1 /* $NetBSD: nfs_syscalls.c,v 1.33 1998/11/08 15:57:44 mycroft Exp $ */ 2 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. All advertising materials mentioning features or use of this software 19 * must display the following acknowledgement: 20 * This product includes software developed by the University of 21 * California, Berkeley and its contributors. 22 * 4. Neither the name of the University nor the names of its contributors 23 * may be used to endorse or promote products derived from this software 24 * without specific prior written permission. 25 * 26 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND 27 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE 28 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE 29 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE 30 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL 31 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS 32 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) 33 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT 34 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY 35 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF 36 * SUCH DAMAGE. 37 * 38 * @(#)nfs_syscalls.c 8.5 (Berkeley) 3/30/95 39 */ 40 41 #include "fs_nfs.h" 42 #include "opt_nfsserver.h" 43 #include "opt_iso.h" 44 45 #include <sys/param.h> 46 #include <sys/systm.h> 47 #include <sys/kernel.h> 48 #include <sys/file.h> 49 #include <sys/stat.h> 50 #include <sys/vnode.h> 51 #include <sys/mount.h> 52 #include <sys/proc.h> 53 #include <sys/uio.h> 54 #include <sys/malloc.h> 55 #include <sys/buf.h> 56 #include <sys/mbuf.h> 57 #include <sys/socket.h> 58 #include <sys/socketvar.h> 59 #include <sys/domain.h> 60 #include <sys/protosw.h> 61 #include <sys/namei.h> 62 #include <sys/syslog.h> 63 #include <sys/filedesc.h> 64 65 #include <sys/syscallargs.h> 66 67 #include <netinet/in.h> 68 #include <netinet/tcp.h> 69 #ifdef ISO 70 #include <netiso/iso.h> 71 #endif 72 #include <nfs/xdr_subs.h> 73 #include <nfs/rpcv2.h> 74 #include <nfs/nfsproto.h> 75 #include <nfs/nfs.h> 76 #include <nfs/nfsm_subs.h> 77 #include <nfs/nfsrvcache.h> 78 #include <nfs/nfsmount.h> 79 #include <nfs/nfsnode.h> 80 #include <nfs/nqnfs.h> 81 #include <nfs/nfsrtt.h> 82 #include <nfs/nfs_var.h> 83 84 void nfsrv_zapsock __P((struct nfssvc_sock *)); 85 86 /* Global defs. */ 87 extern int32_t (*nfsrv3_procs[NFS_NPROCS]) __P((struct nfsrv_descript *, 88 struct nfssvc_sock *, 89 struct proc *, struct mbuf **)); 90 extern int nfs_numasync; 91 extern time_t nqnfsstarttime; 92 extern int nqsrv_writeslack; 93 extern int nfsrtton; 94 extern struct nfsstats nfsstats; 95 extern int nfsrvw_procrastinate; 96 struct nfssvc_sock *nfs_udpsock, *nfs_cltpsock; 97 int nuidhash_max = NFS_MAXUIDHASH; 98 int nfsd_waiting = 0; 99 #ifdef NFSSERVER 100 static int nfs_numnfsd = 0; 101 static int notstarted = 1; 102 static int modify_flag = 0; 103 static struct nfsdrt nfsdrt; 104 extern struct nfs_public nfs_pub; 105 #endif 106 107 #define TRUE 1 108 #define FALSE 0 109 110 #ifdef NFS 111 static int nfs_asyncdaemon[NFS_MAXASYNCDAEMON]; 112 #endif 113 114 #ifdef NFSSERVER 115 static void nfsd_rt __P((int, struct nfsrv_descript *, int)); 116 #endif 117 118 /* 119 * NFS server system calls 120 * getfh() lives here too, but maybe should move to kern/vfs_syscalls.c 121 */ 122 123 /* 124 * Get file handle system call 125 */ 126 int 127 sys_getfh(p, v, retval) 128 struct proc *p; 129 register void *v; 130 register_t *retval; 131 { 132 register struct sys_getfh_args /* { 133 syscallarg(char *) fname; 134 syscallarg(fhandle_t *) fhp; 135 } */ *uap = v; 136 register struct vnode *vp; 137 fhandle_t fh; 138 int error; 139 struct nameidata nd; 140 141 /* 142 * Must be super user 143 */ 144 error = suser(p->p_ucred, &p->p_acflag); 145 if (error) 146 return (error); 147 NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF, UIO_USERSPACE, 148 SCARG(uap, fname), p); 149 error = namei(&nd); 150 if (error) 151 return (error); 152 vp = nd.ni_vp; 153 memset((caddr_t)&fh, 0, sizeof(fh)); 154 fh.fh_fsid = vp->v_mount->mnt_stat.f_fsid; 155 error = VFS_VPTOFH(vp, &fh.fh_fid); 156 vput(vp); 157 if (error) 158 return (error); 159 error = copyout((caddr_t)&fh, (caddr_t)SCARG(uap, fhp), sizeof (fh)); 160 return (error); 161 } 162 163 /* 164 * Nfs server pseudo system call for the nfsd's 165 * Based on the flag value it either: 166 * - adds a socket to the selection list 167 * - remains in the kernel as an nfsd 168 * - remains in the kernel as an nfsiod 169 */ 170 int 171 sys_nfssvc(p, v, retval) 172 struct proc *p; 173 void *v; 174 register_t *retval; 175 { 176 register struct sys_nfssvc_args /* { 177 syscallarg(int) flag; 178 syscallarg(caddr_t) argp; 179 } */ *uap = v; 180 int error; 181 #ifdef NFS 182 struct nameidata nd; 183 struct nfsmount *nmp; 184 struct nfsd_cargs ncd; 185 #endif 186 #ifdef NFSSERVER 187 struct file *fp; 188 struct mbuf *nam; 189 struct nfsd_args nfsdarg; 190 struct nfsd_srvargs nfsd_srvargs, *nsd = &nfsd_srvargs; 191 struct nfsd *nfsd; 192 struct nfssvc_sock *slp; 193 struct nfsuid *nuidp; 194 #endif 195 196 /* 197 * Must be super user 198 */ 199 error = suser(p->p_ucred, &p->p_acflag); 200 if(error) 201 return (error); 202 while (nfssvc_sockhead_flag & SLP_INIT) { 203 nfssvc_sockhead_flag |= SLP_WANTINIT; 204 (void) tsleep((caddr_t)&nfssvc_sockhead, PSOCK, "nfsd init", 0); 205 } 206 if (SCARG(uap, flag) & NFSSVC_BIOD) { 207 #ifdef NFS 208 error = nfssvc_iod(p); 209 #else 210 error = ENOSYS; 211 #endif 212 } else if (SCARG(uap, flag) & NFSSVC_MNTD) { 213 #ifndef NFS 214 error = ENOSYS; 215 #else 216 error = copyin(SCARG(uap, argp), (caddr_t)&ncd, sizeof (ncd)); 217 if (error) 218 return (error); 219 NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF, UIO_USERSPACE, 220 ncd.ncd_dirp, p); 221 error = namei(&nd); 222 if (error) 223 return (error); 224 if ((nd.ni_vp->v_flag & VROOT) == 0) 225 error = EINVAL; 226 nmp = VFSTONFS(nd.ni_vp->v_mount); 227 vput(nd.ni_vp); 228 if (error) 229 return (error); 230 if ((nmp->nm_iflag & NFSMNT_MNTD) && 231 (SCARG(uap, flag) & NFSSVC_GOTAUTH) == 0) 232 return (0); 233 nmp->nm_iflag |= NFSMNT_MNTD; 234 error = nqnfs_clientd(nmp, p->p_ucred, &ncd, SCARG(uap, flag), 235 SCARG(uap, argp), p); 236 #endif /* NFS */ 237 } else if (SCARG(uap, flag) & NFSSVC_ADDSOCK) { 238 #ifndef NFSSERVER 239 error = ENOSYS; 240 #else 241 error = copyin(SCARG(uap, argp), (caddr_t)&nfsdarg, 242 sizeof(nfsdarg)); 243 if (error) 244 return (error); 245 error = getsock(p->p_fd, nfsdarg.sock, &fp); 246 if (error) 247 return (error); 248 /* 249 * Get the client address for connected sockets. 250 */ 251 if (nfsdarg.name == NULL || nfsdarg.namelen == 0) 252 nam = (struct mbuf *)0; 253 else { 254 error = sockargs(&nam, nfsdarg.name, nfsdarg.namelen, 255 MT_SONAME); 256 if (error) 257 return (error); 258 } 259 error = nfssvc_addsock(fp, nam); 260 #endif /* !NFSSERVER */ 261 } else { 262 #ifndef NFSSERVER 263 error = ENOSYS; 264 #else 265 error = copyin(SCARG(uap, argp), (caddr_t)nsd, sizeof (*nsd)); 266 if (error) 267 return (error); 268 if ((SCARG(uap, flag) & NFSSVC_AUTHIN) && 269 ((nfsd = nsd->nsd_nfsd)) != NULL && 270 (nfsd->nfsd_slp->ns_flag & SLP_VALID)) { 271 slp = nfsd->nfsd_slp; 272 273 /* 274 * First check to see if another nfsd has already 275 * added this credential. 276 */ 277 for (nuidp = NUIDHASH(slp,nsd->nsd_cr.cr_uid)->lh_first; 278 nuidp != 0; nuidp = nuidp->nu_hash.le_next) { 279 if (nuidp->nu_cr.cr_uid == nsd->nsd_cr.cr_uid && 280 (!nfsd->nfsd_nd->nd_nam2 || 281 netaddr_match(NU_NETFAM(nuidp), 282 &nuidp->nu_haddr, nfsd->nfsd_nd->nd_nam2))) 283 break; 284 } 285 if (nuidp) { 286 nfsrv_setcred(&nuidp->nu_cr,&nfsd->nfsd_nd->nd_cr); 287 nfsd->nfsd_nd->nd_flag |= ND_KERBFULL; 288 } else { 289 /* 290 * Nope, so we will. 291 */ 292 if (slp->ns_numuids < nuidhash_max) { 293 slp->ns_numuids++; 294 nuidp = (struct nfsuid *) 295 malloc(sizeof (struct nfsuid), M_NFSUID, 296 M_WAITOK); 297 } else 298 nuidp = (struct nfsuid *)0; 299 if ((slp->ns_flag & SLP_VALID) == 0) { 300 if (nuidp) 301 free((caddr_t)nuidp, M_NFSUID); 302 } else { 303 if (nuidp == (struct nfsuid *)0) { 304 nuidp = slp->ns_uidlruhead.tqh_first; 305 LIST_REMOVE(nuidp, nu_hash); 306 TAILQ_REMOVE(&slp->ns_uidlruhead, nuidp, 307 nu_lru); 308 if (nuidp->nu_flag & NU_NAM) 309 m_freem(nuidp->nu_nam); 310 } 311 nuidp->nu_flag = 0; 312 nuidp->nu_cr = nsd->nsd_cr; 313 if (nuidp->nu_cr.cr_ngroups > NGROUPS) 314 nuidp->nu_cr.cr_ngroups = NGROUPS; 315 nuidp->nu_cr.cr_ref = 1; 316 nuidp->nu_timestamp = nsd->nsd_timestamp; 317 nuidp->nu_expire = time.tv_sec + nsd->nsd_ttl; 318 /* 319 * and save the session key in nu_key. 320 */ 321 memcpy(nuidp->nu_key, nsd->nsd_key, 322 sizeof(nsd->nsd_key)); 323 if (nfsd->nfsd_nd->nd_nam2) { 324 struct sockaddr_in *saddr; 325 326 saddr = mtod(nfsd->nfsd_nd->nd_nam2, 327 struct sockaddr_in *); 328 switch (saddr->sin_family) { 329 case AF_INET: 330 nuidp->nu_flag |= NU_INETADDR; 331 nuidp->nu_inetaddr = 332 saddr->sin_addr.s_addr; 333 break; 334 case AF_ISO: 335 default: 336 nuidp->nu_flag |= NU_NAM; 337 nuidp->nu_nam = m_copym( 338 nfsd->nfsd_nd->nd_nam2, 0, 339 M_COPYALL, M_WAIT); 340 break; 341 }; 342 } 343 TAILQ_INSERT_TAIL(&slp->ns_uidlruhead, nuidp, 344 nu_lru); 345 LIST_INSERT_HEAD(NUIDHASH(slp, nsd->nsd_uid), 346 nuidp, nu_hash); 347 nfsrv_setcred(&nuidp->nu_cr, 348 &nfsd->nfsd_nd->nd_cr); 349 nfsd->nfsd_nd->nd_flag |= ND_KERBFULL; 350 } 351 } 352 } 353 if ((SCARG(uap, flag) & NFSSVC_AUTHINFAIL) && 354 (nfsd = nsd->nsd_nfsd)) 355 nfsd->nfsd_flag |= NFSD_AUTHFAIL; 356 error = nfssvc_nfsd(nsd, SCARG(uap, argp), p); 357 #endif /* !NFSSERVER */ 358 } 359 if (error == EINTR || error == ERESTART) 360 error = 0; 361 return (error); 362 } 363 364 #ifdef NFSSERVER 365 /* 366 * Adds a socket to the list for servicing by nfsds. 367 */ 368 int 369 nfssvc_addsock(fp, mynam) 370 struct file *fp; 371 struct mbuf *mynam; 372 { 373 register struct mbuf *m; 374 register int siz; 375 register struct nfssvc_sock *slp; 376 register struct socket *so; 377 struct nfssvc_sock *tslp; 378 int error, s; 379 380 so = (struct socket *)fp->f_data; 381 tslp = (struct nfssvc_sock *)0; 382 /* 383 * Add it to the list, as required. 384 */ 385 if (so->so_proto->pr_protocol == IPPROTO_UDP) { 386 tslp = nfs_udpsock; 387 if (tslp->ns_flag & SLP_VALID) { 388 m_freem(mynam); 389 return (EPERM); 390 } 391 #ifdef ISO 392 } else if (so->so_proto->pr_protocol == ISOPROTO_CLTP) { 393 tslp = nfs_cltpsock; 394 if (tslp->ns_flag & SLP_VALID) { 395 m_freem(mynam); 396 return (EPERM); 397 } 398 #endif /* ISO */ 399 } 400 if (so->so_type == SOCK_STREAM) 401 siz = NFS_MAXPACKET + sizeof (u_long); 402 else 403 siz = NFS_MAXPACKET; 404 error = soreserve(so, siz, siz); 405 if (error) { 406 m_freem(mynam); 407 return (error); 408 } 409 410 /* 411 * Set protocol specific options { for now TCP only } and 412 * reserve some space. For datagram sockets, this can get called 413 * repeatedly for the same socket, but that isn't harmful. 414 */ 415 if (so->so_type == SOCK_STREAM) { 416 MGET(m, M_WAIT, MT_SOOPTS); 417 *mtod(m, int32_t *) = 1; 418 m->m_len = sizeof(int32_t); 419 sosetopt(so, SOL_SOCKET, SO_KEEPALIVE, m); 420 } 421 if (so->so_proto->pr_domain->dom_family == AF_INET && 422 so->so_proto->pr_protocol == IPPROTO_TCP) { 423 MGET(m, M_WAIT, MT_SOOPTS); 424 *mtod(m, int32_t *) = 1; 425 m->m_len = sizeof(int32_t); 426 sosetopt(so, IPPROTO_TCP, TCP_NODELAY, m); 427 } 428 so->so_rcv.sb_flags &= ~SB_NOINTR; 429 so->so_rcv.sb_timeo = 0; 430 so->so_snd.sb_flags &= ~SB_NOINTR; 431 so->so_snd.sb_timeo = 0; 432 if (tslp) 433 slp = tslp; 434 else { 435 slp = (struct nfssvc_sock *) 436 malloc(sizeof (struct nfssvc_sock), M_NFSSVC, M_WAITOK); 437 memset((caddr_t)slp, 0, sizeof (struct nfssvc_sock)); 438 TAILQ_INIT(&slp->ns_uidlruhead); 439 TAILQ_INSERT_TAIL(&nfssvc_sockhead, slp, ns_chain); 440 } 441 slp->ns_so = so; 442 slp->ns_nam = mynam; 443 fp->f_count++; 444 slp->ns_fp = fp; 445 s = splsoftnet(); 446 so->so_upcallarg = (caddr_t)slp; 447 so->so_upcall = nfsrv_rcv; 448 so->so_rcv.sb_flags |= SB_UPCALL; 449 slp->ns_flag = (SLP_VALID | SLP_NEEDQ); 450 nfsrv_wakenfsd(slp); 451 splx(s); 452 return (0); 453 } 454 455 /* 456 * Called by nfssvc() for nfsds. Just loops around servicing rpc requests 457 * until it is killed by a signal. 458 */ 459 int 460 nfssvc_nfsd(nsd, argp, p) 461 struct nfsd_srvargs *nsd; 462 caddr_t argp; 463 struct proc *p; 464 { 465 register struct mbuf *m; 466 register int siz; 467 register struct nfssvc_sock *slp; 468 register struct socket *so; 469 register int *solockp; 470 struct nfsd *nfsd = nsd->nsd_nfsd; 471 struct nfsrv_descript *nd = NULL; 472 struct mbuf *mreq; 473 int error = 0, cacherep, s, sotype, writes_todo; 474 u_quad_t cur_usec; 475 476 #ifndef nolint 477 cacherep = RC_DOIT; 478 writes_todo = 0; 479 #endif 480 s = splsoftnet(); 481 if (nfsd == (struct nfsd *)0) { 482 nsd->nsd_nfsd = nfsd = (struct nfsd *) 483 malloc(sizeof (struct nfsd), M_NFSD, M_WAITOK); 484 memset((caddr_t)nfsd, 0, sizeof (struct nfsd)); 485 nfsd->nfsd_procp = p; 486 TAILQ_INSERT_TAIL(&nfsd_head, nfsd, nfsd_chain); 487 nfs_numnfsd++; 488 } 489 p->p_holdcnt++; 490 /* 491 * Loop getting rpc requests until SIGKILL. 492 */ 493 for (;;) { 494 if ((nfsd->nfsd_flag & NFSD_REQINPROG) == 0) { 495 while (nfsd->nfsd_slp == (struct nfssvc_sock *)0 && 496 (nfsd_head_flag & NFSD_CHECKSLP) == 0) { 497 nfsd->nfsd_flag |= NFSD_WAITING; 498 nfsd_waiting++; 499 error = tsleep((caddr_t)nfsd, PSOCK | PCATCH, 500 "nfsd", 0); 501 nfsd_waiting--; 502 if (error) 503 goto done; 504 } 505 if (nfsd->nfsd_slp == (struct nfssvc_sock *)0 && 506 (nfsd_head_flag & NFSD_CHECKSLP) != 0) { 507 for (slp = nfssvc_sockhead.tqh_first; slp != 0; 508 slp = slp->ns_chain.tqe_next) { 509 if ((slp->ns_flag & (SLP_VALID | SLP_DOREC)) 510 == (SLP_VALID | SLP_DOREC)) { 511 slp->ns_flag &= ~SLP_DOREC; 512 slp->ns_sref++; 513 nfsd->nfsd_slp = slp; 514 break; 515 } 516 } 517 if (slp == 0) 518 nfsd_head_flag &= ~NFSD_CHECKSLP; 519 } 520 if ((slp = nfsd->nfsd_slp) == (struct nfssvc_sock *)0) 521 continue; 522 if (slp->ns_flag & SLP_VALID) { 523 if (slp->ns_flag & SLP_DISCONN) 524 nfsrv_zapsock(slp); 525 else if (slp->ns_flag & SLP_NEEDQ) { 526 slp->ns_flag &= ~SLP_NEEDQ; 527 (void) nfs_sndlock(&slp->ns_solock, 528 (struct nfsreq *)0); 529 nfsrv_rcv(slp->ns_so, (caddr_t)slp, 530 M_WAIT); 531 nfs_sndunlock(&slp->ns_solock); 532 } 533 error = nfsrv_dorec(slp, nfsd, &nd); 534 cur_usec = (u_quad_t)time.tv_sec * 1000000 + 535 (u_quad_t)time.tv_usec; 536 if (error && slp->ns_tq.lh_first && 537 slp->ns_tq.lh_first->nd_time <= cur_usec) { 538 error = 0; 539 cacherep = RC_DOIT; 540 writes_todo = 1; 541 } else 542 writes_todo = 0; 543 nfsd->nfsd_flag |= NFSD_REQINPROG; 544 } 545 } else { 546 error = 0; 547 slp = nfsd->nfsd_slp; 548 } 549 if (error || (slp->ns_flag & SLP_VALID) == 0) { 550 if (nd) { 551 free((caddr_t)nd, M_NFSRVDESC); 552 nd = NULL; 553 } 554 nfsd->nfsd_slp = (struct nfssvc_sock *)0; 555 nfsd->nfsd_flag &= ~NFSD_REQINPROG; 556 nfsrv_slpderef(slp); 557 continue; 558 } 559 splx(s); 560 so = slp->ns_so; 561 sotype = so->so_type; 562 if (so->so_proto->pr_flags & PR_CONNREQUIRED) 563 solockp = &slp->ns_solock; 564 else 565 solockp = (int *)0; 566 if (nd) { 567 nd->nd_starttime = time; 568 if (nd->nd_nam2) 569 nd->nd_nam = nd->nd_nam2; 570 else 571 nd->nd_nam = slp->ns_nam; 572 573 /* 574 * Check to see if authorization is needed. 575 */ 576 if (nfsd->nfsd_flag & NFSD_NEEDAUTH) { 577 nfsd->nfsd_flag &= ~NFSD_NEEDAUTH; 578 nsd->nsd_haddr = mtod(nd->nd_nam, 579 struct sockaddr_in *)->sin_addr.s_addr; 580 nsd->nsd_authlen = nfsd->nfsd_authlen; 581 nsd->nsd_verflen = nfsd->nfsd_verflen; 582 if (!copyout(nfsd->nfsd_authstr,nsd->nsd_authstr, 583 nfsd->nfsd_authlen) && 584 !copyout(nfsd->nfsd_verfstr, nsd->nsd_verfstr, 585 nfsd->nfsd_verflen) && 586 !copyout((caddr_t)nsd, argp, sizeof (*nsd))) { 587 p->p_holdcnt--; 588 return (ENEEDAUTH); 589 } 590 cacherep = RC_DROPIT; 591 } else 592 cacherep = nfsrv_getcache(nd, slp, &mreq); 593 594 /* 595 * Check for just starting up for NQNFS and send 596 * fake "try again later" replies to the NQNFS clients. 597 */ 598 if (notstarted && nqnfsstarttime <= time.tv_sec) { 599 if (modify_flag) { 600 nqnfsstarttime = time.tv_sec + nqsrv_writeslack; 601 modify_flag = 0; 602 } else 603 notstarted = 0; 604 } 605 if (notstarted) { 606 if ((nd->nd_flag & ND_NQNFS) == 0) 607 cacherep = RC_DROPIT; 608 else if (nd->nd_procnum != NFSPROC_WRITE) { 609 nd->nd_procnum = NFSPROC_NOOP; 610 nd->nd_repstat = NQNFS_TRYLATER; 611 cacherep = RC_DOIT; 612 } else 613 modify_flag = 1; 614 } else if (nfsd->nfsd_flag & NFSD_AUTHFAIL) { 615 nfsd->nfsd_flag &= ~NFSD_AUTHFAIL; 616 nd->nd_procnum = NFSPROC_NOOP; 617 nd->nd_repstat = (NFSERR_AUTHERR | AUTH_TOOWEAK); 618 cacherep = RC_DOIT; 619 } 620 } 621 622 /* 623 * Loop to get all the write rpc relies that have been 624 * gathered together. 625 */ 626 do { 627 switch (cacherep) { 628 case RC_DOIT: 629 if (writes_todo || (nd->nd_procnum == NFSPROC_WRITE && 630 nfsrvw_procrastinate > 0 && !notstarted)) 631 error = nfsrv_writegather(&nd, slp, 632 nfsd->nfsd_procp, &mreq); 633 else 634 error = (*(nfsrv3_procs[nd->nd_procnum]))(nd, 635 slp, nfsd->nfsd_procp, &mreq); 636 if (mreq == NULL) 637 break; 638 if (error) { 639 if (nd->nd_procnum != NQNFSPROC_VACATED) 640 nfsstats.srv_errs++; 641 nfsrv_updatecache(nd, FALSE, mreq); 642 if (nd->nd_nam2) 643 m_freem(nd->nd_nam2); 644 break; 645 } 646 nfsstats.srvrpccnt[nd->nd_procnum]++; 647 nfsrv_updatecache(nd, TRUE, mreq); 648 nd->nd_mrep = (struct mbuf *)0; 649 case RC_REPLY: 650 m = mreq; 651 siz = 0; 652 while (m) { 653 siz += m->m_len; 654 m = m->m_next; 655 } 656 if (siz <= 0 || siz > NFS_MAXPACKET) { 657 printf("mbuf siz=%d\n",siz); 658 panic("Bad nfs svc reply"); 659 } 660 m = mreq; 661 m->m_pkthdr.len = siz; 662 m->m_pkthdr.rcvif = (struct ifnet *)0; 663 /* 664 * For stream protocols, prepend a Sun RPC 665 * Record Mark. 666 */ 667 if (sotype == SOCK_STREAM) { 668 M_PREPEND(m, NFSX_UNSIGNED, M_WAIT); 669 *mtod(m, u_int32_t *) = htonl(0x80000000 | siz); 670 } 671 if (solockp) 672 (void) nfs_sndlock(solockp, (struct nfsreq *)0); 673 if (slp->ns_flag & SLP_VALID) 674 error = nfs_send(so, nd->nd_nam2, m, NULL); 675 else { 676 error = EPIPE; 677 m_freem(m); 678 } 679 if (nfsrtton) 680 nfsd_rt(sotype, nd, cacherep); 681 if (nd->nd_nam2) 682 MFREE(nd->nd_nam2, m); 683 if (nd->nd_mrep) 684 m_freem(nd->nd_mrep); 685 if (error == EPIPE) 686 nfsrv_zapsock(slp); 687 if (solockp) 688 nfs_sndunlock(solockp); 689 if (error == EINTR || error == ERESTART) { 690 free((caddr_t)nd, M_NFSRVDESC); 691 nfsrv_slpderef(slp); 692 s = splsoftnet(); 693 goto done; 694 } 695 break; 696 case RC_DROPIT: 697 if (nfsrtton) 698 nfsd_rt(sotype, nd, cacherep); 699 m_freem(nd->nd_mrep); 700 m_freem(nd->nd_nam2); 701 break; 702 }; 703 if (nd) { 704 FREE((caddr_t)nd, M_NFSRVDESC); 705 nd = NULL; 706 } 707 708 /* 709 * Check to see if there are outstanding writes that 710 * need to be serviced. 711 */ 712 cur_usec = (u_quad_t)time.tv_sec * 1000000 + 713 (u_quad_t)time.tv_usec; 714 s = splsoftclock(); 715 if (slp->ns_tq.lh_first && 716 slp->ns_tq.lh_first->nd_time <= cur_usec) { 717 cacherep = RC_DOIT; 718 writes_todo = 1; 719 } else 720 writes_todo = 0; 721 splx(s); 722 } while (writes_todo); 723 s = splsoftnet(); 724 if (nfsrv_dorec(slp, nfsd, &nd)) { 725 nfsd->nfsd_flag &= ~NFSD_REQINPROG; 726 nfsd->nfsd_slp = NULL; 727 nfsrv_slpderef(slp); 728 } 729 } 730 done: 731 p->p_holdcnt--; 732 TAILQ_REMOVE(&nfsd_head, nfsd, nfsd_chain); 733 splx(s); 734 free((caddr_t)nfsd, M_NFSD); 735 nsd->nsd_nfsd = (struct nfsd *)0; 736 if (--nfs_numnfsd == 0) 737 nfsrv_init(TRUE); /* Reinitialize everything */ 738 return (error); 739 } 740 741 /* 742 * Shut down a socket associated with an nfssvc_sock structure. 743 * Should be called with the send lock set, if required. 744 * The trick here is to increment the sref at the start, so that the nfsds 745 * will stop using it and clear ns_flag at the end so that it will not be 746 * reassigned during cleanup. 747 */ 748 void 749 nfsrv_zapsock(slp) 750 register struct nfssvc_sock *slp; 751 { 752 register struct nfsuid *nuidp, *nnuidp; 753 register struct nfsrv_descript *nwp, *nnwp; 754 struct socket *so; 755 struct file *fp; 756 struct mbuf *m; 757 int s; 758 759 slp->ns_flag &= ~SLP_ALLFLAGS; 760 fp = slp->ns_fp; 761 if (fp) { 762 slp->ns_fp = (struct file *)0; 763 so = slp->ns_so; 764 so->so_upcall = NULL; 765 so->so_upcallarg = NULL; 766 so->so_rcv.sb_flags &= ~SB_UPCALL; 767 soshutdown(so, 2); 768 closef(fp, (struct proc *)0); 769 if (slp->ns_nam) 770 MFREE(slp->ns_nam, m); 771 m_freem(slp->ns_raw); 772 m_freem(slp->ns_rec); 773 for (nuidp = slp->ns_uidlruhead.tqh_first; nuidp != 0; 774 nuidp = nnuidp) { 775 nnuidp = nuidp->nu_lru.tqe_next; 776 LIST_REMOVE(nuidp, nu_hash); 777 TAILQ_REMOVE(&slp->ns_uidlruhead, nuidp, nu_lru); 778 if (nuidp->nu_flag & NU_NAM) 779 m_freem(nuidp->nu_nam); 780 free((caddr_t)nuidp, M_NFSUID); 781 } 782 s = splsoftclock(); 783 for (nwp = slp->ns_tq.lh_first; nwp; nwp = nnwp) { 784 nnwp = nwp->nd_tq.le_next; 785 LIST_REMOVE(nwp, nd_tq); 786 free((caddr_t)nwp, M_NFSRVDESC); 787 } 788 LIST_INIT(&slp->ns_tq); 789 splx(s); 790 } 791 } 792 793 /* 794 * Derefence a server socket structure. If it has no more references and 795 * is no longer valid, you can throw it away. 796 */ 797 void 798 nfsrv_slpderef(slp) 799 register struct nfssvc_sock *slp; 800 { 801 if (--(slp->ns_sref) == 0 && (slp->ns_flag & SLP_VALID) == 0) { 802 TAILQ_REMOVE(&nfssvc_sockhead, slp, ns_chain); 803 free((caddr_t)slp, M_NFSSVC); 804 } 805 } 806 807 /* 808 * Initialize the data structures for the server. 809 * Handshake with any new nfsds starting up to avoid any chance of 810 * corruption. 811 */ 812 void 813 nfsrv_init(terminating) 814 int terminating; 815 { 816 register struct nfssvc_sock *slp, *nslp; 817 818 if (nfssvc_sockhead_flag & SLP_INIT) 819 panic("nfsd init"); 820 nfssvc_sockhead_flag |= SLP_INIT; 821 if (terminating) { 822 for (slp = nfssvc_sockhead.tqh_first; slp != 0; slp = nslp) { 823 nslp = slp->ns_chain.tqe_next; 824 if (slp->ns_flag & SLP_VALID) 825 nfsrv_zapsock(slp); 826 TAILQ_REMOVE(&nfssvc_sockhead, slp, ns_chain); 827 free((caddr_t)slp, M_NFSSVC); 828 } 829 nfsrv_cleancache(); /* And clear out server cache */ 830 } else 831 nfs_pub.np_valid = 0; 832 833 TAILQ_INIT(&nfssvc_sockhead); 834 nfssvc_sockhead_flag &= ~SLP_INIT; 835 if (nfssvc_sockhead_flag & SLP_WANTINIT) { 836 nfssvc_sockhead_flag &= ~SLP_WANTINIT; 837 wakeup((caddr_t)&nfssvc_sockhead); 838 } 839 840 TAILQ_INIT(&nfsd_head); 841 nfsd_head_flag &= ~NFSD_CHECKSLP; 842 843 nfs_udpsock = (struct nfssvc_sock *) 844 malloc(sizeof (struct nfssvc_sock), M_NFSSVC, M_WAITOK); 845 memset((caddr_t)nfs_udpsock, 0, sizeof (struct nfssvc_sock)); 846 TAILQ_INIT(&nfs_udpsock->ns_uidlruhead); 847 TAILQ_INSERT_HEAD(&nfssvc_sockhead, nfs_udpsock, ns_chain); 848 849 nfs_cltpsock = (struct nfssvc_sock *) 850 malloc(sizeof (struct nfssvc_sock), M_NFSSVC, M_WAITOK); 851 memset((caddr_t)nfs_cltpsock, 0, sizeof (struct nfssvc_sock)); 852 TAILQ_INIT(&nfs_cltpsock->ns_uidlruhead); 853 TAILQ_INSERT_TAIL(&nfssvc_sockhead, nfs_cltpsock, ns_chain); 854 } 855 856 /* 857 * Add entries to the server monitor log. 858 */ 859 static void 860 nfsd_rt(sotype, nd, cacherep) 861 int sotype; 862 register struct nfsrv_descript *nd; 863 int cacherep; 864 { 865 register struct drt *rt; 866 867 rt = &nfsdrt.drt[nfsdrt.pos]; 868 if (cacherep == RC_DOIT) 869 rt->flag = 0; 870 else if (cacherep == RC_REPLY) 871 rt->flag = DRT_CACHEREPLY; 872 else 873 rt->flag = DRT_CACHEDROP; 874 if (sotype == SOCK_STREAM) 875 rt->flag |= DRT_TCP; 876 if (nd->nd_flag & ND_NQNFS) 877 rt->flag |= DRT_NQNFS; 878 else if (nd->nd_flag & ND_NFSV3) 879 rt->flag |= DRT_NFSV3; 880 rt->proc = nd->nd_procnum; 881 if (mtod(nd->nd_nam, struct sockaddr *)->sa_family == AF_INET) 882 rt->ipadr = mtod(nd->nd_nam, struct sockaddr_in *)->sin_addr.s_addr; 883 else 884 rt->ipadr = INADDR_ANY; 885 rt->resptime = ((time.tv_sec - nd->nd_starttime.tv_sec) * 1000000) + 886 (time.tv_usec - nd->nd_starttime.tv_usec); 887 rt->tstamp = time; 888 nfsdrt.pos = (nfsdrt.pos + 1) % NFSRTTLOGSIZ; 889 } 890 #endif /* NFSSERVER */ 891 892 #ifdef NFS 893 894 int nfs_defect = 0; 895 /* 896 * Asynchronous I/O daemons for client nfs. 897 * They do read-ahead and write-behind operations on the block I/O cache. 898 * Never returns unless it fails or gets killed. 899 */ 900 int 901 nfssvc_iod(p) 902 struct proc *p; 903 { 904 register struct buf *bp; 905 register int i, myiod; 906 struct nfsmount *nmp; 907 int error = 0; 908 909 /* 910 * Assign my position or return error if too many already running 911 */ 912 myiod = -1; 913 for (i = 0; i < NFS_MAXASYNCDAEMON; i++) 914 if (nfs_asyncdaemon[i] == 0) { 915 myiod = i; 916 break; 917 } 918 if (myiod == -1) 919 return (EBUSY); 920 nfs_asyncdaemon[myiod] = 1; 921 nfs_numasync++; 922 p->p_holdcnt++; 923 /* 924 * Just loop around doin our stuff until SIGKILL 925 */ 926 for (;;) { 927 while (((nmp = nfs_iodmount[myiod]) == NULL 928 || nmp->nm_bufq.tqh_first == NULL) 929 && error == 0) { 930 if (nmp) 931 nmp->nm_bufqiods--; 932 nfs_iodwant[myiod] = p; 933 nfs_iodmount[myiod] = NULL; 934 error = tsleep((caddr_t)&nfs_iodwant[myiod], 935 PWAIT | PCATCH, "nfsidl", 0); 936 } 937 if (error) { 938 if (nmp) 939 nmp->nm_bufqiods--; 940 nfs_iodmount[myiod] = NULL; 941 break; 942 } 943 while ((bp = nmp->nm_bufq.tqh_first) != NULL) { 944 /* Take one off the front of the list */ 945 TAILQ_REMOVE(&nmp->nm_bufq, bp, b_freelist); 946 nmp->nm_bufqlen--; 947 if (nmp->nm_bufqwant && nmp->nm_bufqlen < 2 * nfs_numasync) { 948 nmp->nm_bufqwant = FALSE; 949 wakeup(&nmp->nm_bufq); 950 } 951 if (bp->b_flags & B_READ) 952 (void) nfs_doio(bp, bp->b_rcred, (struct proc *)0); 953 else 954 (void) nfs_doio(bp, bp->b_wcred, (struct proc *)0); 955 /* 956 * If there are more than one iod on this mount, then defect 957 * so that the iods can be shared out fairly between the mounts 958 */ 959 if (nfs_defect && nmp->nm_bufqiods > 1) { 960 nfs_iodmount[myiod] = NULL; 961 nmp->nm_bufqiods--; 962 break; 963 } 964 } 965 } 966 p->p_holdcnt--; 967 nfs_asyncdaemon[myiod] = 0; 968 nfs_numasync--; 969 return (error); 970 } 971 972 973 /* 974 * Get an authorization string for the uid by having the mount_nfs sitting 975 * on this mount point porpous out of the kernel and do it. 976 */ 977 int 978 nfs_getauth(nmp, rep, cred, auth_str, auth_len, verf_str, verf_len, key) 979 register struct nfsmount *nmp; 980 struct nfsreq *rep; 981 struct ucred *cred; 982 char **auth_str; 983 int *auth_len; 984 char *verf_str; 985 int *verf_len; 986 NFSKERBKEY_T key; /* return session key */ 987 { 988 int error = 0; 989 990 while ((nmp->nm_iflag & NFSMNT_WAITAUTH) == 0) { 991 nmp->nm_iflag |= NFSMNT_WANTAUTH; 992 (void) tsleep((caddr_t)&nmp->nm_authtype, PSOCK, 993 "nfsauth1", 2 * hz); 994 error = nfs_sigintr(nmp, rep, rep->r_procp); 995 if (error) { 996 nmp->nm_iflag &= ~NFSMNT_WANTAUTH; 997 return (error); 998 } 999 } 1000 nmp->nm_iflag &= ~(NFSMNT_WAITAUTH | NFSMNT_WANTAUTH); 1001 nmp->nm_authstr = *auth_str = (char *)malloc(RPCAUTH_MAXSIZ, M_TEMP, M_WAITOK); 1002 nmp->nm_authlen = RPCAUTH_MAXSIZ; 1003 nmp->nm_verfstr = verf_str; 1004 nmp->nm_verflen = *verf_len; 1005 nmp->nm_authuid = cred->cr_uid; 1006 wakeup((caddr_t)&nmp->nm_authstr); 1007 1008 /* 1009 * And wait for mount_nfs to do its stuff. 1010 */ 1011 while ((nmp->nm_iflag & NFSMNT_HASAUTH) == 0 && error == 0) { 1012 (void) tsleep((caddr_t)&nmp->nm_authlen, PSOCK, 1013 "nfsauth2", 2 * hz); 1014 error = nfs_sigintr(nmp, rep, rep->r_procp); 1015 } 1016 if (nmp->nm_iflag & NFSMNT_AUTHERR) { 1017 nmp->nm_iflag &= ~NFSMNT_AUTHERR; 1018 error = EAUTH; 1019 } 1020 if (error) 1021 free((caddr_t)*auth_str, M_TEMP); 1022 else { 1023 *auth_len = nmp->nm_authlen; 1024 *verf_len = nmp->nm_verflen; 1025 memcpy((caddr_t)key, (caddr_t)nmp->nm_key, sizeof (key)); 1026 } 1027 nmp->nm_iflag &= ~NFSMNT_HASAUTH; 1028 nmp->nm_iflag |= NFSMNT_WAITAUTH; 1029 if (nmp->nm_iflag & NFSMNT_WANTAUTH) { 1030 nmp->nm_iflag &= ~NFSMNT_WANTAUTH; 1031 wakeup((caddr_t)&nmp->nm_authtype); 1032 } 1033 return (error); 1034 } 1035 1036 /* 1037 * Get a nickname authenticator and verifier. 1038 */ 1039 int 1040 nfs_getnickauth(nmp, cred, auth_str, auth_len, verf_str, verf_len) 1041 struct nfsmount *nmp; 1042 struct ucred *cred; 1043 char **auth_str; 1044 int *auth_len; 1045 char *verf_str; 1046 int verf_len; 1047 { 1048 register struct nfsuid *nuidp; 1049 register u_int32_t *nickp, *verfp; 1050 struct timeval ktvin, ktvout; 1051 1052 #ifdef DIAGNOSTIC 1053 if (verf_len < (4 * NFSX_UNSIGNED)) 1054 panic("nfs_getnickauth verf too small"); 1055 #endif 1056 for (nuidp = NMUIDHASH(nmp, cred->cr_uid)->lh_first; 1057 nuidp != 0; nuidp = nuidp->nu_hash.le_next) { 1058 if (nuidp->nu_cr.cr_uid == cred->cr_uid) 1059 break; 1060 } 1061 if (!nuidp || nuidp->nu_expire < time.tv_sec) 1062 return (EACCES); 1063 1064 /* 1065 * Move to the end of the lru list (end of lru == most recently used). 1066 */ 1067 TAILQ_REMOVE(&nmp->nm_uidlruhead, nuidp, nu_lru); 1068 TAILQ_INSERT_TAIL(&nmp->nm_uidlruhead, nuidp, nu_lru); 1069 1070 nickp = (u_int32_t *)malloc(2 * NFSX_UNSIGNED, M_TEMP, M_WAITOK); 1071 *nickp++ = txdr_unsigned(RPCAKN_NICKNAME); 1072 *nickp = txdr_unsigned(nuidp->nu_nickname); 1073 *auth_str = (char *)nickp; 1074 *auth_len = 2 * NFSX_UNSIGNED; 1075 1076 /* 1077 * Now we must encrypt the verifier and package it up. 1078 */ 1079 verfp = (u_int32_t *)verf_str; 1080 *verfp++ = txdr_unsigned(RPCAKN_NICKNAME); 1081 if (time.tv_sec > nuidp->nu_timestamp.tv_sec || 1082 (time.tv_sec == nuidp->nu_timestamp.tv_sec && 1083 time.tv_usec > nuidp->nu_timestamp.tv_usec)) 1084 nuidp->nu_timestamp = time; 1085 else 1086 nuidp->nu_timestamp.tv_usec++; 1087 ktvin.tv_sec = txdr_unsigned(nuidp->nu_timestamp.tv_sec); 1088 ktvin.tv_usec = txdr_unsigned(nuidp->nu_timestamp.tv_usec); 1089 1090 /* 1091 * Now encrypt the timestamp verifier in ecb mode using the session 1092 * key. 1093 */ 1094 #ifdef NFSKERB 1095 XXX 1096 #endif 1097 1098 *verfp++ = ktvout.tv_sec; 1099 *verfp++ = ktvout.tv_usec; 1100 *verfp = 0; 1101 return (0); 1102 } 1103 1104 /* 1105 * Save the current nickname in a hash list entry on the mount point. 1106 */ 1107 int 1108 nfs_savenickauth(nmp, cred, len, key, mdp, dposp, mrep) 1109 register struct nfsmount *nmp; 1110 struct ucred *cred; 1111 int len; 1112 NFSKERBKEY_T key; 1113 struct mbuf **mdp; 1114 char **dposp; 1115 struct mbuf *mrep; 1116 { 1117 register struct nfsuid *nuidp; 1118 register u_int32_t *tl; 1119 register int32_t t1; 1120 struct mbuf *md = *mdp; 1121 struct timeval ktvin, ktvout; 1122 u_int32_t nick; 1123 char *dpos = *dposp, *cp2; 1124 int deltasec, error = 0; 1125 1126 if (len == (3 * NFSX_UNSIGNED)) { 1127 nfsm_dissect(tl, u_int32_t *, 3 * NFSX_UNSIGNED); 1128 ktvin.tv_sec = *tl++; 1129 ktvin.tv_usec = *tl++; 1130 nick = fxdr_unsigned(u_int32_t, *tl); 1131 1132 /* 1133 * Decrypt the timestamp in ecb mode. 1134 */ 1135 #ifdef NFSKERB 1136 XXX 1137 #endif 1138 ktvout.tv_sec = fxdr_unsigned(long, ktvout.tv_sec); 1139 ktvout.tv_usec = fxdr_unsigned(long, ktvout.tv_usec); 1140 deltasec = time.tv_sec - ktvout.tv_sec; 1141 if (deltasec < 0) 1142 deltasec = -deltasec; 1143 /* 1144 * If ok, add it to the hash list for the mount point. 1145 */ 1146 if (deltasec <= NFS_KERBCLOCKSKEW) { 1147 if (nmp->nm_numuids < nuidhash_max) { 1148 nmp->nm_numuids++; 1149 nuidp = (struct nfsuid *) 1150 malloc(sizeof (struct nfsuid), M_NFSUID, 1151 M_WAITOK); 1152 } else { 1153 nuidp = nmp->nm_uidlruhead.tqh_first; 1154 LIST_REMOVE(nuidp, nu_hash); 1155 TAILQ_REMOVE(&nmp->nm_uidlruhead, nuidp, 1156 nu_lru); 1157 } 1158 nuidp->nu_flag = 0; 1159 nuidp->nu_cr.cr_uid = cred->cr_uid; 1160 nuidp->nu_expire = time.tv_sec + NFS_KERBTTL; 1161 nuidp->nu_timestamp = ktvout; 1162 nuidp->nu_nickname = nick; 1163 memcpy(nuidp->nu_key, key, sizeof (key)); 1164 TAILQ_INSERT_TAIL(&nmp->nm_uidlruhead, nuidp, 1165 nu_lru); 1166 LIST_INSERT_HEAD(NMUIDHASH(nmp, cred->cr_uid), 1167 nuidp, nu_hash); 1168 } 1169 } else 1170 nfsm_adv(nfsm_rndup(len)); 1171 nfsmout: 1172 *mdp = md; 1173 *dposp = dpos; 1174 return (error); 1175 } 1176 #endif /* NFS */ 1177