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 1.15 2004/03/11 20:17:36 hmp 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,
71 struct nameidata *, struct thread *);
72 static int ffs_init (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, td)
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 thread *td; /* process requesting mount*/
142 struct ufs_args args;
143 struct ufsmount *ump = 0;
145 int error, flags, ronly = 0;
149 KKASSERT(td->td_proc);
150 cred = td->td_proc->p_ucred;
153 * Use NULL path to flag a root mount
158 * Mounting root file system
162 if ((err = bdevvp(rootdev, &rootvp))) {
163 printf("ffs_mountroot: can't find rootvp\n");
167 if( ( err = ffs_mountfs(rootvp, mp, td, M_FFSNODE)) != 0) {
168 /* fs specific cleanup (if any)*/
172 goto dostatfs; /* success*/
178 * Mounting non-root file system or updating a file system
182 /* copy in user arguments*/
183 err = copyin(data, (caddr_t)&args, sizeof (struct ufs_args));
185 goto error_1; /* can't get arguments*/
188 * If updating, check whether changing from read-only to
189 * read/write; if there is no device name, that's all we do.
191 if (mp->mnt_flag & MNT_UPDATE) {
194 devvp = ump->um_devvp;
196 ronly = fs->fs_ronly; /* MNT_RELOAD might change this */
197 if (ronly == 0 && (mp->mnt_flag & MNT_RDONLY)) {
199 * Flush any dirty data.
201 VFS_SYNC(mp, MNT_WAIT, td);
203 * Check for and optionally get rid of files open
207 if (mp->mnt_flag & MNT_FORCE)
209 if (mp->mnt_flag & MNT_SOFTDEP) {
210 err = softdep_flushfiles(mp, flags, td);
212 err = ffs_flushfiles(mp, flags, td);
216 if (!err && (mp->mnt_flag & MNT_RELOAD))
217 err = ffs_reload(mp, ndp->ni_cnd.cn_cred, td);
221 if (ronly && (mp->mnt_kern_flag & MNTK_WANTRDWR)) {
223 * If upgrade to read-write by non-root, then verify
224 * that user has necessary permissions on the device.
226 if (cred->cr_uid != 0) {
227 vn_lock(devvp, NULL, LK_EXCLUSIVE | LK_RETRY, td);
228 if ((error = VOP_ACCESS(devvp, VREAD | VWRITE,
230 VOP_UNLOCK(devvp, NULL, 0, td);
233 VOP_UNLOCK(devvp, NULL, 0, td);
236 fs->fs_flags &= ~FS_UNCLEAN;
237 if (fs->fs_clean == 0) {
238 fs->fs_flags |= FS_UNCLEAN;
239 if (mp->mnt_flag & MNT_FORCE) {
241 "WARNING: %s was not properly dismounted\n",
245 "WARNING: R/W mount of %s denied. Filesystem is not clean - run fsck\n",
252 /* check to see if we need to start softdep */
253 if (fs->fs_flags & FS_DOSOFTDEP) {
254 err = softdep_mount(devvp, mp, fs);
262 * Soft updates is incompatible with "async",
263 * so if we are doing softupdates stop the user
264 * from setting the async flag in an update.
265 * Softdep_mount() clears it in an initial mount
268 if (mp->mnt_flag & MNT_SOFTDEP) {
269 mp->mnt_flag &= ~MNT_ASYNC;
271 /* if not updating name...*/
272 if (args.fspec == 0) {
274 * Process export requests. Jumping to "success"
275 * will return the vfs_export() error code.
277 err = vfs_export(mp, &ump->um_export, &args.export);
283 * Not an update, or updating the name: look up the name
284 * and verify that it refers to a sensible block device.
286 NDINIT(ndp, NAMEI_LOOKUP, CNP_FOLLOW, UIO_USERSPACE, args.fspec, td);
289 /* can't get devvp!*/
293 NDFREE(ndp, NDF_ONLY_PNBUF);
296 if (!vn_isdisk(devvp, &err))
300 * If mount by non-root, then verify that user has necessary
301 * permissions on the device.
303 if (cred->cr_uid != 0) {
305 if ((mp->mnt_flag & MNT_RDONLY) == 0)
306 accessmode |= VWRITE;
307 vn_lock(devvp, NULL, LK_EXCLUSIVE | LK_RETRY, td);
308 if ((error = VOP_ACCESS(devvp, accessmode, cred, td)) != 0) {
312 VOP_UNLOCK(devvp, NULL, 0, td);
315 if (mp->mnt_flag & MNT_UPDATE) {
319 * If it's not the same vnode, or at least the same device
320 * then it's not correct.
324 if (devvp != ump->um_devvp) {
325 if ( devvp->v_rdev == ump->um_devvp->v_rdev) {
328 err = EINVAL; /* needs translation */
333 * Update device name only on success
336 /* Save "mounted from" info for mount point (NULL pad)*/
337 copyinstr( args.fspec,
338 mp->mnt_stat.f_mntfromname,
341 bzero( mp->mnt_stat.f_mntfromname + size, MNAMELEN - size);
351 * Since this is a new mount, we want the names for
352 * the device and the mount point copied in. If an
353 * error occurs, the mountpoint is discarded by the
356 /* Save "last mounted on" info for mount point (NULL pad)*/
357 copyinstr( path, /* mount point*/
358 mp->mnt_stat.f_mntonname, /* save area*/
359 MNAMELEN - 1, /* max size*/
360 &size); /* real size*/
361 bzero( mp->mnt_stat.f_mntonname + size, MNAMELEN - size);
363 /* Save "mounted from" info for mount point (NULL pad)*/
364 copyinstr( args.fspec, /* device name*/
365 mp->mnt_stat.f_mntfromname, /* save area*/
366 MNAMELEN - 1, /* max size*/
367 &size); /* real size*/
368 bzero( mp->mnt_stat.f_mntfromname + size, MNAMELEN - size);
370 err = ffs_mountfs(devvp, mp, td, M_FFSNODE);
378 * Initialize FS stat information in mount struct; uses both
379 * mp->mnt_stat.f_mntonname and mp->mnt_stat.f_mntfromname
381 * This code is common to root and non-root mounts
383 (void)VFS_STATFS(mp, &mp->mnt_stat, td);
388 error_2: /* error with devvp held*/
390 /* release devvp before failing*/
393 error_1: /* no state to back out*/
396 if (!err && path && (mp->mnt_flag & MNT_UPDATE)) {
397 /* Update clean flag after changing read-onlyness. */
399 if (ronly != fs->fs_ronly) {
400 fs->fs_ronly = ronly;
401 fs->fs_clean = ronly &&
402 (fs->fs_flags & FS_UNCLEAN) == 0 ? 1 : 0;
403 ffs_sbupdate(ump, MNT_WAIT);
410 * Reload all incore data for a filesystem (used after running fsck on
411 * the root filesystem and finding things to fix). The filesystem must
412 * be mounted read-only.
414 * Things to do to update the mount:
415 * 1) invalidate all cached meta-data.
416 * 2) re-read superblock from disk.
417 * 3) re-read summary information from disk.
418 * 4) invalidate all inactive vnodes.
419 * 5) invalidate all cached file data.
420 * 6) re-read inode data for all active vnodes.
423 static int ffs_reload_scan1(struct mount *mp, struct vnode *vp, void *data);
424 static int ffs_reload_scan2(struct mount *mp, struct vnode *vp,
425 lwkt_tokref_t vlock, void *data);
437 ffs_reload(struct mount *mp, struct ucred *cred, struct thread *td)
442 struct fs *fs, *newfs;
443 struct partinfo dpart;
445 int i, blks, size, error;
447 struct scaninfo scaninfo;
450 if ((mp->mnt_flag & MNT_RDONLY) == 0)
453 * Step 1: invalidate all cached meta-data.
455 devvp = VFSTOUFS(mp)->um_devvp;
456 vn_lock(devvp, NULL, LK_EXCLUSIVE | LK_RETRY, td);
457 error = vinvalbuf(devvp, 0, td, 0, 0);
458 VOP_UNLOCK(devvp, NULL, 0, td);
460 panic("ffs_reload: dirty1");
465 * Only VMIO the backing device if the backing device is a real
466 * block device. See ffs_mountmfs() for more details.
468 if (devvp->v_tag != VT_MFS && vn_isdisk(devvp, NULL)) {
469 vn_lock(devvp, NULL, LK_EXCLUSIVE | LK_RETRY, td);
470 vfs_object_create(devvp, td);
471 lwkt_gettoken(&vlock, devvp->v_interlock);
472 VOP_UNLOCK(devvp, &vlock, LK_INTERLOCK, td);
476 * Step 2: re-read superblock from disk.
478 if (VOP_IOCTL(devvp, DIOCGPART, (caddr_t)&dpart, FREAD, NOCRED, td) != 0)
481 size = dpart.disklab->d_secsize;
482 if ((error = bread(devvp, (ufs_daddr_t)(SBOFF/size), SBSIZE, &bp)) != 0)
484 newfs = (struct fs *)bp->b_data;
485 if (newfs->fs_magic != FS_MAGIC || newfs->fs_bsize > MAXBSIZE ||
486 newfs->fs_bsize < sizeof(struct fs)) {
488 return (EIO); /* XXX needs translation */
490 fs = VFSTOUFS(mp)->um_fs;
492 * Copy pointer fields back into superblock before copying in XXX
493 * new superblock. These should really be in the ufsmount. XXX
494 * Note that important parameters (eg fs_ncg) are unchanged.
496 newfs->fs_csp = fs->fs_csp;
497 newfs->fs_maxcluster = fs->fs_maxcluster;
498 newfs->fs_contigdirs = fs->fs_contigdirs;
499 /* The file system is still read-only. */
501 bcopy(newfs, fs, (u_int)fs->fs_sbsize);
502 if (fs->fs_sbsize < SBSIZE)
503 bp->b_flags |= B_INVAL;
505 mp->mnt_maxsymlinklen = fs->fs_maxsymlinklen;
507 /* An old fsck may have zeroed these fields, so recheck them. */
508 if (fs->fs_avgfilesize <= 0) /* XXX */
509 fs->fs_avgfilesize = AVFILESIZ; /* XXX */
510 if (fs->fs_avgfpdir <= 0) /* XXX */
511 fs->fs_avgfpdir = AFPDIR; /* XXX */
514 * Step 3: re-read summary information from disk.
516 blks = howmany(fs->fs_cssize, fs->fs_fsize);
518 for (i = 0; i < blks; i += fs->fs_frag) {
520 if (i + fs->fs_frag > blks)
521 size = (blks - i) * fs->fs_fsize;
522 error = bread(devvp, fsbtodb(fs, fs->fs_csaddr + i), size, &bp);
525 bcopy(bp->b_data, space, (u_int)size);
526 space = (char *)space + size;
530 * We no longer know anything about clusters per cylinder group.
532 if (fs->fs_contigsumsize > 0) {
533 lp = fs->fs_maxcluster;
534 for (i = 0; i < fs->fs_ncg; i++)
535 *lp++ = fs->fs_contigsumsize;
540 scaninfo.devvp = devvp;
542 while (error == 0 && scaninfo.rescan) {
544 error = vmntvnodescan(mp, ffs_reload_scan1,
545 ffs_reload_scan2, &scaninfo);
552 ffs_reload_scan1(struct mount *mp, struct vnode *vp, void *data)
554 struct scaninfo *info = data;
557 * Step 4: invalidate all inactive vnodes.
559 if (vrecycle(vp, NULL, info->td)) {
561 return(-1); /* continue loop, do not call scan2 */
568 ffs_reload_scan2(struct mount *mp, struct vnode *vp, lwkt_tokref_t vlock, void *data)
570 struct scaninfo *info = data;
576 * Step 5: invalidate all cached file data.
578 if (vget(vp, vlock, LK_EXCLUSIVE | LK_INTERLOCK, info->td)) {
582 if (vinvalbuf(vp, 0, info->td, 0, 0))
583 panic("ffs_reload: dirty2");
585 * Step 6: re-read inode data for all active vnodes.
588 error = bread(info->devvp,
589 fsbtodb(info->fs, ino_to_fsba(info->fs, ip->i_number)),
590 (int)info->fs->fs_bsize, &bp);
595 ip->i_din = *((struct dinode *)bp->b_data +
596 ino_to_fsbo(info->fs, ip->i_number));
597 ip->i_effnlink = ip->i_nlink;
604 * Common code for mount and mountroot
607 ffs_mountfs(devvp, mp, td, malloctype)
611 struct malloc_type *malloctype;
613 struct ufsmount *ump;
617 struct partinfo dpart;
619 int error, i, blks, size, ronly;
622 u_int64_t maxfilesize; /* XXX */
628 * Disallow multiple mounts of the same device.
629 * Disallow mounting of a device that is currently in use
630 * (except for root, which might share swap device for miniroot).
631 * Flush out any old buffers remaining from a previous use.
633 error = vfs_mountedon(devvp);
636 ncount = vcount(devvp);
638 if (ncount > 1 && devvp != rootvp)
640 vn_lock(devvp, NULL, LK_EXCLUSIVE | LK_RETRY, td);
641 error = vinvalbuf(devvp, V_SAVE, td, 0, 0);
642 VOP_UNLOCK(devvp, NULL, 0, td);
647 * Only VMIO the backing device if the backing device is a real
648 * block device. This excludes the original MFS implementation.
649 * Note that it is optional that the backing device be VMIOed. This
650 * increases the opportunity for metadata caching.
652 if (devvp->v_tag != VT_MFS && vn_isdisk(devvp, NULL)) {
653 vn_lock(devvp, NULL, LK_EXCLUSIVE | LK_RETRY, td);
654 vfs_object_create(devvp, td);
655 lwkt_gettoken(&vlock, devvp->v_interlock);
656 VOP_UNLOCK(devvp, &vlock, LK_INTERLOCK, td);
659 ronly = (mp->mnt_flag & MNT_RDONLY) != 0;
660 vn_lock(devvp, NULL, LK_EXCLUSIVE | LK_RETRY, td);
661 error = VOP_OPEN(devvp, ronly ? FREAD : FREAD|FWRITE, FSCRED, td);
662 VOP_UNLOCK(devvp, NULL, 0, td);
665 if (devvp->v_rdev->si_iosize_max != 0)
666 mp->mnt_iosize_max = devvp->v_rdev->si_iosize_max;
667 if (mp->mnt_iosize_max > MAXPHYS)
668 mp->mnt_iosize_max = MAXPHYS;
670 if (VOP_IOCTL(devvp, DIOCGPART, (caddr_t)&dpart, FREAD, proc0.p_ucred, td) != 0)
673 size = dpart.disklab->d_secsize;
677 if ((error = bread(devvp, SBLOCK, SBSIZE, &bp)) != 0)
679 fs = (struct fs *)bp->b_data;
680 if (fs->fs_magic != FS_MAGIC || fs->fs_bsize > MAXBSIZE ||
681 fs->fs_bsize < sizeof(struct fs)) {
682 error = EINVAL; /* XXX needs translation */
686 fs->fs_flags &= ~FS_UNCLEAN;
687 if (fs->fs_clean == 0) {
688 fs->fs_flags |= FS_UNCLEAN;
689 if (ronly || (mp->mnt_flag & MNT_FORCE)) {
691 "WARNING: %s was not properly dismounted\n",
695 "WARNING: R/W mount of %s denied. Filesystem is not clean - run fsck\n",
701 /* XXX updating 4.2 FFS superblocks trashes rotational layout tables */
702 if (fs->fs_postblformat == FS_42POSTBLFMT && !ronly) {
703 error = EROFS; /* needs translation */
706 ump = malloc(sizeof *ump, M_UFSMNT, M_WAITOK);
707 bzero((caddr_t)ump, sizeof *ump);
708 ump->um_malloctype = malloctype;
709 ump->um_i_effnlink_valid = 1;
710 ump->um_fs = malloc((u_long)fs->fs_sbsize, M_UFSMNT,
712 ump->um_blkatoff = ffs_blkatoff;
713 ump->um_truncate = ffs_truncate;
714 ump->um_update = ffs_update;
715 ump->um_valloc = ffs_valloc;
716 ump->um_vfree = ffs_vfree;
717 bcopy(bp->b_data, ump->um_fs, (u_int)fs->fs_sbsize);
718 if (fs->fs_sbsize < SBSIZE)
719 bp->b_flags |= B_INVAL;
723 fs->fs_ronly = ronly;
724 size = fs->fs_cssize;
725 blks = howmany(size, fs->fs_fsize);
726 if (fs->fs_contigsumsize > 0)
727 size += fs->fs_ncg * sizeof(int32_t);
728 size += fs->fs_ncg * sizeof(u_int8_t);
729 space = malloc((u_long)size, M_UFSMNT, M_WAITOK);
731 for (i = 0; i < blks; i += fs->fs_frag) {
733 if (i + fs->fs_frag > blks)
734 size = (blks - i) * fs->fs_fsize;
735 if ((error = bread(devvp, fsbtodb(fs, fs->fs_csaddr + i), size,
737 free(fs->fs_csp, M_UFSMNT);
740 bcopy(bp->b_data, space, (u_int)size);
741 space = (char *)space + size;
745 if (fs->fs_contigsumsize > 0) {
746 fs->fs_maxcluster = lp = space;
747 for (i = 0; i < fs->fs_ncg; i++)
748 *lp++ = fs->fs_contigsumsize;
751 size = fs->fs_ncg * sizeof(u_int8_t);
752 fs->fs_contigdirs = (u_int8_t *)space;
753 bzero(fs->fs_contigdirs, size);
754 /* Compatibility for old filesystems XXX */
755 if (fs->fs_avgfilesize <= 0) /* XXX */
756 fs->fs_avgfilesize = AVFILESIZ; /* XXX */
757 if (fs->fs_avgfpdir <= 0) /* XXX */
758 fs->fs_avgfpdir = AFPDIR; /* XXX */
759 mp->mnt_data = (qaddr_t)ump;
760 mp->mnt_stat.f_fsid.val[0] = fs->fs_id[0];
761 mp->mnt_stat.f_fsid.val[1] = fs->fs_id[1];
762 if (fs->fs_id[0] == 0 || fs->fs_id[1] == 0 ||
763 vfs_getvfs(&mp->mnt_stat.f_fsid))
765 mp->mnt_maxsymlinklen = fs->fs_maxsymlinklen;
766 mp->mnt_flag |= MNT_LOCAL;
769 ump->um_devvp = devvp;
770 ump->um_nindir = fs->fs_nindir;
771 ump->um_bptrtodb = fs->fs_fsbtodb;
772 ump->um_seqinc = fs->fs_frag;
773 for (i = 0; i < MAXQUOTAS; i++)
774 ump->um_quotas[i] = NULLVP;
775 devvp->v_specmountpoint = mp;
779 * Set FS local "last mounted on" information (NULL pad)
781 copystr( mp->mnt_stat.f_mntonname, /* mount point*/
782 fs->fs_fsmnt, /* copy area*/
783 sizeof(fs->fs_fsmnt) - 1, /* max size*/
784 &strsize); /* real size*/
785 bzero( fs->fs_fsmnt + strsize, sizeof(fs->fs_fsmnt) - strsize);
787 if( mp->mnt_flag & MNT_ROOTFS) {
789 * Root mount; update timestamp in mount structure.
790 * this will be used by the common root mount code
791 * to update the system clock.
793 mp->mnt_time = fs->fs_time;
796 ump->um_savedmaxfilesize = fs->fs_maxfilesize; /* XXX */
797 maxfilesize = (u_int64_t)0x40000000 * fs->fs_bsize - 1; /* XXX */
798 /* Enforce limit caused by vm object backing (32 bits vm_pindex_t). */
799 if (maxfilesize > (u_int64_t)0x80000000u * PAGE_SIZE - 1)
800 maxfilesize = (u_int64_t)0x80000000u * PAGE_SIZE - 1;
801 if (fs->fs_maxfilesize > maxfilesize) /* XXX */
802 fs->fs_maxfilesize = maxfilesize; /* XXX */
804 if ((fs->fs_flags & FS_DOSOFTDEP) &&
805 (error = softdep_mount(devvp, mp, fs)) != 0) {
806 free(fs->fs_csp, M_UFSMNT);
811 (void) ffs_sbupdate(ump, MNT_WAIT);
815 devvp->v_specmountpoint = NULL;
818 (void)VOP_CLOSE(devvp, ronly ? FREAD : FREAD|FWRITE, td);
820 free(ump->um_fs, M_UFSMNT);
822 mp->mnt_data = (qaddr_t)0;
828 * Sanity checks for old file systems.
830 * XXX - goes away some day.
837 fs->fs_npsect = max(fs->fs_npsect, fs->fs_nsect); /* XXX */
838 fs->fs_interleave = max(fs->fs_interleave, 1); /* XXX */
839 if (fs->fs_postblformat == FS_42POSTBLFMT) /* XXX */
840 fs->fs_nrpos = 8; /* XXX */
841 if (fs->fs_inodefmt < FS_44INODEFMT) { /* XXX */
844 u_int64_t sizepb = fs->fs_bsize; /* XXX */
846 fs->fs_maxfilesize = fs->fs_bsize * NDADDR - 1; /* XXX */
847 for (i = 0; i < NIADDR; i++) { /* XXX */
848 sizepb *= NINDIR(fs); /* XXX */
849 fs->fs_maxfilesize += sizepb; /* XXX */
852 fs->fs_maxfilesize = (u_quad_t) 1LL << 39;
853 fs->fs_qbmask = ~fs->fs_bmask; /* XXX */
854 fs->fs_qfmask = ~fs->fs_fmask; /* XXX */
860 * unmount system call
863 ffs_unmount(struct mount *mp, int mntflags, struct thread *td)
865 struct ufsmount *ump;
870 if (mntflags & MNT_FORCE) {
873 if (mp->mnt_flag & MNT_SOFTDEP) {
874 if ((error = softdep_flushfiles(mp, flags, td)) != 0)
877 if ((error = ffs_flushfiles(mp, flags, td)) != 0)
882 if (fs->fs_ronly == 0) {
883 fs->fs_clean = fs->fs_flags & FS_UNCLEAN ? 0 : 1;
884 error = ffs_sbupdate(ump, MNT_WAIT);
890 ump->um_devvp->v_specmountpoint = NULL;
892 vinvalbuf(ump->um_devvp, V_SAVE, td, 0, 0);
893 error = VOP_CLOSE(ump->um_devvp, fs->fs_ronly ? FREAD : FREAD|FWRITE, td);
895 vrele(ump->um_devvp);
897 free(fs->fs_csp, M_UFSMNT);
900 mp->mnt_data = (qaddr_t)0;
901 mp->mnt_flag &= ~MNT_LOCAL;
906 * Flush out all the files in a filesystem.
909 ffs_flushfiles(struct mount *mp, int flags, struct thread *td)
911 struct ufsmount *ump;
916 if (mp->mnt_flag & MNT_QUOTA) {
918 error = vflush(mp, 0, SKIPSYSTEM|flags);
921 for (i = 0; i < MAXQUOTAS; i++) {
922 if (ump->um_quotas[i] == NULLVP)
927 * Here we fall through to vflush again to ensure
928 * that we have gotten rid of all the system vnodes.
933 * Flush all the files.
935 if ((error = vflush(mp, 0, flags)) != 0)
938 * Flush filesystem metadata.
940 vn_lock(ump->um_devvp, NULL, LK_EXCLUSIVE | LK_RETRY, td);
941 error = VOP_FSYNC(ump->um_devvp, MNT_WAIT, td);
942 VOP_UNLOCK(ump->um_devvp, NULL, 0, td);
947 * Get file system statistics.
950 ffs_statfs(struct mount *mp, struct statfs *sbp, struct thread *td)
952 struct ufsmount *ump;
957 if (fs->fs_magic != FS_MAGIC)
959 sbp->f_bsize = fs->fs_fsize;
960 sbp->f_iosize = fs->fs_bsize;
961 sbp->f_blocks = fs->fs_dsize;
962 sbp->f_bfree = fs->fs_cstotal.cs_nbfree * fs->fs_frag +
963 fs->fs_cstotal.cs_nffree;
964 sbp->f_bavail = freespace(fs, fs->fs_minfree);
965 sbp->f_files = fs->fs_ncg * fs->fs_ipg - ROOTINO;
966 sbp->f_ffree = fs->fs_cstotal.cs_nifree;
967 if (sbp != &mp->mnt_stat) {
968 sbp->f_type = mp->mnt_vfc->vfc_typenum;
969 bcopy((caddr_t)mp->mnt_stat.f_mntonname,
970 (caddr_t)&sbp->f_mntonname[0], MNAMELEN);
971 bcopy((caddr_t)mp->mnt_stat.f_mntfromname,
972 (caddr_t)&sbp->f_mntfromname[0], MNAMELEN);
978 * Go through the disk queues to initiate sandbagged IO;
979 * go through the inodes to write those that have been modified;
980 * initiate the writing of the super block if it has been modified.
982 * Note: we are always called with the filesystem marked `MPBUSY'.
986 static int ffs_sync_scan1(struct mount *mp, struct vnode *vp, void *data);
987 static int ffs_sync_scan2(struct mount *mp, struct vnode *vp,
988 lwkt_tokref_t vlock, void *data);
991 ffs_sync(struct mount *mp, int waitfor, struct thread *td)
993 struct ufsmount *ump = VFSTOUFS(mp);
996 struct scaninfo scaninfo;
999 if (fs->fs_fmod != 0 && fs->fs_ronly != 0) { /* XXX */
1000 printf("fs = %s\n", fs->fs_fsmnt);
1001 panic("ffs_sync: rofs mod");
1005 * Write back each (modified) inode.
1007 scaninfo.allerror = 0;
1008 scaninfo.rescan = 1;
1009 scaninfo.waitfor = waitfor;
1010 while (scaninfo.rescan) {
1011 scaninfo.rescan = 0;
1012 vmntvnodescan(mp, ffs_sync_scan1, ffs_sync_scan2, &scaninfo);
1016 * Force stale file system control information to be flushed.
1018 if (waitfor != MNT_LAZY) {
1019 if (ump->um_mountp->mnt_flag & MNT_SOFTDEP)
1020 waitfor = MNT_NOWAIT;
1021 vn_lock(ump->um_devvp, NULL, LK_EXCLUSIVE | LK_RETRY, td);
1022 if ((error = VOP_FSYNC(ump->um_devvp, waitfor, td)) != 0)
1023 scaninfo.allerror = error;
1024 VOP_UNLOCK(ump->um_devvp, NULL, 0, td);
1030 * Write back modified superblock.
1032 if (fs->fs_fmod != 0 && (error = ffs_sbupdate(ump, waitfor)) != 0)
1033 scaninfo.allerror = error;
1034 return (scaninfo.allerror);
1039 ffs_sync_scan1(struct mount *mp, struct vnode *vp, void *data)
1044 * Depend on the mount list's vnode lock to keep things stable
1045 * enough for a quick test. Since there might be hundreds of
1046 * thousands of vnodes, we cannot afford even a subroutine
1047 * call unless there's a good chance that we have work to do.
1050 if (vp->v_type == VNON || ((ip->i_flag &
1051 (IN_ACCESS | IN_CHANGE | IN_MODIFIED | IN_UPDATE)) == 0 &&
1052 TAILQ_EMPTY(&vp->v_dirtyblkhd))) {
1060 ffs_sync_scan2(struct mount *mp, struct vnode *vp,
1061 lwkt_tokref_t vlock, void *data)
1063 struct scaninfo *info = data;
1064 thread_t td = curthread; /* XXX */
1069 * We have to recheck after having obtained the vnode interlock.
1072 if (vp->v_type == VNON || ((ip->i_flag &
1073 (IN_ACCESS | IN_CHANGE | IN_MODIFIED | IN_UPDATE)) == 0 &&
1074 TAILQ_EMPTY(&vp->v_dirtyblkhd))) {
1075 lwkt_reltoken(vlock);
1078 if (vp->v_type != VCHR) {
1079 error = vget(vp, vlock, LK_INTERLOCK|LK_EXCLUSIVE|LK_NOWAIT, td);
1081 if (error == ENOENT)
1084 if ((error = VOP_FSYNC(vp, info->waitfor, td)) != 0)
1085 info->allerror = error;
1086 VOP_UNLOCK(vp, NULL, 0, td);
1091 * We must reference the vp to prevent it from
1092 * getting ripped out from under UFS_UPDATE, since
1093 * we are not holding a vnode lock.
1096 lwkt_reltoken(vlock);
1097 /* UFS_UPDATE(vp, waitfor == MNT_WAIT); */
1105 * Look up a FFS dinode number to find its incore vnode, otherwise read it
1106 * in from disk. If it is in core, wait for the lock bit to clear, then
1107 * return the inode locked. Detection and handling of mount points must be
1108 * done by the calling routine.
1110 static int ffs_inode_hash_lock;
1113 ffs_vget(mp, ino, vpp)
1120 struct ufsmount *ump;
1129 if ((*vpp = ufs_ihashget(dev, ino)) != NULL) {
1134 * Lock out the creation of new entries in the FFS hash table in
1135 * case getnewvnode() or MALLOC() blocks, otherwise a duplicate
1138 if (ffs_inode_hash_lock) {
1139 while (ffs_inode_hash_lock) {
1140 ffs_inode_hash_lock = -1;
1141 tsleep(&ffs_inode_hash_lock, 0, "ffsvgt", 0);
1145 ffs_inode_hash_lock = 1;
1148 * If this MALLOC() is performed after the getnewvnode()
1149 * it might block, leaving a vnode with a NULL v_data to be
1150 * found by ffs_sync() if a sync happens to fire right then,
1151 * which will cause a panic because ffs_sync() blindly
1152 * dereferences vp->v_data (as well it should).
1154 MALLOC(ip, struct inode *, sizeof(struct inode),
1155 ump->um_malloctype, M_WAITOK);
1157 /* Allocate a new vnode/inode. */
1158 error = getnewvnode(VT_UFS, mp, ffs_vnodeop_p, &vp);
1160 if (ffs_inode_hash_lock < 0)
1161 wakeup(&ffs_inode_hash_lock);
1162 ffs_inode_hash_lock = 0;
1164 FREE(ip, ump->um_malloctype);
1167 bzero((caddr_t)ip, sizeof(struct inode));
1168 lockinit(&ip->i_lock, 0, "inode", VLKTIMEOUT, LK_CANRECURSE);
1171 * FFS supports lock sharing in the stack of vnodes
1173 vp->v_vnlock = &ip->i_lock;
1175 ip->i_fs = fs = ump->um_fs;
1181 for (i = 0; i < MAXQUOTAS; i++)
1182 ip->i_dquot[i] = NODQUOT;
1186 * Put it onto its hash chain and lock it so that other requests for
1187 * this inode will block if they arrive while we are sleeping waiting
1188 * for old data structures to be purged or for the contents of the
1189 * disk portion of this inode to be read.
1193 if (ffs_inode_hash_lock < 0)
1194 wakeup(&ffs_inode_hash_lock);
1195 ffs_inode_hash_lock = 0;
1197 /* Read in the disk contents for the inode, copy into the inode. */
1198 error = bread(ump->um_devvp, fsbtodb(fs, ino_to_fsba(fs, ino)),
1199 (int)fs->fs_bsize, &bp);
1202 * The inode does not contain anything useful, so it would
1203 * be misleading to leave it on its hash chain. With mode
1204 * still zero, it will be unlinked and returned to the free
1212 ip->i_din = *((struct dinode *)bp->b_data + ino_to_fsbo(fs, ino));
1213 if (DOINGSOFTDEP(vp))
1214 softdep_load_inodeblock(ip);
1216 ip->i_effnlink = ip->i_nlink;
1220 * Initialize the vnode from the inode, check for aliases.
1221 * Note that the underlying vnode may have changed.
1223 error = ufs_vinit(mp, ffs_specop_p, ffs_fifoop_p, &vp);
1230 * Finish inode initialization now that aliasing has been resolved.
1232 ip->i_devvp = ump->um_devvp;
1235 * Set up a generation number for this inode if it does not
1236 * already have one. This should only happen on old filesystems.
1238 if (ip->i_gen == 0) {
1239 ip->i_gen = random() / 2 + 1;
1240 if ((vp->v_mount->mnt_flag & MNT_RDONLY) == 0)
1241 ip->i_flag |= IN_MODIFIED;
1244 * Ensure that uid and gid are correct. This is a temporary
1245 * fix until fsck has been changed to do the update.
1247 if (fs->fs_inodefmt < FS_44INODEFMT) { /* XXX */
1248 ip->i_uid = ip->i_din.di_ouid; /* XXX */
1249 ip->i_gid = ip->i_din.di_ogid; /* XXX */
1257 * File handle to vnode
1259 * Have to be really careful about stale file handles:
1260 * - check that the inode number is valid
1261 * - call ffs_vget() to get the locked inode
1262 * - check for an unallocated inode (i_mode == 0)
1263 * - check that the given client host has export rights and return
1264 * those rights via. exflagsp and credanonp
1267 ffs_fhtovp(mp, fhp, vpp)
1275 ufhp = (struct ufid *)fhp;
1276 fs = VFSTOUFS(mp)->um_fs;
1277 if (ufhp->ufid_ino < ROOTINO ||
1278 ufhp->ufid_ino >= fs->fs_ncg * fs->fs_ipg)
1280 return (ufs_fhtovp(mp, ufhp, vpp));
1284 * Vnode pointer to File handle
1296 ufhp = (struct ufid *)fhp;
1297 ufhp->ufid_len = sizeof(struct ufid);
1298 ufhp->ufid_ino = ip->i_number;
1299 ufhp->ufid_gen = ip->i_gen;
1304 * Initialize the filesystem; just use ufs_init.
1308 struct vfsconf *vfsp;
1311 softdep_initialize();
1312 return (ufs_init(vfsp));
1316 * Write a superblock and associated information back to disk.
1319 ffs_sbupdate(mp, waitfor)
1320 struct ufsmount *mp;
1323 struct fs *dfs, *fs = mp->um_fs;
1327 int i, size, error, allerror = 0;
1330 * First write back the summary information.
1332 blks = howmany(fs->fs_cssize, fs->fs_fsize);
1334 for (i = 0; i < blks; i += fs->fs_frag) {
1335 size = fs->fs_bsize;
1336 if (i + fs->fs_frag > blks)
1337 size = (blks - i) * fs->fs_fsize;
1338 bp = getblk(mp->um_devvp, fsbtodb(fs, fs->fs_csaddr + i),
1340 bcopy(space, bp->b_data, (u_int)size);
1341 space = (char *)space + size;
1342 if (waitfor != MNT_WAIT)
1344 else if ((error = bwrite(bp)) != 0)
1348 * Now write back the superblock itself. If any errors occurred
1349 * up to this point, then fail so that the superblock avoids
1350 * being written out as clean.
1354 bp = getblk(mp->um_devvp, SBLOCK, (int)fs->fs_sbsize, 0, 0);
1356 fs->fs_time = time_second;
1357 bcopy((caddr_t)fs, bp->b_data, (u_int)fs->fs_sbsize);
1358 /* Restore compatibility to old file systems. XXX */
1359 dfs = (struct fs *)bp->b_data; /* XXX */
1360 if (fs->fs_postblformat == FS_42POSTBLFMT) /* XXX */
1361 dfs->fs_nrpos = -1; /* XXX */
1362 if (fs->fs_inodefmt < FS_44INODEFMT) { /* XXX */
1363 int32_t *lp, tmp; /* XXX */
1365 lp = (int32_t *)&dfs->fs_qbmask; /* XXX */
1366 tmp = lp[4]; /* XXX */
1367 for (i = 4; i > 0; i--) /* XXX */
1368 lp[i] = lp[i-1]; /* XXX */
1369 lp[0] = tmp; /* XXX */
1371 dfs->fs_maxfilesize = mp->um_savedmaxfilesize; /* XXX */
1372 if (waitfor != MNT_WAIT)
1374 else if ((error = bwrite(bp)) != 0)