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 1.28 2004/11/12 00:09:52 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);
356 * Since this is a new mount, we want the names for
357 * the device and the mount point copied in. If an
358 * error occurs, the mountpoint is discarded by the
361 /* Save "last mounted on" info for mount point (NULL pad)*/
362 copyinstr( path, /* mount point*/
363 mp->mnt_stat.f_mntonname, /* save area*/
364 MNAMELEN - 1, /* max size*/
365 &size); /* real size*/
366 bzero( mp->mnt_stat.f_mntonname + size, MNAMELEN - size);
368 /* Save "mounted from" info for mount point (NULL pad)*/
369 copyinstr( args.fspec, /* device name*/
370 mp->mnt_stat.f_mntfromname, /* save area*/
371 MNAMELEN - 1, /* max size*/
372 &size); /* real size*/
373 bzero( mp->mnt_stat.f_mntfromname + size, MNAMELEN - size);
375 error = ffs_mountfs(devvp, mp, td, M_FFSNODE);
383 * Initialize FS stat information in mount struct; uses both
384 * mp->mnt_stat.f_mntonname and mp->mnt_stat.f_mntfromname
386 * This code is common to root and non-root mounts
388 (void)VFS_STATFS(mp, &mp->mnt_stat, td);
393 error_2: /* error with devvp held*/
395 /* release devvp before failing*/
398 error_1: /* no state to back out*/
401 if (!error && path && (mp->mnt_flag & MNT_UPDATE)) {
402 /* Update clean flag after changing read-onlyness. */
404 if (ronly != fs->fs_ronly) {
405 fs->fs_ronly = ronly;
406 fs->fs_clean = ronly &&
407 (fs->fs_flags & FS_UNCLEAN) == 0 ? 1 : 0;
408 ffs_sbupdate(ump, MNT_WAIT);
415 * Reload all incore data for a filesystem (used after running fsck on
416 * the root filesystem and finding things to fix). The filesystem must
417 * be mounted read-only.
419 * Things to do to update the mount:
420 * 1) invalidate all cached meta-data.
421 * 2) re-read superblock from disk.
422 * 3) re-read summary information from disk.
423 * 4) invalidate all inactive vnodes.
424 * 5) invalidate all cached file data.
425 * 6) re-read inode data for all active vnodes.
428 static int ffs_reload_scan2(struct mount *mp, struct vnode *vp, void *data);
440 ffs_reload(struct mount *mp, struct ucred *cred, struct thread *td)
445 struct fs *fs, *newfs;
446 struct partinfo dpart;
448 int i, blks, size, error;
449 struct scaninfo scaninfo;
452 if ((mp->mnt_flag & MNT_RDONLY) == 0)
455 * Step 1: invalidate all cached meta-data.
457 devvp = VFSTOUFS(mp)->um_devvp;
458 vn_lock(devvp, LK_EXCLUSIVE | LK_RETRY, td);
459 error = vinvalbuf(devvp, 0, td, 0, 0);
460 VOP_UNLOCK(devvp, 0, td);
462 panic("ffs_reload: dirty1");
466 * Only VMIO the backing device if the backing device is a real
467 * block device. See ffs_mountmfs() for more details.
469 if (devvp->v_tag != VT_MFS && vn_isdisk(devvp, NULL)) {
470 vn_lock(devvp, LK_EXCLUSIVE | LK_RETRY, td);
471 vfs_object_create(devvp, td);
472 VOP_UNLOCK(devvp, 0, td);
476 * Step 2: re-read superblock from disk.
478 if (VOP_IOCTL(devvp, DIOCGPART, (caddr_t)&dpart, FREAD, NOCRED, td) != 0)
481 size = dpart.disklab->d_secsize;
482 if ((error = bread(devvp, (ufs_daddr_t)(SBOFF/size), SBSIZE, &bp)) != 0)
487 newfs = (struct fs *)bp->b_data;
488 if (newfs->fs_magic != FS_MAGIC || newfs->fs_bsize > MAXBSIZE ||
489 newfs->fs_bsize < sizeof(struct fs)) {
491 return (EIO); /* XXX needs translation */
493 fs = VFSTOUFS(mp)->um_fs;
495 * Copy pointer fields back into superblock before copying in XXX
496 * new superblock. These should really be in the ufsmount. XXX
497 * Note that important parameters (eg fs_ncg) are unchanged.
499 newfs->fs_csp = fs->fs_csp;
500 newfs->fs_maxcluster = fs->fs_maxcluster;
501 newfs->fs_contigdirs = fs->fs_contigdirs;
502 /* The filesystem is still read-only. */
504 bcopy(newfs, fs, (uint)fs->fs_sbsize);
505 if (fs->fs_sbsize < SBSIZE)
506 bp->b_flags |= B_INVAL;
508 mp->mnt_maxsymlinklen = fs->fs_maxsymlinklen;
510 /* An old fsck may have zeroed these fields, so recheck them. */
511 if (fs->fs_avgfilesize <= 0) /* XXX */
512 fs->fs_avgfilesize = AVFILESIZ; /* XXX */
513 if (fs->fs_avgfpdir <= 0) /* XXX */
514 fs->fs_avgfpdir = AFPDIR; /* XXX */
517 * Step 3: re-read summary information from disk.
519 blks = howmany(fs->fs_cssize, fs->fs_fsize);
521 for (i = 0; i < blks; i += fs->fs_frag) {
523 if (i + fs->fs_frag > blks)
524 size = (blks - i) * fs->fs_fsize;
525 error = bread(devvp, fsbtodb(fs, fs->fs_csaddr + i), size, &bp);
530 bcopy(bp->b_data, space, (uint)size);
531 space = (char *)space + size;
535 * We no longer know anything about clusters per cylinder group.
537 if (fs->fs_contigsumsize > 0) {
538 lp = fs->fs_maxcluster;
539 for (i = 0; i < fs->fs_ncg; i++)
540 *lp++ = fs->fs_contigsumsize;
545 scaninfo.devvp = devvp;
547 while (error == 0 && scaninfo.rescan) {
549 error = vmntvnodescan(mp, VMSC_GETVX,
550 NULL, ffs_reload_scan2, &scaninfo);
556 ffs_reload_scan2(struct mount *mp, struct vnode *vp, void *data)
558 struct scaninfo *info = data;
566 if (vrecycle(vp, info->td))
569 if (vinvalbuf(vp, 0, info->td, 0, 0))
570 panic("ffs_reload: dirty2");
572 * Step 6: re-read inode data for all active vnodes.
575 error = bread(info->devvp,
576 fsbtodb(info->fs, ino_to_fsba(info->fs, ip->i_number)),
577 (int)info->fs->fs_bsize, &bp);
582 ip->i_din = *((struct dinode *)bp->b_data +
583 ino_to_fsbo(info->fs, ip->i_number));
584 ip->i_effnlink = ip->i_nlink;
590 * Common code for mount and mountroot
593 ffs_mountfs(struct vnode *devvp, struct mount *mp, struct thread *td,
594 struct malloc_type *malloctype)
596 struct ufsmount *ump;
600 struct partinfo dpart;
602 int error, i, blks, size, ronly;
604 uint64_t maxfilesize; /* XXX */
608 * Disallow multiple mounts of the same device.
609 * Disallow mounting of a device that is currently in use
610 * Flush out any old buffers remaining from a previous use.
612 error = vfs_mountedon(devvp);
615 if (count_udev(devvp->v_udev) > 0)
617 vn_lock(devvp, LK_EXCLUSIVE | LK_RETRY, td);
618 error = vinvalbuf(devvp, V_SAVE, td, 0, 0);
619 VOP_UNLOCK(devvp, 0, td);
623 ronly = (mp->mnt_flag & MNT_RDONLY) != 0;
624 vn_lock(devvp, LK_EXCLUSIVE | LK_RETRY, td);
625 error = VOP_OPEN(devvp, ronly ? FREAD : FREAD|FWRITE, FSCRED, NULL, td);
626 VOP_UNLOCK(devvp, 0, td);
630 if (dev->si_iosize_max != 0)
631 mp->mnt_iosize_max = dev->si_iosize_max;
632 if (mp->mnt_iosize_max > MAXPHYS)
633 mp->mnt_iosize_max = MAXPHYS;
636 * Only VMIO the backing device if the backing device is a real
637 * block device. This excludes the original MFS implementation.
638 * Note that it is optional that the backing device be VMIOed. This
639 * increases the opportunity for metadata caching.
641 * This call must be made after the VOP_OPEN.
643 if (devvp->v_tag != VT_MFS && vn_isdisk(devvp, NULL)) {
644 vn_lock(devvp, LK_EXCLUSIVE | LK_RETRY, td);
645 vfs_object_create(devvp, td);
646 VOP_UNLOCK(devvp, 0, td);
649 if (VOP_IOCTL(devvp, DIOCGPART, (caddr_t)&dpart, FREAD, proc0.p_ucred, td) != 0)
652 size = dpart.disklab->d_secsize;
656 if ((error = bread(devvp, SBLOCK, SBSIZE, &bp)) != 0)
658 fs = (struct fs *)bp->b_data;
659 if (fs->fs_magic != FS_MAGIC || fs->fs_bsize > MAXBSIZE ||
660 fs->fs_bsize < sizeof(struct fs)) {
661 error = EINVAL; /* XXX needs translation */
665 fs->fs_flags &= ~FS_UNCLEAN;
666 if (fs->fs_clean == 0) {
667 fs->fs_flags |= FS_UNCLEAN;
668 if (ronly || (mp->mnt_flag & MNT_FORCE)) {
670 "WARNING: %s was not properly dismounted\n",
674 "WARNING: R/W mount of %s denied. Filesystem is not clean - run fsck\n",
680 /* XXX updating 4.2 FFS superblocks trashes rotational layout tables */
681 if (fs->fs_postblformat == FS_42POSTBLFMT && !ronly) {
682 error = EROFS; /* needs translation */
685 ump = malloc(sizeof *ump, M_UFSMNT, M_WAITOK);
686 bzero((caddr_t)ump, sizeof *ump);
687 ump->um_malloctype = malloctype;
688 ump->um_i_effnlink_valid = 1;
689 ump->um_fs = malloc((u_long)fs->fs_sbsize, M_UFSMNT,
691 ump->um_blkatoff = ffs_blkatoff;
692 ump->um_truncate = ffs_truncate;
693 ump->um_update = ffs_update;
694 ump->um_valloc = ffs_valloc;
695 ump->um_vfree = ffs_vfree;
696 bcopy(bp->b_data, ump->um_fs, (uint)fs->fs_sbsize);
697 if (fs->fs_sbsize < SBSIZE)
698 bp->b_flags |= B_INVAL;
702 fs->fs_ronly = ronly;
703 size = fs->fs_cssize;
704 blks = howmany(size, fs->fs_fsize);
705 if (fs->fs_contigsumsize > 0)
706 size += fs->fs_ncg * sizeof(int32_t);
707 size += fs->fs_ncg * sizeof(uint8_t);
708 space = malloc((u_long)size, M_UFSMNT, M_WAITOK);
710 for (i = 0; i < blks; i += fs->fs_frag) {
712 if (i + fs->fs_frag > blks)
713 size = (blks - i) * fs->fs_fsize;
714 if ((error = bread(devvp, fsbtodb(fs, fs->fs_csaddr + i), size,
716 free(fs->fs_csp, M_UFSMNT);
719 bcopy(bp->b_data, space, (uint)size);
720 space = (char *)space + size;
724 if (fs->fs_contigsumsize > 0) {
725 fs->fs_maxcluster = lp = space;
726 for (i = 0; i < fs->fs_ncg; i++)
727 *lp++ = fs->fs_contigsumsize;
730 size = fs->fs_ncg * sizeof(uint8_t);
731 fs->fs_contigdirs = (uint8_t *)space;
732 bzero(fs->fs_contigdirs, size);
733 /* Compatibility for old filesystems XXX */
734 if (fs->fs_avgfilesize <= 0) /* XXX */
735 fs->fs_avgfilesize = AVFILESIZ; /* XXX */
736 if (fs->fs_avgfpdir <= 0) /* XXX */
737 fs->fs_avgfpdir = AFPDIR; /* XXX */
738 mp->mnt_data = (qaddr_t)ump;
739 mp->mnt_stat.f_fsid.val[0] = fs->fs_id[0];
740 mp->mnt_stat.f_fsid.val[1] = fs->fs_id[1];
741 if (fs->fs_id[0] == 0 || fs->fs_id[1] == 0 ||
742 vfs_getvfs(&mp->mnt_stat.f_fsid))
744 mp->mnt_maxsymlinklen = fs->fs_maxsymlinklen;
745 mp->mnt_flag |= MNT_LOCAL;
748 ump->um_devvp = devvp;
749 ump->um_nindir = fs->fs_nindir;
750 ump->um_bptrtodb = fs->fs_fsbtodb;
751 ump->um_seqinc = fs->fs_frag;
752 for (i = 0; i < MAXQUOTAS; i++)
753 ump->um_quotas[i] = NULLVP;
754 dev->si_mountpoint = mp;
758 * Set FS local "last mounted on" information (NULL pad)
760 copystr( mp->mnt_stat.f_mntonname, /* mount point*/
761 fs->fs_fsmnt, /* copy area*/
762 sizeof(fs->fs_fsmnt) - 1, /* max size*/
763 &strsize); /* real size*/
764 bzero( fs->fs_fsmnt + strsize, sizeof(fs->fs_fsmnt) - strsize);
766 if( mp->mnt_flag & MNT_ROOTFS) {
768 * Root mount; update timestamp in mount structure.
769 * this will be used by the common root mount code
770 * to update the system clock.
772 mp->mnt_time = fs->fs_time;
775 ump->um_savedmaxfilesize = fs->fs_maxfilesize; /* XXX */
776 maxfilesize = (uint64_t)0x40000000 * fs->fs_bsize - 1; /* XXX */
777 /* Enforce limit caused by vm object backing (32 bits vm_pindex_t). */
778 if (maxfilesize > (uint64_t)0x80000000u * PAGE_SIZE - 1)
779 maxfilesize = (uint64_t)0x80000000u * PAGE_SIZE - 1;
780 if (fs->fs_maxfilesize > maxfilesize) /* XXX */
781 fs->fs_maxfilesize = maxfilesize; /* XXX */
783 if ((fs->fs_flags & FS_DOSOFTDEP) &&
784 (error = softdep_mount(devvp, mp, fs)) != 0) {
785 free(fs->fs_csp, M_UFSMNT);
790 (void) ffs_sbupdate(ump, MNT_WAIT);
792 vfs_add_vnodeops(&mp->mnt_vn_ops, ffs_vnodeop_entries);
793 vfs_add_vnodeops(&mp->mnt_vn_spec_ops, ffs_specop_entries);
794 vfs_add_vnodeops(&mp->mnt_vn_fifo_ops, ffs_fifoop_entries);
798 dev->si_mountpoint = NULL;
801 VOP_CLOSE(devvp, ronly ? FREAD : FREAD|FWRITE, td);
803 free(ump->um_fs, M_UFSMNT);
805 mp->mnt_data = (qaddr_t)0;
811 * Sanity checks for old filesystems.
813 * XXX - goes away some day.
816 ffs_oldfscompat(struct fs *fs)
818 fs->fs_npsect = max(fs->fs_npsect, fs->fs_nsect); /* XXX */
819 fs->fs_interleave = max(fs->fs_interleave, 1); /* XXX */
820 if (fs->fs_postblformat == FS_42POSTBLFMT) /* XXX */
821 fs->fs_nrpos = 8; /* XXX */
822 if (fs->fs_inodefmt < FS_44INODEFMT) { /* XXX */
825 uint64_t sizepb = fs->fs_bsize; /* XXX */
827 fs->fs_maxfilesize = fs->fs_bsize * NDADDR - 1; /* XXX */
828 for (i = 0; i < NIADDR; i++) { /* XXX */
829 sizepb *= NINDIR(fs); /* XXX */
830 fs->fs_maxfilesize += sizepb; /* XXX */
833 fs->fs_maxfilesize = (u_quad_t) 1LL << 39;
834 fs->fs_qbmask = ~fs->fs_bmask; /* XXX */
835 fs->fs_qfmask = ~fs->fs_fmask; /* XXX */
841 * unmount system call
844 ffs_unmount(struct mount *mp, int mntflags, struct thread *td)
846 struct ufsmount *ump;
851 if (mntflags & MNT_FORCE) {
854 if (mp->mnt_flag & MNT_SOFTDEP) {
855 if ((error = softdep_flushfiles(mp, flags, td)) != 0)
858 if ((error = ffs_flushfiles(mp, flags, td)) != 0)
863 if (fs->fs_ronly == 0) {
864 fs->fs_clean = fs->fs_flags & FS_UNCLEAN ? 0 : 1;
865 error = ffs_sbupdate(ump, MNT_WAIT);
871 ump->um_devvp->v_rdev->si_mountpoint = NULL;
873 vinvalbuf(ump->um_devvp, V_SAVE, td, 0, 0);
874 error = VOP_CLOSE(ump->um_devvp, fs->fs_ronly ? FREAD : FREAD|FWRITE, td);
876 vrele(ump->um_devvp);
878 free(fs->fs_csp, M_UFSMNT);
881 mp->mnt_data = (qaddr_t)0;
882 mp->mnt_flag &= ~MNT_LOCAL;
887 * Flush out all the files in a filesystem.
890 ffs_flushfiles(struct mount *mp, int flags, struct thread *td)
892 struct ufsmount *ump;
897 if (mp->mnt_flag & MNT_QUOTA) {
899 error = vflush(mp, 0, SKIPSYSTEM|flags);
902 /* Find out how many quota files we have open. */
903 for (i = 0; i < MAXQUOTAS; i++) {
904 if (ump->um_quotas[i] == NULLVP)
909 * Here we fall through to vflush again to ensure
910 * that we have gotten rid of all the system vnodes.
915 * Flush all the files.
917 if ((error = vflush(mp, 0, flags)) != 0)
920 * Flush filesystem metadata.
922 vn_lock(ump->um_devvp, LK_EXCLUSIVE | LK_RETRY, td);
923 error = VOP_FSYNC(ump->um_devvp, MNT_WAIT, td);
924 VOP_UNLOCK(ump->um_devvp, 0, td);
929 * Get filesystem statistics.
932 ffs_statfs(struct mount *mp, struct statfs *sbp, struct thread *td)
934 struct ufsmount *ump;
939 if (fs->fs_magic != FS_MAGIC)
941 sbp->f_bsize = fs->fs_fsize;
942 sbp->f_iosize = fs->fs_bsize;
943 sbp->f_blocks = fs->fs_dsize;
944 sbp->f_bfree = fs->fs_cstotal.cs_nbfree * fs->fs_frag +
945 fs->fs_cstotal.cs_nffree;
946 sbp->f_bavail = freespace(fs, fs->fs_minfree);
947 sbp->f_files = fs->fs_ncg * fs->fs_ipg - ROOTINO;
948 sbp->f_ffree = fs->fs_cstotal.cs_nifree;
949 if (sbp != &mp->mnt_stat) {
950 sbp->f_type = mp->mnt_vfc->vfc_typenum;
951 bcopy((caddr_t)mp->mnt_stat.f_mntonname,
952 (caddr_t)&sbp->f_mntonname[0], MNAMELEN);
953 bcopy((caddr_t)mp->mnt_stat.f_mntfromname,
954 (caddr_t)&sbp->f_mntfromname[0], MNAMELEN);
960 * Go through the disk queues to initiate sandbagged IO;
961 * go through the inodes to write those that have been modified;
962 * initiate the writing of the super block if it has been modified.
964 * Note: we are always called with the filesystem marked `MPBUSY'.
968 static int ffs_sync_scan1(struct mount *mp, struct vnode *vp, void *data);
969 static int ffs_sync_scan2(struct mount *mp, struct vnode *vp, void *data);
972 ffs_sync(struct mount *mp, int waitfor, struct thread *td)
974 struct ufsmount *ump = VFSTOUFS(mp);
977 struct scaninfo scaninfo;
980 if (fs->fs_fmod != 0 && fs->fs_ronly != 0) { /* XXX */
981 printf("fs = %s\n", fs->fs_fsmnt);
982 panic("ffs_sync: rofs mod");
986 * Write back each (modified) inode.
988 scaninfo.allerror = 0;
990 scaninfo.waitfor = waitfor;
991 while (scaninfo.rescan) {
993 vmntvnodescan(mp, VMSC_GETVP|VMSC_NOWAIT,
994 ffs_sync_scan1, ffs_sync_scan2, &scaninfo);
998 * Force stale filesystem control information to be flushed.
1000 if (waitfor != MNT_LAZY) {
1001 if (ump->um_mountp->mnt_flag & MNT_SOFTDEP)
1002 waitfor = MNT_NOWAIT;
1003 vn_lock(ump->um_devvp, LK_EXCLUSIVE | LK_RETRY, td);
1004 if ((error = VOP_FSYNC(ump->um_devvp, waitfor, td)) != 0)
1005 scaninfo.allerror = error;
1006 VOP_UNLOCK(ump->um_devvp, 0, td);
1012 * Write back modified superblock.
1014 if (fs->fs_fmod != 0 && (error = ffs_sbupdate(ump, waitfor)) != 0)
1015 scaninfo.allerror = error;
1016 return (scaninfo.allerror);
1020 ffs_sync_scan1(struct mount *mp, struct vnode *vp, void *data)
1025 * Depend on the mount list's vnode lock to keep things stable
1026 * enough for a quick test. Since there might be hundreds of
1027 * thousands of vnodes, we cannot afford even a subroutine
1028 * call unless there's a good chance that we have work to do.
1031 /* Restart out whole search if this guy is locked
1032 * or is being reclaimed.
1034 if (vp->v_type == VNON || ((ip->i_flag &
1035 (IN_ACCESS | IN_CHANGE | IN_MODIFIED | IN_UPDATE)) == 0 &&
1036 TAILQ_EMPTY(&vp->v_dirtyblkhd))) {
1043 ffs_sync_scan2(struct mount *mp, struct vnode *vp, void *data)
1045 struct scaninfo *info = data;
1046 thread_t td = curthread; /* XXX */
1051 * We have to recheck after having obtained the vnode interlock.
1054 if (vp->v_type == VNON || ((ip->i_flag &
1055 (IN_ACCESS | IN_CHANGE | IN_MODIFIED | IN_UPDATE)) == 0 &&
1056 TAILQ_EMPTY(&vp->v_dirtyblkhd))) {
1059 if (vp->v_type != VCHR) {
1060 if ((error = VOP_FSYNC(vp, info->waitfor, td)) != 0)
1061 info->allerror = error;
1064 * We must reference the vp to prevent it from
1065 * getting ripped out from under UFS_UPDATE, since
1066 * we are not holding a vnode lock.
1068 /* UFS_UPDATE(vp, waitfor == MNT_WAIT); */
1075 * Look up a FFS dinode number to find its incore vnode, otherwise read it
1076 * in from disk. If it is in core, wait for the lock bit to clear, then
1077 * return the inode locked. Detection and handling of mount points must be
1078 * done by the calling routine.
1082 ffs_vget(struct mount *mp, ino_t ino, struct vnode **vpp)
1086 struct ufsmount *ump;
1095 if ((*vpp = ufs_ihashget(dev, ino)) != NULL) {
1100 * If this MALLOC() is performed after the getnewvnode()
1101 * it might block, leaving a vnode with a NULL v_data to be
1102 * found by ffs_sync() if a sync happens to fire right then,
1103 * which will cause a panic because ffs_sync() blindly
1104 * dereferences vp->v_data (as well it should).
1106 * XXX this may no longer be true since getnewvnode returns a
1107 * VX locked vnode now.
1109 MALLOC(ip, struct inode *, sizeof(struct inode),
1110 ump->um_malloctype, M_WAITOK);
1112 /* Allocate a new vnode/inode. */
1113 error = getnewvnode(VT_UFS, mp, mp->mnt_vn_ops, &vp,
1114 VLKTIMEOUT, LK_CANRECURSE);
1117 free(ip, ump->um_malloctype);
1120 bzero((caddr_t)ip, sizeof(struct inode));
1122 ip->i_fs = fs = ump->um_fs;
1128 for (i = 0; i < MAXQUOTAS; i++)
1129 ip->i_dquot[i] = NODQUOT;
1134 * Insert it into the inode hash table and check for a collision.
1135 * If a collision occurs, throw away the vnode and try again.
1137 if (ufs_ihashins(ip) != 0) {
1138 printf("debug: ufs ihashins collision, retrying inode %ld\n",
1139 (long)ip->i_number);
1141 free(ip, ump->um_malloctype);
1146 /* Read in the disk contents for the inode, copy into the inode. */
1147 error = bread(ump->um_devvp, fsbtodb(fs, ino_to_fsba(fs, ino)),
1148 (int)fs->fs_bsize, &bp);
1151 * The inode does not contain anything useful, so it would
1152 * be misleading to leave it on its hash chain. With mode
1153 * still zero, it will be unlinked and returned to the free
1161 ip->i_din = *((struct dinode *)bp->b_data + ino_to_fsbo(fs, ino));
1162 if (DOINGSOFTDEP(vp))
1163 softdep_load_inodeblock(ip);
1165 ip->i_effnlink = ip->i_nlink;
1169 * Initialize the vnode from the inode, check for aliases.
1170 * Note that the underlying vnode may have changed.
1172 error = ufs_vinit(mp, &vp);
1179 * Finish inode initialization now that aliasing has been resolved.
1181 ip->i_devvp = ump->um_devvp;
1184 * Set up a generation number for this inode if it does not
1185 * already have one. This should only happen on old filesystems.
1187 if (ip->i_gen == 0) {
1188 ip->i_gen = random() / 2 + 1;
1189 if ((vp->v_mount->mnt_flag & MNT_RDONLY) == 0)
1190 ip->i_flag |= IN_MODIFIED;
1193 * Ensure that uid and gid are correct. This is a temporary
1194 * fix until fsck has been changed to do the update.
1196 if (fs->fs_inodefmt < FS_44INODEFMT) { /* XXX */
1197 ip->i_uid = ip->i_din.di_ouid; /* XXX */
1198 ip->i_gid = ip->i_din.di_ogid; /* XXX */
1202 * return a VX locked and refd vnode (VX == same as normal vget()
1203 * vnode so we are ok)
1210 * File handle to vnode
1212 * Have to be really careful about stale file handles:
1213 * - check that the inode number is valid
1214 * - call ffs_vget() to get the locked inode
1215 * - check for an unallocated inode (i_mode == 0)
1216 * - check that the given client host has export rights and return
1217 * those rights via. exflagsp and credanonp
1220 ffs_fhtovp(struct mount *mp, struct fid *fhp, struct vnode **vpp)
1225 ufhp = (struct ufid *)fhp;
1226 fs = VFSTOUFS(mp)->um_fs;
1227 if (ufhp->ufid_ino < ROOTINO ||
1228 ufhp->ufid_ino >= fs->fs_ncg * fs->fs_ipg)
1230 return (ufs_fhtovp(mp, ufhp, vpp));
1234 * Vnode pointer to File handle
1238 ffs_vptofh(struct vnode *vp, struct fid *fhp)
1244 ufhp = (struct ufid *)fhp;
1245 ufhp->ufid_len = sizeof(struct ufid);
1246 ufhp->ufid_ino = ip->i_number;
1247 ufhp->ufid_gen = ip->i_gen;
1252 * Initialize the filesystem; just use ufs_init.
1255 ffs_init(struct vfsconf *vfsp)
1257 softdep_initialize();
1258 return (ufs_init(vfsp));
1262 * Write a superblock and associated information back to disk.
1265 ffs_sbupdate(struct ufsmount *mp, int waitfor)
1267 struct fs *dfs, *fs = mp->um_fs;
1271 int i, size, error, allerror = 0;
1274 * First write back the summary information.
1276 blks = howmany(fs->fs_cssize, fs->fs_fsize);
1278 for (i = 0; i < blks; i += fs->fs_frag) {
1279 size = fs->fs_bsize;
1280 if (i + fs->fs_frag > blks)
1281 size = (blks - i) * fs->fs_fsize;
1282 bp = getblk(mp->um_devvp, fsbtodb(fs, fs->fs_csaddr + i),
1284 bcopy(space, bp->b_data, (uint)size);
1285 space = (char *)space + size;
1286 if (waitfor != MNT_WAIT)
1288 else if ((error = bwrite(bp)) != 0)
1292 * Now write back the superblock itself. If any errors occurred
1293 * up to this point, then fail so that the superblock avoids
1294 * being written out as clean.
1298 bp = getblk(mp->um_devvp, SBLOCK, (int)fs->fs_sbsize, 0, 0);
1300 fs->fs_time = time_second;
1301 bcopy((caddr_t)fs, bp->b_data, (uint)fs->fs_sbsize);
1302 /* Restore compatibility to old filesystems. XXX */
1303 dfs = (struct fs *)bp->b_data; /* XXX */
1304 if (fs->fs_postblformat == FS_42POSTBLFMT) /* XXX */
1305 dfs->fs_nrpos = -1; /* XXX */
1306 if (fs->fs_inodefmt < FS_44INODEFMT) { /* XXX */
1307 int32_t *lp, tmp; /* XXX */
1309 lp = (int32_t *)&dfs->fs_qbmask; /* XXX */
1310 tmp = lp[4]; /* XXX */
1311 for (i = 4; i > 0; i--) /* XXX */
1312 lp[i] = lp[i-1]; /* XXX */
1313 lp[0] = tmp; /* XXX */
1315 dfs->fs_maxfilesize = mp->um_savedmaxfilesize; /* XXX */
1316 if (waitfor != MNT_WAIT)
1318 else if ((error = bwrite(bp)) != 0)