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.
86 * Note that every filesystem has to implement either vop_access
87 * or vop_accessx; failing to do so will result in immediate crash
88 * due to stack overflow, as vop_stdaccess() calls vop_stdaccessx(),
89 * which calls vop_stdaccess() etc.
92 struct vop_vector default_vnodeops = {
94 .vop_bypass = VOP_EOPNOTSUPP,
96 .vop_access = vop_stdaccess,
97 .vop_accessx = vop_stdaccessx,
98 .vop_advlock = vop_stdadvlock,
99 .vop_advlockasync = vop_stdadvlockasync,
100 .vop_bmap = vop_stdbmap,
101 .vop_close = VOP_NULL,
102 .vop_fsync = VOP_NULL,
103 .vop_getpages = vop_stdgetpages,
104 .vop_getwritemount = vop_stdgetwritemount,
105 .vop_inactive = VOP_NULL,
106 .vop_ioctl = VOP_ENOTTY,
107 .vop_kqfilter = vop_stdkqfilter,
108 .vop_islocked = vop_stdislocked,
109 .vop_lock1 = vop_stdlock,
110 .vop_lookup = vop_nolookup,
111 .vop_open = VOP_NULL,
112 .vop_pathconf = VOP_EINVAL,
113 .vop_poll = vop_nopoll,
114 .vop_putpages = vop_stdputpages,
115 .vop_readlink = VOP_EINVAL,
116 .vop_revoke = VOP_PANIC,
117 .vop_strategy = vop_nostrategy,
118 .vop_unlock = vop_stdunlock,
119 .vop_vptocnp = vop_stdvptocnp,
120 .vop_vptofh = vop_stdvptofh,
124 * Series of placeholder functions for various error returns for
129 vop_eopnotsupp(struct vop_generic_args *ap)
132 printf("vop_notsupp[%s]\n", ap->a_desc->vdesc_name);
139 vop_ebadf(struct vop_generic_args *ap)
146 vop_enotty(struct vop_generic_args *ap)
153 vop_einval(struct vop_generic_args *ap)
160 vop_enoent(struct vop_generic_args *ap)
167 vop_null(struct vop_generic_args *ap)
174 * Helper function to panic on some bad VOPs in some filesystems.
177 vop_panic(struct vop_generic_args *ap)
180 panic("filesystem goof: vop_panic[%s]", ap->a_desc->vdesc_name);
184 * vop_std<something> and vop_no<something> are default functions for use by
185 * filesystems that need the "default reasonable" implementation for a
186 * particular operation.
188 * The documentation for the operations they implement exists (if it exists)
189 * in the VOP_<SOMETHING>(9) manpage (all uppercase).
193 * Default vop for filesystems that do not support name lookup
197 struct vop_lookup_args /* {
199 struct vnode **a_vpp;
200 struct componentname *a_cnp;
211 * Strategy routine for VFS devices that have none.
213 * BIO_ERROR and B_INVAL must be cleared prior to calling any strategy
214 * routine. Typically this is done for a BIO_READ strategy call.
215 * Typically B_INVAL is assumed to already be clear prior to a write
216 * and should not be cleared manually unless you just made the buffer
217 * invalid. BIO_ERROR should be cleared either way.
221 vop_nostrategy (struct vop_strategy_args *ap)
223 printf("No strategy for buffer at %p\n", ap->a_bp);
224 vprint("vnode", ap->a_vp);
225 ap->a_bp->b_ioflags |= BIO_ERROR;
226 ap->a_bp->b_error = EOPNOTSUPP;
232 get_next_dirent(struct vnode *vp, struct dirent **dpp, char *dirbuf,
233 int dirbuflen, off_t *off, char **cpos, int *len,
234 int *eofflag, struct thread *td)
241 KASSERT(VOP_ISLOCKED(vp), ("vp %p is not locked", vp));
242 KASSERT(vp->v_type == VDIR, ("vp %p is not a directory", vp));
245 iov.iov_base = dirbuf;
246 iov.iov_len = dirbuflen;
250 uio.uio_offset = *off;
251 uio.uio_resid = dirbuflen;
252 uio.uio_segflg = UIO_SYSSPACE;
253 uio.uio_rw = UIO_READ;
259 error = mac_vnode_check_readdir(td->td_ucred, vp);
262 error = VOP_READDIR(vp, &uio, td->td_ucred, eofflag,
267 *off = uio.uio_offset;
270 *len = (dirbuflen - uio.uio_resid);
273 dp = (struct dirent *)(*cpos);
274 reclen = dp->d_reclen;
277 /* check for malformed directory.. */
278 if (reclen < DIRENT_MINSIZE)
288 * Check if a named file exists in a given directory vnode.
291 dirent_exists(struct vnode *vp, const char *dirname, struct thread *td)
294 int error, eofflag, dirbuflen, len, found;
299 KASSERT(VOP_ISLOCKED(vp), ("vp %p is not locked", vp));
300 KASSERT(vp->v_type == VDIR, ("vp %p is not a directory", vp));
304 error = VOP_GETATTR(vp, &va, td->td_ucred);
308 dirbuflen = DEV_BSIZE;
309 if (dirbuflen < va.va_blocksize)
310 dirbuflen = va.va_blocksize;
311 dirbuf = (char *)malloc(dirbuflen, M_TEMP, M_WAITOK);
316 error = get_next_dirent(vp, &dp, dirbuf, dirbuflen, &off,
317 &cpos, &len, &eofflag, td);
321 if ((dp->d_type != DT_WHT) &&
322 !strcmp(dp->d_name, dirname)) {
326 } while (len > 0 || !eofflag);
329 free(dirbuf, M_TEMP);
334 vop_stdaccess(struct vop_access_args *ap)
337 KASSERT((ap->a_accmode & ~(VEXEC | VWRITE | VREAD | VADMIN |
338 VAPPEND)) == 0, ("invalid bit in accmode"));
340 return (VOP_ACCESSX(ap->a_vp, ap->a_accmode, ap->a_cred, ap->a_td));
344 vop_stdaccessx(struct vop_accessx_args *ap)
347 accmode_t accmode = ap->a_accmode;
349 error = vfs_unixify_accmode(&accmode);
356 return (VOP_ACCESS(ap->a_vp, accmode, ap->a_cred, ap->a_td));
360 * Advisory record locking support
363 vop_stdadvlock(struct vop_advlock_args *ap)
371 cred = curthread->td_ucred;
372 vn_lock(vp, LK_SHARED | LK_RETRY);
373 error = VOP_GETATTR(vp, &vattr, cred);
378 return (lf_advlock(ap, &(vp->v_lockf), vattr.va_size));
382 vop_stdadvlockasync(struct vop_advlockasync_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_advlockasync(ap, &(vp->v_lockf), vattr.va_size));
403 * Standard implementation of POSIX pathconf, to get information about limits
405 * Override per filesystem for the case where the filesystem has smaller
410 struct vop_pathconf_args /* {
417 switch (ap->a_name) {
419 *ap->a_retval = NAME_MAX;
422 *ap->a_retval = PATH_MAX;
425 *ap->a_retval = LINK_MAX;
428 *ap->a_retval = MAX_CANON;
431 *ap->a_retval = MAX_INPUT;
434 *ap->a_retval = PIPE_BUF;
436 case _PC_CHOWN_RESTRICTED:
440 *ap->a_retval = _POSIX_VDISABLE;
449 * Standard lock, unlock and islocked functions.
453 struct vop_lock1_args /* {
460 struct vnode *vp = ap->a_vp;
462 return (_lockmgr_args(vp->v_vnlock, ap->a_flags, VI_MTX(vp),
463 LK_WMESG_DEFAULT, LK_PRIO_DEFAULT, LK_TIMO_DEFAULT, ap->a_file,
470 struct vop_unlock_args /* {
475 struct vnode *vp = ap->a_vp;
477 return (lockmgr(vp->v_vnlock, ap->a_flags | LK_RELEASE, VI_MTX(vp)));
483 struct vop_islocked_args /* {
488 return (lockstatus(ap->a_vp->v_vnlock));
492 * Return true for select/poll.
496 struct vop_poll_args /* {
499 struct ucred *a_cred;
504 return (poll_no_poll(ap->a_events));
508 * Implement poll for local filesystems that support it.
512 struct vop_poll_args /* {
515 struct ucred *a_cred;
519 if (ap->a_events & ~POLLSTANDARD)
520 return (vn_pollrecord(ap->a_vp, ap->a_td, ap->a_events));
521 return (ap->a_events & (POLLIN | POLLOUT | POLLRDNORM | POLLWRNORM));
525 * Return our mount point, as we will take charge of the writes.
528 vop_stdgetwritemount(ap)
529 struct vop_getwritemount_args /* {
531 struct mount **a_mpp;
537 * XXX Since this is called unlocked we may be recycled while
538 * attempting to ref the mount. If this is the case or mountpoint
539 * will be set to NULL. We only have to prevent this call from
540 * returning with a ref to an incorrect mountpoint. It is not
541 * harmful to return with a ref to our previous mountpoint.
543 mp = ap->a_vp->v_mount;
546 if (mp != ap->a_vp->v_mount) {
555 /* XXX Needs good comment and VOP_BMAP(9) manpage */
558 struct vop_bmap_args /* {
561 struct bufobj **a_bop;
568 if (ap->a_bop != NULL)
569 *ap->a_bop = &ap->a_vp->v_bufobj;
570 if (ap->a_bnp != NULL)
571 *ap->a_bnp = ap->a_bn * btodb(ap->a_vp->v_mount->mnt_stat.f_iosize);
572 if (ap->a_runp != NULL)
574 if (ap->a_runb != NULL)
581 struct vop_fsync_args /* {
583 struct ucred *a_cred;
588 struct vnode *vp = ap->a_vp;
593 int maxretry = 1000; /* large, arbitrarily chosen */
599 * MARK/SCAN initialization to avoid infinite loops.
601 TAILQ_FOREACH(bp, &bo->bo_dirty.bv_hd, b_bobufs) {
602 bp->b_vflags &= ~BV_SCANNED;
607 * Flush all dirty buffers associated with a vnode.
610 TAILQ_FOREACH_SAFE(bp, &bo->bo_dirty.bv_hd, b_bobufs, nbp) {
611 if ((bp->b_vflags & BV_SCANNED) != 0)
613 bp->b_vflags |= BV_SCANNED;
614 if (BUF_LOCK(bp, LK_EXCLUSIVE | LK_NOWAIT, NULL))
617 KASSERT(bp->b_bufobj == bo,
618 ("bp %p wrong b_bufobj %p should be %p",
619 bp, bp->b_bufobj, bo));
620 if ((bp->b_flags & B_DELWRI) == 0)
621 panic("fsync: not dirty");
622 if ((vp->v_object != NULL) && (bp->b_flags & B_CLUSTEROK)) {
633 * If synchronous the caller expects us to completely resolve all
634 * dirty buffers in the system. Wait for in-progress I/O to
635 * complete (which could include background bitmap writes), then
636 * retry if dirty blocks still exist.
638 if (ap->a_waitfor == MNT_WAIT) {
639 bufobj_wwait(bo, 0, 0);
640 if (bo->bo_dirty.bv_cnt > 0) {
642 * If we are unable to write any of these buffers
643 * then we fail now rather than trying endlessly
646 TAILQ_FOREACH(bp, &bo->bo_dirty.bv_hd, b_bobufs)
647 if ((error = bp->b_error) == 0)
649 if (error == 0 && --maxretry >= 0)
656 vprint("fsync: giving up on dirty", vp);
661 /* XXX Needs good comment and more info in the manpage (VOP_GETPAGES(9)). */
664 struct vop_getpages_args /* {
669 vm_ooffset_t a_offset;
673 return vnode_pager_generic_getpages(ap->a_vp, ap->a_m,
674 ap->a_count, ap->a_reqpage);
678 vop_stdkqfilter(struct vop_kqfilter_args *ap)
680 return vfs_kqfilter(ap);
683 /* XXX Needs good comment and more info in the manpage (VOP_PUTPAGES(9)). */
686 struct vop_putpages_args /* {
692 vm_ooffset_t a_offset;
696 return vnode_pager_generic_putpages(ap->a_vp, ap->a_m, ap->a_count,
697 ap->a_sync, ap->a_rtvals);
701 vop_stdvptofh(struct vop_vptofh_args *ap)
707 vop_stdvptocnp(struct vop_vptocnp_args *ap)
709 struct vnode *vp = ap->a_vp;
710 struct vnode **dvp = ap->a_vpp;
711 struct ucred *cred = ap->a_cred;
712 char *buf = ap->a_buf;
713 int *buflen = ap->a_buflen;
715 int i, error, eofflag, dirbuflen, flags, locked, len, covered;
729 if (vp->v_type != VDIR)
732 error = VOP_GETATTR(vp, &va, cred);
737 locked = VOP_ISLOCKED(vp);
739 NDINIT_ATVP(&nd, LOOKUP, FOLLOW | LOCKLEAF, UIO_SYSSPACE,
742 error = vn_open_cred(&nd, &flags, 0, VN_OPEN_NOAUDIT, cred, NULL);
744 vn_lock(vp, locked | LK_RETRY);
747 NDFREE(&nd, NDF_ONLY_PNBUF);
749 mvp = *dvp = nd.ni_vp;
751 if (vp->v_mount != (*dvp)->v_mount &&
752 ((*dvp)->v_vflag & VV_ROOT) &&
753 ((*dvp)->v_mount->mnt_flag & MNT_UNION)) {
754 *dvp = (*dvp)->v_mount->mnt_vnodecovered;
757 vn_close(mvp, FREAD, cred, td);
759 vn_lock(*dvp, LK_EXCLUSIVE | LK_RETRY);
763 fileno = va.va_fileid;
765 dirbuflen = DEV_BSIZE;
766 if (dirbuflen < va.va_blocksize)
767 dirbuflen = va.va_blocksize;
768 dirbuf = (char *)malloc(dirbuflen, M_TEMP, M_WAITOK);
770 if ((*dvp)->v_type != VDIR) {
778 /* call VOP_READDIR of parent */
779 error = get_next_dirent(*dvp, &dp, dirbuf, dirbuflen, &off,
780 &cpos, &len, &eofflag, td);
784 if ((dp->d_type != DT_WHT) &&
785 (dp->d_fileno == fileno)) {
788 vn_lock(mvp, LK_EXCLUSIVE | LK_RETRY);
789 if (dirent_exists(mvp, dp->d_name, td)) {
792 vn_lock(*dvp, LK_EXCLUSIVE | LK_RETRY);
796 vn_lock(*dvp, LK_EXCLUSIVE | LK_RETRY);
804 bcopy(dp->d_name, buf + i, dp->d_namlen);
808 } while (len > 0 || !eofflag);
812 free(dirbuf, M_TEMP);
822 vn_close(mvp, FREAD, cred, td);
824 vn_lock(vp, locked | LK_RETRY);
830 * used to fill the vfs function table to get reasonable default return values.
833 vfs_stdroot (mp, flags, vpp)
843 vfs_stdstatfs (mp, sbp)
852 vfs_stdquotactl (mp, cmds, uid, arg)
863 vfs_stdsync(mp, waitfor)
867 struct vnode *vp, *mvp;
869 int error, lockreq, allerror = 0;
872 lockreq = LK_EXCLUSIVE | LK_INTERLOCK;
873 if (waitfor != MNT_WAIT)
874 lockreq |= LK_NOWAIT;
876 * Force stale buffer cache information to be flushed.
880 MNT_VNODE_FOREACH(vp, mp, mvp) {
881 /* bv_cnt is an acceptable race here. */
882 if (vp->v_bufobj.bo_dirty.bv_cnt == 0)
886 if ((error = vget(vp, lockreq, td)) != 0) {
888 if (error == ENOENT) {
889 MNT_VNODE_FOREACH_ABORT_ILOCKED(mp, mvp);
894 error = VOP_FSYNC(vp, waitfor, td);
898 /* Do not turn this into vput. td is not always curthread. */
908 vfs_stdnosync (mp, waitfor)
917 vfs_stdvget (mp, ino, flags, vpp)
928 vfs_stdfhtovp (mp, fhp, vpp)
939 struct vfsconf *vfsp;
947 struct vfsconf *vfsp;
954 vfs_stdextattrctl(mp, cmd, filename_vp, attrnamespace, attrname)
957 struct vnode *filename_vp;
959 const char *attrname;
962 if (filename_vp != NULL)
963 VOP_UNLOCK(filename_vp, 0);
968 vfs_stdsysctl(mp, op, req)
971 struct sysctl_req *req;
977 /* end of vfs default ops */