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 #include <sys/param.h>
39 #include <sys/systm.h>
43 #include <sys/event.h>
44 #include <sys/filio.h>
45 #include <sys/kernel.h>
46 #include <sys/limits.h>
48 #include <sys/lockf.h>
49 #include <sys/malloc.h>
50 #include <sys/mount.h>
51 #include <sys/namei.h>
52 #include <sys/rwlock.h>
53 #include <sys/fcntl.h>
54 #include <sys/unistd.h>
55 #include <sys/vnode.h>
56 #include <sys/dirent.h>
59 #include <security/audit/audit.h>
62 #include <security/mac/mac_framework.h>
65 #include <vm/vm_object.h>
66 #include <vm/vm_extern.h>
68 #include <vm/vm_map.h>
69 #include <vm/vm_page.h>
70 #include <vm/vm_pager.h>
71 #include <vm/vnode_pager.h>
73 static int vop_nolookup(struct vop_lookup_args *);
74 static int vop_norename(struct vop_rename_args *);
75 static int vop_nostrategy(struct vop_strategy_args *);
76 static int dirent_exists(struct vnode *vp, const char *dirname,
79 static int vop_stdis_text(struct vop_is_text_args *ap);
80 static int vop_stdunset_text(struct vop_unset_text_args *ap);
81 static int vop_stdadd_writecount(struct vop_add_writecount_args *ap);
82 static int vop_stdcopy_file_range(struct vop_copy_file_range_args *ap);
83 static int vop_stdfdatasync(struct vop_fdatasync_args *ap);
84 static int vop_stdgetpages_async(struct vop_getpages_async_args *ap);
85 static int vop_stdread_pgcache(struct vop_read_pgcache_args *ap);
86 static int vop_stdstat(struct vop_stat_args *ap);
87 static int vop_stdvput_pair(struct vop_vput_pair_args *ap);
90 * This vnode table stores what we want to do if the filesystem doesn't
91 * implement a particular VOP.
93 * If there is no specific entry here, we will return EOPNOTSUPP.
95 * Note that every filesystem has to implement either vop_access
96 * or vop_accessx; failing to do so will result in immediate crash
97 * due to stack overflow, as vop_stdaccess() calls vop_stdaccessx(),
98 * which calls vop_stdaccess() etc.
101 struct vop_vector default_vnodeops = {
103 .vop_bypass = VOP_EOPNOTSUPP,
105 .vop_access = vop_stdaccess,
106 .vop_accessx = vop_stdaccessx,
107 .vop_advise = vop_stdadvise,
108 .vop_advlock = vop_stdadvlock,
109 .vop_advlockasync = vop_stdadvlockasync,
110 .vop_advlockpurge = vop_stdadvlockpurge,
111 .vop_allocate = vop_stdallocate,
112 .vop_deallocate = vop_stddeallocate,
113 .vop_bmap = vop_stdbmap,
114 .vop_close = VOP_NULL,
115 .vop_fsync = VOP_NULL,
116 .vop_stat = vop_stdstat,
117 .vop_fdatasync = vop_stdfdatasync,
118 .vop_getpages = vop_stdgetpages,
119 .vop_getpages_async = vop_stdgetpages_async,
120 .vop_getwritemount = vop_stdgetwritemount,
121 .vop_inactive = VOP_NULL,
122 .vop_need_inactive = vop_stdneed_inactive,
123 .vop_ioctl = vop_stdioctl,
124 .vop_kqfilter = vop_stdkqfilter,
125 .vop_islocked = vop_stdislocked,
126 .vop_lock1 = vop_stdlock,
127 .vop_lookup = vop_nolookup,
128 .vop_open = VOP_NULL,
129 .vop_pathconf = VOP_EINVAL,
130 .vop_poll = vop_nopoll,
131 .vop_putpages = vop_stdputpages,
132 .vop_readlink = VOP_EINVAL,
133 .vop_read_pgcache = vop_stdread_pgcache,
134 .vop_rename = vop_norename,
135 .vop_revoke = VOP_PANIC,
136 .vop_strategy = vop_nostrategy,
137 .vop_unlock = vop_stdunlock,
138 .vop_vptocnp = vop_stdvptocnp,
139 .vop_vptofh = vop_stdvptofh,
140 .vop_unp_bind = vop_stdunp_bind,
141 .vop_unp_connect = vop_stdunp_connect,
142 .vop_unp_detach = vop_stdunp_detach,
143 .vop_is_text = vop_stdis_text,
144 .vop_set_text = vop_stdset_text,
145 .vop_unset_text = vop_stdunset_text,
146 .vop_add_writecount = vop_stdadd_writecount,
147 .vop_copy_file_range = vop_stdcopy_file_range,
148 .vop_vput_pair = vop_stdvput_pair,
150 VFS_VOP_VECTOR_REGISTER(default_vnodeops);
153 * Series of placeholder functions for various error returns for
158 vop_eopnotsupp(struct vop_generic_args *ap)
161 printf("vop_notsupp[%s]\n", ap->a_desc->vdesc_name);
168 vop_ebadf(struct vop_generic_args *ap)
175 vop_enotty(struct vop_generic_args *ap)
182 vop_einval(struct vop_generic_args *ap)
189 vop_enoent(struct vop_generic_args *ap)
196 vop_eagain(struct vop_generic_args *ap)
203 vop_null(struct vop_generic_args *ap)
210 * Helper function to panic on some bad VOPs in some filesystems.
213 vop_panic(struct vop_generic_args *ap)
216 panic("filesystem goof: vop_panic[%s]", ap->a_desc->vdesc_name);
220 * vop_std<something> and vop_no<something> are default functions for use by
221 * filesystems that need the "default reasonable" implementation for a
222 * particular operation.
224 * The documentation for the operations they implement exists (if it exists)
225 * in the VOP_<SOMETHING>(9) manpage (all uppercase).
229 * Default vop for filesystems that do not support name lookup
232 vop_nolookup(struct vop_lookup_args *ap)
242 * Handle unlock and reference counting for arguments of vop_rename
243 * for filesystems that do not implement rename operation.
246 vop_norename(struct vop_rename_args *ap)
256 * Strategy routine for VFS devices that have none.
258 * BIO_ERROR and B_INVAL must be cleared prior to calling any strategy
259 * routine. Typically this is done for a BIO_READ strategy call.
260 * Typically B_INVAL is assumed to already be clear prior to a write
261 * and should not be cleared manually unless you just made the buffer
262 * invalid. BIO_ERROR should be cleared either way.
266 vop_nostrategy (struct vop_strategy_args *ap)
268 printf("No strategy for buffer at %p\n", ap->a_bp);
269 vn_printf(ap->a_vp, "vnode ");
270 ap->a_bp->b_ioflags |= BIO_ERROR;
271 ap->a_bp->b_error = EOPNOTSUPP;
277 * Check if a named file exists in a given directory vnode
279 * Returns 0 if the file exists, ENOENT if it doesn't, or errors returned by
280 * vn_dir_next_dirent().
283 dirent_exists(struct vnode *vp, const char *dirname, struct thread *td)
287 size_t dirbuflen, len;
292 ASSERT_VOP_LOCKED(vp, "vnode not locked");
293 KASSERT(vp->v_type == VDIR, ("vp %p is not a directory", vp));
295 error = VOP_GETATTR(vp, &va, td->td_ucred);
299 dirbuflen = MAX(DEV_BSIZE, GENERIC_MAXDIRSIZ);
300 if (dirbuflen < va.va_blocksize)
301 dirbuflen = va.va_blocksize;
302 dirbuf = malloc(dirbuflen, M_TEMP, M_WAITOK);
309 error = vn_dir_next_dirent(vp, td, dirbuf, dirbuflen,
310 &dp, &len, &off, &eofflag);
317 if (dp->d_type != DT_WHT && dp->d_fileno != 0 &&
318 strcmp(dp->d_name, dirname) == 0)
325 free(dirbuf, M_TEMP);
330 vop_stdaccess(struct vop_access_args *ap)
333 KASSERT((ap->a_accmode & ~(VEXEC | VWRITE | VREAD | VADMIN |
334 VAPPEND)) == 0, ("invalid bit in accmode"));
336 return (VOP_ACCESSX(ap->a_vp, ap->a_accmode, ap->a_cred, ap->a_td));
340 vop_stdaccessx(struct vop_accessx_args *ap)
343 accmode_t accmode = ap->a_accmode;
345 error = vfs_unixify_accmode(&accmode);
352 return (VOP_ACCESS(ap->a_vp, accmode, ap->a_cred, ap->a_td));
356 * Advisory record locking support
359 vop_stdadvlock(struct vop_advlock_args *ap)
369 * Provide atomicity of open(O_CREAT | O_EXCL | O_EXLOCK) for
370 * local filesystems. See vn_open_cred() for reciprocal part.
373 if (mp != NULL && (mp->mnt_flag & MNT_LOCAL) != 0 &&
374 ap->a_op == F_SETLK && (ap->a_flags & F_FIRSTOPEN) == 0) {
376 while ((vp->v_iflag & VI_FOPENING) != 0)
377 msleep(vp, VI_MTX(vp), PLOCK, "lockfo", 0);
381 if (ap->a_fl->l_whence == SEEK_END) {
383 * The NFSv4 server must avoid doing a vn_lock() here, since it
384 * can deadlock the nfsd threads, due to a LOR. Fortunately
385 * the NFSv4 server always uses SEEK_SET and this code is
386 * only required for the SEEK_END case.
388 vn_lock(vp, LK_SHARED | LK_RETRY);
389 error = VOP_GETATTR(vp, &vattr, curthread->td_ucred);
396 return (lf_advlock(ap, &(vp->v_lockf), vattr.va_size));
400 vop_stdadvlockasync(struct vop_advlockasync_args *ap)
407 if (ap->a_fl->l_whence == SEEK_END) {
408 /* The size argument is only needed for SEEK_END. */
409 vn_lock(vp, LK_SHARED | LK_RETRY);
410 error = VOP_GETATTR(vp, &vattr, curthread->td_ucred);
417 return (lf_advlockasync(ap, &(vp->v_lockf), vattr.va_size));
421 vop_stdadvlockpurge(struct vop_advlockpurge_args *ap)
426 lf_purgelocks(vp, &vp->v_lockf);
433 * Standard implementation of POSIX pathconf, to get information about limits
435 * Override per filesystem for the case where the filesystem has smaller
439 vop_stdpathconf(struct vop_pathconf_args *ap)
442 switch (ap->a_name) {
444 *ap->a_retval = _POSIX_ASYNCHRONOUS_IO;
447 *ap->a_retval = PATH_MAX;
449 case _PC_ACL_EXTENDED:
451 case _PC_CAP_PRESENT:
452 case _PC_DEALLOC_PRESENT:
453 case _PC_INF_PRESENT:
454 case _PC_MAC_PRESENT:
464 * Standard lock, unlock and islocked functions.
467 vop_stdlock(struct vop_lock1_args *ap)
469 struct vnode *vp = ap->a_vp;
473 return (lockmgr_lock_flags(vp->v_vnlock, ap->a_flags,
474 &ilk->lock_object, ap->a_file, ap->a_line));
479 vop_stdunlock(struct vop_unlock_args *ap)
481 struct vnode *vp = ap->a_vp;
483 return (lockmgr_unlock(vp->v_vnlock));
488 vop_stdislocked(struct vop_islocked_args *ap)
491 return (lockstatus(ap->a_vp->v_vnlock));
495 * Variants of the above set.
498 * - shared locking disablement is not supported
499 * - v_vnlock pointer is not honored
502 vop_lock(struct vop_lock1_args *ap)
504 struct vnode *vp = ap->a_vp;
505 int flags = ap->a_flags;
508 MPASS(vp->v_vnlock == &vp->v_lock);
510 if (__predict_false((flags & ~(LK_TYPE_MASK | LK_NODDLKTREAT | LK_RETRY)) != 0))
513 switch (flags & LK_TYPE_MASK) {
515 return (lockmgr_slock(&vp->v_lock, flags, ap->a_file, ap->a_line));
517 return (lockmgr_xlock(&vp->v_lock, flags, ap->a_file, ap->a_line));
521 return (lockmgr_lock_flags(&vp->v_lock, flags,
522 &ilk->lock_object, ap->a_file, ap->a_line));
526 vop_unlock(struct vop_unlock_args *ap)
528 struct vnode *vp = ap->a_vp;
530 MPASS(vp->v_vnlock == &vp->v_lock);
532 return (lockmgr_unlock(&vp->v_lock));
536 vop_islocked(struct vop_islocked_args *ap)
538 struct vnode *vp = ap->a_vp;
540 MPASS(vp->v_vnlock == &vp->v_lock);
542 return (lockstatus(&vp->v_lock));
546 * Return true for select/poll.
549 vop_nopoll(struct vop_poll_args *ap)
552 if (ap->a_events & ~POLLSTANDARD)
554 return (ap->a_events & (POLLIN | POLLOUT | POLLRDNORM | POLLWRNORM));
558 * Implement poll for local filesystems that support it.
561 vop_stdpoll(struct vop_poll_args *ap)
563 if (ap->a_events & ~POLLSTANDARD)
564 return (vn_pollrecord(ap->a_vp, ap->a_td, ap->a_events));
565 return (ap->a_events & (POLLIN | POLLOUT | POLLRDNORM | POLLWRNORM));
569 * Return our mount point, as we will take charge of the writes.
572 vop_stdgetwritemount(struct vop_getwritemount_args *ap)
578 * Note that having a reference does not prevent forced unmount from
579 * setting ->v_mount to NULL after the lock gets released. This is of
580 * no consequence for typical consumers (most notably vn_start_write)
581 * since in this case the vnode is VIRF_DOOMED. Unmount might have
582 * progressed far enough that its completion is only delayed by the
583 * reference obtained here. The consumer only needs to concern itself
587 mp = vfs_ref_from_vp(vp);
593 * If the file system doesn't implement VOP_BMAP, then return sensible defaults:
594 * - Return the vnode's bufobj instead of any underlying device's bufobj
595 * - Calculate the physical block number as if there were equal size
596 * consecutive blocks, but
597 * - Report no contiguous runs of blocks.
600 vop_stdbmap(struct vop_bmap_args *ap)
603 if (ap->a_bop != NULL)
604 *ap->a_bop = &ap->a_vp->v_bufobj;
605 if (ap->a_bnp != NULL)
606 *ap->a_bnp = ap->a_bn * btodb(ap->a_vp->v_mount->mnt_stat.f_iosize);
607 if (ap->a_runp != NULL)
609 if (ap->a_runb != NULL)
615 vop_stdfsync(struct vop_fsync_args *ap)
618 return (vn_fsync_buf(ap->a_vp, ap->a_waitfor));
622 vop_stdfdatasync(struct vop_fdatasync_args *ap)
625 return (VOP_FSYNC(ap->a_vp, MNT_WAIT, ap->a_td));
629 vop_stdfdatasync_buf(struct vop_fdatasync_args *ap)
632 return (vn_fsync_buf(ap->a_vp, MNT_WAIT));
635 /* XXX Needs good comment and more info in the manpage (VOP_GETPAGES(9)). */
637 vop_stdgetpages(struct vop_getpages_args *ap)
640 return vnode_pager_generic_getpages(ap->a_vp, ap->a_m,
641 ap->a_count, ap->a_rbehind, ap->a_rahead, NULL, NULL);
645 vop_stdgetpages_async(struct vop_getpages_async_args *ap)
649 error = VOP_GETPAGES(ap->a_vp, ap->a_m, ap->a_count, ap->a_rbehind,
651 if (ap->a_iodone != NULL)
652 ap->a_iodone(ap->a_arg, ap->a_m, ap->a_count, error);
657 vop_stdkqfilter(struct vop_kqfilter_args *ap)
659 return vfs_kqfilter(ap);
662 /* XXX Needs good comment and more info in the manpage (VOP_PUTPAGES(9)). */
664 vop_stdputpages(struct vop_putpages_args *ap)
667 return vnode_pager_generic_putpages(ap->a_vp, ap->a_m, ap->a_count,
668 ap->a_sync, ap->a_rtvals);
672 vop_stdvptofh(struct vop_vptofh_args *ap)
678 vop_stdvptocnp(struct vop_vptocnp_args *ap)
680 struct vnode *const vp = ap->a_vp;
681 struct vnode **const dvp = ap->a_vpp;
682 char *buf = ap->a_buf;
683 size_t *buflen = ap->a_buflen;
686 int error = 0, covered = 0;
687 int eofflag, flags, locked;
688 size_t dirbuflen, len;
693 struct thread *const td = curthread;
694 struct ucred *const cred = td->td_ucred;
698 if (vp->v_type != VDIR)
701 error = VOP_GETATTR(vp, &va, cred);
706 locked = VOP_ISLOCKED(vp);
708 NDINIT_ATVP(&nd, LOOKUP, FOLLOW | LOCKSHARED | LOCKLEAF, UIO_SYSSPACE,
711 error = vn_open_cred(&nd, &flags, 0, VN_OPEN_NOAUDIT, cred, NULL);
713 vn_lock(vp, locked | LK_RETRY);
718 mvp = *dvp = nd.ni_vp;
720 if (vp->v_mount != (*dvp)->v_mount &&
721 ((*dvp)->v_vflag & VV_ROOT) &&
722 ((*dvp)->v_mount->mnt_flag & MNT_UNION)) {
723 *dvp = (*dvp)->v_mount->mnt_vnodecovered;
726 vn_close(mvp, FREAD, cred, td);
728 vn_lock(*dvp, LK_SHARED | LK_RETRY);
732 fileno = va.va_fileid;
734 dirbuflen = MAX(DEV_BSIZE, GENERIC_MAXDIRSIZ);
735 if (dirbuflen < va.va_blocksize)
736 dirbuflen = va.va_blocksize;
737 dirbuf = malloc(dirbuflen, M_TEMP, M_WAITOK);
739 if ((*dvp)->v_type != VDIR) {
749 /* call VOP_READDIR of parent */
750 error = vn_dir_next_dirent(*dvp, td,
751 dirbuf, dirbuflen, &dp, &len, &off, &eofflag);
760 if ((dp->d_type != DT_WHT) &&
761 (dp->d_fileno == fileno)) {
764 vn_lock(mvp, LK_SHARED | LK_RETRY);
765 if (dirent_exists(mvp, dp->d_name, td) == 0) {
768 vn_lock(*dvp, LK_SHARED | LK_RETRY);
772 vn_lock(*dvp, LK_SHARED | LK_RETRY);
780 if (dp->d_namlen == 1 && dp->d_name[0] == '.') {
783 bcopy(dp->d_name, buf + i, dp->d_namlen);
791 free(dirbuf, M_TEMP);
801 vn_close(mvp, FREAD, cred, td);
803 vn_lock(vp, locked | LK_RETRY);
808 vop_stdallocate(struct vop_allocate_args *ap)
812 off_t maxfilesize = 0;
815 struct vattr vattr, *vap;
817 off_t fsize, len, cur, offset;
830 offset = *ap->a_offset;
832 error = VOP_GETATTR(vp, vap, ap->a_cred);
835 fsize = vap->va_size;
836 iosize = vap->va_blocksize;
838 iosize = BLKDEV_IOSIZE;
839 if (iosize > maxphys)
841 buf = malloc(iosize, M_TEMP, M_WAITOK);
845 * Check if the filesystem sets f_maxfilesize; if not use
846 * VOP_SETATTR to perform the check.
848 sfs = malloc(sizeof(struct statfs), M_STATFS, M_WAITOK);
849 error = VFS_STATFS(vp->v_mount, sfs, td);
851 maxfilesize = sfs->f_maxfilesize;
856 if (offset > maxfilesize || len > maxfilesize ||
857 offset + len > maxfilesize) {
863 if (offset + len > vap->va_size) {
865 * Test offset + len against the filesystem's maxfilesize.
868 vap->va_size = offset + len;
869 error = VOP_SETATTR(vp, vap, ap->a_cred);
873 vap->va_size = fsize;
874 error = VOP_SETATTR(vp, vap, ap->a_cred);
881 * Read and write back anything below the nominal file
882 * size. There's currently no way outside the filesystem
883 * to know whether this area is sparse or not.
886 if ((offset % iosize) != 0)
887 cur -= (offset % iosize);
890 if (offset < fsize) {
893 auio.uio_iov = &aiov;
895 auio.uio_offset = offset;
896 auio.uio_resid = cur;
897 auio.uio_segflg = UIO_SYSSPACE;
898 auio.uio_rw = UIO_READ;
900 error = VOP_READ(vp, &auio, ap->a_ioflag, ap->a_cred);
903 if (auio.uio_resid > 0) {
904 bzero(buf + cur - auio.uio_resid,
913 auio.uio_iov = &aiov;
915 auio.uio_offset = offset;
916 auio.uio_resid = cur;
917 auio.uio_segflg = UIO_SYSSPACE;
918 auio.uio_rw = UIO_WRITE;
921 error = VOP_WRITE(vp, &auio, ap->a_ioflag, ap->a_cred);
935 *ap->a_offset = offset;
941 vp_zerofill(struct vnode *vp, struct vattr *vap, off_t *offsetp, off_t *lenp,
942 int ioflag, struct ucred *cred)
951 iosize = vap->va_blocksize;
957 iosize = BLKDEV_IOSIZE;
958 /* If va_blocksize is 512 bytes, iosize will be 4 kilobytes */
959 iosize = min(iosize * 8, ZERO_REGION_SIZE);
962 int xfersize = iosize;
963 if (offset % iosize != 0)
964 xfersize -= offset % iosize;
968 aiov.iov_base = __DECONST(void *, zero_region);
969 aiov.iov_len = xfersize;
970 auio.uio_iov = &aiov;
972 auio.uio_offset = offset;
973 auio.uio_resid = xfersize;
974 auio.uio_segflg = UIO_SYSSPACE;
975 auio.uio_rw = UIO_WRITE;
978 error = VOP_WRITE(vp, &auio, ioflag, cred);
980 len -= xfersize - auio.uio_resid;
981 offset += xfersize - auio.uio_resid;
995 vop_stddeallocate(struct vop_deallocate_args *ap)
1002 off_t noff, xfersize, rem;
1005 offset = *ap->a_offset;
1008 error = VOP_GETATTR(vp, &va, cred);
1012 len = omin((off_t)va.va_size - offset, *ap->a_len);
1015 error = vn_bmap_seekhole_locked(vp, FIOSEEKDATA, &noff, cred);
1018 /* XXX: Is it okay to fallback further? */
1022 * No more data region to be filled
1029 KASSERT(noff >= offset, ("FIOSEEKDATA going backward"));
1030 if (noff != offset) {
1031 xfersize = omin(noff - offset, len);
1037 error = vn_bmap_seekhole_locked(vp, FIOSEEKHOLE, &noff, cred);
1042 xfersize = rem = omin(noff - offset, len);
1043 error = vp_zerofill(vp, &va, &offset, &rem, ap->a_ioflag, cred);
1045 len -= xfersize - rem;
1053 /* Handle the case when offset is beyond EOF */
1057 *ap->a_offset = offset;
1063 vop_stdadvise(struct vop_advise_args *ap)
1067 daddr_t startn, endn;
1068 off_t bstart, bend, start, end;
1072 switch (ap->a_advice) {
1073 case POSIX_FADV_WILLNEED:
1075 * Do nothing for now. Filesystems should provide a
1076 * custom method which starts an asynchronous read of
1077 * the requested region.
1081 case POSIX_FADV_DONTNEED:
1083 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
1084 if (VN_IS_DOOMED(vp)) {
1090 * Round to block boundaries (and later possibly further to
1091 * page boundaries). Applications cannot reasonably be aware
1092 * of the boundaries, and the rounding must be to expand at
1093 * both extremities to cover enough. It still doesn't cover
1094 * read-ahead. For partial blocks, this gives unnecessary
1095 * discarding of buffers but is efficient enough since the
1096 * pages usually remain in VMIO for some time.
1098 bsize = vp->v_bufobj.bo_bsize;
1099 bstart = rounddown(ap->a_start, bsize);
1100 bend = roundup(ap->a_end, bsize);
1103 * Deactivate pages in the specified range from the backing VM
1104 * object. Pages that are resident in the buffer cache will
1105 * remain wired until their corresponding buffers are released
1108 if (vp->v_object != NULL) {
1109 start = trunc_page(bstart);
1110 end = round_page(bend);
1111 VM_OBJECT_RLOCK(vp->v_object);
1112 vm_object_page_noreuse(vp->v_object, OFF_TO_IDX(start),
1114 VM_OBJECT_RUNLOCK(vp->v_object);
1119 startn = bstart / bsize;
1120 endn = bend / bsize;
1121 error = bnoreuselist(&bo->bo_clean, bo, startn, endn);
1123 error = bnoreuselist(&bo->bo_dirty, bo, startn, endn);
1135 vop_stdunp_bind(struct vop_unp_bind_args *ap)
1138 ap->a_vp->v_unpcb = ap->a_unpcb;
1143 vop_stdunp_connect(struct vop_unp_connect_args *ap)
1146 *ap->a_unpcb = ap->a_vp->v_unpcb;
1151 vop_stdunp_detach(struct vop_unp_detach_args *ap)
1154 ap->a_vp->v_unpcb = NULL;
1159 vop_stdis_text(struct vop_is_text_args *ap)
1162 return (atomic_load_int(&ap->a_vp->v_writecount) < 0);
1166 vop_stdset_text(struct vop_set_text_args *ap)
1174 n = atomic_load_int(&vp->v_writecount);
1176 if (__predict_false(n > 0)) {
1181 * Transition point, we may need to grab a reference on the vnode.
1183 * Take the ref early As a safety measure against bogus calls
1184 * to vop_stdunset_text.
1188 if ((vn_irflag_read(vp) & VIRF_TEXT_REF) != 0) {
1192 if (atomic_fcmpset_int(&vp->v_writecount, &n, -1)) {
1202 if (atomic_fcmpset_int(&vp->v_writecount, &n, n - 1)) {
1206 __assert_unreachable();
1210 vop_stdunset_text(struct vop_unset_text_args *ap)
1217 n = atomic_load_int(&vp->v_writecount);
1219 if (__predict_false(n >= 0)) {
1224 * Transition point, we may need to release a reference on the vnode.
1227 if (atomic_fcmpset_int(&vp->v_writecount, &n, 0)) {
1228 if ((vn_irflag_read(vp) & VIRF_TEXT_REF) != 0) {
1237 if (atomic_fcmpset_int(&vp->v_writecount, &n, n + 1)) {
1241 __assert_unreachable();
1244 static int __always_inline
1245 vop_stdadd_writecount_impl(struct vop_add_writecount_args *ap, bool handle_msync)
1248 struct mount *mp __diagused;
1257 VNPASS((mp->mnt_kern_flag & MNTK_NOMSYNC) == 0, vp);
1259 VNPASS((mp->mnt_kern_flag & MNTK_NOMSYNC) != 0, vp);
1264 n = atomic_load_int(&vp->v_writecount);
1266 if (__predict_false(n < 0)) {
1270 VNASSERT(n + ap->a_inc >= 0, vp,
1271 ("neg writecount increment %d + %d = %d", n, ap->a_inc,
1279 if (atomic_fcmpset_int(&vp->v_writecount, &n, n + ap->a_inc)) {
1283 __assert_unreachable();
1287 vop_stdadd_writecount(struct vop_add_writecount_args *ap)
1290 return (vop_stdadd_writecount_impl(ap, true));
1294 vop_stdadd_writecount_nomsync(struct vop_add_writecount_args *ap)
1297 return (vop_stdadd_writecount_impl(ap, false));
1301 vop_stdneed_inactive(struct vop_need_inactive_args *ap)
1308 vop_stdioctl(struct vop_ioctl_args *ap)
1315 switch (ap->a_command) {
1319 error = vn_lock(vp, LK_SHARED);
1322 if (vp->v_type == VREG)
1323 error = VOP_GETATTR(vp, &va, ap->a_cred);
1328 if (*offp < 0 || *offp >= va.va_size)
1330 else if (ap->a_command == FIOSEEKHOLE)
1344 * used to fill the vfs function table to get reasonable default return values.
1347 vfs_stdroot(struct mount *mp, int flags, struct vnode **vpp)
1350 return (EOPNOTSUPP);
1354 vfs_stdstatfs(struct mount *mp, struct statfs *sbp)
1357 return (EOPNOTSUPP);
1361 vfs_stdquotactl(struct mount *mp, int cmds, uid_t uid, void *arg, bool *mp_busy)
1363 return (EOPNOTSUPP);
1367 vfs_stdsync(struct mount *mp, int waitfor)
1369 struct vnode *vp, *mvp;
1371 int error, lockreq, allerror = 0;
1374 lockreq = LK_EXCLUSIVE | LK_INTERLOCK;
1375 if (waitfor != MNT_WAIT)
1376 lockreq |= LK_NOWAIT;
1378 * Force stale buffer cache information to be flushed.
1381 MNT_VNODE_FOREACH_ALL(vp, mp, mvp) {
1382 if (vp->v_bufobj.bo_dirty.bv_cnt == 0) {
1386 if ((error = vget(vp, lockreq)) != 0) {
1387 if (error == ENOENT) {
1388 MNT_VNODE_FOREACH_ALL_ABORT(mp, mvp);
1393 error = VOP_FSYNC(vp, waitfor, td);
1402 vfs_stdnosync(struct mount *mp, int waitfor)
1409 vop_stdcopy_file_range(struct vop_copy_file_range_args *ap)
1413 error = vn_generic_copy_file_range(ap->a_invp, ap->a_inoffp,
1414 ap->a_outvp, ap->a_outoffp, ap->a_lenp, ap->a_flags, ap->a_incred,
1415 ap->a_outcred, ap->a_fsizetd);
1420 vfs_stdvget(struct mount *mp, ino_t ino, int flags, struct vnode **vpp)
1423 return (EOPNOTSUPP);
1427 vfs_stdfhtovp(struct mount *mp, struct fid *fhp, int flags, struct vnode **vpp)
1430 return (EOPNOTSUPP);
1434 vfs_stdinit(struct vfsconf *vfsp)
1441 vfs_stduninit(struct vfsconf *vfsp)
1448 vfs_stdextattrctl(struct mount *mp, int cmd, struct vnode *filename_vp,
1449 int attrnamespace, const char *attrname)
1452 if (filename_vp != NULL)
1453 VOP_UNLOCK(filename_vp);
1454 return (EOPNOTSUPP);
1458 vfs_stdsysctl(struct mount *mp, fsctlop_t op, struct sysctl_req *req)
1461 return (EOPNOTSUPP);
1464 static vop_bypass_t *
1465 bp_by_off(struct vop_vector *vop, struct vop_generic_args *a)
1468 return (*(vop_bypass_t **)((char *)vop + a->a_desc->vdesc_vop_offset));
1472 vop_sigdefer(struct vop_vector *vop, struct vop_generic_args *a)
1477 bp = bp_by_off(vop, a);
1480 prev_stops = sigdeferstop(SIGDEFERSTOP_SILENT);
1482 sigallowstop(prev_stops);
1487 vop_stdstat(struct vop_stat_args *a)
1499 error = vop_stat_helper_pre(a);
1506 * Initialize defaults for new and unusual fields, so that file
1507 * systems which don't support these fields don't need to know
1510 vap->va_birthtime.tv_sec = -1;
1511 vap->va_birthtime.tv_nsec = 0;
1512 vap->va_fsid = VNOVAL;
1514 vap->va_rdev = NODEV;
1516 error = VOP_GETATTR(vp, vap, a->a_active_cred);
1521 * Zero the spare stat fields
1523 bzero(sb, sizeof *sb);
1526 * Copy from vattr table
1528 if (vap->va_fsid != VNOVAL)
1529 sb->st_dev = vap->va_fsid;
1531 sb->st_dev = vp->v_mount->mnt_stat.f_fsid.val[0];
1532 sb->st_ino = vap->va_fileid;
1533 mode = vap->va_mode;
1534 switch (vap->va_type) {
1561 sb->st_nlink = vap->va_nlink;
1562 sb->st_uid = vap->va_uid;
1563 sb->st_gid = vap->va_gid;
1564 sb->st_rdev = vap->va_rdev;
1565 if (vap->va_size > OFF_MAX) {
1569 sb->st_size = vap->va_size;
1570 sb->st_atim.tv_sec = vap->va_atime.tv_sec;
1571 sb->st_atim.tv_nsec = vap->va_atime.tv_nsec;
1572 sb->st_mtim.tv_sec = vap->va_mtime.tv_sec;
1573 sb->st_mtim.tv_nsec = vap->va_mtime.tv_nsec;
1574 sb->st_ctim.tv_sec = vap->va_ctime.tv_sec;
1575 sb->st_ctim.tv_nsec = vap->va_ctime.tv_nsec;
1576 sb->st_birthtim.tv_sec = vap->va_birthtime.tv_sec;
1577 sb->st_birthtim.tv_nsec = vap->va_birthtime.tv_nsec;
1580 * According to www.opengroup.org, the meaning of st_blksize is
1581 * "a filesystem-specific preferred I/O block size for this
1582 * object. In some filesystem types, this may vary from file
1584 * Use minimum/default of PAGE_SIZE (e.g. for VCHR).
1587 sb->st_blksize = max(PAGE_SIZE, vap->va_blocksize);
1588 sb->st_flags = vap->va_flags;
1589 sb->st_blocks = vap->va_bytes / S_BLKSIZE;
1590 sb->st_gen = vap->va_gen;
1592 return (vop_stat_helper_post(a, error));
1596 vop_stdread_pgcache(struct vop_read_pgcache_args *ap __unused)
1598 return (EJUSTRETURN);
1602 vop_stdvput_pair(struct vop_vput_pair_args *ap)
1604 struct vnode *dvp, *vp, **vpp;
1609 if (vpp != NULL && ap->a_unlock_vp && (vp = *vpp) != NULL)