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_advise = vop_stdadvise,
100 .vop_advlock = vop_stdadvlock,
101 .vop_advlockasync = vop_stdadvlockasync,
102 .vop_allocate = vop_stdallocate,
103 .vop_bmap = vop_stdbmap,
104 .vop_close = VOP_NULL,
105 .vop_fsync = VOP_NULL,
106 .vop_getpages = vop_stdgetpages,
107 .vop_getwritemount = vop_stdgetwritemount,
108 .vop_inactive = VOP_NULL,
109 .vop_ioctl = VOP_ENOTTY,
110 .vop_kqfilter = vop_stdkqfilter,
111 .vop_islocked = vop_stdislocked,
112 .vop_lock1 = vop_stdlock,
113 .vop_lookup = vop_nolookup,
114 .vop_open = VOP_NULL,
115 .vop_pathconf = VOP_EINVAL,
116 .vop_poll = vop_nopoll,
117 .vop_putpages = vop_stdputpages,
118 .vop_readlink = VOP_EINVAL,
119 .vop_rename = vop_norename,
120 .vop_revoke = VOP_PANIC,
121 .vop_strategy = vop_nostrategy,
122 .vop_unlock = vop_stdunlock,
123 .vop_vptocnp = vop_stdvptocnp,
124 .vop_vptofh = vop_stdvptofh,
128 * Series of placeholder functions for various error returns for
133 vop_eopnotsupp(struct vop_generic_args *ap)
136 printf("vop_notsupp[%s]\n", ap->a_desc->vdesc_name);
143 vop_ebadf(struct vop_generic_args *ap)
150 vop_enotty(struct vop_generic_args *ap)
157 vop_einval(struct vop_generic_args *ap)
164 vop_enoent(struct vop_generic_args *ap)
171 vop_null(struct vop_generic_args *ap)
178 * Helper function to panic on some bad VOPs in some filesystems.
181 vop_panic(struct vop_generic_args *ap)
184 panic("filesystem goof: vop_panic[%s]", ap->a_desc->vdesc_name);
188 * vop_std<something> and vop_no<something> are default functions for use by
189 * filesystems that need the "default reasonable" implementation for a
190 * particular operation.
192 * The documentation for the operations they implement exists (if it exists)
193 * in the VOP_<SOMETHING>(9) manpage (all uppercase).
197 * Default vop for filesystems that do not support name lookup
201 struct vop_lookup_args /* {
203 struct vnode **a_vpp;
204 struct componentname *a_cnp;
215 * Handle unlock and reference counting for arguments of vop_rename
216 * for filesystems that do not implement rename operation.
219 vop_norename(struct vop_rename_args *ap)
229 * Strategy routine for VFS devices that have none.
231 * BIO_ERROR and B_INVAL must be cleared prior to calling any strategy
232 * routine. Typically this is done for a BIO_READ strategy call.
233 * Typically B_INVAL is assumed to already be clear prior to a write
234 * and should not be cleared manually unless you just made the buffer
235 * invalid. BIO_ERROR should be cleared either way.
239 vop_nostrategy (struct vop_strategy_args *ap)
241 printf("No strategy for buffer at %p\n", ap->a_bp);
242 vprint("vnode", ap->a_vp);
243 ap->a_bp->b_ioflags |= BIO_ERROR;
244 ap->a_bp->b_error = EOPNOTSUPP;
250 get_next_dirent(struct vnode *vp, struct dirent **dpp, char *dirbuf,
251 int dirbuflen, off_t *off, char **cpos, int *len,
252 int *eofflag, struct thread *td)
259 KASSERT(VOP_ISLOCKED(vp), ("vp %p is not locked", vp));
260 KASSERT(vp->v_type == VDIR, ("vp %p is not a directory", vp));
263 iov.iov_base = dirbuf;
264 iov.iov_len = dirbuflen;
268 uio.uio_offset = *off;
269 uio.uio_resid = dirbuflen;
270 uio.uio_segflg = UIO_SYSSPACE;
271 uio.uio_rw = UIO_READ;
277 error = mac_vnode_check_readdir(td->td_ucred, vp);
280 error = VOP_READDIR(vp, &uio, td->td_ucred, eofflag,
285 *off = uio.uio_offset;
288 *len = (dirbuflen - uio.uio_resid);
294 dp = (struct dirent *)(*cpos);
295 reclen = dp->d_reclen;
298 /* check for malformed directory.. */
299 if (reclen < DIRENT_MINSIZE)
309 * Check if a named file exists in a given directory vnode.
312 dirent_exists(struct vnode *vp, const char *dirname, struct thread *td)
315 int error, eofflag, dirbuflen, len, found;
320 KASSERT(VOP_ISLOCKED(vp), ("vp %p is not locked", vp));
321 KASSERT(vp->v_type == VDIR, ("vp %p is not a directory", vp));
325 error = VOP_GETATTR(vp, &va, td->td_ucred);
329 dirbuflen = DEV_BSIZE;
330 if (dirbuflen < va.va_blocksize)
331 dirbuflen = va.va_blocksize;
332 dirbuf = (char *)malloc(dirbuflen, M_TEMP, M_WAITOK);
337 error = get_next_dirent(vp, &dp, dirbuf, dirbuflen, &off,
338 &cpos, &len, &eofflag, td);
342 if ((dp->d_type != DT_WHT) &&
343 !strcmp(dp->d_name, dirname)) {
347 } while (len > 0 || !eofflag);
350 free(dirbuf, M_TEMP);
355 vop_stdaccess(struct vop_access_args *ap)
358 KASSERT((ap->a_accmode & ~(VEXEC | VWRITE | VREAD | VADMIN |
359 VAPPEND)) == 0, ("invalid bit in accmode"));
361 return (VOP_ACCESSX(ap->a_vp, ap->a_accmode, ap->a_cred, ap->a_td));
365 vop_stdaccessx(struct vop_accessx_args *ap)
368 accmode_t accmode = ap->a_accmode;
370 error = vfs_unixify_accmode(&accmode);
377 return (VOP_ACCESS(ap->a_vp, accmode, ap->a_cred, ap->a_td));
381 * Advisory record locking support
384 vop_stdadvlock(struct vop_advlock_args *ap)
392 cred = curthread->td_ucred;
393 vn_lock(vp, LK_SHARED | LK_RETRY);
394 error = VOP_GETATTR(vp, &vattr, cred);
399 return (lf_advlock(ap, &(vp->v_lockf), vattr.va_size));
403 vop_stdadvlockasync(struct vop_advlockasync_args *ap)
411 cred = curthread->td_ucred;
412 vn_lock(vp, LK_SHARED | LK_RETRY);
413 error = VOP_GETATTR(vp, &vattr, cred);
418 return (lf_advlockasync(ap, &(vp->v_lockf), vattr.va_size));
424 * Standard implementation of POSIX pathconf, to get information about limits
426 * Override per filesystem for the case where the filesystem has smaller
431 struct vop_pathconf_args /* {
438 switch (ap->a_name) {
440 *ap->a_retval = NAME_MAX;
443 *ap->a_retval = PATH_MAX;
446 *ap->a_retval = LINK_MAX;
449 *ap->a_retval = MAX_CANON;
452 *ap->a_retval = MAX_INPUT;
455 *ap->a_retval = PIPE_BUF;
457 case _PC_CHOWN_RESTRICTED:
461 *ap->a_retval = _POSIX_VDISABLE;
470 * Standard lock, unlock and islocked functions.
474 struct vop_lock1_args /* {
481 struct vnode *vp = ap->a_vp;
483 return (_lockmgr_args(vp->v_vnlock, ap->a_flags, VI_MTX(vp),
484 LK_WMESG_DEFAULT, LK_PRIO_DEFAULT, LK_TIMO_DEFAULT, ap->a_file,
491 struct vop_unlock_args /* {
496 struct vnode *vp = ap->a_vp;
498 return (lockmgr(vp->v_vnlock, ap->a_flags | LK_RELEASE, VI_MTX(vp)));
504 struct vop_islocked_args /* {
509 return (lockstatus(ap->a_vp->v_vnlock));
513 * Return true for select/poll.
517 struct vop_poll_args /* {
520 struct ucred *a_cred;
525 return (poll_no_poll(ap->a_events));
529 * Implement poll for local filesystems that support it.
533 struct vop_poll_args /* {
536 struct ucred *a_cred;
540 if (ap->a_events & ~POLLSTANDARD)
541 return (vn_pollrecord(ap->a_vp, ap->a_td, ap->a_events));
542 return (ap->a_events & (POLLIN | POLLOUT | POLLRDNORM | POLLWRNORM));
546 * Return our mount point, as we will take charge of the writes.
549 vop_stdgetwritemount(ap)
550 struct vop_getwritemount_args /* {
552 struct mount **a_mpp;
558 * XXX Since this is called unlocked we may be recycled while
559 * attempting to ref the mount. If this is the case or mountpoint
560 * will be set to NULL. We only have to prevent this call from
561 * returning with a ref to an incorrect mountpoint. It is not
562 * harmful to return with a ref to our previous mountpoint.
564 mp = ap->a_vp->v_mount;
567 if (mp != ap->a_vp->v_mount) {
576 /* XXX Needs good comment and VOP_BMAP(9) manpage */
579 struct vop_bmap_args /* {
582 struct bufobj **a_bop;
589 if (ap->a_bop != NULL)
590 *ap->a_bop = &ap->a_vp->v_bufobj;
591 if (ap->a_bnp != NULL)
592 *ap->a_bnp = ap->a_bn * btodb(ap->a_vp->v_mount->mnt_stat.f_iosize);
593 if (ap->a_runp != NULL)
595 if (ap->a_runb != NULL)
602 struct vop_fsync_args /* {
604 struct ucred *a_cred;
609 struct vnode *vp = ap->a_vp;
614 int maxretry = 1000; /* large, arbitrarily chosen */
620 * MARK/SCAN initialization to avoid infinite loops.
622 TAILQ_FOREACH(bp, &bo->bo_dirty.bv_hd, b_bobufs) {
623 bp->b_vflags &= ~BV_SCANNED;
628 * Flush all dirty buffers associated with a vnode.
631 TAILQ_FOREACH_SAFE(bp, &bo->bo_dirty.bv_hd, b_bobufs, nbp) {
632 if ((bp->b_vflags & BV_SCANNED) != 0)
634 bp->b_vflags |= BV_SCANNED;
635 if (BUF_LOCK(bp, LK_EXCLUSIVE | LK_NOWAIT, NULL))
638 KASSERT(bp->b_bufobj == bo,
639 ("bp %p wrong b_bufobj %p should be %p",
640 bp, bp->b_bufobj, bo));
641 if ((bp->b_flags & B_DELWRI) == 0)
642 panic("fsync: not dirty");
643 if ((vp->v_object != NULL) && (bp->b_flags & B_CLUSTEROK)) {
654 * If synchronous the caller expects us to completely resolve all
655 * dirty buffers in the system. Wait for in-progress I/O to
656 * complete (which could include background bitmap writes), then
657 * retry if dirty blocks still exist.
659 if (ap->a_waitfor == MNT_WAIT) {
660 bufobj_wwait(bo, 0, 0);
661 if (bo->bo_dirty.bv_cnt > 0) {
663 * If we are unable to write any of these buffers
664 * then we fail now rather than trying endlessly
667 TAILQ_FOREACH(bp, &bo->bo_dirty.bv_hd, b_bobufs)
668 if ((error = bp->b_error) == 0)
670 if (error == 0 && --maxretry >= 0)
677 vprint("fsync: giving up on dirty", vp);
682 /* XXX Needs good comment and more info in the manpage (VOP_GETPAGES(9)). */
685 struct vop_getpages_args /* {
690 vm_ooffset_t a_offset;
694 return vnode_pager_generic_getpages(ap->a_vp, ap->a_m,
695 ap->a_count, ap->a_reqpage);
699 vop_stdkqfilter(struct vop_kqfilter_args *ap)
701 return vfs_kqfilter(ap);
704 /* XXX Needs good comment and more info in the manpage (VOP_PUTPAGES(9)). */
707 struct vop_putpages_args /* {
713 vm_ooffset_t a_offset;
717 return vnode_pager_generic_putpages(ap->a_vp, ap->a_m, ap->a_count,
718 ap->a_sync, ap->a_rtvals);
722 vop_stdvptofh(struct vop_vptofh_args *ap)
728 vop_stdvptocnp(struct vop_vptocnp_args *ap)
730 struct vnode *vp = ap->a_vp;
731 struct vnode **dvp = ap->a_vpp;
732 struct ucred *cred = ap->a_cred;
733 char *buf = ap->a_buf;
734 int *buflen = ap->a_buflen;
736 int i, error, eofflag, dirbuflen, flags, locked, len, covered;
750 if (vp->v_type != VDIR)
753 error = VOP_GETATTR(vp, &va, cred);
758 locked = VOP_ISLOCKED(vp);
760 NDINIT_ATVP(&nd, LOOKUP, FOLLOW | LOCKLEAF, UIO_SYSSPACE,
763 error = vn_open_cred(&nd, &flags, 0, VN_OPEN_NOAUDIT, cred, NULL);
765 vn_lock(vp, locked | LK_RETRY);
768 NDFREE(&nd, NDF_ONLY_PNBUF);
770 mvp = *dvp = nd.ni_vp;
772 if (vp->v_mount != (*dvp)->v_mount &&
773 ((*dvp)->v_vflag & VV_ROOT) &&
774 ((*dvp)->v_mount->mnt_flag & MNT_UNION)) {
775 *dvp = (*dvp)->v_mount->mnt_vnodecovered;
778 vn_close(mvp, FREAD, cred, td);
780 vn_lock(*dvp, LK_EXCLUSIVE | LK_RETRY);
784 fileno = va.va_fileid;
786 dirbuflen = DEV_BSIZE;
787 if (dirbuflen < va.va_blocksize)
788 dirbuflen = va.va_blocksize;
789 dirbuf = (char *)malloc(dirbuflen, M_TEMP, M_WAITOK);
791 if ((*dvp)->v_type != VDIR) {
799 /* call VOP_READDIR of parent */
800 error = get_next_dirent(*dvp, &dp, dirbuf, dirbuflen, &off,
801 &cpos, &len, &eofflag, td);
805 if ((dp->d_type != DT_WHT) &&
806 (dp->d_fileno == fileno)) {
809 vn_lock(mvp, LK_EXCLUSIVE | LK_RETRY);
810 if (dirent_exists(mvp, dp->d_name, td)) {
813 vn_lock(*dvp, LK_EXCLUSIVE | LK_RETRY);
817 vn_lock(*dvp, LK_EXCLUSIVE | LK_RETRY);
825 bcopy(dp->d_name, buf + i, dp->d_namlen);
829 } while (len > 0 || !eofflag);
833 free(dirbuf, M_TEMP);
843 vn_close(mvp, FREAD, cred, td);
845 vn_lock(vp, locked | LK_RETRY);
850 vop_stdallocate(struct vop_allocate_args *ap)
856 struct vattr vattr, *vap;
858 off_t fsize, len, cur, offset;
871 offset = *ap->a_offset;
873 error = VOP_GETATTR(vp, vap, td->td_ucred);
876 fsize = vap->va_size;
877 iosize = vap->va_blocksize;
879 iosize = BLKDEV_IOSIZE;
880 if (iosize > MAXPHYS)
882 buf = malloc(iosize, M_TEMP, M_WAITOK);
886 * Check if the filesystem sets f_maxfilesize; if not use
887 * VOP_SETATTR to perform the check.
889 error = VFS_STATFS(vp->v_mount, &sfs, td);
892 if (sfs.f_maxfilesize) {
893 if (offset > sfs.f_maxfilesize || len > sfs.f_maxfilesize ||
894 offset + len > sfs.f_maxfilesize) {
900 if (offset + len > vap->va_size) {
902 * Test offset + len against the filesystem's maxfilesize.
905 vap->va_size = offset + len;
906 error = VOP_SETATTR(vp, vap, td->td_ucred);
910 vap->va_size = fsize;
911 error = VOP_SETATTR(vp, vap, td->td_ucred);
918 * Read and write back anything below the nominal file
919 * size. There's currently no way outside the filesystem
920 * to know whether this area is sparse or not.
923 if ((offset % iosize) != 0)
924 cur -= (offset % iosize);
927 if (offset < fsize) {
930 auio.uio_iov = &aiov;
932 auio.uio_offset = offset;
933 auio.uio_resid = cur;
934 auio.uio_segflg = UIO_SYSSPACE;
935 auio.uio_rw = UIO_READ;
937 error = VOP_READ(vp, &auio, 0, td->td_ucred);
940 if (auio.uio_resid > 0) {
941 bzero(buf + cur - auio.uio_resid,
950 auio.uio_iov = &aiov;
952 auio.uio_offset = offset;
953 auio.uio_resid = cur;
954 auio.uio_segflg = UIO_SYSSPACE;
955 auio.uio_rw = UIO_WRITE;
958 error = VOP_WRITE(vp, &auio, 0, td->td_ucred);
972 *ap->a_offset = offset;
978 vop_stdadvise(struct vop_advise_args *ap)
982 int error, vfslocked;
985 switch (ap->a_advice) {
986 case POSIX_FADV_WILLNEED:
988 * Do nothing for now. Filesystems should provide a
989 * custom method which starts an asynchronous read of
990 * the requested region.
994 case POSIX_FADV_DONTNEED:
996 * Flush any open FS buffers and then remove pages
997 * from the backing VM object. Using vinvalbuf() here
998 * is a bit heavy-handed as it flushes all buffers for
999 * the given vnode, not just the buffers covering the
1003 vfslocked = VFS_LOCK_GIANT(vp->v_mount);
1004 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
1005 if (vp->v_iflag & VI_DOOMED) {
1007 VFS_UNLOCK_GIANT(vfslocked);
1010 vinvalbuf(vp, V_CLEANONLY, 0, 0);
1011 if (vp->v_object != NULL) {
1012 start = trunc_page(ap->a_start);
1013 end = round_page(ap->a_end);
1014 VM_OBJECT_LOCK(vp->v_object);
1015 vm_object_page_cache(vp->v_object, OFF_TO_IDX(start),
1017 VM_OBJECT_UNLOCK(vp->v_object);
1020 VFS_UNLOCK_GIANT(vfslocked);
1031 * used to fill the vfs function table to get reasonable default return values.
1034 vfs_stdroot (mp, flags, vpp)
1040 return (EOPNOTSUPP);
1044 vfs_stdstatfs (mp, sbp)
1049 return (EOPNOTSUPP);
1053 vfs_stdquotactl (mp, cmds, uid, arg)
1060 return (EOPNOTSUPP);
1064 vfs_stdsync(mp, waitfor)
1068 struct vnode *vp, *mvp;
1070 int error, lockreq, allerror = 0;
1073 lockreq = LK_EXCLUSIVE | LK_INTERLOCK;
1074 if (waitfor != MNT_WAIT)
1075 lockreq |= LK_NOWAIT;
1077 * Force stale buffer cache information to be flushed.
1081 MNT_VNODE_FOREACH(vp, mp, mvp) {
1082 /* bv_cnt is an acceptable race here. */
1083 if (vp->v_bufobj.bo_dirty.bv_cnt == 0)
1087 if ((error = vget(vp, lockreq, td)) != 0) {
1089 if (error == ENOENT) {
1090 MNT_VNODE_FOREACH_ABORT_ILOCKED(mp, mvp);
1095 error = VOP_FSYNC(vp, waitfor, td);
1099 /* Do not turn this into vput. td is not always curthread. */
1109 vfs_stdnosync (mp, waitfor)
1118 vfs_stdvget (mp, ino, flags, vpp)
1125 return (EOPNOTSUPP);
1129 vfs_stdfhtovp (mp, fhp, vpp)
1135 return (EOPNOTSUPP);
1140 struct vfsconf *vfsp;
1147 vfs_stduninit (vfsp)
1148 struct vfsconf *vfsp;
1155 vfs_stdextattrctl(mp, cmd, filename_vp, attrnamespace, attrname)
1158 struct vnode *filename_vp;
1160 const char *attrname;
1163 if (filename_vp != NULL)
1164 VOP_UNLOCK(filename_vp, 0);
1165 return (EOPNOTSUPP);
1169 vfs_stdsysctl(mp, op, req)
1172 struct sysctl_req *req;
1175 return (EOPNOTSUPP);
1178 /* end of vfs default ops */