xref: /freebsd/sys/fs/ext2fs/ext2_alloc.c (revision e0c4386e)
1 /*-
2  *  modified for Lites 1.1
3  *
4  *  Aug 1995, Godmar Back (gback@cs.utah.edu)
5  *  University of Utah, Department of Computer Science
6  */
7 /*-
8  * SPDX-License-Identifier: BSD-3-Clause
9  *
10  * Copyright (c) 1982, 1986, 1989, 1993
11  *	The Regents of the University of California.  All rights reserved.
12  *
13  * Redistribution and use in source and binary forms, with or without
14  * modification, are permitted provided that the following conditions
15  * are met:
16  * 1. Redistributions of source code must retain the above copyright
17  *    notice, this list of conditions and the following disclaimer.
18  * 2. Redistributions in binary form must reproduce the above copyright
19  *    notice, this list of conditions and the following disclaimer in the
20  *    documentation and/or other materials provided with the distribution.
21  * 3. Neither the name of the University nor the names of its contributors
22  *    may be used to endorse or promote products derived from this software
23  *    without specific prior written permission.
24  *
25  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
26  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
27  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
28  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
29  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
30  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
31  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
32  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
33  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
34  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
35  * SUCH DAMAGE.
36  */
37 
38 #include <sys/param.h>
39 #include <sys/systm.h>
40 #include <sys/conf.h>
41 #include <sys/vnode.h>
42 #include <sys/sdt.h>
43 #include <sys/stat.h>
44 #include <sys/mount.h>
45 #include <sys/sysctl.h>
46 #include <sys/syslog.h>
47 #include <sys/buf.h>
48 #include <sys/endian.h>
49 
50 #include <fs/ext2fs/fs.h>
51 #include <fs/ext2fs/inode.h>
52 #include <fs/ext2fs/ext2_mount.h>
53 #include <fs/ext2fs/ext2fs.h>
54 #include <fs/ext2fs/ext2_extern.h>
55 
56 SDT_PROVIDER_DEFINE(ext2fs);
57 /*
58  * ext2fs trace probe:
59  * arg0: verbosity. Higher numbers give more verbose messages
60  * arg1: Textual message
61  */
62 SDT_PROBE_DEFINE2(ext2fs, , alloc, trace, "int", "char*");
63 SDT_PROBE_DEFINE3(ext2fs, , alloc, ext2_reallocblks_realloc,
64     "ino_t", "e2fs_lbn_t", "e2fs_lbn_t");
65 SDT_PROBE_DEFINE1(ext2fs, , alloc, ext2_reallocblks_bap, "uint32_t");
66 SDT_PROBE_DEFINE1(ext2fs, , alloc, ext2_reallocblks_blkno, "e2fs_daddr_t");
67 SDT_PROBE_DEFINE2(ext2fs, , alloc, ext2_b_bitmap_validate_error, "char*", "int");
68 SDT_PROBE_DEFINE3(ext2fs, , alloc, ext2_nodealloccg_bmap_corrupted,
69     "int", "daddr_t", "char*");
70 SDT_PROBE_DEFINE2(ext2fs, , alloc, ext2_blkfree_bad_block, "ino_t", "e4fs_daddr_t");
71 SDT_PROBE_DEFINE2(ext2fs, , alloc, ext2_vfree_doublefree, "char*", "ino_t");
72 
73 static daddr_t	ext2_alloccg(struct inode *, int, daddr_t, int);
74 static daddr_t	ext2_clusteralloc(struct inode *, int, daddr_t, int);
75 static u_long	ext2_dirpref(struct inode *);
76 static e4fs_daddr_t ext2_hashalloc(struct inode *, int, long, int,
77     daddr_t (*)(struct inode *, int, daddr_t,
78 						int));
79 static daddr_t	ext2_nodealloccg(struct inode *, int, daddr_t, int);
80 static daddr_t  ext2_mapsearch(struct m_ext2fs *, char *, daddr_t);
81 
82 /*
83  * Allocate a block in the filesystem.
84  *
85  * A preference may be optionally specified. If a preference is given
86  * the following hierarchy is used to allocate a block:
87  *   1) allocate the requested block.
88  *   2) allocate a rotationally optimal block in the same cylinder.
89  *   3) allocate a block in the same cylinder group.
90  *   4) quadratically rehash into other cylinder groups, until an
91  *        available block is located.
92  * If no block preference is given the following hierarchy is used
93  * to allocate a block:
94  *   1) allocate a block in the cylinder group that contains the
95  *        inode for the file.
96  *   2) quadratically rehash into other cylinder groups, until an
97  *        available block is located.
98  */
99 int
100 ext2_alloc(struct inode *ip, daddr_t lbn, e4fs_daddr_t bpref, int size,
101     struct ucred *cred, e4fs_daddr_t *bnp)
102 {
103 	struct m_ext2fs *fs;
104 	struct ext2mount *ump;
105 	e4fs_daddr_t bno;
106 	int cg;
107 
108 	*bnp = 0;
109 	fs = ip->i_e2fs;
110 	ump = ip->i_ump;
111 	mtx_assert(EXT2_MTX(ump), MA_OWNED);
112 #ifdef INVARIANTS
113 	if ((u_int)size > fs->e2fs_bsize || blkoff(fs, size) != 0) {
114 		vn_printf(ip->i_devvp, "bsize = %lu, size = %d, fs = %s\n",
115 		    (long unsigned int)fs->e2fs_bsize, size, fs->e2fs_fsmnt);
116 		panic("ext2_alloc: bad size");
117 	}
118 	if (cred == NOCRED)
119 		panic("ext2_alloc: missing credential");
120 #endif		/* INVARIANTS */
121 	if (size == fs->e2fs_bsize && fs->e2fs_fbcount == 0)
122 		goto nospace;
123 	if (cred->cr_uid != 0 &&
124 	    fs->e2fs_fbcount < fs->e2fs_rbcount)
125 		goto nospace;
126 	if (bpref >= fs->e2fs_bcount)
127 		bpref = 0;
128 	if (bpref == 0)
129 		cg = ino_to_cg(fs, ip->i_number);
130 	else
131 		cg = dtog(fs, bpref);
132 	bno = (daddr_t)ext2_hashalloc(ip, cg, bpref, fs->e2fs_bsize,
133 	    ext2_alloccg);
134 	if (bno > 0) {
135 		/* set next_alloc fields as done in block_getblk */
136 		ip->i_next_alloc_block = lbn;
137 		ip->i_next_alloc_goal = bno;
138 
139 		ip->i_blocks += btodb(fs->e2fs_bsize);
140 		ip->i_flag |= IN_CHANGE | IN_UPDATE;
141 		*bnp = bno;
142 		return (0);
143 	}
144 nospace:
145 	EXT2_UNLOCK(ump);
146 	SDT_PROBE2(ext2fs, , alloc, trace, 1, "cannot allocate data block");
147 	return (ENOSPC);
148 }
149 
150 /*
151  * Allocate EA's block for inode.
152  */
153 e4fs_daddr_t
154 ext2_alloc_meta(struct inode *ip)
155 {
156 	struct m_ext2fs *fs;
157 	daddr_t blk;
158 
159 	fs = ip->i_e2fs;
160 
161 	EXT2_LOCK(ip->i_ump);
162 	blk = ext2_hashalloc(ip, ino_to_cg(fs, ip->i_number), 0, fs->e2fs_bsize,
163 	    ext2_alloccg);
164 	if (0 == blk) {
165 		EXT2_UNLOCK(ip->i_ump);
166 		SDT_PROBE2(ext2fs, , alloc, trace, 1, "cannot allocate meta block");
167 	}
168 
169 	return (blk);
170 }
171 
172 /*
173  * Reallocate a sequence of blocks into a contiguous sequence of blocks.
174  *
175  * The vnode and an array of buffer pointers for a range of sequential
176  * logical blocks to be made contiguous is given. The allocator attempts
177  * to find a range of sequential blocks starting as close as possible to
178  * an fs_rotdelay offset from the end of the allocation for the logical
179  * block immediately preceding the current range. If successful, the
180  * physical block numbers in the buffer pointers and in the inode are
181  * changed to reflect the new allocation. If unsuccessful, the allocation
182  * is left unchanged. The success in doing the reallocation is returned.
183  * Note that the error return is not reflected back to the user. Rather
184  * the previous block allocation will be used.
185  */
186 
187 static SYSCTL_NODE(_vfs, OID_AUTO, ext2fs, CTLFLAG_RW | CTLFLAG_MPSAFE, 0,
188     "EXT2FS filesystem");
189 
190 static int doasyncfree = 1;
191 
192 SYSCTL_INT(_vfs_ext2fs, OID_AUTO, doasyncfree, CTLFLAG_RW, &doasyncfree, 0,
193     "Use asynchronous writes to update block pointers when freeing blocks");
194 
195 static int doreallocblks = 0;
196 
197 SYSCTL_INT(_vfs_ext2fs, OID_AUTO, doreallocblks, CTLFLAG_RW, &doreallocblks, 0, "");
198 
199 int
200 ext2_reallocblks(struct vop_reallocblks_args *ap)
201 {
202 	struct m_ext2fs *fs;
203 	struct inode *ip;
204 	struct vnode *vp;
205 	struct buf *sbp, *ebp;
206 	uint32_t *bap, *sbap, *ebap;
207 	struct ext2mount *ump;
208 	struct cluster_save *buflist;
209 	struct indir start_ap[EXT2_NIADDR + 1], end_ap[EXT2_NIADDR + 1], *idp;
210 	e2fs_lbn_t start_lbn, end_lbn;
211 	int soff;
212 	e2fs_daddr_t newblk, blkno;
213 	int i, len, start_lvl, end_lvl, pref, ssize;
214 
215 	if (doreallocblks == 0)
216 		return (ENOSPC);
217 
218 	vp = ap->a_vp;
219 	ip = VTOI(vp);
220 	fs = ip->i_e2fs;
221 	ump = ip->i_ump;
222 
223 	if (fs->e2fs_contigsumsize <= 0 || ip->i_flag & IN_E4EXTENTS)
224 		return (ENOSPC);
225 
226 	buflist = ap->a_buflist;
227 	len = buflist->bs_nchildren;
228 	start_lbn = buflist->bs_children[0]->b_lblkno;
229 	end_lbn = start_lbn + len - 1;
230 #ifdef INVARIANTS
231 	for (i = 1; i < len; i++)
232 		if (buflist->bs_children[i]->b_lblkno != start_lbn + i)
233 			panic("ext2_reallocblks: non-cluster");
234 #endif
235 	/*
236 	 * If the cluster crosses the boundary for the first indirect
237 	 * block, leave space for the indirect block. Indirect blocks
238 	 * are initially laid out in a position after the last direct
239 	 * block. Block reallocation would usually destroy locality by
240 	 * moving the indirect block out of the way to make room for
241 	 * data blocks if we didn't compensate here. We should also do
242 	 * this for other indirect block boundaries, but it is only
243 	 * important for the first one.
244 	 */
245 	if (start_lbn < EXT2_NDADDR && end_lbn >= EXT2_NDADDR)
246 		return (ENOSPC);
247 	/*
248 	 * If the latest allocation is in a new cylinder group, assume that
249 	 * the filesystem has decided to move and do not force it back to
250 	 * the previous cylinder group.
251 	 */
252 	if (dtog(fs, dbtofsb(fs, buflist->bs_children[0]->b_blkno)) !=
253 	    dtog(fs, dbtofsb(fs, buflist->bs_children[len - 1]->b_blkno)))
254 		return (ENOSPC);
255 	if (ext2_getlbns(vp, start_lbn, start_ap, &start_lvl) ||
256 	    ext2_getlbns(vp, end_lbn, end_ap, &end_lvl))
257 		return (ENOSPC);
258 	/*
259 	 * Get the starting offset and block map for the first block.
260 	 */
261 	if (start_lvl == 0) {
262 		sbap = &ip->i_db[0];
263 		soff = start_lbn;
264 	} else {
265 		idp = &start_ap[start_lvl - 1];
266 		if (bread(vp, idp->in_lbn, (int)fs->e2fs_bsize, NOCRED, &sbp)) {
267 			brelse(sbp);
268 			return (ENOSPC);
269 		}
270 		sbap = (u_int *)sbp->b_data;
271 		soff = idp->in_off;
272 	}
273 	/*
274 	 * If the block range spans two block maps, get the second map.
275 	 */
276 	ebap = NULL;
277 	if (end_lvl == 0 || (idp = &end_ap[end_lvl - 1])->in_off + 1 >= len) {
278 		ssize = len;
279 	} else {
280 #ifdef INVARIANTS
281 		if (start_ap[start_lvl - 1].in_lbn == idp->in_lbn)
282 			panic("ext2_reallocblks: start == end");
283 #endif
284 		ssize = len - (idp->in_off + 1);
285 		if (bread(vp, idp->in_lbn, (int)fs->e2fs_bsize, NOCRED, &ebp))
286 			goto fail;
287 		ebap = (u_int *)ebp->b_data;
288 	}
289 	/*
290 	 * Find the preferred location for the cluster.
291 	 */
292 	EXT2_LOCK(ump);
293 	pref = ext2_blkpref(ip, start_lbn, soff, sbap, 0);
294 	/*
295 	 * Search the block map looking for an allocation of the desired size.
296 	 */
297 	if ((newblk = (e2fs_daddr_t)ext2_hashalloc(ip, dtog(fs, pref), pref,
298 	    len, ext2_clusteralloc)) == 0) {
299 		EXT2_UNLOCK(ump);
300 		goto fail;
301 	}
302 	/*
303 	 * We have found a new contiguous block.
304 	 *
305 	 * First we have to replace the old block pointers with the new
306 	 * block pointers in the inode and indirect blocks associated
307 	 * with the file.
308 	 */
309 	SDT_PROBE3(ext2fs, , alloc, ext2_reallocblks_realloc,
310 	    ip->i_number, start_lbn, end_lbn);
311 	blkno = newblk;
312 	for (bap = &sbap[soff], i = 0; i < len; i++, blkno += fs->e2fs_fpb) {
313 		if (i == ssize) {
314 			bap = ebap;
315 			soff = -i;
316 		}
317 #ifdef INVARIANTS
318 		if (buflist->bs_children[i]->b_blkno != fsbtodb(fs, *bap))
319 			panic("ext2_reallocblks: alloc mismatch");
320 #endif
321 		SDT_PROBE1(ext2fs, , alloc, ext2_reallocblks_bap, *bap);
322 		*bap++ = blkno;
323 	}
324 	/*
325 	 * Next we must write out the modified inode and indirect blocks.
326 	 * For strict correctness, the writes should be synchronous since
327 	 * the old block values may have been written to disk. In practise
328 	 * they are almost never written, but if we are concerned about
329 	 * strict correctness, the `doasyncfree' flag should be set to zero.
330 	 *
331 	 * The test on `doasyncfree' should be changed to test a flag
332 	 * that shows whether the associated buffers and inodes have
333 	 * been written. The flag should be set when the cluster is
334 	 * started and cleared whenever the buffer or inode is flushed.
335 	 * We can then check below to see if it is set, and do the
336 	 * synchronous write only when it has been cleared.
337 	 */
338 	if (sbap != &ip->i_db[0]) {
339 		if (doasyncfree)
340 			bdwrite(sbp);
341 		else
342 			bwrite(sbp);
343 	} else {
344 		ip->i_flag |= IN_CHANGE | IN_UPDATE;
345 		if (!doasyncfree)
346 			ext2_update(vp, 1);
347 	}
348 	if (ssize < len) {
349 		if (doasyncfree)
350 			bdwrite(ebp);
351 		else
352 			bwrite(ebp);
353 	}
354 	/*
355 	 * Last, free the old blocks and assign the new blocks to the buffers.
356 	 */
357 	for (blkno = newblk, i = 0; i < len; i++, blkno += fs->e2fs_fpb) {
358 		ext2_blkfree(ip, dbtofsb(fs, buflist->bs_children[i]->b_blkno),
359 		    fs->e2fs_bsize);
360 		buflist->bs_children[i]->b_blkno = fsbtodb(fs, blkno);
361 		SDT_PROBE1(ext2fs, , alloc, ext2_reallocblks_blkno, blkno);
362 	}
363 
364 	return (0);
365 
366 fail:
367 	if (ssize < len)
368 		brelse(ebp);
369 	if (sbap != &ip->i_db[0])
370 		brelse(sbp);
371 	return (ENOSPC);
372 }
373 
374 /*
375  * Allocate an inode in the filesystem.
376  *
377  */
378 int
379 ext2_valloc(struct vnode *pvp, int mode, struct ucred *cred, struct vnode **vpp)
380 {
381 	struct timespec ts;
382 	struct m_ext2fs *fs;
383 	struct ext2mount *ump;
384 	struct inode *pip;
385 	struct inode *ip;
386 	struct vnode *vp;
387 	struct thread *td;
388 	ino_t ino, ipref;
389 	int error, cg;
390 
391 	*vpp = NULL;
392 	pip = VTOI(pvp);
393 	fs = pip->i_e2fs;
394 	ump = pip->i_ump;
395 
396 	EXT2_LOCK(ump);
397 	if (fs->e2fs_ficount == 0)
398 		goto noinodes;
399 	/*
400 	 * If it is a directory then obtain a cylinder group based on
401 	 * ext2_dirpref else obtain it using ino_to_cg. The preferred inode is
402 	 * always the next inode.
403 	 */
404 	if ((mode & IFMT) == IFDIR) {
405 		cg = ext2_dirpref(pip);
406 		if (fs->e2fs_contigdirs[cg] < 255)
407 			fs->e2fs_contigdirs[cg]++;
408 	} else {
409 		cg = ino_to_cg(fs, pip->i_number);
410 		if (fs->e2fs_contigdirs[cg] > 0)
411 			fs->e2fs_contigdirs[cg]--;
412 	}
413 	ipref = cg * fs->e2fs_ipg + 1;
414 	ino = (ino_t)ext2_hashalloc(pip, cg, (long)ipref, mode, ext2_nodealloccg);
415 	if (ino == 0)
416 		goto noinodes;
417 
418 	td = curthread;
419 	error = vfs_hash_get(ump->um_mountp, ino, LK_EXCLUSIVE, td, vpp, NULL, NULL);
420 	if (error || *vpp != NULL) {
421 		return (error);
422 	}
423 
424 	ip = malloc(sizeof(struct inode), M_EXT2NODE, M_WAITOK | M_ZERO);
425 
426 	/* Allocate a new vnode/inode. */
427 	if ((error = getnewvnode("ext2fs", ump->um_mountp, &ext2_vnodeops, &vp)) != 0) {
428 		free(ip, M_EXT2NODE);
429 		return (error);
430 	}
431 
432 	lockmgr(vp->v_vnlock, LK_EXCLUSIVE, NULL);
433 	vp->v_data = ip;
434 	ip->i_vnode = vp;
435 	ip->i_e2fs = fs = ump->um_e2fs;
436 	ip->i_ump = ump;
437 	ip->i_number = ino;
438 	ip->i_block_group = ino_to_cg(fs, ino);
439 	ip->i_next_alloc_block = 0;
440 	ip->i_next_alloc_goal = 0;
441 
442 	error = insmntque(vp, ump->um_mountp);
443 	if (error) {
444 		free(ip, M_EXT2NODE);
445 		return (error);
446 	}
447 
448 	error = vfs_hash_insert(vp, ino, LK_EXCLUSIVE, td, vpp, NULL, NULL);
449 	if (error || *vpp != NULL) {
450 		*vpp = NULL;
451 		free(ip, M_EXT2NODE);
452 		return (error);
453 	}
454 
455 	if ((error = ext2_vinit(ump->um_mountp, &ext2_fifoops, &vp)) != 0) {
456 		vput(vp);
457 		*vpp = NULL;
458 		free(ip, M_EXT2NODE);
459 		return (error);
460 	}
461 
462 	if (EXT2_HAS_INCOMPAT_FEATURE(fs, EXT2F_INCOMPAT_EXTENTS)
463 	    && (S_ISREG(mode) || S_ISDIR(mode)))
464 		ext4_ext_tree_init(ip);
465 	else
466 		memset(ip->i_data, 0, sizeof(ip->i_data));
467 
468 	/*
469 	 * Set up a new generation number for this inode.
470 	 * Avoid zero values.
471 	 */
472 	do {
473 		ip->i_gen = arc4random();
474 	} while (ip->i_gen == 0);
475 
476 	vfs_timestamp(&ts);
477 	ip->i_birthtime = ts.tv_sec;
478 	ip->i_birthnsec = ts.tv_nsec;
479 
480 	vn_set_state(vp, VSTATE_CONSTRUCTED);
481 	*vpp = vp;
482 
483 	return (0);
484 
485 noinodes:
486 	EXT2_UNLOCK(ump);
487 	SDT_PROBE2(ext2fs, , alloc, trace, 1, "out of inodes");
488 	return (ENOSPC);
489 }
490 
491 /*
492  * 64-bit compatible getters and setters for struct ext2_gd from ext2fs.h
493  */
494 uint64_t
495 e2fs_gd_get_b_bitmap(struct ext2_gd *gd)
496 {
497 
498 	return (((uint64_t)(le32toh(gd->ext4bgd_b_bitmap_hi)) << 32) |
499 	    le32toh(gd->ext2bgd_b_bitmap));
500 }
501 
502 uint64_t
503 e2fs_gd_get_i_bitmap(struct ext2_gd *gd)
504 {
505 
506 	return (((uint64_t)(le32toh(gd->ext4bgd_i_bitmap_hi)) << 32) |
507 	    le32toh(gd->ext2bgd_i_bitmap));
508 }
509 
510 uint64_t
511 e2fs_gd_get_i_tables(struct ext2_gd *gd)
512 {
513 
514 	return (((uint64_t)(le32toh(gd->ext4bgd_i_tables_hi)) << 32) |
515 	    le32toh(gd->ext2bgd_i_tables));
516 }
517 
518 static uint32_t
519 e2fs_gd_get_nbfree(struct ext2_gd *gd)
520 {
521 
522 	return (((uint32_t)(le16toh(gd->ext4bgd_nbfree_hi)) << 16) |
523 	    le16toh(gd->ext2bgd_nbfree));
524 }
525 
526 static void
527 e2fs_gd_set_nbfree(struct ext2_gd *gd, uint32_t val)
528 {
529 
530 	gd->ext2bgd_nbfree = htole16(val & 0xffff);
531 	gd->ext4bgd_nbfree_hi = htole16(val >> 16);
532 }
533 
534 static uint32_t
535 e2fs_gd_get_nifree(struct ext2_gd *gd)
536 {
537 
538 	return (((uint32_t)(le16toh(gd->ext4bgd_nifree_hi)) << 16) |
539 	    le16toh(gd->ext2bgd_nifree));
540 }
541 
542 static void
543 e2fs_gd_set_nifree(struct ext2_gd *gd, uint32_t val)
544 {
545 
546 	gd->ext2bgd_nifree = htole16(val & 0xffff);
547 	gd->ext4bgd_nifree_hi = htole16(val >> 16);
548 }
549 
550 uint32_t
551 e2fs_gd_get_ndirs(struct ext2_gd *gd)
552 {
553 
554 	return (((uint32_t)(le16toh(gd->ext4bgd_ndirs_hi)) << 16) |
555 	    le16toh(gd->ext2bgd_ndirs));
556 }
557 
558 static void
559 e2fs_gd_set_ndirs(struct ext2_gd *gd, uint32_t val)
560 {
561 
562 	gd->ext2bgd_ndirs = htole16(val & 0xffff);
563 	gd->ext4bgd_ndirs_hi = htole16(val >> 16);
564 }
565 
566 static uint32_t
567 e2fs_gd_get_i_unused(struct ext2_gd *gd)
568 {
569 	return ((uint32_t)(le16toh(gd->ext4bgd_i_unused_hi) << 16) |
570 	    le16toh(gd->ext4bgd_i_unused));
571 }
572 
573 static void
574 e2fs_gd_set_i_unused(struct ext2_gd *gd, uint32_t val)
575 {
576 
577 	gd->ext4bgd_i_unused = htole16(val & 0xffff);
578 	gd->ext4bgd_i_unused_hi = htole16(val >> 16);
579 }
580 
581 /*
582  * Find a cylinder to place a directory.
583  *
584  * The policy implemented by this algorithm is to allocate a
585  * directory inode in the same cylinder group as its parent
586  * directory, but also to reserve space for its files inodes
587  * and data. Restrict the number of directories which may be
588  * allocated one after another in the same cylinder group
589  * without intervening allocation of files.
590  *
591  * If we allocate a first level directory then force allocation
592  * in another cylinder group.
593  *
594  */
595 static u_long
596 ext2_dirpref(struct inode *pip)
597 {
598 	struct m_ext2fs *fs;
599 	int cg, prefcg, cgsize;
600 	uint64_t avgbfree, minbfree;
601 	u_int avgifree, avgndir, curdirsize;
602 	u_int minifree, maxndir;
603 	u_int mincg, minndir;
604 	u_int dirsize, maxcontigdirs;
605 
606 	mtx_assert(EXT2_MTX(pip->i_ump), MA_OWNED);
607 	fs = pip->i_e2fs;
608 
609 	avgifree = fs->e2fs_ficount / fs->e2fs_gcount;
610 	avgbfree = fs->e2fs_fbcount / fs->e2fs_gcount;
611 	avgndir = fs->e2fs_total_dir / fs->e2fs_gcount;
612 
613 	/*
614 	 * Force allocation in another cg if creating a first level dir.
615 	 */
616 	ASSERT_VOP_LOCKED(ITOV(pip), "ext2fs_dirpref");
617 	if (ITOV(pip)->v_vflag & VV_ROOT) {
618 		prefcg = arc4random() % fs->e2fs_gcount;
619 		mincg = prefcg;
620 		minndir = fs->e2fs_ipg;
621 		for (cg = prefcg; cg < fs->e2fs_gcount; cg++)
622 			if (e2fs_gd_get_ndirs(&fs->e2fs_gd[cg]) < minndir &&
623 			    e2fs_gd_get_nifree(&fs->e2fs_gd[cg]) >= avgifree &&
624 			    e2fs_gd_get_nbfree(&fs->e2fs_gd[cg]) >= avgbfree) {
625 				mincg = cg;
626 				minndir = e2fs_gd_get_ndirs(&fs->e2fs_gd[cg]);
627 			}
628 		for (cg = 0; cg < prefcg; cg++)
629 			if (e2fs_gd_get_ndirs(&fs->e2fs_gd[cg]) < minndir &&
630 			    e2fs_gd_get_nifree(&fs->e2fs_gd[cg]) >= avgifree &&
631 			    e2fs_gd_get_nbfree(&fs->e2fs_gd[cg]) >= avgbfree) {
632 				mincg = cg;
633 				minndir = e2fs_gd_get_ndirs(&fs->e2fs_gd[cg]);
634 			}
635 		return (mincg);
636 	}
637 	/*
638 	 * Count various limits which used for
639 	 * optimal allocation of a directory inode.
640 	 */
641 	maxndir = min(avgndir + fs->e2fs_ipg / 16, fs->e2fs_ipg);
642 	minifree = avgifree - avgifree / 4;
643 	if (minifree < 1)
644 		minifree = 1;
645 	minbfree = avgbfree - avgbfree / 4;
646 	if (minbfree < 1)
647 		minbfree = 1;
648 	cgsize = fs->e2fs_fsize * fs->e2fs_fpg;
649 	dirsize = AVGDIRSIZE;
650 	curdirsize = avgndir ?
651 	    (cgsize - avgbfree * fs->e2fs_bsize) / avgndir : 0;
652 	if (dirsize < curdirsize)
653 		dirsize = curdirsize;
654 	maxcontigdirs = min((avgbfree * fs->e2fs_bsize) / dirsize, 255);
655 	maxcontigdirs = min(maxcontigdirs, fs->e2fs_ipg / AFPDIR);
656 	if (maxcontigdirs == 0)
657 		maxcontigdirs = 1;
658 
659 	/*
660 	 * Limit number of dirs in one cg and reserve space for
661 	 * regular files, but only if we have no deficit in
662 	 * inodes or space.
663 	 */
664 	prefcg = ino_to_cg(fs, pip->i_number);
665 	for (cg = prefcg; cg < fs->e2fs_gcount; cg++)
666 		if (e2fs_gd_get_ndirs(&fs->e2fs_gd[cg]) < maxndir &&
667 		    e2fs_gd_get_nifree(&fs->e2fs_gd[cg]) >= minifree &&
668 		    e2fs_gd_get_nbfree(&fs->e2fs_gd[cg]) >= minbfree) {
669 			if (fs->e2fs_contigdirs[cg] < maxcontigdirs)
670 				return (cg);
671 		}
672 	for (cg = 0; cg < prefcg; cg++)
673 		if (e2fs_gd_get_ndirs(&fs->e2fs_gd[cg]) < maxndir &&
674 		    e2fs_gd_get_nifree(&fs->e2fs_gd[cg]) >= minifree &&
675 		    e2fs_gd_get_nbfree(&fs->e2fs_gd[cg]) >= minbfree) {
676 			if (fs->e2fs_contigdirs[cg] < maxcontigdirs)
677 				return (cg);
678 		}
679 	/*
680 	 * This is a backstop when we have deficit in space.
681 	 */
682 	for (cg = prefcg; cg < fs->e2fs_gcount; cg++)
683 		if (e2fs_gd_get_nifree(&fs->e2fs_gd[cg]) >= avgifree)
684 			return (cg);
685 	for (cg = 0; cg < prefcg; cg++)
686 		if (e2fs_gd_get_nifree(&fs->e2fs_gd[cg]) >= avgifree)
687 			break;
688 	return (cg);
689 }
690 
691 /*
692  * Select the desired position for the next block in a file.
693  *
694  * we try to mimic what Remy does in inode_getblk/block_getblk
695  *
696  * we note: blocknr == 0 means that we're about to allocate either
697  * a direct block or a pointer block at the first level of indirection
698  * (In other words, stuff that will go in i_db[] or i_ib[])
699  *
700  * blocknr != 0 means that we're allocating a block that is none
701  * of the above. Then, blocknr tells us the number of the block
702  * that will hold the pointer
703  */
704 e4fs_daddr_t
705 ext2_blkpref(struct inode *ip, e2fs_lbn_t lbn, int indx, e2fs_daddr_t *bap,
706     e2fs_daddr_t blocknr)
707 {
708 	struct m_ext2fs *fs;
709 	int tmp;
710 
711 	fs = ip->i_e2fs;
712 
713 	mtx_assert(EXT2_MTX(ip->i_ump), MA_OWNED);
714 
715 	/*
716 	 * If the next block is actually what we thought it is, then set the
717 	 * goal to what we thought it should be.
718 	 */
719 	if (ip->i_next_alloc_block == lbn && ip->i_next_alloc_goal != 0)
720 		return ip->i_next_alloc_goal;
721 
722 	/*
723 	 * Now check whether we were provided with an array that basically
724 	 * tells us previous blocks to which we want to stay close.
725 	 */
726 	if (bap)
727 		for (tmp = indx - 1; tmp >= 0; tmp--)
728 			if (bap[tmp])
729 				return (le32toh(bap[tmp]));
730 
731 	/*
732 	 * Else lets fall back to the blocknr or, if there is none, follow
733 	 * the rule that a block should be allocated near its inode.
734 	 */
735 	return (blocknr ? blocknr :
736 	    (e2fs_daddr_t)(ip->i_block_group *
737 	    EXT2_BLOCKS_PER_GROUP(fs)) + le32toh(fs->e2fs->e2fs_first_dblock));
738 }
739 
740 /*
741  * Implement the cylinder overflow algorithm.
742  *
743  * The policy implemented by this algorithm is:
744  *   1) allocate the block in its requested cylinder group.
745  *   2) quadratically rehash on the cylinder group number.
746  *   3) brute force search for a free block.
747  */
748 static e4fs_daddr_t
749 ext2_hashalloc(struct inode *ip, int cg, long pref, int size,
750     daddr_t (*allocator) (struct inode *, int, daddr_t, int))
751 {
752 	struct m_ext2fs *fs;
753 	e4fs_daddr_t result;
754 	int i, icg = cg;
755 
756 	mtx_assert(EXT2_MTX(ip->i_ump), MA_OWNED);
757 	fs = ip->i_e2fs;
758 	/*
759 	 * 1: preferred cylinder group
760 	 */
761 	result = (*allocator)(ip, cg, pref, size);
762 	if (result)
763 		return (result);
764 	/*
765 	 * 2: quadratic rehash
766 	 */
767 	for (i = 1; i < fs->e2fs_gcount; i *= 2) {
768 		cg += i;
769 		if (cg >= fs->e2fs_gcount)
770 			cg -= fs->e2fs_gcount;
771 		result = (*allocator)(ip, cg, 0, size);
772 		if (result)
773 			return (result);
774 	}
775 	/*
776 	 * 3: brute force search
777 	 * Note that we start at i == 2, since 0 was checked initially,
778 	 * and 1 is always checked in the quadratic rehash.
779 	 */
780 	cg = (icg + 2) % fs->e2fs_gcount;
781 	for (i = 2; i < fs->e2fs_gcount; i++) {
782 		result = (*allocator)(ip, cg, 0, size);
783 		if (result)
784 			return (result);
785 		cg++;
786 		if (cg == fs->e2fs_gcount)
787 			cg = 0;
788 	}
789 	return (0);
790 }
791 
792 static uint64_t
793 ext2_cg_number_gdb_nometa(struct m_ext2fs *fs, int cg)
794 {
795 
796 	if (!ext2_cg_has_sb(fs, cg))
797 		return (0);
798 
799 	if (EXT2_HAS_INCOMPAT_FEATURE(fs, EXT2F_INCOMPAT_META_BG))
800 		return (le32toh(fs->e2fs->e3fs_first_meta_bg));
801 
802 	return ((fs->e2fs_gcount + EXT2_DESCS_PER_BLOCK(fs) - 1) /
803 	    EXT2_DESCS_PER_BLOCK(fs));
804 }
805 
806 static uint64_t
807 ext2_cg_number_gdb_meta(struct m_ext2fs *fs, int cg)
808 {
809 	unsigned long metagroup;
810 	int first, last;
811 
812 	metagroup = cg / EXT2_DESCS_PER_BLOCK(fs);
813 	first = metagroup * EXT2_DESCS_PER_BLOCK(fs);
814 	last = first + EXT2_DESCS_PER_BLOCK(fs) - 1;
815 
816 	if (cg == first || cg == first + 1 || cg == last)
817 		return (1);
818 
819 	return (0);
820 }
821 
822 uint64_t
823 ext2_cg_number_gdb(struct m_ext2fs *fs, int cg)
824 {
825 	unsigned long first_meta_bg, metagroup;
826 
827 	first_meta_bg = le32toh(fs->e2fs->e3fs_first_meta_bg);
828 	metagroup = cg / EXT2_DESCS_PER_BLOCK(fs);
829 
830 	if (!EXT2_HAS_INCOMPAT_FEATURE(fs, EXT2F_INCOMPAT_META_BG) ||
831 	    metagroup < first_meta_bg)
832 		return (ext2_cg_number_gdb_nometa(fs, cg));
833 
834 	return ext2_cg_number_gdb_meta(fs, cg);
835 }
836 
837 static int
838 ext2_number_base_meta_blocks(struct m_ext2fs *fs, int cg)
839 {
840 	int number;
841 
842 	number = ext2_cg_has_sb(fs, cg);
843 
844 	if (!EXT2_HAS_INCOMPAT_FEATURE(fs, EXT2F_INCOMPAT_META_BG) ||
845 	    cg < le32toh(fs->e2fs->e3fs_first_meta_bg) *
846 	    EXT2_DESCS_PER_BLOCK(fs)) {
847 		if (number) {
848 			number += ext2_cg_number_gdb(fs, cg);
849 			number += le16toh(fs->e2fs->e2fs_reserved_ngdb);
850 		}
851 	} else {
852 		number += ext2_cg_number_gdb(fs, cg);
853 	}
854 
855 	return (number);
856 }
857 
858 static void
859 ext2_mark_bitmap_end(int start_bit, int end_bit, char *bitmap)
860 {
861 	int i;
862 
863 	if (start_bit >= end_bit)
864 		return;
865 
866 	for (i = start_bit; i < ((start_bit + 7) & ~7UL); i++)
867 		setbit(bitmap, i);
868 	if (i < end_bit)
869 		memset(bitmap + (i >> 3), 0xff, (end_bit - i) >> 3);
870 }
871 
872 static int
873 ext2_get_group_number(struct m_ext2fs *fs, e4fs_daddr_t block)
874 {
875 
876 	return ((block - le32toh(fs->e2fs->e2fs_first_dblock)) /
877 	    fs->e2fs_bsize);
878 }
879 
880 static int
881 ext2_block_in_group(struct m_ext2fs *fs, e4fs_daddr_t block, int cg)
882 {
883 
884 	return ((ext2_get_group_number(fs, block) == cg) ? 1 : 0);
885 }
886 
887 static int
888 ext2_cg_block_bitmap_init(struct m_ext2fs *fs, int cg, struct buf *bp)
889 {
890 	int bit, bit_max, inodes_per_block;
891 	uint64_t start, tmp;
892 
893 	if (!(le16toh(fs->e2fs_gd[cg].ext4bgd_flags) & EXT2_BG_BLOCK_UNINIT))
894 		return (0);
895 
896 	memset(bp->b_data, 0, fs->e2fs_bsize);
897 
898 	bit_max = ext2_number_base_meta_blocks(fs, cg);
899 	if ((bit_max >> 3) >= fs->e2fs_bsize)
900 		return (EINVAL);
901 
902 	for (bit = 0; bit < bit_max; bit++)
903 		setbit(bp->b_data, bit);
904 
905 	start = (uint64_t)cg * fs->e2fs_bpg +
906 	    le32toh(fs->e2fs->e2fs_first_dblock);
907 
908 	/* Set bits for block and inode bitmaps, and inode table. */
909 	tmp = e2fs_gd_get_b_bitmap(&fs->e2fs_gd[cg]);
910 	if (!EXT2_HAS_INCOMPAT_FEATURE(fs, EXT2F_INCOMPAT_FLEX_BG) ||
911 	    ext2_block_in_group(fs, tmp, cg))
912 		setbit(bp->b_data, tmp - start);
913 
914 	tmp = e2fs_gd_get_i_bitmap(&fs->e2fs_gd[cg]);
915 	if (!EXT2_HAS_INCOMPAT_FEATURE(fs, EXT2F_INCOMPAT_FLEX_BG) ||
916 	    ext2_block_in_group(fs, tmp, cg))
917 		setbit(bp->b_data, tmp - start);
918 
919 	tmp = e2fs_gd_get_i_tables(&fs->e2fs_gd[cg]);
920 	inodes_per_block = fs->e2fs_bsize/EXT2_INODE_SIZE(fs);
921 	while( tmp < e2fs_gd_get_i_tables(&fs->e2fs_gd[cg]) +
922 	    fs->e2fs_ipg / inodes_per_block ) {
923 		if (!EXT2_HAS_INCOMPAT_FEATURE(fs, EXT2F_INCOMPAT_FLEX_BG) ||
924 		    ext2_block_in_group(fs, tmp, cg))
925 			setbit(bp->b_data, tmp - start);
926 		tmp++;
927 	}
928 
929 	/*
930 	 * Also if the number of blocks within the group is less than
931 	 * the blocksize * 8 ( which is the size of bitmap ), set rest
932 	 * of the block bitmap to 1
933 	 */
934 	ext2_mark_bitmap_end(fs->e2fs_bpg, fs->e2fs_bsize * 8,
935 	    bp->b_data);
936 
937 	/* Clean the flag */
938 	fs->e2fs_gd[cg].ext4bgd_flags = htole16(le16toh(
939 	    fs->e2fs_gd[cg].ext4bgd_flags) & ~EXT2_BG_BLOCK_UNINIT);
940 
941 	return (0);
942 }
943 
944 static int
945 ext2_b_bitmap_validate(struct m_ext2fs *fs, struct buf *bp, int cg)
946 {
947 	struct ext2_gd *gd;
948 	uint64_t group_first_block;
949 	unsigned int offset, max_bit;
950 
951 	if (EXT2_HAS_INCOMPAT_FEATURE(fs, EXT2F_INCOMPAT_FLEX_BG)) {
952 		/*
953 		 * It is not possible to check block bitmap in case of this
954 		 * feature, because the inode and block bitmaps and inode table
955 		 * blocks may not be in the group at all.
956 		 * So, skip check in this case.
957 		 */
958 		return (0);
959 	}
960 
961 	gd = &fs->e2fs_gd[cg];
962 	max_bit = fs->e2fs_fpg;
963 	group_first_block = ((uint64_t)cg) * fs->e2fs_fpg +
964 	    le32toh(fs->e2fs->e2fs_first_dblock);
965 
966 	/* Check block bitmap block number */
967 	offset = e2fs_gd_get_b_bitmap(gd) - group_first_block;
968 	if (offset >= max_bit || !isset(bp->b_data, offset)) {
969 		SDT_PROBE2(ext2fs, , alloc, ext2_b_bitmap_validate_error,
970 		    "bad block bitmap, group", cg);
971 		return (EINVAL);
972 	}
973 
974 	/* Check inode bitmap block number */
975 	offset = e2fs_gd_get_i_bitmap(gd) - group_first_block;
976 	if (offset >= max_bit || !isset(bp->b_data, offset)) {
977 		SDT_PROBE2(ext2fs, , alloc, ext2_b_bitmap_validate_error,
978 		    "bad inode bitmap", cg);
979 		return (EINVAL);
980 	}
981 
982 	/* Check inode table */
983 	offset = e2fs_gd_get_i_tables(gd) - group_first_block;
984 	if (offset >= max_bit || offset + fs->e2fs_itpg >= max_bit) {
985 		SDT_PROBE2(ext2fs, , alloc, ext2_b_bitmap_validate_error,
986 		    "bad inode table, group", cg);
987 		return (EINVAL);
988 	}
989 
990 	return (0);
991 }
992 
993 /*
994  * Determine whether a block can be allocated.
995  *
996  * Check to see if a block of the appropriate size is available,
997  * and if it is, allocate it.
998  */
999 static daddr_t
1000 ext2_alloccg(struct inode *ip, int cg, daddr_t bpref, int size)
1001 {
1002 	struct m_ext2fs *fs;
1003 	struct buf *bp;
1004 	struct ext2mount *ump;
1005 	daddr_t bno, runstart, runlen;
1006 	int bit, loc, end, error, start;
1007 	char *bbp;
1008 	/* XXX ondisk32 */
1009 	fs = ip->i_e2fs;
1010 	ump = ip->i_ump;
1011 	if (e2fs_gd_get_nbfree(&fs->e2fs_gd[cg]) == 0)
1012 		return (0);
1013 
1014 	EXT2_UNLOCK(ump);
1015 	error = bread(ip->i_devvp, fsbtodb(fs,
1016 	    e2fs_gd_get_b_bitmap(&fs->e2fs_gd[cg])),
1017 	    (int)fs->e2fs_bsize, NOCRED, &bp);
1018 	if (error)
1019 		goto fail;
1020 
1021 	if (EXT2_HAS_RO_COMPAT_FEATURE(fs, EXT2F_ROCOMPAT_GDT_CSUM) ||
1022 	    EXT2_HAS_RO_COMPAT_FEATURE(fs, EXT2F_ROCOMPAT_METADATA_CKSUM)) {
1023 		error = ext2_cg_block_bitmap_init(fs, cg, bp);
1024 		if (error)
1025 			goto fail;
1026 
1027 		ext2_gd_b_bitmap_csum_set(fs, cg, bp);
1028 	}
1029 	error = ext2_gd_b_bitmap_csum_verify(fs, cg, bp);
1030 	if (error)
1031 		goto fail;
1032 
1033 	error = ext2_b_bitmap_validate(fs,bp, cg);
1034 	if (error)
1035 		goto fail;
1036 
1037 	/*
1038 	 * Check, that another thread did not not allocate the last block in
1039 	 * this group while we were waiting for the buffer.
1040 	 */
1041 	if (e2fs_gd_get_nbfree(&fs->e2fs_gd[cg]) == 0)
1042 		goto fail;
1043 
1044 	bbp = (char *)bp->b_data;
1045 
1046 	if (dtog(fs, bpref) != cg)
1047 		bpref = 0;
1048 	if (bpref != 0) {
1049 		bpref = dtogd(fs, bpref);
1050 		/*
1051 		 * if the requested block is available, use it
1052 		 */
1053 		if (isclr(bbp, bpref)) {
1054 			bno = bpref;
1055 			goto gotit;
1056 		}
1057 	}
1058 	/*
1059 	 * no blocks in the requested cylinder, so take next
1060 	 * available one in this cylinder group.
1061 	 * first try to get 8 contigous blocks, then fall back to a single
1062 	 * block.
1063 	 */
1064 	if (bpref)
1065 		start = dtogd(fs, bpref) / NBBY;
1066 	else
1067 		start = 0;
1068 	end = howmany(fs->e2fs_fpg, NBBY);
1069 retry:
1070 	runlen = 0;
1071 	runstart = 0;
1072 	for (loc = start; loc < end; loc++) {
1073 		if (bbp[loc] == (char)0xff) {
1074 			runlen = 0;
1075 			continue;
1076 		}
1077 
1078 		/* Start of a run, find the number of high clear bits. */
1079 		if (runlen == 0) {
1080 			bit = fls(bbp[loc]);
1081 			runlen = NBBY - bit;
1082 			runstart = loc * NBBY + bit;
1083 		} else if (bbp[loc] == 0) {
1084 			/* Continue a run. */
1085 			runlen += NBBY;
1086 		} else {
1087 			/*
1088 			 * Finish the current run.  If it isn't long
1089 			 * enough, start a new one.
1090 			 */
1091 			bit = ffs(bbp[loc]) - 1;
1092 			runlen += bit;
1093 			if (runlen >= 8) {
1094 				bno = runstart;
1095 				goto gotit;
1096 			}
1097 
1098 			/* Run was too short, start a new one. */
1099 			bit = fls(bbp[loc]);
1100 			runlen = NBBY - bit;
1101 			runstart = loc * NBBY + bit;
1102 		}
1103 
1104 		/* If the current run is long enough, use it. */
1105 		if (runlen >= 8) {
1106 			bno = runstart;
1107 			goto gotit;
1108 		}
1109 	}
1110 	if (start != 0) {
1111 		end = start;
1112 		start = 0;
1113 		goto retry;
1114 	}
1115 	bno = ext2_mapsearch(fs, bbp, bpref);
1116 	if (bno < 0)
1117 		goto fail;
1118 
1119 gotit:
1120 #ifdef INVARIANTS
1121 	if (isset(bbp, bno)) {
1122 		printf("ext2fs_alloccgblk: cg=%d bno=%jd fs=%s\n",
1123 		    cg, (intmax_t)bno, fs->e2fs_fsmnt);
1124 		panic("ext2fs_alloccg: dup alloc");
1125 	}
1126 #endif
1127 	setbit(bbp, bno);
1128 	EXT2_LOCK(ump);
1129 	ext2_clusteracct(fs, bbp, cg, bno, -1);
1130 	fs->e2fs_fbcount--;
1131 	e2fs_gd_set_nbfree(&fs->e2fs_gd[cg],
1132 	    e2fs_gd_get_nbfree(&fs->e2fs_gd[cg]) - 1);
1133 	fs->e2fs_fmod = 1;
1134 	EXT2_UNLOCK(ump);
1135 	ext2_gd_b_bitmap_csum_set(fs, cg, bp);
1136 	bdwrite(bp);
1137 	return (((uint64_t)cg) * fs->e2fs_fpg +
1138 	    le32toh(fs->e2fs->e2fs_first_dblock) + bno);
1139 
1140 fail:
1141 	brelse(bp);
1142 	EXT2_LOCK(ump);
1143 	return (0);
1144 }
1145 
1146 /*
1147  * Determine whether a cluster can be allocated.
1148  */
1149 static daddr_t
1150 ext2_clusteralloc(struct inode *ip, int cg, daddr_t bpref, int len)
1151 {
1152 	struct m_ext2fs *fs;
1153 	struct ext2mount *ump;
1154 	struct buf *bp;
1155 	char *bbp;
1156 	int bit, error, got, i, loc, run;
1157 	int32_t *lp;
1158 	daddr_t bno;
1159 
1160 	fs = ip->i_e2fs;
1161 	ump = ip->i_ump;
1162 
1163 	if (fs->e2fs_maxcluster[cg] < len)
1164 		return (0);
1165 
1166 	EXT2_UNLOCK(ump);
1167 	error = bread(ip->i_devvp,
1168 	    fsbtodb(fs, e2fs_gd_get_b_bitmap(&fs->e2fs_gd[cg])),
1169 	    (int)fs->e2fs_bsize, NOCRED, &bp);
1170 	if (error)
1171 		goto fail_lock;
1172 
1173 	bbp = (char *)bp->b_data;
1174 	EXT2_LOCK(ump);
1175 	/*
1176 	 * Check to see if a cluster of the needed size (or bigger) is
1177 	 * available in this cylinder group.
1178 	 */
1179 	lp = &fs->e2fs_clustersum[cg].cs_sum[len];
1180 	for (i = len; i <= fs->e2fs_contigsumsize; i++)
1181 		if (*lp++ > 0)
1182 			break;
1183 	if (i > fs->e2fs_contigsumsize) {
1184 		/*
1185 		 * Update the cluster summary information to reflect
1186 		 * the true maximum-sized cluster so that future cluster
1187 		 * allocation requests can avoid reading the bitmap only
1188 		 * to find no cluster.
1189 		 */
1190 		lp = &fs->e2fs_clustersum[cg].cs_sum[len - 1];
1191 		for (i = len - 1; i > 0; i--)
1192 			if (*lp-- > 0)
1193 				break;
1194 		fs->e2fs_maxcluster[cg] = i;
1195 		goto fail;
1196 	}
1197 	EXT2_UNLOCK(ump);
1198 
1199 	/* Search the bitmap to find a big enough cluster like in FFS. */
1200 	if (dtog(fs, bpref) != cg)
1201 		bpref = 0;
1202 	if (bpref != 0)
1203 		bpref = dtogd(fs, bpref);
1204 	loc = bpref / NBBY;
1205 	bit = 1 << (bpref % NBBY);
1206 	for (run = 0, got = bpref; got < fs->e2fs_fpg; got++) {
1207 		if ((bbp[loc] & bit) != 0)
1208 			run = 0;
1209 		else {
1210 			run++;
1211 			if (run == len)
1212 				break;
1213 		}
1214 		if ((got & (NBBY - 1)) != (NBBY - 1))
1215 			bit <<= 1;
1216 		else {
1217 			loc++;
1218 			bit = 1;
1219 		}
1220 	}
1221 
1222 	if (got >= fs->e2fs_fpg)
1223 		goto fail_lock;
1224 
1225 	/* Allocate the cluster that we found. */
1226 	for (i = 1; i < len; i++)
1227 		if (!isclr(bbp, got - run + i))
1228 			panic("ext2_clusteralloc: map mismatch");
1229 
1230 	bno = got - run + 1;
1231 	if (bno >= fs->e2fs_fpg)
1232 		panic("ext2_clusteralloc: allocated out of group");
1233 
1234 	EXT2_LOCK(ump);
1235 	for (i = 0; i < len; i += fs->e2fs_fpb) {
1236 		setbit(bbp, bno + i);
1237 		ext2_clusteracct(fs, bbp, cg, bno + i, -1);
1238 		fs->e2fs_fbcount--;
1239 		e2fs_gd_set_nbfree(&fs->e2fs_gd[cg],
1240 		    e2fs_gd_get_nbfree(&fs->e2fs_gd[cg]) - 1);
1241 	}
1242 	fs->e2fs_fmod = 1;
1243 	EXT2_UNLOCK(ump);
1244 
1245 	bdwrite(bp);
1246 	return (cg * fs->e2fs_fpg + le32toh(fs->e2fs->e2fs_first_dblock)
1247 	    + bno);
1248 
1249 fail_lock:
1250 	EXT2_LOCK(ump);
1251 fail:
1252 	brelse(bp);
1253 	return (0);
1254 }
1255 
1256 static int
1257 ext2_zero_inode_table(struct inode *ip, int cg)
1258 {
1259 	struct m_ext2fs *fs;
1260 	struct buf *bp;
1261 	int i, all_blks, used_blks;
1262 
1263 	fs = ip->i_e2fs;
1264 
1265 	if (le16toh(fs->e2fs_gd[cg].ext4bgd_flags) & EXT2_BG_INODE_ZEROED)
1266 		return (0);
1267 
1268 	all_blks = le16toh(fs->e2fs->e2fs_inode_size) * fs->e2fs_ipg /
1269 	    fs->e2fs_bsize;
1270 
1271 	used_blks = howmany(fs->e2fs_ipg -
1272 	    e2fs_gd_get_i_unused(&fs->e2fs_gd[cg]),
1273 	    fs->e2fs_bsize / EXT2_INODE_SIZE(fs));
1274 
1275 	for (i = 0; i < all_blks - used_blks; i++) {
1276 		bp = getblk(ip->i_devvp, fsbtodb(fs,
1277 		    e2fs_gd_get_i_tables(&fs->e2fs_gd[cg]) + used_blks + i),
1278 		    fs->e2fs_bsize, 0, 0, 0);
1279 		if (!bp)
1280 			return (EIO);
1281 
1282 		vfs_bio_bzero_buf(bp, 0, fs->e2fs_bsize);
1283 		bawrite(bp);
1284 	}
1285 
1286 	fs->e2fs_gd[cg].ext4bgd_flags = htole16(le16toh(
1287 	    fs->e2fs_gd[cg].ext4bgd_flags) | EXT2_BG_INODE_ZEROED);
1288 
1289 	return (0);
1290 }
1291 
1292 static void
1293 ext2_fix_bitmap_tail(unsigned char *bitmap, int first, int last)
1294 {
1295 	int i;
1296 
1297 	for (i = first; i <= last; i++)
1298 		bitmap[i] = 0xff;
1299 }
1300 
1301 /*
1302  * Determine whether an inode can be allocated.
1303  *
1304  * Check to see if an inode is available, and if it is,
1305  * allocate it using tode in the specified cylinder group.
1306  */
1307 static daddr_t
1308 ext2_nodealloccg(struct inode *ip, int cg, daddr_t ipref, int mode)
1309 {
1310 	struct m_ext2fs *fs;
1311 	struct buf *bp;
1312 	struct ext2mount *ump;
1313 	int error, start, len, ifree, ibytes;
1314 	char *ibp, *loc;
1315 
1316 	ipref--;	/* to avoid a lot of (ipref -1) */
1317 	if (ipref == -1)
1318 		ipref = 0;
1319 	fs = ip->i_e2fs;
1320 	ump = ip->i_ump;
1321 	if (e2fs_gd_get_nifree(&fs->e2fs_gd[cg]) == 0)
1322 		return (0);
1323 	EXT2_UNLOCK(ump);
1324 	error = bread(ip->i_devvp, fsbtodb(fs,
1325 	    e2fs_gd_get_i_bitmap(&fs->e2fs_gd[cg])),
1326 	    (int)fs->e2fs_bsize, NOCRED, &bp);
1327 	if (error) {
1328 		EXT2_LOCK(ump);
1329 		return (0);
1330 	}
1331 	if (EXT2_HAS_RO_COMPAT_FEATURE(fs, EXT2F_ROCOMPAT_GDT_CSUM) ||
1332 	    EXT2_HAS_RO_COMPAT_FEATURE(fs, EXT2F_ROCOMPAT_METADATA_CKSUM)) {
1333 		if (le16toh(fs->e2fs_gd[cg].ext4bgd_flags) &
1334 		    EXT2_BG_INODE_UNINIT) {
1335 			ibytes = fs->e2fs_ipg / 8;
1336 			memset(bp->b_data, 0, ibytes - 1);
1337 			ext2_fix_bitmap_tail(bp->b_data, ibytes,
1338 			    fs->e2fs_bsize - 1);
1339 			fs->e2fs_gd[cg].ext4bgd_flags = htole16(le16toh(
1340 			    fs->e2fs_gd[cg].ext4bgd_flags) &
1341 			    ~EXT2_BG_INODE_UNINIT);
1342 		}
1343 		ext2_gd_i_bitmap_csum_set(fs, cg, bp);
1344 		error = ext2_zero_inode_table(ip, cg);
1345 		if (error) {
1346 			brelse(bp);
1347 			EXT2_LOCK(ump);
1348 			return (0);
1349 		}
1350 	}
1351 	error = ext2_gd_i_bitmap_csum_verify(fs, cg, bp);
1352 	if (error) {
1353 		brelse(bp);
1354 		EXT2_LOCK(ump);
1355 		return (0);
1356 	}
1357 	if (e2fs_gd_get_nifree(&fs->e2fs_gd[cg]) == 0) {
1358 		/*
1359 		 * Another thread allocated the last i-node in this
1360 		 * group while we were waiting for the buffer.
1361 		 */
1362 		brelse(bp);
1363 		EXT2_LOCK(ump);
1364 		return (0);
1365 	}
1366 	ibp = (char *)bp->b_data;
1367 	if (ipref) {
1368 		ipref %= fs->e2fs_ipg;
1369 		if (isclr(ibp, ipref))
1370 			goto gotit;
1371 	}
1372 	start = ipref / NBBY;
1373 	len = howmany(fs->e2fs_ipg - ipref, NBBY);
1374 	loc = memcchr(&ibp[start], 0xff, len);
1375 	if (loc == NULL) {
1376 		len = start + 1;
1377 		start = 0;
1378 		loc = memcchr(&ibp[start], 0xff, len);
1379 		if (loc == NULL) {
1380 			SDT_PROBE3(ext2fs, , alloc,
1381 			    ext2_nodealloccg_bmap_corrupted, cg, ipref,
1382 			    fs->e2fs_fsmnt);
1383 			brelse(bp);
1384 			EXT2_LOCK(ump);
1385 			return (0);
1386 		}
1387 	}
1388 	ipref = (loc - ibp) * NBBY + ffs(~*loc) - 1;
1389 gotit:
1390 	setbit(ibp, ipref);
1391 	EXT2_LOCK(ump);
1392 	e2fs_gd_set_nifree(&fs->e2fs_gd[cg],
1393 	    e2fs_gd_get_nifree(&fs->e2fs_gd[cg]) - 1);
1394 	if (EXT2_HAS_RO_COMPAT_FEATURE(fs, EXT2F_ROCOMPAT_GDT_CSUM) ||
1395 	    EXT2_HAS_RO_COMPAT_FEATURE(fs, EXT2F_ROCOMPAT_METADATA_CKSUM)) {
1396 		ifree = fs->e2fs_ipg - e2fs_gd_get_i_unused(&fs->e2fs_gd[cg]);
1397 		if (ipref + 1 > ifree)
1398 			e2fs_gd_set_i_unused(&fs->e2fs_gd[cg],
1399 			    fs->e2fs_ipg - (ipref + 1));
1400 	}
1401 	fs->e2fs_ficount--;
1402 	fs->e2fs_fmod = 1;
1403 	if ((mode & IFMT) == IFDIR) {
1404 		e2fs_gd_set_ndirs(&fs->e2fs_gd[cg],
1405 		    e2fs_gd_get_ndirs(&fs->e2fs_gd[cg]) + 1);
1406 		fs->e2fs_total_dir++;
1407 	}
1408 	EXT2_UNLOCK(ump);
1409 	ext2_gd_i_bitmap_csum_set(fs, cg, bp);
1410 	bdwrite(bp);
1411 	return ((uint64_t)cg * fs->e2fs_ipg + ipref + 1);
1412 }
1413 
1414 /*
1415  * Free a block or fragment.
1416  *
1417  */
1418 void
1419 ext2_blkfree(struct inode *ip, e4fs_daddr_t bno, long size)
1420 {
1421 	struct m_ext2fs *fs;
1422 	struct buf *bp;
1423 	struct ext2mount *ump;
1424 	int cg, error;
1425 	char *bbp;
1426 
1427 	fs = ip->i_e2fs;
1428 	ump = ip->i_ump;
1429 	cg = dtog(fs, bno);
1430 	if (bno >= fs->e2fs_bcount) {
1431 		SDT_PROBE2(ext2fs, , alloc, ext2_blkfree_bad_block,
1432 		    ip->i_number, bno);
1433 		return;
1434 	}
1435 	error = bread(ip->i_devvp,
1436 	    fsbtodb(fs, e2fs_gd_get_b_bitmap(&fs->e2fs_gd[cg])),
1437 	    (int)fs->e2fs_bsize, NOCRED, &bp);
1438 	if (error) {
1439 		return;
1440 	}
1441 	bbp = (char *)bp->b_data;
1442 	bno = dtogd(fs, bno);
1443 	if (isclr(bbp, bno)) {
1444 		panic("ext2_blkfree: freeing free block %lld, fs=%s",
1445 		    (long long)bno, fs->e2fs_fsmnt);
1446 	}
1447 	clrbit(bbp, bno);
1448 	EXT2_LOCK(ump);
1449 	ext2_clusteracct(fs, bbp, cg, bno, 1);
1450 	fs->e2fs_fbcount++;
1451 	e2fs_gd_set_nbfree(&fs->e2fs_gd[cg],
1452 	    e2fs_gd_get_nbfree(&fs->e2fs_gd[cg]) + 1);
1453 	fs->e2fs_fmod = 1;
1454 	EXT2_UNLOCK(ump);
1455 	ext2_gd_b_bitmap_csum_set(fs, cg, bp);
1456 	bdwrite(bp);
1457 }
1458 
1459 /*
1460  * Free an inode.
1461  *
1462  */
1463 int
1464 ext2_vfree(struct vnode *pvp, ino_t ino, int mode)
1465 {
1466 	struct m_ext2fs *fs;
1467 	struct inode *pip;
1468 	struct buf *bp;
1469 	struct ext2mount *ump;
1470 	int error, cg;
1471 	char *ibp;
1472 
1473 	pip = VTOI(pvp);
1474 	fs = pip->i_e2fs;
1475 	ump = pip->i_ump;
1476 	if ((u_int)ino > fs->e2fs_ipg * fs->e2fs_gcount)
1477 		panic("ext2_vfree: range: devvp = %p, ino = %ju, fs = %s",
1478 		    pip->i_devvp, (uintmax_t)ino, fs->e2fs_fsmnt);
1479 
1480 	cg = ino_to_cg(fs, ino);
1481 	error = bread(pip->i_devvp,
1482 	    fsbtodb(fs, e2fs_gd_get_i_bitmap(&fs->e2fs_gd[cg])),
1483 	    (int)fs->e2fs_bsize, NOCRED, &bp);
1484 	if (error) {
1485 		return (0);
1486 	}
1487 	ibp = (char *)bp->b_data;
1488 	ino = (ino - 1) % fs->e2fs_ipg;
1489 	if (isclr(ibp, ino)) {
1490 		SDT_PROBE2(ext2fs, , alloc, ext2_vfree_doublefree,
1491 		    fs->e2fs_fsmnt, ino);
1492 		if (fs->e2fs_ronly == 0)
1493 			panic("ext2_vfree: freeing free inode");
1494 	}
1495 	clrbit(ibp, ino);
1496 	EXT2_LOCK(ump);
1497 	fs->e2fs_ficount++;
1498 	e2fs_gd_set_nifree(&fs->e2fs_gd[cg],
1499 	    e2fs_gd_get_nifree(&fs->e2fs_gd[cg]) + 1);
1500 	if ((mode & IFMT) == IFDIR) {
1501 		e2fs_gd_set_ndirs(&fs->e2fs_gd[cg],
1502 		    e2fs_gd_get_ndirs(&fs->e2fs_gd[cg]) - 1);
1503 		fs->e2fs_total_dir--;
1504 	}
1505 	fs->e2fs_fmod = 1;
1506 	EXT2_UNLOCK(ump);
1507 	ext2_gd_i_bitmap_csum_set(fs, cg, bp);
1508 	bdwrite(bp);
1509 	return (0);
1510 }
1511 
1512 /*
1513  * Find a block in the specified cylinder group.
1514  *
1515  * It is a panic if a request is made to find a block if none are
1516  * available.
1517  */
1518 static daddr_t
1519 ext2_mapsearch(struct m_ext2fs *fs, char *bbp, daddr_t bpref)
1520 {
1521 	char *loc;
1522 	int start, len;
1523 
1524 	/*
1525 	 * find the fragment by searching through the free block
1526 	 * map for an appropriate bit pattern
1527 	 */
1528 	if (bpref)
1529 		start = dtogd(fs, bpref) / NBBY;
1530 	else
1531 		start = 0;
1532 	len = howmany(fs->e2fs_fpg, NBBY) - start;
1533 	loc = memcchr(&bbp[start], 0xff, len);
1534 	if (loc == NULL) {
1535 		len = start + 1;
1536 		start = 0;
1537 		loc = memcchr(&bbp[start], 0xff, len);
1538 		if (loc == NULL) {
1539 			panic("ext2_mapsearch: map corrupted: start=%d, len=%d,"
1540 			    "fs=%s", start, len, fs->e2fs_fsmnt);
1541 			/* NOTREACHED */
1542 		}
1543 	}
1544 	return ((loc - bbp) * NBBY + ffs(~*loc) - 1);
1545 }
1546 
1547 int
1548 ext2_cg_has_sb(struct m_ext2fs *fs, int cg)
1549 {
1550 	int a3, a5, a7;
1551 
1552 	if (cg == 0)
1553 		return (1);
1554 
1555 	if (EXT2_HAS_COMPAT_FEATURE(fs, EXT2F_COMPAT_SPARSESUPER2)) {
1556 		if (cg == le32toh(fs->e2fs->e4fs_backup_bgs[0]) ||
1557 		    cg == le32toh(fs->e2fs->e4fs_backup_bgs[1]))
1558 			return (1);
1559 		return (0);
1560 	}
1561 
1562 	if ((cg <= 1) ||
1563 	    !EXT2_HAS_RO_COMPAT_FEATURE(fs, EXT2F_ROCOMPAT_SPARSESUPER))
1564 		return (1);
1565 
1566 	if (!(cg & 1))
1567 		return (0);
1568 
1569 	for (a3 = 3, a5 = 5, a7 = 7;
1570 	    a3 <= cg || a5 <= cg || a7 <= cg;
1571 	    a3 *= 3, a5 *= 5, a7 *= 7)
1572 		if (cg == a3 || cg == a5 || cg == a7)
1573 			return (1);
1574 	return (0);
1575 }
1576