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 dirent_exists(struct vnode *vp, const char *dirname,
81 static int vop_stdis_text(struct vop_is_text_args *ap);
82 static int vop_stdunset_text(struct vop_unset_text_args *ap);
83 static int vop_stdadd_writecount(struct vop_add_writecount_args *ap);
84 static int vop_stdcopy_file_range(struct vop_copy_file_range_args *ap);
85 static int vop_stdfdatasync(struct vop_fdatasync_args *ap);
86 static int vop_stdgetpages_async(struct vop_getpages_async_args *ap);
87 static int vop_stdread_pgcache(struct vop_read_pgcache_args *ap);
88 static int vop_stdstat(struct vop_stat_args *ap);
89 static int vop_stdvput_pair(struct vop_vput_pair_args *ap);
92 * This vnode table stores what we want to do if the filesystem doesn't
93 * implement a particular VOP.
95 * If there is no specific entry here, we will return EOPNOTSUPP.
97 * Note that every filesystem has to implement either vop_access
98 * or vop_accessx; failing to do so will result in immediate crash
99 * due to stack overflow, as vop_stdaccess() calls vop_stdaccessx(),
100 * which calls vop_stdaccess() etc.
103 struct vop_vector default_vnodeops = {
105 .vop_bypass = VOP_EOPNOTSUPP,
107 .vop_access = vop_stdaccess,
108 .vop_accessx = vop_stdaccessx,
109 .vop_advise = vop_stdadvise,
110 .vop_advlock = vop_stdadvlock,
111 .vop_advlockasync = vop_stdadvlockasync,
112 .vop_advlockpurge = vop_stdadvlockpurge,
113 .vop_allocate = vop_stdallocate,
114 .vop_deallocate = vop_stddeallocate,
115 .vop_bmap = vop_stdbmap,
116 .vop_close = VOP_NULL,
117 .vop_fsync = VOP_NULL,
118 .vop_stat = vop_stdstat,
119 .vop_fdatasync = vop_stdfdatasync,
120 .vop_getpages = vop_stdgetpages,
121 .vop_getpages_async = vop_stdgetpages_async,
122 .vop_getwritemount = vop_stdgetwritemount,
123 .vop_inactive = VOP_NULL,
124 .vop_need_inactive = vop_stdneed_inactive,
125 .vop_ioctl = vop_stdioctl,
126 .vop_kqfilter = vop_stdkqfilter,
127 .vop_islocked = vop_stdislocked,
128 .vop_lock1 = vop_stdlock,
129 .vop_lookup = vop_nolookup,
130 .vop_open = VOP_NULL,
131 .vop_pathconf = VOP_EINVAL,
132 .vop_poll = vop_nopoll,
133 .vop_putpages = vop_stdputpages,
134 .vop_readlink = VOP_EINVAL,
135 .vop_read_pgcache = vop_stdread_pgcache,
136 .vop_rename = vop_norename,
137 .vop_revoke = VOP_PANIC,
138 .vop_strategy = vop_nostrategy,
139 .vop_unlock = vop_stdunlock,
140 .vop_vptocnp = vop_stdvptocnp,
141 .vop_vptofh = vop_stdvptofh,
142 .vop_unp_bind = vop_stdunp_bind,
143 .vop_unp_connect = vop_stdunp_connect,
144 .vop_unp_detach = vop_stdunp_detach,
145 .vop_is_text = vop_stdis_text,
146 .vop_set_text = vop_stdset_text,
147 .vop_unset_text = vop_stdunset_text,
148 .vop_add_writecount = vop_stdadd_writecount,
149 .vop_copy_file_range = vop_stdcopy_file_range,
150 .vop_vput_pair = vop_stdvput_pair,
152 VFS_VOP_VECTOR_REGISTER(default_vnodeops);
155 * Series of placeholder functions for various error returns for
160 vop_eopnotsupp(struct vop_generic_args *ap)
163 printf("vop_notsupp[%s]\n", ap->a_desc->vdesc_name);
170 vop_ebadf(struct vop_generic_args *ap)
177 vop_enotty(struct vop_generic_args *ap)
184 vop_einval(struct vop_generic_args *ap)
191 vop_enoent(struct vop_generic_args *ap)
198 vop_eagain(struct vop_generic_args *ap)
205 vop_null(struct vop_generic_args *ap)
212 * Helper function to panic on some bad VOPs in some filesystems.
215 vop_panic(struct vop_generic_args *ap)
218 panic("filesystem goof: vop_panic[%s]", ap->a_desc->vdesc_name);
222 * vop_std<something> and vop_no<something> are default functions for use by
223 * filesystems that need the "default reasonable" implementation for a
224 * particular operation.
226 * The documentation for the operations they implement exists (if it exists)
227 * in the VOP_<SOMETHING>(9) manpage (all uppercase).
231 * Default vop for filesystems that do not support name lookup
234 vop_nolookup(struct vop_lookup_args *ap)
244 * Handle unlock and reference counting for arguments of vop_rename
245 * for filesystems that do not implement rename operation.
248 vop_norename(struct vop_rename_args *ap)
258 * Strategy routine for VFS devices that have none.
260 * BIO_ERROR and B_INVAL must be cleared prior to calling any strategy
261 * routine. Typically this is done for a BIO_READ strategy call.
262 * Typically B_INVAL is assumed to already be clear prior to a write
263 * and should not be cleared manually unless you just made the buffer
264 * invalid. BIO_ERROR should be cleared either way.
268 vop_nostrategy (struct vop_strategy_args *ap)
270 printf("No strategy for buffer at %p\n", ap->a_bp);
271 vn_printf(ap->a_vp, "vnode ");
272 ap->a_bp->b_ioflags |= BIO_ERROR;
273 ap->a_bp->b_error = EOPNOTSUPP;
279 * Check if a named file exists in a given directory vnode
281 * Returns 0 if the file exists, ENOENT if it doesn't, or errors returned by
282 * vn_dir_next_dirent().
285 dirent_exists(struct vnode *vp, const char *dirname, struct thread *td)
289 size_t dirbuflen, len;
294 ASSERT_VOP_LOCKED(vp, "vnode not locked");
295 KASSERT(vp->v_type == VDIR, ("vp %p is not a directory", vp));
297 error = VOP_GETATTR(vp, &va, td->td_ucred);
301 dirbuflen = MAX(DEV_BSIZE, GENERIC_MAXDIRSIZ);
302 if (dirbuflen < va.va_blocksize)
303 dirbuflen = va.va_blocksize;
304 dirbuf = malloc(dirbuflen, M_TEMP, M_WAITOK);
311 error = vn_dir_next_dirent(vp, td, dirbuf, dirbuflen,
312 &dp, &len, &off, &eofflag);
319 if (dp->d_type != DT_WHT && dp->d_fileno != 0 &&
320 strcmp(dp->d_name, dirname) == 0)
327 free(dirbuf, M_TEMP);
332 vop_stdaccess(struct vop_access_args *ap)
335 KASSERT((ap->a_accmode & ~(VEXEC | VWRITE | VREAD | VADMIN |
336 VAPPEND)) == 0, ("invalid bit in accmode"));
338 return (VOP_ACCESSX(ap->a_vp, ap->a_accmode, ap->a_cred, ap->a_td));
342 vop_stdaccessx(struct vop_accessx_args *ap)
345 accmode_t accmode = ap->a_accmode;
347 error = vfs_unixify_accmode(&accmode);
354 return (VOP_ACCESS(ap->a_vp, accmode, ap->a_cred, ap->a_td));
358 * Advisory record locking support
361 vop_stdadvlock(struct vop_advlock_args *ap)
371 * Provide atomicity of open(O_CREAT | O_EXCL | O_EXLOCK) for
372 * local filesystems. See vn_open_cred() for reciprocal part.
375 if (mp != NULL && (mp->mnt_flag & MNT_LOCAL) != 0 &&
376 ap->a_op == F_SETLK && (ap->a_flags & F_FIRSTOPEN) == 0) {
378 while ((vp->v_iflag & VI_FOPENING) != 0)
379 msleep(vp, VI_MTX(vp), PLOCK, "lockfo", 0);
383 if (ap->a_fl->l_whence == SEEK_END) {
385 * The NFSv4 server must avoid doing a vn_lock() here, since it
386 * can deadlock the nfsd threads, due to a LOR. Fortunately
387 * the NFSv4 server always uses SEEK_SET and this code is
388 * only required for the SEEK_END case.
390 vn_lock(vp, LK_SHARED | LK_RETRY);
391 error = VOP_GETATTR(vp, &vattr, curthread->td_ucred);
398 return (lf_advlock(ap, &(vp->v_lockf), vattr.va_size));
402 vop_stdadvlockasync(struct vop_advlockasync_args *ap)
409 if (ap->a_fl->l_whence == SEEK_END) {
410 /* The size argument is only needed for SEEK_END. */
411 vn_lock(vp, LK_SHARED | LK_RETRY);
412 error = VOP_GETATTR(vp, &vattr, curthread->td_ucred);
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
441 vop_stdpathconf(struct vop_pathconf_args *ap)
444 switch (ap->a_name) {
446 *ap->a_retval = _POSIX_ASYNCHRONOUS_IO;
449 *ap->a_retval = PATH_MAX;
451 case _PC_ACL_EXTENDED:
453 case _PC_CAP_PRESENT:
454 case _PC_DEALLOC_PRESENT:
455 case _PC_INF_PRESENT:
456 case _PC_MAC_PRESENT:
466 * Standard lock, unlock and islocked functions.
469 vop_stdlock(struct vop_lock1_args *ap)
471 struct vnode *vp = ap->a_vp;
475 return (lockmgr_lock_flags(vp->v_vnlock, ap->a_flags,
476 &ilk->lock_object, ap->a_file, ap->a_line));
481 vop_stdunlock(struct vop_unlock_args *ap)
483 struct vnode *vp = ap->a_vp;
485 return (lockmgr_unlock(vp->v_vnlock));
490 vop_stdislocked(struct vop_islocked_args *ap)
493 return (lockstatus(ap->a_vp->v_vnlock));
497 * Variants of the above set.
500 * - shared locking disablement is not supported
501 * - v_vnlock pointer is not honored
504 vop_lock(struct vop_lock1_args *ap)
506 struct vnode *vp = ap->a_vp;
507 int flags = ap->a_flags;
510 MPASS(vp->v_vnlock == &vp->v_lock);
512 if (__predict_false((flags & ~(LK_TYPE_MASK | LK_NODDLKTREAT | LK_RETRY)) != 0))
515 switch (flags & LK_TYPE_MASK) {
517 return (lockmgr_slock(&vp->v_lock, flags, ap->a_file, ap->a_line));
519 return (lockmgr_xlock(&vp->v_lock, flags, ap->a_file, ap->a_line));
523 return (lockmgr_lock_flags(&vp->v_lock, flags,
524 &ilk->lock_object, ap->a_file, ap->a_line));
528 vop_unlock(struct vop_unlock_args *ap)
530 struct vnode *vp = ap->a_vp;
532 MPASS(vp->v_vnlock == &vp->v_lock);
534 return (lockmgr_unlock(&vp->v_lock));
538 vop_islocked(struct vop_islocked_args *ap)
540 struct vnode *vp = ap->a_vp;
542 MPASS(vp->v_vnlock == &vp->v_lock);
544 return (lockstatus(&vp->v_lock));
548 * Return true for select/poll.
551 vop_nopoll(struct vop_poll_args *ap)
554 if (ap->a_events & ~POLLSTANDARD)
556 return (ap->a_events & (POLLIN | POLLOUT | POLLRDNORM | POLLWRNORM));
560 * Implement poll for local filesystems that support it.
563 vop_stdpoll(struct vop_poll_args *ap)
565 if (ap->a_events & ~POLLSTANDARD)
566 return (vn_pollrecord(ap->a_vp, ap->a_td, ap->a_events));
567 return (ap->a_events & (POLLIN | POLLOUT | POLLRDNORM | POLLWRNORM));
571 * Return our mount point, as we will take charge of the writes.
574 vop_stdgetwritemount(struct vop_getwritemount_args *ap)
580 * Note that having a reference does not prevent forced unmount from
581 * setting ->v_mount to NULL after the lock gets released. This is of
582 * no consequence for typical consumers (most notably vn_start_write)
583 * since in this case the vnode is VIRF_DOOMED. Unmount might have
584 * progressed far enough that its completion is only delayed by the
585 * reference obtained here. The consumer only needs to concern itself
589 mp = vfs_ref_from_vp(vp);
595 * If the file system doesn't implement VOP_BMAP, then return sensible defaults:
596 * - Return the vnode's bufobj instead of any underlying device's bufobj
597 * - Calculate the physical block number as if there were equal size
598 * consecutive blocks, but
599 * - Report no contiguous runs of blocks.
602 vop_stdbmap(struct vop_bmap_args *ap)
605 if (ap->a_bop != NULL)
606 *ap->a_bop = &ap->a_vp->v_bufobj;
607 if (ap->a_bnp != NULL)
608 *ap->a_bnp = ap->a_bn * btodb(ap->a_vp->v_mount->mnt_stat.f_iosize);
609 if (ap->a_runp != NULL)
611 if (ap->a_runb != NULL)
617 vop_stdfsync(struct vop_fsync_args *ap)
620 return (vn_fsync_buf(ap->a_vp, ap->a_waitfor));
624 vop_stdfdatasync(struct vop_fdatasync_args *ap)
627 return (VOP_FSYNC(ap->a_vp, MNT_WAIT, ap->a_td));
631 vop_stdfdatasync_buf(struct vop_fdatasync_args *ap)
634 return (vn_fsync_buf(ap->a_vp, MNT_WAIT));
637 /* XXX Needs good comment and more info in the manpage (VOP_GETPAGES(9)). */
639 vop_stdgetpages(struct vop_getpages_args *ap)
642 return vnode_pager_generic_getpages(ap->a_vp, ap->a_m,
643 ap->a_count, ap->a_rbehind, ap->a_rahead, NULL, NULL);
647 vop_stdgetpages_async(struct vop_getpages_async_args *ap)
651 error = VOP_GETPAGES(ap->a_vp, ap->a_m, ap->a_count, ap->a_rbehind,
653 if (ap->a_iodone != NULL)
654 ap->a_iodone(ap->a_arg, ap->a_m, ap->a_count, error);
659 vop_stdkqfilter(struct vop_kqfilter_args *ap)
661 return vfs_kqfilter(ap);
664 /* XXX Needs good comment and more info in the manpage (VOP_PUTPAGES(9)). */
666 vop_stdputpages(struct vop_putpages_args *ap)
669 return vnode_pager_generic_putpages(ap->a_vp, ap->a_m, ap->a_count,
670 ap->a_sync, ap->a_rtvals);
674 vop_stdvptofh(struct vop_vptofh_args *ap)
680 vop_stdvptocnp(struct vop_vptocnp_args *ap)
682 struct vnode *const vp = ap->a_vp;
683 struct vnode **const dvp = ap->a_vpp;
684 char *buf = ap->a_buf;
685 size_t *buflen = ap->a_buflen;
688 int error = 0, covered = 0;
689 int eofflag, flags, locked;
690 size_t dirbuflen, len;
695 struct thread *const td = curthread;
696 struct ucred *const cred = td->td_ucred;
700 if (vp->v_type != VDIR)
703 error = VOP_GETATTR(vp, &va, cred);
708 locked = VOP_ISLOCKED(vp);
710 NDINIT_ATVP(&nd, LOOKUP, FOLLOW | LOCKSHARED | LOCKLEAF, UIO_SYSSPACE,
713 error = vn_open_cred(&nd, &flags, 0, VN_OPEN_NOAUDIT, cred, NULL);
715 vn_lock(vp, locked | LK_RETRY);
720 mvp = *dvp = nd.ni_vp;
722 if (vp->v_mount != (*dvp)->v_mount &&
723 ((*dvp)->v_vflag & VV_ROOT) &&
724 ((*dvp)->v_mount->mnt_flag & MNT_UNION)) {
725 *dvp = (*dvp)->v_mount->mnt_vnodecovered;
728 vn_close(mvp, FREAD, cred, td);
730 vn_lock(*dvp, LK_SHARED | LK_RETRY);
734 fileno = va.va_fileid;
736 dirbuflen = MAX(DEV_BSIZE, GENERIC_MAXDIRSIZ);
737 if (dirbuflen < va.va_blocksize)
738 dirbuflen = va.va_blocksize;
739 dirbuf = malloc(dirbuflen, M_TEMP, M_WAITOK);
741 if ((*dvp)->v_type != VDIR) {
751 /* call VOP_READDIR of parent */
752 error = vn_dir_next_dirent(*dvp, td,
753 dirbuf, dirbuflen, &dp, &len, &off, &eofflag);
762 if ((dp->d_type != DT_WHT) &&
763 (dp->d_fileno == fileno)) {
766 vn_lock(mvp, LK_SHARED | LK_RETRY);
767 if (dirent_exists(mvp, dp->d_name, td) == 0) {
770 vn_lock(*dvp, LK_SHARED | LK_RETRY);
774 vn_lock(*dvp, LK_SHARED | LK_RETRY);
782 if (dp->d_namlen == 1 && dp->d_name[0] == '.') {
785 bcopy(dp->d_name, buf + i, dp->d_namlen);
793 free(dirbuf, M_TEMP);
803 vn_close(mvp, FREAD, cred, td);
805 vn_lock(vp, locked | LK_RETRY);
810 vop_stdallocate(struct vop_allocate_args *ap)
814 off_t maxfilesize = 0;
817 struct vattr vattr, *vap;
819 off_t fsize, len, cur, offset;
832 offset = *ap->a_offset;
834 error = VOP_GETATTR(vp, vap, ap->a_cred);
837 fsize = vap->va_size;
838 iosize = vap->va_blocksize;
840 iosize = BLKDEV_IOSIZE;
841 if (iosize > maxphys)
843 buf = malloc(iosize, M_TEMP, M_WAITOK);
847 * Check if the filesystem sets f_maxfilesize; if not use
848 * VOP_SETATTR to perform the check.
850 sfs = malloc(sizeof(struct statfs), M_STATFS, M_WAITOK);
851 error = VFS_STATFS(vp->v_mount, sfs, td);
853 maxfilesize = sfs->f_maxfilesize;
858 if (offset > maxfilesize || len > maxfilesize ||
859 offset + len > maxfilesize) {
865 if (offset + len > vap->va_size) {
867 * Test offset + len against the filesystem's maxfilesize.
870 vap->va_size = offset + len;
871 error = VOP_SETATTR(vp, vap, ap->a_cred);
875 vap->va_size = fsize;
876 error = VOP_SETATTR(vp, vap, ap->a_cred);
883 * Read and write back anything below the nominal file
884 * size. There's currently no way outside the filesystem
885 * to know whether this area is sparse or not.
888 if ((offset % iosize) != 0)
889 cur -= (offset % iosize);
892 if (offset < fsize) {
895 auio.uio_iov = &aiov;
897 auio.uio_offset = offset;
898 auio.uio_resid = cur;
899 auio.uio_segflg = UIO_SYSSPACE;
900 auio.uio_rw = UIO_READ;
902 error = VOP_READ(vp, &auio, ap->a_ioflag, ap->a_cred);
905 if (auio.uio_resid > 0) {
906 bzero(buf + cur - auio.uio_resid,
915 auio.uio_iov = &aiov;
917 auio.uio_offset = offset;
918 auio.uio_resid = cur;
919 auio.uio_segflg = UIO_SYSSPACE;
920 auio.uio_rw = UIO_WRITE;
923 error = VOP_WRITE(vp, &auio, ap->a_ioflag, ap->a_cred);
937 *ap->a_offset = offset;
943 vp_zerofill(struct vnode *vp, struct vattr *vap, off_t *offsetp, off_t *lenp,
944 int ioflag, struct ucred *cred)
953 iosize = vap->va_blocksize;
959 iosize = BLKDEV_IOSIZE;
960 /* If va_blocksize is 512 bytes, iosize will be 4 kilobytes */
961 iosize = min(iosize * 8, ZERO_REGION_SIZE);
964 int xfersize = iosize;
965 if (offset % iosize != 0)
966 xfersize -= offset % iosize;
970 aiov.iov_base = __DECONST(void *, zero_region);
971 aiov.iov_len = xfersize;
972 auio.uio_iov = &aiov;
974 auio.uio_offset = offset;
975 auio.uio_resid = xfersize;
976 auio.uio_segflg = UIO_SYSSPACE;
977 auio.uio_rw = UIO_WRITE;
980 error = VOP_WRITE(vp, &auio, ioflag, cred);
982 len -= xfersize - auio.uio_resid;
983 offset += xfersize - auio.uio_resid;
997 vop_stddeallocate(struct vop_deallocate_args *ap)
1004 off_t noff, xfersize, rem;
1007 offset = *ap->a_offset;
1010 error = VOP_GETATTR(vp, &va, cred);
1014 len = omin((off_t)va.va_size - offset, *ap->a_len);
1017 error = vn_bmap_seekhole_locked(vp, FIOSEEKDATA, &noff, cred);
1020 /* XXX: Is it okay to fallback further? */
1024 * No more data region to be filled
1031 KASSERT(noff >= offset, ("FIOSEEKDATA going backward"));
1032 if (noff != offset) {
1033 xfersize = omin(noff - offset, len);
1039 error = vn_bmap_seekhole_locked(vp, FIOSEEKHOLE, &noff, cred);
1044 xfersize = rem = omin(noff - offset, len);
1045 error = vp_zerofill(vp, &va, &offset, &rem, ap->a_ioflag, cred);
1047 len -= xfersize - rem;
1055 /* Handle the case when offset is beyond EOF */
1059 *ap->a_offset = offset;
1065 vop_stdadvise(struct vop_advise_args *ap)
1069 daddr_t startn, endn;
1070 off_t bstart, bend, start, end;
1074 switch (ap->a_advice) {
1075 case POSIX_FADV_WILLNEED:
1077 * Do nothing for now. Filesystems should provide a
1078 * custom method which starts an asynchronous read of
1079 * the requested region.
1083 case POSIX_FADV_DONTNEED:
1085 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
1086 if (VN_IS_DOOMED(vp)) {
1092 * Round to block boundaries (and later possibly further to
1093 * page boundaries). Applications cannot reasonably be aware
1094 * of the boundaries, and the rounding must be to expand at
1095 * both extremities to cover enough. It still doesn't cover
1096 * read-ahead. For partial blocks, this gives unnecessary
1097 * discarding of buffers but is efficient enough since the
1098 * pages usually remain in VMIO for some time.
1100 bsize = vp->v_bufobj.bo_bsize;
1101 bstart = rounddown(ap->a_start, bsize);
1102 bend = roundup(ap->a_end, bsize);
1105 * Deactivate pages in the specified range from the backing VM
1106 * object. Pages that are resident in the buffer cache will
1107 * remain wired until their corresponding buffers are released
1110 if (vp->v_object != NULL) {
1111 start = trunc_page(bstart);
1112 end = round_page(bend);
1113 VM_OBJECT_RLOCK(vp->v_object);
1114 vm_object_page_noreuse(vp->v_object, OFF_TO_IDX(start),
1116 VM_OBJECT_RUNLOCK(vp->v_object);
1121 startn = bstart / bsize;
1122 endn = bend / bsize;
1123 error = bnoreuselist(&bo->bo_clean, bo, startn, endn);
1125 error = bnoreuselist(&bo->bo_dirty, bo, startn, endn);
1137 vop_stdunp_bind(struct vop_unp_bind_args *ap)
1140 ap->a_vp->v_unpcb = ap->a_unpcb;
1145 vop_stdunp_connect(struct vop_unp_connect_args *ap)
1148 *ap->a_unpcb = ap->a_vp->v_unpcb;
1153 vop_stdunp_detach(struct vop_unp_detach_args *ap)
1156 ap->a_vp->v_unpcb = NULL;
1161 vop_stdis_text(struct vop_is_text_args *ap)
1164 return (atomic_load_int(&ap->a_vp->v_writecount) < 0);
1168 vop_stdset_text(struct vop_set_text_args *ap)
1176 n = atomic_load_int(&vp->v_writecount);
1178 if (__predict_false(n > 0)) {
1183 * Transition point, we may need to grab a reference on the vnode.
1185 * Take the ref early As a safety measure against bogus calls
1186 * to vop_stdunset_text.
1190 if ((vn_irflag_read(vp) & VIRF_TEXT_REF) != 0) {
1194 if (atomic_fcmpset_int(&vp->v_writecount, &n, -1)) {
1204 if (atomic_fcmpset_int(&vp->v_writecount, &n, n - 1)) {
1208 __assert_unreachable();
1212 vop_stdunset_text(struct vop_unset_text_args *ap)
1219 n = atomic_load_int(&vp->v_writecount);
1221 if (__predict_false(n >= 0)) {
1226 * Transition point, we may need to release a reference on the vnode.
1229 if (atomic_fcmpset_int(&vp->v_writecount, &n, 0)) {
1230 if ((vn_irflag_read(vp) & VIRF_TEXT_REF) != 0) {
1239 if (atomic_fcmpset_int(&vp->v_writecount, &n, n + 1)) {
1243 __assert_unreachable();
1246 static int __always_inline
1247 vop_stdadd_writecount_impl(struct vop_add_writecount_args *ap, bool handle_msync)
1250 struct mount *mp __diagused;
1259 VNPASS((mp->mnt_kern_flag & MNTK_NOMSYNC) == 0, vp);
1261 VNPASS((mp->mnt_kern_flag & MNTK_NOMSYNC) != 0, vp);
1266 n = atomic_load_int(&vp->v_writecount);
1268 if (__predict_false(n < 0)) {
1272 VNASSERT(n + ap->a_inc >= 0, vp,
1273 ("neg writecount increment %d + %d = %d", n, ap->a_inc,
1281 if (atomic_fcmpset_int(&vp->v_writecount, &n, n + ap->a_inc)) {
1285 __assert_unreachable();
1289 vop_stdadd_writecount(struct vop_add_writecount_args *ap)
1292 return (vop_stdadd_writecount_impl(ap, true));
1296 vop_stdadd_writecount_nomsync(struct vop_add_writecount_args *ap)
1299 return (vop_stdadd_writecount_impl(ap, false));
1303 vop_stdneed_inactive(struct vop_need_inactive_args *ap)
1310 vop_stdioctl(struct vop_ioctl_args *ap)
1317 switch (ap->a_command) {
1321 error = vn_lock(vp, LK_SHARED);
1324 if (vp->v_type == VREG)
1325 error = VOP_GETATTR(vp, &va, ap->a_cred);
1330 if (*offp < 0 || *offp >= va.va_size)
1332 else if (ap->a_command == FIOSEEKHOLE)
1346 * used to fill the vfs function table to get reasonable default return values.
1349 vfs_stdroot(struct mount *mp, int flags, struct vnode **vpp)
1352 return (EOPNOTSUPP);
1356 vfs_stdstatfs(struct mount *mp, struct statfs *sbp)
1359 return (EOPNOTSUPP);
1363 vfs_stdquotactl(struct mount *mp, int cmds, uid_t uid, void *arg, bool *mp_busy)
1365 return (EOPNOTSUPP);
1369 vfs_stdsync(struct mount *mp, int waitfor)
1371 struct vnode *vp, *mvp;
1373 int error, lockreq, allerror = 0;
1376 lockreq = LK_EXCLUSIVE | LK_INTERLOCK;
1377 if (waitfor != MNT_WAIT)
1378 lockreq |= LK_NOWAIT;
1380 * Force stale buffer cache information to be flushed.
1383 MNT_VNODE_FOREACH_ALL(vp, mp, mvp) {
1384 if (vp->v_bufobj.bo_dirty.bv_cnt == 0) {
1388 if ((error = vget(vp, lockreq)) != 0) {
1389 if (error == ENOENT) {
1390 MNT_VNODE_FOREACH_ALL_ABORT(mp, mvp);
1395 error = VOP_FSYNC(vp, waitfor, td);
1404 vfs_stdnosync(struct mount *mp, int waitfor)
1411 vop_stdcopy_file_range(struct vop_copy_file_range_args *ap)
1415 error = vn_generic_copy_file_range(ap->a_invp, ap->a_inoffp,
1416 ap->a_outvp, ap->a_outoffp, ap->a_lenp, ap->a_flags, ap->a_incred,
1417 ap->a_outcred, ap->a_fsizetd);
1422 vfs_stdvget(struct mount *mp, ino_t ino, int flags, struct vnode **vpp)
1425 return (EOPNOTSUPP);
1429 vfs_stdfhtovp(struct mount *mp, struct fid *fhp, int flags, struct vnode **vpp)
1432 return (EOPNOTSUPP);
1436 vfs_stdinit(struct vfsconf *vfsp)
1443 vfs_stduninit(struct vfsconf *vfsp)
1450 vfs_stdextattrctl(struct mount *mp, int cmd, struct vnode *filename_vp,
1451 int attrnamespace, const char *attrname)
1454 if (filename_vp != NULL)
1455 VOP_UNLOCK(filename_vp);
1456 return (EOPNOTSUPP);
1460 vfs_stdsysctl(struct mount *mp, fsctlop_t op, struct sysctl_req *req)
1463 return (EOPNOTSUPP);
1466 static vop_bypass_t *
1467 bp_by_off(struct vop_vector *vop, struct vop_generic_args *a)
1470 return (*(vop_bypass_t **)((char *)vop + a->a_desc->vdesc_vop_offset));
1474 vop_sigdefer(struct vop_vector *vop, struct vop_generic_args *a)
1479 bp = bp_by_off(vop, a);
1482 prev_stops = sigdeferstop(SIGDEFERSTOP_SILENT);
1484 sigallowstop(prev_stops);
1489 vop_stdstat(struct vop_stat_args *a)
1501 error = vop_stat_helper_pre(a);
1508 * Initialize defaults for new and unusual fields, so that file
1509 * systems which don't support these fields don't need to know
1512 vap->va_birthtime.tv_sec = -1;
1513 vap->va_birthtime.tv_nsec = 0;
1514 vap->va_fsid = VNOVAL;
1516 vap->va_rdev = NODEV;
1518 error = VOP_GETATTR(vp, vap, a->a_active_cred);
1523 * Zero the spare stat fields
1525 bzero(sb, sizeof *sb);
1528 * Copy from vattr table
1530 if (vap->va_fsid != VNOVAL)
1531 sb->st_dev = vap->va_fsid;
1533 sb->st_dev = vp->v_mount->mnt_stat.f_fsid.val[0];
1534 sb->st_ino = vap->va_fileid;
1535 mode = vap->va_mode;
1536 switch (vap->va_type) {
1563 sb->st_nlink = vap->va_nlink;
1564 sb->st_uid = vap->va_uid;
1565 sb->st_gid = vap->va_gid;
1566 sb->st_rdev = vap->va_rdev;
1567 if (vap->va_size > OFF_MAX) {
1571 sb->st_size = vap->va_size;
1572 sb->st_atim.tv_sec = vap->va_atime.tv_sec;
1573 sb->st_atim.tv_nsec = vap->va_atime.tv_nsec;
1574 sb->st_mtim.tv_sec = vap->va_mtime.tv_sec;
1575 sb->st_mtim.tv_nsec = vap->va_mtime.tv_nsec;
1576 sb->st_ctim.tv_sec = vap->va_ctime.tv_sec;
1577 sb->st_ctim.tv_nsec = vap->va_ctime.tv_nsec;
1578 sb->st_birthtim.tv_sec = vap->va_birthtime.tv_sec;
1579 sb->st_birthtim.tv_nsec = vap->va_birthtime.tv_nsec;
1582 * According to www.opengroup.org, the meaning of st_blksize is
1583 * "a filesystem-specific preferred I/O block size for this
1584 * object. In some filesystem types, this may vary from file
1586 * Use minimum/default of PAGE_SIZE (e.g. for VCHR).
1589 sb->st_blksize = max(PAGE_SIZE, vap->va_blocksize);
1590 sb->st_flags = vap->va_flags;
1591 sb->st_blocks = vap->va_bytes / S_BLKSIZE;
1592 sb->st_gen = vap->va_gen;
1594 return (vop_stat_helper_post(a, error));
1598 vop_stdread_pgcache(struct vop_read_pgcache_args *ap __unused)
1600 return (EJUSTRETURN);
1604 vop_stdvput_pair(struct vop_vput_pair_args *ap)
1606 struct vnode *dvp, *vp, **vpp;
1611 if (vpp != NULL && ap->a_unlock_vp && (vp = *vpp) != NULL)