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 1.53 2007/04/20 22:20:12 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 vop_ops ffs_vnode_vops;
91 extern struct vop_ops ffs_spec_vops;
92 extern struct vop_ops ffs_fifo_vops;
97 * Called when mounting local physical media
101 * mp mount point structure
102 * path NULL (flag for root mount!!!)
104 * p process (user credentials check [statfs])
107 * mp mount point structure
108 * path path to mount point
109 * data pointer to argument struct in user space
110 * p process (user credentials check)
113 * !0 error number (errno.h)
118 * mount point is locked
120 * mount point is locked
123 * A NULL path can be used for a flag since the mount
124 * system call will fail with EFAULT in copyinstr in
125 * nlookup() if it is a genuine NULL from the user.
128 ffs_mount(struct mount *mp, /* mount struct pointer */
129 char *path, /* path to mount point */
130 caddr_t data, /* arguments to FS specific mount */
131 struct ucred *cred) /* process requesting mount */
137 struct ufs_args args;
138 struct ufsmount *ump = 0;
140 int flags, ronly = 0;
142 struct nlookupdata nd;
143 struct vnode *rootvp;
149 * Use NULL path to flag a root mount
154 * Mounting root filesystem
158 if ((error = bdevvp(rootdev, &rootvp))) {
159 kprintf("ffs_mountroot: can't find rootvp\n");
163 if( ( error = ffs_mountfs(rootvp, mp, M_FFSNODE)) != 0) {
164 /* fs specific cleanup (if any)*/
169 goto dostatfs; /* success*/
175 * Mounting non-root filesystem or updating a filesystem
179 /* copy in user arguments*/
180 error = copyin(data, (caddr_t)&args, sizeof (struct ufs_args));
182 goto error_1; /* can't get arguments*/
185 * If updating, check whether changing from read-only to
186 * read/write; if there is no device name, that's all we do.
188 if (mp->mnt_flag & MNT_UPDATE) {
191 devvp = ump->um_devvp;
193 ronly = fs->fs_ronly; /* MNT_RELOAD might change this */
194 if (ronly == 0 && (mp->mnt_flag & MNT_RDONLY)) {
196 * Flush any dirty data.
198 VFS_SYNC(mp, MNT_WAIT);
200 * Check for and optionally get rid of files open
204 if (mp->mnt_flag & MNT_FORCE)
206 if (mp->mnt_flag & MNT_SOFTDEP) {
207 error = softdep_flushfiles(mp, flags);
209 error = ffs_flushfiles(mp, flags);
213 if (!error && (mp->mnt_flag & MNT_RELOAD)) {
214 error = ffs_reload(mp, NULL);
219 if (ronly && (mp->mnt_kern_flag & MNTK_WANTRDWR)) {
221 * If upgrade to read-write by non-root, then verify
222 * that user has necessary permissions on the device.
224 if (cred->cr_uid != 0) {
225 vn_lock(devvp, LK_EXCLUSIVE | LK_RETRY);
226 if ((error = VOP_ACCESS(devvp, VREAD | VWRITE,
234 fs->fs_flags &= ~FS_UNCLEAN;
235 if (fs->fs_clean == 0) {
236 fs->fs_flags |= FS_UNCLEAN;
237 if (mp->mnt_flag & MNT_FORCE) {
239 "WARNING: %s was not properly dismounted\n",
243 "WARNING: R/W mount of %s denied. Filesystem is not clean - run fsck\n",
250 /* check to see if we need to start softdep */
251 if (fs->fs_flags & FS_DOSOFTDEP) {
252 error = softdep_mount(devvp, mp, fs);
259 * Soft updates is incompatible with "async",
260 * so if we are doing softupdates stop the user
261 * from setting the async flag in an update.
262 * Softdep_mount() clears it in an initial mount
265 if (mp->mnt_flag & MNT_SOFTDEP) {
266 mp->mnt_flag &= ~MNT_ASYNC;
268 /* if not updating name...*/
269 if (args.fspec == 0) {
271 * Process export requests. Jumping to "success"
272 * will return the vfs_export() error code.
274 error = vfs_export(mp, &ump->um_export, &args.export);
280 * Not an update, or updating the name: look up the name
281 * and verify that it refers to a sensible block device.
284 error = nlookup_init(&nd, args.fspec, UIO_USERSPACE, NLC_FOLLOW);
286 error = nlookup(&nd);
288 error = cache_vref(&nd.nl_nch, nd.nl_cred, &devvp);
293 if (!vn_isdisk(devvp, &error))
297 * If mount by non-root, then verify that user has necessary
298 * permissions on the device.
300 if (cred->cr_uid != 0) {
302 if ((mp->mnt_flag & MNT_RDONLY) == 0)
303 accessmode |= VWRITE;
304 vn_lock(devvp, LK_EXCLUSIVE | LK_RETRY);
305 if ((error = VOP_ACCESS(devvp, accessmode, cred)) != 0) {
312 if (mp->mnt_flag & MNT_UPDATE) {
314 * UPDATE - make sure the resolved vnode represents the same
315 * device. Note that devvp->v_rdev may be NULL since we
316 * haven't opened it, so compare udev instead.
318 * Our current open/writecount state is associated with
319 * um_devvp, so continue using um_devvp and throw away devvp.
321 if (devvp != ump->um_devvp) {
322 if (devvp->v_udev == ump->um_devvp->v_udev) {
324 devvp = ump->um_devvp;
326 kprintf("cannot update mount, udev does"
327 " not match %08x vs %08x\n",
328 devvp->v_udev, ump->um_devvp->v_udev);
329 error = EINVAL; /* needs translation */
335 * Update device name only on success
338 /* Save "mounted from" info for mount point (NULL pad)*/
339 copyinstr( args.fspec,
340 mp->mnt_stat.f_mntfromname,
343 bzero( mp->mnt_stat.f_mntfromname + size, MNAMELEN - size);
352 /* Save "mounted from" info for mount point (NULL pad)*/
353 copyinstr( args.fspec, /* device name*/
354 mp->mnt_stat.f_mntfromname, /* save area*/
355 MNAMELEN - 1, /* max size*/
356 &size); /* real size*/
357 bzero( mp->mnt_stat.f_mntfromname + size, MNAMELEN - size);
359 /* Save "last mounted on" info for mount point (NULL pad)*/
360 bzero(mp->mnt_stat.f_mntonname,
361 sizeof(mp->mnt_stat.f_mntonname));
363 copyinstr(path, mp->mnt_stat.f_mntonname,
364 sizeof(mp->mnt_stat.f_mntonname) - 1,
368 error = ffs_mountfs(devvp, mp, M_FFSNODE);
376 * Initialize FS stat information in mount struct; uses
377 * mp->mnt_stat.f_mntfromname.
379 * This code is common to root and non-root mounts
381 (void)VFS_STATFS(mp, &mp->mnt_stat, cred);
386 error_2: /* error with devvp held*/
388 /* release devvp before failing*/
391 error_1: /* no state to back out*/
394 if (!error && path && (mp->mnt_flag & MNT_UPDATE)) {
395 /* Update clean flag after changing read-onlyness. */
397 if (ronly != fs->fs_ronly) {
398 fs->fs_ronly = ronly;
399 fs->fs_clean = ronly &&
400 (fs->fs_flags & FS_UNCLEAN) == 0 ? 1 : 0;
403 * The device must be re-opened as appropriate or
404 * the device close at unmount time will panic.
406 vn_lock(devvp, LK_EXCLUSIVE | LK_RETRY);
408 VOP_OPEN(devvp, FREAD, FSCRED, NULL);
409 VOP_CLOSE(devvp, FREAD|FWRITE);
411 VOP_OPEN(devvp, FREAD|FWRITE, FSCRED, NULL);
412 VOP_CLOSE(devvp, FREAD);
415 ffs_sbupdate(ump, MNT_WAIT);
422 * Reload all incore data for a filesystem (used after running fsck on
423 * the root filesystem and finding things to fix). The filesystem must
424 * be mounted read-only.
426 * Things to do to update the mount:
427 * 1) invalidate all cached meta-data.
428 * 2) re-read superblock from disk.
429 * 3) re-read summary information from disk.
430 * 4) invalidate all inactive vnodes.
431 * 5) invalidate all cached file data.
432 * 6) re-read inode data for all active vnodes.
435 static int ffs_reload_scan2(struct mount *mp, struct vnode *vp, void *data);
446 ffs_reload(struct mount *mp, struct ucred *cred)
451 struct fs *fs, *newfs;
452 struct partinfo dpart;
454 int i, blks, size, error;
455 struct scaninfo scaninfo;
458 if ((mp->mnt_flag & MNT_RDONLY) == 0)
461 * Step 1: invalidate all cached meta-data.
463 devvp = VFSTOUFS(mp)->um_devvp;
464 vn_lock(devvp, LK_EXCLUSIVE | LK_RETRY);
465 error = vinvalbuf(devvp, 0, 0, 0);
468 panic("ffs_reload: dirty1");
473 * The backing device must be VMIO-capable because we use getblk().
474 * NOTE: the MFS driver now returns a VMIO-enabled descriptor.
476 if (devvp->v_object == NULL)
477 panic("ffs_reload: devvp has no VM object!");
480 * Step 2: re-read superblock from disk.
482 if (VOP_IOCTL(devvp, DIOCGPART, (caddr_t)&dpart, FREAD, cred) != 0)
485 size = dpart.disklab->d_secsize;
486 if ((error = bread(devvp, SBOFF, 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, fsbtodoff(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;
549 while (error == 0 && scaninfo.rescan) {
551 error = vmntvnodescan(mp, VMSC_GETVX,
552 NULL, ffs_reload_scan2, &scaninfo);
558 ffs_reload_scan2(struct mount *mp, struct vnode *vp, void *data)
560 struct scaninfo *info = data;
571 if (vinvalbuf(vp, 0, 0, 0))
572 panic("ffs_reload: dirty2");
574 * Step 6: re-read inode data for all active vnodes.
577 error = bread(info->devvp,
578 fsbtodoff(info->fs, ino_to_fsba(info->fs, ip->i_number)),
579 (int)info->fs->fs_bsize, &bp);
584 ip->i_din = *((struct ufs1_dinode *)bp->b_data +
585 ino_to_fsbo(info->fs, ip->i_number));
586 ip->i_effnlink = ip->i_nlink;
592 * Common code for mount and mountroot
595 ffs_mountfs(struct vnode *devvp, struct mount *mp, struct malloc_type *mtype)
597 struct ufsmount *ump;
601 struct partinfo dpart;
603 int error, i, blks, size, ronly;
605 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);
618 error = vinvalbuf(devvp, V_SAVE, 0, 0);
623 ronly = (mp->mnt_flag & MNT_RDONLY) != 0;
624 vn_lock(devvp, LK_EXCLUSIVE | LK_RETRY);
625 error = VOP_OPEN(devvp, ronly ? FREAD : FREAD|FWRITE, FSCRED, NULL);
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 * Filesystem supports native FSMIDs
638 mp->mnt_kern_flag |= MNTK_FSMID;
641 * The backing device must be VMIO-capable because we use getblk().
642 * NOTE: the MFS driver now returns a VMIO-enabled descriptor.
643 * The VOP_OPEN() call above should have associated a VM object
646 if (devvp->v_object == NULL)
647 panic("ffs_reload: devvp has no VM object!");
649 if (VOP_IOCTL(devvp, DIOCGPART, (caddr_t)&dpart, FREAD, proc0.p_ucred) != 0)
652 size = dpart.disklab->d_secsize;
656 if ((error = bread(devvp, SBOFF, 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 = kmalloc(sizeof *ump, M_UFSMNT, M_WAITOK);
686 bzero((caddr_t)ump, sizeof *ump);
687 ump->um_malloctype = mtype;
688 ump->um_i_effnlink_valid = 1;
689 ump->um_fs = kmalloc((u_long)fs->fs_sbsize, M_UFSMNT,
691 bcopy(bp->b_data, ump->um_fs, (uint)fs->fs_sbsize);
692 if (fs->fs_sbsize < SBSIZE)
693 bp->b_flags |= B_INVAL;
697 fs->fs_ronly = ronly;
698 size = fs->fs_cssize;
699 blks = howmany(size, fs->fs_fsize);
700 if (fs->fs_contigsumsize > 0)
701 size += fs->fs_ncg * sizeof(int32_t);
702 size += fs->fs_ncg * sizeof(uint8_t);
703 space = kmalloc((u_long)size, M_UFSMNT, M_WAITOK);
705 for (i = 0; i < blks; i += fs->fs_frag) {
707 if (i + fs->fs_frag > blks)
708 size = (blks - i) * fs->fs_fsize;
709 if ((error = bread(devvp, fsbtodoff(fs, fs->fs_csaddr + i),
711 kfree(fs->fs_csp, M_UFSMNT);
714 bcopy(bp->b_data, space, (uint)size);
715 space = (char *)space + size;
719 if (fs->fs_contigsumsize > 0) {
720 fs->fs_maxcluster = lp = space;
721 for (i = 0; i < fs->fs_ncg; i++)
722 *lp++ = fs->fs_contigsumsize;
725 size = fs->fs_ncg * sizeof(uint8_t);
726 fs->fs_contigdirs = (uint8_t *)space;
727 bzero(fs->fs_contigdirs, size);
728 /* Compatibility for old filesystems XXX */
729 if (fs->fs_avgfilesize <= 0) /* XXX */
730 fs->fs_avgfilesize = AVFILESIZ; /* XXX */
731 if (fs->fs_avgfpdir <= 0) /* XXX */
732 fs->fs_avgfpdir = AFPDIR; /* XXX */
733 mp->mnt_data = (qaddr_t)ump;
734 mp->mnt_stat.f_fsid.val[0] = fs->fs_id[0];
735 mp->mnt_stat.f_fsid.val[1] = fs->fs_id[1];
736 if (fs->fs_id[0] == 0 || fs->fs_id[1] == 0 ||
737 vfs_getvfs(&mp->mnt_stat.f_fsid))
739 mp->mnt_maxsymlinklen = fs->fs_maxsymlinklen;
740 mp->mnt_flag |= MNT_LOCAL;
743 ump->um_devvp = devvp;
744 ump->um_nindir = fs->fs_nindir;
745 ump->um_bptrtodb = fs->fs_fsbtodb;
746 ump->um_seqinc = fs->fs_frag;
747 for (i = 0; i < MAXQUOTAS; i++)
748 ump->um_quotas[i] = NULLVP;
749 dev->si_mountpoint = mp;
752 /* restore "last mounted on" here */
753 bzero(fs->fs_fsmnt, sizeof(fs->fs_fsmnt));
754 ksnprintf(fs->fs_fsmnt, sizeof(fs->fs_fsmnt),
755 "%s", mp->mnt_stat.f_mntonname);
757 if( mp->mnt_flag & MNT_ROOTFS) {
759 * Root mount; update timestamp in mount structure.
760 * this will be used by the common root mount code
761 * to update the system clock.
763 mp->mnt_time = fs->fs_time;
766 ump->um_savedmaxfilesize = fs->fs_maxfilesize; /* XXX */
767 maxfilesize = (uint64_t)0x40000000 * fs->fs_bsize - 1; /* XXX */
768 /* Enforce limit caused by vm object backing (32 bits vm_pindex_t). */
769 if (maxfilesize > (uint64_t)0x80000000u * PAGE_SIZE - 1)
770 maxfilesize = (uint64_t)0x80000000u * PAGE_SIZE - 1;
771 if (fs->fs_maxfilesize > maxfilesize) /* XXX */
772 fs->fs_maxfilesize = maxfilesize; /* XXX */
774 if ((fs->fs_flags & FS_DOSOFTDEP) &&
775 (error = softdep_mount(devvp, mp, fs)) != 0) {
776 kfree(fs->fs_csp, M_UFSMNT);
781 (void) ffs_sbupdate(ump, MNT_WAIT);
783 vfs_add_vnodeops(mp, &ffs_vnode_vops, &mp->mnt_vn_norm_ops);
784 vfs_add_vnodeops(mp, &ffs_spec_vops, &mp->mnt_vn_spec_ops);
785 vfs_add_vnodeops(mp, &ffs_fifo_vops, &mp->mnt_vn_fifo_ops);
789 dev->si_mountpoint = NULL;
792 VOP_CLOSE(devvp, ronly ? FREAD : FREAD|FWRITE);
794 kfree(ump->um_fs, M_UFSMNT);
795 kfree(ump, M_UFSMNT);
796 mp->mnt_data = (qaddr_t)0;
802 * Sanity checks for old filesystems.
804 * XXX - goes away some day.
807 ffs_oldfscompat(struct fs *fs)
809 fs->fs_npsect = max(fs->fs_npsect, fs->fs_nsect); /* XXX */
810 fs->fs_interleave = max(fs->fs_interleave, 1); /* XXX */
811 if (fs->fs_postblformat == FS_42POSTBLFMT) /* XXX */
812 fs->fs_nrpos = 8; /* XXX */
813 if (fs->fs_inodefmt < FS_44INODEFMT) { /* XXX */
816 uint64_t sizepb = fs->fs_bsize; /* XXX */
818 fs->fs_maxfilesize = fs->fs_bsize * NDADDR - 1; /* XXX */
819 for (i = 0; i < NIADDR; i++) { /* XXX */
820 sizepb *= NINDIR(fs); /* XXX */
821 fs->fs_maxfilesize += sizepb; /* XXX */
824 fs->fs_maxfilesize = (u_quad_t) 1LL << 39;
825 fs->fs_qbmask = ~fs->fs_bmask; /* XXX */
826 fs->fs_qfmask = ~fs->fs_fmask; /* XXX */
832 * unmount system call
835 ffs_unmount(struct mount *mp, int mntflags)
837 struct ufsmount *ump;
842 if (mntflags & MNT_FORCE) {
845 if (mp->mnt_flag & MNT_SOFTDEP) {
846 if ((error = softdep_flushfiles(mp, flags)) != 0)
849 if ((error = ffs_flushfiles(mp, flags)) != 0)
854 if (fs->fs_ronly == 0) {
855 fs->fs_clean = fs->fs_flags & FS_UNCLEAN ? 0 : 1;
856 error = ffs_sbupdate(ump, MNT_WAIT);
862 ump->um_devvp->v_rdev->si_mountpoint = NULL;
864 vinvalbuf(ump->um_devvp, V_SAVE, 0, 0);
865 error = VOP_CLOSE(ump->um_devvp, fs->fs_ronly ? FREAD : FREAD|FWRITE);
867 vrele(ump->um_devvp);
869 kfree(fs->fs_csp, M_UFSMNT);
871 kfree(ump, M_UFSMNT);
872 mp->mnt_data = (qaddr_t)0;
873 mp->mnt_flag &= ~MNT_LOCAL;
878 * Flush out all the files in a filesystem.
881 ffs_flushfiles(struct mount *mp, int flags)
883 struct ufsmount *ump;
888 if (mp->mnt_flag & MNT_QUOTA) {
890 error = vflush(mp, 0, SKIPSYSTEM|flags);
893 /* Find out how many quota files we have open. */
894 for (i = 0; i < MAXQUOTAS; i++) {
895 if (ump->um_quotas[i] == NULLVP)
900 * Here we fall through to vflush again to ensure
901 * that we have gotten rid of all the system vnodes.
906 * Flush all the files.
908 if ((error = vflush(mp, 0, flags)) != 0)
911 * Flush filesystem metadata.
913 vn_lock(ump->um_devvp, LK_EXCLUSIVE | LK_RETRY);
914 error = VOP_FSYNC(ump->um_devvp, MNT_WAIT);
915 vn_unlock(ump->um_devvp);
920 * Get filesystem statistics.
923 ffs_statfs(struct mount *mp, struct statfs *sbp, struct ucred *cred)
925 struct ufsmount *ump;
930 if (fs->fs_magic != FS_MAGIC)
932 sbp->f_bsize = fs->fs_fsize;
933 sbp->f_iosize = fs->fs_bsize;
934 sbp->f_blocks = fs->fs_dsize;
935 sbp->f_bfree = fs->fs_cstotal.cs_nbfree * fs->fs_frag +
936 fs->fs_cstotal.cs_nffree;
937 sbp->f_bavail = freespace(fs, fs->fs_minfree);
938 sbp->f_files = fs->fs_ncg * fs->fs_ipg - ROOTINO;
939 sbp->f_ffree = fs->fs_cstotal.cs_nifree;
940 if (sbp != &mp->mnt_stat) {
941 sbp->f_type = mp->mnt_vfc->vfc_typenum;
942 bcopy((caddr_t)mp->mnt_stat.f_mntfromname,
943 (caddr_t)&sbp->f_mntfromname[0], MNAMELEN);
949 * Go through the disk queues to initiate sandbagged IO;
950 * go through the inodes to write those that have been modified;
951 * initiate the writing of the super block if it has been modified.
953 * Note: we are always called with the filesystem marked `MPBUSY'.
957 static int ffs_sync_scan1(struct mount *mp, struct vnode *vp, void *data);
958 static int ffs_sync_scan2(struct mount *mp, struct vnode *vp, void *data);
961 ffs_sync(struct mount *mp, int waitfor)
963 struct ufsmount *ump = VFSTOUFS(mp);
966 struct scaninfo scaninfo;
969 if (fs->fs_fmod != 0 && fs->fs_ronly != 0) { /* XXX */
970 kprintf("fs = %s\n", fs->fs_fsmnt);
971 panic("ffs_sync: rofs mod");
975 * Write back each (modified) inode.
977 scaninfo.allerror = 0;
979 scaninfo.waitfor = waitfor;
980 while (scaninfo.rescan) {
982 vmntvnodescan(mp, VMSC_GETVP|VMSC_NOWAIT,
983 ffs_sync_scan1, ffs_sync_scan2, &scaninfo);
987 * Force stale filesystem control information to be flushed.
989 if (waitfor != MNT_LAZY) {
990 if (ump->um_mountp->mnt_flag & MNT_SOFTDEP)
991 waitfor = MNT_NOWAIT;
992 vn_lock(ump->um_devvp, LK_EXCLUSIVE | LK_RETRY);
993 if ((error = VOP_FSYNC(ump->um_devvp, waitfor)) != 0)
994 scaninfo.allerror = error;
995 vn_unlock(ump->um_devvp);
1001 * Write back modified superblock.
1003 if (fs->fs_fmod != 0 && (error = ffs_sbupdate(ump, waitfor)) != 0)
1004 scaninfo.allerror = error;
1005 return (scaninfo.allerror);
1009 ffs_sync_scan1(struct mount *mp, struct vnode *vp, void *data)
1014 * Depend on the mount list's vnode lock to keep things stable
1015 * enough for a quick test. Since there might be hundreds of
1016 * thousands of vnodes, we cannot afford even a subroutine
1017 * call unless there's a good chance that we have work to do.
1020 /* Restart out whole search if this guy is locked
1021 * or is being reclaimed.
1023 if (vp->v_type == VNON || ((ip->i_flag &
1024 (IN_ACCESS | IN_CHANGE | IN_MODIFIED | IN_UPDATE)) == 0 &&
1025 RB_EMPTY(&vp->v_rbdirty_tree))) {
1032 ffs_sync_scan2(struct mount *mp, struct vnode *vp, void *data)
1034 struct scaninfo *info = data;
1039 * We have to recheck after having obtained the vnode interlock.
1042 if (vp->v_type == VNON || ((ip->i_flag &
1043 (IN_ACCESS | IN_CHANGE | IN_MODIFIED | IN_UPDATE)) == 0 &&
1044 RB_EMPTY(&vp->v_rbdirty_tree))) {
1047 if (vp->v_type != VCHR) {
1048 if ((error = VOP_FSYNC(vp, info->waitfor)) != 0)
1049 info->allerror = error;
1052 * We must reference the vp to prevent it from
1053 * getting ripped out from under ffs_update, since
1054 * we are not holding a vnode lock.
1056 /* ffs_update(vp, waitfor == MNT_WAIT); */
1063 * Look up a FFS dinode number to find its incore vnode, otherwise read it
1064 * in from disk. If it is in core, wait for the lock bit to clear, then
1065 * return the inode locked. Detection and handling of mount points must be
1066 * done by the calling routine.
1070 ffs_vget(struct mount *mp, ino_t ino, struct vnode **vpp)
1074 struct ufsmount *ump;
1083 if ((*vpp = ufs_ihashget(dev, ino)) != NULL) {
1088 * If this MALLOC() is performed after the getnewvnode()
1089 * it might block, leaving a vnode with a NULL v_data to be
1090 * found by ffs_sync() if a sync happens to fire right then,
1091 * which will cause a panic because ffs_sync() blindly
1092 * dereferences vp->v_data (as well it should).
1094 * XXX this may no longer be true since getnewvnode returns a
1095 * VX locked vnode now.
1097 MALLOC(ip, struct inode *, sizeof(struct inode),
1098 ump->um_malloctype, M_WAITOK);
1100 /* Allocate a new vnode/inode. */
1101 error = getnewvnode(VT_UFS, mp, &vp, VLKTIMEOUT, LK_CANRECURSE);
1104 kfree(ip, ump->um_malloctype);
1107 bzero((caddr_t)ip, sizeof(struct inode));
1109 ip->i_fs = fs = ump->um_fs;
1115 for (i = 0; i < MAXQUOTAS; i++)
1116 ip->i_dquot[i] = NODQUOT;
1121 * Insert it into the inode hash table and check for a collision.
1122 * If a collision occurs, throw away the vnode and try again.
1124 if (ufs_ihashins(ip) != 0) {
1125 kprintf("debug: ufs ihashins collision, retrying inode %ld\n",
1126 (long)ip->i_number);
1128 kfree(ip, ump->um_malloctype);
1133 /* Read in the disk contents for the inode, copy into the inode. */
1134 error = bread(ump->um_devvp, fsbtodoff(fs, ino_to_fsba(fs, ino)),
1135 (int)fs->fs_bsize, &bp);
1138 * The inode does not contain anything useful, so it would
1139 * be misleading to leave it on its hash chain. With mode
1140 * still zero, it will be unlinked and returned to the free
1148 ip->i_din = *((struct ufs1_dinode *)bp->b_data + ino_to_fsbo(fs, ino));
1149 if (DOINGSOFTDEP(vp))
1150 softdep_load_inodeblock(ip);
1152 ip->i_effnlink = ip->i_nlink;
1156 * Initialize the vnode from the inode, check for aliases.
1157 * Note that the underlying vnode may have changed.
1159 error = ufs_vinit(mp, &vp);
1166 * Finish inode initialization now that aliasing has been resolved.
1168 ip->i_devvp = ump->um_devvp;
1171 * Set up a generation number for this inode if it does not
1172 * already have one. This should only happen on old filesystems.
1174 if (ip->i_gen == 0) {
1175 ip->i_gen = krandom() / 2 + 1;
1176 if ((vp->v_mount->mnt_flag & MNT_RDONLY) == 0)
1177 ip->i_flag |= IN_MODIFIED;
1180 * Ensure that uid and gid are correct. This is a temporary
1181 * fix until fsck has been changed to do the update.
1183 if (fs->fs_inodefmt < FS_44INODEFMT) { /* XXX */
1184 ip->i_uid = ip->i_din.di_ouid; /* XXX */
1185 ip->i_gid = ip->i_din.di_ogid; /* XXX */
1189 * return a VX locked and refd vnode (VX == same as normal vget()
1190 * vnode so we are ok)
1197 * File handle to vnode
1199 * Have to be really careful about stale file handles:
1200 * - check that the inode number is valid
1201 * - call ffs_vget() to get the locked inode
1202 * - check for an unallocated inode (i_mode == 0)
1203 * - check that the given client host has export rights and return
1204 * those rights via. exflagsp and credanonp
1207 ffs_fhtovp(struct mount *mp, struct fid *fhp, struct vnode **vpp)
1212 ufhp = (struct ufid *)fhp;
1213 fs = VFSTOUFS(mp)->um_fs;
1214 if (ufhp->ufid_ino < ROOTINO ||
1215 ufhp->ufid_ino >= fs->fs_ncg * fs->fs_ipg)
1217 return (ufs_fhtovp(mp, ufhp, vpp));
1221 * Vnode pointer to File handle
1225 ffs_vptofh(struct vnode *vp, struct fid *fhp)
1231 ufhp = (struct ufid *)fhp;
1232 ufhp->ufid_len = sizeof(struct ufid);
1233 ufhp->ufid_ino = ip->i_number;
1234 ufhp->ufid_gen = ip->i_gen;
1239 * Initialize the filesystem; just use ufs_init.
1242 ffs_init(struct vfsconf *vfsp)
1244 softdep_initialize();
1245 return (ufs_init(vfsp));
1249 * Write a superblock and associated information back to disk.
1252 ffs_sbupdate(struct ufsmount *mp, int waitfor)
1254 struct fs *dfs, *fs = mp->um_fs;
1258 int i, size, error, allerror = 0;
1261 * First write back the summary information.
1263 * NOTE: the getblk is relative to the device vnode so bio1
1264 * contains the device block number.
1266 blks = howmany(fs->fs_cssize, fs->fs_fsize);
1268 for (i = 0; i < blks; i += fs->fs_frag) {
1269 size = fs->fs_bsize;
1270 if (i + fs->fs_frag > blks)
1271 size = (blks - i) * fs->fs_fsize;
1272 bp = getblk(mp->um_devvp, fsbtodoff(fs, fs->fs_csaddr + i),
1274 bcopy(space, bp->b_data, (uint)size);
1275 space = (char *)space + size;
1276 if (waitfor != MNT_WAIT)
1278 else if ((error = bwrite(bp)) != 0)
1282 * Now write back the superblock itself. If any errors occurred
1283 * up to this point, then fail so that the superblock avoids
1284 * being written out as clean.
1288 bp = getblk(mp->um_devvp, SBOFF, (int)fs->fs_sbsize, 0, 0);
1290 fs->fs_time = time_second;
1291 bcopy((caddr_t)fs, bp->b_data, (uint)fs->fs_sbsize);
1292 /* Restore compatibility to old filesystems. XXX */
1293 dfs = (struct fs *)bp->b_data; /* XXX */
1294 if (fs->fs_postblformat == FS_42POSTBLFMT) /* XXX */
1295 dfs->fs_nrpos = -1; /* XXX */
1296 if (fs->fs_inodefmt < FS_44INODEFMT) { /* XXX */
1297 int32_t *lp, tmp; /* XXX */
1299 lp = (int32_t *)&dfs->fs_qbmask; /* XXX */
1300 tmp = lp[4]; /* XXX */
1301 for (i = 4; i > 0; i--) /* XXX */
1302 lp[i] = lp[i-1]; /* XXX */
1303 lp[0] = tmp; /* XXX */
1305 dfs->fs_maxfilesize = mp->um_savedmaxfilesize; /* XXX */
1306 if (waitfor != MNT_WAIT)
1308 else if ((error = bwrite(bp)) != 0)