2 * SPDX-License-Identifier: BSD-3-Clause
4 * Copyright (c) 1989, 1993
5 * The Regents of the University of California. All rights reserved.
7 * This code is derived from software contributed
8 * to Berkeley by John Heidemann of the UCLA Ficus project.
10 * Source: * @(#)i405_init.c 2.10 92/04/27 UCLA Ficus project
12 * Redistribution and use in source and binary forms, with or without
13 * modification, are permitted provided that the following conditions
15 * 1. Redistributions of source code must retain the above copyright
16 * notice, this list of conditions and the following disclaimer.
17 * 2. Redistributions in binary form must reproduce the above copyright
18 * notice, this list of conditions and the following disclaimer in the
19 * documentation and/or other materials provided with the distribution.
20 * 3. Neither the name of the University nor the names of its contributors
21 * may be used to endorse or promote products derived from this software
22 * without specific prior written permission.
24 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
25 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
26 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
27 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
28 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
29 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
30 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
31 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
32 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
33 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
37 #include <sys/cdefs.h>
38 __FBSDID("$FreeBSD$");
40 #include <sys/param.h>
41 #include <sys/systm.h>
45 #include <sys/event.h>
46 #include <sys/filio.h>
47 #include <sys/kernel.h>
48 #include <sys/limits.h>
50 #include <sys/lockf.h>
51 #include <sys/malloc.h>
52 #include <sys/mount.h>
53 #include <sys/namei.h>
54 #include <sys/rwlock.h>
55 #include <sys/fcntl.h>
56 #include <sys/unistd.h>
57 #include <sys/vnode.h>
58 #include <sys/dirent.h>
61 #include <security/audit/audit.h>
64 #include <security/mac/mac_framework.h>
67 #include <vm/vm_object.h>
68 #include <vm/vm_extern.h>
70 #include <vm/vm_map.h>
71 #include <vm/vm_page.h>
72 #include <vm/vm_pager.h>
73 #include <vm/vnode_pager.h>
75 static int vop_nolookup(struct vop_lookup_args *);
76 static int vop_norename(struct vop_rename_args *);
77 static int vop_nostrategy(struct vop_strategy_args *);
78 static int get_next_dirent(struct vnode *vp, struct dirent **dpp,
79 char *dirbuf, int dirbuflen, off_t *off,
80 char **cpos, int *len, int *eofflag,
82 static int dirent_exists(struct vnode *vp, const char *dirname,
85 #define DIRENT_MINSIZE (sizeof(struct dirent) - (MAXNAMLEN+1) + 4)
87 static int vop_stdis_text(struct vop_is_text_args *ap);
88 static int vop_stdunset_text(struct vop_unset_text_args *ap);
89 static int vop_stdadd_writecount(struct vop_add_writecount_args *ap);
90 static int vop_stdcopy_file_range(struct vop_copy_file_range_args *ap);
91 static int vop_stdfdatasync(struct vop_fdatasync_args *ap);
92 static int vop_stdgetpages_async(struct vop_getpages_async_args *ap);
93 static int vop_stdread_pgcache(struct vop_read_pgcache_args *ap);
94 static int vop_stdstat(struct vop_stat_args *ap);
95 static int vop_stdvput_pair(struct vop_vput_pair_args *ap);
96 static int vop_stddeallocate(struct vop_deallocate_args *ap);
99 * This vnode table stores what we want to do if the filesystem doesn't
100 * implement a particular VOP.
102 * If there is no specific entry here, we will return EOPNOTSUPP.
104 * Note that every filesystem has to implement either vop_access
105 * or vop_accessx; failing to do so will result in immediate crash
106 * due to stack overflow, as vop_stdaccess() calls vop_stdaccessx(),
107 * which calls vop_stdaccess() etc.
110 struct vop_vector default_vnodeops = {
112 .vop_bypass = VOP_EOPNOTSUPP,
114 .vop_access = vop_stdaccess,
115 .vop_accessx = vop_stdaccessx,
116 .vop_advise = vop_stdadvise,
117 .vop_advlock = vop_stdadvlock,
118 .vop_advlockasync = vop_stdadvlockasync,
119 .vop_advlockpurge = vop_stdadvlockpurge,
120 .vop_allocate = vop_stdallocate,
121 .vop_deallocate = vop_stddeallocate,
122 .vop_bmap = vop_stdbmap,
123 .vop_close = VOP_NULL,
124 .vop_fsync = VOP_NULL,
125 .vop_stat = vop_stdstat,
126 .vop_fdatasync = vop_stdfdatasync,
127 .vop_getpages = vop_stdgetpages,
128 .vop_getpages_async = vop_stdgetpages_async,
129 .vop_getwritemount = vop_stdgetwritemount,
130 .vop_inactive = VOP_NULL,
131 .vop_need_inactive = vop_stdneed_inactive,
132 .vop_ioctl = vop_stdioctl,
133 .vop_kqfilter = vop_stdkqfilter,
134 .vop_islocked = vop_stdislocked,
135 .vop_lock1 = vop_stdlock,
136 .vop_lookup = vop_nolookup,
137 .vop_open = VOP_NULL,
138 .vop_pathconf = VOP_EINVAL,
139 .vop_poll = vop_nopoll,
140 .vop_putpages = vop_stdputpages,
141 .vop_readlink = VOP_EINVAL,
142 .vop_read_pgcache = vop_stdread_pgcache,
143 .vop_rename = vop_norename,
144 .vop_revoke = VOP_PANIC,
145 .vop_strategy = vop_nostrategy,
146 .vop_unlock = vop_stdunlock,
147 .vop_vptocnp = vop_stdvptocnp,
148 .vop_vptofh = vop_stdvptofh,
149 .vop_unp_bind = vop_stdunp_bind,
150 .vop_unp_connect = vop_stdunp_connect,
151 .vop_unp_detach = vop_stdunp_detach,
152 .vop_is_text = vop_stdis_text,
153 .vop_set_text = vop_stdset_text,
154 .vop_unset_text = vop_stdunset_text,
155 .vop_add_writecount = vop_stdadd_writecount,
156 .vop_copy_file_range = vop_stdcopy_file_range,
157 .vop_vput_pair = vop_stdvput_pair,
159 VFS_VOP_VECTOR_REGISTER(default_vnodeops);
162 * Series of placeholder functions for various error returns for
167 vop_eopnotsupp(struct vop_generic_args *ap)
170 printf("vop_notsupp[%s]\n", ap->a_desc->vdesc_name);
177 vop_ebadf(struct vop_generic_args *ap)
184 vop_enotty(struct vop_generic_args *ap)
191 vop_einval(struct vop_generic_args *ap)
198 vop_enoent(struct vop_generic_args *ap)
205 vop_eagain(struct vop_generic_args *ap)
212 vop_null(struct vop_generic_args *ap)
219 * Helper function to panic on some bad VOPs in some filesystems.
222 vop_panic(struct vop_generic_args *ap)
225 panic("filesystem goof: vop_panic[%s]", ap->a_desc->vdesc_name);
229 * vop_std<something> and vop_no<something> are default functions for use by
230 * filesystems that need the "default reasonable" implementation for a
231 * particular operation.
233 * The documentation for the operations they implement exists (if it exists)
234 * in the VOP_<SOMETHING>(9) manpage (all uppercase).
238 * Default vop for filesystems that do not support name lookup
242 struct vop_lookup_args /* {
244 struct vnode **a_vpp;
245 struct componentname *a_cnp;
256 * Handle unlock and reference counting for arguments of vop_rename
257 * for filesystems that do not implement rename operation.
260 vop_norename(struct vop_rename_args *ap)
270 * Strategy routine for VFS devices that have none.
272 * BIO_ERROR and B_INVAL must be cleared prior to calling any strategy
273 * routine. Typically this is done for a BIO_READ strategy call.
274 * Typically B_INVAL is assumed to already be clear prior to a write
275 * and should not be cleared manually unless you just made the buffer
276 * invalid. BIO_ERROR should be cleared either way.
280 vop_nostrategy (struct vop_strategy_args *ap)
282 printf("No strategy for buffer at %p\n", ap->a_bp);
283 vn_printf(ap->a_vp, "vnode ");
284 ap->a_bp->b_ioflags |= BIO_ERROR;
285 ap->a_bp->b_error = EOPNOTSUPP;
291 get_next_dirent(struct vnode *vp, struct dirent **dpp, char *dirbuf,
292 int dirbuflen, off_t *off, char **cpos, int *len,
293 int *eofflag, struct thread *td)
300 KASSERT(VOP_ISLOCKED(vp), ("vp %p is not locked", vp));
301 KASSERT(vp->v_type == VDIR, ("vp %p is not a directory", vp));
304 iov.iov_base = dirbuf;
305 iov.iov_len = dirbuflen;
309 uio.uio_offset = *off;
310 uio.uio_resid = dirbuflen;
311 uio.uio_segflg = UIO_SYSSPACE;
312 uio.uio_rw = UIO_READ;
318 error = mac_vnode_check_readdir(td->td_ucred, vp);
321 error = VOP_READDIR(vp, &uio, td->td_ucred, eofflag,
326 *off = uio.uio_offset;
329 *len = (dirbuflen - uio.uio_resid);
335 dp = (struct dirent *)(*cpos);
336 reclen = dp->d_reclen;
339 /* check for malformed directory.. */
340 if (reclen < DIRENT_MINSIZE)
350 * Check if a named file exists in a given directory vnode.
353 dirent_exists(struct vnode *vp, const char *dirname, struct thread *td)
356 int error, eofflag, dirbuflen, len, found;
361 KASSERT(VOP_ISLOCKED(vp), ("vp %p is not locked", vp));
362 KASSERT(vp->v_type == VDIR, ("vp %p is not a directory", vp));
366 error = VOP_GETATTR(vp, &va, td->td_ucred);
370 dirbuflen = DEV_BSIZE;
371 if (dirbuflen < va.va_blocksize)
372 dirbuflen = va.va_blocksize;
373 dirbuf = (char *)malloc(dirbuflen, M_TEMP, M_WAITOK);
378 error = get_next_dirent(vp, &dp, dirbuf, dirbuflen, &off,
379 &cpos, &len, &eofflag, td);
383 if (dp->d_type != DT_WHT && dp->d_fileno != 0 &&
384 strcmp(dp->d_name, dirname) == 0) {
388 } while (len > 0 || !eofflag);
391 free(dirbuf, M_TEMP);
396 vop_stdaccess(struct vop_access_args *ap)
399 KASSERT((ap->a_accmode & ~(VEXEC | VWRITE | VREAD | VADMIN |
400 VAPPEND)) == 0, ("invalid bit in accmode"));
402 return (VOP_ACCESSX(ap->a_vp, ap->a_accmode, ap->a_cred, ap->a_td));
406 vop_stdaccessx(struct vop_accessx_args *ap)
409 accmode_t accmode = ap->a_accmode;
411 error = vfs_unixify_accmode(&accmode);
418 return (VOP_ACCESS(ap->a_vp, accmode, ap->a_cred, ap->a_td));
422 * Advisory record locking support
425 vop_stdadvlock(struct vop_advlock_args *ap)
435 * Provide atomicity of open(O_CREAT | O_EXCL | O_EXLOCK) for
436 * local filesystems. See vn_open_cred() for reciprocal part.
439 if (mp != NULL && (mp->mnt_flag & MNT_LOCAL) != 0 &&
440 ap->a_op == F_SETLK && (ap->a_flags & F_FIRSTOPEN) == 0) {
442 while ((vp->v_iflag & VI_FOPENING) != 0)
443 msleep(vp, VI_MTX(vp), PLOCK, "lockfo", 0);
447 if (ap->a_fl->l_whence == SEEK_END) {
449 * The NFSv4 server must avoid doing a vn_lock() here, since it
450 * can deadlock the nfsd threads, due to a LOR. Fortunately
451 * the NFSv4 server always uses SEEK_SET and this code is
452 * only required for the SEEK_END case.
454 vn_lock(vp, LK_SHARED | LK_RETRY);
455 error = VOP_GETATTR(vp, &vattr, curthread->td_ucred);
462 return (lf_advlock(ap, &(vp->v_lockf), vattr.va_size));
466 vop_stdadvlockasync(struct vop_advlockasync_args *ap)
473 if (ap->a_fl->l_whence == SEEK_END) {
474 /* The size argument is only needed for SEEK_END. */
475 vn_lock(vp, LK_SHARED | LK_RETRY);
476 error = VOP_GETATTR(vp, &vattr, curthread->td_ucred);
483 return (lf_advlockasync(ap, &(vp->v_lockf), vattr.va_size));
487 vop_stdadvlockpurge(struct vop_advlockpurge_args *ap)
492 lf_purgelocks(vp, &vp->v_lockf);
499 * Standard implementation of POSIX pathconf, to get information about limits
501 * Override per filesystem for the case where the filesystem has smaller
506 struct vop_pathconf_args /* {
513 switch (ap->a_name) {
515 *ap->a_retval = _POSIX_ASYNCHRONOUS_IO;
518 *ap->a_retval = PATH_MAX;
520 case _PC_ACL_EXTENDED:
522 case _PC_CAP_PRESENT:
523 case _PC_DEALLOC_PRESENT:
524 case _PC_INF_PRESENT:
525 case _PC_MAC_PRESENT:
535 * Standard lock, unlock and islocked functions.
539 struct vop_lock1_args /* {
546 struct vnode *vp = ap->a_vp;
550 return (lockmgr_lock_flags(vp->v_vnlock, ap->a_flags,
551 &ilk->lock_object, ap->a_file, ap->a_line));
557 struct vop_unlock_args /* {
561 struct vnode *vp = ap->a_vp;
563 return (lockmgr_unlock(vp->v_vnlock));
569 struct vop_islocked_args /* {
574 return (lockstatus(ap->a_vp->v_vnlock));
578 * Variants of the above set.
581 * - shared locking disablement is not supported
582 * - v_vnlock pointer is not honored
586 struct vop_lock1_args /* {
593 struct vnode *vp = ap->a_vp;
594 int flags = ap->a_flags;
597 MPASS(vp->v_vnlock == &vp->v_lock);
599 if (__predict_false((flags & ~(LK_TYPE_MASK | LK_NODDLKTREAT | LK_RETRY)) != 0))
602 switch (flags & LK_TYPE_MASK) {
604 return (lockmgr_slock(&vp->v_lock, flags, ap->a_file, ap->a_line));
606 return (lockmgr_xlock(&vp->v_lock, flags, ap->a_file, ap->a_line));
610 return (lockmgr_lock_flags(&vp->v_lock, flags,
611 &ilk->lock_object, ap->a_file, ap->a_line));
616 struct vop_unlock_args /* {
620 struct vnode *vp = ap->a_vp;
622 MPASS(vp->v_vnlock == &vp->v_lock);
624 return (lockmgr_unlock(&vp->v_lock));
629 struct vop_islocked_args /* {
633 struct vnode *vp = ap->a_vp;
635 MPASS(vp->v_vnlock == &vp->v_lock);
637 return (lockstatus(&vp->v_lock));
641 * Return true for select/poll.
645 struct vop_poll_args /* {
648 struct ucred *a_cred;
653 if (ap->a_events & ~POLLSTANDARD)
655 return (ap->a_events & (POLLIN | POLLOUT | POLLRDNORM | POLLWRNORM));
659 * Implement poll for local filesystems that support it.
663 struct vop_poll_args /* {
666 struct ucred *a_cred;
670 if (ap->a_events & ~POLLSTANDARD)
671 return (vn_pollrecord(ap->a_vp, ap->a_td, ap->a_events));
672 return (ap->a_events & (POLLIN | POLLOUT | POLLRDNORM | POLLWRNORM));
676 * Return our mount point, as we will take charge of the writes.
679 vop_stdgetwritemount(ap)
680 struct vop_getwritemount_args /* {
682 struct mount **a_mpp;
689 * Note that having a reference does not prevent forced unmount from
690 * setting ->v_mount to NULL after the lock gets released. This is of
691 * no consequence for typical consumers (most notably vn_start_write)
692 * since in this case the vnode is VIRF_DOOMED. Unmount might have
693 * progressed far enough that its completion is only delayed by the
694 * reference obtained here. The consumer only needs to concern itself
698 mp = vfs_ref_from_vp(vp);
704 * If the file system doesn't implement VOP_BMAP, then return sensible defaults:
705 * - Return the vnode's bufobj instead of any underlying device's bufobj
706 * - Calculate the physical block number as if there were equal size
707 * consecutive blocks, but
708 * - Report no contiguous runs of blocks.
712 struct vop_bmap_args /* {
715 struct bufobj **a_bop;
722 if (ap->a_bop != NULL)
723 *ap->a_bop = &ap->a_vp->v_bufobj;
724 if (ap->a_bnp != NULL)
725 *ap->a_bnp = ap->a_bn * btodb(ap->a_vp->v_mount->mnt_stat.f_iosize);
726 if (ap->a_runp != NULL)
728 if (ap->a_runb != NULL)
735 struct vop_fsync_args /* {
742 return (vn_fsync_buf(ap->a_vp, ap->a_waitfor));
746 vop_stdfdatasync(struct vop_fdatasync_args *ap)
749 return (VOP_FSYNC(ap->a_vp, MNT_WAIT, ap->a_td));
753 vop_stdfdatasync_buf(struct vop_fdatasync_args *ap)
756 return (vn_fsync_buf(ap->a_vp, MNT_WAIT));
759 /* XXX Needs good comment and more info in the manpage (VOP_GETPAGES(9)). */
762 struct vop_getpages_args /* {
771 return vnode_pager_generic_getpages(ap->a_vp, ap->a_m,
772 ap->a_count, ap->a_rbehind, ap->a_rahead, NULL, NULL);
776 vop_stdgetpages_async(struct vop_getpages_async_args *ap)
780 error = VOP_GETPAGES(ap->a_vp, ap->a_m, ap->a_count, ap->a_rbehind,
782 if (ap->a_iodone != NULL)
783 ap->a_iodone(ap->a_arg, ap->a_m, ap->a_count, error);
788 vop_stdkqfilter(struct vop_kqfilter_args *ap)
790 return vfs_kqfilter(ap);
793 /* XXX Needs good comment and more info in the manpage (VOP_PUTPAGES(9)). */
796 struct vop_putpages_args /* {
805 return vnode_pager_generic_putpages(ap->a_vp, ap->a_m, ap->a_count,
806 ap->a_sync, ap->a_rtvals);
810 vop_stdvptofh(struct vop_vptofh_args *ap)
816 vop_stdvptocnp(struct vop_vptocnp_args *ap)
818 struct vnode *vp = ap->a_vp;
819 struct vnode **dvp = ap->a_vpp;
821 char *buf = ap->a_buf;
822 size_t *buflen = ap->a_buflen;
824 int i, error, eofflag, dirbuflen, flags, locked, len, covered;
839 if (vp->v_type != VDIR)
842 error = VOP_GETATTR(vp, &va, cred);
847 locked = VOP_ISLOCKED(vp);
849 NDINIT_ATVP(&nd, LOOKUP, FOLLOW | LOCKSHARED | LOCKLEAF, UIO_SYSSPACE,
852 error = vn_open_cred(&nd, &flags, 0, VN_OPEN_NOAUDIT, cred, NULL);
854 vn_lock(vp, locked | LK_RETRY);
857 NDFREE(&nd, NDF_ONLY_PNBUF);
859 mvp = *dvp = nd.ni_vp;
861 if (vp->v_mount != (*dvp)->v_mount &&
862 ((*dvp)->v_vflag & VV_ROOT) &&
863 ((*dvp)->v_mount->mnt_flag & MNT_UNION)) {
864 *dvp = (*dvp)->v_mount->mnt_vnodecovered;
867 vn_close(mvp, FREAD, cred, td);
869 vn_lock(*dvp, LK_SHARED | LK_RETRY);
873 fileno = va.va_fileid;
875 dirbuflen = DEV_BSIZE;
876 if (dirbuflen < va.va_blocksize)
877 dirbuflen = va.va_blocksize;
878 dirbuf = (char *)malloc(dirbuflen, M_TEMP, M_WAITOK);
880 if ((*dvp)->v_type != VDIR) {
888 /* call VOP_READDIR of parent */
889 error = get_next_dirent(*dvp, &dp, dirbuf, dirbuflen, &off,
890 &cpos, &len, &eofflag, td);
894 if ((dp->d_type != DT_WHT) &&
895 (dp->d_fileno == fileno)) {
898 vn_lock(mvp, LK_SHARED | LK_RETRY);
899 if (dirent_exists(mvp, dp->d_name, td)) {
902 vn_lock(*dvp, LK_SHARED | LK_RETRY);
906 vn_lock(*dvp, LK_SHARED | LK_RETRY);
914 if (dp->d_namlen == 1 && dp->d_name[0] == '.') {
917 bcopy(dp->d_name, buf + i, dp->d_namlen);
922 } while (len > 0 || !eofflag);
926 free(dirbuf, M_TEMP);
936 vn_close(mvp, FREAD, cred, td);
938 vn_lock(vp, locked | LK_RETRY);
943 vop_stdallocate(struct vop_allocate_args *ap)
947 off_t maxfilesize = 0;
950 struct vattr vattr, *vap;
952 off_t fsize, len, cur, offset;
965 offset = *ap->a_offset;
967 error = VOP_GETATTR(vp, vap, td->td_ucred);
970 fsize = vap->va_size;
971 iosize = vap->va_blocksize;
973 iosize = BLKDEV_IOSIZE;
974 if (iosize > maxphys)
976 buf = malloc(iosize, M_TEMP, M_WAITOK);
980 * Check if the filesystem sets f_maxfilesize; if not use
981 * VOP_SETATTR to perform the check.
983 sfs = malloc(sizeof(struct statfs), M_STATFS, M_WAITOK);
984 error = VFS_STATFS(vp->v_mount, sfs, td);
986 maxfilesize = sfs->f_maxfilesize;
991 if (offset > maxfilesize || len > maxfilesize ||
992 offset + len > maxfilesize) {
998 if (offset + len > vap->va_size) {
1000 * Test offset + len against the filesystem's maxfilesize.
1003 vap->va_size = offset + len;
1004 error = VOP_SETATTR(vp, vap, td->td_ucred);
1008 vap->va_size = fsize;
1009 error = VOP_SETATTR(vp, vap, td->td_ucred);
1016 * Read and write back anything below the nominal file
1017 * size. There's currently no way outside the filesystem
1018 * to know whether this area is sparse or not.
1021 if ((offset % iosize) != 0)
1022 cur -= (offset % iosize);
1025 if (offset < fsize) {
1026 aiov.iov_base = buf;
1028 auio.uio_iov = &aiov;
1029 auio.uio_iovcnt = 1;
1030 auio.uio_offset = offset;
1031 auio.uio_resid = cur;
1032 auio.uio_segflg = UIO_SYSSPACE;
1033 auio.uio_rw = UIO_READ;
1035 error = VOP_READ(vp, &auio, 0, td->td_ucred);
1038 if (auio.uio_resid > 0) {
1039 bzero(buf + cur - auio.uio_resid,
1046 aiov.iov_base = buf;
1048 auio.uio_iov = &aiov;
1049 auio.uio_iovcnt = 1;
1050 auio.uio_offset = offset;
1051 auio.uio_resid = cur;
1052 auio.uio_segflg = UIO_SYSSPACE;
1053 auio.uio_rw = UIO_WRITE;
1056 error = VOP_WRITE(vp, &auio, 0, td->td_ucred);
1070 *ap->a_offset = offset;
1076 vp_zerofill(struct vnode *vp, struct vattr *vap, off_t *offsetp, off_t *lenp,
1086 iosize = vap->va_blocksize;
1092 iosize = BLKDEV_IOSIZE;
1093 /* If va_blocksize is 512 bytes, iosize will be 4 kilobytes */
1094 iosize = min(iosize * 8, ZERO_REGION_SIZE);
1097 int xfersize = iosize;
1098 if (offset % iosize != 0)
1099 xfersize -= offset % iosize;
1103 aiov.iov_base = __DECONST(void *, zero_region);
1104 aiov.iov_len = xfersize;
1105 auio.uio_iov = &aiov;
1106 auio.uio_iovcnt = 1;
1107 auio.uio_offset = offset;
1108 auio.uio_resid = xfersize;
1109 auio.uio_segflg = UIO_SYSSPACE;
1110 auio.uio_rw = UIO_WRITE;
1113 error = VOP_WRITE(vp, &auio, 0, cred);
1115 len -= xfersize - auio.uio_resid;
1116 offset += xfersize - auio.uio_resid;
1130 vop_stddeallocate(struct vop_deallocate_args *ap)
1137 off_t noff, xfersize, rem;
1140 offset = *ap->a_offset;
1144 error = VOP_GETATTR(vp, &va, cred);
1148 len = omin(OFF_MAX - offset, *ap->a_len);
1151 error = vn_bmap_seekhole_locked(vp, FIOSEEKDATA, &noff, cred);
1154 /* XXX: Is it okay to fallback further? */
1158 * No more data region to be filled
1164 KASSERT(noff >= offset, ("FIOSEEKDATA going backward"));
1165 if (noff != offset) {
1166 xfersize = omin(noff - offset, len);
1172 error = vn_bmap_seekhole_locked(vp, FIOSEEKHOLE, &noff, cred);
1177 xfersize = rem = omin(noff - offset, len);
1178 error = vp_zerofill(vp, &va, &offset, &rem, cred);
1180 len -= xfersize - rem;
1189 *ap->a_offset = offset;
1195 vop_stdadvise(struct vop_advise_args *ap)
1199 daddr_t startn, endn;
1200 off_t bstart, bend, start, end;
1204 switch (ap->a_advice) {
1205 case POSIX_FADV_WILLNEED:
1207 * Do nothing for now. Filesystems should provide a
1208 * custom method which starts an asynchronous read of
1209 * the requested region.
1213 case POSIX_FADV_DONTNEED:
1215 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
1216 if (VN_IS_DOOMED(vp)) {
1222 * Round to block boundaries (and later possibly further to
1223 * page boundaries). Applications cannot reasonably be aware
1224 * of the boundaries, and the rounding must be to expand at
1225 * both extremities to cover enough. It still doesn't cover
1226 * read-ahead. For partial blocks, this gives unnecessary
1227 * discarding of buffers but is efficient enough since the
1228 * pages usually remain in VMIO for some time.
1230 bsize = vp->v_bufobj.bo_bsize;
1231 bstart = rounddown(ap->a_start, bsize);
1232 bend = roundup(ap->a_end, bsize);
1235 * Deactivate pages in the specified range from the backing VM
1236 * object. Pages that are resident in the buffer cache will
1237 * remain wired until their corresponding buffers are released
1240 if (vp->v_object != NULL) {
1241 start = trunc_page(bstart);
1242 end = round_page(bend);
1243 VM_OBJECT_RLOCK(vp->v_object);
1244 vm_object_page_noreuse(vp->v_object, OFF_TO_IDX(start),
1246 VM_OBJECT_RUNLOCK(vp->v_object);
1251 startn = bstart / bsize;
1252 endn = bend / bsize;
1253 error = bnoreuselist(&bo->bo_clean, bo, startn, endn);
1255 error = bnoreuselist(&bo->bo_dirty, bo, startn, endn);
1267 vop_stdunp_bind(struct vop_unp_bind_args *ap)
1270 ap->a_vp->v_unpcb = ap->a_unpcb;
1275 vop_stdunp_connect(struct vop_unp_connect_args *ap)
1278 *ap->a_unpcb = ap->a_vp->v_unpcb;
1283 vop_stdunp_detach(struct vop_unp_detach_args *ap)
1286 ap->a_vp->v_unpcb = NULL;
1291 vop_stdis_text(struct vop_is_text_args *ap)
1294 return (ap->a_vp->v_writecount < 0);
1298 vop_stdset_text(struct vop_set_text_args *ap)
1307 * Avoid the interlock if execs are already present.
1309 n = atomic_load_int(&vp->v_writecount);
1314 if (atomic_fcmpset_int(&vp->v_writecount, &n, n - 1)) {
1320 if (vp->v_writecount > 0) {
1324 * If requested by fs, keep a use reference to the
1325 * vnode until the last text reference is released.
1328 if (mp != NULL && (mp->mnt_kern_flag & MNTK_TEXT_REFS) != 0 &&
1329 vp->v_writecount == 0) {
1330 VNPASS((vp->v_iflag & VI_TEXT_REF) == 0, vp);
1331 vp->v_iflag |= VI_TEXT_REF;
1335 atomic_subtract_int(&vp->v_writecount, 1);
1343 vop_stdunset_text(struct vop_unset_text_args *ap)
1352 * Avoid the interlock if this is not the last exec.
1354 n = atomic_load_int(&vp->v_writecount);
1359 if (atomic_fcmpset_int(&vp->v_writecount, &n, n + 1)) {
1366 if (vp->v_writecount < 0) {
1367 if ((vp->v_iflag & VI_TEXT_REF) != 0 &&
1368 vp->v_writecount == -1) {
1370 vp->v_iflag &= ~VI_TEXT_REF;
1372 atomic_add_int(&vp->v_writecount, 1);
1384 vop_stdadd_writecount(struct vop_add_writecount_args *ap)
1391 VI_LOCK_FLAGS(vp, MTX_DUPOK);
1392 if (vp->v_writecount < 0) {
1395 VNASSERT(vp->v_writecount + ap->a_inc >= 0, vp,
1396 ("neg writecount increment %d", ap->a_inc));
1397 if (vp->v_writecount == 0) {
1399 if (mp != NULL && (mp->mnt_kern_flag & MNTK_NOMSYNC) == 0)
1402 vp->v_writecount += ap->a_inc;
1410 vop_stdneed_inactive(struct vop_need_inactive_args *ap)
1417 vop_stdioctl(struct vop_ioctl_args *ap)
1424 switch (ap->a_command) {
1428 error = vn_lock(vp, LK_SHARED);
1431 if (vp->v_type == VREG)
1432 error = VOP_GETATTR(vp, &va, ap->a_cred);
1437 if (*offp < 0 || *offp >= va.va_size)
1439 else if (ap->a_command == FIOSEEKHOLE)
1453 * used to fill the vfs function table to get reasonable default return values.
1456 vfs_stdroot (mp, flags, vpp)
1462 return (EOPNOTSUPP);
1466 vfs_stdstatfs (mp, sbp)
1471 return (EOPNOTSUPP);
1475 vfs_stdquotactl (mp, cmds, uid, arg, mp_busy)
1482 return (EOPNOTSUPP);
1486 vfs_stdsync(mp, waitfor)
1490 struct vnode *vp, *mvp;
1492 int error, lockreq, allerror = 0;
1495 lockreq = LK_EXCLUSIVE | LK_INTERLOCK;
1496 if (waitfor != MNT_WAIT)
1497 lockreq |= LK_NOWAIT;
1499 * Force stale buffer cache information to be flushed.
1502 MNT_VNODE_FOREACH_ALL(vp, mp, mvp) {
1503 if (vp->v_bufobj.bo_dirty.bv_cnt == 0) {
1507 if ((error = vget(vp, lockreq)) != 0) {
1508 if (error == ENOENT) {
1509 MNT_VNODE_FOREACH_ALL_ABORT(mp, mvp);
1514 error = VOP_FSYNC(vp, waitfor, td);
1523 vfs_stdnosync (mp, waitfor)
1532 vop_stdcopy_file_range(struct vop_copy_file_range_args *ap)
1536 error = vn_generic_copy_file_range(ap->a_invp, ap->a_inoffp,
1537 ap->a_outvp, ap->a_outoffp, ap->a_lenp, ap->a_flags, ap->a_incred,
1538 ap->a_outcred, ap->a_fsizetd);
1543 vfs_stdvget (mp, ino, flags, vpp)
1550 return (EOPNOTSUPP);
1554 vfs_stdfhtovp (mp, fhp, flags, vpp)
1561 return (EOPNOTSUPP);
1566 struct vfsconf *vfsp;
1573 vfs_stduninit (vfsp)
1574 struct vfsconf *vfsp;
1581 vfs_stdextattrctl(mp, cmd, filename_vp, attrnamespace, attrname)
1584 struct vnode *filename_vp;
1586 const char *attrname;
1589 if (filename_vp != NULL)
1590 VOP_UNLOCK(filename_vp);
1591 return (EOPNOTSUPP);
1595 vfs_stdsysctl(mp, op, req)
1598 struct sysctl_req *req;
1601 return (EOPNOTSUPP);
1604 static vop_bypass_t *
1605 bp_by_off(struct vop_vector *vop, struct vop_generic_args *a)
1608 return (*(vop_bypass_t **)((char *)vop + a->a_desc->vdesc_vop_offset));
1612 vop_sigdefer(struct vop_vector *vop, struct vop_generic_args *a)
1617 bp = bp_by_off(vop, a);
1620 prev_stops = sigdeferstop(SIGDEFERSTOP_SILENT);
1622 sigallowstop(prev_stops);
1627 vop_stdstat(struct vop_stat_args *a)
1639 error = vop_stat_helper_pre(a);
1646 * Initialize defaults for new and unusual fields, so that file
1647 * systems which don't support these fields don't need to know
1650 vap->va_birthtime.tv_sec = -1;
1651 vap->va_birthtime.tv_nsec = 0;
1652 vap->va_fsid = VNOVAL;
1654 vap->va_rdev = NODEV;
1656 error = VOP_GETATTR(vp, vap, a->a_active_cred);
1661 * Zero the spare stat fields
1663 bzero(sb, sizeof *sb);
1666 * Copy from vattr table
1668 if (vap->va_fsid != VNOVAL)
1669 sb->st_dev = vap->va_fsid;
1671 sb->st_dev = vp->v_mount->mnt_stat.f_fsid.val[0];
1672 sb->st_ino = vap->va_fileid;
1673 mode = vap->va_mode;
1674 switch (vap->va_type) {
1701 sb->st_nlink = vap->va_nlink;
1702 sb->st_uid = vap->va_uid;
1703 sb->st_gid = vap->va_gid;
1704 sb->st_rdev = vap->va_rdev;
1705 if (vap->va_size > OFF_MAX) {
1709 sb->st_size = vap->va_size;
1710 sb->st_atim.tv_sec = vap->va_atime.tv_sec;
1711 sb->st_atim.tv_nsec = vap->va_atime.tv_nsec;
1712 sb->st_mtim.tv_sec = vap->va_mtime.tv_sec;
1713 sb->st_mtim.tv_nsec = vap->va_mtime.tv_nsec;
1714 sb->st_ctim.tv_sec = vap->va_ctime.tv_sec;
1715 sb->st_ctim.tv_nsec = vap->va_ctime.tv_nsec;
1716 sb->st_birthtim.tv_sec = vap->va_birthtime.tv_sec;
1717 sb->st_birthtim.tv_nsec = vap->va_birthtime.tv_nsec;
1720 * According to www.opengroup.org, the meaning of st_blksize is
1721 * "a filesystem-specific preferred I/O block size for this
1722 * object. In some filesystem types, this may vary from file
1724 * Use minimum/default of PAGE_SIZE (e.g. for VCHR).
1727 sb->st_blksize = max(PAGE_SIZE, vap->va_blocksize);
1728 sb->st_flags = vap->va_flags;
1729 sb->st_blocks = vap->va_bytes / S_BLKSIZE;
1730 sb->st_gen = vap->va_gen;
1732 return (vop_stat_helper_post(a, error));
1736 vop_stdread_pgcache(struct vop_read_pgcache_args *ap __unused)
1738 return (EJUSTRETURN);
1742 vop_stdvput_pair(struct vop_vput_pair_args *ap)
1744 struct vnode *dvp, *vp, **vpp;
1749 if (vpp != NULL && ap->a_unlock_vp && (vp = *vpp) != NULL)