nfs_nfsiod.c revision 24330
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 1.19 1997/03/22 06:53:11 bde 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 *)); 109static int nfssvc_nfsd __P((struct nfsd_srvargs *,caddr_t,struct proc *)); 110 111static int nfs_privport = 0; 112SYSCTL_INT(_vfs_nfs, NFS_NFSPRIVPORT, nfs_privport, CTLFLAG_RW, &nfs_privport, 0, ""); 113 114/* 115 * NFS server system calls 116 * getfh() lives here too, but maybe should move to kern/vfs_syscalls.c 117 */ 118 119/* 120 * Get file handle system call 121 */ 122#ifndef _SYS_SYSPROTO_H_ 123struct getfh_args { 124 char *fname; 125 fhandle_t *fhp; 126}; 127#endif 128int 129getfh(p, uap, retval) 130 struct proc *p; 131 register struct getfh_args *uap; 132 int *retval; 133{ 134 register struct vnode *vp; 135 fhandle_t fh; 136 int error; 137 struct nameidata nd; 138 139 /* 140 * Must be super user 141 */ 142 error = suser(p->p_ucred, &p->p_acflag); 143 if(error) 144 return (error); 145 NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF, UIO_USERSPACE, uap->fname, p); 146 error = namei(&nd); 147 if (error) 148 return (error); 149 vp = nd.ni_vp; 150 bzero((caddr_t)&fh, sizeof(fh)); 151 fh.fh_fsid = vp->v_mount->mnt_stat.f_fsid; 152 error = VFS_VPTOFH(vp, &fh.fh_fid); 153 vput(vp); 154 if (error) 155 return (error); 156 error = copyout((caddr_t)&fh, (caddr_t)uap->fhp, sizeof (fh)); 157 return (error); 158} 159 160#endif /* NFS_NOSERVER */ 161/* 162 * Nfs server psuedo system call for the nfsd's 163 * Based on the flag value it either: 164 * - adds a socket to the selection list 165 * - remains in the kernel as an nfsd 166 * - remains in the kernel as an nfsiod 167 */ 168#ifndef _SYS_SYSPROTO_H_ 169struct nfssvc_args { 170 int flag; 171 caddr_t argp; 172}; 173#endif 174int 175nfssvc(p, uap, retval) 176 struct proc *p; 177 register struct nfssvc_args *uap; 178 int *retval; 179{ 180#ifndef NFS_NOSERVER 181 struct nameidata nd; 182 struct file *fp; 183 struct mbuf *nam; 184 struct nfsd_args nfsdarg; 185 struct nfsd_srvargs nfsd_srvargs, *nsd = &nfsd_srvargs; 186 struct nfsd_cargs ncd; 187 struct nfsd *nfsd; 188 struct nfssvc_sock *slp; 189 struct nfsuid *nuidp; 190 struct nfsmount *nmp; 191#endif /* NFS_NOSERVER */ 192 int error; 193 194 /* 195 * Must be super user 196 */ 197 error = suser(p->p_ucred, &p->p_acflag); 198 if(error) 199 return (error); 200 while (nfssvc_sockhead_flag & SLP_INIT) { 201 nfssvc_sockhead_flag |= SLP_WANTINIT; 202 (void) tsleep((caddr_t)&nfssvc_sockhead, PSOCK, "nfsd init", 0); 203 } 204 if (uap->flag & NFSSVC_BIOD) 205 error = nfssvc_iod(p); 206#ifdef NFS_NOSERVER 207 else 208 error = ENXIO; 209#else /* !NFS_NOSERVER */ 210 else if (uap->flag & NFSSVC_MNTD) { 211 error = copyin(uap->argp, (caddr_t)&ncd, sizeof (ncd)); 212 if (error) 213 return (error); 214 NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF, UIO_USERSPACE, 215 ncd.ncd_dirp, p); 216 error = namei(&nd); 217 if (error) 218 return (error); 219 if ((nd.ni_vp->v_flag & VROOT) == 0) 220 error = EINVAL; 221 nmp = VFSTONFS(nd.ni_vp->v_mount); 222 vput(nd.ni_vp); 223 if (error) 224 return (error); 225 if ((nmp->nm_flag & NFSMNT_MNTD) && 226 (uap->flag & NFSSVC_GOTAUTH) == 0) 227 return (0); 228 nmp->nm_flag |= NFSMNT_MNTD; 229 error = nqnfs_clientd(nmp, p->p_ucred, &ncd, uap->flag, 230 uap->argp, p); 231 } else if (uap->flag & NFSSVC_ADDSOCK) { 232 error = copyin(uap->argp, (caddr_t)&nfsdarg, sizeof(nfsdarg)); 233 if (error) 234 return (error); 235 error = getsock(p->p_fd, nfsdarg.sock, &fp); 236 if (error) 237 return (error); 238 /* 239 * Get the client address for connected sockets. 240 */ 241 if (nfsdarg.name == NULL || nfsdarg.namelen == 0) 242 nam = (struct mbuf *)0; 243 else { 244 error = sockargs(&nam, nfsdarg.name, nfsdarg.namelen, 245 MT_SONAME); 246 if (error) 247 return (error); 248 } 249 error = nfssvc_addsock(fp, nam); 250 } else { 251 error = copyin(uap->argp, (caddr_t)nsd, sizeof (*nsd)); 252 if (error) 253 return (error); 254 if ((uap->flag & NFSSVC_AUTHIN) && ((nfsd = nsd->nsd_nfsd)) && 255 (nfsd->nfsd_slp->ns_flag & SLP_VALID)) { 256 slp = nfsd->nfsd_slp; 257 258 /* 259 * First check to see if another nfsd has already 260 * added this credential. 261 */ 262 for (nuidp = NUIDHASH(slp,nsd->nsd_cr.cr_uid)->lh_first; 263 nuidp != 0; nuidp = nuidp->nu_hash.le_next) { 264 if (nuidp->nu_cr.cr_uid == nsd->nsd_cr.cr_uid && 265 (!nfsd->nfsd_nd->nd_nam2 || 266 netaddr_match(NU_NETFAM(nuidp), 267 &nuidp->nu_haddr, nfsd->nfsd_nd->nd_nam2))) 268 break; 269 } 270 if (nuidp) { 271 nfsrv_setcred(&nuidp->nu_cr,&nfsd->nfsd_nd->nd_cr); 272 nfsd->nfsd_nd->nd_flag |= ND_KERBFULL; 273 } else { 274 /* 275 * Nope, so we will. 276 */ 277 if (slp->ns_numuids < nuidhash_max) { 278 slp->ns_numuids++; 279 nuidp = (struct nfsuid *) 280 malloc(sizeof (struct nfsuid), M_NFSUID, 281 M_WAITOK); 282 } else 283 nuidp = (struct nfsuid *)0; 284 if ((slp->ns_flag & SLP_VALID) == 0) { 285 if (nuidp) 286 free((caddr_t)nuidp, M_NFSUID); 287 } else { 288 if (nuidp == (struct nfsuid *)0) { 289 nuidp = slp->ns_uidlruhead.tqh_first; 290 LIST_REMOVE(nuidp, nu_hash); 291 TAILQ_REMOVE(&slp->ns_uidlruhead, nuidp, 292 nu_lru); 293 if (nuidp->nu_flag & NU_NAM) 294 m_freem(nuidp->nu_nam); 295 } 296 nuidp->nu_flag = 0; 297 nuidp->nu_cr = nsd->nsd_cr; 298 if (nuidp->nu_cr.cr_ngroups > NGROUPS) 299 nuidp->nu_cr.cr_ngroups = NGROUPS; 300 nuidp->nu_cr.cr_ref = 1; 301 nuidp->nu_timestamp = nsd->nsd_timestamp; 302 nuidp->nu_expire = time.tv_sec + nsd->nsd_ttl; 303 /* 304 * and save the session key in nu_key. 305 */ 306 bcopy(nsd->nsd_key, nuidp->nu_key, 307 sizeof (nsd->nsd_key)); 308 if (nfsd->nfsd_nd->nd_nam2) { 309 struct sockaddr_in *saddr; 310 311 saddr = mtod(nfsd->nfsd_nd->nd_nam2, 312 struct sockaddr_in *); 313 switch (saddr->sin_family) { 314 case AF_INET: 315 nuidp->nu_flag |= NU_INETADDR; 316 nuidp->nu_inetaddr = 317 saddr->sin_addr.s_addr; 318 break; 319 case AF_ISO: 320 default: 321 nuidp->nu_flag |= NU_NAM; 322 nuidp->nu_nam = m_copym( 323 nfsd->nfsd_nd->nd_nam2, 0, 324 M_COPYALL, M_WAIT); 325 break; 326 }; 327 } 328 TAILQ_INSERT_TAIL(&slp->ns_uidlruhead, nuidp, 329 nu_lru); 330 LIST_INSERT_HEAD(NUIDHASH(slp, nsd->nsd_uid), 331 nuidp, nu_hash); 332 nfsrv_setcred(&nuidp->nu_cr, 333 &nfsd->nfsd_nd->nd_cr); 334 nfsd->nfsd_nd->nd_flag |= ND_KERBFULL; 335 } 336 } 337 } 338 if ((uap->flag & NFSSVC_AUTHINFAIL) && (nfsd = nsd->nsd_nfsd)) 339 nfsd->nfsd_flag |= NFSD_AUTHFAIL; 340 error = nfssvc_nfsd(nsd, uap->argp, p); 341 } 342#endif /* NFS_NOSERVER */ 343 if (error == EINTR || error == ERESTART) 344 error = 0; 345 return (error); 346} 347 348#ifndef NFS_NOSERVER 349/* 350 * Adds a socket to the list for servicing by nfsds. 351 */ 352static int 353nfssvc_addsock(fp, mynam) 354 struct file *fp; 355 struct mbuf *mynam; 356{ 357 register struct mbuf *m; 358 register int siz; 359 register struct nfssvc_sock *slp; 360 register struct socket *so; 361 struct nfssvc_sock *tslp; 362 int error, s; 363 364 so = (struct socket *)fp->f_data; 365 tslp = (struct nfssvc_sock *)0; 366 /* 367 * Add it to the list, as required. 368 */ 369 if (so->so_proto->pr_protocol == IPPROTO_UDP) { 370 tslp = nfs_udpsock; 371 if (tslp->ns_flag & SLP_VALID) { 372 m_freem(mynam); 373 return (EPERM); 374 } 375#ifdef ISO 376 } else if (so->so_proto->pr_protocol == ISOPROTO_CLTP) { 377 tslp = nfs_cltpsock; 378 if (tslp->ns_flag & SLP_VALID) { 379 m_freem(mynam); 380 return (EPERM); 381 } 382#endif /* ISO */ 383 } 384 if (so->so_type == SOCK_STREAM) 385 siz = NFS_MAXPACKET + sizeof (u_long); 386 else 387 siz = NFS_MAXPACKET; 388 error = soreserve(so, siz, siz); 389 if (error) { 390 m_freem(mynam); 391 return (error); 392 } 393 394 /* 395 * Set protocol specific options { for now TCP only } and 396 * reserve some space. For datagram sockets, this can get called 397 * repeatedly for the same socket, but that isn't harmful. 398 */ 399 if (so->so_type == SOCK_STREAM) { 400 MGET(m, M_WAIT, MT_SOOPTS); 401 *mtod(m, int *) = 1; 402 m->m_len = sizeof(int); 403 sosetopt(so, SOL_SOCKET, SO_KEEPALIVE, m); 404 } 405 if (so->so_proto->pr_domain->dom_family == AF_INET && 406 so->so_proto->pr_protocol == IPPROTO_TCP) { 407 MGET(m, M_WAIT, MT_SOOPTS); 408 *mtod(m, int *) = 1; 409 m->m_len = sizeof(int); 410 sosetopt(so, IPPROTO_TCP, TCP_NODELAY, m); 411 } 412 so->so_rcv.sb_flags &= ~SB_NOINTR; 413 so->so_rcv.sb_timeo = 0; 414 so->so_snd.sb_flags &= ~SB_NOINTR; 415 so->so_snd.sb_timeo = 0; 416 if (tslp) 417 slp = tslp; 418 else { 419 slp = (struct nfssvc_sock *) 420 malloc(sizeof (struct nfssvc_sock), M_NFSSVC, M_WAITOK); 421 bzero((caddr_t)slp, sizeof (struct nfssvc_sock)); 422 TAILQ_INIT(&slp->ns_uidlruhead); 423 TAILQ_INSERT_TAIL(&nfssvc_sockhead, slp, ns_chain); 424 } 425 slp->ns_so = so; 426 slp->ns_nam = mynam; 427 fp->f_count++; 428 slp->ns_fp = fp; 429 s = splnet(); 430 so->so_upcallarg = (caddr_t)slp; 431 so->so_upcall = nfsrv_rcv; 432 slp->ns_flag = (SLP_VALID | SLP_NEEDQ); 433 nfsrv_wakenfsd(slp); 434 splx(s); 435 return (0); 436} 437 438/* 439 * Called by nfssvc() for nfsds. Just loops around servicing rpc requests 440 * until it is killed by a signal. 441 */ 442static int 443nfssvc_nfsd(nsd, argp, p) 444 struct nfsd_srvargs *nsd; 445 caddr_t argp; 446 struct proc *p; 447{ 448 register struct mbuf *m; 449 register int siz; 450 register struct nfssvc_sock *slp; 451 register struct socket *so; 452 register int *solockp; 453 struct nfsd *nfsd = nsd->nsd_nfsd; 454 struct nfsrv_descript *nd = NULL; 455 struct mbuf *mreq; 456 int error = 0, cacherep, s, sotype, writes_todo; 457 u_quad_t cur_usec; 458 459#ifndef nolint 460 cacherep = RC_DOIT; 461 writes_todo = 0; 462#endif 463 s = splnet(); 464 if (nfsd == (struct nfsd *)0) { 465 nsd->nsd_nfsd = nfsd = (struct nfsd *) 466 malloc(sizeof (struct nfsd), M_NFSD, M_WAITOK); 467 bzero((caddr_t)nfsd, sizeof (struct nfsd)); 468 nfsd->nfsd_procp = p; 469 TAILQ_INSERT_TAIL(&nfsd_head, nfsd, nfsd_chain); 470 nfs_numnfsd++; 471 } 472 /* 473 * Loop getting rpc requests until SIGKILL. 474 */ 475 for (;;) { 476 if ((nfsd->nfsd_flag & NFSD_REQINPROG) == 0) { 477 while (nfsd->nfsd_slp == (struct nfssvc_sock *)0 && 478 (nfsd_head_flag & NFSD_CHECKSLP) == 0) { 479 nfsd->nfsd_flag |= NFSD_WAITING; 480 nfsd_waiting++; 481 error = tsleep((caddr_t)nfsd, PSOCK | PCATCH, 482 "nfsd", 0); 483 nfsd_waiting--; 484 if (error) 485 goto done; 486 } 487 if (nfsd->nfsd_slp == (struct nfssvc_sock *)0 && 488 (nfsd_head_flag & NFSD_CHECKSLP) != 0) { 489 for (slp = nfssvc_sockhead.tqh_first; slp != 0; 490 slp = slp->ns_chain.tqe_next) { 491 if ((slp->ns_flag & (SLP_VALID | SLP_DOREC)) 492 == (SLP_VALID | SLP_DOREC)) { 493 slp->ns_flag &= ~SLP_DOREC; 494 slp->ns_sref++; 495 nfsd->nfsd_slp = slp; 496 break; 497 } 498 } 499 if (slp == 0) 500 nfsd_head_flag &= ~NFSD_CHECKSLP; 501 } 502 if ((slp = nfsd->nfsd_slp) == (struct nfssvc_sock *)0) 503 continue; 504 if (slp->ns_flag & SLP_VALID) { 505 if (slp->ns_flag & SLP_DISCONN) 506 nfsrv_zapsock(slp); 507 else if (slp->ns_flag & SLP_NEEDQ) { 508 slp->ns_flag &= ~SLP_NEEDQ; 509 (void) nfs_sndlock(&slp->ns_solock, 510 (struct nfsreq *)0); 511 nfsrv_rcv(slp->ns_so, (caddr_t)slp, 512 M_WAIT); 513 nfs_sndunlock(&slp->ns_solock); 514 } 515 error = nfsrv_dorec(slp, nfsd, &nd); 516 cur_usec = (u_quad_t)time.tv_sec * 1000000 + 517 (u_quad_t)time.tv_usec; 518 if (error && slp->ns_tq.lh_first && 519 slp->ns_tq.lh_first->nd_time <= cur_usec) { 520 error = 0; 521 cacherep = RC_DOIT; 522 writes_todo = 1; 523 } else 524 writes_todo = 0; 525 nfsd->nfsd_flag |= NFSD_REQINPROG; 526 } 527 } else { 528 error = 0; 529 slp = nfsd->nfsd_slp; 530 } 531 if (error || (slp->ns_flag & SLP_VALID) == 0) { 532 if (nd) { 533 free((caddr_t)nd, M_NFSRVDESC); 534 nd = NULL; 535 } 536 nfsd->nfsd_slp = (struct nfssvc_sock *)0; 537 nfsd->nfsd_flag &= ~NFSD_REQINPROG; 538 nfsrv_slpderef(slp); 539 continue; 540 } 541 splx(s); 542 so = slp->ns_so; 543 sotype = so->so_type; 544 if (so->so_proto->pr_flags & PR_CONNREQUIRED) 545 solockp = &slp->ns_solock; 546 else 547 solockp = (int *)0; 548 if (nd) { 549 gettime(&nd->nd_starttime); 550 if (nd->nd_nam2) 551 nd->nd_nam = nd->nd_nam2; 552 else 553 nd->nd_nam = slp->ns_nam; 554 555 /* 556 * Check to see if authorization is needed. 557 */ 558 if (nfsd->nfsd_flag & NFSD_NEEDAUTH) { 559 nfsd->nfsd_flag &= ~NFSD_NEEDAUTH; 560 nsd->nsd_haddr = mtod(nd->nd_nam, 561 struct sockaddr_in *)->sin_addr.s_addr; 562 nsd->nsd_authlen = nfsd->nfsd_authlen; 563 nsd->nsd_verflen = nfsd->nfsd_verflen; 564 if (!copyout(nfsd->nfsd_authstr,nsd->nsd_authstr, 565 nfsd->nfsd_authlen) && 566 !copyout(nfsd->nfsd_verfstr, nsd->nsd_verfstr, 567 nfsd->nfsd_verflen) && 568 !copyout((caddr_t)nsd, argp, sizeof (*nsd))) 569 return (ENEEDAUTH); 570 cacherep = RC_DROPIT; 571 } else 572 cacherep = nfsrv_getcache(nd, slp, &mreq); 573 574 /* 575 * Check for just starting up for NQNFS and send 576 * fake "try again later" replies to the NQNFS clients. 577 */ 578 if (notstarted && nqnfsstarttime <= time.tv_sec) { 579 if (modify_flag) { 580 nqnfsstarttime = time.tv_sec + nqsrv_writeslack; 581 modify_flag = 0; 582 } else 583 notstarted = 0; 584 } 585 if (notstarted) { 586 if ((nd->nd_flag & ND_NQNFS) == 0) 587 cacherep = RC_DROPIT; 588 else if (nd->nd_procnum != NFSPROC_WRITE) { 589 nd->nd_procnum = NFSPROC_NOOP; 590 nd->nd_repstat = NQNFS_TRYLATER; 591 cacherep = RC_DOIT; 592 } else 593 modify_flag = 1; 594 } else if (nfsd->nfsd_flag & NFSD_AUTHFAIL) { 595 nfsd->nfsd_flag &= ~NFSD_AUTHFAIL; 596 nd->nd_procnum = NFSPROC_NOOP; 597 nd->nd_repstat = (NFSERR_AUTHERR | AUTH_TOOWEAK); 598 cacherep = RC_DOIT; 599 } else if (nfs_privport) { 600 /* Check if source port is privileged */ 601 u_short port; 602 u_long addr; 603 struct mbuf *nam = nd->nd_nam; 604 struct sockaddr_in *sin; 605 606 sin = mtod(nam, struct sockaddr_in *); 607 port = ntohs(sin->sin_port); 608 if (port >= IPPORT_RESERVED) { 609 nd->nd_procnum = NFSPROC_NOOP; 610 nd->nd_repstat = (NFSERR_AUTHERR | AUTH_TOOWEAK); 611 cacherep = RC_DOIT; 612 printf("NFS request from unprivileged port (%s:%d)\n", 613 inet_ntoa(sin->sin_addr), port); 614 } 615 } 616 617 } 618 619 /* 620 * Loop to get all the write rpc relies that have been 621 * gathered together. 622 */ 623 do { 624 switch (cacherep) { 625 case RC_DOIT: 626 if (writes_todo || (nd->nd_procnum == NFSPROC_WRITE && 627 nfsrvw_procrastinate > 0 && !notstarted)) 628 error = nfsrv_writegather(&nd, slp, 629 nfsd->nfsd_procp, &mreq); 630 else 631 error = (*(nfsrv3_procs[nd->nd_procnum]))(nd, 632 slp, nfsd->nfsd_procp, &mreq); 633 if (mreq == NULL) 634 break; 635 if (error) { 636 if (nd->nd_procnum != NQNFSPROC_VACATED) 637 nfsstats.srv_errs++; 638 nfsrv_updatecache(nd, FALSE, mreq); 639 if (nd->nd_nam2) 640 m_freem(nd->nd_nam2); 641 break; 642 } 643 nfsstats.srvrpccnt[nd->nd_procnum]++; 644 nfsrv_updatecache(nd, TRUE, mreq); 645 nd->nd_mrep = (struct mbuf *)0; 646 case RC_REPLY: 647 m = mreq; 648 siz = 0; 649 while (m) { 650 siz += m->m_len; 651 m = m->m_next; 652 } 653 if (siz <= 0 || siz > NFS_MAXPACKET) { 654 printf("mbuf siz=%d\n",siz); 655 panic("Bad nfs svc reply"); 656 } 657 m = mreq; 658 m->m_pkthdr.len = siz; 659 m->m_pkthdr.rcvif = (struct ifnet *)0; 660 /* 661 * For stream protocols, prepend a Sun RPC 662 * Record Mark. 663 */ 664 if (sotype == SOCK_STREAM) { 665 M_PREPEND(m, NFSX_UNSIGNED, M_WAIT); 666 *mtod(m, u_long *) = htonl(0x80000000 | siz); 667 } 668 if (solockp) 669 (void) nfs_sndlock(solockp, (struct nfsreq *)0); 670 if (slp->ns_flag & SLP_VALID) 671 error = nfs_send(so, nd->nd_nam2, m, NULL); 672 else { 673 error = EPIPE; 674 m_freem(m); 675 } 676 if (nfsrtton) 677 nfsd_rt(sotype, nd, cacherep); 678 if (nd->nd_nam2) 679 MFREE(nd->nd_nam2, m); 680 if (nd->nd_mrep) 681 m_freem(nd->nd_mrep); 682 if (error == EPIPE) 683 nfsrv_zapsock(slp); 684 if (solockp) 685 nfs_sndunlock(solockp); 686 if (error == EINTR || error == ERESTART) { 687 free((caddr_t)nd, M_NFSRVDESC); 688 nfsrv_slpderef(slp); 689 s = splnet(); 690 goto done; 691 } 692 break; 693 case RC_DROPIT: 694 if (nfsrtton) 695 nfsd_rt(sotype, nd, cacherep); 696 m_freem(nd->nd_mrep); 697 m_freem(nd->nd_nam2); 698 break; 699 }; 700 if (nd) { 701 FREE((caddr_t)nd, M_NFSRVDESC); 702 nd = NULL; 703 } 704 705 /* 706 * Check to see if there are outstanding writes that 707 * need to be serviced. 708 */ 709 cur_usec = (u_quad_t)time.tv_sec * 1000000 + 710 (u_quad_t)time.tv_usec; 711 s = splsoftclock(); 712 if (slp->ns_tq.lh_first && 713 slp->ns_tq.lh_first->nd_time <= cur_usec) { 714 cacherep = RC_DOIT; 715 writes_todo = 1; 716 } else 717 writes_todo = 0; 718 splx(s); 719 } while (writes_todo); 720 s = splnet(); 721 if (nfsrv_dorec(slp, nfsd, &nd)) { 722 nfsd->nfsd_flag &= ~NFSD_REQINPROG; 723 nfsd->nfsd_slp = NULL; 724 nfsrv_slpderef(slp); 725 } 726 } 727done: 728 TAILQ_REMOVE(&nfsd_head, nfsd, nfsd_chain); 729 splx(s); 730 free((caddr_t)nfsd, M_NFSD); 731 nsd->nsd_nfsd = (struct nfsd *)0; 732 if (--nfs_numnfsd == 0) 733 nfsrv_init(TRUE); /* Reinitialize everything */ 734 return (error); 735} 736#endif /* NFS_NOSERVER */ 737 738int nfs_defect = 0; 739SYSCTL_INT(_vfs_nfs, OID_AUTO, defect, CTLFLAG_RW, &nfs_defect, 0, ""); 740 741/* 742 * Asynchronous I/O daemons for client nfs. 743 * They do read-ahead and write-behind operations on the block I/O cache. 744 * Never returns unless it fails or gets killed. 745 */ 746static int 747nfssvc_iod(p) 748 struct proc *p; 749{ 750 register struct buf *bp, *nbp; 751 register int i, myiod; 752 struct vnode *vp; 753 struct nfsmount *nmp; 754 int error = 0, s; 755 756 /* 757 * Assign my position or return error if too many already running 758 */ 759 myiod = -1; 760 for (i = 0; i < NFS_MAXASYNCDAEMON; i++) 761 if (nfs_asyncdaemon[i] == 0) { 762 nfs_asyncdaemon[i]++; 763 myiod = i; 764 break; 765 } 766 if (myiod == -1) 767 return (EBUSY); 768 nfs_numasync++; 769 /* 770 * Just loop around doin our stuff until SIGKILL 771 */ 772 for (;;) { 773 while (((nmp = nfs_iodmount[myiod]) == NULL 774 || nmp->nm_bufq.tqh_first == NULL) 775 && error == 0) { 776 if (nmp) 777 nmp->nm_bufqiods--; 778 nfs_iodwant[myiod] = p; 779 nfs_iodmount[myiod] = NULL; 780 error = tsleep((caddr_t)&nfs_iodwant[myiod], 781 PWAIT | PCATCH, "nfsidl", 0); 782 } 783 if (error) { 784 nfs_asyncdaemon[myiod] = 0; 785 if (nmp) nmp->nm_bufqiods--; 786 nfs_iodmount[myiod] = NULL; 787 nfs_numasync--; 788 return (error); 789 } 790 while ((bp = nmp->nm_bufq.tqh_first) != NULL) { 791 /* Take one off the front of the list */ 792 TAILQ_REMOVE(&nmp->nm_bufq, bp, b_freelist); 793 nmp->nm_bufqlen--; 794 if (nmp->nm_bufqwant && nmp->nm_bufqlen < 2 * nfs_numasync) { 795 nmp->nm_bufqwant = FALSE; 796 wakeup(&nmp->nm_bufq); 797 } 798 if (bp->b_flags & B_READ) 799 (void) nfs_doio(bp, bp->b_rcred, (struct proc *)0); 800 else 801 (void) nfs_doio(bp, bp->b_wcred, (struct proc *)0); 802 803 /* 804 * If there are more than one iod on this mount, then defect 805 * so that the iods can be shared out fairly between the mounts 806 */ 807 if (nfs_defect && nmp->nm_bufqiods > 1) { 808 NFS_DPF(ASYNCIO, 809 ("nfssvc_iod: iod %d defecting from mount %p\n", 810 myiod, nmp)); 811 nfs_iodmount[myiod] = NULL; 812 nmp->nm_bufqiods--; 813 break; 814 } 815 } 816 } 817} 818 819/* 820 * Shut down a socket associated with an nfssvc_sock structure. 821 * Should be called with the send lock set, if required. 822 * The trick here is to increment the sref at the start, so that the nfsds 823 * will stop using it and clear ns_flag at the end so that it will not be 824 * reassigned during cleanup. 825 */ 826static void 827nfsrv_zapsock(slp) 828 register struct nfssvc_sock *slp; 829{ 830 register struct nfsuid *nuidp, *nnuidp; 831 register struct nfsrv_descript *nwp, *nnwp; 832 struct socket *so; 833 struct file *fp; 834 struct mbuf *m; 835 int s; 836 837 slp->ns_flag &= ~SLP_ALLFLAGS; 838 fp = slp->ns_fp; 839 if (fp) { 840 slp->ns_fp = (struct file *)0; 841 so = slp->ns_so; 842 so->so_upcall = NULL; 843 soshutdown(so, 2); 844 closef(fp, (struct proc *)0); 845 if (slp->ns_nam) 846 MFREE(slp->ns_nam, m); 847 m_freem(slp->ns_raw); 848 m_freem(slp->ns_rec); 849 for (nuidp = slp->ns_uidlruhead.tqh_first; nuidp != 0; 850 nuidp = nnuidp) { 851 nnuidp = nuidp->nu_lru.tqe_next; 852 LIST_REMOVE(nuidp, nu_hash); 853 TAILQ_REMOVE(&slp->ns_uidlruhead, nuidp, nu_lru); 854 if (nuidp->nu_flag & NU_NAM) 855 m_freem(nuidp->nu_nam); 856 free((caddr_t)nuidp, M_NFSUID); 857 } 858 s = splsoftclock(); 859 for (nwp = slp->ns_tq.lh_first; nwp; nwp = nnwp) { 860 nnwp = nwp->nd_tq.le_next; 861 LIST_REMOVE(nwp, nd_tq); 862 free((caddr_t)nwp, M_NFSRVDESC); 863 } 864 LIST_INIT(&slp->ns_tq); 865 splx(s); 866 } 867} 868 869/* 870 * Get an authorization string for the uid by having the mount_nfs sitting 871 * on this mount point porpous out of the kernel and do it. 872 */ 873int 874nfs_getauth(nmp, rep, cred, auth_str, auth_len, verf_str, verf_len, key) 875 register struct nfsmount *nmp; 876 struct nfsreq *rep; 877 struct ucred *cred; 878 char **auth_str; 879 int *auth_len; 880 char *verf_str; 881 int *verf_len; 882 NFSKERBKEY_T key; /* return session key */ 883{ 884 int error = 0; 885 886 while ((nmp->nm_flag & NFSMNT_WAITAUTH) == 0) { 887 nmp->nm_flag |= NFSMNT_WANTAUTH; 888 (void) tsleep((caddr_t)&nmp->nm_authtype, PSOCK, 889 "nfsauth1", 2 * hz); 890 error = nfs_sigintr(nmp, rep, rep->r_procp); 891 if (error) { 892 nmp->nm_flag &= ~NFSMNT_WANTAUTH; 893 return (error); 894 } 895 } 896 nmp->nm_flag &= ~(NFSMNT_WAITAUTH | NFSMNT_WANTAUTH); 897 nmp->nm_authstr = *auth_str = (char *)malloc(RPCAUTH_MAXSIZ, M_TEMP, M_WAITOK); 898 nmp->nm_authlen = RPCAUTH_MAXSIZ; 899 nmp->nm_verfstr = verf_str; 900 nmp->nm_verflen = *verf_len; 901 nmp->nm_authuid = cred->cr_uid; 902 wakeup((caddr_t)&nmp->nm_authstr); 903 904 /* 905 * And wait for mount_nfs to do its stuff. 906 */ 907 while ((nmp->nm_flag & NFSMNT_HASAUTH) == 0 && error == 0) { 908 (void) tsleep((caddr_t)&nmp->nm_authlen, PSOCK, 909 "nfsauth2", 2 * hz); 910 error = nfs_sigintr(nmp, rep, rep->r_procp); 911 } 912 if (nmp->nm_flag & NFSMNT_AUTHERR) { 913 nmp->nm_flag &= ~NFSMNT_AUTHERR; 914 error = EAUTH; 915 } 916 if (error) 917 free((caddr_t)*auth_str, M_TEMP); 918 else { 919 *auth_len = nmp->nm_authlen; 920 *verf_len = nmp->nm_verflen; 921 bcopy((caddr_t)nmp->nm_key, (caddr_t)key, sizeof (key)); 922 } 923 nmp->nm_flag &= ~NFSMNT_HASAUTH; 924 nmp->nm_flag |= NFSMNT_WAITAUTH; 925 if (nmp->nm_flag & NFSMNT_WANTAUTH) { 926 nmp->nm_flag &= ~NFSMNT_WANTAUTH; 927 wakeup((caddr_t)&nmp->nm_authtype); 928 } 929 return (error); 930} 931 932/* 933 * Get a nickname authenticator and verifier. 934 */ 935int 936nfs_getnickauth(nmp, cred, auth_str, auth_len, verf_str, verf_len) 937 struct nfsmount *nmp; 938 struct ucred *cred; 939 char **auth_str; 940 int *auth_len; 941 char *verf_str; 942 int verf_len; 943{ 944 register struct nfsuid *nuidp; 945 register u_long *nickp, *verfp; 946 struct timeval ktvin, ktvout; 947 948#ifdef DIAGNOSTIC 949 if (verf_len < (4 * NFSX_UNSIGNED)) 950 panic("nfs_getnickauth verf too small"); 951#endif 952 for (nuidp = NMUIDHASH(nmp, cred->cr_uid)->lh_first; 953 nuidp != 0; nuidp = nuidp->nu_hash.le_next) { 954 if (nuidp->nu_cr.cr_uid == cred->cr_uid) 955 break; 956 } 957 if (!nuidp || nuidp->nu_expire < time.tv_sec) 958 return (EACCES); 959 960 /* 961 * Move to the end of the lru list (end of lru == most recently used). 962 */ 963 TAILQ_REMOVE(&nmp->nm_uidlruhead, nuidp, nu_lru); 964 TAILQ_INSERT_TAIL(&nmp->nm_uidlruhead, nuidp, nu_lru); 965 966 nickp = (u_long *)malloc(2 * NFSX_UNSIGNED, M_TEMP, M_WAITOK); 967 *nickp++ = txdr_unsigned(RPCAKN_NICKNAME); 968 *nickp = txdr_unsigned(nuidp->nu_nickname); 969 *auth_str = (char *)nickp; 970 *auth_len = 2 * NFSX_UNSIGNED; 971 972 /* 973 * Now we must encrypt the verifier and package it up. 974 */ 975 verfp = (u_long *)verf_str; 976 *verfp++ = txdr_unsigned(RPCAKN_NICKNAME); 977 if (time.tv_sec > nuidp->nu_timestamp.tv_sec || 978 (time.tv_sec == nuidp->nu_timestamp.tv_sec && 979 time.tv_usec > nuidp->nu_timestamp.tv_usec)) 980 gettime(&nuidp->nu_timestamp); 981 else 982 nuidp->nu_timestamp.tv_usec++; 983 ktvin.tv_sec = txdr_unsigned(nuidp->nu_timestamp.tv_sec); 984 ktvin.tv_usec = txdr_unsigned(nuidp->nu_timestamp.tv_usec); 985 986 /* 987 * Now encrypt the timestamp verifier in ecb mode using the session 988 * key. 989 */ 990#ifdef NFSKERB 991 XXX 992#endif 993 994 *verfp++ = ktvout.tv_sec; 995 *verfp++ = ktvout.tv_usec; 996 *verfp = 0; 997 return (0); 998} 999 1000/* 1001 * Save the current nickname in a hash list entry on the mount point. 1002 */ 1003int 1004nfs_savenickauth(nmp, cred, len, key, mdp, dposp, mrep) 1005 register struct nfsmount *nmp; 1006 struct ucred *cred; 1007 int len; 1008 NFSKERBKEY_T key; 1009 struct mbuf **mdp; 1010 char **dposp; 1011 struct mbuf *mrep; 1012{ 1013 register struct nfsuid *nuidp; 1014 register u_long *tl; 1015 register long t1; 1016 struct mbuf *md = *mdp; 1017 struct timeval ktvin, ktvout; 1018 u_long nick; 1019 char *dpos = *dposp, *cp2; 1020 int deltasec, error = 0; 1021 1022 if (len == (3 * NFSX_UNSIGNED)) { 1023 nfsm_dissect(tl, u_long *, 3 * NFSX_UNSIGNED); 1024 ktvin.tv_sec = *tl++; 1025 ktvin.tv_usec = *tl++; 1026 nick = fxdr_unsigned(u_long, *tl); 1027 1028 /* 1029 * Decrypt the timestamp in ecb mode. 1030 */ 1031#ifdef NFSKERB 1032 XXX 1033#endif 1034 ktvout.tv_sec = fxdr_unsigned(long, ktvout.tv_sec); 1035 ktvout.tv_usec = fxdr_unsigned(long, ktvout.tv_usec); 1036 deltasec = time.tv_sec - ktvout.tv_sec; 1037 if (deltasec < 0) 1038 deltasec = -deltasec; 1039 /* 1040 * If ok, add it to the hash list for the mount point. 1041 */ 1042 if (deltasec <= NFS_KERBCLOCKSKEW) { 1043 if (nmp->nm_numuids < nuidhash_max) { 1044 nmp->nm_numuids++; 1045 nuidp = (struct nfsuid *) 1046 malloc(sizeof (struct nfsuid), M_NFSUID, 1047 M_WAITOK); 1048 } else { 1049 nuidp = nmp->nm_uidlruhead.tqh_first; 1050 LIST_REMOVE(nuidp, nu_hash); 1051 TAILQ_REMOVE(&nmp->nm_uidlruhead, nuidp, 1052 nu_lru); 1053 } 1054 nuidp->nu_flag = 0; 1055 nuidp->nu_cr.cr_uid = cred->cr_uid; 1056 nuidp->nu_expire = time.tv_sec + NFS_KERBTTL; 1057 nuidp->nu_timestamp = ktvout; 1058 nuidp->nu_nickname = nick; 1059 bcopy(key, nuidp->nu_key, sizeof (key)); 1060 TAILQ_INSERT_TAIL(&nmp->nm_uidlruhead, nuidp, 1061 nu_lru); 1062 LIST_INSERT_HEAD(NMUIDHASH(nmp, cred->cr_uid), 1063 nuidp, nu_hash); 1064 } 1065 } else 1066 nfsm_adv(nfsm_rndup(len)); 1067nfsmout: 1068 *mdp = md; 1069 *dposp = dpos; 1070 return (error); 1071} 1072 1073#ifndef NFS_NOSERVER 1074 1075/* 1076 * Derefence a server socket structure. If it has no more references and 1077 * is no longer valid, you can throw it away. 1078 */ 1079void 1080nfsrv_slpderef(slp) 1081 register struct nfssvc_sock *slp; 1082{ 1083 if (--(slp->ns_sref) == 0 && (slp->ns_flag & SLP_VALID) == 0) { 1084 TAILQ_REMOVE(&nfssvc_sockhead, slp, ns_chain); 1085 free((caddr_t)slp, M_NFSSVC); 1086 } 1087} 1088 1089/* 1090 * Initialize the data structures for the server. 1091 * Handshake with any new nfsds starting up to avoid any chance of 1092 * corruption. 1093 */ 1094void 1095nfsrv_init(terminating) 1096 int terminating; 1097{ 1098 register struct nfssvc_sock *slp, *nslp; 1099 1100 if (nfssvc_sockhead_flag & SLP_INIT) 1101 panic("nfsd init"); 1102 nfssvc_sockhead_flag |= SLP_INIT; 1103 if (terminating) { 1104 for (slp = nfssvc_sockhead.tqh_first; slp != 0; slp = nslp) { 1105 nslp = slp->ns_chain.tqe_next; 1106 if (slp->ns_flag & SLP_VALID) 1107 nfsrv_zapsock(slp); 1108 TAILQ_REMOVE(&nfssvc_sockhead, slp, ns_chain); 1109 free((caddr_t)slp, M_NFSSVC); 1110 } 1111 nfsrv_cleancache(); /* And clear out server cache */ 1112 } 1113 1114 TAILQ_INIT(&nfssvc_sockhead); 1115 nfssvc_sockhead_flag &= ~SLP_INIT; 1116 if (nfssvc_sockhead_flag & SLP_WANTINIT) { 1117 nfssvc_sockhead_flag &= ~SLP_WANTINIT; 1118 wakeup((caddr_t)&nfssvc_sockhead); 1119 } 1120 1121 TAILQ_INIT(&nfsd_head); 1122 nfsd_head_flag &= ~NFSD_CHECKSLP; 1123 1124 nfs_udpsock = (struct nfssvc_sock *) 1125 malloc(sizeof (struct nfssvc_sock), M_NFSSVC, M_WAITOK); 1126 bzero((caddr_t)nfs_udpsock, sizeof (struct nfssvc_sock)); 1127 TAILQ_INIT(&nfs_udpsock->ns_uidlruhead); 1128 TAILQ_INSERT_HEAD(&nfssvc_sockhead, nfs_udpsock, ns_chain); 1129 1130 nfs_cltpsock = (struct nfssvc_sock *) 1131 malloc(sizeof (struct nfssvc_sock), M_NFSSVC, M_WAITOK); 1132 bzero((caddr_t)nfs_cltpsock, sizeof (struct nfssvc_sock)); 1133 TAILQ_INIT(&nfs_cltpsock->ns_uidlruhead); 1134 TAILQ_INSERT_TAIL(&nfssvc_sockhead, nfs_cltpsock, ns_chain); 1135} 1136 1137/* 1138 * Add entries to the server monitor log. 1139 */ 1140static void 1141nfsd_rt(sotype, nd, cacherep) 1142 int sotype; 1143 register struct nfsrv_descript *nd; 1144 int cacherep; 1145{ 1146 register struct drt *rt; 1147 1148 rt = &nfsdrt.drt[nfsdrt.pos]; 1149 if (cacherep == RC_DOIT) 1150 rt->flag = 0; 1151 else if (cacherep == RC_REPLY) 1152 rt->flag = DRT_CACHEREPLY; 1153 else 1154 rt->flag = DRT_CACHEDROP; 1155 if (sotype == SOCK_STREAM) 1156 rt->flag |= DRT_TCP; 1157 if (nd->nd_flag & ND_NQNFS) 1158 rt->flag |= DRT_NQNFS; 1159 else if (nd->nd_flag & ND_NFSV3) 1160 rt->flag |= DRT_NFSV3; 1161 rt->proc = nd->nd_procnum; 1162 if (mtod(nd->nd_nam, struct sockaddr *)->sa_family == AF_INET) 1163 rt->ipadr = mtod(nd->nd_nam, struct sockaddr_in *)->sin_addr.s_addr; 1164 else 1165 rt->ipadr = INADDR_ANY; 1166 rt->resptime = ((time.tv_sec - nd->nd_starttime.tv_sec) * 1000000) + 1167 (time.tv_usec - nd->nd_starttime.tv_usec); 1168 gettime(&rt->tstamp); 1169 nfsdrt.pos = (nfsdrt.pos + 1) % NFSRTTLOGSIZ; 1170} 1171#endif /* NFS_NOSERVER */ 1172