xref: /netbsd-src/sys/ufs/lfs/lfs_vfsops.c (revision 9fbd88883c38d0c0fbfcbe66d76fe6b0fab3f9de)
1 /*	$NetBSD: lfs_vfsops.c,v 1.71 2001/12/18 07:51:18 chs Exp $	*/
2 
3 /*-
4  * Copyright (c) 1999, 2000 The NetBSD Foundation, Inc.
5  * All rights reserved.
6  *
7  * This code is derived from software contributed to The NetBSD Foundation
8  * by Konrad E. Schroder <perseant@hhhh.org>.
9  *
10  * Redistribution and use in source and binary forms, with or without
11  * modification, are permitted provided that the following conditions
12  * are met:
13  * 1. Redistributions of source code must retain the above copyright
14  *    notice, this list of conditions and the following disclaimer.
15  * 2. Redistributions in binary form must reproduce the above copyright
16  *    notice, this list of conditions and the following disclaimer in the
17  *    documentation and/or other materials provided with the distribution.
18  * 3. All advertising materials mentioning features or use of this software
19  *    must display the following acknowledgement:
20  *      This product includes software developed by the NetBSD
21  *      Foundation, Inc. and its contributors.
22  * 4. Neither the name of The NetBSD Foundation nor the names of its
23  *    contributors may be used to endorse or promote products derived
24  *    from this software without specific prior written permission.
25  *
26  * THIS SOFTWARE IS PROVIDED BY THE NETBSD FOUNDATION, INC. AND CONTRIBUTORS
27  * ``AS IS'' AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED
28  * TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR
29  * PURPOSE ARE DISCLAIMED.  IN NO EVENT SHALL THE FOUNDATION OR CONTRIBUTORS
30  * BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR
31  * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF
32  * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS
33  * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN
34  * CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE)
35  * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE
36  * POSSIBILITY OF SUCH DAMAGE.
37  */
38 /*-
39  * Copyright (c) 1989, 1991, 1993, 1994
40  *	The Regents of the University of California.  All rights reserved.
41  *
42  * Redistribution and use in source and binary forms, with or without
43  * modification, are permitted provided that the following conditions
44  * are met:
45  * 1. Redistributions of source code must retain the above copyright
46  *    notice, this list of conditions and the following disclaimer.
47  * 2. Redistributions in binary form must reproduce the above copyright
48  *    notice, this list of conditions and the following disclaimer in the
49  *    documentation and/or other materials provided with the distribution.
50  * 3. All advertising materials mentioning features or use of this software
51  *    must display the following acknowledgement:
52  *	This product includes software developed by the University of
53  *	California, Berkeley and its contributors.
54  * 4. Neither the name of the University nor the names of its contributors
55  *    may be used to endorse or promote products derived from this software
56  *    without specific prior written permission.
57  *
58  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
59  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
60  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
61  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
62  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
63  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
64  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
65  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
66  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
67  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
68  * SUCH DAMAGE.
69  *
70  *	@(#)lfs_vfsops.c	8.20 (Berkeley) 6/10/95
71  */
72 
73 #include <sys/cdefs.h>
74 __KERNEL_RCSID(0, "$NetBSD: lfs_vfsops.c,v 1.71 2001/12/18 07:51:18 chs Exp $");
75 
76 #if defined(_KERNEL_OPT)
77 #include "opt_quota.h"
78 #endif
79 
80 #include <sys/param.h>
81 #include <sys/systm.h>
82 #include <sys/namei.h>
83 #include <sys/proc.h>
84 #include <sys/kernel.h>
85 #include <sys/vnode.h>
86 #include <sys/mount.h>
87 #include <sys/buf.h>
88 #include <sys/device.h>
89 #include <sys/mbuf.h>
90 #include <sys/file.h>
91 #include <sys/disklabel.h>
92 #include <sys/ioctl.h>
93 #include <sys/errno.h>
94 #include <sys/malloc.h>
95 #include <sys/pool.h>
96 #include <sys/socket.h>
97 #include <uvm/uvm_extern.h>
98 #include <sys/sysctl.h>
99 
100 #include <miscfs/specfs/specdev.h>
101 
102 #include <ufs/ufs/quota.h>
103 #include <ufs/ufs/inode.h>
104 #include <ufs/ufs/ufsmount.h>
105 #include <ufs/ufs/ufs_extern.h>
106 
107 #include <ufs/lfs/lfs.h>
108 #include <ufs/lfs/lfs_extern.h>
109 
110 int lfs_mountfs(struct vnode *, struct mount *, struct proc *);
111 
112 extern const struct vnodeopv_desc lfs_vnodeop_opv_desc;
113 extern const struct vnodeopv_desc lfs_specop_opv_desc;
114 extern const struct vnodeopv_desc lfs_fifoop_opv_desc;
115 
116 const struct vnodeopv_desc * const lfs_vnodeopv_descs[] = {
117 	&lfs_vnodeop_opv_desc,
118 	&lfs_specop_opv_desc,
119 	&lfs_fifoop_opv_desc,
120 	NULL,
121 };
122 
123 struct vfsops lfs_vfsops = {
124 	MOUNT_LFS,
125 	lfs_mount,
126 	ufs_start,
127 	lfs_unmount,
128 	ufs_root,
129 	ufs_quotactl,
130 	lfs_statfs,
131 	lfs_sync,
132 	lfs_vget,
133 	lfs_fhtovp,
134 	lfs_vptofh,
135 	lfs_init,
136 	lfs_reinit,
137 	lfs_done,
138 	lfs_sysctl,
139 	lfs_mountroot,
140 	ufs_check_export,
141 	lfs_vnodeopv_descs,
142 };
143 
144 struct genfs_ops lfs_genfsops = {
145 	NULL,
146 	NULL,
147 	genfs_compat_gop_write,
148 };
149 
150 struct pool lfs_inode_pool;
151 
152 extern int locked_queue_count;
153 extern long locked_queue_bytes;
154 
155 /*
156  * Initialize the filesystem, most work done by ufs_init.
157  */
158 void
159 lfs_init()
160 {
161 	ufs_init();
162 
163 	/*
164 	 * XXX Same structure as FFS inodes?  Should we share a common pool?
165 	 */
166 	pool_init(&lfs_inode_pool, sizeof(struct inode), 0, 0, 0,
167 		  "lfsinopl", 0, pool_page_alloc_nointr, pool_page_free_nointr,
168 		  M_LFSNODE);
169 }
170 
171 void
172 lfs_reinit()
173 {
174 	ufs_reinit();
175 }
176 
177 void
178 lfs_done()
179 {
180 	ufs_done();
181 	pool_destroy(&lfs_inode_pool);
182 }
183 
184 /*
185  * Called by main() when ufs is going to be mounted as root.
186  */
187 int
188 lfs_mountroot()
189 {
190 	extern struct vnode *rootvp;
191 	struct mount *mp;
192 	struct proc *p = curproc;	/* XXX */
193 	int error;
194 
195 	if (root_device->dv_class != DV_DISK)
196 		return (ENODEV);
197 
198 	if (rootdev == NODEV)
199 	  	return (ENODEV);
200 	/*
201 	 * Get vnodes for swapdev and rootdev.
202 	 */
203 	if ((error = bdevvp(rootdev, &rootvp))) {
204 		printf("lfs_mountroot: can't setup bdevvp's");
205 		return (error);
206 	}
207 	if ((error = vfs_rootmountalloc(MOUNT_LFS, "root_device", &mp))) {
208 		vrele(rootvp);
209 		return (error);
210 	}
211 	if ((error = lfs_mountfs(rootvp, mp, p))) {
212 		mp->mnt_op->vfs_refcount--;
213 		vfs_unbusy(mp);
214 		free(mp, M_MOUNT);
215 		vrele(rootvp);
216 		return (error);
217 	}
218 	simple_lock(&mountlist_slock);
219 	CIRCLEQ_INSERT_TAIL(&mountlist, mp, mnt_list);
220 	simple_unlock(&mountlist_slock);
221 	(void)lfs_statfs(mp, &mp->mnt_stat, p);
222 	vfs_unbusy(mp);
223 	inittodr(VFSTOUFS(mp)->um_lfs->lfs_tstamp);
224 	return (0);
225 }
226 
227 /*
228  * VFS Operations.
229  *
230  * mount system call
231  */
232 int
233 lfs_mount(struct mount *mp, const char *path, void *data, struct nameidata *ndp, struct proc *p)
234 {
235 	struct vnode *devvp;
236 	struct ufs_args args;
237 	struct ufsmount *ump = NULL;
238 	struct lfs *fs = NULL;				/* LFS */
239 	size_t size;
240 	int error;
241 	mode_t accessmode;
242 
243 	error = copyin(data, (caddr_t)&args, sizeof (struct ufs_args));
244 	if (error)
245 		return (error);
246 
247 #if 0
248 	/* Until LFS can do NFS right.		XXX */
249 	if (args.export.ex_flags & MNT_EXPORTED)
250 		return (EINVAL);
251 #endif
252 
253 	/*
254 	 * If updating, check whether changing from read-only to
255 	 * read/write; if there is no device name, that's all we do.
256 	 */
257 	if (mp->mnt_flag & MNT_UPDATE) {
258 		ump = VFSTOUFS(mp);
259 		fs = ump->um_lfs;
260 		if (fs->lfs_ronly && (mp->mnt_flag & MNT_WANTRDWR)) {
261 			/*
262 			 * If upgrade to read-write by non-root, then verify
263 			 * that user has necessary permissions on the device.
264 			 */
265 			if (p->p_ucred->cr_uid != 0) {
266 				vn_lock(ump->um_devvp, LK_EXCLUSIVE | LK_RETRY);
267 				error = VOP_ACCESS(ump->um_devvp, VREAD|VWRITE,
268 						   p->p_ucred, p);
269 				VOP_UNLOCK(ump->um_devvp, 0);
270 				if (error)
271 					return (error);
272 			}
273 			fs->lfs_ronly = 0;
274 		}
275 		if (args.fspec == 0) {
276 			/*
277 			 * Process export requests.
278 			 */
279 			return (vfs_export(mp, &ump->um_export, &args.export));
280 		}
281 	}
282 	/*
283 	 * Not an update, or updating the name: look up the name
284 	 * and verify that it refers to a sensible block device.
285 	 */
286 	NDINIT(ndp, LOOKUP, FOLLOW, UIO_USERSPACE, args.fspec, p);
287 	if ((error = namei(ndp)) != 0)
288 		return (error);
289 	devvp = ndp->ni_vp;
290 	if (devvp->v_type != VBLK) {
291 		vrele(devvp);
292 		return (ENOTBLK);
293 	}
294 	if (major(devvp->v_rdev) >= nblkdev) {
295 		vrele(devvp);
296 		return (ENXIO);
297 	}
298 	/*
299 	 * If mount by non-root, then verify that user has necessary
300 	 * permissions on the device.
301 	 */
302 	if (p->p_ucred->cr_uid != 0) {
303 		accessmode = VREAD;
304 		if ((mp->mnt_flag & MNT_RDONLY) == 0)
305 			accessmode |= VWRITE;
306 		vn_lock(devvp, LK_EXCLUSIVE | LK_RETRY);
307 		error = VOP_ACCESS(devvp, accessmode, p->p_ucred, p);
308 		if (error) {
309 			vput(devvp);
310 			return (error);
311 		}
312 		VOP_UNLOCK(devvp, 0);
313 	}
314 	if ((mp->mnt_flag & MNT_UPDATE) == 0)
315 		error = lfs_mountfs(devvp, mp, p);		/* LFS */
316 	else {
317 		if (devvp != ump->um_devvp)
318 			error = EINVAL;	/* needs translation */
319 		else
320 			vrele(devvp);
321 	}
322 	if (error) {
323 		vrele(devvp);
324 		return (error);
325 	}
326 	ump = VFSTOUFS(mp);
327 	fs = ump->um_lfs;					/* LFS */
328 	(void)copyinstr(path, fs->lfs_fsmnt, sizeof(fs->lfs_fsmnt) - 1, &size);
329 	bzero(fs->lfs_fsmnt + size, sizeof(fs->lfs_fsmnt) - size);
330 	bcopy(fs->lfs_fsmnt, mp->mnt_stat.f_mntonname, MNAMELEN);
331 	(void) copyinstr(args.fspec, mp->mnt_stat.f_mntfromname, MNAMELEN - 1,
332 			 &size);
333 	bzero(mp->mnt_stat.f_mntfromname + size, MNAMELEN - size);
334 	return (0);
335 }
336 
337 /*
338  * Roll-forward code.
339  */
340 
341 /*
342  * Load the appropriate indirect block, and change the appropriate pointer.
343  * Mark the block dirty.  Do segment and avail accounting.
344  */
345 static int
346 update_meta(struct lfs *fs, ino_t ino, int version, ufs_daddr_t lbn,
347 	    daddr_t ndaddr, size_t size, struct proc *p)
348 {
349 	int error;
350 	struct vnode *vp;
351 	struct inode *ip;
352 	daddr_t odaddr, ooff;
353 	struct indir a[NIADDR], *ap;
354 	struct buf *bp;
355 	SEGUSE *sup;
356 	int num;
357 
358 	if ((error = lfs_rf_valloc(fs, ino, version, p, &vp)) != 0) {
359 #ifdef DEBUG_LFS_RFW
360 		printf("update_meta: ino %d: lfs_rf_valloc returned %d\n", ino,
361 		       error);
362 #endif
363 		return error;
364 	}
365 
366 	if ((error = VOP_BALLOC(vp, (lbn << fs->lfs_bshift), size,
367 				NOCRED, 0, &bp)) != 0) {
368 		vput(vp);
369 		return (error);
370 	}
371 	/* No need to write, the block is already on disk */
372 	if (bp->b_flags & B_DELWRI) {
373 		LFS_UNLOCK_BUF(bp);
374 		fs->lfs_avail += btofsb(fs, bp->b_bcount);
375 	}
376 	bp->b_flags |= B_INVAL;
377 	brelse(bp);
378 
379 	/*
380 	 * Extend the file, if it is not large enough already.
381 	 * XXX this is not exactly right, we don't know how much of the
382 	 * XXX last block is actually used.  We hope that an inode will
383 	 * XXX appear later to give the correct size.
384 	 */
385 	ip = VTOI(vp);
386 	if (ip->i_ffs_size <= (lbn << fs->lfs_bshift)) {
387 		if (lbn < NDADDR)
388 			ip->i_ffs_size = (lbn << fs->lfs_bshift) +
389 				(size - fs->lfs_fsize) + 1;
390 		else
391 			ip->i_ffs_size = (lbn << fs->lfs_bshift) + 1;
392 	}
393 
394 	error = ufs_bmaparray(vp, lbn, &odaddr, &a[0], &num, NULL);
395 	if (error) {
396 #ifdef DEBUG_LFS_RFW
397 		printf("update_meta: ufs_bmaparray returned %d\n", error);
398 #endif
399 		vput(vp);
400 		return error;
401 	}
402 	switch (num) {
403 	    case 0:
404 		ooff = ip->i_ffs_db[lbn];
405 		if (ooff == UNWRITTEN)
406 			ip->i_ffs_blocks += btofsb(fs, size);
407 		ip->i_ffs_db[lbn] = ndaddr;
408 		break;
409 	    case 1:
410 		ooff = ip->i_ffs_ib[a[0].in_off];
411 		if (ooff == UNWRITTEN)
412 			ip->i_ffs_blocks += btofsb(fs, size);
413 		ip->i_ffs_ib[a[0].in_off] = ndaddr;
414 		break;
415 	    default:
416 		ap = &a[num - 1];
417 		if (bread(vp, ap->in_lbn, fs->lfs_bsize, NOCRED, &bp))
418 			panic("update_meta: bread bno %d", ap->in_lbn);
419 
420 		ooff = ((ufs_daddr_t *)bp->b_data)[ap->in_off];
421 		if (ooff == UNWRITTEN)
422 			ip->i_ffs_blocks += btofsb(fs, size);
423 		((ufs_daddr_t *)bp->b_data)[ap->in_off] = ndaddr;
424 		(void) VOP_BWRITE(bp);
425 	}
426 	LFS_SET_UINO(ip, IN_CHANGE | IN_MODIFIED | IN_UPDATE);
427 
428 	/* Update segment usage information. */
429 	if (odaddr > 0) {
430 		LFS_SEGENTRY(sup, fs, dtosn(fs, dbtofsb(fs, odaddr)), bp);
431 #ifdef DIAGNOSTIC
432 		if (sup->su_nbytes < size) {
433 			panic("update_meta: negative bytes "
434 			      "(segment %d short by %ld)\n",
435 			      dtosn(fs, dbtofsb(fs, odaddr)), (long)size - sup->su_nbytes);
436 			sup->su_nbytes = size;
437 		}
438 #endif
439 		sup->su_nbytes -= size;
440 		VOP_BWRITE(bp);
441 	}
442 	LFS_SEGENTRY(sup, fs, dtosn(fs, ndaddr), bp);
443 	sup->su_nbytes += size;
444 	VOP_BWRITE(bp);
445 
446 	/* Fix this so it can be released */
447 	/* ip->i_lfs_effnblks = ip->i_ffs_blocks; */
448 
449 #ifdef DEBUG_LFS_RFW
450 	/* Now look again to make sure it worked */
451 	ufs_bmaparray(vp, lbn, &odaddr, &a[0], &num, NULL );
452 	if (dbtofsb(fs, odaddr) != ndaddr)
453 		printf("update_meta: failed setting ino %d lbn %d to %x\n",
454 		       ino, lbn, ndaddr);
455 #endif
456 	vput(vp);
457 	return 0;
458 }
459 
460 static int
461 update_inoblk(struct lfs *fs, daddr_t offset, struct ucred *cred,
462 	      struct proc *p)
463 {
464 	struct vnode *devvp, *vp;
465 	struct inode *ip;
466 	struct dinode *dip;
467 	struct buf *dbp, *ibp;
468 	int error;
469 	daddr_t daddr;
470 	IFILE *ifp;
471 	SEGUSE *sup;
472 
473 	devvp = VTOI(fs->lfs_ivnode)->i_devvp;
474 
475 	/*
476 	 * Get the inode, update times and perms.
477 	 * DO NOT update disk blocks, we do that separately.
478 	 */
479 	error = bread(devvp, fsbtodb(fs, offset), fs->lfs_ibsize, cred, &dbp);
480 	if (error) {
481 #ifdef DEBUG_LFS_RFW
482 		printf("update_inoblk: bread returned %d\n", error);
483 #endif
484 		return error;
485 	}
486 	dip = ((struct dinode *)(dbp->b_data)) + INOPB(fs);
487 	while (--dip >= (struct dinode *)dbp->b_data) {
488 		if (dip->di_inumber > LFS_IFILE_INUM) {
489 			/* printf("ino %d version %d\n", dip->di_inumber,
490 			       dip->di_gen); */
491 			error = lfs_rf_valloc(fs, dip->di_inumber, dip->di_gen,
492 					      p, &vp);
493 			if (error) {
494 #ifdef DEBUG_LFS_RFW
495 				printf("update_inoblk: lfs_rf_valloc returned %d\n", error);
496 #endif
497 				continue;
498 			}
499 			ip = VTOI(vp);
500 			if (dip->di_size != ip->i_ffs_size)
501 				VOP_TRUNCATE(vp, dip->di_size, 0, NOCRED, p);
502 			/* Get mode, link count, size, and times */
503 			memcpy(&ip->i_din.ffs_din, dip,
504 			       offsetof(struct dinode, di_db[0]));
505 
506 			/* Then the rest, except di_blocks */
507 			ip->i_ffs_flags = dip->di_flags;
508 			ip->i_ffs_gen = dip->di_gen;
509 			ip->i_ffs_uid = dip->di_uid;
510 			ip->i_ffs_gid = dip->di_gid;
511 
512 			ip->i_ffs_effnlink = dip->di_nlink;
513 
514 			LFS_SET_UINO(ip, IN_CHANGE | IN_MODIFIED | IN_UPDATE);
515 
516 			/* Re-initialize to get type right */
517 			ufs_vinit(vp->v_mount, lfs_specop_p, lfs_fifoop_p,
518 				  &vp);
519 			vput(vp);
520 
521 			/* Record change in location */
522 			LFS_IENTRY(ifp, fs, dip->di_inumber, ibp);
523 			daddr = ifp->if_daddr;
524 			ifp->if_daddr = dbtofsb(fs, dbp->b_blkno);
525 			error = VOP_BWRITE(ibp); /* Ifile */
526 			/* And do segment accounting */
527 			if (dtosn(fs, daddr) != dtosn(fs, dbtofsb(fs, dbp->b_blkno))) {
528 				if (daddr > 0) {
529 					LFS_SEGENTRY(sup, fs, dtosn(fs, daddr),
530 						     ibp);
531 					sup->su_nbytes -= DINODE_SIZE;
532 					VOP_BWRITE(ibp);
533 				}
534 				LFS_SEGENTRY(sup, fs, dtosn(fs, dbtofsb(fs, dbp->b_blkno)),
535 					     ibp);
536 				sup->su_nbytes += DINODE_SIZE;
537 				VOP_BWRITE(ibp);
538 			}
539 		}
540 	}
541 	dbp->b_flags |= B_AGE;
542 	brelse(dbp);
543 
544 	return 0;
545 }
546 
547 #define CHECK_CKSUM   0x0001  /* Check the checksum to make sure it's valid */
548 #define CHECK_UPDATE  0x0002  /* Update Ifile for new data blocks / inodes */
549 
550 static daddr_t
551 check_segsum(struct lfs *fs, daddr_t offset,
552 	     struct ucred *cred, int flags, int *pseg_flags, struct proc *p)
553 {
554 	struct vnode *devvp;
555 	struct buf *bp, *dbp;
556 	int error, nblocks, ninos, i, j;
557 	SEGSUM *ssp;
558 	u_long *dp, *datap; /* XXX u_int32_t */
559 	daddr_t *iaddr, oldoffset;
560 	FINFO *fip;
561 	SEGUSE *sup;
562 	size_t size;
563 	u_int64_t serial;
564 
565 	devvp = VTOI(fs->lfs_ivnode)->i_devvp;
566 	/*
567 	 * If the segment has a superblock and we're at the top
568 	 * of the segment, skip the superblock.
569 	 */
570 	if (sntod(fs, dtosn(fs, offset)) == offset) {
571        		LFS_SEGENTRY(sup, fs, dtosn(fs, offset), bp);
572        		if (sup->su_flags & SEGUSE_SUPERBLOCK)
573 			offset += btofsb(fs, LFS_SBPAD);
574        		brelse(bp);
575 	}
576 
577 	/* Read in the segment summary */
578 	error = bread(devvp, offset, fs->lfs_sumsize, cred, &bp);
579 	if (error)
580 		return -1;
581 
582 	/* Check summary checksum */
583 	ssp = (SEGSUM *)bp->b_data;
584 	if (flags & CHECK_CKSUM) {
585 		if (ssp->ss_sumsum != cksum(&ssp->ss_datasum,
586 					   fs->lfs_sumsize -
587 					   sizeof(ssp->ss_sumsum))) {
588 #ifdef DEBUG_LFS_RFW
589 			printf("Sumsum error at 0x%x\n", offset);
590 #endif
591 			offset = -1;
592 			goto err1;
593 		}
594 		if (ssp->ss_nfinfo == 0 && ssp->ss_ninos == 0) {
595 #ifdef DEBUG_LFS_RFW
596 			printf("Empty pseg at 0x%x\n", offset);
597 #endif
598 			offset = -1;
599 			goto err1;
600 		}
601 		if (ssp->ss_create < fs->lfs_tstamp) {
602 #ifdef DEBUG_LFS_RFW
603 			printf("Old data at 0x%x\n", offset);
604 #endif
605 			offset = -1;
606 			goto err1;
607 		}
608 	}
609 	if (fs->lfs_version > 1) {
610 		serial = ssp->ss_serial;
611 		if (serial != fs->lfs_serial + 1) {
612 #ifdef DEBUG_LFS_RFW
613 			printf("Unexpected serial number at 0x%x\n", offset);
614 #endif
615 			offset = -1;
616 			goto err1;
617 		}
618 		if (ssp->ss_ident != fs->lfs_ident) {
619 #ifdef DEBUG_LFS_RFW
620 			printf("Incorrect fsid (0x%x vs 0x%x) at 0x%x\n",
621 			       ssp->ss_ident, fs->lfs_ident, offset);
622 #endif
623 			offset = -1;
624 			goto err1;
625 		}
626 	}
627 	if (pseg_flags)
628 		*pseg_flags = ssp->ss_flags;
629 	oldoffset = offset;
630 	offset += btofsb(fs, fs->lfs_sumsize);
631 
632 	ninos = howmany(ssp->ss_ninos, INOPB(fs));
633 	iaddr = (daddr_t *)(bp->b_data + fs->lfs_sumsize - sizeof(daddr_t));
634 	if (flags & CHECK_CKSUM) {
635 		/* Count blocks */
636 		nblocks = 0;
637 		fip = (FINFO *)(bp->b_data + SEGSUM_SIZE(fs));
638 		for (i = 0; i < ssp->ss_nfinfo; ++i) {
639 			nblocks += fip->fi_nblocks;
640 			if (fip->fi_nblocks <= 0)
641 				break;
642 			fip = (FINFO *)(((char *)fip) + sizeof(FINFO) +
643 					(fip->fi_nblocks - 1) *
644 					sizeof(ufs_daddr_t));
645 		}
646 		nblocks += ninos;
647 		/* Create the sum array */
648 		datap = dp = (u_long *)malloc(nblocks * sizeof(u_long),
649 					      M_SEGMENT, M_WAITOK);
650 	}
651 
652 	/* Handle individual blocks */
653 	fip = (FINFO *)(bp->b_data + SEGSUM_SIZE(fs));
654 	for (i = 0; i < ssp->ss_nfinfo || ninos; ++i) {
655 		/* Inode block? */
656 		if (ninos && *iaddr == offset) {
657 			if (flags & CHECK_CKSUM) {
658 				/* Read in the head and add to the buffer */
659 				error = bread(devvp, fsbtodb(fs, offset), fs->lfs_bsize,
660 					      cred, &dbp);
661 				if (error) {
662 					offset = -1;
663 					goto err2;
664 				}
665 				(*dp++) = ((u_long *)(dbp->b_data))[0];
666 				dbp->b_flags |= B_AGE;
667 				brelse(dbp);
668 			}
669 			if (flags & CHECK_UPDATE) {
670 				if ((error = update_inoblk(fs, offset, cred, p))
671 				    != 0) {
672 					offset = -1;
673 					goto err2;
674 				}
675 			}
676 			offset += btofsb(fs, fs->lfs_ibsize);
677 			--iaddr;
678 			--ninos;
679 			--i; /* compensate */
680 			continue;
681 		}
682 		/* printf("check: blocks from ino %d version %d\n",
683 		       fip->fi_ino, fip->fi_version); */
684 		size = fs->lfs_bsize;
685 		for (j = 0; j < fip->fi_nblocks; ++j) {
686 			if (j == fip->fi_nblocks - 1)
687 				size = fip->fi_lastlength;
688 			if (flags & CHECK_CKSUM) {
689 				error = bread(devvp, fsbtodb(fs, offset), size, cred, &dbp);
690 				if (error) {
691 					offset = -1;
692 					goto err2;
693 				}
694 				(*dp++) = ((u_long *)(dbp->b_data))[0];
695 				dbp->b_flags |= B_AGE;
696 				brelse(dbp);
697 			}
698 			/* Account for and update any direct blocks */
699 			if ((flags & CHECK_UPDATE) &&
700 			   fip->fi_ino > LFS_IFILE_INUM &&
701 			   fip->fi_blocks[j] >= 0) {
702 				update_meta(fs, fip->fi_ino, fip->fi_version,
703 					    fip->fi_blocks[j], offset, size, p);
704 			}
705 			offset += btofsb(fs, size);
706 		}
707 		fip = (FINFO *)(((char *)fip) + sizeof(FINFO)
708 				+ (fip->fi_nblocks - 1) * sizeof(ufs_daddr_t));
709 	}
710 	/* Checksum the array, compare */
711 	if ((flags & CHECK_CKSUM) &&
712 	   ssp->ss_datasum != cksum(datap, nblocks * sizeof(u_long)))
713 	{
714 #ifdef DEBUG_LFS_RFW
715 		printf("Datasum error at 0x%x (wanted %x got %x)\n", offset,
716 		       ssp->ss_datasum, cksum(datap, nblocks *
717 					      sizeof(u_long)));
718 #endif
719 		offset = -1;
720 		goto err2;
721 	}
722 
723 	/* If we're at the end of the segment, move to the next */
724 	if (dtosn(fs, offset + btofsb(fs, fs->lfs_sumsize + fs->lfs_bsize)) !=
725 	   dtosn(fs, offset)) {
726 		if (dtosn(fs, offset) == dtosn(fs, ssp->ss_next)) {
727 			offset = -1;
728 			goto err2;
729 		}
730 		offset = ssp->ss_next;
731 #ifdef DEBUG_LFS_RFW
732 		printf("LFS roll forward: moving on to offset 0x%x "
733 		       " -> segment %d\n", offset, dtosn(fs,offset));
734 #endif
735 	}
736 
737 	if (flags & CHECK_UPDATE) {
738 		fs->lfs_avail -= (offset - oldoffset);
739 		/* Don't clog the buffer queue */
740 		if (locked_queue_count > LFS_MAX_BUFS ||
741 		    locked_queue_bytes > LFS_MAX_BYTES) {
742 			++fs->lfs_writer;
743 			lfs_flush(fs, SEGM_CKP);
744 			if (--fs->lfs_writer == 0)
745 				wakeup(&fs->lfs_dirops);
746 		}
747 	}
748 
749     err2:
750 	if (flags & CHECK_CKSUM)
751 		free(datap, M_SEGMENT);
752     err1:
753 	bp->b_flags |= B_AGE;
754 	brelse(bp);
755 
756 	/* XXX should we update the serial number even for bad psegs? */
757 	if ((flags & CHECK_UPDATE) && offset > 0 && fs->lfs_version > 1)
758 		fs->lfs_serial = serial;
759 	return offset;
760 }
761 
762 /*
763  * Common code for mount and mountroot
764  * LFS specific
765  */
766 int
767 lfs_mountfs(struct vnode *devvp, struct mount *mp, struct proc *p)
768 {
769 	extern struct vnode *rootvp;
770 	struct dlfs *tdfs, *dfs, *adfs;
771 	struct lfs *fs;
772 	struct ufsmount *ump;
773 	struct vnode *vp;
774 	struct buf *bp, *abp;
775 	struct partinfo dpart;
776 	dev_t dev;
777 	int error, i, ronly, secsize, fsbsize;
778 	struct ucred *cred;
779 	CLEANERINFO *cip;
780         SEGUSE *sup;
781 	int flags, dirty, do_rollforward;
782 	daddr_t offset, oldoffset, lastgoodpseg, sb_addr;
783 	int sn, curseg;
784 
785 	cred = p ? p->p_ucred : NOCRED;
786 	/*
787 	 * Disallow multiple mounts of the same device.
788 	 * Disallow mounting of a device that is currently in use
789 	 * (except for root, which might share swap device for miniroot).
790 	 * Flush out any old buffers remaining from a previous use.
791 	 */
792 	if ((error = vfs_mountedon(devvp)) != 0)
793 		return (error);
794 	if (vcount(devvp) > 1 && devvp != rootvp)
795 		return (EBUSY);
796 	if ((error = vinvalbuf(devvp, V_SAVE, cred, p, 0, 0)) != 0)
797 		return (error);
798 
799 	ronly = (mp->mnt_flag & MNT_RDONLY) != 0;
800 	error = VOP_OPEN(devvp, ronly ? FREAD : FREAD|FWRITE, FSCRED, p);
801 	if (error)
802 		return (error);
803 	if (VOP_IOCTL(devvp, DIOCGPART, (caddr_t)&dpart, FREAD, cred, p) != 0)
804 		secsize = DEV_BSIZE;
805 	else
806 		secsize = dpart.disklab->d_secsize;
807 
808 	/* Don't free random space on error. */
809 	bp = NULL;
810 	abp = NULL;
811 	ump = NULL;
812 
813 	sb_addr = LFS_LABELPAD / secsize;
814 	while (1) {
815 		/* Read in the superblock. */
816 		error = bread(devvp, sb_addr, LFS_SBPAD, cred, &bp);
817 		if (error)
818 			goto out;
819 		dfs = (struct dlfs *)bp->b_data;
820 
821 		/* Check the basics. */
822 		if (dfs->dlfs_magic != LFS_MAGIC || dfs->dlfs_bsize >= MAXBSIZE ||
823 		    dfs->dlfs_version > LFS_VERSION ||
824 		    dfs->dlfs_bsize < sizeof(struct dlfs)) {
825 #ifdef DEBUG_LFS
826 			printf("lfs_mountfs: primary superblock sanity failed\n");
827 #endif
828 			error = EINVAL;		/* XXX needs translation */
829 			goto out;
830 		}
831 		if (dfs->dlfs_inodefmt > LFS_MAXINODEFMT)
832 			printf("lfs_mountfs: warning: unknown inode format %d\n",
833 			       dfs->dlfs_inodefmt);
834 
835 		if (dfs->dlfs_version == 1)
836 			fsbsize = secsize;
837 		else {
838 			fsbsize = 1 << (dfs->dlfs_bshift - dfs->dlfs_blktodb +
839 				dfs->dlfs_fsbtodb);
840 			/*
841 			 * Could be, if the frag size is large enough, that we
842 			 * don't have the "real" primary superblock.  If that's
843 			 * the case, get the real one, and try again.
844 			 */
845 			if (sb_addr != dfs->dlfs_sboffs[0] <<
846                                        dfs->dlfs_fsbtodb) {
847 /* #ifdef DEBUG_LFS */
848 				printf("lfs_mountfs: sb daddr 0x%x is not right, trying 0x%x\n",
849 					sb_addr, dfs->dlfs_sboffs[0] <<
850 						 dfs->dlfs_fsbtodb);
851 /* #endif */
852 				sb_addr = dfs->dlfs_sboffs[0] <<
853 					  dfs->dlfs_fsbtodb;
854 				brelse(bp);
855 				continue;
856 			}
857 		}
858 		break;
859 	}
860 
861 	/*
862 	 * Check the second superblock to see which is newer; then mount
863 	 * using the older of the two.  This is necessary to ensure that
864 	 * the filesystem is valid if it was not unmounted cleanly.
865 	 */
866 
867 	if (dfs->dlfs_sboffs[1] &&
868 	    dfs->dlfs_sboffs[1] - LFS_LABELPAD / fsbsize > LFS_SBPAD / fsbsize)
869 	{
870 		error = bread(devvp, dfs->dlfs_sboffs[1] * (fsbsize / secsize),
871 			LFS_SBPAD, cred, &abp);
872 		if (error)
873 			goto out;
874 		adfs = (struct dlfs *)abp->b_data;
875 
876 		if (dfs->dlfs_version == 1) {
877 			/* 1s resolution comparison */
878 			if (adfs->dlfs_tstamp < dfs->dlfs_tstamp)
879 				tdfs = adfs;
880 			else
881 				tdfs = dfs;
882 		} else {
883 			/* monotonic infinite-resolution comparison */
884 			if (adfs->dlfs_serial < dfs->dlfs_serial)
885 				tdfs = adfs;
886 			else
887 				tdfs = dfs;
888 		}
889 
890 		/* Check the basics. */
891 		if (tdfs->dlfs_magic != LFS_MAGIC ||
892 		    tdfs->dlfs_bsize > MAXBSIZE ||
893 	    	    tdfs->dlfs_version > LFS_VERSION ||
894 	    	    tdfs->dlfs_bsize < sizeof(struct dlfs)) {
895 #ifdef DEBUG_LFS
896 			printf("lfs_mountfs: alt superblock sanity failed\n");
897 #endif
898 			error = EINVAL;		/* XXX needs translation */
899 			goto out;
900 		}
901 	} else {
902 #ifdef DEBUG_LFS
903 		printf("lfs_mountfs: invalid alt superblock daddr=0x%x\n",
904 			dfs->dlfs_sboffs[1]);
905 #endif
906 		error = EINVAL;
907 		goto out;
908 	}
909 
910 	/* Allocate the mount structure, copy the superblock into it. */
911 	fs = malloc(sizeof(struct lfs), M_UFSMNT, M_WAITOK);
912 	memcpy(&fs->lfs_dlfs, tdfs, sizeof(struct dlfs));
913 
914 	/* Compatibility */
915 	if (fs->lfs_version < 2) {
916 		fs->lfs_sumsize = LFS_V1_SUMMARY_SIZE;
917 		fs->lfs_ibsize = fs->lfs_bsize;
918 		fs->lfs_start = fs->lfs_sboffs[0];
919 		fs->lfs_tstamp = fs->lfs_otstamp;
920 		fs->lfs_fsbtodb = 0;
921 	}
922 
923 	/* Before rolling forward, lock so vget will sleep for other procs */
924 	fs->lfs_flags = LFS_NOTYET;
925 	fs->lfs_rfpid = p->p_pid;
926 
927 	ump = malloc(sizeof *ump, M_UFSMNT, M_WAITOK);
928 	memset((caddr_t)ump, 0, sizeof *ump);
929 	ump->um_lfs = fs;
930 	if (sizeof(struct lfs) < LFS_SBPAD) {			/* XXX why? */
931 		bp->b_flags |= B_INVAL;
932 		abp->b_flags |= B_INVAL;
933 	}
934 	brelse(bp);
935 	bp = NULL;
936 	brelse(abp);
937 	abp = NULL;
938 
939 	/* Set up the I/O information */
940 	fs->lfs_devbsize = secsize;
941 	fs->lfs_iocount = 0;
942 	fs->lfs_diropwait = 0;
943 	fs->lfs_activesb = 0;
944 	fs->lfs_uinodes = 0;
945 	fs->lfs_ravail = 0;
946 	fs->lfs_sbactive = 0;
947 #ifdef LFS_TRACK_IOS
948 	for (i = 0; i < LFS_THROTTLE; i++)
949 		fs->lfs_pending[i] = LFS_UNUSED_DADDR;
950 #endif
951 
952 	/* Set up the ifile and lock aflags */
953 	fs->lfs_doifile = 0;
954 	fs->lfs_writer = 0;
955 	fs->lfs_dirops = 0;
956 	fs->lfs_nadirop = 0;
957 	fs->lfs_seglock = 0;
958 	lockinit(&fs->lfs_freelock, PINOD, "lfs_freelock", 0, 0);
959 
960 	/* Set the file system readonly/modify bits. */
961 	fs->lfs_ronly = ronly;
962 	if (ronly == 0)
963 		fs->lfs_fmod = 1;
964 
965 	/* Initialize the mount structure. */
966 	dev = devvp->v_rdev;
967 	mp->mnt_data = (qaddr_t)ump;
968 	mp->mnt_stat.f_fsid.val[0] = (long)dev;
969 	mp->mnt_stat.f_fsid.val[1] = makefstype(MOUNT_LFS);
970 	mp->mnt_stat.f_iosize = fs->lfs_bsize;
971 	mp->mnt_maxsymlinklen = fs->lfs_maxsymlinklen;
972 	mp->mnt_flag |= MNT_LOCAL;
973 	ump->um_flags = 0;
974 	ump->um_mountp = mp;
975 	ump->um_dev = dev;
976 	ump->um_devvp = devvp;
977 	ump->um_bptrtodb = fs->lfs_fsbtodb;
978 	ump->um_seqinc = fragstofsb(fs, fs->lfs_frag);
979 	ump->um_nindir = fs->lfs_nindir;
980 	ump->um_lognindir = ffs(fs->lfs_nindir) - 1;
981 	for (i = 0; i < MAXQUOTAS; i++)
982 		ump->um_quotas[i] = NULLVP;
983 	devvp->v_specmountpoint = mp;
984 
985 	/*
986 	 * We use the ifile vnode for almost every operation.  Instead of
987 	 * retrieving it from the hash table each time we retrieve it here,
988 	 * artificially increment the reference count and keep a pointer
989 	 * to it in the incore copy of the superblock.
990 	 */
991 	if ((error = VFS_VGET(mp, LFS_IFILE_INUM, &vp)) != 0) {
992 #ifdef DEBUG
993 		printf("lfs_mountfs: ifile vget failed, error=%d\n", error);
994 #endif
995 		goto out;
996 	}
997 	fs->lfs_ivnode = vp;
998 	VREF(vp);
999 	vput(vp);
1000 
1001 	/*
1002 	 * Roll forward.
1003 	 *
1004 	 * We don't automatically roll forward for v1 filesystems, because
1005 	 * of the danger that the clock was turned back between the last
1006 	 * checkpoint and crash.  This would roll forward garbage.
1007 	 *
1008 	 * v2 filesystems don't have this problem because they use a
1009 	 * monotonically increasing serial number instead of a timestamp.
1010 	 */
1011 #ifdef LFS_DO_ROLLFORWARD
1012 	do_rollforward = !fs->lfs_ronly;
1013 #else
1014 	do_rollforward = (fs->lfs_version > 1 && !fs->lfs_ronly &&
1015 			  !(fs->lfs_pflags & LFS_PF_CLEAN));
1016 #endif
1017 	if (do_rollforward) {
1018 		/*
1019 		 * Phase I: Find the address of the last good partial
1020 		 * segment that was written after the checkpoint.  Mark
1021 		 * the segments in question dirty, so they won't be
1022 		 * reallocated.
1023 		 */
1024 		lastgoodpseg = oldoffset = offset = fs->lfs_offset;
1025 		flags = 0x0;
1026 #ifdef DEBUG_LFS_RFW
1027 		printf("LFS roll forward phase 1: starting at offset 0x%x\n",
1028 		       offset);
1029 #endif
1030 		LFS_SEGENTRY(sup, fs, dtosn(fs, offset), bp);
1031 		if (!(sup->su_flags & SEGUSE_DIRTY))
1032 			--fs->lfs_nclean;
1033 		sup->su_flags |= SEGUSE_DIRTY;
1034 		(void) VOP_BWRITE(bp);
1035 		while ((offset = check_segsum(fs, offset, cred, CHECK_CKSUM,
1036 					      &flags, p)) > 0)
1037 		{
1038 			if (sntod(fs, oldoffset) != sntod(fs, offset)) {
1039 				LFS_SEGENTRY(sup, fs, dtosn(fs, oldoffset),
1040 					     bp);
1041 				if (!(sup->su_flags & SEGUSE_DIRTY))
1042 					--fs->lfs_nclean;
1043 				sup->su_flags |= SEGUSE_DIRTY;
1044 				(void) VOP_BWRITE(bp);
1045 			}
1046 
1047 #ifdef DEBUG_LFS_RFW
1048 			printf("LFS roll forward phase 1: offset=0x%x\n",
1049 			       offset);
1050 			if (flags & SS_DIROP) {
1051 				printf("lfs_mountfs: dirops at 0x%x\n",
1052 				       oldoffset);
1053 				if (!(flags & SS_CONT))
1054 					printf("lfs_mountfs: dirops end "
1055 					       "at 0x%x\n", oldoffset);
1056 			}
1057 #endif
1058 			if (!(flags & SS_CONT))
1059 				lastgoodpseg = offset;
1060 			oldoffset = offset;
1061 		}
1062 #ifdef DEBUG_LFS_RFW
1063 		if (flags & SS_CONT) {
1064 			printf("LFS roll forward: warning: incomplete "
1065 			       "dirops discarded\n");
1066 		}
1067 		printf("LFS roll forward phase 1: completed: "
1068 		       "lastgoodpseg=0x%x\n", lastgoodpseg);
1069 #endif
1070 		oldoffset = fs->lfs_offset;
1071 		if (fs->lfs_offset != lastgoodpseg) {
1072 			/* Don't overwrite what we're trying to preserve */
1073 			offset = fs->lfs_offset;
1074 			fs->lfs_offset = lastgoodpseg;
1075 			fs->lfs_curseg = sntod(fs, dtosn(fs, fs->lfs_offset));
1076 			for (sn = curseg = dtosn(fs, fs->lfs_curseg);;) {
1077 				sn = (sn + 1) % fs->lfs_nseg;
1078 				if (sn == curseg)
1079 					panic("lfs_mountfs: no clean segments");
1080 				LFS_SEGENTRY(sup, fs, sn, bp);
1081 				dirty = (sup->su_flags & SEGUSE_DIRTY);
1082 				brelse(bp);
1083 				if (!dirty)
1084 					break;
1085 			}
1086 			fs->lfs_nextseg = sntod(fs, sn);
1087 
1088 			/*
1089 			 * Phase II: Roll forward from the first superblock.
1090 			 */
1091 			while (offset != lastgoodpseg) {
1092 #ifdef DEBUG_LFS_RFW
1093 				printf("LFS roll forward phase 2: 0x%x\n",
1094 				       offset);
1095 #endif
1096 				offset = check_segsum(fs, offset, cred,
1097 						      CHECK_UPDATE, NULL, p);
1098 			}
1099 
1100 			/*
1101 			 * Finish: flush our changes to disk.
1102 			 */
1103 			lfs_segwrite(mp, SEGM_CKP | SEGM_SYNC);
1104 			printf("lfs_mountfs: roll forward recovered %d blocks\n",
1105 			       lastgoodpseg - oldoffset);
1106 		}
1107 #ifdef DEBUG_LFS_RFW
1108 		printf("LFS roll forward complete\n");
1109 #endif
1110 	}
1111 	/* If writing, sb is not clean; record in case of immediate crash */
1112 	if (!fs->lfs_ronly) {
1113 		fs->lfs_pflags &= ~LFS_PF_CLEAN;
1114 		lfs_writesuper(fs, fs->lfs_sboffs[0]);
1115 	}
1116 
1117 	/* Allow vget now that roll-forward is complete */
1118 	fs->lfs_flags &= ~(LFS_NOTYET);
1119 	wakeup(&fs->lfs_flags);
1120 
1121 	/*
1122 	 * Initialize the ifile cleaner info with information from
1123 	 * the superblock.
1124 	 */
1125 	LFS_CLEANERINFO(cip, fs, bp);
1126 	cip->clean = fs->lfs_nclean;
1127 	cip->dirty = fs->lfs_nseg - fs->lfs_nclean;
1128 	cip->avail = fs->lfs_avail;
1129 	cip->bfree = fs->lfs_bfree;
1130 	(void) VOP_BWRITE(bp); /* Ifile */
1131 
1132 	/*
1133 	 * Mark the current segment as ACTIVE, since we're going to
1134 	 * be writing to it.
1135 	 */
1136         LFS_SEGENTRY(sup, fs, dtosn(fs, fs->lfs_offset), bp);
1137         sup->su_flags |= SEGUSE_DIRTY | SEGUSE_ACTIVE;
1138         (void) VOP_BWRITE(bp); /* Ifile */
1139 
1140 	return (0);
1141 out:
1142 	if (bp)
1143 		brelse(bp);
1144 	if (abp)
1145 		brelse(abp);
1146 	vn_lock(devvp, LK_EXCLUSIVE | LK_RETRY);
1147 	(void)VOP_CLOSE(devvp, ronly ? FREAD : FREAD|FWRITE, cred, p);
1148 	VOP_UNLOCK(devvp, 0);
1149 	if (ump) {
1150 		free(ump->um_lfs, M_UFSMNT);
1151 		free(ump, M_UFSMNT);
1152 		mp->mnt_data = (qaddr_t)0;
1153 	}
1154 	return (error);
1155 }
1156 
1157 /*
1158  * unmount system call
1159  */
1160 int
1161 lfs_unmount(struct mount *mp, int mntflags, struct proc *p)
1162 {
1163 	struct ufsmount *ump;
1164 	struct lfs *fs;
1165 	int error, flags, ronly, s;
1166 	extern int lfs_allclean_wakeup;
1167 
1168 	flags = 0;
1169 	if (mntflags & MNT_FORCE)
1170 		flags |= FORCECLOSE;
1171 
1172 	ump = VFSTOUFS(mp);
1173 	fs = ump->um_lfs;
1174 #ifdef QUOTA
1175 	if (mp->mnt_flag & MNT_QUOTA) {
1176 		int i;
1177 		error = vflush(mp, fs->lfs_ivnode, SKIPSYSTEM|flags);
1178 		if (error)
1179 			return (error);
1180 		for (i = 0; i < MAXQUOTAS; i++) {
1181 			if (ump->um_quotas[i] == NULLVP)
1182 				continue;
1183 			quotaoff(p, mp, i);
1184 		}
1185 		/*
1186 		 * Here we fall through to vflush again to ensure
1187 		 * that we have gotten rid of all the system vnodes.
1188 		 */
1189 	}
1190 #endif
1191 	if ((error = vflush(mp, fs->lfs_ivnode, flags)) != 0)
1192 		return (error);
1193 	if ((error = VFS_SYNC(mp, 1, p->p_ucred, p)) != 0)
1194 		return (error);
1195 	if (fs->lfs_ivnode->v_dirtyblkhd.lh_first)
1196 		panic("lfs_unmount: still dirty blocks on ifile vnode\n");
1197 
1198 	/* Explicitly write the superblock, to update serial and pflags */
1199 	fs->lfs_pflags |= LFS_PF_CLEAN;
1200 	lfs_writesuper(fs, fs->lfs_sboffs[0]);
1201 	lfs_writesuper(fs, fs->lfs_sboffs[1]);
1202 
1203 	/* Finish with the Ifile, now that we're done with it */
1204 	vrele(fs->lfs_ivnode);
1205 	vgone(fs->lfs_ivnode);
1206 
1207 	/* Wait for superblock writes to complete */
1208 	s = splbio();
1209 	while (fs->lfs_iocount)
1210 		tsleep(&fs->lfs_iocount, PRIBIO + 1, "lfs_umount", 0);
1211 	splx(s);
1212 
1213 	ronly = !fs->lfs_ronly;
1214 	if (ump->um_devvp->v_type != VBAD)
1215 		ump->um_devvp->v_specmountpoint = NULL;
1216 	vn_lock(ump->um_devvp, LK_EXCLUSIVE | LK_RETRY);
1217 	error = VOP_CLOSE(ump->um_devvp,
1218 	    ronly ? FREAD : FREAD|FWRITE, NOCRED, p);
1219 	vput(ump->um_devvp);
1220 
1221 	/* XXX KS - wake up the cleaner so it can die */
1222 	wakeup(&fs->lfs_nextseg);
1223 	wakeup(&lfs_allclean_wakeup);
1224 
1225 	free(fs, M_UFSMNT);
1226 	free(ump, M_UFSMNT);
1227 	mp->mnt_data = (qaddr_t)0;
1228 	mp->mnt_flag &= ~MNT_LOCAL;
1229 	return (error);
1230 }
1231 
1232 /*
1233  * Get file system statistics.
1234  */
1235 int
1236 lfs_statfs(struct mount *mp, struct statfs *sbp, struct proc *p)
1237 {
1238 	struct lfs *fs;
1239 	struct ufsmount *ump;
1240 
1241 	ump = VFSTOUFS(mp);
1242 	fs = ump->um_lfs;
1243 	if (fs->lfs_magic != LFS_MAGIC)
1244 		panic("lfs_statfs: magic");
1245 
1246 	sbp->f_type = 0;
1247 	sbp->f_bsize = fs->lfs_fsize;
1248 	sbp->f_iosize = fs->lfs_bsize;
1249 	sbp->f_blocks = fsbtofrags(fs, LFS_EST_NONMETA(fs));
1250 	sbp->f_bfree = fsbtofrags(fs, LFS_EST_BFREE(fs));
1251 	sbp->f_bavail = fsbtofrags(fs, (long)LFS_EST_BFREE(fs) -
1252 				  (long)LFS_EST_RSVD(fs));
1253 
1254 	sbp->f_files = fs->lfs_bfree / btofsb(fs, fs->lfs_ibsize) * INOPB(fs);
1255 	sbp->f_ffree = sbp->f_files - fs->lfs_nfiles;
1256 	if (sbp != &mp->mnt_stat) {
1257 		bcopy(mp->mnt_stat.f_mntonname, sbp->f_mntonname, MNAMELEN);
1258 		bcopy(mp->mnt_stat.f_mntfromname, sbp->f_mntfromname, MNAMELEN);
1259 	}
1260 	strncpy(sbp->f_fstypename, mp->mnt_op->vfs_name, MFSNAMELEN);
1261 	return (0);
1262 }
1263 
1264 /*
1265  * Go through the disk queues to initiate sandbagged IO;
1266  * go through the inodes to write those that have been modified;
1267  * initiate the writing of the super block if it has been modified.
1268  *
1269  * Note: we are always called with the filesystem marked `MPBUSY'.
1270  */
1271 int
1272 lfs_sync(struct mount *mp, int waitfor, struct ucred *cred, struct proc *p)
1273 {
1274 	int error;
1275 	struct lfs *fs;
1276 
1277 	fs = ((struct ufsmount *)mp->mnt_data)->ufsmount_u.lfs;
1278 	if (fs->lfs_ronly)
1279 		return 0;
1280 	while (fs->lfs_dirops)
1281 		error = tsleep(&fs->lfs_dirops, PRIBIO + 1, "lfs_dirops", 0);
1282 	fs->lfs_writer++;
1283 
1284 	/* All syncs must be checkpoints until roll-forward is implemented. */
1285 	error = lfs_segwrite(mp, SEGM_CKP | (waitfor ? SEGM_SYNC : 0));
1286 	if (--fs->lfs_writer == 0)
1287 		wakeup(&fs->lfs_dirops);
1288 #ifdef QUOTA
1289 	qsync(mp);
1290 #endif
1291 	return (error);
1292 }
1293 
1294 extern struct lock ufs_hashlock;
1295 
1296 /*
1297  * Look up an LFS dinode number to find its incore vnode.  If not already
1298  * in core, read it in from the specified device.  Return the inode locked.
1299  * Detection and handling of mount points must be done by the calling routine.
1300  */
1301 int
1302 lfs_vget(struct mount *mp, ino_t ino, struct vnode **vpp)
1303 {
1304 	struct lfs *fs;
1305 	struct inode *ip;
1306 	struct buf *bp;
1307 	struct ifile *ifp;
1308 	struct vnode *vp;
1309 	struct ufsmount *ump;
1310 	ufs_daddr_t daddr;
1311 	dev_t dev;
1312 	int error;
1313 	struct timespec ts;
1314 
1315 	ump = VFSTOUFS(mp);
1316 	dev = ump->um_dev;
1317 	fs = ump->um_lfs;
1318 
1319 	/*
1320 	 * If the filesystem is not completely mounted yet, suspend
1321 	 * any access requests (wait for roll-forward to complete).
1322 	 */
1323 	while ((fs->lfs_flags & LFS_NOTYET) && curproc->p_pid != fs->lfs_rfpid)
1324 		tsleep(&fs->lfs_flags, PRIBIO+1, "lfs_notyet", 0);
1325 
1326 	if ((*vpp = ufs_ihashget(dev, ino, LK_EXCLUSIVE)) != NULL)
1327 		return (0);
1328 
1329 	if ((error = getnewvnode(VT_LFS, mp, lfs_vnodeop_p, &vp)) != 0) {
1330 		*vpp = NULL;
1331 		 return (error);
1332 	}
1333 
1334 	do {
1335 		if ((*vpp = ufs_ihashget(dev, ino, LK_EXCLUSIVE)) != NULL) {
1336 			ungetnewvnode(vp);
1337 			return (0);
1338 		}
1339 	} while (lockmgr(&ufs_hashlock, LK_EXCLUSIVE|LK_SLEEPFAIL, 0));
1340 
1341 	/* Translate the inode number to a disk address. */
1342 	if (ino == LFS_IFILE_INUM)
1343 		daddr = fs->lfs_idaddr;
1344 	else {
1345 		/* XXX bounds-check this too */
1346 		LFS_IENTRY(ifp, fs, ino, bp);
1347 		daddr = ifp->if_daddr;
1348 		if (fs->lfs_version > 1) {
1349 			ts.tv_sec = ifp->if_atime_sec;
1350 			ts.tv_nsec = ifp->if_atime_nsec;
1351 		}
1352 
1353 		brelse(bp);
1354 		if (daddr == LFS_UNUSED_DADDR) {
1355 			*vpp = NULLVP;
1356 			ungetnewvnode(vp);
1357 			lockmgr(&ufs_hashlock, LK_RELEASE, 0);
1358 			return (ENOENT);
1359 		}
1360 	}
1361 
1362 	/* Allocate/init new vnode/inode. */
1363 	lfs_vcreate(mp, ino, vp);
1364 
1365 	/*
1366 	 * Put it onto its hash chain and lock it so that other requests for
1367 	 * this inode will block if they arrive while we are sleeping waiting
1368 	 * for old data structures to be purged or for the contents of the
1369 	 * disk portion of this inode to be read.
1370 	 */
1371 	ip = VTOI(vp);
1372 	ufs_ihashins(ip);
1373 	lockmgr(&ufs_hashlock, LK_RELEASE, 0);
1374 
1375 	/*
1376 	 * XXX
1377 	 * This may not need to be here, logically it should go down with
1378 	 * the i_devvp initialization.
1379 	 * Ask Kirk.
1380 	 */
1381 	ip->i_lfs = ump->um_lfs;
1382 
1383 	/* Read in the disk contents for the inode, copy into the inode. */
1384 	error = bread(ump->um_devvp, fsbtodb(fs, daddr),
1385 		(fs->lfs_version == 1 ? fs->lfs_bsize : fs->lfs_fsize),
1386 		NOCRED, &bp);
1387 	if (error) {
1388 		/*
1389 		 * The inode does not contain anything useful, so it would
1390 		 * be misleading to leave it on its hash chain. With mode
1391 		 * still zero, it will be unlinked and returned to the free
1392 		 * list by vput().
1393 		 */
1394 		vput(vp);
1395 		brelse(bp);
1396 		*vpp = NULL;
1397 		return (error);
1398 	}
1399 	ip->i_din.ffs_din = *lfs_ifind(fs, ino, bp);
1400 	ip->i_ffs_effnlink = ip->i_ffs_nlink;
1401 	ip->i_lfs_effnblks = ip->i_ffs_blocks;
1402 	if (fs->lfs_version > 1) {
1403 		ip->i_ffs_atime = ts.tv_sec;
1404 		ip->i_ffs_atimensec = ts.tv_nsec;
1405 	}
1406 	brelse(bp);
1407 
1408 	/*
1409 	 * Initialize the vnode from the inode, check for aliases.  In all
1410 	 * cases re-init ip, the underlying vnode/inode may have changed.
1411 	 */
1412 	ufs_vinit(mp, lfs_specop_p, lfs_fifoop_p, &vp);
1413 #ifdef DIAGNOSTIC
1414 	if (vp->v_type == VNON) {
1415 		panic("lfs_vget: ino %d is type VNON! (ifmt %o)\n",
1416 		       ip->i_number, (ip->i_ffs_mode & IFMT) >> 12);
1417 	}
1418 #endif
1419 	/*
1420 	 * Finish inode initialization now that aliasing has been resolved.
1421 	 */
1422 
1423 	genfs_node_init(vp, &lfs_genfsops);
1424 	ip->i_devvp = ump->um_devvp;
1425 	VREF(ip->i_devvp);
1426 	*vpp = vp;
1427 
1428 	uvm_vnp_setsize(vp, ip->i_ffs_size);
1429 
1430 	return (0);
1431 }
1432 
1433 /*
1434  * File handle to vnode
1435  *
1436  * Have to be really careful about stale file handles:
1437  * - check that the inode number is valid
1438  * - call lfs_vget() to get the locked inode
1439  * - check for an unallocated inode (i_mode == 0)
1440  *
1441  * XXX
1442  * use ifile to see if inode is allocated instead of reading off disk
1443  * what is the relationship between my generational number and the NFS
1444  * generational number.
1445  */
1446 int
1447 lfs_fhtovp(struct mount *mp, struct fid *fhp, struct vnode **vpp)
1448 {
1449 	struct ufid *ufhp;
1450 
1451 	ufhp = (struct ufid *)fhp;
1452 	if (ufhp->ufid_ino < ROOTINO)
1453 		return (ESTALE);
1454 	return (ufs_fhtovp(mp, ufhp, vpp));
1455 }
1456 
1457 /*
1458  * Vnode pointer to File handle
1459  */
1460 /* ARGSUSED */
1461 int
1462 lfs_vptofh(struct vnode *vp, struct fid *fhp)
1463 {
1464 	struct inode *ip;
1465 	struct ufid *ufhp;
1466 
1467 	ip = VTOI(vp);
1468 	ufhp = (struct ufid *)fhp;
1469 	ufhp->ufid_len = sizeof(struct ufid);
1470 	ufhp->ufid_ino = ip->i_number;
1471 	ufhp->ufid_gen = ip->i_ffs_gen;
1472 	return (0);
1473 }
1474 
1475 int
1476 lfs_sysctl(int *name, u_int namelen, void *oldp, size_t *oldlenp, void *newp, size_t newlen, struct proc *p)
1477 {
1478 	extern int lfs_writeindir, lfs_dostats, lfs_clean_vnhead;
1479 	extern struct lfs_stats lfs_stats;
1480 	int error;
1481 
1482 	/* all sysctl names at this level are terminal */
1483 	if (namelen != 1)
1484 		return (ENOTDIR);
1485 
1486 	switch (name[0]) {
1487 	case LFS_WRITEINDIR:
1488 		return (sysctl_int(oldp, oldlenp, newp, newlen,
1489 				   &lfs_writeindir));
1490 	case LFS_CLEAN_VNHEAD:
1491 		return (sysctl_int(oldp, oldlenp, newp, newlen,
1492 				   &lfs_clean_vnhead));
1493 	case LFS_DOSTATS:
1494 		if ((error = sysctl_int(oldp, oldlenp, newp, newlen,
1495 				       &lfs_dostats)))
1496 			return error;
1497 		if (lfs_dostats == 0)
1498 			memset(&lfs_stats,0,sizeof(lfs_stats));
1499 		return 0;
1500 	case LFS_STATS:
1501 		return (sysctl_rdstruct(oldp, oldlenp, newp,
1502 					&lfs_stats, sizeof(lfs_stats)));
1503 	default:
1504 		return (EOPNOTSUPP);
1505 	}
1506 	/* NOTREACHED */
1507 }
1508