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_subs.c 8.8 (Berkeley) 5/22/95
37 * $FreeBSD: /repoman/r/ncvs/src/sys/nfsclient/nfs_subs.c,v 1.128 2004/04/14 23:23:55 peadar Exp $
38 * $DragonFly: src/sys/vfs/nfs/nfs_subs.c,v 1.30 2005/05/29 10:08:36 hsu Exp $
42 * These functions support the macros and help fiddle mbuf chains for
43 * the nfs op functions. They do things like create the rpc header and
44 * copy data between mbuf chains and uio lists.
46 #include <sys/param.h>
47 #include <sys/systm.h>
48 #include <sys/kernel.h>
51 #include <sys/mount.h>
52 #include <sys/vnode.h>
53 #include <sys/nlookup.h>
54 #include <sys/namei.h>
56 #include <sys/socket.h>
58 #include <sys/malloc.h>
59 #include <sys/sysent.h>
60 #include <sys/syscall.h>
64 #include <vm/vm_object.h>
65 #include <vm/vm_extern.h>
66 #include <vm/vm_zone.h>
76 #include "nfsm_subs.h"
80 #include <netinet/in.h>
83 * Data items converted to xdr at startup, since they are constant
84 * This is kinda hokey, but may save a little time doing byte swaps
86 u_int32_t nfs_xdrneg1;
87 u_int32_t rpc_call, rpc_vers, rpc_reply, rpc_msgdenied, rpc_autherr,
88 rpc_mismatch, rpc_auth_unix, rpc_msgaccepted,
90 u_int32_t nfs_prog, nqnfs_prog, nfs_true, nfs_false;
92 /* And other global data */
93 static u_int32_t nfs_xid = 0;
94 static enum vtype nv2tov_type[8]= {
95 VNON, VREG, VDIR, VBLK, VCHR, VLNK, VNON, VNON
97 enum vtype nv3tov_type[8]= {
98 VNON, VREG, VDIR, VBLK, VCHR, VLNK, VSOCK, VFIFO
102 int nfs_pbuf_freecnt = -1; /* start out unlimited */
104 struct nfs_reqq nfs_reqq;
105 struct nfssvc_sockhead nfssvc_sockhead;
106 int nfssvc_sockhead_flag;
107 struct nfsd_head nfsd_head;
109 struct nfs_bufq nfs_bufq;
110 struct nqtimerhead nqtimerhead;
111 struct nqfhhashhead *nqfhhashtbl;
114 static void (*nfs_prev_lease_updatetime) (int);
115 static int nfs_prev_nfssvc_sy_narg;
116 static sy_call_t *nfs_prev_nfssvc_sy_call;
120 static int (*nfs_prev_vop_lease_check)(struct vop_lease_args *);
123 * Mapping of old NFS Version 2 RPC numbers to generic numbers.
125 int nfsv3_procid[NFS_NPROCS] = {
154 #endif /* NFS_NOSERVER */
156 * and the reverse mapping from generic to Version 2 procedure numbers
158 int nfsv2_procid[NFS_NPROCS] = {
189 * Maps errno values to nfs error numbers.
190 * Use NFSERR_IO as the catch all for ones not specifically defined in
193 static u_char nfsrv_v2errmap[ELAST] = {
194 NFSERR_PERM, NFSERR_NOENT, NFSERR_IO, NFSERR_IO, NFSERR_IO,
195 NFSERR_NXIO, NFSERR_IO, NFSERR_IO, NFSERR_IO, NFSERR_IO,
196 NFSERR_IO, NFSERR_IO, NFSERR_ACCES, NFSERR_IO, NFSERR_IO,
197 NFSERR_IO, NFSERR_EXIST, NFSERR_IO, NFSERR_NODEV, NFSERR_NOTDIR,
198 NFSERR_ISDIR, NFSERR_IO, NFSERR_IO, NFSERR_IO, NFSERR_IO,
199 NFSERR_IO, NFSERR_FBIG, NFSERR_NOSPC, NFSERR_IO, NFSERR_ROFS,
200 NFSERR_IO, NFSERR_IO, NFSERR_IO, NFSERR_IO, NFSERR_IO,
201 NFSERR_IO, NFSERR_IO, NFSERR_IO, NFSERR_IO, NFSERR_IO,
202 NFSERR_IO, NFSERR_IO, NFSERR_IO, NFSERR_IO, NFSERR_IO,
203 NFSERR_IO, NFSERR_IO, NFSERR_IO, NFSERR_IO, NFSERR_IO,
204 NFSERR_IO, NFSERR_IO, NFSERR_IO, NFSERR_IO, NFSERR_IO,
205 NFSERR_IO, NFSERR_IO, NFSERR_IO, NFSERR_IO, NFSERR_IO,
206 NFSERR_IO, NFSERR_IO, NFSERR_NAMETOL, NFSERR_IO, NFSERR_IO,
207 NFSERR_NOTEMPTY, NFSERR_IO, NFSERR_IO, NFSERR_DQUOT, NFSERR_STALE,
208 NFSERR_IO, NFSERR_IO, NFSERR_IO, NFSERR_IO, NFSERR_IO,
209 NFSERR_IO, NFSERR_IO, NFSERR_IO, NFSERR_IO, NFSERR_IO,
210 NFSERR_IO, NFSERR_IO, NFSERR_IO, NFSERR_IO, NFSERR_IO,
211 NFSERR_IO /* << Last is 86 */
215 * Maps errno values to nfs error numbers.
216 * Although it is not obvious whether or not NFS clients really care if
217 * a returned error value is in the specified list for the procedure, the
218 * safest thing to do is filter them appropriately. For Version 2, the
219 * X/Open XNFS document is the only specification that defines error values
220 * for each RPC (The RFC simply lists all possible error values for all RPCs),
221 * so I have decided to not do this for Version 2.
222 * The first entry is the default error return and the rest are the valid
223 * errors for that RPC in increasing numeric order.
225 static short nfsv3err_null[] = {
230 static short nfsv3err_getattr[] = {
239 static short nfsv3err_setattr[] = {
255 static short nfsv3err_lookup[] = {
268 static short nfsv3err_access[] = {
277 static short nfsv3err_readlink[] = {
289 static short nfsv3err_read[] = {
301 static short nfsv3err_write[] = {
316 static short nfsv3err_create[] = {
333 static short nfsv3err_mkdir[] = {
350 static short nfsv3err_symlink[] = {
367 static short nfsv3err_mknod[] = {
385 static short nfsv3err_remove[] = {
399 static short nfsv3err_rmdir[] = {
417 static short nfsv3err_rename[] = {
440 static short nfsv3err_link[] = {
460 static short nfsv3err_readdir[] = {
473 static short nfsv3err_readdirplus[] = {
487 static short nfsv3err_fsstat[] = {
496 static short nfsv3err_fsinfo[] = {
504 static short nfsv3err_pathconf[] = {
512 static short nfsv3err_commit[] = {
521 static short *nfsrv_v3errmap[] = {
539 nfsv3err_readdirplus,
546 #endif /* NFS_NOSERVER */
548 extern struct nfsrtt nfsrtt;
549 extern time_t nqnfsstarttime;
550 extern int nqsrv_clockskew;
551 extern int nqsrv_writeslack;
552 extern int nqsrv_maxlease;
553 extern struct nfsstats nfsstats;
554 extern int nqnfs_piggy[NFS_NPROCS];
555 extern nfstype nfsv2_type[9];
556 extern nfstype nfsv3_type[9];
557 extern struct nfsnodehashhead *nfsnodehashtbl;
558 extern u_long nfsnodehash;
561 extern int nfssvc(struct proc *, struct nfssvc_args *, int *);
563 LIST_HEAD(nfsnodehashhead, nfsnode);
571 return ((u_quad_t)tv.tv_sec * 1000000 + (u_quad_t)tv.tv_usec);
575 * Create the header for an rpc request packet
576 * The hsiz is the size of the rest of the nfs request header.
577 * (just used to decide if a cluster is a good idea)
580 nfsm_reqh(struct vnode *vp, u_long procid, int hsiz, caddr_t *bposp)
586 struct nfsmount *nmp;
589 mb = m_getl(hsiz, MB_WAIT, MT_DATA, 0, NULL);
591 bpos = mtod(mb, caddr_t);
594 * For NQNFS, add lease request.
597 nmp = VFSTONFS(vp->v_mount);
598 if (nmp->nm_flag & NFSMNT_NQNFS) {
599 nqflag = NQNFS_NEEDLEASE(vp, procid);
601 nfsm_build(tl, u_int32_t *, 2*NFSX_UNSIGNED);
602 *tl++ = txdr_unsigned(nqflag);
603 *tl = txdr_unsigned(nmp->nm_leaseterm);
605 nfsm_build(tl, u_int32_t *, NFSX_UNSIGNED);
610 /* Finally, return values */
616 * Build the RPC header and fill in the authorization info.
617 * The authorization string argument is only used when the credentials
618 * come from outside of the kernel.
619 * Returns the head of the mbuf list.
622 nfsm_rpchead(struct ucred *cr, int nmflag, int procid, int auth_type,
623 int auth_len, char *auth_str, int verf_len, char *verf_str,
624 struct mbuf *mrest, int mrest_len, struct mbuf **mbp,
631 struct mbuf *mreq, *mb2;
632 int siz, grpsiz, authsiz;
634 authsiz = nfsm_rndup(auth_len);
635 MGETHDR(mb, MB_WAIT, MT_DATA);
636 if ((authsiz + 10 * NFSX_UNSIGNED) >= MINCLSIZE) {
638 } else if ((authsiz + 10 * NFSX_UNSIGNED) < MHLEN) {
639 MH_ALIGN(mb, authsiz + 10 * NFSX_UNSIGNED);
641 MH_ALIGN(mb, 8 * NFSX_UNSIGNED);
645 bpos = mtod(mb, caddr_t);
648 * First the RPC header.
650 nfsm_build(tl, u_int32_t *, 8 * NFSX_UNSIGNED);
652 /* Get a pretty random xid to start with */
656 * Skip zero xid if it should ever happen.
661 *tl++ = *xidp = txdr_unsigned(nfs_xid);
664 if (nmflag & NFSMNT_NQNFS) {
665 *tl++ = txdr_unsigned(NQNFS_PROG);
666 *tl++ = txdr_unsigned(NQNFS_VER3);
668 *tl++ = txdr_unsigned(NFS_PROG);
669 if (nmflag & NFSMNT_NFSV3)
670 *tl++ = txdr_unsigned(NFS_VER3);
672 *tl++ = txdr_unsigned(NFS_VER2);
674 if (nmflag & NFSMNT_NFSV3)
675 *tl++ = txdr_unsigned(procid);
677 *tl++ = txdr_unsigned(nfsv2_procid[procid]);
680 * And then the authorization cred.
682 *tl++ = txdr_unsigned(auth_type);
683 *tl = txdr_unsigned(authsiz);
686 nfsm_build(tl, u_int32_t *, auth_len);
687 *tl++ = 0; /* stamp ?? */
688 *tl++ = 0; /* NULL hostname */
689 *tl++ = txdr_unsigned(cr->cr_uid);
690 *tl++ = txdr_unsigned(cr->cr_groups[0]);
691 grpsiz = (auth_len >> 2) - 5;
692 *tl++ = txdr_unsigned(grpsiz);
693 for (i = 1; i <= grpsiz; i++)
694 *tl++ = txdr_unsigned(cr->cr_groups[i]);
699 if (M_TRAILINGSPACE(mb) == 0) {
700 mb2 = m_getl(siz, MB_WAIT, MT_DATA, 0, NULL);
704 bpos = mtod(mb, caddr_t);
706 i = min(siz, M_TRAILINGSPACE(mb));
707 bcopy(auth_str, bpos, i);
713 if ((siz = (nfsm_rndup(auth_len) - auth_len)) > 0) {
714 for (i = 0; i < siz; i++)
722 * And the verifier...
724 nfsm_build(tl, u_int32_t *, 2 * NFSX_UNSIGNED);
726 *tl++ = txdr_unsigned(RPCAUTH_KERB4);
727 *tl = txdr_unsigned(verf_len);
730 if (M_TRAILINGSPACE(mb) == 0) {
731 mb2 = m_getl(siz, MB_WAIT, MT_DATA, 0, NULL);
735 bpos = mtod(mb, caddr_t);
737 i = min(siz, M_TRAILINGSPACE(mb));
738 bcopy(verf_str, bpos, i);
744 if ((siz = (nfsm_rndup(verf_len) - verf_len)) > 0) {
745 for (i = 0; i < siz; i++)
750 *tl++ = txdr_unsigned(RPCAUTH_NULL);
754 mreq->m_pkthdr.len = authsiz + 10 * NFSX_UNSIGNED + mrest_len;
755 mreq->m_pkthdr.rcvif = (struct ifnet *)0;
761 * copies mbuf chain to the uio scatter/gather list
764 nfsm_mbuftouio(struct mbuf **mrep, struct uio *uiop, int siz, caddr_t *dpos)
766 char *mbufcp, *uiocp;
774 len = mtod(mp, caddr_t)+mp->m_len-mbufcp;
775 rem = nfsm_rndup(siz)-siz;
777 if (uiop->uio_iovcnt <= 0 || uiop->uio_iov == NULL)
779 left = uiop->uio_iov->iov_len;
780 uiocp = uiop->uio_iov->iov_base;
789 mbufcp = mtod(mp, caddr_t);
792 xfer = (left > len) ? len : left;
795 if (uiop->uio_iov->iov_op != NULL)
796 (*(uiop->uio_iov->iov_op))
797 (mbufcp, uiocp, xfer);
800 if (uiop->uio_segflg == UIO_SYSSPACE)
801 bcopy(mbufcp, uiocp, xfer);
803 copyout(mbufcp, uiocp, xfer);
808 uiop->uio_offset += xfer;
809 uiop->uio_resid -= xfer;
811 if (uiop->uio_iov->iov_len <= siz) {
815 uiop->uio_iov->iov_base += uiosiz;
816 uiop->uio_iov->iov_len -= uiosiz;
824 error = nfs_adv(mrep, dpos, rem, len);
832 * copies a uio scatter/gather list to an mbuf chain.
833 * NOTE: can ony handle iovcnt == 1
836 nfsm_uiotombuf(struct uio *uiop, struct mbuf **mq, int siz, caddr_t *bpos)
839 struct mbuf *mp, *mp2;
840 int xfer, left, mlen;
842 boolean_t getcluster;
846 if (uiop->uio_iovcnt != 1)
847 panic("nfsm_uiotombuf: iovcnt != 1");
850 if (siz >= MINCLSIZE)
854 rem = nfsm_rndup(siz) - siz;
857 left = uiop->uio_iov->iov_len;
858 uiocp = uiop->uio_iov->iov_base;
863 mlen = M_TRAILINGSPACE(mp);
866 mp = m_getcl(MB_WAIT, MT_DATA, 0);
868 mp = m_get(MB_WAIT, MT_DATA);
872 mlen = M_TRAILINGSPACE(mp);
874 xfer = (left > mlen) ? mlen : left;
877 if (uiop->uio_iov->iov_op != NULL)
878 (*(uiop->uio_iov->iov_op))
879 (uiocp, mtod(mp, caddr_t)+mp->m_len, xfer);
882 if (uiop->uio_segflg == UIO_SYSSPACE)
883 bcopy(uiocp, mtod(mp, caddr_t)+mp->m_len, xfer);
885 copyin(uiocp, mtod(mp, caddr_t)+mp->m_len, xfer);
889 uiop->uio_offset += xfer;
890 uiop->uio_resid -= xfer;
892 uiop->uio_iov->iov_base += uiosiz;
893 uiop->uio_iov->iov_len -= uiosiz;
897 if (rem > M_TRAILINGSPACE(mp)) {
898 MGET(mp, MB_WAIT, MT_DATA);
902 cp = mtod(mp, caddr_t)+mp->m_len;
903 for (left = 0; left < rem; left++)
908 *bpos = mtod(mp, caddr_t)+mp->m_len;
914 * Help break down an mbuf chain by setting the first siz bytes contiguous
915 * pointed to by returned val.
916 * This is used by the macros nfsm_dissect and nfsm_dissecton for tough
917 * cases. (The macros use the vars. dpos and dpos2)
920 nfsm_disct(struct mbuf **mdp, caddr_t *dposp, int siz, int left, caddr_t *cp2)
922 struct mbuf *mp, *mp2;
928 *mdp = mp = mp->m_next;
932 *dposp = mtod(mp, caddr_t);
937 } else if (mp->m_next == NULL) {
939 } else if (siz > MHLEN) {
940 panic("nfs S too big");
942 MGET(mp2, MB_WAIT, MT_DATA);
943 mp2->m_next = mp->m_next;
947 *cp2 = p = mtod(mp, caddr_t);
948 bcopy(*dposp, p, left); /* Copy what was left */
952 /* Loop around copying up the siz2 bytes */
956 xfer = (siz2 > mp2->m_len) ? mp2->m_len : siz2;
958 bcopy(mtod(mp2, caddr_t), p, xfer);
969 *dposp = mtod(mp2, caddr_t);
975 * Advance the position in the mbuf chain.
978 nfs_adv(struct mbuf **mdp, caddr_t *dposp, int offs, int left)
993 *dposp = mtod(m, caddr_t)+offs;
998 * Copy a string into mbufs for the hard cases...
1001 nfsm_strtmbuf(struct mbuf **mb, char **bpos, const char *cp, long siz)
1003 struct mbuf *m1 = NULL, *m2;
1004 long left, xfer, len, tlen;
1010 left = M_TRAILINGSPACE(m2);
1012 tl = ((u_int32_t *)(*bpos));
1013 *tl++ = txdr_unsigned(siz);
1015 left -= NFSX_UNSIGNED;
1016 m2->m_len += NFSX_UNSIGNED;
1018 bcopy(cp, (caddr_t) tl, left);
1025 /* Loop around adding mbufs */
1029 m1 = m_getl(siz, MB_WAIT, MT_DATA, 0, &msize);
1033 tl = mtod(m1, u_int32_t *);
1036 *tl++ = txdr_unsigned(siz);
1037 m1->m_len -= NFSX_UNSIGNED;
1038 tlen = NFSX_UNSIGNED;
1041 if (siz < m1->m_len) {
1042 len = nfsm_rndup(siz);
1045 *(tl+(xfer>>2)) = 0;
1047 xfer = len = m1->m_len;
1049 bcopy(cp, (caddr_t) tl, xfer);
1050 m1->m_len = len+tlen;
1055 *bpos = mtod(m1, caddr_t)+m1->m_len;
1060 * Called once to initialize data structures...
1063 nfs_init(struct vfsconf *vfsp)
1067 callout_init(&nfs_timer_handle);
1068 nfsmount_zone = zinit("NFSMOUNT", sizeof(struct nfsmount), 0, 0, 1);
1070 nfs_mount_type = vfsp->vfc_typenum;
1072 rpc_vers = txdr_unsigned(RPC_VER2);
1073 rpc_call = txdr_unsigned(RPC_CALL);
1074 rpc_reply = txdr_unsigned(RPC_REPLY);
1075 rpc_msgdenied = txdr_unsigned(RPC_MSGDENIED);
1076 rpc_msgaccepted = txdr_unsigned(RPC_MSGACCEPTED);
1077 rpc_mismatch = txdr_unsigned(RPC_MISMATCH);
1078 rpc_autherr = txdr_unsigned(RPC_AUTHERR);
1079 rpc_auth_unix = txdr_unsigned(RPCAUTH_UNIX);
1080 rpc_auth_kerb = txdr_unsigned(RPCAUTH_KERB4);
1081 nfs_prog = txdr_unsigned(NFS_PROG);
1082 nqnfs_prog = txdr_unsigned(NQNFS_PROG);
1083 nfs_true = txdr_unsigned(TRUE);
1084 nfs_false = txdr_unsigned(FALSE);
1085 nfs_xdrneg1 = txdr_unsigned(-1);
1086 nfs_ticks = (hz * NFS_TICKINTVL + 500) / 1000;
1089 /* Ensure async daemons disabled */
1090 for (i = 0; i < NFS_MAXASYNCDAEMON; i++) {
1091 nfs_iodwant[i] = NULL;
1092 nfs_iodmount[i] = (struct nfsmount *)0;
1094 nfs_nhinit(); /* Init the nfsnode table */
1095 #ifndef NFS_NOSERVER
1096 nfsrv_init(0); /* Init server data structures */
1097 nfsrv_initcache(); /* Init the server request cache */
1101 * Initialize the nqnfs server stuff.
1103 if (nqnfsstarttime == 0) {
1104 nqnfsstarttime = boottime.tv_sec + nqsrv_maxlease
1105 + nqsrv_clockskew + nqsrv_writeslack;
1106 NQLOADNOVRAM(nqnfsstarttime);
1107 CIRCLEQ_INIT(&nqtimerhead);
1108 nqfhhashtbl = hashinit(NQLCHSZ, M_NQLEASE, &nqfhhash);
1112 * Initialize reply list and start timer
1114 TAILQ_INIT(&nfs_reqq);
1119 * Set up lease_check and lease_updatetime so that other parts
1120 * of the system can call us, if we are loadable.
1122 #ifndef NFS_NOSERVER
1123 nfs_prev_vop_lease_check = default_vnode_vops->vop_lease;
1124 default_vnode_vops->vop_lease = nqnfs_vop_lease_check;
1126 nfs_prev_lease_updatetime = lease_updatetime;
1127 lease_updatetime = nfs_lease_updatetime;
1128 nfs_prev_nfssvc_sy_narg = sysent[SYS_nfssvc].sy_narg;
1129 sysent[SYS_nfssvc].sy_narg = 2;
1130 nfs_prev_nfssvc_sy_call = sysent[SYS_nfssvc].sy_call;
1131 sysent[SYS_nfssvc].sy_call = (sy_call_t *)nfssvc;
1133 nfs_pbuf_freecnt = nswbuf / 2 + 1;
1139 nfs_uninit(struct vfsconf *vfsp)
1141 callout_stop(&nfs_timer_handle);
1142 nfs_mount_type = -1;
1143 #ifndef NFS_NOSERVER
1144 default_vnode_vops->vop_lease = nfs_prev_vop_lease_check;
1146 lease_updatetime = nfs_prev_lease_updatetime;
1147 sysent[SYS_nfssvc].sy_narg = nfs_prev_nfssvc_sy_narg;
1148 sysent[SYS_nfssvc].sy_call = nfs_prev_nfssvc_sy_call;
1153 * Attribute cache routines.
1154 * nfs_loadattrcache() - loads or updates the cache contents from attributes
1155 * that are on the mbuf list
1156 * nfs_getattrcache() - returns valid attributes if found in cache, returns
1161 * Load the attribute cache (that lives in the nfsnode entry) with
1162 * the values on the mbuf list. Load *vaper with the attributes. vaper
1165 * As a side effect n_mtime, which we use to determine if the file was
1166 * modified by some other host, is set to the attribute timestamp and
1167 * NRMODIFIED is set if the two values differ.
1169 * WARNING: the mtime loaded into vaper does not necessarily represent
1170 * n_mtime or n_attr.mtime due to NACC and NUPD.
1173 nfs_loadattrcache(struct vnode **vpp, struct mbuf **mdp, caddr_t *dposp,
1174 struct vattr *vaper, int lattr_flags)
1176 struct vnode *vp = *vpp;
1178 struct nfs_fattr *fp;
1187 struct timespec mtime;
1188 int v3 = NFS_ISV3(vp);
1191 t1 = (mtod(md, caddr_t) + md->m_len) - *dposp;
1192 if ((error = nfsm_disct(mdp, dposp, NFSX_FATTR(v3), t1, &cp2)) != 0)
1194 fp = (struct nfs_fattr *)cp2;
1196 vtyp = nfsv3tov_type(fp->fa_type);
1197 vmode = fxdr_unsigned(u_short, fp->fa_mode);
1198 rdev = makeudev(fxdr_unsigned(int, fp->fa3_rdev.specdata1),
1199 fxdr_unsigned(int, fp->fa3_rdev.specdata2));
1200 fxdr_nfsv3time(&fp->fa3_mtime, &mtime);
1202 vtyp = nfsv2tov_type(fp->fa_type);
1203 vmode = fxdr_unsigned(u_short, fp->fa_mode);
1207 * The duplicate information returned in fa_type and fa_mode
1208 * is an ambiguity in the NFS version 2 protocol.
1210 * VREG should be taken literally as a regular file. If a
1211 * server intents to return some type information differently
1212 * in the upper bits of the mode field (e.g. for sockets, or
1213 * FIFOs), NFSv2 mandates fa_type to be VNON. Anyway, we
1214 * leave the examination of the mode bits even in the VREG
1215 * case to avoid breakage for bogus servers, but we make sure
1216 * that there are actually type bits set in the upper part of
1217 * fa_mode (and failing that, trust the va_type field).
1219 * NFSv3 cleared the issue, and requires fa_mode to not
1220 * contain any type information (while also introduing sockets
1221 * and FIFOs for fa_type).
1223 if (vtyp == VNON || (vtyp == VREG && (vmode & S_IFMT) != 0))
1224 vtyp = IFTOVT(vmode);
1225 rdev = fxdr_unsigned(int32_t, fp->fa2_rdev);
1226 fxdr_nfsv2time(&fp->fa2_mtime, &mtime);
1229 * Really ugly NFSv2 kludge.
1231 if (vtyp == VCHR && rdev == (udev_t)0xffffffff)
1236 * If v_type == VNON it is a new node, so fill in the v_type,
1237 * n_mtime fields. Check to see if it represents a special
1238 * device, and if so, check for a possible alias. Once the
1239 * correct vnode has been obtained, fill in the rest of the
1243 if (vp->v_type != vtyp) {
1245 if (vp->v_type == VFIFO) {
1246 vp->v_ops = &vp->v_mount->mnt_vn_fifo_ops;
1247 } else if (vp->v_type == VCHR || vp->v_type == VBLK) {
1248 vp->v_ops = &vp->v_mount->mnt_vn_spec_ops;
1249 addaliasu(vp, rdev);
1251 vp->v_ops = &vp->v_mount->mnt_vn_use_ops;
1253 np->n_mtime = mtime.tv_sec;
1254 } else if (np->n_mtime != mtime.tv_sec) {
1256 * If we haven't modified the file locally and the server
1257 * timestamp does not match, then the server probably
1258 * modified the file. We must flag this condition so
1259 * the proper syncnronization can be done. We do not
1260 * try to synchronize the state here because that
1261 * could lead to an endless recursion.
1263 * XXX loadattrcache can be set during the reply to a write,
1264 * before the write timestamp is properly processed. To
1265 * avoid unconditionally setting the rmodified bit (which
1266 * has the effect of flushing the cache), we only do this
1267 * check if the lmodified bit is not set.
1269 np->n_mtime = mtime.tv_sec;
1270 if ((lattr_flags & NFS_LATTR_NOMTIMECHECK) == 0)
1271 np->n_flag |= NRMODIFIED;
1274 vap->va_type = vtyp;
1275 vap->va_mode = (vmode & 07777);
1276 vap->va_rdev = rdev;
1277 vap->va_mtime = mtime;
1278 vap->va_fsid = vp->v_mount->mnt_stat.f_fsid.val[0];
1280 vap->va_nlink = fxdr_unsigned(u_short, fp->fa_nlink);
1281 vap->va_uid = fxdr_unsigned(uid_t, fp->fa_uid);
1282 vap->va_gid = fxdr_unsigned(gid_t, fp->fa_gid);
1283 vap->va_size = fxdr_hyper(&fp->fa3_size);
1284 vap->va_blocksize = NFS_FABLKSIZE;
1285 vap->va_bytes = fxdr_hyper(&fp->fa3_used);
1286 vap->va_fileid = fxdr_unsigned(int32_t,
1287 fp->fa3_fileid.nfsuquad[1]);
1288 fxdr_nfsv3time(&fp->fa3_atime, &vap->va_atime);
1289 fxdr_nfsv3time(&fp->fa3_ctime, &vap->va_ctime);
1291 vap->va_filerev = 0;
1293 vap->va_nlink = fxdr_unsigned(u_short, fp->fa_nlink);
1294 vap->va_uid = fxdr_unsigned(uid_t, fp->fa_uid);
1295 vap->va_gid = fxdr_unsigned(gid_t, fp->fa_gid);
1296 vap->va_size = fxdr_unsigned(u_int32_t, fp->fa2_size);
1297 vap->va_blocksize = fxdr_unsigned(int32_t, fp->fa2_blocksize);
1298 vap->va_bytes = (u_quad_t)fxdr_unsigned(int32_t, fp->fa2_blocks)
1300 vap->va_fileid = fxdr_unsigned(int32_t, fp->fa2_fileid);
1301 fxdr_nfsv2time(&fp->fa2_atime, &vap->va_atime);
1303 vap->va_ctime.tv_sec = fxdr_unsigned(u_int32_t,
1304 fp->fa2_ctime.nfsv2_sec);
1305 vap->va_ctime.tv_nsec = 0;
1306 vap->va_gen = fxdr_unsigned(u_int32_t,fp->fa2_ctime.nfsv2_usec);
1307 vap->va_filerev = 0;
1309 np->n_attrstamp = time_second;
1310 if (vap->va_size != np->n_size) {
1311 if (vap->va_type == VREG) {
1312 if ((lattr_flags & NFS_LATTR_NOSHRINK) &&
1313 vap->va_size < np->n_size) {
1315 * We've been told not to shrink the file;
1316 * zero np->n_attrstamp to indicate that
1317 * the attributes are stale.
1319 * This occurs primarily due to recursive
1320 * NFS ops that are executed during periods
1321 * where we cannot safely reduce the size of
1324 * Additionally, write rpcs are broken down
1325 * into buffers and np->n_size is
1326 * pre-extended. Setting NRMODIFIED here
1327 * can result in n_size getting reset to a
1328 * lower value, which is NOT what we want.
1329 * XXX this needs to be cleaned up a lot
1332 vap->va_size = np->n_size;
1333 np->n_attrstamp = 0;
1334 if ((np->n_flag & NLMODIFIED) == 0)
1335 np->n_flag |= NRMODIFIED;
1336 } else if (np->n_flag & NLMODIFIED) {
1338 * We've modified the file: Use the larger
1339 * of our size, and the server's size. At
1340 * this point the cache coherency is all
1341 * shot to hell. To try to handle multiple
1342 * clients appending to the file at the same
1343 * time mark that the server has changed
1344 * the file if the server's notion of the
1345 * file size is larger then our notion.
1347 * XXX this needs work.
1349 if (vap->va_size < np->n_size) {
1350 vap->va_size = np->n_size;
1352 np->n_size = vap->va_size;
1353 np->n_flag |= NRMODIFIED;
1357 * Someone changed the file's size on the
1358 * server and there are no local changes
1359 * to get in the way, set the size and mark
1362 np->n_size = vap->va_size;
1363 np->n_flag |= NRMODIFIED;
1365 vnode_pager_setsize(vp, np->n_size);
1367 np->n_size = vap->va_size;
1370 if (vaper != NULL) {
1371 bcopy((caddr_t)vap, (caddr_t)vaper, sizeof(*vap));
1372 if (np->n_flag & NCHG) {
1373 if (np->n_flag & NACC)
1374 vaper->va_atime = np->n_atim;
1375 if (np->n_flag & NUPD)
1376 vaper->va_mtime = np->n_mtim;
1383 #include <sys/sysctl.h>
1384 SYSCTL_DECL(_vfs_nfs);
1385 static int nfs_acdebug;
1386 SYSCTL_INT(_vfs_nfs, OID_AUTO, acdebug, CTLFLAG_RW, &nfs_acdebug, 0, "");
1390 * Check the time stamp
1391 * If the cache is valid, copy contents to *vap and return 0
1392 * otherwise return an error
1395 nfs_getattrcache(struct vnode *vp, struct vattr *vaper)
1399 struct nfsmount *nmp;
1404 nmp = VFSTONFS(vp->v_mount);
1407 * Dynamic timeout based on how recently the file was modified.
1408 * n_mtime is always valid.
1410 timeo = (get_approximate_time_t() - np->n_mtime) / 10;
1414 printf("nfs_getattrcache: initial timeo = %d\n", timeo);
1417 if (vap->va_type == VDIR) {
1418 if ((np->n_flag & NLMODIFIED) || timeo < nmp->nm_acdirmin)
1419 timeo = nmp->nm_acdirmin;
1420 else if (timeo > nmp->nm_acdirmax)
1421 timeo = nmp->nm_acdirmax;
1423 if ((np->n_flag & NLMODIFIED) || timeo < nmp->nm_acregmin)
1424 timeo = nmp->nm_acregmin;
1425 else if (timeo > nmp->nm_acregmax)
1426 timeo = nmp->nm_acregmax;
1430 if (nfs_acdebug > 2)
1431 printf("acregmin %d; acregmax %d; acdirmin %d; acdirmax %d\n",
1432 nmp->nm_acregmin, nmp->nm_acregmax,
1433 nmp->nm_acdirmin, nmp->nm_acdirmax);
1436 printf("nfs_getattrcache: age = %d; final timeo = %d\n",
1437 (int)(time_second - np->n_attrstamp), timeo);
1440 if (np->n_attrstamp == 0 || (time_second - np->n_attrstamp) >= timeo) {
1441 nfsstats.attrcache_misses++;
1444 nfsstats.attrcache_hits++;
1447 * Our attribute cache can be stale due to modifications made on
1448 * this host. XXX this is a bad hack. We need a more deterministic
1449 * means of finding out which np fields are valid verses attr cache
1450 * fields. We really should update the vattr info on the fly when
1451 * making local changes.
1453 if (vap->va_size != np->n_size) {
1454 if (vap->va_type == VREG) {
1455 if (np->n_flag & NLMODIFIED) {
1456 if (vap->va_size < np->n_size)
1457 vap->va_size = np->n_size;
1459 np->n_size = vap->va_size;
1461 np->n_size = vap->va_size;
1463 vnode_pager_setsize(vp, np->n_size);
1465 np->n_size = vap->va_size;
1468 bcopy((caddr_t)vap, (caddr_t)vaper, sizeof(struct vattr));
1469 if (np->n_flag & NCHG) {
1470 if (np->n_flag & NACC)
1471 vaper->va_atime = np->n_atim;
1472 if (np->n_flag & NUPD)
1473 vaper->va_mtime = np->n_mtim;
1478 #ifndef NFS_NOSERVER
1481 * Set up nameidata for a lookup() call and do it.
1483 * If pubflag is set, this call is done for a lookup operation on the
1484 * public filehandle. In that case we allow crossing mountpoints and
1485 * absolute pathnames. However, the caller is expected to check that
1486 * the lookup result is within the public fs, and deny access if
1489 * dirp may be set whether an error is returned or not, and must be
1490 * released by the caller.
1492 * On return nd->nl_ncp usually points to the target ncp, which may represent
1495 * NOTE: the caller must call nlookup_done(nd) unconditionally on return
1499 nfs_namei(struct nlookupdata *nd, struct ucred *cred, int nameiop,
1500 struct vnode **dvpp, struct vnode **vpp,
1501 fhandle_t *fhp, int len,
1502 struct nfssvc_sock *slp, struct sockaddr *nam, struct mbuf **mdp,
1503 caddr_t *dposp, struct vnode **dirpp, struct thread *td,
1504 int kerbflag, int pubflag)
1509 char *fromcp, *tocp, *cp;
1511 struct namecache *ncp;
1515 namebuf = zalloc(namei_zone);
1520 * Copy the name from the mbuf list to namebuf.
1525 rem = mtod(md, caddr_t) + md->m_len - fromcp;
1526 for (i = 0; i < len; i++) {
1533 fromcp = mtod(md, caddr_t);
1536 if (*fromcp == '\0' || (!pubflag && *fromcp == '/')) {
1540 *tocp++ = *fromcp++;
1546 len = nfsm_rndup(len)-len;
1550 else if ((error = nfs_adv(mdp, dposp, len, rem)) != 0)
1555 * Extract and set starting directory. The returned dp is refd
1558 error = nfsrv_fhtovp(fhp, FALSE, &dp, cred, slp,
1559 nam, &rdonly, kerbflag, pubflag);
1562 if (dp->v_type != VDIR) {
1569 * Set return directory. Reference to dp is implicitly transfered
1570 * to the returned pointer. This must be set before we potentially
1577 * Oh joy. For WebNFS, handle those pesky '%' escapes,
1578 * and the 'native path' indicator.
1580 cp = zalloc(namei_zone);
1583 if ((unsigned char)*fromcp >= WEBNFS_SPECCHAR_START) {
1584 switch ((unsigned char)*fromcp) {
1585 case WEBNFS_NATIVE_CHAR:
1587 * 'Native' path for us is the same
1588 * as a path according to the NFS spec,
1589 * just skip the escape char.
1594 * More may be added in the future, range 0x80-0xff
1598 zfree(namei_zone, cp);
1603 * Translate the '%' escapes, URL-style.
1605 while (*fromcp != '\0') {
1606 if (*fromcp == WEBNFS_ESC_CHAR) {
1607 if (fromcp[1] != '\0' && fromcp[2] != '\0') {
1609 *tocp++ = HEXSTRTOI(fromcp);
1614 zfree(namei_zone, cp);
1618 *tocp++ = *fromcp++;
1621 zfree(namei_zone, namebuf);
1626 * Setup for search. We need to get a start directory from dp. Note
1627 * that dp is ref'd, but we no longer 'own' the ref (*dirpp owns it).
1630 flags |= NLC_NFS_NOSOFTLINKTRAV;
1631 flags |= NLC_NOCROSSMOUNT;
1634 flags |= NLC_NFS_RDONLY;
1635 if (nameiop == NAMEI_CREATE || nameiop == NAMEI_RENAME)
1636 flags |= NLC_CREATE;
1639 * We need a starting ncp from the directory vnode dp. dp must not
1640 * be locked. The returned ncp will be refd but not locked.
1642 * If no suitable ncp is found we instruct cache_fromdvp() to create
1643 * one. If this fails the directory has probably been removed while
1644 * the target was chdir'd into it and any further lookup will fail.
1646 if ((ncp = cache_fromdvp(dp, cred, 1)) == NULL) {
1650 nlookup_init_raw(nd, namebuf, UIO_SYSSPACE, flags, cred, ncp);
1654 * Ok, do the lookup.
1656 error = nlookup(nd);
1659 * If no error occured return the requested dvpp and vpp. If
1660 * NLC_CREATE was specified nd->nl_ncp may represent a negative
1661 * cache hit in which case we do not attempt to obtain the vp.
1666 if (ncp->nc_parent &&
1667 ncp->nc_parent->nc_mount == ncp->nc_mount) {
1668 error = cache_vget(ncp->nc_parent, nd->nl_cred,
1669 LK_EXCLUSIVE, dvpp);
1674 if (vpp && ncp->nc_vp) {
1675 error = cache_vget(ncp, nd->nl_cred, LK_EXCLUSIVE, vpp);
1678 if (dvpp && *dvpp) {
1693 zfree(namei_zone, namebuf);
1698 * A fiddled version of m_adj() that ensures null fill to a long
1699 * boundary and only trims off the back end
1702 nfsm_adj(struct mbuf *mp, int len, int nul)
1709 * Trim from tail. Scan the mbuf chain,
1710 * calculating its length and finding the last mbuf.
1711 * If the adjustment only affects this mbuf, then just
1712 * adjust and return. Otherwise, rescan and truncate
1713 * after the remaining size.
1719 if (m->m_next == (struct mbuf *)0)
1723 if (m->m_len > len) {
1726 cp = mtod(m, caddr_t)+m->m_len-nul;
1727 for (i = 0; i < nul; i++)
1736 * Correct length for chain is "count".
1737 * Find the mbuf with last data, adjust its length,
1738 * and toss data from remaining mbufs on chain.
1740 for (m = mp; m; m = m->m_next) {
1741 if (m->m_len >= count) {
1744 cp = mtod(m, caddr_t)+m->m_len-nul;
1745 for (i = 0; i < nul; i++)
1752 for (m = m->m_next;m;m = m->m_next)
1757 * Make these functions instead of macros, so that the kernel text size
1758 * doesn't get too big...
1761 nfsm_srvwcc(struct nfsrv_descript *nfsd, int before_ret,
1762 struct vattr *before_vap, int after_ret, struct vattr *after_vap,
1763 struct mbuf **mbp, char **bposp)
1765 struct mbuf *mb = *mbp, *mb2;
1766 char *bpos = *bposp;
1770 * before_ret is 0 if before_vap is valid, non-zero if it isn't.
1773 nfsm_build(tl, u_int32_t *, NFSX_UNSIGNED);
1776 nfsm_build(tl, u_int32_t *, 7 * NFSX_UNSIGNED);
1778 txdr_hyper(before_vap->va_size, tl);
1780 txdr_nfsv3time(&(before_vap->va_mtime), tl);
1782 txdr_nfsv3time(&(before_vap->va_ctime), tl);
1786 nfsm_srvpostopattr(nfsd, after_ret, after_vap, mbp, bposp);
1790 nfsm_srvpostopattr(struct nfsrv_descript *nfsd, int after_ret,
1791 struct vattr *after_vap, struct mbuf **mbp, char **bposp)
1793 struct mbuf *mb = *mbp, *mb2;
1794 char *bpos = *bposp;
1796 struct nfs_fattr *fp;
1799 nfsm_build(tl, u_int32_t *, NFSX_UNSIGNED);
1802 nfsm_build(tl, u_int32_t *, NFSX_UNSIGNED + NFSX_V3FATTR);
1804 fp = (struct nfs_fattr *)tl;
1805 nfsm_srvfattr(nfsd, after_vap, fp);
1812 nfsm_srvfattr(struct nfsrv_descript *nfsd, struct vattr *vap,
1813 struct nfs_fattr *fp)
1816 fp->fa_nlink = txdr_unsigned(vap->va_nlink);
1817 fp->fa_uid = txdr_unsigned(vap->va_uid);
1818 fp->fa_gid = txdr_unsigned(vap->va_gid);
1819 if (nfsd->nd_flag & ND_NFSV3) {
1820 fp->fa_type = vtonfsv3_type(vap->va_type);
1821 fp->fa_mode = vtonfsv3_mode(vap->va_mode);
1822 txdr_hyper(vap->va_size, &fp->fa3_size);
1823 txdr_hyper(vap->va_bytes, &fp->fa3_used);
1824 fp->fa3_rdev.specdata1 = txdr_unsigned(umajor(vap->va_rdev));
1825 fp->fa3_rdev.specdata2 = txdr_unsigned(uminor(vap->va_rdev));
1826 fp->fa3_fsid.nfsuquad[0] = 0;
1827 fp->fa3_fsid.nfsuquad[1] = txdr_unsigned(vap->va_fsid);
1828 fp->fa3_fileid.nfsuquad[0] = 0;
1829 fp->fa3_fileid.nfsuquad[1] = txdr_unsigned(vap->va_fileid);
1830 txdr_nfsv3time(&vap->va_atime, &fp->fa3_atime);
1831 txdr_nfsv3time(&vap->va_mtime, &fp->fa3_mtime);
1832 txdr_nfsv3time(&vap->va_ctime, &fp->fa3_ctime);
1834 fp->fa_type = vtonfsv2_type(vap->va_type);
1835 fp->fa_mode = vtonfsv2_mode(vap->va_type, vap->va_mode);
1836 fp->fa2_size = txdr_unsigned(vap->va_size);
1837 fp->fa2_blocksize = txdr_unsigned(vap->va_blocksize);
1838 if (vap->va_type == VFIFO)
1839 fp->fa2_rdev = 0xffffffff;
1841 fp->fa2_rdev = txdr_unsigned(vap->va_rdev);
1842 fp->fa2_blocks = txdr_unsigned(vap->va_bytes / NFS_FABLKSIZE);
1843 fp->fa2_fsid = txdr_unsigned(vap->va_fsid);
1844 fp->fa2_fileid = txdr_unsigned(vap->va_fileid);
1845 txdr_nfsv2time(&vap->va_atime, &fp->fa2_atime);
1846 txdr_nfsv2time(&vap->va_mtime, &fp->fa2_mtime);
1847 txdr_nfsv2time(&vap->va_ctime, &fp->fa2_ctime);
1852 * nfsrv_fhtovp() - convert a fh to a vnode ptr (optionally locked)
1853 * - look up fsid in mount list (if not found ret error)
1854 * - get vp and export rights by calling VFS_FHTOVP()
1855 * - if cred->cr_uid == 0 or MNT_EXPORTANON set it to credanon
1856 * - if not lockflag unlock it with VOP_UNLOCK()
1859 nfsrv_fhtovp(fhandle_t *fhp, int lockflag, struct vnode **vpp,
1860 struct ucred *cred, struct nfssvc_sock *slp, struct sockaddr *nam,
1861 int *rdonlyp, int kerbflag, int pubflag)
1863 struct thread *td = curthread; /* XXX */
1866 struct ucred *credanon;
1868 #ifdef MNT_EXNORESPORT /* XXX needs mountd and /etc/exports help yet */
1869 struct sockaddr_int *saddr;
1872 *vpp = (struct vnode *)0;
1874 if (nfs_ispublicfh(fhp)) {
1875 if (!pubflag || !nfs_pub.np_valid)
1877 fhp = &nfs_pub.np_handle;
1880 mp = vfs_getvfs(&fhp->fh_fsid);
1883 error = VFS_CHECKEXP(mp, nam, &exflags, &credanon);
1886 error = VFS_FHTOVP(mp, &fhp->fh_fid, vpp);
1889 #ifdef MNT_EXNORESPORT
1890 if (!(exflags & (MNT_EXNORESPORT|MNT_EXPUBLIC))) {
1891 saddr = (struct sockaddr_in *)nam;
1892 if (saddr->sin_family == AF_INET &&
1893 ntohs(saddr->sin_port) >= IPPORT_RESERVED) {
1896 return (NFSERR_AUTHERR | AUTH_TOOWEAK);
1901 * Check/setup credentials.
1903 if (exflags & MNT_EXKERB) {
1907 return (NFSERR_AUTHERR | AUTH_TOOWEAK);
1909 } else if (kerbflag) {
1912 return (NFSERR_AUTHERR | AUTH_TOOWEAK);
1913 } else if (cred->cr_uid == 0 || (exflags & MNT_EXPORTANON)) {
1914 cred->cr_uid = credanon->cr_uid;
1915 for (i = 0; i < credanon->cr_ngroups && i < NGROUPS; i++)
1916 cred->cr_groups[i] = credanon->cr_groups[i];
1917 cred->cr_ngroups = i;
1919 if (exflags & MNT_EXRDONLY)
1924 nfsrv_object_create(*vpp);
1927 VOP_UNLOCK(*vpp, 0, td);
1933 * WebNFS: check if a filehandle is a public filehandle. For v3, this
1934 * means a length of 0, for v2 it means all zeroes. nfsm_srvmtofh has
1935 * transformed this to all zeroes in both cases, so check for it.
1938 nfs_ispublicfh(fhandle_t *fhp)
1940 char *cp = (char *)fhp;
1943 for (i = 0; i < NFSX_V3FH; i++)
1949 #endif /* NFS_NOSERVER */
1951 * This function compares two net addresses by family and returns TRUE
1952 * if they are the same host.
1953 * If there is any doubt, return FALSE.
1954 * The AF_INET family is handled as a special case so that address mbufs
1955 * don't need to be saved to store "struct in_addr", which is only 4 bytes.
1958 netaddr_match(int family, union nethostaddr *haddr, struct sockaddr *nam)
1960 struct sockaddr_in *inetaddr;
1964 inetaddr = (struct sockaddr_in *)nam;
1965 if (inetaddr->sin_family == AF_INET &&
1966 inetaddr->sin_addr.s_addr == haddr->had_inetaddr)
1975 static nfsuint64 nfs_nullcookie = { { 0, 0 } };
1977 * This function finds the directory cookie that corresponds to the
1978 * logical byte offset given.
1981 nfs_getcookie(struct nfsnode *np, off_t off, int add)
1983 struct nfsdmap *dp, *dp2;
1986 pos = (uoff_t)off / NFS_DIRBLKSIZ;
1987 if (pos == 0 || off < 0) {
1990 panic("nfs getcookie add at <= 0");
1992 return (&nfs_nullcookie);
1995 dp = np->n_cookies.lh_first;
1998 MALLOC(dp, struct nfsdmap *, sizeof (struct nfsdmap),
1999 M_NFSDIROFF, M_WAITOK);
2000 dp->ndm_eocookie = 0;
2001 LIST_INSERT_HEAD(&np->n_cookies, dp, ndm_list);
2003 return ((nfsuint64 *)0);
2005 while (pos >= NFSNUMCOOKIES) {
2006 pos -= NFSNUMCOOKIES;
2007 if (dp->ndm_list.le_next) {
2008 if (!add && dp->ndm_eocookie < NFSNUMCOOKIES &&
2009 pos >= dp->ndm_eocookie)
2010 return ((nfsuint64 *)0);
2011 dp = dp->ndm_list.le_next;
2013 MALLOC(dp2, struct nfsdmap *, sizeof (struct nfsdmap),
2014 M_NFSDIROFF, M_WAITOK);
2015 dp2->ndm_eocookie = 0;
2016 LIST_INSERT_AFTER(dp, dp2, ndm_list);
2019 return ((nfsuint64 *)0);
2021 if (pos >= dp->ndm_eocookie) {
2023 dp->ndm_eocookie = pos + 1;
2025 return ((nfsuint64 *)0);
2027 return (&dp->ndm_cookies[pos]);
2031 * Invalidate cached directory information, except for the actual directory
2032 * blocks (which are invalidated separately).
2033 * Done mainly to avoid the use of stale offset cookies.
2036 nfs_invaldir(struct vnode *vp)
2038 struct nfsnode *np = VTONFS(vp);
2041 if (vp->v_type != VDIR)
2042 panic("nfs: invaldir not dir");
2044 np->n_direofoffset = 0;
2045 np->n_cookieverf.nfsuquad[0] = 0;
2046 np->n_cookieverf.nfsuquad[1] = 0;
2047 if (np->n_cookies.lh_first)
2048 np->n_cookies.lh_first->ndm_eocookie = 0;
2052 * The write verifier has changed (probably due to a server reboot), so all
2053 * B_NEEDCOMMIT blocks will have to be written again. Since they are on the
2054 * dirty block list as B_DELWRI, all this takes is clearing the B_NEEDCOMMIT
2055 * and B_CLUSTEROK flags. Once done the new write verifier can be set for the
2058 * B_CLUSTEROK must be cleared along with B_NEEDCOMMIT because stage 1 data
2059 * writes are not clusterable.
2062 static int nfs_clearcommit_bp(struct buf *bp, void *data __unused);
2065 nfs_clearcommit(struct mount *mp)
2067 struct vnode *vp, *nvp;
2071 lwkt_gettoken(&ilock, &mntvnode_token);
2073 for (vp = TAILQ_FIRST(&mp->mnt_nvnodelist); vp; vp = nvp) {
2074 nvp = TAILQ_NEXT(vp, v_nmntvnodes); /* ZZZ */
2075 RB_SCAN(buf_rb_tree, &vp->v_rbdirty_tree, NULL,
2076 nfs_clearcommit_bp, NULL);
2079 lwkt_reltoken(&ilock);
2083 nfs_clearcommit_bp(struct buf *bp, void *data __unused)
2085 if (BUF_REFCNT(bp) == 0 &&
2086 (bp->b_flags & (B_DELWRI | B_NEEDCOMMIT))
2087 == (B_DELWRI | B_NEEDCOMMIT)) {
2088 bp->b_flags &= ~(B_NEEDCOMMIT | B_CLUSTEROK);
2093 #ifndef NFS_NOSERVER
2095 * Map errnos to NFS error numbers. For Version 3 also filter out error
2096 * numbers not specified for the associated procedure.
2099 nfsrv_errmap(struct nfsrv_descript *nd, int err)
2101 short *defaulterrp, *errp;
2103 if (nd->nd_flag & ND_NFSV3) {
2104 if (nd->nd_procnum <= NFSPROC_COMMIT) {
2105 errp = defaulterrp = nfsrv_v3errmap[nd->nd_procnum];
2109 else if (*errp > err)
2112 return ((int)*defaulterrp);
2114 return (err & 0xffff);
2117 return ((int)nfsrv_v2errmap[err - 1]);
2122 nfsrv_object_create(struct vnode *vp)
2124 struct thread *td = curthread;
2126 if (vp == NULL || vp->v_type != VREG)
2128 return (vfs_object_create(vp, td));
2132 * Sort the group list in increasing numerical order.
2133 * (Insertion sort by Chris Torek, who was grossed out by the bubble sort
2134 * that used to be here.)
2137 nfsrvw_sort(gid_t *list, int num)
2142 /* Insertion sort. */
2143 for (i = 1; i < num; i++) {
2145 /* find correct slot for value v, moving others up */
2146 for (j = i; --j >= 0 && v < list[j];)
2147 list[j + 1] = list[j];
2153 * copy credentials making sure that the result can be compared with bcmp().
2156 nfsrv_setcred(struct ucred *incred, struct ucred *outcred)
2160 bzero((caddr_t)outcred, sizeof (struct ucred));
2161 outcred->cr_ref = 1;
2162 outcred->cr_uid = incred->cr_uid;
2163 outcred->cr_ngroups = incred->cr_ngroups;
2164 for (i = 0; i < incred->cr_ngroups; i++)
2165 outcred->cr_groups[i] = incred->cr_groups[i];
2166 nfsrvw_sort(outcred->cr_groups, outcred->cr_ngroups);
2168 #endif /* NFS_NOSERVER */