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 1.44 2006/05/26 17:07:48 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 *);
69 static int ffs_oldfscompat (struct fs *);
70 static int ffs_mount (struct mount *, char *, caddr_t, struct ucred *);
71 static int ffs_init (struct vfsconf *);
73 static struct vfsops ufs_vfsops = {
74 .vfs_mount = ffs_mount,
75 .vfs_unmount = ffs_unmount,
77 .vfs_quotactl = ufs_quotactl,
78 .vfs_statfs = ffs_statfs,
81 .vfs_fhtovp = ffs_fhtovp,
82 .vfs_checkexp = ufs_check_export,
83 .vfs_vptofh = ffs_vptofh,
85 .vfs_uninit = ufs_uninit
88 VFS_SET(ufs_vfsops, ufs, 0);
90 extern struct vnodeopv_entry_desc ffs_vnodeop_entries[];
91 extern struct vnodeopv_entry_desc ffs_specop_entries[];
92 extern struct vnodeopv_entry_desc ffs_fifoop_entries[];
98 * Called when mounting local physical media
102 * mp mount point structure
103 * path NULL (flag for root mount!!!)
105 * p process (user credentials check [statfs])
108 * mp mount point structure
109 * path path to mount point
110 * data pointer to argument struct in user space
111 * p process (user credentials check)
114 * !0 error number (errno.h)
119 * mount point is locked
121 * mount point is locked
124 * A NULL path can be used for a flag since the mount
125 * system call will fail with EFAULT in copyinstr in
126 * nlookup() if it is a genuine NULL from the user.
129 ffs_mount(struct mount *mp, /* mount struct pointer */
130 char *path, /* path to mount point */
131 caddr_t data, /* arguments to FS specific mount */
132 struct ucred *cred) /* process requesting mount */
138 struct ufs_args args;
139 struct ufsmount *ump = 0;
141 int flags, ronly = 0;
143 struct nlookupdata nd;
144 struct vnode *rootvp;
150 * Use NULL path to flag a root mount
155 * Mounting root filesystem
159 if ((error = bdevvp(rootdev, &rootvp))) {
160 printf("ffs_mountroot: can't find rootvp\n");
164 if( ( error = ffs_mountfs(rootvp, mp, M_FFSNODE)) != 0) {
165 /* fs specific cleanup (if any)*/
170 goto dostatfs; /* success*/
176 * Mounting non-root filesystem or updating a filesystem
180 /* copy in user arguments*/
181 error = copyin(data, (caddr_t)&args, sizeof (struct ufs_args));
183 goto error_1; /* can't get arguments*/
186 * If updating, check whether changing from read-only to
187 * read/write; if there is no device name, that's all we do.
189 if (mp->mnt_flag & MNT_UPDATE) {
192 devvp = ump->um_devvp;
194 ronly = fs->fs_ronly; /* MNT_RELOAD might change this */
195 if (ronly == 0 && (mp->mnt_flag & MNT_RDONLY)) {
197 * Flush any dirty data.
199 VFS_SYNC(mp, MNT_WAIT);
201 * Check for and optionally get rid of files open
205 if (mp->mnt_flag & MNT_FORCE)
207 if (mp->mnt_flag & MNT_SOFTDEP) {
208 error = softdep_flushfiles(mp, flags);
210 error = ffs_flushfiles(mp, flags);
214 if (!error && (mp->mnt_flag & MNT_RELOAD)) {
215 error = ffs_reload(mp, NULL);
220 if (ronly && (mp->mnt_kern_flag & MNTK_WANTRDWR)) {
222 * If upgrade to read-write by non-root, then verify
223 * that user has necessary permissions on the device.
225 if (cred->cr_uid != 0) {
226 vn_lock(devvp, LK_EXCLUSIVE | LK_RETRY);
227 if ((error = VOP_ACCESS(devvp, VREAD | VWRITE,
229 VOP_UNLOCK(devvp, 0);
232 VOP_UNLOCK(devvp, 0);
235 fs->fs_flags &= ~FS_UNCLEAN;
236 if (fs->fs_clean == 0) {
237 fs->fs_flags |= FS_UNCLEAN;
238 if (mp->mnt_flag & MNT_FORCE) {
240 "WARNING: %s was not properly dismounted\n",
244 "WARNING: R/W mount of %s denied. Filesystem is not clean - run fsck\n",
251 /* check to see if we need to start softdep */
252 if (fs->fs_flags & FS_DOSOFTDEP) {
253 error = softdep_mount(devvp, mp, fs);
260 * Soft updates is incompatible with "async",
261 * so if we are doing softupdates stop the user
262 * from setting the async flag in an update.
263 * Softdep_mount() clears it in an initial mount
266 if (mp->mnt_flag & MNT_SOFTDEP) {
267 mp->mnt_flag &= ~MNT_ASYNC;
269 /* if not updating name...*/
270 if (args.fspec == 0) {
272 * Process export requests. Jumping to "success"
273 * will return the vfs_export() error code.
275 error = vfs_export(mp, &ump->um_export, &args.export);
281 * Not an update, or updating the name: look up the name
282 * and verify that it refers to a sensible block device.
285 error = nlookup_init(&nd, args.fspec, UIO_USERSPACE, NLC_FOLLOW);
287 error = nlookup(&nd);
289 error = cache_vref(nd.nl_ncp, nd.nl_cred, &devvp);
294 if (!vn_isdisk(devvp, &error))
298 * If mount by non-root, then verify that user has necessary
299 * permissions on the device.
301 if (cred->cr_uid != 0) {
303 if ((mp->mnt_flag & MNT_RDONLY) == 0)
304 accessmode |= VWRITE;
305 vn_lock(devvp, LK_EXCLUSIVE | LK_RETRY);
306 if ((error = VOP_ACCESS(devvp, accessmode, cred)) != 0) {
310 VOP_UNLOCK(devvp, 0);
313 if (mp->mnt_flag & MNT_UPDATE) {
315 * UPDATE - make sure the resolved vnode represents the same
316 * device. Note that devvp->v_rdev may be NULL since we
317 * haven't opened it, so compare udev instead.
319 * Our current open/writecount state is associated with
320 * um_devvp, so continue using um_devvp and throw away devvp.
322 if (devvp != ump->um_devvp) {
323 if (devvp->v_udev == ump->um_devvp->v_udev) {
325 devvp = ump->um_devvp;
327 printf("cannot update mount, udev does"
328 " not match %08x vs %08x\n",
329 devvp->v_udev, ump->um_devvp->v_udev);
330 error = EINVAL; /* needs translation */
336 * Update device name only on success
339 /* Save "mounted from" info for mount point (NULL pad)*/
340 copyinstr( args.fspec,
341 mp->mnt_stat.f_mntfromname,
344 bzero( mp->mnt_stat.f_mntfromname + size, MNAMELEN - size);
353 /* Save "mounted from" info for mount point (NULL pad)*/
354 copyinstr( args.fspec, /* device name*/
355 mp->mnt_stat.f_mntfromname, /* save area*/
356 MNAMELEN - 1, /* max size*/
357 &size); /* real size*/
358 bzero( mp->mnt_stat.f_mntfromname + size, MNAMELEN - size);
360 error = ffs_mountfs(devvp, mp, M_FFSNODE);
368 * Initialize FS stat information in mount struct; uses
369 * mp->mnt_stat.f_mntfromname.
371 * This code is common to root and non-root mounts
373 (void)VFS_STATFS(mp, &mp->mnt_stat, cred);
378 error_2: /* error with devvp held*/
380 /* release devvp before failing*/
383 error_1: /* no state to back out*/
386 if (!error && path && (mp->mnt_flag & MNT_UPDATE)) {
387 /* Update clean flag after changing read-onlyness. */
389 if (ronly != fs->fs_ronly) {
390 fs->fs_ronly = ronly;
391 fs->fs_clean = ronly &&
392 (fs->fs_flags & FS_UNCLEAN) == 0 ? 1 : 0;
395 * The device must be re-opened as appropriate or
396 * the device close at unmount time will panic.
398 vn_lock(devvp, LK_EXCLUSIVE | LK_RETRY);
400 VOP_OPEN(devvp, FREAD, FSCRED, NULL);
401 VOP_CLOSE(devvp, FREAD|FWRITE);
403 VOP_OPEN(devvp, FREAD|FWRITE, FSCRED, NULL);
404 VOP_CLOSE(devvp, FREAD);
406 VOP_UNLOCK(devvp, 0);
407 ffs_sbupdate(ump, MNT_WAIT);
414 * Reload all incore data for a filesystem (used after running fsck on
415 * the root filesystem and finding things to fix). The filesystem must
416 * be mounted read-only.
418 * Things to do to update the mount:
419 * 1) invalidate all cached meta-data.
420 * 2) re-read superblock from disk.
421 * 3) re-read summary information from disk.
422 * 4) invalidate all inactive vnodes.
423 * 5) invalidate all cached file data.
424 * 6) re-read inode data for all active vnodes.
427 static int ffs_reload_scan2(struct mount *mp, struct vnode *vp, void *data);
438 ffs_reload(struct mount *mp, struct ucred *cred)
443 struct fs *fs, *newfs;
444 struct partinfo dpart;
446 int i, blks, size, error;
447 struct scaninfo scaninfo;
450 if ((mp->mnt_flag & MNT_RDONLY) == 0)
453 * Step 1: invalidate all cached meta-data.
455 devvp = VFSTOUFS(mp)->um_devvp;
456 vn_lock(devvp, LK_EXCLUSIVE | LK_RETRY);
457 error = vinvalbuf(devvp, 0, 0, 0);
458 VOP_UNLOCK(devvp, 0);
460 panic("ffs_reload: dirty1");
465 * The backing device must be VMIO-capable because we use getblk().
466 * NOTE: the MFS driver now returns a VMIO-enabled descriptor.
468 if (devvp->v_object == NULL)
469 panic("ffs_reload: devvp has no VM object!");
472 * Step 2: re-read superblock from disk.
474 if (VOP_IOCTL(devvp, DIOCGPART, (caddr_t)&dpart, FREAD, NOCRED) != 0)
477 size = dpart.disklab->d_secsize;
478 if ((error = bread(devvp, SBOFF, SBSIZE, &bp)) != 0) {
482 newfs = (struct fs *)bp->b_data;
483 if (newfs->fs_magic != FS_MAGIC || newfs->fs_bsize > MAXBSIZE ||
484 newfs->fs_bsize < sizeof(struct fs)) {
486 return (EIO); /* XXX needs translation */
488 fs = VFSTOUFS(mp)->um_fs;
490 * Copy pointer fields back into superblock before copying in XXX
491 * new superblock. These should really be in the ufsmount. XXX
492 * Note that important parameters (eg fs_ncg) are unchanged.
494 newfs->fs_csp = fs->fs_csp;
495 newfs->fs_maxcluster = fs->fs_maxcluster;
496 newfs->fs_contigdirs = fs->fs_contigdirs;
497 /* The filesystem is still read-only. */
499 bcopy(newfs, fs, (uint)fs->fs_sbsize);
500 if (fs->fs_sbsize < SBSIZE)
501 bp->b_flags |= B_INVAL;
503 mp->mnt_maxsymlinklen = fs->fs_maxsymlinklen;
505 /* An old fsck may have zeroed these fields, so recheck them. */
506 if (fs->fs_avgfilesize <= 0) /* XXX */
507 fs->fs_avgfilesize = AVFILESIZ; /* XXX */
508 if (fs->fs_avgfpdir <= 0) /* XXX */
509 fs->fs_avgfpdir = AFPDIR; /* XXX */
512 * Step 3: re-read summary information from disk.
514 blks = howmany(fs->fs_cssize, fs->fs_fsize);
516 for (i = 0; i < blks; i += fs->fs_frag) {
518 if (i + fs->fs_frag > blks)
519 size = (blks - i) * fs->fs_fsize;
520 error = bread(devvp, fsbtodoff(fs, fs->fs_csaddr + i), size, &bp);
525 bcopy(bp->b_data, space, (uint)size);
526 space = (char *)space + size;
530 * We no longer know anything about clusters per cylinder group.
532 if (fs->fs_contigsumsize > 0) {
533 lp = fs->fs_maxcluster;
534 for (i = 0; i < fs->fs_ncg; i++)
535 *lp++ = fs->fs_contigsumsize;
540 scaninfo.devvp = devvp;
541 while (error == 0 && scaninfo.rescan) {
543 error = vmntvnodescan(mp, VMSC_GETVX,
544 NULL, ffs_reload_scan2, &scaninfo);
550 ffs_reload_scan2(struct mount *mp, struct vnode *vp, void *data)
552 struct scaninfo *info = data;
563 if (vinvalbuf(vp, 0, 0, 0))
564 panic("ffs_reload: dirty2");
566 * Step 6: re-read inode data for all active vnodes.
569 error = bread(info->devvp,
570 fsbtodoff(info->fs, ino_to_fsba(info->fs, ip->i_number)),
571 (int)info->fs->fs_bsize, &bp);
576 ip->i_din = *((struct ufs1_dinode *)bp->b_data +
577 ino_to_fsbo(info->fs, ip->i_number));
578 ip->i_effnlink = ip->i_nlink;
584 * Common code for mount and mountroot
587 ffs_mountfs(struct vnode *devvp, struct mount *mp, struct malloc_type *mtype)
589 struct ufsmount *ump;
593 struct partinfo dpart;
595 int error, i, blks, size, ronly;
597 uint64_t maxfilesize; /* XXX */
600 * Disallow multiple mounts of the same device.
601 * Disallow mounting of a device that is currently in use
602 * Flush out any old buffers remaining from a previous use.
604 error = vfs_mountedon(devvp);
607 if (count_udev(devvp->v_udev) > 0)
609 vn_lock(devvp, LK_EXCLUSIVE | LK_RETRY);
610 error = vinvalbuf(devvp, V_SAVE, 0, 0);
611 VOP_UNLOCK(devvp, 0);
615 ronly = (mp->mnt_flag & MNT_RDONLY) != 0;
616 vn_lock(devvp, LK_EXCLUSIVE | LK_RETRY);
617 error = VOP_OPEN(devvp, ronly ? FREAD : FREAD|FWRITE, FSCRED, NULL);
618 VOP_UNLOCK(devvp, 0);
622 if (dev->si_iosize_max != 0)
623 mp->mnt_iosize_max = dev->si_iosize_max;
624 if (mp->mnt_iosize_max > MAXPHYS)
625 mp->mnt_iosize_max = MAXPHYS;
628 * The backing device must be VMIO-capable because we use getblk().
629 * NOTE: the MFS driver now returns a VMIO-enabled descriptor.
630 * The VOP_OPEN() call above should have associated a VM object
633 if (devvp->v_object == NULL)
634 panic("ffs_reload: devvp has no VM object!");
636 if (VOP_IOCTL(devvp, DIOCGPART, (caddr_t)&dpart, FREAD, proc0.p_ucred) != 0)
639 size = dpart.disklab->d_secsize;
643 if ((error = bread(devvp, SBOFF, SBSIZE, &bp)) != 0)
645 fs = (struct fs *)bp->b_data;
646 if (fs->fs_magic != FS_MAGIC || fs->fs_bsize > MAXBSIZE ||
647 fs->fs_bsize < sizeof(struct fs)) {
648 error = EINVAL; /* XXX needs translation */
652 fs->fs_flags &= ~FS_UNCLEAN;
653 if (fs->fs_clean == 0) {
654 fs->fs_flags |= FS_UNCLEAN;
655 if (ronly || (mp->mnt_flag & MNT_FORCE)) {
657 "WARNING: %s was not properly dismounted\n",
661 "WARNING: R/W mount of %s denied. Filesystem is not clean - run fsck\n",
667 /* XXX updating 4.2 FFS superblocks trashes rotational layout tables */
668 if (fs->fs_postblformat == FS_42POSTBLFMT && !ronly) {
669 error = EROFS; /* needs translation */
672 ump = malloc(sizeof *ump, M_UFSMNT, M_WAITOK);
673 bzero((caddr_t)ump, sizeof *ump);
674 ump->um_malloctype = mtype;
675 ump->um_i_effnlink_valid = 1;
676 ump->um_fs = malloc((u_long)fs->fs_sbsize, M_UFSMNT,
678 bcopy(bp->b_data, ump->um_fs, (uint)fs->fs_sbsize);
679 if (fs->fs_sbsize < SBSIZE)
680 bp->b_flags |= B_INVAL;
684 fs->fs_ronly = ronly;
685 size = fs->fs_cssize;
686 blks = howmany(size, fs->fs_fsize);
687 if (fs->fs_contigsumsize > 0)
688 size += fs->fs_ncg * sizeof(int32_t);
689 size += fs->fs_ncg * sizeof(uint8_t);
690 space = malloc((u_long)size, M_UFSMNT, M_WAITOK);
692 for (i = 0; i < blks; i += fs->fs_frag) {
694 if (i + fs->fs_frag > blks)
695 size = (blks - i) * fs->fs_fsize;
696 if ((error = bread(devvp, fsbtodoff(fs, fs->fs_csaddr + i),
698 free(fs->fs_csp, M_UFSMNT);
701 bcopy(bp->b_data, space, (uint)size);
702 space = (char *)space + size;
706 if (fs->fs_contigsumsize > 0) {
707 fs->fs_maxcluster = lp = space;
708 for (i = 0; i < fs->fs_ncg; i++)
709 *lp++ = fs->fs_contigsumsize;
712 size = fs->fs_ncg * sizeof(uint8_t);
713 fs->fs_contigdirs = (uint8_t *)space;
714 bzero(fs->fs_contigdirs, size);
715 /* Compatibility for old filesystems XXX */
716 if (fs->fs_avgfilesize <= 0) /* XXX */
717 fs->fs_avgfilesize = AVFILESIZ; /* XXX */
718 if (fs->fs_avgfpdir <= 0) /* XXX */
719 fs->fs_avgfpdir = AFPDIR; /* XXX */
720 mp->mnt_data = (qaddr_t)ump;
721 mp->mnt_stat.f_fsid.val[0] = fs->fs_id[0];
722 mp->mnt_stat.f_fsid.val[1] = fs->fs_id[1];
723 if (fs->fs_id[0] == 0 || fs->fs_id[1] == 0 ||
724 vfs_getvfs(&mp->mnt_stat.f_fsid))
726 mp->mnt_maxsymlinklen = fs->fs_maxsymlinklen;
727 mp->mnt_flag |= MNT_LOCAL;
730 ump->um_devvp = devvp;
731 ump->um_nindir = fs->fs_nindir;
732 ump->um_bptrtodb = fs->fs_fsbtodb;
733 ump->um_seqinc = fs->fs_frag;
734 for (i = 0; i < MAXQUOTAS; i++)
735 ump->um_quotas[i] = NULLVP;
736 dev->si_mountpoint = mp;
739 if( mp->mnt_flag & MNT_ROOTFS) {
741 * Root mount; update timestamp in mount structure.
742 * this will be used by the common root mount code
743 * to update the system clock.
745 mp->mnt_time = fs->fs_time;
748 ump->um_savedmaxfilesize = fs->fs_maxfilesize; /* XXX */
749 maxfilesize = (uint64_t)0x40000000 * fs->fs_bsize - 1; /* XXX */
750 /* Enforce limit caused by vm object backing (32 bits vm_pindex_t). */
751 if (maxfilesize > (uint64_t)0x80000000u * PAGE_SIZE - 1)
752 maxfilesize = (uint64_t)0x80000000u * PAGE_SIZE - 1;
753 if (fs->fs_maxfilesize > maxfilesize) /* XXX */
754 fs->fs_maxfilesize = maxfilesize; /* XXX */
756 if ((fs->fs_flags & FS_DOSOFTDEP) &&
757 (error = softdep_mount(devvp, mp, fs)) != 0) {
758 free(fs->fs_csp, M_UFSMNT);
763 (void) ffs_sbupdate(ump, MNT_WAIT);
765 vfs_add_vnodeops(mp, &mp->mnt_vn_norm_ops, ffs_vnodeop_entries, VVF_SUPPORTS_FSMID);
766 vfs_add_vnodeops(mp, &mp->mnt_vn_spec_ops, ffs_specop_entries, VVF_SUPPORTS_FSMID);
767 vfs_add_vnodeops(mp, &mp->mnt_vn_fifo_ops, ffs_fifoop_entries, VVF_SUPPORTS_FSMID);
771 dev->si_mountpoint = NULL;
774 VOP_CLOSE(devvp, ronly ? FREAD : FREAD|FWRITE);
776 free(ump->um_fs, M_UFSMNT);
778 mp->mnt_data = (qaddr_t)0;
784 * Sanity checks for old filesystems.
786 * XXX - goes away some day.
789 ffs_oldfscompat(struct fs *fs)
791 fs->fs_npsect = max(fs->fs_npsect, fs->fs_nsect); /* XXX */
792 fs->fs_interleave = max(fs->fs_interleave, 1); /* XXX */
793 if (fs->fs_postblformat == FS_42POSTBLFMT) /* XXX */
794 fs->fs_nrpos = 8; /* XXX */
795 if (fs->fs_inodefmt < FS_44INODEFMT) { /* XXX */
798 uint64_t sizepb = fs->fs_bsize; /* XXX */
800 fs->fs_maxfilesize = fs->fs_bsize * NDADDR - 1; /* XXX */
801 for (i = 0; i < NIADDR; i++) { /* XXX */
802 sizepb *= NINDIR(fs); /* XXX */
803 fs->fs_maxfilesize += sizepb; /* XXX */
806 fs->fs_maxfilesize = (u_quad_t) 1LL << 39;
807 fs->fs_qbmask = ~fs->fs_bmask; /* XXX */
808 fs->fs_qfmask = ~fs->fs_fmask; /* XXX */
814 * unmount system call
817 ffs_unmount(struct mount *mp, int mntflags)
819 struct ufsmount *ump;
824 if (mntflags & MNT_FORCE) {
827 if (mp->mnt_flag & MNT_SOFTDEP) {
828 if ((error = softdep_flushfiles(mp, flags)) != 0)
831 if ((error = ffs_flushfiles(mp, flags)) != 0)
836 if (fs->fs_ronly == 0) {
837 fs->fs_clean = fs->fs_flags & FS_UNCLEAN ? 0 : 1;
838 error = ffs_sbupdate(ump, MNT_WAIT);
844 ump->um_devvp->v_rdev->si_mountpoint = NULL;
846 vinvalbuf(ump->um_devvp, V_SAVE, 0, 0);
847 error = VOP_CLOSE(ump->um_devvp, fs->fs_ronly ? FREAD : FREAD|FWRITE);
849 vrele(ump->um_devvp);
851 free(fs->fs_csp, M_UFSMNT);
854 mp->mnt_data = (qaddr_t)0;
855 mp->mnt_flag &= ~MNT_LOCAL;
860 * Flush out all the files in a filesystem.
863 ffs_flushfiles(struct mount *mp, int flags)
865 struct ufsmount *ump;
870 if (mp->mnt_flag & MNT_QUOTA) {
872 error = vflush(mp, 0, SKIPSYSTEM|flags);
875 /* Find out how many quota files we have open. */
876 for (i = 0; i < MAXQUOTAS; i++) {
877 if (ump->um_quotas[i] == NULLVP)
882 * Here we fall through to vflush again to ensure
883 * that we have gotten rid of all the system vnodes.
888 * Flush all the files.
890 if ((error = vflush(mp, 0, flags)) != 0)
893 * Flush filesystem metadata.
895 vn_lock(ump->um_devvp, LK_EXCLUSIVE | LK_RETRY);
896 error = VOP_FSYNC(ump->um_devvp, MNT_WAIT);
897 VOP_UNLOCK(ump->um_devvp, 0);
902 * Get filesystem statistics.
905 ffs_statfs(struct mount *mp, struct statfs *sbp, struct ucred *cred)
907 struct ufsmount *ump;
912 if (fs->fs_magic != FS_MAGIC)
914 sbp->f_bsize = fs->fs_fsize;
915 sbp->f_iosize = fs->fs_bsize;
916 sbp->f_blocks = fs->fs_dsize;
917 sbp->f_bfree = fs->fs_cstotal.cs_nbfree * fs->fs_frag +
918 fs->fs_cstotal.cs_nffree;
919 sbp->f_bavail = freespace(fs, fs->fs_minfree);
920 sbp->f_files = fs->fs_ncg * fs->fs_ipg - ROOTINO;
921 sbp->f_ffree = fs->fs_cstotal.cs_nifree;
922 if (sbp != &mp->mnt_stat) {
923 sbp->f_type = mp->mnt_vfc->vfc_typenum;
924 bcopy((caddr_t)mp->mnt_stat.f_mntfromname,
925 (caddr_t)&sbp->f_mntfromname[0], MNAMELEN);
931 * Go through the disk queues to initiate sandbagged IO;
932 * go through the inodes to write those that have been modified;
933 * initiate the writing of the super block if it has been modified.
935 * Note: we are always called with the filesystem marked `MPBUSY'.
939 static int ffs_sync_scan1(struct mount *mp, struct vnode *vp, void *data);
940 static int ffs_sync_scan2(struct mount *mp, struct vnode *vp, void *data);
943 ffs_sync(struct mount *mp, int waitfor)
945 struct ufsmount *ump = VFSTOUFS(mp);
948 struct scaninfo scaninfo;
951 if (fs->fs_fmod != 0 && fs->fs_ronly != 0) { /* XXX */
952 printf("fs = %s\n", fs->fs_fsmnt);
953 panic("ffs_sync: rofs mod");
957 * Write back each (modified) inode.
959 scaninfo.allerror = 0;
961 scaninfo.waitfor = waitfor;
962 while (scaninfo.rescan) {
964 vmntvnodescan(mp, VMSC_GETVP|VMSC_NOWAIT,
965 ffs_sync_scan1, ffs_sync_scan2, &scaninfo);
969 * Force stale filesystem control information to be flushed.
971 if (waitfor != MNT_LAZY) {
972 if (ump->um_mountp->mnt_flag & MNT_SOFTDEP)
973 waitfor = MNT_NOWAIT;
974 vn_lock(ump->um_devvp, LK_EXCLUSIVE | LK_RETRY);
975 if ((error = VOP_FSYNC(ump->um_devvp, waitfor)) != 0)
976 scaninfo.allerror = error;
977 VOP_UNLOCK(ump->um_devvp, 0);
983 * Write back modified superblock.
985 if (fs->fs_fmod != 0 && (error = ffs_sbupdate(ump, waitfor)) != 0)
986 scaninfo.allerror = error;
987 return (scaninfo.allerror);
991 ffs_sync_scan1(struct mount *mp, struct vnode *vp, void *data)
996 * Depend on the mount list's vnode lock to keep things stable
997 * enough for a quick test. Since there might be hundreds of
998 * thousands of vnodes, we cannot afford even a subroutine
999 * call unless there's a good chance that we have work to do.
1002 /* Restart out whole search if this guy is locked
1003 * or is being reclaimed.
1005 if (vp->v_type == VNON || ((ip->i_flag &
1006 (IN_ACCESS | IN_CHANGE | IN_MODIFIED | IN_UPDATE)) == 0 &&
1007 RB_EMPTY(&vp->v_rbdirty_tree))) {
1014 ffs_sync_scan2(struct mount *mp, struct vnode *vp, void *data)
1016 struct scaninfo *info = data;
1021 * We have to recheck after having obtained the vnode interlock.
1024 if (vp->v_type == VNON || ((ip->i_flag &
1025 (IN_ACCESS | IN_CHANGE | IN_MODIFIED | IN_UPDATE)) == 0 &&
1026 RB_EMPTY(&vp->v_rbdirty_tree))) {
1029 if (vp->v_type != VCHR) {
1030 if ((error = VOP_FSYNC(vp, info->waitfor)) != 0)
1031 info->allerror = error;
1034 * We must reference the vp to prevent it from
1035 * getting ripped out from under ffs_update, since
1036 * we are not holding a vnode lock.
1038 /* ffs_update(vp, waitfor == MNT_WAIT); */
1045 * Look up a FFS dinode number to find its incore vnode, otherwise read it
1046 * in from disk. If it is in core, wait for the lock bit to clear, then
1047 * return the inode locked. Detection and handling of mount points must be
1048 * done by the calling routine.
1052 ffs_vget(struct mount *mp, ino_t ino, struct vnode **vpp)
1056 struct ufsmount *ump;
1065 if ((*vpp = ufs_ihashget(dev, ino)) != NULL) {
1070 * If this MALLOC() is performed after the getnewvnode()
1071 * it might block, leaving a vnode with a NULL v_data to be
1072 * found by ffs_sync() if a sync happens to fire right then,
1073 * which will cause a panic because ffs_sync() blindly
1074 * dereferences vp->v_data (as well it should).
1076 * XXX this may no longer be true since getnewvnode returns a
1077 * VX locked vnode now.
1079 MALLOC(ip, struct inode *, sizeof(struct inode),
1080 ump->um_malloctype, M_WAITOK);
1082 /* Allocate a new vnode/inode. */
1083 error = getnewvnode(VT_UFS, mp, &vp, VLKTIMEOUT, LK_CANRECURSE);
1086 free(ip, ump->um_malloctype);
1089 bzero((caddr_t)ip, sizeof(struct inode));
1091 ip->i_fs = fs = ump->um_fs;
1097 for (i = 0; i < MAXQUOTAS; i++)
1098 ip->i_dquot[i] = NODQUOT;
1103 * Insert it into the inode hash table and check for a collision.
1104 * If a collision occurs, throw away the vnode and try again.
1106 if (ufs_ihashins(ip) != 0) {
1107 printf("debug: ufs ihashins collision, retrying inode %ld\n",
1108 (long)ip->i_number);
1110 free(ip, ump->um_malloctype);
1115 /* Read in the disk contents for the inode, copy into the inode. */
1116 error = bread(ump->um_devvp, fsbtodoff(fs, ino_to_fsba(fs, ino)),
1117 (int)fs->fs_bsize, &bp);
1120 * The inode does not contain anything useful, so it would
1121 * be misleading to leave it on its hash chain. With mode
1122 * still zero, it will be unlinked and returned to the free
1130 ip->i_din = *((struct ufs1_dinode *)bp->b_data + ino_to_fsbo(fs, ino));
1131 if (DOINGSOFTDEP(vp))
1132 softdep_load_inodeblock(ip);
1134 ip->i_effnlink = ip->i_nlink;
1138 * Initialize the vnode from the inode, check for aliases.
1139 * Note that the underlying vnode may have changed.
1141 error = ufs_vinit(mp, &vp);
1148 * Finish inode initialization now that aliasing has been resolved.
1150 ip->i_devvp = ump->um_devvp;
1153 * Set up a generation number for this inode if it does not
1154 * already have one. This should only happen on old filesystems.
1156 if (ip->i_gen == 0) {
1157 ip->i_gen = random() / 2 + 1;
1158 if ((vp->v_mount->mnt_flag & MNT_RDONLY) == 0)
1159 ip->i_flag |= IN_MODIFIED;
1162 * Ensure that uid and gid are correct. This is a temporary
1163 * fix until fsck has been changed to do the update.
1165 if (fs->fs_inodefmt < FS_44INODEFMT) { /* XXX */
1166 ip->i_uid = ip->i_din.di_ouid; /* XXX */
1167 ip->i_gid = ip->i_din.di_ogid; /* XXX */
1171 * return a VX locked and refd vnode (VX == same as normal vget()
1172 * vnode so we are ok)
1179 * File handle to vnode
1181 * Have to be really careful about stale file handles:
1182 * - check that the inode number is valid
1183 * - call ffs_vget() to get the locked inode
1184 * - check for an unallocated inode (i_mode == 0)
1185 * - check that the given client host has export rights and return
1186 * those rights via. exflagsp and credanonp
1189 ffs_fhtovp(struct mount *mp, struct fid *fhp, struct vnode **vpp)
1194 ufhp = (struct ufid *)fhp;
1195 fs = VFSTOUFS(mp)->um_fs;
1196 if (ufhp->ufid_ino < ROOTINO ||
1197 ufhp->ufid_ino >= fs->fs_ncg * fs->fs_ipg)
1199 return (ufs_fhtovp(mp, ufhp, vpp));
1203 * Vnode pointer to File handle
1207 ffs_vptofh(struct vnode *vp, struct fid *fhp)
1213 ufhp = (struct ufid *)fhp;
1214 ufhp->ufid_len = sizeof(struct ufid);
1215 ufhp->ufid_ino = ip->i_number;
1216 ufhp->ufid_gen = ip->i_gen;
1221 * Initialize the filesystem; just use ufs_init.
1224 ffs_init(struct vfsconf *vfsp)
1226 softdep_initialize();
1227 return (ufs_init(vfsp));
1231 * Write a superblock and associated information back to disk.
1234 ffs_sbupdate(struct ufsmount *mp, int waitfor)
1236 struct fs *dfs, *fs = mp->um_fs;
1240 int i, size, error, allerror = 0;
1243 * First write back the summary information.
1245 * NOTE: the getblk is relative to the device vnode so bio1
1246 * contains the device block number.
1248 blks = howmany(fs->fs_cssize, fs->fs_fsize);
1250 for (i = 0; i < blks; i += fs->fs_frag) {
1251 size = fs->fs_bsize;
1252 if (i + fs->fs_frag > blks)
1253 size = (blks - i) * fs->fs_fsize;
1254 bp = getblk(mp->um_devvp, fsbtodoff(fs, fs->fs_csaddr + i),
1256 bcopy(space, bp->b_data, (uint)size);
1257 space = (char *)space + size;
1258 if (waitfor != MNT_WAIT)
1260 else if ((error = bwrite(bp)) != 0)
1264 * Now write back the superblock itself. If any errors occurred
1265 * up to this point, then fail so that the superblock avoids
1266 * being written out as clean.
1270 bp = getblk(mp->um_devvp, SBOFF, (int)fs->fs_sbsize, 0, 0);
1272 fs->fs_time = time_second;
1273 bcopy((caddr_t)fs, bp->b_data, (uint)fs->fs_sbsize);
1274 /* Restore compatibility to old filesystems. XXX */
1275 dfs = (struct fs *)bp->b_data; /* XXX */
1276 if (fs->fs_postblformat == FS_42POSTBLFMT) /* XXX */
1277 dfs->fs_nrpos = -1; /* XXX */
1278 if (fs->fs_inodefmt < FS_44INODEFMT) { /* XXX */
1279 int32_t *lp, tmp; /* XXX */
1281 lp = (int32_t *)&dfs->fs_qbmask; /* XXX */
1282 tmp = lp[4]; /* XXX */
1283 for (i = 4; i > 0; i--) /* XXX */
1284 lp[i] = lp[i-1]; /* XXX */
1285 lp[0] = tmp; /* XXX */
1287 dfs->fs_maxfilesize = mp->um_savedmaxfilesize; /* XXX */
1288 if (waitfor != MNT_WAIT)
1290 else if ((error = bwrite(bp)) != 0)