2 * Copyright (c) 1989, 1993
3 * The Regents of the University of California. All rights reserved.
5 * This code is derived from software contributed
6 * to Berkeley by John Heidemann of the UCLA Ficus project.
8 * Source: * @(#)i405_init.c 2.10 92/04/27 UCLA Ficus project
10 * Redistribution and use in source and binary forms, with or without
11 * modification, are permitted provided that the following conditions
13 * 1. Redistributions of source code must retain the above copyright
14 * notice, this list of conditions and the following disclaimer.
15 * 2. Redistributions in binary form must reproduce the above copyright
16 * notice, this list of conditions and the following disclaimer in the
17 * documentation and/or other materials provided with the distribution.
18 * 4. Neither the name of the University nor the names of its contributors
19 * may be used to endorse or promote products derived from this software
20 * without specific prior written permission.
22 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
23 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
24 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
25 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
26 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
27 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
28 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
29 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
30 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
31 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
35 #include <sys/cdefs.h>
36 __FBSDID("$FreeBSD$");
38 #include <sys/param.h>
39 #include <sys/systm.h>
43 #include <sys/event.h>
44 #include <sys/kernel.h>
45 #include <sys/limits.h>
47 #include <sys/lockf.h>
48 #include <sys/malloc.h>
49 #include <sys/mount.h>
50 #include <sys/mutex.h>
51 #include <sys/namei.h>
52 #include <sys/fcntl.h>
53 #include <sys/unistd.h>
54 #include <sys/vnode.h>
55 #include <sys/dirent.h>
59 #include <vm/vm_object.h>
60 #include <vm/vm_extern.h>
62 #include <vm/vm_map.h>
63 #include <vm/vm_page.h>
64 #include <vm/vm_pager.h>
65 #include <vm/vnode_pager.h>
67 static int vop_nolookup(struct vop_lookup_args *);
68 static int vop_nostrategy(struct vop_strategy_args *);
69 static int get_next_dirent(struct vnode *vp, struct dirent **dpp,
70 char *dirbuf, int dirbuflen, off_t *off,
71 char **cpos, int *len, int *eofflag,
73 static int dirent_exists(struct vnode *vp, const char *dirname,
76 #define DIRENT_MINSIZE (sizeof(struct dirent) - (MAXNAMLEN+1) + 4)
79 * This vnode table stores what we want to do if the filesystem doesn't
80 * implement a particular VOP.
82 * If there is no specific entry here, we will return EOPNOTSUPP.
86 struct vop_vector default_vnodeops = {
88 .vop_bypass = VOP_EOPNOTSUPP,
90 .vop_advlock = vop_stdadvlock,
91 .vop_advlockasync = vop_stdadvlockasync,
92 .vop_bmap = vop_stdbmap,
93 .vop_close = VOP_NULL,
94 .vop_fsync = VOP_NULL,
95 .vop_getpages = vop_stdgetpages,
96 .vop_getwritemount = vop_stdgetwritemount,
97 .vop_inactive = VOP_NULL,
98 .vop_ioctl = VOP_ENOTTY,
99 .vop_kqfilter = vop_stdkqfilter,
100 .vop_islocked = vop_stdislocked,
101 .vop_lease = VOP_NULL,
102 .vop_lock1 = vop_stdlock,
103 .vop_lookup = vop_nolookup,
104 .vop_open = VOP_NULL,
105 .vop_pathconf = VOP_EINVAL,
106 .vop_poll = vop_nopoll,
107 .vop_putpages = vop_stdputpages,
108 .vop_readlink = VOP_EINVAL,
109 .vop_revoke = VOP_PANIC,
110 .vop_strategy = vop_nostrategy,
111 .vop_unlock = vop_stdunlock,
112 .vop_vptocnp = vop_stdvptocnp,
113 .vop_vptofh = vop_stdvptofh,
117 * Series of placeholder functions for various error returns for
122 vop_eopnotsupp(struct vop_generic_args *ap)
125 printf("vop_notsupp[%s]\n", ap->a_desc->vdesc_name);
132 vop_ebadf(struct vop_generic_args *ap)
139 vop_enotty(struct vop_generic_args *ap)
146 vop_einval(struct vop_generic_args *ap)
153 vop_enoent(struct vop_generic_args *ap)
160 vop_null(struct vop_generic_args *ap)
167 * Helper function to panic on some bad VOPs in some filesystems.
170 vop_panic(struct vop_generic_args *ap)
173 panic("filesystem goof: vop_panic[%s]", ap->a_desc->vdesc_name);
177 * vop_std<something> and vop_no<something> are default functions for use by
178 * filesystems that need the "default reasonable" implementation for a
179 * particular operation.
181 * The documentation for the operations they implement exists (if it exists)
182 * in the VOP_<SOMETHING>(9) manpage (all uppercase).
186 * Default vop for filesystems that do not support name lookup
190 struct vop_lookup_args /* {
192 struct vnode **a_vpp;
193 struct componentname *a_cnp;
204 * Strategy routine for VFS devices that have none.
206 * BIO_ERROR and B_INVAL must be cleared prior to calling any strategy
207 * routine. Typically this is done for a BIO_READ strategy call.
208 * Typically B_INVAL is assumed to already be clear prior to a write
209 * and should not be cleared manually unless you just made the buffer
210 * invalid. BIO_ERROR should be cleared either way.
214 vop_nostrategy (struct vop_strategy_args *ap)
216 printf("No strategy for buffer at %p\n", ap->a_bp);
217 vprint("vnode", ap->a_vp);
218 ap->a_bp->b_ioflags |= BIO_ERROR;
219 ap->a_bp->b_error = EOPNOTSUPP;
225 get_next_dirent(struct vnode *vp, struct dirent **dpp, char *dirbuf,
226 int dirbuflen, off_t *off, char **cpos, int *len,
227 int *eofflag, struct thread *td)
234 KASSERT(VOP_ISLOCKED(vp), ("vp %p is not locked", vp));
235 KASSERT(vp->v_type == VDIR, ("vp %p is not a directory", vp));
238 iov.iov_base = dirbuf;
239 iov.iov_len = dirbuflen;
243 uio.uio_offset = *off;
244 uio.uio_resid = dirbuflen;
245 uio.uio_segflg = UIO_SYSSPACE;
246 uio.uio_rw = UIO_READ;
252 error = mac_vnode_check_readdir(td->td_ucred, vp);
255 error = VOP_READDIR(vp, &uio, td->td_ucred, eofflag,
260 *off = uio.uio_offset;
263 *len = (dirbuflen - uio.uio_resid);
266 dp = (struct dirent *)(*cpos);
267 reclen = dp->d_reclen;
270 /* check for malformed directory.. */
271 if (reclen < DIRENT_MINSIZE)
281 * Check if a named file exists in a given directory vnode.
284 dirent_exists(struct vnode *vp, const char *dirname, struct thread *td)
287 int error, eofflag, dirbuflen, len, found;
292 KASSERT(VOP_ISLOCKED(vp), ("vp %p is not locked", vp));
293 KASSERT(vp->v_type == VDIR, ("vp %p is not a directory", vp));
297 error = VOP_GETATTR(vp, &va, td->td_ucred);
301 dirbuflen = DEV_BSIZE;
302 if (dirbuflen < va.va_blocksize)
303 dirbuflen = va.va_blocksize;
304 dirbuf = (char *)malloc(dirbuflen, M_TEMP, M_WAITOK);
309 error = get_next_dirent(vp, &dp, dirbuf, dirbuflen, &off,
310 &cpos, &len, &eofflag, td);
314 if ((dp->d_type != DT_WHT) &&
315 !strcmp(dp->d_name, dirname)) {
319 } while (len > 0 || !eofflag);
322 free(dirbuf, M_TEMP);
327 * Advisory record locking support
330 vop_stdadvlock(struct vop_advlock_args *ap)
338 cred = curthread->td_ucred;
339 vn_lock(vp, LK_SHARED | LK_RETRY);
340 error = VOP_GETATTR(vp, &vattr, cred);
345 return (lf_advlock(ap, &(vp->v_lockf), vattr.va_size));
349 vop_stdadvlockasync(struct vop_advlockasync_args *ap)
357 cred = curthread->td_ucred;
358 vn_lock(vp, LK_SHARED | LK_RETRY);
359 error = VOP_GETATTR(vp, &vattr, cred);
364 return (lf_advlockasync(ap, &(vp->v_lockf), vattr.va_size));
370 * Standard implementation of POSIX pathconf, to get information about limits
372 * Override per filesystem for the case where the filesystem has smaller
377 struct vop_pathconf_args /* {
384 switch (ap->a_name) {
386 *ap->a_retval = NAME_MAX;
389 *ap->a_retval = PATH_MAX;
392 *ap->a_retval = LINK_MAX;
395 *ap->a_retval = MAX_CANON;
398 *ap->a_retval = MAX_INPUT;
401 *ap->a_retval = PIPE_BUF;
403 case _PC_CHOWN_RESTRICTED:
407 *ap->a_retval = _POSIX_VDISABLE;
416 * Standard lock, unlock and islocked functions.
420 struct vop_lock1_args /* {
427 struct vnode *vp = ap->a_vp;
429 return (_lockmgr_args(vp->v_vnlock, ap->a_flags, VI_MTX(vp),
430 LK_WMESG_DEFAULT, LK_PRIO_DEFAULT, LK_TIMO_DEFAULT, ap->a_file,
437 struct vop_unlock_args /* {
442 struct vnode *vp = ap->a_vp;
444 return (lockmgr(vp->v_vnlock, ap->a_flags | LK_RELEASE, VI_MTX(vp)));
450 struct vop_islocked_args /* {
455 return (lockstatus(ap->a_vp->v_vnlock));
459 * Return true for select/poll.
463 struct vop_poll_args /* {
466 struct ucred *a_cred;
471 return (poll_no_poll(ap->a_events));
475 * Implement poll for local filesystems that support it.
479 struct vop_poll_args /* {
482 struct ucred *a_cred;
486 if (ap->a_events & ~POLLSTANDARD)
487 return (vn_pollrecord(ap->a_vp, ap->a_td, ap->a_events));
488 return (ap->a_events & (POLLIN | POLLOUT | POLLRDNORM | POLLWRNORM));
492 * Return our mount point, as we will take charge of the writes.
495 vop_stdgetwritemount(ap)
496 struct vop_getwritemount_args /* {
498 struct mount **a_mpp;
504 * XXX Since this is called unlocked we may be recycled while
505 * attempting to ref the mount. If this is the case or mountpoint
506 * will be set to NULL. We only have to prevent this call from
507 * returning with a ref to an incorrect mountpoint. It is not
508 * harmful to return with a ref to our previous mountpoint.
510 mp = ap->a_vp->v_mount;
513 if (mp != ap->a_vp->v_mount) {
522 /* XXX Needs good comment and VOP_BMAP(9) manpage */
525 struct vop_bmap_args /* {
528 struct bufobj **a_bop;
535 if (ap->a_bop != NULL)
536 *ap->a_bop = &ap->a_vp->v_bufobj;
537 if (ap->a_bnp != NULL)
538 *ap->a_bnp = ap->a_bn * btodb(ap->a_vp->v_mount->mnt_stat.f_iosize);
539 if (ap->a_runp != NULL)
541 if (ap->a_runb != NULL)
548 struct vop_fsync_args /* {
550 struct ucred *a_cred;
555 struct vnode *vp = ap->a_vp;
560 int maxretry = 1000; /* large, arbitrarily chosen */
566 * MARK/SCAN initialization to avoid infinite loops.
568 TAILQ_FOREACH(bp, &bo->bo_dirty.bv_hd, b_bobufs) {
569 bp->b_vflags &= ~BV_SCANNED;
574 * Flush all dirty buffers associated with a vnode.
577 TAILQ_FOREACH_SAFE(bp, &bo->bo_dirty.bv_hd, b_bobufs, nbp) {
578 if ((bp->b_vflags & BV_SCANNED) != 0)
580 bp->b_vflags |= BV_SCANNED;
581 if (BUF_LOCK(bp, LK_EXCLUSIVE | LK_NOWAIT, NULL))
584 KASSERT(bp->b_bufobj == bo,
585 ("bp %p wrong b_bufobj %p should be %p",
586 bp, bp->b_bufobj, bo));
587 if ((bp->b_flags & B_DELWRI) == 0)
588 panic("fsync: not dirty");
589 if ((vp->v_object != NULL) && (bp->b_flags & B_CLUSTEROK)) {
600 * If synchronous the caller expects us to completely resolve all
601 * dirty buffers in the system. Wait for in-progress I/O to
602 * complete (which could include background bitmap writes), then
603 * retry if dirty blocks still exist.
605 if (ap->a_waitfor == MNT_WAIT) {
606 bufobj_wwait(bo, 0, 0);
607 if (bo->bo_dirty.bv_cnt > 0) {
609 * If we are unable to write any of these buffers
610 * then we fail now rather than trying endlessly
613 TAILQ_FOREACH(bp, &bo->bo_dirty.bv_hd, b_bobufs)
614 if ((error = bp->b_error) == 0)
616 if (error == 0 && --maxretry >= 0)
623 vprint("fsync: giving up on dirty", vp);
628 /* XXX Needs good comment and more info in the manpage (VOP_GETPAGES(9)). */
631 struct vop_getpages_args /* {
636 vm_ooffset_t a_offset;
640 return vnode_pager_generic_getpages(ap->a_vp, ap->a_m,
641 ap->a_count, ap->a_reqpage);
645 vop_stdkqfilter(struct vop_kqfilter_args *ap)
647 return vfs_kqfilter(ap);
650 /* XXX Needs good comment and more info in the manpage (VOP_PUTPAGES(9)). */
653 struct vop_putpages_args /* {
659 vm_ooffset_t a_offset;
663 return vnode_pager_generic_putpages(ap->a_vp, ap->a_m, ap->a_count,
664 ap->a_sync, ap->a_rtvals);
668 vop_stdvptofh(struct vop_vptofh_args *ap)
674 vop_stdvptocnp(struct vop_vptocnp_args *ap)
676 struct vnode *vp = ap->a_vp;
677 struct vnode **dvp = ap->a_vpp;
678 char *buf = ap->a_buf;
679 int *buflen = ap->a_buflen;
681 int i, error, eofflag, dirbuflen, flags, locked, len, covered;
695 if (vp->v_type != VDIR)
698 error = VOP_GETATTR(vp, &va, td->td_ucred);
703 locked = VOP_ISLOCKED(vp);
705 NDINIT_ATVP(&nd, LOOKUP, FOLLOW | LOCKLEAF, UIO_SYSSPACE,
708 error = vn_open(&nd, &flags, 0, NULL);
710 vn_lock(vp, locked | LK_RETRY);
713 NDFREE(&nd, NDF_ONLY_PNBUF);
715 mvp = *dvp = nd.ni_vp;
717 if (vp->v_mount != (*dvp)->v_mount &&
718 ((*dvp)->v_vflag & VV_ROOT) &&
719 ((*dvp)->v_mount->mnt_flag & MNT_UNION)) {
720 *dvp = (*dvp)->v_mount->mnt_vnodecovered;
723 vn_close(mvp, FREAD, td->td_ucred, td);
725 vn_lock(*dvp, LK_EXCLUSIVE | LK_RETRY);
729 fileno = va.va_fileid;
731 dirbuflen = DEV_BSIZE;
732 if (dirbuflen < va.va_blocksize)
733 dirbuflen = va.va_blocksize;
734 dirbuf = (char *)malloc(dirbuflen, M_TEMP, M_WAITOK);
736 if ((*dvp)->v_type != VDIR) {
744 /* call VOP_READDIR of parent */
745 error = get_next_dirent(*dvp, &dp, dirbuf, dirbuflen, &off,
746 &cpos, &len, &eofflag, td);
750 if ((dp->d_type != DT_WHT) &&
751 (dp->d_fileno == fileno)) {
754 vn_lock(mvp, LK_EXCLUSIVE | LK_RETRY);
755 if (dirent_exists(mvp, dp->d_name, td)) {
758 vn_lock(*dvp, LK_EXCLUSIVE | LK_RETRY);
762 vn_lock(*dvp, LK_EXCLUSIVE | LK_RETRY);
770 bcopy(dp->d_name, buf + i, dp->d_namlen);
774 } while (len > 0 || !eofflag);
778 free(dirbuf, M_TEMP);
788 vn_close(mvp, FREAD, td->td_ucred, td);
790 vn_lock(vp, locked | LK_RETRY);
796 * used to fill the vfs function table to get reasonable default return values.
799 vfs_stdroot (mp, flags, vpp, td)
810 vfs_stdstatfs (mp, sbp, td)
820 vfs_stdquotactl (mp, cmds, uid, arg, td)
832 vfs_stdsync(mp, waitfor, td)
837 struct vnode *vp, *mvp;
838 int error, lockreq, allerror = 0;
840 lockreq = LK_EXCLUSIVE | LK_INTERLOCK;
841 if (waitfor != MNT_WAIT)
842 lockreq |= LK_NOWAIT;
844 * Force stale buffer cache information to be flushed.
848 MNT_VNODE_FOREACH(vp, mp, mvp) {
849 /* bv_cnt is an acceptable race here. */
850 if (vp->v_bufobj.bo_dirty.bv_cnt == 0)
854 if ((error = vget(vp, lockreq, td)) != 0) {
856 if (error == ENOENT) {
857 MNT_VNODE_FOREACH_ABORT_ILOCKED(mp, mvp);
862 error = VOP_FSYNC(vp, waitfor, td);
866 /* Do not turn this into vput. td is not always curthread. */
876 vfs_stdnosync (mp, waitfor, td)
886 vfs_stdvget (mp, ino, flags, vpp)
897 vfs_stdfhtovp (mp, fhp, vpp)
908 struct vfsconf *vfsp;
916 struct vfsconf *vfsp;
923 vfs_stdextattrctl(mp, cmd, filename_vp, attrnamespace, attrname, td)
926 struct vnode *filename_vp;
928 const char *attrname;
932 if (filename_vp != NULL)
933 VOP_UNLOCK(filename_vp, 0);
938 vfs_stdsysctl(mp, op, req)
941 struct sysctl_req *req;
947 /* end of vfs default ops */