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("$FreeBSD$");
42 #include <sys/param.h>
43 #include <sys/systm.h>
44 #include <sys/fcntl.h>
50 #include <sys/limits.h>
52 #include <sys/mount.h>
53 #include <sys/mutex.h>
54 #include <sys/namei.h>
55 #include <sys/vnode.h>
58 #include <sys/filio.h>
60 #include <sys/ttycom.h>
62 #include <sys/syslog.h>
63 #include <sys/unistd.h>
65 #include <security/mac/mac_framework.h>
67 static fo_rdwr_t vn_read;
68 static fo_rdwr_t vn_write;
69 static fo_truncate_t vn_truncate;
70 static fo_ioctl_t vn_ioctl;
71 static fo_poll_t vn_poll;
72 static fo_kqfilter_t vn_kqfilter;
73 static fo_stat_t vn_statfile;
74 static fo_close_t vn_closefile;
76 struct fileops vnops = {
79 .fo_truncate = vn_truncate,
82 .fo_kqfilter = vn_kqfilter,
83 .fo_stat = vn_statfile,
84 .fo_close = vn_closefile,
85 .fo_flags = DFLAG_PASSABLE | DFLAG_SEEKABLE
89 vn_open(ndp, flagp, cmode, fp)
90 struct nameidata *ndp;
94 struct thread *td = ndp->ni_cnd.cn_thread;
96 return (vn_open_cred(ndp, flagp, cmode, td->td_ucred, fp));
100 * Common code for vnode open operations.
101 * Check permissions, and call the VOP_OPEN or VOP_CREATE routine.
103 * Note that this does NOT free nameidata for the successful case,
104 * due to the NDINIT being done elsewhere.
107 vn_open_cred(ndp, flagp, cmode, cred, fp)
108 struct nameidata *ndp;
115 struct thread *td = ndp->ni_cnd.cn_thread;
117 struct vattr *vap = &vat;
118 int mode, fmode, error;
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 |
128 MPSAFE | AUDITVNODE1;
129 if ((fmode & O_EXCL) == 0 && (fmode & O_NOFOLLOW) == 0)
130 ndp->ni_cnd.cn_flags |= FOLLOW;
132 if ((error = namei(ndp)) != 0)
134 vfslocked = NDHASGIANT(ndp);
136 ndp->ni_cnd.cn_flags &= ~MPSAFE;
137 if (ndp->ni_vp == NULL) {
140 vap->va_mode = cmode;
142 vap->va_vaflags |= VA_EXCLUSIVE;
143 if (vn_start_write(ndp->ni_dvp, &mp, V_NOWAIT) != 0) {
144 NDFREE(ndp, NDF_ONLY_PNBUF);
146 VFS_UNLOCK_GIANT(vfslocked);
147 if ((error = vn_start_write(NULL, &mp,
148 V_XSLEEP | PCATCH)) != 0)
153 error = mac_vnode_check_create(cred, ndp->ni_dvp,
157 VOP_LEASE(ndp->ni_dvp, td, cred, LEASE_WRITE);
158 error = VOP_CREATE(ndp->ni_dvp, &ndp->ni_vp,
164 vn_finished_write(mp);
166 VFS_UNLOCK_GIANT(vfslocked);
167 NDFREE(ndp, NDF_ONLY_PNBUF);
173 if (ndp->ni_dvp == ndp->ni_vp)
179 if (fmode & O_EXCL) {
186 ndp->ni_cnd.cn_nameiop = LOOKUP;
187 ndp->ni_cnd.cn_flags = ISOPEN |
188 ((fmode & O_NOFOLLOW) ? NOFOLLOW : FOLLOW) |
189 LOCKLEAF | MPSAFE | AUDITVNODE1;
190 if ((error = namei(ndp)) != 0)
193 ndp->ni_cnd.cn_flags &= ~MPSAFE;
194 vfslocked = NDHASGIANT(ndp);
197 if (vp->v_type == VLNK) {
201 if (vp->v_type == VSOCK) {
206 if (fmode & (FWRITE | O_TRUNC)) {
207 if (vp->v_type == VDIR) {
217 if (fmode & O_APPEND)
220 error = mac_vnode_check_open(cred, vp, mode);
224 if ((fmode & O_CREAT) == 0) {
226 error = vn_writechk(vp);
231 error = VOP_ACCESS(vp, mode, cred, td);
236 if ((error = VOP_OPEN(vp, fmode, cred, td, fp)) != 0)
242 ASSERT_VOP_ELOCKED(vp, "vn_open_cred");
244 VFS_UNLOCK_GIANT(vfslocked);
247 NDFREE(ndp, NDF_ONLY_PNBUF);
249 VFS_UNLOCK_GIANT(vfslocked);
256 * Check for write permissions on the specified vnode.
257 * Prototype text segments cannot be written.
261 register struct vnode *vp;
264 ASSERT_VOP_LOCKED(vp, "vn_writechk");
266 * If there's shared text associated with
267 * the vnode, try to free it up once. If
268 * we fail, we can't allow writing.
270 if (vp->v_vflag & VV_TEXT)
280 vn_close(vp, flags, file_cred, td)
281 register struct vnode *vp;
283 struct ucred *file_cred;
289 VFS_ASSERT_GIANT(vp->v_mount);
291 vn_start_write(vp, &mp, V_WAIT);
292 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
293 if (flags & FWRITE) {
294 VNASSERT(vp->v_writecount > 0, vp,
295 ("vn_close: negative writecount"));
298 error = VOP_CLOSE(vp, flags, file_cred, td);
300 vn_finished_write(mp);
305 * Heuristic to detect sequential operation.
308 sequential_heuristic(struct uio *uio, struct file *fp)
312 * Offset 0 is handled specially. open() sets f_seqcount to 1 so
313 * that the first I/O is normally considered to be slightly
314 * sequential. Seeking to offset 0 doesn't change sequentiality
315 * unless previous seeks have reduced f_seqcount to 0, in which
316 * case offset 0 is not special.
318 if ((uio->uio_offset == 0 && fp->f_seqcount > 0) ||
319 uio->uio_offset == fp->f_nextoff) {
321 * f_seqcount is in units of fixed-size blocks so that it
322 * depends mainly on the amount of sequential I/O and not
323 * much on the number of sequential I/O's. The fixed size
324 * of 16384 is hard-coded here since it is (not quite) just
325 * a magic size that works well here. This size is more
326 * closely related to the best I/O size for real disks than
327 * to any block size used by software.
329 fp->f_seqcount += howmany(uio->uio_resid, 16384);
330 if (fp->f_seqcount > IO_SEQMAX)
331 fp->f_seqcount = IO_SEQMAX;
332 return (fp->f_seqcount << IO_SEQSHIFT);
335 /* Not sequential. Quickly draw-down sequentiality. */
336 if (fp->f_seqcount > 1)
344 * Package up an I/O request on a vnode into a uio and do it.
347 vn_rdwr(rw, vp, base, len, offset, segflg, ioflg, active_cred, file_cred,
356 struct ucred *active_cred;
357 struct ucred *file_cred;
367 VFS_ASSERT_GIANT(vp->v_mount);
369 if ((ioflg & IO_NODELOCKED) == 0) {
371 if (rw == UIO_WRITE) {
372 if (vp->v_type != VCHR &&
373 (error = vn_start_write(vp, &mp, V_WAIT | PCATCH))
376 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
379 * XXX This should be LK_SHARED but I don't trust VFS
380 * enough to leave it like that until it has been
383 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
387 ASSERT_VOP_LOCKED(vp, "IO_NODELOCKED with no vp lock held");
388 auio.uio_iov = &aiov;
390 aiov.iov_base = base;
392 auio.uio_resid = len;
393 auio.uio_offset = offset;
394 auio.uio_segflg = segflg;
399 if ((ioflg & IO_NOMACCHECK) == 0) {
401 error = mac_vnode_check_read(active_cred, file_cred,
404 error = mac_vnode_check_write(active_cred, file_cred,
414 error = VOP_READ(vp, &auio, ioflg, cred);
416 error = VOP_WRITE(vp, &auio, ioflg, cred);
419 *aresid = auio.uio_resid;
421 if (auio.uio_resid && error == 0)
423 if ((ioflg & IO_NODELOCKED) == 0) {
424 if (rw == UIO_WRITE && vp->v_type != VCHR)
425 vn_finished_write(mp);
432 * Package up an I/O request on a vnode into a uio and do it. The I/O
433 * request is split up into smaller chunks and we try to avoid saturating
434 * the buffer cache while potentially holding a vnode locked, so we
435 * check bwillwrite() before calling vn_rdwr(). We also call uio_yield()
436 * to give other processes a chance to lock the vnode (either other processes
437 * core'ing the same binary, or unrelated processes scanning the directory).
440 vn_rdwr_inchunks(rw, vp, base, len, offset, segflg, ioflg, active_cred,
441 file_cred, aresid, td)
449 struct ucred *active_cred;
450 struct ucred *file_cred;
457 VFS_ASSERT_GIANT(vp->v_mount);
463 * Force `offset' to a multiple of MAXBSIZE except possibly
464 * for the first chunk, so that filesystems only need to
465 * write full blocks except possibly for the first and last
468 chunk = MAXBSIZE - (uoff_t)offset % MAXBSIZE;
472 if (rw != UIO_READ && vp->v_type == VREG)
475 error = vn_rdwr(rw, vp, base, chunk, offset, segflg,
476 ioflg, active_cred, file_cred, &iaresid, td);
477 len -= chunk; /* aresid calc already includes length */
481 base = (char *)base + chunk;
485 *aresid = len + iaresid;
490 * File table vnode read routine.
493 vn_read(fp, uio, active_cred, flags, td)
496 struct ucred *active_cred;
505 KASSERT(uio->uio_td == td, ("uio_td %p is not td %p",
510 if (fp->f_flag & FNONBLOCK)
512 if (fp->f_flag & O_DIRECT)
514 vfslocked = VFS_LOCK_GIANT(vp->v_mount);
515 VOP_LEASE(vp, td, fp->f_cred, LEASE_READ);
517 * According to McKusick the vn lock was protecting f_offset here.
518 * It is now protected by the FOFFSET_LOCKED flag.
520 if ((flags & FOF_OFFSET) == 0) {
521 mtxp = mtx_pool_find(mtxpool_sleep, fp);
523 while(fp->f_vnread_flags & FOFFSET_LOCKED) {
524 fp->f_vnread_flags |= FOFFSET_LOCK_WAITING;
525 msleep(&fp->f_vnread_flags, mtxp, PUSER -1,
526 "vnread offlock", 0);
528 fp->f_vnread_flags |= FOFFSET_LOCKED;
530 vn_lock(vp, LK_SHARED | LK_RETRY);
531 uio->uio_offset = fp->f_offset;
533 vn_lock(vp, LK_SHARED | LK_RETRY);
535 ioflag |= sequential_heuristic(uio, fp);
538 error = mac_vnode_check_read(active_cred, fp->f_cred, vp);
541 error = VOP_READ(vp, uio, ioflag, fp->f_cred);
542 if ((flags & FOF_OFFSET) == 0) {
543 fp->f_offset = uio->uio_offset;
545 if (fp->f_vnread_flags & FOFFSET_LOCK_WAITING)
546 wakeup(&fp->f_vnread_flags);
547 fp->f_vnread_flags = 0;
550 fp->f_nextoff = uio->uio_offset;
552 VFS_UNLOCK_GIANT(vfslocked);
557 * File table vnode write routine.
560 vn_write(fp, uio, active_cred, flags, td)
563 struct ucred *active_cred;
572 KASSERT(uio->uio_td == td, ("uio_td %p is not td %p",
575 vfslocked = VFS_LOCK_GIANT(vp->v_mount);
576 if (vp->v_type == VREG)
579 if (vp->v_type == VREG && (fp->f_flag & O_APPEND))
581 if (fp->f_flag & FNONBLOCK)
583 if (fp->f_flag & O_DIRECT)
585 if ((fp->f_flag & O_FSYNC) ||
586 (vp->v_mount && (vp->v_mount->mnt_flag & MNT_SYNCHRONOUS)))
589 if (vp->v_type != VCHR &&
590 (error = vn_start_write(vp, &mp, V_WAIT | PCATCH)) != 0)
592 VOP_LEASE(vp, td, fp->f_cred, LEASE_WRITE);
593 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
594 if ((flags & FOF_OFFSET) == 0)
595 uio->uio_offset = fp->f_offset;
596 ioflag |= sequential_heuristic(uio, fp);
598 error = mac_vnode_check_write(active_cred, fp->f_cred, vp);
601 error = VOP_WRITE(vp, uio, ioflag, fp->f_cred);
602 if ((flags & FOF_OFFSET) == 0)
603 fp->f_offset = uio->uio_offset;
604 fp->f_nextoff = uio->uio_offset;
606 if (vp->v_type != VCHR)
607 vn_finished_write(mp);
609 VFS_UNLOCK_GIANT(vfslocked);
614 * File table truncate routine.
617 vn_truncate(fp, length, active_cred, td)
620 struct ucred *active_cred;
630 vfslocked = VFS_LOCK_GIANT(vp->v_mount);
631 error = vn_start_write(vp, &mp, V_WAIT | PCATCH);
633 VFS_UNLOCK_GIANT(vfslocked);
636 VOP_LEASE(vp, td, active_cred, LEASE_WRITE);
637 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
638 if (vp->v_type == VDIR) {
643 error = mac_vnode_check_write(active_cred, fp->f_cred, vp);
647 error = vn_writechk(vp);
650 vattr.va_size = length;
651 error = VOP_SETATTR(vp, &vattr, fp->f_cred, td);
655 vn_finished_write(mp);
656 VFS_UNLOCK_GIANT(vfslocked);
661 * File table vnode stat routine.
664 vn_statfile(fp, sb, active_cred, td)
667 struct ucred *active_cred;
670 struct vnode *vp = fp->f_vnode;
674 vfslocked = VFS_LOCK_GIANT(vp->v_mount);
675 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
676 error = vn_stat(vp, sb, active_cred, fp->f_cred, td);
678 VFS_UNLOCK_GIANT(vfslocked);
684 * Stat a vnode; implementation for the stat syscall
687 vn_stat(vp, sb, active_cred, file_cred, td)
689 register struct stat *sb;
690 struct ucred *active_cred;
691 struct ucred *file_cred;
695 register struct vattr *vap;
700 error = mac_vnode_check_stat(active_cred, file_cred, vp);
706 error = VOP_GETATTR(vp, vap, active_cred, td);
711 * Zero the spare stat fields
713 bzero(sb, sizeof *sb);
716 * Copy from vattr table
718 if (vap->va_fsid != VNOVAL)
719 sb->st_dev = vap->va_fsid;
721 sb->st_dev = vp->v_mount->mnt_stat.f_fsid.val[0];
722 sb->st_ino = vap->va_fileid;
724 switch (vap->va_type) {
739 /* This is a cosmetic change, symlinks do not have a mode. */
740 if (vp->v_mount->mnt_flag & MNT_NOSYMFOLLOW)
741 sb->st_mode &= ~ACCESSPERMS; /* 0000 */
743 sb->st_mode |= ACCESSPERMS; /* 0777 */
755 sb->st_nlink = vap->va_nlink;
756 sb->st_uid = vap->va_uid;
757 sb->st_gid = vap->va_gid;
758 sb->st_rdev = vap->va_rdev;
759 if (vap->va_size > OFF_MAX)
761 sb->st_size = vap->va_size;
762 sb->st_atimespec = vap->va_atime;
763 sb->st_mtimespec = vap->va_mtime;
764 sb->st_ctimespec = vap->va_ctime;
765 sb->st_birthtimespec = vap->va_birthtime;
768 * According to www.opengroup.org, the meaning of st_blksize is
769 * "a filesystem-specific preferred I/O block size for this
770 * object. In some filesystem types, this may vary from file
772 * Default to PAGE_SIZE after much discussion.
773 * XXX: min(PAGE_SIZE, vp->v_bufobj.bo_bsize) may be more correct.
776 sb->st_blksize = PAGE_SIZE;
778 sb->st_flags = vap->va_flags;
779 if (priv_check(td, PRIV_VFS_GENERATION))
782 sb->st_gen = vap->va_gen;
784 sb->st_blocks = vap->va_bytes / S_BLKSIZE;
789 * File table vnode ioctl routine.
792 vn_ioctl(fp, com, data, active_cred, td)
796 struct ucred *active_cred;
799 struct vnode *vp = fp->f_vnode;
804 vfslocked = VFS_LOCK_GIANT(vp->v_mount);
806 switch (vp->v_type) {
809 if (com == FIONREAD) {
810 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
811 error = VOP_GETATTR(vp, &vattr, active_cred, td);
814 *(int *)data = vattr.va_size - fp->f_offset;
816 if (com == FIONBIO || com == FIOASYNC) /* XXX */
819 error = VOP_IOCTL(vp, com, data, fp->f_flag,
826 VFS_UNLOCK_GIANT(vfslocked);
831 * File table vnode poll routine.
834 vn_poll(fp, events, active_cred, td)
837 struct ucred *active_cred;
845 vfslocked = VFS_LOCK_GIANT(vp->v_mount);
847 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
848 error = mac_vnode_check_poll(active_cred, fp->f_cred, vp);
853 error = VOP_POLL(vp, events, fp->f_cred, td);
854 VFS_UNLOCK_GIANT(vfslocked);
859 * Acquire the requested lock and then check for validity. LK_RETRY
860 * permits vn_lock to return doomed vnodes.
863 _vn_lock(struct vnode *vp, int flags, char *file, int line)
867 VNASSERT((flags & LK_TYPE_MASK) != 0, vp,
868 ("vn_lock called with no locktype."));
870 error = VOP_LOCK1(vp, flags, file, line);
871 flags &= ~LK_INTERLOCK; /* Interlock is always dropped. */
872 KASSERT((flags & LK_RETRY) == 0 || error == 0,
873 ("LK_RETRY set with incompatible flags %d\n", flags));
875 * Callers specify LK_RETRY if they wish to get dead vnodes.
876 * If RETRY is not set, we return ENOENT instead.
878 if (error == 0 && vp->v_iflag & VI_DOOMED &&
879 (flags & LK_RETRY) == 0) {
884 } while (flags & LK_RETRY && error != 0);
889 * File table vnode close routine.
903 vfslocked = VFS_LOCK_GIANT(vp->v_mount);
904 if (fp->f_type == DTYPE_VNODE && fp->f_flag & FHASLOCK) {
905 lf.l_whence = SEEK_SET;
909 (void) VOP_ADVLOCK(vp, fp, F_UNLCK, &lf, F_FLOCK);
912 fp->f_ops = &badfileops;
914 error = vn_close(vp, fp->f_flag, fp->f_cred, td);
915 VFS_UNLOCK_GIANT(vfslocked);
920 * Preparing to start a filesystem write operation. If the operation is
921 * permitted, then we bump the count of operations in progress and
922 * proceed. If a suspend request is in progress, we wait until the
923 * suspension is over, and then proceed.
926 vn_start_write(vp, mpp, flags)
936 * If a vnode is provided, get and return the mount point that
937 * to which it will write.
940 if ((error = VOP_GETWRITEMOUNT(vp, mpp)) != 0) {
942 if (error != EOPNOTSUPP)
947 if ((mp = *mpp) == NULL)
953 * Check on status of suspension.
955 while ((mp->mnt_kern_flag & MNTK_SUSPEND) != 0) {
956 if (flags & V_NOWAIT) {
960 error = msleep(&mp->mnt_flag, MNT_MTX(mp),
961 (PUSER - 1) | (flags & PCATCH), "suspfs", 0);
965 if (flags & V_XSLEEP)
967 mp->mnt_writeopcount++;
975 * Secondary suspension. Used by operations such as vop_inactive
976 * routines that are needed by the higher level functions. These
977 * are allowed to proceed until all the higher level functions have
978 * completed (indicated by mnt_writeopcount dropping to zero). At that
979 * time, these operations are halted until the suspension is over.
982 vn_write_suspend_wait(vp, mp, flags)
990 if ((error = VOP_GETWRITEMOUNT(vp, &mp)) != 0) {
991 if (error != EOPNOTSUPP)
997 * If we are not suspended or have not yet reached suspended
998 * mode, then let the operation proceed.
1005 if ((mp->mnt_kern_flag & MNTK_SUSPENDED) == 0) {
1010 if (flags & V_NOWAIT) {
1013 return (EWOULDBLOCK);
1016 * Wait for the suspension to finish.
1018 error = msleep(&mp->mnt_flag, MNT_MTX(mp),
1019 (PUSER - 1) | (flags & PCATCH) | PDROP, "suspfs", 0);
1025 * Secondary suspension. Used by operations such as vop_inactive
1026 * routines that are needed by the higher level functions. These
1027 * are allowed to proceed until all the higher level functions have
1028 * completed (indicated by mnt_writeopcount dropping to zero). At that
1029 * time, these operations are halted until the suspension is over.
1032 vn_start_secondary_write(vp, mpp, flags)
1042 if ((error = VOP_GETWRITEMOUNT(vp, mpp)) != 0) {
1044 if (error != EOPNOTSUPP)
1050 * If we are not suspended or have not yet reached suspended
1051 * mode, then let the operation proceed.
1053 if ((mp = *mpp) == NULL)
1058 if ((mp->mnt_kern_flag & (MNTK_SUSPENDED | MNTK_SUSPEND2)) == 0) {
1059 mp->mnt_secondary_writes++;
1060 mp->mnt_secondary_accwrites++;
1065 if (flags & V_NOWAIT) {
1068 return (EWOULDBLOCK);
1071 * Wait for the suspension to finish.
1073 error = msleep(&mp->mnt_flag, MNT_MTX(mp),
1074 (PUSER - 1) | (flags & PCATCH) | PDROP, "suspfs", 0);
1082 * Filesystem write operation has completed. If we are suspending and this
1083 * operation is the last one, notify the suspender that the suspension is
1087 vn_finished_write(mp)
1093 mp->mnt_writeopcount--;
1094 if (mp->mnt_writeopcount < 0)
1095 panic("vn_finished_write: neg cnt");
1096 if ((mp->mnt_kern_flag & MNTK_SUSPEND) != 0 &&
1097 mp->mnt_writeopcount <= 0)
1098 wakeup(&mp->mnt_writeopcount);
1104 * Filesystem secondary write operation has completed. If we are
1105 * suspending and this operation is the last one, notify the suspender
1106 * that the suspension is now in effect.
1109 vn_finished_secondary_write(mp)
1115 mp->mnt_secondary_writes--;
1116 if (mp->mnt_secondary_writes < 0)
1117 panic("vn_finished_secondary_write: neg cnt");
1118 if ((mp->mnt_kern_flag & MNTK_SUSPEND) != 0 &&
1119 mp->mnt_secondary_writes <= 0)
1120 wakeup(&mp->mnt_secondary_writes);
1127 * Request a filesystem to suspend write operations.
1130 vfs_write_suspend(mp)
1133 struct thread *td = curthread;
1137 if (mp->mnt_kern_flag & MNTK_SUSPEND) {
1141 mp->mnt_kern_flag |= MNTK_SUSPEND;
1142 if (mp->mnt_writeopcount > 0)
1143 (void) msleep(&mp->mnt_writeopcount,
1144 MNT_MTX(mp), (PUSER - 1)|PDROP, "suspwt", 0);
1147 if ((error = VFS_SYNC(mp, MNT_SUSPEND, td)) != 0)
1148 vfs_write_resume(mp);
1153 * Request a filesystem to resume write operations.
1156 vfs_write_resume(mp)
1161 if ((mp->mnt_kern_flag & MNTK_SUSPEND) != 0) {
1162 mp->mnt_kern_flag &= ~(MNTK_SUSPEND | MNTK_SUSPEND2 |
1164 wakeup(&mp->mnt_writeopcount);
1165 wakeup(&mp->mnt_flag);
1171 * Implement kqueues for files by translating it to vnode operation.
1174 vn_kqfilter(struct file *fp, struct knote *kn)
1179 vfslocked = VFS_LOCK_GIANT(fp->f_vnode->v_mount);
1180 error = VOP_KQFILTER(fp->f_vnode, kn);
1181 VFS_UNLOCK_GIANT(vfslocked);
1187 * Simplified in-kernel wrapper calls for extended attribute access.
1188 * Both calls pass in a NULL credential, authorizing as "kernel" access.
1189 * Set IO_NODELOCKED in ioflg if the vnode is already locked.
1192 vn_extattr_get(struct vnode *vp, int ioflg, int attrnamespace,
1193 const char *attrname, int *buflen, char *buf, struct thread *td)
1199 iov.iov_len = *buflen;
1202 auio.uio_iov = &iov;
1203 auio.uio_iovcnt = 1;
1204 auio.uio_rw = UIO_READ;
1205 auio.uio_segflg = UIO_SYSSPACE;
1207 auio.uio_offset = 0;
1208 auio.uio_resid = *buflen;
1210 if ((ioflg & IO_NODELOCKED) == 0)
1211 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
1213 ASSERT_VOP_LOCKED(vp, "IO_NODELOCKED with no vp lock held");
1215 /* authorize attribute retrieval as kernel */
1216 error = VOP_GETEXTATTR(vp, attrnamespace, attrname, &auio, NULL, NULL,
1219 if ((ioflg & IO_NODELOCKED) == 0)
1223 *buflen = *buflen - auio.uio_resid;
1230 * XXX failure mode if partially written?
1233 vn_extattr_set(struct vnode *vp, int ioflg, int attrnamespace,
1234 const char *attrname, int buflen, char *buf, struct thread *td)
1241 iov.iov_len = buflen;
1244 auio.uio_iov = &iov;
1245 auio.uio_iovcnt = 1;
1246 auio.uio_rw = UIO_WRITE;
1247 auio.uio_segflg = UIO_SYSSPACE;
1249 auio.uio_offset = 0;
1250 auio.uio_resid = buflen;
1252 if ((ioflg & IO_NODELOCKED) == 0) {
1253 if ((error = vn_start_write(vp, &mp, V_WAIT)) != 0)
1255 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
1258 ASSERT_VOP_LOCKED(vp, "IO_NODELOCKED with no vp lock held");
1260 /* authorize attribute setting as kernel */
1261 error = VOP_SETEXTATTR(vp, attrnamespace, attrname, &auio, NULL, td);
1263 if ((ioflg & IO_NODELOCKED) == 0) {
1264 vn_finished_write(mp);
1272 vn_extattr_rm(struct vnode *vp, int ioflg, int attrnamespace,
1273 const char *attrname, struct thread *td)
1278 if ((ioflg & IO_NODELOCKED) == 0) {
1279 if ((error = vn_start_write(vp, &mp, V_WAIT)) != 0)
1281 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
1284 ASSERT_VOP_LOCKED(vp, "IO_NODELOCKED with no vp lock held");
1286 /* authorize attribute removal as kernel */
1287 error = VOP_DELETEEXTATTR(vp, attrnamespace, attrname, NULL, td);
1288 if (error == EOPNOTSUPP)
1289 error = VOP_SETEXTATTR(vp, attrnamespace, attrname, NULL,
1292 if ((ioflg & IO_NODELOCKED) == 0) {
1293 vn_finished_write(mp);