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