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 1.15 2004/04/19 16:33:49 cpressey 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, "");
119 static int nfs_soreserve = 65535;
120 SYSCTL_INT(_vfs_nfs, OID_AUTO, soreserve, CTLFLAG_RW, &nfs_soreserve, 0, "");
123 * NFS server system calls
126 #endif /* NFS_NOSERVER */
128 * nfssvc_args(int flag, caddr_t argp)
130 * Nfs server psuedo system call for the nfsd's
131 * Based on the flag value it either:
132 * - adds a socket to the selection list
133 * - remains in the kernel as an nfsd
134 * - remains in the kernel as an nfsiod
137 nfssvc(struct nfssvc_args *uap)
142 struct sockaddr *nam;
143 struct nfsd_args nfsdarg;
144 struct nfsd_srvargs nfsd_srvargs, *nsd = &nfsd_srvargs;
145 struct nfsd_cargs ncd;
147 struct nfssvc_sock *slp;
148 struct nfsuid *nuidp;
149 struct nfsmount *nmp;
150 #endif /* NFS_NOSERVER */
152 struct thread *td = curthread;
160 KKASSERT(td->td_proc); /* for ucred and p_fd */
161 while (nfssvc_sockhead_flag & SLP_INIT) {
162 nfssvc_sockhead_flag |= SLP_WANTINIT;
163 (void) tsleep((caddr_t)&nfssvc_sockhead, 0, "nfsd init", 0);
165 if (uap->flag & NFSSVC_BIOD)
166 error = nfssvc_iod(td);
170 #else /* !NFS_NOSERVER */
171 else if (uap->flag & NFSSVC_MNTD) {
172 error = copyin(uap->argp, (caddr_t)&ncd, sizeof (ncd));
175 NDINIT(&nd, NAMEI_LOOKUP, CNP_FOLLOW | CNP_LOCKLEAF,
176 UIO_USERSPACE, ncd.ncd_dirp, td);
180 NDFREE(&nd, NDF_ONLY_PNBUF);
181 if ((nd.ni_vp->v_flag & VROOT) == 0)
183 nmp = VFSTONFS(nd.ni_vp->v_mount);
187 if ((nmp->nm_state & NFSSTA_MNTD) &&
188 (uap->flag & NFSSVC_GOTAUTH) == 0)
190 nmp->nm_state |= NFSSTA_MNTD;
191 error = nqnfs_clientd(nmp, td->td_proc->p_ucred, &ncd, uap->flag,
193 } else if (uap->flag & NFSSVC_ADDSOCK) {
194 error = copyin(uap->argp, (caddr_t)&nfsdarg, sizeof(nfsdarg));
197 error = holdsock(td->td_proc->p_fd, nfsdarg.sock, &fp);
201 * Get the client address for connected sockets.
203 if (nfsdarg.name == NULL || nfsdarg.namelen == 0)
204 nam = (struct sockaddr *)0;
206 error = getsockaddr(&nam, nfsdarg.name,
213 error = nfssvc_addsock(fp, nam, td);
216 error = copyin(uap->argp, (caddr_t)nsd, sizeof (*nsd));
219 if ((uap->flag & NFSSVC_AUTHIN) &&
220 ((nfsd = nsd->nsd_nfsd)) != NULL &&
221 (nfsd->nfsd_slp->ns_flag & SLP_VALID)) {
222 slp = nfsd->nfsd_slp;
225 * First check to see if another nfsd has already
226 * added this credential.
228 for (nuidp = NUIDHASH(slp,nsd->nsd_cr.cr_uid)->lh_first;
229 nuidp != 0; nuidp = nuidp->nu_hash.le_next) {
230 if (nuidp->nu_cr.cr_uid == nsd->nsd_cr.cr_uid &&
231 (!nfsd->nfsd_nd->nd_nam2 ||
232 netaddr_match(NU_NETFAM(nuidp),
233 &nuidp->nu_haddr, nfsd->nfsd_nd->nd_nam2)))
237 nfsrv_setcred(&nuidp->nu_cr,&nfsd->nfsd_nd->nd_cr);
238 nfsd->nfsd_nd->nd_flag |= ND_KERBFULL;
243 if (slp->ns_numuids < nuidhash_max) {
245 nuidp = (struct nfsuid *)
246 malloc(sizeof (struct nfsuid), M_NFSUID,
249 nuidp = (struct nfsuid *)0;
250 if ((slp->ns_flag & SLP_VALID) == 0) {
252 free((caddr_t)nuidp, M_NFSUID);
254 if (nuidp == (struct nfsuid *)0) {
255 nuidp = slp->ns_uidlruhead.tqh_first;
256 LIST_REMOVE(nuidp, nu_hash);
257 TAILQ_REMOVE(&slp->ns_uidlruhead, nuidp,
259 if (nuidp->nu_flag & NU_NAM)
260 FREE(nuidp->nu_nam, M_SONAME);
263 nuidp->nu_cr = nsd->nsd_cr;
264 if (nuidp->nu_cr.cr_ngroups > NGROUPS)
265 nuidp->nu_cr.cr_ngroups = NGROUPS;
266 nuidp->nu_cr.cr_ref = 1;
267 nuidp->nu_timestamp = nsd->nsd_timestamp;
268 nuidp->nu_expire = time_second + nsd->nsd_ttl;
270 * and save the session key in nu_key.
272 bcopy(nsd->nsd_key, nuidp->nu_key,
273 sizeof (nsd->nsd_key));
274 if (nfsd->nfsd_nd->nd_nam2) {
275 struct sockaddr_in *saddr;
277 saddr = (struct sockaddr_in *)
278 nfsd->nfsd_nd->nd_nam2;
279 switch (saddr->sin_family) {
281 nuidp->nu_flag |= NU_INETADDR;
283 saddr->sin_addr.s_addr;
287 nuidp->nu_flag |= NU_NAM;
289 dup_sockaddr(nfsd->nfsd_nd->
294 TAILQ_INSERT_TAIL(&slp->ns_uidlruhead, nuidp,
296 LIST_INSERT_HEAD(NUIDHASH(slp, nsd->nsd_uid),
298 nfsrv_setcred(&nuidp->nu_cr,
299 &nfsd->nfsd_nd->nd_cr);
300 nfsd->nfsd_nd->nd_flag |= ND_KERBFULL;
304 if ((uap->flag & NFSSVC_AUTHINFAIL) && (nfsd = nsd->nsd_nfsd))
305 nfsd->nfsd_flag |= NFSD_AUTHFAIL;
306 error = nfssvc_nfsd(nsd, uap->argp, td);
308 #endif /* NFS_NOSERVER */
309 if (error == EINTR || error == ERESTART)
316 * Adds a socket to the list for servicing by nfsds.
319 nfssvc_addsock(struct file *fp, struct sockaddr *mynam, struct thread *td)
322 struct nfssvc_sock *slp;
326 so = (struct socket *)fp->f_data;
328 tslp = (struct nfssvc_sock *)0;
330 * Add it to the list, as required.
332 if (so->so_proto->pr_protocol == IPPROTO_UDP) {
334 if (tslp->ns_flag & SLP_VALID) {
336 FREE(mynam, M_SONAME);
342 * Reserve buffer space in the socket. Note that due to bugs in
343 * Linux's delayed-ack code, serious performance degredation may
344 * occur with linux hosts if the minimum is used.
346 if (so->so_type == SOCK_STREAM)
347 siz = NFS_MAXPACKET + sizeof (u_long);
350 if (siz < nfs_soreserve)
352 if (siz > sb_max_adj) {
353 printf("Warning: vfs.nfs.soreserve (%d) "
354 "limited to adjusted sb_max (%ld)\n",
355 nfs_soreserve, sb_max_adj);
359 error = soreserve(so, siz, siz, &td->td_proc->p_rlimit[RLIMIT_SBSIZE]);
362 FREE(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 so->so_rcv.sb_flags &= ~SB_NOINTR;
397 so->so_rcv.sb_timeo = 0;
398 so->so_snd.sb_flags &= ~SB_NOINTR;
399 so->so_snd.sb_timeo = 0;
401 slp = (struct nfssvc_sock *)
402 malloc(sizeof (struct nfssvc_sock), M_NFSSVC, M_WAITOK);
403 bzero((caddr_t)slp, sizeof (struct nfssvc_sock));
404 STAILQ_INIT(&slp->ns_rec);
405 TAILQ_INIT(&slp->ns_uidlruhead);
406 TAILQ_INSERT_TAIL(&nfssvc_sockhead, slp, ns_chain);
413 so->so_upcallarg = (caddr_t)slp;
414 so->so_upcall = nfsrv_rcv;
415 so->so_rcv.sb_flags |= SB_UPCALL;
416 slp->ns_flag = (SLP_VALID | SLP_NEEDQ);
423 * Called by nfssvc() for nfsds. Just loops around servicing rpc requests
424 * until it is killed by a signal.
427 nfssvc_nfsd(struct nfsd_srvargs *nsd, caddr_t argp, struct thread *td)
430 struct nfssvc_sock *slp;
431 struct nfsd *nfsd = nsd->nsd_nfsd;
432 struct nfsrv_descript *nd = NULL;
433 struct mbuf *m, *mreq;
434 int error = 0, cacherep, s, sotype, writes_todo;
442 if (nfsd == (struct nfsd *)0) {
443 nsd->nsd_nfsd = nfsd = (struct nfsd *)
444 malloc(sizeof (struct nfsd), M_NFSD, M_WAITOK);
445 bzero((caddr_t)nfsd, sizeof (struct nfsd));
448 TAILQ_INSERT_TAIL(&nfsd_head, nfsd, nfsd_chain);
454 * Loop getting rpc requests until SIGKILL.
457 if ((nfsd->nfsd_flag & NFSD_REQINPROG) == 0) {
458 while (nfsd->nfsd_slp == (struct nfssvc_sock *)0 &&
459 (nfsd_head_flag & NFSD_CHECKSLP) == 0) {
460 nfsd->nfsd_flag |= NFSD_WAITING;
462 error = tsleep((caddr_t)nfsd, PCATCH, "nfsd", 0);
467 if (nfsd->nfsd_slp == (struct nfssvc_sock *)0 &&
468 (nfsd_head_flag & NFSD_CHECKSLP) != 0) {
469 for (slp = nfssvc_sockhead.tqh_first; slp != 0;
470 slp = slp->ns_chain.tqe_next) {
471 if ((slp->ns_flag & (SLP_VALID | SLP_DOREC))
472 == (SLP_VALID | SLP_DOREC)) {
473 slp->ns_flag &= ~SLP_DOREC;
475 nfsd->nfsd_slp = slp;
480 nfsd_head_flag &= ~NFSD_CHECKSLP;
482 if ((slp = nfsd->nfsd_slp) == (struct nfssvc_sock *)0)
484 if (slp->ns_flag & SLP_VALID) {
485 if (slp->ns_flag & SLP_DISCONN)
487 else if (slp->ns_flag & SLP_NEEDQ) {
488 slp->ns_flag &= ~SLP_NEEDQ;
489 (void) nfs_slplock(slp, 1);
490 nfsrv_rcv(slp->ns_so, (caddr_t)slp,
494 error = nfsrv_dorec(slp, nfsd, &nd);
495 cur_usec = nfs_curusec();
496 if (error && slp->ns_tq.lh_first &&
497 slp->ns_tq.lh_first->nd_time <= cur_usec) {
503 nfsd->nfsd_flag |= NFSD_REQINPROG;
507 slp = nfsd->nfsd_slp;
509 if (error || (slp->ns_flag & SLP_VALID) == 0) {
511 free((caddr_t)nd, M_NFSRVDESC);
514 nfsd->nfsd_slp = (struct nfssvc_sock *)0;
515 nfsd->nfsd_flag &= ~NFSD_REQINPROG;
520 sotype = slp->ns_so->so_type;
522 getmicrotime(&nd->nd_starttime);
524 nd->nd_nam = nd->nd_nam2;
526 nd->nd_nam = slp->ns_nam;
529 * Check to see if authorization is needed.
531 if (nfsd->nfsd_flag & NFSD_NEEDAUTH) {
532 nfsd->nfsd_flag &= ~NFSD_NEEDAUTH;
534 ((struct sockaddr_in *)
535 nd->nd_nam)->sin_addr.s_addr;
536 nsd->nsd_authlen = nfsd->nfsd_authlen;
537 nsd->nsd_verflen = nfsd->nfsd_verflen;
538 if (!copyout(nfsd->nfsd_authstr,nsd->nsd_authstr,
539 nfsd->nfsd_authlen) &&
540 !copyout(nfsd->nfsd_verfstr, nsd->nsd_verfstr,
541 nfsd->nfsd_verflen) &&
542 !copyout((caddr_t)nsd, argp, sizeof (*nsd)))
544 cacherep = RC_DROPIT;
546 cacherep = nfsrv_getcache(nd, slp, &mreq);
549 * Check for just starting up for NQNFS and send
550 * fake "try again later" replies to the NQNFS clients.
552 if (notstarted && nqnfsstarttime <= time_second) {
554 nqnfsstarttime = time_second + nqsrv_writeslack;
560 if ((nd->nd_flag & ND_NQNFS) == 0)
561 cacherep = RC_DROPIT;
562 else if (nd->nd_procnum != NFSPROC_WRITE) {
563 nd->nd_procnum = NFSPROC_NOOP;
564 nd->nd_repstat = NQNFS_TRYLATER;
568 } else if (nfsd->nfsd_flag & NFSD_AUTHFAIL) {
569 nfsd->nfsd_flag &= ~NFSD_AUTHFAIL;
570 nd->nd_procnum = NFSPROC_NOOP;
571 nd->nd_repstat = (NFSERR_AUTHERR | AUTH_TOOWEAK);
573 } else if (nfs_privport) {
574 /* Check if source port is privileged */
576 struct sockaddr *nam = nd->nd_nam;
577 struct sockaddr_in *sin;
579 sin = (struct sockaddr_in *)nam;
580 port = ntohs(sin->sin_port);
581 if (port >= IPPORT_RESERVED &&
582 nd->nd_procnum != NFSPROC_NULL) {
583 nd->nd_procnum = NFSPROC_NOOP;
584 nd->nd_repstat = (NFSERR_AUTHERR | AUTH_TOOWEAK);
586 printf("NFS request from unprivileged port (%s:%d)\n",
587 inet_ntoa(sin->sin_addr), port);
594 * Loop to get all the write rpc relies that have been
600 if (nd && (nd->nd_flag & ND_NFSV3))
601 procrastinate = nfsrvw_procrastinate_v3;
603 procrastinate = nfsrvw_procrastinate;
604 if (writes_todo || (nd->nd_procnum == NFSPROC_WRITE &&
605 procrastinate > 0 && !notstarted))
606 error = nfsrv_writegather(&nd, slp,
607 nfsd->nfsd_td, &mreq);
609 error = (*(nfsrv3_procs[nd->nd_procnum]))(nd,
610 slp, nfsd->nfsd_td, &mreq);
613 if (error != 0 && error != NFSERR_RETVOID) {
614 if (nd->nd_procnum != NQNFSPROC_VACATED)
616 nfsrv_updatecache(nd, FALSE, mreq);
618 FREE(nd->nd_nam2, M_SONAME);
621 nfsstats.srvrpccnt[nd->nd_procnum]++;
622 nfsrv_updatecache(nd, TRUE, mreq);
623 nd->nd_mrep = (struct mbuf *)0;
631 if (siz <= 0 || siz > NFS_MAXPACKET) {
632 printf("mbuf siz=%d\n",siz);
633 panic("Bad nfs svc reply");
636 m->m_pkthdr.len = siz;
637 m->m_pkthdr.rcvif = (struct ifnet *)0;
639 * For stream protocols, prepend a Sun RPC
642 if (sotype == SOCK_STREAM) {
643 M_PREPEND(m, NFSX_UNSIGNED, M_WAIT);
646 *mtod(m, u_int32_t *) = htonl(0x80000000 | siz);
648 if (slp->ns_so->so_proto->pr_flags & PR_CONNREQUIRED)
649 (void) nfs_slplock(slp, 1);
650 if (slp->ns_flag & SLP_VALID)
651 error = nfs_send(slp->ns_so, nd->nd_nam2, m, NULL);
657 nfsd_rt(sotype, nd, cacherep);
659 FREE(nd->nd_nam2, M_SONAME);
661 m_freem(nd->nd_mrep);
664 if (slp->ns_so->so_proto->pr_flags & PR_CONNREQUIRED)
666 if (error == EINTR || error == ERESTART) {
667 free((caddr_t)nd, M_NFSRVDESC);
675 nfsd_rt(sotype, nd, cacherep);
676 m_freem(nd->nd_mrep);
678 FREE(nd->nd_nam2, M_SONAME);
682 FREE((caddr_t)nd, M_NFSRVDESC);
687 * Check to see if there are outstanding writes that
688 * need to be serviced.
690 cur_usec = nfs_curusec();
692 if (slp->ns_tq.lh_first &&
693 slp->ns_tq.lh_first->nd_time <= cur_usec) {
699 } while (writes_todo);
701 if (nfsrv_dorec(slp, nfsd, &nd)) {
702 nfsd->nfsd_flag &= ~NFSD_REQINPROG;
703 nfsd->nfsd_slp = NULL;
708 TAILQ_REMOVE(&nfsd_head, nfsd, nfsd_chain);
710 free((caddr_t)nfsd, M_NFSD);
711 nsd->nsd_nfsd = (struct nfsd *)0;
712 if (--nfs_numnfsd == 0)
713 nfsrv_init(TRUE); /* Reinitialize everything */
718 * Shut down a socket associated with an nfssvc_sock structure.
719 * Should be called with the send lock set, if required.
720 * The trick here is to increment the sref at the start, so that the nfsds
721 * will stop using it and clear ns_flag at the end so that it will not be
722 * reassigned during cleanup.
725 nfsrv_zapsock(struct nfssvc_sock *slp)
727 struct nfsuid *nuidp, *nnuidp;
728 struct nfsrv_descript *nwp, *nnwp;
731 struct nfsrv_rec *rec;
734 slp->ns_flag &= ~SLP_ALLFLAGS;
737 slp->ns_fp = (struct file *)0;
739 so->so_rcv.sb_flags &= ~SB_UPCALL;
740 so->so_upcall = NULL;
741 so->so_upcallarg = NULL;
745 FREE(slp->ns_nam, M_SONAME);
746 m_freem(slp->ns_raw);
747 while ((rec = STAILQ_FIRST(&slp->ns_rec)) != NULL) {
748 STAILQ_REMOVE_HEAD(&slp->ns_rec, nr_link);
750 FREE(rec->nr_address, M_SONAME);
751 m_freem(rec->nr_packet);
752 free(rec, M_NFSRVDESC);
754 for (nuidp = slp->ns_uidlruhead.tqh_first; nuidp != 0;
756 nnuidp = nuidp->nu_lru.tqe_next;
757 LIST_REMOVE(nuidp, nu_hash);
758 TAILQ_REMOVE(&slp->ns_uidlruhead, nuidp, nu_lru);
759 if (nuidp->nu_flag & NU_NAM)
760 FREE(nuidp->nu_nam, M_SONAME);
761 free((caddr_t)nuidp, M_NFSUID);
764 for (nwp = slp->ns_tq.lh_first; nwp; nwp = nnwp) {
765 nnwp = nwp->nd_tq.le_next;
766 LIST_REMOVE(nwp, nd_tq);
767 free((caddr_t)nwp, M_NFSRVDESC);
769 LIST_INIT(&slp->ns_tq);
775 * Derefence a server socket structure. If it has no more references and
776 * is no longer valid, you can throw it away.
779 nfsrv_slpderef(struct nfssvc_sock *slp)
781 if (--(slp->ns_sref) == 0 && (slp->ns_flag & SLP_VALID) == 0) {
782 TAILQ_REMOVE(&nfssvc_sockhead, slp, ns_chain);
783 free((caddr_t)slp, M_NFSSVC);
788 * Lock a socket against others.
791 nfs_slplock(struct nfssvc_sock *slp, int wait)
793 int *statep = &slp->ns_solock;
795 if (!wait && (*statep & NFSSTA_SNDLOCK))
796 return(0); /* already locked, fail */
797 while (*statep & NFSSTA_SNDLOCK) {
798 *statep |= NFSSTA_WANTSND;
799 (void) tsleep((caddr_t)statep, 0, "nfsslplck", 0);
801 *statep |= NFSSTA_SNDLOCK;
806 * Unlock the stream socket for others.
809 nfs_slpunlock(struct nfssvc_sock *slp)
811 int *statep = &slp->ns_solock;
813 if ((*statep & NFSSTA_SNDLOCK) == 0)
814 panic("nfs slpunlock");
815 *statep &= ~NFSSTA_SNDLOCK;
816 if (*statep & NFSSTA_WANTSND) {
817 *statep &= ~NFSSTA_WANTSND;
818 wakeup((caddr_t)statep);
823 * Initialize the data structures for the server.
824 * Handshake with any new nfsds starting up to avoid any chance of
828 nfsrv_init(int terminating)
830 struct nfssvc_sock *slp, *nslp;
832 if (nfssvc_sockhead_flag & SLP_INIT)
834 nfssvc_sockhead_flag |= SLP_INIT;
836 for (slp = nfssvc_sockhead.tqh_first; slp != 0; slp = nslp) {
837 nslp = slp->ns_chain.tqe_next;
838 if (slp->ns_flag & SLP_VALID)
840 TAILQ_REMOVE(&nfssvc_sockhead, slp, ns_chain);
841 free((caddr_t)slp, M_NFSSVC);
843 nfsrv_cleancache(); /* And clear out server cache */
845 nfs_pub.np_valid = 0;
847 TAILQ_INIT(&nfssvc_sockhead);
848 nfssvc_sockhead_flag &= ~SLP_INIT;
849 if (nfssvc_sockhead_flag & SLP_WANTINIT) {
850 nfssvc_sockhead_flag &= ~SLP_WANTINIT;
851 wakeup((caddr_t)&nfssvc_sockhead);
854 TAILQ_INIT(&nfsd_head);
855 nfsd_head_flag &= ~NFSD_CHECKSLP;
858 nfs_udpsock = (struct nfssvc_sock *)
859 malloc(sizeof (struct nfssvc_sock), M_NFSSVC, M_WAITOK);
860 bzero((caddr_t)nfs_udpsock, sizeof (struct nfssvc_sock));
861 STAILQ_INIT(&nfs_udpsock->ns_rec);
862 TAILQ_INIT(&nfs_udpsock->ns_uidlruhead);
863 TAILQ_INSERT_HEAD(&nfssvc_sockhead, nfs_udpsock, ns_chain);
865 nfs_cltpsock = (struct nfssvc_sock *)
866 malloc(sizeof (struct nfssvc_sock), M_NFSSVC, M_WAITOK);
867 bzero((caddr_t)nfs_cltpsock, sizeof (struct nfssvc_sock));
868 STAILQ_INIT(&nfs_cltpsock->ns_rec);
869 TAILQ_INIT(&nfs_cltpsock->ns_uidlruhead);
870 TAILQ_INSERT_TAIL(&nfssvc_sockhead, nfs_cltpsock, ns_chain);
875 * Add entries to the server monitor log.
878 nfsd_rt(int sotype, struct nfsrv_descript *nd, int cacherep)
882 rt = &nfsdrt.drt[nfsdrt.pos];
883 if (cacherep == RC_DOIT)
885 else if (cacherep == RC_REPLY)
886 rt->flag = DRT_CACHEREPLY;
888 rt->flag = DRT_CACHEDROP;
889 if (sotype == SOCK_STREAM)
891 if (nd->nd_flag & ND_NQNFS)
892 rt->flag |= DRT_NQNFS;
893 else if (nd->nd_flag & ND_NFSV3)
894 rt->flag |= DRT_NFSV3;
895 rt->proc = nd->nd_procnum;
896 if (nd->nd_nam->sa_family == AF_INET)
897 rt->ipadr = ((struct sockaddr_in *)nd->nd_nam)->sin_addr.s_addr;
899 rt->ipadr = INADDR_ANY;
900 rt->resptime = nfs_curusec() - (nd->nd_starttime.tv_sec * 1000000 + nd->nd_starttime.tv_usec);
901 getmicrotime(&rt->tstamp);
902 nfsdrt.pos = (nfsdrt.pos + 1) % NFSRTTLOGSIZ;
904 #endif /* NFS_NOSERVER */
906 static int nfs_defect = 0;
907 SYSCTL_INT(_vfs_nfs, OID_AUTO, defect, CTLFLAG_RW, &nfs_defect, 0, "");
910 * Asynchronous I/O daemons for client nfs.
911 * They do read-ahead and write-behind operations on the block I/O cache.
912 * Never returns unless it fails or gets killed.
915 nfssvc_iod(struct thread *td)
919 struct nfsmount *nmp;
923 * Assign my position or return error if too many already running
926 for (i = 0; i < NFS_MAXASYNCDAEMON; i++)
927 if (nfs_asyncdaemon[i] == 0) {
928 nfs_asyncdaemon[i]++;
936 * Just loop around doin our stuff until SIGKILL
939 while (((nmp = nfs_iodmount[myiod]) == NULL
940 || nmp->nm_bufq.tqh_first == NULL)
944 nfs_iodwant[myiod] = td;
945 nfs_iodmount[myiod] = NULL;
946 error = tsleep((caddr_t)&nfs_iodwant[myiod],
947 PCATCH, "nfsidl", 0);
950 nfs_asyncdaemon[myiod] = 0;
953 nfs_iodwant[myiod] = NULL;
954 nfs_iodmount[myiod] = NULL;
958 while ((bp = nmp->nm_bufq.tqh_first) != NULL) {
959 /* Take one off the front of the list */
960 TAILQ_REMOVE(&nmp->nm_bufq, bp, b_freelist);
962 if (nmp->nm_bufqwant && nmp->nm_bufqlen <= nfs_numasync) {
963 nmp->nm_bufqwant = FALSE;
964 wakeup(&nmp->nm_bufq);
966 (void) nfs_doio(bp, NULL);
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(struct nfsmount *nmp, struct nfsreq *rep,
990 struct ucred *cred, char **auth_str, int *auth_len, char *verf_str,
991 int *verf_len, NFSKERBKEY_T key /* return session key */)
995 while ((nmp->nm_state & NFSSTA_WAITAUTH) == 0) {
996 nmp->nm_state |= NFSSTA_WANTAUTH;
997 (void) tsleep((caddr_t)&nmp->nm_authtype, 0,
999 error = nfs_sigintr(nmp, rep, rep->r_td);
1001 nmp->nm_state &= ~NFSSTA_WANTAUTH;
1005 nmp->nm_state &= ~(NFSSTA_WAITAUTH | NFSSTA_WANTAUTH);
1006 nmp->nm_authstr = *auth_str = (char *)malloc(RPCAUTH_MAXSIZ, M_TEMP, M_WAITOK);
1007 nmp->nm_authlen = RPCAUTH_MAXSIZ;
1008 nmp->nm_verfstr = verf_str;
1009 nmp->nm_verflen = *verf_len;
1010 nmp->nm_authuid = cred->cr_uid;
1011 wakeup((caddr_t)&nmp->nm_authstr);
1014 * And wait for mount_nfs to do its stuff.
1016 while ((nmp->nm_state & NFSSTA_HASAUTH) == 0 && error == 0) {
1017 (void) tsleep((caddr_t)&nmp->nm_authlen, 0,
1018 "nfsauth2", 2 * hz);
1019 error = nfs_sigintr(nmp, rep, rep->r_td);
1021 if (nmp->nm_state & NFSSTA_AUTHERR) {
1022 nmp->nm_state &= ~NFSSTA_AUTHERR;
1026 free((caddr_t)*auth_str, M_TEMP);
1028 *auth_len = nmp->nm_authlen;
1029 *verf_len = nmp->nm_verflen;
1030 bcopy((caddr_t)nmp->nm_key, (caddr_t)key, sizeof (key));
1032 nmp->nm_state &= ~NFSSTA_HASAUTH;
1033 nmp->nm_state |= NFSSTA_WAITAUTH;
1034 if (nmp->nm_state & NFSSTA_WANTAUTH) {
1035 nmp->nm_state &= ~NFSSTA_WANTAUTH;
1036 wakeup((caddr_t)&nmp->nm_authtype);
1042 * Get a nickname authenticator and verifier.
1045 nfs_getnickauth(struct nfsmount *nmp, struct ucred *cred, char **auth_str,
1046 int *auth_len, char *verf_str, int verf_len)
1048 struct nfsuid *nuidp;
1049 u_int32_t *nickp, *verfp;
1050 struct timeval ktvin, ktvout;
1053 if (verf_len < (4 * NFSX_UNSIGNED))
1054 panic("nfs_getnickauth verf too small");
1056 for (nuidp = NMUIDHASH(nmp, cred->cr_uid)->lh_first;
1057 nuidp != 0; nuidp = nuidp->nu_hash.le_next) {
1058 if (nuidp->nu_cr.cr_uid == cred->cr_uid)
1061 if (!nuidp || nuidp->nu_expire < time_second)
1065 * Move to the end of the lru list (end of lru == most recently used).
1067 TAILQ_REMOVE(&nmp->nm_uidlruhead, nuidp, nu_lru);
1068 TAILQ_INSERT_TAIL(&nmp->nm_uidlruhead, nuidp, nu_lru);
1070 nickp = (u_int32_t *)malloc(2 * NFSX_UNSIGNED, M_TEMP, M_WAITOK);
1071 *nickp++ = txdr_unsigned(RPCAKN_NICKNAME);
1072 *nickp = txdr_unsigned(nuidp->nu_nickname);
1073 *auth_str = (char *)nickp;
1074 *auth_len = 2 * NFSX_UNSIGNED;
1077 * Now we must encrypt the verifier and package it up.
1079 verfp = (u_int32_t *)verf_str;
1080 *verfp++ = txdr_unsigned(RPCAKN_NICKNAME);
1081 if (time_second > nuidp->nu_timestamp.tv_sec ||
1082 (time_second == nuidp->nu_timestamp.tv_sec &&
1083 time_second > nuidp->nu_timestamp.tv_usec))
1084 getmicrotime(&nuidp->nu_timestamp);
1086 nuidp->nu_timestamp.tv_usec++;
1087 ktvin.tv_sec = txdr_unsigned(nuidp->nu_timestamp.tv_sec);
1088 ktvin.tv_usec = txdr_unsigned(nuidp->nu_timestamp.tv_usec);
1091 * Now encrypt the timestamp verifier in ecb mode using the session
1098 *verfp++ = ktvout.tv_sec;
1099 *verfp++ = ktvout.tv_usec;
1105 * Save the current nickname in a hash list entry on the mount point.
1108 nfs_savenickauth(struct nfsmount *nmp, struct ucred *cred, int len,
1109 NFSKERBKEY_T key, struct mbuf **mdp, char **dposp,
1112 struct nfsuid *nuidp;
1115 struct mbuf *md = *mdp;
1116 struct timeval ktvin, ktvout;
1118 char *dpos = *dposp, *cp2;
1119 int deltasec, error = 0;
1121 if (len == (3 * NFSX_UNSIGNED)) {
1122 nfsm_dissect(tl, u_int32_t *, 3 * NFSX_UNSIGNED);
1123 ktvin.tv_sec = *tl++;
1124 ktvin.tv_usec = *tl++;
1125 nick = fxdr_unsigned(u_int32_t, *tl);
1128 * Decrypt the timestamp in ecb mode.
1133 ktvout.tv_sec = fxdr_unsigned(long, ktvout.tv_sec);
1134 ktvout.tv_usec = fxdr_unsigned(long, ktvout.tv_usec);
1135 deltasec = time_second - ktvout.tv_sec;
1137 deltasec = -deltasec;
1139 * If ok, add it to the hash list for the mount point.
1141 if (deltasec <= NFS_KERBCLOCKSKEW) {
1142 if (nmp->nm_numuids < nuidhash_max) {
1144 nuidp = (struct nfsuid *)
1145 malloc(sizeof (struct nfsuid), M_NFSUID,
1148 nuidp = nmp->nm_uidlruhead.tqh_first;
1149 LIST_REMOVE(nuidp, nu_hash);
1150 TAILQ_REMOVE(&nmp->nm_uidlruhead, nuidp,
1154 nuidp->nu_cr.cr_uid = cred->cr_uid;
1155 nuidp->nu_expire = time_second + NFS_KERBTTL;
1156 nuidp->nu_timestamp = ktvout;
1157 nuidp->nu_nickname = nick;
1158 bcopy(key, nuidp->nu_key, sizeof (key));
1159 TAILQ_INSERT_TAIL(&nmp->nm_uidlruhead, nuidp,
1161 LIST_INSERT_HEAD(NMUIDHASH(nmp, cred->cr_uid),
1165 nfsm_adv(nfsm_rndup(len));