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);
98 * This vnode table stores what we want to do if the filesystem doesn't
99 * implement a particular VOP.
101 * If there is no specific entry here, we will return EOPNOTSUPP.
103 * Note that every filesystem has to implement either vop_access
104 * or vop_accessx; failing to do so will result in immediate crash
105 * due to stack overflow, as vop_stdaccess() calls vop_stdaccessx(),
106 * which calls vop_stdaccess() etc.
109 struct vop_vector default_vnodeops = {
111 .vop_bypass = VOP_EOPNOTSUPP,
113 .vop_access = vop_stdaccess,
114 .vop_accessx = vop_stdaccessx,
115 .vop_advise = vop_stdadvise,
116 .vop_advlock = vop_stdadvlock,
117 .vop_advlockasync = vop_stdadvlockasync,
118 .vop_advlockpurge = vop_stdadvlockpurge,
119 .vop_allocate = vop_stdallocate,
120 .vop_deallocate = vop_stddeallocate,
121 .vop_bmap = vop_stdbmap,
122 .vop_close = VOP_NULL,
123 .vop_fsync = VOP_NULL,
124 .vop_stat = vop_stdstat,
125 .vop_fdatasync = vop_stdfdatasync,
126 .vop_getpages = vop_stdgetpages,
127 .vop_getpages_async = vop_stdgetpages_async,
128 .vop_getwritemount = vop_stdgetwritemount,
129 .vop_inactive = VOP_NULL,
130 .vop_need_inactive = vop_stdneed_inactive,
131 .vop_ioctl = vop_stdioctl,
132 .vop_kqfilter = vop_stdkqfilter,
133 .vop_islocked = vop_stdislocked,
134 .vop_lock1 = vop_stdlock,
135 .vop_lookup = vop_nolookup,
136 .vop_open = VOP_NULL,
137 .vop_pathconf = VOP_EINVAL,
138 .vop_poll = vop_nopoll,
139 .vop_putpages = vop_stdputpages,
140 .vop_readlink = VOP_EINVAL,
141 .vop_read_pgcache = vop_stdread_pgcache,
142 .vop_rename = vop_norename,
143 .vop_revoke = VOP_PANIC,
144 .vop_strategy = vop_nostrategy,
145 .vop_unlock = vop_stdunlock,
146 .vop_vptocnp = vop_stdvptocnp,
147 .vop_vptofh = vop_stdvptofh,
148 .vop_unp_bind = vop_stdunp_bind,
149 .vop_unp_connect = vop_stdunp_connect,
150 .vop_unp_detach = vop_stdunp_detach,
151 .vop_is_text = vop_stdis_text,
152 .vop_set_text = vop_stdset_text,
153 .vop_unset_text = vop_stdunset_text,
154 .vop_add_writecount = vop_stdadd_writecount,
155 .vop_copy_file_range = vop_stdcopy_file_range,
156 .vop_vput_pair = vop_stdvput_pair,
158 VFS_VOP_VECTOR_REGISTER(default_vnodeops);
161 * Series of placeholder functions for various error returns for
166 vop_eopnotsupp(struct vop_generic_args *ap)
169 printf("vop_notsupp[%s]\n", ap->a_desc->vdesc_name);
176 vop_ebadf(struct vop_generic_args *ap)
183 vop_enotty(struct vop_generic_args *ap)
190 vop_einval(struct vop_generic_args *ap)
197 vop_enoent(struct vop_generic_args *ap)
204 vop_eagain(struct vop_generic_args *ap)
211 vop_null(struct vop_generic_args *ap)
218 * Helper function to panic on some bad VOPs in some filesystems.
221 vop_panic(struct vop_generic_args *ap)
224 panic("filesystem goof: vop_panic[%s]", ap->a_desc->vdesc_name);
228 * vop_std<something> and vop_no<something> are default functions for use by
229 * filesystems that need the "default reasonable" implementation for a
230 * particular operation.
232 * The documentation for the operations they implement exists (if it exists)
233 * in the VOP_<SOMETHING>(9) manpage (all uppercase).
237 * Default vop for filesystems that do not support name lookup
240 vop_nolookup(struct vop_lookup_args *ap)
250 * Handle unlock and reference counting for arguments of vop_rename
251 * for filesystems that do not implement rename operation.
254 vop_norename(struct vop_rename_args *ap)
264 * Strategy routine for VFS devices that have none.
266 * BIO_ERROR and B_INVAL must be cleared prior to calling any strategy
267 * routine. Typically this is done for a BIO_READ strategy call.
268 * Typically B_INVAL is assumed to already be clear prior to a write
269 * and should not be cleared manually unless you just made the buffer
270 * invalid. BIO_ERROR should be cleared either way.
274 vop_nostrategy (struct vop_strategy_args *ap)
276 printf("No strategy for buffer at %p\n", ap->a_bp);
277 vn_printf(ap->a_vp, "vnode ");
278 ap->a_bp->b_ioflags |= BIO_ERROR;
279 ap->a_bp->b_error = EOPNOTSUPP;
285 get_next_dirent(struct vnode *vp, struct dirent **dpp, char *dirbuf,
286 int dirbuflen, off_t *off, char **cpos, int *len,
287 int *eofflag, struct thread *td)
294 KASSERT(VOP_ISLOCKED(vp), ("vp %p is not locked", vp));
295 KASSERT(vp->v_type == VDIR, ("vp %p is not a directory", vp));
298 iov.iov_base = dirbuf;
299 iov.iov_len = dirbuflen;
303 uio.uio_offset = *off;
304 uio.uio_resid = dirbuflen;
305 uio.uio_segflg = UIO_SYSSPACE;
306 uio.uio_rw = UIO_READ;
312 error = mac_vnode_check_readdir(td->td_ucred, vp);
315 error = VOP_READDIR(vp, &uio, td->td_ucred, eofflag,
320 *off = uio.uio_offset;
323 *len = (dirbuflen - uio.uio_resid);
329 dp = (struct dirent *)(*cpos);
330 reclen = dp->d_reclen;
333 /* check for malformed directory.. */
334 if (reclen < DIRENT_MINSIZE)
344 * Check if a named file exists in a given directory vnode.
347 dirent_exists(struct vnode *vp, const char *dirname, struct thread *td)
350 int error, eofflag, dirbuflen, len, found;
355 KASSERT(VOP_ISLOCKED(vp), ("vp %p is not locked", vp));
356 KASSERT(vp->v_type == VDIR, ("vp %p is not a directory", vp));
360 error = VOP_GETATTR(vp, &va, td->td_ucred);
364 dirbuflen = DEV_BSIZE;
365 if (dirbuflen < va.va_blocksize)
366 dirbuflen = va.va_blocksize;
367 dirbuf = (char *)malloc(dirbuflen, M_TEMP, M_WAITOK);
372 error = get_next_dirent(vp, &dp, dirbuf, dirbuflen, &off,
373 &cpos, &len, &eofflag, td);
377 if (dp->d_type != DT_WHT && dp->d_fileno != 0 &&
378 strcmp(dp->d_name, dirname) == 0) {
382 } while (len > 0 || !eofflag);
385 free(dirbuf, M_TEMP);
390 vop_stdaccess(struct vop_access_args *ap)
393 KASSERT((ap->a_accmode & ~(VEXEC | VWRITE | VREAD | VADMIN |
394 VAPPEND)) == 0, ("invalid bit in accmode"));
396 return (VOP_ACCESSX(ap->a_vp, ap->a_accmode, ap->a_cred, ap->a_td));
400 vop_stdaccessx(struct vop_accessx_args *ap)
403 accmode_t accmode = ap->a_accmode;
405 error = vfs_unixify_accmode(&accmode);
412 return (VOP_ACCESS(ap->a_vp, accmode, ap->a_cred, ap->a_td));
416 * Advisory record locking support
419 vop_stdadvlock(struct vop_advlock_args *ap)
429 * Provide atomicity of open(O_CREAT | O_EXCL | O_EXLOCK) for
430 * local filesystems. See vn_open_cred() for reciprocal part.
433 if (mp != NULL && (mp->mnt_flag & MNT_LOCAL) != 0 &&
434 ap->a_op == F_SETLK && (ap->a_flags & F_FIRSTOPEN) == 0) {
436 while ((vp->v_iflag & VI_FOPENING) != 0)
437 msleep(vp, VI_MTX(vp), PLOCK, "lockfo", 0);
441 if (ap->a_fl->l_whence == SEEK_END) {
443 * The NFSv4 server must avoid doing a vn_lock() here, since it
444 * can deadlock the nfsd threads, due to a LOR. Fortunately
445 * the NFSv4 server always uses SEEK_SET and this code is
446 * only required for the SEEK_END case.
448 vn_lock(vp, LK_SHARED | LK_RETRY);
449 error = VOP_GETATTR(vp, &vattr, curthread->td_ucred);
456 return (lf_advlock(ap, &(vp->v_lockf), vattr.va_size));
460 vop_stdadvlockasync(struct vop_advlockasync_args *ap)
467 if (ap->a_fl->l_whence == SEEK_END) {
468 /* The size argument is only needed for SEEK_END. */
469 vn_lock(vp, LK_SHARED | LK_RETRY);
470 error = VOP_GETATTR(vp, &vattr, curthread->td_ucred);
477 return (lf_advlockasync(ap, &(vp->v_lockf), vattr.va_size));
481 vop_stdadvlockpurge(struct vop_advlockpurge_args *ap)
486 lf_purgelocks(vp, &vp->v_lockf);
493 * Standard implementation of POSIX pathconf, to get information about limits
495 * Override per filesystem for the case where the filesystem has smaller
499 vop_stdpathconf(struct vop_pathconf_args *ap)
502 switch (ap->a_name) {
504 *ap->a_retval = _POSIX_ASYNCHRONOUS_IO;
507 *ap->a_retval = PATH_MAX;
509 case _PC_ACL_EXTENDED:
511 case _PC_CAP_PRESENT:
512 case _PC_DEALLOC_PRESENT:
513 case _PC_INF_PRESENT:
514 case _PC_MAC_PRESENT:
524 * Standard lock, unlock and islocked functions.
527 vop_stdlock(struct vop_lock1_args *ap)
529 struct vnode *vp = ap->a_vp;
533 return (lockmgr_lock_flags(vp->v_vnlock, ap->a_flags,
534 &ilk->lock_object, ap->a_file, ap->a_line));
539 vop_stdunlock(struct vop_unlock_args *ap)
541 struct vnode *vp = ap->a_vp;
543 return (lockmgr_unlock(vp->v_vnlock));
548 vop_stdislocked(struct vop_islocked_args *ap)
551 return (lockstatus(ap->a_vp->v_vnlock));
555 * Variants of the above set.
558 * - shared locking disablement is not supported
559 * - v_vnlock pointer is not honored
562 vop_lock(struct vop_lock1_args *ap)
564 struct vnode *vp = ap->a_vp;
565 int flags = ap->a_flags;
568 MPASS(vp->v_vnlock == &vp->v_lock);
570 if (__predict_false((flags & ~(LK_TYPE_MASK | LK_NODDLKTREAT | LK_RETRY)) != 0))
573 switch (flags & LK_TYPE_MASK) {
575 return (lockmgr_slock(&vp->v_lock, flags, ap->a_file, ap->a_line));
577 return (lockmgr_xlock(&vp->v_lock, flags, ap->a_file, ap->a_line));
581 return (lockmgr_lock_flags(&vp->v_lock, flags,
582 &ilk->lock_object, ap->a_file, ap->a_line));
586 vop_unlock(struct vop_unlock_args *ap)
588 struct vnode *vp = ap->a_vp;
590 MPASS(vp->v_vnlock == &vp->v_lock);
592 return (lockmgr_unlock(&vp->v_lock));
596 vop_islocked(struct vop_islocked_args *ap)
598 struct vnode *vp = ap->a_vp;
600 MPASS(vp->v_vnlock == &vp->v_lock);
602 return (lockstatus(&vp->v_lock));
606 * Return true for select/poll.
609 vop_nopoll(struct vop_poll_args *ap)
612 if (ap->a_events & ~POLLSTANDARD)
614 return (ap->a_events & (POLLIN | POLLOUT | POLLRDNORM | POLLWRNORM));
618 * Implement poll for local filesystems that support it.
621 vop_stdpoll(struct vop_poll_args *ap)
623 if (ap->a_events & ~POLLSTANDARD)
624 return (vn_pollrecord(ap->a_vp, ap->a_td, ap->a_events));
625 return (ap->a_events & (POLLIN | POLLOUT | POLLRDNORM | POLLWRNORM));
629 * Return our mount point, as we will take charge of the writes.
632 vop_stdgetwritemount(struct vop_getwritemount_args *ap)
638 * Note that having a reference does not prevent forced unmount from
639 * setting ->v_mount to NULL after the lock gets released. This is of
640 * no consequence for typical consumers (most notably vn_start_write)
641 * since in this case the vnode is VIRF_DOOMED. Unmount might have
642 * progressed far enough that its completion is only delayed by the
643 * reference obtained here. The consumer only needs to concern itself
647 mp = vfs_ref_from_vp(vp);
653 * If the file system doesn't implement VOP_BMAP, then return sensible defaults:
654 * - Return the vnode's bufobj instead of any underlying device's bufobj
655 * - Calculate the physical block number as if there were equal size
656 * consecutive blocks, but
657 * - Report no contiguous runs of blocks.
660 vop_stdbmap(struct vop_bmap_args *ap)
663 if (ap->a_bop != NULL)
664 *ap->a_bop = &ap->a_vp->v_bufobj;
665 if (ap->a_bnp != NULL)
666 *ap->a_bnp = ap->a_bn * btodb(ap->a_vp->v_mount->mnt_stat.f_iosize);
667 if (ap->a_runp != NULL)
669 if (ap->a_runb != NULL)
675 vop_stdfsync(struct vop_fsync_args *ap)
678 return (vn_fsync_buf(ap->a_vp, ap->a_waitfor));
682 vop_stdfdatasync(struct vop_fdatasync_args *ap)
685 return (VOP_FSYNC(ap->a_vp, MNT_WAIT, ap->a_td));
689 vop_stdfdatasync_buf(struct vop_fdatasync_args *ap)
692 return (vn_fsync_buf(ap->a_vp, MNT_WAIT));
695 /* XXX Needs good comment and more info in the manpage (VOP_GETPAGES(9)). */
697 vop_stdgetpages(struct vop_getpages_args *ap)
700 return vnode_pager_generic_getpages(ap->a_vp, ap->a_m,
701 ap->a_count, ap->a_rbehind, ap->a_rahead, NULL, NULL);
705 vop_stdgetpages_async(struct vop_getpages_async_args *ap)
709 error = VOP_GETPAGES(ap->a_vp, ap->a_m, ap->a_count, ap->a_rbehind,
711 if (ap->a_iodone != NULL)
712 ap->a_iodone(ap->a_arg, ap->a_m, ap->a_count, error);
717 vop_stdkqfilter(struct vop_kqfilter_args *ap)
719 return vfs_kqfilter(ap);
722 /* XXX Needs good comment and more info in the manpage (VOP_PUTPAGES(9)). */
724 vop_stdputpages(struct vop_putpages_args *ap)
727 return vnode_pager_generic_putpages(ap->a_vp, ap->a_m, ap->a_count,
728 ap->a_sync, ap->a_rtvals);
732 vop_stdvptofh(struct vop_vptofh_args *ap)
738 vop_stdvptocnp(struct vop_vptocnp_args *ap)
740 struct vnode *vp = ap->a_vp;
741 struct vnode **dvp = ap->a_vpp;
743 char *buf = ap->a_buf;
744 size_t *buflen = ap->a_buflen;
746 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 | LOCKSHARED | LOCKLEAF, UIO_SYSSPACE,
774 error = vn_open_cred(&nd, &flags, 0, VN_OPEN_NOAUDIT, cred, NULL);
776 vn_lock(vp, locked | LK_RETRY);
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_SHARED | 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_SHARED | LK_RETRY);
821 if (dirent_exists(mvp, dp->d_name, td)) {
824 vn_lock(*dvp, LK_SHARED | LK_RETRY);
828 vn_lock(*dvp, LK_SHARED | LK_RETRY);
836 if (dp->d_namlen == 1 && dp->d_name[0] == '.') {
839 bcopy(dp->d_name, buf + i, dp->d_namlen);
844 } while (len > 0 || !eofflag);
848 free(dirbuf, M_TEMP);
858 vn_close(mvp, FREAD, cred, td);
860 vn_lock(vp, locked | LK_RETRY);
865 vop_stdallocate(struct vop_allocate_args *ap)
869 off_t maxfilesize = 0;
872 struct vattr vattr, *vap;
874 off_t fsize, len, cur, offset;
887 offset = *ap->a_offset;
889 error = VOP_GETATTR(vp, vap, ap->a_cred);
892 fsize = vap->va_size;
893 iosize = vap->va_blocksize;
895 iosize = BLKDEV_IOSIZE;
896 if (iosize > maxphys)
898 buf = malloc(iosize, M_TEMP, M_WAITOK);
902 * Check if the filesystem sets f_maxfilesize; if not use
903 * VOP_SETATTR to perform the check.
905 sfs = malloc(sizeof(struct statfs), M_STATFS, M_WAITOK);
906 error = VFS_STATFS(vp->v_mount, sfs, td);
908 maxfilesize = sfs->f_maxfilesize;
913 if (offset > maxfilesize || len > maxfilesize ||
914 offset + len > maxfilesize) {
920 if (offset + len > vap->va_size) {
922 * Test offset + len against the filesystem's maxfilesize.
925 vap->va_size = offset + len;
926 error = VOP_SETATTR(vp, vap, ap->a_cred);
930 vap->va_size = fsize;
931 error = VOP_SETATTR(vp, vap, ap->a_cred);
938 * Read and write back anything below the nominal file
939 * size. There's currently no way outside the filesystem
940 * to know whether this area is sparse or not.
943 if ((offset % iosize) != 0)
944 cur -= (offset % iosize);
947 if (offset < fsize) {
950 auio.uio_iov = &aiov;
952 auio.uio_offset = offset;
953 auio.uio_resid = cur;
954 auio.uio_segflg = UIO_SYSSPACE;
955 auio.uio_rw = UIO_READ;
957 error = VOP_READ(vp, &auio, ap->a_ioflag, ap->a_cred);
960 if (auio.uio_resid > 0) {
961 bzero(buf + cur - auio.uio_resid,
970 auio.uio_iov = &aiov;
972 auio.uio_offset = offset;
973 auio.uio_resid = cur;
974 auio.uio_segflg = UIO_SYSSPACE;
975 auio.uio_rw = UIO_WRITE;
978 error = VOP_WRITE(vp, &auio, ap->a_ioflag, ap->a_cred);
992 *ap->a_offset = offset;
998 vp_zerofill(struct vnode *vp, struct vattr *vap, off_t *offsetp, off_t *lenp,
999 int ioflag, struct ucred *cred)
1008 iosize = vap->va_blocksize;
1014 iosize = BLKDEV_IOSIZE;
1015 /* If va_blocksize is 512 bytes, iosize will be 4 kilobytes */
1016 iosize = min(iosize * 8, ZERO_REGION_SIZE);
1019 int xfersize = iosize;
1020 if (offset % iosize != 0)
1021 xfersize -= offset % iosize;
1025 aiov.iov_base = __DECONST(void *, zero_region);
1026 aiov.iov_len = xfersize;
1027 auio.uio_iov = &aiov;
1028 auio.uio_iovcnt = 1;
1029 auio.uio_offset = offset;
1030 auio.uio_resid = xfersize;
1031 auio.uio_segflg = UIO_SYSSPACE;
1032 auio.uio_rw = UIO_WRITE;
1035 error = VOP_WRITE(vp, &auio, ioflag, cred);
1037 len -= xfersize - auio.uio_resid;
1038 offset += xfersize - auio.uio_resid;
1052 vop_stddeallocate(struct vop_deallocate_args *ap)
1059 off_t noff, xfersize, rem;
1062 offset = *ap->a_offset;
1065 error = VOP_GETATTR(vp, &va, cred);
1069 len = omin((off_t)va.va_size - offset, *ap->a_len);
1072 error = vn_bmap_seekhole_locked(vp, FIOSEEKDATA, &noff, cred);
1075 /* XXX: Is it okay to fallback further? */
1079 * No more data region to be filled
1086 KASSERT(noff >= offset, ("FIOSEEKDATA going backward"));
1087 if (noff != offset) {
1088 xfersize = omin(noff - offset, len);
1094 error = vn_bmap_seekhole_locked(vp, FIOSEEKHOLE, &noff, cred);
1099 xfersize = rem = omin(noff - offset, len);
1100 error = vp_zerofill(vp, &va, &offset, &rem, ap->a_ioflag, cred);
1102 len -= xfersize - rem;
1110 /* Handle the case when offset is beyond EOF */
1114 *ap->a_offset = offset;
1120 vop_stdadvise(struct vop_advise_args *ap)
1124 daddr_t startn, endn;
1125 off_t bstart, bend, start, end;
1129 switch (ap->a_advice) {
1130 case POSIX_FADV_WILLNEED:
1132 * Do nothing for now. Filesystems should provide a
1133 * custom method which starts an asynchronous read of
1134 * the requested region.
1138 case POSIX_FADV_DONTNEED:
1140 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
1141 if (VN_IS_DOOMED(vp)) {
1147 * Round to block boundaries (and later possibly further to
1148 * page boundaries). Applications cannot reasonably be aware
1149 * of the boundaries, and the rounding must be to expand at
1150 * both extremities to cover enough. It still doesn't cover
1151 * read-ahead. For partial blocks, this gives unnecessary
1152 * discarding of buffers but is efficient enough since the
1153 * pages usually remain in VMIO for some time.
1155 bsize = vp->v_bufobj.bo_bsize;
1156 bstart = rounddown(ap->a_start, bsize);
1157 bend = roundup(ap->a_end, bsize);
1160 * Deactivate pages in the specified range from the backing VM
1161 * object. Pages that are resident in the buffer cache will
1162 * remain wired until their corresponding buffers are released
1165 if (vp->v_object != NULL) {
1166 start = trunc_page(bstart);
1167 end = round_page(bend);
1168 VM_OBJECT_RLOCK(vp->v_object);
1169 vm_object_page_noreuse(vp->v_object, OFF_TO_IDX(start),
1171 VM_OBJECT_RUNLOCK(vp->v_object);
1176 startn = bstart / bsize;
1177 endn = bend / bsize;
1178 error = bnoreuselist(&bo->bo_clean, bo, startn, endn);
1180 error = bnoreuselist(&bo->bo_dirty, bo, startn, endn);
1192 vop_stdunp_bind(struct vop_unp_bind_args *ap)
1195 ap->a_vp->v_unpcb = ap->a_unpcb;
1200 vop_stdunp_connect(struct vop_unp_connect_args *ap)
1203 *ap->a_unpcb = ap->a_vp->v_unpcb;
1208 vop_stdunp_detach(struct vop_unp_detach_args *ap)
1211 ap->a_vp->v_unpcb = NULL;
1216 vop_stdis_text(struct vop_is_text_args *ap)
1219 return (atomic_load_int(&ap->a_vp->v_writecount) < 0);
1223 vop_stdset_text(struct vop_set_text_args *ap)
1231 n = atomic_load_int(&vp->v_writecount);
1233 if (__predict_false(n > 0)) {
1238 * Transition point, we may need to grab a reference on the vnode.
1240 * Take the ref early As a safety measure against bogus calls
1241 * to vop_stdunset_text.
1245 if ((vn_irflag_read(vp) & VIRF_TEXT_REF) != 0) {
1249 if (atomic_fcmpset_int(&vp->v_writecount, &n, -1)) {
1259 if (atomic_fcmpset_int(&vp->v_writecount, &n, n - 1)) {
1263 __assert_unreachable();
1267 vop_stdunset_text(struct vop_unset_text_args *ap)
1274 n = atomic_load_int(&vp->v_writecount);
1276 if (__predict_false(n >= 0)) {
1281 * Transition point, we may need to release a reference on the vnode.
1284 if (atomic_fcmpset_int(&vp->v_writecount, &n, 0)) {
1285 if ((vn_irflag_read(vp) & VIRF_TEXT_REF) != 0) {
1294 if (atomic_fcmpset_int(&vp->v_writecount, &n, n + 1)) {
1298 __assert_unreachable();
1301 static int __always_inline
1302 vop_stdadd_writecount_impl(struct vop_add_writecount_args *ap, bool handle_msync)
1305 struct mount *mp __diagused;
1314 VNPASS((mp->mnt_kern_flag & MNTK_NOMSYNC) == 0, vp);
1316 VNPASS((mp->mnt_kern_flag & MNTK_NOMSYNC) != 0, vp);
1321 n = atomic_load_int(&vp->v_writecount);
1323 if (__predict_false(n < 0)) {
1327 VNASSERT(n + ap->a_inc >= 0, vp,
1328 ("neg writecount increment %d + %d = %d", n, ap->a_inc,
1336 if (atomic_fcmpset_int(&vp->v_writecount, &n, n + ap->a_inc)) {
1340 __assert_unreachable();
1344 vop_stdadd_writecount(struct vop_add_writecount_args *ap)
1347 return (vop_stdadd_writecount_impl(ap, true));
1351 vop_stdadd_writecount_nomsync(struct vop_add_writecount_args *ap)
1354 return (vop_stdadd_writecount_impl(ap, false));
1358 vop_stdneed_inactive(struct vop_need_inactive_args *ap)
1365 vop_stdioctl(struct vop_ioctl_args *ap)
1372 switch (ap->a_command) {
1376 error = vn_lock(vp, LK_SHARED);
1379 if (vp->v_type == VREG)
1380 error = VOP_GETATTR(vp, &va, ap->a_cred);
1385 if (*offp < 0 || *offp >= va.va_size)
1387 else if (ap->a_command == FIOSEEKHOLE)
1401 * used to fill the vfs function table to get reasonable default return values.
1404 vfs_stdroot(struct mount *mp, int flags, struct vnode **vpp)
1407 return (EOPNOTSUPP);
1411 vfs_stdstatfs(struct mount *mp, struct statfs *sbp)
1414 return (EOPNOTSUPP);
1418 vfs_stdquotactl(struct mount *mp, int cmds, uid_t uid, void *arg, bool *mp_busy)
1420 return (EOPNOTSUPP);
1424 vfs_stdsync(struct mount *mp, int waitfor)
1426 struct vnode *vp, *mvp;
1428 int error, lockreq, allerror = 0;
1431 lockreq = LK_EXCLUSIVE | LK_INTERLOCK;
1432 if (waitfor != MNT_WAIT)
1433 lockreq |= LK_NOWAIT;
1435 * Force stale buffer cache information to be flushed.
1438 MNT_VNODE_FOREACH_ALL(vp, mp, mvp) {
1439 if (vp->v_bufobj.bo_dirty.bv_cnt == 0) {
1443 if ((error = vget(vp, lockreq)) != 0) {
1444 if (error == ENOENT) {
1445 MNT_VNODE_FOREACH_ALL_ABORT(mp, mvp);
1450 error = VOP_FSYNC(vp, waitfor, td);
1459 vfs_stdnosync(struct mount *mp, int waitfor)
1466 vop_stdcopy_file_range(struct vop_copy_file_range_args *ap)
1470 error = vn_generic_copy_file_range(ap->a_invp, ap->a_inoffp,
1471 ap->a_outvp, ap->a_outoffp, ap->a_lenp, ap->a_flags, ap->a_incred,
1472 ap->a_outcred, ap->a_fsizetd);
1477 vfs_stdvget(struct mount *mp, ino_t ino, int flags, struct vnode **vpp)
1480 return (EOPNOTSUPP);
1484 vfs_stdfhtovp(struct mount *mp, struct fid *fhp, int flags, struct vnode **vpp)
1487 return (EOPNOTSUPP);
1491 vfs_stdinit(struct vfsconf *vfsp)
1498 vfs_stduninit(struct vfsconf *vfsp)
1505 vfs_stdextattrctl(struct mount *mp, int cmd, struct vnode *filename_vp,
1506 int attrnamespace, const char *attrname)
1509 if (filename_vp != NULL)
1510 VOP_UNLOCK(filename_vp);
1511 return (EOPNOTSUPP);
1515 vfs_stdsysctl(struct mount *mp, fsctlop_t op, struct sysctl_req *req)
1518 return (EOPNOTSUPP);
1521 static vop_bypass_t *
1522 bp_by_off(struct vop_vector *vop, struct vop_generic_args *a)
1525 return (*(vop_bypass_t **)((char *)vop + a->a_desc->vdesc_vop_offset));
1529 vop_sigdefer(struct vop_vector *vop, struct vop_generic_args *a)
1534 bp = bp_by_off(vop, a);
1537 prev_stops = sigdeferstop(SIGDEFERSTOP_SILENT);
1539 sigallowstop(prev_stops);
1544 vop_stdstat(struct vop_stat_args *a)
1556 error = vop_stat_helper_pre(a);
1563 * Initialize defaults for new and unusual fields, so that file
1564 * systems which don't support these fields don't need to know
1567 vap->va_birthtime.tv_sec = -1;
1568 vap->va_birthtime.tv_nsec = 0;
1569 vap->va_fsid = VNOVAL;
1571 vap->va_rdev = NODEV;
1573 error = VOP_GETATTR(vp, vap, a->a_active_cred);
1578 * Zero the spare stat fields
1580 bzero(sb, sizeof *sb);
1583 * Copy from vattr table
1585 if (vap->va_fsid != VNOVAL)
1586 sb->st_dev = vap->va_fsid;
1588 sb->st_dev = vp->v_mount->mnt_stat.f_fsid.val[0];
1589 sb->st_ino = vap->va_fileid;
1590 mode = vap->va_mode;
1591 switch (vap->va_type) {
1618 sb->st_nlink = vap->va_nlink;
1619 sb->st_uid = vap->va_uid;
1620 sb->st_gid = vap->va_gid;
1621 sb->st_rdev = vap->va_rdev;
1622 if (vap->va_size > OFF_MAX) {
1626 sb->st_size = vap->va_size;
1627 sb->st_atim.tv_sec = vap->va_atime.tv_sec;
1628 sb->st_atim.tv_nsec = vap->va_atime.tv_nsec;
1629 sb->st_mtim.tv_sec = vap->va_mtime.tv_sec;
1630 sb->st_mtim.tv_nsec = vap->va_mtime.tv_nsec;
1631 sb->st_ctim.tv_sec = vap->va_ctime.tv_sec;
1632 sb->st_ctim.tv_nsec = vap->va_ctime.tv_nsec;
1633 sb->st_birthtim.tv_sec = vap->va_birthtime.tv_sec;
1634 sb->st_birthtim.tv_nsec = vap->va_birthtime.tv_nsec;
1637 * According to www.opengroup.org, the meaning of st_blksize is
1638 * "a filesystem-specific preferred I/O block size for this
1639 * object. In some filesystem types, this may vary from file
1641 * Use minimum/default of PAGE_SIZE (e.g. for VCHR).
1644 sb->st_blksize = max(PAGE_SIZE, vap->va_blocksize);
1645 sb->st_flags = vap->va_flags;
1646 sb->st_blocks = vap->va_bytes / S_BLKSIZE;
1647 sb->st_gen = vap->va_gen;
1649 return (vop_stat_helper_post(a, error));
1653 vop_stdread_pgcache(struct vop_read_pgcache_args *ap __unused)
1655 return (EJUSTRETURN);
1659 vop_stdvput_pair(struct vop_vput_pair_args *ap)
1661 struct vnode *dvp, *vp, **vpp;
1666 if (vpp != NULL && ap->a_unlock_vp && (vp = *vpp) != NULL)