VM: only single page chunks
[minix.git] / sys / ufs / ext2fs / ext2fs_vfsops.c
blob76f6dd5a5f6ca3142eae67b3e9404d9d9ba4495d
1 /* $NetBSD: ext2fs_vfsops.c,v 1.162 2011/11/14 18:35:14 hannken Exp $ */
3 /*
4 * Copyright (c) 1989, 1991, 1993, 1994
5 * The Regents of the University of California. All rights reserved.
7 * Redistribution and use in source and binary forms, with or without
8 * modification, are permitted provided that the following conditions
9 * are met:
10 * 1. Redistributions of source code must retain the above copyright
11 * notice, this list of conditions and the following disclaimer.
12 * 2. Redistributions in binary form must reproduce the above copyright
13 * notice, this list of conditions and the following disclaimer in the
14 * documentation and/or other materials provided with the distribution.
15 * 3. Neither the name of the University nor the names of its contributors
16 * may be used to endorse or promote products derived from this software
17 * without specific prior written permission.
19 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
20 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
21 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
22 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
23 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
24 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
25 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
26 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
27 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
28 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
29 * SUCH DAMAGE.
31 * @(#)ffs_vfsops.c 8.14 (Berkeley) 11/28/94
32 * Modified for ext2fs by Manuel Bouyer.
36 * Copyright (c) 1997 Manuel Bouyer.
38 * Redistribution and use in source and binary forms, with or without
39 * modification, are permitted provided that the following conditions
40 * are met:
41 * 1. Redistributions of source code must retain the above copyright
42 * notice, this list of conditions and the following disclaimer.
43 * 2. Redistributions in binary form must reproduce the above copyright
44 * notice, this list of conditions and the following disclaimer in the
45 * documentation and/or other materials provided with the distribution.
47 * THIS SOFTWARE IS PROVIDED BY THE AUTHOR ``AS IS'' AND ANY EXPRESS OR
48 * IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES
49 * OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE DISCLAIMED.
50 * IN NO EVENT SHALL THE AUTHOR BE LIABLE FOR ANY DIRECT, INDIRECT,
51 * INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT
52 * NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
53 * DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
54 * THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
55 * (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF
56 * THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
58 * @(#)ffs_vfsops.c 8.14 (Berkeley) 11/28/94
59 * Modified for ext2fs by Manuel Bouyer.
62 #include <sys/cdefs.h>
63 __KERNEL_RCSID(0, "$NetBSD: ext2fs_vfsops.c,v 1.162 2011/11/14 18:35:14 hannken Exp $");
65 #if defined(_KERNEL_OPT)
66 #include "opt_compat_netbsd.h"
67 #endif
69 #include <sys/param.h>
70 #include <sys/systm.h>
71 #include <sys/sysctl.h>
72 #include <sys/namei.h>
73 #include <sys/proc.h>
74 #include <sys/kernel.h>
75 #include <sys/vnode.h>
76 #include <sys/socket.h>
77 #include <sys/mount.h>
78 #include <sys/buf.h>
79 #include <sys/device.h>
80 #include <sys/mbuf.h>
81 #include <sys/file.h>
82 #include <sys/disklabel.h>
83 #include <sys/ioctl.h>
84 #include <sys/errno.h>
85 #include <sys/malloc.h>
86 #include <sys/pool.h>
87 #include <sys/lock.h>
88 #include <sys/conf.h>
89 #include <sys/kauth.h>
90 #include <sys/module.h>
92 #include <miscfs/genfs/genfs.h>
93 #include <miscfs/specfs/specdev.h>
95 #include <ufs/ufs/quota.h>
96 #include <ufs/ufs/ufsmount.h>
97 #include <ufs/ufs/inode.h>
98 #include <ufs/ufs/dir.h>
99 #include <ufs/ufs/ufs_extern.h>
101 #include <ufs/ext2fs/ext2fs.h>
102 #include <ufs/ext2fs/ext2fs_dir.h>
103 #include <ufs/ext2fs/ext2fs_extern.h>
105 MODULE(MODULE_CLASS_VFS, ext2fs, "ffs");
107 int ext2fs_sbupdate(struct ufsmount *, int);
108 static int ext2fs_checksb(struct ext2fs *, int);
110 static struct sysctllog *ext2fs_sysctl_log;
112 extern const struct vnodeopv_desc ext2fs_vnodeop_opv_desc;
113 extern const struct vnodeopv_desc ext2fs_specop_opv_desc;
114 extern const struct vnodeopv_desc ext2fs_fifoop_opv_desc;
116 const struct vnodeopv_desc * const ext2fs_vnodeopv_descs[] = {
117 &ext2fs_vnodeop_opv_desc,
118 &ext2fs_specop_opv_desc,
119 &ext2fs_fifoop_opv_desc,
120 NULL,
123 struct vfsops ext2fs_vfsops = {
124 MOUNT_EXT2FS,
125 sizeof (struct ufs_args),
126 ext2fs_mount,
127 ufs_start,
128 ext2fs_unmount,
129 ufs_root,
130 ufs_quotactl,
131 ext2fs_statvfs,
132 ext2fs_sync,
133 ext2fs_vget,
134 ext2fs_fhtovp,
135 ext2fs_vptofh,
136 ext2fs_init,
137 ext2fs_reinit,
138 ext2fs_done,
139 ext2fs_mountroot,
140 (int (*)(struct mount *, struct vnode *, struct timespec *)) eopnotsupp,
141 vfs_stdextattrctl,
142 (void *)eopnotsupp, /* vfs_suspendctl */
143 genfs_renamelock_enter,
144 genfs_renamelock_exit,
145 (void *)eopnotsupp,
146 ext2fs_vnodeopv_descs,
148 { NULL, NULL },
151 static const struct genfs_ops ext2fs_genfsops = {
152 .gop_size = genfs_size,
153 .gop_alloc = ext2fs_gop_alloc,
154 .gop_write = genfs_gop_write,
155 .gop_markupdate = ufs_gop_markupdate,
158 static const struct ufs_ops ext2fs_ufsops = {
159 .uo_itimes = ext2fs_itimes,
160 .uo_update = ext2fs_update,
161 .uo_vfree = ext2fs_vfree,
162 .uo_unmark_vnode = (void (*)(vnode_t *))nullop,
165 /* Fill in the inode uid/gid from ext2 halves. */
166 void
167 ext2fs_set_inode_guid(struct inode *ip)
170 ip->i_gid = ip->i_e2fs_gid;
171 ip->i_uid = ip->i_e2fs_uid;
172 if (ip->i_e2fs->e2fs.e2fs_rev > E2FS_REV0) {
173 ip->i_gid |= ip->i_e2fs_gid_high << 16;
174 ip->i_uid |= ip->i_e2fs_uid_high << 16;
178 static int
179 ext2fs_modcmd(modcmd_t cmd, void *arg)
181 int error;
183 switch (cmd) {
184 case MODULE_CMD_INIT:
185 error = vfs_attach(&ext2fs_vfsops);
186 if (error != 0)
187 break;
188 sysctl_createv(&ext2fs_sysctl_log, 0, NULL, NULL,
189 CTLFLAG_PERMANENT,
190 CTLTYPE_NODE, "vfs", NULL,
191 NULL, 0, NULL, 0,
192 CTL_VFS, CTL_EOL);
193 sysctl_createv(&ext2fs_sysctl_log, 0, NULL, NULL,
194 CTLFLAG_PERMANENT,
195 CTLTYPE_NODE, "ext2fs",
196 SYSCTL_DESCR("Linux EXT2FS file system"),
197 NULL, 0, NULL, 0,
198 CTL_VFS, 17, CTL_EOL);
200 * XXX the "17" above could be dynamic, thereby eliminating
201 * one more instance of the "number to vfs" mapping problem,
202 * but "17" is the order as taken from sys/mount.h
204 break;
205 case MODULE_CMD_FINI:
206 error = vfs_detach(&ext2fs_vfsops);
207 if (error != 0)
208 break;
209 sysctl_teardown(&ext2fs_sysctl_log);
210 break;
211 default:
212 error = ENOTTY;
213 break;
216 return (error);
220 * XXX Same structure as FFS inodes? Should we share a common pool?
222 struct pool ext2fs_inode_pool;
223 struct pool ext2fs_dinode_pool;
225 extern u_long ext2gennumber;
227 void
228 ext2fs_init(void)
231 pool_init(&ext2fs_inode_pool, sizeof(struct inode), 0, 0, 0,
232 "ext2fsinopl", &pool_allocator_nointr, IPL_NONE);
233 pool_init(&ext2fs_dinode_pool, sizeof(struct ext2fs_dinode), 0, 0, 0,
234 "ext2dinopl", &pool_allocator_nointr, IPL_NONE);
235 ufs_init();
238 void
239 ext2fs_reinit(void)
241 ufs_reinit();
244 void
245 ext2fs_done(void)
248 ufs_done();
249 pool_destroy(&ext2fs_inode_pool);
250 pool_destroy(&ext2fs_dinode_pool);
254 * Called by main() when ext2fs is going to be mounted as root.
256 * Name is updated by mount(8) after booting.
258 #define ROOTNAME "root_device"
261 ext2fs_mountroot(void)
263 extern struct vnode *rootvp;
264 struct m_ext2fs *fs;
265 struct mount *mp;
266 struct ufsmount *ump;
267 int error;
269 if (device_class(root_device) != DV_DISK)
270 return (ENODEV);
272 if ((error = vfs_rootmountalloc(MOUNT_EXT2FS, "root_device", &mp))) {
273 vrele(rootvp);
274 return (error);
277 if ((error = ext2fs_mountfs(rootvp, mp)) != 0) {
278 vfs_unbusy(mp, false, NULL);
279 vfs_destroy(mp);
280 return (error);
282 mutex_enter(&mountlist_lock);
283 CIRCLEQ_INSERT_TAIL(&mountlist, mp, mnt_list);
284 mutex_exit(&mountlist_lock);
285 ump = VFSTOUFS(mp);
286 fs = ump->um_e2fs;
287 memset(fs->e2fs_fsmnt, 0, sizeof(fs->e2fs_fsmnt));
288 (void) copystr(mp->mnt_stat.f_mntonname, fs->e2fs_fsmnt,
289 sizeof(fs->e2fs_fsmnt) - 1, 0);
290 if (fs->e2fs.e2fs_rev > E2FS_REV0) {
291 memset(fs->e2fs.e2fs_fsmnt, 0, sizeof(fs->e2fs.e2fs_fsmnt));
292 (void) copystr(mp->mnt_stat.f_mntonname, fs->e2fs.e2fs_fsmnt,
293 sizeof(fs->e2fs.e2fs_fsmnt) - 1, 0);
295 (void)ext2fs_statvfs(mp, &mp->mnt_stat);
296 vfs_unbusy(mp, false, NULL);
297 setrootfstime((time_t)fs->e2fs.e2fs_wtime);
298 return (0);
302 * VFS Operations.
304 * mount system call
307 ext2fs_mount(struct mount *mp, const char *path, void *data, size_t *data_len)
309 struct lwp *l = curlwp;
310 struct vnode *devvp;
311 struct ufs_args *args = data;
312 struct ufsmount *ump = NULL;
313 struct m_ext2fs *fs;
314 size_t size;
315 int error = 0, flags, update;
316 mode_t accessmode;
318 if (*data_len < sizeof *args)
319 return EINVAL;
321 if (mp->mnt_flag & MNT_GETARGS) {
322 ump = VFSTOUFS(mp);
323 if (ump == NULL)
324 return EIO;
325 memset(args, 0, sizeof *args);
326 args->fspec = NULL;
327 *data_len = sizeof *args;
328 return 0;
331 update = mp->mnt_flag & MNT_UPDATE;
333 /* Check arguments */
334 if (args->fspec != NULL) {
336 * Look up the name and verify that it's sane.
338 error = namei_simple_user(args->fspec,
339 NSM_FOLLOW_NOEMULROOT, &devvp);
340 if (error != 0)
341 return (error);
343 if (!update) {
345 * Be sure this is a valid block device
347 if (devvp->v_type != VBLK)
348 error = ENOTBLK;
349 else if (bdevsw_lookup(devvp->v_rdev) == NULL)
350 error = ENXIO;
351 } else {
353 * Be sure we're still naming the same device
354 * used for our initial mount
356 ump = VFSTOUFS(mp);
357 if (devvp != ump->um_devvp) {
358 if (devvp->v_rdev != ump->um_devvp->v_rdev)
359 error = EINVAL;
360 else {
361 vrele(devvp);
362 devvp = ump->um_devvp;
363 vref(devvp);
367 } else {
368 if (!update) {
369 /* New mounts must have a filename for the device */
370 return (EINVAL);
371 } else {
372 ump = VFSTOUFS(mp);
373 devvp = ump->um_devvp;
374 vref(devvp);
379 * If mount by non-root, then verify that user has necessary
380 * permissions on the device.
382 * Permission to update a mount is checked higher, so here we presume
383 * updating the mount is okay (for example, as far as securelevel goes)
384 * which leaves us with the normal check.
386 if (error == 0) {
387 accessmode = VREAD;
388 if (update ?
389 (mp->mnt_iflag & IMNT_WANTRDWR) != 0 :
390 (mp->mnt_flag & MNT_RDONLY) == 0)
391 accessmode |= VWRITE;
392 vn_lock(devvp, LK_EXCLUSIVE | LK_RETRY);
393 error = genfs_can_mount(devvp, accessmode, l->l_cred);
394 VOP_UNLOCK(devvp);
397 if (error) {
398 vrele(devvp);
399 return (error);
402 if (!update) {
403 int xflags;
405 if (mp->mnt_flag & MNT_RDONLY)
406 xflags = FREAD;
407 else
408 xflags = FREAD|FWRITE;
409 vn_lock(devvp, LK_EXCLUSIVE | LK_RETRY);
410 error = VOP_OPEN(devvp, xflags, FSCRED);
411 VOP_UNLOCK(devvp);
412 if (error)
413 goto fail;
414 error = ext2fs_mountfs(devvp, mp);
415 if (error) {
416 vn_lock(devvp, LK_EXCLUSIVE | LK_RETRY);
417 (void)VOP_CLOSE(devvp, xflags, NOCRED);
418 VOP_UNLOCK(devvp);
419 goto fail;
422 ump = VFSTOUFS(mp);
423 fs = ump->um_e2fs;
424 } else {
426 * Update the mount.
430 * The initial mount got a reference on this
431 * device, so drop the one obtained via
432 * namei(), above.
434 vrele(devvp);
436 ump = VFSTOUFS(mp);
437 fs = ump->um_e2fs;
438 if (fs->e2fs_ronly == 0 && (mp->mnt_flag & MNT_RDONLY)) {
440 * Changing from r/w to r/o
442 flags = WRITECLOSE;
443 if (mp->mnt_flag & MNT_FORCE)
444 flags |= FORCECLOSE;
445 error = ext2fs_flushfiles(mp, flags);
446 if (error == 0 &&
447 ext2fs_cgupdate(ump, MNT_WAIT) == 0 &&
448 (fs->e2fs.e2fs_state & E2FS_ERRORS) == 0) {
449 fs->e2fs.e2fs_state = E2FS_ISCLEAN;
450 (void) ext2fs_sbupdate(ump, MNT_WAIT);
452 if (error)
453 return (error);
454 fs->e2fs_ronly = 1;
457 if (mp->mnt_flag & MNT_RELOAD) {
458 error = ext2fs_reload(mp, l->l_cred, l);
459 if (error)
460 return (error);
463 if (fs->e2fs_ronly && (mp->mnt_iflag & IMNT_WANTRDWR)) {
465 * Changing from read-only to read/write
467 fs->e2fs_ronly = 0;
468 if (fs->e2fs.e2fs_state == E2FS_ISCLEAN)
469 fs->e2fs.e2fs_state = 0;
470 else
471 fs->e2fs.e2fs_state = E2FS_ERRORS;
472 fs->e2fs_fmod = 1;
474 if (args->fspec == NULL)
475 return 0;
478 error = set_statvfs_info(path, UIO_USERSPACE, args->fspec,
479 UIO_USERSPACE, mp->mnt_op->vfs_name, mp, l);
480 (void) copystr(mp->mnt_stat.f_mntonname, fs->e2fs_fsmnt,
481 sizeof(fs->e2fs_fsmnt) - 1, &size);
482 memset(fs->e2fs_fsmnt + size, 0, sizeof(fs->e2fs_fsmnt) - size);
483 if (fs->e2fs.e2fs_rev > E2FS_REV0) {
484 (void) copystr(mp->mnt_stat.f_mntonname, fs->e2fs.e2fs_fsmnt,
485 sizeof(fs->e2fs.e2fs_fsmnt) - 1, &size);
486 memset(fs->e2fs.e2fs_fsmnt, 0,
487 sizeof(fs->e2fs.e2fs_fsmnt) - size);
489 if (fs->e2fs_fmod != 0) { /* XXX */
490 fs->e2fs_fmod = 0;
491 if (fs->e2fs.e2fs_state == 0)
492 fs->e2fs.e2fs_wtime = time_second;
493 else
494 printf("%s: file system not clean; please fsck(8)\n",
495 mp->mnt_stat.f_mntfromname);
496 (void) ext2fs_cgupdate(ump, MNT_WAIT);
498 return (error);
500 fail:
501 vrele(devvp);
502 return (error);
506 * Reload all incore data for a filesystem (used after running fsck on
507 * the root filesystem and finding things to fix). The filesystem must
508 * be mounted read-only.
510 * Things to do to update the mount:
511 * 1) invalidate all cached meta-data.
512 * 2) re-read superblock from disk.
513 * 3) re-read summary information from disk.
514 * 4) invalidate all inactive vnodes.
515 * 5) invalidate all cached file data.
516 * 6) re-read inode data for all active vnodes.
519 ext2fs_reload(struct mount *mp, kauth_cred_t cred, struct lwp *l)
521 struct vnode *vp, *mvp, *devvp;
522 struct inode *ip;
523 struct buf *bp;
524 struct m_ext2fs *fs;
525 struct ext2fs *newfs;
526 int i, error;
527 void *cp;
528 struct ufsmount *ump;
530 if ((mp->mnt_flag & MNT_RDONLY) == 0)
531 return (EINVAL);
533 ump = VFSTOUFS(mp);
535 * Step 1: invalidate all cached meta-data.
537 devvp = ump->um_devvp;
538 vn_lock(devvp, LK_EXCLUSIVE | LK_RETRY);
539 error = vinvalbuf(devvp, 0, cred, l, 0, 0);
540 VOP_UNLOCK(devvp);
541 if (error)
542 panic("ext2fs_reload: dirty1");
544 * Step 2: re-read superblock from disk.
546 error = bread(devvp, SBLOCK, SBSIZE, NOCRED, 0, &bp);
547 if (error) {
548 brelse(bp, 0);
549 return (error);
551 newfs = (struct ext2fs *)bp->b_data;
552 error = ext2fs_checksb(newfs, (mp->mnt_flag & MNT_RDONLY) != 0);
553 if (error) {
554 brelse(bp, 0);
555 return (error);
558 fs = ump->um_e2fs;
560 * copy in new superblock, and compute in-memory values
562 e2fs_sbload(newfs, &fs->e2fs);
563 fs->e2fs_ncg =
564 howmany(fs->e2fs.e2fs_bcount - fs->e2fs.e2fs_first_dblock,
565 fs->e2fs.e2fs_bpg);
566 fs->e2fs_fsbtodb = fs->e2fs.e2fs_log_bsize + LOG_MINBSIZE - DEV_BSHIFT;
567 fs->e2fs_bsize = MINBSIZE << fs->e2fs.e2fs_log_bsize;
568 fs->e2fs_bshift = LOG_MINBSIZE + fs->e2fs.e2fs_log_bsize;
569 fs->e2fs_qbmask = fs->e2fs_bsize - 1;
570 fs->e2fs_bmask = ~fs->e2fs_qbmask;
571 fs->e2fs_ngdb =
572 howmany(fs->e2fs_ncg, fs->e2fs_bsize / sizeof(struct ext2_gd));
573 fs->e2fs_ipb = fs->e2fs_bsize / EXT2_DINODE_SIZE(fs);
574 fs->e2fs_itpg = fs->e2fs.e2fs_ipg / fs->e2fs_ipb;
575 brelse(bp, 0);
578 * Step 3: re-read summary information from disk.
581 for (i = 0; i < fs->e2fs_ngdb; i++) {
582 error = bread(devvp ,
583 fsbtodb(fs, fs->e2fs.e2fs_first_dblock +
584 1 /* superblock */ + i),
585 fs->e2fs_bsize, NOCRED, 0, &bp);
586 if (error) {
587 brelse(bp, 0);
588 return (error);
590 e2fs_cgload((struct ext2_gd *)bp->b_data,
591 &fs->e2fs_gd[i * fs->e2fs_bsize / sizeof(struct ext2_gd)],
592 fs->e2fs_bsize);
593 brelse(bp, 0);
596 /* Allocate a marker vnode. */
597 mvp = vnalloc(mp);
599 * NOTE: not using the TAILQ_FOREACH here since in this loop vgone()
600 * and vclean() can be called indirectly
602 mutex_enter(&mntvnode_lock);
603 loop:
604 for (vp = TAILQ_FIRST(&mp->mnt_vnodelist); vp; vp = vunmark(mvp)) {
605 vmark(mvp, vp);
606 if (vp->v_mount != mp || vismarker(vp))
607 continue;
609 * Step 4: invalidate all inactive vnodes.
611 if (vrecycle(vp, &mntvnode_lock, l)) {
612 mutex_enter(&mntvnode_lock);
613 (void)vunmark(mvp);
614 goto loop;
617 * Step 5: invalidate all cached file data.
619 mutex_enter(vp->v_interlock);
620 mutex_exit(&mntvnode_lock);
621 if (vget(vp, LK_EXCLUSIVE)) {
622 mutex_enter(&mntvnode_lock);
623 (void)vunmark(mvp);
624 goto loop;
626 if (vinvalbuf(vp, 0, cred, l, 0, 0))
627 panic("ext2fs_reload: dirty2");
629 * Step 6: re-read inode data for all active vnodes.
631 ip = VTOI(vp);
632 error = bread(devvp, fsbtodb(fs, ino_to_fsba(fs, ip->i_number)),
633 (int)fs->e2fs_bsize, NOCRED, 0, &bp);
634 if (error) {
635 vput(vp);
636 mutex_enter(&mntvnode_lock);
637 (void)vunmark(mvp);
638 break;
640 cp = (char *)bp->b_data +
641 (ino_to_fsbo(fs, ip->i_number) * EXT2_DINODE_SIZE(fs));
642 e2fs_iload((struct ext2fs_dinode *)cp, ip->i_din.e2fs_din);
643 ext2fs_set_inode_guid(ip);
644 brelse(bp, 0);
645 vput(vp);
646 mutex_enter(&mntvnode_lock);
648 mutex_exit(&mntvnode_lock);
649 vnfree(mvp);
650 return (error);
654 * Common code for mount and mountroot
657 ext2fs_mountfs(struct vnode *devvp, struct mount *mp)
659 struct lwp *l = curlwp;
660 struct ufsmount *ump;
661 struct buf *bp;
662 struct ext2fs *fs;
663 struct m_ext2fs *m_fs;
664 dev_t dev;
665 int error, i, ronly;
666 kauth_cred_t cred;
667 struct proc *p;
669 dev = devvp->v_rdev;
670 p = l ? l->l_proc : NULL;
671 cred = l ? l->l_cred : NOCRED;
673 /* Flush out any old buffers remaining from a previous use. */
674 vn_lock(devvp, LK_EXCLUSIVE | LK_RETRY);
675 error = vinvalbuf(devvp, V_SAVE, cred, l, 0, 0);
676 VOP_UNLOCK(devvp);
677 if (error)
678 return (error);
680 ronly = (mp->mnt_flag & MNT_RDONLY) != 0;
682 bp = NULL;
683 ump = NULL;
685 #ifdef DEBUG_EXT2
686 printf("ext2 sb size: %zu\n", sizeof(struct ext2fs));
687 #endif
688 error = bread(devvp, SBLOCK, SBSIZE, cred, 0, &bp);
689 if (error)
690 goto out;
691 fs = (struct ext2fs *)bp->b_data;
692 error = ext2fs_checksb(fs, ronly);
693 if (error)
694 goto out;
695 ump = malloc(sizeof(*ump), M_UFSMNT, M_WAITOK);
696 memset(ump, 0, sizeof(*ump));
697 ump->um_fstype = UFS1;
698 ump->um_ops = &ext2fs_ufsops;
699 ump->um_e2fs = malloc(sizeof(struct m_ext2fs), M_UFSMNT, M_WAITOK);
700 memset(ump->um_e2fs, 0, sizeof(struct m_ext2fs));
701 e2fs_sbload((struct ext2fs *)bp->b_data, &ump->um_e2fs->e2fs);
702 brelse(bp, 0);
703 bp = NULL;
704 m_fs = ump->um_e2fs;
705 m_fs->e2fs_ronly = ronly;
707 #ifdef DEBUG_EXT2
708 printf("ext2 ino size %zu\n", EXT2_DINODE_SIZE(m_fs));
709 #endif
710 if (ronly == 0) {
711 if (m_fs->e2fs.e2fs_state == E2FS_ISCLEAN)
712 m_fs->e2fs.e2fs_state = 0;
713 else
714 m_fs->e2fs.e2fs_state = E2FS_ERRORS;
715 m_fs->e2fs_fmod = 1;
718 /* compute dynamic sb infos */
719 m_fs->e2fs_ncg =
720 howmany(m_fs->e2fs.e2fs_bcount - m_fs->e2fs.e2fs_first_dblock,
721 m_fs->e2fs.e2fs_bpg);
722 m_fs->e2fs_fsbtodb = m_fs->e2fs.e2fs_log_bsize + LOG_MINBSIZE - DEV_BSHIFT;
723 m_fs->e2fs_bsize = MINBSIZE << m_fs->e2fs.e2fs_log_bsize;
724 m_fs->e2fs_bshift = LOG_MINBSIZE + m_fs->e2fs.e2fs_log_bsize;
725 m_fs->e2fs_qbmask = m_fs->e2fs_bsize - 1;
726 m_fs->e2fs_bmask = ~m_fs->e2fs_qbmask;
727 m_fs->e2fs_ngdb =
728 howmany(m_fs->e2fs_ncg, m_fs->e2fs_bsize / sizeof(struct ext2_gd));
729 m_fs->e2fs_ipb = m_fs->e2fs_bsize / EXT2_DINODE_SIZE(m_fs);
730 m_fs->e2fs_itpg = m_fs->e2fs.e2fs_ipg / m_fs->e2fs_ipb;
732 m_fs->e2fs_gd = malloc(m_fs->e2fs_ngdb * m_fs->e2fs_bsize,
733 M_UFSMNT, M_WAITOK);
734 for (i = 0; i < m_fs->e2fs_ngdb; i++) {
735 error = bread(devvp ,
736 fsbtodb(m_fs, m_fs->e2fs.e2fs_first_dblock +
737 1 /* superblock */ + i),
738 m_fs->e2fs_bsize, NOCRED, 0, &bp);
739 if (error) {
740 free(m_fs->e2fs_gd, M_UFSMNT);
741 goto out;
743 e2fs_cgload((struct ext2_gd *)bp->b_data,
744 &m_fs->e2fs_gd[
745 i * m_fs->e2fs_bsize / sizeof(struct ext2_gd)],
746 m_fs->e2fs_bsize);
747 brelse(bp, 0);
748 bp = NULL;
751 mp->mnt_data = ump;
752 mp->mnt_stat.f_fsidx.__fsid_val[0] = (long)dev;
753 mp->mnt_stat.f_fsidx.__fsid_val[1] = makefstype(MOUNT_EXT2FS);
754 mp->mnt_stat.f_fsid = mp->mnt_stat.f_fsidx.__fsid_val[0];
755 mp->mnt_stat.f_namemax = EXT2FS_MAXNAMLEN;
756 mp->mnt_flag |= MNT_LOCAL;
757 mp->mnt_dev_bshift = DEV_BSHIFT; /* XXX */
758 mp->mnt_fs_bshift = m_fs->e2fs_bshift;
759 mp->mnt_iflag |= IMNT_DTYPE;
760 ump->um_flags = 0;
761 ump->um_mountp = mp;
762 ump->um_dev = dev;
763 ump->um_devvp = devvp;
764 ump->um_nindir = NINDIR(m_fs);
765 ump->um_lognindir = ffs(NINDIR(m_fs)) - 1;
766 ump->um_bptrtodb = m_fs->e2fs_fsbtodb;
767 ump->um_seqinc = 1; /* no frags */
768 ump->um_maxsymlinklen = EXT2_MAXSYMLINKLEN;
769 ump->um_dirblksiz = m_fs->e2fs_bsize;
770 ump->um_maxfilesize = ((uint64_t)0x80000000 * m_fs->e2fs_bsize - 1);
771 devvp->v_specmountpoint = mp;
772 return (0);
774 out:
775 KASSERT(bp != NULL);
776 brelse(bp, 0);
777 if (ump) {
778 free(ump->um_e2fs, M_UFSMNT);
779 free(ump, M_UFSMNT);
780 mp->mnt_data = NULL;
782 return (error);
786 * unmount system call
789 ext2fs_unmount(struct mount *mp, int mntflags)
791 struct ufsmount *ump;
792 struct m_ext2fs *fs;
793 int error, flags;
795 flags = 0;
796 if (mntflags & MNT_FORCE)
797 flags |= FORCECLOSE;
798 if ((error = ext2fs_flushfiles(mp, flags)) != 0)
799 return (error);
800 ump = VFSTOUFS(mp);
801 fs = ump->um_e2fs;
802 if (fs->e2fs_ronly == 0 &&
803 ext2fs_cgupdate(ump, MNT_WAIT) == 0 &&
804 (fs->e2fs.e2fs_state & E2FS_ERRORS) == 0) {
805 fs->e2fs.e2fs_state = E2FS_ISCLEAN;
806 (void) ext2fs_sbupdate(ump, MNT_WAIT);
808 if (ump->um_devvp->v_type != VBAD)
809 ump->um_devvp->v_specmountpoint = NULL;
810 vn_lock(ump->um_devvp, LK_EXCLUSIVE | LK_RETRY);
811 error = VOP_CLOSE(ump->um_devvp, fs->e2fs_ronly ? FREAD : FREAD|FWRITE,
812 NOCRED);
813 vput(ump->um_devvp);
814 free(fs->e2fs_gd, M_UFSMNT);
815 free(fs, M_UFSMNT);
816 free(ump, M_UFSMNT);
817 mp->mnt_data = NULL;
818 mp->mnt_flag &= ~MNT_LOCAL;
819 return (error);
823 * Flush out all the files in a filesystem.
826 ext2fs_flushfiles(struct mount *mp, int flags)
828 extern int doforce;
829 int error;
831 if (!doforce)
832 flags &= ~FORCECLOSE;
833 error = vflush(mp, NULLVP, flags);
834 return (error);
838 * Get file system statistics.
841 ext2fs_statvfs(struct mount *mp, struct statvfs *sbp)
843 struct ufsmount *ump;
844 struct m_ext2fs *fs;
845 uint32_t overhead, overhead_per_group, ngdb;
846 int i, ngroups;
848 ump = VFSTOUFS(mp);
849 fs = ump->um_e2fs;
850 if (fs->e2fs.e2fs_magic != E2FS_MAGIC)
851 panic("ext2fs_statvfs");
854 * Compute the overhead (FS structures)
856 overhead_per_group =
857 1 /* block bitmap */ +
858 1 /* inode bitmap */ +
859 fs->e2fs_itpg;
860 overhead = fs->e2fs.e2fs_first_dblock +
861 fs->e2fs_ncg * overhead_per_group;
862 if (fs->e2fs.e2fs_rev > E2FS_REV0 &&
863 fs->e2fs.e2fs_features_rocompat & EXT2F_ROCOMPAT_SPARSESUPER) {
864 for (i = 0, ngroups = 0; i < fs->e2fs_ncg; i++) {
865 if (cg_has_sb(i))
866 ngroups++;
868 } else {
869 ngroups = fs->e2fs_ncg;
871 ngdb = fs->e2fs_ngdb;
872 if (fs->e2fs.e2fs_rev > E2FS_REV0 &&
873 fs->e2fs.e2fs_features_compat & EXT2F_COMPAT_RESIZE)
874 ngdb += fs->e2fs.e2fs_reserved_ngdb;
875 overhead += ngroups * (1 /* superblock */ + ngdb);
877 sbp->f_bsize = fs->e2fs_bsize;
878 sbp->f_frsize = MINBSIZE << fs->e2fs.e2fs_fsize;
879 sbp->f_iosize = fs->e2fs_bsize;
880 sbp->f_blocks = fs->e2fs.e2fs_bcount - overhead;
881 sbp->f_bfree = fs->e2fs.e2fs_fbcount;
882 sbp->f_bresvd = fs->e2fs.e2fs_rbcount;
883 if (sbp->f_bfree > sbp->f_bresvd)
884 sbp->f_bavail = sbp->f_bfree - sbp->f_bresvd;
885 else
886 sbp->f_bavail = 0;
887 sbp->f_files = fs->e2fs.e2fs_icount;
888 sbp->f_ffree = fs->e2fs.e2fs_ficount;
889 sbp->f_favail = fs->e2fs.e2fs_ficount;
890 sbp->f_fresvd = 0;
891 copy_statvfs_info(sbp, mp);
892 return (0);
896 * Go through the disk queues to initiate sandbagged IO;
897 * go through the inodes to write those that have been modified;
898 * initiate the writing of the super block if it has been modified.
900 * Note: we are always called with the filesystem marked `MPBUSY'.
903 ext2fs_sync(struct mount *mp, int waitfor, kauth_cred_t cred)
905 struct vnode *vp, *mvp;
906 struct inode *ip;
907 struct ufsmount *ump = VFSTOUFS(mp);
908 struct m_ext2fs *fs;
909 int error, allerror = 0;
911 fs = ump->um_e2fs;
912 if (fs->e2fs_fmod != 0 && fs->e2fs_ronly != 0) { /* XXX */
913 printf("fs = %s\n", fs->e2fs_fsmnt);
914 panic("update: rofs mod");
917 /* Allocate a marker vnode. */
918 mvp = vnalloc(mp);
921 * Write back each (modified) inode.
923 mutex_enter(&mntvnode_lock);
924 loop:
926 * NOTE: not using the TAILQ_FOREACH here since in this loop vgone()
927 * and vclean() can be called indirectly
929 for (vp = TAILQ_FIRST(&mp->mnt_vnodelist); vp; vp = vunmark(mvp)) {
930 vmark(mvp, vp);
931 if (vp->v_mount != mp || vismarker(vp))
932 continue;
933 mutex_enter(vp->v_interlock);
934 ip = VTOI(vp);
935 if (ip == NULL || (vp->v_iflag & (VI_XLOCK|VI_CLEAN)) != 0 ||
936 vp->v_type == VNON ||
937 ((ip->i_flag &
938 (IN_CHANGE | IN_UPDATE | IN_MODIFIED)) == 0 &&
939 LIST_EMPTY(&vp->v_dirtyblkhd) &&
940 UVM_OBJ_IS_CLEAN(&vp->v_uobj)))
942 mutex_exit(vp->v_interlock);
943 continue;
945 mutex_exit(&mntvnode_lock);
946 error = vget(vp, LK_EXCLUSIVE | LK_NOWAIT);
947 if (error) {
948 mutex_enter(&mntvnode_lock);
949 if (error == ENOENT) {
950 mutex_enter(&mntvnode_lock);
951 (void)vunmark(mvp);
952 goto loop;
954 continue;
956 if (vp->v_type == VREG && waitfor == MNT_LAZY)
957 error = ext2fs_update(vp, NULL, NULL, 0);
958 else
959 error = VOP_FSYNC(vp, cred,
960 waitfor == MNT_WAIT ? FSYNC_WAIT : 0, 0, 0);
961 if (error)
962 allerror = error;
963 vput(vp);
964 mutex_enter(&mntvnode_lock);
966 mutex_exit(&mntvnode_lock);
967 vnfree(mvp);
969 * Force stale file system control information to be flushed.
971 if (waitfor != MNT_LAZY) {
972 vn_lock(ump->um_devvp, LK_EXCLUSIVE | LK_RETRY);
973 if ((error = VOP_FSYNC(ump->um_devvp, cred,
974 waitfor == MNT_WAIT ? FSYNC_WAIT : 0, 0, 0)) != 0)
975 allerror = error;
976 VOP_UNLOCK(ump->um_devvp);
979 * Write back modified superblock.
981 if (fs->e2fs_fmod != 0) {
982 fs->e2fs_fmod = 0;
983 fs->e2fs.e2fs_wtime = time_second;
984 if ((error = ext2fs_cgupdate(ump, waitfor)))
985 allerror = error;
987 return (allerror);
991 * Look up a EXT2FS dinode number to find its incore vnode, otherwise read it
992 * in from disk. If it is in core, wait for the lock bit to clear, then
993 * return the inode locked. Detection and handling of mount points must be
994 * done by the calling routine.
997 ext2fs_vget(struct mount *mp, ino_t ino, struct vnode **vpp)
999 struct m_ext2fs *fs;
1000 struct inode *ip;
1001 struct ufsmount *ump;
1002 struct buf *bp;
1003 struct vnode *vp;
1004 dev_t dev;
1005 int error;
1006 void *cp;
1008 ump = VFSTOUFS(mp);
1009 dev = ump->um_dev;
1010 retry:
1011 if ((*vpp = ufs_ihashget(dev, ino, LK_EXCLUSIVE)) != NULL)
1012 return (0);
1014 /* Allocate a new vnode/inode. */
1015 error = getnewvnode(VT_EXT2FS, mp, ext2fs_vnodeop_p, NULL, &vp);
1016 if (error) {
1017 *vpp = NULL;
1018 return (error);
1020 ip = pool_get(&ext2fs_inode_pool, PR_WAITOK);
1022 mutex_enter(&ufs_hashlock);
1023 if ((*vpp = ufs_ihashget(dev, ino, 0)) != NULL) {
1024 mutex_exit(&ufs_hashlock);
1025 ungetnewvnode(vp);
1026 pool_put(&ext2fs_inode_pool, ip);
1027 goto retry;
1030 vp->v_vflag |= VV_LOCKSWORK;
1032 memset(ip, 0, sizeof(struct inode));
1033 vp->v_data = ip;
1034 ip->i_vnode = vp;
1035 ip->i_ump = ump;
1036 ip->i_e2fs = fs = ump->um_e2fs;
1037 ip->i_dev = dev;
1038 ip->i_number = ino;
1039 ip->i_e2fs_last_lblk = 0;
1040 ip->i_e2fs_last_blk = 0;
1041 genfs_node_init(vp, &ext2fs_genfsops);
1044 * Put it onto its hash chain and lock it so that other requests for
1045 * this inode will block if they arrive while we are sleeping waiting
1046 * for old data structures to be purged or for the contents of the
1047 * disk portion of this inode to be read.
1050 ufs_ihashins(ip);
1051 mutex_exit(&ufs_hashlock);
1053 /* Read in the disk contents for the inode, copy into the inode. */
1054 error = bread(ump->um_devvp, fsbtodb(fs, ino_to_fsba(fs, ino)),
1055 (int)fs->e2fs_bsize, NOCRED, 0, &bp);
1056 if (error) {
1059 * The inode does not contain anything useful, so it would
1060 * be misleading to leave it on its hash chain. With mode
1061 * still zero, it will be unlinked and returned to the free
1062 * list by vput().
1065 vput(vp);
1066 brelse(bp, 0);
1067 *vpp = NULL;
1068 return (error);
1070 cp = (char *)bp->b_data + (ino_to_fsbo(fs, ino) * EXT2_DINODE_SIZE(fs));
1071 ip->i_din.e2fs_din = pool_get(&ext2fs_dinode_pool, PR_WAITOK);
1072 e2fs_iload((struct ext2fs_dinode *)cp, ip->i_din.e2fs_din);
1073 ext2fs_set_inode_guid(ip);
1074 brelse(bp, 0);
1076 /* If the inode was deleted, reset all fields */
1077 if (ip->i_e2fs_dtime != 0) {
1078 ip->i_e2fs_mode = ip->i_e2fs_nblock = 0;
1079 (void)ext2fs_setsize(ip, 0);
1080 memset(ip->i_e2fs_blocks, 0, sizeof(ip->i_e2fs_blocks));
1084 * Initialize the vnode from the inode, check for aliases.
1087 error = ext2fs_vinit(mp, ext2fs_specop_p, ext2fs_fifoop_p, &vp);
1088 if (error) {
1089 vput(vp);
1090 *vpp = NULL;
1091 return (error);
1094 * Finish inode initialization now that aliasing has been resolved.
1097 ip->i_devvp = ump->um_devvp;
1098 vref(ip->i_devvp);
1101 * Set up a generation number for this inode if it does not
1102 * already have one. This should only happen on old filesystems.
1105 if (ip->i_e2fs_gen == 0) {
1106 if (++ext2gennumber < (u_long)time_second)
1107 ext2gennumber = time_second;
1108 ip->i_e2fs_gen = ext2gennumber;
1109 if ((vp->v_mount->mnt_flag & MNT_RDONLY) == 0)
1110 ip->i_flag |= IN_MODIFIED;
1112 uvm_vnp_setsize(vp, ext2fs_size(ip));
1113 *vpp = vp;
1114 return (0);
1118 * File handle to vnode
1120 * Have to be really careful about stale file handles:
1121 * - check that the inode number is valid
1122 * - call ext2fs_vget() to get the locked inode
1123 * - check for an unallocated inode (i_mode == 0)
1126 ext2fs_fhtovp(struct mount *mp, struct fid *fhp, struct vnode **vpp)
1128 struct inode *ip;
1129 struct vnode *nvp;
1130 int error;
1131 struct ufid ufh;
1132 struct m_ext2fs *fs;
1134 if (fhp->fid_len != sizeof(struct ufid))
1135 return EINVAL;
1137 memcpy(&ufh, fhp, sizeof(struct ufid));
1138 fs = VFSTOUFS(mp)->um_e2fs;
1139 if ((ufh.ufid_ino < EXT2_FIRSTINO && ufh.ufid_ino != EXT2_ROOTINO) ||
1140 ufh.ufid_ino >= fs->e2fs_ncg * fs->e2fs.e2fs_ipg)
1141 return (ESTALE);
1143 if ((error = VFS_VGET(mp, ufh.ufid_ino, &nvp)) != 0) {
1144 *vpp = NULLVP;
1145 return (error);
1147 ip = VTOI(nvp);
1148 if (ip->i_e2fs_mode == 0 || ip->i_e2fs_dtime != 0 ||
1149 ip->i_e2fs_gen != ufh.ufid_gen) {
1150 vput(nvp);
1151 *vpp = NULLVP;
1152 return (ESTALE);
1154 *vpp = nvp;
1155 return (0);
1159 * Vnode pointer to File handle
1161 /* ARGSUSED */
1163 ext2fs_vptofh(struct vnode *vp, struct fid *fhp, size_t *fh_size)
1165 struct inode *ip;
1166 struct ufid ufh;
1168 if (*fh_size < sizeof(struct ufid)) {
1169 *fh_size = sizeof(struct ufid);
1170 return E2BIG;
1172 *fh_size = sizeof(struct ufid);
1174 ip = VTOI(vp);
1175 memset(&ufh, 0, sizeof(ufh));
1176 ufh.ufid_len = sizeof(struct ufid);
1177 ufh.ufid_ino = ip->i_number;
1178 ufh.ufid_gen = ip->i_e2fs_gen;
1179 memcpy(fhp, &ufh, sizeof(ufh));
1180 return (0);
1184 * Write a superblock and associated information back to disk.
1187 ext2fs_sbupdate(struct ufsmount *mp, int waitfor)
1189 struct m_ext2fs *fs = mp->um_e2fs;
1190 struct buf *bp;
1191 int error = 0;
1193 bp = getblk(mp->um_devvp, SBLOCK, SBSIZE, 0, 0);
1194 e2fs_sbsave(&fs->e2fs, (struct ext2fs*)bp->b_data);
1195 if (waitfor == MNT_WAIT)
1196 error = bwrite(bp);
1197 else
1198 bawrite(bp);
1199 return (error);
1203 ext2fs_cgupdate(struct ufsmount *mp, int waitfor)
1205 struct m_ext2fs *fs = mp->um_e2fs;
1206 struct buf *bp;
1207 int i, error = 0, allerror = 0;
1209 allerror = ext2fs_sbupdate(mp, waitfor);
1210 for (i = 0; i < fs->e2fs_ngdb; i++) {
1211 bp = getblk(mp->um_devvp, fsbtodb(fs,
1212 fs->e2fs.e2fs_first_dblock +
1213 1 /* superblock */ + i), fs->e2fs_bsize, 0, 0);
1214 e2fs_cgsave(&fs->e2fs_gd[
1215 i * fs->e2fs_bsize / sizeof(struct ext2_gd)],
1216 (struct ext2_gd *)bp->b_data, fs->e2fs_bsize);
1217 if (waitfor == MNT_WAIT)
1218 error = bwrite(bp);
1219 else
1220 bawrite(bp);
1223 if (!allerror && error)
1224 allerror = error;
1225 return (allerror);
1228 static int
1229 ext2fs_checksb(struct ext2fs *fs, int ronly)
1232 if (fs2h16(fs->e2fs_magic) != E2FS_MAGIC) {
1233 return (EINVAL); /* XXX needs translation */
1235 if (fs2h32(fs->e2fs_rev) > E2FS_REV1) {
1236 #ifdef DIAGNOSTIC
1237 printf("Ext2 fs: unsupported revision number: %x\n",
1238 fs2h32(fs->e2fs_rev));
1239 #endif
1240 return (EINVAL); /* XXX needs translation */
1242 if (fs2h32(fs->e2fs_log_bsize) > 2) { /* block size = 1024|2048|4096 */
1243 #ifdef DIAGNOSTIC
1244 printf("Ext2 fs: bad block size: %d "
1245 "(expected <= 2 for ext2 fs)\n",
1246 fs2h32(fs->e2fs_log_bsize));
1247 #endif
1248 return (EINVAL); /* XXX needs translation */
1250 if (fs2h32(fs->e2fs_rev) > E2FS_REV0) {
1251 if (fs2h32(fs->e2fs_first_ino) != EXT2_FIRSTINO) {
1252 printf("Ext2 fs: unsupported first inode position\n");
1253 return (EINVAL); /* XXX needs translation */
1255 if (fs2h32(fs->e2fs_features_incompat) &
1256 ~EXT2F_INCOMPAT_SUPP) {
1257 printf("Ext2 fs: unsupported optional feature\n");
1258 return (EINVAL); /* XXX needs translation */
1260 if (!ronly && fs2h32(fs->e2fs_features_rocompat) &
1261 ~EXT2F_ROCOMPAT_SUPP) {
1262 return (EROFS); /* XXX needs translation */
1265 return (0);