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 1.34 2005/09/17 07:43:12 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 *,struct thread *);
69 static int ffs_oldfscompat (struct fs *);
70 static int ffs_mount (struct mount *, char *, caddr_t, struct thread *);
71 static int ffs_init (struct vfsconf *);
73 static struct vfsops ufs_vfsops = {
74 .vfs_mount = ffs_mount,
75 .vfs_unmount = ffs_unmount,
77 .vfs_quotactl = ufs_quotactl,
78 .vfs_statfs = ffs_statfs,
81 .vfs_fhtovp = ffs_fhtovp,
82 .vfs_checkexp = ufs_check_export,
83 .vfs_vptofh = ffs_vptofh,
85 .vfs_uninit = ufs_uninit
88 VFS_SET(ufs_vfsops, ufs, 0);
90 extern struct vnodeopv_entry_desc ffs_vnodeop_entries[];
91 extern struct vnodeopv_entry_desc ffs_specop_entries[];
92 extern struct vnodeopv_entry_desc ffs_fifoop_entries[];
98 * Called when mounting local physical media
102 * mp mount point structure
103 * path NULL (flag for root mount!!!)
105 * p process (user credentials check [statfs])
108 * mp mount point structure
109 * path path to mount point
110 * data pointer to argument struct in user space
111 * p process (user credentials check)
114 * !0 error number (errno.h)
119 * mount point is locked
121 * mount point is locked
124 * A NULL path can be used for a flag since the mount
125 * system call will fail with EFAULT in copyinstr in
126 * nlookup() if it is a genuine NULL from the user.
129 ffs_mount(struct mount *mp, /* mount struct pointer */
130 char *path, /* path to mount point */
131 caddr_t data, /* arguments to FS specific mount */
132 struct thread *td) /* process requesting mount */
138 struct ufs_args args;
139 struct ufsmount *ump = 0;
141 int flags, ronly = 0;
144 struct nlookupdata nd;
145 struct vnode *rootvp;
147 KKASSERT(td->td_proc);
148 cred = td->td_proc->p_ucred;
152 * Use NULL path to flag a root mount
157 * Mounting root filesystem
161 if ((error = bdevvp(rootdev, &rootvp))) {
162 printf("ffs_mountroot: can't find rootvp\n");
166 if( ( error = ffs_mountfs(rootvp, mp, td, M_FFSNODE)) != 0) {
167 /* fs specific cleanup (if any)*/
171 goto dostatfs; /* success*/
177 * Mounting non-root filesystem or updating a filesystem
181 /* copy in user arguments*/
182 error = copyin(data, (caddr_t)&args, sizeof (struct ufs_args));
184 goto error_1; /* can't get arguments*/
187 * If updating, check whether changing from read-only to
188 * read/write; if there is no device name, that's all we do.
190 if (mp->mnt_flag & MNT_UPDATE) {
193 devvp = ump->um_devvp;
195 ronly = fs->fs_ronly; /* MNT_RELOAD might change this */
196 if (ronly == 0 && (mp->mnt_flag & MNT_RDONLY)) {
198 * Flush any dirty data.
200 VFS_SYNC(mp, MNT_WAIT, td);
202 * Check for and optionally get rid of files open
206 if (mp->mnt_flag & MNT_FORCE)
208 if (mp->mnt_flag & MNT_SOFTDEP) {
209 error = softdep_flushfiles(mp, flags, td);
211 error = ffs_flushfiles(mp, flags, td);
215 if (!error && (mp->mnt_flag & MNT_RELOAD))
216 error = ffs_reload(mp, NULL, td);
220 if (ronly && (mp->mnt_kern_flag & MNTK_WANTRDWR)) {
222 * If upgrade to read-write by non-root, then verify
223 * that user has necessary permissions on the device.
225 if (cred->cr_uid != 0) {
226 vn_lock(devvp, LK_EXCLUSIVE | LK_RETRY, td);
227 if ((error = VOP_ACCESS(devvp, VREAD | VWRITE,
229 VOP_UNLOCK(devvp, 0, td);
232 VOP_UNLOCK(devvp, 0, td);
235 fs->fs_flags &= ~FS_UNCLEAN;
236 if (fs->fs_clean == 0) {
237 fs->fs_flags |= FS_UNCLEAN;
238 if (mp->mnt_flag & MNT_FORCE) {
240 "WARNING: %s was not properly dismounted\n",
244 "WARNING: R/W mount of %s denied. Filesystem is not clean - run fsck\n",
251 /* check to see if we need to start softdep */
252 if (fs->fs_flags & FS_DOSOFTDEP) {
253 error = softdep_mount(devvp, mp, fs);
261 * Soft updates is incompatible with "async",
262 * so if we are doing softupdates stop the user
263 * from setting the async flag in an update.
264 * Softdep_mount() clears it in an initial mount
267 if (mp->mnt_flag & MNT_SOFTDEP) {
268 mp->mnt_flag &= ~MNT_ASYNC;
270 /* if not updating name...*/
271 if (args.fspec == 0) {
273 * Process export requests. Jumping to "success"
274 * will return the vfs_export() error code.
276 error = vfs_export(mp, &ump->um_export, &args.export);
282 * Not an update, or updating the name: look up the name
283 * and verify that it refers to a sensible block device.
286 error = nlookup_init(&nd, args.fspec, UIO_USERSPACE, NLC_FOLLOW);
288 error = nlookup(&nd);
290 error = cache_vref(nd.nl_ncp, nd.nl_cred, &devvp);
295 if (!vn_isdisk(devvp, &error))
299 * If mount by non-root, then verify that user has necessary
300 * permissions on the device.
302 if (cred->cr_uid != 0) {
304 if ((mp->mnt_flag & MNT_RDONLY) == 0)
305 accessmode |= VWRITE;
306 vn_lock(devvp, LK_EXCLUSIVE | LK_RETRY, td);
307 if ((error = VOP_ACCESS(devvp, accessmode, cred, td)) != 0) {
311 VOP_UNLOCK(devvp, 0, td);
314 if (mp->mnt_flag & MNT_UPDATE) {
318 * If it's not the same vnode, or at least the same device
319 * then it's not correct. NOTE: devvp->v_rdev may be NULL
320 * since we haven't opened it, so we compare udev instead.
323 if (devvp != ump->um_devvp) {
324 if (devvp->v_udev == ump->um_devvp->v_udev) {
327 printf("cannot update mount, udev does"
328 " not match %08x vs %08x\n",
329 devvp->v_udev, ump->um_devvp->v_udev);
330 error = EINVAL; /* needs translation */
336 * Update device name only on success
339 /* Save "mounted from" info for mount point (NULL pad)*/
340 copyinstr( args.fspec,
341 mp->mnt_stat.f_mntfromname,
344 bzero( mp->mnt_stat.f_mntfromname + size, MNAMELEN - size);
353 /* Save "mounted from" info for mount point (NULL pad)*/
354 copyinstr( args.fspec, /* device name*/
355 mp->mnt_stat.f_mntfromname, /* save area*/
356 MNAMELEN - 1, /* max size*/
357 &size); /* real size*/
358 bzero( mp->mnt_stat.f_mntfromname + size, MNAMELEN - size);
360 error = ffs_mountfs(devvp, mp, td, M_FFSNODE);
368 * Initialize FS stat information in mount struct; uses
369 * mp->mnt_stat.f_mntfromname.
371 * This code is common to root and non-root mounts
373 (void)VFS_STATFS(mp, &mp->mnt_stat, td);
378 error_2: /* error with devvp held*/
380 /* release devvp before failing*/
383 error_1: /* no state to back out*/
386 if (!error && path && (mp->mnt_flag & MNT_UPDATE)) {
387 /* Update clean flag after changing read-onlyness. */
389 if (ronly != fs->fs_ronly) {
390 fs->fs_ronly = ronly;
391 fs->fs_clean = ronly &&
392 (fs->fs_flags & FS_UNCLEAN) == 0 ? 1 : 0;
393 ffs_sbupdate(ump, MNT_WAIT);
400 * Reload all incore data for a filesystem (used after running fsck on
401 * the root filesystem and finding things to fix). The filesystem must
402 * be mounted read-only.
404 * Things to do to update the mount:
405 * 1) invalidate all cached meta-data.
406 * 2) re-read superblock from disk.
407 * 3) re-read summary information from disk.
408 * 4) invalidate all inactive vnodes.
409 * 5) invalidate all cached file data.
410 * 6) re-read inode data for all active vnodes.
413 static int ffs_reload_scan2(struct mount *mp, struct vnode *vp, void *data);
425 ffs_reload(struct mount *mp, struct ucred *cred, struct thread *td)
430 struct fs *fs, *newfs;
431 struct partinfo dpart;
433 int i, blks, size, error;
434 struct scaninfo scaninfo;
437 if ((mp->mnt_flag & MNT_RDONLY) == 0)
440 * Step 1: invalidate all cached meta-data.
442 devvp = VFSTOUFS(mp)->um_devvp;
443 vn_lock(devvp, LK_EXCLUSIVE | LK_RETRY, td);
444 error = vinvalbuf(devvp, 0, td, 0, 0);
445 VOP_UNLOCK(devvp, 0, td);
447 panic("ffs_reload: dirty1");
451 * Only VMIO the backing device if the backing device is a real
452 * block device. See ffs_mountmfs() for more details.
454 if (devvp->v_tag != VT_MFS && vn_isdisk(devvp, NULL)) {
455 vn_lock(devvp, LK_EXCLUSIVE | LK_RETRY, td);
456 vfs_object_create(devvp, td);
457 VOP_UNLOCK(devvp, 0, td);
461 * Step 2: re-read superblock from disk.
463 if (VOP_IOCTL(devvp, DIOCGPART, (caddr_t)&dpart, FREAD, NOCRED, td) != 0)
466 size = dpart.disklab->d_secsize;
467 if ((error = bread(devvp, (ufs_daddr_t)(SBOFF/size), SBSIZE, &bp)) != 0)
472 newfs = (struct fs *)bp->b_data;
473 if (newfs->fs_magic != FS_MAGIC || newfs->fs_bsize > MAXBSIZE ||
474 newfs->fs_bsize < sizeof(struct fs)) {
476 return (EIO); /* XXX needs translation */
478 fs = VFSTOUFS(mp)->um_fs;
480 * Copy pointer fields back into superblock before copying in XXX
481 * new superblock. These should really be in the ufsmount. XXX
482 * Note that important parameters (eg fs_ncg) are unchanged.
484 newfs->fs_csp = fs->fs_csp;
485 newfs->fs_maxcluster = fs->fs_maxcluster;
486 newfs->fs_contigdirs = fs->fs_contigdirs;
487 /* The filesystem is still read-only. */
489 bcopy(newfs, fs, (uint)fs->fs_sbsize);
490 if (fs->fs_sbsize < SBSIZE)
491 bp->b_flags |= B_INVAL;
493 mp->mnt_maxsymlinklen = fs->fs_maxsymlinklen;
495 /* An old fsck may have zeroed these fields, so recheck them. */
496 if (fs->fs_avgfilesize <= 0) /* XXX */
497 fs->fs_avgfilesize = AVFILESIZ; /* XXX */
498 if (fs->fs_avgfpdir <= 0) /* XXX */
499 fs->fs_avgfpdir = AFPDIR; /* XXX */
502 * Step 3: re-read summary information from disk.
504 blks = howmany(fs->fs_cssize, fs->fs_fsize);
506 for (i = 0; i < blks; i += fs->fs_frag) {
508 if (i + fs->fs_frag > blks)
509 size = (blks - i) * fs->fs_fsize;
510 error = bread(devvp, fsbtodb(fs, fs->fs_csaddr + i), size, &bp);
515 bcopy(bp->b_data, space, (uint)size);
516 space = (char *)space + size;
520 * We no longer know anything about clusters per cylinder group.
522 if (fs->fs_contigsumsize > 0) {
523 lp = fs->fs_maxcluster;
524 for (i = 0; i < fs->fs_ncg; i++)
525 *lp++ = fs->fs_contigsumsize;
530 scaninfo.devvp = devvp;
532 while (error == 0 && scaninfo.rescan) {
534 error = vmntvnodescan(mp, VMSC_GETVX,
535 NULL, ffs_reload_scan2, &scaninfo);
541 ffs_reload_scan2(struct mount *mp, struct vnode *vp, void *data)
543 struct scaninfo *info = data;
551 if (vrecycle(vp, info->td))
554 if (vinvalbuf(vp, 0, info->td, 0, 0))
555 panic("ffs_reload: dirty2");
557 * Step 6: re-read inode data for all active vnodes.
560 error = bread(info->devvp,
561 fsbtodb(info->fs, ino_to_fsba(info->fs, ip->i_number)),
562 (int)info->fs->fs_bsize, &bp);
567 ip->i_din = *((struct dinode *)bp->b_data +
568 ino_to_fsbo(info->fs, ip->i_number));
569 ip->i_effnlink = ip->i_nlink;
575 * Common code for mount and mountroot
578 ffs_mountfs(struct vnode *devvp, struct mount *mp, struct thread *td,
579 struct malloc_type *malloctype)
581 struct ufsmount *ump;
585 struct partinfo dpart;
587 int error, i, blks, size, ronly;
589 uint64_t maxfilesize; /* XXX */
592 * Disallow multiple mounts of the same device.
593 * Disallow mounting of a device that is currently in use
594 * Flush out any old buffers remaining from a previous use.
596 error = vfs_mountedon(devvp);
599 if (count_udev(devvp->v_udev) > 0)
601 vn_lock(devvp, LK_EXCLUSIVE | LK_RETRY, td);
602 error = vinvalbuf(devvp, V_SAVE, td, 0, 0);
603 VOP_UNLOCK(devvp, 0, td);
607 ronly = (mp->mnt_flag & MNT_RDONLY) != 0;
608 vn_lock(devvp, LK_EXCLUSIVE | LK_RETRY, td);
609 error = VOP_OPEN(devvp, ronly ? FREAD : FREAD|FWRITE, FSCRED, NULL, td);
610 VOP_UNLOCK(devvp, 0, td);
614 if (dev->si_iosize_max != 0)
615 mp->mnt_iosize_max = dev->si_iosize_max;
616 if (mp->mnt_iosize_max > MAXPHYS)
617 mp->mnt_iosize_max = MAXPHYS;
620 * Only VMIO the backing device if the backing device is a real
621 * block device. This excludes the original MFS implementation.
622 * Note that it is optional that the backing device be VMIOed. This
623 * increases the opportunity for metadata caching.
625 * This call must be made after the VOP_OPEN.
627 if (devvp->v_tag != VT_MFS && vn_isdisk(devvp, NULL)) {
628 vn_lock(devvp, LK_EXCLUSIVE | LK_RETRY, td);
629 vfs_object_create(devvp, td);
630 VOP_UNLOCK(devvp, 0, td);
633 if (VOP_IOCTL(devvp, DIOCGPART, (caddr_t)&dpart, FREAD, proc0.p_ucred, td) != 0)
636 size = dpart.disklab->d_secsize;
640 if ((error = bread(devvp, SBLOCK, SBSIZE, &bp)) != 0)
642 fs = (struct fs *)bp->b_data;
643 if (fs->fs_magic != FS_MAGIC || fs->fs_bsize > MAXBSIZE ||
644 fs->fs_bsize < sizeof(struct fs)) {
645 error = EINVAL; /* XXX needs translation */
649 fs->fs_flags &= ~FS_UNCLEAN;
650 if (fs->fs_clean == 0) {
651 fs->fs_flags |= FS_UNCLEAN;
652 if (ronly || (mp->mnt_flag & MNT_FORCE)) {
654 "WARNING: %s was not properly dismounted\n",
658 "WARNING: R/W mount of %s denied. Filesystem is not clean - run fsck\n",
664 /* XXX updating 4.2 FFS superblocks trashes rotational layout tables */
665 if (fs->fs_postblformat == FS_42POSTBLFMT && !ronly) {
666 error = EROFS; /* needs translation */
669 ump = malloc(sizeof *ump, M_UFSMNT, M_WAITOK);
670 bzero((caddr_t)ump, sizeof *ump);
671 ump->um_malloctype = malloctype;
672 ump->um_i_effnlink_valid = 1;
673 ump->um_fs = malloc((u_long)fs->fs_sbsize, M_UFSMNT,
675 ump->um_blkatoff = ffs_blkatoff;
676 ump->um_truncate = ffs_truncate;
677 ump->um_update = ffs_update;
678 ump->um_valloc = ffs_valloc;
679 ump->um_vfree = ffs_vfree;
680 bcopy(bp->b_data, ump->um_fs, (uint)fs->fs_sbsize);
681 if (fs->fs_sbsize < SBSIZE)
682 bp->b_flags |= B_INVAL;
686 fs->fs_ronly = ronly;
687 size = fs->fs_cssize;
688 blks = howmany(size, fs->fs_fsize);
689 if (fs->fs_contigsumsize > 0)
690 size += fs->fs_ncg * sizeof(int32_t);
691 size += fs->fs_ncg * sizeof(uint8_t);
692 space = malloc((u_long)size, M_UFSMNT, M_WAITOK);
694 for (i = 0; i < blks; i += fs->fs_frag) {
696 if (i + fs->fs_frag > blks)
697 size = (blks - i) * fs->fs_fsize;
698 if ((error = bread(devvp, fsbtodb(fs, fs->fs_csaddr + i), size,
700 free(fs->fs_csp, M_UFSMNT);
703 bcopy(bp->b_data, space, (uint)size);
704 space = (char *)space + size;
708 if (fs->fs_contigsumsize > 0) {
709 fs->fs_maxcluster = lp = space;
710 for (i = 0; i < fs->fs_ncg; i++)
711 *lp++ = fs->fs_contigsumsize;
714 size = fs->fs_ncg * sizeof(uint8_t);
715 fs->fs_contigdirs = (uint8_t *)space;
716 bzero(fs->fs_contigdirs, size);
717 /* Compatibility for old filesystems XXX */
718 if (fs->fs_avgfilesize <= 0) /* XXX */
719 fs->fs_avgfilesize = AVFILESIZ; /* XXX */
720 if (fs->fs_avgfpdir <= 0) /* XXX */
721 fs->fs_avgfpdir = AFPDIR; /* XXX */
722 mp->mnt_data = (qaddr_t)ump;
723 mp->mnt_stat.f_fsid.val[0] = fs->fs_id[0];
724 mp->mnt_stat.f_fsid.val[1] = fs->fs_id[1];
725 if (fs->fs_id[0] == 0 || fs->fs_id[1] == 0 ||
726 vfs_getvfs(&mp->mnt_stat.f_fsid))
728 mp->mnt_maxsymlinklen = fs->fs_maxsymlinklen;
729 mp->mnt_flag |= MNT_LOCAL;
732 ump->um_devvp = devvp;
733 ump->um_nindir = fs->fs_nindir;
734 ump->um_bptrtodb = fs->fs_fsbtodb;
735 ump->um_seqinc = fs->fs_frag;
736 for (i = 0; i < MAXQUOTAS; i++)
737 ump->um_quotas[i] = NULLVP;
738 dev->si_mountpoint = mp;
741 if( mp->mnt_flag & MNT_ROOTFS) {
743 * Root mount; update timestamp in mount structure.
744 * this will be used by the common root mount code
745 * to update the system clock.
747 mp->mnt_time = fs->fs_time;
750 ump->um_savedmaxfilesize = fs->fs_maxfilesize; /* XXX */
751 maxfilesize = (uint64_t)0x40000000 * fs->fs_bsize - 1; /* XXX */
752 /* Enforce limit caused by vm object backing (32 bits vm_pindex_t). */
753 if (maxfilesize > (uint64_t)0x80000000u * PAGE_SIZE - 1)
754 maxfilesize = (uint64_t)0x80000000u * PAGE_SIZE - 1;
755 if (fs->fs_maxfilesize > maxfilesize) /* XXX */
756 fs->fs_maxfilesize = maxfilesize; /* XXX */
758 if ((fs->fs_flags & FS_DOSOFTDEP) &&
759 (error = softdep_mount(devvp, mp, fs)) != 0) {
760 free(fs->fs_csp, M_UFSMNT);
765 (void) ffs_sbupdate(ump, MNT_WAIT);
767 vfs_add_vnodeops(mp, &mp->mnt_vn_norm_ops, ffs_vnodeop_entries, VVF_SUPPORTS_FSMID);
768 vfs_add_vnodeops(mp, &mp->mnt_vn_spec_ops, ffs_specop_entries, VVF_SUPPORTS_FSMID);
769 vfs_add_vnodeops(mp, &mp->mnt_vn_fifo_ops, ffs_fifoop_entries, VVF_SUPPORTS_FSMID);
773 dev->si_mountpoint = NULL;
776 VOP_CLOSE(devvp, ronly ? FREAD : FREAD|FWRITE, td);
778 free(ump->um_fs, M_UFSMNT);
780 mp->mnt_data = (qaddr_t)0;
786 * Sanity checks for old filesystems.
788 * XXX - goes away some day.
791 ffs_oldfscompat(struct fs *fs)
793 fs->fs_npsect = max(fs->fs_npsect, fs->fs_nsect); /* XXX */
794 fs->fs_interleave = max(fs->fs_interleave, 1); /* XXX */
795 if (fs->fs_postblformat == FS_42POSTBLFMT) /* XXX */
796 fs->fs_nrpos = 8; /* XXX */
797 if (fs->fs_inodefmt < FS_44INODEFMT) { /* XXX */
800 uint64_t sizepb = fs->fs_bsize; /* XXX */
802 fs->fs_maxfilesize = fs->fs_bsize * NDADDR - 1; /* XXX */
803 for (i = 0; i < NIADDR; i++) { /* XXX */
804 sizepb *= NINDIR(fs); /* XXX */
805 fs->fs_maxfilesize += sizepb; /* XXX */
808 fs->fs_maxfilesize = (u_quad_t) 1LL << 39;
809 fs->fs_qbmask = ~fs->fs_bmask; /* XXX */
810 fs->fs_qfmask = ~fs->fs_fmask; /* XXX */
816 * unmount system call
819 ffs_unmount(struct mount *mp, int mntflags, struct thread *td)
821 struct ufsmount *ump;
826 if (mntflags & MNT_FORCE) {
829 if (mp->mnt_flag & MNT_SOFTDEP) {
830 if ((error = softdep_flushfiles(mp, flags, td)) != 0)
833 if ((error = ffs_flushfiles(mp, flags, td)) != 0)
838 if (fs->fs_ronly == 0) {
839 fs->fs_clean = fs->fs_flags & FS_UNCLEAN ? 0 : 1;
840 error = ffs_sbupdate(ump, MNT_WAIT);
846 ump->um_devvp->v_rdev->si_mountpoint = NULL;
848 vinvalbuf(ump->um_devvp, V_SAVE, td, 0, 0);
849 error = VOP_CLOSE(ump->um_devvp, fs->fs_ronly ? FREAD : FREAD|FWRITE, td);
851 vrele(ump->um_devvp);
853 free(fs->fs_csp, M_UFSMNT);
856 mp->mnt_data = (qaddr_t)0;
857 mp->mnt_flag &= ~MNT_LOCAL;
862 * Flush out all the files in a filesystem.
865 ffs_flushfiles(struct mount *mp, int flags, struct thread *td)
867 struct ufsmount *ump;
872 if (mp->mnt_flag & MNT_QUOTA) {
874 error = vflush(mp, 0, SKIPSYSTEM|flags);
877 /* Find out how many quota files we have open. */
878 for (i = 0; i < MAXQUOTAS; i++) {
879 if (ump->um_quotas[i] == NULLVP)
884 * Here we fall through to vflush again to ensure
885 * that we have gotten rid of all the system vnodes.
890 * Flush all the files.
892 if ((error = vflush(mp, 0, flags)) != 0)
895 * Flush filesystem metadata.
897 vn_lock(ump->um_devvp, LK_EXCLUSIVE | LK_RETRY, td);
898 error = VOP_FSYNC(ump->um_devvp, MNT_WAIT, td);
899 VOP_UNLOCK(ump->um_devvp, 0, td);
904 * Get filesystem statistics.
907 ffs_statfs(struct mount *mp, struct statfs *sbp, struct thread *td)
909 struct ufsmount *ump;
914 if (fs->fs_magic != FS_MAGIC)
916 sbp->f_bsize = fs->fs_fsize;
917 sbp->f_iosize = fs->fs_bsize;
918 sbp->f_blocks = fs->fs_dsize;
919 sbp->f_bfree = fs->fs_cstotal.cs_nbfree * fs->fs_frag +
920 fs->fs_cstotal.cs_nffree;
921 sbp->f_bavail = freespace(fs, fs->fs_minfree);
922 sbp->f_files = fs->fs_ncg * fs->fs_ipg - ROOTINO;
923 sbp->f_ffree = fs->fs_cstotal.cs_nifree;
924 if (sbp != &mp->mnt_stat) {
925 sbp->f_type = mp->mnt_vfc->vfc_typenum;
926 bcopy((caddr_t)mp->mnt_stat.f_mntfromname,
927 (caddr_t)&sbp->f_mntfromname[0], MNAMELEN);
933 * Go through the disk queues to initiate sandbagged IO;
934 * go through the inodes to write those that have been modified;
935 * initiate the writing of the super block if it has been modified.
937 * Note: we are always called with the filesystem marked `MPBUSY'.
941 static int ffs_sync_scan1(struct mount *mp, struct vnode *vp, void *data);
942 static int ffs_sync_scan2(struct mount *mp, struct vnode *vp, void *data);
945 ffs_sync(struct mount *mp, int waitfor, struct thread *td)
947 struct ufsmount *ump = VFSTOUFS(mp);
950 struct scaninfo scaninfo;
953 if (fs->fs_fmod != 0 && fs->fs_ronly != 0) { /* XXX */
954 printf("fs = %s\n", fs->fs_fsmnt);
955 panic("ffs_sync: rofs mod");
959 * Write back each (modified) inode.
961 scaninfo.allerror = 0;
963 scaninfo.waitfor = waitfor;
964 while (scaninfo.rescan) {
966 vmntvnodescan(mp, VMSC_GETVP|VMSC_NOWAIT,
967 ffs_sync_scan1, ffs_sync_scan2, &scaninfo);
971 * Force stale filesystem control information to be flushed.
973 if (waitfor != MNT_LAZY) {
974 if (ump->um_mountp->mnt_flag & MNT_SOFTDEP)
975 waitfor = MNT_NOWAIT;
976 vn_lock(ump->um_devvp, LK_EXCLUSIVE | LK_RETRY, td);
977 if ((error = VOP_FSYNC(ump->um_devvp, waitfor, td)) != 0)
978 scaninfo.allerror = error;
979 VOP_UNLOCK(ump->um_devvp, 0, td);
985 * Write back modified superblock.
987 if (fs->fs_fmod != 0 && (error = ffs_sbupdate(ump, waitfor)) != 0)
988 scaninfo.allerror = error;
989 return (scaninfo.allerror);
993 ffs_sync_scan1(struct mount *mp, struct vnode *vp, void *data)
998 * Depend on the mount list's vnode lock to keep things stable
999 * enough for a quick test. Since there might be hundreds of
1000 * thousands of vnodes, we cannot afford even a subroutine
1001 * call unless there's a good chance that we have work to do.
1004 /* Restart out whole search if this guy is locked
1005 * or is being reclaimed.
1007 if (vp->v_type == VNON || ((ip->i_flag &
1008 (IN_ACCESS | IN_CHANGE | IN_MODIFIED | IN_UPDATE)) == 0 &&
1009 RB_EMPTY(&vp->v_rbdirty_tree))) {
1016 ffs_sync_scan2(struct mount *mp, struct vnode *vp, void *data)
1018 struct scaninfo *info = data;
1019 thread_t td = curthread; /* XXX */
1024 * We have to recheck after having obtained the vnode interlock.
1027 if (vp->v_type == VNON || ((ip->i_flag &
1028 (IN_ACCESS | IN_CHANGE | IN_MODIFIED | IN_UPDATE)) == 0 &&
1029 RB_EMPTY(&vp->v_rbdirty_tree))) {
1032 if (vp->v_type != VCHR) {
1033 if ((error = VOP_FSYNC(vp, info->waitfor, td)) != 0)
1034 info->allerror = error;
1037 * We must reference the vp to prevent it from
1038 * getting ripped out from under UFS_UPDATE, since
1039 * we are not holding a vnode lock.
1041 /* UFS_UPDATE(vp, waitfor == MNT_WAIT); */
1048 * Look up a FFS dinode number to find its incore vnode, otherwise read it
1049 * in from disk. If it is in core, wait for the lock bit to clear, then
1050 * return the inode locked. Detection and handling of mount points must be
1051 * done by the calling routine.
1055 ffs_vget(struct mount *mp, ino_t ino, struct vnode **vpp)
1059 struct ufsmount *ump;
1068 if ((*vpp = ufs_ihashget(dev, ino)) != NULL) {
1073 * If this MALLOC() is performed after the getnewvnode()
1074 * it might block, leaving a vnode with a NULL v_data to be
1075 * found by ffs_sync() if a sync happens to fire right then,
1076 * which will cause a panic because ffs_sync() blindly
1077 * dereferences vp->v_data (as well it should).
1079 * XXX this may no longer be true since getnewvnode returns a
1080 * VX locked vnode now.
1082 MALLOC(ip, struct inode *, sizeof(struct inode),
1083 ump->um_malloctype, M_WAITOK);
1085 /* Allocate a new vnode/inode. */
1086 error = getnewvnode(VT_UFS, mp, &vp, VLKTIMEOUT, LK_CANRECURSE);
1089 free(ip, ump->um_malloctype);
1092 bzero((caddr_t)ip, sizeof(struct inode));
1094 ip->i_fs = fs = ump->um_fs;
1100 for (i = 0; i < MAXQUOTAS; i++)
1101 ip->i_dquot[i] = NODQUOT;
1106 * Insert it into the inode hash table and check for a collision.
1107 * If a collision occurs, throw away the vnode and try again.
1109 if (ufs_ihashins(ip) != 0) {
1110 printf("debug: ufs ihashins collision, retrying inode %ld\n",
1111 (long)ip->i_number);
1113 free(ip, ump->um_malloctype);
1118 /* Read in the disk contents for the inode, copy into the inode. */
1119 error = bread(ump->um_devvp, fsbtodb(fs, ino_to_fsba(fs, ino)),
1120 (int)fs->fs_bsize, &bp);
1123 * The inode does not contain anything useful, so it would
1124 * be misleading to leave it on its hash chain. With mode
1125 * still zero, it will be unlinked and returned to the free
1133 ip->i_din = *((struct dinode *)bp->b_data + ino_to_fsbo(fs, ino));
1134 if (DOINGSOFTDEP(vp))
1135 softdep_load_inodeblock(ip);
1137 ip->i_effnlink = ip->i_nlink;
1141 * Initialize the vnode from the inode, check for aliases.
1142 * Note that the underlying vnode may have changed.
1144 error = ufs_vinit(mp, &vp);
1151 * Finish inode initialization now that aliasing has been resolved.
1153 ip->i_devvp = ump->um_devvp;
1156 * Set up a generation number for this inode if it does not
1157 * already have one. This should only happen on old filesystems.
1159 if (ip->i_gen == 0) {
1160 ip->i_gen = random() / 2 + 1;
1161 if ((vp->v_mount->mnt_flag & MNT_RDONLY) == 0)
1162 ip->i_flag |= IN_MODIFIED;
1165 * Ensure that uid and gid are correct. This is a temporary
1166 * fix until fsck has been changed to do the update.
1168 if (fs->fs_inodefmt < FS_44INODEFMT) { /* XXX */
1169 ip->i_uid = ip->i_din.di_ouid; /* XXX */
1170 ip->i_gid = ip->i_din.di_ogid; /* XXX */
1174 * return a VX locked and refd vnode (VX == same as normal vget()
1175 * vnode so we are ok)
1182 * File handle to vnode
1184 * Have to be really careful about stale file handles:
1185 * - check that the inode number is valid
1186 * - call ffs_vget() to get the locked inode
1187 * - check for an unallocated inode (i_mode == 0)
1188 * - check that the given client host has export rights and return
1189 * those rights via. exflagsp and credanonp
1192 ffs_fhtovp(struct mount *mp, struct fid *fhp, struct vnode **vpp)
1197 ufhp = (struct ufid *)fhp;
1198 fs = VFSTOUFS(mp)->um_fs;
1199 if (ufhp->ufid_ino < ROOTINO ||
1200 ufhp->ufid_ino >= fs->fs_ncg * fs->fs_ipg)
1202 return (ufs_fhtovp(mp, ufhp, vpp));
1206 * Vnode pointer to File handle
1210 ffs_vptofh(struct vnode *vp, struct fid *fhp)
1216 ufhp = (struct ufid *)fhp;
1217 ufhp->ufid_len = sizeof(struct ufid);
1218 ufhp->ufid_ino = ip->i_number;
1219 ufhp->ufid_gen = ip->i_gen;
1224 * Initialize the filesystem; just use ufs_init.
1227 ffs_init(struct vfsconf *vfsp)
1229 softdep_initialize();
1230 return (ufs_init(vfsp));
1234 * Write a superblock and associated information back to disk.
1237 ffs_sbupdate(struct ufsmount *mp, int waitfor)
1239 struct fs *dfs, *fs = mp->um_fs;
1243 int i, size, error, allerror = 0;
1246 * First write back the summary information.
1248 blks = howmany(fs->fs_cssize, fs->fs_fsize);
1250 for (i = 0; i < blks; i += fs->fs_frag) {
1251 size = fs->fs_bsize;
1252 if (i + fs->fs_frag > blks)
1253 size = (blks - i) * fs->fs_fsize;
1254 bp = getblk(mp->um_devvp, fsbtodb(fs, fs->fs_csaddr + i),
1256 bcopy(space, bp->b_data, (uint)size);
1257 space = (char *)space + size;
1258 if (waitfor != MNT_WAIT)
1260 else if ((error = bwrite(bp)) != 0)
1264 * Now write back the superblock itself. If any errors occurred
1265 * up to this point, then fail so that the superblock avoids
1266 * being written out as clean.
1270 bp = getblk(mp->um_devvp, SBLOCK, (int)fs->fs_sbsize, 0, 0);
1272 fs->fs_time = time_second;
1273 bcopy((caddr_t)fs, bp->b_data, (uint)fs->fs_sbsize);
1274 /* Restore compatibility to old filesystems. XXX */
1275 dfs = (struct fs *)bp->b_data; /* XXX */
1276 if (fs->fs_postblformat == FS_42POSTBLFMT) /* XXX */
1277 dfs->fs_nrpos = -1; /* XXX */
1278 if (fs->fs_inodefmt < FS_44INODEFMT) { /* XXX */
1279 int32_t *lp, tmp; /* XXX */
1281 lp = (int32_t *)&dfs->fs_qbmask; /* XXX */
1282 tmp = lp[4]; /* XXX */
1283 for (i = 4; i > 0; i--) /* XXX */
1284 lp[i] = lp[i-1]; /* XXX */
1285 lp[0] = tmp; /* XXX */
1287 dfs->fs_maxfilesize = mp->um_savedmaxfilesize; /* XXX */
1288 if (waitfor != MNT_WAIT)
1290 else if ((error = bwrite(bp)) != 0)