1 /*- 2 * SPDX-License-Identifier: BSD-3-Clause 3 * 4 * Copyright (c) 1989, 1993 5 * The Regents of the University of California. All rights reserved. 6 * 7 * This code is derived from software contributed to Berkeley by 8 * Rick Macklem at The University of Guelph. 9 * 10 * Redistribution and use in source and binary forms, with or without 11 * modification, are permitted provided that the following conditions 12 * are met: 13 * 1. Redistributions of source code must retain the above copyright 14 * notice, this list of conditions and the following disclaimer. 15 * 2. Redistributions in binary form must reproduce the above copyright 16 * notice, this list of conditions and the following disclaimer in the 17 * documentation and/or other materials provided with the distribution. 18 * 3. Neither the name of the University nor the names of its contributors 19 * may be used to endorse or promote products derived from this software 20 * without specific prior written permission. 21 * 22 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND 23 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE 24 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE 25 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE 26 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL 27 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS 28 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) 29 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT 30 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY 31 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF 32 * SUCH DAMAGE. 33 * 34 */ 35 36 #include <sys/cdefs.h> 37 __FBSDID("$FreeBSD$"); 38 39 /* 40 * Functions that need to be different for different versions of BSD 41 * kernel should be kept here, along with any global storage specific 42 * to this BSD variant. 43 */ 44 #include <fs/nfs/nfsport.h> 45 #include <sys/smp.h> 46 #include <sys/sysctl.h> 47 #include <sys/taskqueue.h> 48 #include <rpc/rpc_com.h> 49 #include <vm/vm.h> 50 #include <vm/vm_object.h> 51 #include <vm/vm_page.h> 52 #include <vm/vm_param.h> 53 #include <vm/vm_map.h> 54 #include <vm/vm_kern.h> 55 #include <vm/vm_extern.h> 56 #include <vm/uma.h> 57 58 extern int nfscl_ticks; 59 extern int nfsrv_nfsuserd; 60 extern struct nfssockreq nfsrv_nfsuserdsock; 61 extern void (*nfsd_call_recall)(struct vnode *, int, struct ucred *, 62 struct thread *); 63 extern int nfsrv_useacl; 64 struct mount nfsv4root_mnt; 65 int newnfs_numnfsd = 0; 66 struct nfsstatsv1 nfsstatsv1; 67 int nfs_numnfscbd = 0; 68 int nfscl_debuglevel = 0; 69 char nfsv4_callbackaddr[INET6_ADDRSTRLEN]; 70 struct callout newnfsd_callout; 71 int nfsrv_lughashsize = 100; 72 struct mtx nfsrv_dslock_mtx; 73 struct nfsdevicehead nfsrv_devidhead; 74 volatile int nfsrv_devidcnt = 0; 75 void (*nfsd_call_servertimer)(void) = NULL; 76 void (*ncl_call_invalcaches)(struct vnode *) = NULL; 77 78 int nfs_pnfsio(task_fn_t *, void *); 79 80 static int nfs_realign_test; 81 static int nfs_realign_count; 82 static struct ext_nfsstats oldnfsstats; 83 84 SYSCTL_NODE(_vfs, OID_AUTO, nfs, CTLFLAG_RW, 0, "NFS filesystem"); 85 SYSCTL_INT(_vfs_nfs, OID_AUTO, realign_test, CTLFLAG_RW, &nfs_realign_test, 86 0, "Number of realign tests done"); 87 SYSCTL_INT(_vfs_nfs, OID_AUTO, realign_count, CTLFLAG_RW, &nfs_realign_count, 88 0, "Number of mbuf realignments done"); 89 SYSCTL_STRING(_vfs_nfs, OID_AUTO, callback_addr, CTLFLAG_RW, 90 nfsv4_callbackaddr, sizeof(nfsv4_callbackaddr), 91 "NFSv4 callback addr for server to use"); 92 SYSCTL_INT(_vfs_nfs, OID_AUTO, debuglevel, CTLFLAG_RW, &nfscl_debuglevel, 93 0, "Debug level for NFS client"); 94 SYSCTL_INT(_vfs_nfs, OID_AUTO, userhashsize, CTLFLAG_RDTUN, &nfsrv_lughashsize, 95 0, "Size of hash tables for uid/name mapping"); 96 int nfs_pnfsiothreads = -1; 97 SYSCTL_INT(_vfs_nfs, OID_AUTO, pnfsiothreads, CTLFLAG_RW, &nfs_pnfsiothreads, 98 0, "Number of pNFS mirror I/O threads"); 99 100 /* 101 * Defines for malloc 102 * (Here for FreeBSD, since they allocate storage.) 103 */ 104 MALLOC_DEFINE(M_NEWNFSRVCACHE, "NFSD srvcache", "NFSD Server Request Cache"); 105 MALLOC_DEFINE(M_NEWNFSDCLIENT, "NFSD V4client", "NFSD V4 Client Id"); 106 MALLOC_DEFINE(M_NEWNFSDSTATE, "NFSD V4state", 107 "NFSD V4 State (Openowner, Open, Lockowner, Delegation"); 108 MALLOC_DEFINE(M_NEWNFSDLOCK, "NFSD V4lock", "NFSD V4 byte range lock"); 109 MALLOC_DEFINE(M_NEWNFSDLOCKFILE, "NFSD lckfile", "NFSD Open/Lock file"); 110 MALLOC_DEFINE(M_NEWNFSSTRING, "NFSD string", "NFSD V4 long string"); 111 MALLOC_DEFINE(M_NEWNFSUSERGROUP, "NFSD usrgroup", "NFSD V4 User/group map"); 112 MALLOC_DEFINE(M_NEWNFSDREQ, "NFS req", "NFS request header"); 113 MALLOC_DEFINE(M_NEWNFSFH, "NFS fh", "NFS file handle"); 114 MALLOC_DEFINE(M_NEWNFSCLOWNER, "NFSCL owner", "NFSCL Open Owner"); 115 MALLOC_DEFINE(M_NEWNFSCLOPEN, "NFSCL open", "NFSCL Open"); 116 MALLOC_DEFINE(M_NEWNFSCLDELEG, "NFSCL deleg", "NFSCL Delegation"); 117 MALLOC_DEFINE(M_NEWNFSCLCLIENT, "NFSCL client", "NFSCL Client"); 118 MALLOC_DEFINE(M_NEWNFSCLLOCKOWNER, "NFSCL lckown", "NFSCL Lock Owner"); 119 MALLOC_DEFINE(M_NEWNFSCLLOCK, "NFSCL lck", "NFSCL Lock"); 120 MALLOC_DEFINE(M_NEWNFSV4NODE, "NEWNFSnode", "NFS vnode"); 121 MALLOC_DEFINE(M_NEWNFSDIRECTIO, "NEWdirectio", "NFS Direct IO buffer"); 122 MALLOC_DEFINE(M_NEWNFSDIROFF, "NFSCL diroffdiroff", 123 "NFS directory offset data"); 124 MALLOC_DEFINE(M_NEWNFSDROLLBACK, "NFSD rollback", 125 "NFS local lock rollback"); 126 MALLOC_DEFINE(M_NEWNFSLAYOUT, "NFSCL layout", "NFSv4.1 Layout"); 127 MALLOC_DEFINE(M_NEWNFSFLAYOUT, "NFSCL flayout", "NFSv4.1 File Layout"); 128 MALLOC_DEFINE(M_NEWNFSDEVINFO, "NFSCL devinfo", "NFSv4.1 Device Info"); 129 MALLOC_DEFINE(M_NEWNFSSOCKREQ, "NFSCL sockreq", "NFS Sock Req"); 130 MALLOC_DEFINE(M_NEWNFSCLDS, "NFSCL session", "NFSv4.1 Session"); 131 MALLOC_DEFINE(M_NEWNFSLAYRECALL, "NFSCL layrecall", "NFSv4.1 Layout Recall"); 132 MALLOC_DEFINE(M_NEWNFSDSESSION, "NFSD session", "NFSD Session for a client"); 133 134 /* 135 * Definition of mutex locks. 136 * newnfsd_mtx is used in nfsrvd_nfsd() to protect the nfs socket list 137 * and assorted other nfsd structures. 138 */ 139 struct mtx newnfsd_mtx; 140 struct mtx nfs_sockl_mutex; 141 struct mtx nfs_state_mutex; 142 struct mtx nfs_nameid_mutex; 143 struct mtx nfs_req_mutex; 144 struct mtx nfs_slock_mutex; 145 struct mtx nfs_clstate_mutex; 146 147 /* local functions */ 148 static int nfssvc_call(struct thread *, struct nfssvc_args *, struct ucred *); 149 150 #ifdef __NO_STRICT_ALIGNMENT 151 /* 152 * These architectures don't need re-alignment, so just return. 153 */ 154 int 155 newnfs_realign(struct mbuf **pm, int how) 156 { 157 158 return (0); 159 } 160 #else /* !__NO_STRICT_ALIGNMENT */ 161 /* 162 * newnfs_realign: 163 * 164 * Check for badly aligned mbuf data and realign by copying the unaligned 165 * portion of the data into a new mbuf chain and freeing the portions 166 * of the old chain that were replaced. 167 * 168 * We cannot simply realign the data within the existing mbuf chain 169 * because the underlying buffers may contain other rpc commands and 170 * we cannot afford to overwrite them. 171 * 172 * We would prefer to avoid this situation entirely. The situation does 173 * not occur with NFS/UDP and is supposed to only occasionally occur 174 * with TCP. Use vfs.nfs.realign_count and realign_test to check this. 175 * 176 */ 177 int 178 newnfs_realign(struct mbuf **pm, int how) 179 { 180 struct mbuf *m, *n; 181 int off, space; 182 183 ++nfs_realign_test; 184 while ((m = *pm) != NULL) { 185 if ((m->m_len & 0x3) || (mtod(m, intptr_t) & 0x3)) { 186 /* 187 * NB: we can't depend on m_pkthdr.len to help us 188 * decide what to do here. May not be worth doing 189 * the m_length calculation as m_copyback will 190 * expand the mbuf chain below as needed. 191 */ 192 space = m_length(m, NULL); 193 if (space >= MINCLSIZE) { 194 /* NB: m_copyback handles space > MCLBYTES */ 195 n = m_getcl(how, MT_DATA, 0); 196 } else 197 n = m_get(how, MT_DATA); 198 if (n == NULL) 199 return (ENOMEM); 200 /* 201 * Align the remainder of the mbuf chain. 202 */ 203 n->m_len = 0; 204 off = 0; 205 while (m != NULL) { 206 m_copyback(n, off, m->m_len, mtod(m, caddr_t)); 207 off += m->m_len; 208 m = m->m_next; 209 } 210 m_freem(*pm); 211 *pm = n; 212 ++nfs_realign_count; 213 break; 214 } 215 pm = &m->m_next; 216 } 217 218 return (0); 219 } 220 #endif /* __NO_STRICT_ALIGNMENT */ 221 222 #ifdef notdef 223 static void 224 nfsrv_object_create(struct vnode *vp, struct thread *td) 225 { 226 227 if (vp == NULL || vp->v_type != VREG) 228 return; 229 (void) vfs_object_create(vp, td, td->td_ucred); 230 } 231 #endif 232 233 /* 234 * Look up a file name. Basically just initialize stuff and call namei(). 235 */ 236 int 237 nfsrv_lookupfilename(struct nameidata *ndp, char *fname, NFSPROC_T *p) 238 { 239 int error; 240 241 NDINIT(ndp, LOOKUP, FOLLOW | LOCKLEAF, UIO_USERSPACE, fname, 242 p); 243 error = namei(ndp); 244 if (!error) { 245 NDFREE(ndp, NDF_ONLY_PNBUF); 246 } 247 return (error); 248 } 249 250 /* 251 * Copy NFS uid, gids to the cred structure. 252 */ 253 void 254 newnfs_copycred(struct nfscred *nfscr, struct ucred *cr) 255 { 256 257 KASSERT(nfscr->nfsc_ngroups >= 0, 258 ("newnfs_copycred: negative nfsc_ngroups")); 259 cr->cr_uid = nfscr->nfsc_uid; 260 crsetgroups(cr, nfscr->nfsc_ngroups, nfscr->nfsc_groups); 261 } 262 263 /* 264 * Map args from nfsmsleep() to msleep(). 265 */ 266 int 267 nfsmsleep(void *chan, void *mutex, int prio, const char *wmesg, 268 struct timespec *ts) 269 { 270 u_int64_t nsecval; 271 int error, timeo; 272 273 if (ts) { 274 timeo = hz * ts->tv_sec; 275 nsecval = (u_int64_t)ts->tv_nsec; 276 nsecval = ((nsecval * ((u_int64_t)hz)) + 500000000) / 277 1000000000; 278 timeo += (int)nsecval; 279 } else { 280 timeo = 0; 281 } 282 error = msleep(chan, (struct mtx *)mutex, prio, wmesg, timeo); 283 return (error); 284 } 285 286 /* 287 * Get the file system info for the server. For now, just assume FFS. 288 */ 289 void 290 nfsvno_getfs(struct nfsfsinfo *sip, int isdgram) 291 { 292 int pref; 293 294 /* 295 * XXX 296 * There should be file system VFS OP(s) to get this information. 297 * For now, assume ufs. 298 */ 299 if (isdgram) 300 pref = NFS_MAXDGRAMDATA; 301 else 302 pref = NFS_SRVMAXIO; 303 sip->fs_rtmax = NFS_SRVMAXIO; 304 sip->fs_rtpref = pref; 305 sip->fs_rtmult = NFS_FABLKSIZE; 306 sip->fs_wtmax = NFS_SRVMAXIO; 307 sip->fs_wtpref = pref; 308 sip->fs_wtmult = NFS_FABLKSIZE; 309 sip->fs_dtpref = pref; 310 sip->fs_maxfilesize = 0xffffffffffffffffull; 311 sip->fs_timedelta.tv_sec = 0; 312 sip->fs_timedelta.tv_nsec = 1; 313 sip->fs_properties = (NFSV3FSINFO_LINK | 314 NFSV3FSINFO_SYMLINK | NFSV3FSINFO_HOMOGENEOUS | 315 NFSV3FSINFO_CANSETTIME); 316 } 317 318 /* 319 * Do the pathconf vnode op. 320 */ 321 int 322 nfsvno_pathconf(struct vnode *vp, int flag, long *retf, 323 struct ucred *cred, struct thread *p) 324 { 325 int error; 326 327 error = VOP_PATHCONF(vp, flag, retf); 328 if (error == EOPNOTSUPP || error == EINVAL) { 329 /* 330 * Some file systems return EINVAL for name arguments not 331 * supported and some return EOPNOTSUPP for this case. 332 * So the NFSv3 Pathconf RPC doesn't fail for these cases, 333 * just fake them. 334 */ 335 switch (flag) { 336 case _PC_LINK_MAX: 337 *retf = NFS_LINK_MAX; 338 break; 339 case _PC_NAME_MAX: 340 *retf = NAME_MAX; 341 break; 342 case _PC_CHOWN_RESTRICTED: 343 *retf = 1; 344 break; 345 case _PC_NO_TRUNC: 346 *retf = 1; 347 break; 348 default: 349 /* 350 * Only happens if a _PC_xxx is added to the server, 351 * but this isn't updated. 352 */ 353 *retf = 0; 354 printf("nfsrvd pathconf flag=%d not supp\n", flag); 355 } 356 error = 0; 357 } 358 NFSEXITCODE(error); 359 return (error); 360 } 361 362 /* Fake nfsrv_atroot. Just return 0 */ 363 int 364 nfsrv_atroot(struct vnode *vp, uint64_t *retp) 365 { 366 367 return (0); 368 } 369 370 /* 371 * Set the credentials to refer to root. 372 * If only the various BSDen could agree on whether cr_gid is a separate 373 * field or cr_groups[0]... 374 */ 375 void 376 newnfs_setroot(struct ucred *cred) 377 { 378 379 cred->cr_uid = 0; 380 cred->cr_groups[0] = 0; 381 cred->cr_ngroups = 1; 382 } 383 384 /* 385 * Get the client credential. Used for Renew and recovery. 386 */ 387 struct ucred * 388 newnfs_getcred(void) 389 { 390 struct ucred *cred; 391 struct thread *td = curthread; 392 393 cred = crdup(td->td_ucred); 394 newnfs_setroot(cred); 395 return (cred); 396 } 397 398 /* 399 * Nfs timer routine 400 * Call the nfsd's timer function once/sec. 401 */ 402 void 403 newnfs_timer(void *arg) 404 { 405 static time_t lasttime = 0; 406 /* 407 * Call the server timer, if set up. 408 * The argument indicates if it is the next second and therefore 409 * leases should be checked. 410 */ 411 if (lasttime != NFSD_MONOSEC) { 412 lasttime = NFSD_MONOSEC; 413 if (nfsd_call_servertimer != NULL) 414 (*nfsd_call_servertimer)(); 415 } 416 callout_reset(&newnfsd_callout, nfscl_ticks, newnfs_timer, NULL); 417 } 418 419 420 /* 421 * Sleep for a short period of time unless errval == NFSERR_GRACE, where 422 * the sleep should be for 5 seconds. 423 * Since lbolt doesn't exist in FreeBSD-CURRENT, just use a timeout on 424 * an event that never gets a wakeup. Only return EINTR or 0. 425 */ 426 int 427 nfs_catnap(int prio, int errval, const char *wmesg) 428 { 429 static int non_event; 430 int ret; 431 432 if (errval == NFSERR_GRACE) 433 ret = tsleep(&non_event, prio, wmesg, 5 * hz); 434 else 435 ret = tsleep(&non_event, prio, wmesg, 1); 436 if (ret != EINTR) 437 ret = 0; 438 return (ret); 439 } 440 441 /* 442 * Get referral. For now, just fail. 443 */ 444 struct nfsreferral * 445 nfsv4root_getreferral(struct vnode *vp, struct vnode *dvp, u_int32_t fileno) 446 { 447 448 return (NULL); 449 } 450 451 static int 452 nfssvc_nfscommon(struct thread *td, struct nfssvc_args *uap) 453 { 454 int error; 455 456 error = nfssvc_call(td, uap, td->td_ucred); 457 NFSEXITCODE(error); 458 return (error); 459 } 460 461 static int 462 nfssvc_call(struct thread *p, struct nfssvc_args *uap, struct ucred *cred) 463 { 464 int error = EINVAL, i, j; 465 struct nfsd_idargs nid; 466 struct nfsd_oidargs onid; 467 struct { 468 int vers; /* Just the first field of nfsstats. */ 469 } nfsstatver; 470 471 if (uap->flag & NFSSVC_IDNAME) { 472 if ((uap->flag & NFSSVC_NEWSTRUCT) != 0) 473 error = copyin(uap->argp, &nid, sizeof(nid)); 474 else { 475 error = copyin(uap->argp, &onid, sizeof(onid)); 476 if (error == 0) { 477 nid.nid_flag = onid.nid_flag; 478 nid.nid_uid = onid.nid_uid; 479 nid.nid_gid = onid.nid_gid; 480 nid.nid_usermax = onid.nid_usermax; 481 nid.nid_usertimeout = onid.nid_usertimeout; 482 nid.nid_name = onid.nid_name; 483 nid.nid_namelen = onid.nid_namelen; 484 nid.nid_ngroup = 0; 485 nid.nid_grps = NULL; 486 } 487 } 488 if (error) 489 goto out; 490 error = nfssvc_idname(&nid); 491 goto out; 492 } else if (uap->flag & NFSSVC_GETSTATS) { 493 if ((uap->flag & NFSSVC_NEWSTRUCT) == 0) { 494 /* Copy fields to the old ext_nfsstat structure. */ 495 oldnfsstats.attrcache_hits = 496 nfsstatsv1.attrcache_hits; 497 oldnfsstats.attrcache_misses = 498 nfsstatsv1.attrcache_misses; 499 oldnfsstats.lookupcache_hits = 500 nfsstatsv1.lookupcache_hits; 501 oldnfsstats.lookupcache_misses = 502 nfsstatsv1.lookupcache_misses; 503 oldnfsstats.direofcache_hits = 504 nfsstatsv1.direofcache_hits; 505 oldnfsstats.direofcache_misses = 506 nfsstatsv1.direofcache_misses; 507 oldnfsstats.accesscache_hits = 508 nfsstatsv1.accesscache_hits; 509 oldnfsstats.accesscache_misses = 510 nfsstatsv1.accesscache_misses; 511 oldnfsstats.biocache_reads = 512 nfsstatsv1.biocache_reads; 513 oldnfsstats.read_bios = 514 nfsstatsv1.read_bios; 515 oldnfsstats.read_physios = 516 nfsstatsv1.read_physios; 517 oldnfsstats.biocache_writes = 518 nfsstatsv1.biocache_writes; 519 oldnfsstats.write_bios = 520 nfsstatsv1.write_bios; 521 oldnfsstats.write_physios = 522 nfsstatsv1.write_physios; 523 oldnfsstats.biocache_readlinks = 524 nfsstatsv1.biocache_readlinks; 525 oldnfsstats.readlink_bios = 526 nfsstatsv1.readlink_bios; 527 oldnfsstats.biocache_readdirs = 528 nfsstatsv1.biocache_readdirs; 529 oldnfsstats.readdir_bios = 530 nfsstatsv1.readdir_bios; 531 for (i = 0; i < NFSV4_NPROCS; i++) 532 oldnfsstats.rpccnt[i] = nfsstatsv1.rpccnt[i]; 533 oldnfsstats.rpcretries = nfsstatsv1.rpcretries; 534 for (i = 0; i < NFSV4OP_NOPS; i++) 535 oldnfsstats.srvrpccnt[i] = 536 nfsstatsv1.srvrpccnt[i]; 537 for (i = NFSV42_NOPS, j = NFSV4OP_NOPS; 538 i < NFSV42_NOPS + NFSV4OP_FAKENOPS; i++, j++) 539 oldnfsstats.srvrpccnt[j] = 540 nfsstatsv1.srvrpccnt[i]; 541 oldnfsstats.srvrpc_errs = nfsstatsv1.srvrpc_errs; 542 oldnfsstats.srv_errs = nfsstatsv1.srv_errs; 543 oldnfsstats.rpcrequests = nfsstatsv1.rpcrequests; 544 oldnfsstats.rpctimeouts = nfsstatsv1.rpctimeouts; 545 oldnfsstats.rpcunexpected = nfsstatsv1.rpcunexpected; 546 oldnfsstats.rpcinvalid = nfsstatsv1.rpcinvalid; 547 oldnfsstats.srvcache_inproghits = 548 nfsstatsv1.srvcache_inproghits; 549 oldnfsstats.srvcache_idemdonehits = 550 nfsstatsv1.srvcache_idemdonehits; 551 oldnfsstats.srvcache_nonidemdonehits = 552 nfsstatsv1.srvcache_nonidemdonehits; 553 oldnfsstats.srvcache_misses = 554 nfsstatsv1.srvcache_misses; 555 oldnfsstats.srvcache_tcppeak = 556 nfsstatsv1.srvcache_tcppeak; 557 oldnfsstats.srvcache_size = nfsstatsv1.srvcache_size; 558 oldnfsstats.srvclients = nfsstatsv1.srvclients; 559 oldnfsstats.srvopenowners = nfsstatsv1.srvopenowners; 560 oldnfsstats.srvopens = nfsstatsv1.srvopens; 561 oldnfsstats.srvlockowners = nfsstatsv1.srvlockowners; 562 oldnfsstats.srvlocks = nfsstatsv1.srvlocks; 563 oldnfsstats.srvdelegates = nfsstatsv1.srvdelegates; 564 for (i = 0; i < NFSV4OP_CBNOPS; i++) 565 oldnfsstats.cbrpccnt[i] = 566 nfsstatsv1.cbrpccnt[i]; 567 oldnfsstats.clopenowners = nfsstatsv1.clopenowners; 568 oldnfsstats.clopens = nfsstatsv1.clopens; 569 oldnfsstats.cllockowners = nfsstatsv1.cllockowners; 570 oldnfsstats.cllocks = nfsstatsv1.cllocks; 571 oldnfsstats.cldelegates = nfsstatsv1.cldelegates; 572 oldnfsstats.cllocalopenowners = 573 nfsstatsv1.cllocalopenowners; 574 oldnfsstats.cllocalopens = nfsstatsv1.cllocalopens; 575 oldnfsstats.cllocallockowners = 576 nfsstatsv1.cllocallockowners; 577 oldnfsstats.cllocallocks = nfsstatsv1.cllocallocks; 578 error = copyout(&oldnfsstats, uap->argp, 579 sizeof (oldnfsstats)); 580 } else { 581 error = copyin(uap->argp, &nfsstatver, 582 sizeof(nfsstatver)); 583 if (error == 0 && nfsstatver.vers != NFSSTATS_V1) 584 error = EPERM; 585 if (error == 0) 586 error = copyout(&nfsstatsv1, uap->argp, 587 sizeof (nfsstatsv1)); 588 } 589 if (error == 0) { 590 if ((uap->flag & NFSSVC_ZEROCLTSTATS) != 0) { 591 nfsstatsv1.attrcache_hits = 0; 592 nfsstatsv1.attrcache_misses = 0; 593 nfsstatsv1.lookupcache_hits = 0; 594 nfsstatsv1.lookupcache_misses = 0; 595 nfsstatsv1.direofcache_hits = 0; 596 nfsstatsv1.direofcache_misses = 0; 597 nfsstatsv1.accesscache_hits = 0; 598 nfsstatsv1.accesscache_misses = 0; 599 nfsstatsv1.biocache_reads = 0; 600 nfsstatsv1.read_bios = 0; 601 nfsstatsv1.read_physios = 0; 602 nfsstatsv1.biocache_writes = 0; 603 nfsstatsv1.write_bios = 0; 604 nfsstatsv1.write_physios = 0; 605 nfsstatsv1.biocache_readlinks = 0; 606 nfsstatsv1.readlink_bios = 0; 607 nfsstatsv1.biocache_readdirs = 0; 608 nfsstatsv1.readdir_bios = 0; 609 nfsstatsv1.rpcretries = 0; 610 nfsstatsv1.rpcrequests = 0; 611 nfsstatsv1.rpctimeouts = 0; 612 nfsstatsv1.rpcunexpected = 0; 613 nfsstatsv1.rpcinvalid = 0; 614 bzero(nfsstatsv1.rpccnt, 615 sizeof(nfsstatsv1.rpccnt)); 616 } 617 if ((uap->flag & NFSSVC_ZEROSRVSTATS) != 0) { 618 nfsstatsv1.srvrpc_errs = 0; 619 nfsstatsv1.srv_errs = 0; 620 nfsstatsv1.srvcache_inproghits = 0; 621 nfsstatsv1.srvcache_idemdonehits = 0; 622 nfsstatsv1.srvcache_nonidemdonehits = 0; 623 nfsstatsv1.srvcache_misses = 0; 624 nfsstatsv1.srvcache_tcppeak = 0; 625 bzero(nfsstatsv1.srvrpccnt, 626 sizeof(nfsstatsv1.srvrpccnt)); 627 bzero(nfsstatsv1.cbrpccnt, 628 sizeof(nfsstatsv1.cbrpccnt)); 629 } 630 } 631 goto out; 632 } else if (uap->flag & NFSSVC_NFSUSERDPORT) { 633 u_short sockport; 634 struct sockaddr *sad; 635 struct sockaddr_un *sun; 636 637 if ((uap->flag & NFSSVC_NEWSTRUCT) != 0) { 638 /* New nfsuserd using an AF_LOCAL socket. */ 639 sun = malloc(sizeof(struct sockaddr_un), M_SONAME, 640 M_WAITOK | M_ZERO); 641 error = copyinstr(uap->argp, sun->sun_path, 642 sizeof(sun->sun_path), NULL); 643 if (error != 0) { 644 free(sun, M_SONAME); 645 return (error); 646 } 647 sun->sun_family = AF_LOCAL; 648 sun->sun_len = SUN_LEN(sun); 649 sockport = 0; 650 sad = (struct sockaddr *)sun; 651 } else { 652 error = copyin(uap->argp, (caddr_t)&sockport, 653 sizeof (u_short)); 654 sad = NULL; 655 } 656 if (error == 0) 657 error = nfsrv_nfsuserdport(sad, sockport, p); 658 } else if (uap->flag & NFSSVC_NFSUSERDDELPORT) { 659 nfsrv_nfsuserddelport(); 660 error = 0; 661 } 662 663 out: 664 NFSEXITCODE(error); 665 return (error); 666 } 667 668 /* 669 * called by all three modevent routines, so that it gets things 670 * initialized soon enough. 671 */ 672 void 673 newnfs_portinit(void) 674 { 675 static int inited = 0; 676 677 if (inited) 678 return; 679 inited = 1; 680 /* Initialize SMP locks used by both client and server. */ 681 mtx_init(&newnfsd_mtx, "newnfsd_mtx", NULL, MTX_DEF); 682 mtx_init(&nfs_state_mutex, "nfs_state_mutex", NULL, MTX_DEF); 683 mtx_init(&nfs_clstate_mutex, "nfs_clstate_mutex", NULL, MTX_DEF); 684 } 685 686 /* 687 * Determine if the file system supports NFSv4 ACLs. 688 * Return 1 if it does, 0 otherwise. 689 */ 690 int 691 nfs_supportsnfsv4acls(struct vnode *vp) 692 { 693 int error; 694 long retval; 695 696 ASSERT_VOP_LOCKED(vp, "nfs supports nfsv4acls"); 697 698 if (nfsrv_useacl == 0) 699 return (0); 700 error = VOP_PATHCONF(vp, _PC_ACL_NFS4, &retval); 701 if (error == 0 && retval != 0) 702 return (1); 703 return (0); 704 } 705 706 /* 707 * These are the first fields of all the context structures passed into 708 * nfs_pnfsio(). 709 */ 710 struct pnfsio { 711 int done; 712 int inprog; 713 struct task tsk; 714 }; 715 716 /* 717 * Do a mirror I/O on a pNFS thread. 718 */ 719 int 720 nfs_pnfsio(task_fn_t *func, void *context) 721 { 722 struct pnfsio *pio; 723 int ret; 724 static struct taskqueue *pnfsioq = NULL; 725 726 pio = (struct pnfsio *)context; 727 if (pnfsioq == NULL) { 728 if (nfs_pnfsiothreads == 0) 729 return (EPERM); 730 if (nfs_pnfsiothreads < 0) 731 nfs_pnfsiothreads = mp_ncpus * 4; 732 pnfsioq = taskqueue_create("pnfsioq", M_WAITOK, 733 taskqueue_thread_enqueue, &pnfsioq); 734 if (pnfsioq == NULL) 735 return (ENOMEM); 736 ret = taskqueue_start_threads(&pnfsioq, nfs_pnfsiothreads, 737 0, "pnfsiot"); 738 if (ret != 0) { 739 taskqueue_free(pnfsioq); 740 pnfsioq = NULL; 741 return (ret); 742 } 743 } 744 pio->inprog = 1; 745 TASK_INIT(&pio->tsk, 0, func, context); 746 ret = taskqueue_enqueue(pnfsioq, &pio->tsk); 747 if (ret != 0) 748 pio->inprog = 0; 749 return (ret); 750 } 751 752 extern int (*nfsd_call_nfscommon)(struct thread *, struct nfssvc_args *); 753 754 /* 755 * Called once to initialize data structures... 756 */ 757 static int 758 nfscommon_modevent(module_t mod, int type, void *data) 759 { 760 int error = 0; 761 static int loaded = 0; 762 763 switch (type) { 764 case MOD_LOAD: 765 if (loaded) 766 goto out; 767 newnfs_portinit(); 768 mtx_init(&nfs_nameid_mutex, "nfs_nameid_mutex", NULL, MTX_DEF); 769 mtx_init(&nfs_sockl_mutex, "nfs_sockl_mutex", NULL, MTX_DEF); 770 mtx_init(&nfs_slock_mutex, "nfs_slock_mutex", NULL, MTX_DEF); 771 mtx_init(&nfs_req_mutex, "nfs_req_mutex", NULL, MTX_DEF); 772 mtx_init(&nfsrv_nfsuserdsock.nr_mtx, "nfsuserd", NULL, 773 MTX_DEF); 774 mtx_init(&nfsrv_dslock_mtx, "nfs4ds", NULL, MTX_DEF); 775 TAILQ_INIT(&nfsrv_devidhead); 776 callout_init(&newnfsd_callout, 1); 777 newnfs_init(); 778 nfsd_call_nfscommon = nfssvc_nfscommon; 779 loaded = 1; 780 break; 781 782 case MOD_UNLOAD: 783 if (newnfs_numnfsd != 0 || nfsrv_nfsuserd != 0 || 784 nfs_numnfscbd != 0) { 785 error = EBUSY; 786 break; 787 } 788 789 nfsd_call_nfscommon = NULL; 790 callout_drain(&newnfsd_callout); 791 /* Clean out the name<-->id cache. */ 792 nfsrv_cleanusergroup(); 793 /* and get rid of the mutexes */ 794 mtx_destroy(&nfs_nameid_mutex); 795 mtx_destroy(&newnfsd_mtx); 796 mtx_destroy(&nfs_state_mutex); 797 mtx_destroy(&nfs_clstate_mutex); 798 mtx_destroy(&nfs_sockl_mutex); 799 mtx_destroy(&nfs_slock_mutex); 800 mtx_destroy(&nfs_req_mutex); 801 mtx_destroy(&nfsrv_nfsuserdsock.nr_mtx); 802 mtx_destroy(&nfsrv_dslock_mtx); 803 loaded = 0; 804 break; 805 default: 806 error = EOPNOTSUPP; 807 break; 808 } 809 810 out: 811 NFSEXITCODE(error); 812 return error; 813 } 814 static moduledata_t nfscommon_mod = { 815 "nfscommon", 816 nfscommon_modevent, 817 NULL, 818 }; 819 DECLARE_MODULE(nfscommon, nfscommon_mod, SI_SUB_VFS, SI_ORDER_ANY); 820 821 /* So that loader and kldload(2) can find us, wherever we are.. */ 822 MODULE_VERSION(nfscommon, 1); 823 MODULE_DEPEND(nfscommon, nfssvc, 1, 1, 1); 824 MODULE_DEPEND(nfscommon, krpc, 1, 1, 1); 825 826