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 $
37 #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/diskslice.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>
67 static MALLOC_DEFINE(M_FFSNODE, "FFS node", "FFS vnode private part");
69 static int ffs_sbupdate (struct ufsmount *, int);
70 static int ffs_reload (struct mount *, struct ucred *);
71 static int ffs_oldfscompat (struct fs *);
72 static int ffs_mount (struct mount *, char *, caddr_t, struct ucred *);
73 static int ffs_init (struct vfsconf *);
75 static struct vfsops ufs_vfsops = {
76 .vfs_mount = ffs_mount,
77 .vfs_unmount = ffs_unmount,
79 .vfs_quotactl = ufs_quotactl,
80 .vfs_statfs = ffs_statfs,
83 .vfs_fhtovp = ffs_fhtovp,
84 .vfs_checkexp = ufs_check_export,
85 .vfs_vptofh = ffs_vptofh,
87 .vfs_uninit = ufs_uninit
90 VFS_SET(ufs_vfsops, ufs, 0);
91 MODULE_VERSION(ufs, 1);
93 extern struct vop_ops ffs_vnode_vops;
94 extern struct vop_ops ffs_spec_vops;
95 extern struct vop_ops ffs_fifo_vops;
100 * Called when mounting local physical media
104 * mp mount point structure
105 * path NULL (flag for root mount!!!)
107 * p process (user credentials check [statfs])
110 * mp mount point structure
111 * path path to mount point
112 * data pointer to argument struct in user space
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 * nlookup() if it is a genuine NULL from the user.
131 ffs_mount(struct mount *mp, /* mount struct pointer */
132 char *path, /* path to mount point */
133 caddr_t data, /* arguments to FS specific mount */
134 struct ucred *cred) /* process requesting mount */
140 struct ufs_args args;
141 struct ufsmount *ump = 0;
143 int flags, ronly = 0;
145 struct nlookupdata nd;
146 struct vnode *rootvp;
152 * Use NULL path to flag a root mount
157 * Mounting root filesystem
161 if ((error = bdevvp(rootdev, &rootvp))) {
162 kprintf("ffs_mountroot: can't find rootvp\n");
166 if( ( error = ffs_mountfs(rootvp, mp, M_FFSNODE)) != 0) {
167 /* fs specific cleanup (if any)*/
172 goto dostatfs; /* success*/
178 * Mounting non-root filesystem or updating a filesystem
182 /* copy in user arguments*/
183 error = 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);
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 error = softdep_flushfiles(mp, flags);
212 error = ffs_flushfiles(mp, flags);
216 if (!error && (mp->mnt_flag & MNT_RELOAD)) {
217 error = ffs_reload(mp, NULL);
222 if (ronly && (mp->mnt_kern_flag & MNTK_WANTRDWR)) {
224 * If upgrade to read-write by non-root, then verify
225 * that user has necessary permissions on the device.
227 if (cred->cr_uid != 0) {
228 vn_lock(devvp, LK_EXCLUSIVE | LK_RETRY);
229 if ((error = VOP_EACCESS(devvp, VREAD | VWRITE,
237 fs->fs_flags &= ~FS_UNCLEAN;
238 if (fs->fs_clean == 0) {
239 fs->fs_flags |= FS_UNCLEAN;
240 if (mp->mnt_flag & MNT_FORCE) {
242 "WARNING: %s was not properly dismounted\n",
246 "WARNING: R/W mount of %s denied. Filesystem is not clean - run fsck\n",
253 /* check to see if we need to start softdep */
254 if (fs->fs_flags & FS_DOSOFTDEP) {
255 error = 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 error = 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.
287 error = nlookup_init(&nd, args.fspec, UIO_USERSPACE, NLC_FOLLOW);
289 error = nlookup(&nd);
291 error = cache_vref(&nd.nl_nch, nd.nl_cred, &devvp);
296 if (!vn_isdisk(devvp, &error))
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, LK_EXCLUSIVE | LK_RETRY);
308 if ((error = VOP_EACCESS(devvp, accessmode, cred)) != 0) {
315 if (mp->mnt_flag & MNT_UPDATE) {
317 * UPDATE - make sure the resolved vnode represents the same
318 * device. Note that devvp->v_rdev may be NULL since we
319 * haven't opened it, so compare udev instead.
321 * Our current open/writecount state is associated with
322 * um_devvp, so continue using um_devvp and throw away devvp.
324 if (devvp != ump->um_devvp) {
325 if (devvp->v_umajor == ump->um_devvp->v_umajor &&
326 devvp->v_uminor == ump->um_devvp->v_uminor) {
328 devvp = ump->um_devvp;
330 kprintf("cannot update mount, udev does"
331 " not match %08x:%08x vs %08x:%08x\n",
332 devvp->v_umajor, devvp->v_uminor,
333 ump->um_devvp->v_umajor,
334 ump->um_devvp->v_uminor);
335 error = 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);
358 /* Save "mounted from" info for mount point (NULL pad)*/
359 copyinstr( args.fspec, /* device name*/
360 mp->mnt_stat.f_mntfromname, /* save area*/
361 MNAMELEN - 1, /* max size*/
362 &size); /* real size*/
363 bzero( mp->mnt_stat.f_mntfromname + size, MNAMELEN - size);
365 /* Save "last mounted on" info for mount point (NULL pad)*/
366 bzero(mp->mnt_stat.f_mntonname,
367 sizeof(mp->mnt_stat.f_mntonname));
369 copyinstr(path, mp->mnt_stat.f_mntonname,
370 sizeof(mp->mnt_stat.f_mntonname) - 1,
372 } else { /* Root mount */
373 mp->mnt_stat.f_mntonname[0] = '/';
376 error = ffs_mountfs(devvp, mp, M_FFSNODE);
384 * Initialize FS stat information in mount struct; uses
385 * mp->mnt_stat.f_mntfromname.
387 * This code is common to root and non-root mounts
389 (void)VFS_STATFS(mp, &mp->mnt_stat, cred);
394 error_2: /* error with devvp held*/
396 /* release devvp before failing*/
399 error_1: /* no state to back out*/
402 if (!error && path && (mp->mnt_flag & MNT_UPDATE)) {
403 /* Update clean flag after changing read-onlyness. */
405 if (ronly != fs->fs_ronly) {
406 fs->fs_ronly = ronly;
407 fs->fs_clean = ronly &&
408 (fs->fs_flags & FS_UNCLEAN) == 0 ? 1 : 0;
411 * The device must be re-opened as appropriate or
412 * the device close at unmount time will panic.
414 vn_lock(devvp, LK_EXCLUSIVE | LK_RETRY);
416 VOP_OPEN(devvp, FREAD, FSCRED, NULL);
417 VOP_CLOSE(devvp, FREAD|FWRITE);
419 VOP_OPEN(devvp, FREAD|FWRITE, FSCRED, NULL);
420 VOP_CLOSE(devvp, FREAD);
423 ffs_sbupdate(ump, MNT_WAIT);
430 * Reload all incore data for a filesystem (used after running fsck on
431 * the root filesystem and finding things to fix). The filesystem must
432 * be mounted read-only.
434 * Things to do to update the mount:
435 * 1) invalidate all cached meta-data.
436 * 2) re-read superblock from disk.
437 * 3) re-read summary information from disk.
438 * 4) invalidate all inactive vnodes.
439 * 5) invalidate all cached file data.
440 * 6) re-read inode data for all active vnodes.
443 static int ffs_reload_scan2(struct mount *mp, struct vnode *vp, void *data);
454 ffs_reload(struct mount *mp, struct ucred *cred)
459 struct fs *fs, *newfs;
460 struct partinfo dpart;
462 int i, blks, size, error;
463 struct scaninfo scaninfo;
466 if ((mp->mnt_flag & MNT_RDONLY) == 0)
469 * Step 1: invalidate all cached meta-data.
471 devvp = VFSTOUFS(mp)->um_devvp;
472 vn_lock(devvp, LK_EXCLUSIVE | LK_RETRY);
473 error = vinvalbuf(devvp, 0, 0, 0);
476 panic("ffs_reload: dirty1");
481 * The backing device must be VMIO-capable because we use getblk().
482 * NOTE: the MFS driver now returns a VMIO-enabled descriptor.
484 if (devvp->v_object == NULL)
485 panic("ffs_reload: devvp has no VM object!");
488 * Step 2: re-read superblock from disk.
490 if (VOP_IOCTL(devvp, DIOCGPART, (caddr_t)&dpart, FREAD,
494 size = dpart.media_blksize;
496 if ((error = bread(devvp, SBOFF, SBSIZE, &bp)) != 0) {
500 newfs = (struct fs *)bp->b_data;
501 if (newfs->fs_magic != FS_MAGIC || newfs->fs_bsize > MAXBSIZE ||
502 newfs->fs_bsize < sizeof(struct fs)) {
504 return (EIO); /* XXX needs translation */
506 fs = VFSTOUFS(mp)->um_fs;
508 * Copy pointer fields back into superblock before copying in XXX
509 * new superblock. These should really be in the ufsmount. XXX
510 * Note that important parameters (eg fs_ncg) are unchanged.
512 newfs->fs_csp = fs->fs_csp;
513 newfs->fs_maxcluster = fs->fs_maxcluster;
514 newfs->fs_contigdirs = fs->fs_contigdirs;
515 /* The filesystem is still read-only. */
517 bcopy(newfs, fs, (uint)fs->fs_sbsize);
518 if (fs->fs_sbsize < SBSIZE)
519 bp->b_flags |= B_INVAL;
521 mp->mnt_maxsymlinklen = fs->fs_maxsymlinklen;
523 /* An old fsck may have zeroed these fields, so recheck them. */
524 if (fs->fs_avgfilesize <= 0) /* XXX */
525 fs->fs_avgfilesize = AVFILESIZ; /* XXX */
526 if (fs->fs_avgfpdir <= 0) /* XXX */
527 fs->fs_avgfpdir = AFPDIR; /* XXX */
530 * Step 3: re-read summary information from disk.
532 blks = howmany(fs->fs_cssize, fs->fs_fsize);
534 for (i = 0; i < blks; i += fs->fs_frag) {
536 if (i + fs->fs_frag > blks)
537 size = (blks - i) * fs->fs_fsize;
538 error = bread(devvp, fsbtodoff(fs, fs->fs_csaddr + i), size, &bp);
543 bcopy(bp->b_data, space, (uint)size);
544 space = (char *)space + size;
548 * We no longer know anything about clusters per cylinder group.
550 if (fs->fs_contigsumsize > 0) {
551 lp = fs->fs_maxcluster;
552 for (i = 0; i < fs->fs_ncg; i++)
553 *lp++ = fs->fs_contigsumsize;
558 scaninfo.devvp = devvp;
559 while (error == 0 && scaninfo.rescan) {
561 error = vmntvnodescan(mp, VMSC_GETVX,
562 NULL, ffs_reload_scan2, &scaninfo);
568 ffs_reload_scan2(struct mount *mp, struct vnode *vp, void *data)
570 struct scaninfo *info = data;
581 if (vinvalbuf(vp, 0, 0, 0))
582 panic("ffs_reload: dirty2");
584 * Step 6: re-read inode data for all active vnodes.
587 error = bread(info->devvp,
588 fsbtodoff(info->fs, ino_to_fsba(info->fs, ip->i_number)),
589 (int)info->fs->fs_bsize, &bp);
594 ip->i_din = *((struct ufs1_dinode *)bp->b_data +
595 ino_to_fsbo(info->fs, ip->i_number));
596 ip->i_effnlink = ip->i_nlink;
602 * Common code for mount and mountroot
605 ffs_mountfs(struct vnode *devvp, struct mount *mp, struct malloc_type *mtype)
607 struct ufsmount *ump;
611 struct partinfo dpart;
613 int error, i, blks, size, ronly;
615 uint64_t maxfilesize; /* XXX */
618 * Disallow multiple mounts of the same device.
619 * Disallow mounting of a device that is currently in use
620 * Flush out any old buffers remaining from a previous use.
622 error = vfs_mountedon(devvp);
625 if (vcount(devvp) > 0)
627 vn_lock(devvp, LK_EXCLUSIVE | LK_RETRY);
628 error = vinvalbuf(devvp, V_SAVE, 0, 0);
633 ronly = (mp->mnt_flag & MNT_RDONLY) != 0;
634 vn_lock(devvp, LK_EXCLUSIVE | LK_RETRY);
635 error = VOP_OPEN(devvp, ronly ? FREAD : FREAD|FWRITE, FSCRED, NULL);
640 if (dev->si_iosize_max != 0)
641 mp->mnt_iosize_max = dev->si_iosize_max;
642 if (mp->mnt_iosize_max > MAXPHYS)
643 mp->mnt_iosize_max = MAXPHYS;
646 * Filesystem supports native FSMIDs
648 mp->mnt_kern_flag |= MNTK_FSMID;
651 * The backing device must be VMIO-capable because we use getblk().
652 * NOTE: the MFS driver now returns a VMIO-enabled descriptor.
653 * The VOP_OPEN() call above should have associated a VM object
656 if (devvp->v_object == NULL)
657 panic("ffs_reload: devvp has no VM object!");
659 if (VOP_IOCTL(devvp, DIOCGPART, (caddr_t)&dpart, FREAD,
660 proc0.p_ucred, NULL) != 0) {
663 size = dpart.media_blksize;
668 if ((error = bread(devvp, SBOFF, SBSIZE, &bp)) != 0)
670 fs = (struct fs *)bp->b_data;
671 if (fs->fs_magic != FS_MAGIC || fs->fs_bsize > MAXBSIZE ||
672 fs->fs_bsize < sizeof(struct fs)) {
673 error = EINVAL; /* XXX needs translation */
677 fs->fs_flags &= ~FS_UNCLEAN;
678 if (fs->fs_clean == 0) {
679 fs->fs_flags |= FS_UNCLEAN;
680 if (ronly || (mp->mnt_flag & MNT_FORCE)) {
682 "WARNING: %s was not properly dismounted\n",
686 "WARNING: R/W mount of %s denied. Filesystem is not clean - run fsck\n",
692 /* XXX updating 4.2 FFS superblocks trashes rotational layout tables */
693 if (fs->fs_postblformat == FS_42POSTBLFMT && !ronly) {
694 error = EROFS; /* needs translation */
697 ump = kmalloc(sizeof *ump, M_UFSMNT, M_WAITOK | M_ZERO);
698 ump->um_malloctype = mtype;
699 ump->um_i_effnlink_valid = 1;
700 ump->um_fs = kmalloc((u_long)fs->fs_sbsize, M_UFSMNT,
702 bcopy(bp->b_data, ump->um_fs, (uint)fs->fs_sbsize);
703 if (fs->fs_sbsize < SBSIZE)
704 bp->b_flags |= B_INVAL;
708 fs->fs_ronly = ronly;
709 size = fs->fs_cssize;
710 blks = howmany(size, fs->fs_fsize);
711 if (fs->fs_contigsumsize > 0)
712 size += fs->fs_ncg * sizeof(int32_t);
713 size += fs->fs_ncg * sizeof(uint8_t);
714 space = kmalloc((u_long)size, M_UFSMNT, M_WAITOK);
716 for (i = 0; i < blks; i += fs->fs_frag) {
718 if (i + fs->fs_frag > blks)
719 size = (blks - i) * fs->fs_fsize;
720 if ((error = bread(devvp, fsbtodoff(fs, fs->fs_csaddr + i),
722 kfree(fs->fs_csp, M_UFSMNT);
725 bcopy(bp->b_data, space, (uint)size);
726 space = (char *)space + size;
730 if (fs->fs_contigsumsize > 0) {
731 fs->fs_maxcluster = lp = space;
732 for (i = 0; i < fs->fs_ncg; i++)
733 *lp++ = fs->fs_contigsumsize;
736 size = fs->fs_ncg * sizeof(uint8_t);
737 fs->fs_contigdirs = (uint8_t *)space;
738 bzero(fs->fs_contigdirs, size);
739 /* Compatibility for old filesystems XXX */
740 if (fs->fs_avgfilesize <= 0) /* XXX */
741 fs->fs_avgfilesize = AVFILESIZ; /* XXX */
742 if (fs->fs_avgfpdir <= 0) /* XXX */
743 fs->fs_avgfpdir = AFPDIR; /* XXX */
744 mp->mnt_data = (qaddr_t)ump;
745 mp->mnt_stat.f_fsid.val[0] = fs->fs_id[0];
746 mp->mnt_stat.f_fsid.val[1] = fs->fs_id[1];
747 if (fs->fs_id[0] == 0 || fs->fs_id[1] == 0 ||
748 vfs_getvfs(&mp->mnt_stat.f_fsid))
750 mp->mnt_maxsymlinklen = fs->fs_maxsymlinklen;
751 mp->mnt_flag |= MNT_LOCAL;
754 ump->um_devvp = devvp;
755 ump->um_nindir = fs->fs_nindir;
756 ump->um_bptrtodb = fs->fs_fsbtodb;
757 ump->um_seqinc = fs->fs_frag;
758 for (i = 0; i < MAXQUOTAS; i++)
759 ump->um_quotas[i] = NULLVP;
760 dev->si_mountpoint = mp;
763 /* restore "last mounted on" here */
764 bzero(fs->fs_fsmnt, sizeof(fs->fs_fsmnt));
765 ksnprintf(fs->fs_fsmnt, sizeof(fs->fs_fsmnt),
766 "%s", mp->mnt_stat.f_mntonname);
768 if( mp->mnt_flag & MNT_ROOTFS) {
770 * Root mount; update timestamp in mount structure.
771 * this will be used by the common root mount code
772 * to update the system clock.
774 mp->mnt_time = fs->fs_time;
777 ump->um_savedmaxfilesize = fs->fs_maxfilesize; /* XXX */
778 maxfilesize = (uint64_t)0x40000000 * fs->fs_bsize - 1; /* XXX */
779 /* Enforce limit caused by vm object backing (32 bits vm_pindex_t). */
780 if (maxfilesize > (uint64_t)0x80000000u * PAGE_SIZE - 1)
781 maxfilesize = (uint64_t)0x80000000u * PAGE_SIZE - 1;
782 if (fs->fs_maxfilesize > maxfilesize) /* XXX */
783 fs->fs_maxfilesize = maxfilesize; /* XXX */
785 if ((fs->fs_flags & FS_DOSOFTDEP) &&
786 (error = softdep_mount(devvp, mp, fs)) != 0) {
787 kfree(fs->fs_csp, M_UFSMNT);
792 (void) ffs_sbupdate(ump, MNT_WAIT);
794 vfs_add_vnodeops(mp, &ffs_vnode_vops, &mp->mnt_vn_norm_ops);
795 vfs_add_vnodeops(mp, &ffs_spec_vops, &mp->mnt_vn_spec_ops);
796 vfs_add_vnodeops(mp, &ffs_fifo_vops, &mp->mnt_vn_fifo_ops);
800 dev->si_mountpoint = NULL;
803 VOP_CLOSE(devvp, ronly ? FREAD : FREAD|FWRITE);
805 kfree(ump->um_fs, M_UFSMNT);
806 kfree(ump, M_UFSMNT);
807 mp->mnt_data = (qaddr_t)0;
813 * Sanity checks for old filesystems.
815 * XXX - goes away some day.
818 ffs_oldfscompat(struct fs *fs)
820 fs->fs_npsect = max(fs->fs_npsect, fs->fs_nsect); /* XXX */
821 fs->fs_interleave = max(fs->fs_interleave, 1); /* XXX */
822 if (fs->fs_postblformat == FS_42POSTBLFMT) /* XXX */
823 fs->fs_nrpos = 8; /* XXX */
824 if (fs->fs_inodefmt < FS_44INODEFMT) { /* XXX */
827 uint64_t sizepb = fs->fs_bsize; /* XXX */
829 fs->fs_maxfilesize = fs->fs_bsize * NDADDR - 1; /* XXX */
830 for (i = 0; i < NIADDR; i++) { /* XXX */
831 sizepb *= NINDIR(fs); /* XXX */
832 fs->fs_maxfilesize += sizepb; /* XXX */
835 fs->fs_maxfilesize = (u_quad_t) 1LL << 39;
836 fs->fs_qbmask = ~fs->fs_bmask; /* XXX */
837 fs->fs_qfmask = ~fs->fs_fmask; /* XXX */
843 * unmount system call
846 ffs_unmount(struct mount *mp, int mntflags)
848 struct ufsmount *ump;
853 if (mntflags & MNT_FORCE) {
856 if (mp->mnt_flag & MNT_SOFTDEP) {
857 if ((error = softdep_flushfiles(mp, flags)) != 0)
860 if ((error = ffs_flushfiles(mp, flags)) != 0)
865 if (fs->fs_ronly == 0) {
866 fs->fs_clean = fs->fs_flags & FS_UNCLEAN ? 0 : 1;
867 error = ffs_sbupdate(ump, MNT_WAIT);
873 ump->um_devvp->v_rdev->si_mountpoint = NULL;
875 vinvalbuf(ump->um_devvp, V_SAVE, 0, 0);
876 error = VOP_CLOSE(ump->um_devvp, fs->fs_ronly ? FREAD : FREAD|FWRITE);
878 vrele(ump->um_devvp);
880 kfree(fs->fs_csp, M_UFSMNT);
882 kfree(ump, M_UFSMNT);
883 mp->mnt_data = (qaddr_t)0;
884 mp->mnt_flag &= ~MNT_LOCAL;
889 * Flush out all the files in a filesystem.
892 ffs_flushfiles(struct mount *mp, int flags)
894 struct ufsmount *ump;
899 if (mp->mnt_flag & MNT_QUOTA) {
901 error = vflush(mp, 0, SKIPSYSTEM|flags);
904 /* Find out how many quota files we have open. */
905 for (i = 0; i < MAXQUOTAS; i++) {
906 if (ump->um_quotas[i] == NULLVP)
911 * Here we fall through to vflush again to ensure
912 * that we have gotten rid of all the system vnodes.
917 * Flush all the files.
919 if ((error = vflush(mp, 0, flags)) != 0)
922 * Flush filesystem metadata.
924 vn_lock(ump->um_devvp, LK_EXCLUSIVE | LK_RETRY);
925 error = VOP_FSYNC(ump->um_devvp, MNT_WAIT, 0);
926 vn_unlock(ump->um_devvp);
931 * Get filesystem statistics.
934 ffs_statfs(struct mount *mp, struct statfs *sbp, struct ucred *cred)
936 struct ufsmount *ump;
941 if (fs->fs_magic != FS_MAGIC)
943 sbp->f_bsize = fs->fs_fsize;
944 sbp->f_iosize = fs->fs_bsize;
945 sbp->f_blocks = fs->fs_dsize;
946 sbp->f_bfree = fs->fs_cstotal.cs_nbfree * fs->fs_frag +
947 fs->fs_cstotal.cs_nffree;
948 sbp->f_bavail = freespace(fs, fs->fs_minfree);
949 sbp->f_files = fs->fs_ncg * fs->fs_ipg - ROOTINO;
950 sbp->f_ffree = fs->fs_cstotal.cs_nifree;
951 if (sbp != &mp->mnt_stat) {
952 sbp->f_type = mp->mnt_vfc->vfc_typenum;
953 bcopy((caddr_t)mp->mnt_stat.f_mntfromname,
954 (caddr_t)&sbp->f_mntfromname[0], MNAMELEN);
960 * Go through the disk queues to initiate sandbagged IO;
961 * go through the inodes to write those that have been modified;
962 * initiate the writing of the super block if it has been modified.
964 * Note: we are always called with the filesystem marked `MPBUSY'.
968 static int ffs_sync_scan1(struct mount *mp, struct vnode *vp, void *data);
969 static int ffs_sync_scan2(struct mount *mp, struct vnode *vp, void *data);
972 ffs_sync(struct mount *mp, int waitfor)
974 struct ufsmount *ump = VFSTOUFS(mp);
977 struct scaninfo scaninfo;
980 if (fs->fs_fmod != 0 && fs->fs_ronly != 0) { /* XXX */
981 kprintf("fs = %s\n", fs->fs_fsmnt);
982 panic("ffs_sync: rofs mod");
986 * Write back each (modified) inode.
988 scaninfo.allerror = 0;
990 scaninfo.waitfor = waitfor;
991 while (scaninfo.rescan) {
993 vmntvnodescan(mp, VMSC_GETVP|VMSC_NOWAIT,
994 ffs_sync_scan1, ffs_sync_scan2, &scaninfo);
998 * Force stale filesystem control information to be flushed.
1000 if ((waitfor & MNT_LAZY) == 0) {
1001 if (ump->um_mountp->mnt_flag & MNT_SOFTDEP)
1002 waitfor = MNT_NOWAIT;
1003 vn_lock(ump->um_devvp, LK_EXCLUSIVE | LK_RETRY);
1004 if ((error = VOP_FSYNC(ump->um_devvp, waitfor, 0)) != 0)
1005 scaninfo.allerror = error;
1006 vn_unlock(ump->um_devvp);
1012 * Write back modified superblock.
1014 if (fs->fs_fmod != 0 && (error = ffs_sbupdate(ump, waitfor)) != 0)
1015 scaninfo.allerror = error;
1016 return (scaninfo.allerror);
1020 ffs_sync_scan1(struct mount *mp, struct vnode *vp, void *data)
1025 * Depend on the mount list's vnode lock to keep things stable
1026 * enough for a quick test. Since there might be hundreds of
1027 * thousands of vnodes, we cannot afford even a subroutine
1028 * call unless there's a good chance that we have work to do.
1031 /* Restart out whole search if this guy is locked
1032 * or is being reclaimed.
1034 if (vp->v_type == VNON || ((ip->i_flag &
1035 (IN_ACCESS | IN_CHANGE | IN_MODIFIED | IN_UPDATE)) == 0 &&
1036 RB_EMPTY(&vp->v_rbdirty_tree))) {
1043 ffs_sync_scan2(struct mount *mp, struct vnode *vp, void *data)
1045 struct scaninfo *info = data;
1050 * We have to recheck after having obtained the vnode interlock.
1053 if (vp->v_type == VNON || vp->v_type == VBAD ||
1055 (IN_ACCESS | IN_CHANGE | IN_MODIFIED | IN_UPDATE)) == 0 &&
1056 RB_EMPTY(&vp->v_rbdirty_tree))) {
1059 if (vp->v_type != VCHR) {
1060 if ((error = VOP_FSYNC(vp, info->waitfor, 0)) != 0)
1061 info->allerror = error;
1064 * We must reference the vp to prevent it from
1065 * getting ripped out from under ffs_update, since
1066 * we are not holding a vnode lock.
1068 /* ffs_update(vp, waitfor == MNT_WAIT); */
1075 * Look up a FFS dinode number to find its incore vnode, otherwise read it
1076 * in from disk. If it is in core, wait for the lock bit to clear, then
1077 * return the inode locked. Detection and handling of mount points must be
1078 * done by the calling routine.
1082 ffs_vget(struct mount *mp, struct vnode *dvp, ino_t ino, struct vnode **vpp)
1086 struct ufsmount *ump;
1095 if ((*vpp = ufs_ihashget(dev, ino)) != NULL) {
1100 * If this MALLOC() is performed after the getnewvnode()
1101 * it might block, leaving a vnode with a NULL v_data to be
1102 * found by ffs_sync() if a sync happens to fire right then,
1103 * which will cause a panic because ffs_sync() blindly
1104 * dereferences vp->v_data (as well it should).
1106 * XXX this may no longer be true since getnewvnode returns a
1107 * VX locked vnode now.
1109 ip = kmalloc(sizeof(struct inode), ump->um_malloctype, M_WAITOK);
1111 /* Allocate a new vnode/inode. */
1112 error = getnewvnode(VT_UFS, mp, &vp, VLKTIMEOUT, LK_CANRECURSE);
1115 kfree(ip, ump->um_malloctype);
1118 bzero((caddr_t)ip, sizeof(struct inode));
1120 ip->i_fs = fs = ump->um_fs;
1126 for (i = 0; i < MAXQUOTAS; i++)
1127 ip->i_dquot[i] = NODQUOT;
1132 * Insert it into the inode hash table and check for a collision.
1133 * If a collision occurs, throw away the vnode and try again.
1135 if (ufs_ihashins(ip) != 0) {
1136 kprintf("debug: ufs ihashins collision, retrying inode %ld\n",
1137 (long)ip->i_number);
1140 kfree(ip, ump->um_malloctype);
1145 /* Read in the disk contents for the inode, copy into the inode. */
1146 error = bread(ump->um_devvp, fsbtodoff(fs, ino_to_fsba(fs, ino)),
1147 (int)fs->fs_bsize, &bp);
1150 * The inode does not contain anything useful, so it would
1151 * be misleading to leave it on its hash chain. With mode
1152 * still zero, it will be unlinked and returned to the free
1161 ip->i_din = *((struct ufs1_dinode *)bp->b_data + ino_to_fsbo(fs, ino));
1162 if (DOINGSOFTDEP(vp))
1163 softdep_load_inodeblock(ip);
1165 ip->i_effnlink = ip->i_nlink;
1169 * Initialize the vnode from the inode, check for aliases.
1170 * Note that the underlying vnode may have changed.
1172 error = ufs_vinit(mp, &vp);
1180 * Finish inode initialization now that aliasing has been resolved.
1182 ip->i_devvp = ump->um_devvp;
1185 * Set up a generation number for this inode if it does not
1186 * already have one. This should only happen on old filesystems.
1188 if (ip->i_gen == 0) {
1189 ip->i_gen = krandom() / 2 + 1;
1190 if ((vp->v_mount->mnt_flag & MNT_RDONLY) == 0)
1191 ip->i_flag |= IN_MODIFIED;
1194 * Ensure that uid and gid are correct. This is a temporary
1195 * fix until fsck has been changed to do the update.
1197 if (fs->fs_inodefmt < FS_44INODEFMT) { /* XXX */
1198 ip->i_uid = ip->i_din.di_ouid; /* XXX */
1199 ip->i_gid = ip->i_din.di_ogid; /* XXX */
1203 * return a VX locked and refd vnode (VX == same as normal vget()
1204 * vnode so we are ok)
1211 * File handle to vnode
1213 * Have to be really careful about stale file handles:
1214 * - check that the inode number is valid
1215 * - call ffs_vget() to get the locked inode
1216 * - check for an unallocated inode (i_mode == 0)
1217 * - check that the given client host has export rights and return
1218 * those rights via. exflagsp and credanonp
1221 ffs_fhtovp(struct mount *mp, struct vnode *rootvp,
1222 struct fid *fhp, struct vnode **vpp)
1227 ufhp = (struct ufid *)fhp;
1228 fs = VFSTOUFS(mp)->um_fs;
1229 if (ufhp->ufid_ino < ROOTINO ||
1230 ufhp->ufid_ino >= fs->fs_ncg * fs->fs_ipg)
1232 return (ufs_fhtovp(mp, rootvp, ufhp, vpp));
1236 * Vnode pointer to File handle
1240 ffs_vptofh(struct vnode *vp, struct fid *fhp)
1246 ufhp = (struct ufid *)fhp;
1247 ufhp->ufid_len = sizeof(struct ufid);
1248 ufhp->ufid_ino = ip->i_number;
1249 ufhp->ufid_gen = ip->i_gen;
1254 * Initialize the filesystem; just use ufs_init.
1257 ffs_init(struct vfsconf *vfsp)
1259 softdep_initialize();
1260 kmalloc_raise_limit(M_FFSNODE, 0);
1261 return (ufs_init(vfsp));
1265 * Write a superblock and associated information back to disk.
1268 ffs_sbupdate(struct ufsmount *mp, int waitfor)
1270 struct fs *dfs, *fs = mp->um_fs;
1274 int i, size, error, allerror = 0;
1277 * First write back the summary information.
1279 * NOTE: the getblk is relative to the device vnode so bio1
1280 * contains the device block number.
1282 blks = howmany(fs->fs_cssize, fs->fs_fsize);
1284 for (i = 0; i < blks; i += fs->fs_frag) {
1285 size = fs->fs_bsize;
1286 if (i + fs->fs_frag > blks)
1287 size = (blks - i) * fs->fs_fsize;
1288 bp = getblk(mp->um_devvp, fsbtodoff(fs, fs->fs_csaddr + i),
1290 bcopy(space, bp->b_data, (uint)size);
1291 space = (char *)space + size;
1292 if (waitfor != MNT_WAIT)
1294 else if ((error = bwrite(bp)) != 0)
1298 * Now write back the superblock itself. If any errors occurred
1299 * up to this point, then fail so that the superblock avoids
1300 * being written out as clean.
1304 bp = getblk(mp->um_devvp, SBOFF, (int)fs->fs_sbsize, 0, 0);
1306 fs->fs_time = time_second;
1307 bcopy((caddr_t)fs, bp->b_data, (uint)fs->fs_sbsize);
1308 /* Restore compatibility to old filesystems. XXX */
1309 dfs = (struct fs *)bp->b_data; /* XXX */
1310 if (fs->fs_postblformat == FS_42POSTBLFMT) /* XXX */
1311 dfs->fs_nrpos = -1; /* XXX */
1312 if (fs->fs_inodefmt < FS_44INODEFMT) { /* XXX */
1313 int32_t *lp, tmp; /* XXX */
1315 lp = (int32_t *)&dfs->fs_qbmask; /* XXX */
1316 tmp = lp[4]; /* XXX */
1317 for (i = 4; i > 0; i--) /* XXX */
1318 lp[i] = lp[i-1]; /* XXX */
1319 lp[0] = tmp; /* XXX */
1321 dfs->fs_maxfilesize = mp->um_savedmaxfilesize; /* XXX */
1322 if (waitfor != MNT_WAIT)
1324 else if ((error = bwrite(bp)) != 0)