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 1.26 2004/10/12 19:21:12 dillon Exp $
38 #include "opt_quota.h"
40 #include <sys/param.h>
41 #include <sys/systm.h>
43 #include <sys/namei.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!!!)
108 * p process (user credentials check [statfs])
111 * mp mount point structure
112 * path path to mount point
113 * data pointer to argument struct in user space
114 * ndp mount point namei() return (used for
115 * credentials on reload), reused to look
117 * p process (user credentials check)
120 * !0 error number (errno.h)
125 * mount point is locked
127 * mount point is locked
130 * A NULL path can be used for a flag since the mount
131 * system call will fail with EFAULT in copyinstr in
132 * namei() if it is a genuine NULL from the user.
135 ffs_mount(struct mount *mp, /* mount struct pointer */
136 char *path, /* path to mount point */
137 caddr_t data, /* arguments to FS specific mount */
138 struct thread *td) /* process requesting mount */
144 struct ufs_args args;
145 struct ufsmount *ump = 0;
147 int error, flags, ronly = 0;
151 struct vnode *rootvp;
153 KKASSERT(td->td_proc);
154 cred = td->td_proc->p_ucred;
157 * Use NULL path to flag a root mount
162 * Mounting root filesystem
166 if ((err = bdevvp(rootdev, &rootvp))) {
167 printf("ffs_mountroot: can't find rootvp\n");
171 if( ( err = ffs_mountfs(rootvp, mp, td, M_FFSNODE)) != 0) {
172 /* fs specific cleanup (if any)*/
176 goto dostatfs; /* success*/
182 * Mounting non-root filesystem or updating a filesystem
186 /* copy in user arguments*/
187 err = copyin(data, (caddr_t)&args, sizeof (struct ufs_args));
189 goto error_1; /* can't get arguments*/
192 * If updating, check whether changing from read-only to
193 * read/write; if there is no device name, that's all we do.
195 if (mp->mnt_flag & MNT_UPDATE) {
198 devvp = ump->um_devvp;
200 ronly = fs->fs_ronly; /* MNT_RELOAD might change this */
201 if (ronly == 0 && (mp->mnt_flag & MNT_RDONLY)) {
203 * Flush any dirty data.
205 VFS_SYNC(mp, MNT_WAIT, td);
207 * Check for and optionally get rid of files open
211 if (mp->mnt_flag & MNT_FORCE)
213 if (mp->mnt_flag & MNT_SOFTDEP) {
214 err = softdep_flushfiles(mp, flags, td);
216 err = ffs_flushfiles(mp, flags, td);
220 if (!err && (mp->mnt_flag & MNT_RELOAD))
221 err = ffs_reload(mp, NULL, td);
225 if (ronly && (mp->mnt_kern_flag & MNTK_WANTRDWR)) {
227 * If upgrade to read-write by non-root, then verify
228 * that user has necessary permissions on the device.
230 if (cred->cr_uid != 0) {
231 vn_lock(devvp, LK_EXCLUSIVE | LK_RETRY, td);
232 if ((error = VOP_ACCESS(devvp, VREAD | VWRITE,
234 VOP_UNLOCK(devvp, 0, td);
237 VOP_UNLOCK(devvp, 0, td);
240 fs->fs_flags &= ~FS_UNCLEAN;
241 if (fs->fs_clean == 0) {
242 fs->fs_flags |= FS_UNCLEAN;
243 if (mp->mnt_flag & MNT_FORCE) {
245 "WARNING: %s was not properly dismounted\n",
249 "WARNING: R/W mount of %s denied. Filesystem is not clean - run fsck\n",
256 /* check to see if we need to start softdep */
257 if (fs->fs_flags & FS_DOSOFTDEP) {
258 err = softdep_mount(devvp, mp, fs);
266 * Soft updates is incompatible with "async",
267 * so if we are doing softupdates stop the user
268 * from setting the async flag in an update.
269 * Softdep_mount() clears it in an initial mount
272 if (mp->mnt_flag & MNT_SOFTDEP) {
273 mp->mnt_flag &= ~MNT_ASYNC;
275 /* if not updating name...*/
276 if (args.fspec == 0) {
278 * Process export requests. Jumping to "success"
279 * will return the vfs_export() error code.
281 err = vfs_export(mp, &ump->um_export, &args.export);
287 * Not an update, or updating the name: look up the name
288 * and verify that it refers to a sensible block device.
290 NDINIT(&nd, NAMEI_LOOKUP, CNP_FOLLOW, UIO_USERSPACE, args.fspec, td);
293 /* can't get devvp!*/
297 NDFREE(&nd, NDF_ONLY_PNBUF);
300 if (!vn_isdisk(devvp, &err))
304 * If mount by non-root, then verify that user has necessary
305 * permissions on the device.
307 if (cred->cr_uid != 0) {
309 if ((mp->mnt_flag & MNT_RDONLY) == 0)
310 accessmode |= VWRITE;
311 vn_lock(devvp, LK_EXCLUSIVE | LK_RETRY, td);
312 if ((error = VOP_ACCESS(devvp, accessmode, cred, td)) != 0) {
316 VOP_UNLOCK(devvp, 0, td);
319 if (mp->mnt_flag & MNT_UPDATE) {
323 * If it's not the same vnode, or at least the same device
324 * then it's not correct. NOTE: devvp->v_rdev may be NULL
325 * since we haven't opened it, so we compare udev instead.
328 if (devvp != ump->um_devvp) {
329 if (devvp->v_udev == ump->um_devvp->v_udev) {
332 printf("cannot update mount, udev does"
333 " not match %08x vs %08x\n",
334 devvp->v_udev, ump->um_devvp->v_udev);
335 err = EINVAL; /* needs translation */
341 * Update device name only on success
344 /* Save "mounted from" info for mount point (NULL pad)*/
345 copyinstr( args.fspec,
346 mp->mnt_stat.f_mntfromname,
349 bzero( mp->mnt_stat.f_mntfromname + size, MNAMELEN - size);
359 * Since this is a new mount, we want the names for
360 * the device and the mount point copied in. If an
361 * error occurs, the mountpoint is discarded by the
364 /* Save "last mounted on" info for mount point (NULL pad)*/
365 copyinstr( path, /* mount point*/
366 mp->mnt_stat.f_mntonname, /* save area*/
367 MNAMELEN - 1, /* max size*/
368 &size); /* real size*/
369 bzero( mp->mnt_stat.f_mntonname + size, MNAMELEN - size);
371 /* Save "mounted from" info for mount point (NULL pad)*/
372 copyinstr( args.fspec, /* device name*/
373 mp->mnt_stat.f_mntfromname, /* save area*/
374 MNAMELEN - 1, /* max size*/
375 &size); /* real size*/
376 bzero( mp->mnt_stat.f_mntfromname + size, MNAMELEN - size);
378 err = ffs_mountfs(devvp, mp, td, M_FFSNODE);
386 * Initialize FS stat information in mount struct; uses both
387 * mp->mnt_stat.f_mntonname and mp->mnt_stat.f_mntfromname
389 * This code is common to root and non-root mounts
391 (void)VFS_STATFS(mp, &mp->mnt_stat, td);
396 error_2: /* error with devvp held*/
398 /* release devvp before failing*/
401 error_1: /* no state to back out*/
404 if (!err && path && (mp->mnt_flag & MNT_UPDATE)) {
405 /* Update clean flag after changing read-onlyness. */
407 if (ronly != fs->fs_ronly) {
408 fs->fs_ronly = ronly;
409 fs->fs_clean = ronly &&
410 (fs->fs_flags & FS_UNCLEAN) == 0 ? 1 : 0;
411 ffs_sbupdate(ump, MNT_WAIT);
418 * Reload all incore data for a filesystem (used after running fsck on
419 * the root filesystem and finding things to fix). The filesystem must
420 * be mounted read-only.
422 * Things to do to update the mount:
423 * 1) invalidate all cached meta-data.
424 * 2) re-read superblock from disk.
425 * 3) re-read summary information from disk.
426 * 4) invalidate all inactive vnodes.
427 * 5) invalidate all cached file data.
428 * 6) re-read inode data for all active vnodes.
431 static int ffs_reload_scan2(struct mount *mp, struct vnode *vp, void *data);
443 ffs_reload(struct mount *mp, struct ucred *cred, struct thread *td)
448 struct fs *fs, *newfs;
449 struct partinfo dpart;
451 int i, blks, size, error;
452 struct scaninfo scaninfo;
455 if ((mp->mnt_flag & MNT_RDONLY) == 0)
458 * Step 1: invalidate all cached meta-data.
460 devvp = VFSTOUFS(mp)->um_devvp;
461 vn_lock(devvp, LK_EXCLUSIVE | LK_RETRY, td);
462 error = vinvalbuf(devvp, 0, td, 0, 0);
463 VOP_UNLOCK(devvp, 0, td);
465 panic("ffs_reload: dirty1");
469 * Only VMIO the backing device if the backing device is a real
470 * block device. See ffs_mountmfs() for more details.
472 if (devvp->v_tag != VT_MFS && vn_isdisk(devvp, NULL)) {
473 vn_lock(devvp, LK_EXCLUSIVE | LK_RETRY, td);
474 vfs_object_create(devvp, td);
475 VOP_UNLOCK(devvp, 0, td);
479 * Step 2: re-read superblock from disk.
481 if (VOP_IOCTL(devvp, DIOCGPART, (caddr_t)&dpart, FREAD, NOCRED, td) != 0)
484 size = dpart.disklab->d_secsize;
485 if ((error = bread(devvp, (ufs_daddr_t)(SBOFF/size), SBSIZE, &bp)) != 0)
490 newfs = (struct fs *)bp->b_data;
491 if (newfs->fs_magic != FS_MAGIC || newfs->fs_bsize > MAXBSIZE ||
492 newfs->fs_bsize < sizeof(struct fs)) {
494 return (EIO); /* XXX needs translation */
496 fs = VFSTOUFS(mp)->um_fs;
498 * Copy pointer fields back into superblock before copying in XXX
499 * new superblock. These should really be in the ufsmount. XXX
500 * Note that important parameters (eg fs_ncg) are unchanged.
502 newfs->fs_csp = fs->fs_csp;
503 newfs->fs_maxcluster = fs->fs_maxcluster;
504 newfs->fs_contigdirs = fs->fs_contigdirs;
505 /* The filesystem is still read-only. */
507 bcopy(newfs, fs, (uint)fs->fs_sbsize);
508 if (fs->fs_sbsize < SBSIZE)
509 bp->b_flags |= B_INVAL;
511 mp->mnt_maxsymlinklen = fs->fs_maxsymlinklen;
513 /* An old fsck may have zeroed these fields, so recheck them. */
514 if (fs->fs_avgfilesize <= 0) /* XXX */
515 fs->fs_avgfilesize = AVFILESIZ; /* XXX */
516 if (fs->fs_avgfpdir <= 0) /* XXX */
517 fs->fs_avgfpdir = AFPDIR; /* XXX */
520 * Step 3: re-read summary information from disk.
522 blks = howmany(fs->fs_cssize, fs->fs_fsize);
524 for (i = 0; i < blks; i += fs->fs_frag) {
526 if (i + fs->fs_frag > blks)
527 size = (blks - i) * fs->fs_fsize;
528 error = bread(devvp, fsbtodb(fs, fs->fs_csaddr + i), size, &bp);
533 bcopy(bp->b_data, space, (uint)size);
534 space = (char *)space + size;
538 * We no longer know anything about clusters per cylinder group.
540 if (fs->fs_contigsumsize > 0) {
541 lp = fs->fs_maxcluster;
542 for (i = 0; i < fs->fs_ncg; i++)
543 *lp++ = fs->fs_contigsumsize;
548 scaninfo.devvp = devvp;
550 while (error == 0 && scaninfo.rescan) {
552 error = vmntvnodescan(mp, VMSC_GETVX,
553 NULL, ffs_reload_scan2, &scaninfo);
559 ffs_reload_scan2(struct mount *mp, struct vnode *vp, void *data)
561 struct scaninfo *info = data;
569 if (vrecycle(vp, info->td))
572 if (vinvalbuf(vp, 0, info->td, 0, 0))
573 panic("ffs_reload: dirty2");
575 * Step 6: re-read inode data for all active vnodes.
578 error = bread(info->devvp,
579 fsbtodb(info->fs, ino_to_fsba(info->fs, ip->i_number)),
580 (int)info->fs->fs_bsize, &bp);
585 ip->i_din = *((struct dinode *)bp->b_data +
586 ino_to_fsbo(info->fs, ip->i_number));
587 ip->i_effnlink = ip->i_nlink;
593 * Common code for mount and mountroot
596 ffs_mountfs(struct vnode *devvp, struct mount *mp, struct thread *td,
597 struct malloc_type *malloctype)
599 struct ufsmount *ump;
603 struct partinfo dpart;
605 int error, i, blks, size, ronly;
607 uint64_t maxfilesize; /* XXX */
611 * Disallow multiple mounts of the same device.
612 * Disallow mounting of a device that is currently in use
613 * Flush out any old buffers remaining from a previous use.
615 error = vfs_mountedon(devvp);
618 if (count_udev(devvp->v_udev) > 0)
620 vn_lock(devvp, LK_EXCLUSIVE | LK_RETRY, td);
621 error = vinvalbuf(devvp, V_SAVE, td, 0, 0);
622 VOP_UNLOCK(devvp, 0, td);
627 * Only VMIO the backing device if the backing device is a real
628 * block device. This excludes the original MFS implementation.
629 * Note that it is optional that the backing device be VMIOed. This
630 * increases the opportunity for metadata caching.
632 if (devvp->v_tag != VT_MFS && vn_isdisk(devvp, NULL)) {
633 vn_lock(devvp, LK_EXCLUSIVE | LK_RETRY, td);
634 vfs_object_create(devvp, td);
635 VOP_UNLOCK(devvp, 0, td);
638 ronly = (mp->mnt_flag & MNT_RDONLY) != 0;
639 vn_lock(devvp, LK_EXCLUSIVE | LK_RETRY, td);
640 error = VOP_OPEN(devvp, ronly ? FREAD : FREAD|FWRITE, FSCRED, td);
641 VOP_UNLOCK(devvp, 0, td);
645 if (dev->si_iosize_max != 0)
646 mp->mnt_iosize_max = dev->si_iosize_max;
647 if (mp->mnt_iosize_max > MAXPHYS)
648 mp->mnt_iosize_max = MAXPHYS;
650 if (VOP_IOCTL(devvp, DIOCGPART, (caddr_t)&dpart, FREAD, proc0.p_ucred, td) != 0)
653 size = dpart.disklab->d_secsize;
657 if ((error = bread(devvp, SBLOCK, SBSIZE, &bp)) != 0)
659 fs = (struct fs *)bp->b_data;
660 if (fs->fs_magic != FS_MAGIC || fs->fs_bsize > MAXBSIZE ||
661 fs->fs_bsize < sizeof(struct fs)) {
662 error = EINVAL; /* XXX needs translation */
666 fs->fs_flags &= ~FS_UNCLEAN;
667 if (fs->fs_clean == 0) {
668 fs->fs_flags |= FS_UNCLEAN;
669 if (ronly || (mp->mnt_flag & MNT_FORCE)) {
671 "WARNING: %s was not properly dismounted\n",
675 "WARNING: R/W mount of %s denied. Filesystem is not clean - run fsck\n",
681 /* XXX updating 4.2 FFS superblocks trashes rotational layout tables */
682 if (fs->fs_postblformat == FS_42POSTBLFMT && !ronly) {
683 error = EROFS; /* needs translation */
686 ump = malloc(sizeof *ump, M_UFSMNT, M_WAITOK);
687 bzero((caddr_t)ump, sizeof *ump);
688 ump->um_malloctype = malloctype;
689 ump->um_i_effnlink_valid = 1;
690 ump->um_fs = malloc((u_long)fs->fs_sbsize, M_UFSMNT,
692 ump->um_blkatoff = ffs_blkatoff;
693 ump->um_truncate = ffs_truncate;
694 ump->um_update = ffs_update;
695 ump->um_valloc = ffs_valloc;
696 ump->um_vfree = ffs_vfree;
697 bcopy(bp->b_data, ump->um_fs, (uint)fs->fs_sbsize);
698 if (fs->fs_sbsize < SBSIZE)
699 bp->b_flags |= B_INVAL;
703 fs->fs_ronly = ronly;
704 size = fs->fs_cssize;
705 blks = howmany(size, fs->fs_fsize);
706 if (fs->fs_contigsumsize > 0)
707 size += fs->fs_ncg * sizeof(int32_t);
708 size += fs->fs_ncg * sizeof(uint8_t);
709 space = malloc((u_long)size, M_UFSMNT, M_WAITOK);
711 for (i = 0; i < blks; i += fs->fs_frag) {
713 if (i + fs->fs_frag > blks)
714 size = (blks - i) * fs->fs_fsize;
715 if ((error = bread(devvp, fsbtodb(fs, fs->fs_csaddr + i), size,
717 free(fs->fs_csp, M_UFSMNT);
720 bcopy(bp->b_data, space, (uint)size);
721 space = (char *)space + size;
725 if (fs->fs_contigsumsize > 0) {
726 fs->fs_maxcluster = lp = space;
727 for (i = 0; i < fs->fs_ncg; i++)
728 *lp++ = fs->fs_contigsumsize;
731 size = fs->fs_ncg * sizeof(uint8_t);
732 fs->fs_contigdirs = (uint8_t *)space;
733 bzero(fs->fs_contigdirs, size);
734 /* Compatibility for old filesystems XXX */
735 if (fs->fs_avgfilesize <= 0) /* XXX */
736 fs->fs_avgfilesize = AVFILESIZ; /* XXX */
737 if (fs->fs_avgfpdir <= 0) /* XXX */
738 fs->fs_avgfpdir = AFPDIR; /* XXX */
739 mp->mnt_data = (qaddr_t)ump;
740 mp->mnt_stat.f_fsid.val[0] = fs->fs_id[0];
741 mp->mnt_stat.f_fsid.val[1] = fs->fs_id[1];
742 if (fs->fs_id[0] == 0 || fs->fs_id[1] == 0 ||
743 vfs_getvfs(&mp->mnt_stat.f_fsid))
745 mp->mnt_maxsymlinklen = fs->fs_maxsymlinklen;
746 mp->mnt_flag |= MNT_LOCAL;
749 ump->um_devvp = devvp;
750 ump->um_nindir = fs->fs_nindir;
751 ump->um_bptrtodb = fs->fs_fsbtodb;
752 ump->um_seqinc = fs->fs_frag;
753 for (i = 0; i < MAXQUOTAS; i++)
754 ump->um_quotas[i] = NULLVP;
755 dev->si_mountpoint = mp;
759 * Set FS local "last mounted on" information (NULL pad)
761 copystr( mp->mnt_stat.f_mntonname, /* mount point*/
762 fs->fs_fsmnt, /* copy area*/
763 sizeof(fs->fs_fsmnt) - 1, /* max size*/
764 &strsize); /* real size*/
765 bzero( fs->fs_fsmnt + strsize, sizeof(fs->fs_fsmnt) - strsize);
767 if( mp->mnt_flag & MNT_ROOTFS) {
769 * Root mount; update timestamp in mount structure.
770 * this will be used by the common root mount code
771 * to update the system clock.
773 mp->mnt_time = fs->fs_time;
776 ump->um_savedmaxfilesize = fs->fs_maxfilesize; /* XXX */
777 maxfilesize = (uint64_t)0x40000000 * fs->fs_bsize - 1; /* XXX */
778 /* Enforce limit caused by vm object backing (32 bits vm_pindex_t). */
779 if (maxfilesize > (uint64_t)0x80000000u * PAGE_SIZE - 1)
780 maxfilesize = (uint64_t)0x80000000u * PAGE_SIZE - 1;
781 if (fs->fs_maxfilesize > maxfilesize) /* XXX */
782 fs->fs_maxfilesize = maxfilesize; /* XXX */
784 if ((fs->fs_flags & FS_DOSOFTDEP) &&
785 (error = softdep_mount(devvp, mp, fs)) != 0) {
786 free(fs->fs_csp, M_UFSMNT);
791 (void) ffs_sbupdate(ump, MNT_WAIT);
793 vfs_add_vnodeops(&mp->mnt_vn_ops, ffs_vnodeop_entries);
794 vfs_add_vnodeops(&mp->mnt_vn_spec_ops, ffs_specop_entries);
795 vfs_add_vnodeops(&mp->mnt_vn_fifo_ops, ffs_fifoop_entries);
799 dev->si_mountpoint = NULL;
802 VOP_CLOSE(devvp, ronly ? FREAD : FREAD|FWRITE, td);
804 free(ump->um_fs, M_UFSMNT);
806 mp->mnt_data = (qaddr_t)0;
812 * Sanity checks for old filesystems.
814 * XXX - goes away some day.
817 ffs_oldfscompat(struct fs *fs)
819 fs->fs_npsect = max(fs->fs_npsect, fs->fs_nsect); /* XXX */
820 fs->fs_interleave = max(fs->fs_interleave, 1); /* XXX */
821 if (fs->fs_postblformat == FS_42POSTBLFMT) /* XXX */
822 fs->fs_nrpos = 8; /* XXX */
823 if (fs->fs_inodefmt < FS_44INODEFMT) { /* XXX */
826 uint64_t sizepb = fs->fs_bsize; /* XXX */
828 fs->fs_maxfilesize = fs->fs_bsize * NDADDR - 1; /* XXX */
829 for (i = 0; i < NIADDR; i++) { /* XXX */
830 sizepb *= NINDIR(fs); /* XXX */
831 fs->fs_maxfilesize += sizepb; /* XXX */
834 fs->fs_maxfilesize = (u_quad_t) 1LL << 39;
835 fs->fs_qbmask = ~fs->fs_bmask; /* XXX */
836 fs->fs_qfmask = ~fs->fs_fmask; /* XXX */
842 * unmount system call
845 ffs_unmount(struct mount *mp, int mntflags, struct thread *td)
847 struct ufsmount *ump;
852 if (mntflags & MNT_FORCE) {
855 if (mp->mnt_flag & MNT_SOFTDEP) {
856 if ((error = softdep_flushfiles(mp, flags, td)) != 0)
859 if ((error = ffs_flushfiles(mp, flags, td)) != 0)
864 if (fs->fs_ronly == 0) {
865 fs->fs_clean = fs->fs_flags & FS_UNCLEAN ? 0 : 1;
866 error = ffs_sbupdate(ump, MNT_WAIT);
872 ump->um_devvp->v_rdev->si_mountpoint = NULL;
874 vinvalbuf(ump->um_devvp, V_SAVE, td, 0, 0);
875 error = VOP_CLOSE(ump->um_devvp, fs->fs_ronly ? FREAD : FREAD|FWRITE, td);
877 vrele(ump->um_devvp);
879 free(fs->fs_csp, M_UFSMNT);
882 mp->mnt_data = (qaddr_t)0;
883 mp->mnt_flag &= ~MNT_LOCAL;
888 * Flush out all the files in a filesystem.
891 ffs_flushfiles(struct mount *mp, int flags, struct thread *td)
893 struct ufsmount *ump;
898 if (mp->mnt_flag & MNT_QUOTA) {
900 error = vflush(mp, 0, SKIPSYSTEM|flags);
903 /* Find out how many quota files we have open. */
904 for (i = 0; i < MAXQUOTAS; i++) {
905 if (ump->um_quotas[i] == NULLVP)
910 * Here we fall through to vflush again to ensure
911 * that we have gotten rid of all the system vnodes.
916 * Flush all the files.
918 if ((error = vflush(mp, 0, flags)) != 0)
921 * Flush filesystem metadata.
923 vn_lock(ump->um_devvp, LK_EXCLUSIVE | LK_RETRY, td);
924 error = VOP_FSYNC(ump->um_devvp, MNT_WAIT, td);
925 VOP_UNLOCK(ump->um_devvp, 0, td);
930 * Get filesystem statistics.
933 ffs_statfs(struct mount *mp, struct statfs *sbp, struct thread *td)
935 struct ufsmount *ump;
940 if (fs->fs_magic != FS_MAGIC)
942 sbp->f_bsize = fs->fs_fsize;
943 sbp->f_iosize = fs->fs_bsize;
944 sbp->f_blocks = fs->fs_dsize;
945 sbp->f_bfree = fs->fs_cstotal.cs_nbfree * fs->fs_frag +
946 fs->fs_cstotal.cs_nffree;
947 sbp->f_bavail = freespace(fs, fs->fs_minfree);
948 sbp->f_files = fs->fs_ncg * fs->fs_ipg - ROOTINO;
949 sbp->f_ffree = fs->fs_cstotal.cs_nifree;
950 if (sbp != &mp->mnt_stat) {
951 sbp->f_type = mp->mnt_vfc->vfc_typenum;
952 bcopy((caddr_t)mp->mnt_stat.f_mntonname,
953 (caddr_t)&sbp->f_mntonname[0], MNAMELEN);
954 bcopy((caddr_t)mp->mnt_stat.f_mntfromname,
955 (caddr_t)&sbp->f_mntfromname[0], MNAMELEN);
961 * Go through the disk queues to initiate sandbagged IO;
962 * go through the inodes to write those that have been modified;
963 * initiate the writing of the super block if it has been modified.
965 * Note: we are always called with the filesystem marked `MPBUSY'.
969 static int ffs_sync_scan1(struct mount *mp, struct vnode *vp, void *data);
970 static int ffs_sync_scan2(struct mount *mp, struct vnode *vp, void *data);
973 ffs_sync(struct mount *mp, int waitfor, struct thread *td)
975 struct ufsmount *ump = VFSTOUFS(mp);
978 struct scaninfo scaninfo;
981 if (fs->fs_fmod != 0 && fs->fs_ronly != 0) { /* XXX */
982 printf("fs = %s\n", fs->fs_fsmnt);
983 panic("ffs_sync: rofs mod");
987 * Write back each (modified) inode.
989 scaninfo.allerror = 0;
991 scaninfo.waitfor = waitfor;
992 while (scaninfo.rescan) {
994 vmntvnodescan(mp, VMSC_GETVP|VMSC_NOWAIT,
995 ffs_sync_scan1, ffs_sync_scan2, &scaninfo);
999 * Force stale filesystem control information to be flushed.
1001 if (waitfor != MNT_LAZY) {
1002 if (ump->um_mountp->mnt_flag & MNT_SOFTDEP)
1003 waitfor = MNT_NOWAIT;
1004 vn_lock(ump->um_devvp, LK_EXCLUSIVE | LK_RETRY, td);
1005 if ((error = VOP_FSYNC(ump->um_devvp, waitfor, td)) != 0)
1006 scaninfo.allerror = error;
1007 VOP_UNLOCK(ump->um_devvp, 0, td);
1013 * Write back modified superblock.
1015 if (fs->fs_fmod != 0 && (error = ffs_sbupdate(ump, waitfor)) != 0)
1016 scaninfo.allerror = error;
1017 return (scaninfo.allerror);
1021 ffs_sync_scan1(struct mount *mp, struct vnode *vp, void *data)
1026 * Depend on the mount list's vnode lock to keep things stable
1027 * enough for a quick test. Since there might be hundreds of
1028 * thousands of vnodes, we cannot afford even a subroutine
1029 * call unless there's a good chance that we have work to do.
1032 /* Restart out whole search if this guy is locked
1033 * or is being reclaimed.
1035 if (vp->v_type == VNON || ((ip->i_flag &
1036 (IN_ACCESS | IN_CHANGE | IN_MODIFIED | IN_UPDATE)) == 0 &&
1037 TAILQ_EMPTY(&vp->v_dirtyblkhd))) {
1044 ffs_sync_scan2(struct mount *mp, struct vnode *vp, void *data)
1046 struct scaninfo *info = data;
1047 thread_t td = curthread; /* XXX */
1052 * We have to recheck after having obtained the vnode interlock.
1055 if (vp->v_type == VNON || ((ip->i_flag &
1056 (IN_ACCESS | IN_CHANGE | IN_MODIFIED | IN_UPDATE)) == 0 &&
1057 TAILQ_EMPTY(&vp->v_dirtyblkhd))) {
1060 if (vp->v_type != VCHR) {
1061 if ((error = VOP_FSYNC(vp, info->waitfor, td)) != 0)
1062 info->allerror = error;
1065 * We must reference the vp to prevent it from
1066 * getting ripped out from under UFS_UPDATE, since
1067 * we are not holding a vnode lock.
1069 /* UFS_UPDATE(vp, waitfor == MNT_WAIT); */
1076 * Look up a FFS dinode number to find its incore vnode, otherwise read it
1077 * in from disk. If it is in core, wait for the lock bit to clear, then
1078 * return the inode locked. Detection and handling of mount points must be
1079 * done by the calling routine.
1083 ffs_vget(struct mount *mp, ino_t ino, struct vnode **vpp)
1087 struct ufsmount *ump;
1096 if ((*vpp = ufs_ihashget(dev, ino)) != NULL) {
1101 * If this MALLOC() is performed after the getnewvnode()
1102 * it might block, leaving a vnode with a NULL v_data to be
1103 * found by ffs_sync() if a sync happens to fire right then,
1104 * which will cause a panic because ffs_sync() blindly
1105 * dereferences vp->v_data (as well it should).
1107 * XXX this may no longer be true since getnewvnode returns a
1108 * VX locked vnode now.
1110 MALLOC(ip, struct inode *, sizeof(struct inode),
1111 ump->um_malloctype, M_WAITOK);
1113 /* Allocate a new vnode/inode. */
1114 error = getnewvnode(VT_UFS, mp, mp->mnt_vn_ops, &vp,
1115 VLKTIMEOUT, LK_CANRECURSE);
1118 free(ip, ump->um_malloctype);
1121 bzero((caddr_t)ip, sizeof(struct inode));
1123 ip->i_fs = fs = ump->um_fs;
1129 for (i = 0; i < MAXQUOTAS; i++)
1130 ip->i_dquot[i] = NODQUOT;
1135 * Insert it into the inode hash table and check for a collision.
1136 * If a collision occurs, throw away the vnode and try again.
1138 if (ufs_ihashins(ip) != 0) {
1139 printf("debug: ufs ihashins collision, retrying inode %ld\n",
1140 (long)ip->i_number);
1142 free(ip, ump->um_malloctype);
1147 /* Read in the disk contents for the inode, copy into the inode. */
1148 error = bread(ump->um_devvp, fsbtodb(fs, ino_to_fsba(fs, ino)),
1149 (int)fs->fs_bsize, &bp);
1152 * The inode does not contain anything useful, so it would
1153 * be misleading to leave it on its hash chain. With mode
1154 * still zero, it will be unlinked and returned to the free
1162 ip->i_din = *((struct dinode *)bp->b_data + ino_to_fsbo(fs, ino));
1163 if (DOINGSOFTDEP(vp))
1164 softdep_load_inodeblock(ip);
1166 ip->i_effnlink = ip->i_nlink;
1170 * Initialize the vnode from the inode, check for aliases.
1171 * Note that the underlying vnode may have changed.
1173 error = ufs_vinit(mp, &vp);
1180 * Finish inode initialization now that aliasing has been resolved.
1182 ip->i_devvp = ump->um_devvp;
1185 * Set up a generation number for this inode if it does not
1186 * already have one. This should only happen on old filesystems.
1188 if (ip->i_gen == 0) {
1189 ip->i_gen = random() / 2 + 1;
1190 if ((vp->v_mount->mnt_flag & MNT_RDONLY) == 0)
1191 ip->i_flag |= IN_MODIFIED;
1194 * Ensure that uid and gid are correct. This is a temporary
1195 * fix until fsck has been changed to do the update.
1197 if (fs->fs_inodefmt < FS_44INODEFMT) { /* XXX */
1198 ip->i_uid = ip->i_din.di_ouid; /* XXX */
1199 ip->i_gid = ip->i_din.di_ogid; /* XXX */
1203 * return a VX locked and refd vnode (VX == same as normal vget()
1204 * vnode so we are ok)
1211 * File handle to vnode
1213 * Have to be really careful about stale file handles:
1214 * - check that the inode number is valid
1215 * - call ffs_vget() to get the locked inode
1216 * - check for an unallocated inode (i_mode == 0)
1217 * - check that the given client host has export rights and return
1218 * those rights via. exflagsp and credanonp
1221 ffs_fhtovp(struct mount *mp, struct fid *fhp, struct vnode **vpp)
1226 ufhp = (struct ufid *)fhp;
1227 fs = VFSTOUFS(mp)->um_fs;
1228 if (ufhp->ufid_ino < ROOTINO ||
1229 ufhp->ufid_ino >= fs->fs_ncg * fs->fs_ipg)
1231 return (ufs_fhtovp(mp, ufhp, vpp));
1235 * Vnode pointer to File handle
1239 ffs_vptofh(struct vnode *vp, struct fid *fhp)
1245 ufhp = (struct ufid *)fhp;
1246 ufhp->ufid_len = sizeof(struct ufid);
1247 ufhp->ufid_ino = ip->i_number;
1248 ufhp->ufid_gen = ip->i_gen;
1253 * Initialize the filesystem; just use ufs_init.
1256 ffs_init(struct vfsconf *vfsp)
1258 softdep_initialize();
1259 return (ufs_init(vfsp));
1263 * Write a superblock and associated information back to disk.
1266 ffs_sbupdate(struct ufsmount *mp, int waitfor)
1268 struct fs *dfs, *fs = mp->um_fs;
1272 int i, size, error, allerror = 0;
1275 * First write back the summary information.
1277 blks = howmany(fs->fs_cssize, fs->fs_fsize);
1279 for (i = 0; i < blks; i += fs->fs_frag) {
1280 size = fs->fs_bsize;
1281 if (i + fs->fs_frag > blks)
1282 size = (blks - i) * fs->fs_fsize;
1283 bp = getblk(mp->um_devvp, fsbtodb(fs, fs->fs_csaddr + i),
1285 bcopy(space, bp->b_data, (uint)size);
1286 space = (char *)space + size;
1287 if (waitfor != MNT_WAIT)
1289 else if ((error = bwrite(bp)) != 0)
1293 * Now write back the superblock itself. If any errors occurred
1294 * up to this point, then fail so that the superblock avoids
1295 * being written out as clean.
1299 bp = getblk(mp->um_devvp, SBLOCK, (int)fs->fs_sbsize, 0, 0);
1301 fs->fs_time = time_second;
1302 bcopy((caddr_t)fs, bp->b_data, (uint)fs->fs_sbsize);
1303 /* Restore compatibility to old filesystems. XXX */
1304 dfs = (struct fs *)bp->b_data; /* XXX */
1305 if (fs->fs_postblformat == FS_42POSTBLFMT) /* XXX */
1306 dfs->fs_nrpos = -1; /* XXX */
1307 if (fs->fs_inodefmt < FS_44INODEFMT) { /* XXX */
1308 int32_t *lp, tmp; /* XXX */
1310 lp = (int32_t *)&dfs->fs_qbmask; /* XXX */
1311 tmp = lp[4]; /* XXX */
1312 for (i = 4; i > 0; i--) /* XXX */
1313 lp[i] = lp[i-1]; /* XXX */
1314 lp[0] = tmp; /* XXX */
1316 dfs->fs_maxfilesize = mp->um_savedmaxfilesize; /* XXX */
1317 if (waitfor != MNT_WAIT)
1319 else if ((error = bwrite(bp)) != 0)