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 * 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 * 3. 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/param.h>
36 #include <sys/systm.h>
40 #include <sys/event.h>
41 #include <sys/filio.h>
42 #include <sys/kernel.h>
43 #include <sys/limits.h>
45 #include <sys/lockf.h>
46 #include <sys/malloc.h>
47 #include <sys/mount.h>
48 #include <sys/namei.h>
49 #include <sys/rwlock.h>
50 #include <sys/fcntl.h>
51 #include <sys/unistd.h>
52 #include <sys/vnode.h>
53 #include <sys/dirent.h>
56 #include <security/audit/audit.h>
59 #include <security/mac/mac_framework.h>
62 #include <vm/vm_object.h>
63 #include <vm/vm_extern.h>
65 #include <vm/vm_map.h>
66 #include <vm/vm_page.h>
67 #include <vm/vm_pager.h>
68 #include <vm/vnode_pager.h>
70 static int vop_nolookup(struct vop_lookup_args *);
71 static int vop_norename(struct vop_rename_args *);
72 static int vop_nostrategy(struct vop_strategy_args *);
73 static int dirent_exists(struct vnode *vp, const char *dirname,
76 static int vop_stdis_text(struct vop_is_text_args *ap);
77 static int vop_stdunset_text(struct vop_unset_text_args *ap);
78 static int vop_stdadd_writecount(struct vop_add_writecount_args *ap);
79 static int vop_stdcopy_file_range(struct vop_copy_file_range_args *ap);
80 static int vop_stdfdatasync(struct vop_fdatasync_args *ap);
81 static int vop_stdgetpages_async(struct vop_getpages_async_args *ap);
82 static int vop_stdread_pgcache(struct vop_read_pgcache_args *ap);
83 static int vop_stdstat(struct vop_stat_args *ap);
84 static int vop_stdvput_pair(struct vop_vput_pair_args *ap);
85 static int vop_stdgetlowvnode(struct vop_getlowvnode_args *ap);
88 * This vnode table stores what we want to do if the filesystem doesn't
89 * implement a particular VOP.
91 * If there is no specific entry here, we will return EOPNOTSUPP.
93 * Note that every filesystem has to implement either vop_access
94 * or vop_accessx; failing to do so will result in immediate crash
95 * due to stack overflow, as vop_stdaccess() calls vop_stdaccessx(),
96 * which calls vop_stdaccess() etc.
99 struct vop_vector default_vnodeops = {
101 .vop_bypass = VOP_EOPNOTSUPP,
103 .vop_access = vop_stdaccess,
104 .vop_accessx = vop_stdaccessx,
105 .vop_advise = vop_stdadvise,
106 .vop_advlock = vop_stdadvlock,
107 .vop_advlockasync = vop_stdadvlockasync,
108 .vop_advlockpurge = vop_stdadvlockpurge,
109 .vop_allocate = vop_stdallocate,
110 .vop_deallocate = vop_stddeallocate,
111 .vop_bmap = vop_stdbmap,
112 .vop_close = VOP_NULL,
113 .vop_fsync = VOP_NULL,
114 .vop_stat = vop_stdstat,
115 .vop_fdatasync = vop_stdfdatasync,
116 .vop_getlowvnode = vop_stdgetlowvnode,
117 .vop_getpages = vop_stdgetpages,
118 .vop_getpages_async = vop_stdgetpages_async,
119 .vop_getwritemount = vop_stdgetwritemount,
120 .vop_inactive = VOP_NULL,
121 .vop_need_inactive = vop_stdneed_inactive,
122 .vop_ioctl = vop_stdioctl,
123 .vop_kqfilter = vop_stdkqfilter,
124 .vop_islocked = vop_stdislocked,
125 .vop_lock1 = vop_stdlock,
126 .vop_lookup = vop_nolookup,
127 .vop_open = VOP_NULL,
128 .vop_pathconf = VOP_EINVAL,
129 .vop_poll = vop_nopoll,
130 .vop_putpages = vop_stdputpages,
131 .vop_readlink = VOP_EINVAL,
132 .vop_read_pgcache = vop_stdread_pgcache,
133 .vop_rename = vop_norename,
134 .vop_revoke = VOP_PANIC,
135 .vop_strategy = vop_nostrategy,
136 .vop_unlock = vop_stdunlock,
137 .vop_vptocnp = vop_stdvptocnp,
138 .vop_vptofh = vop_stdvptofh,
139 .vop_unp_bind = vop_stdunp_bind,
140 .vop_unp_connect = vop_stdunp_connect,
141 .vop_unp_detach = vop_stdunp_detach,
142 .vop_is_text = vop_stdis_text,
143 .vop_set_text = vop_stdset_text,
144 .vop_unset_text = vop_stdunset_text,
145 .vop_add_writecount = vop_stdadd_writecount,
146 .vop_copy_file_range = vop_stdcopy_file_range,
147 .vop_vput_pair = vop_stdvput_pair,
149 VFS_VOP_VECTOR_REGISTER(default_vnodeops);
152 * Series of placeholder functions for various error returns for
157 vop_eopnotsupp(struct vop_generic_args *ap)
160 printf("vop_notsupp[%s]\n", ap->a_desc->vdesc_name);
167 vop_ebadf(struct vop_generic_args *ap)
174 vop_enotty(struct vop_generic_args *ap)
181 vop_einval(struct vop_generic_args *ap)
188 vop_enoent(struct vop_generic_args *ap)
195 vop_eagain(struct vop_generic_args *ap)
202 vop_null(struct vop_generic_args *ap)
209 * Helper function to panic on some bad VOPs in some filesystems.
212 vop_panic(struct vop_generic_args *ap)
215 panic("filesystem goof: vop_panic[%s]", ap->a_desc->vdesc_name);
219 * vop_std<something> and vop_no<something> are default functions for use by
220 * filesystems that need the "default reasonable" implementation for a
221 * particular operation.
223 * The documentation for the operations they implement exists (if it exists)
224 * in the VOP_<SOMETHING>(9) manpage (all uppercase).
228 * Default vop for filesystems that do not support name lookup
231 vop_nolookup(struct vop_lookup_args *ap)
241 * Handle unlock and reference counting for arguments of vop_rename
242 * for filesystems that do not implement rename operation.
245 vop_norename(struct vop_rename_args *ap)
255 * Strategy routine for VFS devices that have none.
257 * BIO_ERROR and B_INVAL must be cleared prior to calling any strategy
258 * routine. Typically this is done for a BIO_READ strategy call.
259 * Typically B_INVAL is assumed to already be clear prior to a write
260 * and should not be cleared manually unless you just made the buffer
261 * invalid. BIO_ERROR should be cleared either way.
265 vop_nostrategy (struct vop_strategy_args *ap)
267 printf("No strategy for buffer at %p\n", ap->a_bp);
268 vn_printf(ap->a_vp, "vnode ");
269 ap->a_bp->b_ioflags |= BIO_ERROR;
270 ap->a_bp->b_error = EOPNOTSUPP;
276 * Check if a named file exists in a given directory vnode
278 * Returns 0 if the file exists, ENOENT if it doesn't, or errors returned by
279 * vn_dir_next_dirent().
282 dirent_exists(struct vnode *vp, const char *dirname, struct thread *td)
286 size_t dirbuflen, len;
291 ASSERT_VOP_LOCKED(vp, "vnode not locked");
292 KASSERT(vp->v_type == VDIR, ("vp %p is not a directory", vp));
294 error = VOP_GETATTR(vp, &va, td->td_ucred);
298 dirbuflen = MAX(DEV_BSIZE, GENERIC_MAXDIRSIZ);
299 if (dirbuflen < va.va_blocksize)
300 dirbuflen = va.va_blocksize;
301 dirbuf = malloc(dirbuflen, M_TEMP, M_WAITOK);
308 error = vn_dir_next_dirent(vp, td, dirbuf, dirbuflen,
309 &dp, &len, &off, &eofflag);
316 if (dp->d_type != DT_WHT && dp->d_fileno != 0 &&
317 strcmp(dp->d_name, dirname) == 0)
324 free(dirbuf, M_TEMP);
329 vop_stdaccess(struct vop_access_args *ap)
332 KASSERT((ap->a_accmode & ~(VEXEC | VWRITE | VREAD | VADMIN |
333 VAPPEND)) == 0, ("invalid bit in accmode"));
335 return (VOP_ACCESSX(ap->a_vp, ap->a_accmode, ap->a_cred, ap->a_td));
339 vop_stdaccessx(struct vop_accessx_args *ap)
342 accmode_t accmode = ap->a_accmode;
344 error = vfs_unixify_accmode(&accmode);
351 return (VOP_ACCESS(ap->a_vp, accmode, ap->a_cred, ap->a_td));
355 * Advisory record locking support
358 vop_stdadvlock(struct vop_advlock_args *ap)
368 * Provide atomicity of open(O_CREAT | O_EXCL | O_EXLOCK) for
369 * local filesystems. See vn_open_cred() for reciprocal part.
372 if (mp != NULL && (mp->mnt_flag & MNT_LOCAL) != 0 &&
373 ap->a_op == F_SETLK && (ap->a_flags & F_FIRSTOPEN) == 0) {
375 while ((vp->v_iflag & VI_FOPENING) != 0)
376 msleep(vp, VI_MTX(vp), PLOCK, "lockfo", 0);
380 if (ap->a_fl->l_whence == SEEK_END) {
382 * The NFSv4 server must avoid doing a vn_lock() here, since it
383 * can deadlock the nfsd threads, due to a LOR. Fortunately
384 * the NFSv4 server always uses SEEK_SET and this code is
385 * only required for the SEEK_END case.
387 vn_lock(vp, LK_SHARED | LK_RETRY);
388 error = VOP_GETATTR(vp, &vattr, curthread->td_ucred);
395 return (lf_advlock(ap, &(vp->v_lockf), vattr.va_size));
399 vop_stdadvlockasync(struct vop_advlockasync_args *ap)
406 if (ap->a_fl->l_whence == SEEK_END) {
407 /* The size argument is only needed for SEEK_END. */
408 vn_lock(vp, LK_SHARED | LK_RETRY);
409 error = VOP_GETATTR(vp, &vattr, curthread->td_ucred);
416 return (lf_advlockasync(ap, &(vp->v_lockf), vattr.va_size));
420 vop_stdadvlockpurge(struct vop_advlockpurge_args *ap)
425 lf_purgelocks(vp, &vp->v_lockf);
432 * Standard implementation of POSIX pathconf, to get information about limits
434 * Override per filesystem for the case where the filesystem has smaller
438 vop_stdpathconf(struct vop_pathconf_args *ap)
441 switch (ap->a_name) {
443 *ap->a_retval = _POSIX_ASYNCHRONOUS_IO;
446 *ap->a_retval = PATH_MAX;
448 case _PC_ACL_EXTENDED:
450 case _PC_CAP_PRESENT:
451 case _PC_DEALLOC_PRESENT:
452 case _PC_INF_PRESENT:
453 case _PC_MAC_PRESENT:
463 * Standard lock, unlock and islocked functions.
466 vop_stdlock(struct vop_lock1_args *ap)
468 struct vnode *vp = ap->a_vp;
472 return (lockmgr_lock_flags(vp->v_vnlock, ap->a_flags,
473 &ilk->lock_object, ap->a_file, ap->a_line));
478 vop_stdunlock(struct vop_unlock_args *ap)
480 struct vnode *vp = ap->a_vp;
482 return (lockmgr_unlock(vp->v_vnlock));
487 vop_stdislocked(struct vop_islocked_args *ap)
490 return (lockstatus(ap->a_vp->v_vnlock));
494 * Variants of the above set.
497 * - shared locking disablement is not supported
498 * - v_vnlock pointer is not honored
501 vop_lock(struct vop_lock1_args *ap)
503 struct vnode *vp = ap->a_vp;
504 int flags = ap->a_flags;
507 MPASS(vp->v_vnlock == &vp->v_lock);
509 if (__predict_false((flags & ~(LK_TYPE_MASK | LK_NODDLKTREAT | LK_RETRY)) != 0))
512 switch (flags & LK_TYPE_MASK) {
514 return (lockmgr_slock(&vp->v_lock, flags, ap->a_file, ap->a_line));
516 return (lockmgr_xlock(&vp->v_lock, flags, ap->a_file, ap->a_line));
520 return (lockmgr_lock_flags(&vp->v_lock, flags,
521 &ilk->lock_object, ap->a_file, ap->a_line));
525 vop_unlock(struct vop_unlock_args *ap)
527 struct vnode *vp = ap->a_vp;
529 MPASS(vp->v_vnlock == &vp->v_lock);
531 return (lockmgr_unlock(&vp->v_lock));
535 vop_islocked(struct vop_islocked_args *ap)
537 struct vnode *vp = ap->a_vp;
539 MPASS(vp->v_vnlock == &vp->v_lock);
541 return (lockstatus(&vp->v_lock));
545 * Return true for select/poll.
548 vop_nopoll(struct vop_poll_args *ap)
551 if (ap->a_events & ~POLLSTANDARD)
553 return (ap->a_events & (POLLIN | POLLOUT | POLLRDNORM | POLLWRNORM));
557 * Implement poll for local filesystems that support it.
560 vop_stdpoll(struct vop_poll_args *ap)
562 if (ap->a_events & ~POLLSTANDARD)
563 return (vn_pollrecord(ap->a_vp, ap->a_td, ap->a_events));
564 return (ap->a_events & (POLLIN | POLLOUT | POLLRDNORM | POLLWRNORM));
568 * Return our mount point, as we will take charge of the writes.
571 vop_stdgetwritemount(struct vop_getwritemount_args *ap)
577 * Note that having a reference does not prevent forced unmount from
578 * setting ->v_mount to NULL after the lock gets released. This is of
579 * no consequence for typical consumers (most notably vn_start_write)
580 * since in this case the vnode is VIRF_DOOMED. Unmount might have
581 * progressed far enough that its completion is only delayed by the
582 * reference obtained here. The consumer only needs to concern itself
586 mp = vfs_ref_from_vp(vp);
592 * If the file system doesn't implement VOP_BMAP, then return sensible defaults:
593 * - Return the vnode's bufobj instead of any underlying device's bufobj
594 * - Calculate the physical block number as if there were equal size
595 * consecutive blocks, but
596 * - Report no contiguous runs of blocks.
599 vop_stdbmap(struct vop_bmap_args *ap)
602 if (ap->a_bop != NULL)
603 *ap->a_bop = &ap->a_vp->v_bufobj;
604 if (ap->a_bnp != NULL)
605 *ap->a_bnp = ap->a_bn * btodb(ap->a_vp->v_mount->mnt_stat.f_iosize);
606 if (ap->a_runp != NULL)
608 if (ap->a_runb != NULL)
614 vop_stdfsync(struct vop_fsync_args *ap)
617 return (vn_fsync_buf(ap->a_vp, ap->a_waitfor));
621 vop_stdfdatasync(struct vop_fdatasync_args *ap)
624 return (VOP_FSYNC(ap->a_vp, MNT_WAIT, ap->a_td));
628 vop_stdfdatasync_buf(struct vop_fdatasync_args *ap)
631 return (vn_fsync_buf(ap->a_vp, MNT_WAIT));
634 /* XXX Needs good comment and more info in the manpage (VOP_GETPAGES(9)). */
636 vop_stdgetpages(struct vop_getpages_args *ap)
639 return vnode_pager_generic_getpages(ap->a_vp, ap->a_m,
640 ap->a_count, ap->a_rbehind, ap->a_rahead, NULL, NULL);
644 vop_stdgetpages_async(struct vop_getpages_async_args *ap)
648 error = VOP_GETPAGES(ap->a_vp, ap->a_m, ap->a_count, ap->a_rbehind,
650 if (ap->a_iodone != NULL)
651 ap->a_iodone(ap->a_arg, ap->a_m, ap->a_count, error);
656 vop_stdkqfilter(struct vop_kqfilter_args *ap)
658 return vfs_kqfilter(ap);
661 /* XXX Needs good comment and more info in the manpage (VOP_PUTPAGES(9)). */
663 vop_stdputpages(struct vop_putpages_args *ap)
666 return vnode_pager_generic_putpages(ap->a_vp, ap->a_m, ap->a_count,
667 ap->a_sync, ap->a_rtvals);
671 vop_stdvptofh(struct vop_vptofh_args *ap)
677 vop_stdvptocnp(struct vop_vptocnp_args *ap)
679 struct vnode *const vp = ap->a_vp;
680 struct vnode **const dvp = ap->a_vpp;
681 char *buf = ap->a_buf;
682 size_t *buflen = ap->a_buflen;
685 int error = 0, covered = 0;
686 int eofflag, flags, locked;
687 size_t dirbuflen, len;
692 struct thread *const td = curthread;
693 struct ucred *const cred = td->td_ucred;
697 if (vp->v_type != VDIR)
700 error = VOP_GETATTR(vp, &va, cred);
705 locked = VOP_ISLOCKED(vp);
707 NDINIT_ATVP(&nd, LOOKUP, FOLLOW | LOCKSHARED | LOCKLEAF, UIO_SYSSPACE,
710 error = vn_open_cred(&nd, &flags, 0, VN_OPEN_NOAUDIT, cred, NULL);
712 vn_lock(vp, locked | LK_RETRY);
717 mvp = *dvp = nd.ni_vp;
719 if (vp->v_mount != (*dvp)->v_mount &&
720 ((*dvp)->v_vflag & VV_ROOT) &&
721 ((*dvp)->v_mount->mnt_flag & MNT_UNION)) {
722 *dvp = (*dvp)->v_mount->mnt_vnodecovered;
725 vn_close(mvp, FREAD, cred, td);
727 vn_lock(*dvp, LK_SHARED | LK_RETRY);
731 fileno = va.va_fileid;
733 dirbuflen = MAX(DEV_BSIZE, GENERIC_MAXDIRSIZ);
734 if (dirbuflen < va.va_blocksize)
735 dirbuflen = va.va_blocksize;
736 dirbuf = malloc(dirbuflen, M_TEMP, M_WAITOK);
738 if ((*dvp)->v_type != VDIR) {
748 /* call VOP_READDIR of parent */
749 error = vn_dir_next_dirent(*dvp, td,
750 dirbuf, dirbuflen, &dp, &len, &off, &eofflag);
759 if ((dp->d_type != DT_WHT) &&
760 (dp->d_fileno == fileno)) {
763 vn_lock(mvp, LK_SHARED | LK_RETRY);
764 if (dirent_exists(mvp, dp->d_name, td) == 0) {
767 vn_lock(*dvp, LK_SHARED | LK_RETRY);
771 vn_lock(*dvp, LK_SHARED | LK_RETRY);
779 if (dp->d_namlen == 1 && dp->d_name[0] == '.') {
782 bcopy(dp->d_name, buf + i, dp->d_namlen);
790 free(dirbuf, M_TEMP);
800 vn_close(mvp, FREAD, cred, td);
802 vn_lock(vp, locked | LK_RETRY);
807 vop_stdallocate(struct vop_allocate_args *ap)
811 off_t maxfilesize = 0;
814 struct vattr vattr, *vap;
816 off_t fsize, len, cur, offset;
829 offset = *ap->a_offset;
831 error = VOP_GETATTR(vp, vap, ap->a_cred);
834 fsize = vap->va_size;
835 iosize = vap->va_blocksize;
837 iosize = BLKDEV_IOSIZE;
838 if (iosize > maxphys)
840 buf = malloc(iosize, M_TEMP, M_WAITOK);
844 * Check if the filesystem sets f_maxfilesize; if not use
845 * VOP_SETATTR to perform the check.
847 sfs = malloc(sizeof(struct statfs), M_STATFS, M_WAITOK);
848 error = VFS_STATFS(vp->v_mount, sfs, td);
850 maxfilesize = sfs->f_maxfilesize;
855 if (offset > maxfilesize || len > maxfilesize ||
856 offset + len > maxfilesize) {
862 if (offset + len > vap->va_size) {
864 * Test offset + len against the filesystem's maxfilesize.
867 vap->va_size = offset + len;
868 error = VOP_SETATTR(vp, vap, ap->a_cred);
872 vap->va_size = fsize;
873 error = VOP_SETATTR(vp, vap, ap->a_cred);
880 * Read and write back anything below the nominal file
881 * size. There's currently no way outside the filesystem
882 * to know whether this area is sparse or not.
885 if ((offset % iosize) != 0)
886 cur -= (offset % iosize);
889 if (offset < fsize) {
892 auio.uio_iov = &aiov;
894 auio.uio_offset = offset;
895 auio.uio_resid = cur;
896 auio.uio_segflg = UIO_SYSSPACE;
897 auio.uio_rw = UIO_READ;
899 error = VOP_READ(vp, &auio, ap->a_ioflag, ap->a_cred);
902 if (auio.uio_resid > 0) {
903 bzero(buf + cur - auio.uio_resid,
912 auio.uio_iov = &aiov;
914 auio.uio_offset = offset;
915 auio.uio_resid = cur;
916 auio.uio_segflg = UIO_SYSSPACE;
917 auio.uio_rw = UIO_WRITE;
920 error = VOP_WRITE(vp, &auio, ap->a_ioflag, ap->a_cred);
934 *ap->a_offset = offset;
940 vp_zerofill(struct vnode *vp, struct vattr *vap, off_t *offsetp, off_t *lenp,
941 int ioflag, struct ucred *cred)
950 iosize = vap->va_blocksize;
956 iosize = BLKDEV_IOSIZE;
957 /* If va_blocksize is 512 bytes, iosize will be 4 kilobytes */
958 iosize = min(iosize * 8, ZERO_REGION_SIZE);
961 int xfersize = iosize;
962 if (offset % iosize != 0)
963 xfersize -= offset % iosize;
967 aiov.iov_base = __DECONST(void *, zero_region);
968 aiov.iov_len = xfersize;
969 auio.uio_iov = &aiov;
971 auio.uio_offset = offset;
972 auio.uio_resid = xfersize;
973 auio.uio_segflg = UIO_SYSSPACE;
974 auio.uio_rw = UIO_WRITE;
977 error = VOP_WRITE(vp, &auio, ioflag, cred);
979 len -= xfersize - auio.uio_resid;
980 offset += xfersize - auio.uio_resid;
994 vop_stddeallocate(struct vop_deallocate_args *ap)
1001 off_t noff, xfersize, rem;
1004 offset = *ap->a_offset;
1007 error = VOP_GETATTR(vp, &va, cred);
1011 len = omin((off_t)va.va_size - offset, *ap->a_len);
1014 error = vn_bmap_seekhole_locked(vp, FIOSEEKDATA, &noff, cred);
1017 /* XXX: Is it okay to fallback further? */
1021 * No more data region to be filled
1028 KASSERT(noff >= offset, ("FIOSEEKDATA going backward"));
1029 if (noff != offset) {
1030 xfersize = omin(noff - offset, len);
1036 error = vn_bmap_seekhole_locked(vp, FIOSEEKHOLE, &noff, cred);
1041 xfersize = rem = omin(noff - offset, len);
1042 error = vp_zerofill(vp, &va, &offset, &rem, ap->a_ioflag, cred);
1044 len -= xfersize - rem;
1052 /* Handle the case when offset is beyond EOF */
1056 *ap->a_offset = offset;
1062 vop_stdadvise(struct vop_advise_args *ap)
1066 daddr_t startn, endn;
1067 off_t bstart, bend, start, end;
1071 switch (ap->a_advice) {
1072 case POSIX_FADV_WILLNEED:
1074 * Do nothing for now. Filesystems should provide a
1075 * custom method which starts an asynchronous read of
1076 * the requested region.
1080 case POSIX_FADV_DONTNEED:
1082 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
1083 if (VN_IS_DOOMED(vp)) {
1089 * Round to block boundaries (and later possibly further to
1090 * page boundaries). Applications cannot reasonably be aware
1091 * of the boundaries, and the rounding must be to expand at
1092 * both extremities to cover enough. It still doesn't cover
1093 * read-ahead. For partial blocks, this gives unnecessary
1094 * discarding of buffers but is efficient enough since the
1095 * pages usually remain in VMIO for some time.
1097 bsize = vp->v_bufobj.bo_bsize;
1098 bstart = rounddown(ap->a_start, bsize);
1099 bend = roundup(ap->a_end, bsize);
1102 * Deactivate pages in the specified range from the backing VM
1103 * object. Pages that are resident in the buffer cache will
1104 * remain wired until their corresponding buffers are released
1107 if (vp->v_object != NULL) {
1108 start = trunc_page(bstart);
1109 end = round_page(bend);
1110 VM_OBJECT_RLOCK(vp->v_object);
1111 vm_object_page_noreuse(vp->v_object, OFF_TO_IDX(start),
1113 VM_OBJECT_RUNLOCK(vp->v_object);
1118 startn = bstart / bsize;
1119 endn = bend / bsize;
1120 error = bnoreuselist(&bo->bo_clean, bo, startn, endn);
1122 error = bnoreuselist(&bo->bo_dirty, bo, startn, endn);
1134 vop_stdunp_bind(struct vop_unp_bind_args *ap)
1137 ap->a_vp->v_unpcb = ap->a_unpcb;
1142 vop_stdunp_connect(struct vop_unp_connect_args *ap)
1145 *ap->a_unpcb = ap->a_vp->v_unpcb;
1150 vop_stdunp_detach(struct vop_unp_detach_args *ap)
1153 ap->a_vp->v_unpcb = NULL;
1158 vop_stdis_text(struct vop_is_text_args *ap)
1161 return (atomic_load_int(&ap->a_vp->v_writecount) < 0);
1165 vop_stdset_text(struct vop_set_text_args *ap)
1173 n = atomic_load_int(&vp->v_writecount);
1175 if (__predict_false(n > 0)) {
1180 * Transition point, we may need to grab a reference on the vnode.
1182 * Take the ref early As a safety measure against bogus calls
1183 * to vop_stdunset_text.
1187 if ((vn_irflag_read(vp) & VIRF_TEXT_REF) != 0) {
1191 if (atomic_fcmpset_int(&vp->v_writecount, &n, -1)) {
1201 if (atomic_fcmpset_int(&vp->v_writecount, &n, n - 1)) {
1205 __assert_unreachable();
1209 vop_stdunset_text(struct vop_unset_text_args *ap)
1216 n = atomic_load_int(&vp->v_writecount);
1218 if (__predict_false(n >= 0)) {
1223 * Transition point, we may need to release a reference on the vnode.
1226 if (atomic_fcmpset_int(&vp->v_writecount, &n, 0)) {
1227 if ((vn_irflag_read(vp) & VIRF_TEXT_REF) != 0) {
1236 if (atomic_fcmpset_int(&vp->v_writecount, &n, n + 1)) {
1240 __assert_unreachable();
1243 static int __always_inline
1244 vop_stdadd_writecount_impl(struct vop_add_writecount_args *ap, bool handle_msync)
1247 struct mount *mp __diagused;
1256 VNPASS((mp->mnt_kern_flag & MNTK_NOMSYNC) == 0, vp);
1258 VNPASS((mp->mnt_kern_flag & MNTK_NOMSYNC) != 0, vp);
1263 n = atomic_load_int(&vp->v_writecount);
1265 if (__predict_false(n < 0)) {
1269 VNASSERT(n + ap->a_inc >= 0, vp,
1270 ("neg writecount increment %d + %d = %d", n, ap->a_inc,
1278 if (atomic_fcmpset_int(&vp->v_writecount, &n, n + ap->a_inc)) {
1282 __assert_unreachable();
1286 vop_stdadd_writecount(struct vop_add_writecount_args *ap)
1289 return (vop_stdadd_writecount_impl(ap, true));
1293 vop_stdadd_writecount_nomsync(struct vop_add_writecount_args *ap)
1296 return (vop_stdadd_writecount_impl(ap, false));
1300 vop_stdneed_inactive(struct vop_need_inactive_args *ap)
1307 vop_stdioctl(struct vop_ioctl_args *ap)
1314 switch (ap->a_command) {
1318 error = vn_lock(vp, LK_SHARED);
1321 if (vp->v_type == VREG)
1322 error = VOP_GETATTR(vp, &va, ap->a_cred);
1327 if (*offp < 0 || *offp >= va.va_size)
1329 else if (ap->a_command == FIOSEEKHOLE)
1343 * used to fill the vfs function table to get reasonable default return values.
1346 vfs_stdroot(struct mount *mp, int flags, struct vnode **vpp)
1349 return (EOPNOTSUPP);
1353 vfs_stdstatfs(struct mount *mp, struct statfs *sbp)
1356 return (EOPNOTSUPP);
1360 vfs_stdquotactl(struct mount *mp, int cmds, uid_t uid, void *arg, bool *mp_busy)
1362 return (EOPNOTSUPP);
1366 vfs_stdsync(struct mount *mp, int waitfor)
1368 struct vnode *vp, *mvp;
1370 int error, lockreq, allerror = 0;
1373 lockreq = LK_EXCLUSIVE | LK_INTERLOCK;
1374 if (waitfor != MNT_WAIT)
1375 lockreq |= LK_NOWAIT;
1377 * Force stale buffer cache information to be flushed.
1380 MNT_VNODE_FOREACH_ALL(vp, mp, mvp) {
1381 if (vp->v_bufobj.bo_dirty.bv_cnt == 0) {
1385 if ((error = vget(vp, lockreq)) != 0) {
1386 if (error == ENOENT) {
1387 MNT_VNODE_FOREACH_ALL_ABORT(mp, mvp);
1392 error = VOP_FSYNC(vp, waitfor, td);
1401 vfs_stdnosync(struct mount *mp, int waitfor)
1408 vop_stdcopy_file_range(struct vop_copy_file_range_args *ap)
1412 error = vn_generic_copy_file_range(ap->a_invp, ap->a_inoffp,
1413 ap->a_outvp, ap->a_outoffp, ap->a_lenp, ap->a_flags, ap->a_incred,
1414 ap->a_outcred, ap->a_fsizetd);
1419 vfs_stdvget(struct mount *mp, ino_t ino, int flags, struct vnode **vpp)
1422 return (EOPNOTSUPP);
1426 vfs_stdfhtovp(struct mount *mp, struct fid *fhp, int flags, struct vnode **vpp)
1429 return (EOPNOTSUPP);
1433 vfs_stdinit(struct vfsconf *vfsp)
1440 vfs_stduninit(struct vfsconf *vfsp)
1447 vfs_stdextattrctl(struct mount *mp, int cmd, struct vnode *filename_vp,
1448 int attrnamespace, const char *attrname)
1451 if (filename_vp != NULL)
1452 VOP_UNLOCK(filename_vp);
1453 return (EOPNOTSUPP);
1457 vfs_stdsysctl(struct mount *mp, fsctlop_t op, struct sysctl_req *req)
1460 return (EOPNOTSUPP);
1463 static vop_bypass_t *
1464 bp_by_off(struct vop_vector *vop, struct vop_generic_args *a)
1467 return (*(vop_bypass_t **)((char *)vop + a->a_desc->vdesc_vop_offset));
1471 vop_sigdefer(struct vop_vector *vop, struct vop_generic_args *a)
1476 bp = bp_by_off(vop, a);
1479 prev_stops = sigdeferstop(SIGDEFERSTOP_SILENT);
1481 sigallowstop(prev_stops);
1486 vop_stdstat(struct vop_stat_args *a)
1498 error = vop_stat_helper_pre(a);
1505 * Initialize defaults for new and unusual fields, so that file
1506 * systems which don't support these fields don't need to know
1509 vap->va_birthtime.tv_sec = -1;
1510 vap->va_birthtime.tv_nsec = 0;
1511 vap->va_fsid = VNOVAL;
1513 vap->va_rdev = NODEV;
1515 error = VOP_GETATTR(vp, vap, a->a_active_cred);
1520 * Zero the spare stat fields
1522 bzero(sb, sizeof *sb);
1525 * Copy from vattr table
1527 if (vap->va_fsid != VNOVAL)
1528 sb->st_dev = vap->va_fsid;
1530 sb->st_dev = vp->v_mount->mnt_stat.f_fsid.val[0];
1531 sb->st_ino = vap->va_fileid;
1532 mode = vap->va_mode;
1533 switch (vap->va_type) {
1560 sb->st_nlink = vap->va_nlink;
1561 sb->st_uid = vap->va_uid;
1562 sb->st_gid = vap->va_gid;
1563 sb->st_rdev = vap->va_rdev;
1564 if (vap->va_size > OFF_MAX) {
1568 sb->st_size = vap->va_size;
1569 sb->st_atim.tv_sec = vap->va_atime.tv_sec;
1570 sb->st_atim.tv_nsec = vap->va_atime.tv_nsec;
1571 sb->st_mtim.tv_sec = vap->va_mtime.tv_sec;
1572 sb->st_mtim.tv_nsec = vap->va_mtime.tv_nsec;
1573 sb->st_ctim.tv_sec = vap->va_ctime.tv_sec;
1574 sb->st_ctim.tv_nsec = vap->va_ctime.tv_nsec;
1575 sb->st_birthtim.tv_sec = vap->va_birthtime.tv_sec;
1576 sb->st_birthtim.tv_nsec = vap->va_birthtime.tv_nsec;
1579 * According to www.opengroup.org, the meaning of st_blksize is
1580 * "a filesystem-specific preferred I/O block size for this
1581 * object. In some filesystem types, this may vary from file
1583 * Use minimum/default of PAGE_SIZE (e.g. for VCHR).
1586 sb->st_blksize = max(PAGE_SIZE, vap->va_blocksize);
1587 sb->st_flags = vap->va_flags;
1588 sb->st_blocks = vap->va_bytes / S_BLKSIZE;
1589 sb->st_gen = vap->va_gen;
1591 return (vop_stat_helper_post(a, error));
1595 vop_stdread_pgcache(struct vop_read_pgcache_args *ap __unused)
1597 return (EJUSTRETURN);
1601 vop_stdvput_pair(struct vop_vput_pair_args *ap)
1603 struct vnode *dvp, *vp, **vpp;
1608 if (vpp != NULL && ap->a_unlock_vp && (vp = *vpp) != NULL)
1614 vop_stdgetlowvnode(struct vop_getlowvnode_args *ap)
1617 *ap->a_vplp = ap->a_vp;