIN, NFS_VERSMAX, 1842 "NFS", rfs_disptable); 1843 } 1844 1845 static char *aclcallnames_v2[] = { 1846 "ACL2_NULL", 1847 "ACL2_GETACL", 1848 "ACL2_SETACL", 1849 "ACL2_GETATTR", 1850 "ACL2_ACCESS", 1851 "ACL2_GETXATTRDIR" 1852 }; 1853 1854 static struct rpcdisp acldisptab_v2[] = { 1855 /* 1856 * ACL VERSION 2 1857 */ 1858 1859 /* ACL2_NULL = 0 */ 1860 {rpc_null, 1861 xdr_void, NULL_xdrproc_t, 0, 1862 xdr_void, NULL_xdrproc_t, 0, 1863 nullfree, RPC_IDEMPOTENT, 1864 0}, 1865 1866 /* ACL2_GETACL = 1 */ 1867 {acl2_getacl, 1868 xdr_GETACL2args, xdr_fastGETACL2args, sizeof (GETACL2args), 1869 xdr_GETACL2res, NULL_xdrproc_t, sizeof (GETACL2res), 1870 acl2_getacl_free, RPC_IDEMPOTENT, 1871 acl2_getacl_getfh}, 1872 1873 /* ACL2_SETACL = 2 */ 1874 {acl2_setacl, 1875 xdr_SETACL2args, NULL_xdrproc_t, sizeof (SETACL2args), 1876 #ifdef _LITTLE_ENDIAN 1877 xdr_SETACL2res, xdr_fastSETACL2res, sizeof (SETACL2res), 1878 #else 1879 xdr_SETACL2res, NULL_xdrproc_t, sizeof (SETACL2res), 1880 #endif 1881 nullfree, RPC_MAPRESP, 1882 acl2_setacl_getfh}, 1883 1884 /* ACL2_GETATTR = 3 */ 1885 {acl2_getattr, 1886 xdr_GETATTR2args, xdr_fastGETATTR2args, sizeof (GETATTR2args), 1887 #ifdef _LITTLE_ENDIAN 1888 xdr_GETATTR2res, xdr_fastGETATTR2res, sizeof (GETATTR2res), 1889 #else 1890 xdr_GETATTR2res, NULL_xdrproc_t, sizeof (GETATTR2res), 1891 #endif 1892 nullfree, RPC_IDEMPOTENT|RPC_ALLOWANON|RPC_MAPRESP, 1893 acl2_getattr_getfh}, 1894 1895 /* ACL2_ACCESS = 4 */ 1896 {acl2_access, 1897 xdr_ACCESS2args, xdr_fastACCESS2args, sizeof (ACCESS2args), 1898 #ifdef _LITTLE_ENDIAN 1899 xdr_ACCESS2res, xdr_fastACCESS2res, sizeof (ACCESS2res), 1900 #else 1901 xdr_ACCESS2res, NULL_xdrproc_t, sizeof (ACCESS2res), 1902 #endif 1903 nullfree, RPC_IDEMPOTENT|RPC_MAPRESP, 1904 acl2_access_getfh}, 1905 1906 /* ACL2_GETXATTRDIR = 5 */ 1907 {acl2_getxattrdir, 1908 xdr_GETXATTRDIR2args, NULL_xdrproc_t, sizeof (GETXATTRDIR2args), 1909 xdr_GETXATTRDIR2res, NULL_xdrproc_t, sizeof (GETXATTRDIR2res), 1910 nullfree, RPC_IDEMPOTENT, 1911 acl2_getxattrdir_getfh}, 1912 }; 1913 1914 static char *aclcallnames_v3[] = { 1915 "ACL3_NULL", 1916 "ACL3_GETACL", 1917 "ACL3_SETACL", 1918 "ACL3_GETXATTRDIR" 1919 }; 1920 1921 static struct rpcdisp acldisptab_v3[] = { 1922 /* 1923 * ACL VERSION 3 1924 */ 1925 1926 /* ACL3_NULL = 0 */ 1927 {rpc_null, 1928 xdr_void, NULL_xdrproc_t, 0, 1929 xdr_void, NULL_xdrproc_t, 0, 1930 nullfree, RPC_IDEMPOTENT, 1931 0}, 1932 1933 /* ACL3_GETACL = 1 */ 1934 {acl3_getacl, 1935 xdr_GETACL3args, NULL_xdrproc_t, sizeof (GETACL3args), 1936 xdr_GETACL3res, NULL_xdrproc_t, sizeof (GETACL3res), 1937 acl3_getacl_free, RPC_IDEMPOTENT, 1938 acl3_getacl_getfh}, 1939 1940 /* ACL3_SETACL = 2 */ 1941 {acl3_setacl, 1942 xdr_SETACL3args, NULL_xdrproc_t, sizeof (SETACL3args), 1943 xdr_SETACL3res, NULL_xdrproc_t, sizeof (SETACL3res), 1944 nullfree, 0, 1945 acl3_setacl_getfh}, 1946 1947 /* ACL3_GETXATTRDIR = 3 */ 1948 {acl3_getxattrdir, 1949 xdr_GETXATTRDIR3args, NULL_xdrproc_t, sizeof (GETXATTRDIR3args), 1950 xdr_GETXATTRDIR3res, NULL_xdrproc_t, sizeof (GETXATTRDIR3res), 1951 nullfree, RPC_IDEMPOTENT, 1952 acl3_getxattrdir_getfh}, 1953 }; 1954 1955 static struct rpc_disptable acl_disptable[] = { 1956 {sizeof (acldisptab_v2) / sizeof (acldisptab_v2[0]), 1957 aclcallnames_v2, 1958 acldisptab_v2}, 1959 {sizeof (acldisptab_v3) / sizeof (acldisptab_v3[0]), 1960 aclcallnames_v3, 1961 acldisptab_v3}, 1962 }; 1963 1964 static void 1965 acl_dispatch(struct svc_req *req, SVCXPRT *xprt) 1966 { 1967 common_dispatch(req, xprt, NFS_ACL_VERSMIN, NFS_ACL_VERSMAX, 1968 "ACL", acl_disptable); 1969 } 1970 1971 int 1972 checkwin(int flavor, int window, struct svc_req *req) 1973 { 1974 struct authdes_cred *adc; 1975 1976 switch (flavor) { 1977 case AUTH_DES: 1978 adc = (struct authdes_cred *)req->rq_clntcred; 1979 CTASSERT(sizeof (struct authdes_cred) <= RQCRED_SIZE); 1980 if (adc->adc_fullname.window > window) 1981 return (0); 1982 break; 1983 1984 default: 1985 break; 1986 } 1987 return (1); 1988 } 1989 1990 1991 /* 1992 * checkauth() will check the access permission against the export 1993 * information. Then map root uid/gid to appropriate uid/gid. 1994 * 1995 * This routine is used by NFS V3 and V2 code. 1996 */ 1997 static int 1998 checkauth(struct exportinfo *exi, struct svc_req *req, cred_t *cr, int anon_ok, 1999 bool_t publicfh_ok, bool_t *ro) 2000 { 2001 int i, nfsflavor, rpcflavor, stat, access; 2002 struct secinfo *secp; 2003 caddr_t principal; 2004 char buf[INET6_ADDRSTRLEN]; /* to hold both IPv4 and IPv6 addr */ 2005 int anon_res = 0; 2006 2007 uid_t uid; 2008 gid_t gid; 2009 uint_t ngids; 2010 gid_t *gids; 2011 2012 /* 2013 * Check for privileged port number 2014 * N.B.: this assumes that we know the format of a netbuf. 2015 */ 2016 if (nfs_portmon) { 2017 struct sockaddr *ca; 2018 ca = (struct sockaddr *)svc_getrpccaller(req->rq_xprt)->buf; 2019 2020 if (ca == NULL) 2021 return (0); 2022 2023 if ((ca->sa_family == AF_INET && 2024 ntohs(((struct sockaddr_in *)ca)->sin_port) >= 2025 IPPORT_RESERVED) || 2026 (ca->sa_family == AF_INET6 && 2027 ntohs(((struct sockaddr_in6 *)ca)->sin6_port) >= 2028 IPPORT_RESERVED)) { 2029 cmn_err(CE_NOTE, 2030 "nfs_server: client %s%ssent NFS request from " 2031 "unprivileged port", 2032 client_name(req), client_addr(req, buf)); 2033 return (0); 2034 } 2035 } 2036 2037 /* 2038 * return 1 on success or 0 on failure 2039 */ 2040 stat = sec_svc_getcred(req, cr, &principal, &nfsflavor); 2041 2042 /* 2043 * A failed AUTH_UNIX sec_svc_getcred() implies we couldn't set 2044 * the credentials; below we map that to anonymous. 2045 */ 2046 if (!stat && nfsflavor != AUTH_UNIX) { 2047 cmn_err(CE_NOTE, 2048 "nfs_server: couldn't get unix cred for %s", 2049 client_name(req)); 2050 return (0); 2051 } 2052 2053 /* 2054 * Short circuit checkauth() on operations that support the 2055 * public filehandle, and if the request for that operation 2056 * is using the public filehandle. Note that we must call 2057 * sec_svc_getcred() first so that xp_cookie is set to the 2058 * right value. Normally xp_cookie is just the RPC flavor 2059 * of the the request, but in the case of RPCSEC_GSS it 2060 * could be a pseudo flavor. 2061 */ 2062 if (publicfh_ok) 2063 return (1); 2064 2065 rpcflavor = req->rq_cred.oa_flavor; 2066 /* 2067 * Check if the auth flavor is valid for this export 2068 */ 2069 access = nfsauth_access(exi, req, cr, &uid, &gid, &ngids, &gids); 2070 if (access & NFSAUTH_DROP) 2071 return (-1); /* drop the request */ 2072 2073 if (access & NFSAUTH_RO) 2074 *ro = TRUE; 2075 2076 if (access & NFSAUTH_DENIED) { 2077 /* 2078 * If anon_ok == 1 and we got NFSAUTH_DENIED, it was 2079 * probably due to the flavor not matching during 2080 * the mount attempt. So map the flavor to AUTH_NONE 2081 * so that the credentials get mapped to the anonymous 2082 * user. 2083 */ 2084 if (anon_ok == 1) 2085 rpcflavor = AUTH_NONE; 2086 else 2087 return (0); /* deny access */ 2088 2089 } else if (access & NFSAUTH_MAPNONE) { 2090 /* 2091 * Access was granted even though the flavor mismatched 2092 * because AUTH_NONE was one of the exported flavors. 2093 */ 2094 rpcflavor = AUTH_NONE; 2095 2096 } else if (access & NFSAUTH_WRONGSEC) { 2097 /* 2098 * NFSAUTH_WRONGSEC is used for NFSv4. If we get here, 2099 * it means a client ignored the list of allowed flavors 2100 * returned via the MOUNT protocol. So we just disallow it! 2101 */ 2102 return (0); 2103 } 2104 2105 if (rpcflavor != AUTH_SYS) 2106 kmem_free(gids, ngids * sizeof (gid_t)); 2107 2108 switch (rpcflavor) { 2109 case AUTH_NONE: 2110 anon_res = crsetugid(cr, exi->exi_export.ex_anon, 2111 exi->exi_export.ex_anon); 2112 (void) crsetgroups(cr, 0, NULL); 2113 break; 2114 2115 case AUTH_UNIX: 2116 if (!stat || crgetuid(cr) == 0 && !(access & NFSAUTH_UIDMAP)) { 2117 anon_res = crsetugid(cr, exi->exi_export.ex_anon, 2118 exi->exi_export.ex_anon); 2119 (void) crsetgroups(cr, 0, NULL); 2120 } else if (crgetuid(cr) == 0 && access & NFSAUTH_ROOT) { 2121 /* 2122 * It is root, so apply rootid to get real UID 2123 * Find the secinfo structure. We should be able 2124 * to find it by the time we reach here. 2125 * nfsauth_access() has done the checking. 2126 */ 2127 secp = NULL; 2128 for (i = 0; i < exi->exi_export.ex_seccnt; i++) { 2129 struct secinfo *sptr; 2130 sptr = &exi->exi_export.ex_secinfo[i]; 2131 if (sptr->s_secinfo.sc_nfsnum == nfsflavor) { 2132 secp = sptr; 2133 break; 2134 } 2135 } 2136 if (secp != NULL) { 2137 (void) crsetugid(cr, secp->s_rootid, 2138 secp->s_rootid); 2139 (void) crsetgroups(cr, 0, NULL); 2140 } 2141 } else if (crgetuid(cr) != uid || crgetgid(cr) != gid) { 2142 if (crsetugid(cr, uid, gid) != 0) 2143 anon_res = crsetugid(cr, 2144 exi->exi_export.ex_anon, 2145 exi->exi_export.ex_anon); 2146 (void) crsetgroups(cr, 0, NULL); 2147 } else if (access & NFSAUTH_GROUPS) { 2148 (void) crsetgroups(cr, ngids, gids); 2149 } 2150 2151 kmem_free(gids, ngids * sizeof (gid_t)); 2152 2153 break; 2154 2155 case AUTH_DES: 2156 case RPCSEC_GSS: 2157 /* 2158 * Find the secinfo structure. We should be able 2159 * to find it by the time we reach here. 2160 * nfsauth_access() has done the checking. 2161 */ 2162 secp = NULL; 2163 for (i = 0; i < exi->exi_export.ex_seccnt; i++) { 2164 if (exi->exi_export.ex_secinfo[i].s_secinfo.sc_nfsnum == 2165 nfsflavor) { 2166 secp = &exi->exi_export.ex_secinfo[i]; 2167 break; 2168 } 2169 } 2170 2171 if (!secp) { 2172 cmn_err(CE_NOTE, "nfs_server: client %s%shad " 2173 "no secinfo data for flavor %d", 2174 client_name(req), client_addr(req, buf), 2175 nfsflavor); 2176 return (0); 2177 } 2178 2179 if (!checkwin(rpcflavor, secp->s_window, req)) { 2180 cmn_err(CE_NOTE, 2181 "nfs_server: client %s%sused invalid " 2182 "auth window value", 2183 client_name(req), client_addr(req, buf)); 2184 return (0); 2185 } 2186 2187 /* 2188 * Map root principals listed in the share's root= list to root, 2189 * and map any others principals that were mapped to root by RPC 2190 * to anon. 2191 */ 2192 if (principal && sec_svc_inrootlist(rpcflavor, principal, 2193 secp->s_rootcnt, secp->s_rootnames)) { 2194 if (crgetuid(cr) == 0 && secp->s_rootid == 0) 2195 return (1); 2196 2197 2198 (void) crsetugid(cr, secp->s_rootid, secp->s_rootid); 2199 2200 /* 2201 * NOTE: If and when kernel-land privilege tracing is 2202 * added this may have to be replaced with code that 2203 * retrieves root's supplementary groups (e.g., using 2204 * kgss_get_group_info(). In the meantime principals 2205 * mapped to uid 0 get all privileges, so setting cr's 2206 * supplementary groups for them does nothing. 2207 */ 2208 (void) crsetgroups(cr, 0, NULL); 2209 2210 return (1); 2211 } 2212 2213 /* 2214 * Not a root princ, or not in root list, map UID 0/nobody to 2215 * the anon ID for the share. (RPC sets cr's UIDs and GIDs to 2216 * UID_NOBODY and GID_NOBODY, respectively.) 2217 */ 2218 if (crgetuid(cr) != 0 && 2219 (crgetuid(cr) != UID_NOBODY || crgetgid(cr) != GID_NOBODY)) 2220 return (1); 2221 2222 anon_res = crsetugid(cr, exi->exi_export.ex_anon, 2223 exi->exi_export.ex_anon); 2224 (void) crsetgroups(cr, 0, NULL); 2225 break; 2226 default: 2227 return (0); 2228 } /* switch on rpcflavor */ 2229 2230 /* 2231 * Even if anon access is disallowed via ex_anon == -1, we allow 2232 * this access if anon_ok is set. So set creds to the default 2233 * "nobody" id. 2234 */ 2235 if (anon_res != 0) { 2236 if (anon_ok == 0) { 2237 cmn_err(CE_NOTE, 2238 "nfs_server: client %s%ssent wrong " 2239 "authentication for %s", 2240 client_name(req), client_addr(req, buf), 2241 exi->exi_export.ex_path ? 2242 exi->exi_export.ex_path : "?"); 2243 return (0); 2244 } 2245 2246 if (crsetugid(cr, UID_NOBODY, GID_NOBODY) != 0) 2247 return (0); 2248 } 2249 2250 return (1); 2251 } 2252 2253 /* 2254 * returns 0 on failure, -1 on a drop, -2 on wrong security flavor, 2255 * and 1 on success 2256 */ 2257 int 2258 checkauth4(struct compound_state *cs, struct svc_req *req) 2259 { 2260 int i, rpcflavor, access; 2261 struct secinfo *secp; 2262 char buf[MAXHOST + 1]; 2263 int anon_res = 0, nfsflavor; 2264 struct exportinfo *exi; 2265 cred_t *cr; 2266 caddr_t principal; 2267 2268 uid_t uid; 2269 gid_t gid; 2270 uint_t ngids; 2271 gid_t *gids; 2272 2273 exi = cs->exi; 2274 cr = cs->cr; 2275 principal = cs->principal; 2276 nfsflavor = cs->nfsflavor; 2277 2278 ASSERT(cr != NULL); 2279 2280 rpcflavor = req->rq_cred.oa_flavor; 2281 cs->access &= ~CS_ACCESS_LIMITED; 2282 2283 /* 2284 * Check for privileged port number 2285 * N.B.: this assumes that we know the format of a netbuf. 2286 */ 2287 if (nfs_portmon) { 2288 struct sockaddr *ca; 2289 ca = (struct sockaddr *)svc_getrpccaller(req->rq_xprt)->buf; 2290 2291 if (ca == NULL) 2292 return (0); 2293 2294 if ((ca->sa_family == AF_INET && 2295 ntohs(((struct sockaddr_in *)ca)->sin_port) >= 2296 IPPORT_RESERVED) || 2297 (ca->sa_family == AF_INET6 && 2298 ntohs(((struct sockaddr_in6 *)ca)->sin6_port) >= 2299 IPPORT_RESERVED)) { 2300 cmn_err(CE_NOTE, 2301 "nfs_server: client %s%ssent NFSv4 request from " 2302 "unprivileged port", 2303 client_name(req), client_addr(req, buf)); 2304 return (0); 2305 } 2306 } 2307 2308 /* 2309 * Check the access right per auth flavor on the vnode of 2310 * this export for the given request. 2311 */ 2312 access = nfsauth4_access(cs->exi, cs->vp, req, cr, &uid, &gid, &ngids, 2313 &gids); 2314 2315 if (access & NFSAUTH_WRONGSEC) 2316 return (-2); /* no access for this security flavor */ 2317 2318 if (access & NFSAUTH_DROP) 2319 return (-1); /* drop the request */ 2320 2321 if (access & NFSAUTH_DENIED) { 2322 2323 if (exi->exi_export.ex_seccnt > 0) 2324 return (0); /* deny access */ 2325 2326 } else if (access & NFSAUTH_LIMITED) { 2327 2328 cs->access |= CS_ACCESS_LIMITED; 2329 2330 } else if (access & NFSAUTH_MAPNONE) { 2331 /* 2332 * Access was granted even though the flavor mismatched 2333 * because AUTH_NONE was one of the exported flavors. 2334 */ 2335 rpcflavor = AUTH_NONE; 2336 } 2337 2338 /* 2339 * XXX probably need to redo some of it for nfsv4? 2340 * return 1 on success or 0 on failure 2341 */ 2342 2343 if (rpcflavor != AUTH_SYS) 2344 kmem_free(gids, ngids * sizeof (gid_t)); 2345 2346 switch (rpcflavor) { 2347 case AUTH_NONE: 2348 anon_res = crsetugid(cr, exi->exi_export.ex_anon, 2349 exi->exi_export.ex_anon); 2350 (void) crsetgroups(cr, 0, NULL); 2351 break; 2352 2353 case AUTH_UNIX: 2354 if (crgetuid(cr) == 0 && !(access & NFSAUTH_UIDMAP)) { 2355 anon_res = crsetugid(cr, exi->exi_export.ex_anon, 2356 exi->exi_export.ex_anon); 2357 (void) crsetgroups(cr, 0, NULL); 2358 } else if (crgetuid(cr) == 0 && access & NFSAUTH_ROOT) { 2359 /* 2360 * It is root, so apply rootid to get real UID 2361 * Find the secinfo structure. We should be able 2362 * to find it by the time we reach here. 2363 * nfsauth_access() has done the checking. 2364 */ 2365 secp = NULL; 2366 for (i = 0; i < exi->exi_export.ex_seccnt; i++) { 2367 struct secinfo *sptr; 2368 sptr = &exi->exi_export.ex_secinfo[i]; 2369 if (sptr->s_secinfo.sc_nfsnum == nfsflavor) { 2370 secp = &exi->exi_export.ex_secinfo[i]; 2371 break; 2372 } 2373 } 2374 if (secp != NULL) { 2375 (void) crsetugid(cr, secp->s_rootid, 2376 secp->s_rootid); 2377 (void) crsetgroups(cr, 0, NULL); 2378 } 2379 } else if (crgetuid(cr) != uid || crgetgid(cr) != gid) { 2380 if (crsetugid(cr, uid, gid) != 0) 2381 anon_res = crsetugid(cr, 2382 exi->exi_export.ex_anon, 2383 exi->exi_export.ex_anon); 2384 (void) crsetgroups(cr, 0, NULL); 2385 } if (access & NFSAUTH_GROUPS) { 2386 (void) crsetgroups(cr, ngids, gids); 2387 } 2388 2389 kmem_free(gids, ngids * sizeof (gid_t)); 2390 2391 break; 2392 2393 default: 2394 /* 2395 * Find the secinfo structure. We should be able 2396 * to find it by the time we reach here. 2397 * nfsauth_access() has done the checking. 2398 */ 2399 secp = NULL; 2400 for (i = 0; i < exi->exi_export.ex_seccnt; i++) { 2401 if (exi->exi_export.ex_secinfo[i].s_secinfo.sc_nfsnum == 2402 nfsflavor) { 2403 secp = &exi->exi_export.ex_secinfo[i]; 2404 break; 2405 } 2406 } 2407 2408 if (!secp) { 2409 cmn_err(CE_NOTE, "nfs_server: client %s%shad " 2410 "no secinfo data for flavor %d", 2411 client_name(req), client_addr(req, buf), 2412 nfsflavor); 2413 return (0); 2414 } 2415 2416 if (!checkwin(rpcflavor, secp->s_window, req)) { 2417 cmn_err(CE_NOTE, 2418 "nfs_server: client %s%sused invalid " 2419 "auth window value", 2420 client_name(req), client_addr(req, buf)); 2421 return (0); 2422 } 2423 2424 /* 2425 * Map root principals listed in the share's root= list to root, 2426 * and map any others principals that were mapped to root by RPC 2427 * to anon. If not going to anon, set to rootid (root_mapping). 2428 */ 2429 if (principal && sec_svc_inrootlist(rpcflavor, principal, 2430 secp->s_rootcnt, secp->s_rootnames)) { 2431 if (crgetuid(cr) == 0 && secp->s_rootid == 0) 2432 return (1); 2433 2434 (void) crsetugid(cr, secp->s_rootid, secp->s_rootid); 2435 2436 /* 2437 * NOTE: If and when kernel-land privilege tracing is 2438 * added this may have to be replaced with code that 2439 * retrieves root's supplementary groups (e.g., using 2440 * kgss_get_group_info(). In the meantime principals 2441 * mapped to uid 0 get all privileges, so setting cr's 2442 * supplementary groups for them does nothing. 2443 */ 2444 (void) crsetgroups(cr, 0, NULL); 2445 2446 return (1); 2447 } 2448 2449 /* 2450 * Not a root princ, or not in root list, map UID 0/nobody to 2451 * the anon ID for the share. (RPC sets cr's UIDs and GIDs to 2452 * UID_NOBODY and GID_NOBODY, respectively.) 2453 */ 2454 if (crgetuid(cr) != 0 && 2455 (crgetuid(cr) != UID_NOBODY || crgetgid(cr) != GID_NOBODY)) 2456 return (1); 2457 2458 anon_res = crsetugid(cr, exi->exi_export.ex_anon, 2459 exi->exi_export.ex_anon); 2460 (void) crsetgroups(cr, 0, NULL); 2461 break; 2462 } /* switch on rpcflavor */ 2463 2464 /* 2465 * Even if anon access is disallowed via ex_anon == -1, we allow 2466 * this access if anon_ok is set. So set creds to the default 2467 * "nobody" id. 2468 */ 2469 2470 if (anon_res != 0) { 2471 cmn_err(CE_NOTE, 2472 "nfs_server: client %s%ssent wrong " 2473 "authentication for %s", 2474 client_name(req), client_addr(req, buf), 2475 exi->exi_export.ex_path ? 2476 exi->exi_export.ex_path : "?"); 2477 return (0); 2478 } 2479 2480 return (1); 2481 } 2482 2483 2484 static char * 2485 client_name(struct svc_req *req) 2486 { 2487 char *hostname = NULL; 2488 2489 /* 2490 * If it's a Unix cred then use the 2491 * hostname from the credential. 2492 */ 2493 if (req->rq_cred.oa_flavor == AUTH_UNIX) { 2494 hostname = ((struct authunix_parms *) 2495 req->rq_clntcred)->aup_machname; 2496 } 2497 if (hostname == NULL) 2498 hostname = ""; 2499 2500 return (hostname); 2501 } 2502 2503 static char * 2504 client_addr(struct svc_req *req, char *buf) 2505 { 2506 struct sockaddr *ca; 2507 uchar_t *b; 2508 char *frontspace = ""; 2509 2510 /* 2511 * We assume we are called in tandem with client_name and the 2512 * format string looks like "...client %s%sblah blah..." 2513 * 2514 * If it's a Unix cred then client_name returned 2515 * a host name, so we need insert a space between host name 2516 * and IP address. 2517 */ 2518 if (req->rq_cred.oa_flavor == AUTH_UNIX) 2519 frontspace = " "; 2520 2521 /* 2522 * Convert the caller's IP address to a dotted string 2523 */ 2524 ca = (struct sockaddr *)svc_getrpccaller(req->rq_xprt)->buf; 2525 2526 if (ca->sa_family == AF_INET) { 2527 b = (uchar_t *)&((struct sockaddr_in *)ca)->sin_addr; 2528 (void) sprintf(buf, "%s(%d.%d.%d.%d) ", frontspace, 2529 b[0] & 0xFF, b[1] & 0xFF, b[2] & 0xFF, b[3] & 0xFF); 2530 } else if (ca->sa_family == AF_INET6) { 2531 struct sockaddr_in6 *sin6; 2532 sin6 = (struct sockaddr_in6 *)ca; 2533 (void) kinet_ntop6((uchar_t *)&sin6->sin6_addr, 2534 buf, INET6_ADDRSTRLEN); 2535 2536 } else { 2537 2538 /* 2539 * No IP address to print. If there was a host name 2540 * printed, then we print a space. 2541 */ 2542 (void) sprintf(buf, frontspace); 2543 } 2544 2545 return (buf); 2546 } 2547 2548 /* 2549 * NFS Server initialization routine. This routine should only be called 2550 * once. It performs the following tasks: 2551 * - Call sub-initialization routines (localize access to variables) 2552 * - Initialize all locks 2553 * - initialize the version 3 write verifier 2554 */ 2555 void 2556 nfs_srvinit(void) 2557 { 2558 2559 /* Truly global stuff in this module (not per zone) */ 2560 rw_init(&nfssrv_globals_rwl, NULL, RW_DEFAULT, NULL); 2561 list_create(&nfssrv_globals_list, sizeof (nfs_globals_t), 2562 offsetof(nfs_globals_t, nfs_g_link)); 2563 tsd_create(&nfs_server_tsd_key, NULL); 2564 2565 /* The order here is important */ 2566 nfs_exportinit(); 2567 rfs_srvrinit(); 2568 rfs3_srvrinit(); 2569 rfs4_srvrinit(); 2570 nfsauth_init(); 2571 2572 /* 2573 * NFS server zone-specific global variables 2574 * Note the zone_init is called for the GZ here. 2575 */ 2576 zone_key_create(&nfssrv_zone_key, nfs_server_zone_init, 2577 nfs_server_zone_shutdown, nfs_server_zone_fini); 2578 } 2579 2580 /* 2581 * NFS Server finalization routine. This routine is called to cleanup the 2582 * initialization work previously performed if the NFS server module could 2583 * not be loaded correctly. 2584 */ 2585 void 2586 nfs_srvfini(void) 2587 { 2588 2589 /* 2590 * NFS server zone-specific global variables 2591 * Note the zone_fini is called for the GZ here. 2592 */ 2593 (void) zone_key_delete(nfssrv_zone_key); 2594 2595 /* The order here is important (reverse of init) */ 2596 nfsauth_fini(); 2597 rfs4_srvrfini(); 2598 rfs3_srvrfini(); 2599 rfs_srvrfini(); 2600 nfs_exportfini(); 2601 2602 /* Truly global stuff in this module (not per zone) */ 2603 tsd_destroy(&nfs_server_tsd_key); 2604 list_destroy(&nfssrv_globals_list); 2605 rw_destroy(&nfssrv_globals_rwl); 2606 } 2607 2608 /* 2609 * Zone init, shutdown, fini functions for the NFS server 2610 * 2611 * This design is careful to create the entire hierarhcy of 2612 * NFS server "globals" (including those created by various 2613 * per-module *_zone_init functions, etc.) so that all these 2614 * objects have exactly the same lifetime. 2615 * 2616 * These objects are also kept on a list for two reasons: 2617 * 1: It makes finding these in mdb _much_ easier. 2618 * 2: It allows operating across all zone globals for 2619 * functions like nfs_auth.c:exi_cache_reclaim 2620 */ 2621 static void * 2622 nfs_server_zone_init(zoneid_t zoneid) 2623 { 2624 nfs_globals_t *ng; 2625 2626 ng = kmem_zalloc(sizeof (*ng), KM_SLEEP); 2627 2628 ng->nfs_versmin = NFS_VERSMIN_DEFAULT; 2629 ng->nfs_versmax = NFS_VERSMAX_DEFAULT; 2630 2631 /* Init the stuff to control start/stop */ 2632 ng->nfs_server_upordown = NFS_SERVER_STOPPED; 2633 mutex_init(&ng->nfs_server_upordown_lock, NULL, MUTEX_DEFAULT, NULL); 2634 cv_init(&ng->nfs_server_upordown_cv, NULL, CV_DEFAULT, NULL); 2635 mutex_init(&ng->rdma_wait_mutex, NULL, MUTEX_DEFAULT, NULL); 2636 cv_init(&ng->rdma_wait_cv, NULL, CV_DEFAULT, NULL); 2637 2638 ng->nfs_zoneid = zoneid; 2639 2640 /* 2641 * Order here is important. 2642 * export init must precede srv init calls. 2643 */ 2644 nfs_export_zone_init(ng); 2645 rfs_stat_zone_init(ng); 2646 rfs_srv_zone_init(ng); 2647 rfs3_srv_zone_init(ng); 2648 rfs4_srv_zone_init(ng); 2649 nfsauth_zone_init(ng); 2650 2651 rw_enter(&nfssrv_globals_rwl, RW_WRITER); 2652 list_insert_tail(&nfssrv_globals_list, ng); 2653 rw_exit(&nfssrv_globals_rwl); 2654 2655 return (ng); 2656 } 2657 2658 /* ARGSUSED */ 2659 static void 2660 nfs_server_zone_shutdown(zoneid_t zoneid, void *data) 2661 { 2662 nfs_globals_t *ng; 2663 2664 ng = (nfs_globals_t *)data; 2665 2666 /* 2667 * Order is like _fini, but only 2668 * some modules need this hook. 2669 */ 2670 nfsauth_zone_shutdown(ng); 2671 nfs_export_zone_shutdown(ng); 2672 } 2673 2674 /* ARGSUSED */ 2675 static void 2676 nfs_server_zone_fini(zoneid_t zoneid, void *data) 2677 { 2678 nfs_globals_t *ng; 2679 2680 ng = (nfs_globals_t *)data; 2681 2682 rw_enter(&nfssrv_globals_rwl, RW_WRITER); 2683 list_remove(&nfssrv_globals_list, ng); 2684 rw_exit(&nfssrv_globals_rwl); 2685 2686 /* 2687 * Order here is important. 2688 * reverse order from init 2689 */ 2690 nfsauth_zone_fini(ng); 2691 rfs4_srv_zone_fini(ng); 2692 rfs3_srv_zone_fini(ng); 2693 rfs_srv_zone_fini(ng); 2694 rfs_stat_zone_fini(ng); 2695 nfs_export_zone_fini(ng); 2696 2697 mutex_destroy(&ng->nfs_server_upordown_lock); 2698 cv_destroy(&ng->nfs_server_upordown_cv); 2699 mutex_destroy(&ng->rdma_wait_mutex); 2700 cv_destroy(&ng->rdma_wait_cv); 2701 2702 kmem_free(ng, sizeof (*ng)); 2703 } 2704 2705 /* 2706 * Set up an iovec array of up to cnt pointers. 2707 */ 2708 void 2709 mblk_to_iov(mblk_t *m, int cnt, struct iovec *iovp) 2710 { 2711 while (m != NULL && cnt-- > 0) { 2712 iovp->iov_base = (caddr_t)m->b_rptr; 2713 iovp->iov_len = (m->b_wptr - m->b_rptr); 2714 iovp++; 2715 m = m->b_cont; 2716 } 2717 } 2718 2719 /* 2720 * Common code between NFS Version 2 and NFS Version 3 for the public 2721 * filehandle multicomponent lookups. 2722 */ 2723 2724 /* 2725 * Public filehandle evaluation of a multi-component lookup, following 2726 * symbolic links, if necessary. This may result in a vnode in another 2727 * filesystem, which is OK as long as the other filesystem is exported. 2728 * 2729 * Note that the exi will be set either to NULL or a new reference to the 2730 * exportinfo struct that corresponds to the vnode of the multi-component path. 2731 * It is the callers responsibility to release this reference. 2732 */ 2733 int 2734 rfs_publicfh_mclookup(char *p, vnode_t *dvp, cred_t *cr, vnode_t **vpp, 2735 struct exportinfo **exi, struct sec_ol *sec) 2736 { 2737 int pathflag; 2738 vnode_t *mc_dvp = NULL; 2739 vnode_t *realvp; 2740 int error; 2741 2742 *exi = NULL; 2743 2744 /* 2745 * check if the given path is a url or native path. Since p is 2746 * modified by MCLpath(), it may be empty after returning from 2747 * there, and should be checked. 2748 */ 2749 if ((pathflag = MCLpath(&p)) == -1) 2750 return (EIO); 2751 2752 /* 2753 * If pathflag is SECURITY_QUERY, turn the SEC_QUERY bit 2754 * on in sec->sec_flags. This bit will later serve as an 2755 * indication in makefh_ol() or makefh3_ol() to overload the 2756 * filehandle to contain the sec modes used by the server for 2757 * the path. 2758 */ 2759 if (pathflag == SECURITY_QUERY) { 2760 if ((sec->sec_index = (uint_t)(*p)) > 0) { 2761 sec->sec_flags |= SEC_QUERY; 2762 p++; 2763 if ((pathflag = MCLpath(&p)) == -1) 2764 return (EIO); 2765 } else { 2766 cmn_err(CE_NOTE, 2767 "nfs_server: invalid security index %d, " 2768 "violating WebNFS SNEGO protocol.", sec->sec_index); 2769 return (EIO); 2770 } 2771 } 2772 2773 if (p[0] == '\0') { 2774 error = ENOENT; 2775 goto publicfh_done; 2776 } 2777 2778 error = rfs_pathname(p, &mc_dvp, vpp, dvp, cr, pathflag); 2779 2780 /* 2781 * If name resolves to "/" we get EINVAL since we asked for 2782 * the vnode of the directory that the file is in. Try again 2783 * with NULL directory vnode. 2784 */ 2785 if (error == EINVAL) { 2786 error = rfs_pathname(p, NULL, vpp, dvp, cr, pathflag); 2787 if (!error) { 2788 ASSERT(*vpp != NULL); 2789 if ((*vpp)->v_type == VDIR) { 2790 VN_HOLD(*vpp); 2791 mc_dvp = *vpp; 2792 } else { 2793 /* 2794 * This should not happen, the filesystem is 2795 * in an inconsistent state. Fail the lookup 2796 * at this point. 2797 */ 2798 VN_RELE(*vpp); 2799 error = EINVAL; 2800 } 2801 } 2802 } 2803 2804 if (error) 2805 goto publicfh_done; 2806 2807 if (*vpp == NULL) { 2808 error = ENOENT; 2809 goto publicfh_done; 2810 } 2811 2812 ASSERT(mc_dvp != NULL); 2813 ASSERT(*vpp != NULL); 2814 2815 if ((*vpp)->v_type == VDIR) { 2816 do { 2817 /* 2818 * *vpp may be an AutoFS node, so we perform 2819 * a VOP_ACCESS() to trigger the mount of the intended 2820 * filesystem, so we can perform the lookup in the 2821 * intended filesystem. 2822 */ 2823 (void) VOP_ACCESS(*vpp, 0, 0, cr, NULL); 2824 2825 /* 2826 * If vnode is covered, get the 2827 * the topmost vnode. 2828 */ 2829 if (vn_mountedvfs(*vpp) != NULL) { 2830 error = traverse(vpp); 2831 if (error) { 2832 VN_RELE(*vpp); 2833 goto publicfh_done; 2834 } 2835 } 2836 2837 if (VOP_REALVP(*vpp, &realvp, NULL) == 0 && 2838 realvp != *vpp) { 2839 /* 2840 * If realvp is different from *vpp 2841 * then release our reference on *vpp, so that 2842 * the export access check be performed on the 2843 * real filesystem instead. 2844 */ 2845 VN_HOLD(realvp); 2846 VN_RELE(*vpp); 2847 *vpp = realvp; 2848 } else { 2849 break; 2850 } 2851 /* LINTED */ 2852 } while (TRUE); 2853 2854 /* 2855 * Let nfs_vptexi() figure what the real parent is. 2856 */ 2857 VN_RELE(mc_dvp); 2858 mc_dvp = NULL; 2859 2860 } else { 2861 /* 2862 * If vnode is covered, get the 2863 * the topmost vnode. 2864 */ 2865 if (vn_mountedvfs(mc_dvp) != NULL) { 2866 error = traverse(&mc_dvp); 2867 if (error) { 2868 VN_RELE(*vpp); 2869 goto publicfh_done; 2870 } 2871 } 2872 2873 if (VOP_REALVP(mc_dvp, &realvp, NULL) == 0 && 2874 realvp != mc_dvp) { 2875 /* 2876 * *vpp is a file, obtain realvp of the parent 2877 * directory vnode. 2878 */ 2879 VN_HOLD(realvp); 2880 VN_RELE(mc_dvp); 2881 mc_dvp = realvp; 2882 } 2883 } 2884 2885 /* 2886 * The pathname may take us from the public filesystem to another. 2887 * If that's the case then just set the exportinfo to the new export 2888 * and build filehandle for it. Thanks to per-access checking there's 2889 * no security issues with doing this. If the client is not allowed 2890 * access to this new export then it will get an access error when it 2891 * tries to use the filehandle 2892 */ 2893 if (error = nfs_check_vpexi(mc_dvp, *vpp, kcred, exi)) { 2894 VN_RELE(*vpp); 2895 goto publicfh_done; 2896 } 2897 2898 /* 2899 * Not allowed access to pseudo exports. 2900 */ 2901 if (PSEUDO(*exi)) { 2902 error = ENOENT; 2903 VN_RELE(*vpp); 2904 goto publicfh_done; 2905 } 2906 2907 /* 2908 * Do a lookup for the index file. We know the index option doesn't 2909 * allow paths through handling in the share command, so mc_dvp will 2910 * be the parent for the index file vnode, if its present. Use 2911 * temporary pointers to preserve and reuse the vnode pointers of the 2912 * original directory in case there's no index file. Note that the 2913 * index file is a native path, and should not be interpreted by 2914 * the URL parser in rfs_pathname() 2915 */ 2916 if (((*exi)->exi_export.ex_flags & EX_INDEX) && 2917 ((*vpp)->v_type == VDIR) && (pathflag == URLPATH)) { 2918 vnode_t *tvp, *tmc_dvp; /* temporary vnode pointers */ 2919 2920 tmc_dvp = mc_dvp; 2921 mc_dvp = tvp = *vpp; 2922 2923 error = rfs_pathname((*exi)->exi_export.ex_index, NULL, vpp, 2924 mc_dvp, cr, NATIVEPATH); 2925 2926 if (error == ENOENT) { 2927 *vpp = tvp; 2928 mc_dvp = tmc_dvp; 2929 error = 0; 2930 } else { /* ok or error other than ENOENT */ 2931 if (tmc_dvp) 2932 VN_RELE(tmc_dvp); 2933 if (error) 2934 goto publicfh_done; 2935 2936 /* 2937 * Found a valid vp for index "filename". Sanity check 2938 * for odd case where a directory is provided as index 2939 * option argument and leads us to another filesystem 2940 */ 2941 2942 /* Release the reference on the old exi value */ 2943 ASSERT(*exi != NULL); 2944 exi_rele(*exi); 2945 *exi = NULL; 2946 2947 if (error = nfs_check_vpexi(mc_dvp, *vpp, kcred, exi)) { 2948 VN_RELE(*vpp); 2949 goto publicfh_done; 2950 } 2951 /* Have a new *exi */ 2952 } 2953 } 2954 2955 publicfh_done: 2956 if (mc_dvp) 2957 VN_RELE(mc_dvp); 2958 2959 return (error); 2960 } 2961 2962 /* 2963 * Evaluate a multi-component path 2964 */ 2965 int 2966 rfs_pathname( 2967 char *path, /* pathname to evaluate */ 2968 vnode_t **dirvpp, /* ret for ptr to parent dir vnode */ 2969 vnode_t **compvpp, /* ret for ptr to component vnode */ 2970 vnode_t *startdvp, /* starting vnode */ 2971 cred_t *cr, /* user's credential */ 2972 int pathflag) /* flag to identify path, e.g. URL */ 2973 { 2974 char namebuf[TYPICALMAXPATHLEN]; 2975 struct pathname pn; 2976 int error; 2977 2978 ASSERT3U(crgetzoneid(cr), ==, curzone->zone_id); 2979 2980 /* 2981 * If pathname starts with '/', then set startdvp to root. 2982 */ 2983 if (*path == '/') { 2984 while (*path == '/') 2985 path++; 2986 2987 startdvp = ZONE_ROOTVP(); 2988 } 2989 2990 error = pn_get_buf(path, UIO_SYSSPACE, &pn, namebuf, sizeof (namebuf)); 2991 if (error == 0) { 2992 /* 2993 * Call the URL parser for URL paths to modify the original 2994 * string to handle any '%' encoded characters that exist. 2995 * Done here to avoid an extra bcopy in the lookup. 2996 * We need to be careful about pathlen's. We know that 2997 * rfs_pathname() is called with a non-empty path. However, 2998 * it could be emptied due to the path simply being all /'s, 2999 * which is valid to proceed with the lookup, or due to the 3000 * URL parser finding an encoded null character at the 3001 * beginning of path which should not proceed with the lookup. 3002 */ 3003 if (pn.pn_pathlen != 0 && pathflag == URLPATH) { 3004 URLparse(pn.pn_path); 3005 if ((pn.pn_pathlen = strlen(pn.pn_path)) == 0) 3006 return (ENOENT); 3007 } 3008 VN_HOLD(startdvp); 3009 error = lookuppnvp(&pn, NULL, NO_FOLLOW, dirvpp, compvpp, 3010 ZONE_ROOTVP(), startdvp, cr); 3011 } 3012 if (error == ENAMETOOLONG) { 3013 /* 3014 * This thread used a pathname > TYPICALMAXPATHLEN bytes long. 3015 */ 3016 if (error = pn_get(path, UIO_SYSSPACE, &pn)) 3017 return (error); 3018 if (pn.pn_pathlen != 0 && pathflag == URLPATH) { 3019 URLparse(pn.pn_path); 3020 if ((pn.pn_pathlen = strlen(pn.pn_path)) == 0) { 3021 pn_free(&pn); 3022 return (ENOENT); 3023 } 3024 } 3025 VN_HOLD(startdvp); 3026 error = lookuppnvp(&pn, NULL, NO_FOLLOW, dirvpp, compvpp, 3027 ZONE_ROOTVP(), startdvp, cr); 3028 pn_free(&pn); 3029 } 3030 3031 return (error); 3032 } 3033 3034 /* 3035 * Adapt the multicomponent lookup path depending on the pathtype 3036 */ 3037 static int 3038 MCLpath(char **path) 3039 { 3040 unsigned char c = (unsigned char)**path; 3041 3042 /* 3043 * If the MCL path is between 0x20 and 0x7E (graphic printable 3044 * character of the US-ASCII coded character set), its a URL path, 3045 * per RFC 1738. 3046 */ 3047 if (c >= 0x20 && c <= 0x7E) 3048 return (URLPATH); 3049 3050 /* 3051 * If the first octet of the MCL path is not an ASCII character 3052 * then it must be interpreted as a tag value that describes the 3053 * format of the remaining octets of the MCL path. 3054 * 3055 * If the first octet of the MCL path is 0x81 it is a query 3056 * for the security info. 3057 */ 3058 switch (c) { 3059 case 0x80: /* native path, i.e. MCL via mount protocol */ 3060 (*path)++; 3061 return (NATIVEPATH); 3062 case 0x81: /* security query */ 3063 (*path)++; 3064 return (SECURITY_QUERY); 3065 default: 3066 return (-1); 3067 } 3068 } 3069 3070 #define fromhex(c) ((c >= '0' && c <= '9') ? (c - '0') : \ 3071 ((c >= 'A' && c <= 'F') ? (c - 'A' + 10) :\ 3072 ((c >= 'a' && c <= 'f') ? (c - 'a' + 10) : 0))) 3073 3074 /* 3075 * The implementation of URLparse guarantees that the final string will 3076 * fit in the original one. Replaces '%' occurrences followed by 2 characters 3077 * with its corresponding hexadecimal character. 3078 */ 3079 static void 3080 URLparse(char *str) 3081 { 3082 char *p, *q; 3083 3084 p = q = str; 3085 while (*p) { 3086 *q = *p; 3087 if (*p++ == '%') { 3088 if (*p) { 3089 *q = fromhex(*p) * 16; 3090 p++; 3091 if (*p) { 3092 *q += fromhex(*p); 3093 p++; 3094 } 3095 } 3096 } 3097 q++; 3098 } 3099 *q = '\0'; 3100 } 3101 3102 3103 /* 3104 * Get the export information for the lookup vnode, and verify its 3105 * useable. 3106 */ 3107 int 3108 nfs_check_vpexi(vnode_t *mc_dvp, vnode_t *vp, cred_t *cr, 3109 struct exportinfo **exi) 3110 { 3111 int walk; 3112 int error = 0; 3113 3114 *exi = nfs_vptoexi(mc_dvp, vp, cr, &walk, NULL, FALSE); 3115 if (*exi == NULL) 3116 error = EACCES; 3117 else { 3118 /* 3119 * If nosub is set for this export then 3120 * a lookup relative to the public fh 3121 * must not terminate below the 3122 * exported directory. 3123 */ 3124 if ((*exi)->exi_export.ex_flags & EX_NOSUB && walk > 0) 3125 error = EACCES; 3126 } 3127 3128 return (error); 3129 } 3130 3131 /* 3132 * Used by NFSv3 and NFSv4 server to query label of 3133 * a pathname component during lookup/access ops. 3134 */ 3135 ts_label_t * 3136 nfs_getflabel(vnode_t *vp, struct exportinfo *exi) 3137 { 3138 zone_t *zone; 3139 ts_label_t *zone_label; 3140 char *path; 3141 3142 mutex_enter(&vp->v_lock); 3143 if (vp->v_path != vn_vpath_empty) { 3144 zone = zone_find_by_any_path(vp->v_path, B_FALSE); 3145 mutex_exit(&vp->v_lock); 3146 } else { 3147 /* 3148 * v_path not cached. Fall back on pathname of exported 3149 * file system as we rely on pathname from which we can 3150 * derive a label. The exported file system portion of 3151 * path is sufficient to obtain a label. 3152 */ 3153 path = exi->exi_export.ex_path; 3154 if (path == NULL) { 3155 mutex_exit(&vp->v_lock); 3156 return (NULL); 3157 } 3158 zone = zone_find_by_any_path(path, B_FALSE); 3159 mutex_exit(&vp->v_lock); 3160 } 3161 /* 3162 * Caller has verified that the file is either 3163 * exported or visible. So if the path falls in 3164 * global zone, admin_low is returned; otherwise 3165 * the zone's label is returned. 3166 */ 3167 zone_label = zone->zone_slabel; 3168 label_hold(zone_label); 3169 zone_rele(zone); 3170 return (zone_label); 3171 } 3172 3173 /* 3174 * TX NFS routine used by NFSv3 and NFSv4 to do label check 3175 * on client label and server's file object lable. 3176 */ 3177 boolean_t 3178 do_rfs_label_check(bslabel_t *clabel, vnode_t *vp, int flag, 3179 struct exportinfo *exi) 3180 { 3181 bslabel_t *slabel; 3182 ts_label_t *tslabel; 3183 boolean_t result; 3184 3185 if ((tslabel = nfs_getflabel(vp, exi)) == NULL) { 3186 return (B_FALSE); 3187 } 3188 slabel = label2bslabel(tslabel); 3189 DTRACE_PROBE4(tx__rfs__log__info__labelcheck, char *, 3190 "comparing server's file label(1) with client label(2) (vp(3))", 3191 bslabel_t *, slabel, bslabel_t *, clabel, vnode_t *, vp); 3192 3193 if (flag == EQUALITY_CHECK) 3194 result = blequal(clabel, slabel); 3195 else 3196 result = bldominates(clabel, slabel); 3197 label_rele(tslabel); 3198 return (result); 3199 } 3200 3201 /* 3202 * Callback function to return the loaned buffers. 3203 * Calls VOP_RETZCBUF() only after all uio_iov[] 3204 * buffers are returned. nu_ref maintains the count. 3205 */ 3206 void 3207 rfs_free_xuio(void *free_arg) 3208 { 3209 uint_t ref; 3210 nfs_xuio_t *nfsuiop = (nfs_xuio_t *)free_arg; 3211 3212 ref = atomic_dec_uint_nv(&nfsuiop->nu_ref); 3213 3214 /* 3215 * Call VOP_RETZCBUF() only when all the iov buffers 3216 * are sent OTW. 3217 */ 3218 if (ref != 0) 3219 return; 3220 3221 if (((uio_t *)nfsuiop)->uio_extflg & UIO_XUIO) { 3222 (void) VOP_RETZCBUF(nfsuiop->nu_vp, (xuio_t *)free_arg, NULL, 3223 NULL); 3224 VN_RELE(nfsuiop->nu_vp); 3225 } 3226 3227 kmem_cache_free(nfs_xuio_cache, free_arg); 3228 } 3229 3230 xuio_t * 3231 rfs_setup_xuio(vnode_t *vp) 3232 { 3233 nfs_xuio_t *nfsuiop; 3234 3235 nfsuiop = kmem_cache_alloc(nfs_xuio_cache, KM_SLEEP); 3236 3237 bzero(nfsuiop, sizeof (nfs_xuio_t)); 3238 nfsuiop->nu_vp = vp; 3239 3240 /* 3241 * ref count set to 1. more may be added 3242 * if multiple mblks refer to multiple iov's. 3243 * This is done in uio_to_mblk(). 3244 */ 3245 3246 nfsuiop->nu_ref = 1; 3247 3248 nfsuiop->nu_frtn.free_func = rfs_free_xuio; 3249 nfsuiop->nu_frtn.free_arg = (char *)nfsuiop; 3250 3251 nfsuiop->nu_uio.xu_type = UIOTYPE_ZEROCOPY; 3252 3253 return (&nfsuiop->nu_uio); 3254 } 3255 3256 mblk_t * 3257 uio_to_mblk(uio_t *uiop) 3258 { 3259 struct iovec *iovp; 3260 int i; 3261 mblk_t *mp, *mp1; 3262 nfs_xuio_t *nfsuiop = (nfs_xuio_t *)uiop; 3263 3264 if (uiop->uio_iovcnt == 0) 3265 return (NULL); 3266 3267 iovp = uiop->uio_iov; 3268 mp = mp1 = esballoca((uchar_t *)iovp->iov_base, iovp->iov_len, 3269 BPRI_MED, &nfsuiop->nu_frtn); 3270 ASSERT(mp != NULL); 3271 3272 mp->b_wptr += iovp->iov_len; 3273 mp->b_datap->db_type = M_DATA; 3274 3275 for (i = 1; i < uiop->uio_iovcnt; i++) { 3276 iovp = (uiop->uio_iov + i); 3277 3278 mp1->b_cont = esballoca( 3279 (uchar_t *)iovp->iov_base, iovp->iov_len, BPRI_MED, 3280 &nfsuiop->nu_frtn); 3281 3282 mp1 = mp1->b_cont; 3283 ASSERT(mp1 != NULL); 3284 mp1->b_wptr += iovp->iov_len; 3285 mp1->b_datap->db_type = M_DATA; 3286 } 3287 3288 nfsuiop->nu_ref = uiop->uio_iovcnt; 3289 3290 return (mp); 3291 } 3292 3293 /* 3294 * Allocate memory to hold data for a read request of len bytes. 3295 * 3296 * We don't allocate buffers greater than kmem_max_cached in size to avoid 3297 * allocating memory from the kmem_oversized arena. If we allocate oversized 3298 * buffers, we incur heavy cross-call activity when freeing these large buffers 3299 * in the TCP receive path. Note that we can't set b_wptr here since the 3300 * length of the data returned may differ from the length requested when 3301 * reading the end of a file; we set b_wptr in rfs_rndup_mblks() once the 3302 * length of the read is known. 3303 */ 3304 mblk_t * 3305 rfs_read_alloc(uint_t len, struct iovec **iov, int *iovcnt) 3306 { 3307 struct iovec *iovarr; 3308 mblk_t *mp, **mpp = ∓ 3309 size_t mpsize; 3310 uint_t remain = len; 3311 int i, err = 0; 3312 3313 *iovcnt = howmany(len, kmem_max_cached); 3314 3315 iovarr = kmem_alloc(*iovcnt * sizeof (struct iovec), KM_SLEEP); 3316 *iov = iovarr; 3317 3318 for (i = 0; i < *iovcnt; remain -= mpsize, i++) { 3319 ASSERT(remain <= len); 3320 /* 3321 * We roundup the size we allocate to a multiple of 3322 * BYTES_PER_XDR_UNIT (4 bytes) so that the call to 3323 * xdrmblk_putmblk() never fails. 3324 */ 3325 ASSERT(kmem_max_cached % BYTES_PER_XDR_UNIT == 0); 3326 mpsize = MIN(kmem_max_cached, remain); 3327 *mpp = allocb_wait(RNDUP(mpsize), BPRI_MED, STR_NOSIG, &err); 3328 ASSERT(*mpp != NULL); 3329 ASSERT(err == 0); 3330 3331 iovarr[i].iov_base = (caddr_t)(*mpp)->b_rptr; 3332 iovarr[i].iov_len = mpsize; 3333 mpp = &(*mpp)->b_cont; 3334 } 3335 return (mp); 3336 } 3337 3338 void 3339 rfs_rndup_mblks(mblk_t *mp, uint_t len, int buf_loaned) 3340 { 3341 int i; 3342 int alloc_err = 0; 3343 mblk_t *rmp; 3344 uint_t mpsize, remainder; 3345 3346 remainder = P2NPHASE(len, BYTES_PER_XDR_UNIT); 3347 3348 /* 3349 * Non copy-reduction case. This function assumes that blocks were 3350 * allocated in multiples of BYTES_PER_XDR_UNIT bytes, which makes this 3351 * padding safe without bounds checking. 3352 */ 3353 if (!buf_loaned) { 3354 /* 3355 * Set the size of each mblk in the chain until we've consumed 3356 * the specified length for all but the last one. 3357 */ 3358 while ((mpsize = MBLKSIZE(mp)) < len) { 3359 ASSERT(mpsize % BYTES_PER_XDR_UNIT == 0); 3360 mp->b_wptr += mpsize; 3361 len -= mpsize; 3362 mp = mp->b_cont; 3363 ASSERT(mp != NULL); 3364 } 3365 3366 ASSERT(len + remainder <= mpsize); 3367 mp->b_wptr += len; 3368 for (i = 0; i < remainder; i++) 3369 *mp->b_wptr++ = '\0'; 3370 return; 3371 } 3372 3373 /* 3374 * No remainder mblk required. 3375 */ 3376 if (remainder == 0) 3377 return; 3378 3379 /* 3380 * Get to the last mblk in the chain. 3381 */ 3382 while (mp->b_cont != NULL) 3383 mp = mp->b_cont; 3384 3385 /* 3386 * In case of copy-reduction mblks, the size of the mblks are fixed 3387 * and are of the size of the loaned buffers. Allocate a remainder 3388 * mblk and chain it to the data buffers. This is sub-optimal, but not 3389 * expected to happen commonly. 3390 */ 3391 rmp = allocb_wait(remainder, BPRI_MED, STR_NOSIG, &alloc_err); 3392 ASSERT(rmp != NULL); 3393 ASSERT(alloc_err == 0); 3394 3395 for (i = 0; i < remainder; i++) 3396 *rmp->b_wptr++ = '\0'; 3397 3398 rmp->b_datap->db_type = M_DATA; 3399 mp->b_cont = rmp; 3400 } 3401