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 1.2 2003/06/17 04:28:59 dillon Exp $
38 #include "opt_quota.h"
40 #include <sys/param.h>
41 #include <sys/systm.h>
42 #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>
53 #include <ufs/ufs/quota.h>
54 #include <ufs/ufs/ufsmount.h>
55 #include <ufs/ufs/inode.h>
56 #include <ufs/ufs/ufs_extern.h>
58 #include <ufs/ffs/fs.h>
59 #include <ufs/ffs/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 __P((struct ufsmount *, int));
68 static int ffs_reload __P((struct mount *,struct ucred *,struct proc *));
69 static int ffs_oldfscompat __P((struct fs *));
70 static int ffs_mount __P((struct mount *, char *, caddr_t,
71 struct nameidata *, struct proc *));
72 static int ffs_init __P((struct vfsconf *));
74 static struct vfsops ufs_vfsops = {
91 VFS_SET(ufs_vfsops, ufs, 0);
96 * Called when mounting local physical media
100 * mp mount point structure
101 * 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 * ndp mount point namei() return (used for
111 * credentials on reload), reused to look
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 * namei() if it is a genuine NULL from the user.
131 ffs_mount( mp, path, data, ndp, p)
132 struct mount *mp; /* mount struct pointer*/
133 char *path; /* path to mount point*/
134 caddr_t data; /* arguments to FS specific mount*/
135 struct nameidata *ndp; /* mount point credentials*/
136 struct proc *p; /* process requesting mount*/
142 struct ufs_args args;
143 struct ufsmount *ump = 0;
144 register struct fs *fs;
145 int error, flags, ronly = 0;
149 * Use NULL path to flag a root mount
154 * Mounting root file system
158 if ((err = bdevvp(rootdev, &rootvp))) {
159 printf("ffs_mountroot: can't find rootvp\n");
163 if( ( err = ffs_mountfs(rootvp, mp, p, M_FFSNODE)) != 0) {
164 /* fs specific cleanup (if any)*/
168 goto dostatfs; /* success*/
174 * Mounting non-root file system or updating a file system
178 /* copy in user arguments*/
179 err = copyin(data, (caddr_t)&args, sizeof (struct ufs_args));
181 goto error_1; /* can't get arguments*/
184 * If updating, check whether changing from read-only to
185 * read/write; if there is no device name, that's all we do.
187 if (mp->mnt_flag & MNT_UPDATE) {
190 devvp = ump->um_devvp;
192 ronly = fs->fs_ronly; /* MNT_RELOAD might change this */
193 if (ronly == 0 && (mp->mnt_flag & MNT_RDONLY)) {
195 * Flush any dirty data.
197 VFS_SYNC(mp, MNT_WAIT, p->p_ucred, p);
199 * Check for and optionally get rid of files open
203 if (mp->mnt_flag & MNT_FORCE)
205 if (mp->mnt_flag & MNT_SOFTDEP) {
206 err = softdep_flushfiles(mp, flags, p);
208 err = ffs_flushfiles(mp, flags, p);
212 if (!err && (mp->mnt_flag & MNT_RELOAD))
213 err = ffs_reload(mp, ndp->ni_cnd.cn_cred, p);
217 if (ronly && (mp->mnt_kern_flag & MNTK_WANTRDWR)) {
219 * If upgrade to read-write by non-root, then verify
220 * that user has necessary permissions on the device.
222 if (p->p_ucred->cr_uid != 0) {
223 vn_lock(devvp, LK_EXCLUSIVE | LK_RETRY, p);
224 if ((error = VOP_ACCESS(devvp, VREAD | VWRITE,
225 p->p_ucred, p)) != 0) {
226 VOP_UNLOCK(devvp, 0, p);
229 VOP_UNLOCK(devvp, 0, p);
232 fs->fs_flags &= ~FS_UNCLEAN;
233 if (fs->fs_clean == 0) {
234 fs->fs_flags |= FS_UNCLEAN;
235 if (mp->mnt_flag & MNT_FORCE) {
237 "WARNING: %s was not properly dismounted\n",
241 "WARNING: R/W mount of %s denied. Filesystem is not clean - run fsck\n",
248 /* check to see if we need to start softdep */
249 if (fs->fs_flags & FS_DOSOFTDEP) {
250 err = softdep_mount(devvp, mp, fs, p->p_ucred);
258 * Soft updates is incompatible with "async",
259 * so if we are doing softupdates stop the user
260 * from setting the async flag in an update.
261 * Softdep_mount() clears it in an initial mount
264 if (mp->mnt_flag & MNT_SOFTDEP) {
265 mp->mnt_flag &= ~MNT_ASYNC;
267 /* if not updating name...*/
268 if (args.fspec == 0) {
270 * Process export requests. Jumping to "success"
271 * will return the vfs_export() error code.
273 err = vfs_export(mp, &ump->um_export, &args.export);
279 * Not an update, or updating the name: look up the name
280 * and verify that it refers to a sensible block device.
282 NDINIT(ndp, LOOKUP, FOLLOW, UIO_USERSPACE, args.fspec, p);
285 /* can't get devvp!*/
289 NDFREE(ndp, NDF_ONLY_PNBUF);
292 if (!vn_isdisk(devvp, &err))
296 * If mount by non-root, then verify that user has necessary
297 * permissions on the device.
299 if (p->p_ucred->cr_uid != 0) {
301 if ((mp->mnt_flag & MNT_RDONLY) == 0)
302 accessmode |= VWRITE;
303 vn_lock(devvp, LK_EXCLUSIVE | LK_RETRY, p);
304 if ((error = VOP_ACCESS(devvp, accessmode, p->p_ucred, p)) != 0) {
308 VOP_UNLOCK(devvp, 0, p);
311 if (mp->mnt_flag & MNT_UPDATE) {
315 * If it's not the same vnode, or at least the same device
316 * then it's not correct.
320 if (devvp != ump->um_devvp) {
321 if ( devvp->v_rdev == ump->um_devvp->v_rdev) {
324 err = EINVAL; /* needs translation */
329 * Update device name only on success
332 /* Save "mounted from" info for mount point (NULL pad)*/
333 copyinstr( args.fspec,
334 mp->mnt_stat.f_mntfromname,
337 bzero( mp->mnt_stat.f_mntfromname + size, MNAMELEN - size);
347 * Since this is a new mount, we want the names for
348 * the device and the mount point copied in. If an
349 * error occurs, the mountpoint is discarded by the
352 /* Save "last mounted on" info for mount point (NULL pad)*/
353 copyinstr( path, /* mount point*/
354 mp->mnt_stat.f_mntonname, /* save area*/
355 MNAMELEN - 1, /* max size*/
356 &size); /* real size*/
357 bzero( mp->mnt_stat.f_mntonname + size, MNAMELEN - size);
359 /* Save "mounted from" info for mount point (NULL pad)*/
360 copyinstr( args.fspec, /* device name*/
361 mp->mnt_stat.f_mntfromname, /* save area*/
362 MNAMELEN - 1, /* max size*/
363 &size); /* real size*/
364 bzero( mp->mnt_stat.f_mntfromname + size, MNAMELEN - size);
366 err = ffs_mountfs(devvp, mp, p, M_FFSNODE);
374 * Initialize FS stat information in mount struct; uses both
375 * mp->mnt_stat.f_mntonname and mp->mnt_stat.f_mntfromname
377 * This code is common to root and non-root mounts
379 (void)VFS_STATFS(mp, &mp->mnt_stat, p);
384 error_2: /* error with devvp held*/
386 /* release devvp before failing*/
389 error_1: /* no state to back out*/
392 if (!err && path && (mp->mnt_flag & MNT_UPDATE)) {
393 /* Update clean flag after changing read-onlyness. */
395 if (ronly != fs->fs_ronly) {
396 fs->fs_ronly = ronly;
397 fs->fs_clean = ronly &&
398 (fs->fs_flags & FS_UNCLEAN) == 0 ? 1 : 0;
399 ffs_sbupdate(ump, MNT_WAIT);
406 * Reload all incore data for a filesystem (used after running fsck on
407 * the root filesystem and finding things to fix). The filesystem must
408 * be mounted read-only.
410 * Things to do to update the mount:
411 * 1) invalidate all cached meta-data.
412 * 2) re-read superblock from disk.
413 * 3) re-read summary information from disk.
414 * 4) invalidate all inactive vnodes.
415 * 5) invalidate all cached file data.
416 * 6) re-read inode data for all active vnodes.
419 ffs_reload(mp, cred, p)
420 register struct mount *mp;
424 register struct vnode *vp, *nvp, *devvp;
428 struct fs *fs, *newfs;
429 struct partinfo dpart;
431 int i, blks, size, error;
434 if ((mp->mnt_flag & MNT_RDONLY) == 0)
437 * Step 1: invalidate all cached meta-data.
439 devvp = VFSTOUFS(mp)->um_devvp;
440 vn_lock(devvp, LK_EXCLUSIVE | LK_RETRY, p);
441 error = vinvalbuf(devvp, 0, cred, p, 0, 0);
442 VOP_UNLOCK(devvp, 0, p);
444 panic("ffs_reload: dirty1");
449 * Only VMIO the backing device if the backing device is a real
450 * block device. See ffs_mountmfs() for more details.
452 if (devvp->v_tag != VT_MFS && vn_isdisk(devvp, NULL)) {
453 vn_lock(devvp, LK_EXCLUSIVE | LK_RETRY, p);
454 vfs_object_create(devvp, p, p->p_ucred);
455 simple_lock(&devvp->v_interlock);
456 VOP_UNLOCK(devvp, LK_INTERLOCK, p);
460 * Step 2: re-read superblock from disk.
462 if (VOP_IOCTL(devvp, DIOCGPART, (caddr_t)&dpart, FREAD, NOCRED, p) != 0)
465 size = dpart.disklab->d_secsize;
466 if ((error = bread(devvp, (ufs_daddr_t)(SBOFF/size), SBSIZE, NOCRED,&bp)) != 0)
468 newfs = (struct fs *)bp->b_data;
469 if (newfs->fs_magic != FS_MAGIC || newfs->fs_bsize > MAXBSIZE ||
470 newfs->fs_bsize < sizeof(struct fs)) {
472 return (EIO); /* XXX needs translation */
474 fs = VFSTOUFS(mp)->um_fs;
476 * Copy pointer fields back into superblock before copying in XXX
477 * new superblock. These should really be in the ufsmount. XXX
478 * Note that important parameters (eg fs_ncg) are unchanged.
480 newfs->fs_csp = fs->fs_csp;
481 newfs->fs_maxcluster = fs->fs_maxcluster;
482 newfs->fs_contigdirs = fs->fs_contigdirs;
483 bcopy(newfs, fs, (u_int)fs->fs_sbsize);
484 if (fs->fs_sbsize < SBSIZE)
485 bp->b_flags |= B_INVAL;
487 mp->mnt_maxsymlinklen = fs->fs_maxsymlinklen;
489 /* An old fsck may have zeroed these fields, so recheck them. */
490 if (fs->fs_avgfilesize <= 0) /* XXX */
491 fs->fs_avgfilesize = AVFILESIZ; /* XXX */
492 if (fs->fs_avgfpdir <= 0) /* XXX */
493 fs->fs_avgfpdir = AFPDIR; /* XXX */
496 * Step 3: re-read summary information from disk.
498 blks = howmany(fs->fs_cssize, fs->fs_fsize);
500 for (i = 0; i < blks; i += fs->fs_frag) {
502 if (i + fs->fs_frag > blks)
503 size = (blks - i) * fs->fs_fsize;
504 error = bread(devvp, fsbtodb(fs, fs->fs_csaddr + i), size,
508 bcopy(bp->b_data, space, (u_int)size);
509 space = (char *)space + size;
513 * We no longer know anything about clusters per cylinder group.
515 if (fs->fs_contigsumsize > 0) {
516 lp = fs->fs_maxcluster;
517 for (i = 0; i < fs->fs_ncg; i++)
518 *lp++ = fs->fs_contigsumsize;
522 simple_lock(&mntvnode_slock);
523 for (vp = TAILQ_FIRST(&mp->mnt_nvnodelist); vp != NULL; vp = nvp) {
524 if (vp->v_mount != mp) {
525 simple_unlock(&mntvnode_slock);
528 nvp = TAILQ_NEXT(vp, v_nmntvnodes);
530 * Step 4: invalidate all inactive vnodes.
532 if (vrecycle(vp, &mntvnode_slock, p))
535 * Step 5: invalidate all cached file data.
537 simple_lock(&vp->v_interlock);
538 simple_unlock(&mntvnode_slock);
539 if (vget(vp, LK_EXCLUSIVE | LK_INTERLOCK, p)) {
542 if (vinvalbuf(vp, 0, cred, p, 0, 0))
543 panic("ffs_reload: dirty2");
545 * Step 6: re-read inode data for all active vnodes.
549 bread(devvp, fsbtodb(fs, ino_to_fsba(fs, ip->i_number)),
550 (int)fs->fs_bsize, NOCRED, &bp);
555 ip->i_din = *((struct dinode *)bp->b_data +
556 ino_to_fsbo(fs, ip->i_number));
557 ip->i_effnlink = ip->i_nlink;
560 simple_lock(&mntvnode_slock);
562 simple_unlock(&mntvnode_slock);
567 * Common code for mount and mountroot
570 ffs_mountfs(devvp, mp, p, malloctype)
571 register struct vnode *devvp;
574 struct malloc_type *malloctype;
576 register struct ufsmount *ump;
578 register struct fs *fs;
580 struct partinfo dpart;
582 int error, i, blks, size, ronly;
585 u_int64_t maxfilesize; /* XXX */
590 cred = p ? p->p_ucred : NOCRED;
592 * Disallow multiple mounts of the same device.
593 * Disallow mounting of a device that is currently in use
594 * (except for root, which might share swap device for miniroot).
595 * Flush out any old buffers remaining from a previous use.
597 error = vfs_mountedon(devvp);
600 ncount = vcount(devvp);
602 if (ncount > 1 && devvp != rootvp)
604 vn_lock(devvp, LK_EXCLUSIVE | LK_RETRY, p);
605 error = vinvalbuf(devvp, V_SAVE, cred, p, 0, 0);
606 VOP_UNLOCK(devvp, 0, p);
611 * Only VMIO the backing device if the backing device is a real
612 * block device. This excludes the original MFS implementation.
613 * Note that it is optional that the backing device be VMIOed. This
614 * increases the opportunity for metadata caching.
616 if (devvp->v_tag != VT_MFS && vn_isdisk(devvp, NULL)) {
617 vn_lock(devvp, LK_EXCLUSIVE | LK_RETRY, p);
618 vfs_object_create(devvp, p, p->p_ucred);
619 simple_lock(&devvp->v_interlock);
620 VOP_UNLOCK(devvp, LK_INTERLOCK, p);
623 ronly = (mp->mnt_flag & MNT_RDONLY) != 0;
624 vn_lock(devvp, LK_EXCLUSIVE | LK_RETRY, p);
625 error = VOP_OPEN(devvp, ronly ? FREAD : FREAD|FWRITE, FSCRED, p);
626 VOP_UNLOCK(devvp, 0, p);
629 if (devvp->v_rdev->si_iosize_max != 0)
630 mp->mnt_iosize_max = devvp->v_rdev->si_iosize_max;
631 if (mp->mnt_iosize_max > MAXPHYS)
632 mp->mnt_iosize_max = MAXPHYS;
634 if (VOP_IOCTL(devvp, DIOCGPART, (caddr_t)&dpart, FREAD, cred, p) != 0)
637 size = dpart.disklab->d_secsize;
641 if ((error = bread(devvp, SBLOCK, SBSIZE, cred, &bp)) != 0)
643 fs = (struct fs *)bp->b_data;
644 if (fs->fs_magic != FS_MAGIC || fs->fs_bsize > MAXBSIZE ||
645 fs->fs_bsize < sizeof(struct fs)) {
646 error = EINVAL; /* XXX needs translation */
650 fs->fs_flags &= ~FS_UNCLEAN;
651 if (fs->fs_clean == 0) {
652 fs->fs_flags |= FS_UNCLEAN;
653 if (ronly || (mp->mnt_flag & MNT_FORCE)) {
655 "WARNING: %s was not properly dismounted\n",
659 "WARNING: R/W mount of %s denied. Filesystem is not clean - run fsck\n",
665 /* XXX updating 4.2 FFS superblocks trashes rotational layout tables */
666 if (fs->fs_postblformat == FS_42POSTBLFMT && !ronly) {
667 error = EROFS; /* needs translation */
670 ump = malloc(sizeof *ump, M_UFSMNT, M_WAITOK);
671 bzero((caddr_t)ump, sizeof *ump);
672 ump->um_malloctype = malloctype;
673 ump->um_i_effnlink_valid = 1;
674 ump->um_fs = malloc((u_long)fs->fs_sbsize, M_UFSMNT,
676 ump->um_blkatoff = ffs_blkatoff;
677 ump->um_truncate = ffs_truncate;
678 ump->um_update = ffs_update;
679 ump->um_valloc = ffs_valloc;
680 ump->um_vfree = ffs_vfree;
681 bcopy(bp->b_data, ump->um_fs, (u_int)fs->fs_sbsize);
682 if (fs->fs_sbsize < SBSIZE)
683 bp->b_flags |= B_INVAL;
687 fs->fs_ronly = ronly;
688 size = fs->fs_cssize;
689 blks = howmany(size, fs->fs_fsize);
690 if (fs->fs_contigsumsize > 0)
691 size += fs->fs_ncg * sizeof(int32_t);
692 size += fs->fs_ncg * sizeof(u_int8_t);
693 space = malloc((u_long)size, M_UFSMNT, M_WAITOK);
695 for (i = 0; i < blks; i += fs->fs_frag) {
697 if (i + fs->fs_frag > blks)
698 size = (blks - i) * fs->fs_fsize;
699 if ((error = bread(devvp, fsbtodb(fs, fs->fs_csaddr + i), size,
701 free(fs->fs_csp, M_UFSMNT);
704 bcopy(bp->b_data, space, (u_int)size);
705 space = (char *)space + size;
709 if (fs->fs_contigsumsize > 0) {
710 fs->fs_maxcluster = lp = space;
711 for (i = 0; i < fs->fs_ncg; i++)
712 *lp++ = fs->fs_contigsumsize;
715 size = fs->fs_ncg * sizeof(u_int8_t);
716 fs->fs_contigdirs = (u_int8_t *)space;
717 bzero(fs->fs_contigdirs, size);
718 /* Compatibility for old filesystems XXX */
719 if (fs->fs_avgfilesize <= 0) /* XXX */
720 fs->fs_avgfilesize = AVFILESIZ; /* XXX */
721 if (fs->fs_avgfpdir <= 0) /* XXX */
722 fs->fs_avgfpdir = AFPDIR; /* XXX */
723 mp->mnt_data = (qaddr_t)ump;
724 mp->mnt_stat.f_fsid.val[0] = fs->fs_id[0];
725 mp->mnt_stat.f_fsid.val[1] = fs->fs_id[1];
726 if (fs->fs_id[0] == 0 || fs->fs_id[1] == 0 ||
727 vfs_getvfs(&mp->mnt_stat.f_fsid))
729 mp->mnt_maxsymlinklen = fs->fs_maxsymlinklen;
730 mp->mnt_flag |= MNT_LOCAL;
733 ump->um_devvp = devvp;
734 ump->um_nindir = fs->fs_nindir;
735 ump->um_bptrtodb = fs->fs_fsbtodb;
736 ump->um_seqinc = fs->fs_frag;
737 for (i = 0; i < MAXQUOTAS; i++)
738 ump->um_quotas[i] = NULLVP;
739 devvp->v_specmountpoint = mp;
743 * Set FS local "last mounted on" information (NULL pad)
745 copystr( mp->mnt_stat.f_mntonname, /* mount point*/
746 fs->fs_fsmnt, /* copy area*/
747 sizeof(fs->fs_fsmnt) - 1, /* max size*/
748 &strsize); /* real size*/
749 bzero( fs->fs_fsmnt + strsize, sizeof(fs->fs_fsmnt) - strsize);
751 if( mp->mnt_flag & MNT_ROOTFS) {
753 * Root mount; update timestamp in mount structure.
754 * this will be used by the common root mount code
755 * to update the system clock.
757 mp->mnt_time = fs->fs_time;
760 ump->um_savedmaxfilesize = fs->fs_maxfilesize; /* XXX */
761 maxfilesize = (u_int64_t)0x40000000 * fs->fs_bsize - 1; /* XXX */
762 /* Enforce limit caused by vm object backing (32 bits vm_pindex_t). */
763 if (maxfilesize > (u_int64_t)0x80000000u * PAGE_SIZE - 1)
764 maxfilesize = (u_int64_t)0x80000000u * PAGE_SIZE - 1;
765 if (fs->fs_maxfilesize > maxfilesize) /* XXX */
766 fs->fs_maxfilesize = maxfilesize; /* XXX */
768 if ((fs->fs_flags & FS_DOSOFTDEP) &&
769 (error = softdep_mount(devvp, mp, fs, cred)) != 0) {
770 free(fs->fs_csp, M_UFSMNT);
775 (void) ffs_sbupdate(ump, MNT_WAIT);
779 devvp->v_specmountpoint = NULL;
782 (void)VOP_CLOSE(devvp, ronly ? FREAD : FREAD|FWRITE, cred, p);
784 free(ump->um_fs, M_UFSMNT);
786 mp->mnt_data = (qaddr_t)0;
792 * Sanity checks for old file systems.
794 * XXX - goes away some day.
801 fs->fs_npsect = max(fs->fs_npsect, fs->fs_nsect); /* XXX */
802 fs->fs_interleave = max(fs->fs_interleave, 1); /* XXX */
803 if (fs->fs_postblformat == FS_42POSTBLFMT) /* XXX */
804 fs->fs_nrpos = 8; /* XXX */
805 if (fs->fs_inodefmt < FS_44INODEFMT) { /* XXX */
808 u_int64_t sizepb = fs->fs_bsize; /* XXX */
810 fs->fs_maxfilesize = fs->fs_bsize * NDADDR - 1; /* XXX */
811 for (i = 0; i < NIADDR; i++) { /* XXX */
812 sizepb *= NINDIR(fs); /* XXX */
813 fs->fs_maxfilesize += sizepb; /* XXX */
816 fs->fs_maxfilesize = (u_quad_t) 1LL << 39;
817 fs->fs_qbmask = ~fs->fs_bmask; /* XXX */
818 fs->fs_qfmask = ~fs->fs_fmask; /* XXX */
824 * unmount system call
827 ffs_unmount(mp, mntflags, p)
832 register struct ufsmount *ump;
833 register struct fs *fs;
837 if (mntflags & MNT_FORCE) {
840 if (mp->mnt_flag & MNT_SOFTDEP) {
841 if ((error = softdep_flushfiles(mp, flags, p)) != 0)
844 if ((error = ffs_flushfiles(mp, flags, p)) != 0)
849 if (fs->fs_ronly == 0) {
850 fs->fs_clean = fs->fs_flags & FS_UNCLEAN ? 0 : 1;
851 error = ffs_sbupdate(ump, MNT_WAIT);
857 ump->um_devvp->v_specmountpoint = NULL;
859 vinvalbuf(ump->um_devvp, V_SAVE, NOCRED, p, 0, 0);
860 error = VOP_CLOSE(ump->um_devvp, fs->fs_ronly ? FREAD : FREAD|FWRITE,
863 vrele(ump->um_devvp);
865 free(fs->fs_csp, M_UFSMNT);
868 mp->mnt_data = (qaddr_t)0;
869 mp->mnt_flag &= ~MNT_LOCAL;
874 * Flush out all the files in a filesystem.
877 ffs_flushfiles(mp, flags, p)
878 register struct mount *mp;
882 register struct ufsmount *ump;
887 if (mp->mnt_flag & MNT_QUOTA) {
889 error = vflush(mp, 0, SKIPSYSTEM|flags);
892 for (i = 0; i < MAXQUOTAS; i++) {
893 if (ump->um_quotas[i] == NULLVP)
898 * Here we fall through to vflush again to ensure
899 * that we have gotten rid of all the system vnodes.
904 * Flush all the files.
906 if ((error = vflush(mp, 0, flags)) != 0)
909 * Flush filesystem metadata.
911 vn_lock(ump->um_devvp, LK_EXCLUSIVE | LK_RETRY, p);
912 error = VOP_FSYNC(ump->um_devvp, p->p_ucred, MNT_WAIT, p);
913 VOP_UNLOCK(ump->um_devvp, 0, p);
918 * Get file system statistics.
921 ffs_statfs(mp, sbp, p)
923 register struct statfs *sbp;
926 register struct ufsmount *ump;
927 register struct fs *fs;
931 if (fs->fs_magic != FS_MAGIC)
933 sbp->f_bsize = fs->fs_fsize;
934 sbp->f_iosize = fs->fs_bsize;
935 sbp->f_blocks = fs->fs_dsize;
936 sbp->f_bfree = fs->fs_cstotal.cs_nbfree * fs->fs_frag +
937 fs->fs_cstotal.cs_nffree;
938 sbp->f_bavail = freespace(fs, fs->fs_minfree);
939 sbp->f_files = fs->fs_ncg * fs->fs_ipg - ROOTINO;
940 sbp->f_ffree = fs->fs_cstotal.cs_nifree;
941 if (sbp != &mp->mnt_stat) {
942 sbp->f_type = mp->mnt_vfc->vfc_typenum;
943 bcopy((caddr_t)mp->mnt_stat.f_mntonname,
944 (caddr_t)&sbp->f_mntonname[0], MNAMELEN);
945 bcopy((caddr_t)mp->mnt_stat.f_mntfromname,
946 (caddr_t)&sbp->f_mntfromname[0], MNAMELEN);
952 * Go through the disk queues to initiate sandbagged IO;
953 * go through the inodes to write those that have been modified;
954 * initiate the writing of the super block if it has been modified.
956 * Note: we are always called with the filesystem marked `MPBUSY'.
959 ffs_sync(mp, waitfor, cred, p)
965 struct vnode *nvp, *vp;
967 struct ufsmount *ump = VFSTOUFS(mp);
969 int error, allerror = 0;
972 if (fs->fs_fmod != 0 && fs->fs_ronly != 0) { /* XXX */
973 printf("fs = %s\n", fs->fs_fsmnt);
974 panic("ffs_sync: rofs mod");
977 * Write back each (modified) inode.
979 simple_lock(&mntvnode_slock);
981 for (vp = TAILQ_FIRST(&mp->mnt_nvnodelist); vp != NULL; vp = nvp) {
983 * If the vnode that we are about to sync is no longer
984 * associated with this mount point, start over.
986 if (vp->v_mount != mp)
990 * Depend on the mntvnode_slock to keep things stable enough
991 * for a quick test. Since there might be hundreds of
992 * thousands of vnodes, we cannot afford even a subroutine
993 * call unless there's a good chance that we have work to do.
995 nvp = TAILQ_NEXT(vp, v_nmntvnodes);
997 if (vp->v_type == VNON || ((ip->i_flag &
998 (IN_ACCESS | IN_CHANGE | IN_MODIFIED | IN_UPDATE)) == 0 &&
999 TAILQ_EMPTY(&vp->v_dirtyblkhd))) {
1002 if (vp->v_type != VCHR) {
1003 simple_unlock(&mntvnode_slock);
1004 error = vget(vp, LK_EXCLUSIVE | LK_NOWAIT, p);
1006 simple_lock(&mntvnode_slock);
1007 if (error == ENOENT)
1010 if ((error = VOP_FSYNC(vp, cred, waitfor, p)) != 0)
1012 VOP_UNLOCK(vp, 0, p);
1014 simple_lock(&mntvnode_slock);
1018 * We must reference the vp to prevent it from
1019 * getting ripped out from under UFS_UPDATE, since
1020 * we are not holding a vnode lock. XXX why aren't
1021 * we holding a vnode lock?
1024 simple_unlock(&mntvnode_slock);
1025 /* UFS_UPDATE(vp, waitfor == MNT_WAIT); */
1028 simple_lock(&mntvnode_slock);
1030 if (TAILQ_NEXT(vp, v_nmntvnodes) != nvp)
1033 simple_unlock(&mntvnode_slock);
1035 * Force stale file system control information to be flushed.
1037 if (waitfor != MNT_LAZY) {
1038 if (ump->um_mountp->mnt_flag & MNT_SOFTDEP)
1039 waitfor = MNT_NOWAIT;
1040 vn_lock(ump->um_devvp, LK_EXCLUSIVE | LK_RETRY, p);
1041 if ((error = VOP_FSYNC(ump->um_devvp, cred, waitfor, p)) != 0)
1043 VOP_UNLOCK(ump->um_devvp, 0, p);
1049 * Write back modified superblock.
1051 if (fs->fs_fmod != 0 && (error = ffs_sbupdate(ump, waitfor)) != 0)
1057 * Look up a FFS dinode number to find its incore vnode, otherwise read it
1058 * in from disk. If it is in core, wait for the lock bit to clear, then
1059 * return the inode locked. Detection and handling of mount points must be
1060 * done by the calling routine.
1062 static int ffs_inode_hash_lock;
1065 ffs_vget(mp, ino, vpp)
1072 struct ufsmount *ump;
1081 if ((*vpp = ufs_ihashget(dev, ino)) != NULL) {
1086 * Lock out the creation of new entries in the FFS hash table in
1087 * case getnewvnode() or MALLOC() blocks, otherwise a duplicate
1090 if (ffs_inode_hash_lock) {
1091 while (ffs_inode_hash_lock) {
1092 ffs_inode_hash_lock = -1;
1093 tsleep(&ffs_inode_hash_lock, PVM, "ffsvgt", 0);
1097 ffs_inode_hash_lock = 1;
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 MALLOC(ip, struct inode *, sizeof(struct inode),
1107 ump->um_malloctype, M_WAITOK);
1109 /* Allocate a new vnode/inode. */
1110 error = getnewvnode(VT_UFS, mp, ffs_vnodeop_p, &vp);
1112 if (ffs_inode_hash_lock < 0)
1113 wakeup(&ffs_inode_hash_lock);
1114 ffs_inode_hash_lock = 0;
1116 FREE(ip, ump->um_malloctype);
1119 bzero((caddr_t)ip, sizeof(struct inode));
1120 lockinit(&ip->i_lock, PINOD, "inode", VLKTIMEOUT, LK_CANRECURSE);
1123 * FFS supports lock sharing in the stack of vnodes
1125 vp->v_vnlock = &ip->i_lock;
1127 ip->i_fs = fs = ump->um_fs;
1133 for (i = 0; i < MAXQUOTAS; i++)
1134 ip->i_dquot[i] = NODQUOT;
1138 * Put it onto its hash chain and lock it so that other requests for
1139 * this inode will block if they arrive while we are sleeping waiting
1140 * for old data structures to be purged or for the contents of the
1141 * disk portion of this inode to be read.
1145 if (ffs_inode_hash_lock < 0)
1146 wakeup(&ffs_inode_hash_lock);
1147 ffs_inode_hash_lock = 0;
1149 /* Read in the disk contents for the inode, copy into the inode. */
1150 error = bread(ump->um_devvp, fsbtodb(fs, ino_to_fsba(fs, ino)),
1151 (int)fs->fs_bsize, NOCRED, &bp);
1154 * The inode does not contain anything useful, so it would
1155 * be misleading to leave it on its hash chain. With mode
1156 * still zero, it will be unlinked and returned to the free
1164 ip->i_din = *((struct dinode *)bp->b_data + ino_to_fsbo(fs, ino));
1165 if (DOINGSOFTDEP(vp))
1166 softdep_load_inodeblock(ip);
1168 ip->i_effnlink = ip->i_nlink;
1172 * Initialize the vnode from the inode, check for aliases.
1173 * Note that the underlying vnode may have changed.
1175 error = ufs_vinit(mp, ffs_specop_p, ffs_fifoop_p, &vp);
1182 * Finish inode initialization now that aliasing has been resolved.
1184 ip->i_devvp = ump->um_devvp;
1187 * Set up a generation number for this inode if it does not
1188 * already have one. This should only happen on old filesystems.
1190 if (ip->i_gen == 0) {
1191 ip->i_gen = random() / 2 + 1;
1192 if ((vp->v_mount->mnt_flag & MNT_RDONLY) == 0)
1193 ip->i_flag |= IN_MODIFIED;
1196 * Ensure that uid and gid are correct. This is a temporary
1197 * fix until fsck has been changed to do the update.
1199 if (fs->fs_inodefmt < FS_44INODEFMT) { /* XXX */
1200 ip->i_uid = ip->i_din.di_ouid; /* XXX */
1201 ip->i_gid = ip->i_din.di_ogid; /* XXX */
1209 * File handle to vnode
1211 * Have to be really careful about stale file handles:
1212 * - check that the inode number is valid
1213 * - call ffs_vget() to get the locked inode
1214 * - check for an unallocated inode (i_mode == 0)
1215 * - check that the given client host has export rights and return
1216 * those rights via. exflagsp and credanonp
1219 ffs_fhtovp(mp, fhp, vpp)
1220 register struct mount *mp;
1224 register struct ufid *ufhp;
1227 ufhp = (struct ufid *)fhp;
1228 fs = VFSTOUFS(mp)->um_fs;
1229 if (ufhp->ufid_ino < ROOTINO ||
1230 ufhp->ufid_ino >= fs->fs_ncg * fs->fs_ipg)
1232 return (ufs_fhtovp(mp, ufhp, vpp));
1236 * Vnode pointer to File handle
1244 register struct inode *ip;
1245 register struct ufid *ufhp;
1248 ufhp = (struct ufid *)fhp;
1249 ufhp->ufid_len = sizeof(struct ufid);
1250 ufhp->ufid_ino = ip->i_number;
1251 ufhp->ufid_gen = ip->i_gen;
1256 * Initialize the filesystem; just use ufs_init.
1260 struct vfsconf *vfsp;
1263 softdep_initialize();
1264 return (ufs_init(vfsp));
1268 * Write a superblock and associated information back to disk.
1271 ffs_sbupdate(mp, waitfor)
1272 struct ufsmount *mp;
1275 register struct fs *dfs, *fs = mp->um_fs;
1276 register struct buf *bp;
1279 int i, size, error, allerror = 0;
1282 * First write back the summary information.
1284 blks = howmany(fs->fs_cssize, fs->fs_fsize);
1286 for (i = 0; i < blks; i += fs->fs_frag) {
1287 size = fs->fs_bsize;
1288 if (i + fs->fs_frag > blks)
1289 size = (blks - i) * fs->fs_fsize;
1290 bp = getblk(mp->um_devvp, fsbtodb(fs, fs->fs_csaddr + i),
1292 bcopy(space, bp->b_data, (u_int)size);
1293 space = (char *)space + size;
1294 if (waitfor != MNT_WAIT)
1296 else if ((error = bwrite(bp)) != 0)
1300 * Now write back the superblock itself. If any errors occurred
1301 * up to this point, then fail so that the superblock avoids
1302 * being written out as clean.
1306 bp = getblk(mp->um_devvp, SBLOCK, (int)fs->fs_sbsize, 0, 0);
1308 fs->fs_time = time_second;
1309 bcopy((caddr_t)fs, bp->b_data, (u_int)fs->fs_sbsize);
1310 /* Restore compatibility to old file systems. XXX */
1311 dfs = (struct fs *)bp->b_data; /* XXX */
1312 if (fs->fs_postblformat == FS_42POSTBLFMT) /* XXX */
1313 dfs->fs_nrpos = -1; /* XXX */
1314 if (fs->fs_inodefmt < FS_44INODEFMT) { /* XXX */
1315 int32_t *lp, tmp; /* XXX */
1317 lp = (int32_t *)&dfs->fs_qbmask; /* XXX */
1318 tmp = lp[4]; /* XXX */
1319 for (i = 4; i > 0; i--) /* XXX */
1320 lp[i] = lp[i-1]; /* XXX */
1321 lp[0] = tmp; /* XXX */
1323 dfs->fs_maxfilesize = mp->um_savedmaxfilesize; /* XXX */
1324 if (waitfor != MNT_WAIT)
1326 else if ((error = bwrite(bp)) != 0)