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_nostrategy(struct vop_strategy_args *);
71 static int get_next_dirent(struct vnode *vp, struct dirent **dpp,
72 char *dirbuf, int dirbuflen, off_t *off,
73 char **cpos, int *len, int *eofflag,
75 static int dirent_exists(struct vnode *vp, const char *dirname,
78 #define DIRENT_MINSIZE (sizeof(struct dirent) - (MAXNAMLEN+1) + 4)
81 * This vnode table stores what we want to do if the filesystem doesn't
82 * implement a particular VOP.
84 * If there is no specific entry here, we will return EOPNOTSUPP.
88 struct vop_vector default_vnodeops = {
90 .vop_bypass = VOP_EOPNOTSUPP,
92 .vop_accessx = vop_stdaccessx,
93 .vop_advlock = vop_stdadvlock,
94 .vop_advlockasync = vop_stdadvlockasync,
95 .vop_bmap = vop_stdbmap,
96 .vop_close = VOP_NULL,
97 .vop_fsync = VOP_NULL,
98 .vop_getpages = vop_stdgetpages,
99 .vop_getwritemount = vop_stdgetwritemount,
100 .vop_inactive = VOP_NULL,
101 .vop_ioctl = VOP_ENOTTY,
102 .vop_kqfilter = vop_stdkqfilter,
103 .vop_islocked = vop_stdislocked,
104 .vop_lock1 = vop_stdlock,
105 .vop_lookup = vop_nolookup,
106 .vop_open = VOP_NULL,
107 .vop_pathconf = VOP_EINVAL,
108 .vop_poll = vop_nopoll,
109 .vop_putpages = vop_stdputpages,
110 .vop_readlink = VOP_EINVAL,
111 .vop_revoke = VOP_PANIC,
112 .vop_strategy = vop_nostrategy,
113 .vop_unlock = vop_stdunlock,
114 .vop_vptocnp = vop_stdvptocnp,
115 .vop_vptofh = vop_stdvptofh,
119 * Series of placeholder functions for various error returns for
124 vop_eopnotsupp(struct vop_generic_args *ap)
127 printf("vop_notsupp[%s]\n", ap->a_desc->vdesc_name);
134 vop_ebadf(struct vop_generic_args *ap)
141 vop_enotty(struct vop_generic_args *ap)
148 vop_einval(struct vop_generic_args *ap)
155 vop_enoent(struct vop_generic_args *ap)
162 vop_null(struct vop_generic_args *ap)
169 * Helper function to panic on some bad VOPs in some filesystems.
172 vop_panic(struct vop_generic_args *ap)
175 panic("filesystem goof: vop_panic[%s]", ap->a_desc->vdesc_name);
179 * vop_std<something> and vop_no<something> are default functions for use by
180 * filesystems that need the "default reasonable" implementation for a
181 * particular operation.
183 * The documentation for the operations they implement exists (if it exists)
184 * in the VOP_<SOMETHING>(9) manpage (all uppercase).
188 * Default vop for filesystems that do not support name lookup
192 struct vop_lookup_args /* {
194 struct vnode **a_vpp;
195 struct componentname *a_cnp;
206 * Strategy routine for VFS devices that have none.
208 * BIO_ERROR and B_INVAL must be cleared prior to calling any strategy
209 * routine. Typically this is done for a BIO_READ strategy call.
210 * Typically B_INVAL is assumed to already be clear prior to a write
211 * and should not be cleared manually unless you just made the buffer
212 * invalid. BIO_ERROR should be cleared either way.
216 vop_nostrategy (struct vop_strategy_args *ap)
218 printf("No strategy for buffer at %p\n", ap->a_bp);
219 vprint("vnode", ap->a_vp);
220 ap->a_bp->b_ioflags |= BIO_ERROR;
221 ap->a_bp->b_error = EOPNOTSUPP;
227 get_next_dirent(struct vnode *vp, struct dirent **dpp, char *dirbuf,
228 int dirbuflen, off_t *off, char **cpos, int *len,
229 int *eofflag, struct thread *td)
236 KASSERT(VOP_ISLOCKED(vp), ("vp %p is not locked", vp));
237 KASSERT(vp->v_type == VDIR, ("vp %p is not a directory", vp));
240 iov.iov_base = dirbuf;
241 iov.iov_len = dirbuflen;
245 uio.uio_offset = *off;
246 uio.uio_resid = dirbuflen;
247 uio.uio_segflg = UIO_SYSSPACE;
248 uio.uio_rw = UIO_READ;
254 error = mac_vnode_check_readdir(td->td_ucred, vp);
257 error = VOP_READDIR(vp, &uio, td->td_ucred, eofflag,
262 *off = uio.uio_offset;
265 *len = (dirbuflen - uio.uio_resid);
268 dp = (struct dirent *)(*cpos);
269 reclen = dp->d_reclen;
272 /* check for malformed directory.. */
273 if (reclen < DIRENT_MINSIZE)
283 * Check if a named file exists in a given directory vnode.
286 dirent_exists(struct vnode *vp, const char *dirname, struct thread *td)
289 int error, eofflag, dirbuflen, len, found;
294 KASSERT(VOP_ISLOCKED(vp), ("vp %p is not locked", vp));
295 KASSERT(vp->v_type == VDIR, ("vp %p is not a directory", vp));
299 error = VOP_GETATTR(vp, &va, td->td_ucred);
303 dirbuflen = DEV_BSIZE;
304 if (dirbuflen < va.va_blocksize)
305 dirbuflen = va.va_blocksize;
306 dirbuf = (char *)malloc(dirbuflen, M_TEMP, M_WAITOK);
311 error = get_next_dirent(vp, &dp, dirbuf, dirbuflen, &off,
312 &cpos, &len, &eofflag, td);
316 if ((dp->d_type != DT_WHT) &&
317 !strcmp(dp->d_name, dirname)) {
321 } while (len > 0 || !eofflag);
324 free(dirbuf, M_TEMP);
329 vop_stdaccessx(struct vop_accessx_args *ap)
332 accmode_t accmode = ap->a_accmode;
334 error = vfs_unixify_accmode(&accmode);
341 return (VOP_ACCESS(ap->a_vp, accmode, ap->a_cred, ap->a_td));
345 * Advisory record locking support
348 vop_stdadvlock(struct vop_advlock_args *ap)
356 cred = curthread->td_ucred;
357 vn_lock(vp, LK_SHARED | LK_RETRY);
358 error = VOP_GETATTR(vp, &vattr, cred);
363 return (lf_advlock(ap, &(vp->v_lockf), vattr.va_size));
367 vop_stdadvlockasync(struct vop_advlockasync_args *ap)
375 cred = curthread->td_ucred;
376 vn_lock(vp, LK_SHARED | LK_RETRY);
377 error = VOP_GETATTR(vp, &vattr, cred);
382 return (lf_advlockasync(ap, &(vp->v_lockf), vattr.va_size));
388 * Standard implementation of POSIX pathconf, to get information about limits
390 * Override per filesystem for the case where the filesystem has smaller
395 struct vop_pathconf_args /* {
402 switch (ap->a_name) {
404 *ap->a_retval = NAME_MAX;
407 *ap->a_retval = PATH_MAX;
410 *ap->a_retval = LINK_MAX;
413 *ap->a_retval = MAX_CANON;
416 *ap->a_retval = MAX_INPUT;
419 *ap->a_retval = PIPE_BUF;
421 case _PC_CHOWN_RESTRICTED:
425 *ap->a_retval = _POSIX_VDISABLE;
434 * Standard lock, unlock and islocked functions.
438 struct vop_lock1_args /* {
445 struct vnode *vp = ap->a_vp;
447 return (_lockmgr_args(vp->v_vnlock, ap->a_flags, VI_MTX(vp),
448 LK_WMESG_DEFAULT, LK_PRIO_DEFAULT, LK_TIMO_DEFAULT, ap->a_file,
455 struct vop_unlock_args /* {
460 struct vnode *vp = ap->a_vp;
462 return (lockmgr(vp->v_vnlock, ap->a_flags | LK_RELEASE, VI_MTX(vp)));
468 struct vop_islocked_args /* {
473 return (lockstatus(ap->a_vp->v_vnlock));
477 * Return true for select/poll.
481 struct vop_poll_args /* {
484 struct ucred *a_cred;
489 return (poll_no_poll(ap->a_events));
493 * Implement poll for local filesystems that support it.
497 struct vop_poll_args /* {
500 struct ucred *a_cred;
504 if (ap->a_events & ~POLLSTANDARD)
505 return (vn_pollrecord(ap->a_vp, ap->a_td, ap->a_events));
506 return (ap->a_events & (POLLIN | POLLOUT | POLLRDNORM | POLLWRNORM));
510 * Return our mount point, as we will take charge of the writes.
513 vop_stdgetwritemount(ap)
514 struct vop_getwritemount_args /* {
516 struct mount **a_mpp;
522 * XXX Since this is called unlocked we may be recycled while
523 * attempting to ref the mount. If this is the case or mountpoint
524 * will be set to NULL. We only have to prevent this call from
525 * returning with a ref to an incorrect mountpoint. It is not
526 * harmful to return with a ref to our previous mountpoint.
528 mp = ap->a_vp->v_mount;
531 if (mp != ap->a_vp->v_mount) {
540 /* XXX Needs good comment and VOP_BMAP(9) manpage */
543 struct vop_bmap_args /* {
546 struct bufobj **a_bop;
553 if (ap->a_bop != NULL)
554 *ap->a_bop = &ap->a_vp->v_bufobj;
555 if (ap->a_bnp != NULL)
556 *ap->a_bnp = ap->a_bn * btodb(ap->a_vp->v_mount->mnt_stat.f_iosize);
557 if (ap->a_runp != NULL)
559 if (ap->a_runb != NULL)
566 struct vop_fsync_args /* {
568 struct ucred *a_cred;
573 struct vnode *vp = ap->a_vp;
578 int maxretry = 1000; /* large, arbitrarily chosen */
584 * MARK/SCAN initialization to avoid infinite loops.
586 TAILQ_FOREACH(bp, &bo->bo_dirty.bv_hd, b_bobufs) {
587 bp->b_vflags &= ~BV_SCANNED;
592 * Flush all dirty buffers associated with a vnode.
595 TAILQ_FOREACH_SAFE(bp, &bo->bo_dirty.bv_hd, b_bobufs, nbp) {
596 if ((bp->b_vflags & BV_SCANNED) != 0)
598 bp->b_vflags |= BV_SCANNED;
599 if (BUF_LOCK(bp, LK_EXCLUSIVE | LK_NOWAIT, NULL))
602 KASSERT(bp->b_bufobj == bo,
603 ("bp %p wrong b_bufobj %p should be %p",
604 bp, bp->b_bufobj, bo));
605 if ((bp->b_flags & B_DELWRI) == 0)
606 panic("fsync: not dirty");
607 if ((vp->v_object != NULL) && (bp->b_flags & B_CLUSTEROK)) {
618 * If synchronous the caller expects us to completely resolve all
619 * dirty buffers in the system. Wait for in-progress I/O to
620 * complete (which could include background bitmap writes), then
621 * retry if dirty blocks still exist.
623 if (ap->a_waitfor == MNT_WAIT) {
624 bufobj_wwait(bo, 0, 0);
625 if (bo->bo_dirty.bv_cnt > 0) {
627 * If we are unable to write any of these buffers
628 * then we fail now rather than trying endlessly
631 TAILQ_FOREACH(bp, &bo->bo_dirty.bv_hd, b_bobufs)
632 if ((error = bp->b_error) == 0)
634 if (error == 0 && --maxretry >= 0)
641 vprint("fsync: giving up on dirty", vp);
646 /* XXX Needs good comment and more info in the manpage (VOP_GETPAGES(9)). */
649 struct vop_getpages_args /* {
654 vm_ooffset_t a_offset;
658 return vnode_pager_generic_getpages(ap->a_vp, ap->a_m,
659 ap->a_count, ap->a_reqpage);
663 vop_stdkqfilter(struct vop_kqfilter_args *ap)
665 return vfs_kqfilter(ap);
668 /* XXX Needs good comment and more info in the manpage (VOP_PUTPAGES(9)). */
671 struct vop_putpages_args /* {
677 vm_ooffset_t a_offset;
681 return vnode_pager_generic_putpages(ap->a_vp, ap->a_m, ap->a_count,
682 ap->a_sync, ap->a_rtvals);
686 vop_stdvptofh(struct vop_vptofh_args *ap)
692 vop_stdvptocnp(struct vop_vptocnp_args *ap)
694 struct vnode *vp = ap->a_vp;
695 struct vnode **dvp = ap->a_vpp;
696 struct ucred *cred = ap->a_cred;
697 char *buf = ap->a_buf;
698 int *buflen = ap->a_buflen;
700 int i, error, eofflag, dirbuflen, flags, locked, len, covered;
714 if (vp->v_type != VDIR)
717 error = VOP_GETATTR(vp, &va, cred);
722 locked = VOP_ISLOCKED(vp);
724 NDINIT_ATVP(&nd, LOOKUP, FOLLOW | LOCKLEAF, UIO_SYSSPACE,
727 error = vn_open_cred(&nd, &flags, 0, VN_OPEN_NOAUDIT, cred, NULL);
729 vn_lock(vp, locked | LK_RETRY);
732 NDFREE(&nd, NDF_ONLY_PNBUF);
734 mvp = *dvp = nd.ni_vp;
736 if (vp->v_mount != (*dvp)->v_mount &&
737 ((*dvp)->v_vflag & VV_ROOT) &&
738 ((*dvp)->v_mount->mnt_flag & MNT_UNION)) {
739 *dvp = (*dvp)->v_mount->mnt_vnodecovered;
742 vn_close(mvp, FREAD, cred, td);
744 vn_lock(*dvp, LK_EXCLUSIVE | LK_RETRY);
748 fileno = va.va_fileid;
750 dirbuflen = DEV_BSIZE;
751 if (dirbuflen < va.va_blocksize)
752 dirbuflen = va.va_blocksize;
753 dirbuf = (char *)malloc(dirbuflen, M_TEMP, M_WAITOK);
755 if ((*dvp)->v_type != VDIR) {
763 /* call VOP_READDIR of parent */
764 error = get_next_dirent(*dvp, &dp, dirbuf, dirbuflen, &off,
765 &cpos, &len, &eofflag, td);
769 if ((dp->d_type != DT_WHT) &&
770 (dp->d_fileno == fileno)) {
773 vn_lock(mvp, LK_EXCLUSIVE | LK_RETRY);
774 if (dirent_exists(mvp, dp->d_name, td)) {
777 vn_lock(*dvp, LK_EXCLUSIVE | LK_RETRY);
781 vn_lock(*dvp, LK_EXCLUSIVE | LK_RETRY);
789 bcopy(dp->d_name, buf + i, dp->d_namlen);
793 } while (len > 0 || !eofflag);
797 free(dirbuf, M_TEMP);
807 vn_close(mvp, FREAD, cred, td);
809 vn_lock(vp, locked | LK_RETRY);
815 * used to fill the vfs function table to get reasonable default return values.
818 vfs_stdroot (mp, flags, vpp)
828 vfs_stdstatfs (mp, sbp)
837 vfs_stdquotactl (mp, cmds, uid, arg)
848 vfs_stdsync(mp, waitfor)
852 struct vnode *vp, *mvp;
854 int error, lockreq, allerror = 0;
857 lockreq = LK_EXCLUSIVE | LK_INTERLOCK;
858 if (waitfor != MNT_WAIT)
859 lockreq |= LK_NOWAIT;
861 * Force stale buffer cache information to be flushed.
865 MNT_VNODE_FOREACH(vp, mp, mvp) {
866 /* bv_cnt is an acceptable race here. */
867 if (vp->v_bufobj.bo_dirty.bv_cnt == 0)
871 if ((error = vget(vp, lockreq, td)) != 0) {
873 if (error == ENOENT) {
874 MNT_VNODE_FOREACH_ABORT_ILOCKED(mp, mvp);
879 error = VOP_FSYNC(vp, waitfor, td);
883 /* Do not turn this into vput. td is not always curthread. */
893 vfs_stdnosync (mp, waitfor)
902 vfs_stdvget (mp, ino, flags, vpp)
913 vfs_stdfhtovp (mp, fhp, vpp)
924 struct vfsconf *vfsp;
932 struct vfsconf *vfsp;
939 vfs_stdextattrctl(mp, cmd, filename_vp, attrnamespace, attrname)
942 struct vnode *filename_vp;
944 const char *attrname;
947 if (filename_vp != NULL)
948 VOP_UNLOCK(filename_vp, 0);
953 vfs_stdsysctl(mp, op, req)
956 struct sysctl_req *req;
962 /* end of vfs default ops */