2 * Copyright (c) 1989, 1991, 1993, 1994
3 * The Regents of the University of California. All rights reserved.
5 * Redistribution and use in source and binary forms, with or without
6 * modification, are permitted provided that the following conditions
8 * 1. Redistributions of source code must retain the above copyright
9 * notice, this list of conditions and the following disclaimer.
10 * 2. Redistributions in binary form must reproduce the above copyright
11 * notice, this list of conditions and the following disclaimer in the
12 * documentation and/or other materials provided with the distribution.
13 * 3. All advertising materials mentioning features or use of this software
14 * must display the following acknowledgement:
15 * This product includes software developed by the University of
16 * California, Berkeley and its contributors.
17 * 4. Neither the name of the University nor the names of its contributors
18 * may be used to endorse or promote products derived from this software
19 * without specific prior written permission.
21 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
22 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
23 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
24 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
25 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
26 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
27 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
28 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
29 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
30 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
33 * @(#)ffs_vfsops.c 8.31 (Berkeley) 5/20/95
34 * $FreeBSD: src/sys/ufs/ffs/ffs_vfsops.c,v 1.117.2.10 2002/06/23 22:34:52 iedowse Exp $
35 * $DragonFly: src/sys/vfs/ufs/ffs_vfsops.c,v 1.32 2005/04/15 19:08:32 dillon Exp $
38 #include "opt_quota.h"
40 #include <sys/param.h>
41 #include <sys/systm.h>
43 #include <sys/nlookup.h>
44 #include <sys/kernel.h>
45 #include <sys/vnode.h>
46 #include <sys/mount.h>
49 #include <sys/fcntl.h>
50 #include <sys/disklabel.h>
51 #include <sys/malloc.h>
56 #include "ufs_extern.h"
59 #include "ffs_extern.h"
62 #include <vm/vm_page.h>
63 #include <vm/vm_zone.h>
65 static MALLOC_DEFINE(M_FFSNODE, "FFS node", "FFS vnode private part");
67 static int ffs_sbupdate (struct ufsmount *, int);
68 static int ffs_reload (struct mount *,struct ucred *,struct thread *);
69 static int ffs_oldfscompat (struct fs *);
70 static int ffs_mount (struct mount *, char *, caddr_t, struct thread *);
71 static int ffs_init (struct vfsconf *);
73 static struct vfsops ufs_vfsops = {
90 VFS_SET(ufs_vfsops, ufs, 0);
92 extern struct vnodeopv_entry_desc ffs_vnodeop_entries[];
93 extern struct vnodeopv_entry_desc ffs_specop_entries[];
94 extern struct vnodeopv_entry_desc ffs_fifoop_entries[];
100 * Called when mounting local physical media
104 * mp mount point structure
105 * path NULL (flag for root mount!!!)
107 * p process (user credentials check [statfs])
110 * mp mount point structure
111 * path path to mount point
112 * data pointer to argument struct in user space
113 * p process (user credentials check)
116 * !0 error number (errno.h)
121 * mount point is locked
123 * mount point is locked
126 * A NULL path can be used for a flag since the mount
127 * system call will fail with EFAULT in copyinstr in
128 * nlookup() if it is a genuine NULL from the user.
131 ffs_mount(struct mount *mp, /* mount struct pointer */
132 char *path, /* path to mount point */
133 caddr_t data, /* arguments to FS specific mount */
134 struct thread *td) /* process requesting mount */
140 struct ufs_args args;
141 struct ufsmount *ump = 0;
143 int flags, ronly = 0;
146 struct nlookupdata nd;
147 struct vnode *rootvp;
149 KKASSERT(td->td_proc);
150 cred = td->td_proc->p_ucred;
154 * Use NULL path to flag a root mount
159 * Mounting root filesystem
163 if ((error = bdevvp(rootdev, &rootvp))) {
164 printf("ffs_mountroot: can't find rootvp\n");
168 if( ( error = ffs_mountfs(rootvp, mp, td, M_FFSNODE)) != 0) {
169 /* fs specific cleanup (if any)*/
173 goto dostatfs; /* success*/
179 * Mounting non-root filesystem or updating a filesystem
183 /* copy in user arguments*/
184 error = copyin(data, (caddr_t)&args, sizeof (struct ufs_args));
186 goto error_1; /* can't get arguments*/
189 * If updating, check whether changing from read-only to
190 * read/write; if there is no device name, that's all we do.
192 if (mp->mnt_flag & MNT_UPDATE) {
195 devvp = ump->um_devvp;
197 ronly = fs->fs_ronly; /* MNT_RELOAD might change this */
198 if (ronly == 0 && (mp->mnt_flag & MNT_RDONLY)) {
200 * Flush any dirty data.
202 VFS_SYNC(mp, MNT_WAIT, td);
204 * Check for and optionally get rid of files open
208 if (mp->mnt_flag & MNT_FORCE)
210 if (mp->mnt_flag & MNT_SOFTDEP) {
211 error = softdep_flushfiles(mp, flags, td);
213 error = ffs_flushfiles(mp, flags, td);
217 if (!error && (mp->mnt_flag & MNT_RELOAD))
218 error = ffs_reload(mp, NULL, td);
222 if (ronly && (mp->mnt_kern_flag & MNTK_WANTRDWR)) {
224 * If upgrade to read-write by non-root, then verify
225 * that user has necessary permissions on the device.
227 if (cred->cr_uid != 0) {
228 vn_lock(devvp, LK_EXCLUSIVE | LK_RETRY, td);
229 if ((error = VOP_ACCESS(devvp, VREAD | VWRITE,
231 VOP_UNLOCK(devvp, 0, td);
234 VOP_UNLOCK(devvp, 0, td);
237 fs->fs_flags &= ~FS_UNCLEAN;
238 if (fs->fs_clean == 0) {
239 fs->fs_flags |= FS_UNCLEAN;
240 if (mp->mnt_flag & MNT_FORCE) {
242 "WARNING: %s was not properly dismounted\n",
246 "WARNING: R/W mount of %s denied. Filesystem is not clean - run fsck\n",
253 /* check to see if we need to start softdep */
254 if (fs->fs_flags & FS_DOSOFTDEP) {
255 error = softdep_mount(devvp, mp, fs);
263 * Soft updates is incompatible with "async",
264 * so if we are doing softupdates stop the user
265 * from setting the async flag in an update.
266 * Softdep_mount() clears it in an initial mount
269 if (mp->mnt_flag & MNT_SOFTDEP) {
270 mp->mnt_flag &= ~MNT_ASYNC;
272 /* if not updating name...*/
273 if (args.fspec == 0) {
275 * Process export requests. Jumping to "success"
276 * will return the vfs_export() error code.
278 error = vfs_export(mp, &ump->um_export, &args.export);
284 * Not an update, or updating the name: look up the name
285 * and verify that it refers to a sensible block device.
288 error = nlookup_init(&nd, args.fspec, UIO_USERSPACE, NLC_FOLLOW);
290 error = nlookup(&nd);
292 error = cache_vref(nd.nl_ncp, nd.nl_cred, &devvp);
297 if (!vn_isdisk(devvp, &error))
301 * If mount by non-root, then verify that user has necessary
302 * permissions on the device.
304 if (cred->cr_uid != 0) {
306 if ((mp->mnt_flag & MNT_RDONLY) == 0)
307 accessmode |= VWRITE;
308 vn_lock(devvp, LK_EXCLUSIVE | LK_RETRY, td);
309 if ((error = VOP_ACCESS(devvp, accessmode, cred, td)) != 0) {
313 VOP_UNLOCK(devvp, 0, td);
316 if (mp->mnt_flag & MNT_UPDATE) {
320 * If it's not the same vnode, or at least the same device
321 * then it's not correct. NOTE: devvp->v_rdev may be NULL
322 * since we haven't opened it, so we compare udev instead.
325 if (devvp != ump->um_devvp) {
326 if (devvp->v_udev == ump->um_devvp->v_udev) {
329 printf("cannot update mount, udev does"
330 " not match %08x vs %08x\n",
331 devvp->v_udev, ump->um_devvp->v_udev);
332 error = EINVAL; /* needs translation */
338 * Update device name only on success
341 /* Save "mounted from" info for mount point (NULL pad)*/
342 copyinstr( args.fspec,
343 mp->mnt_stat.f_mntfromname,
346 bzero( mp->mnt_stat.f_mntfromname + size, MNAMELEN - size);
355 /* Save "mounted from" info for mount point (NULL pad)*/
356 copyinstr( args.fspec, /* device name*/
357 mp->mnt_stat.f_mntfromname, /* save area*/
358 MNAMELEN - 1, /* max size*/
359 &size); /* real size*/
360 bzero( mp->mnt_stat.f_mntfromname + size, MNAMELEN - size);
362 error = ffs_mountfs(devvp, mp, td, M_FFSNODE);
370 * Initialize FS stat information in mount struct; uses
371 * mp->mnt_stat.f_mntfromname.
373 * This code is common to root and non-root mounts
375 (void)VFS_STATFS(mp, &mp->mnt_stat, td);
380 error_2: /* error with devvp held*/
382 /* release devvp before failing*/
385 error_1: /* no state to back out*/
388 if (!error && path && (mp->mnt_flag & MNT_UPDATE)) {
389 /* Update clean flag after changing read-onlyness. */
391 if (ronly != fs->fs_ronly) {
392 fs->fs_ronly = ronly;
393 fs->fs_clean = ronly &&
394 (fs->fs_flags & FS_UNCLEAN) == 0 ? 1 : 0;
395 ffs_sbupdate(ump, MNT_WAIT);
402 * Reload all incore data for a filesystem (used after running fsck on
403 * the root filesystem and finding things to fix). The filesystem must
404 * be mounted read-only.
406 * Things to do to update the mount:
407 * 1) invalidate all cached meta-data.
408 * 2) re-read superblock from disk.
409 * 3) re-read summary information from disk.
410 * 4) invalidate all inactive vnodes.
411 * 5) invalidate all cached file data.
412 * 6) re-read inode data for all active vnodes.
415 static int ffs_reload_scan2(struct mount *mp, struct vnode *vp, void *data);
427 ffs_reload(struct mount *mp, struct ucred *cred, struct thread *td)
432 struct fs *fs, *newfs;
433 struct partinfo dpart;
435 int i, blks, size, error;
436 struct scaninfo scaninfo;
439 if ((mp->mnt_flag & MNT_RDONLY) == 0)
442 * Step 1: invalidate all cached meta-data.
444 devvp = VFSTOUFS(mp)->um_devvp;
445 vn_lock(devvp, LK_EXCLUSIVE | LK_RETRY, td);
446 error = vinvalbuf(devvp, 0, td, 0, 0);
447 VOP_UNLOCK(devvp, 0, td);
449 panic("ffs_reload: dirty1");
453 * Only VMIO the backing device if the backing device is a real
454 * block device. See ffs_mountmfs() for more details.
456 if (devvp->v_tag != VT_MFS && vn_isdisk(devvp, NULL)) {
457 vn_lock(devvp, LK_EXCLUSIVE | LK_RETRY, td);
458 vfs_object_create(devvp, td);
459 VOP_UNLOCK(devvp, 0, td);
463 * Step 2: re-read superblock from disk.
465 if (VOP_IOCTL(devvp, DIOCGPART, (caddr_t)&dpart, FREAD, NOCRED, td) != 0)
468 size = dpart.disklab->d_secsize;
469 if ((error = bread(devvp, (ufs_daddr_t)(SBOFF/size), SBSIZE, &bp)) != 0)
474 newfs = (struct fs *)bp->b_data;
475 if (newfs->fs_magic != FS_MAGIC || newfs->fs_bsize > MAXBSIZE ||
476 newfs->fs_bsize < sizeof(struct fs)) {
478 return (EIO); /* XXX needs translation */
480 fs = VFSTOUFS(mp)->um_fs;
482 * Copy pointer fields back into superblock before copying in XXX
483 * new superblock. These should really be in the ufsmount. XXX
484 * Note that important parameters (eg fs_ncg) are unchanged.
486 newfs->fs_csp = fs->fs_csp;
487 newfs->fs_maxcluster = fs->fs_maxcluster;
488 newfs->fs_contigdirs = fs->fs_contigdirs;
489 /* The filesystem is still read-only. */
491 bcopy(newfs, fs, (uint)fs->fs_sbsize);
492 if (fs->fs_sbsize < SBSIZE)
493 bp->b_flags |= B_INVAL;
495 mp->mnt_maxsymlinklen = fs->fs_maxsymlinklen;
497 /* An old fsck may have zeroed these fields, so recheck them. */
498 if (fs->fs_avgfilesize <= 0) /* XXX */
499 fs->fs_avgfilesize = AVFILESIZ; /* XXX */
500 if (fs->fs_avgfpdir <= 0) /* XXX */
501 fs->fs_avgfpdir = AFPDIR; /* XXX */
504 * Step 3: re-read summary information from disk.
506 blks = howmany(fs->fs_cssize, fs->fs_fsize);
508 for (i = 0; i < blks; i += fs->fs_frag) {
510 if (i + fs->fs_frag > blks)
511 size = (blks - i) * fs->fs_fsize;
512 error = bread(devvp, fsbtodb(fs, fs->fs_csaddr + i), size, &bp);
517 bcopy(bp->b_data, space, (uint)size);
518 space = (char *)space + size;
522 * We no longer know anything about clusters per cylinder group.
524 if (fs->fs_contigsumsize > 0) {
525 lp = fs->fs_maxcluster;
526 for (i = 0; i < fs->fs_ncg; i++)
527 *lp++ = fs->fs_contigsumsize;
532 scaninfo.devvp = devvp;
534 while (error == 0 && scaninfo.rescan) {
536 error = vmntvnodescan(mp, VMSC_GETVX,
537 NULL, ffs_reload_scan2, &scaninfo);
543 ffs_reload_scan2(struct mount *mp, struct vnode *vp, void *data)
545 struct scaninfo *info = data;
553 if (vrecycle(vp, info->td))
556 if (vinvalbuf(vp, 0, info->td, 0, 0))
557 panic("ffs_reload: dirty2");
559 * Step 6: re-read inode data for all active vnodes.
562 error = bread(info->devvp,
563 fsbtodb(info->fs, ino_to_fsba(info->fs, ip->i_number)),
564 (int)info->fs->fs_bsize, &bp);
569 ip->i_din = *((struct dinode *)bp->b_data +
570 ino_to_fsbo(info->fs, ip->i_number));
571 ip->i_effnlink = ip->i_nlink;
577 * Common code for mount and mountroot
580 ffs_mountfs(struct vnode *devvp, struct mount *mp, struct thread *td,
581 struct malloc_type *malloctype)
583 struct ufsmount *ump;
587 struct partinfo dpart;
589 int error, i, blks, size, ronly;
591 uint64_t maxfilesize; /* XXX */
594 * Disallow multiple mounts of the same device.
595 * Disallow mounting of a device that is currently in use
596 * Flush out any old buffers remaining from a previous use.
598 error = vfs_mountedon(devvp);
601 if (count_udev(devvp->v_udev) > 0)
603 vn_lock(devvp, LK_EXCLUSIVE | LK_RETRY, td);
604 error = vinvalbuf(devvp, V_SAVE, td, 0, 0);
605 VOP_UNLOCK(devvp, 0, td);
609 ronly = (mp->mnt_flag & MNT_RDONLY) != 0;
610 vn_lock(devvp, LK_EXCLUSIVE | LK_RETRY, td);
611 error = VOP_OPEN(devvp, ronly ? FREAD : FREAD|FWRITE, FSCRED, NULL, td);
612 VOP_UNLOCK(devvp, 0, td);
616 if (dev->si_iosize_max != 0)
617 mp->mnt_iosize_max = dev->si_iosize_max;
618 if (mp->mnt_iosize_max > MAXPHYS)
619 mp->mnt_iosize_max = MAXPHYS;
622 * Only VMIO the backing device if the backing device is a real
623 * block device. This excludes the original MFS implementation.
624 * Note that it is optional that the backing device be VMIOed. This
625 * increases the opportunity for metadata caching.
627 * This call must be made after the VOP_OPEN.
629 if (devvp->v_tag != VT_MFS && vn_isdisk(devvp, NULL)) {
630 vn_lock(devvp, LK_EXCLUSIVE | LK_RETRY, td);
631 vfs_object_create(devvp, td);
632 VOP_UNLOCK(devvp, 0, td);
635 if (VOP_IOCTL(devvp, DIOCGPART, (caddr_t)&dpart, FREAD, proc0.p_ucred, td) != 0)
638 size = dpart.disklab->d_secsize;
642 if ((error = bread(devvp, SBLOCK, SBSIZE, &bp)) != 0)
644 fs = (struct fs *)bp->b_data;
645 if (fs->fs_magic != FS_MAGIC || fs->fs_bsize > MAXBSIZE ||
646 fs->fs_bsize < sizeof(struct fs)) {
647 error = EINVAL; /* XXX needs translation */
651 fs->fs_flags &= ~FS_UNCLEAN;
652 if (fs->fs_clean == 0) {
653 fs->fs_flags |= FS_UNCLEAN;
654 if (ronly || (mp->mnt_flag & MNT_FORCE)) {
656 "WARNING: %s was not properly dismounted\n",
660 "WARNING: R/W mount of %s denied. Filesystem is not clean - run fsck\n",
666 /* XXX updating 4.2 FFS superblocks trashes rotational layout tables */
667 if (fs->fs_postblformat == FS_42POSTBLFMT && !ronly) {
668 error = EROFS; /* needs translation */
671 ump = malloc(sizeof *ump, M_UFSMNT, M_WAITOK);
672 bzero((caddr_t)ump, sizeof *ump);
673 ump->um_malloctype = malloctype;
674 ump->um_i_effnlink_valid = 1;
675 ump->um_fs = malloc((u_long)fs->fs_sbsize, M_UFSMNT,
677 ump->um_blkatoff = ffs_blkatoff;
678 ump->um_truncate = ffs_truncate;
679 ump->um_update = ffs_update;
680 ump->um_valloc = ffs_valloc;
681 ump->um_vfree = ffs_vfree;
682 bcopy(bp->b_data, ump->um_fs, (uint)fs->fs_sbsize);
683 if (fs->fs_sbsize < SBSIZE)
684 bp->b_flags |= B_INVAL;
688 fs->fs_ronly = ronly;
689 size = fs->fs_cssize;
690 blks = howmany(size, fs->fs_fsize);
691 if (fs->fs_contigsumsize > 0)
692 size += fs->fs_ncg * sizeof(int32_t);
693 size += fs->fs_ncg * sizeof(uint8_t);
694 space = malloc((u_long)size, M_UFSMNT, M_WAITOK);
696 for (i = 0; i < blks; i += fs->fs_frag) {
698 if (i + fs->fs_frag > blks)
699 size = (blks - i) * fs->fs_fsize;
700 if ((error = bread(devvp, fsbtodb(fs, fs->fs_csaddr + i), size,
702 free(fs->fs_csp, M_UFSMNT);
705 bcopy(bp->b_data, space, (uint)size);
706 space = (char *)space + size;
710 if (fs->fs_contigsumsize > 0) {
711 fs->fs_maxcluster = lp = space;
712 for (i = 0; i < fs->fs_ncg; i++)
713 *lp++ = fs->fs_contigsumsize;
716 size = fs->fs_ncg * sizeof(uint8_t);
717 fs->fs_contigdirs = (uint8_t *)space;
718 bzero(fs->fs_contigdirs, size);
719 /* Compatibility for old filesystems XXX */
720 if (fs->fs_avgfilesize <= 0) /* XXX */
721 fs->fs_avgfilesize = AVFILESIZ; /* XXX */
722 if (fs->fs_avgfpdir <= 0) /* XXX */
723 fs->fs_avgfpdir = AFPDIR; /* XXX */
724 mp->mnt_data = (qaddr_t)ump;
725 mp->mnt_stat.f_fsid.val[0] = fs->fs_id[0];
726 mp->mnt_stat.f_fsid.val[1] = fs->fs_id[1];
727 if (fs->fs_id[0] == 0 || fs->fs_id[1] == 0 ||
728 vfs_getvfs(&mp->mnt_stat.f_fsid))
730 mp->mnt_maxsymlinklen = fs->fs_maxsymlinklen;
731 mp->mnt_flag |= MNT_LOCAL;
734 ump->um_devvp = devvp;
735 ump->um_nindir = fs->fs_nindir;
736 ump->um_bptrtodb = fs->fs_fsbtodb;
737 ump->um_seqinc = fs->fs_frag;
738 for (i = 0; i < MAXQUOTAS; i++)
739 ump->um_quotas[i] = NULLVP;
740 dev->si_mountpoint = mp;
743 if( mp->mnt_flag & MNT_ROOTFS) {
745 * Root mount; update timestamp in mount structure.
746 * this will be used by the common root mount code
747 * to update the system clock.
749 mp->mnt_time = fs->fs_time;
752 ump->um_savedmaxfilesize = fs->fs_maxfilesize; /* XXX */
753 maxfilesize = (uint64_t)0x40000000 * fs->fs_bsize - 1; /* XXX */
754 /* Enforce limit caused by vm object backing (32 bits vm_pindex_t). */
755 if (maxfilesize > (uint64_t)0x80000000u * PAGE_SIZE - 1)
756 maxfilesize = (uint64_t)0x80000000u * PAGE_SIZE - 1;
757 if (fs->fs_maxfilesize > maxfilesize) /* XXX */
758 fs->fs_maxfilesize = maxfilesize; /* XXX */
760 if ((fs->fs_flags & FS_DOSOFTDEP) &&
761 (error = softdep_mount(devvp, mp, fs)) != 0) {
762 free(fs->fs_csp, M_UFSMNT);
767 (void) ffs_sbupdate(ump, MNT_WAIT);
769 vfs_add_vnodeops(mp, &mp->mnt_vn_norm_ops, ffs_vnodeop_entries);
770 vfs_add_vnodeops(mp, &mp->mnt_vn_spec_ops, ffs_specop_entries);
771 vfs_add_vnodeops(mp, &mp->mnt_vn_fifo_ops, ffs_fifoop_entries);
775 dev->si_mountpoint = NULL;
778 VOP_CLOSE(devvp, ronly ? FREAD : FREAD|FWRITE, td);
780 free(ump->um_fs, M_UFSMNT);
782 mp->mnt_data = (qaddr_t)0;
788 * Sanity checks for old filesystems.
790 * XXX - goes away some day.
793 ffs_oldfscompat(struct fs *fs)
795 fs->fs_npsect = max(fs->fs_npsect, fs->fs_nsect); /* XXX */
796 fs->fs_interleave = max(fs->fs_interleave, 1); /* XXX */
797 if (fs->fs_postblformat == FS_42POSTBLFMT) /* XXX */
798 fs->fs_nrpos = 8; /* XXX */
799 if (fs->fs_inodefmt < FS_44INODEFMT) { /* XXX */
802 uint64_t sizepb = fs->fs_bsize; /* XXX */
804 fs->fs_maxfilesize = fs->fs_bsize * NDADDR - 1; /* XXX */
805 for (i = 0; i < NIADDR; i++) { /* XXX */
806 sizepb *= NINDIR(fs); /* XXX */
807 fs->fs_maxfilesize += sizepb; /* XXX */
810 fs->fs_maxfilesize = (u_quad_t) 1LL << 39;
811 fs->fs_qbmask = ~fs->fs_bmask; /* XXX */
812 fs->fs_qfmask = ~fs->fs_fmask; /* XXX */
818 * unmount system call
821 ffs_unmount(struct mount *mp, int mntflags, struct thread *td)
823 struct ufsmount *ump;
828 if (mntflags & MNT_FORCE) {
831 if (mp->mnt_flag & MNT_SOFTDEP) {
832 if ((error = softdep_flushfiles(mp, flags, td)) != 0)
835 if ((error = ffs_flushfiles(mp, flags, td)) != 0)
840 if (fs->fs_ronly == 0) {
841 fs->fs_clean = fs->fs_flags & FS_UNCLEAN ? 0 : 1;
842 error = ffs_sbupdate(ump, MNT_WAIT);
848 ump->um_devvp->v_rdev->si_mountpoint = NULL;
850 vinvalbuf(ump->um_devvp, V_SAVE, td, 0, 0);
851 error = VOP_CLOSE(ump->um_devvp, fs->fs_ronly ? FREAD : FREAD|FWRITE, td);
853 vrele(ump->um_devvp);
855 free(fs->fs_csp, M_UFSMNT);
858 mp->mnt_data = (qaddr_t)0;
859 mp->mnt_flag &= ~MNT_LOCAL;
864 * Flush out all the files in a filesystem.
867 ffs_flushfiles(struct mount *mp, int flags, struct thread *td)
869 struct ufsmount *ump;
874 if (mp->mnt_flag & MNT_QUOTA) {
876 error = vflush(mp, 0, SKIPSYSTEM|flags);
879 /* Find out how many quota files we have open. */
880 for (i = 0; i < MAXQUOTAS; i++) {
881 if (ump->um_quotas[i] == NULLVP)
886 * Here we fall through to vflush again to ensure
887 * that we have gotten rid of all the system vnodes.
892 * Flush all the files.
894 if ((error = vflush(mp, 0, flags)) != 0)
897 * Flush filesystem metadata.
899 vn_lock(ump->um_devvp, LK_EXCLUSIVE | LK_RETRY, td);
900 error = VOP_FSYNC(ump->um_devvp, MNT_WAIT, td);
901 VOP_UNLOCK(ump->um_devvp, 0, td);
906 * Get filesystem statistics.
909 ffs_statfs(struct mount *mp, struct statfs *sbp, struct thread *td)
911 struct ufsmount *ump;
916 if (fs->fs_magic != FS_MAGIC)
918 sbp->f_bsize = fs->fs_fsize;
919 sbp->f_iosize = fs->fs_bsize;
920 sbp->f_blocks = fs->fs_dsize;
921 sbp->f_bfree = fs->fs_cstotal.cs_nbfree * fs->fs_frag +
922 fs->fs_cstotal.cs_nffree;
923 sbp->f_bavail = freespace(fs, fs->fs_minfree);
924 sbp->f_files = fs->fs_ncg * fs->fs_ipg - ROOTINO;
925 sbp->f_ffree = fs->fs_cstotal.cs_nifree;
926 if (sbp != &mp->mnt_stat) {
927 sbp->f_type = mp->mnt_vfc->vfc_typenum;
928 bcopy((caddr_t)mp->mnt_stat.f_mntfromname,
929 (caddr_t)&sbp->f_mntfromname[0], MNAMELEN);
935 * Go through the disk queues to initiate sandbagged IO;
936 * go through the inodes to write those that have been modified;
937 * initiate the writing of the super block if it has been modified.
939 * Note: we are always called with the filesystem marked `MPBUSY'.
943 static int ffs_sync_scan1(struct mount *mp, struct vnode *vp, void *data);
944 static int ffs_sync_scan2(struct mount *mp, struct vnode *vp, void *data);
947 ffs_sync(struct mount *mp, int waitfor, struct thread *td)
949 struct ufsmount *ump = VFSTOUFS(mp);
952 struct scaninfo scaninfo;
955 if (fs->fs_fmod != 0 && fs->fs_ronly != 0) { /* XXX */
956 printf("fs = %s\n", fs->fs_fsmnt);
957 panic("ffs_sync: rofs mod");
961 * Write back each (modified) inode.
963 scaninfo.allerror = 0;
965 scaninfo.waitfor = waitfor;
966 while (scaninfo.rescan) {
968 vmntvnodescan(mp, VMSC_GETVP|VMSC_NOWAIT,
969 ffs_sync_scan1, ffs_sync_scan2, &scaninfo);
973 * Force stale filesystem control information to be flushed.
975 if (waitfor != MNT_LAZY) {
976 if (ump->um_mountp->mnt_flag & MNT_SOFTDEP)
977 waitfor = MNT_NOWAIT;
978 vn_lock(ump->um_devvp, LK_EXCLUSIVE | LK_RETRY, td);
979 if ((error = VOP_FSYNC(ump->um_devvp, waitfor, td)) != 0)
980 scaninfo.allerror = error;
981 VOP_UNLOCK(ump->um_devvp, 0, td);
987 * Write back modified superblock.
989 if (fs->fs_fmod != 0 && (error = ffs_sbupdate(ump, waitfor)) != 0)
990 scaninfo.allerror = error;
991 return (scaninfo.allerror);
995 ffs_sync_scan1(struct mount *mp, struct vnode *vp, void *data)
1000 * Depend on the mount list's vnode lock to keep things stable
1001 * enough for a quick test. Since there might be hundreds of
1002 * thousands of vnodes, we cannot afford even a subroutine
1003 * call unless there's a good chance that we have work to do.
1006 /* Restart out whole search if this guy is locked
1007 * or is being reclaimed.
1009 if (vp->v_type == VNON || ((ip->i_flag &
1010 (IN_ACCESS | IN_CHANGE | IN_MODIFIED | IN_UPDATE)) == 0 &&
1011 RB_EMPTY(&vp->v_rbdirty_tree))) {
1018 ffs_sync_scan2(struct mount *mp, struct vnode *vp, void *data)
1020 struct scaninfo *info = data;
1021 thread_t td = curthread; /* XXX */
1026 * We have to recheck after having obtained the vnode interlock.
1029 if (vp->v_type == VNON || ((ip->i_flag &
1030 (IN_ACCESS | IN_CHANGE | IN_MODIFIED | IN_UPDATE)) == 0 &&
1031 RB_EMPTY(&vp->v_rbdirty_tree))) {
1034 if (vp->v_type != VCHR) {
1035 if ((error = VOP_FSYNC(vp, info->waitfor, td)) != 0)
1036 info->allerror = error;
1039 * We must reference the vp to prevent it from
1040 * getting ripped out from under UFS_UPDATE, since
1041 * we are not holding a vnode lock.
1043 /* UFS_UPDATE(vp, waitfor == MNT_WAIT); */
1050 * Look up a FFS dinode number to find its incore vnode, otherwise read it
1051 * in from disk. If it is in core, wait for the lock bit to clear, then
1052 * return the inode locked. Detection and handling of mount points must be
1053 * done by the calling routine.
1057 ffs_vget(struct mount *mp, ino_t ino, struct vnode **vpp)
1061 struct ufsmount *ump;
1070 if ((*vpp = ufs_ihashget(dev, ino)) != NULL) {
1075 * If this MALLOC() is performed after the getnewvnode()
1076 * it might block, leaving a vnode with a NULL v_data to be
1077 * found by ffs_sync() if a sync happens to fire right then,
1078 * which will cause a panic because ffs_sync() blindly
1079 * dereferences vp->v_data (as well it should).
1081 * XXX this may no longer be true since getnewvnode returns a
1082 * VX locked vnode now.
1084 MALLOC(ip, struct inode *, sizeof(struct inode),
1085 ump->um_malloctype, M_WAITOK);
1087 /* Allocate a new vnode/inode. */
1088 error = getnewvnode(VT_UFS, mp, &vp, VLKTIMEOUT, LK_CANRECURSE);
1091 free(ip, ump->um_malloctype);
1094 bzero((caddr_t)ip, sizeof(struct inode));
1096 ip->i_fs = fs = ump->um_fs;
1102 for (i = 0; i < MAXQUOTAS; i++)
1103 ip->i_dquot[i] = NODQUOT;
1108 * Insert it into the inode hash table and check for a collision.
1109 * If a collision occurs, throw away the vnode and try again.
1111 if (ufs_ihashins(ip) != 0) {
1112 printf("debug: ufs ihashins collision, retrying inode %ld\n",
1113 (long)ip->i_number);
1115 free(ip, ump->um_malloctype);
1120 /* Read in the disk contents for the inode, copy into the inode. */
1121 error = bread(ump->um_devvp, fsbtodb(fs, ino_to_fsba(fs, ino)),
1122 (int)fs->fs_bsize, &bp);
1125 * The inode does not contain anything useful, so it would
1126 * be misleading to leave it on its hash chain. With mode
1127 * still zero, it will be unlinked and returned to the free
1135 ip->i_din = *((struct dinode *)bp->b_data + ino_to_fsbo(fs, ino));
1136 if (DOINGSOFTDEP(vp))
1137 softdep_load_inodeblock(ip);
1139 ip->i_effnlink = ip->i_nlink;
1143 * Initialize the vnode from the inode, check for aliases.
1144 * Note that the underlying vnode may have changed.
1146 error = ufs_vinit(mp, &vp);
1153 * Finish inode initialization now that aliasing has been resolved.
1155 ip->i_devvp = ump->um_devvp;
1158 * Set up a generation number for this inode if it does not
1159 * already have one. This should only happen on old filesystems.
1161 if (ip->i_gen == 0) {
1162 ip->i_gen = random() / 2 + 1;
1163 if ((vp->v_mount->mnt_flag & MNT_RDONLY) == 0)
1164 ip->i_flag |= IN_MODIFIED;
1167 * Ensure that uid and gid are correct. This is a temporary
1168 * fix until fsck has been changed to do the update.
1170 if (fs->fs_inodefmt < FS_44INODEFMT) { /* XXX */
1171 ip->i_uid = ip->i_din.di_ouid; /* XXX */
1172 ip->i_gid = ip->i_din.di_ogid; /* XXX */
1176 * return a VX locked and refd vnode (VX == same as normal vget()
1177 * vnode so we are ok)
1184 * File handle to vnode
1186 * Have to be really careful about stale file handles:
1187 * - check that the inode number is valid
1188 * - call ffs_vget() to get the locked inode
1189 * - check for an unallocated inode (i_mode == 0)
1190 * - check that the given client host has export rights and return
1191 * those rights via. exflagsp and credanonp
1194 ffs_fhtovp(struct mount *mp, struct fid *fhp, struct vnode **vpp)
1199 ufhp = (struct ufid *)fhp;
1200 fs = VFSTOUFS(mp)->um_fs;
1201 if (ufhp->ufid_ino < ROOTINO ||
1202 ufhp->ufid_ino >= fs->fs_ncg * fs->fs_ipg)
1204 return (ufs_fhtovp(mp, ufhp, vpp));
1208 * Vnode pointer to File handle
1212 ffs_vptofh(struct vnode *vp, struct fid *fhp)
1218 ufhp = (struct ufid *)fhp;
1219 ufhp->ufid_len = sizeof(struct ufid);
1220 ufhp->ufid_ino = ip->i_number;
1221 ufhp->ufid_gen = ip->i_gen;
1226 * Initialize the filesystem; just use ufs_init.
1229 ffs_init(struct vfsconf *vfsp)
1231 softdep_initialize();
1232 return (ufs_init(vfsp));
1236 * Write a superblock and associated information back to disk.
1239 ffs_sbupdate(struct ufsmount *mp, int waitfor)
1241 struct fs *dfs, *fs = mp->um_fs;
1245 int i, size, error, allerror = 0;
1248 * First write back the summary information.
1250 blks = howmany(fs->fs_cssize, fs->fs_fsize);
1252 for (i = 0; i < blks; i += fs->fs_frag) {
1253 size = fs->fs_bsize;
1254 if (i + fs->fs_frag > blks)
1255 size = (blks - i) * fs->fs_fsize;
1256 bp = getblk(mp->um_devvp, fsbtodb(fs, fs->fs_csaddr + i),
1258 bcopy(space, bp->b_data, (uint)size);
1259 space = (char *)space + size;
1260 if (waitfor != MNT_WAIT)
1262 else if ((error = bwrite(bp)) != 0)
1266 * Now write back the superblock itself. If any errors occurred
1267 * up to this point, then fail so that the superblock avoids
1268 * being written out as clean.
1272 bp = getblk(mp->um_devvp, SBLOCK, (int)fs->fs_sbsize, 0, 0);
1274 fs->fs_time = time_second;
1275 bcopy((caddr_t)fs, bp->b_data, (uint)fs->fs_sbsize);
1276 /* Restore compatibility to old filesystems. XXX */
1277 dfs = (struct fs *)bp->b_data; /* XXX */
1278 if (fs->fs_postblformat == FS_42POSTBLFMT) /* XXX */
1279 dfs->fs_nrpos = -1; /* XXX */
1280 if (fs->fs_inodefmt < FS_44INODEFMT) { /* XXX */
1281 int32_t *lp, tmp; /* XXX */
1283 lp = (int32_t *)&dfs->fs_qbmask; /* XXX */
1284 tmp = lp[4]; /* XXX */
1285 for (i = 4; i > 0; i--) /* XXX */
1286 lp[i] = lp[i-1]; /* XXX */
1287 lp[0] = tmp; /* XXX */
1289 dfs->fs_maxfilesize = mp->um_savedmaxfilesize; /* XXX */
1290 if (waitfor != MNT_WAIT)
1292 else if ((error = bwrite(bp)) != 0)