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 * 4. Neither the name of the University nor the names of its contributors
17 * may be used to endorse or promote products derived from this software
18 * without specific prior written permission.
20 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
21 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
22 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
23 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
24 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
25 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
26 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
27 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
28 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
29 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
32 * @(#)nfs_subs.c 8.8 (Berkeley) 5/22/95
35 #include <sys/cdefs.h>
36 __FBSDID("$FreeBSD$");
39 * These functions support the macros and help fiddle mbuf chains for
40 * the nfs op functions. They do things like create the rpc header and
41 * copy data between mbuf chains and uio lists.
44 #include "opt_kdtrace.h"
46 #include <sys/param.h>
47 #include <sys/systm.h>
48 #include <sys/kernel.h>
52 #include <sys/mount.h>
53 #include <sys/vnode.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>
61 #include <sys/sysproto.h>
62 #include <sys/taskqueue.h>
65 #include <vm/vm_object.h>
66 #include <vm/vm_extern.h>
69 #include <nfs/nfsproto.h>
70 #include <nfsclient/nfs.h>
71 #include <nfsclient/nfsnode.h>
72 #include <nfs/nfs_kdtrace.h>
73 #include <nfs/xdr_subs.h>
74 #include <nfsclient/nfsm_subs.h>
75 #include <nfsclient/nfsmount.h>
77 #include <netinet/in.h>
80 * Note that stdarg.h and the ANSI style va_start macro is used for both
81 * ANSI and traditional C compilers.
83 #include <machine/stdarg.h>
86 dtrace_nfsclient_attrcache_flush_probe_func_t
87 dtrace_nfsclient_attrcache_flush_done_probe;
88 uint32_t nfsclient_attrcache_flush_done_id;
90 dtrace_nfsclient_attrcache_get_hit_probe_func_t
91 dtrace_nfsclient_attrcache_get_hit_probe;
92 uint32_t nfsclient_attrcache_get_hit_id;
94 dtrace_nfsclient_attrcache_get_miss_probe_func_t
95 dtrace_nfsclient_attrcache_get_miss_probe;
96 uint32_t nfsclient_attrcache_get_miss_id;
98 dtrace_nfsclient_attrcache_load_probe_func_t
99 dtrace_nfsclient_attrcache_load_done_probe;
100 uint32_t nfsclient_attrcache_load_done_id;
101 #endif /* !KDTRACE_HOOKS */
104 * Data items converted to xdr at startup, since they are constant
105 * This is kinda hokey, but may save a little time doing byte swaps
107 u_int32_t nfs_xdrneg1;
108 u_int32_t nfs_true, nfs_false;
110 /* And other global data */
111 static u_int32_t nfs_xid = 0;
112 static enum vtype nv2tov_type[8]= {
113 VNON, VREG, VDIR, VBLK, VCHR, VLNK, VNON, VNON
117 int nfs_pbuf_freecnt = -1; /* start out unlimited */
119 struct nfs_bufq nfs_bufq;
120 static struct mtx nfs_xid_mtx;
121 struct task nfs_nfsiodnew_task;
124 * and the reverse mapping from generic to Version 2 procedure numbers
126 int nfsv2_procid[NFS_NPROCS] = {
152 LIST_HEAD(nfsnodehashhead, nfsnode);
159 mtx_lock(&nfs_xid_mtx);
161 /* Get a pretty random xid to start with */
165 * Skip zero xid if it should ever happen.
170 mtx_unlock(&nfs_xid_mtx);
175 * Create the header for an rpc request packet
176 * The hsiz is the size of the rest of the nfs request header.
177 * (just used to decide if a cluster is a good idea)
180 nfsm_reqhead(struct vnode *vp, u_long procid, int hsiz)
184 MGET(mb, M_WAIT, MT_DATA);
185 if (hsiz >= MINCLSIZE)
192 * copies a uio scatter/gather list to an mbuf chain.
193 * NOTE: can ony handle iovcnt == 1
196 nfsm_uiotombuf(struct uio *uiop, struct mbuf **mq, int siz, caddr_t *bpos)
199 struct mbuf *mp, *mp2;
200 int xfer, left, mlen;
201 int uiosiz, clflg, rem;
204 KASSERT(uiop->uio_iovcnt == 1, ("nfsm_uiotombuf: iovcnt != 1"));
206 if (siz > MLEN) /* or should it >= MCLBYTES ?? */
210 rem = nfsm_rndup(siz)-siz;
213 left = uiop->uio_iov->iov_len;
214 uiocp = uiop->uio_iov->iov_base;
219 mlen = M_TRAILINGSPACE(mp);
221 MGET(mp, M_WAIT, MT_DATA);
227 mlen = M_TRAILINGSPACE(mp);
229 xfer = (left > mlen) ? mlen : left;
232 if (uiop->uio_iov->iov_op != NULL)
233 (*(uiop->uio_iov->iov_op))
234 (uiocp, mtod(mp, caddr_t)+mp->m_len, xfer);
237 if (uiop->uio_segflg == UIO_SYSSPACE)
238 bcopy(uiocp, mtod(mp, caddr_t)+mp->m_len, xfer);
240 copyin(uiocp, mtod(mp, caddr_t)+mp->m_len, xfer);
244 uiop->uio_offset += xfer;
245 uiop->uio_resid -= xfer;
247 uiop->uio_iov->iov_base =
248 (char *)uiop->uio_iov->iov_base + uiosiz;
249 uiop->uio_iov->iov_len -= uiosiz;
253 if (rem > M_TRAILINGSPACE(mp)) {
254 MGET(mp, M_WAIT, MT_DATA);
258 cp = mtod(mp, caddr_t)+mp->m_len;
259 for (left = 0; left < rem; left++)
264 *bpos = mtod(mp, caddr_t)+mp->m_len;
270 * Copy a string into mbufs for the hard cases...
273 nfsm_strtmbuf(struct mbuf **mb, char **bpos, const char *cp, long siz)
275 struct mbuf *m1 = NULL, *m2;
276 long left, xfer, len, tlen;
282 left = M_TRAILINGSPACE(m2);
284 tl = ((u_int32_t *)(*bpos));
285 *tl++ = txdr_unsigned(siz);
287 left -= NFSX_UNSIGNED;
288 m2->m_len += NFSX_UNSIGNED;
290 bcopy(cp, (caddr_t) tl, left);
297 /* Loop around adding mbufs */
299 MGET(m1, M_WAIT, MT_DATA);
302 m1->m_len = NFSMSIZ(m1);
305 tl = mtod(m1, u_int32_t *);
308 *tl++ = txdr_unsigned(siz);
309 m1->m_len -= NFSX_UNSIGNED;
310 tlen = NFSX_UNSIGNED;
313 if (siz < m1->m_len) {
314 len = nfsm_rndup(siz);
319 xfer = len = m1->m_len;
321 bcopy(cp, (caddr_t) tl, xfer);
322 m1->m_len = len+tlen;
327 *bpos = mtod(m1, caddr_t)+m1->m_len;
332 * Called once to initialize data structures...
335 nfs_init(struct vfsconf *vfsp)
339 nfsmount_zone = uma_zcreate("NFSMOUNT", sizeof(struct nfsmount),
340 NULL, NULL, NULL, NULL, UMA_ALIGN_PTR, 0);
341 nfs_true = txdr_unsigned(TRUE);
342 nfs_false = txdr_unsigned(FALSE);
343 nfs_xdrneg1 = txdr_unsigned(-1);
344 nfs_ticks = (hz * NFS_TICKINTVL + 500) / 1000;
347 /* Ensure async daemons disabled */
348 for (i = 0; i < NFS_MAXASYNCDAEMON; i++) {
349 nfs_iodwant[i] = NFSIOD_NOT_AVAILABLE;
350 nfs_iodmount[i] = NULL;
352 nfs_nhinit(); /* Init the nfsnode table */
355 * Initialize reply list and start timer
357 mtx_init(&nfs_iod_mtx, "NFS iod lock", NULL, MTX_DEF);
358 mtx_init(&nfs_xid_mtx, "NFS xid lock", NULL, MTX_DEF);
359 TASK_INIT(&nfs_nfsiodnew_task, 0, nfs_nfsiodnew_tq, NULL);
361 nfs_pbuf_freecnt = nswbuf / 2 + 1;
367 nfs_uninit(struct vfsconf *vfsp)
372 * Tell all nfsiod processes to exit. Clear nfs_iodmax, and wakeup
373 * any sleeping nfsiods so they check nfs_iodmax and exit.
374 * Drain nfsiodnew task before we wait for them to finish.
376 mtx_lock(&nfs_iod_mtx);
378 mtx_unlock(&nfs_iod_mtx);
379 taskqueue_drain(taskqueue_thread, &nfs_nfsiodnew_task);
380 mtx_lock(&nfs_iod_mtx);
381 for (i = 0; i < nfs_numasync; i++)
382 if (nfs_iodwant[i] == NFSIOD_AVAILABLE)
383 wakeup(&nfs_iodwant[i]);
384 /* The last nfsiod to exit will wake us up when nfs_numasync hits 0 */
386 msleep(&nfs_numasync, &nfs_iod_mtx, PWAIT, "ioddie", 0);
387 mtx_unlock(&nfs_iod_mtx);
389 uma_zdestroy(nfsmount_zone);
394 nfs_dircookie_lock(struct nfsnode *np)
396 mtx_lock(&np->n_mtx);
397 while (np->n_flag & NDIRCOOKIELK)
398 (void) msleep(&np->n_flag, &np->n_mtx, PZERO, "nfsdirlk", 0);
399 np->n_flag |= NDIRCOOKIELK;
400 mtx_unlock(&np->n_mtx);
404 nfs_dircookie_unlock(struct nfsnode *np)
406 mtx_lock(&np->n_mtx);
407 np->n_flag &= ~NDIRCOOKIELK;
409 mtx_unlock(&np->n_mtx);
413 nfs_upgrade_vnlock(struct vnode *vp)
417 ASSERT_VOP_LOCKED(vp, "nfs_upgrade_vnlock");
418 old_lock = VOP_ISLOCKED(vp);
419 if (old_lock != LK_EXCLUSIVE) {
420 KASSERT(old_lock == LK_SHARED,
421 ("nfs_upgrade_vnlock: wrong old_lock %d", old_lock));
422 /* Upgrade to exclusive lock, this might block */
423 vn_lock(vp, LK_UPGRADE | LK_RETRY);
429 nfs_downgrade_vnlock(struct vnode *vp, int old_lock)
431 if (old_lock != LK_EXCLUSIVE) {
432 KASSERT(old_lock == LK_SHARED, ("wrong old_lock %d", old_lock));
433 /* Downgrade from exclusive lock. */
434 vn_lock(vp, LK_DOWNGRADE | LK_RETRY);
439 nfs_printf(const char *fmt, ...)
451 * Attribute cache routines.
452 * nfs_loadattrcache() - loads or updates the cache contents from attributes
453 * that are on the mbuf list
454 * nfs_getattrcache() - returns valid attributes if found in cache, returns
459 * Load the attribute cache (that lives in the nfsnode entry) with
460 * the values on the mbuf list and
462 * copy the attributes to *vaper
465 nfs_loadattrcache(struct vnode **vpp, struct mbuf **mdp, caddr_t *dposp,
466 struct vattr *vaper, int dontshrink)
468 struct vnode *vp = *vpp;
470 struct nfs_fattr *fp;
471 struct nfsnode *np = NULL;
478 struct timespec mtime, mtime_save;
479 int v3 = NFS_ISV3(vp);
483 t1 = (mtod(md, caddr_t) + md->m_len) - *dposp;
484 cp2 = nfsm_disct(mdp, dposp, NFSX_FATTR(v3), t1, M_WAIT);
489 fp = (struct nfs_fattr *)cp2;
491 vtyp = nfsv3tov_type(fp->fa_type);
492 vmode = fxdr_unsigned(u_short, fp->fa_mode);
493 rdev = makedev(fxdr_unsigned(int, fp->fa3_rdev.specdata1),
494 fxdr_unsigned(int, fp->fa3_rdev.specdata2));
495 fxdr_nfsv3time(&fp->fa3_mtime, &mtime);
497 vtyp = nfsv2tov_type(fp->fa_type);
498 vmode = fxdr_unsigned(u_short, fp->fa_mode);
502 * The duplicate information returned in fa_type and fa_mode
503 * is an ambiguity in the NFS version 2 protocol.
505 * VREG should be taken literally as a regular file. If a
506 * server intents to return some type information differently
507 * in the upper bits of the mode field (e.g. for sockets, or
508 * FIFOs), NFSv2 mandates fa_type to be VNON. Anyway, we
509 * leave the examination of the mode bits even in the VREG
510 * case to avoid breakage for bogus servers, but we make sure
511 * that there are actually type bits set in the upper part of
512 * fa_mode (and failing that, trust the va_type field).
514 * NFSv3 cleared the issue, and requires fa_mode to not
515 * contain any type information (while also introduing sockets
516 * and FIFOs for fa_type).
518 if (vtyp == VNON || (vtyp == VREG && (vmode & S_IFMT) != 0))
519 vtyp = IFTOVT(vmode);
520 rdev = fxdr_unsigned(int32_t, fp->fa2_rdev);
521 fxdr_nfsv2time(&fp->fa2_mtime, &mtime);
524 * Really ugly NFSv2 kludge.
526 if (vtyp == VCHR && rdev == 0xffffffff)
531 * If v_type == VNON it is a new node, so fill in the v_type,
532 * n_mtime fields. Check to see if it represents a special
533 * device, and if so, check for a possible alias. Once the
534 * correct vnode has been obtained, fill in the rest of the
538 mtx_lock(&np->n_mtx);
539 if (vp->v_type != vtyp) {
541 if (vp->v_type == VFIFO)
542 vp->v_op = &nfs_fifoops;
547 vap->va_mode = (vmode & 07777);
549 mtime_save = vap->va_mtime;
550 vap->va_mtime = mtime;
551 vap->va_fsid = vp->v_mount->mnt_stat.f_fsid.val[0];
553 vap->va_nlink = fxdr_unsigned(u_short, fp->fa_nlink);
554 vap->va_uid = fxdr_unsigned(uid_t, fp->fa_uid);
555 vap->va_gid = fxdr_unsigned(gid_t, fp->fa_gid);
556 vap->va_size = fxdr_hyper(&fp->fa3_size);
557 vap->va_blocksize = NFS_FABLKSIZE;
558 vap->va_bytes = fxdr_hyper(&fp->fa3_used);
559 vap->va_fileid = fxdr_unsigned(int32_t,
560 fp->fa3_fileid.nfsuquad[1]);
561 fxdr_nfsv3time(&fp->fa3_atime, &vap->va_atime);
562 fxdr_nfsv3time(&fp->fa3_ctime, &vap->va_ctime);
566 vap->va_nlink = fxdr_unsigned(u_short, fp->fa_nlink);
567 vap->va_uid = fxdr_unsigned(uid_t, fp->fa_uid);
568 vap->va_gid = fxdr_unsigned(gid_t, fp->fa_gid);
569 vap->va_size = fxdr_unsigned(u_int32_t, fp->fa2_size);
570 vap->va_blocksize = fxdr_unsigned(int32_t, fp->fa2_blocksize);
571 vap->va_bytes = (u_quad_t)fxdr_unsigned(int32_t, fp->fa2_blocks)
573 vap->va_fileid = fxdr_unsigned(int32_t, fp->fa2_fileid);
574 fxdr_nfsv2time(&fp->fa2_atime, &vap->va_atime);
576 vap->va_ctime.tv_sec = fxdr_unsigned(u_int32_t,
577 fp->fa2_ctime.nfsv2_sec);
578 vap->va_ctime.tv_nsec = 0;
579 vap->va_gen = fxdr_unsigned(u_int32_t, fp->fa2_ctime.nfsv2_usec);
582 np->n_attrstamp = time_second;
583 if (vap->va_size != np->n_size) {
584 if (vap->va_type == VREG) {
585 if (dontshrink && vap->va_size < np->n_size) {
587 * We've been told not to shrink the file;
588 * zero np->n_attrstamp to indicate that
589 * the attributes are stale.
591 vap->va_size = np->n_size;
593 KDTRACE_NFS_ATTRCACHE_FLUSH_DONE(vp);
594 } else if (np->n_flag & NMODIFIED) {
596 * We've modified the file: Use the larger
597 * of our size, and the server's size.
599 if (vap->va_size < np->n_size) {
600 vap->va_size = np->n_size;
602 np->n_size = vap->va_size;
603 np->n_flag |= NSIZECHANGED;
606 np->n_size = vap->va_size;
607 np->n_flag |= NSIZECHANGED;
609 vnode_pager_setsize(vp, np->n_size);
611 np->n_size = vap->va_size;
615 * The following checks are added to prevent a race between (say)
616 * a READDIR+ and a WRITE.
617 * READDIR+, WRITE requests sent out.
618 * READDIR+ resp, WRITE resp received on client.
619 * However, the WRITE resp was handled before the READDIR+ resp
620 * causing the post op attrs from the write to be loaded first
621 * and the attrs from the READDIR+ to be loaded later. If this
622 * happens, we have stale attrs loaded into the attrcache.
623 * We detect this by for the mtime moving back. We invalidate the
624 * attrcache when this happens.
626 if (timespeccmp(&mtime_save, &vap->va_mtime, >)) {
627 /* Size changed or mtime went backwards */
629 KDTRACE_NFS_ATTRCACHE_FLUSH_DONE(vp);
632 bcopy((caddr_t)vap, (caddr_t)vaper, sizeof(*vap));
633 if (np->n_flag & NCHG) {
634 if (np->n_flag & NACC)
635 vaper->va_atime = np->n_atim;
636 if (np->n_flag & NUPD)
637 vaper->va_mtime = np->n_mtim;
642 if (np->n_attrstamp != 0)
643 KDTRACE_NFS_ATTRCACHE_LOAD_DONE(vp, &np->n_vattr, 0);
645 mtx_unlock(&np->n_mtx);
649 KDTRACE_NFS_ATTRCACHE_LOAD_DONE(vp, NULL, error);
655 #include <sys/sysctl.h>
656 SYSCTL_DECL(_vfs_oldnfs);
657 static int nfs_acdebug;
658 SYSCTL_INT(_vfs_oldnfs, OID_AUTO, acdebug, CTLFLAG_RW, &nfs_acdebug, 0,
659 "Toggle acdebug (attribute cache debug) flag");
663 * Check the time stamp
664 * If the cache is valid, copy contents to *vap and return 0
665 * otherwise return an error
668 nfs_getattrcache(struct vnode *vp, struct vattr *vaper)
672 struct nfsmount *nmp;
677 nmp = VFSTONFS(vp->v_mount);
679 mtx_lock(&Giant); /* nfs_printf() */
681 mtx_lock(&np->n_mtx);
682 /* XXX n_mtime doesn't seem to be updated on a miss-and-reload */
683 timeo = (time_second - np->n_mtime.tv_sec) / 10;
687 nfs_printf("nfs_getattrcache: initial timeo = %d\n", timeo);
690 if (vap->va_type == VDIR) {
691 if ((np->n_flag & NMODIFIED) || timeo < nmp->nm_acdirmin)
692 timeo = nmp->nm_acdirmin;
693 else if (timeo > nmp->nm_acdirmax)
694 timeo = nmp->nm_acdirmax;
696 if ((np->n_flag & NMODIFIED) || timeo < nmp->nm_acregmin)
697 timeo = nmp->nm_acregmin;
698 else if (timeo > nmp->nm_acregmax)
699 timeo = nmp->nm_acregmax;
704 nfs_printf("acregmin %d; acregmax %d; acdirmin %d; acdirmax %d\n",
705 nmp->nm_acregmin, nmp->nm_acregmax,
706 nmp->nm_acdirmin, nmp->nm_acdirmax);
709 nfs_printf("nfs_getattrcache: age = %d; final timeo = %d\n",
710 (time_second - np->n_attrstamp), timeo);
713 if ((time_second - np->n_attrstamp) >= timeo) {
714 nfsstats.attrcache_misses++;
715 mtx_unlock(&np->n_mtx);
717 mtx_unlock(&Giant); /* nfs_printf() */
719 KDTRACE_NFS_ATTRCACHE_GET_MISS(vp);
722 nfsstats.attrcache_hits++;
723 if (vap->va_size != np->n_size) {
724 if (vap->va_type == VREG) {
725 if (np->n_flag & NMODIFIED) {
726 if (vap->va_size < np->n_size)
727 vap->va_size = np->n_size;
729 np->n_size = vap->va_size;
731 np->n_size = vap->va_size;
733 vnode_pager_setsize(vp, np->n_size);
735 np->n_size = vap->va_size;
738 bcopy((caddr_t)vap, (caddr_t)vaper, sizeof(struct vattr));
739 if (np->n_flag & NCHG) {
740 if (np->n_flag & NACC)
741 vaper->va_atime = np->n_atim;
742 if (np->n_flag & NUPD)
743 vaper->va_mtime = np->n_mtim;
745 mtx_unlock(&np->n_mtx);
747 mtx_unlock(&Giant); /* nfs_printf() */
749 KDTRACE_NFS_ATTRCACHE_GET_HIT(vp, vap);
754 * Purge all cached information about an NFS vnode including name
755 * cache entries, the attribute cache, and the access cache. This is
756 * called when an NFS request for a node fails with a stale
760 nfs_purgecache(struct vnode *vp)
767 mtx_lock(&np->n_mtx);
769 KDTRACE_NFS_ATTRCACHE_FLUSH_DONE(vp);
770 for (i = 0; i < NFS_ACCESSCACHESIZE; i++)
771 np->n_accesscache[i].stamp = 0;
772 KDTRACE_NFS_ACCESSCACHE_FLUSH_DONE(vp);
773 mtx_unlock(&np->n_mtx);
776 static nfsuint64 nfs_nullcookie = { { 0, 0 } };
778 * This function finds the directory cookie that corresponds to the
779 * logical byte offset given.
782 nfs_getcookie(struct nfsnode *np, off_t off, int add)
784 struct nfsdmap *dp, *dp2;
786 nfsuint64 *retval = NULL;
788 pos = (uoff_t)off / NFS_DIRBLKSIZ;
789 if (pos == 0 || off < 0) {
790 KASSERT(!add, ("nfs getcookie add at <= 0"));
791 return (&nfs_nullcookie);
794 dp = LIST_FIRST(&np->n_cookies);
797 dp = malloc(sizeof (struct nfsdmap),
798 M_NFSDIROFF, M_WAITOK);
799 dp->ndm_eocookie = 0;
800 LIST_INSERT_HEAD(&np->n_cookies, dp, ndm_list);
804 while (pos >= NFSNUMCOOKIES) {
805 pos -= NFSNUMCOOKIES;
806 if (LIST_NEXT(dp, ndm_list)) {
807 if (!add && dp->ndm_eocookie < NFSNUMCOOKIES &&
808 pos >= dp->ndm_eocookie)
810 dp = LIST_NEXT(dp, ndm_list);
812 dp2 = malloc(sizeof (struct nfsdmap),
813 M_NFSDIROFF, M_WAITOK);
814 dp2->ndm_eocookie = 0;
815 LIST_INSERT_AFTER(dp, dp2, ndm_list);
820 if (pos >= dp->ndm_eocookie) {
822 dp->ndm_eocookie = pos + 1;
826 retval = &dp->ndm_cookies[pos];
832 * Invalidate cached directory information, except for the actual directory
833 * blocks (which are invalidated separately).
834 * Done mainly to avoid the use of stale offset cookies.
837 nfs_invaldir(struct vnode *vp)
839 struct nfsnode *np = VTONFS(vp);
841 KASSERT(vp->v_type == VDIR, ("nfs: invaldir not dir"));
842 nfs_dircookie_lock(np);
843 np->n_direofoffset = 0;
844 np->n_cookieverf.nfsuquad[0] = 0;
845 np->n_cookieverf.nfsuquad[1] = 0;
846 if (LIST_FIRST(&np->n_cookies))
847 LIST_FIRST(&np->n_cookies)->ndm_eocookie = 0;
848 nfs_dircookie_unlock(np);
852 * The write verifier has changed (probably due to a server reboot), so all
853 * B_NEEDCOMMIT blocks will have to be written again. Since they are on the
854 * dirty block list as B_DELWRI, all this takes is clearing the B_NEEDCOMMIT
855 * and B_CLUSTEROK flags. Once done the new write verifier can be set for the
858 * B_CLUSTEROK must be cleared along with B_NEEDCOMMIT because stage 1 data
859 * writes are not clusterable.
862 nfs_clearcommit(struct mount *mp)
864 struct vnode *vp, *nvp;
865 struct buf *bp, *nbp;
869 MNT_VNODE_FOREACH(vp, mp, nvp) {
872 if (vp->v_iflag & VI_DOOMED) {
880 TAILQ_FOREACH_SAFE(bp, &bo->bo_dirty.bv_hd, b_bobufs, nbp) {
881 if (!BUF_ISLOCKED(bp) &&
882 (bp->b_flags & (B_DELWRI | B_NEEDCOMMIT))
883 == (B_DELWRI | B_NEEDCOMMIT))
884 bp->b_flags &= ~(B_NEEDCOMMIT | B_CLUSTEROK);
894 * Helper functions for former macros. Some of these should be
895 * moved to their callers.
899 nfsm_mtofh_xx(struct vnode *d, struct vnode **v, int v3, int *f,
900 struct mbuf **md, caddr_t *dpos)
902 struct nfsnode *ttnp;
910 tl = nfsm_dissect_xx(NFSX_UNSIGNED, md, dpos);
913 *f = fxdr_unsigned(int, *tl);
917 t1 = nfsm_getfh_xx(&ttfhp, &ttfhsize, (v3), md, dpos);
920 t1 = nfs_nget(d->v_mount, ttfhp, ttfhsize, &ttnp, LK_EXCLUSIVE);
926 tl = nfsm_dissect_xx(NFSX_UNSIGNED, md, dpos);
930 *f = fxdr_unsigned(int, *tl);
931 else if (fxdr_unsigned(int, *tl))
932 nfsm_adv_xx(NFSX_V3FATTR, md, dpos);
936 t1 = nfs_loadattrcache(&ttvp, md, dpos, NULL, 0);
945 nfsm_getfh_xx(nfsfh_t **f, int *s, int v3, struct mbuf **md, caddr_t *dpos)
950 tl = nfsm_dissect_xx(NFSX_UNSIGNED, md, dpos);
953 *s = fxdr_unsigned(int, *tl);
954 if (*s <= 0 || *s > NFSX_V3FHMAX)
958 *f = nfsm_dissect_xx(nfsm_rndup(*s), md, dpos);
967 nfsm_loadattr_xx(struct vnode **v, struct vattr *va, struct mbuf **md,
972 struct vnode *ttvp = *v;
973 t1 = nfs_loadattrcache(&ttvp, md, dpos, va, 0);
981 nfsm_postop_attr_xx(struct vnode **v, int *f, struct mbuf **md,
987 struct vnode *ttvp = *v;
988 tl = nfsm_dissect_xx(NFSX_UNSIGNED, md, dpos);
991 *f = fxdr_unsigned(int, *tl);
993 t1 = nfs_loadattrcache(&ttvp, md, dpos, NULL, 1);
1004 nfsm_wcc_data_xx(struct vnode **v, int *f, struct mbuf **md, caddr_t *dpos)
1007 int ttattrf, ttretf = 0;
1010 tl = nfsm_dissect_xx(NFSX_UNSIGNED, md, dpos);
1013 if (*tl == nfs_true) {
1014 tl = nfsm_dissect_xx(6 * NFSX_UNSIGNED, md, dpos);
1017 mtx_lock(&(VTONFS(*v))->n_mtx);
1019 ttretf = (VTONFS(*v)->n_mtime.tv_sec == fxdr_unsigned(u_int32_t, *(tl + 2)) &&
1020 VTONFS(*v)->n_mtime.tv_nsec == fxdr_unsigned(u_int32_t, *(tl + 3)));
1021 mtx_unlock(&(VTONFS(*v))->n_mtx);
1023 t1 = nfsm_postop_attr_xx(v, &ttattrf, md, dpos);
1034 nfsm_strtom_xx(const char *a, int s, int m, struct mbuf **mb, caddr_t *bpos)
1040 return ENAMETOOLONG;
1041 t1 = nfsm_rndup(s) + NFSX_UNSIGNED;
1042 if (t1 <= M_TRAILINGSPACE(*mb)) {
1043 tl = nfsm_build_xx(t1, mb, bpos);
1044 *tl++ = txdr_unsigned(s);
1045 *(tl + ((t1 >> 2) - 2)) = 0;
1048 t1 = nfsm_strtmbuf(mb, bpos, a, s);
1056 nfsm_fhtom_xx(struct vnode *v, int v3, struct mbuf **mb, caddr_t *bpos)
1063 t1 = nfsm_rndup(VTONFS(v)->n_fhsize) + NFSX_UNSIGNED;
1064 if (t1 < M_TRAILINGSPACE(*mb)) {
1065 tl = nfsm_build_xx(t1, mb, bpos);
1066 *tl++ = txdr_unsigned(VTONFS(v)->n_fhsize);
1067 *(tl + ((t1 >> 2) - 2)) = 0;
1068 bcopy(VTONFS(v)->n_fhp, tl, VTONFS(v)->n_fhsize);
1070 t1 = nfsm_strtmbuf(mb, bpos,
1071 (const char *)VTONFS(v)->n_fhp,
1072 VTONFS(v)->n_fhsize);
1077 cp = nfsm_build_xx(NFSX_V2FH, mb, bpos);
1078 bcopy(VTONFS(v)->n_fhp, cp, NFSX_V2FH);
1084 nfsm_v3attrbuild_xx(struct vattr *va, int full, struct mbuf **mb,
1089 if (va->va_mode != (mode_t)VNOVAL) {
1090 tl = nfsm_build_xx(2 * NFSX_UNSIGNED, mb, bpos);
1092 *tl = txdr_unsigned(va->va_mode);
1094 tl = nfsm_build_xx(NFSX_UNSIGNED, mb, bpos);
1097 if (full && va->va_uid != (uid_t)VNOVAL) {
1098 tl = nfsm_build_xx(2 * NFSX_UNSIGNED, mb, bpos);
1100 *tl = txdr_unsigned(va->va_uid);
1102 tl = nfsm_build_xx(NFSX_UNSIGNED, mb, bpos);
1105 if (full && va->va_gid != (gid_t)VNOVAL) {
1106 tl = nfsm_build_xx(2 * NFSX_UNSIGNED, mb, bpos);
1108 *tl = txdr_unsigned(va->va_gid);
1110 tl = nfsm_build_xx(NFSX_UNSIGNED, mb, bpos);
1113 if (full && va->va_size != VNOVAL) {
1114 tl = nfsm_build_xx(3 * NFSX_UNSIGNED, mb, bpos);
1116 txdr_hyper(va->va_size, tl);
1118 tl = nfsm_build_xx(NFSX_UNSIGNED, mb, bpos);
1121 if (va->va_atime.tv_sec != VNOVAL) {
1122 if (va->va_atime.tv_sec != time_second) {
1123 tl = nfsm_build_xx(3 * NFSX_UNSIGNED, mb, bpos);
1124 *tl++ = txdr_unsigned(NFSV3SATTRTIME_TOCLIENT);
1125 txdr_nfsv3time(&va->va_atime, tl);
1127 tl = nfsm_build_xx(NFSX_UNSIGNED, mb, bpos);
1128 *tl = txdr_unsigned(NFSV3SATTRTIME_TOSERVER);
1131 tl = nfsm_build_xx(NFSX_UNSIGNED, mb, bpos);
1132 *tl = txdr_unsigned(NFSV3SATTRTIME_DONTCHANGE);
1134 if (va->va_mtime.tv_sec != VNOVAL) {
1135 if (va->va_mtime.tv_sec != time_second) {
1136 tl = nfsm_build_xx(3 * NFSX_UNSIGNED, mb, bpos);
1137 *tl++ = txdr_unsigned(NFSV3SATTRTIME_TOCLIENT);
1138 txdr_nfsv3time(&va->va_mtime, tl);
1140 tl = nfsm_build_xx(NFSX_UNSIGNED, mb, bpos);
1141 *tl = txdr_unsigned(NFSV3SATTRTIME_TOSERVER);
1144 tl = nfsm_build_xx(NFSX_UNSIGNED, mb, bpos);
1145 *tl = txdr_unsigned(NFSV3SATTRTIME_DONTCHANGE);