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 1.3 2003/06/23 17:55:48 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 * nfssvc_args(int flag, caddr_t argp)
127 * Nfs server psuedo system call for the nfsd's
128 * Based on the flag value it either:
129 * - adds a socket to the selection list
130 * - remains in the kernel as an nfsd
131 * - remains in the kernel as an nfsiod
134 nfssvc(struct nfssvc_args *uap)
139 struct sockaddr *nam;
140 struct nfsd_args nfsdarg;
141 struct nfsd_srvargs nfsd_srvargs, *nsd = &nfsd_srvargs;
142 struct nfsd_cargs ncd;
144 struct nfssvc_sock *slp;
145 struct nfsuid *nuidp;
146 struct nfsmount *nmp;
147 #endif /* NFS_NOSERVER */
149 struct proc *p = curproc;
157 while (nfssvc_sockhead_flag & SLP_INIT) {
158 nfssvc_sockhead_flag |= SLP_WANTINIT;
159 (void) tsleep((caddr_t)&nfssvc_sockhead, PSOCK, "nfsd init", 0);
161 if (uap->flag & NFSSVC_BIOD)
162 error = nfssvc_iod(p);
166 #else /* !NFS_NOSERVER */
167 else if (uap->flag & NFSSVC_MNTD) {
168 error = copyin(uap->argp, (caddr_t)&ncd, sizeof (ncd));
171 NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF, UIO_USERSPACE,
176 NDFREE(&nd, NDF_ONLY_PNBUF);
177 if ((nd.ni_vp->v_flag & VROOT) == 0)
179 nmp = VFSTONFS(nd.ni_vp->v_mount);
183 if ((nmp->nm_state & NFSSTA_MNTD) &&
184 (uap->flag & NFSSVC_GOTAUTH) == 0)
186 nmp->nm_state |= NFSSTA_MNTD;
187 error = nqnfs_clientd(nmp, p->p_ucred, &ncd, uap->flag,
189 } else if (uap->flag & NFSSVC_ADDSOCK) {
190 error = copyin(uap->argp, (caddr_t)&nfsdarg, sizeof(nfsdarg));
193 error = holdsock(p->p_fd, nfsdarg.sock, &fp);
197 * Get the client address for connected sockets.
199 if (nfsdarg.name == NULL || nfsdarg.namelen == 0)
200 nam = (struct sockaddr *)0;
202 error = getsockaddr(&nam, nfsdarg.name,
209 error = nfssvc_addsock(fp, nam, p);
212 error = copyin(uap->argp, (caddr_t)nsd, sizeof (*nsd));
215 if ((uap->flag & NFSSVC_AUTHIN) &&
216 ((nfsd = nsd->nsd_nfsd)) != NULL &&
217 (nfsd->nfsd_slp->ns_flag & SLP_VALID)) {
218 slp = nfsd->nfsd_slp;
221 * First check to see if another nfsd has already
222 * added this credential.
224 for (nuidp = NUIDHASH(slp,nsd->nsd_cr.cr_uid)->lh_first;
225 nuidp != 0; nuidp = nuidp->nu_hash.le_next) {
226 if (nuidp->nu_cr.cr_uid == nsd->nsd_cr.cr_uid &&
227 (!nfsd->nfsd_nd->nd_nam2 ||
228 netaddr_match(NU_NETFAM(nuidp),
229 &nuidp->nu_haddr, nfsd->nfsd_nd->nd_nam2)))
233 nfsrv_setcred(&nuidp->nu_cr,&nfsd->nfsd_nd->nd_cr);
234 nfsd->nfsd_nd->nd_flag |= ND_KERBFULL;
239 if (slp->ns_numuids < nuidhash_max) {
241 nuidp = (struct nfsuid *)
242 malloc(sizeof (struct nfsuid), M_NFSUID,
245 nuidp = (struct nfsuid *)0;
246 if ((slp->ns_flag & SLP_VALID) == 0) {
248 free((caddr_t)nuidp, M_NFSUID);
250 if (nuidp == (struct nfsuid *)0) {
251 nuidp = slp->ns_uidlruhead.tqh_first;
252 LIST_REMOVE(nuidp, nu_hash);
253 TAILQ_REMOVE(&slp->ns_uidlruhead, nuidp,
255 if (nuidp->nu_flag & NU_NAM)
256 FREE(nuidp->nu_nam, M_SONAME);
259 nuidp->nu_cr = nsd->nsd_cr;
260 if (nuidp->nu_cr.cr_ngroups > NGROUPS)
261 nuidp->nu_cr.cr_ngroups = NGROUPS;
262 nuidp->nu_cr.cr_ref = 1;
263 nuidp->nu_timestamp = nsd->nsd_timestamp;
264 nuidp->nu_expire = time_second + nsd->nsd_ttl;
266 * and save the session key in nu_key.
268 bcopy(nsd->nsd_key, nuidp->nu_key,
269 sizeof (nsd->nsd_key));
270 if (nfsd->nfsd_nd->nd_nam2) {
271 struct sockaddr_in *saddr;
273 saddr = (struct sockaddr_in *)
274 nfsd->nfsd_nd->nd_nam2;
275 switch (saddr->sin_family) {
277 nuidp->nu_flag |= NU_INETADDR;
279 saddr->sin_addr.s_addr;
283 nuidp->nu_flag |= NU_NAM;
285 dup_sockaddr(nfsd->nfsd_nd->
290 TAILQ_INSERT_TAIL(&slp->ns_uidlruhead, nuidp,
292 LIST_INSERT_HEAD(NUIDHASH(slp, nsd->nsd_uid),
294 nfsrv_setcred(&nuidp->nu_cr,
295 &nfsd->nfsd_nd->nd_cr);
296 nfsd->nfsd_nd->nd_flag |= ND_KERBFULL;
300 if ((uap->flag & NFSSVC_AUTHINFAIL) && (nfsd = nsd->nsd_nfsd))
301 nfsd->nfsd_flag |= NFSD_AUTHFAIL;
302 error = nfssvc_nfsd(nsd, uap->argp, p);
304 #endif /* NFS_NOSERVER */
305 if (error == EINTR || error == ERESTART)
312 * Adds a socket to the list for servicing by nfsds.
315 nfssvc_addsock(fp, mynam, p)
317 struct sockaddr *mynam;
321 register struct nfssvc_sock *slp;
322 register struct socket *so;
325 so = (struct socket *)fp->f_data;
327 tslp = (struct nfssvc_sock *)0;
329 * Add it to the list, as required.
331 if (so->so_proto->pr_protocol == IPPROTO_UDP) {
333 if (tslp->ns_flag & SLP_VALID) {
335 FREE(mynam, M_SONAME);
340 if (so->so_type == SOCK_STREAM)
341 siz = NFS_MAXPACKET + sizeof (u_long);
344 error = soreserve(so, siz, siz);
347 FREE(mynam, M_SONAME);
352 * Set protocol specific options { for now TCP only } and
353 * reserve some space. For datagram sockets, this can get called
354 * repeatedly for the same socket, but that isn't harmful.
356 if (so->so_type == SOCK_STREAM) {
360 bzero(&sopt, sizeof sopt);
361 sopt.sopt_level = SOL_SOCKET;
362 sopt.sopt_name = SO_KEEPALIVE;
363 sopt.sopt_val = &val;
364 sopt.sopt_valsize = sizeof val;
368 if (so->so_proto->pr_domain->dom_family == AF_INET &&
369 so->so_proto->pr_protocol == IPPROTO_TCP) {
373 bzero(&sopt, sizeof sopt);
374 sopt.sopt_level = IPPROTO_TCP;
375 sopt.sopt_name = TCP_NODELAY;
376 sopt.sopt_val = &val;
377 sopt.sopt_valsize = sizeof val;
381 so->so_rcv.sb_flags &= ~SB_NOINTR;
382 so->so_rcv.sb_timeo = 0;
383 so->so_snd.sb_flags &= ~SB_NOINTR;
384 so->so_snd.sb_timeo = 0;
386 slp = (struct nfssvc_sock *)
387 malloc(sizeof (struct nfssvc_sock), M_NFSSVC, M_WAITOK);
388 bzero((caddr_t)slp, sizeof (struct nfssvc_sock));
389 STAILQ_INIT(&slp->ns_rec);
390 TAILQ_INIT(&slp->ns_uidlruhead);
391 TAILQ_INSERT_TAIL(&nfssvc_sockhead, slp, ns_chain);
398 so->so_upcallarg = (caddr_t)slp;
399 so->so_upcall = nfsrv_rcv;
400 so->so_rcv.sb_flags |= SB_UPCALL;
401 slp->ns_flag = (SLP_VALID | SLP_NEEDQ);
408 * Called by nfssvc() for nfsds. Just loops around servicing rpc requests
409 * until it is killed by a signal.
412 nfssvc_nfsd(nsd, argp, p)
413 struct nfsd_srvargs *nsd;
418 register struct nfssvc_sock *slp;
419 struct nfsd *nfsd = nsd->nsd_nfsd;
420 struct nfsrv_descript *nd = NULL;
421 struct mbuf *m, *mreq;
422 int error = 0, cacherep, s, sotype, writes_todo;
430 if (nfsd == (struct nfsd *)0) {
431 nsd->nsd_nfsd = nfsd = (struct nfsd *)
432 malloc(sizeof (struct nfsd), M_NFSD, M_WAITOK);
433 bzero((caddr_t)nfsd, sizeof (struct nfsd));
435 nfsd->nfsd_procp = p;
436 TAILQ_INSERT_TAIL(&nfsd_head, nfsd, nfsd_chain);
442 * Loop getting rpc requests until SIGKILL.
445 if ((nfsd->nfsd_flag & NFSD_REQINPROG) == 0) {
446 while (nfsd->nfsd_slp == (struct nfssvc_sock *)0 &&
447 (nfsd_head_flag & NFSD_CHECKSLP) == 0) {
448 nfsd->nfsd_flag |= NFSD_WAITING;
450 error = tsleep((caddr_t)nfsd, PSOCK | PCATCH,
456 if (nfsd->nfsd_slp == (struct nfssvc_sock *)0 &&
457 (nfsd_head_flag & NFSD_CHECKSLP) != 0) {
458 for (slp = nfssvc_sockhead.tqh_first; slp != 0;
459 slp = slp->ns_chain.tqe_next) {
460 if ((slp->ns_flag & (SLP_VALID | SLP_DOREC))
461 == (SLP_VALID | SLP_DOREC)) {
462 slp->ns_flag &= ~SLP_DOREC;
464 nfsd->nfsd_slp = slp;
469 nfsd_head_flag &= ~NFSD_CHECKSLP;
471 if ((slp = nfsd->nfsd_slp) == (struct nfssvc_sock *)0)
473 if (slp->ns_flag & SLP_VALID) {
474 if (slp->ns_flag & SLP_DISCONN)
476 else if (slp->ns_flag & SLP_NEEDQ) {
477 slp->ns_flag &= ~SLP_NEEDQ;
478 (void) nfs_slplock(slp, 1);
479 nfsrv_rcv(slp->ns_so, (caddr_t)slp,
483 error = nfsrv_dorec(slp, nfsd, &nd);
484 cur_usec = nfs_curusec();
485 if (error && slp->ns_tq.lh_first &&
486 slp->ns_tq.lh_first->nd_time <= cur_usec) {
492 nfsd->nfsd_flag |= NFSD_REQINPROG;
496 slp = nfsd->nfsd_slp;
498 if (error || (slp->ns_flag & SLP_VALID) == 0) {
500 free((caddr_t)nd, M_NFSRVDESC);
503 nfsd->nfsd_slp = (struct nfssvc_sock *)0;
504 nfsd->nfsd_flag &= ~NFSD_REQINPROG;
509 sotype = slp->ns_so->so_type;
511 getmicrotime(&nd->nd_starttime);
513 nd->nd_nam = nd->nd_nam2;
515 nd->nd_nam = slp->ns_nam;
518 * Check to see if authorization is needed.
520 if (nfsd->nfsd_flag & NFSD_NEEDAUTH) {
521 nfsd->nfsd_flag &= ~NFSD_NEEDAUTH;
523 ((struct sockaddr_in *)
524 nd->nd_nam)->sin_addr.s_addr;
525 nsd->nsd_authlen = nfsd->nfsd_authlen;
526 nsd->nsd_verflen = nfsd->nfsd_verflen;
527 if (!copyout(nfsd->nfsd_authstr,nsd->nsd_authstr,
528 nfsd->nfsd_authlen) &&
529 !copyout(nfsd->nfsd_verfstr, nsd->nsd_verfstr,
530 nfsd->nfsd_verflen) &&
531 !copyout((caddr_t)nsd, argp, sizeof (*nsd)))
533 cacherep = RC_DROPIT;
535 cacherep = nfsrv_getcache(nd, slp, &mreq);
538 * Check for just starting up for NQNFS and send
539 * fake "try again later" replies to the NQNFS clients.
541 if (notstarted && nqnfsstarttime <= time_second) {
543 nqnfsstarttime = time_second + nqsrv_writeslack;
549 if ((nd->nd_flag & ND_NQNFS) == 0)
550 cacherep = RC_DROPIT;
551 else if (nd->nd_procnum != NFSPROC_WRITE) {
552 nd->nd_procnum = NFSPROC_NOOP;
553 nd->nd_repstat = NQNFS_TRYLATER;
557 } else if (nfsd->nfsd_flag & NFSD_AUTHFAIL) {
558 nfsd->nfsd_flag &= ~NFSD_AUTHFAIL;
559 nd->nd_procnum = NFSPROC_NOOP;
560 nd->nd_repstat = (NFSERR_AUTHERR | AUTH_TOOWEAK);
562 } else if (nfs_privport) {
563 /* Check if source port is privileged */
565 struct sockaddr *nam = nd->nd_nam;
566 struct sockaddr_in *sin;
568 sin = (struct sockaddr_in *)nam;
569 port = ntohs(sin->sin_port);
570 if (port >= IPPORT_RESERVED &&
571 nd->nd_procnum != NFSPROC_NULL) {
572 nd->nd_procnum = NFSPROC_NOOP;
573 nd->nd_repstat = (NFSERR_AUTHERR | AUTH_TOOWEAK);
575 printf("NFS request from unprivileged port (%s:%d)\n",
576 inet_ntoa(sin->sin_addr), port);
583 * Loop to get all the write rpc relies that have been
589 if (nd && (nd->nd_flag & ND_NFSV3))
590 procrastinate = nfsrvw_procrastinate_v3;
592 procrastinate = nfsrvw_procrastinate;
593 if (writes_todo || (nd->nd_procnum == NFSPROC_WRITE &&
594 procrastinate > 0 && !notstarted))
595 error = nfsrv_writegather(&nd, slp,
596 nfsd->nfsd_procp, &mreq);
598 error = (*(nfsrv3_procs[nd->nd_procnum]))(nd,
599 slp, nfsd->nfsd_procp, &mreq);
602 if (error != 0 && error != NFSERR_RETVOID) {
603 if (nd->nd_procnum != NQNFSPROC_VACATED)
605 nfsrv_updatecache(nd, FALSE, mreq);
607 FREE(nd->nd_nam2, M_SONAME);
610 nfsstats.srvrpccnt[nd->nd_procnum]++;
611 nfsrv_updatecache(nd, TRUE, mreq);
612 nd->nd_mrep = (struct mbuf *)0;
620 if (siz <= 0 || siz > NFS_MAXPACKET) {
621 printf("mbuf siz=%d\n",siz);
622 panic("Bad nfs svc reply");
625 m->m_pkthdr.len = siz;
626 m->m_pkthdr.rcvif = (struct ifnet *)0;
628 * For stream protocols, prepend a Sun RPC
631 if (sotype == SOCK_STREAM) {
632 M_PREPEND(m, NFSX_UNSIGNED, M_WAIT);
633 *mtod(m, u_int32_t *) = htonl(0x80000000 | siz);
635 if (slp->ns_so->so_proto->pr_flags & PR_CONNREQUIRED)
636 (void) nfs_slplock(slp, 1);
637 if (slp->ns_flag & SLP_VALID)
638 error = nfs_send(slp->ns_so, nd->nd_nam2, m, NULL);
644 nfsd_rt(sotype, nd, cacherep);
646 FREE(nd->nd_nam2, M_SONAME);
648 m_freem(nd->nd_mrep);
651 if (slp->ns_so->so_proto->pr_flags & PR_CONNREQUIRED)
653 if (error == EINTR || error == ERESTART) {
654 free((caddr_t)nd, M_NFSRVDESC);
662 nfsd_rt(sotype, nd, cacherep);
663 m_freem(nd->nd_mrep);
665 FREE(nd->nd_nam2, M_SONAME);
669 FREE((caddr_t)nd, M_NFSRVDESC);
674 * Check to see if there are outstanding writes that
675 * need to be serviced.
677 cur_usec = nfs_curusec();
679 if (slp->ns_tq.lh_first &&
680 slp->ns_tq.lh_first->nd_time <= cur_usec) {
686 } while (writes_todo);
688 if (nfsrv_dorec(slp, nfsd, &nd)) {
689 nfsd->nfsd_flag &= ~NFSD_REQINPROG;
690 nfsd->nfsd_slp = NULL;
695 TAILQ_REMOVE(&nfsd_head, nfsd, nfsd_chain);
697 free((caddr_t)nfsd, M_NFSD);
698 nsd->nsd_nfsd = (struct nfsd *)0;
699 if (--nfs_numnfsd == 0)
700 nfsrv_init(TRUE); /* Reinitialize everything */
705 * Shut down a socket associated with an nfssvc_sock structure.
706 * Should be called with the send lock set, if required.
707 * The trick here is to increment the sref at the start, so that the nfsds
708 * will stop using it and clear ns_flag at the end so that it will not be
709 * reassigned during cleanup.
713 register struct nfssvc_sock *slp;
715 register struct nfsuid *nuidp, *nnuidp;
716 register struct nfsrv_descript *nwp, *nnwp;
719 struct nfsrv_rec *rec;
722 slp->ns_flag &= ~SLP_ALLFLAGS;
725 slp->ns_fp = (struct file *)0;
727 so->so_rcv.sb_flags &= ~SB_UPCALL;
728 so->so_upcall = NULL;
729 so->so_upcallarg = NULL;
731 closef(fp, (struct proc *)0);
733 FREE(slp->ns_nam, M_SONAME);
734 m_freem(slp->ns_raw);
735 while ((rec = STAILQ_FIRST(&slp->ns_rec)) != NULL) {
736 STAILQ_REMOVE_HEAD(&slp->ns_rec, nr_link);
738 FREE(rec->nr_address, M_SONAME);
739 m_freem(rec->nr_packet);
740 free(rec, M_NFSRVDESC);
742 for (nuidp = slp->ns_uidlruhead.tqh_first; nuidp != 0;
744 nnuidp = nuidp->nu_lru.tqe_next;
745 LIST_REMOVE(nuidp, nu_hash);
746 TAILQ_REMOVE(&slp->ns_uidlruhead, nuidp, nu_lru);
747 if (nuidp->nu_flag & NU_NAM)
748 FREE(nuidp->nu_nam, M_SONAME);
749 free((caddr_t)nuidp, M_NFSUID);
752 for (nwp = slp->ns_tq.lh_first; nwp; nwp = nnwp) {
753 nnwp = nwp->nd_tq.le_next;
754 LIST_REMOVE(nwp, nd_tq);
755 free((caddr_t)nwp, M_NFSRVDESC);
757 LIST_INIT(&slp->ns_tq);
763 * Derefence a server socket structure. If it has no more references and
764 * is no longer valid, you can throw it away.
768 register struct nfssvc_sock *slp;
770 if (--(slp->ns_sref) == 0 && (slp->ns_flag & SLP_VALID) == 0) {
771 TAILQ_REMOVE(&nfssvc_sockhead, slp, ns_chain);
772 free((caddr_t)slp, M_NFSSVC);
777 * Lock a socket against others.
780 nfs_slplock(slp, wait)
781 register struct nfssvc_sock *slp;
784 int *statep = &slp->ns_solock;
786 if (!wait && (*statep & NFSSTA_SNDLOCK))
787 return(0); /* already locked, fail */
788 while (*statep & NFSSTA_SNDLOCK) {
789 *statep |= NFSSTA_WANTSND;
790 (void) tsleep((caddr_t)statep, PZERO - 1, "nfsslplck", 0);
792 *statep |= NFSSTA_SNDLOCK;
797 * Unlock the stream socket for others.
801 register struct nfssvc_sock *slp;
803 int *statep = &slp->ns_solock;
805 if ((*statep & NFSSTA_SNDLOCK) == 0)
806 panic("nfs slpunlock");
807 *statep &= ~NFSSTA_SNDLOCK;
808 if (*statep & NFSSTA_WANTSND) {
809 *statep &= ~NFSSTA_WANTSND;
810 wakeup((caddr_t)statep);
815 * Initialize the data structures for the server.
816 * Handshake with any new nfsds starting up to avoid any chance of
820 nfsrv_init(terminating)
823 register struct nfssvc_sock *slp, *nslp;
825 if (nfssvc_sockhead_flag & SLP_INIT)
827 nfssvc_sockhead_flag |= SLP_INIT;
829 for (slp = nfssvc_sockhead.tqh_first; slp != 0; slp = nslp) {
830 nslp = slp->ns_chain.tqe_next;
831 if (slp->ns_flag & SLP_VALID)
833 TAILQ_REMOVE(&nfssvc_sockhead, slp, ns_chain);
834 free((caddr_t)slp, M_NFSSVC);
836 nfsrv_cleancache(); /* And clear out server cache */
838 nfs_pub.np_valid = 0;
840 TAILQ_INIT(&nfssvc_sockhead);
841 nfssvc_sockhead_flag &= ~SLP_INIT;
842 if (nfssvc_sockhead_flag & SLP_WANTINIT) {
843 nfssvc_sockhead_flag &= ~SLP_WANTINIT;
844 wakeup((caddr_t)&nfssvc_sockhead);
847 TAILQ_INIT(&nfsd_head);
848 nfsd_head_flag &= ~NFSD_CHECKSLP;
851 nfs_udpsock = (struct nfssvc_sock *)
852 malloc(sizeof (struct nfssvc_sock), M_NFSSVC, M_WAITOK);
853 bzero((caddr_t)nfs_udpsock, sizeof (struct nfssvc_sock));
854 STAILQ_INIT(&nfs_udpsock->ns_rec);
855 TAILQ_INIT(&nfs_udpsock->ns_uidlruhead);
856 TAILQ_INSERT_HEAD(&nfssvc_sockhead, nfs_udpsock, ns_chain);
858 nfs_cltpsock = (struct nfssvc_sock *)
859 malloc(sizeof (struct nfssvc_sock), M_NFSSVC, M_WAITOK);
860 bzero((caddr_t)nfs_cltpsock, sizeof (struct nfssvc_sock));
861 STAILQ_INIT(&nfs_cltpsock->ns_rec);
862 TAILQ_INIT(&nfs_cltpsock->ns_uidlruhead);
863 TAILQ_INSERT_TAIL(&nfssvc_sockhead, nfs_cltpsock, ns_chain);
868 * Add entries to the server monitor log.
871 nfsd_rt(sotype, nd, cacherep)
873 register struct nfsrv_descript *nd;
876 register struct drt *rt;
878 rt = &nfsdrt.drt[nfsdrt.pos];
879 if (cacherep == RC_DOIT)
881 else if (cacherep == RC_REPLY)
882 rt->flag = DRT_CACHEREPLY;
884 rt->flag = DRT_CACHEDROP;
885 if (sotype == SOCK_STREAM)
887 if (nd->nd_flag & ND_NQNFS)
888 rt->flag |= DRT_NQNFS;
889 else if (nd->nd_flag & ND_NFSV3)
890 rt->flag |= DRT_NFSV3;
891 rt->proc = nd->nd_procnum;
892 if (nd->nd_nam->sa_family == AF_INET)
893 rt->ipadr = ((struct sockaddr_in *)nd->nd_nam)->sin_addr.s_addr;
895 rt->ipadr = INADDR_ANY;
896 rt->resptime = nfs_curusec() - (nd->nd_starttime.tv_sec * 1000000 + nd->nd_starttime.tv_usec);
897 getmicrotime(&rt->tstamp);
898 nfsdrt.pos = (nfsdrt.pos + 1) % NFSRTTLOGSIZ;
900 #endif /* NFS_NOSERVER */
902 static int nfs_defect = 0;
903 SYSCTL_INT(_vfs_nfs, OID_AUTO, defect, CTLFLAG_RW, &nfs_defect, 0, "");
906 * Asynchronous I/O daemons for client nfs.
907 * They do read-ahead and write-behind operations on the block I/O cache.
908 * Never returns unless it fails or gets killed.
914 register struct buf *bp;
915 register int i, myiod;
916 struct nfsmount *nmp;
920 * Assign my position or return error if too many already running
923 for (i = 0; i < NFS_MAXASYNCDAEMON; i++)
924 if (nfs_asyncdaemon[i] == 0) {
925 nfs_asyncdaemon[i]++;
933 * Just loop around doin our stuff until SIGKILL
936 while (((nmp = nfs_iodmount[myiod]) == NULL
937 || nmp->nm_bufq.tqh_first == NULL)
941 nfs_iodwant[myiod] = p;
942 nfs_iodmount[myiod] = NULL;
943 error = tsleep((caddr_t)&nfs_iodwant[myiod],
944 PWAIT | PCATCH, "nfsidl", 0);
947 nfs_asyncdaemon[myiod] = 0;
950 nfs_iodwant[myiod] = NULL;
951 nfs_iodmount[myiod] = NULL;
955 while ((bp = nmp->nm_bufq.tqh_first) != NULL) {
956 /* Take one off the front of the list */
957 TAILQ_REMOVE(&nmp->nm_bufq, bp, b_freelist);
959 if (nmp->nm_bufqwant && nmp->nm_bufqlen <= nfs_numasync) {
960 nmp->nm_bufqwant = FALSE;
961 wakeup(&nmp->nm_bufq);
963 if (bp->b_flags & B_READ)
964 (void) nfs_doio(bp, bp->b_rcred, (struct proc *)0);
966 (void) nfs_doio(bp, bp->b_wcred, (struct proc *)0);
968 * If there are more than one iod on this mount, then defect
969 * so that the iods can be shared out fairly between the mounts
971 if (nfs_defect && nmp->nm_bufqiods > 1) {
973 ("nfssvc_iod: iod %d defecting from mount %p\n",
975 nfs_iodmount[myiod] = NULL;
985 * Get an authorization string for the uid by having the mount_nfs sitting
986 * on this mount point porpous out of the kernel and do it.
989 nfs_getauth(nmp, rep, cred, auth_str, auth_len, verf_str, verf_len, key)
990 register struct nfsmount *nmp;
997 NFSKERBKEY_T key; /* return session key */
1001 while ((nmp->nm_state & NFSSTA_WAITAUTH) == 0) {
1002 nmp->nm_state |= NFSSTA_WANTAUTH;
1003 (void) tsleep((caddr_t)&nmp->nm_authtype, PSOCK,
1004 "nfsauth1", 2 * hz);
1005 error = nfs_sigintr(nmp, rep, rep->r_procp);
1007 nmp->nm_state &= ~NFSSTA_WANTAUTH;
1011 nmp->nm_state &= ~(NFSSTA_WAITAUTH | NFSSTA_WANTAUTH);
1012 nmp->nm_authstr = *auth_str = (char *)malloc(RPCAUTH_MAXSIZ, M_TEMP, M_WAITOK);
1013 nmp->nm_authlen = RPCAUTH_MAXSIZ;
1014 nmp->nm_verfstr = verf_str;
1015 nmp->nm_verflen = *verf_len;
1016 nmp->nm_authuid = cred->cr_uid;
1017 wakeup((caddr_t)&nmp->nm_authstr);
1020 * And wait for mount_nfs to do its stuff.
1022 while ((nmp->nm_state & NFSSTA_HASAUTH) == 0 && error == 0) {
1023 (void) tsleep((caddr_t)&nmp->nm_authlen, PSOCK,
1024 "nfsauth2", 2 * hz);
1025 error = nfs_sigintr(nmp, rep, rep->r_procp);
1027 if (nmp->nm_state & NFSSTA_AUTHERR) {
1028 nmp->nm_state &= ~NFSSTA_AUTHERR;
1032 free((caddr_t)*auth_str, M_TEMP);
1034 *auth_len = nmp->nm_authlen;
1035 *verf_len = nmp->nm_verflen;
1036 bcopy((caddr_t)nmp->nm_key, (caddr_t)key, sizeof (key));
1038 nmp->nm_state &= ~NFSSTA_HASAUTH;
1039 nmp->nm_state |= NFSSTA_WAITAUTH;
1040 if (nmp->nm_state & NFSSTA_WANTAUTH) {
1041 nmp->nm_state &= ~NFSSTA_WANTAUTH;
1042 wakeup((caddr_t)&nmp->nm_authtype);
1048 * Get a nickname authenticator and verifier.
1051 nfs_getnickauth(nmp, cred, auth_str, auth_len, verf_str, verf_len)
1052 struct nfsmount *nmp;
1059 register struct nfsuid *nuidp;
1060 register u_int32_t *nickp, *verfp;
1061 struct timeval ktvin, ktvout;
1064 if (verf_len < (4 * NFSX_UNSIGNED))
1065 panic("nfs_getnickauth verf too small");
1067 for (nuidp = NMUIDHASH(nmp, cred->cr_uid)->lh_first;
1068 nuidp != 0; nuidp = nuidp->nu_hash.le_next) {
1069 if (nuidp->nu_cr.cr_uid == cred->cr_uid)
1072 if (!nuidp || nuidp->nu_expire < time_second)
1076 * Move to the end of the lru list (end of lru == most recently used).
1078 TAILQ_REMOVE(&nmp->nm_uidlruhead, nuidp, nu_lru);
1079 TAILQ_INSERT_TAIL(&nmp->nm_uidlruhead, nuidp, nu_lru);
1081 nickp = (u_int32_t *)malloc(2 * NFSX_UNSIGNED, M_TEMP, M_WAITOK);
1082 *nickp++ = txdr_unsigned(RPCAKN_NICKNAME);
1083 *nickp = txdr_unsigned(nuidp->nu_nickname);
1084 *auth_str = (char *)nickp;
1085 *auth_len = 2 * NFSX_UNSIGNED;
1088 * Now we must encrypt the verifier and package it up.
1090 verfp = (u_int32_t *)verf_str;
1091 *verfp++ = txdr_unsigned(RPCAKN_NICKNAME);
1092 if (time_second > nuidp->nu_timestamp.tv_sec ||
1093 (time_second == nuidp->nu_timestamp.tv_sec &&
1094 time_second > nuidp->nu_timestamp.tv_usec))
1095 getmicrotime(&nuidp->nu_timestamp);
1097 nuidp->nu_timestamp.tv_usec++;
1098 ktvin.tv_sec = txdr_unsigned(nuidp->nu_timestamp.tv_sec);
1099 ktvin.tv_usec = txdr_unsigned(nuidp->nu_timestamp.tv_usec);
1102 * Now encrypt the timestamp verifier in ecb mode using the session
1109 *verfp++ = ktvout.tv_sec;
1110 *verfp++ = ktvout.tv_usec;
1116 * Save the current nickname in a hash list entry on the mount point.
1119 nfs_savenickauth(nmp, cred, len, key, mdp, dposp, mrep)
1120 register struct nfsmount *nmp;
1128 register struct nfsuid *nuidp;
1129 register u_int32_t *tl;
1130 register int32_t t1;
1131 struct mbuf *md = *mdp;
1132 struct timeval ktvin, ktvout;
1134 char *dpos = *dposp, *cp2;
1135 int deltasec, error = 0;
1137 if (len == (3 * NFSX_UNSIGNED)) {
1138 nfsm_dissect(tl, u_int32_t *, 3 * NFSX_UNSIGNED);
1139 ktvin.tv_sec = *tl++;
1140 ktvin.tv_usec = *tl++;
1141 nick = fxdr_unsigned(u_int32_t, *tl);
1144 * Decrypt the timestamp in ecb mode.
1149 ktvout.tv_sec = fxdr_unsigned(long, ktvout.tv_sec);
1150 ktvout.tv_usec = fxdr_unsigned(long, ktvout.tv_usec);
1151 deltasec = time_second - ktvout.tv_sec;
1153 deltasec = -deltasec;
1155 * If ok, add it to the hash list for the mount point.
1157 if (deltasec <= NFS_KERBCLOCKSKEW) {
1158 if (nmp->nm_numuids < nuidhash_max) {
1160 nuidp = (struct nfsuid *)
1161 malloc(sizeof (struct nfsuid), M_NFSUID,
1164 nuidp = nmp->nm_uidlruhead.tqh_first;
1165 LIST_REMOVE(nuidp, nu_hash);
1166 TAILQ_REMOVE(&nmp->nm_uidlruhead, nuidp,
1170 nuidp->nu_cr.cr_uid = cred->cr_uid;
1171 nuidp->nu_expire = time_second + NFS_KERBTTL;
1172 nuidp->nu_timestamp = ktvout;
1173 nuidp->nu_nickname = nick;
1174 bcopy(key, nuidp->nu_key, sizeof (key));
1175 TAILQ_INSERT_TAIL(&nmp->nm_uidlruhead, nuidp,
1177 LIST_INSERT_HEAD(NMUIDHASH(nmp, cred->cr_uid),
1181 nfsm_adv(nfsm_rndup(len));