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/audit/audit.h>
65 #include <security/mac/mac_framework.h>
68 #include <vm/vm_object.h>
70 static fo_rdwr_t vn_read;
71 static fo_rdwr_t vn_write;
72 static fo_truncate_t vn_truncate;
73 static fo_ioctl_t vn_ioctl;
74 static fo_poll_t vn_poll;
75 static fo_kqfilter_t vn_kqfilter;
76 static fo_stat_t vn_statfile;
77 static fo_close_t vn_closefile;
79 struct fileops vnops = {
82 .fo_truncate = vn_truncate,
85 .fo_kqfilter = vn_kqfilter,
86 .fo_stat = vn_statfile,
87 .fo_close = vn_closefile,
90 .fo_flags = DFLAG_PASSABLE | DFLAG_SEEKABLE
94 vn_open(ndp, flagp, cmode, fp)
95 struct nameidata *ndp;
99 struct thread *td = ndp->ni_cnd.cn_thread;
101 return (vn_open_cred(ndp, flagp, cmode, 0, td->td_ucred, fp));
105 * Common code for vnode open operations.
106 * Check permissions, and call the VOP_OPEN or VOP_CREATE routine.
108 * Note that this does NOT free nameidata for the successful case,
109 * due to the NDINIT being done elsewhere.
112 vn_open_cred(struct nameidata *ndp, int *flagp, int cmode, u_int vn_open_flags,
113 struct ucred *cred, struct file *fp)
117 struct thread *td = ndp->ni_cnd.cn_thread;
119 struct vattr *vap = &vat;
122 int vfslocked, mpsafe;
124 mpsafe = ndp->ni_cnd.cn_flags & MPSAFE;
128 if (fmode & O_CREAT) {
129 ndp->ni_cnd.cn_nameiop = CREATE;
130 ndp->ni_cnd.cn_flags = ISOPEN | LOCKPARENT | LOCKLEAF |
132 if ((fmode & O_EXCL) == 0 && (fmode & O_NOFOLLOW) == 0)
133 ndp->ni_cnd.cn_flags |= FOLLOW;
134 if (!(vn_open_flags & VN_OPEN_NOAUDIT))
135 ndp->ni_cnd.cn_flags |= AUDITVNODE1;
137 if ((error = namei(ndp)) != 0)
139 vfslocked = NDHASGIANT(ndp);
141 ndp->ni_cnd.cn_flags &= ~MPSAFE;
142 if (ndp->ni_vp == NULL) {
145 vap->va_mode = cmode;
147 vap->va_vaflags |= VA_EXCLUSIVE;
148 if (vn_start_write(ndp->ni_dvp, &mp, V_NOWAIT) != 0) {
149 NDFREE(ndp, NDF_ONLY_PNBUF);
151 VFS_UNLOCK_GIANT(vfslocked);
152 if ((error = vn_start_write(NULL, &mp,
153 V_XSLEEP | PCATCH)) != 0)
158 error = mac_vnode_check_create(cred, ndp->ni_dvp,
162 error = VOP_CREATE(ndp->ni_dvp, &ndp->ni_vp,
165 vn_finished_write(mp);
167 VFS_UNLOCK_GIANT(vfslocked);
168 NDFREE(ndp, NDF_ONLY_PNBUF);
174 if (ndp->ni_dvp == ndp->ni_vp)
180 if (fmode & O_EXCL) {
187 ndp->ni_cnd.cn_nameiop = LOOKUP;
188 ndp->ni_cnd.cn_flags = ISOPEN |
189 ((fmode & O_NOFOLLOW) ? NOFOLLOW : FOLLOW) |
191 if (!(fmode & FWRITE))
192 ndp->ni_cnd.cn_flags |= LOCKSHARED;
193 if (!(vn_open_flags & VN_OPEN_NOAUDIT))
194 ndp->ni_cnd.cn_flags |= AUDITVNODE1;
195 if ((error = namei(ndp)) != 0)
198 ndp->ni_cnd.cn_flags &= ~MPSAFE;
199 vfslocked = NDHASGIANT(ndp);
202 if (vp->v_type == VLNK) {
206 if (vp->v_type == VSOCK) {
210 if (vp->v_type != VDIR && fmode & O_DIRECTORY) {
215 if (fmode & (FWRITE | O_TRUNC)) {
216 if (vp->v_type == VDIR) {
226 if ((fmode & O_APPEND) && (fmode & FWRITE))
229 error = mac_vnode_check_open(cred, vp, accmode);
233 if ((fmode & O_CREAT) == 0) {
234 if (accmode & VWRITE) {
235 error = vn_writechk(vp);
240 error = VOP_ACCESS(vp, accmode, cred, td);
245 if ((error = VOP_OPEN(vp, fmode, cred, td, fp)) != 0)
248 if (fmode & FWRITE) {
250 CTR3(KTR_VFS, "%s: vp %p v_writecount increased to %d",
251 __func__, vp, vp->v_writecount);
254 ASSERT_VOP_LOCKED(vp, "vn_open_cred");
256 VFS_UNLOCK_GIANT(vfslocked);
259 NDFREE(ndp, NDF_ONLY_PNBUF);
261 VFS_UNLOCK_GIANT(vfslocked);
268 * Check for write permissions on the specified vnode.
269 * Prototype text segments cannot be written.
273 register struct vnode *vp;
276 ASSERT_VOP_LOCKED(vp, "vn_writechk");
278 * If there's shared text associated with
279 * the vnode, try to free it up once. If
280 * we fail, we can't allow writing.
282 if (vp->v_vflag & VV_TEXT)
292 vn_close(vp, flags, file_cred, td)
293 register struct vnode *vp;
295 struct ucred *file_cred;
299 int error, lock_flags;
301 if (!(flags & FWRITE) && vp->v_mount != NULL &&
302 vp->v_mount->mnt_kern_flag & MNTK_EXTENDED_SHARED)
303 lock_flags = LK_SHARED;
305 lock_flags = LK_EXCLUSIVE;
307 VFS_ASSERT_GIANT(vp->v_mount);
309 vn_start_write(vp, &mp, V_WAIT);
310 vn_lock(vp, lock_flags | LK_RETRY);
311 if (flags & FWRITE) {
312 VNASSERT(vp->v_writecount > 0, vp,
313 ("vn_close: negative writecount"));
315 CTR3(KTR_VFS, "%s: vp %p v_writecount decreased to %d",
316 __func__, vp, vp->v_writecount);
318 error = VOP_CLOSE(vp, flags, file_cred, td);
320 vn_finished_write(mp);
325 * Heuristic to detect sequential operation.
328 sequential_heuristic(struct uio *uio, struct file *fp)
331 if (atomic_load_acq_int(&(fp->f_flag)) & FRDAHEAD)
332 return (fp->f_seqcount << IO_SEQSHIFT);
335 * Offset 0 is handled specially. open() sets f_seqcount to 1 so
336 * that the first I/O is normally considered to be slightly
337 * sequential. Seeking to offset 0 doesn't change sequentiality
338 * unless previous seeks have reduced f_seqcount to 0, in which
339 * case offset 0 is not special.
341 if ((uio->uio_offset == 0 && fp->f_seqcount > 0) ||
342 uio->uio_offset == fp->f_nextoff) {
344 * f_seqcount is in units of fixed-size blocks so that it
345 * depends mainly on the amount of sequential I/O and not
346 * much on the number of sequential I/O's. The fixed size
347 * of 16384 is hard-coded here since it is (not quite) just
348 * a magic size that works well here. This size is more
349 * closely related to the best I/O size for real disks than
350 * to any block size used by software.
352 fp->f_seqcount += howmany(uio->uio_resid, 16384);
353 if (fp->f_seqcount > IO_SEQMAX)
354 fp->f_seqcount = IO_SEQMAX;
355 return (fp->f_seqcount << IO_SEQSHIFT);
358 /* Not sequential. Quickly draw-down sequentiality. */
359 if (fp->f_seqcount > 1)
367 * Package up an I/O request on a vnode into a uio and do it.
370 vn_rdwr(rw, vp, base, len, offset, segflg, ioflg, active_cred, file_cred,
379 struct ucred *active_cred;
380 struct ucred *file_cred;
388 int error, lock_flags;
390 VFS_ASSERT_GIANT(vp->v_mount);
392 if ((ioflg & IO_NODELOCKED) == 0) {
394 if (rw == UIO_WRITE) {
395 if (vp->v_type != VCHR &&
396 (error = vn_start_write(vp, &mp, V_WAIT | PCATCH))
399 if (MNT_SHARED_WRITES(mp) ||
400 ((mp == NULL) && MNT_SHARED_WRITES(vp->v_mount))) {
401 lock_flags = LK_SHARED;
403 lock_flags = LK_EXCLUSIVE;
405 vn_lock(vp, lock_flags | LK_RETRY);
407 vn_lock(vp, LK_SHARED | LK_RETRY);
410 ASSERT_VOP_LOCKED(vp, "IO_NODELOCKED with no vp lock held");
411 auio.uio_iov = &aiov;
413 aiov.iov_base = base;
415 auio.uio_resid = len;
416 auio.uio_offset = offset;
417 auio.uio_segflg = segflg;
422 if ((ioflg & IO_NOMACCHECK) == 0) {
424 error = mac_vnode_check_read(active_cred, file_cred,
427 error = mac_vnode_check_write(active_cred, file_cred,
437 error = VOP_READ(vp, &auio, ioflg, cred);
439 error = VOP_WRITE(vp, &auio, ioflg, cred);
442 *aresid = auio.uio_resid;
444 if (auio.uio_resid && error == 0)
446 if ((ioflg & IO_NODELOCKED) == 0) {
447 if (rw == UIO_WRITE && vp->v_type != VCHR)
448 vn_finished_write(mp);
455 * Package up an I/O request on a vnode into a uio and do it. The I/O
456 * request is split up into smaller chunks and we try to avoid saturating
457 * the buffer cache while potentially holding a vnode locked, so we
458 * check bwillwrite() before calling vn_rdwr(). We also call kern_yield()
459 * to give other processes a chance to lock the vnode (either other processes
460 * core'ing the same binary, or unrelated processes scanning the directory).
463 vn_rdwr_inchunks(rw, vp, base, len, offset, segflg, ioflg, active_cred,
464 file_cred, aresid, td)
472 struct ucred *active_cred;
473 struct ucred *file_cred;
480 VFS_ASSERT_GIANT(vp->v_mount);
486 * Force `offset' to a multiple of MAXBSIZE except possibly
487 * for the first chunk, so that filesystems only need to
488 * write full blocks except possibly for the first and last
491 chunk = MAXBSIZE - (uoff_t)offset % MAXBSIZE;
495 if (rw != UIO_READ && vp->v_type == VREG)
498 error = vn_rdwr(rw, vp, base, chunk, offset, segflg,
499 ioflg, active_cred, file_cred, &iaresid, td);
500 len -= chunk; /* aresid calc already includes length */
504 base = (char *)base + chunk;
505 kern_yield(PRI_USER);
508 *aresid = len + iaresid;
513 * File table vnode read routine.
516 vn_read(fp, uio, active_cred, flags, td)
519 struct ucred *active_cred;
526 int advice, vfslocked;
529 KASSERT(uio->uio_td == td, ("uio_td %p is not td %p",
534 if (fp->f_flag & FNONBLOCK)
536 if (fp->f_flag & O_DIRECT)
538 advice = POSIX_FADV_NORMAL;
539 vfslocked = VFS_LOCK_GIANT(vp->v_mount);
541 * According to McKusick the vn lock was protecting f_offset here.
542 * It is now protected by the FOFFSET_LOCKED flag.
544 if ((flags & FOF_OFFSET) == 0 || fp->f_advice != NULL) {
545 mtxp = mtx_pool_find(mtxpool_sleep, fp);
547 if ((flags & FOF_OFFSET) == 0) {
548 while (fp->f_vnread_flags & FOFFSET_LOCKED) {
549 fp->f_vnread_flags |= FOFFSET_LOCK_WAITING;
550 msleep(&fp->f_vnread_flags, mtxp, PUSER -1,
551 "vnread offlock", 0);
553 fp->f_vnread_flags |= FOFFSET_LOCKED;
554 uio->uio_offset = fp->f_offset;
556 if (fp->f_advice != NULL &&
557 uio->uio_offset >= fp->f_advice->fa_start &&
558 uio->uio_offset + uio->uio_resid <= fp->f_advice->fa_end)
559 advice = fp->f_advice->fa_advice;
562 vn_lock(vp, LK_SHARED | LK_RETRY);
565 case POSIX_FADV_NORMAL:
566 case POSIX_FADV_SEQUENTIAL:
567 case POSIX_FADV_NOREUSE:
568 ioflag |= sequential_heuristic(uio, fp);
570 case POSIX_FADV_RANDOM:
571 /* Disable read-ahead for random I/O. */
574 offset = uio->uio_offset;
577 error = mac_vnode_check_read(active_cred, fp->f_cred, vp);
580 error = VOP_READ(vp, uio, ioflag, fp->f_cred);
581 if ((flags & FOF_OFFSET) == 0) {
582 fp->f_offset = uio->uio_offset;
584 if (fp->f_vnread_flags & FOFFSET_LOCK_WAITING)
585 wakeup(&fp->f_vnread_flags);
586 fp->f_vnread_flags = 0;
589 fp->f_nextoff = uio->uio_offset;
591 if (error == 0 && advice == POSIX_FADV_NOREUSE &&
592 offset != uio->uio_offset)
593 error = VOP_ADVISE(vp, offset, uio->uio_offset - 1,
594 POSIX_FADV_DONTNEED);
595 VFS_UNLOCK_GIANT(vfslocked);
600 * File table vnode write routine.
603 vn_write(fp, uio, active_cred, flags, td)
606 struct ucred *active_cred;
612 int error, ioflag, lock_flags;
614 int advice, vfslocked;
616 KASSERT(uio->uio_td == td, ("uio_td %p is not td %p",
619 vfslocked = VFS_LOCK_GIANT(vp->v_mount);
620 if (vp->v_type == VREG)
623 if (vp->v_type == VREG && (fp->f_flag & O_APPEND))
625 if (fp->f_flag & FNONBLOCK)
627 if (fp->f_flag & O_DIRECT)
629 if ((fp->f_flag & O_FSYNC) ||
630 (vp->v_mount && (vp->v_mount->mnt_flag & MNT_SYNCHRONOUS)))
633 if (vp->v_type != VCHR &&
634 (error = vn_start_write(vp, &mp, V_WAIT | PCATCH)) != 0)
637 if ((MNT_SHARED_WRITES(mp) ||
638 ((mp == NULL) && MNT_SHARED_WRITES(vp->v_mount))) &&
639 (flags & FOF_OFFSET) != 0) {
640 lock_flags = LK_SHARED;
642 lock_flags = LK_EXCLUSIVE;
645 vn_lock(vp, lock_flags | LK_RETRY);
646 if ((flags & FOF_OFFSET) == 0)
647 uio->uio_offset = fp->f_offset;
648 advice = POSIX_FADV_NORMAL;
649 if (fp->f_advice != NULL) {
650 mtxp = mtx_pool_find(mtxpool_sleep, fp);
652 if (fp->f_advice != NULL &&
653 uio->uio_offset >= fp->f_advice->fa_start &&
654 uio->uio_offset + uio->uio_resid <= fp->f_advice->fa_end)
655 advice = fp->f_advice->fa_advice;
659 case POSIX_FADV_NORMAL:
660 case POSIX_FADV_SEQUENTIAL:
661 ioflag |= sequential_heuristic(uio, fp);
663 case POSIX_FADV_RANDOM:
664 /* XXX: Is this correct? */
666 case POSIX_FADV_NOREUSE:
668 * Request the underlying FS to discard the buffers
669 * and pages after the I/O is complete.
676 error = mac_vnode_check_write(active_cred, fp->f_cred, vp);
679 error = VOP_WRITE(vp, uio, ioflag, fp->f_cred);
680 if ((flags & FOF_OFFSET) == 0)
681 fp->f_offset = uio->uio_offset;
682 fp->f_nextoff = uio->uio_offset;
684 if (vp->v_type != VCHR)
685 vn_finished_write(mp);
687 VFS_UNLOCK_GIANT(vfslocked);
692 * File table truncate routine.
695 vn_truncate(fp, length, active_cred, td)
698 struct ucred *active_cred;
708 vfslocked = VFS_LOCK_GIANT(vp->v_mount);
709 error = vn_start_write(vp, &mp, V_WAIT | PCATCH);
711 VFS_UNLOCK_GIANT(vfslocked);
714 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
715 if (vp->v_type == VDIR) {
720 error = mac_vnode_check_write(active_cred, fp->f_cred, vp);
724 error = vn_writechk(vp);
727 vattr.va_size = length;
728 error = VOP_SETATTR(vp, &vattr, fp->f_cred);
732 vn_finished_write(mp);
733 VFS_UNLOCK_GIANT(vfslocked);
738 * File table vnode stat routine.
741 vn_statfile(fp, sb, active_cred, td)
744 struct ucred *active_cred;
747 struct vnode *vp = fp->f_vnode;
751 vfslocked = VFS_LOCK_GIANT(vp->v_mount);
752 vn_lock(vp, LK_SHARED | LK_RETRY);
753 error = vn_stat(vp, sb, active_cred, fp->f_cred, td);
755 VFS_UNLOCK_GIANT(vfslocked);
761 * Stat a vnode; implementation for the stat syscall
764 vn_stat(vp, sb, active_cred, file_cred, td)
766 register struct stat *sb;
767 struct ucred *active_cred;
768 struct ucred *file_cred;
772 register struct vattr *vap;
777 error = mac_vnode_check_stat(active_cred, file_cred, vp);
785 * Initialize defaults for new and unusual fields, so that file
786 * systems which don't support these fields don't need to know
789 vap->va_birthtime.tv_sec = -1;
790 vap->va_birthtime.tv_nsec = 0;
791 vap->va_fsid = VNOVAL;
792 vap->va_rdev = NODEV;
794 error = VOP_GETATTR(vp, vap, active_cred);
799 * Zero the spare stat fields
801 bzero(sb, sizeof *sb);
804 * Copy from vattr table
806 if (vap->va_fsid != VNOVAL)
807 sb->st_dev = vap->va_fsid;
809 sb->st_dev = vp->v_mount->mnt_stat.f_fsid.val[0];
810 sb->st_ino = vap->va_fileid;
812 switch (vap->va_type) {
838 sb->st_nlink = vap->va_nlink;
839 sb->st_uid = vap->va_uid;
840 sb->st_gid = vap->va_gid;
841 sb->st_rdev = vap->va_rdev;
842 if (vap->va_size > OFF_MAX)
844 sb->st_size = vap->va_size;
845 sb->st_atim = vap->va_atime;
846 sb->st_mtim = vap->va_mtime;
847 sb->st_ctim = vap->va_ctime;
848 sb->st_birthtim = vap->va_birthtime;
851 * According to www.opengroup.org, the meaning of st_blksize is
852 * "a filesystem-specific preferred I/O block size for this
853 * object. In some filesystem types, this may vary from file
855 * Use miminum/default of PAGE_SIZE (e.g. for VCHR).
858 sb->st_blksize = max(PAGE_SIZE, vap->va_blocksize);
860 sb->st_flags = vap->va_flags;
861 if (priv_check(td, PRIV_VFS_GENERATION))
864 sb->st_gen = vap->va_gen;
866 sb->st_blocks = vap->va_bytes / S_BLKSIZE;
871 * File table vnode ioctl routine.
874 vn_ioctl(fp, com, data, active_cred, td)
878 struct ucred *active_cred;
881 struct vnode *vp = fp->f_vnode;
886 vfslocked = VFS_LOCK_GIANT(vp->v_mount);
888 switch (vp->v_type) {
891 if (com == FIONREAD) {
892 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
893 error = VOP_GETATTR(vp, &vattr, active_cred);
896 *(int *)data = vattr.va_size - fp->f_offset;
898 if (com == FIONBIO || com == FIOASYNC) /* XXX */
901 error = VOP_IOCTL(vp, com, data, fp->f_flag,
908 VFS_UNLOCK_GIANT(vfslocked);
913 * File table vnode poll routine.
916 vn_poll(fp, events, active_cred, td)
919 struct ucred *active_cred;
927 vfslocked = VFS_LOCK_GIANT(vp->v_mount);
929 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
930 error = mac_vnode_check_poll(active_cred, fp->f_cred, vp);
935 error = VOP_POLL(vp, events, fp->f_cred, td);
936 VFS_UNLOCK_GIANT(vfslocked);
941 * Acquire the requested lock and then check for validity. LK_RETRY
942 * permits vn_lock to return doomed vnodes.
945 _vn_lock(struct vnode *vp, int flags, char *file, int line)
949 VNASSERT((flags & LK_TYPE_MASK) != 0, vp,
950 ("vn_lock called with no locktype."));
952 #ifdef DEBUG_VFS_LOCKS
953 KASSERT(vp->v_holdcnt != 0,
954 ("vn_lock %p: zero hold count", vp));
956 error = VOP_LOCK1(vp, flags, file, line);
957 flags &= ~LK_INTERLOCK; /* Interlock is always dropped. */
958 KASSERT((flags & LK_RETRY) == 0 || error == 0,
959 ("LK_RETRY set with incompatible flags (0x%x) or an error occured (%d)",
962 * Callers specify LK_RETRY if they wish to get dead vnodes.
963 * If RETRY is not set, we return ENOENT instead.
965 if (error == 0 && vp->v_iflag & VI_DOOMED &&
966 (flags & LK_RETRY) == 0) {
971 } while (flags & LK_RETRY && error != 0);
976 * File table vnode close routine.
990 vfslocked = VFS_LOCK_GIANT(vp->v_mount);
991 if (fp->f_type == DTYPE_VNODE && fp->f_flag & FHASLOCK) {
992 lf.l_whence = SEEK_SET;
996 (void) VOP_ADVLOCK(vp, fp, F_UNLCK, &lf, F_FLOCK);
999 fp->f_ops = &badfileops;
1001 error = vn_close(vp, fp->f_flag, fp->f_cred, td);
1002 VFS_UNLOCK_GIANT(vfslocked);
1007 * Preparing to start a filesystem write operation. If the operation is
1008 * permitted, then we bump the count of operations in progress and
1009 * proceed. If a suspend request is in progress, we wait until the
1010 * suspension is over, and then proceed.
1013 vn_start_write(vp, mpp, flags)
1023 * If a vnode is provided, get and return the mount point that
1024 * to which it will write.
1027 if ((error = VOP_GETWRITEMOUNT(vp, mpp)) != 0) {
1029 if (error != EOPNOTSUPP)
1034 if ((mp = *mpp) == NULL)
1038 * VOP_GETWRITEMOUNT() returns with the mp refcount held through
1040 * As long as a vnode is not provided we need to acquire a
1041 * refcount for the provided mountpoint too, in order to
1042 * emulate a vfs_ref().
1049 * Check on status of suspension.
1051 if ((curthread->td_pflags & TDP_IGNSUSP) == 0 ||
1052 mp->mnt_susp_owner != curthread) {
1053 while ((mp->mnt_kern_flag & MNTK_SUSPEND) != 0) {
1054 if (flags & V_NOWAIT) {
1055 error = EWOULDBLOCK;
1058 error = msleep(&mp->mnt_flag, MNT_MTX(mp),
1059 (PUSER - 1) | (flags & PCATCH), "suspfs", 0);
1064 if (flags & V_XSLEEP)
1066 mp->mnt_writeopcount++;
1068 if (error != 0 || (flags & V_XSLEEP) != 0)
1075 * Secondary suspension. Used by operations such as vop_inactive
1076 * routines that are needed by the higher level functions. These
1077 * are allowed to proceed until all the higher level functions have
1078 * completed (indicated by mnt_writeopcount dropping to zero). At that
1079 * time, these operations are halted until the suspension is over.
1082 vn_start_secondary_write(vp, mpp, flags)
1092 if ((error = VOP_GETWRITEMOUNT(vp, mpp)) != 0) {
1094 if (error != EOPNOTSUPP)
1100 * If we are not suspended or have not yet reached suspended
1101 * mode, then let the operation proceed.
1103 if ((mp = *mpp) == NULL)
1107 * VOP_GETWRITEMOUNT() returns with the mp refcount held through
1109 * As long as a vnode is not provided we need to acquire a
1110 * refcount for the provided mountpoint too, in order to
1111 * emulate a vfs_ref().
1116 if ((mp->mnt_kern_flag & (MNTK_SUSPENDED | MNTK_SUSPEND2)) == 0) {
1117 mp->mnt_secondary_writes++;
1118 mp->mnt_secondary_accwrites++;
1122 if (flags & V_NOWAIT) {
1125 return (EWOULDBLOCK);
1128 * Wait for the suspension to finish.
1130 error = msleep(&mp->mnt_flag, MNT_MTX(mp),
1131 (PUSER - 1) | (flags & PCATCH) | PDROP, "suspfs", 0);
1139 * Filesystem write operation has completed. If we are suspending and this
1140 * operation is the last one, notify the suspender that the suspension is
1144 vn_finished_write(mp)
1151 mp->mnt_writeopcount--;
1152 if (mp->mnt_writeopcount < 0)
1153 panic("vn_finished_write: neg cnt");
1154 if ((mp->mnt_kern_flag & MNTK_SUSPEND) != 0 &&
1155 mp->mnt_writeopcount <= 0)
1156 wakeup(&mp->mnt_writeopcount);
1162 * Filesystem secondary write operation has completed. If we are
1163 * suspending and this operation is the last one, notify the suspender
1164 * that the suspension is now in effect.
1167 vn_finished_secondary_write(mp)
1174 mp->mnt_secondary_writes--;
1175 if (mp->mnt_secondary_writes < 0)
1176 panic("vn_finished_secondary_write: neg cnt");
1177 if ((mp->mnt_kern_flag & MNTK_SUSPEND) != 0 &&
1178 mp->mnt_secondary_writes <= 0)
1179 wakeup(&mp->mnt_secondary_writes);
1186 * Request a filesystem to suspend write operations.
1189 vfs_write_suspend(mp)
1195 if (mp->mnt_susp_owner == curthread) {
1199 while (mp->mnt_kern_flag & MNTK_SUSPEND)
1200 msleep(&mp->mnt_flag, MNT_MTX(mp), PUSER - 1, "wsuspfs", 0);
1201 mp->mnt_kern_flag |= MNTK_SUSPEND;
1202 mp->mnt_susp_owner = curthread;
1203 if (mp->mnt_writeopcount > 0)
1204 (void) msleep(&mp->mnt_writeopcount,
1205 MNT_MTX(mp), (PUSER - 1)|PDROP, "suspwt", 0);
1208 if ((error = VFS_SYNC(mp, MNT_SUSPEND)) != 0)
1209 vfs_write_resume(mp);
1214 * Request a filesystem to resume write operations.
1217 vfs_write_resume(mp)
1222 if ((mp->mnt_kern_flag & MNTK_SUSPEND) != 0) {
1223 KASSERT(mp->mnt_susp_owner == curthread, ("mnt_susp_owner"));
1224 mp->mnt_kern_flag &= ~(MNTK_SUSPEND | MNTK_SUSPEND2 |
1226 mp->mnt_susp_owner = NULL;
1227 wakeup(&mp->mnt_writeopcount);
1228 wakeup(&mp->mnt_flag);
1229 curthread->td_pflags &= ~TDP_IGNSUSP;
1237 * Implement kqueues for files by translating it to vnode operation.
1240 vn_kqfilter(struct file *fp, struct knote *kn)
1245 vfslocked = VFS_LOCK_GIANT(fp->f_vnode->v_mount);
1246 error = VOP_KQFILTER(fp->f_vnode, kn);
1247 VFS_UNLOCK_GIANT(vfslocked);
1253 * Simplified in-kernel wrapper calls for extended attribute access.
1254 * Both calls pass in a NULL credential, authorizing as "kernel" access.
1255 * Set IO_NODELOCKED in ioflg if the vnode is already locked.
1258 vn_extattr_get(struct vnode *vp, int ioflg, int attrnamespace,
1259 const char *attrname, int *buflen, char *buf, struct thread *td)
1265 iov.iov_len = *buflen;
1268 auio.uio_iov = &iov;
1269 auio.uio_iovcnt = 1;
1270 auio.uio_rw = UIO_READ;
1271 auio.uio_segflg = UIO_SYSSPACE;
1273 auio.uio_offset = 0;
1274 auio.uio_resid = *buflen;
1276 if ((ioflg & IO_NODELOCKED) == 0)
1277 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
1279 ASSERT_VOP_LOCKED(vp, "IO_NODELOCKED with no vp lock held");
1281 /* authorize attribute retrieval as kernel */
1282 error = VOP_GETEXTATTR(vp, attrnamespace, attrname, &auio, NULL, NULL,
1285 if ((ioflg & IO_NODELOCKED) == 0)
1289 *buflen = *buflen - auio.uio_resid;
1296 * XXX failure mode if partially written?
1299 vn_extattr_set(struct vnode *vp, int ioflg, int attrnamespace,
1300 const char *attrname, int buflen, char *buf, struct thread *td)
1307 iov.iov_len = buflen;
1310 auio.uio_iov = &iov;
1311 auio.uio_iovcnt = 1;
1312 auio.uio_rw = UIO_WRITE;
1313 auio.uio_segflg = UIO_SYSSPACE;
1315 auio.uio_offset = 0;
1316 auio.uio_resid = buflen;
1318 if ((ioflg & IO_NODELOCKED) == 0) {
1319 if ((error = vn_start_write(vp, &mp, V_WAIT)) != 0)
1321 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
1324 ASSERT_VOP_LOCKED(vp, "IO_NODELOCKED with no vp lock held");
1326 /* authorize attribute setting as kernel */
1327 error = VOP_SETEXTATTR(vp, attrnamespace, attrname, &auio, NULL, td);
1329 if ((ioflg & IO_NODELOCKED) == 0) {
1330 vn_finished_write(mp);
1338 vn_extattr_rm(struct vnode *vp, int ioflg, int attrnamespace,
1339 const char *attrname, struct thread *td)
1344 if ((ioflg & IO_NODELOCKED) == 0) {
1345 if ((error = vn_start_write(vp, &mp, V_WAIT)) != 0)
1347 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
1350 ASSERT_VOP_LOCKED(vp, "IO_NODELOCKED with no vp lock held");
1352 /* authorize attribute removal as kernel */
1353 error = VOP_DELETEEXTATTR(vp, attrnamespace, attrname, NULL, td);
1354 if (error == EOPNOTSUPP)
1355 error = VOP_SETEXTATTR(vp, attrnamespace, attrname, NULL,
1358 if ((ioflg & IO_NODELOCKED) == 0) {
1359 vn_finished_write(mp);
1367 vn_vget_ino(struct vnode *vp, ino_t ino, int lkflags, struct vnode **rvp)
1373 ltype = VOP_ISLOCKED(vp);
1374 KASSERT(ltype == LK_EXCLUSIVE || ltype == LK_SHARED,
1375 ("vn_vget_ino: vp not locked"));
1376 error = vfs_busy(mp, MBF_NOWAIT);
1380 error = vfs_busy(mp, 0);
1381 vn_lock(vp, ltype | LK_RETRY);
1385 if (vp->v_iflag & VI_DOOMED) {
1391 error = VFS_VGET(mp, ino, lkflags, rvp);
1393 vn_lock(vp, ltype | LK_RETRY);
1394 if (vp->v_iflag & VI_DOOMED) {
1403 vn_rlimit_fsize(const struct vnode *vp, const struct uio *uio,
1404 const struct thread *td)
1407 if (vp->v_type != VREG || td == NULL)
1409 PROC_LOCK(td->td_proc);
1410 if ((uoff_t)uio->uio_offset + uio->uio_resid >
1411 lim_cur(td->td_proc, RLIMIT_FSIZE)) {
1412 kern_psignal(td->td_proc, SIGXFSZ);
1413 PROC_UNLOCK(td->td_proc);
1416 PROC_UNLOCK(td->td_proc);
1421 vn_chmod(struct file *fp, mode_t mode, struct ucred *active_cred,
1425 int error, vfslocked;
1428 vfslocked = VFS_LOCK_GIANT(vp->v_mount);
1430 vn_lock(vp, LK_SHARED | LK_RETRY);
1431 AUDIT_ARG_VNODE1(vp);
1434 error = setfmode(td, active_cred, vp, mode);
1435 VFS_UNLOCK_GIANT(vfslocked);
1440 vn_chown(struct file *fp, uid_t uid, gid_t gid, struct ucred *active_cred,
1444 int error, vfslocked;
1447 vfslocked = VFS_LOCK_GIANT(vp->v_mount);
1449 vn_lock(vp, LK_SHARED | LK_RETRY);
1450 AUDIT_ARG_VNODE1(vp);
1453 error = setfown(td, active_cred, vp, uid, gid);
1454 VFS_UNLOCK_GIANT(vfslocked);
1459 vn_pages_remove(struct vnode *vp, vm_pindex_t start, vm_pindex_t end)
1463 if ((object = vp->v_object) == NULL)
1465 VM_OBJECT_LOCK(object);
1466 vm_object_page_remove(object, start, end, 0);
1467 VM_OBJECT_UNLOCK(object);