nfs_nfsiod.c revision 25307
1/* 2 * Copyright (c) 1989, 1993 3 * The Regents of the University of California. All rights reserved. 4 * 5 * This code is derived from software contributed to Berkeley by 6 * Rick Macklem at The University of Guelph. 7 * 8 * Redistribution and use in source and binary forms, with or without 9 * modification, are permitted provided that the following conditions 10 * are met: 11 * 1. Redistributions of source code must retain the above copyright 12 * notice, this list of conditions and the following disclaimer. 13 * 2. Redistributions in binary form must reproduce the above copyright 14 * notice, this list of conditions and the following disclaimer in the 15 * documentation and/or other materials provided with the distribution. 16 * 3. All advertising materials mentioning features or use of this software 17 * must display the following acknowledgement: 18 * This product includes software developed by the University of 19 * California, Berkeley and its contributors. 20 * 4. Neither the name of the University nor the names of its contributors 21 * may be used to endorse or promote products derived from this software 22 * without specific prior written permission. 23 * 24 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND 25 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE 26 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE 27 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE 28 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL 29 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS 30 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) 31 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT 32 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY 33 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF 34 * SUCH DAMAGE. 35 * 36 * @(#)nfs_syscalls.c 8.5 (Berkeley) 3/30/95 37 * $Id: nfs_syscalls.c,v 1.21 1997/04/27 20:01:23 wollman Exp $ 38 */ 39 40#include <sys/param.h> 41#include <sys/systm.h> 42#include <sys/sysproto.h> 43#include <sys/kernel.h> 44#include <sys/sysctl.h> 45#include <sys/file.h> 46#include <sys/filedesc.h> 47#include <sys/stat.h> 48#include <sys/vnode.h> 49#include <sys/mount.h> 50#include <sys/proc.h> 51#include <sys/uio.h> 52#include <sys/malloc.h> 53#include <sys/buf.h> 54#include <sys/mbuf.h> 55#include <sys/socket.h> 56#include <sys/socketvar.h> 57#include <sys/domain.h> 58#include <sys/protosw.h> 59#include <sys/namei.h> 60#include <sys/syslog.h> 61 62#include <netinet/in.h> 63#include <netinet/tcp.h> 64#ifdef ISO 65#include <netiso/iso.h> 66#endif 67#include <nfs/xdr_subs.h> 68#include <nfs/rpcv2.h> 69#include <nfs/nfsproto.h> 70#include <nfs/nfs.h> 71#include <nfs/nfsm_subs.h> 72#include <nfs/nfsrvcache.h> 73#include <nfs/nfsmount.h> 74#include <nfs/nfsnode.h> 75#include <nfs/nqnfs.h> 76#include <nfs/nfsrtt.h> 77 78/* Global defs. */ 79extern int (*nfsrv3_procs[NFS_NPROCS]) __P((struct nfsrv_descript *nd, 80 struct nfssvc_sock *slp, 81 struct proc *procp, 82 struct mbuf **mreqp)); 83extern int nfs_numasync; 84extern time_t nqnfsstarttime; 85extern int nqsrv_writeslack; 86extern int nfsrtton; 87extern struct nfsstats nfsstats; 88extern int nfsrvw_procrastinate; 89struct nfssvc_sock *nfs_udpsock, *nfs_cltpsock; 90static int nuidhash_max = NFS_MAXUIDHASH; 91 92static void nfsrv_zapsock __P((struct nfssvc_sock *slp)); 93static int nfssvc_iod __P((struct proc *)); 94 95#define TRUE 1 96#define FALSE 0 97 98static int nfs_asyncdaemon[NFS_MAXASYNCDAEMON]; 99 100#ifndef NFS_NOSERVER 101int nfsd_waiting = 0; 102static struct nfsdrt nfsdrt; 103static int nfs_numnfsd = 0; 104static int notstarted = 1; 105static int modify_flag = 0; 106static void nfsd_rt __P((int sotype, struct nfsrv_descript *nd, 107 int cacherep)); 108static int nfssvc_addsock __P((struct file *, struct mbuf *, 109 struct proc *)); 110static int nfssvc_nfsd __P((struct nfsd_srvargs *,caddr_t,struct proc *)); 111 112static int nfs_privport = 0; 113SYSCTL_INT(_vfs_nfs, NFS_NFSPRIVPORT, nfs_privport, CTLFLAG_RW, &nfs_privport, 0, ""); 114 115/* 116 * NFS server system calls 117 * getfh() lives here too, but maybe should move to kern/vfs_syscalls.c 118 */ 119 120/* 121 * Get file handle system call 122 */ 123#ifndef _SYS_SYSPROTO_H_ 124struct getfh_args { 125 char *fname; 126 fhandle_t *fhp; 127}; 128#endif 129int 130getfh(p, uap, retval) 131 struct proc *p; 132 register struct getfh_args *uap; 133 int *retval; 134{ 135 register struct vnode *vp; 136 fhandle_t fh; 137 int error; 138 struct nameidata nd; 139 140 /* 141 * Must be super user 142 */ 143 error = suser(p->p_ucred, &p->p_acflag); 144 if(error) 145 return (error); 146 NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF, UIO_USERSPACE, uap->fname, p); 147 error = namei(&nd); 148 if (error) 149 return (error); 150 vp = nd.ni_vp; 151 bzero((caddr_t)&fh, sizeof(fh)); 152 fh.fh_fsid = vp->v_mount->mnt_stat.f_fsid; 153 error = VFS_VPTOFH(vp, &fh.fh_fid); 154 vput(vp); 155 if (error) 156 return (error); 157 error = copyout((caddr_t)&fh, (caddr_t)uap->fhp, sizeof (fh)); 158 return (error); 159} 160 161#endif /* NFS_NOSERVER */ 162/* 163 * Nfs server psuedo system call for the nfsd's 164 * Based on the flag value it either: 165 * - adds a socket to the selection list 166 * - remains in the kernel as an nfsd 167 * - remains in the kernel as an nfsiod 168 */ 169#ifndef _SYS_SYSPROTO_H_ 170struct nfssvc_args { 171 int flag; 172 caddr_t argp; 173}; 174#endif 175int 176nfssvc(p, uap, retval) 177 struct proc *p; 178 register struct nfssvc_args *uap; 179 int *retval; 180{ 181#ifndef NFS_NOSERVER 182 struct nameidata nd; 183 struct file *fp; 184 struct mbuf *nam; 185 struct nfsd_args nfsdarg; 186 struct nfsd_srvargs nfsd_srvargs, *nsd = &nfsd_srvargs; 187 struct nfsd_cargs ncd; 188 struct nfsd *nfsd; 189 struct nfssvc_sock *slp; 190 struct nfsuid *nuidp; 191 struct nfsmount *nmp; 192#endif /* NFS_NOSERVER */ 193 int error; 194 195 /* 196 * Must be super user 197 */ 198 error = suser(p->p_ucred, &p->p_acflag); 199 if(error) 200 return (error); 201 while (nfssvc_sockhead_flag & SLP_INIT) { 202 nfssvc_sockhead_flag |= SLP_WANTINIT; 203 (void) tsleep((caddr_t)&nfssvc_sockhead, PSOCK, "nfsd init", 0); 204 } 205 if (uap->flag & NFSSVC_BIOD) 206 error = nfssvc_iod(p); 207#ifdef NFS_NOSERVER 208 else 209 error = ENXIO; 210#else /* !NFS_NOSERVER */ 211 else if (uap->flag & NFSSVC_MNTD) { 212 error = copyin(uap->argp, (caddr_t)&ncd, sizeof (ncd)); 213 if (error) 214 return (error); 215 NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF, UIO_USERSPACE, 216 ncd.ncd_dirp, p); 217 error = namei(&nd); 218 if (error) 219 return (error); 220 if ((nd.ni_vp->v_flag & VROOT) == 0) 221 error = EINVAL; 222 nmp = VFSTONFS(nd.ni_vp->v_mount); 223 vput(nd.ni_vp); 224 if (error) 225 return (error); 226 if ((nmp->nm_flag & NFSMNT_MNTD) && 227 (uap->flag & NFSSVC_GOTAUTH) == 0) 228 return (0); 229 nmp->nm_flag |= NFSMNT_MNTD; 230 error = nqnfs_clientd(nmp, p->p_ucred, &ncd, uap->flag, 231 uap->argp, p); 232 } else if (uap->flag & NFSSVC_ADDSOCK) { 233 error = copyin(uap->argp, (caddr_t)&nfsdarg, sizeof(nfsdarg)); 234 if (error) 235 return (error); 236 error = getsock(p->p_fd, nfsdarg.sock, &fp); 237 if (error) 238 return (error); 239 /* 240 * Get the client address for connected sockets. 241 */ 242 if (nfsdarg.name == NULL || nfsdarg.namelen == 0) 243 nam = (struct mbuf *)0; 244 else { 245 error = sockargs(&nam, nfsdarg.name, nfsdarg.namelen, 246 MT_SONAME); 247 if (error) 248 return (error); 249 } 250 error = nfssvc_addsock(fp, nam, p); 251 } else { 252 error = copyin(uap->argp, (caddr_t)nsd, sizeof (*nsd)); 253 if (error) 254 return (error); 255 if ((uap->flag & NFSSVC_AUTHIN) && ((nfsd = nsd->nsd_nfsd)) && 256 (nfsd->nfsd_slp->ns_flag & SLP_VALID)) { 257 slp = nfsd->nfsd_slp; 258 259 /* 260 * First check to see if another nfsd has already 261 * added this credential. 262 */ 263 for (nuidp = NUIDHASH(slp,nsd->nsd_cr.cr_uid)->lh_first; 264 nuidp != 0; nuidp = nuidp->nu_hash.le_next) { 265 if (nuidp->nu_cr.cr_uid == nsd->nsd_cr.cr_uid && 266 (!nfsd->nfsd_nd->nd_nam2 || 267 netaddr_match(NU_NETFAM(nuidp), 268 &nuidp->nu_haddr, nfsd->nfsd_nd->nd_nam2))) 269 break; 270 } 271 if (nuidp) { 272 nfsrv_setcred(&nuidp->nu_cr,&nfsd->nfsd_nd->nd_cr); 273 nfsd->nfsd_nd->nd_flag |= ND_KERBFULL; 274 } else { 275 /* 276 * Nope, so we will. 277 */ 278 if (slp->ns_numuids < nuidhash_max) { 279 slp->ns_numuids++; 280 nuidp = (struct nfsuid *) 281 malloc(sizeof (struct nfsuid), M_NFSUID, 282 M_WAITOK); 283 } else 284 nuidp = (struct nfsuid *)0; 285 if ((slp->ns_flag & SLP_VALID) == 0) { 286 if (nuidp) 287 free((caddr_t)nuidp, M_NFSUID); 288 } else { 289 if (nuidp == (struct nfsuid *)0) { 290 nuidp = slp->ns_uidlruhead.tqh_first; 291 LIST_REMOVE(nuidp, nu_hash); 292 TAILQ_REMOVE(&slp->ns_uidlruhead, nuidp, 293 nu_lru); 294 if (nuidp->nu_flag & NU_NAM) 295 m_freem(nuidp->nu_nam); 296 } 297 nuidp->nu_flag = 0; 298 nuidp->nu_cr = nsd->nsd_cr; 299 if (nuidp->nu_cr.cr_ngroups > NGROUPS) 300 nuidp->nu_cr.cr_ngroups = NGROUPS; 301 nuidp->nu_cr.cr_ref = 1; 302 nuidp->nu_timestamp = nsd->nsd_timestamp; 303 nuidp->nu_expire = time.tv_sec + nsd->nsd_ttl; 304 /* 305 * and save the session key in nu_key. 306 */ 307 bcopy(nsd->nsd_key, nuidp->nu_key, 308 sizeof (nsd->nsd_key)); 309 if (nfsd->nfsd_nd->nd_nam2) { 310 struct sockaddr_in *saddr; 311 312 saddr = mtod(nfsd->nfsd_nd->nd_nam2, 313 struct sockaddr_in *); 314 switch (saddr->sin_family) { 315 case AF_INET: 316 nuidp->nu_flag |= NU_INETADDR; 317 nuidp->nu_inetaddr = 318 saddr->sin_addr.s_addr; 319 break; 320 case AF_ISO: 321 default: 322 nuidp->nu_flag |= NU_NAM; 323 nuidp->nu_nam = m_copym( 324 nfsd->nfsd_nd->nd_nam2, 0, 325 M_COPYALL, M_WAIT); 326 break; 327 }; 328 } 329 TAILQ_INSERT_TAIL(&slp->ns_uidlruhead, nuidp, 330 nu_lru); 331 LIST_INSERT_HEAD(NUIDHASH(slp, nsd->nsd_uid), 332 nuidp, nu_hash); 333 nfsrv_setcred(&nuidp->nu_cr, 334 &nfsd->nfsd_nd->nd_cr); 335 nfsd->nfsd_nd->nd_flag |= ND_KERBFULL; 336 } 337 } 338 } 339 if ((uap->flag & NFSSVC_AUTHINFAIL) && (nfsd = nsd->nsd_nfsd)) 340 nfsd->nfsd_flag |= NFSD_AUTHFAIL; 341 error = nfssvc_nfsd(nsd, uap->argp, p); 342 } 343#endif /* NFS_NOSERVER */ 344 if (error == EINTR || error == ERESTART) 345 error = 0; 346 return (error); 347} 348 349#ifndef NFS_NOSERVER 350/* 351 * Adds a socket to the list for servicing by nfsds. 352 */ 353static int 354nfssvc_addsock(fp, mynam, p) 355 struct file *fp; 356 struct mbuf *mynam; 357 struct proc *p; 358{ 359 register struct mbuf *m; 360 register int siz; 361 register struct nfssvc_sock *slp; 362 register struct socket *so; 363 struct nfssvc_sock *tslp; 364 int error, s; 365 366 so = (struct socket *)fp->f_data; 367 tslp = (struct nfssvc_sock *)0; 368 /* 369 * Add it to the list, as required. 370 */ 371 if (so->so_proto->pr_protocol == IPPROTO_UDP) { 372 tslp = nfs_udpsock; 373 if (tslp->ns_flag & SLP_VALID) { 374 m_freem(mynam); 375 return (EPERM); 376 } 377#ifdef ISO 378 } else if (so->so_proto->pr_protocol == ISOPROTO_CLTP) { 379 tslp = nfs_cltpsock; 380 if (tslp->ns_flag & SLP_VALID) { 381 m_freem(mynam); 382 return (EPERM); 383 } 384#endif /* ISO */ 385 } 386 if (so->so_type == SOCK_STREAM) 387 siz = NFS_MAXPACKET + sizeof (u_long); 388 else 389 siz = NFS_MAXPACKET; 390 error = soreserve(so, siz, siz); 391 if (error) { 392 m_freem(mynam); 393 return (error); 394 } 395 396 /* 397 * Set protocol specific options { for now TCP only } and 398 * reserve some space. For datagram sockets, this can get called 399 * repeatedly for the same socket, but that isn't harmful. 400 */ 401 if (so->so_type == SOCK_STREAM) { 402 MGET(m, M_WAIT, MT_SOOPTS); 403 *mtod(m, int *) = 1; 404 m->m_len = sizeof(int); 405 sosetopt(so, SOL_SOCKET, SO_KEEPALIVE, m, p); 406 } 407 if (so->so_proto->pr_domain->dom_family == AF_INET && 408 so->so_proto->pr_protocol == IPPROTO_TCP) { 409 MGET(m, M_WAIT, MT_SOOPTS); 410 *mtod(m, int *) = 1; 411 m->m_len = sizeof(int); 412 sosetopt(so, IPPROTO_TCP, TCP_NODELAY, m, p); 413 } 414 so->so_rcv.sb_flags &= ~SB_NOINTR; 415 so->so_rcv.sb_timeo = 0; 416 so->so_snd.sb_flags &= ~SB_NOINTR; 417 so->so_snd.sb_timeo = 0; 418 if (tslp) 419 slp = tslp; 420 else { 421 slp = (struct nfssvc_sock *) 422 malloc(sizeof (struct nfssvc_sock), M_NFSSVC, M_WAITOK); 423 bzero((caddr_t)slp, sizeof (struct nfssvc_sock)); 424 TAILQ_INIT(&slp->ns_uidlruhead); 425 TAILQ_INSERT_TAIL(&nfssvc_sockhead, slp, ns_chain); 426 } 427 slp->ns_so = so; 428 slp->ns_nam = mynam; 429 fp->f_count++; 430 slp->ns_fp = fp; 431 s = splnet(); 432 so->so_upcallarg = (caddr_t)slp; 433 so->so_upcall = nfsrv_rcv; 434 slp->ns_flag = (SLP_VALID | SLP_NEEDQ); 435 nfsrv_wakenfsd(slp); 436 splx(s); 437 return (0); 438} 439 440/* 441 * Called by nfssvc() for nfsds. Just loops around servicing rpc requests 442 * until it is killed by a signal. 443 */ 444static int 445nfssvc_nfsd(nsd, argp, p) 446 struct nfsd_srvargs *nsd; 447 caddr_t argp; 448 struct proc *p; 449{ 450 register struct mbuf *m; 451 register int siz; 452 register struct nfssvc_sock *slp; 453 register struct socket *so; 454 register int *solockp; 455 struct nfsd *nfsd = nsd->nsd_nfsd; 456 struct nfsrv_descript *nd = NULL; 457 struct mbuf *mreq; 458 int error = 0, cacherep, s, sotype, writes_todo; 459 u_quad_t cur_usec; 460 461#ifndef nolint 462 cacherep = RC_DOIT; 463 writes_todo = 0; 464#endif 465 s = splnet(); 466 if (nfsd == (struct nfsd *)0) { 467 nsd->nsd_nfsd = nfsd = (struct nfsd *) 468 malloc(sizeof (struct nfsd), M_NFSD, M_WAITOK); 469 bzero((caddr_t)nfsd, sizeof (struct nfsd)); 470 nfsd->nfsd_procp = p; 471 TAILQ_INSERT_TAIL(&nfsd_head, nfsd, nfsd_chain); 472 nfs_numnfsd++; 473 } 474 /* 475 * Loop getting rpc requests until SIGKILL. 476 */ 477 for (;;) { 478 if ((nfsd->nfsd_flag & NFSD_REQINPROG) == 0) { 479 while (nfsd->nfsd_slp == (struct nfssvc_sock *)0 && 480 (nfsd_head_flag & NFSD_CHECKSLP) == 0) { 481 nfsd->nfsd_flag |= NFSD_WAITING; 482 nfsd_waiting++; 483 error = tsleep((caddr_t)nfsd, PSOCK | PCATCH, 484 "nfsd", 0); 485 nfsd_waiting--; 486 if (error) 487 goto done; 488 } 489 if (nfsd->nfsd_slp == (struct nfssvc_sock *)0 && 490 (nfsd_head_flag & NFSD_CHECKSLP) != 0) { 491 for (slp = nfssvc_sockhead.tqh_first; slp != 0; 492 slp = slp->ns_chain.tqe_next) { 493 if ((slp->ns_flag & (SLP_VALID | SLP_DOREC)) 494 == (SLP_VALID | SLP_DOREC)) { 495 slp->ns_flag &= ~SLP_DOREC; 496 slp->ns_sref++; 497 nfsd->nfsd_slp = slp; 498 break; 499 } 500 } 501 if (slp == 0) 502 nfsd_head_flag &= ~NFSD_CHECKSLP; 503 } 504 if ((slp = nfsd->nfsd_slp) == (struct nfssvc_sock *)0) 505 continue; 506 if (slp->ns_flag & SLP_VALID) { 507 if (slp->ns_flag & SLP_DISCONN) 508 nfsrv_zapsock(slp); 509 else if (slp->ns_flag & SLP_NEEDQ) { 510 slp->ns_flag &= ~SLP_NEEDQ; 511 (void) nfs_sndlock(&slp->ns_solock, 512 (struct nfsreq *)0); 513 nfsrv_rcv(slp->ns_so, (caddr_t)slp, 514 M_WAIT); 515 nfs_sndunlock(&slp->ns_solock); 516 } 517 error = nfsrv_dorec(slp, nfsd, &nd); 518 cur_usec = (u_quad_t)time.tv_sec * 1000000 + 519 (u_quad_t)time.tv_usec; 520 if (error && slp->ns_tq.lh_first && 521 slp->ns_tq.lh_first->nd_time <= cur_usec) { 522 error = 0; 523 cacherep = RC_DOIT; 524 writes_todo = 1; 525 } else 526 writes_todo = 0; 527 nfsd->nfsd_flag |= NFSD_REQINPROG; 528 } 529 } else { 530 error = 0; 531 slp = nfsd->nfsd_slp; 532 } 533 if (error || (slp->ns_flag & SLP_VALID) == 0) { 534 if (nd) { 535 free((caddr_t)nd, M_NFSRVDESC); 536 nd = NULL; 537 } 538 nfsd->nfsd_slp = (struct nfssvc_sock *)0; 539 nfsd->nfsd_flag &= ~NFSD_REQINPROG; 540 nfsrv_slpderef(slp); 541 continue; 542 } 543 splx(s); 544 so = slp->ns_so; 545 sotype = so->so_type; 546 if (so->so_proto->pr_flags & PR_CONNREQUIRED) 547 solockp = &slp->ns_solock; 548 else 549 solockp = (int *)0; 550 if (nd) { 551 gettime(&nd->nd_starttime); 552 if (nd->nd_nam2) 553 nd->nd_nam = nd->nd_nam2; 554 else 555 nd->nd_nam = slp->ns_nam; 556 557 /* 558 * Check to see if authorization is needed. 559 */ 560 if (nfsd->nfsd_flag & NFSD_NEEDAUTH) { 561 nfsd->nfsd_flag &= ~NFSD_NEEDAUTH; 562 nsd->nsd_haddr = mtod(nd->nd_nam, 563 struct sockaddr_in *)->sin_addr.s_addr; 564 nsd->nsd_authlen = nfsd->nfsd_authlen; 565 nsd->nsd_verflen = nfsd->nfsd_verflen; 566 if (!copyout(nfsd->nfsd_authstr,nsd->nsd_authstr, 567 nfsd->nfsd_authlen) && 568 !copyout(nfsd->nfsd_verfstr, nsd->nsd_verfstr, 569 nfsd->nfsd_verflen) && 570 !copyout((caddr_t)nsd, argp, sizeof (*nsd))) 571 return (ENEEDAUTH); 572 cacherep = RC_DROPIT; 573 } else 574 cacherep = nfsrv_getcache(nd, slp, &mreq); 575 576 /* 577 * Check for just starting up for NQNFS and send 578 * fake "try again later" replies to the NQNFS clients. 579 */ 580 if (notstarted && nqnfsstarttime <= time.tv_sec) { 581 if (modify_flag) { 582 nqnfsstarttime = time.tv_sec + nqsrv_writeslack; 583 modify_flag = 0; 584 } else 585 notstarted = 0; 586 } 587 if (notstarted) { 588 if ((nd->nd_flag & ND_NQNFS) == 0) 589 cacherep = RC_DROPIT; 590 else if (nd->nd_procnum != NFSPROC_WRITE) { 591 nd->nd_procnum = NFSPROC_NOOP; 592 nd->nd_repstat = NQNFS_TRYLATER; 593 cacherep = RC_DOIT; 594 } else 595 modify_flag = 1; 596 } else if (nfsd->nfsd_flag & NFSD_AUTHFAIL) { 597 nfsd->nfsd_flag &= ~NFSD_AUTHFAIL; 598 nd->nd_procnum = NFSPROC_NOOP; 599 nd->nd_repstat = (NFSERR_AUTHERR | AUTH_TOOWEAK); 600 cacherep = RC_DOIT; 601 } else if (nfs_privport) { 602 /* Check if source port is privileged */ 603 u_short port; 604 u_long addr; 605 struct mbuf *nam = nd->nd_nam; 606 struct sockaddr_in *sin; 607 608 sin = mtod(nam, struct sockaddr_in *); 609 port = ntohs(sin->sin_port); 610 if (port >= IPPORT_RESERVED && 611 nd->nd_procnum != NFSPROC_NULL) { 612 nd->nd_procnum = NFSPROC_NOOP; 613 nd->nd_repstat = (NFSERR_AUTHERR | AUTH_TOOWEAK); 614 cacherep = RC_DOIT; 615 printf("NFS request from unprivileged port (%s:%d)\n", 616 inet_ntoa(sin->sin_addr), port); 617 } 618 } 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_long *) = 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 = splnet(); 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 = splnet(); 724 if (nfsrv_dorec(slp, nfsd, &nd)) { 725 nfsd->nfsd_flag &= ~NFSD_REQINPROG; 726 nfsd->nfsd_slp = NULL; 727 nfsrv_slpderef(slp); 728 } 729 } 730done: 731 TAILQ_REMOVE(&nfsd_head, nfsd, nfsd_chain); 732 splx(s); 733 free((caddr_t)nfsd, M_NFSD); 734 nsd->nsd_nfsd = (struct nfsd *)0; 735 if (--nfs_numnfsd == 0) 736 nfsrv_init(TRUE); /* Reinitialize everything */ 737 return (error); 738} 739#endif /* NFS_NOSERVER */ 740 741int nfs_defect = 0; 742SYSCTL_INT(_vfs_nfs, OID_AUTO, defect, CTLFLAG_RW, &nfs_defect, 0, ""); 743 744/* 745 * Asynchronous I/O daemons for client nfs. 746 * They do read-ahead and write-behind operations on the block I/O cache. 747 * Never returns unless it fails or gets killed. 748 */ 749static int 750nfssvc_iod(p) 751 struct proc *p; 752{ 753 register struct buf *bp, *nbp; 754 register int i, myiod; 755 struct vnode *vp; 756 struct nfsmount *nmp; 757 int error = 0, s; 758 759 /* 760 * Assign my position or return error if too many already running 761 */ 762 myiod = -1; 763 for (i = 0; i < NFS_MAXASYNCDAEMON; i++) 764 if (nfs_asyncdaemon[i] == 0) { 765 nfs_asyncdaemon[i]++; 766 myiod = i; 767 break; 768 } 769 if (myiod == -1) 770 return (EBUSY); 771 nfs_numasync++; 772 /* 773 * Just loop around doin our stuff until SIGKILL 774 */ 775 for (;;) { 776 while (((nmp = nfs_iodmount[myiod]) == NULL 777 || nmp->nm_bufq.tqh_first == NULL) 778 && error == 0) { 779 if (nmp) 780 nmp->nm_bufqiods--; 781 nfs_iodwant[myiod] = p; 782 nfs_iodmount[myiod] = NULL; 783 error = tsleep((caddr_t)&nfs_iodwant[myiod], 784 PWAIT | PCATCH, "nfsidl", 0); 785 } 786 if (error) { 787 nfs_asyncdaemon[myiod] = 0; 788 if (nmp) nmp->nm_bufqiods--; 789 nfs_iodmount[myiod] = NULL; 790 nfs_numasync--; 791 return (error); 792 } 793 while ((bp = nmp->nm_bufq.tqh_first) != NULL) { 794 /* Take one off the front of the list */ 795 TAILQ_REMOVE(&nmp->nm_bufq, bp, b_freelist); 796 nmp->nm_bufqlen--; 797 if (nmp->nm_bufqwant && nmp->nm_bufqlen < 2 * nfs_numasync) { 798 nmp->nm_bufqwant = FALSE; 799 wakeup(&nmp->nm_bufq); 800 } 801 if (bp->b_flags & B_READ) 802 (void) nfs_doio(bp, bp->b_rcred, (struct proc *)0); 803 else 804 (void) nfs_doio(bp, bp->b_wcred, (struct proc *)0); 805 806 /* 807 * If there are more than one iod on this mount, then defect 808 * so that the iods can be shared out fairly between the mounts 809 */ 810 if (nfs_defect && nmp->nm_bufqiods > 1) { 811 NFS_DPF(ASYNCIO, 812 ("nfssvc_iod: iod %d defecting from mount %p\n", 813 myiod, nmp)); 814 nfs_iodmount[myiod] = NULL; 815 nmp->nm_bufqiods--; 816 break; 817 } 818 } 819 } 820} 821 822/* 823 * Shut down a socket associated with an nfssvc_sock structure. 824 * Should be called with the send lock set, if required. 825 * The trick here is to increment the sref at the start, so that the nfsds 826 * will stop using it and clear ns_flag at the end so that it will not be 827 * reassigned during cleanup. 828 */ 829static void 830nfsrv_zapsock(slp) 831 register struct nfssvc_sock *slp; 832{ 833 register struct nfsuid *nuidp, *nnuidp; 834 register struct nfsrv_descript *nwp, *nnwp; 835 struct socket *so; 836 struct file *fp; 837 struct mbuf *m; 838 int s; 839 840 slp->ns_flag &= ~SLP_ALLFLAGS; 841 fp = slp->ns_fp; 842 if (fp) { 843 slp->ns_fp = (struct file *)0; 844 so = slp->ns_so; 845 so->so_upcall = NULL; 846 soshutdown(so, 2); 847 closef(fp, (struct proc *)0); 848 if (slp->ns_nam) 849 MFREE(slp->ns_nam, m); 850 m_freem(slp->ns_raw); 851 m_freem(slp->ns_rec); 852 for (nuidp = slp->ns_uidlruhead.tqh_first; nuidp != 0; 853 nuidp = nnuidp) { 854 nnuidp = nuidp->nu_lru.tqe_next; 855 LIST_REMOVE(nuidp, nu_hash); 856 TAILQ_REMOVE(&slp->ns_uidlruhead, nuidp, nu_lru); 857 if (nuidp->nu_flag & NU_NAM) 858 m_freem(nuidp->nu_nam); 859 free((caddr_t)nuidp, M_NFSUID); 860 } 861 s = splsoftclock(); 862 for (nwp = slp->ns_tq.lh_first; nwp; nwp = nnwp) { 863 nnwp = nwp->nd_tq.le_next; 864 LIST_REMOVE(nwp, nd_tq); 865 free((caddr_t)nwp, M_NFSRVDESC); 866 } 867 LIST_INIT(&slp->ns_tq); 868 splx(s); 869 } 870} 871 872/* 873 * Get an authorization string for the uid by having the mount_nfs sitting 874 * on this mount point porpous out of the kernel and do it. 875 */ 876int 877nfs_getauth(nmp, rep, cred, auth_str, auth_len, verf_str, verf_len, key) 878 register struct nfsmount *nmp; 879 struct nfsreq *rep; 880 struct ucred *cred; 881 char **auth_str; 882 int *auth_len; 883 char *verf_str; 884 int *verf_len; 885 NFSKERBKEY_T key; /* return session key */ 886{ 887 int error = 0; 888 889 while ((nmp->nm_flag & NFSMNT_WAITAUTH) == 0) { 890 nmp->nm_flag |= NFSMNT_WANTAUTH; 891 (void) tsleep((caddr_t)&nmp->nm_authtype, PSOCK, 892 "nfsauth1", 2 * hz); 893 error = nfs_sigintr(nmp, rep, rep->r_procp); 894 if (error) { 895 nmp->nm_flag &= ~NFSMNT_WANTAUTH; 896 return (error); 897 } 898 } 899 nmp->nm_flag &= ~(NFSMNT_WAITAUTH | NFSMNT_WANTAUTH); 900 nmp->nm_authstr = *auth_str = (char *)malloc(RPCAUTH_MAXSIZ, M_TEMP, M_WAITOK); 901 nmp->nm_authlen = RPCAUTH_MAXSIZ; 902 nmp->nm_verfstr = verf_str; 903 nmp->nm_verflen = *verf_len; 904 nmp->nm_authuid = cred->cr_uid; 905 wakeup((caddr_t)&nmp->nm_authstr); 906 907 /* 908 * And wait for mount_nfs to do its stuff. 909 */ 910 while ((nmp->nm_flag & NFSMNT_HASAUTH) == 0 && error == 0) { 911 (void) tsleep((caddr_t)&nmp->nm_authlen, PSOCK, 912 "nfsauth2", 2 * hz); 913 error = nfs_sigintr(nmp, rep, rep->r_procp); 914 } 915 if (nmp->nm_flag & NFSMNT_AUTHERR) { 916 nmp->nm_flag &= ~NFSMNT_AUTHERR; 917 error = EAUTH; 918 } 919 if (error) 920 free((caddr_t)*auth_str, M_TEMP); 921 else { 922 *auth_len = nmp->nm_authlen; 923 *verf_len = nmp->nm_verflen; 924 bcopy((caddr_t)nmp->nm_key, (caddr_t)key, sizeof (key)); 925 } 926 nmp->nm_flag &= ~NFSMNT_HASAUTH; 927 nmp->nm_flag |= NFSMNT_WAITAUTH; 928 if (nmp->nm_flag & NFSMNT_WANTAUTH) { 929 nmp->nm_flag &= ~NFSMNT_WANTAUTH; 930 wakeup((caddr_t)&nmp->nm_authtype); 931 } 932 return (error); 933} 934 935/* 936 * Get a nickname authenticator and verifier. 937 */ 938int 939nfs_getnickauth(nmp, cred, auth_str, auth_len, verf_str, verf_len) 940 struct nfsmount *nmp; 941 struct ucred *cred; 942 char **auth_str; 943 int *auth_len; 944 char *verf_str; 945 int verf_len; 946{ 947 register struct nfsuid *nuidp; 948 register u_long *nickp, *verfp; 949 struct timeval ktvin, ktvout; 950 951#ifdef DIAGNOSTIC 952 if (verf_len < (4 * NFSX_UNSIGNED)) 953 panic("nfs_getnickauth verf too small"); 954#endif 955 for (nuidp = NMUIDHASH(nmp, cred->cr_uid)->lh_first; 956 nuidp != 0; nuidp = nuidp->nu_hash.le_next) { 957 if (nuidp->nu_cr.cr_uid == cred->cr_uid) 958 break; 959 } 960 if (!nuidp || nuidp->nu_expire < time.tv_sec) 961 return (EACCES); 962 963 /* 964 * Move to the end of the lru list (end of lru == most recently used). 965 */ 966 TAILQ_REMOVE(&nmp->nm_uidlruhead, nuidp, nu_lru); 967 TAILQ_INSERT_TAIL(&nmp->nm_uidlruhead, nuidp, nu_lru); 968 969 nickp = (u_long *)malloc(2 * NFSX_UNSIGNED, M_TEMP, M_WAITOK); 970 *nickp++ = txdr_unsigned(RPCAKN_NICKNAME); 971 *nickp = txdr_unsigned(nuidp->nu_nickname); 972 *auth_str = (char *)nickp; 973 *auth_len = 2 * NFSX_UNSIGNED; 974 975 /* 976 * Now we must encrypt the verifier and package it up. 977 */ 978 verfp = (u_long *)verf_str; 979 *verfp++ = txdr_unsigned(RPCAKN_NICKNAME); 980 if (time.tv_sec > nuidp->nu_timestamp.tv_sec || 981 (time.tv_sec == nuidp->nu_timestamp.tv_sec && 982 time.tv_usec > nuidp->nu_timestamp.tv_usec)) 983 gettime(&nuidp->nu_timestamp); 984 else 985 nuidp->nu_timestamp.tv_usec++; 986 ktvin.tv_sec = txdr_unsigned(nuidp->nu_timestamp.tv_sec); 987 ktvin.tv_usec = txdr_unsigned(nuidp->nu_timestamp.tv_usec); 988 989 /* 990 * Now encrypt the timestamp verifier in ecb mode using the session 991 * key. 992 */ 993#ifdef NFSKERB 994 XXX 995#endif 996 997 *verfp++ = ktvout.tv_sec; 998 *verfp++ = ktvout.tv_usec; 999 *verfp = 0; 1000 return (0); 1001} 1002 1003/* 1004 * Save the current nickname in a hash list entry on the mount point. 1005 */ 1006int 1007nfs_savenickauth(nmp, cred, len, key, mdp, dposp, mrep) 1008 register struct nfsmount *nmp; 1009 struct ucred *cred; 1010 int len; 1011 NFSKERBKEY_T key; 1012 struct mbuf **mdp; 1013 char **dposp; 1014 struct mbuf *mrep; 1015{ 1016 register struct nfsuid *nuidp; 1017 register u_long *tl; 1018 register long t1; 1019 struct mbuf *md = *mdp; 1020 struct timeval ktvin, ktvout; 1021 u_long nick; 1022 char *dpos = *dposp, *cp2; 1023 int deltasec, error = 0; 1024 1025 if (len == (3 * NFSX_UNSIGNED)) { 1026 nfsm_dissect(tl, u_long *, 3 * NFSX_UNSIGNED); 1027 ktvin.tv_sec = *tl++; 1028 ktvin.tv_usec = *tl++; 1029 nick = fxdr_unsigned(u_long, *tl); 1030 1031 /* 1032 * Decrypt the timestamp in ecb mode. 1033 */ 1034#ifdef NFSKERB 1035 XXX 1036#endif 1037 ktvout.tv_sec = fxdr_unsigned(long, ktvout.tv_sec); 1038 ktvout.tv_usec = fxdr_unsigned(long, ktvout.tv_usec); 1039 deltasec = time.tv_sec - ktvout.tv_sec; 1040 if (deltasec < 0) 1041 deltasec = -deltasec; 1042 /* 1043 * If ok, add it to the hash list for the mount point. 1044 */ 1045 if (deltasec <= NFS_KERBCLOCKSKEW) { 1046 if (nmp->nm_numuids < nuidhash_max) { 1047 nmp->nm_numuids++; 1048 nuidp = (struct nfsuid *) 1049 malloc(sizeof (struct nfsuid), M_NFSUID, 1050 M_WAITOK); 1051 } else { 1052 nuidp = nmp->nm_uidlruhead.tqh_first; 1053 LIST_REMOVE(nuidp, nu_hash); 1054 TAILQ_REMOVE(&nmp->nm_uidlruhead, nuidp, 1055 nu_lru); 1056 } 1057 nuidp->nu_flag = 0; 1058 nuidp->nu_cr.cr_uid = cred->cr_uid; 1059 nuidp->nu_expire = time.tv_sec + NFS_KERBTTL; 1060 nuidp->nu_timestamp = ktvout; 1061 nuidp->nu_nickname = nick; 1062 bcopy(key, nuidp->nu_key, sizeof (key)); 1063 TAILQ_INSERT_TAIL(&nmp->nm_uidlruhead, nuidp, 1064 nu_lru); 1065 LIST_INSERT_HEAD(NMUIDHASH(nmp, cred->cr_uid), 1066 nuidp, nu_hash); 1067 } 1068 } else 1069 nfsm_adv(nfsm_rndup(len)); 1070nfsmout: 1071 *mdp = md; 1072 *dposp = dpos; 1073 return (error); 1074} 1075 1076#ifndef NFS_NOSERVER 1077 1078/* 1079 * Derefence a server socket structure. If it has no more references and 1080 * is no longer valid, you can throw it away. 1081 */ 1082void 1083nfsrv_slpderef(slp) 1084 register struct nfssvc_sock *slp; 1085{ 1086 if (--(slp->ns_sref) == 0 && (slp->ns_flag & SLP_VALID) == 0) { 1087 TAILQ_REMOVE(&nfssvc_sockhead, slp, ns_chain); 1088 free((caddr_t)slp, M_NFSSVC); 1089 } 1090} 1091 1092/* 1093 * Initialize the data structures for the server. 1094 * Handshake with any new nfsds starting up to avoid any chance of 1095 * corruption. 1096 */ 1097void 1098nfsrv_init(terminating) 1099 int terminating; 1100{ 1101 register struct nfssvc_sock *slp, *nslp; 1102 1103 if (nfssvc_sockhead_flag & SLP_INIT) 1104 panic("nfsd init"); 1105 nfssvc_sockhead_flag |= SLP_INIT; 1106 if (terminating) { 1107 for (slp = nfssvc_sockhead.tqh_first; slp != 0; slp = nslp) { 1108 nslp = slp->ns_chain.tqe_next; 1109 if (slp->ns_flag & SLP_VALID) 1110 nfsrv_zapsock(slp); 1111 TAILQ_REMOVE(&nfssvc_sockhead, slp, ns_chain); 1112 free((caddr_t)slp, M_NFSSVC); 1113 } 1114 nfsrv_cleancache(); /* And clear out server cache */ 1115 } 1116 1117 TAILQ_INIT(&nfssvc_sockhead); 1118 nfssvc_sockhead_flag &= ~SLP_INIT; 1119 if (nfssvc_sockhead_flag & SLP_WANTINIT) { 1120 nfssvc_sockhead_flag &= ~SLP_WANTINIT; 1121 wakeup((caddr_t)&nfssvc_sockhead); 1122 } 1123 1124 TAILQ_INIT(&nfsd_head); 1125 nfsd_head_flag &= ~NFSD_CHECKSLP; 1126 1127 nfs_udpsock = (struct nfssvc_sock *) 1128 malloc(sizeof (struct nfssvc_sock), M_NFSSVC, M_WAITOK); 1129 bzero((caddr_t)nfs_udpsock, sizeof (struct nfssvc_sock)); 1130 TAILQ_INIT(&nfs_udpsock->ns_uidlruhead); 1131 TAILQ_INSERT_HEAD(&nfssvc_sockhead, nfs_udpsock, ns_chain); 1132 1133 nfs_cltpsock = (struct nfssvc_sock *) 1134 malloc(sizeof (struct nfssvc_sock), M_NFSSVC, M_WAITOK); 1135 bzero((caddr_t)nfs_cltpsock, sizeof (struct nfssvc_sock)); 1136 TAILQ_INIT(&nfs_cltpsock->ns_uidlruhead); 1137 TAILQ_INSERT_TAIL(&nfssvc_sockhead, nfs_cltpsock, ns_chain); 1138} 1139 1140/* 1141 * Add entries to the server monitor log. 1142 */ 1143static void 1144nfsd_rt(sotype, nd, cacherep) 1145 int sotype; 1146 register struct nfsrv_descript *nd; 1147 int cacherep; 1148{ 1149 register struct drt *rt; 1150 1151 rt = &nfsdrt.drt[nfsdrt.pos]; 1152 if (cacherep == RC_DOIT) 1153 rt->flag = 0; 1154 else if (cacherep == RC_REPLY) 1155 rt->flag = DRT_CACHEREPLY; 1156 else 1157 rt->flag = DRT_CACHEDROP; 1158 if (sotype == SOCK_STREAM) 1159 rt->flag |= DRT_TCP; 1160 if (nd->nd_flag & ND_NQNFS) 1161 rt->flag |= DRT_NQNFS; 1162 else if (nd->nd_flag & ND_NFSV3) 1163 rt->flag |= DRT_NFSV3; 1164 rt->proc = nd->nd_procnum; 1165 if (mtod(nd->nd_nam, struct sockaddr *)->sa_family == AF_INET) 1166 rt->ipadr = mtod(nd->nd_nam, struct sockaddr_in *)->sin_addr.s_addr; 1167 else 1168 rt->ipadr = INADDR_ANY; 1169 rt->resptime = ((time.tv_sec - nd->nd_starttime.tv_sec) * 1000000) + 1170 (time.tv_usec - nd->nd_starttime.tv_usec); 1171 gettime(&rt->tstamp); 1172 nfsdrt.pos = (nfsdrt.pos + 1) % NFSRTTLOGSIZ; 1173} 1174#endif /* NFS_NOSERVER */ 1175