2 * Copyright (c) 1989, 1993
3 * The Regents of the University of California. All rights reserved.
5 * This code is derived from software contributed to Berkeley by
6 * Rick Macklem at The University of Guelph.
8 * Redistribution and use in source and binary forms, with or without
9 * modification, are permitted provided that the following conditions
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.
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
36 * @(#)nfs_syscalls.c 8.5 (Berkeley) 3/30/95
37 * $FreeBSD: src/sys/nfs/nfs_syscalls.c,v 1.58.2.1 2000/11/26 02:30:06 dillon Exp $
38 * $DragonFly: src/sys/vfs/nfs/nfs_syscalls.c,v 1.2 2003/06/17 04:28:54 dillon Exp $
41 #include <sys/param.h>
42 #include <sys/systm.h>
43 #include <sys/sysproto.h>
44 #include <sys/kernel.h>
45 #include <sys/sysctl.h>
47 #include <sys/filedesc.h>
48 #include <sys/vnode.h>
49 #include <sys/malloc.h>
50 #include <sys/mount.h>
54 #include <sys/socket.h>
55 #include <sys/socketvar.h>
56 #include <sys/domain.h>
57 #include <sys/protosw.h>
58 #include <sys/namei.h>
59 #include <vm/vm_zone.h>
61 #include <netinet/in.h>
62 #include <netinet/tcp.h>
63 #include <nfs/xdr_subs.h>
64 #include <nfs/rpcv2.h>
65 #include <nfs/nfsproto.h>
67 #include <nfs/nfsm_subs.h>
68 #include <nfs/nfsrvcache.h>
69 #include <nfs/nfsmount.h>
70 #include <nfs/nfsnode.h>
71 #include <nfs/nqnfs.h>
72 #include <nfs/nfsrtt.h>
74 static MALLOC_DEFINE(M_NFSSVC, "NFS srvsock", "Nfs server structure");
77 extern int32_t (*nfsrv3_procs[NFS_NPROCS]) __P((struct nfsrv_descript *nd,
78 struct nfssvc_sock *slp,
80 struct mbuf **mreqp));
81 extern int nfs_numasync;
82 extern time_t nqnfsstarttime;
83 extern int nqsrv_writeslack;
85 extern struct nfsstats nfsstats;
86 extern int nfsrvw_procrastinate;
87 extern int nfsrvw_procrastinate_v3;
88 static int nuidhash_max = NFS_MAXUIDHASH;
91 static void nfsrv_zapsock __P((struct nfssvc_sock *slp));
93 static int nfssvc_iod __P((struct proc *));
98 static int nfs_asyncdaemon[NFS_MAXASYNCDAEMON];
100 SYSCTL_DECL(_vfs_nfs);
103 int nfsd_waiting = 0;
104 static struct nfsdrt nfsdrt;
105 static int nfs_numnfsd = 0;
106 static int notstarted = 1;
107 static int modify_flag = 0;
108 static void nfsd_rt __P((int sotype, struct nfsrv_descript *nd,
110 static int nfssvc_addsock __P((struct file *, struct sockaddr *,
112 static int nfssvc_nfsd __P((struct nfsd_srvargs *,caddr_t,struct proc *));
114 static int nfs_privport = 0;
115 SYSCTL_INT(_vfs_nfs, NFS_NFSPRIVPORT, nfs_privport, CTLFLAG_RW, &nfs_privport, 0, "");
116 SYSCTL_INT(_vfs_nfs, OID_AUTO, gatherdelay, CTLFLAG_RW, &nfsrvw_procrastinate, 0, "");
117 SYSCTL_INT(_vfs_nfs, OID_AUTO, gatherdelay_v3, CTLFLAG_RW, &nfsrvw_procrastinate_v3, 0, "");
120 * NFS server system calls
123 #endif /* NFS_NOSERVER */
125 * Nfs server psuedo system call for the nfsd's
126 * Based on the flag value it either:
127 * - adds a socket to the selection list
128 * - remains in the kernel as an nfsd
129 * - remains in the kernel as an nfsiod
131 #ifndef _SYS_SYSPROTO_H_
140 register struct nfssvc_args *uap;
145 struct sockaddr *nam;
146 struct nfsd_args nfsdarg;
147 struct nfsd_srvargs nfsd_srvargs, *nsd = &nfsd_srvargs;
148 struct nfsd_cargs ncd;
150 struct nfssvc_sock *slp;
151 struct nfsuid *nuidp;
152 struct nfsmount *nmp;
153 #endif /* NFS_NOSERVER */
162 while (nfssvc_sockhead_flag & SLP_INIT) {
163 nfssvc_sockhead_flag |= SLP_WANTINIT;
164 (void) tsleep((caddr_t)&nfssvc_sockhead, PSOCK, "nfsd init", 0);
166 if (uap->flag & NFSSVC_BIOD)
167 error = nfssvc_iod(p);
171 #else /* !NFS_NOSERVER */
172 else if (uap->flag & NFSSVC_MNTD) {
173 error = copyin(uap->argp, (caddr_t)&ncd, sizeof (ncd));
176 NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF, UIO_USERSPACE,
181 NDFREE(&nd, NDF_ONLY_PNBUF);
182 if ((nd.ni_vp->v_flag & VROOT) == 0)
184 nmp = VFSTONFS(nd.ni_vp->v_mount);
188 if ((nmp->nm_state & NFSSTA_MNTD) &&
189 (uap->flag & NFSSVC_GOTAUTH) == 0)
191 nmp->nm_state |= NFSSTA_MNTD;
192 error = nqnfs_clientd(nmp, p->p_ucred, &ncd, uap->flag,
194 } else if (uap->flag & NFSSVC_ADDSOCK) {
195 error = copyin(uap->argp, (caddr_t)&nfsdarg, sizeof(nfsdarg));
198 error = holdsock(p->p_fd, nfsdarg.sock, &fp);
202 * Get the client address for connected sockets.
204 if (nfsdarg.name == NULL || nfsdarg.namelen == 0)
205 nam = (struct sockaddr *)0;
207 error = getsockaddr(&nam, nfsdarg.name,
214 error = nfssvc_addsock(fp, nam, p);
217 error = copyin(uap->argp, (caddr_t)nsd, sizeof (*nsd));
220 if ((uap->flag & NFSSVC_AUTHIN) &&
221 ((nfsd = nsd->nsd_nfsd)) != NULL &&
222 (nfsd->nfsd_slp->ns_flag & SLP_VALID)) {
223 slp = nfsd->nfsd_slp;
226 * First check to see if another nfsd has already
227 * added this credential.
229 for (nuidp = NUIDHASH(slp,nsd->nsd_cr.cr_uid)->lh_first;
230 nuidp != 0; nuidp = nuidp->nu_hash.le_next) {
231 if (nuidp->nu_cr.cr_uid == nsd->nsd_cr.cr_uid &&
232 (!nfsd->nfsd_nd->nd_nam2 ||
233 netaddr_match(NU_NETFAM(nuidp),
234 &nuidp->nu_haddr, nfsd->nfsd_nd->nd_nam2)))
238 nfsrv_setcred(&nuidp->nu_cr,&nfsd->nfsd_nd->nd_cr);
239 nfsd->nfsd_nd->nd_flag |= ND_KERBFULL;
244 if (slp->ns_numuids < nuidhash_max) {
246 nuidp = (struct nfsuid *)
247 malloc(sizeof (struct nfsuid), M_NFSUID,
250 nuidp = (struct nfsuid *)0;
251 if ((slp->ns_flag & SLP_VALID) == 0) {
253 free((caddr_t)nuidp, M_NFSUID);
255 if (nuidp == (struct nfsuid *)0) {
256 nuidp = slp->ns_uidlruhead.tqh_first;
257 LIST_REMOVE(nuidp, nu_hash);
258 TAILQ_REMOVE(&slp->ns_uidlruhead, nuidp,
260 if (nuidp->nu_flag & NU_NAM)
261 FREE(nuidp->nu_nam, M_SONAME);
264 nuidp->nu_cr = nsd->nsd_cr;
265 if (nuidp->nu_cr.cr_ngroups > NGROUPS)
266 nuidp->nu_cr.cr_ngroups = NGROUPS;
267 nuidp->nu_cr.cr_ref = 1;
268 nuidp->nu_timestamp = nsd->nsd_timestamp;
269 nuidp->nu_expire = time_second + nsd->nsd_ttl;
271 * and save the session key in nu_key.
273 bcopy(nsd->nsd_key, nuidp->nu_key,
274 sizeof (nsd->nsd_key));
275 if (nfsd->nfsd_nd->nd_nam2) {
276 struct sockaddr_in *saddr;
278 saddr = (struct sockaddr_in *)
279 nfsd->nfsd_nd->nd_nam2;
280 switch (saddr->sin_family) {
282 nuidp->nu_flag |= NU_INETADDR;
284 saddr->sin_addr.s_addr;
288 nuidp->nu_flag |= NU_NAM;
290 dup_sockaddr(nfsd->nfsd_nd->
295 TAILQ_INSERT_TAIL(&slp->ns_uidlruhead, nuidp,
297 LIST_INSERT_HEAD(NUIDHASH(slp, nsd->nsd_uid),
299 nfsrv_setcred(&nuidp->nu_cr,
300 &nfsd->nfsd_nd->nd_cr);
301 nfsd->nfsd_nd->nd_flag |= ND_KERBFULL;
305 if ((uap->flag & NFSSVC_AUTHINFAIL) && (nfsd = nsd->nsd_nfsd))
306 nfsd->nfsd_flag |= NFSD_AUTHFAIL;
307 error = nfssvc_nfsd(nsd, uap->argp, p);
309 #endif /* NFS_NOSERVER */
310 if (error == EINTR || error == ERESTART)
317 * Adds a socket to the list for servicing by nfsds.
320 nfssvc_addsock(fp, mynam, p)
322 struct sockaddr *mynam;
326 register struct nfssvc_sock *slp;
327 register struct socket *so;
330 so = (struct socket *)fp->f_data;
332 tslp = (struct nfssvc_sock *)0;
334 * Add it to the list, as required.
336 if (so->so_proto->pr_protocol == IPPROTO_UDP) {
338 if (tslp->ns_flag & SLP_VALID) {
340 FREE(mynam, M_SONAME);
345 if (so->so_type == SOCK_STREAM)
346 siz = NFS_MAXPACKET + sizeof (u_long);
349 error = soreserve(so, siz, siz);
352 FREE(mynam, M_SONAME);
357 * Set protocol specific options { for now TCP only } and
358 * reserve some space. For datagram sockets, this can get called
359 * repeatedly for the same socket, but that isn't harmful.
361 if (so->so_type == SOCK_STREAM) {
365 bzero(&sopt, sizeof sopt);
366 sopt.sopt_level = SOL_SOCKET;
367 sopt.sopt_name = SO_KEEPALIVE;
368 sopt.sopt_val = &val;
369 sopt.sopt_valsize = sizeof val;
373 if (so->so_proto->pr_domain->dom_family == AF_INET &&
374 so->so_proto->pr_protocol == IPPROTO_TCP) {
378 bzero(&sopt, sizeof sopt);
379 sopt.sopt_level = IPPROTO_TCP;
380 sopt.sopt_name = TCP_NODELAY;
381 sopt.sopt_val = &val;
382 sopt.sopt_valsize = sizeof val;
386 so->so_rcv.sb_flags &= ~SB_NOINTR;
387 so->so_rcv.sb_timeo = 0;
388 so->so_snd.sb_flags &= ~SB_NOINTR;
389 so->so_snd.sb_timeo = 0;
391 slp = (struct nfssvc_sock *)
392 malloc(sizeof (struct nfssvc_sock), M_NFSSVC, M_WAITOK);
393 bzero((caddr_t)slp, sizeof (struct nfssvc_sock));
394 STAILQ_INIT(&slp->ns_rec);
395 TAILQ_INIT(&slp->ns_uidlruhead);
396 TAILQ_INSERT_TAIL(&nfssvc_sockhead, slp, ns_chain);
403 so->so_upcallarg = (caddr_t)slp;
404 so->so_upcall = nfsrv_rcv;
405 so->so_rcv.sb_flags |= SB_UPCALL;
406 slp->ns_flag = (SLP_VALID | SLP_NEEDQ);
413 * Called by nfssvc() for nfsds. Just loops around servicing rpc requests
414 * until it is killed by a signal.
417 nfssvc_nfsd(nsd, argp, p)
418 struct nfsd_srvargs *nsd;
423 register struct nfssvc_sock *slp;
424 struct nfsd *nfsd = nsd->nsd_nfsd;
425 struct nfsrv_descript *nd = NULL;
426 struct mbuf *m, *mreq;
427 int error = 0, cacherep, s, sotype, writes_todo;
435 if (nfsd == (struct nfsd *)0) {
436 nsd->nsd_nfsd = nfsd = (struct nfsd *)
437 malloc(sizeof (struct nfsd), M_NFSD, M_WAITOK);
438 bzero((caddr_t)nfsd, sizeof (struct nfsd));
440 nfsd->nfsd_procp = p;
441 TAILQ_INSERT_TAIL(&nfsd_head, nfsd, nfsd_chain);
447 * Loop getting rpc requests until SIGKILL.
450 if ((nfsd->nfsd_flag & NFSD_REQINPROG) == 0) {
451 while (nfsd->nfsd_slp == (struct nfssvc_sock *)0 &&
452 (nfsd_head_flag & NFSD_CHECKSLP) == 0) {
453 nfsd->nfsd_flag |= NFSD_WAITING;
455 error = tsleep((caddr_t)nfsd, PSOCK | PCATCH,
461 if (nfsd->nfsd_slp == (struct nfssvc_sock *)0 &&
462 (nfsd_head_flag & NFSD_CHECKSLP) != 0) {
463 for (slp = nfssvc_sockhead.tqh_first; slp != 0;
464 slp = slp->ns_chain.tqe_next) {
465 if ((slp->ns_flag & (SLP_VALID | SLP_DOREC))
466 == (SLP_VALID | SLP_DOREC)) {
467 slp->ns_flag &= ~SLP_DOREC;
469 nfsd->nfsd_slp = slp;
474 nfsd_head_flag &= ~NFSD_CHECKSLP;
476 if ((slp = nfsd->nfsd_slp) == (struct nfssvc_sock *)0)
478 if (slp->ns_flag & SLP_VALID) {
479 if (slp->ns_flag & SLP_DISCONN)
481 else if (slp->ns_flag & SLP_NEEDQ) {
482 slp->ns_flag &= ~SLP_NEEDQ;
483 (void) nfs_slplock(slp, 1);
484 nfsrv_rcv(slp->ns_so, (caddr_t)slp,
488 error = nfsrv_dorec(slp, nfsd, &nd);
489 cur_usec = nfs_curusec();
490 if (error && slp->ns_tq.lh_first &&
491 slp->ns_tq.lh_first->nd_time <= cur_usec) {
497 nfsd->nfsd_flag |= NFSD_REQINPROG;
501 slp = nfsd->nfsd_slp;
503 if (error || (slp->ns_flag & SLP_VALID) == 0) {
505 free((caddr_t)nd, M_NFSRVDESC);
508 nfsd->nfsd_slp = (struct nfssvc_sock *)0;
509 nfsd->nfsd_flag &= ~NFSD_REQINPROG;
514 sotype = slp->ns_so->so_type;
516 getmicrotime(&nd->nd_starttime);
518 nd->nd_nam = nd->nd_nam2;
520 nd->nd_nam = slp->ns_nam;
523 * Check to see if authorization is needed.
525 if (nfsd->nfsd_flag & NFSD_NEEDAUTH) {
526 nfsd->nfsd_flag &= ~NFSD_NEEDAUTH;
528 ((struct sockaddr_in *)
529 nd->nd_nam)->sin_addr.s_addr;
530 nsd->nsd_authlen = nfsd->nfsd_authlen;
531 nsd->nsd_verflen = nfsd->nfsd_verflen;
532 if (!copyout(nfsd->nfsd_authstr,nsd->nsd_authstr,
533 nfsd->nfsd_authlen) &&
534 !copyout(nfsd->nfsd_verfstr, nsd->nsd_verfstr,
535 nfsd->nfsd_verflen) &&
536 !copyout((caddr_t)nsd, argp, sizeof (*nsd)))
538 cacherep = RC_DROPIT;
540 cacherep = nfsrv_getcache(nd, slp, &mreq);
543 * Check for just starting up for NQNFS and send
544 * fake "try again later" replies to the NQNFS clients.
546 if (notstarted && nqnfsstarttime <= time_second) {
548 nqnfsstarttime = time_second + nqsrv_writeslack;
554 if ((nd->nd_flag & ND_NQNFS) == 0)
555 cacherep = RC_DROPIT;
556 else if (nd->nd_procnum != NFSPROC_WRITE) {
557 nd->nd_procnum = NFSPROC_NOOP;
558 nd->nd_repstat = NQNFS_TRYLATER;
562 } else if (nfsd->nfsd_flag & NFSD_AUTHFAIL) {
563 nfsd->nfsd_flag &= ~NFSD_AUTHFAIL;
564 nd->nd_procnum = NFSPROC_NOOP;
565 nd->nd_repstat = (NFSERR_AUTHERR | AUTH_TOOWEAK);
567 } else if (nfs_privport) {
568 /* Check if source port is privileged */
570 struct sockaddr *nam = nd->nd_nam;
571 struct sockaddr_in *sin;
573 sin = (struct sockaddr_in *)nam;
574 port = ntohs(sin->sin_port);
575 if (port >= IPPORT_RESERVED &&
576 nd->nd_procnum != NFSPROC_NULL) {
577 nd->nd_procnum = NFSPROC_NOOP;
578 nd->nd_repstat = (NFSERR_AUTHERR | AUTH_TOOWEAK);
580 printf("NFS request from unprivileged port (%s:%d)\n",
581 inet_ntoa(sin->sin_addr), port);
588 * Loop to get all the write rpc relies that have been
594 if (nd && (nd->nd_flag & ND_NFSV3))
595 procrastinate = nfsrvw_procrastinate_v3;
597 procrastinate = nfsrvw_procrastinate;
598 if (writes_todo || (nd->nd_procnum == NFSPROC_WRITE &&
599 procrastinate > 0 && !notstarted))
600 error = nfsrv_writegather(&nd, slp,
601 nfsd->nfsd_procp, &mreq);
603 error = (*(nfsrv3_procs[nd->nd_procnum]))(nd,
604 slp, nfsd->nfsd_procp, &mreq);
607 if (error != 0 && error != NFSERR_RETVOID) {
608 if (nd->nd_procnum != NQNFSPROC_VACATED)
610 nfsrv_updatecache(nd, FALSE, mreq);
612 FREE(nd->nd_nam2, M_SONAME);
615 nfsstats.srvrpccnt[nd->nd_procnum]++;
616 nfsrv_updatecache(nd, TRUE, mreq);
617 nd->nd_mrep = (struct mbuf *)0;
625 if (siz <= 0 || siz > NFS_MAXPACKET) {
626 printf("mbuf siz=%d\n",siz);
627 panic("Bad nfs svc reply");
630 m->m_pkthdr.len = siz;
631 m->m_pkthdr.rcvif = (struct ifnet *)0;
633 * For stream protocols, prepend a Sun RPC
636 if (sotype == SOCK_STREAM) {
637 M_PREPEND(m, NFSX_UNSIGNED, M_WAIT);
638 *mtod(m, u_int32_t *) = htonl(0x80000000 | siz);
640 if (slp->ns_so->so_proto->pr_flags & PR_CONNREQUIRED)
641 (void) nfs_slplock(slp, 1);
642 if (slp->ns_flag & SLP_VALID)
643 error = nfs_send(slp->ns_so, nd->nd_nam2, m, NULL);
649 nfsd_rt(sotype, nd, cacherep);
651 FREE(nd->nd_nam2, M_SONAME);
653 m_freem(nd->nd_mrep);
656 if (slp->ns_so->so_proto->pr_flags & PR_CONNREQUIRED)
658 if (error == EINTR || error == ERESTART) {
659 free((caddr_t)nd, M_NFSRVDESC);
667 nfsd_rt(sotype, nd, cacherep);
668 m_freem(nd->nd_mrep);
670 FREE(nd->nd_nam2, M_SONAME);
674 FREE((caddr_t)nd, M_NFSRVDESC);
679 * Check to see if there are outstanding writes that
680 * need to be serviced.
682 cur_usec = nfs_curusec();
684 if (slp->ns_tq.lh_first &&
685 slp->ns_tq.lh_first->nd_time <= cur_usec) {
691 } while (writes_todo);
693 if (nfsrv_dorec(slp, nfsd, &nd)) {
694 nfsd->nfsd_flag &= ~NFSD_REQINPROG;
695 nfsd->nfsd_slp = NULL;
700 TAILQ_REMOVE(&nfsd_head, nfsd, nfsd_chain);
702 free((caddr_t)nfsd, M_NFSD);
703 nsd->nsd_nfsd = (struct nfsd *)0;
704 if (--nfs_numnfsd == 0)
705 nfsrv_init(TRUE); /* Reinitialize everything */
710 * Shut down a socket associated with an nfssvc_sock structure.
711 * Should be called with the send lock set, if required.
712 * The trick here is to increment the sref at the start, so that the nfsds
713 * will stop using it and clear ns_flag at the end so that it will not be
714 * reassigned during cleanup.
718 register struct nfssvc_sock *slp;
720 register struct nfsuid *nuidp, *nnuidp;
721 register struct nfsrv_descript *nwp, *nnwp;
724 struct nfsrv_rec *rec;
727 slp->ns_flag &= ~SLP_ALLFLAGS;
730 slp->ns_fp = (struct file *)0;
732 so->so_rcv.sb_flags &= ~SB_UPCALL;
733 so->so_upcall = NULL;
734 so->so_upcallarg = NULL;
736 closef(fp, (struct proc *)0);
738 FREE(slp->ns_nam, M_SONAME);
739 m_freem(slp->ns_raw);
740 while ((rec = STAILQ_FIRST(&slp->ns_rec)) != NULL) {
741 STAILQ_REMOVE_HEAD(&slp->ns_rec, nr_link);
743 FREE(rec->nr_address, M_SONAME);
744 m_freem(rec->nr_packet);
745 free(rec, M_NFSRVDESC);
747 for (nuidp = slp->ns_uidlruhead.tqh_first; nuidp != 0;
749 nnuidp = nuidp->nu_lru.tqe_next;
750 LIST_REMOVE(nuidp, nu_hash);
751 TAILQ_REMOVE(&slp->ns_uidlruhead, nuidp, nu_lru);
752 if (nuidp->nu_flag & NU_NAM)
753 FREE(nuidp->nu_nam, M_SONAME);
754 free((caddr_t)nuidp, M_NFSUID);
757 for (nwp = slp->ns_tq.lh_first; nwp; nwp = nnwp) {
758 nnwp = nwp->nd_tq.le_next;
759 LIST_REMOVE(nwp, nd_tq);
760 free((caddr_t)nwp, M_NFSRVDESC);
762 LIST_INIT(&slp->ns_tq);
768 * Derefence a server socket structure. If it has no more references and
769 * is no longer valid, you can throw it away.
773 register struct nfssvc_sock *slp;
775 if (--(slp->ns_sref) == 0 && (slp->ns_flag & SLP_VALID) == 0) {
776 TAILQ_REMOVE(&nfssvc_sockhead, slp, ns_chain);
777 free((caddr_t)slp, M_NFSSVC);
782 * Lock a socket against others.
785 nfs_slplock(slp, wait)
786 register struct nfssvc_sock *slp;
789 int *statep = &slp->ns_solock;
791 if (!wait && (*statep & NFSSTA_SNDLOCK))
792 return(0); /* already locked, fail */
793 while (*statep & NFSSTA_SNDLOCK) {
794 *statep |= NFSSTA_WANTSND;
795 (void) tsleep((caddr_t)statep, PZERO - 1, "nfsslplck", 0);
797 *statep |= NFSSTA_SNDLOCK;
802 * Unlock the stream socket for others.
806 register struct nfssvc_sock *slp;
808 int *statep = &slp->ns_solock;
810 if ((*statep & NFSSTA_SNDLOCK) == 0)
811 panic("nfs slpunlock");
812 *statep &= ~NFSSTA_SNDLOCK;
813 if (*statep & NFSSTA_WANTSND) {
814 *statep &= ~NFSSTA_WANTSND;
815 wakeup((caddr_t)statep);
820 * Initialize the data structures for the server.
821 * Handshake with any new nfsds starting up to avoid any chance of
825 nfsrv_init(terminating)
828 register struct nfssvc_sock *slp, *nslp;
830 if (nfssvc_sockhead_flag & SLP_INIT)
832 nfssvc_sockhead_flag |= SLP_INIT;
834 for (slp = nfssvc_sockhead.tqh_first; slp != 0; slp = nslp) {
835 nslp = slp->ns_chain.tqe_next;
836 if (slp->ns_flag & SLP_VALID)
838 TAILQ_REMOVE(&nfssvc_sockhead, slp, ns_chain);
839 free((caddr_t)slp, M_NFSSVC);
841 nfsrv_cleancache(); /* And clear out server cache */
843 nfs_pub.np_valid = 0;
845 TAILQ_INIT(&nfssvc_sockhead);
846 nfssvc_sockhead_flag &= ~SLP_INIT;
847 if (nfssvc_sockhead_flag & SLP_WANTINIT) {
848 nfssvc_sockhead_flag &= ~SLP_WANTINIT;
849 wakeup((caddr_t)&nfssvc_sockhead);
852 TAILQ_INIT(&nfsd_head);
853 nfsd_head_flag &= ~NFSD_CHECKSLP;
856 nfs_udpsock = (struct nfssvc_sock *)
857 malloc(sizeof (struct nfssvc_sock), M_NFSSVC, M_WAITOK);
858 bzero((caddr_t)nfs_udpsock, sizeof (struct nfssvc_sock));
859 STAILQ_INIT(&nfs_udpsock->ns_rec);
860 TAILQ_INIT(&nfs_udpsock->ns_uidlruhead);
861 TAILQ_INSERT_HEAD(&nfssvc_sockhead, nfs_udpsock, ns_chain);
863 nfs_cltpsock = (struct nfssvc_sock *)
864 malloc(sizeof (struct nfssvc_sock), M_NFSSVC, M_WAITOK);
865 bzero((caddr_t)nfs_cltpsock, sizeof (struct nfssvc_sock));
866 STAILQ_INIT(&nfs_cltpsock->ns_rec);
867 TAILQ_INIT(&nfs_cltpsock->ns_uidlruhead);
868 TAILQ_INSERT_TAIL(&nfssvc_sockhead, nfs_cltpsock, ns_chain);
873 * Add entries to the server monitor log.
876 nfsd_rt(sotype, nd, cacherep)
878 register struct nfsrv_descript *nd;
881 register struct drt *rt;
883 rt = &nfsdrt.drt[nfsdrt.pos];
884 if (cacherep == RC_DOIT)
886 else if (cacherep == RC_REPLY)
887 rt->flag = DRT_CACHEREPLY;
889 rt->flag = DRT_CACHEDROP;
890 if (sotype == SOCK_STREAM)
892 if (nd->nd_flag & ND_NQNFS)
893 rt->flag |= DRT_NQNFS;
894 else if (nd->nd_flag & ND_NFSV3)
895 rt->flag |= DRT_NFSV3;
896 rt->proc = nd->nd_procnum;
897 if (nd->nd_nam->sa_family == AF_INET)
898 rt->ipadr = ((struct sockaddr_in *)nd->nd_nam)->sin_addr.s_addr;
900 rt->ipadr = INADDR_ANY;
901 rt->resptime = nfs_curusec() - (nd->nd_starttime.tv_sec * 1000000 + nd->nd_starttime.tv_usec);
902 getmicrotime(&rt->tstamp);
903 nfsdrt.pos = (nfsdrt.pos + 1) % NFSRTTLOGSIZ;
905 #endif /* NFS_NOSERVER */
907 static int nfs_defect = 0;
908 SYSCTL_INT(_vfs_nfs, OID_AUTO, defect, CTLFLAG_RW, &nfs_defect, 0, "");
911 * Asynchronous I/O daemons for client nfs.
912 * They do read-ahead and write-behind operations on the block I/O cache.
913 * Never returns unless it fails or gets killed.
919 register struct buf *bp;
920 register int i, myiod;
921 struct nfsmount *nmp;
925 * Assign my position or return error if too many already running
928 for (i = 0; i < NFS_MAXASYNCDAEMON; i++)
929 if (nfs_asyncdaemon[i] == 0) {
930 nfs_asyncdaemon[i]++;
938 * Just loop around doin our stuff until SIGKILL
941 while (((nmp = nfs_iodmount[myiod]) == NULL
942 || nmp->nm_bufq.tqh_first == NULL)
946 nfs_iodwant[myiod] = p;
947 nfs_iodmount[myiod] = NULL;
948 error = tsleep((caddr_t)&nfs_iodwant[myiod],
949 PWAIT | PCATCH, "nfsidl", 0);
952 nfs_asyncdaemon[myiod] = 0;
955 nfs_iodwant[myiod] = NULL;
956 nfs_iodmount[myiod] = NULL;
960 while ((bp = nmp->nm_bufq.tqh_first) != NULL) {
961 /* Take one off the front of the list */
962 TAILQ_REMOVE(&nmp->nm_bufq, bp, b_freelist);
964 if (nmp->nm_bufqwant && nmp->nm_bufqlen <= nfs_numasync) {
965 nmp->nm_bufqwant = FALSE;
966 wakeup(&nmp->nm_bufq);
968 if (bp->b_flags & B_READ)
969 (void) nfs_doio(bp, bp->b_rcred, (struct proc *)0);
971 (void) nfs_doio(bp, bp->b_wcred, (struct proc *)0);
973 * If there are more than one iod on this mount, then defect
974 * so that the iods can be shared out fairly between the mounts
976 if (nfs_defect && nmp->nm_bufqiods > 1) {
978 ("nfssvc_iod: iod %d defecting from mount %p\n",
980 nfs_iodmount[myiod] = NULL;
990 * Get an authorization string for the uid by having the mount_nfs sitting
991 * on this mount point porpous out of the kernel and do it.
994 nfs_getauth(nmp, rep, cred, auth_str, auth_len, verf_str, verf_len, key)
995 register struct nfsmount *nmp;
1002 NFSKERBKEY_T key; /* return session key */
1006 while ((nmp->nm_state & NFSSTA_WAITAUTH) == 0) {
1007 nmp->nm_state |= NFSSTA_WANTAUTH;
1008 (void) tsleep((caddr_t)&nmp->nm_authtype, PSOCK,
1009 "nfsauth1", 2 * hz);
1010 error = nfs_sigintr(nmp, rep, rep->r_procp);
1012 nmp->nm_state &= ~NFSSTA_WANTAUTH;
1016 nmp->nm_state &= ~(NFSSTA_WAITAUTH | NFSSTA_WANTAUTH);
1017 nmp->nm_authstr = *auth_str = (char *)malloc(RPCAUTH_MAXSIZ, M_TEMP, M_WAITOK);
1018 nmp->nm_authlen = RPCAUTH_MAXSIZ;
1019 nmp->nm_verfstr = verf_str;
1020 nmp->nm_verflen = *verf_len;
1021 nmp->nm_authuid = cred->cr_uid;
1022 wakeup((caddr_t)&nmp->nm_authstr);
1025 * And wait for mount_nfs to do its stuff.
1027 while ((nmp->nm_state & NFSSTA_HASAUTH) == 0 && error == 0) {
1028 (void) tsleep((caddr_t)&nmp->nm_authlen, PSOCK,
1029 "nfsauth2", 2 * hz);
1030 error = nfs_sigintr(nmp, rep, rep->r_procp);
1032 if (nmp->nm_state & NFSSTA_AUTHERR) {
1033 nmp->nm_state &= ~NFSSTA_AUTHERR;
1037 free((caddr_t)*auth_str, M_TEMP);
1039 *auth_len = nmp->nm_authlen;
1040 *verf_len = nmp->nm_verflen;
1041 bcopy((caddr_t)nmp->nm_key, (caddr_t)key, sizeof (key));
1043 nmp->nm_state &= ~NFSSTA_HASAUTH;
1044 nmp->nm_state |= NFSSTA_WAITAUTH;
1045 if (nmp->nm_state & NFSSTA_WANTAUTH) {
1046 nmp->nm_state &= ~NFSSTA_WANTAUTH;
1047 wakeup((caddr_t)&nmp->nm_authtype);
1053 * Get a nickname authenticator and verifier.
1056 nfs_getnickauth(nmp, cred, auth_str, auth_len, verf_str, verf_len)
1057 struct nfsmount *nmp;
1064 register struct nfsuid *nuidp;
1065 register u_int32_t *nickp, *verfp;
1066 struct timeval ktvin, ktvout;
1069 if (verf_len < (4 * NFSX_UNSIGNED))
1070 panic("nfs_getnickauth verf too small");
1072 for (nuidp = NMUIDHASH(nmp, cred->cr_uid)->lh_first;
1073 nuidp != 0; nuidp = nuidp->nu_hash.le_next) {
1074 if (nuidp->nu_cr.cr_uid == cred->cr_uid)
1077 if (!nuidp || nuidp->nu_expire < time_second)
1081 * Move to the end of the lru list (end of lru == most recently used).
1083 TAILQ_REMOVE(&nmp->nm_uidlruhead, nuidp, nu_lru);
1084 TAILQ_INSERT_TAIL(&nmp->nm_uidlruhead, nuidp, nu_lru);
1086 nickp = (u_int32_t *)malloc(2 * NFSX_UNSIGNED, M_TEMP, M_WAITOK);
1087 *nickp++ = txdr_unsigned(RPCAKN_NICKNAME);
1088 *nickp = txdr_unsigned(nuidp->nu_nickname);
1089 *auth_str = (char *)nickp;
1090 *auth_len = 2 * NFSX_UNSIGNED;
1093 * Now we must encrypt the verifier and package it up.
1095 verfp = (u_int32_t *)verf_str;
1096 *verfp++ = txdr_unsigned(RPCAKN_NICKNAME);
1097 if (time_second > nuidp->nu_timestamp.tv_sec ||
1098 (time_second == nuidp->nu_timestamp.tv_sec &&
1099 time_second > nuidp->nu_timestamp.tv_usec))
1100 getmicrotime(&nuidp->nu_timestamp);
1102 nuidp->nu_timestamp.tv_usec++;
1103 ktvin.tv_sec = txdr_unsigned(nuidp->nu_timestamp.tv_sec);
1104 ktvin.tv_usec = txdr_unsigned(nuidp->nu_timestamp.tv_usec);
1107 * Now encrypt the timestamp verifier in ecb mode using the session
1114 *verfp++ = ktvout.tv_sec;
1115 *verfp++ = ktvout.tv_usec;
1121 * Save the current nickname in a hash list entry on the mount point.
1124 nfs_savenickauth(nmp, cred, len, key, mdp, dposp, mrep)
1125 register struct nfsmount *nmp;
1133 register struct nfsuid *nuidp;
1134 register u_int32_t *tl;
1135 register int32_t t1;
1136 struct mbuf *md = *mdp;
1137 struct timeval ktvin, ktvout;
1139 char *dpos = *dposp, *cp2;
1140 int deltasec, error = 0;
1142 if (len == (3 * NFSX_UNSIGNED)) {
1143 nfsm_dissect(tl, u_int32_t *, 3 * NFSX_UNSIGNED);
1144 ktvin.tv_sec = *tl++;
1145 ktvin.tv_usec = *tl++;
1146 nick = fxdr_unsigned(u_int32_t, *tl);
1149 * Decrypt the timestamp in ecb mode.
1154 ktvout.tv_sec = fxdr_unsigned(long, ktvout.tv_sec);
1155 ktvout.tv_usec = fxdr_unsigned(long, ktvout.tv_usec);
1156 deltasec = time_second - ktvout.tv_sec;
1158 deltasec = -deltasec;
1160 * If ok, add it to the hash list for the mount point.
1162 if (deltasec <= NFS_KERBCLOCKSKEW) {
1163 if (nmp->nm_numuids < nuidhash_max) {
1165 nuidp = (struct nfsuid *)
1166 malloc(sizeof (struct nfsuid), M_NFSUID,
1169 nuidp = nmp->nm_uidlruhead.tqh_first;
1170 LIST_REMOVE(nuidp, nu_hash);
1171 TAILQ_REMOVE(&nmp->nm_uidlruhead, nuidp,
1175 nuidp->nu_cr.cr_uid = cred->cr_uid;
1176 nuidp->nu_expire = time_second + NFS_KERBTTL;
1177 nuidp->nu_timestamp = ktvout;
1178 nuidp->nu_nickname = nick;
1179 bcopy(key, nuidp->nu_key, sizeof (key));
1180 TAILQ_INSERT_TAIL(&nmp->nm_uidlruhead, nuidp,
1182 LIST_INSERT_HEAD(NMUIDHASH(nmp, cred->cr_uid),
1186 nfsm_adv(nfsm_rndup(len));