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 1.23 2004/08/24 14:01:57 drhodus 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);
93 extern struct vnodeopv_entry_desc ffs_vnodeop_entries[];
94 extern struct vnodeopv_entry_desc ffs_specop_entries[];
95 extern struct vnodeopv_entry_desc ffs_fifoop_entries[];
101 * Called when mounting local physical media
105 * mp mount point structure
106 * path NULL (flag for root mount!!!)
109 * p process (user credentials check [statfs])
112 * mp mount point structure
113 * path path to mount point
114 * data pointer to argument struct in user space
115 * ndp mount point namei() return (used for
116 * credentials on reload), reused to look
118 * p process (user credentials check)
121 * !0 error number (errno.h)
126 * mount point is locked
128 * mount point is locked
131 * A NULL path can be used for a flag since the mount
132 * system call will fail with EFAULT in copyinstr in
133 * namei() if it is a genuine NULL from the user.
136 ffs_mount(struct mount *mp, /* mount struct pointer */
137 char *path, /* path to mount point */
138 caddr_t data, /* arguments to FS specific mount */
139 struct nameidata *ndp, /* mount point credentials */
140 struct thread *td) /* process requesting mount */
146 struct ufs_args args;
147 struct ufsmount *ump = 0;
149 int error, flags, ronly = 0;
153 KKASSERT(td->td_proc);
154 cred = td->td_proc->p_ucred;
157 * Use NULL path to flag a root mount
162 * Mounting root filesystem
166 if ((err = bdevvp(rootdev, &rootvp))) {
167 printf("ffs_mountroot: can't find rootvp\n");
171 if( ( err = ffs_mountfs(rootvp, mp, td, M_FFSNODE)) != 0) {
172 /* fs specific cleanup (if any)*/
176 goto dostatfs; /* success*/
182 * Mounting non-root filesystem or updating a filesystem
186 /* copy in user arguments*/
187 err = copyin(data, (caddr_t)&args, sizeof (struct ufs_args));
189 goto error_1; /* can't get arguments*/
192 * If updating, check whether changing from read-only to
193 * read/write; if there is no device name, that's all we do.
195 if (mp->mnt_flag & MNT_UPDATE) {
198 devvp = ump->um_devvp;
200 ronly = fs->fs_ronly; /* MNT_RELOAD might change this */
201 if (ronly == 0 && (mp->mnt_flag & MNT_RDONLY)) {
203 * Flush any dirty data.
205 VFS_SYNC(mp, MNT_WAIT, td);
207 * Check for and optionally get rid of files open
211 if (mp->mnt_flag & MNT_FORCE)
213 if (mp->mnt_flag & MNT_SOFTDEP) {
214 err = softdep_flushfiles(mp, flags, td);
216 err = ffs_flushfiles(mp, flags, td);
220 if (!err && (mp->mnt_flag & MNT_RELOAD))
221 err = ffs_reload(mp, ndp->ni_cnd.cn_cred, td);
225 if (ronly && (mp->mnt_kern_flag & MNTK_WANTRDWR)) {
227 * If upgrade to read-write by non-root, then verify
228 * that user has necessary permissions on the device.
230 if (cred->cr_uid != 0) {
231 vn_lock(devvp, NULL, LK_EXCLUSIVE | LK_RETRY, td);
232 if ((error = VOP_ACCESS(devvp, VREAD | VWRITE,
234 VOP_UNLOCK(devvp, NULL, 0, td);
237 VOP_UNLOCK(devvp, NULL, 0, td);
240 fs->fs_flags &= ~FS_UNCLEAN;
241 if (fs->fs_clean == 0) {
242 fs->fs_flags |= FS_UNCLEAN;
243 if (mp->mnt_flag & MNT_FORCE) {
245 "WARNING: %s was not properly dismounted\n",
249 "WARNING: R/W mount of %s denied. Filesystem is not clean - run fsck\n",
256 /* check to see if we need to start softdep */
257 if (fs->fs_flags & FS_DOSOFTDEP) {
258 err = softdep_mount(devvp, mp, fs);
266 * Soft updates is incompatible with "async",
267 * so if we are doing softupdates stop the user
268 * from setting the async flag in an update.
269 * Softdep_mount() clears it in an initial mount
272 if (mp->mnt_flag & MNT_SOFTDEP) {
273 mp->mnt_flag &= ~MNT_ASYNC;
275 /* if not updating name...*/
276 if (args.fspec == 0) {
278 * Process export requests. Jumping to "success"
279 * will return the vfs_export() error code.
281 err = vfs_export(mp, &ump->um_export, &args.export);
287 * Not an update, or updating the name: look up the name
288 * and verify that it refers to a sensible block device.
290 NDINIT(ndp, NAMEI_LOOKUP, CNP_FOLLOW, UIO_USERSPACE, args.fspec, td);
293 /* can't get devvp!*/
297 NDFREE(ndp, NDF_ONLY_PNBUF);
300 if (!vn_isdisk(devvp, &err))
304 * If mount by non-root, then verify that user has necessary
305 * permissions on the device.
307 if (cred->cr_uid != 0) {
309 if ((mp->mnt_flag & MNT_RDONLY) == 0)
310 accessmode |= VWRITE;
311 vn_lock(devvp, NULL, LK_EXCLUSIVE | LK_RETRY, td);
312 if ((error = VOP_ACCESS(devvp, accessmode, cred, td)) != 0) {
316 VOP_UNLOCK(devvp, NULL, 0, td);
319 if (mp->mnt_flag & MNT_UPDATE) {
323 * If it's not the same vnode, or at least the same device
324 * then it's not correct. NOTE: devvp->v_rdev may be NULL
325 * since we haven't opened it, so we compare udev instead.
328 if (devvp != ump->um_devvp) {
329 if (devvp->v_udev == ump->um_devvp->v_udev) {
332 printf("cannot update mount, udev does"
333 " not match %08x vs %08x\n",
334 devvp->v_udev, ump->um_devvp->v_udev);
335 err = EINVAL; /* needs translation */
341 * Update device name only on success
344 /* Save "mounted from" info for mount point (NULL pad)*/
345 copyinstr( args.fspec,
346 mp->mnt_stat.f_mntfromname,
349 bzero( mp->mnt_stat.f_mntfromname + size, MNAMELEN - size);
359 * Since this is a new mount, we want the names for
360 * the device and the mount point copied in. If an
361 * error occurs, the mountpoint is discarded by the
364 /* Save "last mounted on" info for mount point (NULL pad)*/
365 copyinstr( path, /* mount point*/
366 mp->mnt_stat.f_mntonname, /* save area*/
367 MNAMELEN - 1, /* max size*/
368 &size); /* real size*/
369 bzero( mp->mnt_stat.f_mntonname + size, MNAMELEN - size);
371 /* Save "mounted from" info for mount point (NULL pad)*/
372 copyinstr( args.fspec, /* device name*/
373 mp->mnt_stat.f_mntfromname, /* save area*/
374 MNAMELEN - 1, /* max size*/
375 &size); /* real size*/
376 bzero( mp->mnt_stat.f_mntfromname + size, MNAMELEN - size);
378 err = ffs_mountfs(devvp, mp, td, M_FFSNODE);
386 * Initialize FS stat information in mount struct; uses both
387 * mp->mnt_stat.f_mntonname and mp->mnt_stat.f_mntfromname
389 * This code is common to root and non-root mounts
391 (void)VFS_STATFS(mp, &mp->mnt_stat, td);
396 error_2: /* error with devvp held*/
398 /* release devvp before failing*/
401 error_1: /* no state to back out*/
404 if (!err && path && (mp->mnt_flag & MNT_UPDATE)) {
405 /* Update clean flag after changing read-onlyness. */
407 if (ronly != fs->fs_ronly) {
408 fs->fs_ronly = ronly;
409 fs->fs_clean = ronly &&
410 (fs->fs_flags & FS_UNCLEAN) == 0 ? 1 : 0;
411 ffs_sbupdate(ump, MNT_WAIT);
418 * Reload all incore data for a filesystem (used after running fsck on
419 * the root filesystem and finding things to fix). The filesystem must
420 * be mounted read-only.
422 * Things to do to update the mount:
423 * 1) invalidate all cached meta-data.
424 * 2) re-read superblock from disk.
425 * 3) re-read summary information from disk.
426 * 4) invalidate all inactive vnodes.
427 * 5) invalidate all cached file data.
428 * 6) re-read inode data for all active vnodes.
431 static int ffs_reload_scan1(struct mount *mp, struct vnode *vp, void *data);
432 static int ffs_reload_scan2(struct mount *mp, struct vnode *vp,
433 lwkt_tokref_t vlock, void *data);
445 ffs_reload(struct mount *mp, struct ucred *cred, struct thread *td)
450 struct fs *fs, *newfs;
451 struct partinfo dpart;
453 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, NULL, LK_EXCLUSIVE | LK_RETRY, td);
465 error = vinvalbuf(devvp, 0, td, 0, 0);
466 VOP_UNLOCK(devvp, NULL, 0, td);
468 panic("ffs_reload: dirty1");
472 * Only VMIO the backing device if the backing device is a real
473 * block device. See ffs_mountmfs() for more details.
475 if (devvp->v_tag != VT_MFS && vn_isdisk(devvp, NULL)) {
476 vn_lock(devvp, NULL, LK_EXCLUSIVE | LK_RETRY, td);
477 vfs_object_create(devvp, td);
478 lwkt_gettoken(&vlock, devvp->v_interlock);
479 VOP_UNLOCK(devvp, &vlock, LK_INTERLOCK, td);
483 * Step 2: re-read superblock from disk.
485 if (VOP_IOCTL(devvp, DIOCGPART, (caddr_t)&dpart, FREAD, NOCRED, td) != 0)
488 size = dpart.disklab->d_secsize;
489 if ((error = bread(devvp, (ufs_daddr_t)(SBOFF/size), SBSIZE, &bp)) != 0)
494 newfs = (struct fs *)bp->b_data;
495 if (newfs->fs_magic != FS_MAGIC || newfs->fs_bsize > MAXBSIZE ||
496 newfs->fs_bsize < sizeof(struct fs)) {
498 return (EIO); /* XXX needs translation */
500 fs = VFSTOUFS(mp)->um_fs;
502 * Copy pointer fields back into superblock before copying in XXX
503 * new superblock. These should really be in the ufsmount. XXX
504 * Note that important parameters (eg fs_ncg) are unchanged.
506 newfs->fs_csp = fs->fs_csp;
507 newfs->fs_maxcluster = fs->fs_maxcluster;
508 newfs->fs_contigdirs = fs->fs_contigdirs;
509 /* The filesystem is still read-only. */
511 bcopy(newfs, fs, (uint)fs->fs_sbsize);
512 if (fs->fs_sbsize < SBSIZE)
513 bp->b_flags |= B_INVAL;
515 mp->mnt_maxsymlinklen = fs->fs_maxsymlinklen;
517 /* An old fsck may have zeroed these fields, so recheck them. */
518 if (fs->fs_avgfilesize <= 0) /* XXX */
519 fs->fs_avgfilesize = AVFILESIZ; /* XXX */
520 if (fs->fs_avgfpdir <= 0) /* XXX */
521 fs->fs_avgfpdir = AFPDIR; /* XXX */
524 * Step 3: re-read summary information from disk.
526 blks = howmany(fs->fs_cssize, fs->fs_fsize);
528 for (i = 0; i < blks; i += fs->fs_frag) {
530 if (i + fs->fs_frag > blks)
531 size = (blks - i) * fs->fs_fsize;
532 if (error = bread(devvp, fsbtodb(fs, fs->fs_csaddr + i), size,
537 bcopy(bp->b_data, space, (uint)size);
538 space = (char *)space + size;
542 * We no longer know anything about clusters per cylinder group.
544 if (fs->fs_contigsumsize > 0) {
545 lp = fs->fs_maxcluster;
546 for (i = 0; i < fs->fs_ncg; i++)
547 *lp++ = fs->fs_contigsumsize;
552 scaninfo.devvp = devvp;
554 while (error == 0 && scaninfo.rescan) {
556 error = vmntvnodescan(mp, ffs_reload_scan1,
557 ffs_reload_scan2, &scaninfo);
563 ffs_reload_scan1(struct mount *mp, struct vnode *vp, void *data)
565 struct scaninfo *info = data;
568 * Step 4: invalidate all inactive vnodes.
570 if (vrecycle(vp, NULL, info->td)) {
572 return(-1); /* continue loop, do not call scan2 */
578 ffs_reload_scan2(struct mount *mp, struct vnode *vp, lwkt_tokref_t vlock, void *data)
580 struct scaninfo *info = data;
586 * Step 5: invalidate all cached file data.
588 if (vget(vp, vlock, LK_EXCLUSIVE | LK_INTERLOCK, info->td)) {
592 if (vinvalbuf(vp, 0, info->td, 0, 0))
593 panic("ffs_reload: dirty2");
595 * Step 6: re-read inode data for all active vnodes.
598 error = bread(info->devvp,
599 fsbtodb(info->fs, ino_to_fsba(info->fs, ip->i_number)),
600 (int)info->fs->fs_bsize, &bp);
606 ip->i_din = *((struct dinode *)bp->b_data +
607 ino_to_fsbo(info->fs, ip->i_number));
608 ip->i_effnlink = ip->i_nlink;
615 * Common code for mount and mountroot
618 ffs_mountfs(struct vnode *devvp, struct mount *mp, struct thread *td,
619 struct malloc_type *malloctype)
621 struct ufsmount *ump;
625 struct partinfo dpart;
627 int error, i, blks, size, ronly;
630 uint64_t maxfilesize; /* XXX */
634 * Disallow multiple mounts of the same device.
635 * Disallow mounting of a device that is currently in use
636 * (except for root, which might share swap device for miniroot).
637 * Flush out any old buffers remaining from a previous use.
639 error = vfs_mountedon(devvp);
642 if (count_udev(devvp->v_udev) > 0 && devvp != rootvp)
644 vn_lock(devvp, NULL, LK_EXCLUSIVE | LK_RETRY, td);
645 error = vinvalbuf(devvp, V_SAVE, td, 0, 0);
646 VOP_UNLOCK(devvp, NULL, 0, td);
651 * Only VMIO the backing device if the backing device is a real
652 * block device. This excludes the original MFS implementation.
653 * Note that it is optional that the backing device be VMIOed. This
654 * increases the opportunity for metadata caching.
656 if (devvp->v_tag != VT_MFS && vn_isdisk(devvp, NULL)) {
657 vn_lock(devvp, NULL, LK_EXCLUSIVE | LK_RETRY, td);
658 vfs_object_create(devvp, td);
659 lwkt_gettoken(&vlock, devvp->v_interlock);
660 VOP_UNLOCK(devvp, &vlock, LK_INTERLOCK, td);
663 ronly = (mp->mnt_flag & MNT_RDONLY) != 0;
664 vn_lock(devvp, NULL, LK_EXCLUSIVE | LK_RETRY, td);
665 error = VOP_OPEN(devvp, ronly ? FREAD : FREAD|FWRITE, FSCRED, td);
666 VOP_UNLOCK(devvp, NULL, 0, td);
670 if (dev->si_iosize_max != 0)
671 mp->mnt_iosize_max = dev->si_iosize_max;
672 if (mp->mnt_iosize_max > MAXPHYS)
673 mp->mnt_iosize_max = MAXPHYS;
675 if (VOP_IOCTL(devvp, DIOCGPART, (caddr_t)&dpart, FREAD, proc0.p_ucred, td) != 0)
678 size = dpart.disklab->d_secsize;
682 if ((error = bread(devvp, SBLOCK, SBSIZE, &bp)) != 0)
684 fs = (struct fs *)bp->b_data;
685 if (fs->fs_magic != FS_MAGIC || fs->fs_bsize > MAXBSIZE ||
686 fs->fs_bsize < sizeof(struct fs)) {
687 error = EINVAL; /* XXX needs translation */
691 fs->fs_flags &= ~FS_UNCLEAN;
692 if (fs->fs_clean == 0) {
693 fs->fs_flags |= FS_UNCLEAN;
694 if (ronly || (mp->mnt_flag & MNT_FORCE)) {
696 "WARNING: %s was not properly dismounted\n",
700 "WARNING: R/W mount of %s denied. Filesystem is not clean - run fsck\n",
706 /* XXX updating 4.2 FFS superblocks trashes rotational layout tables */
707 if (fs->fs_postblformat == FS_42POSTBLFMT && !ronly) {
708 error = EROFS; /* needs translation */
711 ump = malloc(sizeof *ump, M_UFSMNT, M_WAITOK);
712 bzero((caddr_t)ump, sizeof *ump);
713 ump->um_malloctype = malloctype;
714 ump->um_i_effnlink_valid = 1;
715 ump->um_fs = malloc((u_long)fs->fs_sbsize, M_UFSMNT,
717 ump->um_blkatoff = ffs_blkatoff;
718 ump->um_truncate = ffs_truncate;
719 ump->um_update = ffs_update;
720 ump->um_valloc = ffs_valloc;
721 ump->um_vfree = ffs_vfree;
722 bcopy(bp->b_data, ump->um_fs, (uint)fs->fs_sbsize);
723 if (fs->fs_sbsize < SBSIZE)
724 bp->b_flags |= B_INVAL;
728 fs->fs_ronly = ronly;
729 size = fs->fs_cssize;
730 blks = howmany(size, fs->fs_fsize);
731 if (fs->fs_contigsumsize > 0)
732 size += fs->fs_ncg * sizeof(int32_t);
733 size += fs->fs_ncg * sizeof(uint8_t);
734 space = malloc((u_long)size, M_UFSMNT, M_WAITOK);
736 for (i = 0; i < blks; i += fs->fs_frag) {
738 if (i + fs->fs_frag > blks)
739 size = (blks - i) * fs->fs_fsize;
740 if ((error = bread(devvp, fsbtodb(fs, fs->fs_csaddr + i), size,
742 free(fs->fs_csp, M_UFSMNT);
745 bcopy(bp->b_data, space, (uint)size);
746 space = (char *)space + size;
750 if (fs->fs_contigsumsize > 0) {
751 fs->fs_maxcluster = lp = space;
752 for (i = 0; i < fs->fs_ncg; i++)
753 *lp++ = fs->fs_contigsumsize;
756 size = fs->fs_ncg * sizeof(uint8_t);
757 fs->fs_contigdirs = (uint8_t *)space;
758 bzero(fs->fs_contigdirs, size);
759 /* Compatibility for old filesystems XXX */
760 if (fs->fs_avgfilesize <= 0) /* XXX */
761 fs->fs_avgfilesize = AVFILESIZ; /* XXX */
762 if (fs->fs_avgfpdir <= 0) /* XXX */
763 fs->fs_avgfpdir = AFPDIR; /* XXX */
764 mp->mnt_data = (qaddr_t)ump;
765 mp->mnt_stat.f_fsid.val[0] = fs->fs_id[0];
766 mp->mnt_stat.f_fsid.val[1] = fs->fs_id[1];
767 if (fs->fs_id[0] == 0 || fs->fs_id[1] == 0 ||
768 vfs_getvfs(&mp->mnt_stat.f_fsid))
770 mp->mnt_maxsymlinklen = fs->fs_maxsymlinklen;
771 mp->mnt_flag |= MNT_LOCAL;
774 ump->um_devvp = devvp;
775 ump->um_nindir = fs->fs_nindir;
776 ump->um_bptrtodb = fs->fs_fsbtodb;
777 ump->um_seqinc = fs->fs_frag;
778 for (i = 0; i < MAXQUOTAS; i++)
779 ump->um_quotas[i] = NULLVP;
780 dev->si_mountpoint = mp;
784 * Set FS local "last mounted on" information (NULL pad)
786 copystr( mp->mnt_stat.f_mntonname, /* mount point*/
787 fs->fs_fsmnt, /* copy area*/
788 sizeof(fs->fs_fsmnt) - 1, /* max size*/
789 &strsize); /* real size*/
790 bzero( fs->fs_fsmnt + strsize, sizeof(fs->fs_fsmnt) - strsize);
792 if( mp->mnt_flag & MNT_ROOTFS) {
794 * Root mount; update timestamp in mount structure.
795 * this will be used by the common root mount code
796 * to update the system clock.
798 mp->mnt_time = fs->fs_time;
801 ump->um_savedmaxfilesize = fs->fs_maxfilesize; /* XXX */
802 maxfilesize = (uint64_t)0x40000000 * fs->fs_bsize - 1; /* XXX */
803 /* Enforce limit caused by vm object backing (32 bits vm_pindex_t). */
804 if (maxfilesize > (uint64_t)0x80000000u * PAGE_SIZE - 1)
805 maxfilesize = (uint64_t)0x80000000u * PAGE_SIZE - 1;
806 if (fs->fs_maxfilesize > maxfilesize) /* XXX */
807 fs->fs_maxfilesize = maxfilesize; /* XXX */
809 if ((fs->fs_flags & FS_DOSOFTDEP) &&
810 (error = softdep_mount(devvp, mp, fs)) != 0) {
811 free(fs->fs_csp, M_UFSMNT);
816 (void) ffs_sbupdate(ump, MNT_WAIT);
818 vfs_add_vnodeops(&mp->mnt_vn_ops, ffs_vnodeop_entries);
819 vfs_add_vnodeops(&mp->mnt_vn_spec_ops, ffs_specop_entries);
820 vfs_add_vnodeops(&mp->mnt_vn_fifo_ops, ffs_fifoop_entries);
824 dev->si_mountpoint = NULL;
827 VOP_CLOSE(devvp, ronly ? FREAD : FREAD|FWRITE, td);
829 free(ump->um_fs, M_UFSMNT);
831 mp->mnt_data = (qaddr_t)0;
837 * Sanity checks for old filesystems.
839 * XXX - goes away some day.
842 ffs_oldfscompat(struct fs *fs)
844 fs->fs_npsect = max(fs->fs_npsect, fs->fs_nsect); /* XXX */
845 fs->fs_interleave = max(fs->fs_interleave, 1); /* XXX */
846 if (fs->fs_postblformat == FS_42POSTBLFMT) /* XXX */
847 fs->fs_nrpos = 8; /* XXX */
848 if (fs->fs_inodefmt < FS_44INODEFMT) { /* XXX */
851 uint64_t sizepb = fs->fs_bsize; /* XXX */
853 fs->fs_maxfilesize = fs->fs_bsize * NDADDR - 1; /* XXX */
854 for (i = 0; i < NIADDR; i++) { /* XXX */
855 sizepb *= NINDIR(fs); /* XXX */
856 fs->fs_maxfilesize += sizepb; /* XXX */
859 fs->fs_maxfilesize = (u_quad_t) 1LL << 39;
860 fs->fs_qbmask = ~fs->fs_bmask; /* XXX */
861 fs->fs_qfmask = ~fs->fs_fmask; /* XXX */
867 * unmount system call
870 ffs_unmount(struct mount *mp, int mntflags, struct thread *td)
872 struct ufsmount *ump;
877 if (mntflags & MNT_FORCE) {
880 if (mp->mnt_flag & MNT_SOFTDEP) {
881 if ((error = softdep_flushfiles(mp, flags, td)) != 0)
884 if ((error = ffs_flushfiles(mp, flags, td)) != 0)
889 if (fs->fs_ronly == 0) {
890 fs->fs_clean = fs->fs_flags & FS_UNCLEAN ? 0 : 1;
891 error = ffs_sbupdate(ump, MNT_WAIT);
897 ump->um_devvp->v_rdev->si_mountpoint = NULL;
899 vinvalbuf(ump->um_devvp, V_SAVE, td, 0, 0);
900 error = VOP_CLOSE(ump->um_devvp, fs->fs_ronly ? FREAD : FREAD|FWRITE, td);
902 vrele(ump->um_devvp);
904 free(fs->fs_csp, M_UFSMNT);
907 mp->mnt_data = (qaddr_t)0;
908 mp->mnt_flag &= ~MNT_LOCAL;
913 * Flush out all the files in a filesystem.
916 ffs_flushfiles(struct mount *mp, int flags, struct thread *td)
918 struct ufsmount *ump;
923 if (mp->mnt_flag & MNT_QUOTA) {
925 error = vflush(mp, 0, SKIPSYSTEM|flags);
928 /* Find out how many quota files we have open. */
929 for (i = 0; i < MAXQUOTAS; i++) {
930 if (ump->um_quotas[i] == NULLVP)
935 * Here we fall through to vflush again to ensure
936 * that we have gotten rid of all the system vnodes.
941 * Flush all the files.
943 if ((error = vflush(mp, 0, flags)) != 0)
946 * Flush filesystem metadata.
948 vn_lock(ump->um_devvp, NULL, LK_EXCLUSIVE | LK_RETRY, td);
949 error = VOP_FSYNC(ump->um_devvp, MNT_WAIT, td);
950 VOP_UNLOCK(ump->um_devvp, NULL, 0, td);
955 * Get filesystem statistics.
958 ffs_statfs(struct mount *mp, struct statfs *sbp, struct thread *td)
960 struct ufsmount *ump;
965 if (fs->fs_magic != FS_MAGIC)
967 sbp->f_bsize = fs->fs_fsize;
968 sbp->f_iosize = fs->fs_bsize;
969 sbp->f_blocks = fs->fs_dsize;
970 sbp->f_bfree = fs->fs_cstotal.cs_nbfree * fs->fs_frag +
971 fs->fs_cstotal.cs_nffree;
972 sbp->f_bavail = freespace(fs, fs->fs_minfree);
973 sbp->f_files = fs->fs_ncg * fs->fs_ipg - ROOTINO;
974 sbp->f_ffree = fs->fs_cstotal.cs_nifree;
975 if (sbp != &mp->mnt_stat) {
976 sbp->f_type = mp->mnt_vfc->vfc_typenum;
977 bcopy((caddr_t)mp->mnt_stat.f_mntonname,
978 (caddr_t)&sbp->f_mntonname[0], MNAMELEN);
979 bcopy((caddr_t)mp->mnt_stat.f_mntfromname,
980 (caddr_t)&sbp->f_mntfromname[0], MNAMELEN);
986 * Go through the disk queues to initiate sandbagged IO;
987 * go through the inodes to write those that have been modified;
988 * initiate the writing of the super block if it has been modified.
990 * Note: we are always called with the filesystem marked `MPBUSY'.
994 static int ffs_sync_scan1(struct mount *mp, struct vnode *vp, void *data);
995 static int ffs_sync_scan2(struct mount *mp, struct vnode *vp,
996 lwkt_tokref_t vlock, void *data);
999 ffs_sync(struct mount *mp, int waitfor, struct thread *td)
1001 struct ufsmount *ump = VFSTOUFS(mp);
1004 struct scaninfo scaninfo;
1007 if (fs->fs_fmod != 0 && fs->fs_ronly != 0) { /* XXX */
1008 printf("fs = %s\n", fs->fs_fsmnt);
1009 panic("ffs_sync: rofs mod");
1013 * Write back each (modified) inode.
1015 scaninfo.allerror = 0;
1016 scaninfo.rescan = 1;
1017 scaninfo.waitfor = waitfor;
1018 while (scaninfo.rescan) {
1019 scaninfo.rescan = 0;
1020 vmntvnodescan(mp, ffs_sync_scan1, ffs_sync_scan2, &scaninfo);
1024 * Force stale filesystem control information to be flushed.
1026 if (waitfor != MNT_LAZY) {
1027 if (ump->um_mountp->mnt_flag & MNT_SOFTDEP)
1028 waitfor = MNT_NOWAIT;
1029 vn_lock(ump->um_devvp, NULL, LK_EXCLUSIVE | LK_RETRY, td);
1030 if ((error = VOP_FSYNC(ump->um_devvp, waitfor, td)) != 0)
1031 scaninfo.allerror = error;
1032 VOP_UNLOCK(ump->um_devvp, NULL, 0, td);
1038 * Write back modified superblock.
1040 if (fs->fs_fmod != 0 && (error = ffs_sbupdate(ump, waitfor)) != 0)
1041 scaninfo.allerror = error;
1042 return (scaninfo.allerror);
1046 ffs_sync_scan1(struct mount *mp, struct vnode *vp, void *data)
1051 * Depend on the mount list's vnode lock to keep things stable
1052 * enough for a quick test. Since there might be hundreds of
1053 * thousands of vnodes, we cannot afford even a subroutine
1054 * call unless there's a good chance that we have work to do.
1057 /* Restart out whole search if this guy is locked
1058 * or is being reclaimed.
1060 if (vp->v_type == VNON || ((ip->i_flag &
1061 (IN_ACCESS | IN_CHANGE | IN_MODIFIED | IN_UPDATE)) == 0 &&
1062 TAILQ_EMPTY(&vp->v_dirtyblkhd))) {
1069 ffs_sync_scan2(struct mount *mp, struct vnode *vp,
1070 lwkt_tokref_t vlock, void *data)
1072 struct scaninfo *info = data;
1073 thread_t td = curthread; /* XXX */
1078 * We have to recheck after having obtained the vnode interlock.
1081 if (vp->v_type == VNON || ((ip->i_flag &
1082 (IN_ACCESS | IN_CHANGE | IN_MODIFIED | IN_UPDATE)) == 0 &&
1083 TAILQ_EMPTY(&vp->v_dirtyblkhd))) {
1084 lwkt_reltoken(vlock);
1087 if (vp->v_type != VCHR) {
1088 error = vget(vp, vlock, LK_INTERLOCK|LK_EXCLUSIVE|LK_NOWAIT, td);
1090 if (error == ENOENT)
1093 if ((error = VOP_FSYNC(vp, info->waitfor, td)) != 0)
1094 info->allerror = error;
1095 VOP_UNLOCK(vp, NULL, 0, td);
1100 * We must reference the vp to prevent it from
1101 * getting ripped out from under UFS_UPDATE, since
1102 * we are not holding a vnode lock.
1105 lwkt_reltoken(vlock);
1106 /* UFS_UPDATE(vp, waitfor == MNT_WAIT); */
1114 * Look up a FFS dinode number to find its incore vnode, otherwise read it
1115 * in from disk. If it is in core, wait for the lock bit to clear, then
1116 * return the inode locked. Detection and handling of mount points must be
1117 * done by the calling routine.
1119 static int ffs_inode_hash_lock;
1122 ffs_vget(struct mount *mp, ino_t ino, struct vnode **vpp)
1126 struct ufsmount *ump;
1135 if ((*vpp = ufs_ihashget(dev, ino)) != NULL) {
1140 * Lock out the creation of new entries in the FFS hash table in
1141 * case getnewvnode() or MALLOC() blocks, otherwise a duplicate
1144 if (ffs_inode_hash_lock) {
1145 while (ffs_inode_hash_lock) {
1146 ffs_inode_hash_lock = -1;
1147 tsleep(&ffs_inode_hash_lock, 0, "ffsvgt", 0);
1151 ffs_inode_hash_lock = 1;
1154 * If this MALLOC() is performed after the getnewvnode()
1155 * it might block, leaving a vnode with a NULL v_data to be
1156 * found by ffs_sync() if a sync happens to fire right then,
1157 * which will cause a panic because ffs_sync() blindly
1158 * dereferences vp->v_data (as well it should).
1160 MALLOC(ip, struct inode *, sizeof(struct inode),
1161 ump->um_malloctype, M_WAITOK);
1163 /* Allocate a new vnode/inode. */
1164 error = getnewvnode(VT_UFS, mp, mp->mnt_vn_ops, &vp);
1166 if (ffs_inode_hash_lock < 0)
1167 wakeup(&ffs_inode_hash_lock);
1168 ffs_inode_hash_lock = 0;
1170 FREE(ip, ump->um_malloctype);
1173 bzero((caddr_t)ip, sizeof(struct inode));
1174 lockinit(&ip->i_lock, 0, "inode", VLKTIMEOUT, LK_CANRECURSE);
1177 * FFS supports lock sharing in the stack of vnodes
1179 vp->v_vnlock = &ip->i_lock;
1181 ip->i_fs = fs = ump->um_fs;
1187 for (i = 0; i < MAXQUOTAS; i++)
1188 ip->i_dquot[i] = NODQUOT;
1193 * Put it onto its hash chain and lock it so that other requests for
1194 * this inode will block if they arrive while we are sleeping waiting
1195 * for old data structures to be purged or for the contents of the
1196 * disk portion of this inode to be read.
1200 if (ffs_inode_hash_lock < 0)
1201 wakeup(&ffs_inode_hash_lock);
1202 ffs_inode_hash_lock = 0;
1204 /* Read in the disk contents for the inode, copy into the inode. */
1205 error = bread(ump->um_devvp, fsbtodb(fs, ino_to_fsba(fs, ino)),
1206 (int)fs->fs_bsize, &bp);
1209 * The inode does not contain anything useful, so it would
1210 * be misleading to leave it on its hash chain. With mode
1211 * still zero, it will be unlinked and returned to the free
1219 ip->i_din = *((struct dinode *)bp->b_data + ino_to_fsbo(fs, ino));
1220 if (DOINGSOFTDEP(vp))
1221 softdep_load_inodeblock(ip);
1223 ip->i_effnlink = ip->i_nlink;
1227 * Initialize the vnode from the inode, check for aliases.
1228 * Note that the underlying vnode may have changed.
1230 error = ufs_vinit(mp, &vp);
1237 * Finish inode initialization now that aliasing has been resolved.
1239 ip->i_devvp = ump->um_devvp;
1242 * Set up a generation number for this inode if it does not
1243 * already have one. This should only happen on old filesystems.
1245 if (ip->i_gen == 0) {
1246 ip->i_gen = random() / 2 + 1;
1247 if ((vp->v_mount->mnt_flag & MNT_RDONLY) == 0)
1248 ip->i_flag |= IN_MODIFIED;
1251 * Ensure that uid and gid are correct. This is a temporary
1252 * fix until fsck has been changed to do the update.
1254 if (fs->fs_inodefmt < FS_44INODEFMT) { /* XXX */
1255 ip->i_uid = ip->i_din.di_ouid; /* XXX */
1256 ip->i_gid = ip->i_din.di_ogid; /* XXX */
1264 * File handle to vnode
1266 * Have to be really careful about stale file handles:
1267 * - check that the inode number is valid
1268 * - call ffs_vget() to get the locked inode
1269 * - check for an unallocated inode (i_mode == 0)
1270 * - check that the given client host has export rights and return
1271 * those rights via. exflagsp and credanonp
1274 ffs_fhtovp(struct mount *mp, struct fid *fhp, struct vnode **vpp)
1279 ufhp = (struct ufid *)fhp;
1280 fs = VFSTOUFS(mp)->um_fs;
1281 if (ufhp->ufid_ino < ROOTINO ||
1282 ufhp->ufid_ino >= fs->fs_ncg * fs->fs_ipg)
1284 return (ufs_fhtovp(mp, ufhp, vpp));
1288 * Vnode pointer to File handle
1292 ffs_vptofh(struct vnode *vp, struct fid *fhp)
1298 ufhp = (struct ufid *)fhp;
1299 ufhp->ufid_len = sizeof(struct ufid);
1300 ufhp->ufid_ino = ip->i_number;
1301 ufhp->ufid_gen = ip->i_gen;
1306 * Initialize the filesystem; just use ufs_init.
1309 ffs_init(struct vfsconf *vfsp)
1311 softdep_initialize();
1312 return (ufs_init(vfsp));
1316 * Write a superblock and associated information back to disk.
1319 ffs_sbupdate(struct ufsmount *mp, int waitfor)
1321 struct fs *dfs, *fs = mp->um_fs;
1325 int i, size, error, allerror = 0;
1328 * First write back the summary information.
1330 blks = howmany(fs->fs_cssize, fs->fs_fsize);
1332 for (i = 0; i < blks; i += fs->fs_frag) {
1333 size = fs->fs_bsize;
1334 if (i + fs->fs_frag > blks)
1335 size = (blks - i) * fs->fs_fsize;
1336 bp = getblk(mp->um_devvp, fsbtodb(fs, fs->fs_csaddr + i),
1338 bcopy(space, bp->b_data, (uint)size);
1339 space = (char *)space + size;
1340 if (waitfor != MNT_WAIT)
1342 else if ((error = bwrite(bp)) != 0)
1346 * Now write back the superblock itself. If any errors occurred
1347 * up to this point, then fail so that the superblock avoids
1348 * being written out as clean.
1352 bp = getblk(mp->um_devvp, SBLOCK, (int)fs->fs_sbsize, 0, 0);
1354 fs->fs_time = time_second;
1355 bcopy((caddr_t)fs, bp->b_data, (uint)fs->fs_sbsize);
1356 /* Restore compatibility to old filesystems. XXX */
1357 dfs = (struct fs *)bp->b_data; /* XXX */
1358 if (fs->fs_postblformat == FS_42POSTBLFMT) /* XXX */
1359 dfs->fs_nrpos = -1; /* XXX */
1360 if (fs->fs_inodefmt < FS_44INODEFMT) { /* XXX */
1361 int32_t *lp, tmp; /* XXX */
1363 lp = (int32_t *)&dfs->fs_qbmask; /* XXX */
1364 tmp = lp[4]; /* XXX */
1365 for (i = 4; i > 0; i--) /* XXX */
1366 lp[i] = lp[i-1]; /* XXX */
1367 lp[0] = tmp; /* XXX */
1369 dfs->fs_maxfilesize = mp->um_savedmaxfilesize; /* XXX */
1370 if (waitfor != MNT_WAIT)
1372 else if ((error = bwrite(bp)) != 0)