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 * 3. All advertising materials mentioning features or use of this software
19 * must display the following acknowledgement:
20 * This product includes software developed by the University of
21 * California, Berkeley and its contributors.
22 * 4. Neither the name of the University nor the names of its contributors
23 * may be used to endorse or promote products derived from this software
24 * without specific prior written permission.
26 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
27 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
28 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
29 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
30 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
31 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
32 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
33 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
34 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
35 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
38 * @(#)vfs_vnops.c 8.2 (Berkeley) 1/21/94
39 * $FreeBSD: src/sys/kern/vfs_vnops.c,v 1.87.2.13 2002/12/29 18:19:53 dillon Exp $
40 * $DragonFly: src/sys/kern/vfs_vnops.c,v 1.23 2004/10/12 19:20:46 dillon Exp $
43 #include <sys/param.h>
44 #include <sys/systm.h>
45 #include <sys/fcntl.h>
49 #include <sys/mount.h>
50 #include <sys/namei.h>
51 #include <sys/vnode.h>
53 #include <sys/filio.h>
54 #include <sys/ttycom.h>
56 #include <sys/syslog.h>
58 static int vn_closefile (struct file *fp, struct thread *td);
59 static int vn_ioctl (struct file *fp, u_long com, caddr_t data,
61 static int vn_read (struct file *fp, struct uio *uio,
62 struct ucred *cred, int flags, struct thread *td);
63 static int vn_poll (struct file *fp, int events, struct ucred *cred,
65 static int vn_kqfilter (struct file *fp, struct knote *kn);
66 static int vn_statfile (struct file *fp, struct stat *sb, struct thread *td);
67 static int vn_write (struct file *fp, struct uio *uio,
68 struct ucred *cred, int flags, struct thread *td);
70 struct fileops vnops = {
73 vn_read, vn_write, vn_ioctl, vn_poll, vn_kqfilter,
74 vn_statfile, vn_closefile
78 * Common code for vnode open operations.
79 * Check permissions, and call the VOP_OPEN or VOP_CREATE routine.
81 * Note that this does NOT free nameidata for the successful case,
82 * due to the NDINIT being done elsewhere.
85 vn_open(ndp, fmode, cmode)
86 struct nameidata *ndp;
90 struct thread *td = ndp->ni_cnd.cn_td;
91 struct ucred *cred = ndp->ni_cnd.cn_cred;
93 struct vattr *vap = &vat;
96 KKASSERT(cred == td->td_proc->p_ucred);
98 if (fmode & O_CREAT) {
99 ndp->ni_cnd.cn_nameiop = NAMEI_CREATE;
100 ndp->ni_cnd.cn_flags = CNP_LOCKPARENT | CNP_LOCKLEAF;
101 if ((fmode & O_EXCL) == 0 && (fmode & O_NOFOLLOW) == 0)
102 ndp->ni_cnd.cn_flags |= CNP_FOLLOW;
107 if (ndp->ni_vp == NULL) {
110 vap->va_mode = cmode;
112 vap->va_vaflags |= VA_EXCLUSIVE;
113 VOP_LEASE(ndp->ni_dvp, td, cred, LEASE_WRITE);
114 error = VOP_CREATE(ndp->ni_dvp, NCPNULL, &ndp->ni_vp,
117 NDFREE(ndp, NDF_ONLY_PNBUF);
122 ASSERT_VOP_UNLOCKED(ndp->ni_dvp, "create");
123 ASSERT_VOP_LOCKED(ndp->ni_vp, "create");
127 if (ndp->ni_dvp == ndp->ni_vp)
133 if (fmode & O_EXCL) {
140 ndp->ni_cnd.cn_nameiop = NAMEI_LOOKUP;
141 ndp->ni_cnd.cn_flags = CNP_LOCKLEAF |
142 ((fmode & O_NOFOLLOW) ? 0 : CNP_FOLLOW);
148 if (vp->v_type == VLNK) {
152 if (vp->v_type == VSOCK) {
156 if ((fmode & O_CREAT) == 0) {
158 if (fmode & (FWRITE | O_TRUNC)) {
159 if (vp->v_type == VDIR) {
163 error = vn_writechk(vp);
171 error = VOP_ACCESS(vp, mode, cred, td);
176 if (fmode & O_TRUNC) {
177 VOP_UNLOCK(vp, 0, td); /* XXX */
178 VOP_LEASE(vp, td, cred, LEASE_WRITE);
179 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY, td); /* XXX */
182 error = VOP_SETATTR(vp, vap, cred, td);
186 error = VOP_OPEN(vp, fmode, cred, td);
190 * Make sure that a VM object is created for VMIO support.
192 if (vn_canvmio(vp) == TRUE) {
193 if ((error = vfs_object_create(vp, td)) != 0)
201 NDFREE(ndp, NDF_ONLY_PNBUF);
207 * Check for write permissions on the specified vnode.
208 * Prototype text segments cannot be written.
216 * If there's shared text associated with
217 * the vnode, try to free it up once. If
218 * we fail, we can't allow writing.
220 if (vp->v_flag & VTEXT)
229 vn_close(struct vnode *vp, int flags, struct thread *td)
235 if ((error = vn_lock(vp, LK_EXCLUSIVE | LK_RETRY, td)) == 0) {
236 error = VOP_CLOSE(vp, flags, td);
237 VOP_UNLOCK(vp, 0, td);
245 sequential_heuristic(struct uio *uio, struct file *fp)
248 * Sequential heuristic - detect sequential operation
250 if ((uio->uio_offset == 0 && fp->f_seqcount > 0) ||
251 uio->uio_offset == fp->f_nextoff) {
252 int tmpseq = fp->f_seqcount;
254 * XXX we assume that the filesystem block size is
255 * the default. Not true, but still gives us a pretty
256 * good indicator of how sequential the read operations
259 tmpseq += (uio->uio_resid + BKVASIZE - 1) / BKVASIZE;
260 if (tmpseq > IO_SEQMAX)
262 fp->f_seqcount = tmpseq;
263 return(fp->f_seqcount << IO_SEQSHIFT);
267 * Not sequential, quick draw-down of seqcount
269 if (fp->f_seqcount > 1)
277 * Package up an I/O request on a vnode into a uio and do it.
279 * We are going to assume the caller has done the appropriate
280 * VOP_LEASE() call before calling vn_rdwr()
283 vn_rdwr(rw, vp, base, len, offset, segflg, ioflg, cred, aresid, td)
299 if ((ioflg & IO_NODELOCKED) == 0)
300 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY, td);
301 auio.uio_iov = &aiov;
303 aiov.iov_base = base;
305 auio.uio_resid = len;
306 auio.uio_offset = offset;
307 auio.uio_segflg = segflg;
310 if (rw == UIO_READ) {
311 error = VOP_READ(vp, &auio, ioflg, cred);
313 error = VOP_WRITE(vp, &auio, ioflg, cred);
316 *aresid = auio.uio_resid;
318 if (auio.uio_resid && error == 0)
320 if ((ioflg & IO_NODELOCKED) == 0)
321 VOP_UNLOCK(vp, 0, td);
326 * Package up an I/O request on a vnode into a uio and do it. The I/O
327 * request is split up into smaller chunks and we try to avoid saturating
328 * the buffer cache while potentially holding a vnode locked, so we
329 * check bwillwrite() before calling vn_rdwr(). We also call uio_yield()
330 * to give other processes a chance to lock the vnode (either other processes
331 * core'ing the same binary, or unrelated processes scanning the directory).
334 vn_rdwr_inchunks(rw, vp, base, len, offset, segflg, ioflg, cred, aresid, td)
352 * Force `offset' to a multiple of MAXBSIZE except possibly
353 * for the first chunk, so that filesystems only need to
354 * write full blocks except possibly for the first and last
357 chunk = MAXBSIZE - (uoff_t)offset % MAXBSIZE;
361 if (rw != UIO_READ && vp->v_type == VREG)
363 error = vn_rdwr(rw, vp, base, chunk, offset, segflg,
364 ioflg, cred, aresid, td);
365 len -= chunk; /* aresid calc already includes length */
378 * File table vnode read routine.
381 vn_read(fp, uio, cred, flags, td)
391 KASSERT(uio->uio_td == td, ("uio_td %p is not td %p", uio->uio_td, td));
392 vp = (struct vnode *)fp->f_data;
394 if (fp->f_flag & FNONBLOCK)
396 if (fp->f_flag & O_DIRECT)
398 VOP_LEASE(vp, td, cred, LEASE_READ);
399 vn_lock(vp, LK_SHARED | LK_NOPAUSE | LK_RETRY, td);
400 if ((flags & FOF_OFFSET) == 0)
401 uio->uio_offset = fp->f_offset;
403 ioflag |= sequential_heuristic(uio, fp);
405 error = VOP_READ(vp, uio, ioflag, cred);
406 if ((flags & FOF_OFFSET) == 0)
407 fp->f_offset = uio->uio_offset;
408 fp->f_nextoff = uio->uio_offset;
409 VOP_UNLOCK(vp, 0, td);
414 * File table vnode write routine.
417 vn_write(fp, uio, cred, flags, td)
427 KASSERT(uio->uio_td == td, ("uio_procp %p is not p %p",
429 vp = (struct vnode *)fp->f_data;
430 if (vp->v_type == VREG)
432 vp = (struct vnode *)fp->f_data; /* XXX needed? */
434 if (vp->v_type == VREG && (fp->f_flag & O_APPEND))
436 if (fp->f_flag & FNONBLOCK)
438 if (fp->f_flag & O_DIRECT)
440 if ((fp->f_flag & O_FSYNC) ||
441 (vp->v_mount && (vp->v_mount->mnt_flag & MNT_SYNCHRONOUS)))
443 VOP_LEASE(vp, td, cred, LEASE_WRITE);
444 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY, td);
445 if ((flags & FOF_OFFSET) == 0)
446 uio->uio_offset = fp->f_offset;
447 ioflag |= sequential_heuristic(uio, fp);
448 error = VOP_WRITE(vp, uio, ioflag, cred);
449 if ((flags & FOF_OFFSET) == 0)
450 fp->f_offset = uio->uio_offset;
451 fp->f_nextoff = uio->uio_offset;
452 VOP_UNLOCK(vp, 0, td);
457 * File table vnode stat routine.
460 vn_statfile(struct file *fp, struct stat *sb, struct thread *td)
462 struct vnode *vp = (struct vnode *)fp->f_data;
464 return vn_stat(vp, sb, td);
468 vn_stat(struct vnode *vp, struct stat *sb, struct thread *td)
476 error = VOP_GETATTR(vp, vap, td);
481 * Zero the spare stat fields
484 sb->st_qspare[0] = 0;
485 sb->st_qspare[1] = 0;
488 * Copy from vattr table
490 if (vap->va_fsid != VNOVAL)
491 sb->st_dev = vap->va_fsid;
493 sb->st_dev = vp->v_mount->mnt_stat.f_fsid.val[0];
494 sb->st_ino = vap->va_fileid;
496 switch (vap->va_type) {
511 /* This is a cosmetic change, symlinks do not have a mode. */
512 if (vp->v_mount->mnt_flag & MNT_NOSYMFOLLOW)
513 sb->st_mode &= ~ACCESSPERMS; /* 0000 */
515 sb->st_mode |= ACCESSPERMS; /* 0777 */
527 sb->st_nlink = vap->va_nlink;
528 sb->st_uid = vap->va_uid;
529 sb->st_gid = vap->va_gid;
530 sb->st_rdev = vap->va_rdev;
531 sb->st_size = vap->va_size;
532 sb->st_atimespec = vap->va_atime;
533 sb->st_mtimespec = vap->va_mtime;
534 sb->st_ctimespec = vap->va_ctime;
537 * According to www.opengroup.org, the meaning of st_blksize is
538 * "a filesystem-specific preferred I/O block size for this
539 * object. In some filesystem types, this may vary from file
541 * Default to PAGE_SIZE after much discussion.
544 if (vap->va_type == VREG) {
545 sb->st_blksize = vap->va_blocksize;
546 } else if (vn_isdisk(vp, NULL)) {
548 * XXX this is broken. If the device is not yet open (aka
549 * stat() call, aka v_rdev == NULL), how are we supposed
550 * to get a valid block size out of it?
554 if ((dev = vp->v_rdev) == NULL)
555 dev = udev2dev(vp->v_udev, vp->v_type == VBLK);
556 sb->st_blksize = dev->si_bsize_best;
557 if (sb->st_blksize < dev->si_bsize_phys)
558 sb->st_blksize = dev->si_bsize_phys;
559 if (sb->st_blksize < BLKDEV_IOSIZE)
560 sb->st_blksize = BLKDEV_IOSIZE;
562 sb->st_blksize = PAGE_SIZE;
565 sb->st_flags = vap->va_flags;
569 sb->st_gen = vap->va_gen;
571 #if (S_BLKSIZE == 512)
572 /* Optimize this case */
573 sb->st_blocks = vap->va_bytes >> 9;
575 sb->st_blocks = vap->va_bytes / S_BLKSIZE;
581 * File table vnode ioctl routine.
584 vn_ioctl(struct file *fp, u_long com, caddr_t data, struct thread *td)
586 struct vnode *vp = ((struct vnode *)fp->f_data);
592 KKASSERT(td->td_proc != NULL);
593 ucred = td->td_proc->p_ucred;
595 switch (vp->v_type) {
598 if (com == FIONREAD) {
599 error = VOP_GETATTR(vp, &vattr, td);
602 *(int *)data = vattr.va_size - fp->f_offset;
605 if (com == FIONBIO || com == FIOASYNC) /* XXX */
606 return (0); /* XXX */
615 if (com == FIODTYPE) {
616 if (vp->v_type != VCHR && vp->v_type != VBLK)
618 *(int *)data = dev_dflags(vp->v_rdev) & D_TYPEMASK;
621 error = VOP_IOCTL(vp, com, data, fp->f_flag, ucred, td);
622 if (error == 0 && com == TIOCSCTTY) {
623 struct session *sess = td->td_proc->p_session;
625 /* Do nothing if reassigning same control tty */
626 if (sess->s_ttyvp == vp)
629 /* Get rid of reference to old control tty */
641 * File table vnode poll routine.
644 vn_poll(struct file *fp, int events, struct ucred *cred, struct thread *td)
646 return (VOP_POLL(((struct vnode *)fp->f_data), events, cred, td));
650 * Check that the vnode is still valid, and if so
651 * acquire requested lock.
655 vn_lock(struct vnode *vp, int flags, struct thread *td)
657 debug_vn_lock(struct vnode *vp, int flags, struct thread *td,
658 const char *filename, int line)
665 vp->filename = filename;
668 error = VOP_LOCK(vp, flags | LK_NOPAUSE, td);
671 } while (flags & LK_RETRY);
674 * Because we (had better!) have a ref on the vnode, once it
675 * goes to VRECLAIMED state it will not be recycled until all
676 * refs go away. So we can just check the flag.
678 if (error == 0 && (vp->v_flag & VRECLAIMED)) {
679 VOP_UNLOCK(vp, 0, td);
686 * File table vnode close routine.
689 vn_closefile(struct file *fp, struct thread *td)
693 fp->f_ops = &badfileops;
694 err = vn_close(((struct vnode *)fp->f_data), fp->f_flag, td);
699 vn_kqfilter(struct file *fp, struct knote *kn)
702 return (VOP_KQFILTER(((struct vnode *)fp->f_data), kn));