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_bmap = vop_stdbmap,
121 .vop_close = VOP_NULL,
122 .vop_fsync = VOP_NULL,
123 .vop_stat = vop_stdstat,
124 .vop_fdatasync = vop_stdfdatasync,
125 .vop_getpages = vop_stdgetpages,
126 .vop_getpages_async = vop_stdgetpages_async,
127 .vop_getwritemount = vop_stdgetwritemount,
128 .vop_inactive = VOP_NULL,
129 .vop_need_inactive = vop_stdneed_inactive,
130 .vop_ioctl = vop_stdioctl,
131 .vop_kqfilter = vop_stdkqfilter,
132 .vop_islocked = vop_stdislocked,
133 .vop_lock1 = vop_stdlock,
134 .vop_lookup = vop_nolookup,
135 .vop_open = VOP_NULL,
136 .vop_pathconf = VOP_EINVAL,
137 .vop_poll = vop_nopoll,
138 .vop_putpages = vop_stdputpages,
139 .vop_readlink = VOP_EINVAL,
140 .vop_read_pgcache = vop_stdread_pgcache,
141 .vop_rename = vop_norename,
142 .vop_revoke = VOP_PANIC,
143 .vop_strategy = vop_nostrategy,
144 .vop_unlock = vop_stdunlock,
145 .vop_vptocnp = vop_stdvptocnp,
146 .vop_vptofh = vop_stdvptofh,
147 .vop_unp_bind = vop_stdunp_bind,
148 .vop_unp_connect = vop_stdunp_connect,
149 .vop_unp_detach = vop_stdunp_detach,
150 .vop_is_text = vop_stdis_text,
151 .vop_set_text = vop_stdset_text,
152 .vop_unset_text = vop_stdunset_text,
153 .vop_add_writecount = vop_stdadd_writecount,
154 .vop_copy_file_range = vop_stdcopy_file_range,
155 .vop_vput_pair = vop_stdvput_pair,
157 VFS_VOP_VECTOR_REGISTER(default_vnodeops);
160 * Series of placeholder functions for various error returns for
165 vop_eopnotsupp(struct vop_generic_args *ap)
168 printf("vop_notsupp[%s]\n", ap->a_desc->vdesc_name);
175 vop_ebadf(struct vop_generic_args *ap)
182 vop_enotty(struct vop_generic_args *ap)
189 vop_einval(struct vop_generic_args *ap)
196 vop_enoent(struct vop_generic_args *ap)
203 vop_eagain(struct vop_generic_args *ap)
210 vop_null(struct vop_generic_args *ap)
217 * Helper function to panic on some bad VOPs in some filesystems.
220 vop_panic(struct vop_generic_args *ap)
223 panic("filesystem goof: vop_panic[%s]", ap->a_desc->vdesc_name);
227 * vop_std<something> and vop_no<something> are default functions for use by
228 * filesystems that need the "default reasonable" implementation for a
229 * particular operation.
231 * The documentation for the operations they implement exists (if it exists)
232 * in the VOP_<SOMETHING>(9) manpage (all uppercase).
236 * Default vop for filesystems that do not support name lookup
239 vop_nolookup(struct vop_lookup_args *ap)
249 * Handle unlock and reference counting for arguments of vop_rename
250 * for filesystems that do not implement rename operation.
253 vop_norename(struct vop_rename_args *ap)
263 * Strategy routine for VFS devices that have none.
265 * BIO_ERROR and B_INVAL must be cleared prior to calling any strategy
266 * routine. Typically this is done for a BIO_READ strategy call.
267 * Typically B_INVAL is assumed to already be clear prior to a write
268 * and should not be cleared manually unless you just made the buffer
269 * invalid. BIO_ERROR should be cleared either way.
273 vop_nostrategy (struct vop_strategy_args *ap)
275 printf("No strategy for buffer at %p\n", ap->a_bp);
276 vn_printf(ap->a_vp, "vnode ");
277 ap->a_bp->b_ioflags |= BIO_ERROR;
278 ap->a_bp->b_error = EOPNOTSUPP;
284 get_next_dirent(struct vnode *vp, struct dirent **dpp, char *dirbuf,
285 int dirbuflen, off_t *off, char **cpos, int *len,
286 int *eofflag, struct thread *td)
293 KASSERT(VOP_ISLOCKED(vp), ("vp %p is not locked", vp));
294 KASSERT(vp->v_type == VDIR, ("vp %p is not a directory", vp));
297 iov.iov_base = dirbuf;
298 iov.iov_len = dirbuflen;
302 uio.uio_offset = *off;
303 uio.uio_resid = dirbuflen;
304 uio.uio_segflg = UIO_SYSSPACE;
305 uio.uio_rw = UIO_READ;
311 error = mac_vnode_check_readdir(td->td_ucred, vp);
314 error = VOP_READDIR(vp, &uio, td->td_ucred, eofflag,
319 *off = uio.uio_offset;
322 *len = (dirbuflen - uio.uio_resid);
328 dp = (struct dirent *)(*cpos);
329 reclen = dp->d_reclen;
332 /* check for malformed directory.. */
333 if (reclen < DIRENT_MINSIZE)
343 * Check if a named file exists in a given directory vnode.
346 dirent_exists(struct vnode *vp, const char *dirname, struct thread *td)
349 int error, eofflag, dirbuflen, len, found;
354 KASSERT(VOP_ISLOCKED(vp), ("vp %p is not locked", vp));
355 KASSERT(vp->v_type == VDIR, ("vp %p is not a directory", vp));
359 error = VOP_GETATTR(vp, &va, td->td_ucred);
363 dirbuflen = DEV_BSIZE;
364 if (dirbuflen < va.va_blocksize)
365 dirbuflen = va.va_blocksize;
366 dirbuf = (char *)malloc(dirbuflen, M_TEMP, M_WAITOK);
371 error = get_next_dirent(vp, &dp, dirbuf, dirbuflen, &off,
372 &cpos, &len, &eofflag, td);
376 if (dp->d_type != DT_WHT && dp->d_fileno != 0 &&
377 strcmp(dp->d_name, dirname) == 0) {
381 } while (len > 0 || !eofflag);
384 free(dirbuf, M_TEMP);
389 vop_stdaccess(struct vop_access_args *ap)
392 KASSERT((ap->a_accmode & ~(VEXEC | VWRITE | VREAD | VADMIN |
393 VAPPEND)) == 0, ("invalid bit in accmode"));
395 return (VOP_ACCESSX(ap->a_vp, ap->a_accmode, ap->a_cred, ap->a_td));
399 vop_stdaccessx(struct vop_accessx_args *ap)
402 accmode_t accmode = ap->a_accmode;
404 error = vfs_unixify_accmode(&accmode);
411 return (VOP_ACCESS(ap->a_vp, accmode, ap->a_cred, ap->a_td));
415 * Advisory record locking support
418 vop_stdadvlock(struct vop_advlock_args *ap)
428 * Provide atomicity of open(O_CREAT | O_EXCL | O_EXLOCK) for
429 * local filesystems. See vn_open_cred() for reciprocal part.
432 if (mp != NULL && (mp->mnt_flag & MNT_LOCAL) != 0 &&
433 ap->a_op == F_SETLK && (ap->a_flags & F_FIRSTOPEN) == 0) {
435 while ((vp->v_iflag & VI_FOPENING) != 0)
436 msleep(vp, VI_MTX(vp), PLOCK, "lockfo", 0);
440 if (ap->a_fl->l_whence == SEEK_END) {
442 * The NFSv4 server must avoid doing a vn_lock() here, since it
443 * can deadlock the nfsd threads, due to a LOR. Fortunately
444 * the NFSv4 server always uses SEEK_SET and this code is
445 * only required for the SEEK_END case.
447 vn_lock(vp, LK_SHARED | LK_RETRY);
448 error = VOP_GETATTR(vp, &vattr, curthread->td_ucred);
455 return (lf_advlock(ap, &(vp->v_lockf), vattr.va_size));
459 vop_stdadvlockasync(struct vop_advlockasync_args *ap)
466 if (ap->a_fl->l_whence == SEEK_END) {
467 /* The size argument is only needed for SEEK_END. */
468 vn_lock(vp, LK_SHARED | LK_RETRY);
469 error = VOP_GETATTR(vp, &vattr, curthread->td_ucred);
476 return (lf_advlockasync(ap, &(vp->v_lockf), vattr.va_size));
480 vop_stdadvlockpurge(struct vop_advlockpurge_args *ap)
485 lf_purgelocks(vp, &vp->v_lockf);
492 * Standard implementation of POSIX pathconf, to get information about limits
494 * Override per filesystem for the case where the filesystem has smaller
498 vop_stdpathconf(struct vop_pathconf_args *ap)
501 switch (ap->a_name) {
503 *ap->a_retval = _POSIX_ASYNCHRONOUS_IO;
506 *ap->a_retval = PATH_MAX;
508 case _PC_ACL_EXTENDED:
510 case _PC_CAP_PRESENT:
511 case _PC_INF_PRESENT:
512 case _PC_MAC_PRESENT:
522 * Standard lock, unlock and islocked functions.
525 vop_stdlock(struct vop_lock1_args *ap)
527 struct vnode *vp = ap->a_vp;
531 return (lockmgr_lock_flags(vp->v_vnlock, ap->a_flags,
532 &ilk->lock_object, ap->a_file, ap->a_line));
537 vop_stdunlock(struct vop_unlock_args *ap)
539 struct vnode *vp = ap->a_vp;
541 return (lockmgr_unlock(vp->v_vnlock));
546 vop_stdislocked(struct vop_islocked_args *ap)
549 return (lockstatus(ap->a_vp->v_vnlock));
553 * Variants of the above set.
556 * - shared locking disablement is not supported
557 * - v_vnlock pointer is not honored
560 vop_lock(struct vop_lock1_args *ap)
562 struct vnode *vp = ap->a_vp;
563 int flags = ap->a_flags;
566 MPASS(vp->v_vnlock == &vp->v_lock);
568 if (__predict_false((flags & ~(LK_TYPE_MASK | LK_NODDLKTREAT | LK_RETRY)) != 0))
571 switch (flags & LK_TYPE_MASK) {
573 return (lockmgr_slock(&vp->v_lock, flags, ap->a_file, ap->a_line));
575 return (lockmgr_xlock(&vp->v_lock, flags, ap->a_file, ap->a_line));
579 return (lockmgr_lock_flags(&vp->v_lock, flags,
580 &ilk->lock_object, ap->a_file, ap->a_line));
584 vop_unlock(struct vop_unlock_args *ap)
586 struct vnode *vp = ap->a_vp;
588 MPASS(vp->v_vnlock == &vp->v_lock);
590 return (lockmgr_unlock(&vp->v_lock));
594 vop_islocked(struct vop_islocked_args *ap)
596 struct vnode *vp = ap->a_vp;
598 MPASS(vp->v_vnlock == &vp->v_lock);
600 return (lockstatus(&vp->v_lock));
604 * Return true for select/poll.
607 vop_nopoll(struct vop_poll_args *ap)
610 if (ap->a_events & ~POLLSTANDARD)
612 return (ap->a_events & (POLLIN | POLLOUT | POLLRDNORM | POLLWRNORM));
616 * Implement poll for local filesystems that support it.
619 vop_stdpoll(struct vop_poll_args *ap)
621 if (ap->a_events & ~POLLSTANDARD)
622 return (vn_pollrecord(ap->a_vp, ap->a_td, ap->a_events));
623 return (ap->a_events & (POLLIN | POLLOUT | POLLRDNORM | POLLWRNORM));
627 * Return our mount point, as we will take charge of the writes.
630 vop_stdgetwritemount(struct vop_getwritemount_args *ap)
636 * Note that having a reference does not prevent forced unmount from
637 * setting ->v_mount to NULL after the lock gets released. This is of
638 * no consequence for typical consumers (most notably vn_start_write)
639 * since in this case the vnode is VIRF_DOOMED. Unmount might have
640 * progressed far enough that its completion is only delayed by the
641 * reference obtained here. The consumer only needs to concern itself
645 mp = vfs_ref_from_vp(vp);
651 * If the file system doesn't implement VOP_BMAP, then return sensible defaults:
652 * - Return the vnode's bufobj instead of any underlying device's bufobj
653 * - Calculate the physical block number as if there were equal size
654 * consecutive blocks, but
655 * - Report no contiguous runs of blocks.
658 vop_stdbmap(struct vop_bmap_args *ap)
661 if (ap->a_bop != NULL)
662 *ap->a_bop = &ap->a_vp->v_bufobj;
663 if (ap->a_bnp != NULL)
664 *ap->a_bnp = ap->a_bn * btodb(ap->a_vp->v_mount->mnt_stat.f_iosize);
665 if (ap->a_runp != NULL)
667 if (ap->a_runb != NULL)
673 vop_stdfsync(struct vop_fsync_args *ap)
676 return (vn_fsync_buf(ap->a_vp, ap->a_waitfor));
680 vop_stdfdatasync(struct vop_fdatasync_args *ap)
683 return (VOP_FSYNC(ap->a_vp, MNT_WAIT, ap->a_td));
687 vop_stdfdatasync_buf(struct vop_fdatasync_args *ap)
690 return (vn_fsync_buf(ap->a_vp, MNT_WAIT));
693 /* XXX Needs good comment and more info in the manpage (VOP_GETPAGES(9)). */
695 vop_stdgetpages(struct vop_getpages_args *ap)
698 return vnode_pager_generic_getpages(ap->a_vp, ap->a_m,
699 ap->a_count, ap->a_rbehind, ap->a_rahead, NULL, NULL);
703 vop_stdgetpages_async(struct vop_getpages_async_args *ap)
707 error = VOP_GETPAGES(ap->a_vp, ap->a_m, ap->a_count, ap->a_rbehind,
709 if (ap->a_iodone != NULL)
710 ap->a_iodone(ap->a_arg, ap->a_m, ap->a_count, error);
715 vop_stdkqfilter(struct vop_kqfilter_args *ap)
717 return vfs_kqfilter(ap);
720 /* XXX Needs good comment and more info in the manpage (VOP_PUTPAGES(9)). */
722 vop_stdputpages(struct vop_putpages_args *ap)
725 return vnode_pager_generic_putpages(ap->a_vp, ap->a_m, ap->a_count,
726 ap->a_sync, ap->a_rtvals);
730 vop_stdvptofh(struct vop_vptofh_args *ap)
736 vop_stdvptocnp(struct vop_vptocnp_args *ap)
738 struct vnode *vp = ap->a_vp;
739 struct vnode **dvp = ap->a_vpp;
741 char *buf = ap->a_buf;
742 size_t *buflen = ap->a_buflen;
744 int i, error, eofflag, dirbuflen, flags, locked, len, covered;
759 if (vp->v_type != VDIR)
762 error = VOP_GETATTR(vp, &va, cred);
767 locked = VOP_ISLOCKED(vp);
769 NDINIT_ATVP(&nd, LOOKUP, FOLLOW | LOCKSHARED | LOCKLEAF, UIO_SYSSPACE,
772 error = vn_open_cred(&nd, &flags, 0, VN_OPEN_NOAUDIT, cred, NULL);
774 vn_lock(vp, locked | LK_RETRY);
777 NDFREE(&nd, NDF_ONLY_PNBUF);
779 mvp = *dvp = nd.ni_vp;
781 if (vp->v_mount != (*dvp)->v_mount &&
782 ((*dvp)->v_vflag & VV_ROOT) &&
783 ((*dvp)->v_mount->mnt_flag & MNT_UNION)) {
784 *dvp = (*dvp)->v_mount->mnt_vnodecovered;
787 vn_close(mvp, FREAD, cred, td);
789 vn_lock(*dvp, LK_SHARED | LK_RETRY);
793 fileno = va.va_fileid;
795 dirbuflen = DEV_BSIZE;
796 if (dirbuflen < va.va_blocksize)
797 dirbuflen = va.va_blocksize;
798 dirbuf = (char *)malloc(dirbuflen, M_TEMP, M_WAITOK);
800 if ((*dvp)->v_type != VDIR) {
808 /* call VOP_READDIR of parent */
809 error = get_next_dirent(*dvp, &dp, dirbuf, dirbuflen, &off,
810 &cpos, &len, &eofflag, td);
814 if ((dp->d_type != DT_WHT) &&
815 (dp->d_fileno == fileno)) {
818 vn_lock(mvp, LK_SHARED | LK_RETRY);
819 if (dirent_exists(mvp, dp->d_name, td)) {
822 vn_lock(*dvp, LK_SHARED | LK_RETRY);
826 vn_lock(*dvp, LK_SHARED | LK_RETRY);
834 if (dp->d_namlen == 1 && dp->d_name[0] == '.') {
837 bcopy(dp->d_name, buf + i, dp->d_namlen);
842 } while (len > 0 || !eofflag);
846 free(dirbuf, M_TEMP);
856 vn_close(mvp, FREAD, cred, td);
858 vn_lock(vp, locked | LK_RETRY);
863 vop_stdallocate(struct vop_allocate_args *ap)
867 off_t maxfilesize = 0;
870 struct vattr vattr, *vap;
872 off_t fsize, len, cur, offset;
885 offset = *ap->a_offset;
887 error = VOP_GETATTR(vp, vap, ap->a_cred);
890 fsize = vap->va_size;
891 iosize = vap->va_blocksize;
893 iosize = BLKDEV_IOSIZE;
894 if (iosize > maxphys)
896 buf = malloc(iosize, M_TEMP, M_WAITOK);
900 * Check if the filesystem sets f_maxfilesize; if not use
901 * VOP_SETATTR to perform the check.
903 sfs = malloc(sizeof(struct statfs), M_STATFS, M_WAITOK);
904 error = VFS_STATFS(vp->v_mount, sfs, td);
906 maxfilesize = sfs->f_maxfilesize;
911 if (offset > maxfilesize || len > maxfilesize ||
912 offset + len > maxfilesize) {
918 if (offset + len > vap->va_size) {
920 * Test offset + len against the filesystem's maxfilesize.
923 vap->va_size = offset + len;
924 error = VOP_SETATTR(vp, vap, ap->a_cred);
928 vap->va_size = fsize;
929 error = VOP_SETATTR(vp, vap, ap->a_cred);
936 * Read and write back anything below the nominal file
937 * size. There's currently no way outside the filesystem
938 * to know whether this area is sparse or not.
941 if ((offset % iosize) != 0)
942 cur -= (offset % iosize);
945 if (offset < fsize) {
948 auio.uio_iov = &aiov;
950 auio.uio_offset = offset;
951 auio.uio_resid = cur;
952 auio.uio_segflg = UIO_SYSSPACE;
953 auio.uio_rw = UIO_READ;
955 error = VOP_READ(vp, &auio, ap->a_ioflag, ap->a_cred);
958 if (auio.uio_resid > 0) {
959 bzero(buf + cur - auio.uio_resid,
968 auio.uio_iov = &aiov;
970 auio.uio_offset = offset;
971 auio.uio_resid = cur;
972 auio.uio_segflg = UIO_SYSSPACE;
973 auio.uio_rw = UIO_WRITE;
976 error = VOP_WRITE(vp, &auio, ap->a_ioflag, ap->a_cred);
990 *ap->a_offset = offset;
996 vop_stdadvise(struct vop_advise_args *ap)
1000 daddr_t startn, endn;
1001 off_t bstart, bend, start, end;
1005 switch (ap->a_advice) {
1006 case POSIX_FADV_WILLNEED:
1008 * Do nothing for now. Filesystems should provide a
1009 * custom method which starts an asynchronous read of
1010 * the requested region.
1014 case POSIX_FADV_DONTNEED:
1016 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
1017 if (VN_IS_DOOMED(vp)) {
1023 * Round to block boundaries (and later possibly further to
1024 * page boundaries). Applications cannot reasonably be aware
1025 * of the boundaries, and the rounding must be to expand at
1026 * both extremities to cover enough. It still doesn't cover
1027 * read-ahead. For partial blocks, this gives unnecessary
1028 * discarding of buffers but is efficient enough since the
1029 * pages usually remain in VMIO for some time.
1031 bsize = vp->v_bufobj.bo_bsize;
1032 bstart = rounddown(ap->a_start, bsize);
1033 bend = roundup(ap->a_end, bsize);
1036 * Deactivate pages in the specified range from the backing VM
1037 * object. Pages that are resident in the buffer cache will
1038 * remain wired until their corresponding buffers are released
1041 if (vp->v_object != NULL) {
1042 start = trunc_page(bstart);
1043 end = round_page(bend);
1044 VM_OBJECT_RLOCK(vp->v_object);
1045 vm_object_page_noreuse(vp->v_object, OFF_TO_IDX(start),
1047 VM_OBJECT_RUNLOCK(vp->v_object);
1052 startn = bstart / bsize;
1053 endn = bend / bsize;
1054 error = bnoreuselist(&bo->bo_clean, bo, startn, endn);
1056 error = bnoreuselist(&bo->bo_dirty, bo, startn, endn);
1068 vop_stdunp_bind(struct vop_unp_bind_args *ap)
1071 ap->a_vp->v_unpcb = ap->a_unpcb;
1076 vop_stdunp_connect(struct vop_unp_connect_args *ap)
1079 *ap->a_unpcb = ap->a_vp->v_unpcb;
1084 vop_stdunp_detach(struct vop_unp_detach_args *ap)
1087 ap->a_vp->v_unpcb = NULL;
1092 vop_stdis_text(struct vop_is_text_args *ap)
1095 return (ap->a_vp->v_writecount < 0);
1099 vop_stdset_text(struct vop_set_text_args *ap)
1108 * Avoid the interlock if execs are already present.
1110 n = atomic_load_int(&vp->v_writecount);
1115 if (atomic_fcmpset_int(&vp->v_writecount, &n, n - 1)) {
1121 if (vp->v_writecount > 0) {
1125 * If requested by fs, keep a use reference to the
1126 * vnode until the last text reference is released.
1129 if (mp != NULL && (mp->mnt_kern_flag & MNTK_TEXT_REFS) != 0 &&
1130 vp->v_writecount == 0) {
1131 VNPASS((vp->v_iflag & VI_TEXT_REF) == 0, vp);
1132 vp->v_iflag |= VI_TEXT_REF;
1136 atomic_subtract_int(&vp->v_writecount, 1);
1144 vop_stdunset_text(struct vop_unset_text_args *ap)
1153 * Avoid the interlock if this is not the last exec.
1155 n = atomic_load_int(&vp->v_writecount);
1160 if (atomic_fcmpset_int(&vp->v_writecount, &n, n + 1)) {
1167 if (vp->v_writecount < 0) {
1168 if ((vp->v_iflag & VI_TEXT_REF) != 0 &&
1169 vp->v_writecount == -1) {
1171 vp->v_iflag &= ~VI_TEXT_REF;
1173 atomic_add_int(&vp->v_writecount, 1);
1185 vop_stdadd_writecount(struct vop_add_writecount_args *ap)
1192 VI_LOCK_FLAGS(vp, MTX_DUPOK);
1193 if (vp->v_writecount < 0) {
1196 VNASSERT(vp->v_writecount + ap->a_inc >= 0, vp,
1197 ("neg writecount increment %d", ap->a_inc));
1198 if (vp->v_writecount == 0) {
1200 if (mp != NULL && (mp->mnt_kern_flag & MNTK_NOMSYNC) == 0)
1203 vp->v_writecount += ap->a_inc;
1211 vop_stdneed_inactive(struct vop_need_inactive_args *ap)
1218 vop_stdioctl(struct vop_ioctl_args *ap)
1225 switch (ap->a_command) {
1229 error = vn_lock(vp, LK_SHARED);
1232 if (vp->v_type == VREG)
1233 error = VOP_GETATTR(vp, &va, ap->a_cred);
1238 if (*offp < 0 || *offp >= va.va_size)
1240 else if (ap->a_command == FIOSEEKHOLE)
1254 * used to fill the vfs function table to get reasonable default return values.
1257 vfs_stdroot(struct mount *mp, int flags, struct vnode **vpp)
1260 return (EOPNOTSUPP);
1264 vfs_stdstatfs(struct mount *mp, struct statfs *sbp)
1267 return (EOPNOTSUPP);
1271 vfs_stdquotactl(struct mount *mp, int cmds, uid_t uid, void *arg)
1274 return (EOPNOTSUPP);
1278 vfs_stdsync(struct mount *mp, int waitfor)
1280 struct vnode *vp, *mvp;
1282 int error, lockreq, allerror = 0;
1285 lockreq = LK_EXCLUSIVE | LK_INTERLOCK;
1286 if (waitfor != MNT_WAIT)
1287 lockreq |= LK_NOWAIT;
1289 * Force stale buffer cache information to be flushed.
1292 MNT_VNODE_FOREACH_ALL(vp, mp, mvp) {
1293 if (vp->v_bufobj.bo_dirty.bv_cnt == 0) {
1297 if ((error = vget(vp, lockreq)) != 0) {
1298 if (error == ENOENT) {
1299 MNT_VNODE_FOREACH_ALL_ABORT(mp, mvp);
1304 error = VOP_FSYNC(vp, waitfor, td);
1313 vfs_stdnosync(struct mount *mp, int waitfor)
1320 vop_stdcopy_file_range(struct vop_copy_file_range_args *ap)
1324 error = vn_generic_copy_file_range(ap->a_invp, ap->a_inoffp,
1325 ap->a_outvp, ap->a_outoffp, ap->a_lenp, ap->a_flags, ap->a_incred,
1326 ap->a_outcred, ap->a_fsizetd);
1331 vfs_stdvget(struct mount *mp, ino_t ino, int flags, struct vnode **vpp)
1334 return (EOPNOTSUPP);
1338 vfs_stdfhtovp(struct mount *mp, struct fid *fhp, int flags, struct vnode **vpp)
1341 return (EOPNOTSUPP);
1345 vfs_stdinit(struct vfsconf *vfsp)
1352 vfs_stduninit(struct vfsconf *vfsp)
1359 vfs_stdextattrctl(struct mount *mp, int cmd, struct vnode *filename_vp,
1360 int attrnamespace, const char *attrname)
1363 if (filename_vp != NULL)
1364 VOP_UNLOCK(filename_vp);
1365 return (EOPNOTSUPP);
1369 vfs_stdsysctl(struct mount *mp, fsctlop_t op, struct sysctl_req *req)
1372 return (EOPNOTSUPP);
1375 static vop_bypass_t *
1376 bp_by_off(struct vop_vector *vop, struct vop_generic_args *a)
1379 return (*(vop_bypass_t **)((char *)vop + a->a_desc->vdesc_vop_offset));
1383 vop_sigdefer(struct vop_vector *vop, struct vop_generic_args *a)
1388 bp = bp_by_off(vop, a);
1391 prev_stops = sigdeferstop(SIGDEFERSTOP_SILENT);
1393 sigallowstop(prev_stops);
1398 vop_stdstat(struct vop_stat_args *a)
1410 error = vop_stat_helper_pre(a);
1417 * Initialize defaults for new and unusual fields, so that file
1418 * systems which don't support these fields don't need to know
1421 vap->va_birthtime.tv_sec = -1;
1422 vap->va_birthtime.tv_nsec = 0;
1423 vap->va_fsid = VNOVAL;
1424 vap->va_rdev = NODEV;
1426 error = VOP_GETATTR(vp, vap, a->a_active_cred);
1431 * Zero the spare stat fields
1433 bzero(sb, sizeof *sb);
1436 * Copy from vattr table
1438 if (vap->va_fsid != VNOVAL)
1439 sb->st_dev = vap->va_fsid;
1441 sb->st_dev = vp->v_mount->mnt_stat.f_fsid.val[0];
1442 sb->st_ino = vap->va_fileid;
1443 mode = vap->va_mode;
1444 switch (vap->va_type) {
1471 sb->st_nlink = vap->va_nlink;
1472 sb->st_uid = vap->va_uid;
1473 sb->st_gid = vap->va_gid;
1474 sb->st_rdev = vap->va_rdev;
1475 if (vap->va_size > OFF_MAX) {
1479 sb->st_size = vap->va_size;
1480 sb->st_atim.tv_sec = vap->va_atime.tv_sec;
1481 sb->st_atim.tv_nsec = vap->va_atime.tv_nsec;
1482 sb->st_mtim.tv_sec = vap->va_mtime.tv_sec;
1483 sb->st_mtim.tv_nsec = vap->va_mtime.tv_nsec;
1484 sb->st_ctim.tv_sec = vap->va_ctime.tv_sec;
1485 sb->st_ctim.tv_nsec = vap->va_ctime.tv_nsec;
1486 sb->st_birthtim.tv_sec = vap->va_birthtime.tv_sec;
1487 sb->st_birthtim.tv_nsec = vap->va_birthtime.tv_nsec;
1490 * According to www.opengroup.org, the meaning of st_blksize is
1491 * "a filesystem-specific preferred I/O block size for this
1492 * object. In some filesystem types, this may vary from file
1494 * Use minimum/default of PAGE_SIZE (e.g. for VCHR).
1497 sb->st_blksize = max(PAGE_SIZE, vap->va_blocksize);
1498 sb->st_flags = vap->va_flags;
1499 sb->st_blocks = vap->va_bytes / S_BLKSIZE;
1500 sb->st_gen = vap->va_gen;
1502 return (vop_stat_helper_post(a, error));
1506 vop_stdread_pgcache(struct vop_read_pgcache_args *ap __unused)
1508 return (EJUSTRETURN);
1512 vop_stdvput_pair(struct vop_vput_pair_args *ap)
1514 struct vnode *dvp, *vp, **vpp;
1519 if (vpp != NULL && ap->a_unlock_vp && (vp = *vpp) != NULL)