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_advlockpurge = vop_stdadvlockpurge,
103 .vop_allocate = vop_stdallocate,
104 .vop_bmap = vop_stdbmap,
105 .vop_close = VOP_NULL,
106 .vop_fsync = VOP_NULL,
107 .vop_getpages = vop_stdgetpages,
108 .vop_getwritemount = vop_stdgetwritemount,
109 .vop_inactive = VOP_NULL,
110 .vop_ioctl = VOP_ENOTTY,
111 .vop_kqfilter = vop_stdkqfilter,
112 .vop_islocked = vop_stdislocked,
113 .vop_lock1 = vop_stdlock,
114 .vop_lookup = vop_nolookup,
115 .vop_open = VOP_NULL,
116 .vop_pathconf = VOP_EINVAL,
117 .vop_poll = vop_nopoll,
118 .vop_putpages = vop_stdputpages,
119 .vop_readlink = VOP_EINVAL,
120 .vop_rename = vop_norename,
121 .vop_revoke = VOP_PANIC,
122 .vop_strategy = vop_nostrategy,
123 .vop_unlock = vop_stdunlock,
124 .vop_vptocnp = vop_stdvptocnp,
125 .vop_vptofh = vop_stdvptofh,
129 * Series of placeholder functions for various error returns for
134 vop_eopnotsupp(struct vop_generic_args *ap)
137 printf("vop_notsupp[%s]\n", ap->a_desc->vdesc_name);
144 vop_ebadf(struct vop_generic_args *ap)
151 vop_enotty(struct vop_generic_args *ap)
158 vop_einval(struct vop_generic_args *ap)
165 vop_enoent(struct vop_generic_args *ap)
172 vop_null(struct vop_generic_args *ap)
179 * Helper function to panic on some bad VOPs in some filesystems.
182 vop_panic(struct vop_generic_args *ap)
185 panic("filesystem goof: vop_panic[%s]", ap->a_desc->vdesc_name);
189 * vop_std<something> and vop_no<something> are default functions for use by
190 * filesystems that need the "default reasonable" implementation for a
191 * particular operation.
193 * The documentation for the operations they implement exists (if it exists)
194 * in the VOP_<SOMETHING>(9) manpage (all uppercase).
198 * Default vop for filesystems that do not support name lookup
202 struct vop_lookup_args /* {
204 struct vnode **a_vpp;
205 struct componentname *a_cnp;
216 * Handle unlock and reference counting for arguments of vop_rename
217 * for filesystems that do not implement rename operation.
220 vop_norename(struct vop_rename_args *ap)
230 * Strategy routine for VFS devices that have none.
232 * BIO_ERROR and B_INVAL must be cleared prior to calling any strategy
233 * routine. Typically this is done for a BIO_READ strategy call.
234 * Typically B_INVAL is assumed to already be clear prior to a write
235 * and should not be cleared manually unless you just made the buffer
236 * invalid. BIO_ERROR should be cleared either way.
240 vop_nostrategy (struct vop_strategy_args *ap)
242 printf("No strategy for buffer at %p\n", ap->a_bp);
243 vprint("vnode", ap->a_vp);
244 ap->a_bp->b_ioflags |= BIO_ERROR;
245 ap->a_bp->b_error = EOPNOTSUPP;
251 get_next_dirent(struct vnode *vp, struct dirent **dpp, char *dirbuf,
252 int dirbuflen, off_t *off, char **cpos, int *len,
253 int *eofflag, struct thread *td)
260 KASSERT(VOP_ISLOCKED(vp), ("vp %p is not locked", vp));
261 KASSERT(vp->v_type == VDIR, ("vp %p is not a directory", vp));
264 iov.iov_base = dirbuf;
265 iov.iov_len = dirbuflen;
269 uio.uio_offset = *off;
270 uio.uio_resid = dirbuflen;
271 uio.uio_segflg = UIO_SYSSPACE;
272 uio.uio_rw = UIO_READ;
278 error = mac_vnode_check_readdir(td->td_ucred, vp);
281 error = VOP_READDIR(vp, &uio, td->td_ucred, eofflag,
286 *off = uio.uio_offset;
289 *len = (dirbuflen - uio.uio_resid);
295 dp = (struct dirent *)(*cpos);
296 reclen = dp->d_reclen;
299 /* check for malformed directory.. */
300 if (reclen < DIRENT_MINSIZE)
310 * Check if a named file exists in a given directory vnode.
313 dirent_exists(struct vnode *vp, const char *dirname, struct thread *td)
316 int error, eofflag, dirbuflen, len, found;
321 KASSERT(VOP_ISLOCKED(vp), ("vp %p is not locked", vp));
322 KASSERT(vp->v_type == VDIR, ("vp %p is not a directory", vp));
326 error = VOP_GETATTR(vp, &va, td->td_ucred);
330 dirbuflen = DEV_BSIZE;
331 if (dirbuflen < va.va_blocksize)
332 dirbuflen = va.va_blocksize;
333 dirbuf = (char *)malloc(dirbuflen, M_TEMP, M_WAITOK);
338 error = get_next_dirent(vp, &dp, dirbuf, dirbuflen, &off,
339 &cpos, &len, &eofflag, td);
343 if ((dp->d_type != DT_WHT) &&
344 !strcmp(dp->d_name, dirname)) {
348 } while (len > 0 || !eofflag);
351 free(dirbuf, M_TEMP);
356 vop_stdaccess(struct vop_access_args *ap)
359 KASSERT((ap->a_accmode & ~(VEXEC | VWRITE | VREAD | VADMIN |
360 VAPPEND)) == 0, ("invalid bit in accmode"));
362 return (VOP_ACCESSX(ap->a_vp, ap->a_accmode, ap->a_cred, ap->a_td));
366 vop_stdaccessx(struct vop_accessx_args *ap)
369 accmode_t accmode = ap->a_accmode;
371 error = vfs_unixify_accmode(&accmode);
378 return (VOP_ACCESS(ap->a_vp, accmode, ap->a_cred, ap->a_td));
382 * Advisory record locking support
385 vop_stdadvlock(struct vop_advlock_args *ap)
393 cred = curthread->td_ucred;
394 vn_lock(vp, LK_SHARED | LK_RETRY);
395 error = VOP_GETATTR(vp, &vattr, cred);
400 return (lf_advlock(ap, &(vp->v_lockf), vattr.va_size));
404 vop_stdadvlockasync(struct vop_advlockasync_args *ap)
412 cred = curthread->td_ucred;
413 vn_lock(vp, LK_SHARED | LK_RETRY);
414 error = VOP_GETATTR(vp, &vattr, cred);
419 return (lf_advlockasync(ap, &(vp->v_lockf), vattr.va_size));
423 vop_stdadvlockpurge(struct vop_advlockpurge_args *ap)
428 lf_purgelocks(vp, &vp->v_lockf);
435 * Standard implementation of POSIX pathconf, to get information about limits
437 * Override per filesystem for the case where the filesystem has smaller
442 struct vop_pathconf_args /* {
449 switch (ap->a_name) {
451 *ap->a_retval = NAME_MAX;
454 *ap->a_retval = PATH_MAX;
457 *ap->a_retval = LINK_MAX;
460 *ap->a_retval = MAX_CANON;
463 *ap->a_retval = MAX_INPUT;
466 *ap->a_retval = PIPE_BUF;
468 case _PC_CHOWN_RESTRICTED:
472 *ap->a_retval = _POSIX_VDISABLE;
481 * Standard lock, unlock and islocked functions.
485 struct vop_lock1_args /* {
492 struct vnode *vp = ap->a_vp;
494 return (_lockmgr_args(vp->v_vnlock, ap->a_flags, VI_MTX(vp),
495 LK_WMESG_DEFAULT, LK_PRIO_DEFAULT, LK_TIMO_DEFAULT, ap->a_file,
502 struct vop_unlock_args /* {
507 struct vnode *vp = ap->a_vp;
509 return (lockmgr(vp->v_vnlock, ap->a_flags | LK_RELEASE, VI_MTX(vp)));
515 struct vop_islocked_args /* {
520 return (lockstatus(ap->a_vp->v_vnlock));
524 * Return true for select/poll.
528 struct vop_poll_args /* {
531 struct ucred *a_cred;
536 return (poll_no_poll(ap->a_events));
540 * Implement poll for local filesystems that support it.
544 struct vop_poll_args /* {
547 struct ucred *a_cred;
551 if (ap->a_events & ~POLLSTANDARD)
552 return (vn_pollrecord(ap->a_vp, ap->a_td, ap->a_events));
553 return (ap->a_events & (POLLIN | POLLOUT | POLLRDNORM | POLLWRNORM));
557 * Return our mount point, as we will take charge of the writes.
560 vop_stdgetwritemount(ap)
561 struct vop_getwritemount_args /* {
563 struct mount **a_mpp;
569 * XXX Since this is called unlocked we may be recycled while
570 * attempting to ref the mount. If this is the case or mountpoint
571 * will be set to NULL. We only have to prevent this call from
572 * returning with a ref to an incorrect mountpoint. It is not
573 * harmful to return with a ref to our previous mountpoint.
575 mp = ap->a_vp->v_mount;
578 if (mp != ap->a_vp->v_mount) {
587 /* XXX Needs good comment and VOP_BMAP(9) manpage */
590 struct vop_bmap_args /* {
593 struct bufobj **a_bop;
600 if (ap->a_bop != NULL)
601 *ap->a_bop = &ap->a_vp->v_bufobj;
602 if (ap->a_bnp != NULL)
603 *ap->a_bnp = ap->a_bn * btodb(ap->a_vp->v_mount->mnt_stat.f_iosize);
604 if (ap->a_runp != NULL)
606 if (ap->a_runb != NULL)
613 struct vop_fsync_args /* {
615 struct ucred *a_cred;
620 struct vnode *vp = ap->a_vp;
625 int maxretry = 1000; /* large, arbitrarily chosen */
631 * MARK/SCAN initialization to avoid infinite loops.
633 TAILQ_FOREACH(bp, &bo->bo_dirty.bv_hd, b_bobufs) {
634 bp->b_vflags &= ~BV_SCANNED;
639 * Flush all dirty buffers associated with a vnode.
642 TAILQ_FOREACH_SAFE(bp, &bo->bo_dirty.bv_hd, b_bobufs, nbp) {
643 if ((bp->b_vflags & BV_SCANNED) != 0)
645 bp->b_vflags |= BV_SCANNED;
646 if (BUF_LOCK(bp, LK_EXCLUSIVE | LK_NOWAIT, NULL))
649 KASSERT(bp->b_bufobj == bo,
650 ("bp %p wrong b_bufobj %p should be %p",
651 bp, bp->b_bufobj, bo));
652 if ((bp->b_flags & B_DELWRI) == 0)
653 panic("fsync: not dirty");
654 if ((vp->v_object != NULL) && (bp->b_flags & B_CLUSTEROK)) {
665 * If synchronous the caller expects us to completely resolve all
666 * dirty buffers in the system. Wait for in-progress I/O to
667 * complete (which could include background bitmap writes), then
668 * retry if dirty blocks still exist.
670 if (ap->a_waitfor == MNT_WAIT) {
671 bufobj_wwait(bo, 0, 0);
672 if (bo->bo_dirty.bv_cnt > 0) {
674 * If we are unable to write any of these buffers
675 * then we fail now rather than trying endlessly
678 TAILQ_FOREACH(bp, &bo->bo_dirty.bv_hd, b_bobufs)
679 if ((error = bp->b_error) == 0)
681 if (error == 0 && --maxretry >= 0)
688 vprint("fsync: giving up on dirty", vp);
693 /* XXX Needs good comment and more info in the manpage (VOP_GETPAGES(9)). */
696 struct vop_getpages_args /* {
701 vm_ooffset_t a_offset;
705 return vnode_pager_generic_getpages(ap->a_vp, ap->a_m,
706 ap->a_count, ap->a_reqpage);
710 vop_stdkqfilter(struct vop_kqfilter_args *ap)
712 return vfs_kqfilter(ap);
715 /* XXX Needs good comment and more info in the manpage (VOP_PUTPAGES(9)). */
718 struct vop_putpages_args /* {
724 vm_ooffset_t a_offset;
728 return vnode_pager_generic_putpages(ap->a_vp, ap->a_m, ap->a_count,
729 ap->a_sync, ap->a_rtvals);
733 vop_stdvptofh(struct vop_vptofh_args *ap)
739 vop_stdvptocnp(struct vop_vptocnp_args *ap)
741 struct vnode *vp = ap->a_vp;
742 struct vnode **dvp = ap->a_vpp;
743 struct ucred *cred = ap->a_cred;
744 char *buf = ap->a_buf;
745 int *buflen = ap->a_buflen;
747 int i, error, eofflag, dirbuflen, flags, locked, len, covered;
761 if (vp->v_type != VDIR)
764 error = VOP_GETATTR(vp, &va, cred);
769 locked = VOP_ISLOCKED(vp);
771 NDINIT_ATVP(&nd, LOOKUP, FOLLOW | LOCKLEAF, UIO_SYSSPACE,
774 error = vn_open_cred(&nd, &flags, 0, VN_OPEN_NOAUDIT, cred, NULL);
776 vn_lock(vp, locked | LK_RETRY);
779 NDFREE(&nd, NDF_ONLY_PNBUF);
781 mvp = *dvp = nd.ni_vp;
783 if (vp->v_mount != (*dvp)->v_mount &&
784 ((*dvp)->v_vflag & VV_ROOT) &&
785 ((*dvp)->v_mount->mnt_flag & MNT_UNION)) {
786 *dvp = (*dvp)->v_mount->mnt_vnodecovered;
789 vn_close(mvp, FREAD, cred, td);
791 vn_lock(*dvp, LK_EXCLUSIVE | LK_RETRY);
795 fileno = va.va_fileid;
797 dirbuflen = DEV_BSIZE;
798 if (dirbuflen < va.va_blocksize)
799 dirbuflen = va.va_blocksize;
800 dirbuf = (char *)malloc(dirbuflen, M_TEMP, M_WAITOK);
802 if ((*dvp)->v_type != VDIR) {
810 /* call VOP_READDIR of parent */
811 error = get_next_dirent(*dvp, &dp, dirbuf, dirbuflen, &off,
812 &cpos, &len, &eofflag, td);
816 if ((dp->d_type != DT_WHT) &&
817 (dp->d_fileno == fileno)) {
820 vn_lock(mvp, LK_EXCLUSIVE | LK_RETRY);
821 if (dirent_exists(mvp, dp->d_name, td)) {
824 vn_lock(*dvp, LK_EXCLUSIVE | LK_RETRY);
828 vn_lock(*dvp, LK_EXCLUSIVE | LK_RETRY);
836 bcopy(dp->d_name, buf + i, dp->d_namlen);
840 } while (len > 0 || !eofflag);
844 free(dirbuf, M_TEMP);
854 vn_close(mvp, FREAD, cred, td);
856 vn_lock(vp, locked | LK_RETRY);
861 vop_stdallocate(struct vop_allocate_args *ap)
867 struct vattr vattr, *vap;
869 off_t fsize, len, cur, offset;
882 offset = *ap->a_offset;
884 error = VOP_GETATTR(vp, vap, td->td_ucred);
887 fsize = vap->va_size;
888 iosize = vap->va_blocksize;
890 iosize = BLKDEV_IOSIZE;
891 if (iosize > MAXPHYS)
893 buf = malloc(iosize, M_TEMP, M_WAITOK);
897 * Check if the filesystem sets f_maxfilesize; if not use
898 * VOP_SETATTR to perform the check.
900 error = VFS_STATFS(vp->v_mount, &sfs, td);
903 if (sfs.f_maxfilesize) {
904 if (offset > sfs.f_maxfilesize || len > sfs.f_maxfilesize ||
905 offset + len > sfs.f_maxfilesize) {
911 if (offset + len > vap->va_size) {
913 * Test offset + len against the filesystem's maxfilesize.
916 vap->va_size = offset + len;
917 error = VOP_SETATTR(vp, vap, td->td_ucred);
921 vap->va_size = fsize;
922 error = VOP_SETATTR(vp, vap, td->td_ucred);
929 * Read and write back anything below the nominal file
930 * size. There's currently no way outside the filesystem
931 * to know whether this area is sparse or not.
934 if ((offset % iosize) != 0)
935 cur -= (offset % iosize);
938 if (offset < fsize) {
941 auio.uio_iov = &aiov;
943 auio.uio_offset = offset;
944 auio.uio_resid = cur;
945 auio.uio_segflg = UIO_SYSSPACE;
946 auio.uio_rw = UIO_READ;
948 error = VOP_READ(vp, &auio, 0, td->td_ucred);
951 if (auio.uio_resid > 0) {
952 bzero(buf + cur - auio.uio_resid,
961 auio.uio_iov = &aiov;
963 auio.uio_offset = offset;
964 auio.uio_resid = cur;
965 auio.uio_segflg = UIO_SYSSPACE;
966 auio.uio_rw = UIO_WRITE;
969 error = VOP_WRITE(vp, &auio, 0, td->td_ucred);
983 *ap->a_offset = offset;
989 vop_stdadvise(struct vop_advise_args *ap)
993 int error, vfslocked;
996 switch (ap->a_advice) {
997 case POSIX_FADV_WILLNEED:
999 * Do nothing for now. Filesystems should provide a
1000 * custom method which starts an asynchronous read of
1001 * the requested region.
1005 case POSIX_FADV_DONTNEED:
1007 * Flush any open FS buffers and then remove pages
1008 * from the backing VM object. Using vinvalbuf() here
1009 * is a bit heavy-handed as it flushes all buffers for
1010 * the given vnode, not just the buffers covering the
1014 vfslocked = VFS_LOCK_GIANT(vp->v_mount);
1015 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
1016 if (vp->v_iflag & VI_DOOMED) {
1018 VFS_UNLOCK_GIANT(vfslocked);
1021 vinvalbuf(vp, V_CLEANONLY, 0, 0);
1022 if (vp->v_object != NULL) {
1023 start = trunc_page(ap->a_start);
1024 end = round_page(ap->a_end);
1025 VM_OBJECT_LOCK(vp->v_object);
1026 vm_object_page_cache(vp->v_object, OFF_TO_IDX(start),
1028 VM_OBJECT_UNLOCK(vp->v_object);
1031 VFS_UNLOCK_GIANT(vfslocked);
1042 * used to fill the vfs function table to get reasonable default return values.
1045 vfs_stdroot (mp, flags, vpp)
1051 return (EOPNOTSUPP);
1055 vfs_stdstatfs (mp, sbp)
1060 return (EOPNOTSUPP);
1064 vfs_stdquotactl (mp, cmds, uid, arg)
1071 return (EOPNOTSUPP);
1075 vfs_stdsync(mp, waitfor)
1079 struct vnode *vp, *mvp;
1081 int error, lockreq, allerror = 0;
1084 lockreq = LK_EXCLUSIVE | LK_INTERLOCK;
1085 if (waitfor != MNT_WAIT)
1086 lockreq |= LK_NOWAIT;
1088 * Force stale buffer cache information to be flushed.
1092 MNT_VNODE_FOREACH(vp, mp, mvp) {
1093 /* bv_cnt is an acceptable race here. */
1094 if (vp->v_bufobj.bo_dirty.bv_cnt == 0)
1098 if ((error = vget(vp, lockreq, td)) != 0) {
1100 if (error == ENOENT) {
1101 MNT_VNODE_FOREACH_ABORT_ILOCKED(mp, mvp);
1106 error = VOP_FSYNC(vp, waitfor, td);
1117 vfs_stdnosync (mp, waitfor)
1126 vfs_stdvget (mp, ino, flags, vpp)
1133 return (EOPNOTSUPP);
1137 vfs_stdfhtovp (mp, fhp, flags, vpp)
1144 return (EOPNOTSUPP);
1149 struct vfsconf *vfsp;
1156 vfs_stduninit (vfsp)
1157 struct vfsconf *vfsp;
1164 vfs_stdextattrctl(mp, cmd, filename_vp, attrnamespace, attrname)
1167 struct vnode *filename_vp;
1169 const char *attrname;
1172 if (filename_vp != NULL)
1173 VOP_UNLOCK(filename_vp, 0);
1174 return (EOPNOTSUPP);
1178 vfs_stdsysctl(mp, op, req)
1181 struct sysctl_req *req;
1184 return (EOPNOTSUPP);
1187 /* end of vfs default ops */