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);
289 dp = (struct dirent *)(*cpos);
290 reclen = dp->d_reclen;
293 /* check for malformed directory.. */
294 if (reclen < DIRENT_MINSIZE)
304 * Check if a named file exists in a given directory vnode.
307 dirent_exists(struct vnode *vp, const char *dirname, struct thread *td)
310 int error, eofflag, dirbuflen, len, found;
315 KASSERT(VOP_ISLOCKED(vp), ("vp %p is not locked", vp));
316 KASSERT(vp->v_type == VDIR, ("vp %p is not a directory", vp));
320 error = VOP_GETATTR(vp, &va, td->td_ucred);
324 dirbuflen = DEV_BSIZE;
325 if (dirbuflen < va.va_blocksize)
326 dirbuflen = va.va_blocksize;
327 dirbuf = (char *)malloc(dirbuflen, M_TEMP, M_WAITOK);
332 error = get_next_dirent(vp, &dp, dirbuf, dirbuflen, &off,
333 &cpos, &len, &eofflag, td);
337 if ((dp->d_type != DT_WHT) &&
338 !strcmp(dp->d_name, dirname)) {
342 } while (len > 0 || !eofflag);
345 free(dirbuf, M_TEMP);
350 vop_stdaccess(struct vop_access_args *ap)
353 KASSERT((ap->a_accmode & ~(VEXEC | VWRITE | VREAD | VADMIN |
354 VAPPEND)) == 0, ("invalid bit in accmode"));
356 return (VOP_ACCESSX(ap->a_vp, ap->a_accmode, ap->a_cred, ap->a_td));
360 vop_stdaccessx(struct vop_accessx_args *ap)
363 accmode_t accmode = ap->a_accmode;
365 error = vfs_unixify_accmode(&accmode);
372 return (VOP_ACCESS(ap->a_vp, accmode, ap->a_cred, ap->a_td));
376 * Advisory record locking support
379 vop_stdadvlock(struct vop_advlock_args *ap)
387 cred = curthread->td_ucred;
388 vn_lock(vp, LK_SHARED | LK_RETRY);
389 error = VOP_GETATTR(vp, &vattr, cred);
394 return (lf_advlock(ap, &(vp->v_lockf), vattr.va_size));
398 vop_stdadvlockasync(struct vop_advlockasync_args *ap)
406 cred = curthread->td_ucred;
407 vn_lock(vp, LK_SHARED | LK_RETRY);
408 error = VOP_GETATTR(vp, &vattr, cred);
413 return (lf_advlockasync(ap, &(vp->v_lockf), vattr.va_size));
419 * Standard implementation of POSIX pathconf, to get information about limits
421 * Override per filesystem for the case where the filesystem has smaller
426 struct vop_pathconf_args /* {
433 switch (ap->a_name) {
435 *ap->a_retval = NAME_MAX;
438 *ap->a_retval = PATH_MAX;
441 *ap->a_retval = LINK_MAX;
444 *ap->a_retval = MAX_CANON;
447 *ap->a_retval = MAX_INPUT;
450 *ap->a_retval = PIPE_BUF;
452 case _PC_CHOWN_RESTRICTED:
456 *ap->a_retval = _POSIX_VDISABLE;
465 * Standard lock, unlock and islocked functions.
469 struct vop_lock1_args /* {
476 struct vnode *vp = ap->a_vp;
478 return (_lockmgr_args(vp->v_vnlock, ap->a_flags, VI_MTX(vp),
479 LK_WMESG_DEFAULT, LK_PRIO_DEFAULT, LK_TIMO_DEFAULT, ap->a_file,
486 struct vop_unlock_args /* {
491 struct vnode *vp = ap->a_vp;
493 return (lockmgr(vp->v_vnlock, ap->a_flags | LK_RELEASE, VI_MTX(vp)));
499 struct vop_islocked_args /* {
504 return (lockstatus(ap->a_vp->v_vnlock));
508 * Return true for select/poll.
512 struct vop_poll_args /* {
515 struct ucred *a_cred;
520 return (poll_no_poll(ap->a_events));
524 * Implement poll for local filesystems that support it.
528 struct vop_poll_args /* {
531 struct ucred *a_cred;
535 if (ap->a_events & ~POLLSTANDARD)
536 return (vn_pollrecord(ap->a_vp, ap->a_td, ap->a_events));
537 return (ap->a_events & (POLLIN | POLLOUT | POLLRDNORM | POLLWRNORM));
541 * Return our mount point, as we will take charge of the writes.
544 vop_stdgetwritemount(ap)
545 struct vop_getwritemount_args /* {
547 struct mount **a_mpp;
553 * XXX Since this is called unlocked we may be recycled while
554 * attempting to ref the mount. If this is the case or mountpoint
555 * will be set to NULL. We only have to prevent this call from
556 * returning with a ref to an incorrect mountpoint. It is not
557 * harmful to return with a ref to our previous mountpoint.
559 mp = ap->a_vp->v_mount;
562 if (mp != ap->a_vp->v_mount) {
571 /* XXX Needs good comment and VOP_BMAP(9) manpage */
574 struct vop_bmap_args /* {
577 struct bufobj **a_bop;
584 if (ap->a_bop != NULL)
585 *ap->a_bop = &ap->a_vp->v_bufobj;
586 if (ap->a_bnp != NULL)
587 *ap->a_bnp = ap->a_bn * btodb(ap->a_vp->v_mount->mnt_stat.f_iosize);
588 if (ap->a_runp != NULL)
590 if (ap->a_runb != NULL)
597 struct vop_fsync_args /* {
599 struct ucred *a_cred;
604 struct vnode *vp = ap->a_vp;
609 int maxretry = 1000; /* large, arbitrarily chosen */
615 * MARK/SCAN initialization to avoid infinite loops.
617 TAILQ_FOREACH(bp, &bo->bo_dirty.bv_hd, b_bobufs) {
618 bp->b_vflags &= ~BV_SCANNED;
623 * Flush all dirty buffers associated with a vnode.
626 TAILQ_FOREACH_SAFE(bp, &bo->bo_dirty.bv_hd, b_bobufs, nbp) {
627 if ((bp->b_vflags & BV_SCANNED) != 0)
629 bp->b_vflags |= BV_SCANNED;
630 if (BUF_LOCK(bp, LK_EXCLUSIVE | LK_NOWAIT, NULL))
633 KASSERT(bp->b_bufobj == bo,
634 ("bp %p wrong b_bufobj %p should be %p",
635 bp, bp->b_bufobj, bo));
636 if ((bp->b_flags & B_DELWRI) == 0)
637 panic("fsync: not dirty");
638 if ((vp->v_object != NULL) && (bp->b_flags & B_CLUSTEROK)) {
649 * If synchronous the caller expects us to completely resolve all
650 * dirty buffers in the system. Wait for in-progress I/O to
651 * complete (which could include background bitmap writes), then
652 * retry if dirty blocks still exist.
654 if (ap->a_waitfor == MNT_WAIT) {
655 bufobj_wwait(bo, 0, 0);
656 if (bo->bo_dirty.bv_cnt > 0) {
658 * If we are unable to write any of these buffers
659 * then we fail now rather than trying endlessly
662 TAILQ_FOREACH(bp, &bo->bo_dirty.bv_hd, b_bobufs)
663 if ((error = bp->b_error) == 0)
665 if (error == 0 && --maxretry >= 0)
672 vprint("fsync: giving up on dirty", vp);
677 /* XXX Needs good comment and more info in the manpage (VOP_GETPAGES(9)). */
680 struct vop_getpages_args /* {
685 vm_ooffset_t a_offset;
689 return vnode_pager_generic_getpages(ap->a_vp, ap->a_m,
690 ap->a_count, ap->a_reqpage);
694 vop_stdkqfilter(struct vop_kqfilter_args *ap)
696 return vfs_kqfilter(ap);
699 /* XXX Needs good comment and more info in the manpage (VOP_PUTPAGES(9)). */
702 struct vop_putpages_args /* {
708 vm_ooffset_t a_offset;
712 return vnode_pager_generic_putpages(ap->a_vp, ap->a_m, ap->a_count,
713 ap->a_sync, ap->a_rtvals);
717 vop_stdvptofh(struct vop_vptofh_args *ap)
723 vop_stdvptocnp(struct vop_vptocnp_args *ap)
725 struct vnode *vp = ap->a_vp;
726 struct vnode **dvp = ap->a_vpp;
727 struct ucred *cred = ap->a_cred;
728 char *buf = ap->a_buf;
729 int *buflen = ap->a_buflen;
731 int i, error, eofflag, dirbuflen, flags, locked, len, covered;
745 if (vp->v_type != VDIR)
748 error = VOP_GETATTR(vp, &va, cred);
753 locked = VOP_ISLOCKED(vp);
755 NDINIT_ATVP(&nd, LOOKUP, FOLLOW | LOCKLEAF, UIO_SYSSPACE,
758 error = vn_open_cred(&nd, &flags, 0, VN_OPEN_NOAUDIT, cred, NULL);
760 vn_lock(vp, locked | LK_RETRY);
763 NDFREE(&nd, NDF_ONLY_PNBUF);
765 mvp = *dvp = nd.ni_vp;
767 if (vp->v_mount != (*dvp)->v_mount &&
768 ((*dvp)->v_vflag & VV_ROOT) &&
769 ((*dvp)->v_mount->mnt_flag & MNT_UNION)) {
770 *dvp = (*dvp)->v_mount->mnt_vnodecovered;
773 vn_close(mvp, FREAD, cred, td);
775 vn_lock(*dvp, LK_EXCLUSIVE | LK_RETRY);
779 fileno = va.va_fileid;
781 dirbuflen = DEV_BSIZE;
782 if (dirbuflen < va.va_blocksize)
783 dirbuflen = va.va_blocksize;
784 dirbuf = (char *)malloc(dirbuflen, M_TEMP, M_WAITOK);
786 if ((*dvp)->v_type != VDIR) {
794 /* call VOP_READDIR of parent */
795 error = get_next_dirent(*dvp, &dp, dirbuf, dirbuflen, &off,
796 &cpos, &len, &eofflag, td);
800 if ((dp->d_type != DT_WHT) &&
801 (dp->d_fileno == fileno)) {
804 vn_lock(mvp, LK_EXCLUSIVE | LK_RETRY);
805 if (dirent_exists(mvp, dp->d_name, td)) {
808 vn_lock(*dvp, LK_EXCLUSIVE | LK_RETRY);
812 vn_lock(*dvp, LK_EXCLUSIVE | LK_RETRY);
820 bcopy(dp->d_name, buf + i, dp->d_namlen);
824 } while (len > 0 || !eofflag);
828 free(dirbuf, M_TEMP);
838 vn_close(mvp, FREAD, cred, td);
840 vn_lock(vp, locked | LK_RETRY);
846 * used to fill the vfs function table to get reasonable default return values.
849 vfs_stdroot (mp, flags, vpp)
859 vfs_stdstatfs (mp, sbp)
868 vfs_stdquotactl (mp, cmds, uid, arg)
879 vfs_stdsync(mp, waitfor)
883 struct vnode *vp, *mvp;
885 int error, lockreq, allerror = 0;
888 lockreq = LK_EXCLUSIVE | LK_INTERLOCK;
889 if (waitfor != MNT_WAIT)
890 lockreq |= LK_NOWAIT;
892 * Force stale buffer cache information to be flushed.
896 MNT_VNODE_FOREACH(vp, mp, mvp) {
897 /* bv_cnt is an acceptable race here. */
898 if (vp->v_bufobj.bo_dirty.bv_cnt == 0)
902 if ((error = vget(vp, lockreq, td)) != 0) {
904 if (error == ENOENT) {
905 MNT_VNODE_FOREACH_ABORT_ILOCKED(mp, mvp);
910 error = VOP_FSYNC(vp, waitfor, td);
914 /* Do not turn this into vput. td is not always curthread. */
924 vfs_stdnosync (mp, waitfor)
933 vfs_stdvget (mp, ino, flags, vpp)
944 vfs_stdfhtovp (mp, fhp, vpp)
955 struct vfsconf *vfsp;
963 struct vfsconf *vfsp;
970 vfs_stdextattrctl(mp, cmd, filename_vp, attrnamespace, attrname)
973 struct vnode *filename_vp;
975 const char *attrname;
978 if (filename_vp != NULL)
979 VOP_UNLOCK(filename_vp, 0);
984 vfs_stdsysctl(mp, op, req)
987 struct sysctl_req *req;
993 /* end of vfs default ops */