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>
58 #include <security/mac/mac_framework.h>
61 #include <vm/vm_object.h>
62 #include <vm/vm_extern.h>
64 #include <vm/vm_map.h>
65 #include <vm/vm_page.h>
66 #include <vm/vm_pager.h>
67 #include <vm/vnode_pager.h>
69 static int vop_nolookup(struct vop_lookup_args *);
70 static int vop_norename(struct vop_rename_args *);
71 static int vop_nostrategy(struct vop_strategy_args *);
72 static int get_next_dirent(struct vnode *vp, struct dirent **dpp,
73 char *dirbuf, int dirbuflen, off_t *off,
74 char **cpos, int *len, int *eofflag,
76 static int dirent_exists(struct vnode *vp, const char *dirname,
79 #define DIRENT_MINSIZE (sizeof(struct dirent) - (MAXNAMLEN+1) + 4)
82 * This vnode table stores what we want to do if the filesystem doesn't
83 * implement a particular VOP.
85 * If there is no specific entry here, we will return EOPNOTSUPP.
87 * Note that every filesystem has to implement either vop_access
88 * or vop_accessx; failing to do so will result in immediate crash
89 * due to stack overflow, as vop_stdaccess() calls vop_stdaccessx(),
90 * which calls vop_stdaccess() etc.
93 struct vop_vector default_vnodeops = {
95 .vop_bypass = VOP_EOPNOTSUPP,
97 .vop_access = vop_stdaccess,
98 .vop_accessx = vop_stdaccessx,
99 .vop_advlock = vop_stdadvlock,
100 .vop_advlockasync = vop_stdadvlockasync,
101 .vop_bmap = vop_stdbmap,
102 .vop_close = VOP_NULL,
103 .vop_fsync = VOP_NULL,
104 .vop_getpages = vop_stdgetpages,
105 .vop_getwritemount = vop_stdgetwritemount,
106 .vop_inactive = VOP_NULL,
107 .vop_ioctl = VOP_ENOTTY,
108 .vop_kqfilter = vop_stdkqfilter,
109 .vop_islocked = vop_stdislocked,
110 .vop_lock1 = vop_stdlock,
111 .vop_lookup = vop_nolookup,
112 .vop_open = VOP_NULL,
113 .vop_pathconf = VOP_EINVAL,
114 .vop_poll = vop_nopoll,
115 .vop_putpages = vop_stdputpages,
116 .vop_readlink = VOP_EINVAL,
117 .vop_rename = vop_norename,
118 .vop_revoke = VOP_PANIC,
119 .vop_strategy = vop_nostrategy,
120 .vop_unlock = vop_stdunlock,
121 .vop_vptocnp = vop_stdvptocnp,
122 .vop_vptofh = vop_stdvptofh,
126 * Series of placeholder functions for various error returns for
131 vop_eopnotsupp(struct vop_generic_args *ap)
134 printf("vop_notsupp[%s]\n", ap->a_desc->vdesc_name);
141 vop_ebadf(struct vop_generic_args *ap)
148 vop_enotty(struct vop_generic_args *ap)
155 vop_einval(struct vop_generic_args *ap)
162 vop_enoent(struct vop_generic_args *ap)
169 vop_null(struct vop_generic_args *ap)
176 * Helper function to panic on some bad VOPs in some filesystems.
179 vop_panic(struct vop_generic_args *ap)
182 panic("filesystem goof: vop_panic[%s]", ap->a_desc->vdesc_name);
186 * vop_std<something> and vop_no<something> are default functions for use by
187 * filesystems that need the "default reasonable" implementation for a
188 * particular operation.
190 * The documentation for the operations they implement exists (if it exists)
191 * in the VOP_<SOMETHING>(9) manpage (all uppercase).
195 * Default vop for filesystems that do not support name lookup
199 struct vop_lookup_args /* {
201 struct vnode **a_vpp;
202 struct componentname *a_cnp;
213 * Handle unlock and reference counting for arguments of vop_rename
214 * for filesystems that do not implement rename operation.
217 vop_norename(struct vop_rename_args *ap)
227 * Strategy routine for VFS devices that have none.
229 * BIO_ERROR and B_INVAL must be cleared prior to calling any strategy
230 * routine. Typically this is done for a BIO_READ strategy call.
231 * Typically B_INVAL is assumed to already be clear prior to a write
232 * and should not be cleared manually unless you just made the buffer
233 * invalid. BIO_ERROR should be cleared either way.
237 vop_nostrategy (struct vop_strategy_args *ap)
239 printf("No strategy for buffer at %p\n", ap->a_bp);
240 vprint("vnode", ap->a_vp);
241 ap->a_bp->b_ioflags |= BIO_ERROR;
242 ap->a_bp->b_error = EOPNOTSUPP;
248 get_next_dirent(struct vnode *vp, struct dirent **dpp, char *dirbuf,
249 int dirbuflen, off_t *off, char **cpos, int *len,
250 int *eofflag, struct thread *td)
257 KASSERT(VOP_ISLOCKED(vp), ("vp %p is not locked", vp));
258 KASSERT(vp->v_type == VDIR, ("vp %p is not a directory", vp));
261 iov.iov_base = dirbuf;
262 iov.iov_len = dirbuflen;
266 uio.uio_offset = *off;
267 uio.uio_resid = dirbuflen;
268 uio.uio_segflg = UIO_SYSSPACE;
269 uio.uio_rw = UIO_READ;
275 error = mac_vnode_check_readdir(td->td_ucred, vp);
278 error = VOP_READDIR(vp, &uio, td->td_ucred, eofflag,
283 *off = uio.uio_offset;
286 *len = (dirbuflen - uio.uio_resid);
292 dp = (struct dirent *)(*cpos);
293 reclen = dp->d_reclen;
296 /* check for malformed directory.. */
297 if (reclen < DIRENT_MINSIZE)
307 * Check if a named file exists in a given directory vnode.
310 dirent_exists(struct vnode *vp, const char *dirname, struct thread *td)
313 int error, eofflag, dirbuflen, len, found;
318 KASSERT(VOP_ISLOCKED(vp), ("vp %p is not locked", vp));
319 KASSERT(vp->v_type == VDIR, ("vp %p is not a directory", vp));
323 error = VOP_GETATTR(vp, &va, td->td_ucred);
327 dirbuflen = DEV_BSIZE;
328 if (dirbuflen < va.va_blocksize)
329 dirbuflen = va.va_blocksize;
330 dirbuf = (char *)malloc(dirbuflen, M_TEMP, M_WAITOK);
335 error = get_next_dirent(vp, &dp, dirbuf, dirbuflen, &off,
336 &cpos, &len, &eofflag, td);
340 if ((dp->d_type != DT_WHT) &&
341 !strcmp(dp->d_name, dirname)) {
345 } while (len > 0 || !eofflag);
348 free(dirbuf, M_TEMP);
353 vop_stdaccess(struct vop_access_args *ap)
356 KASSERT((ap->a_accmode & ~(VEXEC | VWRITE | VREAD | VADMIN |
357 VAPPEND)) == 0, ("invalid bit in accmode"));
359 return (VOP_ACCESSX(ap->a_vp, ap->a_accmode, ap->a_cred, ap->a_td));
363 vop_stdaccessx(struct vop_accessx_args *ap)
366 accmode_t accmode = ap->a_accmode;
368 error = vfs_unixify_accmode(&accmode);
375 return (VOP_ACCESS(ap->a_vp, accmode, ap->a_cred, ap->a_td));
379 * Advisory record locking support
382 vop_stdadvlock(struct vop_advlock_args *ap)
390 cred = curthread->td_ucred;
391 vn_lock(vp, LK_SHARED | LK_RETRY);
392 error = VOP_GETATTR(vp, &vattr, cred);
397 return (lf_advlock(ap, &(vp->v_lockf), vattr.va_size));
401 vop_stdadvlockasync(struct vop_advlockasync_args *ap)
409 cred = curthread->td_ucred;
410 vn_lock(vp, LK_SHARED | LK_RETRY);
411 error = VOP_GETATTR(vp, &vattr, cred);
416 return (lf_advlockasync(ap, &(vp->v_lockf), vattr.va_size));
422 * Standard implementation of POSIX pathconf, to get information about limits
424 * Override per filesystem for the case where the filesystem has smaller
429 struct vop_pathconf_args /* {
436 switch (ap->a_name) {
438 *ap->a_retval = NAME_MAX;
441 *ap->a_retval = PATH_MAX;
444 *ap->a_retval = LINK_MAX;
447 *ap->a_retval = MAX_CANON;
450 *ap->a_retval = MAX_INPUT;
453 *ap->a_retval = PIPE_BUF;
455 case _PC_CHOWN_RESTRICTED:
459 *ap->a_retval = _POSIX_VDISABLE;
468 * Standard lock, unlock and islocked functions.
472 struct vop_lock1_args /* {
479 struct vnode *vp = ap->a_vp;
481 return (_lockmgr_args(vp->v_vnlock, ap->a_flags, VI_MTX(vp),
482 LK_WMESG_DEFAULT, LK_PRIO_DEFAULT, LK_TIMO_DEFAULT, ap->a_file,
489 struct vop_unlock_args /* {
494 struct vnode *vp = ap->a_vp;
496 return (lockmgr(vp->v_vnlock, ap->a_flags | LK_RELEASE, VI_MTX(vp)));
502 struct vop_islocked_args /* {
507 return (lockstatus(ap->a_vp->v_vnlock));
511 * Return true for select/poll.
515 struct vop_poll_args /* {
518 struct ucred *a_cred;
523 return (poll_no_poll(ap->a_events));
527 * Implement poll for local filesystems that support it.
531 struct vop_poll_args /* {
534 struct ucred *a_cred;
538 if (ap->a_events & ~POLLSTANDARD)
539 return (vn_pollrecord(ap->a_vp, ap->a_td, ap->a_events));
540 return (ap->a_events & (POLLIN | POLLOUT | POLLRDNORM | POLLWRNORM));
544 * Return our mount point, as we will take charge of the writes.
547 vop_stdgetwritemount(ap)
548 struct vop_getwritemount_args /* {
550 struct mount **a_mpp;
556 * XXX Since this is called unlocked we may be recycled while
557 * attempting to ref the mount. If this is the case or mountpoint
558 * will be set to NULL. We only have to prevent this call from
559 * returning with a ref to an incorrect mountpoint. It is not
560 * harmful to return with a ref to our previous mountpoint.
562 mp = ap->a_vp->v_mount;
565 if (mp != ap->a_vp->v_mount) {
574 /* XXX Needs good comment and VOP_BMAP(9) manpage */
577 struct vop_bmap_args /* {
580 struct bufobj **a_bop;
587 if (ap->a_bop != NULL)
588 *ap->a_bop = &ap->a_vp->v_bufobj;
589 if (ap->a_bnp != NULL)
590 *ap->a_bnp = ap->a_bn * btodb(ap->a_vp->v_mount->mnt_stat.f_iosize);
591 if (ap->a_runp != NULL)
593 if (ap->a_runb != NULL)
600 struct vop_fsync_args /* {
602 struct ucred *a_cred;
607 struct vnode *vp = ap->a_vp;
612 int maxretry = 1000; /* large, arbitrarily chosen */
618 * MARK/SCAN initialization to avoid infinite loops.
620 TAILQ_FOREACH(bp, &bo->bo_dirty.bv_hd, b_bobufs) {
621 bp->b_vflags &= ~BV_SCANNED;
626 * Flush all dirty buffers associated with a vnode.
629 TAILQ_FOREACH_SAFE(bp, &bo->bo_dirty.bv_hd, b_bobufs, nbp) {
630 if ((bp->b_vflags & BV_SCANNED) != 0)
632 bp->b_vflags |= BV_SCANNED;
633 if (BUF_LOCK(bp, LK_EXCLUSIVE | LK_NOWAIT, NULL))
636 KASSERT(bp->b_bufobj == bo,
637 ("bp %p wrong b_bufobj %p should be %p",
638 bp, bp->b_bufobj, bo));
639 if ((bp->b_flags & B_DELWRI) == 0)
640 panic("fsync: not dirty");
641 if ((vp->v_object != NULL) && (bp->b_flags & B_CLUSTEROK)) {
652 * If synchronous the caller expects us to completely resolve all
653 * dirty buffers in the system. Wait for in-progress I/O to
654 * complete (which could include background bitmap writes), then
655 * retry if dirty blocks still exist.
657 if (ap->a_waitfor == MNT_WAIT) {
658 bufobj_wwait(bo, 0, 0);
659 if (bo->bo_dirty.bv_cnt > 0) {
661 * If we are unable to write any of these buffers
662 * then we fail now rather than trying endlessly
665 TAILQ_FOREACH(bp, &bo->bo_dirty.bv_hd, b_bobufs)
666 if ((error = bp->b_error) == 0)
668 if (error == 0 && --maxretry >= 0)
675 vprint("fsync: giving up on dirty", vp);
680 /* XXX Needs good comment and more info in the manpage (VOP_GETPAGES(9)). */
683 struct vop_getpages_args /* {
688 vm_ooffset_t a_offset;
692 return vnode_pager_generic_getpages(ap->a_vp, ap->a_m,
693 ap->a_count, ap->a_reqpage);
697 vop_stdkqfilter(struct vop_kqfilter_args *ap)
699 return vfs_kqfilter(ap);
702 /* XXX Needs good comment and more info in the manpage (VOP_PUTPAGES(9)). */
705 struct vop_putpages_args /* {
711 vm_ooffset_t a_offset;
715 return vnode_pager_generic_putpages(ap->a_vp, ap->a_m, ap->a_count,
716 ap->a_sync, ap->a_rtvals);
720 vop_stdvptofh(struct vop_vptofh_args *ap)
726 vop_stdvptocnp(struct vop_vptocnp_args *ap)
728 struct vnode *vp = ap->a_vp;
729 struct vnode **dvp = ap->a_vpp;
730 struct ucred *cred = ap->a_cred;
731 char *buf = ap->a_buf;
732 int *buflen = ap->a_buflen;
734 int i, error, eofflag, dirbuflen, flags, locked, len, covered;
748 if (vp->v_type != VDIR)
751 error = VOP_GETATTR(vp, &va, cred);
756 locked = VOP_ISLOCKED(vp);
758 NDINIT_ATVP(&nd, LOOKUP, FOLLOW | LOCKLEAF, UIO_SYSSPACE,
761 error = vn_open_cred(&nd, &flags, 0, VN_OPEN_NOAUDIT, cred, NULL);
763 vn_lock(vp, locked | LK_RETRY);
766 NDFREE(&nd, NDF_ONLY_PNBUF);
768 mvp = *dvp = nd.ni_vp;
770 if (vp->v_mount != (*dvp)->v_mount &&
771 ((*dvp)->v_vflag & VV_ROOT) &&
772 ((*dvp)->v_mount->mnt_flag & MNT_UNION)) {
773 *dvp = (*dvp)->v_mount->mnt_vnodecovered;
776 vn_close(mvp, FREAD, cred, td);
778 vn_lock(*dvp, LK_EXCLUSIVE | LK_RETRY);
782 fileno = va.va_fileid;
784 dirbuflen = DEV_BSIZE;
785 if (dirbuflen < va.va_blocksize)
786 dirbuflen = va.va_blocksize;
787 dirbuf = (char *)malloc(dirbuflen, M_TEMP, M_WAITOK);
789 if ((*dvp)->v_type != VDIR) {
797 /* call VOP_READDIR of parent */
798 error = get_next_dirent(*dvp, &dp, dirbuf, dirbuflen, &off,
799 &cpos, &len, &eofflag, td);
803 if ((dp->d_type != DT_WHT) &&
804 (dp->d_fileno == fileno)) {
807 vn_lock(mvp, LK_EXCLUSIVE | LK_RETRY);
808 if (dirent_exists(mvp, dp->d_name, td)) {
811 vn_lock(*dvp, LK_EXCLUSIVE | LK_RETRY);
815 vn_lock(*dvp, LK_EXCLUSIVE | LK_RETRY);
823 bcopy(dp->d_name, buf + i, dp->d_namlen);
827 } while (len > 0 || !eofflag);
831 free(dirbuf, M_TEMP);
841 vn_close(mvp, FREAD, cred, td);
843 vn_lock(vp, locked | LK_RETRY);
849 * used to fill the vfs function table to get reasonable default return values.
852 vfs_stdroot (mp, flags, vpp)
862 vfs_stdstatfs (mp, sbp)
871 vfs_stdquotactl (mp, cmds, uid, arg)
882 vfs_stdsync(mp, waitfor)
886 struct vnode *vp, *mvp;
888 int error, lockreq, allerror = 0;
891 lockreq = LK_EXCLUSIVE | LK_INTERLOCK;
892 if (waitfor != MNT_WAIT)
893 lockreq |= LK_NOWAIT;
895 * Force stale buffer cache information to be flushed.
899 MNT_VNODE_FOREACH(vp, mp, mvp) {
900 /* bv_cnt is an acceptable race here. */
901 if (vp->v_bufobj.bo_dirty.bv_cnt == 0)
905 if ((error = vget(vp, lockreq, td)) != 0) {
907 if (error == ENOENT) {
908 MNT_VNODE_FOREACH_ABORT_ILOCKED(mp, mvp);
913 error = VOP_FSYNC(vp, waitfor, td);
917 /* Do not turn this into vput. td is not always curthread. */
927 vfs_stdnosync (mp, waitfor)
936 vfs_stdvget (mp, ino, flags, vpp)
947 vfs_stdfhtovp (mp, fhp, vpp)
958 struct vfsconf *vfsp;
966 struct vfsconf *vfsp;
973 vfs_stdextattrctl(mp, cmd, filename_vp, attrnamespace, attrname)
976 struct vnode *filename_vp;
978 const char *attrname;
981 if (filename_vp != NULL)
982 VOP_UNLOCK(filename_vp, 0);
987 vfs_stdsysctl(mp, op, req)
990 struct sysctl_req *req;
996 /* end of vfs default ops */