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 1.42 2006/05/06 16:20:19 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 thread *);
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 thread *td) /* process requesting mount */
138 struct ufs_args args;
139 struct ufsmount *ump = 0;
141 int flags, ronly = 0;
144 struct nlookupdata nd;
145 struct vnode *rootvp;
147 KKASSERT(td->td_proc);
148 cred = td->td_proc->p_ucred;
153 * Use NULL path to flag a root mount
158 * Mounting root filesystem
162 if ((error = bdevvp(rootdev, &rootvp))) {
163 printf("ffs_mountroot: can't find rootvp\n");
167 if( ( error = ffs_mountfs(rootvp, mp, M_FFSNODE)) != 0) {
168 /* 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);
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);
213 error = ffs_flushfiles(mp, flags);
217 if (!error && (mp->mnt_flag & MNT_RELOAD)) {
218 error = ffs_reload(mp, NULL);
223 if (ronly && (mp->mnt_kern_flag & MNTK_WANTRDWR)) {
225 * If upgrade to read-write by non-root, then verify
226 * that user has necessary permissions on the device.
228 if (cred->cr_uid != 0) {
229 vn_lock(devvp, LK_EXCLUSIVE | LK_RETRY);
230 if ((error = VOP_ACCESS(devvp, VREAD | VWRITE,
232 VOP_UNLOCK(devvp, 0);
235 VOP_UNLOCK(devvp, 0);
238 fs->fs_flags &= ~FS_UNCLEAN;
239 if (fs->fs_clean == 0) {
240 fs->fs_flags |= FS_UNCLEAN;
241 if (mp->mnt_flag & MNT_FORCE) {
243 "WARNING: %s was not properly dismounted\n",
247 "WARNING: R/W mount of %s denied. Filesystem is not clean - run fsck\n",
254 /* check to see if we need to start softdep */
255 if (fs->fs_flags & FS_DOSOFTDEP) {
256 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);
309 if ((error = VOP_ACCESS(devvp, accessmode, cred)) != 0) {
313 VOP_UNLOCK(devvp, 0);
316 if (mp->mnt_flag & MNT_UPDATE) {
318 * UPDATE - make sure the resolved vnode represents the same
319 * device. Note that devvp->v_rdev may be NULL since we
320 * haven't opened it, so compare udev instead.
322 * Our current open/writecount state is associated with
323 * um_devvp, so continue using um_devvp and throw away devvp.
325 if (devvp != ump->um_devvp) {
326 if (devvp->v_udev == ump->um_devvp->v_udev) {
328 devvp = ump->um_devvp;
330 printf("cannot update mount, udev does"
331 " not match %08x vs %08x\n",
332 devvp->v_udev, ump->um_devvp->v_udev);
333 error = EINVAL; /* needs translation */
339 * Update device name only on success
342 /* Save "mounted from" info for mount point (NULL pad)*/
343 copyinstr( args.fspec,
344 mp->mnt_stat.f_mntfromname,
347 bzero( mp->mnt_stat.f_mntfromname + size, MNAMELEN - size);
356 /* Save "mounted from" info for mount point (NULL pad)*/
357 copyinstr( args.fspec, /* device name*/
358 mp->mnt_stat.f_mntfromname, /* save area*/
359 MNAMELEN - 1, /* max size*/
360 &size); /* real size*/
361 bzero( mp->mnt_stat.f_mntfromname + size, MNAMELEN - size);
363 error = ffs_mountfs(devvp, mp, M_FFSNODE);
371 * Initialize FS stat information in mount struct; uses
372 * mp->mnt_stat.f_mntfromname.
374 * This code is common to root and non-root mounts
376 (void)VFS_STATFS(mp, &mp->mnt_stat, td);
381 error_2: /* error with devvp held*/
383 /* release devvp before failing*/
386 error_1: /* no state to back out*/
389 if (!error && path && (mp->mnt_flag & MNT_UPDATE)) {
390 /* Update clean flag after changing read-onlyness. */
392 if (ronly != fs->fs_ronly) {
393 fs->fs_ronly = ronly;
394 fs->fs_clean = ronly &&
395 (fs->fs_flags & FS_UNCLEAN) == 0 ? 1 : 0;
398 * The device must be re-opened as appropriate or
399 * the device close at unmount time will panic.
401 vn_lock(devvp, LK_EXCLUSIVE | LK_RETRY);
403 VOP_OPEN(devvp, FREAD, FSCRED, NULL);
404 VOP_CLOSE(devvp, FREAD|FWRITE);
406 VOP_OPEN(devvp, FREAD|FWRITE, FSCRED, NULL);
407 VOP_CLOSE(devvp, FREAD);
409 VOP_UNLOCK(devvp, 0);
410 ffs_sbupdate(ump, MNT_WAIT);
417 * Reload all incore data for a filesystem (used after running fsck on
418 * the root filesystem and finding things to fix). The filesystem must
419 * be mounted read-only.
421 * Things to do to update the mount:
422 * 1) invalidate all cached meta-data.
423 * 2) re-read superblock from disk.
424 * 3) re-read summary information from disk.
425 * 4) invalidate all inactive vnodes.
426 * 5) invalidate all cached file data.
427 * 6) re-read inode data for all active vnodes.
430 static int ffs_reload_scan2(struct mount *mp, struct vnode *vp, void *data);
441 ffs_reload(struct mount *mp, struct ucred *cred)
446 struct fs *fs, *newfs;
447 struct partinfo dpart;
449 int i, blks, size, error;
450 struct scaninfo scaninfo;
453 if ((mp->mnt_flag & MNT_RDONLY) == 0)
456 * Step 1: invalidate all cached meta-data.
458 devvp = VFSTOUFS(mp)->um_devvp;
459 vn_lock(devvp, LK_EXCLUSIVE | LK_RETRY);
460 error = vinvalbuf(devvp, 0, 0, 0);
461 VOP_UNLOCK(devvp, 0);
463 panic("ffs_reload: dirty1");
468 * The backing device must be VMIO-capable because we use getblk().
469 * NOTE: the MFS driver now returns a VMIO-enabled descriptor.
471 if (devvp->v_object == NULL)
472 panic("ffs_reload: devvp has no VM object!");
475 * Step 2: re-read superblock from disk.
477 if (VOP_IOCTL(devvp, DIOCGPART, (caddr_t)&dpart, FREAD, NOCRED) != 0)
480 size = dpart.disklab->d_secsize;
481 if ((error = bread(devvp, SBOFF, SBSIZE, &bp)) != 0) {
485 newfs = (struct fs *)bp->b_data;
486 if (newfs->fs_magic != FS_MAGIC || newfs->fs_bsize > MAXBSIZE ||
487 newfs->fs_bsize < sizeof(struct fs)) {
489 return (EIO); /* XXX needs translation */
491 fs = VFSTOUFS(mp)->um_fs;
493 * Copy pointer fields back into superblock before copying in XXX
494 * new superblock. These should really be in the ufsmount. XXX
495 * Note that important parameters (eg fs_ncg) are unchanged.
497 newfs->fs_csp = fs->fs_csp;
498 newfs->fs_maxcluster = fs->fs_maxcluster;
499 newfs->fs_contigdirs = fs->fs_contigdirs;
500 /* The filesystem is still read-only. */
502 bcopy(newfs, fs, (uint)fs->fs_sbsize);
503 if (fs->fs_sbsize < SBSIZE)
504 bp->b_flags |= B_INVAL;
506 mp->mnt_maxsymlinklen = fs->fs_maxsymlinklen;
508 /* An old fsck may have zeroed these fields, so recheck them. */
509 if (fs->fs_avgfilesize <= 0) /* XXX */
510 fs->fs_avgfilesize = AVFILESIZ; /* XXX */
511 if (fs->fs_avgfpdir <= 0) /* XXX */
512 fs->fs_avgfpdir = AFPDIR; /* XXX */
515 * Step 3: re-read summary information from disk.
517 blks = howmany(fs->fs_cssize, fs->fs_fsize);
519 for (i = 0; i < blks; i += fs->fs_frag) {
521 if (i + fs->fs_frag > blks)
522 size = (blks - i) * fs->fs_fsize;
523 error = bread(devvp, fsbtodoff(fs, fs->fs_csaddr + i), size, &bp);
528 bcopy(bp->b_data, space, (uint)size);
529 space = (char *)space + size;
533 * We no longer know anything about clusters per cylinder group.
535 if (fs->fs_contigsumsize > 0) {
536 lp = fs->fs_maxcluster;
537 for (i = 0; i < fs->fs_ncg; i++)
538 *lp++ = fs->fs_contigsumsize;
543 scaninfo.devvp = devvp;
544 while (error == 0 && scaninfo.rescan) {
546 error = vmntvnodescan(mp, VMSC_GETVX,
547 NULL, ffs_reload_scan2, &scaninfo);
553 ffs_reload_scan2(struct mount *mp, struct vnode *vp, void *data)
555 struct scaninfo *info = data;
566 if (vinvalbuf(vp, 0, 0, 0))
567 panic("ffs_reload: dirty2");
569 * Step 6: re-read inode data for all active vnodes.
572 error = bread(info->devvp,
573 fsbtodoff(info->fs, ino_to_fsba(info->fs, ip->i_number)),
574 (int)info->fs->fs_bsize, &bp);
579 ip->i_din = *((struct ufs1_dinode *)bp->b_data +
580 ino_to_fsbo(info->fs, ip->i_number));
581 ip->i_effnlink = ip->i_nlink;
587 * Common code for mount and mountroot
590 ffs_mountfs(struct vnode *devvp, struct mount *mp, struct malloc_type *mtype)
592 struct ufsmount *ump;
596 struct partinfo dpart;
598 int error, i, blks, size, ronly;
600 uint64_t maxfilesize; /* XXX */
603 * Disallow multiple mounts of the same device.
604 * Disallow mounting of a device that is currently in use
605 * Flush out any old buffers remaining from a previous use.
607 error = vfs_mountedon(devvp);
610 if (count_udev(devvp->v_udev) > 0)
612 vn_lock(devvp, LK_EXCLUSIVE | LK_RETRY);
613 error = vinvalbuf(devvp, V_SAVE, 0, 0);
614 VOP_UNLOCK(devvp, 0);
618 ronly = (mp->mnt_flag & MNT_RDONLY) != 0;
619 vn_lock(devvp, LK_EXCLUSIVE | LK_RETRY);
620 error = VOP_OPEN(devvp, ronly ? FREAD : FREAD|FWRITE, FSCRED, NULL);
621 VOP_UNLOCK(devvp, 0);
625 if (dev->si_iosize_max != 0)
626 mp->mnt_iosize_max = dev->si_iosize_max;
627 if (mp->mnt_iosize_max > MAXPHYS)
628 mp->mnt_iosize_max = MAXPHYS;
631 * The backing device must be VMIO-capable because we use getblk().
632 * NOTE: the MFS driver now returns a VMIO-enabled descriptor.
633 * The VOP_OPEN() call above should have associated a VM object
636 if (devvp->v_object == NULL)
637 panic("ffs_reload: devvp has no VM object!");
639 if (VOP_IOCTL(devvp, DIOCGPART, (caddr_t)&dpart, FREAD, proc0.p_ucred) != 0)
642 size = dpart.disklab->d_secsize;
646 if ((error = bread(devvp, SBOFF, SBSIZE, &bp)) != 0)
648 fs = (struct fs *)bp->b_data;
649 if (fs->fs_magic != FS_MAGIC || fs->fs_bsize > MAXBSIZE ||
650 fs->fs_bsize < sizeof(struct fs)) {
651 error = EINVAL; /* XXX needs translation */
655 fs->fs_flags &= ~FS_UNCLEAN;
656 if (fs->fs_clean == 0) {
657 fs->fs_flags |= FS_UNCLEAN;
658 if (ronly || (mp->mnt_flag & MNT_FORCE)) {
660 "WARNING: %s was not properly dismounted\n",
664 "WARNING: R/W mount of %s denied. Filesystem is not clean - run fsck\n",
670 /* XXX updating 4.2 FFS superblocks trashes rotational layout tables */
671 if (fs->fs_postblformat == FS_42POSTBLFMT && !ronly) {
672 error = EROFS; /* needs translation */
675 ump = malloc(sizeof *ump, M_UFSMNT, M_WAITOK);
676 bzero((caddr_t)ump, sizeof *ump);
677 ump->um_malloctype = mtype;
678 ump->um_i_effnlink_valid = 1;
679 ump->um_fs = malloc((u_long)fs->fs_sbsize, M_UFSMNT,
681 ump->um_blkatoff = ffs_blkatoff;
682 ump->um_truncate = ffs_truncate;
683 ump->um_update = ffs_update;
684 ump->um_valloc = ffs_valloc;
685 ump->um_vfree = ffs_vfree;
686 bcopy(bp->b_data, ump->um_fs, (uint)fs->fs_sbsize);
687 if (fs->fs_sbsize < SBSIZE)
688 bp->b_flags |= B_INVAL;
692 fs->fs_ronly = ronly;
693 size = fs->fs_cssize;
694 blks = howmany(size, fs->fs_fsize);
695 if (fs->fs_contigsumsize > 0)
696 size += fs->fs_ncg * sizeof(int32_t);
697 size += fs->fs_ncg * sizeof(uint8_t);
698 space = malloc((u_long)size, M_UFSMNT, M_WAITOK);
700 for (i = 0; i < blks; i += fs->fs_frag) {
702 if (i + fs->fs_frag > blks)
703 size = (blks - i) * fs->fs_fsize;
704 if ((error = bread(devvp, fsbtodoff(fs, fs->fs_csaddr + i),
706 free(fs->fs_csp, M_UFSMNT);
709 bcopy(bp->b_data, space, (uint)size);
710 space = (char *)space + size;
714 if (fs->fs_contigsumsize > 0) {
715 fs->fs_maxcluster = lp = space;
716 for (i = 0; i < fs->fs_ncg; i++)
717 *lp++ = fs->fs_contigsumsize;
720 size = fs->fs_ncg * sizeof(uint8_t);
721 fs->fs_contigdirs = (uint8_t *)space;
722 bzero(fs->fs_contigdirs, size);
723 /* Compatibility for old filesystems XXX */
724 if (fs->fs_avgfilesize <= 0) /* XXX */
725 fs->fs_avgfilesize = AVFILESIZ; /* XXX */
726 if (fs->fs_avgfpdir <= 0) /* XXX */
727 fs->fs_avgfpdir = AFPDIR; /* XXX */
728 mp->mnt_data = (qaddr_t)ump;
729 mp->mnt_stat.f_fsid.val[0] = fs->fs_id[0];
730 mp->mnt_stat.f_fsid.val[1] = fs->fs_id[1];
731 if (fs->fs_id[0] == 0 || fs->fs_id[1] == 0 ||
732 vfs_getvfs(&mp->mnt_stat.f_fsid))
734 mp->mnt_maxsymlinklen = fs->fs_maxsymlinklen;
735 mp->mnt_flag |= MNT_LOCAL;
738 ump->um_devvp = devvp;
739 ump->um_nindir = fs->fs_nindir;
740 ump->um_bptrtodb = fs->fs_fsbtodb;
741 ump->um_seqinc = fs->fs_frag;
742 for (i = 0; i < MAXQUOTAS; i++)
743 ump->um_quotas[i] = NULLVP;
744 dev->si_mountpoint = mp;
747 if( mp->mnt_flag & MNT_ROOTFS) {
749 * Root mount; update timestamp in mount structure.
750 * this will be used by the common root mount code
751 * to update the system clock.
753 mp->mnt_time = fs->fs_time;
756 ump->um_savedmaxfilesize = fs->fs_maxfilesize; /* XXX */
757 maxfilesize = (uint64_t)0x40000000 * fs->fs_bsize - 1; /* XXX */
758 /* Enforce limit caused by vm object backing (32 bits vm_pindex_t). */
759 if (maxfilesize > (uint64_t)0x80000000u * PAGE_SIZE - 1)
760 maxfilesize = (uint64_t)0x80000000u * PAGE_SIZE - 1;
761 if (fs->fs_maxfilesize > maxfilesize) /* XXX */
762 fs->fs_maxfilesize = maxfilesize; /* XXX */
764 if ((fs->fs_flags & FS_DOSOFTDEP) &&
765 (error = softdep_mount(devvp, mp, fs)) != 0) {
766 free(fs->fs_csp, M_UFSMNT);
771 (void) ffs_sbupdate(ump, MNT_WAIT);
773 vfs_add_vnodeops(mp, &mp->mnt_vn_norm_ops, ffs_vnodeop_entries, VVF_SUPPORTS_FSMID);
774 vfs_add_vnodeops(mp, &mp->mnt_vn_spec_ops, ffs_specop_entries, VVF_SUPPORTS_FSMID);
775 vfs_add_vnodeops(mp, &mp->mnt_vn_fifo_ops, ffs_fifoop_entries, VVF_SUPPORTS_FSMID);
779 dev->si_mountpoint = NULL;
782 VOP_CLOSE(devvp, ronly ? FREAD : FREAD|FWRITE);
784 free(ump->um_fs, M_UFSMNT);
786 mp->mnt_data = (qaddr_t)0;
792 * Sanity checks for old filesystems.
794 * XXX - goes away some day.
797 ffs_oldfscompat(struct fs *fs)
799 fs->fs_npsect = max(fs->fs_npsect, fs->fs_nsect); /* XXX */
800 fs->fs_interleave = max(fs->fs_interleave, 1); /* XXX */
801 if (fs->fs_postblformat == FS_42POSTBLFMT) /* XXX */
802 fs->fs_nrpos = 8; /* XXX */
803 if (fs->fs_inodefmt < FS_44INODEFMT) { /* XXX */
806 uint64_t sizepb = fs->fs_bsize; /* XXX */
808 fs->fs_maxfilesize = fs->fs_bsize * NDADDR - 1; /* XXX */
809 for (i = 0; i < NIADDR; i++) { /* XXX */
810 sizepb *= NINDIR(fs); /* XXX */
811 fs->fs_maxfilesize += sizepb; /* XXX */
814 fs->fs_maxfilesize = (u_quad_t) 1LL << 39;
815 fs->fs_qbmask = ~fs->fs_bmask; /* XXX */
816 fs->fs_qfmask = ~fs->fs_fmask; /* XXX */
822 * unmount system call
825 ffs_unmount(struct mount *mp, int mntflags, struct thread *td)
827 struct ufsmount *ump;
832 if (mntflags & MNT_FORCE) {
835 if (mp->mnt_flag & MNT_SOFTDEP) {
836 if ((error = softdep_flushfiles(mp, flags)) != 0)
839 if ((error = ffs_flushfiles(mp, flags)) != 0)
844 if (fs->fs_ronly == 0) {
845 fs->fs_clean = fs->fs_flags & FS_UNCLEAN ? 0 : 1;
846 error = ffs_sbupdate(ump, MNT_WAIT);
852 ump->um_devvp->v_rdev->si_mountpoint = NULL;
854 vinvalbuf(ump->um_devvp, V_SAVE, 0, 0);
855 error = VOP_CLOSE(ump->um_devvp, fs->fs_ronly ? FREAD : FREAD|FWRITE);
857 vrele(ump->um_devvp);
859 free(fs->fs_csp, M_UFSMNT);
862 mp->mnt_data = (qaddr_t)0;
863 mp->mnt_flag &= ~MNT_LOCAL;
868 * Flush out all the files in a filesystem.
871 ffs_flushfiles(struct mount *mp, int flags)
873 struct ufsmount *ump;
878 if (mp->mnt_flag & MNT_QUOTA) {
880 error = vflush(mp, 0, SKIPSYSTEM|flags);
883 /* Find out how many quota files we have open. */
884 for (i = 0; i < MAXQUOTAS; i++) {
885 if (ump->um_quotas[i] == NULLVP)
890 * Here we fall through to vflush again to ensure
891 * that we have gotten rid of all the system vnodes.
896 * Flush all the files.
898 if ((error = vflush(mp, 0, flags)) != 0)
901 * Flush filesystem metadata.
903 vn_lock(ump->um_devvp, LK_EXCLUSIVE | LK_RETRY);
904 error = VOP_FSYNC(ump->um_devvp, MNT_WAIT);
905 VOP_UNLOCK(ump->um_devvp, 0);
910 * Get filesystem statistics.
913 ffs_statfs(struct mount *mp, struct statfs *sbp, struct thread *td)
915 struct ufsmount *ump;
920 if (fs->fs_magic != FS_MAGIC)
922 sbp->f_bsize = fs->fs_fsize;
923 sbp->f_iosize = fs->fs_bsize;
924 sbp->f_blocks = fs->fs_dsize;
925 sbp->f_bfree = fs->fs_cstotal.cs_nbfree * fs->fs_frag +
926 fs->fs_cstotal.cs_nffree;
927 sbp->f_bavail = freespace(fs, fs->fs_minfree);
928 sbp->f_files = fs->fs_ncg * fs->fs_ipg - ROOTINO;
929 sbp->f_ffree = fs->fs_cstotal.cs_nifree;
930 if (sbp != &mp->mnt_stat) {
931 sbp->f_type = mp->mnt_vfc->vfc_typenum;
932 bcopy((caddr_t)mp->mnt_stat.f_mntfromname,
933 (caddr_t)&sbp->f_mntfromname[0], MNAMELEN);
939 * Go through the disk queues to initiate sandbagged IO;
940 * go through the inodes to write those that have been modified;
941 * initiate the writing of the super block if it has been modified.
943 * Note: we are always called with the filesystem marked `MPBUSY'.
947 static int ffs_sync_scan1(struct mount *mp, struct vnode *vp, void *data);
948 static int ffs_sync_scan2(struct mount *mp, struct vnode *vp, void *data);
951 ffs_sync(struct mount *mp, int waitfor)
953 struct ufsmount *ump = VFSTOUFS(mp);
956 struct scaninfo scaninfo;
959 if (fs->fs_fmod != 0 && fs->fs_ronly != 0) { /* XXX */
960 printf("fs = %s\n", fs->fs_fsmnt);
961 panic("ffs_sync: rofs mod");
965 * Write back each (modified) inode.
967 scaninfo.allerror = 0;
969 scaninfo.waitfor = waitfor;
970 while (scaninfo.rescan) {
972 vmntvnodescan(mp, VMSC_GETVP|VMSC_NOWAIT,
973 ffs_sync_scan1, ffs_sync_scan2, &scaninfo);
977 * Force stale filesystem control information to be flushed.
979 if (waitfor != MNT_LAZY) {
980 if (ump->um_mountp->mnt_flag & MNT_SOFTDEP)
981 waitfor = MNT_NOWAIT;
982 vn_lock(ump->um_devvp, LK_EXCLUSIVE | LK_RETRY);
983 if ((error = VOP_FSYNC(ump->um_devvp, waitfor)) != 0)
984 scaninfo.allerror = error;
985 VOP_UNLOCK(ump->um_devvp, 0);
991 * Write back modified superblock.
993 if (fs->fs_fmod != 0 && (error = ffs_sbupdate(ump, waitfor)) != 0)
994 scaninfo.allerror = error;
995 return (scaninfo.allerror);
999 ffs_sync_scan1(struct mount *mp, struct vnode *vp, void *data)
1004 * Depend on the mount list's vnode lock to keep things stable
1005 * enough for a quick test. Since there might be hundreds of
1006 * thousands of vnodes, we cannot afford even a subroutine
1007 * call unless there's a good chance that we have work to do.
1010 /* Restart out whole search if this guy is locked
1011 * or is being reclaimed.
1013 if (vp->v_type == VNON || ((ip->i_flag &
1014 (IN_ACCESS | IN_CHANGE | IN_MODIFIED | IN_UPDATE)) == 0 &&
1015 RB_EMPTY(&vp->v_rbdirty_tree))) {
1022 ffs_sync_scan2(struct mount *mp, struct vnode *vp, void *data)
1024 struct scaninfo *info = data;
1029 * We have to recheck after having obtained the vnode interlock.
1032 if (vp->v_type == VNON || ((ip->i_flag &
1033 (IN_ACCESS | IN_CHANGE | IN_MODIFIED | IN_UPDATE)) == 0 &&
1034 RB_EMPTY(&vp->v_rbdirty_tree))) {
1037 if (vp->v_type != VCHR) {
1038 if ((error = VOP_FSYNC(vp, info->waitfor)) != 0)
1039 info->allerror = error;
1042 * We must reference the vp to prevent it from
1043 * getting ripped out from under UFS_UPDATE, since
1044 * we are not holding a vnode lock.
1046 /* UFS_UPDATE(vp, waitfor == MNT_WAIT); */
1053 * Look up a FFS dinode number to find its incore vnode, otherwise read it
1054 * in from disk. If it is in core, wait for the lock bit to clear, then
1055 * return the inode locked. Detection and handling of mount points must be
1056 * done by the calling routine.
1060 ffs_vget(struct mount *mp, ino_t ino, struct vnode **vpp)
1064 struct ufsmount *ump;
1073 if ((*vpp = ufs_ihashget(dev, ino)) != NULL) {
1078 * If this MALLOC() is performed after the getnewvnode()
1079 * it might block, leaving a vnode with a NULL v_data to be
1080 * found by ffs_sync() if a sync happens to fire right then,
1081 * which will cause a panic because ffs_sync() blindly
1082 * dereferences vp->v_data (as well it should).
1084 * XXX this may no longer be true since getnewvnode returns a
1085 * VX locked vnode now.
1087 MALLOC(ip, struct inode *, sizeof(struct inode),
1088 ump->um_malloctype, M_WAITOK);
1090 /* Allocate a new vnode/inode. */
1091 error = getnewvnode(VT_UFS, mp, &vp, VLKTIMEOUT, LK_CANRECURSE);
1094 free(ip, ump->um_malloctype);
1097 bzero((caddr_t)ip, sizeof(struct inode));
1099 ip->i_fs = fs = ump->um_fs;
1105 for (i = 0; i < MAXQUOTAS; i++)
1106 ip->i_dquot[i] = NODQUOT;
1111 * Insert it into the inode hash table and check for a collision.
1112 * If a collision occurs, throw away the vnode and try again.
1114 if (ufs_ihashins(ip) != 0) {
1115 printf("debug: ufs ihashins collision, retrying inode %ld\n",
1116 (long)ip->i_number);
1118 free(ip, ump->um_malloctype);
1123 /* Read in the disk contents for the inode, copy into the inode. */
1124 error = bread(ump->um_devvp, fsbtodoff(fs, ino_to_fsba(fs, ino)),
1125 (int)fs->fs_bsize, &bp);
1128 * The inode does not contain anything useful, so it would
1129 * be misleading to leave it on its hash chain. With mode
1130 * still zero, it will be unlinked and returned to the free
1138 ip->i_din = *((struct ufs1_dinode *)bp->b_data + ino_to_fsbo(fs, ino));
1139 if (DOINGSOFTDEP(vp))
1140 softdep_load_inodeblock(ip);
1142 ip->i_effnlink = ip->i_nlink;
1146 * Initialize the vnode from the inode, check for aliases.
1147 * Note that the underlying vnode may have changed.
1149 error = ufs_vinit(mp, &vp);
1156 * Finish inode initialization now that aliasing has been resolved.
1158 ip->i_devvp = ump->um_devvp;
1161 * Set up a generation number for this inode if it does not
1162 * already have one. This should only happen on old filesystems.
1164 if (ip->i_gen == 0) {
1165 ip->i_gen = random() / 2 + 1;
1166 if ((vp->v_mount->mnt_flag & MNT_RDONLY) == 0)
1167 ip->i_flag |= IN_MODIFIED;
1170 * Ensure that uid and gid are correct. This is a temporary
1171 * fix until fsck has been changed to do the update.
1173 if (fs->fs_inodefmt < FS_44INODEFMT) { /* XXX */
1174 ip->i_uid = ip->i_din.di_ouid; /* XXX */
1175 ip->i_gid = ip->i_din.di_ogid; /* XXX */
1179 * return a VX locked and refd vnode (VX == same as normal vget()
1180 * vnode so we are ok)
1187 * File handle to vnode
1189 * Have to be really careful about stale file handles:
1190 * - check that the inode number is valid
1191 * - call ffs_vget() to get the locked inode
1192 * - check for an unallocated inode (i_mode == 0)
1193 * - check that the given client host has export rights and return
1194 * those rights via. exflagsp and credanonp
1197 ffs_fhtovp(struct mount *mp, struct fid *fhp, struct vnode **vpp)
1202 ufhp = (struct ufid *)fhp;
1203 fs = VFSTOUFS(mp)->um_fs;
1204 if (ufhp->ufid_ino < ROOTINO ||
1205 ufhp->ufid_ino >= fs->fs_ncg * fs->fs_ipg)
1207 return (ufs_fhtovp(mp, ufhp, vpp));
1211 * Vnode pointer to File handle
1215 ffs_vptofh(struct vnode *vp, struct fid *fhp)
1221 ufhp = (struct ufid *)fhp;
1222 ufhp->ufid_len = sizeof(struct ufid);
1223 ufhp->ufid_ino = ip->i_number;
1224 ufhp->ufid_gen = ip->i_gen;
1229 * Initialize the filesystem; just use ufs_init.
1232 ffs_init(struct vfsconf *vfsp)
1234 softdep_initialize();
1235 return (ufs_init(vfsp));
1239 * Write a superblock and associated information back to disk.
1242 ffs_sbupdate(struct ufsmount *mp, int waitfor)
1244 struct fs *dfs, *fs = mp->um_fs;
1248 int i, size, error, allerror = 0;
1251 * First write back the summary information.
1253 * NOTE: the getblk is relative to the device vnode so bio1
1254 * contains the device block number.
1256 blks = howmany(fs->fs_cssize, fs->fs_fsize);
1258 for (i = 0; i < blks; i += fs->fs_frag) {
1259 size = fs->fs_bsize;
1260 if (i + fs->fs_frag > blks)
1261 size = (blks - i) * fs->fs_fsize;
1262 bp = getblk(mp->um_devvp, fsbtodoff(fs, fs->fs_csaddr + i),
1264 bcopy(space, bp->b_data, (uint)size);
1265 space = (char *)space + size;
1266 if (waitfor != MNT_WAIT)
1268 else if ((error = bwrite(bp)) != 0)
1272 * Now write back the superblock itself. If any errors occurred
1273 * up to this point, then fail so that the superblock avoids
1274 * being written out as clean.
1278 bp = getblk(mp->um_devvp, SBOFF, (int)fs->fs_sbsize, 0, 0);
1280 fs->fs_time = time_second;
1281 bcopy((caddr_t)fs, bp->b_data, (uint)fs->fs_sbsize);
1282 /* Restore compatibility to old filesystems. XXX */
1283 dfs = (struct fs *)bp->b_data; /* XXX */
1284 if (fs->fs_postblformat == FS_42POSTBLFMT) /* XXX */
1285 dfs->fs_nrpos = -1; /* XXX */
1286 if (fs->fs_inodefmt < FS_44INODEFMT) { /* XXX */
1287 int32_t *lp, tmp; /* XXX */
1289 lp = (int32_t *)&dfs->fs_qbmask; /* XXX */
1290 tmp = lp[4]; /* XXX */
1291 for (i = 4; i > 0; i--) /* XXX */
1292 lp[i] = lp[i-1]; /* XXX */
1293 lp[0] = tmp; /* XXX */
1295 dfs->fs_maxfilesize = mp->um_savedmaxfilesize; /* XXX */
1296 if (waitfor != MNT_WAIT)
1298 else if ((error = bwrite(bp)) != 0)