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 1.12 2004/03/05 16:57:16 hsu 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/resourcevar.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 <vm/vm_zone.h>
62 #include <netinet/in.h>
63 #include <netinet/tcp.h>
68 #include "nfsm_subs.h"
69 #include "nfsrvcache.h"
75 static MALLOC_DEFINE(M_NFSSVC, "NFS srvsock", "Nfs server structure");
78 extern int32_t (*nfsrv3_procs[NFS_NPROCS]) (struct nfsrv_descript *nd,
79 struct nfssvc_sock *slp,
82 extern int nfs_numasync;
83 extern time_t nqnfsstarttime;
84 extern int nqsrv_writeslack;
86 extern struct nfsstats nfsstats;
87 extern int nfsrvw_procrastinate;
88 extern int nfsrvw_procrastinate_v3;
89 static int nuidhash_max = NFS_MAXUIDHASH;
92 static void nfsrv_zapsock (struct nfssvc_sock *slp);
94 static int nfssvc_iod (struct thread *);
99 static int nfs_asyncdaemon[NFS_MAXASYNCDAEMON];
101 SYSCTL_DECL(_vfs_nfs);
104 int nfsd_waiting = 0;
105 static struct nfsdrt nfsdrt;
106 static int nfs_numnfsd = 0;
107 static int notstarted = 1;
108 static int modify_flag = 0;
109 static void nfsd_rt (int sotype, struct nfsrv_descript *nd,
111 static int nfssvc_addsock (struct file *, struct sockaddr *,
113 static int nfssvc_nfsd (struct nfsd_srvargs *,caddr_t,struct thread *);
115 static int nfs_privport = 0;
116 SYSCTL_INT(_vfs_nfs, NFS_NFSPRIVPORT, nfs_privport, CTLFLAG_RW, &nfs_privport, 0, "");
117 SYSCTL_INT(_vfs_nfs, OID_AUTO, gatherdelay, CTLFLAG_RW, &nfsrvw_procrastinate, 0, "");
118 SYSCTL_INT(_vfs_nfs, OID_AUTO, gatherdelay_v3, CTLFLAG_RW, &nfsrvw_procrastinate_v3, 0, "");
121 * NFS server system calls
124 #endif /* NFS_NOSERVER */
126 * nfssvc_args(int flag, caddr_t argp)
128 * Nfs server psuedo system call for the nfsd's
129 * Based on the flag value it either:
130 * - adds a socket to the selection list
131 * - remains in the kernel as an nfsd
132 * - remains in the kernel as an nfsiod
135 nfssvc(struct nfssvc_args *uap)
140 struct sockaddr *nam;
141 struct nfsd_args nfsdarg;
142 struct nfsd_srvargs nfsd_srvargs, *nsd = &nfsd_srvargs;
143 struct nfsd_cargs ncd;
145 struct nfssvc_sock *slp;
146 struct nfsuid *nuidp;
147 struct nfsmount *nmp;
148 #endif /* NFS_NOSERVER */
150 struct thread *td = curthread;
158 KKASSERT(td->td_proc); /* for ucred and p_fd */
159 while (nfssvc_sockhead_flag & SLP_INIT) {
160 nfssvc_sockhead_flag |= SLP_WANTINIT;
161 (void) tsleep((caddr_t)&nfssvc_sockhead, 0, "nfsd init", 0);
163 if (uap->flag & NFSSVC_BIOD)
164 error = nfssvc_iod(td);
168 #else /* !NFS_NOSERVER */
169 else if (uap->flag & NFSSVC_MNTD) {
170 error = copyin(uap->argp, (caddr_t)&ncd, sizeof (ncd));
173 NDINIT(&nd, NAMEI_LOOKUP, CNP_FOLLOW | CNP_LOCKLEAF,
174 UIO_USERSPACE, ncd.ncd_dirp, td);
178 NDFREE(&nd, NDF_ONLY_PNBUF);
179 if ((nd.ni_vp->v_flag & VROOT) == 0)
181 nmp = VFSTONFS(nd.ni_vp->v_mount);
185 if ((nmp->nm_state & NFSSTA_MNTD) &&
186 (uap->flag & NFSSVC_GOTAUTH) == 0)
188 nmp->nm_state |= NFSSTA_MNTD;
189 error = nqnfs_clientd(nmp, td->td_proc->p_ucred, &ncd, uap->flag,
191 } else if (uap->flag & NFSSVC_ADDSOCK) {
192 error = copyin(uap->argp, (caddr_t)&nfsdarg, sizeof(nfsdarg));
195 error = holdsock(td->td_proc->p_fd, nfsdarg.sock, &fp);
199 * Get the client address for connected sockets.
201 if (nfsdarg.name == NULL || nfsdarg.namelen == 0)
202 nam = (struct sockaddr *)0;
204 error = getsockaddr(&nam, nfsdarg.name,
211 error = nfssvc_addsock(fp, nam, td);
214 error = copyin(uap->argp, (caddr_t)nsd, sizeof (*nsd));
217 if ((uap->flag & NFSSVC_AUTHIN) &&
218 ((nfsd = nsd->nsd_nfsd)) != NULL &&
219 (nfsd->nfsd_slp->ns_flag & SLP_VALID)) {
220 slp = nfsd->nfsd_slp;
223 * First check to see if another nfsd has already
224 * added this credential.
226 for (nuidp = NUIDHASH(slp,nsd->nsd_cr.cr_uid)->lh_first;
227 nuidp != 0; nuidp = nuidp->nu_hash.le_next) {
228 if (nuidp->nu_cr.cr_uid == nsd->nsd_cr.cr_uid &&
229 (!nfsd->nfsd_nd->nd_nam2 ||
230 netaddr_match(NU_NETFAM(nuidp),
231 &nuidp->nu_haddr, nfsd->nfsd_nd->nd_nam2)))
235 nfsrv_setcred(&nuidp->nu_cr,&nfsd->nfsd_nd->nd_cr);
236 nfsd->nfsd_nd->nd_flag |= ND_KERBFULL;
241 if (slp->ns_numuids < nuidhash_max) {
243 nuidp = (struct nfsuid *)
244 malloc(sizeof (struct nfsuid), M_NFSUID,
247 nuidp = (struct nfsuid *)0;
248 if ((slp->ns_flag & SLP_VALID) == 0) {
250 free((caddr_t)nuidp, M_NFSUID);
252 if (nuidp == (struct nfsuid *)0) {
253 nuidp = slp->ns_uidlruhead.tqh_first;
254 LIST_REMOVE(nuidp, nu_hash);
255 TAILQ_REMOVE(&slp->ns_uidlruhead, nuidp,
257 if (nuidp->nu_flag & NU_NAM)
258 FREE(nuidp->nu_nam, M_SONAME);
261 nuidp->nu_cr = nsd->nsd_cr;
262 if (nuidp->nu_cr.cr_ngroups > NGROUPS)
263 nuidp->nu_cr.cr_ngroups = NGROUPS;
264 nuidp->nu_cr.cr_ref = 1;
265 nuidp->nu_timestamp = nsd->nsd_timestamp;
266 nuidp->nu_expire = time_second + nsd->nsd_ttl;
268 * and save the session key in nu_key.
270 bcopy(nsd->nsd_key, nuidp->nu_key,
271 sizeof (nsd->nsd_key));
272 if (nfsd->nfsd_nd->nd_nam2) {
273 struct sockaddr_in *saddr;
275 saddr = (struct sockaddr_in *)
276 nfsd->nfsd_nd->nd_nam2;
277 switch (saddr->sin_family) {
279 nuidp->nu_flag |= NU_INETADDR;
281 saddr->sin_addr.s_addr;
285 nuidp->nu_flag |= NU_NAM;
287 dup_sockaddr(nfsd->nfsd_nd->
292 TAILQ_INSERT_TAIL(&slp->ns_uidlruhead, nuidp,
294 LIST_INSERT_HEAD(NUIDHASH(slp, nsd->nsd_uid),
296 nfsrv_setcred(&nuidp->nu_cr,
297 &nfsd->nfsd_nd->nd_cr);
298 nfsd->nfsd_nd->nd_flag |= ND_KERBFULL;
302 if ((uap->flag & NFSSVC_AUTHINFAIL) && (nfsd = nsd->nsd_nfsd))
303 nfsd->nfsd_flag |= NFSD_AUTHFAIL;
304 error = nfssvc_nfsd(nsd, uap->argp, td);
306 #endif /* NFS_NOSERVER */
307 if (error == EINTR || error == ERESTART)
314 * Adds a socket to the list for servicing by nfsds.
317 nfssvc_addsock(struct file *fp, struct sockaddr *mynam, struct thread *td)
320 struct nfssvc_sock *slp;
324 so = (struct socket *)fp->f_data;
326 tslp = (struct nfssvc_sock *)0;
328 * Add it to the list, as required.
330 if (so->so_proto->pr_protocol == IPPROTO_UDP) {
332 if (tslp->ns_flag & SLP_VALID) {
334 FREE(mynam, M_SONAME);
339 if (so->so_type == SOCK_STREAM)
340 siz = NFS_MAXPACKET + sizeof (u_long);
343 error = soreserve(so, siz, siz, &td->td_proc->p_rlimit[RLIMIT_SBSIZE]);
346 FREE(mynam, M_SONAME);
351 * Set protocol specific options { for now TCP only } and
352 * reserve some space. For datagram sockets, this can get called
353 * repeatedly for the same socket, but that isn't harmful.
355 if (so->so_type == SOCK_STREAM) {
359 bzero(&sopt, sizeof sopt);
360 sopt.sopt_level = SOL_SOCKET;
361 sopt.sopt_name = SO_KEEPALIVE;
362 sopt.sopt_val = &val;
363 sopt.sopt_valsize = sizeof val;
367 if (so->so_proto->pr_domain->dom_family == AF_INET &&
368 so->so_proto->pr_protocol == IPPROTO_TCP) {
372 bzero(&sopt, sizeof sopt);
373 sopt.sopt_level = IPPROTO_TCP;
374 sopt.sopt_name = TCP_NODELAY;
375 sopt.sopt_val = &val;
376 sopt.sopt_valsize = sizeof val;
380 so->so_rcv.sb_flags &= ~SB_NOINTR;
381 so->so_rcv.sb_timeo = 0;
382 so->so_snd.sb_flags &= ~SB_NOINTR;
383 so->so_snd.sb_timeo = 0;
385 slp = (struct nfssvc_sock *)
386 malloc(sizeof (struct nfssvc_sock), M_NFSSVC, M_WAITOK);
387 bzero((caddr_t)slp, sizeof (struct nfssvc_sock));
388 STAILQ_INIT(&slp->ns_rec);
389 TAILQ_INIT(&slp->ns_uidlruhead);
390 TAILQ_INSERT_TAIL(&nfssvc_sockhead, slp, ns_chain);
397 so->so_upcallarg = (caddr_t)slp;
398 so->so_upcall = nfsrv_rcv;
399 so->so_rcv.sb_flags |= SB_UPCALL;
400 slp->ns_flag = (SLP_VALID | SLP_NEEDQ);
407 * Called by nfssvc() for nfsds. Just loops around servicing rpc requests
408 * until it is killed by a signal.
411 nfssvc_nfsd(struct nfsd_srvargs *nsd, caddr_t argp, struct thread *td)
414 struct nfssvc_sock *slp;
415 struct nfsd *nfsd = nsd->nsd_nfsd;
416 struct nfsrv_descript *nd = NULL;
417 struct mbuf *m, *mreq;
418 int error = 0, cacherep, s, sotype, writes_todo;
426 if (nfsd == (struct nfsd *)0) {
427 nsd->nsd_nfsd = nfsd = (struct nfsd *)
428 malloc(sizeof (struct nfsd), M_NFSD, M_WAITOK);
429 bzero((caddr_t)nfsd, sizeof (struct nfsd));
432 TAILQ_INSERT_TAIL(&nfsd_head, nfsd, nfsd_chain);
438 * Loop getting rpc requests until SIGKILL.
441 if ((nfsd->nfsd_flag & NFSD_REQINPROG) == 0) {
442 while (nfsd->nfsd_slp == (struct nfssvc_sock *)0 &&
443 (nfsd_head_flag & NFSD_CHECKSLP) == 0) {
444 nfsd->nfsd_flag |= NFSD_WAITING;
446 error = tsleep((caddr_t)nfsd, PCATCH, "nfsd", 0);
451 if (nfsd->nfsd_slp == (struct nfssvc_sock *)0 &&
452 (nfsd_head_flag & NFSD_CHECKSLP) != 0) {
453 for (slp = nfssvc_sockhead.tqh_first; slp != 0;
454 slp = slp->ns_chain.tqe_next) {
455 if ((slp->ns_flag & (SLP_VALID | SLP_DOREC))
456 == (SLP_VALID | SLP_DOREC)) {
457 slp->ns_flag &= ~SLP_DOREC;
459 nfsd->nfsd_slp = slp;
464 nfsd_head_flag &= ~NFSD_CHECKSLP;
466 if ((slp = nfsd->nfsd_slp) == (struct nfssvc_sock *)0)
468 if (slp->ns_flag & SLP_VALID) {
469 if (slp->ns_flag & SLP_DISCONN)
471 else if (slp->ns_flag & SLP_NEEDQ) {
472 slp->ns_flag &= ~SLP_NEEDQ;
473 (void) nfs_slplock(slp, 1);
474 nfsrv_rcv(slp->ns_so, (caddr_t)slp,
478 error = nfsrv_dorec(slp, nfsd, &nd);
479 cur_usec = nfs_curusec();
480 if (error && slp->ns_tq.lh_first &&
481 slp->ns_tq.lh_first->nd_time <= cur_usec) {
487 nfsd->nfsd_flag |= NFSD_REQINPROG;
491 slp = nfsd->nfsd_slp;
493 if (error || (slp->ns_flag & SLP_VALID) == 0) {
495 free((caddr_t)nd, M_NFSRVDESC);
498 nfsd->nfsd_slp = (struct nfssvc_sock *)0;
499 nfsd->nfsd_flag &= ~NFSD_REQINPROG;
504 sotype = slp->ns_so->so_type;
506 getmicrotime(&nd->nd_starttime);
508 nd->nd_nam = nd->nd_nam2;
510 nd->nd_nam = slp->ns_nam;
513 * Check to see if authorization is needed.
515 if (nfsd->nfsd_flag & NFSD_NEEDAUTH) {
516 nfsd->nfsd_flag &= ~NFSD_NEEDAUTH;
518 ((struct sockaddr_in *)
519 nd->nd_nam)->sin_addr.s_addr;
520 nsd->nsd_authlen = nfsd->nfsd_authlen;
521 nsd->nsd_verflen = nfsd->nfsd_verflen;
522 if (!copyout(nfsd->nfsd_authstr,nsd->nsd_authstr,
523 nfsd->nfsd_authlen) &&
524 !copyout(nfsd->nfsd_verfstr, nsd->nsd_verfstr,
525 nfsd->nfsd_verflen) &&
526 !copyout((caddr_t)nsd, argp, sizeof (*nsd)))
528 cacherep = RC_DROPIT;
530 cacherep = nfsrv_getcache(nd, slp, &mreq);
533 * Check for just starting up for NQNFS and send
534 * fake "try again later" replies to the NQNFS clients.
536 if (notstarted && nqnfsstarttime <= time_second) {
538 nqnfsstarttime = time_second + nqsrv_writeslack;
544 if ((nd->nd_flag & ND_NQNFS) == 0)
545 cacherep = RC_DROPIT;
546 else if (nd->nd_procnum != NFSPROC_WRITE) {
547 nd->nd_procnum = NFSPROC_NOOP;
548 nd->nd_repstat = NQNFS_TRYLATER;
552 } else if (nfsd->nfsd_flag & NFSD_AUTHFAIL) {
553 nfsd->nfsd_flag &= ~NFSD_AUTHFAIL;
554 nd->nd_procnum = NFSPROC_NOOP;
555 nd->nd_repstat = (NFSERR_AUTHERR | AUTH_TOOWEAK);
557 } else if (nfs_privport) {
558 /* Check if source port is privileged */
560 struct sockaddr *nam = nd->nd_nam;
561 struct sockaddr_in *sin;
563 sin = (struct sockaddr_in *)nam;
564 port = ntohs(sin->sin_port);
565 if (port >= IPPORT_RESERVED &&
566 nd->nd_procnum != NFSPROC_NULL) {
567 nd->nd_procnum = NFSPROC_NOOP;
568 nd->nd_repstat = (NFSERR_AUTHERR | AUTH_TOOWEAK);
570 printf("NFS request from unprivileged port (%s:%d)\n",
571 inet_ntoa(sin->sin_addr), port);
578 * Loop to get all the write rpc relies that have been
584 if (nd && (nd->nd_flag & ND_NFSV3))
585 procrastinate = nfsrvw_procrastinate_v3;
587 procrastinate = nfsrvw_procrastinate;
588 if (writes_todo || (nd->nd_procnum == NFSPROC_WRITE &&
589 procrastinate > 0 && !notstarted))
590 error = nfsrv_writegather(&nd, slp,
591 nfsd->nfsd_td, &mreq);
593 error = (*(nfsrv3_procs[nd->nd_procnum]))(nd,
594 slp, nfsd->nfsd_td, &mreq);
597 if (error != 0 && error != NFSERR_RETVOID) {
598 if (nd->nd_procnum != NQNFSPROC_VACATED)
600 nfsrv_updatecache(nd, FALSE, mreq);
602 FREE(nd->nd_nam2, M_SONAME);
605 nfsstats.srvrpccnt[nd->nd_procnum]++;
606 nfsrv_updatecache(nd, TRUE, mreq);
607 nd->nd_mrep = (struct mbuf *)0;
615 if (siz <= 0 || siz > NFS_MAXPACKET) {
616 printf("mbuf siz=%d\n",siz);
617 panic("Bad nfs svc reply");
620 m->m_pkthdr.len = siz;
621 m->m_pkthdr.rcvif = (struct ifnet *)0;
623 * For stream protocols, prepend a Sun RPC
626 if (sotype == SOCK_STREAM) {
627 M_PREPEND(m, NFSX_UNSIGNED, M_WAIT);
630 *mtod(m, u_int32_t *) = htonl(0x80000000 | siz);
632 if (slp->ns_so->so_proto->pr_flags & PR_CONNREQUIRED)
633 (void) nfs_slplock(slp, 1);
634 if (slp->ns_flag & SLP_VALID)
635 error = nfs_send(slp->ns_so, nd->nd_nam2, m, NULL);
641 nfsd_rt(sotype, nd, cacherep);
643 FREE(nd->nd_nam2, M_SONAME);
645 m_freem(nd->nd_mrep);
648 if (slp->ns_so->so_proto->pr_flags & PR_CONNREQUIRED)
650 if (error == EINTR || error == ERESTART) {
651 free((caddr_t)nd, M_NFSRVDESC);
659 nfsd_rt(sotype, nd, cacherep);
660 m_freem(nd->nd_mrep);
662 FREE(nd->nd_nam2, M_SONAME);
666 FREE((caddr_t)nd, M_NFSRVDESC);
671 * Check to see if there are outstanding writes that
672 * need to be serviced.
674 cur_usec = nfs_curusec();
676 if (slp->ns_tq.lh_first &&
677 slp->ns_tq.lh_first->nd_time <= cur_usec) {
683 } while (writes_todo);
685 if (nfsrv_dorec(slp, nfsd, &nd)) {
686 nfsd->nfsd_flag &= ~NFSD_REQINPROG;
687 nfsd->nfsd_slp = NULL;
692 TAILQ_REMOVE(&nfsd_head, nfsd, nfsd_chain);
694 free((caddr_t)nfsd, M_NFSD);
695 nsd->nsd_nfsd = (struct nfsd *)0;
696 if (--nfs_numnfsd == 0)
697 nfsrv_init(TRUE); /* Reinitialize everything */
702 * Shut down a socket associated with an nfssvc_sock structure.
703 * Should be called with the send lock set, if required.
704 * The trick here is to increment the sref at the start, so that the nfsds
705 * will stop using it and clear ns_flag at the end so that it will not be
706 * reassigned during cleanup.
710 struct nfssvc_sock *slp;
712 struct nfsuid *nuidp, *nnuidp;
713 struct nfsrv_descript *nwp, *nnwp;
716 struct nfsrv_rec *rec;
719 slp->ns_flag &= ~SLP_ALLFLAGS;
722 slp->ns_fp = (struct file *)0;
724 so->so_rcv.sb_flags &= ~SB_UPCALL;
725 so->so_upcall = NULL;
726 so->so_upcallarg = NULL;
730 FREE(slp->ns_nam, M_SONAME);
731 m_freem(slp->ns_raw);
732 while ((rec = STAILQ_FIRST(&slp->ns_rec)) != NULL) {
733 STAILQ_REMOVE_HEAD(&slp->ns_rec, nr_link);
735 FREE(rec->nr_address, M_SONAME);
736 m_freem(rec->nr_packet);
737 free(rec, M_NFSRVDESC);
739 for (nuidp = slp->ns_uidlruhead.tqh_first; nuidp != 0;
741 nnuidp = nuidp->nu_lru.tqe_next;
742 LIST_REMOVE(nuidp, nu_hash);
743 TAILQ_REMOVE(&slp->ns_uidlruhead, nuidp, nu_lru);
744 if (nuidp->nu_flag & NU_NAM)
745 FREE(nuidp->nu_nam, M_SONAME);
746 free((caddr_t)nuidp, M_NFSUID);
749 for (nwp = slp->ns_tq.lh_first; nwp; nwp = nnwp) {
750 nnwp = nwp->nd_tq.le_next;
751 LIST_REMOVE(nwp, nd_tq);
752 free((caddr_t)nwp, M_NFSRVDESC);
754 LIST_INIT(&slp->ns_tq);
760 * Derefence a server socket structure. If it has no more references and
761 * is no longer valid, you can throw it away.
765 struct nfssvc_sock *slp;
767 if (--(slp->ns_sref) == 0 && (slp->ns_flag & SLP_VALID) == 0) {
768 TAILQ_REMOVE(&nfssvc_sockhead, slp, ns_chain);
769 free((caddr_t)slp, M_NFSSVC);
774 * Lock a socket against others.
777 nfs_slplock(slp, wait)
778 struct nfssvc_sock *slp;
781 int *statep = &slp->ns_solock;
783 if (!wait && (*statep & NFSSTA_SNDLOCK))
784 return(0); /* already locked, fail */
785 while (*statep & NFSSTA_SNDLOCK) {
786 *statep |= NFSSTA_WANTSND;
787 (void) tsleep((caddr_t)statep, 0, "nfsslplck", 0);
789 *statep |= NFSSTA_SNDLOCK;
794 * Unlock the stream socket for others.
798 struct nfssvc_sock *slp;
800 int *statep = &slp->ns_solock;
802 if ((*statep & NFSSTA_SNDLOCK) == 0)
803 panic("nfs slpunlock");
804 *statep &= ~NFSSTA_SNDLOCK;
805 if (*statep & NFSSTA_WANTSND) {
806 *statep &= ~NFSSTA_WANTSND;
807 wakeup((caddr_t)statep);
812 * Initialize the data structures for the server.
813 * Handshake with any new nfsds starting up to avoid any chance of
817 nfsrv_init(terminating)
820 struct nfssvc_sock *slp, *nslp;
822 if (nfssvc_sockhead_flag & SLP_INIT)
824 nfssvc_sockhead_flag |= SLP_INIT;
826 for (slp = nfssvc_sockhead.tqh_first; slp != 0; slp = nslp) {
827 nslp = slp->ns_chain.tqe_next;
828 if (slp->ns_flag & SLP_VALID)
830 TAILQ_REMOVE(&nfssvc_sockhead, slp, ns_chain);
831 free((caddr_t)slp, M_NFSSVC);
833 nfsrv_cleancache(); /* And clear out server cache */
835 nfs_pub.np_valid = 0;
837 TAILQ_INIT(&nfssvc_sockhead);
838 nfssvc_sockhead_flag &= ~SLP_INIT;
839 if (nfssvc_sockhead_flag & SLP_WANTINIT) {
840 nfssvc_sockhead_flag &= ~SLP_WANTINIT;
841 wakeup((caddr_t)&nfssvc_sockhead);
844 TAILQ_INIT(&nfsd_head);
845 nfsd_head_flag &= ~NFSD_CHECKSLP;
848 nfs_udpsock = (struct nfssvc_sock *)
849 malloc(sizeof (struct nfssvc_sock), M_NFSSVC, M_WAITOK);
850 bzero((caddr_t)nfs_udpsock, sizeof (struct nfssvc_sock));
851 STAILQ_INIT(&nfs_udpsock->ns_rec);
852 TAILQ_INIT(&nfs_udpsock->ns_uidlruhead);
853 TAILQ_INSERT_HEAD(&nfssvc_sockhead, nfs_udpsock, ns_chain);
855 nfs_cltpsock = (struct nfssvc_sock *)
856 malloc(sizeof (struct nfssvc_sock), M_NFSSVC, M_WAITOK);
857 bzero((caddr_t)nfs_cltpsock, sizeof (struct nfssvc_sock));
858 STAILQ_INIT(&nfs_cltpsock->ns_rec);
859 TAILQ_INIT(&nfs_cltpsock->ns_uidlruhead);
860 TAILQ_INSERT_TAIL(&nfssvc_sockhead, nfs_cltpsock, ns_chain);
865 * Add entries to the server monitor log.
868 nfsd_rt(sotype, nd, cacherep)
870 struct nfsrv_descript *nd;
875 rt = &nfsdrt.drt[nfsdrt.pos];
876 if (cacherep == RC_DOIT)
878 else if (cacherep == RC_REPLY)
879 rt->flag = DRT_CACHEREPLY;
881 rt->flag = DRT_CACHEDROP;
882 if (sotype == SOCK_STREAM)
884 if (nd->nd_flag & ND_NQNFS)
885 rt->flag |= DRT_NQNFS;
886 else if (nd->nd_flag & ND_NFSV3)
887 rt->flag |= DRT_NFSV3;
888 rt->proc = nd->nd_procnum;
889 if (nd->nd_nam->sa_family == AF_INET)
890 rt->ipadr = ((struct sockaddr_in *)nd->nd_nam)->sin_addr.s_addr;
892 rt->ipadr = INADDR_ANY;
893 rt->resptime = nfs_curusec() - (nd->nd_starttime.tv_sec * 1000000 + nd->nd_starttime.tv_usec);
894 getmicrotime(&rt->tstamp);
895 nfsdrt.pos = (nfsdrt.pos + 1) % NFSRTTLOGSIZ;
897 #endif /* NFS_NOSERVER */
899 static int nfs_defect = 0;
900 SYSCTL_INT(_vfs_nfs, OID_AUTO, defect, CTLFLAG_RW, &nfs_defect, 0, "");
903 * Asynchronous I/O daemons for client nfs.
904 * They do read-ahead and write-behind operations on the block I/O cache.
905 * Never returns unless it fails or gets killed.
908 nfssvc_iod(struct thread *td)
912 struct nfsmount *nmp;
916 * Assign my position or return error if too many already running
919 for (i = 0; i < NFS_MAXASYNCDAEMON; i++)
920 if (nfs_asyncdaemon[i] == 0) {
921 nfs_asyncdaemon[i]++;
929 * Just loop around doin our stuff until SIGKILL
932 while (((nmp = nfs_iodmount[myiod]) == NULL
933 || nmp->nm_bufq.tqh_first == NULL)
937 nfs_iodwant[myiod] = td;
938 nfs_iodmount[myiod] = NULL;
939 error = tsleep((caddr_t)&nfs_iodwant[myiod],
940 PCATCH, "nfsidl", 0);
943 nfs_asyncdaemon[myiod] = 0;
946 nfs_iodwant[myiod] = NULL;
947 nfs_iodmount[myiod] = NULL;
951 while ((bp = nmp->nm_bufq.tqh_first) != NULL) {
952 /* Take one off the front of the list */
953 TAILQ_REMOVE(&nmp->nm_bufq, bp, b_freelist);
955 if (nmp->nm_bufqwant && nmp->nm_bufqlen <= nfs_numasync) {
956 nmp->nm_bufqwant = FALSE;
957 wakeup(&nmp->nm_bufq);
959 (void) nfs_doio(bp, NULL);
961 * If there are more than one iod on this mount, then defect
962 * so that the iods can be shared out fairly between the mounts
964 if (nfs_defect && nmp->nm_bufqiods > 1) {
966 ("nfssvc_iod: iod %d defecting from mount %p\n",
968 nfs_iodmount[myiod] = NULL;
978 * Get an authorization string for the uid by having the mount_nfs sitting
979 * on this mount point porpous out of the kernel and do it.
982 nfs_getauth(nmp, rep, cred, auth_str, auth_len, verf_str, verf_len, key)
983 struct nfsmount *nmp;
990 NFSKERBKEY_T key; /* return session key */
994 while ((nmp->nm_state & NFSSTA_WAITAUTH) == 0) {
995 nmp->nm_state |= NFSSTA_WANTAUTH;
996 (void) tsleep((caddr_t)&nmp->nm_authtype, 0,
998 error = nfs_sigintr(nmp, rep, rep->r_td);
1000 nmp->nm_state &= ~NFSSTA_WANTAUTH;
1004 nmp->nm_state &= ~(NFSSTA_WAITAUTH | NFSSTA_WANTAUTH);
1005 nmp->nm_authstr = *auth_str = (char *)malloc(RPCAUTH_MAXSIZ, M_TEMP, M_WAITOK);
1006 nmp->nm_authlen = RPCAUTH_MAXSIZ;
1007 nmp->nm_verfstr = verf_str;
1008 nmp->nm_verflen = *verf_len;
1009 nmp->nm_authuid = cred->cr_uid;
1010 wakeup((caddr_t)&nmp->nm_authstr);
1013 * And wait for mount_nfs to do its stuff.
1015 while ((nmp->nm_state & NFSSTA_HASAUTH) == 0 && error == 0) {
1016 (void) tsleep((caddr_t)&nmp->nm_authlen, 0,
1017 "nfsauth2", 2 * hz);
1018 error = nfs_sigintr(nmp, rep, rep->r_td);
1020 if (nmp->nm_state & NFSSTA_AUTHERR) {
1021 nmp->nm_state &= ~NFSSTA_AUTHERR;
1025 free((caddr_t)*auth_str, M_TEMP);
1027 *auth_len = nmp->nm_authlen;
1028 *verf_len = nmp->nm_verflen;
1029 bcopy((caddr_t)nmp->nm_key, (caddr_t)key, sizeof (key));
1031 nmp->nm_state &= ~NFSSTA_HASAUTH;
1032 nmp->nm_state |= NFSSTA_WAITAUTH;
1033 if (nmp->nm_state & NFSSTA_WANTAUTH) {
1034 nmp->nm_state &= ~NFSSTA_WANTAUTH;
1035 wakeup((caddr_t)&nmp->nm_authtype);
1041 * Get a nickname authenticator and verifier.
1044 nfs_getnickauth(nmp, cred, auth_str, auth_len, verf_str, verf_len)
1045 struct nfsmount *nmp;
1052 struct nfsuid *nuidp;
1053 u_int32_t *nickp, *verfp;
1054 struct timeval ktvin, ktvout;
1057 if (verf_len < (4 * NFSX_UNSIGNED))
1058 panic("nfs_getnickauth verf too small");
1060 for (nuidp = NMUIDHASH(nmp, cred->cr_uid)->lh_first;
1061 nuidp != 0; nuidp = nuidp->nu_hash.le_next) {
1062 if (nuidp->nu_cr.cr_uid == cred->cr_uid)
1065 if (!nuidp || nuidp->nu_expire < time_second)
1069 * Move to the end of the lru list (end of lru == most recently used).
1071 TAILQ_REMOVE(&nmp->nm_uidlruhead, nuidp, nu_lru);
1072 TAILQ_INSERT_TAIL(&nmp->nm_uidlruhead, nuidp, nu_lru);
1074 nickp = (u_int32_t *)malloc(2 * NFSX_UNSIGNED, M_TEMP, M_WAITOK);
1075 *nickp++ = txdr_unsigned(RPCAKN_NICKNAME);
1076 *nickp = txdr_unsigned(nuidp->nu_nickname);
1077 *auth_str = (char *)nickp;
1078 *auth_len = 2 * NFSX_UNSIGNED;
1081 * Now we must encrypt the verifier and package it up.
1083 verfp = (u_int32_t *)verf_str;
1084 *verfp++ = txdr_unsigned(RPCAKN_NICKNAME);
1085 if (time_second > nuidp->nu_timestamp.tv_sec ||
1086 (time_second == nuidp->nu_timestamp.tv_sec &&
1087 time_second > nuidp->nu_timestamp.tv_usec))
1088 getmicrotime(&nuidp->nu_timestamp);
1090 nuidp->nu_timestamp.tv_usec++;
1091 ktvin.tv_sec = txdr_unsigned(nuidp->nu_timestamp.tv_sec);
1092 ktvin.tv_usec = txdr_unsigned(nuidp->nu_timestamp.tv_usec);
1095 * Now encrypt the timestamp verifier in ecb mode using the session
1102 *verfp++ = ktvout.tv_sec;
1103 *verfp++ = ktvout.tv_usec;
1109 * Save the current nickname in a hash list entry on the mount point.
1112 nfs_savenickauth(nmp, cred, len, key, mdp, dposp, mrep)
1113 struct nfsmount *nmp;
1121 struct nfsuid *nuidp;
1124 struct mbuf *md = *mdp;
1125 struct timeval ktvin, ktvout;
1127 char *dpos = *dposp, *cp2;
1128 int deltasec, error = 0;
1130 if (len == (3 * NFSX_UNSIGNED)) {
1131 nfsm_dissect(tl, u_int32_t *, 3 * NFSX_UNSIGNED);
1132 ktvin.tv_sec = *tl++;
1133 ktvin.tv_usec = *tl++;
1134 nick = fxdr_unsigned(u_int32_t, *tl);
1137 * Decrypt the timestamp in ecb mode.
1142 ktvout.tv_sec = fxdr_unsigned(long, ktvout.tv_sec);
1143 ktvout.tv_usec = fxdr_unsigned(long, ktvout.tv_usec);
1144 deltasec = time_second - ktvout.tv_sec;
1146 deltasec = -deltasec;
1148 * If ok, add it to the hash list for the mount point.
1150 if (deltasec <= NFS_KERBCLOCKSKEW) {
1151 if (nmp->nm_numuids < nuidhash_max) {
1153 nuidp = (struct nfsuid *)
1154 malloc(sizeof (struct nfsuid), M_NFSUID,
1157 nuidp = nmp->nm_uidlruhead.tqh_first;
1158 LIST_REMOVE(nuidp, nu_hash);
1159 TAILQ_REMOVE(&nmp->nm_uidlruhead, nuidp,
1163 nuidp->nu_cr.cr_uid = cred->cr_uid;
1164 nuidp->nu_expire = time_second + NFS_KERBTTL;
1165 nuidp->nu_timestamp = ktvout;
1166 nuidp->nu_nickname = nick;
1167 bcopy(key, nuidp->nu_key, sizeof (key));
1168 TAILQ_INSERT_TAIL(&nmp->nm_uidlruhead, nuidp,
1170 LIST_INSERT_HEAD(NMUIDHASH(nmp, cred->cr_uid),
1174 nfsm_adv(nfsm_rndup(len));