xref: /netbsd-src/sys/ufs/ext2fs/ext2fs_vfsops.c (revision 3b01aba77a7a698587faaae455bbfe740923c1f5)
1 /*	$NetBSD: ext2fs_vfsops.c,v 1.43 2001/05/30 11:57:19 mrg Exp $	*/
2 
3 /*
4  * Copyright (c) 1997 Manuel Bouyer.
5  * Copyright (c) 1989, 1991, 1993, 1994
6  *	The Regents of the University of California.  All rights reserved.
7  *
8  * Redistribution and use in source and binary forms, with or without
9  * modification, are permitted provided that the following conditions
10  * are met:
11  * 1. Redistributions of source code must retain the above copyright
12  *    notice, this list of conditions and the following disclaimer.
13  * 2. Redistributions in binary form must reproduce the above copyright
14  *    notice, this list of conditions and the following disclaimer in the
15  *    documentation and/or other materials provided with the distribution.
16  * 3. All advertising materials mentioning features or use of this software
17  *    must display the following acknowledgement:
18  *	This product includes software developed by the University of
19  *	California, Berkeley and its contributors.
20  * 4. Neither the name of the University nor the names of its contributors
21  *    may be used to endorse or promote products derived from this software
22  *    without specific prior written permission.
23  *
24  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
25  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
26  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
27  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
28  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
29  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
30  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
31  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
32  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
33  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
34  * SUCH DAMAGE.
35  *
36  *	@(#)ffs_vfsops.c	8.14 (Berkeley) 11/28/94
37  * Modified for ext2fs by Manuel Bouyer.
38  */
39 
40 #if defined(_KERNEL_OPT)
41 #include "opt_compat_netbsd.h"
42 #endif
43 
44 #include <sys/param.h>
45 #include <sys/systm.h>
46 #include <sys/namei.h>
47 #include <sys/proc.h>
48 #include <sys/kernel.h>
49 #include <sys/vnode.h>
50 #include <sys/socket.h>
51 #include <sys/mount.h>
52 #include <sys/buf.h>
53 #include <sys/device.h>
54 #include <sys/mbuf.h>
55 #include <sys/file.h>
56 #include <sys/disklabel.h>
57 #include <sys/ioctl.h>
58 #include <sys/errno.h>
59 #include <sys/malloc.h>
60 #include <sys/pool.h>
61 #include <sys/lock.h>
62 
63 #include <miscfs/specfs/specdev.h>
64 
65 #include <ufs/ufs/quota.h>
66 #include <ufs/ufs/ufsmount.h>
67 #include <ufs/ufs/inode.h>
68 #include <ufs/ufs/dir.h>
69 #include <ufs/ufs/ufs_extern.h>
70 
71 #include <ufs/ext2fs/ext2fs.h>
72 #include <ufs/ext2fs/ext2fs_extern.h>
73 
74 extern struct lock ufs_hashlock;
75 
76 int ext2fs_sbupdate __P((struct ufsmount *, int));
77 static int ext2fs_checksb __P((struct ext2fs *, int));
78 
79 extern const struct vnodeopv_desc ext2fs_vnodeop_opv_desc;
80 extern const struct vnodeopv_desc ext2fs_specop_opv_desc;
81 extern const struct vnodeopv_desc ext2fs_fifoop_opv_desc;
82 
83 const struct vnodeopv_desc * const ext2fs_vnodeopv_descs[] = {
84 	&ext2fs_vnodeop_opv_desc,
85 	&ext2fs_specop_opv_desc,
86 	&ext2fs_fifoop_opv_desc,
87 	NULL,
88 };
89 
90 struct vfsops ext2fs_vfsops = {
91 	MOUNT_EXT2FS,
92 	ext2fs_mount,
93 	ufs_start,
94 	ext2fs_unmount,
95 	ufs_root,
96 	ufs_quotactl,
97 	ext2fs_statfs,
98 	ext2fs_sync,
99 	ext2fs_vget,
100 	ext2fs_fhtovp,
101 	ext2fs_vptofh,
102 	ext2fs_init,
103 	ext2fs_done,
104 	ext2fs_sysctl,
105 	ext2fs_mountroot,
106 	ufs_check_export,
107 	ext2fs_vnodeopv_descs,
108 };
109 
110 struct pool ext2fs_inode_pool;
111 
112 extern u_long ext2gennumber;
113 
114 void
115 ext2fs_init()
116 {
117 	ufs_init();
118 
119 	/*
120 	 * XXX Same structure as FFS inodes?  Should we share a common pool?
121 	 */
122 	pool_init(&ext2fs_inode_pool, sizeof(struct inode), 0, 0, 0,
123 	    "ext2fsinopl", 0, pool_page_alloc_nointr, pool_page_free_nointr,
124 	    M_EXT2FSNODE);
125 }
126 
127 void
128 ext2fs_done()
129 {
130 	ufs_done();
131 	pool_destroy(&ext2fs_inode_pool);
132 }
133 
134 /*
135  * Called by main() when ext2fs is going to be mounted as root.
136  *
137  * Name is updated by mount(8) after booting.
138  */
139 #define ROOTNAME	"root_device"
140 
141 int
142 ext2fs_mountroot()
143 {
144 	extern struct vnode *rootvp;
145 	struct m_ext2fs *fs;
146 	struct mount *mp;
147 	struct proc *p = curproc;	/* XXX */
148 	struct ufsmount *ump;
149 	int error;
150 
151 	if (root_device->dv_class != DV_DISK)
152 		return (ENODEV);
153 
154 	/*
155 	 * Get vnodes for rootdev.
156 	 */
157 	if (bdevvp(rootdev, &rootvp))
158 		panic("ext2fs_mountroot: can't setup bdevvp's");
159 
160 	if ((error = vfs_rootmountalloc(MOUNT_EXT2FS, "root_device", &mp))) {
161 		vrele(rootvp);
162 		return (error);
163 	}
164 
165 	if ((error = ext2fs_mountfs(rootvp, mp, p)) != 0) {
166 		mp->mnt_op->vfs_refcount--;
167 		vfs_unbusy(mp);
168 		free(mp, M_MOUNT);
169 		vrele(rootvp);
170 		return (error);
171 	}
172 	simple_lock(&mountlist_slock);
173 	CIRCLEQ_INSERT_TAIL(&mountlist, mp, mnt_list);
174 	simple_unlock(&mountlist_slock);
175 	ump = VFSTOUFS(mp);
176 	fs = ump->um_e2fs;
177 	memset(fs->e2fs_fsmnt, 0, sizeof(fs->e2fs_fsmnt));
178 	(void) copystr(mp->mnt_stat.f_mntonname, fs->e2fs_fsmnt,
179 	    sizeof(fs->e2fs_fsmnt) - 1, 0);
180 	if (fs->e2fs.e2fs_rev > E2FS_REV0) {
181 		memset(fs->e2fs.e2fs_fsmnt, 0, sizeof(fs->e2fs.e2fs_fsmnt));
182 		(void) copystr(mp->mnt_stat.f_mntonname, fs->e2fs.e2fs_fsmnt,
183 		    sizeof(fs->e2fs.e2fs_fsmnt) - 1, 0);
184 	}
185 	(void)ext2fs_statfs(mp, &mp->mnt_stat, p);
186 	vfs_unbusy(mp);
187 	inittodr(fs->e2fs.e2fs_wtime);
188 	return (0);
189 }
190 
191 /*
192  * VFS Operations.
193  *
194  * mount system call
195  */
196 int
197 ext2fs_mount(mp, path, data, ndp, p)
198 	struct mount *mp;
199 	const char *path;
200 	void * data;
201 	struct nameidata *ndp;
202 	struct proc *p;
203 {
204 	struct vnode *devvp;
205 	struct ufs_args args;
206 	struct ufsmount *ump = NULL;
207 	struct m_ext2fs *fs;
208 	size_t size;
209 	int error, flags;
210 	mode_t accessmode;
211 
212 	error = copyin(data, (caddr_t)&args, sizeof (struct ufs_args));
213 	if (error)
214 		return (error);
215 	/*
216 	 * If updating, check whether changing from read-only to
217 	 * read/write; if there is no device name, that's all we do.
218 	 */
219 	if (mp->mnt_flag & MNT_UPDATE) {
220 		ump = VFSTOUFS(mp);
221 		fs = ump->um_e2fs;
222 		if (fs->e2fs_ronly == 0 && (mp->mnt_flag & MNT_RDONLY)) {
223 			flags = WRITECLOSE;
224 			if (mp->mnt_flag & MNT_FORCE)
225 				flags |= FORCECLOSE;
226 			error = ext2fs_flushfiles(mp, flags, p);
227 			if (error == 0 &&
228 				ext2fs_cgupdate(ump, MNT_WAIT) == 0 &&
229 				(fs->e2fs.e2fs_state & E2FS_ERRORS) == 0) {
230 				fs->e2fs.e2fs_state = E2FS_ISCLEAN;
231 				(void) ext2fs_sbupdate(ump, MNT_WAIT);
232 			}
233 			if (error)
234 				return (error);
235 			fs->e2fs_ronly = 1;
236 		}
237 		if (mp->mnt_flag & MNT_RELOAD) {
238 			error = ext2fs_reload(mp, ndp->ni_cnd.cn_cred, p);
239 			if (error)
240 				return (error);
241 		}
242 		if (fs->e2fs_ronly && (mp->mnt_flag & MNT_WANTRDWR)) {
243 			/*
244 			 * If upgrade to read-write by non-root, then verify
245 			 * that user has necessary permissions on the device.
246 			 */
247 			if (p->p_ucred->cr_uid != 0) {
248 				devvp = ump->um_devvp;
249 				vn_lock(devvp, LK_EXCLUSIVE | LK_RETRY);
250 				error = VOP_ACCESS(devvp, VREAD | VWRITE,
251 						   p->p_ucred, p);
252 				VOP_UNLOCK(devvp, 0);
253 				if (error)
254 					return (error);
255 			}
256 			fs->e2fs_ronly = 0;
257 			if (fs->e2fs.e2fs_state == E2FS_ISCLEAN)
258 				fs->e2fs.e2fs_state = 0;
259 			else
260 				fs->e2fs.e2fs_state = E2FS_ERRORS;
261 			fs->e2fs_fmod = 1;
262 		}
263 		if (args.fspec == 0) {
264 			/*
265 			 * Process export requests.
266 			 */
267 			return (vfs_export(mp, &ump->um_export, &args.export));
268 		}
269 	}
270 	/*
271 	 * Not an update, or updating the name: look up the name
272 	 * and verify that it refers to a sensible block device.
273 	 */
274 	NDINIT(ndp, LOOKUP, FOLLOW, UIO_USERSPACE, args.fspec, p);
275 	if ((error = namei(ndp)) != 0)
276 		return (error);
277 	devvp = ndp->ni_vp;
278 
279 	if (devvp->v_type != VBLK) {
280 		vrele(devvp);
281 		return (ENOTBLK);
282 	}
283 	if (major(devvp->v_rdev) >= nblkdev) {
284 		vrele(devvp);
285 		return (ENXIO);
286 	}
287 	/*
288 	 * If mount by non-root, then verify that user has necessary
289 	 * permissions on the device.
290 	 */
291 	if (p->p_ucred->cr_uid != 0) {
292 		accessmode = VREAD;
293 		if ((mp->mnt_flag & MNT_RDONLY) == 0)
294 			accessmode |= VWRITE;
295 		vn_lock(devvp, LK_EXCLUSIVE | LK_RETRY);
296 		error = VOP_ACCESS(devvp, accessmode, p->p_ucred, p);
297 		VOP_UNLOCK(devvp, 0);
298 		if (error) {
299 			vrele(devvp);
300 			return (error);
301 		}
302 	}
303 	if ((mp->mnt_flag & MNT_UPDATE) == 0)
304 		error = ext2fs_mountfs(devvp, mp, p);
305 	else {
306 		if (devvp != ump->um_devvp)
307 			error = EINVAL;	/* needs translation */
308 		else
309 			vrele(devvp);
310 	}
311 	if (error) {
312 		vrele(devvp);
313 		return (error);
314 	}
315 	ump = VFSTOUFS(mp);
316 	fs = ump->um_e2fs;
317 	(void) copyinstr(path, fs->e2fs_fsmnt, sizeof(fs->e2fs_fsmnt) - 1,
318 	    &size);
319 	memset(fs->e2fs_fsmnt + size, 0, sizeof(fs->e2fs_fsmnt) - size);
320 	if (fs->e2fs.e2fs_rev > E2FS_REV0) {
321 		(void) copystr(mp->mnt_stat.f_mntonname, fs->e2fs.e2fs_fsmnt,
322 		    sizeof(fs->e2fs.e2fs_fsmnt) - 1, &size);
323 		memset(fs->e2fs.e2fs_fsmnt, 0,
324 		    sizeof(fs->e2fs.e2fs_fsmnt) - size);
325 	}
326 	memcpy(mp->mnt_stat.f_mntonname, fs->e2fs_fsmnt, MNAMELEN);
327 	(void) copyinstr(args.fspec, mp->mnt_stat.f_mntfromname, MNAMELEN - 1,
328 		&size);
329 	memset(mp->mnt_stat.f_mntfromname + size, 0, MNAMELEN - size);
330 	if (fs->e2fs_fmod != 0) {	/* XXX */
331 		fs->e2fs_fmod = 0;
332 		if (fs->e2fs.e2fs_state == 0)
333 			fs->e2fs.e2fs_wtime = time.tv_sec;
334 		else
335 			printf("%s: file system not clean; please fsck(8)\n",
336 				mp->mnt_stat.f_mntfromname);
337 		(void) ext2fs_cgupdate(ump, MNT_WAIT);
338 	}
339 	return (0);
340 }
341 
342 /*
343  * Reload all incore data for a filesystem (used after running fsck on
344  * the root filesystem and finding things to fix). The filesystem must
345  * be mounted read-only.
346  *
347  * Things to do to update the mount:
348  *	1) invalidate all cached meta-data.
349  *	2) re-read superblock from disk.
350  *	3) re-read summary information from disk.
351  *	4) invalidate all inactive vnodes.
352  *	5) invalidate all cached file data.
353  *	6) re-read inode data for all active vnodes.
354  */
355 int
356 ext2fs_reload(mountp, cred, p)
357 	struct mount *mountp;
358 	struct ucred *cred;
359 	struct proc *p;
360 {
361 	struct vnode *vp, *nvp, *devvp;
362 	struct inode *ip;
363 	struct buf *bp;
364 	struct m_ext2fs *fs;
365 	struct ext2fs *newfs;
366 	struct partinfo dpart;
367 	int i, size, error;
368 	caddr_t cp;
369 
370 	if ((mountp->mnt_flag & MNT_RDONLY) == 0)
371 		return (EINVAL);
372 	/*
373 	 * Step 1: invalidate all cached meta-data.
374 	 */
375 	devvp = VFSTOUFS(mountp)->um_devvp;
376 	vn_lock(devvp, LK_EXCLUSIVE | LK_RETRY);
377 	error = vinvalbuf(devvp, 0, cred, p, 0, 0);
378 	VOP_UNLOCK(devvp, 0);
379 	if (error)
380 		panic("ext2fs_reload: dirty1");
381 	/*
382 	 * Step 2: re-read superblock from disk.
383 	 */
384 	if (VOP_IOCTL(devvp, DIOCGPART, (caddr_t)&dpart, FREAD, NOCRED, p) != 0)
385 		size = DEV_BSIZE;
386 	else
387 		size = dpart.disklab->d_secsize;
388 	error = bread(devvp, (ufs_daddr_t)(SBOFF / size), SBSIZE, NOCRED, &bp);
389 	if (error) {
390 		brelse(bp);
391 		return (error);
392 	}
393 	newfs = (struct ext2fs *)bp->b_data;
394 	error = ext2fs_checksb(newfs, (mountp->mnt_flag & MNT_RDONLY) != 0);
395 	if (error) {
396 		brelse(bp);
397 		return (error);
398 	}
399 
400 	fs = VFSTOUFS(mountp)->um_e2fs;
401 	/*
402 	 * copy in new superblock, and compute in-memory values
403 	 */
404 	e2fs_sbload(newfs, &fs->e2fs);
405 	fs->e2fs_ncg =
406 	    howmany(fs->e2fs.e2fs_bcount - fs->e2fs.e2fs_first_dblock,
407 	    fs->e2fs.e2fs_bpg);
408 	/* XXX assume hw bsize = 512 */
409 	fs->e2fs_fsbtodb = fs->e2fs.e2fs_log_bsize + 1;
410 	fs->e2fs_bsize = 1024 << fs->e2fs.e2fs_log_bsize;
411 	fs->e2fs_bshift = LOG_MINBSIZE + fs->e2fs.e2fs_log_bsize;
412 	fs->e2fs_qbmask = fs->e2fs_bsize - 1;
413 	fs->e2fs_bmask = ~fs->e2fs_qbmask;
414 	fs->e2fs_ngdb = howmany(fs->e2fs_ncg,
415 			fs->e2fs_bsize / sizeof(struct ext2_gd));
416 	fs->e2fs_ipb = fs->e2fs_bsize / EXT2_DINODE_SIZE;
417 	fs->e2fs_itpg = fs->e2fs.e2fs_ipg/fs->e2fs_ipb;
418 
419 	/*
420 	 * Step 3: re-read summary information from disk.
421 	 */
422 
423 	for (i=0; i < fs->e2fs_ngdb; i++) {
424 		error = bread(devvp ,
425 		    fsbtodb(fs, ((fs->e2fs_bsize>1024)? 0 : 1) + i + 1),
426 		    fs->e2fs_bsize, NOCRED, &bp);
427 		if (error) {
428 			brelse(bp);
429 			return (error);
430 		}
431 		e2fs_cgload((struct ext2_gd*)bp->b_data,
432 		    &fs->e2fs_gd[i* fs->e2fs_bsize / sizeof(struct ext2_gd)],
433 		    fs->e2fs_bsize);
434 		brelse(bp);
435 	}
436 
437 loop:
438 	simple_lock(&mntvnode_slock);
439 	for (vp = mountp->mnt_vnodelist.lh_first; vp != NULL; vp = nvp) {
440 		if (vp->v_mount != mountp) {
441 			simple_unlock(&mntvnode_slock);
442 			goto loop;
443 		}
444 		nvp = vp->v_mntvnodes.le_next;
445 		/*
446 		 * Step 4: invalidate all inactive vnodes.
447 		 */
448 		if (vrecycle(vp, &mntvnode_slock, p))
449 			goto loop;
450 		/*
451 		 * Step 5: invalidate all cached file data.
452 		 */
453 		simple_lock(&vp->v_interlock);
454 		simple_unlock(&mntvnode_slock);
455 		if (vget(vp, LK_EXCLUSIVE | LK_INTERLOCK))
456 			goto loop;
457 		if (vinvalbuf(vp, 0, cred, p, 0, 0))
458 			panic("ext2fs_reload: dirty2");
459 		/*
460 		 * Step 6: re-read inode data for all active vnodes.
461 		 */
462 		ip = VTOI(vp);
463 		error = bread(devvp, fsbtodb(fs, ino_to_fsba(fs, ip->i_number)),
464 				  (int)fs->e2fs_bsize, NOCRED, &bp);
465 		if (error) {
466 			vput(vp);
467 			return (error);
468 		}
469 		cp = (caddr_t)bp->b_data +
470 		    (ino_to_fsbo(fs, ip->i_number) * EXT2_DINODE_SIZE);
471 		e2fs_iload((struct ext2fs_dinode *)cp, &ip->i_din.e2fs_din);
472 		brelse(bp);
473 		vput(vp);
474 		simple_lock(&mntvnode_slock);
475 	}
476 	simple_unlock(&mntvnode_slock);
477 	return (0);
478 }
479 
480 /*
481  * Common code for mount and mountroot
482  */
483 int
484 ext2fs_mountfs(devvp, mp, p)
485 	struct vnode *devvp;
486 	struct mount *mp;
487 	struct proc *p;
488 {
489 	struct ufsmount *ump;
490 	struct buf *bp;
491 	struct ext2fs *fs;
492 	struct m_ext2fs *m_fs;
493 	dev_t dev;
494 	struct partinfo dpart;
495 	int error, i, size, ronly;
496 	struct ucred *cred;
497 	extern struct vnode *rootvp;
498 
499 	dev = devvp->v_rdev;
500 	cred = p ? p->p_ucred : NOCRED;
501 	/*
502 	 * Disallow multiple mounts of the same device.
503 	 * Disallow mounting of a device that is currently in use
504 	 * (except for root, which might share swap device for miniroot).
505 	 * Flush out any old buffers remaining from a previous use.
506 	 */
507 	if ((error = vfs_mountedon(devvp)) != 0)
508 		return (error);
509 	if (vcount(devvp) > 1 && devvp != rootvp)
510 		return (EBUSY);
511 	vn_lock(devvp, LK_EXCLUSIVE | LK_RETRY);
512 	error = vinvalbuf(devvp, V_SAVE, cred, p, 0, 0);
513 	VOP_UNLOCK(devvp, 0);
514 	if (error)
515 		return (error);
516 
517 	ronly = (mp->mnt_flag & MNT_RDONLY) != 0;
518 	error = VOP_OPEN(devvp, ronly ? FREAD : FREAD|FWRITE, FSCRED, p);
519 	if (error)
520 		return (error);
521 	if (VOP_IOCTL(devvp, DIOCGPART, (caddr_t)&dpart, FREAD, cred, p) != 0)
522 		size = DEV_BSIZE;
523 	else
524 		size = dpart.disklab->d_secsize;
525 
526 	bp = NULL;
527 	ump = NULL;
528 
529 #ifdef DEBUG_EXT2
530 	printf("sb size: %d ino size %d\n", sizeof(struct ext2fs),
531 	    EXT2_DINODE_SIZE);
532 #endif
533 	error = bread(devvp, (SBOFF / DEV_BSIZE), SBSIZE, cred, &bp);
534 	if (error)
535 		goto out;
536 	fs = (struct ext2fs *)bp->b_data;
537 	error = ext2fs_checksb(fs, ronly);
538 	if (error)
539 		goto out;
540 	ump = malloc(sizeof *ump, M_UFSMNT, M_WAITOK);
541 	memset((caddr_t)ump, 0, sizeof *ump);
542 	ump->um_e2fs = malloc(sizeof(struct m_ext2fs), M_UFSMNT, M_WAITOK);
543 	memset((caddr_t)ump->um_e2fs, 0, sizeof(struct m_ext2fs));
544 	e2fs_sbload((struct ext2fs*)bp->b_data, &ump->um_e2fs->e2fs);
545 	brelse(bp);
546 	bp = NULL;
547 	m_fs = ump->um_e2fs;
548 	m_fs->e2fs_ronly = ronly;
549 	if (ronly == 0) {
550 		if (m_fs->e2fs.e2fs_state == E2FS_ISCLEAN)
551 			m_fs->e2fs.e2fs_state = 0;
552 		else
553 			m_fs->e2fs.e2fs_state = E2FS_ERRORS;
554 		m_fs->e2fs_fmod = 1;
555 	}
556 
557 	/* compute dynamic sb infos */
558 	m_fs->e2fs_ncg =
559 		howmany(m_fs->e2fs.e2fs_bcount - m_fs->e2fs.e2fs_first_dblock,
560 		m_fs->e2fs.e2fs_bpg);
561 	/* XXX assume hw bsize = 512 */
562 	m_fs->e2fs_fsbtodb = m_fs->e2fs.e2fs_log_bsize + 1;
563 	m_fs->e2fs_bsize = 1024 << m_fs->e2fs.e2fs_log_bsize;
564 	m_fs->e2fs_bshift = LOG_MINBSIZE + m_fs->e2fs.e2fs_log_bsize;
565 	m_fs->e2fs_qbmask = m_fs->e2fs_bsize - 1;
566 	m_fs->e2fs_bmask = ~m_fs->e2fs_qbmask;
567 	m_fs->e2fs_ngdb = howmany(m_fs->e2fs_ncg,
568 		m_fs->e2fs_bsize / sizeof(struct ext2_gd));
569 	m_fs->e2fs_ipb = m_fs->e2fs_bsize / EXT2_DINODE_SIZE;
570 	m_fs->e2fs_itpg = m_fs->e2fs.e2fs_ipg/m_fs->e2fs_ipb;
571 
572 	m_fs->e2fs_gd = malloc(m_fs->e2fs_ngdb * m_fs->e2fs_bsize,
573 		M_UFSMNT, M_WAITOK);
574 	for (i=0; i < m_fs->e2fs_ngdb; i++) {
575 		error = bread(devvp ,
576 		    fsbtodb(m_fs, ((m_fs->e2fs_bsize>1024)? 0 : 1) + i + 1),
577 		    m_fs->e2fs_bsize, NOCRED, &bp);
578 		if (error) {
579 			free(m_fs->e2fs_gd, M_UFSMNT);
580 			goto out;
581 		}
582 		e2fs_cgload((struct ext2_gd*)bp->b_data,
583 		    &m_fs->e2fs_gd[
584 			i * m_fs->e2fs_bsize / sizeof(struct ext2_gd)],
585 		    m_fs->e2fs_bsize);
586 		brelse(bp);
587 		bp = NULL;
588 	}
589 
590 	mp->mnt_data = (qaddr_t)ump;
591 	mp->mnt_stat.f_fsid.val[0] = (long)dev;
592 	mp->mnt_stat.f_fsid.val[1] = makefstype(MOUNT_EXT2FS);
593 	mp->mnt_maxsymlinklen = EXT2_MAXSYMLINKLEN;
594 	mp->mnt_flag |= MNT_LOCAL;
595 	mp->mnt_dev_bshift = DEV_BSHIFT;	/* XXX */
596 	mp->mnt_fs_bshift = m_fs->e2fs_bshift;
597 	ump->um_flags = 0;
598 	ump->um_mountp = mp;
599 	ump->um_dev = dev;
600 	ump->um_devvp = devvp;
601 	ump->um_nindir = NINDIR(m_fs);
602 	ump->um_lognindir = ffs(NINDIR(m_fs)) - 1;
603 	ump->um_bptrtodb = m_fs->e2fs_fsbtodb;
604 	ump->um_seqinc = 1; /* no frags */
605 	devvp->v_specmountpoint = mp;
606 	return (0);
607 
608 out:
609 	if (bp)
610 		brelse(bp);
611 	vn_lock(devvp, LK_EXCLUSIVE | LK_RETRY);
612 	(void)VOP_CLOSE(devvp, ronly ? FREAD : FREAD|FWRITE, cred, p);
613 	VOP_UNLOCK(devvp, 0);
614 	if (ump) {
615 		free(ump->um_e2fs, M_UFSMNT);
616 		free(ump, M_UFSMNT);
617 		mp->mnt_data = (qaddr_t)0;
618 	}
619 	return (error);
620 }
621 
622 /*
623  * unmount system call
624  */
625 int
626 ext2fs_unmount(mp, mntflags, p)
627 	struct mount *mp;
628 	int mntflags;
629 	struct proc *p;
630 {
631 	struct ufsmount *ump;
632 	struct m_ext2fs *fs;
633 	int error, flags;
634 
635 	flags = 0;
636 	if (mntflags & MNT_FORCE)
637 		flags |= FORCECLOSE;
638 	if ((error = ext2fs_flushfiles(mp, flags, p)) != 0)
639 		return (error);
640 	ump = VFSTOUFS(mp);
641 	fs = ump->um_e2fs;
642 	if (fs->e2fs_ronly == 0 &&
643 		ext2fs_cgupdate(ump, MNT_WAIT) == 0 &&
644 		(fs->e2fs.e2fs_state & E2FS_ERRORS) == 0) {
645 		fs->e2fs.e2fs_state = E2FS_ISCLEAN;
646 		(void) ext2fs_sbupdate(ump, MNT_WAIT);
647 	}
648 	if (ump->um_devvp->v_type != VBAD)
649 		ump->um_devvp->v_specmountpoint = NULL;
650 	vn_lock(ump->um_devvp, LK_EXCLUSIVE | LK_RETRY);
651 	error = VOP_CLOSE(ump->um_devvp, fs->e2fs_ronly ? FREAD : FREAD|FWRITE,
652 		NOCRED, p);
653 	vput(ump->um_devvp);
654 	free(fs->e2fs_gd, M_UFSMNT);
655 	free(fs, M_UFSMNT);
656 	free(ump, M_UFSMNT);
657 	mp->mnt_data = (qaddr_t)0;
658 	mp->mnt_flag &= ~MNT_LOCAL;
659 	return (error);
660 }
661 
662 /*
663  * Flush out all the files in a filesystem.
664  */
665 int
666 ext2fs_flushfiles(mp, flags, p)
667 	struct mount *mp;
668 	int flags;
669 	struct proc *p;
670 {
671 	extern int doforce;
672 	struct ufsmount *ump;
673 	int error;
674 
675 	if (!doforce)
676 		flags &= ~FORCECLOSE;
677 	ump = VFSTOUFS(mp);
678 	error = vflush(mp, NULLVP, flags);
679 	return (error);
680 }
681 
682 /*
683  * Get file system statistics.
684  */
685 int
686 ext2fs_statfs(mp, sbp, p)
687 	struct mount *mp;
688 	struct statfs *sbp;
689 	struct proc *p;
690 {
691 	struct ufsmount *ump;
692 	struct m_ext2fs *fs;
693 	u_int32_t overhead, overhead_per_group;
694 	int i, ngroups;
695 
696 	ump = VFSTOUFS(mp);
697 	fs = ump->um_e2fs;
698 	if (fs->e2fs.e2fs_magic != E2FS_MAGIC)
699 		panic("ext2fs_statfs");
700 
701 #ifdef COMPAT_09
702 	sbp->f_type = 1;
703 #else
704 	sbp->f_type = 0;
705 #endif
706 
707 	/*
708 	 * Compute the overhead (FS structures)
709 	 */
710 	overhead_per_group = 1 /* block bitmap */ +
711 				 1 /* inode bitmap */ +
712 				 fs->e2fs_itpg;
713 	overhead = fs->e2fs.e2fs_first_dblock +
714 		   fs->e2fs_ncg * overhead_per_group;
715 	if (fs->e2fs.e2fs_rev > E2FS_REV0 &&
716 	    fs->e2fs.e2fs_features_rocompat & EXT2F_ROCOMPAT_SPARSESUPER) {
717 		for (i = 0, ngroups = 0; i < fs->e2fs_ncg; i++) {
718 			if (cg_has_sb(i))
719 				ngroups++;
720 		}
721 	} else {
722 		ngroups = fs->e2fs_ncg;
723 	}
724 	overhead += ngroups * (1 + fs->e2fs_ngdb);
725 
726 	sbp->f_bsize = fs->e2fs_bsize;
727 	sbp->f_iosize = fs->e2fs_bsize;
728 	sbp->f_blocks = fs->e2fs.e2fs_bcount - overhead;
729 	sbp->f_bfree = fs->e2fs.e2fs_fbcount;
730 	sbp->f_bavail = sbp->f_bfree - fs->e2fs.e2fs_rbcount;
731 	sbp->f_files =  fs->e2fs.e2fs_icount;
732 	sbp->f_ffree = fs->e2fs.e2fs_ficount;
733 	if (sbp != &mp->mnt_stat) {
734 		memcpy(sbp->f_mntonname, mp->mnt_stat.f_mntonname, MNAMELEN);
735 		memcpy(sbp->f_mntfromname, mp->mnt_stat.f_mntfromname, MNAMELEN);
736 	}
737 	strncpy(sbp->f_fstypename, mp->mnt_op->vfs_name, MFSNAMELEN);
738 	return (0);
739 }
740 
741 /*
742  * Go through the disk queues to initiate sandbagged IO;
743  * go through the inodes to write those that have been modified;
744  * initiate the writing of the super block if it has been modified.
745  *
746  * Note: we are always called with the filesystem marked `MPBUSY'.
747  */
748 int
749 ext2fs_sync(mp, waitfor, cred, p)
750 	struct mount *mp;
751 	int waitfor;
752 	struct ucred *cred;
753 	struct proc *p;
754 {
755 	struct vnode *vp, *nvp;
756 	struct inode *ip;
757 	struct ufsmount *ump = VFSTOUFS(mp);
758 	struct m_ext2fs *fs;
759 	int error, allerror = 0;
760 
761 	fs = ump->um_e2fs;
762 	if (fs->e2fs_fmod != 0 && fs->e2fs_ronly != 0) {	/* XXX */
763 		printf("fs = %s\n", fs->e2fs_fsmnt);
764 		panic("update: rofs mod");
765 	}
766 	/*
767 	 * Write back each (modified) inode.
768 	 */
769 	simple_lock(&mntvnode_slock);
770 loop:
771 	for (vp = LIST_FIRST(&mp->mnt_vnodelist); vp != NULL; vp = nvp) {
772 		/*
773 		 * If the vnode that we are about to sync is no longer
774 		 * associated with this mount point, start over.
775 		 */
776 		if (vp->v_mount != mp)
777 			goto loop;
778 		simple_lock(&vp->v_interlock);
779 		nvp = LIST_NEXT(vp, v_mntvnodes);
780 		ip = VTOI(vp);
781 		if (waitfor == MNT_LAZY || vp->v_type == VNON ||
782 		    ((ip->i_flag &
783 		      (IN_ACCESS | IN_CHANGE | IN_UPDATE | IN_MODIFIED | IN_ACCESSED)) == 0 &&
784 		     LIST_EMPTY(&vp->v_dirtyblkhd) &&
785 		     vp->v_uvm.u_obj.uo_npages == 0))
786 		{
787 			simple_unlock(&vp->v_interlock);
788 			continue;
789 		}
790 		simple_unlock(&mntvnode_slock);
791 		error = vget(vp, LK_EXCLUSIVE | LK_NOWAIT | LK_INTERLOCK);
792 		if (error) {
793 			simple_lock(&mntvnode_slock);
794 			if (error == ENOENT)
795 				goto loop;
796 			continue;
797 		}
798 		if ((error = VOP_FSYNC(vp, cred,
799 		    waitfor == MNT_WAIT ? FSYNC_WAIT : 0, 0, 0, p)) != 0)
800 			allerror = error;
801 		vput(vp);
802 		simple_lock(&mntvnode_slock);
803 	}
804 	simple_unlock(&mntvnode_slock);
805 	/*
806 	 * Force stale file system control information to be flushed.
807 	 */
808 	if (waitfor != MNT_LAZY) {
809 		vn_lock(ump->um_devvp, LK_EXCLUSIVE | LK_RETRY);
810 		if ((error = VOP_FSYNC(ump->um_devvp, cred,
811 		    waitfor == MNT_WAIT ? FSYNC_WAIT : 0, 0, 0, p)) != 0)
812 			allerror = error;
813 		VOP_UNLOCK(ump->um_devvp, 0);
814 	}
815 	/*
816 	 * Write back modified superblock.
817 	 */
818 	if (fs->e2fs_fmod != 0) {
819 		fs->e2fs_fmod = 0;
820 		fs->e2fs.e2fs_wtime = time.tv_sec;
821 		if ((error = ext2fs_cgupdate(ump, waitfor)))
822 			allerror = error;
823 	}
824 	return (allerror);
825 }
826 
827 /*
828  * Look up a EXT2FS dinode number to find its incore vnode, otherwise read it
829  * in from disk.  If it is in core, wait for the lock bit to clear, then
830  * return the inode locked.  Detection and handling of mount points must be
831  * done by the calling routine.
832  */
833 int
834 ext2fs_vget(mp, ino, vpp)
835 	struct mount *mp;
836 	ino_t ino;
837 	struct vnode **vpp;
838 {
839 	struct m_ext2fs *fs;
840 	struct inode *ip;
841 	struct ufsmount *ump;
842 	struct buf *bp;
843 	struct vnode *vp;
844 	dev_t dev;
845 	int error;
846 	caddr_t cp;
847 
848 	ump = VFSTOUFS(mp);
849 	dev = ump->um_dev;
850 
851 	if ((*vpp = ufs_ihashget(dev, ino, LK_EXCLUSIVE)) != NULL)
852 		return (0);
853 
854 	/* Allocate a new vnode/inode. */
855 	if ((error = getnewvnode(VT_EXT2FS, mp, ext2fs_vnodeop_p, &vp)) != 0) {
856 		*vpp = NULL;
857 		return (error);
858 	}
859 
860 	do {
861 		if ((*vpp = ufs_ihashget(dev, ino, LK_EXCLUSIVE)) != NULL) {
862 			ungetnewvnode(vp);
863 			return (0);
864 		}
865 	} while (lockmgr(&ufs_hashlock, LK_EXCLUSIVE|LK_SLEEPFAIL, 0));
866 
867 	ip = pool_get(&ext2fs_inode_pool, PR_WAITOK);
868 	memset((caddr_t)ip, 0, sizeof(struct inode));
869 	vp->v_data = ip;
870 	ip->i_vnode = vp;
871 	ip->i_e2fs = fs = ump->um_e2fs;
872 	ip->i_dev = dev;
873 	ip->i_number = ino;
874 	ip->i_e2fs_last_lblk = 0;
875 	ip->i_e2fs_last_blk = 0;
876 
877 	/*
878 	 * Put it onto its hash chain and lock it so that other requests for
879 	 * this inode will block if they arrive while we are sleeping waiting
880 	 * for old data structures to be purged or for the contents of the
881 	 * disk portion of this inode to be read.
882 	 */
883 	ufs_ihashins(ip);
884 	lockmgr(&ufs_hashlock, LK_RELEASE, 0);
885 
886 	/* Read in the disk contents for the inode, copy into the inode. */
887 	error = bread(ump->um_devvp, fsbtodb(fs, ino_to_fsba(fs, ino)),
888 			  (int)fs->e2fs_bsize, NOCRED, &bp);
889 	if (error) {
890 		/*
891 		 * The inode does not contain anything useful, so it would
892 		 * be misleading to leave it on its hash chain. With mode
893 		 * still zero, it will be unlinked and returned to the free
894 		 * list by vput().
895 		 */
896 		vput(vp);
897 		brelse(bp);
898 		*vpp = NULL;
899 		return (error);
900 	}
901 	cp = (caddr_t)bp->b_data +
902 	    (ino_to_fsbo(fs, ino) * EXT2_DINODE_SIZE);
903 	e2fs_iload((struct ext2fs_dinode *)cp, &ip->i_din.e2fs_din);
904 	brelse(bp);
905 
906 	/* If the inode was deleted, reset all fields */
907 	if (ip->i_e2fs_dtime != 0) {
908 		ip->i_e2fs_mode = ip->i_e2fs_size = ip->i_e2fs_nblock = 0;
909 		memset(ip->i_e2fs_blocks, 0, sizeof(ip->i_e2fs_blocks));
910 	}
911 
912 	/*
913 	 * Initialize the vnode from the inode, check for aliases.
914 	 * Note that the underlying vnode may have changed.
915 	 */
916 	error = ext2fs_vinit(mp, ext2fs_specop_p, ext2fs_fifoop_p, &vp);
917 	if (error) {
918 		vput(vp);
919 		*vpp = NULL;
920 		return (error);
921 	}
922 	/*
923 	 * Finish inode initialization now that aliasing has been resolved.
924 	 */
925 	ip->i_devvp = ump->um_devvp;
926 	VREF(ip->i_devvp);
927 	/*
928 	 * Set up a generation number for this inode if it does not
929 	 * already have one. This should only happen on old filesystems.
930 	 */
931 	if (ip->i_e2fs_gen == 0) {
932 		if (++ext2gennumber < (u_long)time.tv_sec)
933 			ext2gennumber = time.tv_sec;
934 		ip->i_e2fs_gen = ext2gennumber;
935 		if ((vp->v_mount->mnt_flag & MNT_RDONLY) == 0)
936 			ip->i_flag |= IN_MODIFIED;
937 	}
938 
939 	vp->v_uvm.u_size = ip->i_e2fs_size;
940 	*vpp = vp;
941 	return (0);
942 }
943 
944 /*
945  * File handle to vnode
946  *
947  * Have to be really careful about stale file handles:
948  * - check that the inode number is valid
949  * - call ext2fs_vget() to get the locked inode
950  * - check for an unallocated inode (i_mode == 0)
951  */
952 int
953 ext2fs_fhtovp(mp, fhp, vpp)
954 	struct mount *mp;
955 	struct fid *fhp;
956 	struct vnode **vpp;
957 {
958 	struct inode *ip;
959 	struct vnode *nvp;
960 	int error;
961 	struct ufid *ufhp;
962 	struct m_ext2fs *fs;
963 
964 	ufhp = (struct ufid *)fhp;
965 	fs = VFSTOUFS(mp)->um_e2fs;
966 	if ((ufhp->ufid_ino < EXT2_FIRSTINO && ufhp->ufid_ino != EXT2_ROOTINO) ||
967 		ufhp->ufid_ino >= fs->e2fs_ncg * fs->e2fs.e2fs_ipg)
968 		return (ESTALE);
969 
970 	if ((error = VFS_VGET(mp, ufhp->ufid_ino, &nvp)) != 0) {
971 		*vpp = NULLVP;
972 		return (error);
973 	}
974 	ip = VTOI(nvp);
975 	if (ip->i_e2fs_mode == 0 || ip->i_e2fs_dtime != 0 ||
976 		ip->i_e2fs_gen != ufhp->ufid_gen) {
977 		vput(nvp);
978 		*vpp = NULLVP;
979 		return (ESTALE);
980 	}
981 	*vpp = nvp;
982 	return (0);
983 }
984 
985 /*
986  * Vnode pointer to File handle
987  */
988 /* ARGSUSED */
989 int
990 ext2fs_vptofh(vp, fhp)
991 	struct vnode *vp;
992 	struct fid *fhp;
993 {
994 	struct inode *ip;
995 	struct ufid *ufhp;
996 
997 	ip = VTOI(vp);
998 	ufhp = (struct ufid *)fhp;
999 	ufhp->ufid_len = sizeof(struct ufid);
1000 	ufhp->ufid_ino = ip->i_number;
1001 	ufhp->ufid_gen = ip->i_e2fs_gen;
1002 	return (0);
1003 }
1004 
1005 int
1006 ext2fs_sysctl(name, namelen, oldp, oldlenp, newp, newlen, p)
1007 	int *name;
1008 	u_int namelen;
1009 	void *oldp;
1010 	size_t *oldlenp;
1011 	void *newp;
1012 	size_t newlen;
1013 	struct proc *p;
1014 {
1015 	return (EOPNOTSUPP);
1016 }
1017 
1018 /*
1019  * Write a superblock and associated information back to disk.
1020  */
1021 int
1022 ext2fs_sbupdate(mp, waitfor)
1023 	struct ufsmount *mp;
1024 	int waitfor;
1025 {
1026 	struct m_ext2fs *fs = mp->um_e2fs;
1027 	struct buf *bp;
1028 	int error = 0;
1029 
1030 	bp = getblk(mp->um_devvp, SBLOCK, SBSIZE, 0, 0);
1031 	e2fs_sbsave(&fs->e2fs, (struct ext2fs*)bp->b_data);
1032 	if (waitfor == MNT_WAIT)
1033 		error = bwrite(bp);
1034 	else
1035 		bawrite(bp);
1036 	return (error);
1037 }
1038 
1039 int
1040 ext2fs_cgupdate(mp, waitfor)
1041 	struct ufsmount *mp;
1042 	int waitfor;
1043 {
1044 	struct m_ext2fs *fs = mp->um_e2fs;
1045 	struct buf *bp;
1046 	int i, error = 0, allerror = 0;
1047 
1048 	allerror = ext2fs_sbupdate(mp, waitfor);
1049 	for (i = 0; i < fs->e2fs_ngdb; i++) {
1050 		bp = getblk(mp->um_devvp, fsbtodb(fs, ((fs->e2fs_bsize>1024)?0:1)+i+1),
1051 			fs->e2fs_bsize, 0, 0);
1052 		e2fs_cgsave(&fs->e2fs_gd[i* fs->e2fs_bsize / sizeof(struct ext2_gd)],
1053 				(struct ext2_gd*)bp->b_data, fs->e2fs_bsize);
1054 		if (waitfor == MNT_WAIT)
1055 			error = bwrite(bp);
1056 		else
1057 			bawrite(bp);
1058 	}
1059 
1060 	if (!allerror && error)
1061 		allerror = error;
1062 	return (allerror);
1063 }
1064 
1065 static int
1066 ext2fs_checksb(fs, ronly)
1067 	struct ext2fs *fs;
1068 	int ronly;
1069 {
1070 	if (fs2h16(fs->e2fs_magic) != E2FS_MAGIC) {
1071 		return (EIO);		/* XXX needs translation */
1072 	}
1073 	if (fs2h32(fs->e2fs_rev) > E2FS_REV1) {
1074 #ifdef DIAGNOSTIC
1075 		printf("Ext2 fs: unsupported revision number: %x\n",
1076 					fs2h32(fs->e2fs_rev));
1077 #endif
1078 		return (EIO);		/* XXX needs translation */
1079 	}
1080 	if (fs2h32(fs->e2fs_log_bsize) > 2) { /* block size = 1024|2048|4096 */
1081 #ifdef DIAGNOSTIC
1082 		printf("Ext2 fs: bad block size: %d (expected <=2 for ext2 fs)\n",
1083 			fs2h32(fs->e2fs_log_bsize));
1084 #endif
1085 		return (EIO);	   /* XXX needs translation */
1086 	}
1087 	if (fs2h32(fs->e2fs_rev) > E2FS_REV0) {
1088 		if (fs2h32(fs->e2fs_first_ino) != EXT2_FIRSTINO ||
1089 		    fs2h16(fs->e2fs_inode_size) != EXT2_DINODE_SIZE) {
1090 			printf("Ext2 fs: unsupported inode size\n");
1091 			return (EINVAL);      /* XXX needs translation */
1092 		}
1093 		if (fs2h32(fs->e2fs_features_incompat) &
1094 		    ~EXT2F_INCOMPAT_SUPP) {
1095 			printf("Ext2 fs: unsupported optionnal feature\n");
1096 			return (EINVAL);      /* XXX needs translation */
1097 		}
1098 		if (!ronly && fs2h32(fs->e2fs_features_rocompat) &
1099 		    ~EXT2F_ROCOMPAT_SUPP) {
1100 			return (EROFS);      /* XXX needs translation */
1101 		}
1102 	}
1103 	return (0);
1104 }
1105