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 $
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>
46 #include <sys/filedesc.h>
47 #include <sys/vnode.h>
48 #include <sys/malloc.h>
49 #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/nlookup.h>
61 #include <sys/mutex2.h>
62 #include <sys/thread2.h>
64 #include <netinet/in.h>
65 #include <netinet/tcp.h>
70 #include "nfsm_subs.h"
71 #include "nfsrvcache.h"
76 #include <sys/thread2.h>
78 static MALLOC_DEFINE(M_NFSSVC, "NFS srvsock", "Nfs server structure");
80 static int nuidhash_max = NFS_MAXUIDHASH;
83 static void nfsrv_zapsock (struct nfssvc_sock *slp);
89 SYSCTL_DECL(_vfs_nfs);
93 static struct nfsdrt nfsdrt;
94 static int nfs_numnfsd = 0;
95 static void nfsd_rt (int sotype, struct nfsrv_descript *nd,
97 static int nfssvc_addsock (struct file *, struct sockaddr *,
99 static int nfssvc_nfsd (struct nfsd_srvargs *,caddr_t,struct thread *);
101 static int nfs_privport = 0;
102 SYSCTL_INT(_vfs_nfs, NFS_NFSPRIVPORT, nfs_privport, CTLFLAG_RW, &nfs_privport,
103 0, "Enable privileged source port checks");
104 SYSCTL_INT(_vfs_nfs, OID_AUTO, gatherdelay, CTLFLAG_RW, &nfsrvw_procrastinate, 0,
105 "Enable NFS request procrastination");
106 SYSCTL_INT(_vfs_nfs, OID_AUTO, gatherdelay_v3, CTLFLAG_RW, &nfsrvw_procrastinate_v3, 0,
107 "Enable NFSv3 request procrastination");
108 int nfs_soreserve = NFS_MAXPACKET * NFS_MAXASYNCBIO;
109 SYSCTL_INT(_vfs_nfs, OID_AUTO, soreserve, CTLFLAG_RW, &nfs_soreserve, 0,
110 "Minimum NFS socket buffer size reservation");
113 * NFS server system calls
116 #endif /* NFS_NOSERVER */
118 * nfssvc_args(int flag, caddr_t argp)
120 * Nfs server psuedo system call for the nfsd's
121 * Based on the flag value it either:
122 * - adds a socket to the selection list
123 * - remains in the kernel as an nfsd
124 * - remains in the kernel as an nfsiod
129 sys_nfssvc(struct nfssvc_args *uap)
132 struct nlookupdata nd;
134 struct sockaddr *nam;
135 struct nfsd_args nfsdarg;
136 struct nfsd_srvargs nfsd_srvargs, *nsd = &nfsd_srvargs;
137 struct nfsd_cargs ncd;
139 struct nfssvc_sock *slp;
140 struct nfsuid *nuidp;
141 struct nfsmount *nmp;
143 #endif /* NFS_NOSERVER */
145 struct thread *td = curthread;
150 error = priv_check(td, PRIV_ROOT);
154 lwkt_gettoken(&nfs_token);
156 while (nfssvc_sockhead_flag & SLP_INIT) {
157 nfssvc_sockhead_flag |= SLP_WANTINIT;
158 tsleep((caddr_t)&nfssvc_sockhead, 0, "nfsd init", 0);
160 if (uap->flag & NFSSVC_BIOD)
161 error = ENXIO; /* no longer need nfsiod's */
165 #else /* !NFS_NOSERVER */
166 else if (uap->flag & NFSSVC_MNTD) {
167 error = copyin(uap->argp, (caddr_t)&ncd, sizeof (ncd));
171 error = nlookup_init(&nd, ncd.ncd_dirp, UIO_USERSPACE,
174 error = nlookup(&nd);
176 error = cache_vget(&nd.nl_nch, nd.nl_cred, LK_EXCLUSIVE, &vp);
181 if ((vp->v_flag & VROOT) == 0)
183 nmp = VFSTONFS(vp->v_mount);
187 if ((nmp->nm_state & NFSSTA_MNTD) &&
188 (uap->flag & NFSSVC_GOTAUTH) == 0) {
192 nmp->nm_state |= NFSSTA_MNTD;
193 error = nfs_clientd(nmp, td->td_ucred, &ncd, uap->flag,
195 } else if (uap->flag & NFSSVC_ADDSOCK) {
196 error = copyin(uap->argp, (caddr_t)&nfsdarg, sizeof(nfsdarg));
199 error = holdsock(td->td_proc->p_fd, nfsdarg.sock, &fp);
203 * Get the client address for connected sockets.
205 if (nfsdarg.name == NULL || nfsdarg.namelen == 0)
208 error = getsockaddr(&nam, nfsdarg.name,
215 error = nfssvc_addsock(fp, nam, td);
218 error = copyin(uap->argp, (caddr_t)nsd, sizeof (*nsd));
221 if ((uap->flag & NFSSVC_AUTHIN) &&
222 ((nfsd = nsd->nsd_nfsd)) != NULL &&
223 (nfsd->nfsd_slp->ns_flag & SLP_VALID)) {
224 slp = nfsd->nfsd_slp;
227 * First check to see if another nfsd has already
228 * added this credential.
230 for (nuidp = NUIDHASH(slp,nsd->nsd_cr.cr_uid)->lh_first;
231 nuidp != 0; nuidp = nuidp->nu_hash.le_next) {
232 if (nuidp->nu_cr.cr_uid == nsd->nsd_cr.cr_uid &&
233 (!nfsd->nfsd_nd->nd_nam2 ||
234 netaddr_match(NU_NETFAM(nuidp),
235 &nuidp->nu_haddr, nfsd->nfsd_nd->nd_nam2)))
239 nfsrv_setcred(&nuidp->nu_cr,&nfsd->nfsd_nd->nd_cr);
240 nfsd->nfsd_nd->nd_flag |= ND_KERBFULL;
245 if (slp->ns_numuids < nuidhash_max) {
247 nuidp = (struct nfsuid *)
248 kmalloc(sizeof (struct nfsuid), M_NFSUID,
252 if ((slp->ns_flag & SLP_VALID) == 0) {
254 kfree((caddr_t)nuidp, M_NFSUID);
257 nuidp = TAILQ_FIRST(&slp->ns_uidlruhead);
258 LIST_REMOVE(nuidp, nu_hash);
259 TAILQ_REMOVE(&slp->ns_uidlruhead, nuidp,
261 if (nuidp->nu_flag & NU_NAM)
262 kfree(nuidp->nu_nam, M_SONAME);
265 nuidp->nu_cr = nsd->nsd_cr;
266 if (nuidp->nu_cr.cr_ngroups > NGROUPS)
267 nuidp->nu_cr.cr_ngroups = NGROUPS;
268 nuidp->nu_cr.cr_ref = 1;
269 nuidp->nu_timestamp = nsd->nsd_timestamp;
270 nuidp->nu_expire = time_second + nsd->nsd_ttl;
272 * and save the session key in nu_key.
274 bcopy(nsd->nsd_key, nuidp->nu_key,
275 sizeof (nsd->nsd_key));
276 if (nfsd->nfsd_nd->nd_nam2) {
277 struct sockaddr_in *saddr;
279 saddr = (struct sockaddr_in *)
280 nfsd->nfsd_nd->nd_nam2;
281 switch (saddr->sin_family) {
283 nuidp->nu_flag |= NU_INETADDR;
285 saddr->sin_addr.s_addr;
289 nuidp->nu_flag |= NU_NAM;
291 dup_sockaddr(nfsd->nfsd_nd->nd_nam2);
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, td);
309 #endif /* NFS_NOSERVER */
310 if (error == EINTR || error == ERESTART)
313 lwkt_reltoken(&nfs_token);
319 * Adds a socket to the list for servicing by nfsds.
322 nfssvc_addsock(struct file *fp, struct sockaddr *mynam, struct thread *td)
325 struct nfssvc_sock *slp;
329 so = (struct socket *)fp->f_data;
333 * Add it to the list, as required.
335 if (so->so_proto->pr_protocol == IPPROTO_UDP) {
337 if (tslp->ns_flag & SLP_VALID) {
339 kfree(mynam, M_SONAME);
345 * Reserve buffer space in the socket. Note that due to bugs in
346 * Linux's delayed-ack code, serious performance degredation may
347 * occur with linux hosts if the minimum is used.
349 * NFS sockets are not limited to the standard sb_max or by
352 if (so->so_type == SOCK_STREAM)
353 siz = NFS_MAXPACKET + sizeof (u_long);
356 if (siz < nfs_soreserve)
359 error = soreserve(so, siz, siz, NULL);
362 kfree(mynam, M_SONAME);
367 * Set protocol specific options { for now TCP only } and
368 * reserve some space. For datagram sockets, this can get called
369 * repeatedly for the same socket, but that isn't harmful.
371 if (so->so_type == SOCK_STREAM) {
375 bzero(&sopt, sizeof sopt);
376 sopt.sopt_level = SOL_SOCKET;
377 sopt.sopt_name = SO_KEEPALIVE;
378 sopt.sopt_val = &val;
379 sopt.sopt_valsize = sizeof val;
383 if (so->so_proto->pr_domain->dom_family == AF_INET &&
384 so->so_proto->pr_protocol == IPPROTO_TCP) {
388 bzero(&sopt, sizeof sopt);
389 sopt.sopt_level = IPPROTO_TCP;
390 sopt.sopt_name = TCP_NODELAY;
391 sopt.sopt_val = &val;
392 sopt.sopt_valsize = sizeof val;
396 bzero(&sopt, sizeof sopt);
397 sopt.sopt_level = IPPROTO_TCP;
398 sopt.sopt_name = TCP_FASTKEEP;
399 sopt.sopt_val = &val;
400 sopt.sopt_valsize = sizeof val;
404 atomic_clear_int(&so->so_rcv.ssb_flags, SSB_NOINTR);
405 so->so_rcv.ssb_timeo = 0;
406 atomic_clear_int(&so->so_snd.ssb_flags, SSB_NOINTR);
407 so->so_snd.ssb_timeo = 0;
409 slp = kmalloc(sizeof (struct nfssvc_sock), M_NFSSVC, M_WAITOK | M_ZERO);
410 mtx_init(&slp->ns_solock);
411 STAILQ_INIT(&slp->ns_rec);
412 TAILQ_INIT(&slp->ns_uidlruhead);
413 lwkt_token_init(&slp->ns_token, "nfssrv_token");
415 lwkt_gettoken(&nfs_token);
417 TAILQ_INSERT_TAIL(&nfssvc_sockhead, slp, ns_chain);
418 lwkt_gettoken(&slp->ns_token);
425 so->so_upcallarg = (caddr_t)slp;
426 so->so_upcall = nfsrv_rcv_upcall;
427 atomic_set_int(&so->so_rcv.ssb_flags, SSB_UPCALL);
428 slp->ns_flag = (SLP_VALID | SLP_NEEDQ);
429 nfsrv_wakenfsd(slp, 1);
431 lwkt_reltoken(&slp->ns_token);
432 lwkt_reltoken(&nfs_token);
438 * Called by nfssvc() for nfsds. Just loops around servicing rpc requests
439 * until it is killed by a signal.
442 nfssvc_nfsd(struct nfsd_srvargs *nsd, caddr_t argp, struct thread *td)
445 struct nfssvc_sock *slp;
446 struct nfsd *nfsd = nsd->nsd_nfsd;
447 struct nfsrv_descript *nd = NULL;
448 struct mbuf *m, *mreq;
449 int error, cacherep, sotype, writes_todo;
457 lwkt_gettoken(&nfs_token);
460 nsd->nsd_nfsd = nfsd = (struct nfsd *)
461 kmalloc(sizeof (struct nfsd), M_NFSD, M_WAITOK|M_ZERO);
463 TAILQ_INSERT_TAIL(&nfsd_head, nfsd, nfsd_chain);
468 * Loop getting rpc requests until SIGKILL.
471 if ((nfsd->nfsd_flag & NFSD_REQINPROG) == 0) {
472 while (nfsd->nfsd_slp == NULL &&
473 (nfsd_head_flag & NFSD_CHECKSLP) == 0) {
474 nfsd->nfsd_flag |= NFSD_WAITING;
476 error = tsleep(nfsd, PCATCH, "nfsd", 0);
478 if (error && nfsd->nfsd_slp == NULL)
481 if (nfsd->nfsd_slp == NULL &&
482 (nfsd_head_flag & NFSD_CHECKSLP)) {
483 TAILQ_FOREACH(slp, &nfssvc_sockhead, ns_chain) {
484 if ((slp->ns_flag & SLP_ACTION_MASK) ||
485 slp->ns_needq_upcall) {
487 nfsd->nfsd_slp = slp;
492 nfsd_head_flag &= ~NFSD_CHECKSLP;
494 if ((slp = nfsd->nfsd_slp) == NULL)
497 lwkt_reltoken(&nfs_token);
498 lwkt_gettoken(&slp->ns_token);
500 if (slp->ns_needq_upcall) {
501 slp->ns_needq_upcall = 0;
502 slp->ns_flag |= SLP_NEEDQ;
505 if (slp->ns_flag & SLP_VALID) {
507 * We can both process additional received
508 * data into new records and process existing
509 * records. This keeps the pipeline hot by
510 * allowing the tcp socket to continue to
511 * drain while we are processing records.
513 if (slp->ns_flag & SLP_DISCONN) {
515 } else if (slp->ns_flag & SLP_NEEDQ) {
516 (void)nfs_slplock(slp, 1);
517 nfsrv_rcv(slp->ns_so, (caddr_t)slp,
521 error = nfsrv_dorec(slp, nfsd, &nd);
522 cur_usec = nfs_curusec();
523 if (error && slp->ns_tq.lh_first &&
524 slp->ns_tq.lh_first->nd_time <= cur_usec) {
531 nfsd->nfsd_flag |= NFSD_REQINPROG;
533 slp->ns_flag &= ~SLP_ACTION_MASK;
538 slp = nfsd->nfsd_slp;
540 lwkt_reltoken(&nfs_token);
541 lwkt_gettoken(&slp->ns_token);
543 if (slp->ns_needq_upcall) {
544 slp->ns_needq_upcall = 0;
545 slp->ns_flag |= SLP_NEEDQ;
547 if (NFSRV_RECLIMIT(slp) == 0 &&
548 (slp->ns_flag & SLP_NEEDQ)) {
549 (void)nfs_slplock(slp, 1);
550 nfsrv_rcv(slp->ns_so, (caddr_t)slp,
557 * nfs_token not held here. slp token is held.
559 if (error || (slp->ns_flag & SLP_VALID) == 0) {
561 kfree((caddr_t)nd, M_NFSRVDESC);
564 nfsd->nfsd_flag &= ~NFSD_REQINPROG;
565 if (slp->ns_flag & SLP_ACTION_MASK) {
566 lwkt_reltoken(&slp->ns_token);
567 lwkt_gettoken(&nfs_token);
569 nfsd->nfsd_slp = NULL;
570 lwkt_reltoken(&slp->ns_token);
571 lwkt_gettoken(&nfs_token);
578 * Execute the NFS request - handle the server side cache
580 * nfs_token not held here. slp token is held.
582 sotype = slp->ns_so->so_type;
584 getmicrotime(&nd->nd_starttime);
586 nd->nd_nam = nd->nd_nam2;
588 nd->nd_nam = slp->ns_nam;
591 * Check to see if authorization is needed.
593 if (nfsd->nfsd_flag & NFSD_NEEDAUTH) {
594 nfsd->nfsd_flag &= ~NFSD_NEEDAUTH;
596 ((struct sockaddr_in *)
597 nd->nd_nam)->sin_addr.s_addr;
598 nsd->nsd_authlen = nfsd->nfsd_authlen;
599 nsd->nsd_verflen = nfsd->nfsd_verflen;
600 if (!copyout(nfsd->nfsd_authstr,nsd->nsd_authstr,
601 nfsd->nfsd_authlen) &&
602 !copyout(nfsd->nfsd_verfstr, nsd->nsd_verfstr,
603 nfsd->nfsd_verflen) &&
604 !copyout((caddr_t)nsd, argp, sizeof (*nsd)))
606 lwkt_reltoken(&slp->ns_token);
609 cacherep = RC_DROPIT;
611 cacherep = nfsrv_getcache(nd, slp, &mreq);
614 if (nfsd->nfsd_flag & NFSD_AUTHFAIL) {
615 nfsd->nfsd_flag &= ~NFSD_AUTHFAIL;
616 nd->nd_procnum = NFSPROC_NOOP;
617 nd->nd_repstat = (NFSERR_AUTHERR | AUTH_TOOWEAK);
619 } else if (nfs_privport) {
620 /* Check if source port is privileged */
622 struct sockaddr *nam = nd->nd_nam;
623 struct sockaddr_in *sin;
625 sin = (struct sockaddr_in *)nam;
626 port = ntohs(sin->sin_port);
627 if (port >= IPPORT_RESERVED &&
628 nd->nd_procnum != NFSPROC_NULL) {
629 nd->nd_procnum = NFSPROC_NOOP;
630 nd->nd_repstat = (NFSERR_AUTHERR | AUTH_TOOWEAK);
632 kprintf("NFS request from unprivileged port (%s:%d)\n",
633 inet_ntoa(sin->sin_addr), port);
640 * Execute the NFS request - direct execution
642 * Loop to get all the write rpc replies that have been
645 * nfs_token not held here. slp token is held.
650 if (nd && (nd->nd_flag & ND_NFSV3))
651 procrastinate = nfsrvw_procrastinate_v3;
653 procrastinate = nfsrvw_procrastinate;
654 if (writes_todo || (nd->nd_procnum == NFSPROC_WRITE &&
657 error = nfsrv_writegather(&nd, slp,
658 nfsd->nfsd_td, &mreq);
660 /* NOT YET lwkt_reltoken(&slp->ns_token); */
661 error = (*(nfsrv3_procs[nd->nd_procnum]))(nd,
662 slp, nfsd->nfsd_td, &mreq);
663 /* NOT YET lwkt_gettoken(&slp->ns_token); */
664 lwpkthreaddeferred(); /* vnlru issues */
668 if (error != 0 && error != NFSERR_RETVOID) {
669 if (nd->nd_procnum != NQNFSPROC_VACATED)
671 nfsrv_updatecache(nd, FALSE, mreq);
673 kfree(nd->nd_nam2, M_SONAME);
676 nfsstats.srvrpccnt[nd->nd_procnum]++;
677 nfsrv_updatecache(nd, TRUE, mreq);
687 if (siz <= 0 || siz > NFS_MAXPACKET) {
688 kprintf("mbuf siz=%d\n",siz);
689 panic("Bad nfs svc reply");
692 m->m_pkthdr.len = siz;
693 m->m_pkthdr.rcvif = NULL;
695 * For stream protocols, prepend a Sun RPC
698 if (sotype == SOCK_STREAM) {
699 M_PREPEND(m, NFSX_UNSIGNED, MB_WAIT);
704 *mtod(m, u_int32_t *) = htonl(0x80000000 | siz);
706 if (slp->ns_so->so_proto->pr_flags & PR_CONNREQUIRED)
707 (void) nfs_slplock(slp, 1);
708 if (slp->ns_flag & SLP_VALID)
709 error = nfs_send(slp->ns_so, nd->nd_nam2, m, NULL);
716 nfsd_rt(sotype, nd, cacherep);
718 kfree(nd->nd_nam2, M_SONAME);
720 m_freem(nd->nd_mrep);
721 if (error == EPIPE || error == ENOBUFS)
723 if (slp->ns_so->so_proto->pr_flags & PR_CONNREQUIRED)
725 if (error == EINTR || error == ERESTART) {
726 kfree((caddr_t)nd, M_NFSRVDESC);
727 lwkt_reltoken(&slp->ns_token);
728 lwkt_gettoken(&nfs_token);
729 nfsd->nfsd_slp = NULL;
736 nfsd_rt(sotype, nd, cacherep);
737 m_freem(nd->nd_mrep);
739 kfree(nd->nd_nam2, M_SONAME);
743 kfree((caddr_t)nd, M_NFSRVDESC);
748 * Check to see if there are outstanding writes that
749 * need to be serviced.
751 cur_usec = nfs_curusec();
752 if (slp->ns_tq.lh_first &&
753 slp->ns_tq.lh_first->nd_time <= cur_usec) {
759 } while (writes_todo);
762 * nfs_token not held here. slp token is held.
764 if (nfsrv_dorec(slp, nfsd, &nd)) {
765 nfsd->nfsd_flag &= ~NFSD_REQINPROG;
766 if (slp->ns_flag & SLP_ACTION_MASK) {
767 lwkt_reltoken(&slp->ns_token);
768 lwkt_gettoken(&nfs_token);
770 nfsd->nfsd_slp = NULL;
771 lwkt_reltoken(&slp->ns_token);
772 lwkt_gettoken(&nfs_token);
776 lwkt_reltoken(&slp->ns_token);
777 lwkt_gettoken(&nfs_token);
781 TAILQ_REMOVE(&nfsd_head, nfsd, nfsd_chain);
782 kfree((caddr_t)nfsd, M_NFSD);
783 nsd->nsd_nfsd = NULL;
784 if (--nfs_numnfsd == 0)
785 nfsrv_init(TRUE); /* Reinitialize everything */
787 lwkt_reltoken(&nfs_token);
792 * Shut down a socket associated with an nfssvc_sock structure.
793 * Should be called with the send lock set, if required.
794 * The trick here is to increment the sref at the start, so that the nfsds
795 * will stop using it and clear ns_flag at the end so that it will not be
796 * reassigned during cleanup.
799 nfsrv_zapsock(struct nfssvc_sock *slp)
801 struct nfsuid *nuidp, *nnuidp;
802 struct nfsrv_descript *nwp, *nnwp;
805 struct nfsrv_rec *rec;
807 slp->ns_flag &= ~SLP_ALLFLAGS;
812 atomic_clear_int(&so->so_rcv.ssb_flags, SSB_UPCALL);
813 so->so_upcall = NULL;
814 so->so_upcallarg = NULL;
815 soshutdown(so, SHUT_RDWR);
818 kfree(slp->ns_nam, M_SONAME);
819 m_freem(slp->ns_raw);
820 while ((rec = STAILQ_FIRST(&slp->ns_rec)) != NULL) {
822 STAILQ_REMOVE_HEAD(&slp->ns_rec, nr_link);
824 kfree(rec->nr_address, M_SONAME);
825 m_freem(rec->nr_packet);
826 kfree(rec, M_NFSRVDESC);
828 KKASSERT(slp->ns_numrec == 0);
830 TAILQ_FOREACH_MUTABLE(nuidp, &slp->ns_uidlruhead, nu_lru,
832 LIST_REMOVE(nuidp, nu_hash);
833 TAILQ_REMOVE(&slp->ns_uidlruhead, nuidp, nu_lru);
834 if (nuidp->nu_flag & NU_NAM)
835 kfree(nuidp->nu_nam, M_SONAME);
836 kfree((caddr_t)nuidp, M_NFSUID);
839 for (nwp = slp->ns_tq.lh_first; nwp; nwp = nnwp) {
840 nnwp = nwp->nd_tq.le_next;
841 LIST_REMOVE(nwp, nd_tq);
842 kfree((caddr_t)nwp, M_NFSRVDESC);
844 LIST_INIT(&slp->ns_tq);
851 * Derefence a server socket structure. If it has no more references and
852 * is no longer valid, you can throw it away.
854 * Must be holding nfs_token!
857 nfsrv_slpderef(struct nfssvc_sock *slp)
859 ASSERT_LWKT_TOKEN_HELD(&nfs_token);
860 if (slp->ns_sref == 1) {
861 KKASSERT((slp->ns_flag & SLP_VALID) == 0);
862 TAILQ_REMOVE(&nfssvc_sockhead, slp, ns_chain);
864 kfree((caddr_t)slp, M_NFSSVC);
871 nfsrv_slpref(struct nfssvc_sock *slp)
873 ASSERT_LWKT_TOKEN_HELD(&nfs_token);
878 * Lock a socket against others.
880 * Returns 0 on failure, 1 on success.
883 nfs_slplock(struct nfssvc_sock *slp, int wait)
885 mtx_t mtx = &slp->ns_solock;
888 mtx_lock_ex(mtx, "nfsslplck", 0, 0);
890 } else if (mtx_lock_ex_try(mtx) == 0) {
898 * Unlock the stream socket for others.
901 nfs_slpunlock(struct nfssvc_sock *slp)
903 mtx_t mtx = &slp->ns_solock;
909 * Initialize the data structures for the server.
910 * Handshake with any new nfsds starting up to avoid any chance of
914 nfsrv_init(int terminating)
916 struct nfssvc_sock *slp, *nslp;
918 lwkt_gettoken(&nfs_token);
919 if (nfssvc_sockhead_flag & SLP_INIT)
921 nfssvc_sockhead_flag |= SLP_INIT;
924 TAILQ_FOREACH_MUTABLE(slp, &nfssvc_sockhead, ns_chain, nslp) {
925 if (slp->ns_flag & SLP_VALID)
928 TAILQ_REMOVE(&nfssvc_sockhead, slp, ns_chain);
929 kfree((caddr_t)slp, M_NFSSVC);
932 nfsrv_cleancache(); /* And clear out server cache */
934 nfs_pub.np_valid = 0;
937 TAILQ_INIT(&nfssvc_sockhead);
938 nfssvc_sockhead_flag &= ~SLP_INIT;
939 if (nfssvc_sockhead_flag & SLP_WANTINIT) {
940 nfssvc_sockhead_flag &= ~SLP_WANTINIT;
941 wakeup((caddr_t)&nfssvc_sockhead);
944 TAILQ_INIT(&nfsd_head);
945 nfsd_head_flag &= ~NFSD_CHECKSLP;
947 lwkt_reltoken(&nfs_token);
950 nfs_udpsock = (struct nfssvc_sock *)
951 kmalloc(sizeof (struct nfssvc_sock), M_NFSSVC, M_WAITOK | M_ZERO);
952 mtx_init(&nfs_udpsock->ns_solock);
953 STAILQ_INIT(&nfs_udpsock->ns_rec);
954 TAILQ_INIT(&nfs_udpsock->ns_uidlruhead);
955 TAILQ_INSERT_HEAD(&nfssvc_sockhead, nfs_udpsock, ns_chain);
957 nfs_cltpsock = (struct nfssvc_sock *)
958 kmalloc(sizeof (struct nfssvc_sock), M_NFSSVC, M_WAITOK | M_ZERO);
959 mtx_init(&nfs_cltpsock->ns_solock);
960 STAILQ_INIT(&nfs_cltpsock->ns_rec);
961 TAILQ_INIT(&nfs_cltpsock->ns_uidlruhead);
962 TAILQ_INSERT_TAIL(&nfssvc_sockhead, nfs_cltpsock, ns_chain);
967 * Add entries to the server monitor log.
970 nfsd_rt(int sotype, struct nfsrv_descript *nd, int cacherep)
974 rt = &nfsdrt.drt[nfsdrt.pos];
975 if (cacherep == RC_DOIT)
977 else if (cacherep == RC_REPLY)
978 rt->flag = DRT_CACHEREPLY;
980 rt->flag = DRT_CACHEDROP;
981 if (sotype == SOCK_STREAM)
983 if (nd->nd_flag & ND_NFSV3)
984 rt->flag |= DRT_NFSV3;
985 rt->proc = nd->nd_procnum;
986 if (nd->nd_nam->sa_family == AF_INET)
987 rt->ipadr = ((struct sockaddr_in *)nd->nd_nam)->sin_addr.s_addr;
989 rt->ipadr = INADDR_ANY;
990 rt->resptime = nfs_curusec() - (nd->nd_starttime.tv_sec * 1000000 + nd->nd_starttime.tv_usec);
991 getmicrotime(&rt->tstamp);
992 nfsdrt.pos = (nfsdrt.pos + 1) % NFSRTTLOGSIZ;
994 #endif /* NFS_NOSERVER */
997 * Get an authorization string for the uid by having the mount_nfs sitting
998 * on this mount point porpous out of the kernel and do it.
1001 nfs_getauth(struct nfsmount *nmp, struct nfsreq *rep,
1002 struct ucred *cred, char **auth_str, int *auth_len, char *verf_str,
1003 int *verf_len, NFSKERBKEY_T key /* return session key */)
1007 while ((nmp->nm_state & NFSSTA_WAITAUTH) == 0) {
1008 nmp->nm_state |= NFSSTA_WANTAUTH;
1009 (void) tsleep((caddr_t)&nmp->nm_authtype, 0,
1010 "nfsauth1", 2 * hz);
1011 error = nfs_sigintr(nmp, rep, rep->r_td);
1013 nmp->nm_state &= ~NFSSTA_WANTAUTH;
1017 nmp->nm_state &= ~(NFSSTA_WAITAUTH | NFSSTA_WANTAUTH);
1018 nmp->nm_authstr = *auth_str = (char *)kmalloc(RPCAUTH_MAXSIZ, M_TEMP, M_WAITOK);
1019 nmp->nm_authlen = RPCAUTH_MAXSIZ;
1020 nmp->nm_verfstr = verf_str;
1021 nmp->nm_verflen = *verf_len;
1022 nmp->nm_authuid = cred->cr_uid;
1023 wakeup((caddr_t)&nmp->nm_authstr);
1026 * And wait for mount_nfs to do its stuff.
1028 while ((nmp->nm_state & NFSSTA_HASAUTH) == 0 && error == 0) {
1029 (void) tsleep((caddr_t)&nmp->nm_authlen, 0,
1030 "nfsauth2", 2 * hz);
1031 error = nfs_sigintr(nmp, rep, rep->r_td);
1033 if (nmp->nm_state & NFSSTA_AUTHERR) {
1034 nmp->nm_state &= ~NFSSTA_AUTHERR;
1038 kfree((caddr_t)*auth_str, M_TEMP);
1040 *auth_len = nmp->nm_authlen;
1041 *verf_len = nmp->nm_verflen;
1042 bcopy((caddr_t)nmp->nm_key, (caddr_t)key, sizeof (NFSKERBKEY_T));
1044 nmp->nm_state &= ~NFSSTA_HASAUTH;
1045 nmp->nm_state |= NFSSTA_WAITAUTH;
1046 if (nmp->nm_state & NFSSTA_WANTAUTH) {
1047 nmp->nm_state &= ~NFSSTA_WANTAUTH;
1048 wakeup((caddr_t)&nmp->nm_authtype);
1054 * Get a nickname authenticator and verifier.
1057 nfs_getnickauth(struct nfsmount *nmp, struct ucred *cred, char **auth_str,
1058 int *auth_len, char *verf_str, int verf_len)
1060 struct nfsuid *nuidp;
1061 u_int32_t *nickp, *verfp;
1062 struct timeval ktvout;
1065 if (verf_len < (4 * NFSX_UNSIGNED))
1066 panic("nfs_getnickauth verf too small");
1068 for (nuidp = NMUIDHASH(nmp, cred->cr_uid)->lh_first;
1069 nuidp != NULL; nuidp = nuidp->nu_hash.le_next) {
1070 if (nuidp->nu_cr.cr_uid == cred->cr_uid)
1073 if (!nuidp || nuidp->nu_expire < time_second)
1077 * Move to the end of the lru list (end of lru == most recently used).
1079 TAILQ_REMOVE(&nmp->nm_uidlruhead, nuidp, nu_lru);
1080 TAILQ_INSERT_TAIL(&nmp->nm_uidlruhead, nuidp, nu_lru);
1082 nickp = (u_int32_t *)kmalloc(2 * NFSX_UNSIGNED, M_TEMP, M_WAITOK);
1083 *nickp++ = txdr_unsigned(RPCAKN_NICKNAME);
1084 *nickp = txdr_unsigned(nuidp->nu_nickname);
1085 *auth_str = (char *)nickp;
1086 *auth_len = 2 * NFSX_UNSIGNED;
1089 * Now we must encrypt the verifier and package it up.
1091 verfp = (u_int32_t *)verf_str;
1092 *verfp++ = txdr_unsigned(RPCAKN_NICKNAME);
1093 if (time_second > nuidp->nu_timestamp.tv_sec ||
1094 (time_second == nuidp->nu_timestamp.tv_sec &&
1095 time_second > nuidp->nu_timestamp.tv_usec))
1096 getmicrotime(&nuidp->nu_timestamp);
1098 nuidp->nu_timestamp.tv_usec++;
1101 * Now encrypt the timestamp verifier in ecb mode using the session
1111 *verfp++ = ktvout.tv_sec;
1112 *verfp++ = ktvout.tv_usec;
1118 * Save the current nickname in a hash list entry on the mount point.
1121 nfs_savenickauth(struct nfsmount *nmp, struct ucred *cred, int len,
1122 NFSKERBKEY_T key, struct mbuf **mdp, char **dposp,
1125 struct nfsuid *nuidp;
1127 struct timeval ktvin, ktvout;
1129 int deltasec, error = 0;
1130 struct nfsm_info info;
1136 if (len == (3 * NFSX_UNSIGNED)) {
1137 NULLOUT(tl = nfsm_dissect(&info, 3 * NFSX_UNSIGNED));
1138 ktvin.tv_sec = *tl++;
1139 ktvin.tv_usec = *tl++;
1140 nick = fxdr_unsigned(u_int32_t, *tl);
1143 * Decrypt the timestamp in ecb mode.
1151 ktvout.tv_sec = fxdr_unsigned(long, ktvout.tv_sec);
1152 ktvout.tv_usec = fxdr_unsigned(long, ktvout.tv_usec);
1153 deltasec = time_second - ktvout.tv_sec;
1155 deltasec = -deltasec;
1157 * If ok, add it to the hash list for the mount point.
1159 if (deltasec <= NFS_KERBCLOCKSKEW) {
1160 if (nmp->nm_numuids < nuidhash_max) {
1162 nuidp = (struct nfsuid *)
1163 kmalloc(sizeof (struct nfsuid), M_NFSUID,
1166 nuidp = TAILQ_FIRST(&nmp->nm_uidlruhead);
1167 LIST_REMOVE(nuidp, nu_hash);
1168 TAILQ_REMOVE(&nmp->nm_uidlruhead, nuidp,
1172 nuidp->nu_cr.cr_uid = cred->cr_uid;
1173 nuidp->nu_expire = time_second + NFS_KERBTTL;
1174 nuidp->nu_timestamp = ktvout;
1175 nuidp->nu_nickname = nick;
1176 bcopy(key, nuidp->nu_key, sizeof (NFSKERBKEY_T));
1177 TAILQ_INSERT_TAIL(&nmp->nm_uidlruhead, nuidp,
1179 LIST_INSERT_HEAD(NMUIDHASH(nmp, cred->cr_uid),
1183 ERROROUT(nfsm_adv(&info, nfsm_rndup(len)));