2 * modified for Lites 1.1
4 * Aug 1995, Godmar Back (gback@cs.utah.edu)
5 * University of Utah, Department of Computer Science
8 * Copyright (c) 1982, 1986, 1989, 1993
9 * The Regents of the University of California. All rights reserved.
11 * Redistribution and use in source and binary forms, with or without
12 * modification, are permitted provided that the following conditions
14 * 1. Redistributions of source code must retain the above copyright
15 * notice, this list of conditions and the following disclaimer.
16 * 2. Redistributions in binary form must reproduce the above copyright
17 * notice, this list of conditions and the following disclaimer in the
18 * documentation and/or other materials provided with the distribution.
19 * 3. Neither the name of the University nor the names of its contributors
20 * may be used to endorse or promote products derived from this software
21 * without specific prior written permission.
23 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
24 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
25 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
26 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
27 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
28 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
29 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
30 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
31 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
32 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
35 * @(#)ffs_alloc.c 8.8 (Berkeley) 2/21/94
39 #include <sys/param.h>
40 #include <sys/systm.h>
42 #include <sys/vnode.h>
44 #include <sys/mount.h>
45 #include <sys/sysctl.h>
46 #include <sys/syslog.h>
48 #include <sys/endian.h>
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>
56 static daddr_t ext2_alloccg(struct inode *, int, daddr_t, int);
57 static daddr_t ext2_clusteralloc(struct inode *, int, daddr_t, int);
58 static u_long ext2_dirpref(struct inode *);
59 static e4fs_daddr_t ext2_hashalloc(struct inode *, int, long, int,
60 daddr_t (*)(struct inode *, int, daddr_t,
62 static daddr_t ext2_nodealloccg(struct inode *, int, daddr_t, int);
63 static daddr_t ext2_mapsearch(struct m_ext2fs *, char *, daddr_t);
66 * Allocate a block in the filesystem.
68 * A preference may be optionally specified. If a preference is given
69 * the following hierarchy is used to allocate a block:
70 * 1) allocate the requested block.
71 * 2) allocate a rotationally optimal block in the same cylinder.
72 * 3) allocate a block in the same cylinder group.
73 * 4) quadradically rehash into other cylinder groups, until an
74 * available block is located.
75 * If no block preference is given the following hierarchy is used
76 * to allocate a block:
77 * 1) allocate a block in the cylinder group that contains the
79 * 2) quadradically rehash into other cylinder groups, until an
80 * available block is located.
83 ext2_alloc(struct inode *ip, daddr_t lbn, e4fs_daddr_t bpref, int size,
84 struct ucred *cred, e4fs_daddr_t *bnp)
87 struct ext2mount *ump;
94 mtx_assert(EXT2_MTX(ump), MA_OWNED);
96 if ((u_int)size > fs->e2fs_bsize || blkoff(fs, size) != 0) {
97 vn_printf(ip->i_devvp, "bsize = %lu, size = %d, fs = %s\n",
98 (long unsigned int)fs->e2fs_bsize, size, fs->e2fs_fsmnt);
99 panic("ext2_alloc: bad size");
102 panic("ext2_alloc: missing credential");
103 #endif /* INVARIANTS */
104 if (size == fs->e2fs_bsize && fs->e2fs->e2fs_fbcount == 0)
106 if (cred->cr_uid != 0 &&
107 fs->e2fs->e2fs_fbcount < fs->e2fs->e2fs_rbcount)
109 if (bpref >= fs->e2fs->e2fs_bcount)
112 cg = ino_to_cg(fs, ip->i_number);
114 cg = dtog(fs, bpref);
115 bno = (daddr_t)ext2_hashalloc(ip, cg, bpref, fs->e2fs_bsize,
118 /* set next_alloc fields as done in block_getblk */
119 ip->i_next_alloc_block = lbn;
120 ip->i_next_alloc_goal = bno;
122 ip->i_blocks += btodb(fs->e2fs_bsize);
123 ip->i_flag |= IN_CHANGE | IN_UPDATE;
129 ext2_fserr(fs, cred->cr_uid, "filesystem full");
130 uprintf("\n%s: write failed, filesystem is full\n", fs->e2fs_fsmnt);
135 * Allocate EA's block for inode.
138 ext2_alloc_meta(struct inode *ip)
145 EXT2_LOCK(ip->i_ump);
146 blk = ext2_hashalloc(ip, ino_to_cg(fs, ip->i_number), 0, fs->e2fs_bsize,
149 EXT2_UNLOCK(ip->i_ump);
155 * Reallocate a sequence of blocks into a contiguous sequence of blocks.
157 * The vnode and an array of buffer pointers for a range of sequential
158 * logical blocks to be made contiguous is given. The allocator attempts
159 * to find a range of sequential blocks starting as close as possible to
160 * an fs_rotdelay offset from the end of the allocation for the logical
161 * block immediately preceding the current range. If successful, the
162 * physical block numbers in the buffer pointers and in the inode are
163 * changed to reflect the new allocation. If unsuccessful, the allocation
164 * is left unchanged. The success in doing the reallocation is returned.
165 * Note that the error return is not reflected back to the user. Rather
166 * the previous block allocation will be used.
169 static SYSCTL_NODE(_vfs, OID_AUTO, ext2fs, CTLFLAG_RW, 0, "EXT2FS filesystem");
171 static int doasyncfree = 1;
173 SYSCTL_INT(_vfs_ext2fs, OID_AUTO, doasyncfree, CTLFLAG_RW, &doasyncfree, 0,
174 "Use asychronous writes to update block pointers when freeing blocks");
176 static int doreallocblks = 0;
178 SYSCTL_INT(_vfs_ext2fs, OID_AUTO, doreallocblks, CTLFLAG_RW, &doreallocblks, 0, "");
181 ext2_reallocblks(struct vop_reallocblks_args *ap)
186 struct buf *sbp, *ebp;
187 uint32_t *bap, *sbap, *ebap;
188 struct ext2mount *ump;
189 struct cluster_save *buflist;
190 struct indir start_ap[EXT2_NIADDR + 1], end_ap[EXT2_NIADDR + 1], *idp;
191 e2fs_lbn_t start_lbn, end_lbn;
193 e2fs_daddr_t newblk, blkno;
194 int i, len, start_lvl, end_lvl, pref, ssize;
196 if (doreallocblks == 0)
204 if (fs->e2fs_contigsumsize <= 0 || ip->i_flag & IN_E4EXTENTS)
207 buflist = ap->a_buflist;
208 len = buflist->bs_nchildren;
209 start_lbn = buflist->bs_children[0]->b_lblkno;
210 end_lbn = start_lbn + len - 1;
212 for (i = 1; i < len; i++)
213 if (buflist->bs_children[i]->b_lblkno != start_lbn + i)
214 panic("ext2_reallocblks: non-cluster");
217 * If the cluster crosses the boundary for the first indirect
218 * block, leave space for the indirect block. Indirect blocks
219 * are initially laid out in a position after the last direct
220 * block. Block reallocation would usually destroy locality by
221 * moving the indirect block out of the way to make room for
222 * data blocks if we didn't compensate here. We should also do
223 * this for other indirect block boundaries, but it is only
224 * important for the first one.
226 if (start_lbn < EXT2_NDADDR && end_lbn >= EXT2_NDADDR)
229 * If the latest allocation is in a new cylinder group, assume that
230 * the filesystem has decided to move and do not force it back to
231 * the previous cylinder group.
233 if (dtog(fs, dbtofsb(fs, buflist->bs_children[0]->b_blkno)) !=
234 dtog(fs, dbtofsb(fs, buflist->bs_children[len - 1]->b_blkno)))
236 if (ext2_getlbns(vp, start_lbn, start_ap, &start_lvl) ||
237 ext2_getlbns(vp, end_lbn, end_ap, &end_lvl))
240 * Get the starting offset and block map for the first block.
242 if (start_lvl == 0) {
246 idp = &start_ap[start_lvl - 1];
247 if (bread(vp, idp->in_lbn, (int)fs->e2fs_bsize, NOCRED, &sbp)) {
251 sbap = (u_int *)sbp->b_data;
255 * If the block range spans two block maps, get the second map.
258 if (end_lvl == 0 || (idp = &end_ap[end_lvl - 1])->in_off + 1 >= len) {
262 if (start_ap[start_lvl - 1].in_lbn == idp->in_lbn)
263 panic("ext2_reallocblks: start == end");
265 ssize = len - (idp->in_off + 1);
266 if (bread(vp, idp->in_lbn, (int)fs->e2fs_bsize, NOCRED, &ebp))
268 ebap = (u_int *)ebp->b_data;
271 * Find the preferred location for the cluster.
274 pref = ext2_blkpref(ip, start_lbn, soff, sbap, 0);
276 * Search the block map looking for an allocation of the desired size.
278 if ((newblk = (e2fs_daddr_t)ext2_hashalloc(ip, dtog(fs, pref), pref,
279 len, ext2_clusteralloc)) == 0) {
284 * We have found a new contiguous block.
286 * First we have to replace the old block pointers with the new
287 * block pointers in the inode and indirect blocks associated
291 printf("realloc: ino %ju, lbns %jd-%jd\n\told:",
292 (uintmax_t)ip->i_number, (intmax_t)start_lbn, (intmax_t)end_lbn);
295 for (bap = &sbap[soff], i = 0; i < len; i++, blkno += fs->e2fs_fpb) {
301 if (buflist->bs_children[i]->b_blkno != fsbtodb(fs, *bap))
302 panic("ext2_reallocblks: alloc mismatch");
305 printf(" %d,", *bap);
310 * Next we must write out the modified inode and indirect blocks.
311 * For strict correctness, the writes should be synchronous since
312 * the old block values may have been written to disk. In practise
313 * they are almost never written, but if we are concerned about
314 * strict correctness, the `doasyncfree' flag should be set to zero.
316 * The test on `doasyncfree' should be changed to test a flag
317 * that shows whether the associated buffers and inodes have
318 * been written. The flag should be set when the cluster is
319 * started and cleared whenever the buffer or inode is flushed.
320 * We can then check below to see if it is set, and do the
321 * synchronous write only when it has been cleared.
323 if (sbap != &ip->i_db[0]) {
329 ip->i_flag |= IN_CHANGE | IN_UPDATE;
340 * Last, free the old blocks and assign the new blocks to the buffers.
345 for (blkno = newblk, i = 0; i < len; i++, blkno += fs->e2fs_fpb) {
346 ext2_blkfree(ip, dbtofsb(fs, buflist->bs_children[i]->b_blkno),
348 buflist->bs_children[i]->b_blkno = fsbtodb(fs, blkno);
350 printf(" %d,", blkno);
361 if (sbap != &ip->i_db[0])
367 * Allocate an inode in the filesystem.
371 ext2_valloc(struct vnode *pvp, int mode, struct ucred *cred, struct vnode **vpp)
377 struct ext2mount *ump;
387 if (fs->e2fs->e2fs_ficount == 0)
390 * If it is a directory then obtain a cylinder group based on
391 * ext2_dirpref else obtain it using ino_to_cg. The preferred inode is
392 * always the next inode.
394 if ((mode & IFMT) == IFDIR) {
395 cg = ext2_dirpref(pip);
396 if (fs->e2fs_contigdirs[cg] < 255)
397 fs->e2fs_contigdirs[cg]++;
399 cg = ino_to_cg(fs, pip->i_number);
400 if (fs->e2fs_contigdirs[cg] > 0)
401 fs->e2fs_contigdirs[cg]--;
403 ipref = cg * fs->e2fs->e2fs_ipg + 1;
404 ino = (ino_t)ext2_hashalloc(pip, cg, (long)ipref, mode, ext2_nodealloccg);
408 error = VFS_VGET(pvp->v_mount, ino, LK_EXCLUSIVE, vpp);
410 ext2_vfree(pvp, ino, mode);
416 * The question is whether using VGET was such good idea at all:
417 * Linux doesn't read the old inode in when it is allocating a
418 * new one. I will set at least i_size and i_blocks to zero.
425 if (EXT2_HAS_INCOMPAT_FEATURE(fs, EXT2F_INCOMPAT_EXTENTS)
426 && (S_ISREG(mode) || S_ISDIR(mode)))
427 ext4_ext_tree_init(ip);
429 memset(ip->i_data, 0, sizeof(ip->i_data));
433 * Set up a new generation number for this inode.
437 ip->i_gen = arc4random();
438 } while (ip->i_gen == 0);
441 ip->i_birthtime = ts.tv_sec;
442 ip->i_birthnsec = ts.tv_nsec;
445 printf("ext2_valloc: allocated inode %d\n", ino);
450 ext2_fserr(fs, cred->cr_uid, "out of inodes");
451 uprintf("\n%s: create/symlink failed, no inodes free\n", fs->e2fs_fsmnt);
456 * Find a cylinder to place a directory.
458 * The policy implemented by this algorithm is to allocate a
459 * directory inode in the same cylinder group as its parent
460 * directory, but also to reserve space for its files inodes
461 * and data. Restrict the number of directories which may be
462 * allocated one after another in the same cylinder group
463 * without intervening allocation of files.
465 * If we allocate a first level directory then force allocation
466 * in another cylinder group.
470 ext2_dirpref(struct inode *pip)
473 int cg, prefcg, cgsize;
474 u_int avgifree, avgbfree, avgndir, curdirsize;
475 u_int minifree, minbfree, maxndir;
476 u_int mincg, minndir;
477 u_int dirsize, maxcontigdirs;
479 mtx_assert(EXT2_MTX(pip->i_ump), MA_OWNED);
482 avgifree = fs->e2fs->e2fs_ficount / fs->e2fs_gcount;
483 avgbfree = fs->e2fs->e2fs_fbcount / fs->e2fs_gcount;
484 avgndir = fs->e2fs_total_dir / fs->e2fs_gcount;
487 * Force allocation in another cg if creating a first level dir.
489 ASSERT_VOP_LOCKED(ITOV(pip), "ext2fs_dirpref");
490 if (ITOV(pip)->v_vflag & VV_ROOT) {
491 prefcg = arc4random() % fs->e2fs_gcount;
493 minndir = fs->e2fs_ipg;
494 for (cg = prefcg; cg < fs->e2fs_gcount; cg++)
495 if (fs->e2fs_gd[cg].ext2bgd_ndirs < minndir &&
496 fs->e2fs_gd[cg].ext2bgd_nifree >= avgifree &&
497 fs->e2fs_gd[cg].ext2bgd_nbfree >= avgbfree) {
499 minndir = fs->e2fs_gd[cg].ext2bgd_ndirs;
501 for (cg = 0; cg < prefcg; cg++)
502 if (fs->e2fs_gd[cg].ext2bgd_ndirs < minndir &&
503 fs->e2fs_gd[cg].ext2bgd_nifree >= avgifree &&
504 fs->e2fs_gd[cg].ext2bgd_nbfree >= avgbfree) {
506 minndir = fs->e2fs_gd[cg].ext2bgd_ndirs;
511 * Count various limits which used for
512 * optimal allocation of a directory inode.
514 maxndir = min(avgndir + fs->e2fs_ipg / 16, fs->e2fs_ipg);
515 minifree = avgifree - avgifree / 4;
518 minbfree = avgbfree - avgbfree / 4;
521 cgsize = fs->e2fs_fsize * fs->e2fs_fpg;
522 dirsize = AVGDIRSIZE;
523 curdirsize = avgndir ? (cgsize - avgbfree * fs->e2fs_bsize) / avgndir : 0;
524 if (dirsize < curdirsize)
525 dirsize = curdirsize;
526 maxcontigdirs = min((avgbfree * fs->e2fs_bsize) / dirsize, 255);
527 maxcontigdirs = min(maxcontigdirs, fs->e2fs_ipg / AFPDIR);
528 if (maxcontigdirs == 0)
532 * Limit number of dirs in one cg and reserve space for
533 * regular files, but only if we have no deficit in
536 prefcg = ino_to_cg(fs, pip->i_number);
537 for (cg = prefcg; cg < fs->e2fs_gcount; cg++)
538 if (fs->e2fs_gd[cg].ext2bgd_ndirs < maxndir &&
539 fs->e2fs_gd[cg].ext2bgd_nifree >= minifree &&
540 fs->e2fs_gd[cg].ext2bgd_nbfree >= minbfree) {
541 if (fs->e2fs_contigdirs[cg] < maxcontigdirs)
544 for (cg = 0; cg < prefcg; cg++)
545 if (fs->e2fs_gd[cg].ext2bgd_ndirs < maxndir &&
546 fs->e2fs_gd[cg].ext2bgd_nifree >= minifree &&
547 fs->e2fs_gd[cg].ext2bgd_nbfree >= minbfree) {
548 if (fs->e2fs_contigdirs[cg] < maxcontigdirs)
552 * This is a backstop when we have deficit in space.
554 for (cg = prefcg; cg < fs->e2fs_gcount; cg++)
555 if (fs->e2fs_gd[cg].ext2bgd_nifree >= avgifree)
557 for (cg = 0; cg < prefcg; cg++)
558 if (fs->e2fs_gd[cg].ext2bgd_nifree >= avgifree)
564 * Select the desired position for the next block in a file.
566 * we try to mimic what Remy does in inode_getblk/block_getblk
568 * we note: blocknr == 0 means that we're about to allocate either
569 * a direct block or a pointer block at the first level of indirection
570 * (In other words, stuff that will go in i_db[] or i_ib[])
572 * blocknr != 0 means that we're allocating a block that is none
573 * of the above. Then, blocknr tells us the number of the block
574 * that will hold the pointer
577 ext2_blkpref(struct inode *ip, e2fs_lbn_t lbn, int indx, e2fs_daddr_t *bap,
578 e2fs_daddr_t blocknr)
585 mtx_assert(EXT2_MTX(ip->i_ump), MA_OWNED);
588 * If the next block is actually what we thought it is, then set the
589 * goal to what we thought it should be.
591 if (ip->i_next_alloc_block == lbn && ip->i_next_alloc_goal != 0)
592 return ip->i_next_alloc_goal;
595 * Now check whether we were provided with an array that basically
596 * tells us previous blocks to which we want to stay close.
599 for (tmp = indx - 1; tmp >= 0; tmp--)
604 * Else lets fall back to the blocknr or, if there is none, follow
605 * the rule that a block should be allocated near its inode.
607 return (blocknr ? blocknr :
608 (e2fs_daddr_t)(ip->i_block_group *
609 EXT2_BLOCKS_PER_GROUP(fs)) + fs->e2fs->e2fs_first_dblock);
613 * Implement the cylinder overflow algorithm.
615 * The policy implemented by this algorithm is:
616 * 1) allocate the block in its requested cylinder group.
617 * 2) quadradically rehash on the cylinder group number.
618 * 3) brute force search for a free block.
621 ext2_hashalloc(struct inode *ip, int cg, long pref, int size,
622 daddr_t (*allocator) (struct inode *, int, daddr_t, int))
628 mtx_assert(EXT2_MTX(ip->i_ump), MA_OWNED);
631 * 1: preferred cylinder group
633 result = (*allocator)(ip, cg, pref, size);
637 * 2: quadratic rehash
639 for (i = 1; i < fs->e2fs_gcount; i *= 2) {
641 if (cg >= fs->e2fs_gcount)
642 cg -= fs->e2fs_gcount;
643 result = (*allocator)(ip, cg, 0, size);
648 * 3: brute force search
649 * Note that we start at i == 2, since 0 was checked initially,
650 * and 1 is always checked in the quadratic rehash.
652 cg = (icg + 2) % fs->e2fs_gcount;
653 for (i = 2; i < fs->e2fs_gcount; i++) {
654 result = (*allocator)(ip, cg, 0, size);
658 if (cg == fs->e2fs_gcount)
665 ext2_cg_num_gdb(struct m_ext2fs *fs, int cg)
667 int gd_per_block, metagroup, first, last;
669 gd_per_block = fs->e2fs_bsize / sizeof(struct ext2_gd);
670 metagroup = cg / gd_per_block;
671 first = metagroup * gd_per_block;
672 last = first + gd_per_block - 1;
674 if (!EXT2_HAS_INCOMPAT_FEATURE(fs, EXT2F_INCOMPAT_META_BG) ||
675 metagroup < fs->e2fs->e3fs_first_meta_bg) {
676 if (!ext2_cg_has_sb(fs, cg))
678 if (EXT2_HAS_INCOMPAT_FEATURE(fs, EXT2F_INCOMPAT_META_BG))
679 return (fs->e2fs->e3fs_first_meta_bg);
680 return (fs->e2fs_gdbcount);
683 if (cg == first || cg == first + 1 || cg == last)
690 ext2_num_base_meta_blocks(struct m_ext2fs *fs, int cg)
692 int num, gd_per_block;
694 gd_per_block = fs->e2fs_bsize / sizeof(struct ext2_gd);
695 num = ext2_cg_has_sb(fs, cg);
697 if (!EXT2_HAS_INCOMPAT_FEATURE(fs, EXT2F_INCOMPAT_META_BG) ||
698 cg < fs->e2fs->e3fs_first_meta_bg * gd_per_block) {
700 num += ext2_cg_num_gdb(fs, cg);
701 num += fs->e2fs->e2fs_reserved_ngdb;
704 num += ext2_cg_num_gdb(fs, cg);
711 ext2_get_cg_number(struct m_ext2fs *fs, daddr_t blk)
715 if (fs->e2fs->e2fs_bpg == fs->e2fs_bsize * 8)
716 cg = (blk - fs->e2fs->e2fs_first_dblock) / (fs->e2fs_bsize * 8);
718 cg = blk - fs->e2fs->e2fs_first_dblock;
724 ext2_mark_bitmap_end(int start_bit, int end_bit, char *bitmap)
728 if (start_bit >= end_bit)
731 for (i = start_bit; i < ((start_bit + 7) & ~7UL); i++)
734 memset(bitmap + (i >> 3), 0xff, (end_bit - i) >> 3);
738 ext2_cg_block_bitmap_init(struct m_ext2fs *fs, int cg, struct buf *bp)
740 int bit, bit_max, inodes_per_block;
743 if (!EXT2_HAS_RO_COMPAT_FEATURE(fs, EXT2F_ROCOMPAT_GDT_CSUM) ||
744 !(fs->e2fs_gd[cg].ext4bgd_flags & EXT2_BG_BLOCK_UNINIT))
747 memset(bp->b_data, 0, fs->e2fs_bsize);
749 bit_max = ext2_num_base_meta_blocks(fs, cg);
750 if ((bit_max >> 3) >= fs->e2fs_bsize)
753 for (bit = 0; bit < bit_max; bit++)
754 setbit(bp->b_data, bit);
756 start = cg * fs->e2fs->e2fs_bpg + fs->e2fs->e2fs_first_dblock;
758 /* Set bits for block and inode bitmaps, and inode table */
759 tmp = fs->e2fs_gd[cg].ext2bgd_b_bitmap;
760 if (!EXT2_HAS_INCOMPAT_FEATURE(fs, EXT2F_INCOMPAT_FLEX_BG) ||
761 tmp == ext2_get_cg_number(fs, cg))
762 setbit(bp->b_data, tmp - start);
764 tmp = fs->e2fs_gd[cg].ext2bgd_i_bitmap;
765 if (!EXT2_HAS_INCOMPAT_FEATURE(fs, EXT2F_INCOMPAT_FLEX_BG) ||
766 tmp == ext2_get_cg_number(fs, cg))
767 setbit(bp->b_data, tmp - start);
769 tmp = fs->e2fs_gd[cg].ext2bgd_i_tables;
770 inodes_per_block = fs->e2fs_bsize/EXT2_INODE_SIZE(fs);
771 while( tmp < fs->e2fs_gd[cg].ext2bgd_i_tables +
772 fs->e2fs->e2fs_ipg / inodes_per_block ) {
773 if (!EXT2_HAS_INCOMPAT_FEATURE(fs, EXT2F_INCOMPAT_FLEX_BG) ||
774 tmp == ext2_get_cg_number(fs, cg))
775 setbit(bp->b_data, tmp - start);
780 * Also if the number of blocks within the group is less than
781 * the blocksize * 8 ( which is the size of bitmap ), set rest
782 * of the block bitmap to 1
784 ext2_mark_bitmap_end(fs->e2fs->e2fs_bpg, fs->e2fs_bsize * 8,
788 fs->e2fs_gd[cg].ext4bgd_flags &= ~EXT2_BG_BLOCK_UNINIT;
794 * Determine whether a block can be allocated.
796 * Check to see if a block of the appropriate size is available,
797 * and if it is, allocate it.
800 ext2_alloccg(struct inode *ip, int cg, daddr_t bpref, int size)
804 struct ext2mount *ump;
805 daddr_t bno, runstart, runlen;
806 int bit, loc, end, error, start;
811 if (fs->e2fs_gd[cg].ext2bgd_nbfree == 0)
814 error = bread(ip->i_devvp, fsbtodb(fs,
815 fs->e2fs_gd[cg].ext2bgd_b_bitmap),
816 (int)fs->e2fs_bsize, NOCRED, &bp);
822 if (EXT2_HAS_RO_COMPAT_FEATURE(fs, EXT2F_ROCOMPAT_GDT_CSUM)) {
823 error = ext2_cg_block_bitmap_init(fs, cg, bp);
830 if (fs->e2fs_gd[cg].ext2bgd_nbfree == 0) {
832 * Another thread allocated the last block in this
833 * group while we were waiting for the buffer.
839 bbp = (char *)bp->b_data;
841 if (dtog(fs, bpref) != cg)
844 bpref = dtogd(fs, bpref);
846 * if the requested block is available, use it
848 if (isclr(bbp, bpref)) {
854 * no blocks in the requested cylinder, so take next
855 * available one in this cylinder group.
856 * first try to get 8 contigous blocks, then fall back to a single
860 start = dtogd(fs, bpref) / NBBY;
863 end = howmany(fs->e2fs->e2fs_fpg, NBBY) - start;
867 for (loc = start; loc < end; loc++) {
868 if (bbp[loc] == (char)0xff) {
873 /* Start of a run, find the number of high clear bits. */
877 runstart = loc * NBBY + bit;
878 } else if (bbp[loc] == 0) {
879 /* Continue a run. */
883 * Finish the current run. If it isn't long
884 * enough, start a new one.
886 bit = ffs(bbp[loc]) - 1;
893 /* Run was too short, start a new one. */
896 runstart = loc * NBBY + bit;
899 /* If the current run is long enough, use it. */
910 bno = ext2_mapsearch(fs, bbp, bpref);
918 if (isset(bbp, bno)) {
919 printf("ext2fs_alloccgblk: cg=%d bno=%jd fs=%s\n",
920 cg, (intmax_t)bno, fs->e2fs_fsmnt);
921 panic("ext2fs_alloccg: dup alloc");
926 ext2_clusteracct(fs, bbp, cg, bno, -1);
927 fs->e2fs->e2fs_fbcount--;
928 fs->e2fs_gd[cg].ext2bgd_nbfree--;
932 return (cg * fs->e2fs->e2fs_fpg + fs->e2fs->e2fs_first_dblock + bno);
936 * Determine whether a cluster can be allocated.
939 ext2_clusteralloc(struct inode *ip, int cg, daddr_t bpref, int len)
942 struct ext2mount *ump;
945 int bit, error, got, i, loc, run;
952 if (fs->e2fs_maxcluster[cg] < len)
956 error = bread(ip->i_devvp,
957 fsbtodb(fs, fs->e2fs_gd[cg].ext2bgd_b_bitmap),
958 (int)fs->e2fs_bsize, NOCRED, &bp);
962 bbp = (char *)bp->b_data;
965 * Check to see if a cluster of the needed size (or bigger) is
966 * available in this cylinder group.
968 lp = &fs->e2fs_clustersum[cg].cs_sum[len];
969 for (i = len; i <= fs->e2fs_contigsumsize; i++)
972 if (i > fs->e2fs_contigsumsize) {
974 * Update the cluster summary information to reflect
975 * the true maximum-sized cluster so that future cluster
976 * allocation requests can avoid reading the bitmap only
977 * to find no cluster.
979 lp = &fs->e2fs_clustersum[cg].cs_sum[len - 1];
980 for (i = len - 1; i > 0; i--)
983 fs->e2fs_maxcluster[cg] = i;
988 /* Search the bitmap to find a big enough cluster like in FFS. */
989 if (dtog(fs, bpref) != cg)
992 bpref = dtogd(fs, bpref);
994 bit = 1 << (bpref % NBBY);
995 for (run = 0, got = bpref; got < fs->e2fs->e2fs_fpg; got++) {
996 if ((bbp[loc] & bit) != 0)
1003 if ((got & (NBBY - 1)) != (NBBY - 1))
1011 if (got >= fs->e2fs->e2fs_fpg)
1014 /* Allocate the cluster that we found. */
1015 for (i = 1; i < len; i++)
1016 if (!isclr(bbp, got - run + i))
1017 panic("ext2_clusteralloc: map mismatch");
1019 bno = got - run + 1;
1020 if (bno >= fs->e2fs->e2fs_fpg)
1021 panic("ext2_clusteralloc: allocated out of group");
1024 for (i = 0; i < len; i += fs->e2fs_fpb) {
1025 setbit(bbp, bno + i);
1026 ext2_clusteracct(fs, bbp, cg, bno + i, -1);
1027 fs->e2fs->e2fs_fbcount--;
1028 fs->e2fs_gd[cg].ext2bgd_nbfree--;
1034 return (cg * fs->e2fs->e2fs_fpg + fs->e2fs->e2fs_first_dblock + bno);
1044 ext2_zero_inode_table(struct inode *ip, int cg)
1046 struct m_ext2fs *fs;
1048 int i, all_blks, used_blks;
1052 if (fs->e2fs_gd[cg].ext4bgd_flags & EXT2_BG_INODE_ZEROED)
1055 all_blks = fs->e2fs->e2fs_inode_size * fs->e2fs->e2fs_ipg /
1058 used_blks = howmany(fs->e2fs->e2fs_ipg -
1059 fs->e2fs_gd[cg].ext4bgd_i_unused,
1060 fs->e2fs_bsize / EXT2_INODE_SIZE(fs));
1062 for (i = 0; i < all_blks - used_blks; i++) {
1063 bp = getblk(ip->i_devvp, fsbtodb(fs,
1064 fs->e2fs_gd[cg].ext2bgd_i_tables + used_blks + i),
1065 fs->e2fs_bsize, 0, 0, 0);
1069 vfs_bio_bzero_buf(bp, 0, fs->e2fs_bsize);
1073 fs->e2fs_gd[cg].ext4bgd_flags |= EXT2_BG_INODE_ZEROED;
1079 * Determine whether an inode can be allocated.
1081 * Check to see if an inode is available, and if it is,
1082 * allocate it using tode in the specified cylinder group.
1085 ext2_nodealloccg(struct inode *ip, int cg, daddr_t ipref, int mode)
1087 struct m_ext2fs *fs;
1089 struct ext2mount *ump;
1090 int error, start, len;
1093 ipref--; /* to avoid a lot of (ipref -1) */
1098 if (fs->e2fs_gd[cg].ext2bgd_nifree == 0)
1101 error = bread(ip->i_devvp, fsbtodb(fs,
1102 fs->e2fs_gd[cg].ext2bgd_i_bitmap),
1103 (int)fs->e2fs_bsize, NOCRED, &bp);
1109 if (EXT2_HAS_RO_COMPAT_FEATURE(fs, EXT2F_ROCOMPAT_GDT_CSUM)) {
1110 if (fs->e2fs_gd[cg].ext4bgd_flags & EXT2_BG_INODE_UNINIT) {
1111 memset(bp->b_data, 0, fs->e2fs_bsize);
1112 fs->e2fs_gd[cg].ext4bgd_flags &= ~EXT2_BG_INODE_UNINIT;
1114 error = ext2_zero_inode_table(ip, cg);
1121 if (fs->e2fs_gd[cg].ext2bgd_nifree == 0) {
1123 * Another thread allocated the last i-node in this
1124 * group while we were waiting for the buffer.
1130 ibp = (char *)bp->b_data;
1132 ipref %= fs->e2fs->e2fs_ipg;
1133 if (isclr(ibp, ipref))
1136 start = ipref / NBBY;
1137 len = howmany(fs->e2fs->e2fs_ipg - ipref, NBBY);
1138 loc = memcchr(&ibp[start], 0xff, len);
1142 loc = memcchr(&ibp[start], 0xff, len);
1144 printf("cg = %d, ipref = %lld, fs = %s\n",
1145 cg, (long long)ipref, fs->e2fs_fsmnt);
1146 panic("ext2fs_nodealloccg: map corrupted");
1150 ipref = (loc - ibp) * NBBY + ffs(~*loc) - 1;
1154 fs->e2fs_gd[cg].ext2bgd_nifree--;
1155 if (EXT2_HAS_RO_COMPAT_FEATURE(fs, EXT2F_ROCOMPAT_GDT_CSUM))
1156 fs->e2fs_gd[cg].ext4bgd_i_unused--;
1157 fs->e2fs->e2fs_ficount--;
1159 if ((mode & IFMT) == IFDIR) {
1160 fs->e2fs_gd[cg].ext2bgd_ndirs++;
1161 fs->e2fs_total_dir++;
1165 return (cg * fs->e2fs->e2fs_ipg + ipref + 1);
1169 * Free a block or fragment.
1173 ext2_blkfree(struct inode *ip, e4fs_daddr_t bno, long size)
1175 struct m_ext2fs *fs;
1177 struct ext2mount *ump;
1184 if (bno >= fs->e2fs->e2fs_bcount) {
1185 printf("bad block %lld, ino %ju\n", (long long)bno,
1186 (uintmax_t)ip->i_number);
1187 ext2_fserr(fs, ip->i_uid, "bad block");
1190 error = bread(ip->i_devvp,
1191 fsbtodb(fs, fs->e2fs_gd[cg].ext2bgd_b_bitmap),
1192 (int)fs->e2fs_bsize, NOCRED, &bp);
1197 bbp = (char *)bp->b_data;
1198 bno = dtogd(fs, bno);
1199 if (isclr(bbp, bno)) {
1200 printf("block = %lld, fs = %s\n",
1201 (long long)bno, fs->e2fs_fsmnt);
1202 panic("ext2_blkfree: freeing free block");
1206 ext2_clusteracct(fs, bbp, cg, bno, 1);
1207 fs->e2fs->e2fs_fbcount++;
1208 fs->e2fs_gd[cg].ext2bgd_nbfree++;
1219 ext2_vfree(struct vnode *pvp, ino_t ino, int mode)
1221 struct m_ext2fs *fs;
1224 struct ext2mount *ump;
1231 if ((u_int)ino > fs->e2fs_ipg * fs->e2fs_gcount)
1232 panic("ext2_vfree: range: devvp = %p, ino = %ju, fs = %s",
1233 pip->i_devvp, (uintmax_t)ino, fs->e2fs_fsmnt);
1235 cg = ino_to_cg(fs, ino);
1236 error = bread(pip->i_devvp,
1237 fsbtodb(fs, fs->e2fs_gd[cg].ext2bgd_i_bitmap),
1238 (int)fs->e2fs_bsize, NOCRED, &bp);
1243 ibp = (char *)bp->b_data;
1244 ino = (ino - 1) % fs->e2fs->e2fs_ipg;
1245 if (isclr(ibp, ino)) {
1246 printf("ino = %llu, fs = %s\n",
1247 (unsigned long long)ino, fs->e2fs_fsmnt);
1248 if (fs->e2fs_ronly == 0)
1249 panic("ext2_vfree: freeing free inode");
1253 fs->e2fs->e2fs_ficount++;
1254 fs->e2fs_gd[cg].ext2bgd_nifree++;
1255 if (EXT2_HAS_RO_COMPAT_FEATURE(fs, EXT2F_ROCOMPAT_GDT_CSUM))
1256 fs->e2fs_gd[cg].ext4bgd_i_unused++;
1257 if ((mode & IFMT) == IFDIR) {
1258 fs->e2fs_gd[cg].ext2bgd_ndirs--;
1259 fs->e2fs_total_dir--;
1268 * Find a block in the specified cylinder group.
1270 * It is a panic if a request is made to find a block if none are
1274 ext2_mapsearch(struct m_ext2fs *fs, char *bbp, daddr_t bpref)
1280 * find the fragment by searching through the free block
1281 * map for an appropriate bit pattern
1284 start = dtogd(fs, bpref) / NBBY;
1287 len = howmany(fs->e2fs->e2fs_fpg, NBBY) - start;
1288 loc = memcchr(&bbp[start], 0xff, len);
1292 loc = memcchr(&bbp[start], 0xff, len);
1294 printf("start = %d, len = %d, fs = %s\n",
1295 start, len, fs->e2fs_fsmnt);
1296 panic("ext2_mapsearch: map corrupted");
1300 return ((loc - bbp) * NBBY + ffs(~*loc) - 1);
1304 * Fserr prints the name of a filesystem with an error diagnostic.
1306 * The form of the error message is:
1310 ext2_fserr(struct m_ext2fs *fs, uid_t uid, char *cp)
1313 log(LOG_ERR, "uid %u on %s: %s\n", uid, fs->e2fs_fsmnt, cp);
1317 ext2_cg_has_sb(struct m_ext2fs *fs, int cg)
1324 if (EXT2_HAS_COMPAT_FEATURE(fs, EXT2F_COMPAT_SPARSESUPER2)) {
1325 if (cg == fs->e2fs->e4fs_backup_bgs[0] ||
1326 cg == fs->e2fs->e4fs_backup_bgs[1])
1332 !EXT2_HAS_RO_COMPAT_FEATURE(fs, EXT2F_ROCOMPAT_SPARSESUPER))
1338 for (a3 = 3, a5 = 5, a7 = 7;
1339 a3 <= cg || a5 <= cg || a7 <= cg;
1340 a3 *= 3, a5 *= 5, a7 *= 7)
1341 if (cg == a3 || cg == a5 || cg == a7)