2 * Copyright (c) 1982, 1986, 1989, 1993
3 * The Regents of the University of California. All rights reserved.
4 * (c) UNIX System Laboratories, Inc.
5 * All or some portions of this file are derived from material licensed
6 * to the University of California by American Telephone and Telegraph
7 * Co. or Unix System Laboratories, Inc. and are reproduced herein with
8 * the permission of UNIX System Laboratories, Inc.
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 * 4. 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
34 * @(#)vfs_vnops.c 8.2 (Berkeley) 1/21/94
37 #include <sys/cdefs.h>
38 __FBSDID("$FreeBSD$");
40 #include <sys/param.h>
41 #include <sys/systm.h>
42 #include <sys/fcntl.h>
48 #include <sys/limits.h>
50 #include <sys/mount.h>
51 #include <sys/mutex.h>
52 #include <sys/namei.h>
53 #include <sys/vnode.h>
56 #include <sys/filio.h>
57 #include <sys/resourcevar.h>
59 #include <sys/ttycom.h>
61 #include <sys/syslog.h>
62 #include <sys/unistd.h>
64 #include <security/mac/mac_framework.h>
67 #include <vm/vm_object.h>
69 static fo_rdwr_t vn_read;
70 static fo_rdwr_t vn_write;
71 static fo_truncate_t vn_truncate;
72 static fo_ioctl_t vn_ioctl;
73 static fo_poll_t vn_poll;
74 static fo_kqfilter_t vn_kqfilter;
75 static fo_stat_t vn_statfile;
76 static fo_close_t vn_closefile;
78 struct fileops vnops = {
81 .fo_truncate = vn_truncate,
84 .fo_kqfilter = vn_kqfilter,
85 .fo_stat = vn_statfile,
86 .fo_close = vn_closefile,
87 .fo_flags = DFLAG_PASSABLE | DFLAG_SEEKABLE
91 vn_open(ndp, flagp, cmode, fp)
92 struct nameidata *ndp;
96 struct thread *td = ndp->ni_cnd.cn_thread;
98 return (vn_open_cred(ndp, flagp, cmode, 0, td->td_ucred, fp));
102 * Common code for vnode open operations.
103 * Check permissions, and call the VOP_OPEN or VOP_CREATE routine.
105 * Note that this does NOT free nameidata for the successful case,
106 * due to the NDINIT being done elsewhere.
109 vn_open_cred(struct nameidata *ndp, int *flagp, int cmode, u_int vn_open_flags,
110 struct ucred *cred, struct file *fp)
114 struct thread *td = ndp->ni_cnd.cn_thread;
116 struct vattr *vap = &vat;
119 int vfslocked, mpsafe;
121 mpsafe = ndp->ni_cnd.cn_flags & MPSAFE;
125 if (fmode & O_CREAT) {
126 ndp->ni_cnd.cn_nameiop = CREATE;
127 ndp->ni_cnd.cn_flags = ISOPEN | LOCKPARENT | LOCKLEAF |
129 if ((fmode & O_EXCL) == 0 && (fmode & O_NOFOLLOW) == 0)
130 ndp->ni_cnd.cn_flags |= FOLLOW;
131 if (!(vn_open_flags & VN_OPEN_NOAUDIT))
132 ndp->ni_cnd.cn_flags |= AUDITVNODE1;
134 if ((error = namei(ndp)) != 0)
136 vfslocked = NDHASGIANT(ndp);
138 ndp->ni_cnd.cn_flags &= ~MPSAFE;
139 if (ndp->ni_vp == NULL) {
142 vap->va_mode = cmode;
144 vap->va_vaflags |= VA_EXCLUSIVE;
145 if (vn_start_write(ndp->ni_dvp, &mp, V_NOWAIT) != 0) {
146 NDFREE(ndp, NDF_ONLY_PNBUF);
148 VFS_UNLOCK_GIANT(vfslocked);
149 if ((error = vn_start_write(NULL, &mp,
150 V_XSLEEP | PCATCH)) != 0)
155 error = mac_vnode_check_create(cred, ndp->ni_dvp,
159 error = VOP_CREATE(ndp->ni_dvp, &ndp->ni_vp,
162 vn_finished_write(mp);
164 VFS_UNLOCK_GIANT(vfslocked);
165 NDFREE(ndp, NDF_ONLY_PNBUF);
171 if (ndp->ni_dvp == ndp->ni_vp)
177 if (fmode & O_EXCL) {
184 ndp->ni_cnd.cn_nameiop = LOOKUP;
185 ndp->ni_cnd.cn_flags = ISOPEN |
186 ((fmode & O_NOFOLLOW) ? NOFOLLOW : FOLLOW) |
188 if (!(fmode & FWRITE))
189 ndp->ni_cnd.cn_flags |= LOCKSHARED;
190 if (!(vn_open_flags & VN_OPEN_NOAUDIT))
191 ndp->ni_cnd.cn_flags |= AUDITVNODE1;
192 if ((error = namei(ndp)) != 0)
195 ndp->ni_cnd.cn_flags &= ~MPSAFE;
196 vfslocked = NDHASGIANT(ndp);
199 if (vp->v_type == VLNK) {
203 if (vp->v_type == VSOCK) {
208 if (fmode & (FWRITE | O_TRUNC)) {
209 if (vp->v_type == VDIR) {
219 if ((fmode & O_APPEND) && (fmode & FWRITE))
222 error = mac_vnode_check_open(cred, vp, accmode);
226 if ((fmode & O_CREAT) == 0) {
227 if (accmode & VWRITE) {
228 error = vn_writechk(vp);
233 error = VOP_ACCESS(vp, accmode, cred, td);
238 if ((error = VOP_OPEN(vp, fmode, cred, td, fp)) != 0)
244 ASSERT_VOP_LOCKED(vp, "vn_open_cred");
246 VFS_UNLOCK_GIANT(vfslocked);
249 NDFREE(ndp, NDF_ONLY_PNBUF);
251 VFS_UNLOCK_GIANT(vfslocked);
258 * Check for write permissions on the specified vnode.
259 * Prototype text segments cannot be written.
263 register struct vnode *vp;
266 ASSERT_VOP_LOCKED(vp, "vn_writechk");
268 * If there's shared text associated with
269 * the vnode, try to free it up once. If
270 * we fail, we can't allow writing.
272 if (vp->v_vflag & VV_TEXT)
282 vn_close(vp, flags, file_cred, td)
283 register struct vnode *vp;
285 struct ucred *file_cred;
289 int error, lock_flags;
291 if (!(flags & FWRITE) && vp->v_mount != NULL &&
292 vp->v_mount->mnt_kern_flag & MNTK_EXTENDED_SHARED)
293 lock_flags = LK_SHARED;
295 lock_flags = LK_EXCLUSIVE;
297 VFS_ASSERT_GIANT(vp->v_mount);
299 vn_start_write(vp, &mp, V_WAIT);
300 vn_lock(vp, lock_flags | LK_RETRY);
301 if (flags & FWRITE) {
302 VNASSERT(vp->v_writecount > 0, vp,
303 ("vn_close: negative writecount"));
306 error = VOP_CLOSE(vp, flags, file_cred, td);
308 vn_finished_write(mp);
313 * Heuristic to detect sequential operation.
316 sequential_heuristic(struct uio *uio, struct file *fp)
319 if (atomic_load_acq_int(&(fp->f_flag)) & FRDAHEAD)
320 return (fp->f_seqcount << IO_SEQSHIFT);
323 * Offset 0 is handled specially. open() sets f_seqcount to 1 so
324 * that the first I/O is normally considered to be slightly
325 * sequential. Seeking to offset 0 doesn't change sequentiality
326 * unless previous seeks have reduced f_seqcount to 0, in which
327 * case offset 0 is not special.
329 if ((uio->uio_offset == 0 && fp->f_seqcount > 0) ||
330 uio->uio_offset == fp->f_nextoff) {
332 * f_seqcount is in units of fixed-size blocks so that it
333 * depends mainly on the amount of sequential I/O and not
334 * much on the number of sequential I/O's. The fixed size
335 * of 16384 is hard-coded here since it is (not quite) just
336 * a magic size that works well here. This size is more
337 * closely related to the best I/O size for real disks than
338 * to any block size used by software.
340 fp->f_seqcount += howmany(uio->uio_resid, 16384);
341 if (fp->f_seqcount > IO_SEQMAX)
342 fp->f_seqcount = IO_SEQMAX;
343 return (fp->f_seqcount << IO_SEQSHIFT);
346 /* Not sequential. Quickly draw-down sequentiality. */
347 if (fp->f_seqcount > 1)
355 * Package up an I/O request on a vnode into a uio and do it.
358 vn_rdwr(rw, vp, base, len, offset, segflg, ioflg, active_cred, file_cred,
367 struct ucred *active_cred;
368 struct ucred *file_cred;
376 int error, lock_flags;
378 VFS_ASSERT_GIANT(vp->v_mount);
380 if ((ioflg & IO_NODELOCKED) == 0) {
382 if (rw == UIO_WRITE) {
383 if (vp->v_type != VCHR &&
384 (error = vn_start_write(vp, &mp, V_WAIT | PCATCH))
387 if (MNT_SHARED_WRITES(mp) ||
388 ((mp == NULL) && MNT_SHARED_WRITES(vp->v_mount))) {
389 lock_flags = LK_SHARED;
391 lock_flags = LK_EXCLUSIVE;
393 vn_lock(vp, lock_flags | LK_RETRY);
395 vn_lock(vp, LK_SHARED | LK_RETRY);
398 ASSERT_VOP_LOCKED(vp, "IO_NODELOCKED with no vp lock held");
399 auio.uio_iov = &aiov;
401 aiov.iov_base = base;
403 auio.uio_resid = len;
404 auio.uio_offset = offset;
405 auio.uio_segflg = segflg;
410 if ((ioflg & IO_NOMACCHECK) == 0) {
412 error = mac_vnode_check_read(active_cred, file_cred,
415 error = mac_vnode_check_write(active_cred, file_cred,
425 error = VOP_READ(vp, &auio, ioflg, cred);
427 error = VOP_WRITE(vp, &auio, ioflg, cred);
430 *aresid = auio.uio_resid;
432 if (auio.uio_resid && error == 0)
434 if ((ioflg & IO_NODELOCKED) == 0) {
435 if (rw == UIO_WRITE && vp->v_type != VCHR)
436 vn_finished_write(mp);
443 * Package up an I/O request on a vnode into a uio and do it. The I/O
444 * request is split up into smaller chunks and we try to avoid saturating
445 * the buffer cache while potentially holding a vnode locked, so we
446 * check bwillwrite() before calling vn_rdwr(). We also call uio_yield()
447 * to give other processes a chance to lock the vnode (either other processes
448 * core'ing the same binary, or unrelated processes scanning the directory).
451 vn_rdwr_inchunks(rw, vp, base, len, offset, segflg, ioflg, active_cred,
452 file_cred, aresid, td)
460 struct ucred *active_cred;
461 struct ucred *file_cred;
468 VFS_ASSERT_GIANT(vp->v_mount);
474 * Force `offset' to a multiple of MAXBSIZE except possibly
475 * for the first chunk, so that filesystems only need to
476 * write full blocks except possibly for the first and last
479 chunk = MAXBSIZE - (uoff_t)offset % MAXBSIZE;
483 if (rw != UIO_READ && vp->v_type == VREG)
486 error = vn_rdwr(rw, vp, base, chunk, offset, segflg,
487 ioflg, active_cred, file_cred, &iaresid, td);
488 len -= chunk; /* aresid calc already includes length */
492 base = (char *)base + chunk;
496 *aresid = len + iaresid;
501 * File table vnode read routine.
504 vn_read(fp, uio, active_cred, flags, td)
507 struct ucred *active_cred;
514 int advice, vfslocked;
516 KASSERT(uio->uio_td == td, ("uio_td %p is not td %p",
521 if (fp->f_flag & FNONBLOCK)
523 if (fp->f_flag & O_DIRECT)
525 advice = POSIX_FADV_NORMAL;
526 vfslocked = VFS_LOCK_GIANT(vp->v_mount);
528 * According to McKusick the vn lock was protecting f_offset here.
529 * It is now protected by the FOFFSET_LOCKED flag.
531 if ((flags & FOF_OFFSET) == 0 || fp->f_advice != NULL) {
532 mtxp = mtx_pool_find(mtxpool_sleep, fp);
534 if ((flags & FOF_OFFSET) == 0) {
535 while (fp->f_vnread_flags & FOFFSET_LOCKED) {
536 fp->f_vnread_flags |= FOFFSET_LOCK_WAITING;
537 msleep(&fp->f_vnread_flags, mtxp, PUSER -1,
538 "vnread offlock", 0);
540 fp->f_vnread_flags |= FOFFSET_LOCKED;
541 uio->uio_offset = fp->f_offset;
543 if (fp->f_advice != NULL &&
544 uio->uio_offset >= fp->f_advice->fa_start &&
545 uio->uio_offset + uio->uio_resid <= fp->f_advice->fa_end)
546 advice = fp->f_advice->fa_advice;
549 vn_lock(vp, LK_SHARED | LK_RETRY);
552 case POSIX_FADV_NORMAL:
553 case POSIX_FADV_SEQUENTIAL:
554 ioflag |= sequential_heuristic(uio, fp);
556 case POSIX_FADV_RANDOM:
557 /* Disable read-ahead for random I/O. */
559 case POSIX_FADV_NOREUSE:
561 * Request the underlying FS to discard the buffers
562 * and pages after the I/O is complete.
569 error = mac_vnode_check_read(active_cred, fp->f_cred, vp);
572 error = VOP_READ(vp, uio, ioflag, fp->f_cred);
573 if ((flags & FOF_OFFSET) == 0) {
574 fp->f_offset = uio->uio_offset;
576 if (fp->f_vnread_flags & FOFFSET_LOCK_WAITING)
577 wakeup(&fp->f_vnread_flags);
578 fp->f_vnread_flags = 0;
581 fp->f_nextoff = uio->uio_offset;
583 VFS_UNLOCK_GIANT(vfslocked);
588 * File table vnode write routine.
591 vn_write(fp, uio, active_cred, flags, td)
594 struct ucred *active_cred;
600 int error, ioflag, lock_flags;
602 int advice, vfslocked;
604 KASSERT(uio->uio_td == td, ("uio_td %p is not td %p",
607 vfslocked = VFS_LOCK_GIANT(vp->v_mount);
608 if (vp->v_type == VREG)
611 if (vp->v_type == VREG && (fp->f_flag & O_APPEND))
613 if (fp->f_flag & FNONBLOCK)
615 if (fp->f_flag & O_DIRECT)
617 if ((fp->f_flag & O_FSYNC) ||
618 (vp->v_mount && (vp->v_mount->mnt_flag & MNT_SYNCHRONOUS)))
621 if (vp->v_type != VCHR &&
622 (error = vn_start_write(vp, &mp, V_WAIT | PCATCH)) != 0)
625 if ((MNT_SHARED_WRITES(mp) ||
626 ((mp == NULL) && MNT_SHARED_WRITES(vp->v_mount))) &&
627 (flags & FOF_OFFSET) != 0) {
628 lock_flags = LK_SHARED;
630 lock_flags = LK_EXCLUSIVE;
633 vn_lock(vp, lock_flags | LK_RETRY);
634 if ((flags & FOF_OFFSET) == 0)
635 uio->uio_offset = fp->f_offset;
636 advice = POSIX_FADV_NORMAL;
637 if (fp->f_advice != NULL) {
638 mtxp = mtx_pool_find(mtxpool_sleep, fp);
640 if (fp->f_advice != NULL &&
641 uio->uio_offset >= fp->f_advice->fa_start &&
642 uio->uio_offset + uio->uio_resid <= fp->f_advice->fa_end)
643 advice = fp->f_advice->fa_advice;
647 case POSIX_FADV_NORMAL:
648 case POSIX_FADV_SEQUENTIAL:
649 ioflag |= sequential_heuristic(uio, fp);
651 case POSIX_FADV_RANDOM:
652 /* XXX: Is this correct? */
654 case POSIX_FADV_NOREUSE:
656 * Request the underlying FS to discard the buffers
657 * and pages after the I/O is complete.
664 error = mac_vnode_check_write(active_cred, fp->f_cred, vp);
667 error = VOP_WRITE(vp, uio, ioflag, fp->f_cred);
668 if ((flags & FOF_OFFSET) == 0)
669 fp->f_offset = uio->uio_offset;
670 fp->f_nextoff = uio->uio_offset;
672 if (vp->v_type != VCHR)
673 vn_finished_write(mp);
675 VFS_UNLOCK_GIANT(vfslocked);
680 * File table truncate routine.
683 vn_truncate(fp, length, active_cred, td)
686 struct ucred *active_cred;
696 vfslocked = VFS_LOCK_GIANT(vp->v_mount);
697 error = vn_start_write(vp, &mp, V_WAIT | PCATCH);
699 VFS_UNLOCK_GIANT(vfslocked);
702 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
703 if (vp->v_type == VDIR) {
708 error = mac_vnode_check_write(active_cred, fp->f_cred, vp);
712 error = vn_writechk(vp);
715 vattr.va_size = length;
716 error = VOP_SETATTR(vp, &vattr, fp->f_cred);
720 vn_finished_write(mp);
721 VFS_UNLOCK_GIANT(vfslocked);
726 * File table vnode stat routine.
729 vn_statfile(fp, sb, active_cred, td)
732 struct ucred *active_cred;
735 struct vnode *vp = fp->f_vnode;
739 vfslocked = VFS_LOCK_GIANT(vp->v_mount);
740 vn_lock(vp, LK_SHARED | LK_RETRY);
741 error = vn_stat(vp, sb, active_cred, fp->f_cred, td);
743 VFS_UNLOCK_GIANT(vfslocked);
749 * Stat a vnode; implementation for the stat syscall
752 vn_stat(vp, sb, active_cred, file_cred, td)
754 register struct stat *sb;
755 struct ucred *active_cred;
756 struct ucred *file_cred;
760 register struct vattr *vap;
765 error = mac_vnode_check_stat(active_cred, file_cred, vp);
773 * Initialize defaults for new and unusual fields, so that file
774 * systems which don't support these fields don't need to know
777 vap->va_birthtime.tv_sec = -1;
778 vap->va_birthtime.tv_nsec = 0;
779 vap->va_fsid = VNOVAL;
780 vap->va_rdev = NODEV;
782 error = VOP_GETATTR(vp, vap, active_cred);
787 * Zero the spare stat fields
789 bzero(sb, sizeof *sb);
792 * Copy from vattr table
794 if (vap->va_fsid != VNOVAL)
795 sb->st_dev = vap->va_fsid;
797 sb->st_dev = vp->v_mount->mnt_stat.f_fsid.val[0];
798 sb->st_ino = vap->va_fileid;
800 switch (vap->va_type) {
826 sb->st_nlink = vap->va_nlink;
827 sb->st_uid = vap->va_uid;
828 sb->st_gid = vap->va_gid;
829 sb->st_rdev = vap->va_rdev;
830 if (vap->va_size > OFF_MAX)
832 sb->st_size = vap->va_size;
833 sb->st_atimespec = vap->va_atime;
834 sb->st_mtimespec = vap->va_mtime;
835 sb->st_ctimespec = vap->va_ctime;
836 sb->st_birthtimespec = vap->va_birthtime;
839 * According to www.opengroup.org, the meaning of st_blksize is
840 * "a filesystem-specific preferred I/O block size for this
841 * object. In some filesystem types, this may vary from file
843 * Use miminum/default of PAGE_SIZE (e.g. for VCHR).
846 sb->st_blksize = max(PAGE_SIZE, vap->va_blocksize);
848 sb->st_flags = vap->va_flags;
849 if (priv_check(td, PRIV_VFS_GENERATION))
852 sb->st_gen = vap->va_gen;
854 sb->st_blocks = vap->va_bytes / S_BLKSIZE;
859 * File table vnode ioctl routine.
862 vn_ioctl(fp, com, data, active_cred, td)
866 struct ucred *active_cred;
869 struct vnode *vp = fp->f_vnode;
874 vfslocked = VFS_LOCK_GIANT(vp->v_mount);
876 switch (vp->v_type) {
879 if (com == FIONREAD) {
880 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
881 error = VOP_GETATTR(vp, &vattr, active_cred);
884 *(int *)data = vattr.va_size - fp->f_offset;
886 if (com == FIONBIO || com == FIOASYNC) /* XXX */
889 error = VOP_IOCTL(vp, com, data, fp->f_flag,
896 VFS_UNLOCK_GIANT(vfslocked);
901 * File table vnode poll routine.
904 vn_poll(fp, events, active_cred, td)
907 struct ucred *active_cred;
915 vfslocked = VFS_LOCK_GIANT(vp->v_mount);
917 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
918 error = mac_vnode_check_poll(active_cred, fp->f_cred, vp);
923 error = VOP_POLL(vp, events, fp->f_cred, td);
924 VFS_UNLOCK_GIANT(vfslocked);
929 * Acquire the requested lock and then check for validity. LK_RETRY
930 * permits vn_lock to return doomed vnodes.
933 _vn_lock(struct vnode *vp, int flags, char *file, int line)
937 VNASSERT((flags & LK_TYPE_MASK) != 0, vp,
938 ("vn_lock called with no locktype."));
940 #ifdef DEBUG_VFS_LOCKS
941 KASSERT(vp->v_holdcnt != 0,
942 ("vn_lock %p: zero hold count", vp));
944 error = VOP_LOCK1(vp, flags, file, line);
945 flags &= ~LK_INTERLOCK; /* Interlock is always dropped. */
946 KASSERT((flags & LK_RETRY) == 0 || error == 0,
947 ("LK_RETRY set with incompatible flags (0x%x) or an error occured (%d)",
950 * Callers specify LK_RETRY if they wish to get dead vnodes.
951 * If RETRY is not set, we return ENOENT instead.
953 if (error == 0 && vp->v_iflag & VI_DOOMED &&
954 (flags & LK_RETRY) == 0) {
959 } while (flags & LK_RETRY && error != 0);
964 * File table vnode close routine.
978 vfslocked = VFS_LOCK_GIANT(vp->v_mount);
979 if (fp->f_type == DTYPE_VNODE && fp->f_flag & FHASLOCK) {
980 lf.l_whence = SEEK_SET;
984 (void) VOP_ADVLOCK(vp, fp, F_UNLCK, &lf, F_FLOCK);
987 fp->f_ops = &badfileops;
989 error = vn_close(vp, fp->f_flag, fp->f_cred, td);
990 VFS_UNLOCK_GIANT(vfslocked);
995 * Preparing to start a filesystem write operation. If the operation is
996 * permitted, then we bump the count of operations in progress and
997 * proceed. If a suspend request is in progress, we wait until the
998 * suspension is over, and then proceed.
1001 vn_start_write(vp, mpp, flags)
1011 * If a vnode is provided, get and return the mount point that
1012 * to which it will write.
1015 if ((error = VOP_GETWRITEMOUNT(vp, mpp)) != 0) {
1017 if (error != EOPNOTSUPP)
1022 if ((mp = *mpp) == NULL)
1026 * VOP_GETWRITEMOUNT() returns with the mp refcount held through
1028 * As long as a vnode is not provided we need to acquire a
1029 * refcount for the provided mountpoint too, in order to
1030 * emulate a vfs_ref().
1037 * Check on status of suspension.
1039 if ((curthread->td_pflags & TDP_IGNSUSP) == 0 ||
1040 mp->mnt_susp_owner != curthread) {
1041 while ((mp->mnt_kern_flag & MNTK_SUSPEND) != 0) {
1042 if (flags & V_NOWAIT) {
1043 error = EWOULDBLOCK;
1046 error = msleep(&mp->mnt_flag, MNT_MTX(mp),
1047 (PUSER - 1) | (flags & PCATCH), "suspfs", 0);
1052 if (flags & V_XSLEEP)
1054 mp->mnt_writeopcount++;
1056 if (error != 0 || (flags & V_XSLEEP) != 0)
1063 * Secondary suspension. Used by operations such as vop_inactive
1064 * routines that are needed by the higher level functions. These
1065 * are allowed to proceed until all the higher level functions have
1066 * completed (indicated by mnt_writeopcount dropping to zero). At that
1067 * time, these operations are halted until the suspension is over.
1070 vn_start_secondary_write(vp, mpp, flags)
1080 if ((error = VOP_GETWRITEMOUNT(vp, mpp)) != 0) {
1082 if (error != EOPNOTSUPP)
1088 * If we are not suspended or have not yet reached suspended
1089 * mode, then let the operation proceed.
1091 if ((mp = *mpp) == NULL)
1095 * VOP_GETWRITEMOUNT() returns with the mp refcount held through
1097 * As long as a vnode is not provided we need to acquire a
1098 * refcount for the provided mountpoint too, in order to
1099 * emulate a vfs_ref().
1104 if ((mp->mnt_kern_flag & (MNTK_SUSPENDED | MNTK_SUSPEND2)) == 0) {
1105 mp->mnt_secondary_writes++;
1106 mp->mnt_secondary_accwrites++;
1110 if (flags & V_NOWAIT) {
1113 return (EWOULDBLOCK);
1116 * Wait for the suspension to finish.
1118 error = msleep(&mp->mnt_flag, MNT_MTX(mp),
1119 (PUSER - 1) | (flags & PCATCH) | PDROP, "suspfs", 0);
1127 * Filesystem write operation has completed. If we are suspending and this
1128 * operation is the last one, notify the suspender that the suspension is
1132 vn_finished_write(mp)
1139 mp->mnt_writeopcount--;
1140 if (mp->mnt_writeopcount < 0)
1141 panic("vn_finished_write: neg cnt");
1142 if ((mp->mnt_kern_flag & MNTK_SUSPEND) != 0 &&
1143 mp->mnt_writeopcount <= 0)
1144 wakeup(&mp->mnt_writeopcount);
1150 * Filesystem secondary write operation has completed. If we are
1151 * suspending and this operation is the last one, notify the suspender
1152 * that the suspension is now in effect.
1155 vn_finished_secondary_write(mp)
1162 mp->mnt_secondary_writes--;
1163 if (mp->mnt_secondary_writes < 0)
1164 panic("vn_finished_secondary_write: neg cnt");
1165 if ((mp->mnt_kern_flag & MNTK_SUSPEND) != 0 &&
1166 mp->mnt_secondary_writes <= 0)
1167 wakeup(&mp->mnt_secondary_writes);
1174 * Request a filesystem to suspend write operations.
1177 vfs_write_suspend(mp)
1183 if (mp->mnt_susp_owner == curthread) {
1187 while (mp->mnt_kern_flag & MNTK_SUSPEND)
1188 msleep(&mp->mnt_flag, MNT_MTX(mp), PUSER - 1, "wsuspfs", 0);
1189 mp->mnt_kern_flag |= MNTK_SUSPEND;
1190 mp->mnt_susp_owner = curthread;
1191 if (mp->mnt_writeopcount > 0)
1192 (void) msleep(&mp->mnt_writeopcount,
1193 MNT_MTX(mp), (PUSER - 1)|PDROP, "suspwt", 0);
1196 if ((error = VFS_SYNC(mp, MNT_SUSPEND)) != 0)
1197 vfs_write_resume(mp);
1202 * Request a filesystem to resume write operations.
1205 vfs_write_resume(mp)
1210 if ((mp->mnt_kern_flag & MNTK_SUSPEND) != 0) {
1211 KASSERT(mp->mnt_susp_owner == curthread, ("mnt_susp_owner"));
1212 mp->mnt_kern_flag &= ~(MNTK_SUSPEND | MNTK_SUSPEND2 |
1214 mp->mnt_susp_owner = NULL;
1215 wakeup(&mp->mnt_writeopcount);
1216 wakeup(&mp->mnt_flag);
1217 curthread->td_pflags &= ~TDP_IGNSUSP;
1225 * Implement kqueues for files by translating it to vnode operation.
1228 vn_kqfilter(struct file *fp, struct knote *kn)
1233 vfslocked = VFS_LOCK_GIANT(fp->f_vnode->v_mount);
1234 error = VOP_KQFILTER(fp->f_vnode, kn);
1235 VFS_UNLOCK_GIANT(vfslocked);
1241 * Simplified in-kernel wrapper calls for extended attribute access.
1242 * Both calls pass in a NULL credential, authorizing as "kernel" access.
1243 * Set IO_NODELOCKED in ioflg if the vnode is already locked.
1246 vn_extattr_get(struct vnode *vp, int ioflg, int attrnamespace,
1247 const char *attrname, int *buflen, char *buf, struct thread *td)
1253 iov.iov_len = *buflen;
1256 auio.uio_iov = &iov;
1257 auio.uio_iovcnt = 1;
1258 auio.uio_rw = UIO_READ;
1259 auio.uio_segflg = UIO_SYSSPACE;
1261 auio.uio_offset = 0;
1262 auio.uio_resid = *buflen;
1264 if ((ioflg & IO_NODELOCKED) == 0)
1265 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
1267 ASSERT_VOP_LOCKED(vp, "IO_NODELOCKED with no vp lock held");
1269 /* authorize attribute retrieval as kernel */
1270 error = VOP_GETEXTATTR(vp, attrnamespace, attrname, &auio, NULL, NULL,
1273 if ((ioflg & IO_NODELOCKED) == 0)
1277 *buflen = *buflen - auio.uio_resid;
1284 * XXX failure mode if partially written?
1287 vn_extattr_set(struct vnode *vp, int ioflg, int attrnamespace,
1288 const char *attrname, int buflen, char *buf, struct thread *td)
1295 iov.iov_len = buflen;
1298 auio.uio_iov = &iov;
1299 auio.uio_iovcnt = 1;
1300 auio.uio_rw = UIO_WRITE;
1301 auio.uio_segflg = UIO_SYSSPACE;
1303 auio.uio_offset = 0;
1304 auio.uio_resid = buflen;
1306 if ((ioflg & IO_NODELOCKED) == 0) {
1307 if ((error = vn_start_write(vp, &mp, V_WAIT)) != 0)
1309 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
1312 ASSERT_VOP_LOCKED(vp, "IO_NODELOCKED with no vp lock held");
1314 /* authorize attribute setting as kernel */
1315 error = VOP_SETEXTATTR(vp, attrnamespace, attrname, &auio, NULL, td);
1317 if ((ioflg & IO_NODELOCKED) == 0) {
1318 vn_finished_write(mp);
1326 vn_extattr_rm(struct vnode *vp, int ioflg, int attrnamespace,
1327 const char *attrname, struct thread *td)
1332 if ((ioflg & IO_NODELOCKED) == 0) {
1333 if ((error = vn_start_write(vp, &mp, V_WAIT)) != 0)
1335 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
1338 ASSERT_VOP_LOCKED(vp, "IO_NODELOCKED with no vp lock held");
1340 /* authorize attribute removal as kernel */
1341 error = VOP_DELETEEXTATTR(vp, attrnamespace, attrname, NULL, td);
1342 if (error == EOPNOTSUPP)
1343 error = VOP_SETEXTATTR(vp, attrnamespace, attrname, NULL,
1346 if ((ioflg & IO_NODELOCKED) == 0) {
1347 vn_finished_write(mp);
1355 vn_vget_ino(struct vnode *vp, ino_t ino, int lkflags, struct vnode **rvp)
1361 ltype = VOP_ISLOCKED(vp);
1362 KASSERT(ltype == LK_EXCLUSIVE || ltype == LK_SHARED,
1363 ("vn_vget_ino: vp not locked"));
1364 error = vfs_busy(mp, MBF_NOWAIT);
1368 error = vfs_busy(mp, 0);
1369 vn_lock(vp, ltype | LK_RETRY);
1373 if (vp->v_iflag & VI_DOOMED) {
1379 error = VFS_VGET(mp, ino, lkflags, rvp);
1381 vn_lock(vp, ltype | LK_RETRY);
1382 if (vp->v_iflag & VI_DOOMED) {
1391 vn_rlimit_fsize(const struct vnode *vp, const struct uio *uio, const struct thread *td)
1393 if (vp->v_type != VREG || td == NULL)
1396 PROC_LOCK(td->td_proc);
1397 if (uio->uio_offset + uio->uio_resid >
1398 lim_cur(td->td_proc, RLIMIT_FSIZE)) {
1399 psignal(td->td_proc, SIGXFSZ);
1400 PROC_UNLOCK(td->td_proc);
1403 PROC_UNLOCK(td->td_proc);
1409 vn_pages_remove(struct vnode *vp, vm_pindex_t start, vm_pindex_t end)
1413 if ((object = vp->v_object) == NULL)
1415 VM_OBJECT_LOCK(object);
1416 vm_object_page_remove(object, start, end, 0);
1417 VM_OBJECT_UNLOCK(object);