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_allocate = vop_stdallocate,
102 .vop_bmap = vop_stdbmap,
103 .vop_close = VOP_NULL,
104 .vop_fsync = VOP_NULL,
105 .vop_getpages = vop_stdgetpages,
106 .vop_getwritemount = vop_stdgetwritemount,
107 .vop_inactive = VOP_NULL,
108 .vop_ioctl = VOP_ENOTTY,
109 .vop_kqfilter = vop_stdkqfilter,
110 .vop_islocked = vop_stdislocked,
111 .vop_lock1 = vop_stdlock,
112 .vop_lookup = vop_nolookup,
113 .vop_open = VOP_NULL,
114 .vop_pathconf = VOP_EINVAL,
115 .vop_poll = vop_nopoll,
116 .vop_putpages = vop_stdputpages,
117 .vop_readlink = VOP_EINVAL,
118 .vop_rename = vop_norename,
119 .vop_revoke = VOP_PANIC,
120 .vop_strategy = vop_nostrategy,
121 .vop_unlock = vop_stdunlock,
122 .vop_vptocnp = vop_stdvptocnp,
123 .vop_vptofh = vop_stdvptofh,
127 * Series of placeholder functions for various error returns for
132 vop_eopnotsupp(struct vop_generic_args *ap)
135 printf("vop_notsupp[%s]\n", ap->a_desc->vdesc_name);
142 vop_ebadf(struct vop_generic_args *ap)
149 vop_enotty(struct vop_generic_args *ap)
156 vop_einval(struct vop_generic_args *ap)
163 vop_enoent(struct vop_generic_args *ap)
170 vop_null(struct vop_generic_args *ap)
177 * Helper function to panic on some bad VOPs in some filesystems.
180 vop_panic(struct vop_generic_args *ap)
183 panic("filesystem goof: vop_panic[%s]", ap->a_desc->vdesc_name);
187 * vop_std<something> and vop_no<something> are default functions for use by
188 * filesystems that need the "default reasonable" implementation for a
189 * particular operation.
191 * The documentation for the operations they implement exists (if it exists)
192 * in the VOP_<SOMETHING>(9) manpage (all uppercase).
196 * Default vop for filesystems that do not support name lookup
200 struct vop_lookup_args /* {
202 struct vnode **a_vpp;
203 struct componentname *a_cnp;
214 * Handle unlock and reference counting for arguments of vop_rename
215 * for filesystems that do not implement rename operation.
218 vop_norename(struct vop_rename_args *ap)
228 * Strategy routine for VFS devices that have none.
230 * BIO_ERROR and B_INVAL must be cleared prior to calling any strategy
231 * routine. Typically this is done for a BIO_READ strategy call.
232 * Typically B_INVAL is assumed to already be clear prior to a write
233 * and should not be cleared manually unless you just made the buffer
234 * invalid. BIO_ERROR should be cleared either way.
238 vop_nostrategy (struct vop_strategy_args *ap)
240 printf("No strategy for buffer at %p\n", ap->a_bp);
241 vprint("vnode", ap->a_vp);
242 ap->a_bp->b_ioflags |= BIO_ERROR;
243 ap->a_bp->b_error = EOPNOTSUPP;
249 get_next_dirent(struct vnode *vp, struct dirent **dpp, char *dirbuf,
250 int dirbuflen, off_t *off, char **cpos, int *len,
251 int *eofflag, struct thread *td)
258 KASSERT(VOP_ISLOCKED(vp), ("vp %p is not locked", vp));
259 KASSERT(vp->v_type == VDIR, ("vp %p is not a directory", vp));
262 iov.iov_base = dirbuf;
263 iov.iov_len = dirbuflen;
267 uio.uio_offset = *off;
268 uio.uio_resid = dirbuflen;
269 uio.uio_segflg = UIO_SYSSPACE;
270 uio.uio_rw = UIO_READ;
276 error = mac_vnode_check_readdir(td->td_ucred, vp);
279 error = VOP_READDIR(vp, &uio, td->td_ucred, eofflag,
284 *off = uio.uio_offset;
287 *len = (dirbuflen - uio.uio_resid);
293 dp = (struct dirent *)(*cpos);
294 reclen = dp->d_reclen;
297 /* check for malformed directory.. */
298 if (reclen < DIRENT_MINSIZE)
308 * Check if a named file exists in a given directory vnode.
311 dirent_exists(struct vnode *vp, const char *dirname, struct thread *td)
314 int error, eofflag, dirbuflen, len, found;
319 KASSERT(VOP_ISLOCKED(vp), ("vp %p is not locked", vp));
320 KASSERT(vp->v_type == VDIR, ("vp %p is not a directory", vp));
324 error = VOP_GETATTR(vp, &va, td->td_ucred);
328 dirbuflen = DEV_BSIZE;
329 if (dirbuflen < va.va_blocksize)
330 dirbuflen = va.va_blocksize;
331 dirbuf = (char *)malloc(dirbuflen, M_TEMP, M_WAITOK);
336 error = get_next_dirent(vp, &dp, dirbuf, dirbuflen, &off,
337 &cpos, &len, &eofflag, td);
341 if ((dp->d_type != DT_WHT) &&
342 !strcmp(dp->d_name, dirname)) {
346 } while (len > 0 || !eofflag);
349 free(dirbuf, M_TEMP);
354 vop_stdaccess(struct vop_access_args *ap)
357 KASSERT((ap->a_accmode & ~(VEXEC | VWRITE | VREAD | VADMIN |
358 VAPPEND)) == 0, ("invalid bit in accmode"));
360 return (VOP_ACCESSX(ap->a_vp, ap->a_accmode, ap->a_cred, ap->a_td));
364 vop_stdaccessx(struct vop_accessx_args *ap)
367 accmode_t accmode = ap->a_accmode;
369 error = vfs_unixify_accmode(&accmode);
376 return (VOP_ACCESS(ap->a_vp, accmode, ap->a_cred, ap->a_td));
380 * Advisory record locking support
383 vop_stdadvlock(struct vop_advlock_args *ap)
391 cred = curthread->td_ucred;
392 vn_lock(vp, LK_SHARED | LK_RETRY);
393 error = VOP_GETATTR(vp, &vattr, cred);
398 return (lf_advlock(ap, &(vp->v_lockf), vattr.va_size));
402 vop_stdadvlockasync(struct vop_advlockasync_args *ap)
410 cred = curthread->td_ucred;
411 vn_lock(vp, LK_SHARED | LK_RETRY);
412 error = VOP_GETATTR(vp, &vattr, cred);
417 return (lf_advlockasync(ap, &(vp->v_lockf), vattr.va_size));
423 * Standard implementation of POSIX pathconf, to get information about limits
425 * Override per filesystem for the case where the filesystem has smaller
430 struct vop_pathconf_args /* {
437 switch (ap->a_name) {
439 *ap->a_retval = NAME_MAX;
442 *ap->a_retval = PATH_MAX;
445 *ap->a_retval = LINK_MAX;
448 *ap->a_retval = MAX_CANON;
451 *ap->a_retval = MAX_INPUT;
454 *ap->a_retval = PIPE_BUF;
456 case _PC_CHOWN_RESTRICTED:
460 *ap->a_retval = _POSIX_VDISABLE;
469 * Standard lock, unlock and islocked functions.
473 struct vop_lock1_args /* {
480 struct vnode *vp = ap->a_vp;
482 return (_lockmgr_args(vp->v_vnlock, ap->a_flags, VI_MTX(vp),
483 LK_WMESG_DEFAULT, LK_PRIO_DEFAULT, LK_TIMO_DEFAULT, ap->a_file,
490 struct vop_unlock_args /* {
495 struct vnode *vp = ap->a_vp;
497 return (lockmgr(vp->v_vnlock, ap->a_flags | LK_RELEASE, VI_MTX(vp)));
503 struct vop_islocked_args /* {
508 return (lockstatus(ap->a_vp->v_vnlock));
512 * Return true for select/poll.
516 struct vop_poll_args /* {
519 struct ucred *a_cred;
524 return (poll_no_poll(ap->a_events));
528 * Implement poll for local filesystems that support it.
532 struct vop_poll_args /* {
535 struct ucred *a_cred;
539 if (ap->a_events & ~POLLSTANDARD)
540 return (vn_pollrecord(ap->a_vp, ap->a_td, ap->a_events));
541 return (ap->a_events & (POLLIN | POLLOUT | POLLRDNORM | POLLWRNORM));
545 * Return our mount point, as we will take charge of the writes.
548 vop_stdgetwritemount(ap)
549 struct vop_getwritemount_args /* {
551 struct mount **a_mpp;
557 * XXX Since this is called unlocked we may be recycled while
558 * attempting to ref the mount. If this is the case or mountpoint
559 * will be set to NULL. We only have to prevent this call from
560 * returning with a ref to an incorrect mountpoint. It is not
561 * harmful to return with a ref to our previous mountpoint.
563 mp = ap->a_vp->v_mount;
566 if (mp != ap->a_vp->v_mount) {
575 /* XXX Needs good comment and VOP_BMAP(9) manpage */
578 struct vop_bmap_args /* {
581 struct bufobj **a_bop;
588 if (ap->a_bop != NULL)
589 *ap->a_bop = &ap->a_vp->v_bufobj;
590 if (ap->a_bnp != NULL)
591 *ap->a_bnp = ap->a_bn * btodb(ap->a_vp->v_mount->mnt_stat.f_iosize);
592 if (ap->a_runp != NULL)
594 if (ap->a_runb != NULL)
601 struct vop_fsync_args /* {
603 struct ucred *a_cred;
608 struct vnode *vp = ap->a_vp;
613 int maxretry = 1000; /* large, arbitrarily chosen */
619 * MARK/SCAN initialization to avoid infinite loops.
621 TAILQ_FOREACH(bp, &bo->bo_dirty.bv_hd, b_bobufs) {
622 bp->b_vflags &= ~BV_SCANNED;
627 * Flush all dirty buffers associated with a vnode.
630 TAILQ_FOREACH_SAFE(bp, &bo->bo_dirty.bv_hd, b_bobufs, nbp) {
631 if ((bp->b_vflags & BV_SCANNED) != 0)
633 bp->b_vflags |= BV_SCANNED;
634 if (BUF_LOCK(bp, LK_EXCLUSIVE | LK_NOWAIT, NULL))
637 KASSERT(bp->b_bufobj == bo,
638 ("bp %p wrong b_bufobj %p should be %p",
639 bp, bp->b_bufobj, bo));
640 if ((bp->b_flags & B_DELWRI) == 0)
641 panic("fsync: not dirty");
642 if ((vp->v_object != NULL) && (bp->b_flags & B_CLUSTEROK)) {
653 * If synchronous the caller expects us to completely resolve all
654 * dirty buffers in the system. Wait for in-progress I/O to
655 * complete (which could include background bitmap writes), then
656 * retry if dirty blocks still exist.
658 if (ap->a_waitfor == MNT_WAIT) {
659 bufobj_wwait(bo, 0, 0);
660 if (bo->bo_dirty.bv_cnt > 0) {
662 * If we are unable to write any of these buffers
663 * then we fail now rather than trying endlessly
666 TAILQ_FOREACH(bp, &bo->bo_dirty.bv_hd, b_bobufs)
667 if ((error = bp->b_error) == 0)
669 if (error == 0 && --maxretry >= 0)
676 vprint("fsync: giving up on dirty", vp);
681 /* XXX Needs good comment and more info in the manpage (VOP_GETPAGES(9)). */
684 struct vop_getpages_args /* {
689 vm_ooffset_t a_offset;
693 return vnode_pager_generic_getpages(ap->a_vp, ap->a_m,
694 ap->a_count, ap->a_reqpage);
698 vop_stdkqfilter(struct vop_kqfilter_args *ap)
700 return vfs_kqfilter(ap);
703 /* XXX Needs good comment and more info in the manpage (VOP_PUTPAGES(9)). */
706 struct vop_putpages_args /* {
712 vm_ooffset_t a_offset;
716 return vnode_pager_generic_putpages(ap->a_vp, ap->a_m, ap->a_count,
717 ap->a_sync, ap->a_rtvals);
721 vop_stdvptofh(struct vop_vptofh_args *ap)
727 vop_stdvptocnp(struct vop_vptocnp_args *ap)
729 struct vnode *vp = ap->a_vp;
730 struct vnode **dvp = ap->a_vpp;
731 struct ucred *cred = ap->a_cred;
732 char *buf = ap->a_buf;
733 int *buflen = ap->a_buflen;
735 int i, error, eofflag, dirbuflen, flags, locked, len, covered;
749 if (vp->v_type != VDIR)
752 error = VOP_GETATTR(vp, &va, cred);
757 locked = VOP_ISLOCKED(vp);
759 NDINIT_ATVP(&nd, LOOKUP, FOLLOW | LOCKLEAF, UIO_SYSSPACE,
762 error = vn_open_cred(&nd, &flags, 0, VN_OPEN_NOAUDIT, cred, NULL);
764 vn_lock(vp, locked | LK_RETRY);
767 NDFREE(&nd, NDF_ONLY_PNBUF);
769 mvp = *dvp = nd.ni_vp;
771 if (vp->v_mount != (*dvp)->v_mount &&
772 ((*dvp)->v_vflag & VV_ROOT) &&
773 ((*dvp)->v_mount->mnt_flag & MNT_UNION)) {
774 *dvp = (*dvp)->v_mount->mnt_vnodecovered;
777 vn_close(mvp, FREAD, cred, td);
779 vn_lock(*dvp, LK_EXCLUSIVE | LK_RETRY);
783 fileno = va.va_fileid;
785 dirbuflen = DEV_BSIZE;
786 if (dirbuflen < va.va_blocksize)
787 dirbuflen = va.va_blocksize;
788 dirbuf = (char *)malloc(dirbuflen, M_TEMP, M_WAITOK);
790 if ((*dvp)->v_type != VDIR) {
798 /* call VOP_READDIR of parent */
799 error = get_next_dirent(*dvp, &dp, dirbuf, dirbuflen, &off,
800 &cpos, &len, &eofflag, td);
804 if ((dp->d_type != DT_WHT) &&
805 (dp->d_fileno == fileno)) {
808 vn_lock(mvp, LK_EXCLUSIVE | LK_RETRY);
809 if (dirent_exists(mvp, dp->d_name, td)) {
812 vn_lock(*dvp, LK_EXCLUSIVE | LK_RETRY);
816 vn_lock(*dvp, LK_EXCLUSIVE | LK_RETRY);
824 bcopy(dp->d_name, buf + i, dp->d_namlen);
828 } while (len > 0 || !eofflag);
832 free(dirbuf, M_TEMP);
842 vn_close(mvp, FREAD, cred, td);
844 vn_lock(vp, locked | LK_RETRY);
849 vop_stdallocate(struct vop_allocate_args *ap)
855 struct vattr vattr, *vap;
857 off_t fsize, len, cur, offset;
870 offset = *ap->a_offset;
872 error = VOP_GETATTR(vp, vap, td->td_ucred);
875 fsize = vap->va_size;
876 iosize = vap->va_blocksize;
878 iosize = BLKDEV_IOSIZE;
879 if (iosize > MAXPHYS)
881 buf = malloc(iosize, M_TEMP, M_WAITOK);
885 * Check if the filesystem sets f_maxfilesize; if not use
886 * VOP_SETATTR to perform the check.
888 error = VFS_STATFS(vp->v_mount, &sfs, td);
891 if (sfs.f_maxfilesize) {
892 if (offset > sfs.f_maxfilesize || len > sfs.f_maxfilesize ||
893 offset + len > sfs.f_maxfilesize) {
899 if (offset + len > vap->va_size) {
901 * Test offset + len against the filesystem's maxfilesize.
904 vap->va_size = offset + len;
905 error = VOP_SETATTR(vp, vap, td->td_ucred);
909 vap->va_size = fsize;
910 error = VOP_SETATTR(vp, vap, td->td_ucred);
917 * Read and write back anything below the nominal file
918 * size. There's currently no way outside the filesystem
919 * to know whether this area is sparse or not.
922 if ((offset % iosize) != 0)
923 cur -= (offset % iosize);
926 if (offset < fsize) {
929 auio.uio_iov = &aiov;
931 auio.uio_offset = offset;
932 auio.uio_resid = cur;
933 auio.uio_segflg = UIO_SYSSPACE;
934 auio.uio_rw = UIO_READ;
936 error = VOP_READ(vp, &auio, 0, td->td_ucred);
939 if (auio.uio_resid > 0) {
940 bzero(buf + cur - auio.uio_resid,
949 auio.uio_iov = &aiov;
951 auio.uio_offset = offset;
952 auio.uio_resid = cur;
953 auio.uio_segflg = UIO_SYSSPACE;
954 auio.uio_rw = UIO_WRITE;
957 error = VOP_WRITE(vp, &auio, 0, td->td_ucred);
971 *ap->a_offset = offset;
978 * used to fill the vfs function table to get reasonable default return values.
981 vfs_stdroot (mp, flags, vpp)
991 vfs_stdstatfs (mp, sbp)
1000 vfs_stdquotactl (mp, cmds, uid, arg)
1007 return (EOPNOTSUPP);
1011 vfs_stdsync(mp, waitfor)
1015 struct vnode *vp, *mvp;
1017 int error, lockreq, allerror = 0;
1020 lockreq = LK_EXCLUSIVE | LK_INTERLOCK;
1021 if (waitfor != MNT_WAIT)
1022 lockreq |= LK_NOWAIT;
1024 * Force stale buffer cache information to be flushed.
1028 MNT_VNODE_FOREACH(vp, mp, mvp) {
1029 /* bv_cnt is an acceptable race here. */
1030 if (vp->v_bufobj.bo_dirty.bv_cnt == 0)
1034 if ((error = vget(vp, lockreq, td)) != 0) {
1036 if (error == ENOENT) {
1037 MNT_VNODE_FOREACH_ABORT_ILOCKED(mp, mvp);
1042 error = VOP_FSYNC(vp, waitfor, td);
1046 /* Do not turn this into vput. td is not always curthread. */
1056 vfs_stdnosync (mp, waitfor)
1065 vfs_stdvget (mp, ino, flags, vpp)
1072 return (EOPNOTSUPP);
1076 vfs_stdfhtovp (mp, fhp, vpp)
1082 return (EOPNOTSUPP);
1087 struct vfsconf *vfsp;
1094 vfs_stduninit (vfsp)
1095 struct vfsconf *vfsp;
1102 vfs_stdextattrctl(mp, cmd, filename_vp, attrnamespace, attrname)
1105 struct vnode *filename_vp;
1107 const char *attrname;
1110 if (filename_vp != NULL)
1111 VOP_UNLOCK(filename_vp, 0);
1112 return (EOPNOTSUPP);
1116 vfs_stdsysctl(mp, op, req)
1119 struct sysctl_req *req;
1122 return (EOPNOTSUPP);
1125 /* end of vfs default ops */