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 1.54 2007/05/09 00:53:36 dillon Exp $
38 #include "opt_quota.h"
40 #include <sys/param.h>
41 #include <sys/systm.h>
43 #include <sys/nlookup.h>
44 #include <sys/kernel.h>
45 #include <sys/vnode.h>
46 #include <sys/mount.h>
49 #include <sys/fcntl.h>
50 #include <sys/disklabel.h>
51 #include <sys/malloc.h>
56 #include "ufs_extern.h"
59 #include "ffs_extern.h"
62 #include <vm/vm_page.h>
63 #include <vm/vm_zone.h>
65 static MALLOC_DEFINE(M_FFSNODE, "FFS node", "FFS vnode private part");
67 static int ffs_sbupdate (struct ufsmount *, int);
68 static int ffs_reload (struct mount *, struct ucred *);
69 static int ffs_oldfscompat (struct fs *);
70 static int ffs_mount (struct mount *, char *, caddr_t, struct ucred *);
71 static int ffs_init (struct vfsconf *);
73 static struct vfsops ufs_vfsops = {
74 .vfs_mount = ffs_mount,
75 .vfs_unmount = ffs_unmount,
77 .vfs_quotactl = ufs_quotactl,
78 .vfs_statfs = ffs_statfs,
81 .vfs_fhtovp = ffs_fhtovp,
82 .vfs_checkexp = ufs_check_export,
83 .vfs_vptofh = ffs_vptofh,
85 .vfs_uninit = ufs_uninit
88 VFS_SET(ufs_vfsops, ufs, 0);
90 extern struct vop_ops ffs_vnode_vops;
91 extern struct vop_ops ffs_spec_vops;
92 extern struct vop_ops ffs_fifo_vops;
97 * Called when mounting local physical media
101 * mp mount point structure
102 * path NULL (flag for root mount!!!)
104 * p process (user credentials check [statfs])
107 * mp mount point structure
108 * path path to mount point
109 * data pointer to argument struct in user space
110 * p process (user credentials check)
113 * !0 error number (errno.h)
118 * mount point is locked
120 * mount point is locked
123 * A NULL path can be used for a flag since the mount
124 * system call will fail with EFAULT in copyinstr in
125 * nlookup() if it is a genuine NULL from the user.
128 ffs_mount(struct mount *mp, /* mount struct pointer */
129 char *path, /* path to mount point */
130 caddr_t data, /* arguments to FS specific mount */
131 struct ucred *cred) /* process requesting mount */
137 struct ufs_args args;
138 struct ufsmount *ump = 0;
140 int flags, ronly = 0;
142 struct nlookupdata nd;
143 struct vnode *rootvp;
149 * Use NULL path to flag a root mount
154 * Mounting root filesystem
158 if ((error = bdevvp(rootdev, &rootvp))) {
159 kprintf("ffs_mountroot: can't find rootvp\n");
163 if( ( error = ffs_mountfs(rootvp, mp, M_FFSNODE)) != 0) {
164 /* fs specific cleanup (if any)*/
169 goto dostatfs; /* success*/
175 * Mounting non-root filesystem or updating a filesystem
179 /* copy in user arguments*/
180 error = copyin(data, (caddr_t)&args, sizeof (struct ufs_args));
182 goto error_1; /* can't get arguments*/
185 * If updating, check whether changing from read-only to
186 * read/write; if there is no device name, that's all we do.
188 if (mp->mnt_flag & MNT_UPDATE) {
191 devvp = ump->um_devvp;
193 ronly = fs->fs_ronly; /* MNT_RELOAD might change this */
194 if (ronly == 0 && (mp->mnt_flag & MNT_RDONLY)) {
196 * Flush any dirty data.
198 VFS_SYNC(mp, MNT_WAIT);
200 * Check for and optionally get rid of files open
204 if (mp->mnt_flag & MNT_FORCE)
206 if (mp->mnt_flag & MNT_SOFTDEP) {
207 error = softdep_flushfiles(mp, flags);
209 error = ffs_flushfiles(mp, flags);
213 if (!error && (mp->mnt_flag & MNT_RELOAD)) {
214 error = ffs_reload(mp, NULL);
219 if (ronly && (mp->mnt_kern_flag & MNTK_WANTRDWR)) {
221 * If upgrade to read-write by non-root, then verify
222 * that user has necessary permissions on the device.
224 if (cred->cr_uid != 0) {
225 vn_lock(devvp, LK_EXCLUSIVE | LK_RETRY);
226 if ((error = VOP_ACCESS(devvp, VREAD | VWRITE,
234 fs->fs_flags &= ~FS_UNCLEAN;
235 if (fs->fs_clean == 0) {
236 fs->fs_flags |= FS_UNCLEAN;
237 if (mp->mnt_flag & MNT_FORCE) {
239 "WARNING: %s was not properly dismounted\n",
243 "WARNING: R/W mount of %s denied. Filesystem is not clean - run fsck\n",
250 /* check to see if we need to start softdep */
251 if (fs->fs_flags & FS_DOSOFTDEP) {
252 error = softdep_mount(devvp, mp, fs);
259 * Soft updates is incompatible with "async",
260 * so if we are doing softupdates stop the user
261 * from setting the async flag in an update.
262 * Softdep_mount() clears it in an initial mount
265 if (mp->mnt_flag & MNT_SOFTDEP) {
266 mp->mnt_flag &= ~MNT_ASYNC;
268 /* if not updating name...*/
269 if (args.fspec == 0) {
271 * Process export requests. Jumping to "success"
272 * will return the vfs_export() error code.
274 error = vfs_export(mp, &ump->um_export, &args.export);
280 * Not an update, or updating the name: look up the name
281 * and verify that it refers to a sensible block device.
284 error = nlookup_init(&nd, args.fspec, UIO_USERSPACE, NLC_FOLLOW);
286 error = nlookup(&nd);
288 error = cache_vref(&nd.nl_nch, nd.nl_cred, &devvp);
293 if (!vn_isdisk(devvp, &error))
297 * If mount by non-root, then verify that user has necessary
298 * permissions on the device.
300 if (cred->cr_uid != 0) {
302 if ((mp->mnt_flag & MNT_RDONLY) == 0)
303 accessmode |= VWRITE;
304 vn_lock(devvp, LK_EXCLUSIVE | LK_RETRY);
305 if ((error = VOP_ACCESS(devvp, accessmode, cred)) != 0) {
312 if (mp->mnt_flag & MNT_UPDATE) {
314 * UPDATE - make sure the resolved vnode represents the same
315 * device. Note that devvp->v_rdev may be NULL since we
316 * haven't opened it, so compare udev instead.
318 * Our current open/writecount state is associated with
319 * um_devvp, so continue using um_devvp and throw away devvp.
321 if (devvp != ump->um_devvp) {
322 if (devvp->v_umajor == ump->um_devvp->v_umajor &&
323 devvp->v_uminor == ump->um_devvp->v_uminor) {
325 devvp = ump->um_devvp;
327 kprintf("cannot update mount, udev does"
328 " not match %08x:%08x vs %08x:%08x\n",
329 devvp->v_umajor, devvp->v_uminor,
330 ump->um_devvp->v_umajor,
331 ump->um_devvp->v_uminor);
332 error = EINVAL; /* needs translation */
338 * Update device name only on success
341 /* Save "mounted from" info for mount point (NULL pad)*/
342 copyinstr( args.fspec,
343 mp->mnt_stat.f_mntfromname,
346 bzero( mp->mnt_stat.f_mntfromname + size, MNAMELEN - size);
355 /* Save "mounted from" info for mount point (NULL pad)*/
356 copyinstr( args.fspec, /* device name*/
357 mp->mnt_stat.f_mntfromname, /* save area*/
358 MNAMELEN - 1, /* max size*/
359 &size); /* real size*/
360 bzero( mp->mnt_stat.f_mntfromname + size, MNAMELEN - size);
362 /* Save "last mounted on" info for mount point (NULL pad)*/
363 bzero(mp->mnt_stat.f_mntonname,
364 sizeof(mp->mnt_stat.f_mntonname));
366 copyinstr(path, mp->mnt_stat.f_mntonname,
367 sizeof(mp->mnt_stat.f_mntonname) - 1,
371 error = ffs_mountfs(devvp, mp, M_FFSNODE);
379 * Initialize FS stat information in mount struct; uses
380 * mp->mnt_stat.f_mntfromname.
382 * This code is common to root and non-root mounts
384 (void)VFS_STATFS(mp, &mp->mnt_stat, cred);
389 error_2: /* error with devvp held*/
391 /* release devvp before failing*/
394 error_1: /* no state to back out*/
397 if (!error && path && (mp->mnt_flag & MNT_UPDATE)) {
398 /* Update clean flag after changing read-onlyness. */
400 if (ronly != fs->fs_ronly) {
401 fs->fs_ronly = ronly;
402 fs->fs_clean = ronly &&
403 (fs->fs_flags & FS_UNCLEAN) == 0 ? 1 : 0;
406 * The device must be re-opened as appropriate or
407 * the device close at unmount time will panic.
409 vn_lock(devvp, LK_EXCLUSIVE | LK_RETRY);
411 VOP_OPEN(devvp, FREAD, FSCRED, NULL);
412 VOP_CLOSE(devvp, FREAD|FWRITE);
414 VOP_OPEN(devvp, FREAD|FWRITE, FSCRED, NULL);
415 VOP_CLOSE(devvp, FREAD);
418 ffs_sbupdate(ump, MNT_WAIT);
425 * Reload all incore data for a filesystem (used after running fsck on
426 * the root filesystem and finding things to fix). The filesystem must
427 * be mounted read-only.
429 * Things to do to update the mount:
430 * 1) invalidate all cached meta-data.
431 * 2) re-read superblock from disk.
432 * 3) re-read summary information from disk.
433 * 4) invalidate all inactive vnodes.
434 * 5) invalidate all cached file data.
435 * 6) re-read inode data for all active vnodes.
438 static int ffs_reload_scan2(struct mount *mp, struct vnode *vp, void *data);
449 ffs_reload(struct mount *mp, struct ucred *cred)
454 struct fs *fs, *newfs;
455 struct partinfo dpart;
457 int i, blks, size, error;
458 struct scaninfo scaninfo;
461 if ((mp->mnt_flag & MNT_RDONLY) == 0)
464 * Step 1: invalidate all cached meta-data.
466 devvp = VFSTOUFS(mp)->um_devvp;
467 vn_lock(devvp, LK_EXCLUSIVE | LK_RETRY);
468 error = vinvalbuf(devvp, 0, 0, 0);
471 panic("ffs_reload: dirty1");
476 * The backing device must be VMIO-capable because we use getblk().
477 * NOTE: the MFS driver now returns a VMIO-enabled descriptor.
479 if (devvp->v_object == NULL)
480 panic("ffs_reload: devvp has no VM object!");
483 * Step 2: re-read superblock from disk.
485 if (VOP_IOCTL(devvp, DIOCGPART, (caddr_t)&dpart, FREAD, cred) != 0)
488 size = dpart.disklab->d_secsize;
489 if ((error = bread(devvp, SBOFF, SBSIZE, &bp)) != 0) {
493 newfs = (struct fs *)bp->b_data;
494 if (newfs->fs_magic != FS_MAGIC || newfs->fs_bsize > MAXBSIZE ||
495 newfs->fs_bsize < sizeof(struct fs)) {
497 return (EIO); /* XXX needs translation */
499 fs = VFSTOUFS(mp)->um_fs;
501 * Copy pointer fields back into superblock before copying in XXX
502 * new superblock. These should really be in the ufsmount. XXX
503 * Note that important parameters (eg fs_ncg) are unchanged.
505 newfs->fs_csp = fs->fs_csp;
506 newfs->fs_maxcluster = fs->fs_maxcluster;
507 newfs->fs_contigdirs = fs->fs_contigdirs;
508 /* The filesystem is still read-only. */
510 bcopy(newfs, fs, (uint)fs->fs_sbsize);
511 if (fs->fs_sbsize < SBSIZE)
512 bp->b_flags |= B_INVAL;
514 mp->mnt_maxsymlinklen = fs->fs_maxsymlinklen;
516 /* An old fsck may have zeroed these fields, so recheck them. */
517 if (fs->fs_avgfilesize <= 0) /* XXX */
518 fs->fs_avgfilesize = AVFILESIZ; /* XXX */
519 if (fs->fs_avgfpdir <= 0) /* XXX */
520 fs->fs_avgfpdir = AFPDIR; /* XXX */
523 * Step 3: re-read summary information from disk.
525 blks = howmany(fs->fs_cssize, fs->fs_fsize);
527 for (i = 0; i < blks; i += fs->fs_frag) {
529 if (i + fs->fs_frag > blks)
530 size = (blks - i) * fs->fs_fsize;
531 error = bread(devvp, fsbtodoff(fs, fs->fs_csaddr + i), size, &bp);
536 bcopy(bp->b_data, space, (uint)size);
537 space = (char *)space + size;
541 * We no longer know anything about clusters per cylinder group.
543 if (fs->fs_contigsumsize > 0) {
544 lp = fs->fs_maxcluster;
545 for (i = 0; i < fs->fs_ncg; i++)
546 *lp++ = fs->fs_contigsumsize;
551 scaninfo.devvp = devvp;
552 while (error == 0 && scaninfo.rescan) {
554 error = vmntvnodescan(mp, VMSC_GETVX,
555 NULL, ffs_reload_scan2, &scaninfo);
561 ffs_reload_scan2(struct mount *mp, struct vnode *vp, void *data)
563 struct scaninfo *info = data;
574 if (vinvalbuf(vp, 0, 0, 0))
575 panic("ffs_reload: dirty2");
577 * Step 6: re-read inode data for all active vnodes.
580 error = bread(info->devvp,
581 fsbtodoff(info->fs, ino_to_fsba(info->fs, ip->i_number)),
582 (int)info->fs->fs_bsize, &bp);
587 ip->i_din = *((struct ufs1_dinode *)bp->b_data +
588 ino_to_fsbo(info->fs, ip->i_number));
589 ip->i_effnlink = ip->i_nlink;
595 * Common code for mount and mountroot
598 ffs_mountfs(struct vnode *devvp, struct mount *mp, struct malloc_type *mtype)
600 struct ufsmount *ump;
604 struct partinfo dpart;
606 int error, i, blks, size, ronly;
608 uint64_t maxfilesize; /* XXX */
611 * Disallow multiple mounts of the same device.
612 * Disallow mounting of a device that is currently in use
613 * Flush out any old buffers remaining from a previous use.
615 error = vfs_mountedon(devvp);
618 if (count_udev(devvp->v_umajor, devvp->v_uminor) > 0)
620 vn_lock(devvp, LK_EXCLUSIVE | LK_RETRY);
621 error = vinvalbuf(devvp, V_SAVE, 0, 0);
626 ronly = (mp->mnt_flag & MNT_RDONLY) != 0;
627 vn_lock(devvp, LK_EXCLUSIVE | LK_RETRY);
628 error = VOP_OPEN(devvp, ronly ? FREAD : FREAD|FWRITE, FSCRED, NULL);
633 if (dev->si_iosize_max != 0)
634 mp->mnt_iosize_max = dev->si_iosize_max;
635 if (mp->mnt_iosize_max > MAXPHYS)
636 mp->mnt_iosize_max = MAXPHYS;
639 * Filesystem supports native FSMIDs
641 mp->mnt_kern_flag |= MNTK_FSMID;
644 * The backing device must be VMIO-capable because we use getblk().
645 * NOTE: the MFS driver now returns a VMIO-enabled descriptor.
646 * The VOP_OPEN() call above should have associated a VM object
649 if (devvp->v_object == NULL)
650 panic("ffs_reload: devvp has no VM object!");
652 if (VOP_IOCTL(devvp, DIOCGPART, (caddr_t)&dpart, FREAD, proc0.p_ucred) != 0)
655 size = dpart.disklab->d_secsize;
659 if ((error = bread(devvp, SBOFF, SBSIZE, &bp)) != 0)
661 fs = (struct fs *)bp->b_data;
662 if (fs->fs_magic != FS_MAGIC || fs->fs_bsize > MAXBSIZE ||
663 fs->fs_bsize < sizeof(struct fs)) {
664 error = EINVAL; /* XXX needs translation */
668 fs->fs_flags &= ~FS_UNCLEAN;
669 if (fs->fs_clean == 0) {
670 fs->fs_flags |= FS_UNCLEAN;
671 if (ronly || (mp->mnt_flag & MNT_FORCE)) {
673 "WARNING: %s was not properly dismounted\n",
677 "WARNING: R/W mount of %s denied. Filesystem is not clean - run fsck\n",
683 /* XXX updating 4.2 FFS superblocks trashes rotational layout tables */
684 if (fs->fs_postblformat == FS_42POSTBLFMT && !ronly) {
685 error = EROFS; /* needs translation */
688 ump = kmalloc(sizeof *ump, M_UFSMNT, M_WAITOK);
689 bzero((caddr_t)ump, sizeof *ump);
690 ump->um_malloctype = mtype;
691 ump->um_i_effnlink_valid = 1;
692 ump->um_fs = kmalloc((u_long)fs->fs_sbsize, M_UFSMNT,
694 bcopy(bp->b_data, ump->um_fs, (uint)fs->fs_sbsize);
695 if (fs->fs_sbsize < SBSIZE)
696 bp->b_flags |= B_INVAL;
700 fs->fs_ronly = ronly;
701 size = fs->fs_cssize;
702 blks = howmany(size, fs->fs_fsize);
703 if (fs->fs_contigsumsize > 0)
704 size += fs->fs_ncg * sizeof(int32_t);
705 size += fs->fs_ncg * sizeof(uint8_t);
706 space = kmalloc((u_long)size, M_UFSMNT, M_WAITOK);
708 for (i = 0; i < blks; i += fs->fs_frag) {
710 if (i + fs->fs_frag > blks)
711 size = (blks - i) * fs->fs_fsize;
712 if ((error = bread(devvp, fsbtodoff(fs, fs->fs_csaddr + i),
714 kfree(fs->fs_csp, M_UFSMNT);
717 bcopy(bp->b_data, space, (uint)size);
718 space = (char *)space + size;
722 if (fs->fs_contigsumsize > 0) {
723 fs->fs_maxcluster = lp = space;
724 for (i = 0; i < fs->fs_ncg; i++)
725 *lp++ = fs->fs_contigsumsize;
728 size = fs->fs_ncg * sizeof(uint8_t);
729 fs->fs_contigdirs = (uint8_t *)space;
730 bzero(fs->fs_contigdirs, size);
731 /* Compatibility for old filesystems XXX */
732 if (fs->fs_avgfilesize <= 0) /* XXX */
733 fs->fs_avgfilesize = AVFILESIZ; /* XXX */
734 if (fs->fs_avgfpdir <= 0) /* XXX */
735 fs->fs_avgfpdir = AFPDIR; /* XXX */
736 mp->mnt_data = (qaddr_t)ump;
737 mp->mnt_stat.f_fsid.val[0] = fs->fs_id[0];
738 mp->mnt_stat.f_fsid.val[1] = fs->fs_id[1];
739 if (fs->fs_id[0] == 0 || fs->fs_id[1] == 0 ||
740 vfs_getvfs(&mp->mnt_stat.f_fsid))
742 mp->mnt_maxsymlinklen = fs->fs_maxsymlinklen;
743 mp->mnt_flag |= MNT_LOCAL;
746 ump->um_devvp = devvp;
747 ump->um_nindir = fs->fs_nindir;
748 ump->um_bptrtodb = fs->fs_fsbtodb;
749 ump->um_seqinc = fs->fs_frag;
750 for (i = 0; i < MAXQUOTAS; i++)
751 ump->um_quotas[i] = NULLVP;
752 dev->si_mountpoint = mp;
755 /* restore "last mounted on" here */
756 bzero(fs->fs_fsmnt, sizeof(fs->fs_fsmnt));
757 ksnprintf(fs->fs_fsmnt, sizeof(fs->fs_fsmnt),
758 "%s", mp->mnt_stat.f_mntonname);
760 if( mp->mnt_flag & MNT_ROOTFS) {
762 * Root mount; update timestamp in mount structure.
763 * this will be used by the common root mount code
764 * to update the system clock.
766 mp->mnt_time = fs->fs_time;
769 ump->um_savedmaxfilesize = fs->fs_maxfilesize; /* XXX */
770 maxfilesize = (uint64_t)0x40000000 * fs->fs_bsize - 1; /* XXX */
771 /* Enforce limit caused by vm object backing (32 bits vm_pindex_t). */
772 if (maxfilesize > (uint64_t)0x80000000u * PAGE_SIZE - 1)
773 maxfilesize = (uint64_t)0x80000000u * PAGE_SIZE - 1;
774 if (fs->fs_maxfilesize > maxfilesize) /* XXX */
775 fs->fs_maxfilesize = maxfilesize; /* XXX */
777 if ((fs->fs_flags & FS_DOSOFTDEP) &&
778 (error = softdep_mount(devvp, mp, fs)) != 0) {
779 kfree(fs->fs_csp, M_UFSMNT);
784 (void) ffs_sbupdate(ump, MNT_WAIT);
786 vfs_add_vnodeops(mp, &ffs_vnode_vops, &mp->mnt_vn_norm_ops);
787 vfs_add_vnodeops(mp, &ffs_spec_vops, &mp->mnt_vn_spec_ops);
788 vfs_add_vnodeops(mp, &ffs_fifo_vops, &mp->mnt_vn_fifo_ops);
792 dev->si_mountpoint = NULL;
795 VOP_CLOSE(devvp, ronly ? FREAD : FREAD|FWRITE);
797 kfree(ump->um_fs, M_UFSMNT);
798 kfree(ump, M_UFSMNT);
799 mp->mnt_data = (qaddr_t)0;
805 * Sanity checks for old filesystems.
807 * XXX - goes away some day.
810 ffs_oldfscompat(struct fs *fs)
812 fs->fs_npsect = max(fs->fs_npsect, fs->fs_nsect); /* XXX */
813 fs->fs_interleave = max(fs->fs_interleave, 1); /* XXX */
814 if (fs->fs_postblformat == FS_42POSTBLFMT) /* XXX */
815 fs->fs_nrpos = 8; /* XXX */
816 if (fs->fs_inodefmt < FS_44INODEFMT) { /* XXX */
819 uint64_t sizepb = fs->fs_bsize; /* XXX */
821 fs->fs_maxfilesize = fs->fs_bsize * NDADDR - 1; /* XXX */
822 for (i = 0; i < NIADDR; i++) { /* XXX */
823 sizepb *= NINDIR(fs); /* XXX */
824 fs->fs_maxfilesize += sizepb; /* XXX */
827 fs->fs_maxfilesize = (u_quad_t) 1LL << 39;
828 fs->fs_qbmask = ~fs->fs_bmask; /* XXX */
829 fs->fs_qfmask = ~fs->fs_fmask; /* XXX */
835 * unmount system call
838 ffs_unmount(struct mount *mp, int mntflags)
840 struct ufsmount *ump;
845 if (mntflags & MNT_FORCE) {
848 if (mp->mnt_flag & MNT_SOFTDEP) {
849 if ((error = softdep_flushfiles(mp, flags)) != 0)
852 if ((error = ffs_flushfiles(mp, flags)) != 0)
857 if (fs->fs_ronly == 0) {
858 fs->fs_clean = fs->fs_flags & FS_UNCLEAN ? 0 : 1;
859 error = ffs_sbupdate(ump, MNT_WAIT);
865 ump->um_devvp->v_rdev->si_mountpoint = NULL;
867 vinvalbuf(ump->um_devvp, V_SAVE, 0, 0);
868 error = VOP_CLOSE(ump->um_devvp, fs->fs_ronly ? FREAD : FREAD|FWRITE);
870 vrele(ump->um_devvp);
872 kfree(fs->fs_csp, M_UFSMNT);
874 kfree(ump, M_UFSMNT);
875 mp->mnt_data = (qaddr_t)0;
876 mp->mnt_flag &= ~MNT_LOCAL;
881 * Flush out all the files in a filesystem.
884 ffs_flushfiles(struct mount *mp, int flags)
886 struct ufsmount *ump;
891 if (mp->mnt_flag & MNT_QUOTA) {
893 error = vflush(mp, 0, SKIPSYSTEM|flags);
896 /* Find out how many quota files we have open. */
897 for (i = 0; i < MAXQUOTAS; i++) {
898 if (ump->um_quotas[i] == NULLVP)
903 * Here we fall through to vflush again to ensure
904 * that we have gotten rid of all the system vnodes.
909 * Flush all the files.
911 if ((error = vflush(mp, 0, flags)) != 0)
914 * Flush filesystem metadata.
916 vn_lock(ump->um_devvp, LK_EXCLUSIVE | LK_RETRY);
917 error = VOP_FSYNC(ump->um_devvp, MNT_WAIT);
918 vn_unlock(ump->um_devvp);
923 * Get filesystem statistics.
926 ffs_statfs(struct mount *mp, struct statfs *sbp, struct ucred *cred)
928 struct ufsmount *ump;
933 if (fs->fs_magic != FS_MAGIC)
935 sbp->f_bsize = fs->fs_fsize;
936 sbp->f_iosize = fs->fs_bsize;
937 sbp->f_blocks = fs->fs_dsize;
938 sbp->f_bfree = fs->fs_cstotal.cs_nbfree * fs->fs_frag +
939 fs->fs_cstotal.cs_nffree;
940 sbp->f_bavail = freespace(fs, fs->fs_minfree);
941 sbp->f_files = fs->fs_ncg * fs->fs_ipg - ROOTINO;
942 sbp->f_ffree = fs->fs_cstotal.cs_nifree;
943 if (sbp != &mp->mnt_stat) {
944 sbp->f_type = mp->mnt_vfc->vfc_typenum;
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'.
960 static int ffs_sync_scan1(struct mount *mp, struct vnode *vp, void *data);
961 static int ffs_sync_scan2(struct mount *mp, struct vnode *vp, void *data);
964 ffs_sync(struct mount *mp, int waitfor)
966 struct ufsmount *ump = VFSTOUFS(mp);
969 struct scaninfo scaninfo;
972 if (fs->fs_fmod != 0 && fs->fs_ronly != 0) { /* XXX */
973 kprintf("fs = %s\n", fs->fs_fsmnt);
974 panic("ffs_sync: rofs mod");
978 * Write back each (modified) inode.
980 scaninfo.allerror = 0;
982 scaninfo.waitfor = waitfor;
983 while (scaninfo.rescan) {
985 vmntvnodescan(mp, VMSC_GETVP|VMSC_NOWAIT,
986 ffs_sync_scan1, ffs_sync_scan2, &scaninfo);
990 * Force stale filesystem control information to be flushed.
992 if (waitfor != MNT_LAZY) {
993 if (ump->um_mountp->mnt_flag & MNT_SOFTDEP)
994 waitfor = MNT_NOWAIT;
995 vn_lock(ump->um_devvp, LK_EXCLUSIVE | LK_RETRY);
996 if ((error = VOP_FSYNC(ump->um_devvp, waitfor)) != 0)
997 scaninfo.allerror = error;
998 vn_unlock(ump->um_devvp);
1004 * Write back modified superblock.
1006 if (fs->fs_fmod != 0 && (error = ffs_sbupdate(ump, waitfor)) != 0)
1007 scaninfo.allerror = error;
1008 return (scaninfo.allerror);
1012 ffs_sync_scan1(struct mount *mp, struct vnode *vp, void *data)
1017 * Depend on the mount list's vnode lock to keep things stable
1018 * enough for a quick test. Since there might be hundreds of
1019 * thousands of vnodes, we cannot afford even a subroutine
1020 * call unless there's a good chance that we have work to do.
1023 /* Restart out whole search if this guy is locked
1024 * or is being reclaimed.
1026 if (vp->v_type == VNON || ((ip->i_flag &
1027 (IN_ACCESS | IN_CHANGE | IN_MODIFIED | IN_UPDATE)) == 0 &&
1028 RB_EMPTY(&vp->v_rbdirty_tree))) {
1035 ffs_sync_scan2(struct mount *mp, struct vnode *vp, void *data)
1037 struct scaninfo *info = data;
1042 * We have to recheck after having obtained the vnode interlock.
1045 if (vp->v_type == VNON || ((ip->i_flag &
1046 (IN_ACCESS | IN_CHANGE | IN_MODIFIED | IN_UPDATE)) == 0 &&
1047 RB_EMPTY(&vp->v_rbdirty_tree))) {
1050 if (vp->v_type != VCHR) {
1051 if ((error = VOP_FSYNC(vp, info->waitfor)) != 0)
1052 info->allerror = error;
1055 * We must reference the vp to prevent it from
1056 * getting ripped out from under ffs_update, since
1057 * we are not holding a vnode lock.
1059 /* ffs_update(vp, waitfor == MNT_WAIT); */
1066 * Look up a FFS dinode number to find its incore vnode, otherwise read it
1067 * in from disk. If it is in core, wait for the lock bit to clear, then
1068 * return the inode locked. Detection and handling of mount points must be
1069 * done by the calling routine.
1073 ffs_vget(struct mount *mp, ino_t ino, struct vnode **vpp)
1077 struct ufsmount *ump;
1086 if ((*vpp = ufs_ihashget(dev, ino)) != NULL) {
1091 * If this MALLOC() is performed after the getnewvnode()
1092 * it might block, leaving a vnode with a NULL v_data to be
1093 * found by ffs_sync() if a sync happens to fire right then,
1094 * which will cause a panic because ffs_sync() blindly
1095 * dereferences vp->v_data (as well it should).
1097 * XXX this may no longer be true since getnewvnode returns a
1098 * VX locked vnode now.
1100 MALLOC(ip, struct inode *, sizeof(struct inode),
1101 ump->um_malloctype, M_WAITOK);
1103 /* Allocate a new vnode/inode. */
1104 error = getnewvnode(VT_UFS, mp, &vp, VLKTIMEOUT, LK_CANRECURSE);
1107 kfree(ip, ump->um_malloctype);
1110 bzero((caddr_t)ip, sizeof(struct inode));
1112 ip->i_fs = fs = ump->um_fs;
1118 for (i = 0; i < MAXQUOTAS; i++)
1119 ip->i_dquot[i] = NODQUOT;
1124 * Insert it into the inode hash table and check for a collision.
1125 * If a collision occurs, throw away the vnode and try again.
1127 if (ufs_ihashins(ip) != 0) {
1128 kprintf("debug: ufs ihashins collision, retrying inode %ld\n",
1129 (long)ip->i_number);
1131 kfree(ip, ump->um_malloctype);
1136 /* Read in the disk contents for the inode, copy into the inode. */
1137 error = bread(ump->um_devvp, fsbtodoff(fs, ino_to_fsba(fs, ino)),
1138 (int)fs->fs_bsize, &bp);
1141 * The inode does not contain anything useful, so it would
1142 * be misleading to leave it on its hash chain. With mode
1143 * still zero, it will be unlinked and returned to the free
1151 ip->i_din = *((struct ufs1_dinode *)bp->b_data + ino_to_fsbo(fs, ino));
1152 if (DOINGSOFTDEP(vp))
1153 softdep_load_inodeblock(ip);
1155 ip->i_effnlink = ip->i_nlink;
1159 * Initialize the vnode from the inode, check for aliases.
1160 * Note that the underlying vnode may have changed.
1162 error = ufs_vinit(mp, &vp);
1169 * Finish inode initialization now that aliasing has been resolved.
1171 ip->i_devvp = ump->um_devvp;
1174 * Set up a generation number for this inode if it does not
1175 * already have one. This should only happen on old filesystems.
1177 if (ip->i_gen == 0) {
1178 ip->i_gen = krandom() / 2 + 1;
1179 if ((vp->v_mount->mnt_flag & MNT_RDONLY) == 0)
1180 ip->i_flag |= IN_MODIFIED;
1183 * Ensure that uid and gid are correct. This is a temporary
1184 * fix until fsck has been changed to do the update.
1186 if (fs->fs_inodefmt < FS_44INODEFMT) { /* XXX */
1187 ip->i_uid = ip->i_din.di_ouid; /* XXX */
1188 ip->i_gid = ip->i_din.di_ogid; /* XXX */
1192 * return a VX locked and refd vnode (VX == same as normal vget()
1193 * vnode so we are ok)
1200 * File handle to vnode
1202 * Have to be really careful about stale file handles:
1203 * - check that the inode number is valid
1204 * - call ffs_vget() to get the locked inode
1205 * - check for an unallocated inode (i_mode == 0)
1206 * - check that the given client host has export rights and return
1207 * those rights via. exflagsp and credanonp
1210 ffs_fhtovp(struct mount *mp, struct fid *fhp, struct vnode **vpp)
1215 ufhp = (struct ufid *)fhp;
1216 fs = VFSTOUFS(mp)->um_fs;
1217 if (ufhp->ufid_ino < ROOTINO ||
1218 ufhp->ufid_ino >= fs->fs_ncg * fs->fs_ipg)
1220 return (ufs_fhtovp(mp, ufhp, vpp));
1224 * Vnode pointer to File handle
1228 ffs_vptofh(struct vnode *vp, struct fid *fhp)
1234 ufhp = (struct ufid *)fhp;
1235 ufhp->ufid_len = sizeof(struct ufid);
1236 ufhp->ufid_ino = ip->i_number;
1237 ufhp->ufid_gen = ip->i_gen;
1242 * Initialize the filesystem; just use ufs_init.
1245 ffs_init(struct vfsconf *vfsp)
1247 softdep_initialize();
1248 return (ufs_init(vfsp));
1252 * Write a superblock and associated information back to disk.
1255 ffs_sbupdate(struct ufsmount *mp, int waitfor)
1257 struct fs *dfs, *fs = mp->um_fs;
1261 int i, size, error, allerror = 0;
1264 * First write back the summary information.
1266 * NOTE: the getblk is relative to the device vnode so bio1
1267 * contains the device block number.
1269 blks = howmany(fs->fs_cssize, fs->fs_fsize);
1271 for (i = 0; i < blks; i += fs->fs_frag) {
1272 size = fs->fs_bsize;
1273 if (i + fs->fs_frag > blks)
1274 size = (blks - i) * fs->fs_fsize;
1275 bp = getblk(mp->um_devvp, fsbtodoff(fs, fs->fs_csaddr + i),
1277 bcopy(space, bp->b_data, (uint)size);
1278 space = (char *)space + size;
1279 if (waitfor != MNT_WAIT)
1281 else if ((error = bwrite(bp)) != 0)
1285 * Now write back the superblock itself. If any errors occurred
1286 * up to this point, then fail so that the superblock avoids
1287 * being written out as clean.
1291 bp = getblk(mp->um_devvp, SBOFF, (int)fs->fs_sbsize, 0, 0);
1293 fs->fs_time = time_second;
1294 bcopy((caddr_t)fs, bp->b_data, (uint)fs->fs_sbsize);
1295 /* Restore compatibility to old filesystems. XXX */
1296 dfs = (struct fs *)bp->b_data; /* XXX */
1297 if (fs->fs_postblformat == FS_42POSTBLFMT) /* XXX */
1298 dfs->fs_nrpos = -1; /* XXX */
1299 if (fs->fs_inodefmt < FS_44INODEFMT) { /* XXX */
1300 int32_t *lp, tmp; /* XXX */
1302 lp = (int32_t *)&dfs->fs_qbmask; /* XXX */
1303 tmp = lp[4]; /* XXX */
1304 for (i = 4; i > 0; i--) /* XXX */
1305 lp[i] = lp[i-1]; /* XXX */
1306 lp[0] = tmp; /* XXX */
1308 dfs->fs_maxfilesize = mp->um_savedmaxfilesize; /* XXX */
1309 if (waitfor != MNT_WAIT)
1311 else if ((error = bwrite(bp)) != 0)