xref: /freebsd/sys/fs/ext2fs/ext2_alloc.c (revision 1323ec57)
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  *	@(#)ffs_alloc.c	8.8 (Berkeley) 2/21/94
38  * $FreeBSD$
39  */
40 
41 #include <sys/param.h>
42 #include <sys/systm.h>
43 #include <sys/conf.h>
44 #include <sys/vnode.h>
45 #include <sys/sdt.h>
46 #include <sys/stat.h>
47 #include <sys/mount.h>
48 #include <sys/sysctl.h>
49 #include <sys/syslog.h>
50 #include <sys/buf.h>
51 #include <sys/endian.h>
52 
53 #include <fs/ext2fs/fs.h>
54 #include <fs/ext2fs/inode.h>
55 #include <fs/ext2fs/ext2_mount.h>
56 #include <fs/ext2fs/ext2fs.h>
57 #include <fs/ext2fs/ext2_extern.h>
58 
59 SDT_PROVIDER_DEFINE(ext2fs);
60 /*
61  * ext2fs trace probe:
62  * arg0: verbosity. Higher numbers give more verbose messages
63  * arg1: Textual message
64  */
65 SDT_PROBE_DEFINE2(ext2fs, , alloc, trace, "int", "char*");
66 SDT_PROBE_DEFINE3(ext2fs, , alloc, ext2_reallocblks_realloc,
67     "ino_t", "e2fs_lbn_t", "e2fs_lbn_t");
68 SDT_PROBE_DEFINE1(ext2fs, , alloc, ext2_reallocblks_bap, "uint32_t");
69 SDT_PROBE_DEFINE1(ext2fs, , alloc, ext2_reallocblks_blkno, "e2fs_daddr_t");
70 SDT_PROBE_DEFINE2(ext2fs, , alloc, ext2_b_bitmap_validate_error, "char*", "int");
71 SDT_PROBE_DEFINE3(ext2fs, , alloc, ext2_nodealloccg_bmap_corrupted,
72     "int", "daddr_t", "char*");
73 SDT_PROBE_DEFINE2(ext2fs, , alloc, ext2_blkfree_bad_block, "ino_t", "e4fs_daddr_t");
74 SDT_PROBE_DEFINE2(ext2fs, , alloc, ext2_vfree_doublefree, "char*", "ino_t");
75 
76 static daddr_t	ext2_alloccg(struct inode *, int, daddr_t, int);
77 static daddr_t	ext2_clusteralloc(struct inode *, int, daddr_t, int);
78 static u_long	ext2_dirpref(struct inode *);
79 static e4fs_daddr_t ext2_hashalloc(struct inode *, int, long, int,
80     daddr_t (*)(struct inode *, int, daddr_t,
81 						int));
82 static daddr_t	ext2_nodealloccg(struct inode *, int, daddr_t, int);
83 static daddr_t  ext2_mapsearch(struct m_ext2fs *, char *, daddr_t);
84 
85 /*
86  * Allocate a block in the filesystem.
87  *
88  * A preference may be optionally specified. If a preference is given
89  * the following hierarchy is used to allocate a block:
90  *   1) allocate the requested block.
91  *   2) allocate a rotationally optimal block in the same cylinder.
92  *   3) allocate a block in the same cylinder group.
93  *   4) quadratically rehash into other cylinder groups, until an
94  *        available block is located.
95  * If no block preference is given the following hierarchy is used
96  * to allocate a block:
97  *   1) allocate a block in the cylinder group that contains the
98  *        inode for the file.
99  *   2) quadratically rehash into other cylinder groups, until an
100  *        available block is located.
101  */
102 int
103 ext2_alloc(struct inode *ip, daddr_t lbn, e4fs_daddr_t bpref, int size,
104     struct ucred *cred, e4fs_daddr_t *bnp)
105 {
106 	struct m_ext2fs *fs;
107 	struct ext2mount *ump;
108 	e4fs_daddr_t bno;
109 	int cg;
110 
111 	*bnp = 0;
112 	fs = ip->i_e2fs;
113 	ump = ip->i_ump;
114 	mtx_assert(EXT2_MTX(ump), MA_OWNED);
115 #ifdef INVARIANTS
116 	if ((u_int)size > fs->e2fs_bsize || blkoff(fs, size) != 0) {
117 		vn_printf(ip->i_devvp, "bsize = %lu, size = %d, fs = %s\n",
118 		    (long unsigned int)fs->e2fs_bsize, size, fs->e2fs_fsmnt);
119 		panic("ext2_alloc: bad size");
120 	}
121 	if (cred == NOCRED)
122 		panic("ext2_alloc: missing credential");
123 #endif		/* INVARIANTS */
124 	if (size == fs->e2fs_bsize && fs->e2fs_fbcount == 0)
125 		goto nospace;
126 	if (cred->cr_uid != 0 &&
127 	    fs->e2fs_fbcount < fs->e2fs_rbcount)
128 		goto nospace;
129 	if (bpref >= fs->e2fs_bcount)
130 		bpref = 0;
131 	if (bpref == 0)
132 		cg = ino_to_cg(fs, ip->i_number);
133 	else
134 		cg = dtog(fs, bpref);
135 	bno = (daddr_t)ext2_hashalloc(ip, cg, bpref, fs->e2fs_bsize,
136 	    ext2_alloccg);
137 	if (bno > 0) {
138 		/* set next_alloc fields as done in block_getblk */
139 		ip->i_next_alloc_block = lbn;
140 		ip->i_next_alloc_goal = bno;
141 
142 		ip->i_blocks += btodb(fs->e2fs_bsize);
143 		ip->i_flag |= IN_CHANGE | IN_UPDATE;
144 		*bnp = bno;
145 		return (0);
146 	}
147 nospace:
148 	EXT2_UNLOCK(ump);
149 	SDT_PROBE2(ext2fs, , alloc, trace, 1, "cannot allocate data block");
150 	return (ENOSPC);
151 }
152 
153 /*
154  * Allocate EA's block for inode.
155  */
156 e4fs_daddr_t
157 ext2_alloc_meta(struct inode *ip)
158 {
159 	struct m_ext2fs *fs;
160 	daddr_t blk;
161 
162 	fs = ip->i_e2fs;
163 
164 	EXT2_LOCK(ip->i_ump);
165 	blk = ext2_hashalloc(ip, ino_to_cg(fs, ip->i_number), 0, fs->e2fs_bsize,
166 	    ext2_alloccg);
167 	if (0 == blk) {
168 		EXT2_UNLOCK(ip->i_ump);
169 		SDT_PROBE2(ext2fs, , alloc, trace, 1, "cannot allocate meta block");
170 	}
171 
172 	return (blk);
173 }
174 
175 /*
176  * Reallocate a sequence of blocks into a contiguous sequence of blocks.
177  *
178  * The vnode and an array of buffer pointers for a range of sequential
179  * logical blocks to be made contiguous is given. The allocator attempts
180  * to find a range of sequential blocks starting as close as possible to
181  * an fs_rotdelay offset from the end of the allocation for the logical
182  * block immediately preceding the current range. If successful, the
183  * physical block numbers in the buffer pointers and in the inode are
184  * changed to reflect the new allocation. If unsuccessful, the allocation
185  * is left unchanged. The success in doing the reallocation is returned.
186  * Note that the error return is not reflected back to the user. Rather
187  * the previous block allocation will be used.
188  */
189 
190 static SYSCTL_NODE(_vfs, OID_AUTO, ext2fs, CTLFLAG_RW | CTLFLAG_MPSAFE, 0,
191     "EXT2FS filesystem");
192 
193 static int doasyncfree = 1;
194 
195 SYSCTL_INT(_vfs_ext2fs, OID_AUTO, doasyncfree, CTLFLAG_RW, &doasyncfree, 0,
196     "Use asynchronous writes to update block pointers when freeing blocks");
197 
198 static int doreallocblks = 0;
199 
200 SYSCTL_INT(_vfs_ext2fs, OID_AUTO, doreallocblks, CTLFLAG_RW, &doreallocblks, 0, "");
201 
202 int
203 ext2_reallocblks(struct vop_reallocblks_args *ap)
204 {
205 	struct m_ext2fs *fs;
206 	struct inode *ip;
207 	struct vnode *vp;
208 	struct buf *sbp, *ebp;
209 	uint32_t *bap, *sbap, *ebap;
210 	struct ext2mount *ump;
211 	struct cluster_save *buflist;
212 	struct indir start_ap[EXT2_NIADDR + 1], end_ap[EXT2_NIADDR + 1], *idp;
213 	e2fs_lbn_t start_lbn, end_lbn;
214 	int soff;
215 	e2fs_daddr_t newblk, blkno;
216 	int i, len, start_lvl, end_lvl, pref, ssize;
217 
218 	if (doreallocblks == 0)
219 		return (ENOSPC);
220 
221 	vp = ap->a_vp;
222 	ip = VTOI(vp);
223 	fs = ip->i_e2fs;
224 	ump = ip->i_ump;
225 
226 	if (fs->e2fs_contigsumsize <= 0 || ip->i_flag & IN_E4EXTENTS)
227 		return (ENOSPC);
228 
229 	buflist = ap->a_buflist;
230 	len = buflist->bs_nchildren;
231 	start_lbn = buflist->bs_children[0]->b_lblkno;
232 	end_lbn = start_lbn + len - 1;
233 #ifdef INVARIANTS
234 	for (i = 1; i < len; i++)
235 		if (buflist->bs_children[i]->b_lblkno != start_lbn + i)
236 			panic("ext2_reallocblks: non-cluster");
237 #endif
238 	/*
239 	 * If the cluster crosses the boundary for the first indirect
240 	 * block, leave space for the indirect block. Indirect blocks
241 	 * are initially laid out in a position after the last direct
242 	 * block. Block reallocation would usually destroy locality by
243 	 * moving the indirect block out of the way to make room for
244 	 * data blocks if we didn't compensate here. We should also do
245 	 * this for other indirect block boundaries, but it is only
246 	 * important for the first one.
247 	 */
248 	if (start_lbn < EXT2_NDADDR && end_lbn >= EXT2_NDADDR)
249 		return (ENOSPC);
250 	/*
251 	 * If the latest allocation is in a new cylinder group, assume that
252 	 * the filesystem has decided to move and do not force it back to
253 	 * the previous cylinder group.
254 	 */
255 	if (dtog(fs, dbtofsb(fs, buflist->bs_children[0]->b_blkno)) !=
256 	    dtog(fs, dbtofsb(fs, buflist->bs_children[len - 1]->b_blkno)))
257 		return (ENOSPC);
258 	if (ext2_getlbns(vp, start_lbn, start_ap, &start_lvl) ||
259 	    ext2_getlbns(vp, end_lbn, end_ap, &end_lvl))
260 		return (ENOSPC);
261 	/*
262 	 * Get the starting offset and block map for the first block.
263 	 */
264 	if (start_lvl == 0) {
265 		sbap = &ip->i_db[0];
266 		soff = start_lbn;
267 	} else {
268 		idp = &start_ap[start_lvl - 1];
269 		if (bread(vp, idp->in_lbn, (int)fs->e2fs_bsize, NOCRED, &sbp)) {
270 			brelse(sbp);
271 			return (ENOSPC);
272 		}
273 		sbap = (u_int *)sbp->b_data;
274 		soff = idp->in_off;
275 	}
276 	/*
277 	 * If the block range spans two block maps, get the second map.
278 	 */
279 	ebap = NULL;
280 	if (end_lvl == 0 || (idp = &end_ap[end_lvl - 1])->in_off + 1 >= len) {
281 		ssize = len;
282 	} else {
283 #ifdef INVARIANTS
284 		if (start_ap[start_lvl - 1].in_lbn == idp->in_lbn)
285 			panic("ext2_reallocblks: start == end");
286 #endif
287 		ssize = len - (idp->in_off + 1);
288 		if (bread(vp, idp->in_lbn, (int)fs->e2fs_bsize, NOCRED, &ebp))
289 			goto fail;
290 		ebap = (u_int *)ebp->b_data;
291 	}
292 	/*
293 	 * Find the preferred location for the cluster.
294 	 */
295 	EXT2_LOCK(ump);
296 	pref = ext2_blkpref(ip, start_lbn, soff, sbap, 0);
297 	/*
298 	 * Search the block map looking for an allocation of the desired size.
299 	 */
300 	if ((newblk = (e2fs_daddr_t)ext2_hashalloc(ip, dtog(fs, pref), pref,
301 	    len, ext2_clusteralloc)) == 0) {
302 		EXT2_UNLOCK(ump);
303 		goto fail;
304 	}
305 	/*
306 	 * We have found a new contiguous block.
307 	 *
308 	 * First we have to replace the old block pointers with the new
309 	 * block pointers in the inode and indirect blocks associated
310 	 * with the file.
311 	 */
312 	SDT_PROBE3(ext2fs, , alloc, ext2_reallocblks_realloc,
313 	    ip->i_number, start_lbn, end_lbn);
314 	blkno = newblk;
315 	for (bap = &sbap[soff], i = 0; i < len; i++, blkno += fs->e2fs_fpb) {
316 		if (i == ssize) {
317 			bap = ebap;
318 			soff = -i;
319 		}
320 #ifdef INVARIANTS
321 		if (buflist->bs_children[i]->b_blkno != fsbtodb(fs, *bap))
322 			panic("ext2_reallocblks: alloc mismatch");
323 #endif
324 		SDT_PROBE1(ext2fs, , alloc, ext2_reallocblks_bap, *bap);
325 		*bap++ = blkno;
326 	}
327 	/*
328 	 * Next we must write out the modified inode and indirect blocks.
329 	 * For strict correctness, the writes should be synchronous since
330 	 * the old block values may have been written to disk. In practise
331 	 * they are almost never written, but if we are concerned about
332 	 * strict correctness, the `doasyncfree' flag should be set to zero.
333 	 *
334 	 * The test on `doasyncfree' should be changed to test a flag
335 	 * that shows whether the associated buffers and inodes have
336 	 * been written. The flag should be set when the cluster is
337 	 * started and cleared whenever the buffer or inode is flushed.
338 	 * We can then check below to see if it is set, and do the
339 	 * synchronous write only when it has been cleared.
340 	 */
341 	if (sbap != &ip->i_db[0]) {
342 		if (doasyncfree)
343 			bdwrite(sbp);
344 		else
345 			bwrite(sbp);
346 	} else {
347 		ip->i_flag |= IN_CHANGE | IN_UPDATE;
348 		if (!doasyncfree)
349 			ext2_update(vp, 1);
350 	}
351 	if (ssize < len) {
352 		if (doasyncfree)
353 			bdwrite(ebp);
354 		else
355 			bwrite(ebp);
356 	}
357 	/*
358 	 * Last, free the old blocks and assign the new blocks to the buffers.
359 	 */
360 	for (blkno = newblk, i = 0; i < len; i++, blkno += fs->e2fs_fpb) {
361 		ext2_blkfree(ip, dbtofsb(fs, buflist->bs_children[i]->b_blkno),
362 		    fs->e2fs_bsize);
363 		buflist->bs_children[i]->b_blkno = fsbtodb(fs, blkno);
364 		SDT_PROBE1(ext2fs, , alloc, ext2_reallocblks_blkno, blkno);
365 	}
366 
367 	return (0);
368 
369 fail:
370 	if (ssize < len)
371 		brelse(ebp);
372 	if (sbap != &ip->i_db[0])
373 		brelse(sbp);
374 	return (ENOSPC);
375 }
376 
377 /*
378  * Allocate an inode in the filesystem.
379  *
380  */
381 int
382 ext2_valloc(struct vnode *pvp, int mode, struct ucred *cred, struct vnode **vpp)
383 {
384 	struct timespec ts;
385 	struct m_ext2fs *fs;
386 	struct ext2mount *ump;
387 	struct inode *pip;
388 	struct inode *ip;
389 	struct vnode *vp;
390 	struct thread *td;
391 	ino_t ino, ipref;
392 	int error, cg;
393 
394 	*vpp = NULL;
395 	pip = VTOI(pvp);
396 	fs = pip->i_e2fs;
397 	ump = pip->i_ump;
398 
399 	EXT2_LOCK(ump);
400 	if (fs->e2fs_ficount == 0)
401 		goto noinodes;
402 	/*
403 	 * If it is a directory then obtain a cylinder group based on
404 	 * ext2_dirpref else obtain it using ino_to_cg. The preferred inode is
405 	 * always the next inode.
406 	 */
407 	if ((mode & IFMT) == IFDIR) {
408 		cg = ext2_dirpref(pip);
409 		if (fs->e2fs_contigdirs[cg] < 255)
410 			fs->e2fs_contigdirs[cg]++;
411 	} else {
412 		cg = ino_to_cg(fs, pip->i_number);
413 		if (fs->e2fs_contigdirs[cg] > 0)
414 			fs->e2fs_contigdirs[cg]--;
415 	}
416 	ipref = cg * fs->e2fs_ipg + 1;
417 	ino = (ino_t)ext2_hashalloc(pip, cg, (long)ipref, mode, ext2_nodealloccg);
418 	if (ino == 0)
419 		goto noinodes;
420 
421 	td = curthread;
422 	error = vfs_hash_get(ump->um_mountp, ino, LK_EXCLUSIVE, td, vpp, NULL, NULL);
423 	if (error || *vpp != NULL) {
424 		return (error);
425 	}
426 
427 	ip = malloc(sizeof(struct inode), M_EXT2NODE, M_WAITOK | M_ZERO);
428 
429 	/* Allocate a new vnode/inode. */
430 	if ((error = getnewvnode("ext2fs", ump->um_mountp, &ext2_vnodeops, &vp)) != 0) {
431 		free(ip, M_EXT2NODE);
432 		return (error);
433 	}
434 
435 	lockmgr(vp->v_vnlock, LK_EXCLUSIVE, NULL);
436 	vp->v_data = ip;
437 	ip->i_vnode = vp;
438 	ip->i_e2fs = fs = ump->um_e2fs;
439 	ip->i_ump = ump;
440 	ip->i_number = ino;
441 	ip->i_block_group = ino_to_cg(fs, ino);
442 	ip->i_next_alloc_block = 0;
443 	ip->i_next_alloc_goal = 0;
444 
445 	error = insmntque(vp, ump->um_mountp);
446 	if (error) {
447 		free(ip, M_EXT2NODE);
448 		return (error);
449 	}
450 
451 	error = vfs_hash_insert(vp, ino, LK_EXCLUSIVE, td, vpp, NULL, NULL);
452 	if (error || *vpp != NULL) {
453 		*vpp = NULL;
454 		free(ip, M_EXT2NODE);
455 		return (error);
456 	}
457 
458 	if ((error = ext2_vinit(ump->um_mountp, &ext2_fifoops, &vp)) != 0) {
459 		vput(vp);
460 		*vpp = NULL;
461 		free(ip, M_EXT2NODE);
462 		return (error);
463 	}
464 
465 	if (EXT2_HAS_INCOMPAT_FEATURE(fs, EXT2F_INCOMPAT_EXTENTS)
466 	    && (S_ISREG(mode) || S_ISDIR(mode)))
467 		ext4_ext_tree_init(ip);
468 	else
469 		memset(ip->i_data, 0, sizeof(ip->i_data));
470 
471 	/*
472 	 * Set up a new generation number for this inode.
473 	 * Avoid zero values.
474 	 */
475 	do {
476 		ip->i_gen = arc4random();
477 	} while (ip->i_gen == 0);
478 
479 	vfs_timestamp(&ts);
480 	ip->i_birthtime = ts.tv_sec;
481 	ip->i_birthnsec = ts.tv_nsec;
482 
483 	*vpp = vp;
484 
485 	return (0);
486 
487 noinodes:
488 	EXT2_UNLOCK(ump);
489 	SDT_PROBE2(ext2fs, , alloc, trace, 1, "out of inodes");
490 	return (ENOSPC);
491 }
492 
493 /*
494  * 64-bit compatible getters and setters for struct ext2_gd from ext2fs.h
495  */
496 uint64_t
497 e2fs_gd_get_b_bitmap(struct ext2_gd *gd)
498 {
499 
500 	return (((uint64_t)(le32toh(gd->ext4bgd_b_bitmap_hi)) << 32) |
501 	    le32toh(gd->ext2bgd_b_bitmap));
502 }
503 
504 uint64_t
505 e2fs_gd_get_i_bitmap(struct ext2_gd *gd)
506 {
507 
508 	return (((uint64_t)(le32toh(gd->ext4bgd_i_bitmap_hi)) << 32) |
509 	    le32toh(gd->ext2bgd_i_bitmap));
510 }
511 
512 uint64_t
513 e2fs_gd_get_i_tables(struct ext2_gd *gd)
514 {
515 
516 	return (((uint64_t)(le32toh(gd->ext4bgd_i_tables_hi)) << 32) |
517 	    le32toh(gd->ext2bgd_i_tables));
518 }
519 
520 static uint32_t
521 e2fs_gd_get_nbfree(struct ext2_gd *gd)
522 {
523 
524 	return (((uint32_t)(le16toh(gd->ext4bgd_nbfree_hi)) << 16) |
525 	    le16toh(gd->ext2bgd_nbfree));
526 }
527 
528 static void
529 e2fs_gd_set_nbfree(struct ext2_gd *gd, uint32_t val)
530 {
531 
532 	gd->ext2bgd_nbfree = htole16(val & 0xffff);
533 	gd->ext4bgd_nbfree_hi = htole16(val >> 16);
534 }
535 
536 static uint32_t
537 e2fs_gd_get_nifree(struct ext2_gd *gd)
538 {
539 
540 	return (((uint32_t)(le16toh(gd->ext4bgd_nifree_hi)) << 16) |
541 	    le16toh(gd->ext2bgd_nifree));
542 }
543 
544 static void
545 e2fs_gd_set_nifree(struct ext2_gd *gd, uint32_t val)
546 {
547 
548 	gd->ext2bgd_nifree = htole16(val & 0xffff);
549 	gd->ext4bgd_nifree_hi = htole16(val >> 16);
550 }
551 
552 uint32_t
553 e2fs_gd_get_ndirs(struct ext2_gd *gd)
554 {
555 
556 	return (((uint32_t)(le16toh(gd->ext4bgd_ndirs_hi)) << 16) |
557 	    le16toh(gd->ext2bgd_ndirs));
558 }
559 
560 static void
561 e2fs_gd_set_ndirs(struct ext2_gd *gd, uint32_t val)
562 {
563 
564 	gd->ext2bgd_ndirs = htole16(val & 0xffff);
565 	gd->ext4bgd_ndirs_hi = htole16(val >> 16);
566 }
567 
568 static uint32_t
569 e2fs_gd_get_i_unused(struct ext2_gd *gd)
570 {
571 	return ((uint32_t)(le16toh(gd->ext4bgd_i_unused_hi) << 16) |
572 	    le16toh(gd->ext4bgd_i_unused));
573 }
574 
575 static void
576 e2fs_gd_set_i_unused(struct ext2_gd *gd, uint32_t val)
577 {
578 
579 	gd->ext4bgd_i_unused = htole16(val & 0xffff);
580 	gd->ext4bgd_i_unused_hi = htole16(val >> 16);
581 }
582 
583 /*
584  * Find a cylinder to place a directory.
585  *
586  * The policy implemented by this algorithm is to allocate a
587  * directory inode in the same cylinder group as its parent
588  * directory, but also to reserve space for its files inodes
589  * and data. Restrict the number of directories which may be
590  * allocated one after another in the same cylinder group
591  * without intervening allocation of files.
592  *
593  * If we allocate a first level directory then force allocation
594  * in another cylinder group.
595  *
596  */
597 static u_long
598 ext2_dirpref(struct inode *pip)
599 {
600 	struct m_ext2fs *fs;
601 	int cg, prefcg, cgsize;
602 	uint64_t avgbfree, minbfree;
603 	u_int avgifree, avgndir, curdirsize;
604 	u_int minifree, maxndir;
605 	u_int mincg, minndir;
606 	u_int dirsize, maxcontigdirs;
607 
608 	mtx_assert(EXT2_MTX(pip->i_ump), MA_OWNED);
609 	fs = pip->i_e2fs;
610 
611 	avgifree = fs->e2fs_ficount / fs->e2fs_gcount;
612 	avgbfree = fs->e2fs_fbcount / fs->e2fs_gcount;
613 	avgndir = fs->e2fs_total_dir / fs->e2fs_gcount;
614 
615 	/*
616 	 * Force allocation in another cg if creating a first level dir.
617 	 */
618 	ASSERT_VOP_LOCKED(ITOV(pip), "ext2fs_dirpref");
619 	if (ITOV(pip)->v_vflag & VV_ROOT) {
620 		prefcg = arc4random() % fs->e2fs_gcount;
621 		mincg = prefcg;
622 		minndir = fs->e2fs_ipg;
623 		for (cg = prefcg; cg < fs->e2fs_gcount; cg++)
624 			if (e2fs_gd_get_ndirs(&fs->e2fs_gd[cg]) < minndir &&
625 			    e2fs_gd_get_nifree(&fs->e2fs_gd[cg]) >= avgifree &&
626 			    e2fs_gd_get_nbfree(&fs->e2fs_gd[cg]) >= avgbfree) {
627 				mincg = cg;
628 				minndir = e2fs_gd_get_ndirs(&fs->e2fs_gd[cg]);
629 			}
630 		for (cg = 0; cg < prefcg; cg++)
631 			if (e2fs_gd_get_ndirs(&fs->e2fs_gd[cg]) < minndir &&
632 			    e2fs_gd_get_nifree(&fs->e2fs_gd[cg]) >= avgifree &&
633 			    e2fs_gd_get_nbfree(&fs->e2fs_gd[cg]) >= avgbfree) {
634 				mincg = cg;
635 				minndir = e2fs_gd_get_ndirs(&fs->e2fs_gd[cg]);
636 			}
637 		return (mincg);
638 	}
639 	/*
640 	 * Count various limits which used for
641 	 * optimal allocation of a directory inode.
642 	 */
643 	maxndir = min(avgndir + fs->e2fs_ipg / 16, fs->e2fs_ipg);
644 	minifree = avgifree - avgifree / 4;
645 	if (minifree < 1)
646 		minifree = 1;
647 	minbfree = avgbfree - avgbfree / 4;
648 	if (minbfree < 1)
649 		minbfree = 1;
650 	cgsize = fs->e2fs_fsize * fs->e2fs_fpg;
651 	dirsize = AVGDIRSIZE;
652 	curdirsize = avgndir ?
653 	    (cgsize - avgbfree * fs->e2fs_bsize) / avgndir : 0;
654 	if (dirsize < curdirsize)
655 		dirsize = curdirsize;
656 	maxcontigdirs = min((avgbfree * fs->e2fs_bsize) / dirsize, 255);
657 	maxcontigdirs = min(maxcontigdirs, fs->e2fs_ipg / AFPDIR);
658 	if (maxcontigdirs == 0)
659 		maxcontigdirs = 1;
660 
661 	/*
662 	 * Limit number of dirs in one cg and reserve space for
663 	 * regular files, but only if we have no deficit in
664 	 * inodes or space.
665 	 */
666 	prefcg = ino_to_cg(fs, pip->i_number);
667 	for (cg = prefcg; cg < fs->e2fs_gcount; cg++)
668 		if (e2fs_gd_get_ndirs(&fs->e2fs_gd[cg]) < maxndir &&
669 		    e2fs_gd_get_nifree(&fs->e2fs_gd[cg]) >= minifree &&
670 		    e2fs_gd_get_nbfree(&fs->e2fs_gd[cg]) >= minbfree) {
671 			if (fs->e2fs_contigdirs[cg] < maxcontigdirs)
672 				return (cg);
673 		}
674 	for (cg = 0; cg < prefcg; cg++)
675 		if (e2fs_gd_get_ndirs(&fs->e2fs_gd[cg]) < maxndir &&
676 		    e2fs_gd_get_nifree(&fs->e2fs_gd[cg]) >= minifree &&
677 		    e2fs_gd_get_nbfree(&fs->e2fs_gd[cg]) >= minbfree) {
678 			if (fs->e2fs_contigdirs[cg] < maxcontigdirs)
679 				return (cg);
680 		}
681 	/*
682 	 * This is a backstop when we have deficit in space.
683 	 */
684 	for (cg = prefcg; cg < fs->e2fs_gcount; cg++)
685 		if (e2fs_gd_get_nifree(&fs->e2fs_gd[cg]) >= avgifree)
686 			return (cg);
687 	for (cg = 0; cg < prefcg; cg++)
688 		if (e2fs_gd_get_nifree(&fs->e2fs_gd[cg]) >= avgifree)
689 			break;
690 	return (cg);
691 }
692 
693 /*
694  * Select the desired position for the next block in a file.
695  *
696  * we try to mimic what Remy does in inode_getblk/block_getblk
697  *
698  * we note: blocknr == 0 means that we're about to allocate either
699  * a direct block or a pointer block at the first level of indirection
700  * (In other words, stuff that will go in i_db[] or i_ib[])
701  *
702  * blocknr != 0 means that we're allocating a block that is none
703  * of the above. Then, blocknr tells us the number of the block
704  * that will hold the pointer
705  */
706 e4fs_daddr_t
707 ext2_blkpref(struct inode *ip, e2fs_lbn_t lbn, int indx, e2fs_daddr_t *bap,
708     e2fs_daddr_t blocknr)
709 {
710 	struct m_ext2fs *fs;
711 	int tmp;
712 
713 	fs = ip->i_e2fs;
714 
715 	mtx_assert(EXT2_MTX(ip->i_ump), MA_OWNED);
716 
717 	/*
718 	 * If the next block is actually what we thought it is, then set the
719 	 * goal to what we thought it should be.
720 	 */
721 	if (ip->i_next_alloc_block == lbn && ip->i_next_alloc_goal != 0)
722 		return ip->i_next_alloc_goal;
723 
724 	/*
725 	 * Now check whether we were provided with an array that basically
726 	 * tells us previous blocks to which we want to stay close.
727 	 */
728 	if (bap)
729 		for (tmp = indx - 1; tmp >= 0; tmp--)
730 			if (bap[tmp])
731 				return (le32toh(bap[tmp]));
732 
733 	/*
734 	 * Else lets fall back to the blocknr or, if there is none, follow
735 	 * the rule that a block should be allocated near its inode.
736 	 */
737 	return (blocknr ? blocknr :
738 	    (e2fs_daddr_t)(ip->i_block_group *
739 	    EXT2_BLOCKS_PER_GROUP(fs)) + le32toh(fs->e2fs->e2fs_first_dblock));
740 }
741 
742 /*
743  * Implement the cylinder overflow algorithm.
744  *
745  * The policy implemented by this algorithm is:
746  *   1) allocate the block in its requested cylinder group.
747  *   2) quadratically rehash on the cylinder group number.
748  *   3) brute force search for a free block.
749  */
750 static e4fs_daddr_t
751 ext2_hashalloc(struct inode *ip, int cg, long pref, int size,
752     daddr_t (*allocator) (struct inode *, int, daddr_t, int))
753 {
754 	struct m_ext2fs *fs;
755 	e4fs_daddr_t result;
756 	int i, icg = cg;
757 
758 	mtx_assert(EXT2_MTX(ip->i_ump), MA_OWNED);
759 	fs = ip->i_e2fs;
760 	/*
761 	 * 1: preferred cylinder group
762 	 */
763 	result = (*allocator)(ip, cg, pref, size);
764 	if (result)
765 		return (result);
766 	/*
767 	 * 2: quadratic rehash
768 	 */
769 	for (i = 1; i < fs->e2fs_gcount; i *= 2) {
770 		cg += i;
771 		if (cg >= fs->e2fs_gcount)
772 			cg -= fs->e2fs_gcount;
773 		result = (*allocator)(ip, cg, 0, size);
774 		if (result)
775 			return (result);
776 	}
777 	/*
778 	 * 3: brute force search
779 	 * Note that we start at i == 2, since 0 was checked initially,
780 	 * and 1 is always checked in the quadratic rehash.
781 	 */
782 	cg = (icg + 2) % fs->e2fs_gcount;
783 	for (i = 2; i < fs->e2fs_gcount; i++) {
784 		result = (*allocator)(ip, cg, 0, size);
785 		if (result)
786 			return (result);
787 		cg++;
788 		if (cg == fs->e2fs_gcount)
789 			cg = 0;
790 	}
791 	return (0);
792 }
793 
794 static uint64_t
795 ext2_cg_number_gdb_nometa(struct m_ext2fs *fs, int cg)
796 {
797 
798 	if (!ext2_cg_has_sb(fs, cg))
799 		return (0);
800 
801 	if (EXT2_HAS_INCOMPAT_FEATURE(fs, EXT2F_INCOMPAT_META_BG))
802 		return (le32toh(fs->e2fs->e3fs_first_meta_bg));
803 
804 	return ((fs->e2fs_gcount + EXT2_DESCS_PER_BLOCK(fs) - 1) /
805 	    EXT2_DESCS_PER_BLOCK(fs));
806 }
807 
808 static uint64_t
809 ext2_cg_number_gdb_meta(struct m_ext2fs *fs, int cg)
810 {
811 	unsigned long metagroup;
812 	int first, last;
813 
814 	metagroup = cg / EXT2_DESCS_PER_BLOCK(fs);
815 	first = metagroup * EXT2_DESCS_PER_BLOCK(fs);
816 	last = first + EXT2_DESCS_PER_BLOCK(fs) - 1;
817 
818 	if (cg == first || cg == first + 1 || cg == last)
819 		return (1);
820 
821 	return (0);
822 }
823 
824 uint64_t
825 ext2_cg_number_gdb(struct m_ext2fs *fs, int cg)
826 {
827 	unsigned long first_meta_bg, metagroup;
828 
829 	first_meta_bg = le32toh(fs->e2fs->e3fs_first_meta_bg);
830 	metagroup = cg / EXT2_DESCS_PER_BLOCK(fs);
831 
832 	if (!EXT2_HAS_INCOMPAT_FEATURE(fs, EXT2F_INCOMPAT_META_BG) ||
833 	    metagroup < first_meta_bg)
834 		return (ext2_cg_number_gdb_nometa(fs, cg));
835 
836 	return ext2_cg_number_gdb_meta(fs, cg);
837 }
838 
839 static int
840 ext2_number_base_meta_blocks(struct m_ext2fs *fs, int cg)
841 {
842 	int number;
843 
844 	number = ext2_cg_has_sb(fs, cg);
845 
846 	if (!EXT2_HAS_INCOMPAT_FEATURE(fs, EXT2F_INCOMPAT_META_BG) ||
847 	    cg < le32toh(fs->e2fs->e3fs_first_meta_bg) *
848 	    EXT2_DESCS_PER_BLOCK(fs)) {
849 		if (number) {
850 			number += ext2_cg_number_gdb(fs, cg);
851 			number += le16toh(fs->e2fs->e2fs_reserved_ngdb);
852 		}
853 	} else {
854 		number += ext2_cg_number_gdb(fs, cg);
855 	}
856 
857 	return (number);
858 }
859 
860 static void
861 ext2_mark_bitmap_end(int start_bit, int end_bit, char *bitmap)
862 {
863 	int i;
864 
865 	if (start_bit >= end_bit)
866 		return;
867 
868 	for (i = start_bit; i < ((start_bit + 7) & ~7UL); i++)
869 		setbit(bitmap, i);
870 	if (i < end_bit)
871 		memset(bitmap + (i >> 3), 0xff, (end_bit - i) >> 3);
872 }
873 
874 static int
875 ext2_get_group_number(struct m_ext2fs *fs, e4fs_daddr_t block)
876 {
877 
878 	return ((block - le32toh(fs->e2fs->e2fs_first_dblock)) /
879 	    fs->e2fs_bsize);
880 }
881 
882 static int
883 ext2_block_in_group(struct m_ext2fs *fs, e4fs_daddr_t block, int cg)
884 {
885 
886 	return ((ext2_get_group_number(fs, block) == cg) ? 1 : 0);
887 }
888 
889 static int
890 ext2_cg_block_bitmap_init(struct m_ext2fs *fs, int cg, struct buf *bp)
891 {
892 	int bit, bit_max, inodes_per_block;
893 	uint64_t start, tmp;
894 
895 	if (!(le16toh(fs->e2fs_gd[cg].ext4bgd_flags) & EXT2_BG_BLOCK_UNINIT))
896 		return (0);
897 
898 	memset(bp->b_data, 0, fs->e2fs_bsize);
899 
900 	bit_max = ext2_number_base_meta_blocks(fs, cg);
901 	if ((bit_max >> 3) >= fs->e2fs_bsize)
902 		return (EINVAL);
903 
904 	for (bit = 0; bit < bit_max; bit++)
905 		setbit(bp->b_data, bit);
906 
907 	start = (uint64_t)cg * fs->e2fs_bpg +
908 	    le32toh(fs->e2fs->e2fs_first_dblock);
909 
910 	/* Set bits for block and inode bitmaps, and inode table. */
911 	tmp = e2fs_gd_get_b_bitmap(&fs->e2fs_gd[cg]);
912 	if (!EXT2_HAS_INCOMPAT_FEATURE(fs, EXT2F_INCOMPAT_FLEX_BG) ||
913 	    ext2_block_in_group(fs, tmp, cg))
914 		setbit(bp->b_data, tmp - start);
915 
916 	tmp = e2fs_gd_get_i_bitmap(&fs->e2fs_gd[cg]);
917 	if (!EXT2_HAS_INCOMPAT_FEATURE(fs, EXT2F_INCOMPAT_FLEX_BG) ||
918 	    ext2_block_in_group(fs, tmp, cg))
919 		setbit(bp->b_data, tmp - start);
920 
921 	tmp = e2fs_gd_get_i_tables(&fs->e2fs_gd[cg]);
922 	inodes_per_block = fs->e2fs_bsize/EXT2_INODE_SIZE(fs);
923 	while( tmp < e2fs_gd_get_i_tables(&fs->e2fs_gd[cg]) +
924 	    fs->e2fs_ipg / inodes_per_block ) {
925 		if (!EXT2_HAS_INCOMPAT_FEATURE(fs, EXT2F_INCOMPAT_FLEX_BG) ||
926 		    ext2_block_in_group(fs, tmp, cg))
927 			setbit(bp->b_data, tmp - start);
928 		tmp++;
929 	}
930 
931 	/*
932 	 * Also if the number of blocks within the group is less than
933 	 * the blocksize * 8 ( which is the size of bitmap ), set rest
934 	 * of the block bitmap to 1
935 	 */
936 	ext2_mark_bitmap_end(fs->e2fs_bpg, fs->e2fs_bsize * 8,
937 	    bp->b_data);
938 
939 	/* Clean the flag */
940 	fs->e2fs_gd[cg].ext4bgd_flags = htole16(le16toh(
941 	    fs->e2fs_gd[cg].ext4bgd_flags) & ~EXT2_BG_BLOCK_UNINIT);
942 
943 	return (0);
944 }
945 
946 static int
947 ext2_b_bitmap_validate(struct m_ext2fs *fs, struct buf *bp, int cg)
948 {
949 	struct ext2_gd *gd;
950 	uint64_t group_first_block;
951 	unsigned int offset, max_bit;
952 
953 	if (EXT2_HAS_INCOMPAT_FEATURE(fs, EXT2F_INCOMPAT_FLEX_BG)) {
954 		/*
955 		 * It is not possible to check block bitmap in case of this
956 		 * feature, because the inode and block bitmaps and inode table
957 		 * blocks may not be in the group at all.
958 		 * So, skip check in this case.
959 		 */
960 		return (0);
961 	}
962 
963 	gd = &fs->e2fs_gd[cg];
964 	max_bit = fs->e2fs_fpg;
965 	group_first_block = ((uint64_t)cg) * fs->e2fs_fpg +
966 	    le32toh(fs->e2fs->e2fs_first_dblock);
967 
968 	/* Check block bitmap block number */
969 	offset = e2fs_gd_get_b_bitmap(gd) - group_first_block;
970 	if (offset >= max_bit || !isset(bp->b_data, offset)) {
971 		SDT_PROBE2(ext2fs, , alloc, ext2_b_bitmap_validate_error,
972 		    "bad block bitmap, group", cg);
973 		return (EINVAL);
974 	}
975 
976 	/* Check inode bitmap block number */
977 	offset = e2fs_gd_get_i_bitmap(gd) - group_first_block;
978 	if (offset >= max_bit || !isset(bp->b_data, offset)) {
979 		SDT_PROBE2(ext2fs, , alloc, ext2_b_bitmap_validate_error,
980 		    "bad inode bitmap", cg);
981 		return (EINVAL);
982 	}
983 
984 	/* Check inode table */
985 	offset = e2fs_gd_get_i_tables(gd) - group_first_block;
986 	if (offset >= max_bit || offset + fs->e2fs_itpg >= max_bit) {
987 		SDT_PROBE2(ext2fs, , alloc, ext2_b_bitmap_validate_error,
988 		    "bad inode table, group", cg);
989 		return (EINVAL);
990 	}
991 
992 	return (0);
993 }
994 
995 /*
996  * Determine whether a block can be allocated.
997  *
998  * Check to see if a block of the appropriate size is available,
999  * and if it is, allocate it.
1000  */
1001 static daddr_t
1002 ext2_alloccg(struct inode *ip, int cg, daddr_t bpref, int size)
1003 {
1004 	struct m_ext2fs *fs;
1005 	struct buf *bp;
1006 	struct ext2mount *ump;
1007 	daddr_t bno, runstart, runlen;
1008 	int bit, loc, end, error, start;
1009 	char *bbp;
1010 	/* XXX ondisk32 */
1011 	fs = ip->i_e2fs;
1012 	ump = ip->i_ump;
1013 	if (e2fs_gd_get_nbfree(&fs->e2fs_gd[cg]) == 0)
1014 		return (0);
1015 
1016 	EXT2_UNLOCK(ump);
1017 	error = bread(ip->i_devvp, fsbtodb(fs,
1018 	    e2fs_gd_get_b_bitmap(&fs->e2fs_gd[cg])),
1019 	    (int)fs->e2fs_bsize, NOCRED, &bp);
1020 	if (error)
1021 		goto fail;
1022 
1023 	if (EXT2_HAS_RO_COMPAT_FEATURE(fs, EXT2F_ROCOMPAT_GDT_CSUM) ||
1024 	    EXT2_HAS_RO_COMPAT_FEATURE(fs, EXT2F_ROCOMPAT_METADATA_CKSUM)) {
1025 		error = ext2_cg_block_bitmap_init(fs, cg, bp);
1026 		if (error)
1027 			goto fail;
1028 
1029 		ext2_gd_b_bitmap_csum_set(fs, cg, bp);
1030 	}
1031 	error = ext2_gd_b_bitmap_csum_verify(fs, cg, bp);
1032 	if (error)
1033 		goto fail;
1034 
1035 	error = ext2_b_bitmap_validate(fs,bp, cg);
1036 	if (error)
1037 		goto fail;
1038 
1039 	/*
1040 	 * Check, that another thread did not not allocate the last block in
1041 	 * this group while we were waiting for the buffer.
1042 	 */
1043 	if (e2fs_gd_get_nbfree(&fs->e2fs_gd[cg]) == 0)
1044 		goto fail;
1045 
1046 	bbp = (char *)bp->b_data;
1047 
1048 	if (dtog(fs, bpref) != cg)
1049 		bpref = 0;
1050 	if (bpref != 0) {
1051 		bpref = dtogd(fs, bpref);
1052 		/*
1053 		 * if the requested block is available, use it
1054 		 */
1055 		if (isclr(bbp, bpref)) {
1056 			bno = bpref;
1057 			goto gotit;
1058 		}
1059 	}
1060 	/*
1061 	 * no blocks in the requested cylinder, so take next
1062 	 * available one in this cylinder group.
1063 	 * first try to get 8 contigous blocks, then fall back to a single
1064 	 * block.
1065 	 */
1066 	if (bpref)
1067 		start = dtogd(fs, bpref) / NBBY;
1068 	else
1069 		start = 0;
1070 	end = howmany(fs->e2fs_fpg, NBBY) - start;
1071 retry:
1072 	runlen = 0;
1073 	runstart = 0;
1074 	for (loc = start; loc < end; loc++) {
1075 		if (bbp[loc] == (char)0xff) {
1076 			runlen = 0;
1077 			continue;
1078 		}
1079 
1080 		/* Start of a run, find the number of high clear bits. */
1081 		if (runlen == 0) {
1082 			bit = fls(bbp[loc]);
1083 			runlen = NBBY - bit;
1084 			runstart = loc * NBBY + bit;
1085 		} else if (bbp[loc] == 0) {
1086 			/* Continue a run. */
1087 			runlen += NBBY;
1088 		} else {
1089 			/*
1090 			 * Finish the current run.  If it isn't long
1091 			 * enough, start a new one.
1092 			 */
1093 			bit = ffs(bbp[loc]) - 1;
1094 			runlen += bit;
1095 			if (runlen >= 8) {
1096 				bno = runstart;
1097 				goto gotit;
1098 			}
1099 
1100 			/* Run was too short, start a new one. */
1101 			bit = fls(bbp[loc]);
1102 			runlen = NBBY - bit;
1103 			runstart = loc * NBBY + bit;
1104 		}
1105 
1106 		/* If the current run is long enough, use it. */
1107 		if (runlen >= 8) {
1108 			bno = runstart;
1109 			goto gotit;
1110 		}
1111 	}
1112 	if (start != 0) {
1113 		end = start;
1114 		start = 0;
1115 		goto retry;
1116 	}
1117 	bno = ext2_mapsearch(fs, bbp, bpref);
1118 	if (bno < 0)
1119 		goto fail;
1120 
1121 gotit:
1122 #ifdef INVARIANTS
1123 	if (isset(bbp, bno)) {
1124 		printf("ext2fs_alloccgblk: cg=%d bno=%jd fs=%s\n",
1125 		    cg, (intmax_t)bno, fs->e2fs_fsmnt);
1126 		panic("ext2fs_alloccg: dup alloc");
1127 	}
1128 #endif
1129 	setbit(bbp, bno);
1130 	EXT2_LOCK(ump);
1131 	ext2_clusteracct(fs, bbp, cg, bno, -1);
1132 	fs->e2fs_fbcount--;
1133 	e2fs_gd_set_nbfree(&fs->e2fs_gd[cg],
1134 	    e2fs_gd_get_nbfree(&fs->e2fs_gd[cg]) - 1);
1135 	fs->e2fs_fmod = 1;
1136 	EXT2_UNLOCK(ump);
1137 	ext2_gd_b_bitmap_csum_set(fs, cg, bp);
1138 	bdwrite(bp);
1139 	return (((uint64_t)cg) * fs->e2fs_fpg +
1140 	    le32toh(fs->e2fs->e2fs_first_dblock) + bno);
1141 
1142 fail:
1143 	brelse(bp);
1144 	EXT2_LOCK(ump);
1145 	return (0);
1146 }
1147 
1148 /*
1149  * Determine whether a cluster can be allocated.
1150  */
1151 static daddr_t
1152 ext2_clusteralloc(struct inode *ip, int cg, daddr_t bpref, int len)
1153 {
1154 	struct m_ext2fs *fs;
1155 	struct ext2mount *ump;
1156 	struct buf *bp;
1157 	char *bbp;
1158 	int bit, error, got, i, loc, run;
1159 	int32_t *lp;
1160 	daddr_t bno;
1161 
1162 	fs = ip->i_e2fs;
1163 	ump = ip->i_ump;
1164 
1165 	if (fs->e2fs_maxcluster[cg] < len)
1166 		return (0);
1167 
1168 	EXT2_UNLOCK(ump);
1169 	error = bread(ip->i_devvp,
1170 	    fsbtodb(fs, e2fs_gd_get_b_bitmap(&fs->e2fs_gd[cg])),
1171 	    (int)fs->e2fs_bsize, NOCRED, &bp);
1172 	if (error)
1173 		goto fail_lock;
1174 
1175 	bbp = (char *)bp->b_data;
1176 	EXT2_LOCK(ump);
1177 	/*
1178 	 * Check to see if a cluster of the needed size (or bigger) is
1179 	 * available in this cylinder group.
1180 	 */
1181 	lp = &fs->e2fs_clustersum[cg].cs_sum[len];
1182 	for (i = len; i <= fs->e2fs_contigsumsize; i++)
1183 		if (*lp++ > 0)
1184 			break;
1185 	if (i > fs->e2fs_contigsumsize) {
1186 		/*
1187 		 * Update the cluster summary information to reflect
1188 		 * the true maximum-sized cluster so that future cluster
1189 		 * allocation requests can avoid reading the bitmap only
1190 		 * to find no cluster.
1191 		 */
1192 		lp = &fs->e2fs_clustersum[cg].cs_sum[len - 1];
1193 		for (i = len - 1; i > 0; i--)
1194 			if (*lp-- > 0)
1195 				break;
1196 		fs->e2fs_maxcluster[cg] = i;
1197 		goto fail;
1198 	}
1199 	EXT2_UNLOCK(ump);
1200 
1201 	/* Search the bitmap to find a big enough cluster like in FFS. */
1202 	if (dtog(fs, bpref) != cg)
1203 		bpref = 0;
1204 	if (bpref != 0)
1205 		bpref = dtogd(fs, bpref);
1206 	loc = bpref / NBBY;
1207 	bit = 1 << (bpref % NBBY);
1208 	for (run = 0, got = bpref; got < fs->e2fs_fpg; got++) {
1209 		if ((bbp[loc] & bit) != 0)
1210 			run = 0;
1211 		else {
1212 			run++;
1213 			if (run == len)
1214 				break;
1215 		}
1216 		if ((got & (NBBY - 1)) != (NBBY - 1))
1217 			bit <<= 1;
1218 		else {
1219 			loc++;
1220 			bit = 1;
1221 		}
1222 	}
1223 
1224 	if (got >= fs->e2fs_fpg)
1225 		goto fail_lock;
1226 
1227 	/* Allocate the cluster that we found. */
1228 	for (i = 1; i < len; i++)
1229 		if (!isclr(bbp, got - run + i))
1230 			panic("ext2_clusteralloc: map mismatch");
1231 
1232 	bno = got - run + 1;
1233 	if (bno >= fs->e2fs_fpg)
1234 		panic("ext2_clusteralloc: allocated out of group");
1235 
1236 	EXT2_LOCK(ump);
1237 	for (i = 0; i < len; i += fs->e2fs_fpb) {
1238 		setbit(bbp, bno + i);
1239 		ext2_clusteracct(fs, bbp, cg, bno + i, -1);
1240 		fs->e2fs_fbcount--;
1241 		e2fs_gd_set_nbfree(&fs->e2fs_gd[cg],
1242 		    e2fs_gd_get_nbfree(&fs->e2fs_gd[cg]) - 1);
1243 	}
1244 	fs->e2fs_fmod = 1;
1245 	EXT2_UNLOCK(ump);
1246 
1247 	bdwrite(bp);
1248 	return (cg * fs->e2fs_fpg + le32toh(fs->e2fs->e2fs_first_dblock)
1249 	    + bno);
1250 
1251 fail_lock:
1252 	EXT2_LOCK(ump);
1253 fail:
1254 	brelse(bp);
1255 	return (0);
1256 }
1257 
1258 static int
1259 ext2_zero_inode_table(struct inode *ip, int cg)
1260 {
1261 	struct m_ext2fs *fs;
1262 	struct buf *bp;
1263 	int i, all_blks, used_blks;
1264 
1265 	fs = ip->i_e2fs;
1266 
1267 	if (le16toh(fs->e2fs_gd[cg].ext4bgd_flags) & EXT2_BG_INODE_ZEROED)
1268 		return (0);
1269 
1270 	all_blks = le16toh(fs->e2fs->e2fs_inode_size) * fs->e2fs_ipg /
1271 	    fs->e2fs_bsize;
1272 
1273 	used_blks = howmany(fs->e2fs_ipg -
1274 	    e2fs_gd_get_i_unused(&fs->e2fs_gd[cg]),
1275 	    fs->e2fs_bsize / EXT2_INODE_SIZE(fs));
1276 
1277 	for (i = 0; i < all_blks - used_blks; i++) {
1278 		bp = getblk(ip->i_devvp, fsbtodb(fs,
1279 		    e2fs_gd_get_i_tables(&fs->e2fs_gd[cg]) + used_blks + i),
1280 		    fs->e2fs_bsize, 0, 0, 0);
1281 		if (!bp)
1282 			return (EIO);
1283 
1284 		vfs_bio_bzero_buf(bp, 0, fs->e2fs_bsize);
1285 		bawrite(bp);
1286 	}
1287 
1288 	fs->e2fs_gd[cg].ext4bgd_flags = htole16(le16toh(
1289 	    fs->e2fs_gd[cg].ext4bgd_flags) | EXT2_BG_INODE_ZEROED);
1290 
1291 	return (0);
1292 }
1293 
1294 static void
1295 ext2_fix_bitmap_tail(unsigned char *bitmap, int first, int last)
1296 {
1297 	int i;
1298 
1299 	for (i = first; i <= last; i++)
1300 		bitmap[i] = 0xff;
1301 }
1302 
1303 /*
1304  * Determine whether an inode can be allocated.
1305  *
1306  * Check to see if an inode is available, and if it is,
1307  * allocate it using tode in the specified cylinder group.
1308  */
1309 static daddr_t
1310 ext2_nodealloccg(struct inode *ip, int cg, daddr_t ipref, int mode)
1311 {
1312 	struct m_ext2fs *fs;
1313 	struct buf *bp;
1314 	struct ext2mount *ump;
1315 	int error, start, len, ifree, ibytes;
1316 	char *ibp, *loc;
1317 
1318 	ipref--;	/* to avoid a lot of (ipref -1) */
1319 	if (ipref == -1)
1320 		ipref = 0;
1321 	fs = ip->i_e2fs;
1322 	ump = ip->i_ump;
1323 	if (e2fs_gd_get_nifree(&fs->e2fs_gd[cg]) == 0)
1324 		return (0);
1325 	EXT2_UNLOCK(ump);
1326 	error = bread(ip->i_devvp, fsbtodb(fs,
1327 	    e2fs_gd_get_i_bitmap(&fs->e2fs_gd[cg])),
1328 	    (int)fs->e2fs_bsize, NOCRED, &bp);
1329 	if (error) {
1330 		EXT2_LOCK(ump);
1331 		return (0);
1332 	}
1333 	if (EXT2_HAS_RO_COMPAT_FEATURE(fs, EXT2F_ROCOMPAT_GDT_CSUM) ||
1334 	    EXT2_HAS_RO_COMPAT_FEATURE(fs, EXT2F_ROCOMPAT_METADATA_CKSUM)) {
1335 		if (le16toh(fs->e2fs_gd[cg].ext4bgd_flags) &
1336 		    EXT2_BG_INODE_UNINIT) {
1337 			ibytes = fs->e2fs_ipg / 8;
1338 			memset(bp->b_data, 0, ibytes - 1);
1339 			ext2_fix_bitmap_tail(bp->b_data, ibytes,
1340 			    fs->e2fs_bsize - 1);
1341 			fs->e2fs_gd[cg].ext4bgd_flags = htole16(le16toh(
1342 			    fs->e2fs_gd[cg].ext4bgd_flags) &
1343 			    ~EXT2_BG_INODE_UNINIT);
1344 		}
1345 		ext2_gd_i_bitmap_csum_set(fs, cg, bp);
1346 		error = ext2_zero_inode_table(ip, cg);
1347 		if (error) {
1348 			brelse(bp);
1349 			EXT2_LOCK(ump);
1350 			return (0);
1351 		}
1352 	}
1353 	error = ext2_gd_i_bitmap_csum_verify(fs, cg, bp);
1354 	if (error) {
1355 		brelse(bp);
1356 		EXT2_LOCK(ump);
1357 		return (0);
1358 	}
1359 	if (e2fs_gd_get_nifree(&fs->e2fs_gd[cg]) == 0) {
1360 		/*
1361 		 * Another thread allocated the last i-node in this
1362 		 * group while we were waiting for the buffer.
1363 		 */
1364 		brelse(bp);
1365 		EXT2_LOCK(ump);
1366 		return (0);
1367 	}
1368 	ibp = (char *)bp->b_data;
1369 	if (ipref) {
1370 		ipref %= fs->e2fs_ipg;
1371 		if (isclr(ibp, ipref))
1372 			goto gotit;
1373 	}
1374 	start = ipref / NBBY;
1375 	len = howmany(fs->e2fs_ipg - ipref, NBBY);
1376 	loc = memcchr(&ibp[start], 0xff, len);
1377 	if (loc == NULL) {
1378 		len = start + 1;
1379 		start = 0;
1380 		loc = memcchr(&ibp[start], 0xff, len);
1381 		if (loc == NULL) {
1382 			SDT_PROBE3(ext2fs, , alloc,
1383 			    ext2_nodealloccg_bmap_corrupted, cg, ipref,
1384 			    fs->e2fs_fsmnt);
1385 			brelse(bp);
1386 			EXT2_LOCK(ump);
1387 			return (0);
1388 		}
1389 	}
1390 	ipref = (loc - ibp) * NBBY + ffs(~*loc) - 1;
1391 gotit:
1392 	setbit(ibp, ipref);
1393 	EXT2_LOCK(ump);
1394 	e2fs_gd_set_nifree(&fs->e2fs_gd[cg],
1395 	    e2fs_gd_get_nifree(&fs->e2fs_gd[cg]) - 1);
1396 	if (EXT2_HAS_RO_COMPAT_FEATURE(fs, EXT2F_ROCOMPAT_GDT_CSUM) ||
1397 	    EXT2_HAS_RO_COMPAT_FEATURE(fs, EXT2F_ROCOMPAT_METADATA_CKSUM)) {
1398 		ifree = fs->e2fs_ipg - e2fs_gd_get_i_unused(&fs->e2fs_gd[cg]);
1399 		if (ipref + 1 > ifree)
1400 			e2fs_gd_set_i_unused(&fs->e2fs_gd[cg],
1401 			    fs->e2fs_ipg - (ipref + 1));
1402 	}
1403 	fs->e2fs_ficount--;
1404 	fs->e2fs_fmod = 1;
1405 	if ((mode & IFMT) == IFDIR) {
1406 		e2fs_gd_set_ndirs(&fs->e2fs_gd[cg],
1407 		    e2fs_gd_get_ndirs(&fs->e2fs_gd[cg]) + 1);
1408 		fs->e2fs_total_dir++;
1409 	}
1410 	EXT2_UNLOCK(ump);
1411 	ext2_gd_i_bitmap_csum_set(fs, cg, bp);
1412 	bdwrite(bp);
1413 	return ((uint64_t)cg * fs->e2fs_ipg + ipref + 1);
1414 }
1415 
1416 /*
1417  * Free a block or fragment.
1418  *
1419  */
1420 void
1421 ext2_blkfree(struct inode *ip, e4fs_daddr_t bno, long size)
1422 {
1423 	struct m_ext2fs *fs;
1424 	struct buf *bp;
1425 	struct ext2mount *ump;
1426 	int cg, error;
1427 	char *bbp;
1428 
1429 	fs = ip->i_e2fs;
1430 	ump = ip->i_ump;
1431 	cg = dtog(fs, bno);
1432 	if (bno >= fs->e2fs_bcount) {
1433 		SDT_PROBE2(ext2fs, , alloc, ext2_blkfree_bad_block,
1434 		    ip->i_number, bno);
1435 		return;
1436 	}
1437 	error = bread(ip->i_devvp,
1438 	    fsbtodb(fs, e2fs_gd_get_b_bitmap(&fs->e2fs_gd[cg])),
1439 	    (int)fs->e2fs_bsize, NOCRED, &bp);
1440 	if (error) {
1441 		return;
1442 	}
1443 	bbp = (char *)bp->b_data;
1444 	bno = dtogd(fs, bno);
1445 	if (isclr(bbp, bno)) {
1446 		panic("ext2_blkfree: freeing free block %lld, fs=%s",
1447 		    (long long)bno, fs->e2fs_fsmnt);
1448 	}
1449 	clrbit(bbp, bno);
1450 	EXT2_LOCK(ump);
1451 	ext2_clusteracct(fs, bbp, cg, bno, 1);
1452 	fs->e2fs_fbcount++;
1453 	e2fs_gd_set_nbfree(&fs->e2fs_gd[cg],
1454 	    e2fs_gd_get_nbfree(&fs->e2fs_gd[cg]) + 1);
1455 	fs->e2fs_fmod = 1;
1456 	EXT2_UNLOCK(ump);
1457 	ext2_gd_b_bitmap_csum_set(fs, cg, bp);
1458 	bdwrite(bp);
1459 }
1460 
1461 /*
1462  * Free an inode.
1463  *
1464  */
1465 int
1466 ext2_vfree(struct vnode *pvp, ino_t ino, int mode)
1467 {
1468 	struct m_ext2fs *fs;
1469 	struct inode *pip;
1470 	struct buf *bp;
1471 	struct ext2mount *ump;
1472 	int error, cg;
1473 	char *ibp;
1474 
1475 	pip = VTOI(pvp);
1476 	fs = pip->i_e2fs;
1477 	ump = pip->i_ump;
1478 	if ((u_int)ino > fs->e2fs_ipg * fs->e2fs_gcount)
1479 		panic("ext2_vfree: range: devvp = %p, ino = %ju, fs = %s",
1480 		    pip->i_devvp, (uintmax_t)ino, fs->e2fs_fsmnt);
1481 
1482 	cg = ino_to_cg(fs, ino);
1483 	error = bread(pip->i_devvp,
1484 	    fsbtodb(fs, e2fs_gd_get_i_bitmap(&fs->e2fs_gd[cg])),
1485 	    (int)fs->e2fs_bsize, NOCRED, &bp);
1486 	if (error) {
1487 		return (0);
1488 	}
1489 	ibp = (char *)bp->b_data;
1490 	ino = (ino - 1) % fs->e2fs_ipg;
1491 	if (isclr(ibp, ino)) {
1492 		SDT_PROBE2(ext2fs, , alloc, ext2_vfree_doublefree,
1493 		    fs->e2fs_fsmnt, ino);
1494 		if (fs->e2fs_ronly == 0)
1495 			panic("ext2_vfree: freeing free inode");
1496 	}
1497 	clrbit(ibp, ino);
1498 	EXT2_LOCK(ump);
1499 	fs->e2fs_ficount++;
1500 	e2fs_gd_set_nifree(&fs->e2fs_gd[cg],
1501 	    e2fs_gd_get_nifree(&fs->e2fs_gd[cg]) + 1);
1502 	if ((mode & IFMT) == IFDIR) {
1503 		e2fs_gd_set_ndirs(&fs->e2fs_gd[cg],
1504 		    e2fs_gd_get_ndirs(&fs->e2fs_gd[cg]) - 1);
1505 		fs->e2fs_total_dir--;
1506 	}
1507 	fs->e2fs_fmod = 1;
1508 	EXT2_UNLOCK(ump);
1509 	ext2_gd_i_bitmap_csum_set(fs, cg, bp);
1510 	bdwrite(bp);
1511 	return (0);
1512 }
1513 
1514 /*
1515  * Find a block in the specified cylinder group.
1516  *
1517  * It is a panic if a request is made to find a block if none are
1518  * available.
1519  */
1520 static daddr_t
1521 ext2_mapsearch(struct m_ext2fs *fs, char *bbp, daddr_t bpref)
1522 {
1523 	char *loc;
1524 	int start, len;
1525 
1526 	/*
1527 	 * find the fragment by searching through the free block
1528 	 * map for an appropriate bit pattern
1529 	 */
1530 	if (bpref)
1531 		start = dtogd(fs, bpref) / NBBY;
1532 	else
1533 		start = 0;
1534 	len = howmany(fs->e2fs_fpg, NBBY) - start;
1535 	loc = memcchr(&bbp[start], 0xff, len);
1536 	if (loc == NULL) {
1537 		len = start + 1;
1538 		start = 0;
1539 		loc = memcchr(&bbp[start], 0xff, len);
1540 		if (loc == NULL) {
1541 			panic("ext2_mapsearch: map corrupted: start=%d, len=%d,"
1542 			    "fs=%s", start, len, fs->e2fs_fsmnt);
1543 			/* NOTREACHED */
1544 		}
1545 	}
1546 	return ((loc - bbp) * NBBY + ffs(~*loc) - 1);
1547 }
1548 
1549 int
1550 ext2_cg_has_sb(struct m_ext2fs *fs, int cg)
1551 {
1552 	int a3, a5, a7;
1553 
1554 	if (cg == 0)
1555 		return (1);
1556 
1557 	if (EXT2_HAS_COMPAT_FEATURE(fs, EXT2F_COMPAT_SPARSESUPER2)) {
1558 		if (cg == le32toh(fs->e2fs->e4fs_backup_bgs[0]) ||
1559 		    cg == le32toh(fs->e2fs->e4fs_backup_bgs[1]))
1560 			return (1);
1561 		return (0);
1562 	}
1563 
1564 	if ((cg <= 1) ||
1565 	    !EXT2_HAS_RO_COMPAT_FEATURE(fs, EXT2F_ROCOMPAT_SPARSESUPER))
1566 		return (1);
1567 
1568 	if (!(cg & 1))
1569 		return (0);
1570 
1571 	for (a3 = 3, a5 = 5, a7 = 7;
1572 	    a3 <= cg || a5 <= cg || a7 <= cg;
1573 	    a3 *= 3, a5 *= 5, a7 *= 7)
1574 		if (cg == a3 || cg == a5 || cg == a7)
1575 			return (1);
1576 	return (0);
1577 }
1578