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 void ext2_fserr(struct m_ext2fs *, uid_t, char *);
60 static u_long ext2_hashalloc(struct inode *, int, long, int,
61 daddr_t (*)(struct inode *, int, daddr_t,
63 static daddr_t ext2_nodealloccg(struct inode *, int, daddr_t, int);
64 static daddr_t ext2_mapsearch(struct m_ext2fs *, char *, daddr_t);
67 * Allocate a block in the filesystem.
69 * A preference may be optionally specified. If a preference is given
70 * the following hierarchy is used to allocate a block:
71 * 1) allocate the requested block.
72 * 2) allocate a rotationally optimal block in the same cylinder.
73 * 3) allocate a block in the same cylinder group.
74 * 4) quadradically rehash into other cylinder groups, until an
75 * available block is located.
76 * If no block preference is given the following hierarchy is used
77 * to allocate a block:
78 * 1) allocate a block in the cylinder group that contains the
80 * 2) quadradically rehash into other cylinder groups, until an
81 * available block is located.
84 ext2_alloc(struct inode *ip, daddr_t lbn, e4fs_daddr_t bpref, int size,
85 struct ucred *cred, e4fs_daddr_t *bnp)
88 struct ext2mount *ump;
95 mtx_assert(EXT2_MTX(ump), MA_OWNED);
97 if ((u_int)size > fs->e2fs_bsize || blkoff(fs, size) != 0) {
98 vn_printf(ip->i_devvp, "bsize = %lu, size = %d, fs = %s\n",
99 (long unsigned int)fs->e2fs_bsize, size, fs->e2fs_fsmnt);
100 panic("ext2_alloc: bad size");
103 panic("ext2_alloc: missing credential");
104 #endif /* INVARIANTS */
105 if (size == fs->e2fs_bsize && fs->e2fs->e2fs_fbcount == 0)
107 if (cred->cr_uid != 0 &&
108 fs->e2fs->e2fs_fbcount < fs->e2fs->e2fs_rbcount)
110 if (bpref >= fs->e2fs->e2fs_bcount)
113 cg = ino_to_cg(fs, ip->i_number);
115 cg = dtog(fs, bpref);
116 bno = (daddr_t)ext2_hashalloc(ip, cg, bpref, fs->e2fs_bsize,
119 /* set next_alloc fields as done in block_getblk */
120 ip->i_next_alloc_block = lbn;
121 ip->i_next_alloc_goal = bno;
123 ip->i_blocks += btodb(fs->e2fs_bsize);
124 ip->i_flag |= IN_CHANGE | IN_UPDATE;
130 ext2_fserr(fs, cred->cr_uid, "filesystem full");
131 uprintf("\n%s: write failed, filesystem is full\n", fs->e2fs_fsmnt);
136 * Allocate EA's block for inode.
139 ext2_allocfacl(struct inode *ip)
146 EXT2_LOCK(ip->i_ump);
147 facl = ext2_alloccg(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 = 1;
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)
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 /* now we want to make sure that the block pointers are zeroed out */
426 for (i = 0; i < EXT2_NDADDR; i++)
428 for (i = 0; i < EXT2_NIADDR; i++)
432 * Set up a new generation number for this inode.
436 ip->i_gen = arc4random();
437 } while (ip->i_gen == 0);
440 ip->i_birthtime = ts.tv_sec;
441 ip->i_birthnsec = ts.tv_nsec;
444 printf("ext2_valloc: allocated inode %d\n", ino);
449 ext2_fserr(fs, cred->cr_uid, "out of inodes");
450 uprintf("\n%s: create/symlink failed, no inodes free\n", fs->e2fs_fsmnt);
455 * Find a cylinder to place a directory.
457 * The policy implemented by this algorithm is to allocate a
458 * directory inode in the same cylinder group as its parent
459 * directory, but also to reserve space for its files inodes
460 * and data. Restrict the number of directories which may be
461 * allocated one after another in the same cylinder group
462 * without intervening allocation of files.
464 * If we allocate a first level directory then force allocation
465 * in another cylinder group.
469 ext2_dirpref(struct inode *pip)
472 int cg, prefcg, cgsize;
473 u_int avgifree, avgbfree, avgndir, curdirsize;
474 u_int minifree, minbfree, maxndir;
475 u_int mincg, minndir;
476 u_int dirsize, maxcontigdirs;
478 mtx_assert(EXT2_MTX(pip->i_ump), MA_OWNED);
481 avgifree = fs->e2fs->e2fs_ficount / fs->e2fs_gcount;
482 avgbfree = fs->e2fs->e2fs_fbcount / fs->e2fs_gcount;
483 avgndir = fs->e2fs_total_dir / fs->e2fs_gcount;
486 * Force allocation in another cg if creating a first level dir.
488 ASSERT_VOP_LOCKED(ITOV(pip), "ext2fs_dirpref");
489 if (ITOV(pip)->v_vflag & VV_ROOT) {
490 prefcg = arc4random() % fs->e2fs_gcount;
492 minndir = fs->e2fs_ipg;
493 for (cg = prefcg; cg < fs->e2fs_gcount; cg++)
494 if (fs->e2fs_gd[cg].ext2bgd_ndirs < minndir &&
495 fs->e2fs_gd[cg].ext2bgd_nifree >= avgifree &&
496 fs->e2fs_gd[cg].ext2bgd_nbfree >= avgbfree) {
498 minndir = fs->e2fs_gd[cg].ext2bgd_ndirs;
500 for (cg = 0; cg < prefcg; cg++)
501 if (fs->e2fs_gd[cg].ext2bgd_ndirs < minndir &&
502 fs->e2fs_gd[cg].ext2bgd_nifree >= avgifree &&
503 fs->e2fs_gd[cg].ext2bgd_nbfree >= avgbfree) {
505 minndir = fs->e2fs_gd[cg].ext2bgd_ndirs;
510 * Count various limits which used for
511 * optimal allocation of a directory inode.
513 maxndir = min(avgndir + fs->e2fs_ipg / 16, fs->e2fs_ipg);
514 minifree = avgifree - avgifree / 4;
517 minbfree = avgbfree - avgbfree / 4;
520 cgsize = fs->e2fs_fsize * fs->e2fs_fpg;
521 dirsize = AVGDIRSIZE;
522 curdirsize = avgndir ? (cgsize - avgbfree * fs->e2fs_bsize) / avgndir : 0;
523 if (dirsize < curdirsize)
524 dirsize = curdirsize;
525 maxcontigdirs = min((avgbfree * fs->e2fs_bsize) / dirsize, 255);
526 maxcontigdirs = min(maxcontigdirs, fs->e2fs_ipg / AFPDIR);
527 if (maxcontigdirs == 0)
531 * Limit number of dirs in one cg and reserve space for
532 * regular files, but only if we have no deficit in
535 prefcg = ino_to_cg(fs, pip->i_number);
536 for (cg = prefcg; cg < fs->e2fs_gcount; cg++)
537 if (fs->e2fs_gd[cg].ext2bgd_ndirs < maxndir &&
538 fs->e2fs_gd[cg].ext2bgd_nifree >= minifree &&
539 fs->e2fs_gd[cg].ext2bgd_nbfree >= minbfree) {
540 if (fs->e2fs_contigdirs[cg] < maxcontigdirs)
543 for (cg = 0; cg < prefcg; cg++)
544 if (fs->e2fs_gd[cg].ext2bgd_ndirs < maxndir &&
545 fs->e2fs_gd[cg].ext2bgd_nifree >= minifree &&
546 fs->e2fs_gd[cg].ext2bgd_nbfree >= minbfree) {
547 if (fs->e2fs_contigdirs[cg] < maxcontigdirs)
551 * This is a backstop when we have deficit in space.
553 for (cg = prefcg; cg < fs->e2fs_gcount; cg++)
554 if (fs->e2fs_gd[cg].ext2bgd_nifree >= avgifree)
556 for (cg = 0; cg < prefcg; cg++)
557 if (fs->e2fs_gd[cg].ext2bgd_nifree >= avgifree)
563 * Select the desired position for the next block in a file.
565 * we try to mimic what Remy does in inode_getblk/block_getblk
567 * we note: blocknr == 0 means that we're about to allocate either
568 * a direct block or a pointer block at the first level of indirection
569 * (In other words, stuff that will go in i_db[] or i_ib[])
571 * blocknr != 0 means that we're allocating a block that is none
572 * of the above. Then, blocknr tells us the number of the block
573 * that will hold the pointer
576 ext2_blkpref(struct inode *ip, e2fs_lbn_t lbn, int indx, e2fs_daddr_t *bap,
577 e2fs_daddr_t blocknr)
581 mtx_assert(EXT2_MTX(ip->i_ump), MA_OWNED);
584 * If the next block is actually what we thought it is, then set the
585 * goal to what we thought it should be.
587 if (ip->i_next_alloc_block == lbn && ip->i_next_alloc_goal != 0)
588 return ip->i_next_alloc_goal;
591 * Now check whether we were provided with an array that basically
592 * tells us previous blocks to which we want to stay close.
595 for (tmp = indx - 1; tmp >= 0; tmp--)
600 * Else lets fall back to the blocknr or, if there is none, follow
601 * the rule that a block should be allocated near its inode.
603 return blocknr ? blocknr :
604 (e2fs_daddr_t)(ip->i_block_group *
605 EXT2_BLOCKS_PER_GROUP(ip->i_e2fs)) +
606 ip->i_e2fs->e2fs->e2fs_first_dblock;
610 * Implement the cylinder overflow algorithm.
612 * The policy implemented by this algorithm is:
613 * 1) allocate the block in its requested cylinder group.
614 * 2) quadradically rehash on the cylinder group number.
615 * 3) brute force search for a free block.
618 ext2_hashalloc(struct inode *ip, int cg, long pref, int size,
619 daddr_t (*allocator) (struct inode *, int, daddr_t, int))
625 mtx_assert(EXT2_MTX(ip->i_ump), MA_OWNED);
628 * 1: preferred cylinder group
630 result = (*allocator)(ip, cg, pref, size);
634 * 2: quadratic rehash
636 for (i = 1; i < fs->e2fs_gcount; i *= 2) {
638 if (cg >= fs->e2fs_gcount)
639 cg -= fs->e2fs_gcount;
640 result = (*allocator)(ip, cg, 0, size);
645 * 3: brute force search
646 * Note that we start at i == 2, since 0 was checked initially,
647 * and 1 is always checked in the quadratic rehash.
649 cg = (icg + 2) % fs->e2fs_gcount;
650 for (i = 2; i < fs->e2fs_gcount; i++) {
651 result = (*allocator)(ip, cg, 0, size);
655 if (cg == fs->e2fs_gcount)
662 ext2_cg_num_gdb(struct m_ext2fs *fs, int cg)
664 int gd_per_block, metagroup, first, last;
666 gd_per_block = fs->e2fs_bsize / sizeof(struct ext2_gd);
667 metagroup = cg / gd_per_block;
668 first = metagroup * gd_per_block;
669 last = first + gd_per_block - 1;
671 if (!EXT2_HAS_INCOMPAT_FEATURE(fs, EXT2F_INCOMPAT_META_BG) ||
672 metagroup < fs->e2fs->e3fs_first_meta_bg) {
673 if (!ext2_cg_has_sb(fs, cg))
675 if (EXT2_HAS_INCOMPAT_FEATURE(fs, EXT2F_INCOMPAT_META_BG))
676 return (fs->e2fs->e3fs_first_meta_bg);
677 return (fs->e2fs_gdbcount);
680 if (cg == first || cg == first + 1 || cg == last)
687 ext2_num_base_meta_blocks(struct m_ext2fs *fs, int cg)
689 int num, gd_per_block;
691 gd_per_block = fs->e2fs_bsize / sizeof(struct ext2_gd);
692 num = ext2_cg_has_sb(fs, cg);
694 if (!EXT2_HAS_INCOMPAT_FEATURE(fs, EXT2F_INCOMPAT_META_BG) ||
695 cg < fs->e2fs->e3fs_first_meta_bg * gd_per_block) {
697 num += ext2_cg_num_gdb(fs, cg);
698 num += fs->e2fs->e2fs_reserved_ngdb;
701 num += ext2_cg_num_gdb(fs, cg);
708 ext2_get_cg_number(struct m_ext2fs *fs, daddr_t blk)
712 if (fs->e2fs->e2fs_bpg == fs->e2fs_bsize * 8)
713 cg = (blk - fs->e2fs->e2fs_first_dblock) / (fs->e2fs_bsize * 8);
715 cg = blk - fs->e2fs->e2fs_first_dblock;
721 ext2_mark_bitmap_end(int start_bit, int end_bit, char *bitmap)
725 if (start_bit >= end_bit)
728 for (i = start_bit; i < ((start_bit + 7) & ~7UL); i++)
731 memset(bitmap + (i >> 3), 0xff, (end_bit - i) >> 3);
735 ext2_cg_block_bitmap_init(struct m_ext2fs *fs, int cg, struct buf *bp)
737 int bit, bit_max, inodes_per_block;
740 if (!EXT2_HAS_RO_COMPAT_FEATURE(fs, EXT2F_ROCOMPAT_GDT_CSUM) ||
741 !(fs->e2fs_gd[cg].ext4bgd_flags & EXT2_BG_BLOCK_UNINIT))
744 memset(bp->b_data, 0, fs->e2fs_bsize);
746 bit_max = ext2_num_base_meta_blocks(fs, cg);
747 if ((bit_max >> 3) >= fs->e2fs_bsize)
750 for (bit = 0; bit < bit_max; bit++)
751 setbit(bp->b_data, bit);
753 start = cg * fs->e2fs->e2fs_bpg + fs->e2fs->e2fs_first_dblock;
755 /* Set bits for block and inode bitmaps, and inode table */
756 tmp = fs->e2fs_gd[cg].ext2bgd_b_bitmap;
757 if (!EXT2_HAS_INCOMPAT_FEATURE(fs, EXT2F_INCOMPAT_FLEX_BG) ||
758 tmp == ext2_get_cg_number(fs, cg))
759 setbit(bp->b_data, tmp - start);
761 tmp = fs->e2fs_gd[cg].ext2bgd_i_bitmap;
762 if (!EXT2_HAS_INCOMPAT_FEATURE(fs, EXT2F_INCOMPAT_FLEX_BG) ||
763 tmp == ext2_get_cg_number(fs, cg))
764 setbit(bp->b_data, tmp - start);
766 tmp = fs->e2fs_gd[cg].ext2bgd_i_tables;
767 inodes_per_block = fs->e2fs_bsize/EXT2_INODE_SIZE(fs);
768 while( tmp < fs->e2fs_gd[cg].ext2bgd_i_tables +
769 fs->e2fs->e2fs_ipg / inodes_per_block ) {
770 if (!EXT2_HAS_INCOMPAT_FEATURE(fs, EXT2F_INCOMPAT_FLEX_BG) ||
771 tmp == ext2_get_cg_number(fs, cg))
772 setbit(bp->b_data, tmp - start);
777 * Also if the number of blocks within the group is less than
778 * the blocksize * 8 ( which is the size of bitmap ), set rest
779 * of the block bitmap to 1
781 ext2_mark_bitmap_end(fs->e2fs->e2fs_bpg, fs->e2fs_bsize * 8,
785 fs->e2fs_gd[cg].ext4bgd_flags &= ~EXT2_BG_BLOCK_UNINIT;
791 * Determine whether a block can be allocated.
793 * Check to see if a block of the appropriate size is available,
794 * and if it is, allocate it.
797 ext2_alloccg(struct inode *ip, int cg, daddr_t bpref, int size)
801 struct ext2mount *ump;
802 daddr_t bno, runstart, runlen;
803 int bit, loc, end, error, start;
808 if (fs->e2fs_gd[cg].ext2bgd_nbfree == 0)
811 error = bread(ip->i_devvp, fsbtodb(fs,
812 fs->e2fs_gd[cg].ext2bgd_b_bitmap),
813 (int)fs->e2fs_bsize, NOCRED, &bp);
819 if (EXT2_HAS_RO_COMPAT_FEATURE(fs, EXT2F_ROCOMPAT_GDT_CSUM)) {
820 error = ext2_cg_block_bitmap_init(fs, cg, bp);
827 if (fs->e2fs_gd[cg].ext2bgd_nbfree == 0) {
829 * Another thread allocated the last block in this
830 * group while we were waiting for the buffer.
836 bbp = (char *)bp->b_data;
838 if (dtog(fs, bpref) != cg)
841 bpref = dtogd(fs, bpref);
843 * if the requested block is available, use it
845 if (isclr(bbp, bpref)) {
851 * no blocks in the requested cylinder, so take next
852 * available one in this cylinder group.
853 * first try to get 8 contigous blocks, then fall back to a single
857 start = dtogd(fs, bpref) / NBBY;
860 end = howmany(fs->e2fs->e2fs_fpg, NBBY) - start;
864 for (loc = start; loc < end; loc++) {
865 if (bbp[loc] == (char)0xff) {
870 /* Start of a run, find the number of high clear bits. */
874 runstart = loc * NBBY + bit;
875 } else if (bbp[loc] == 0) {
876 /* Continue a run. */
880 * Finish the current run. If it isn't long
881 * enough, start a new one.
883 bit = ffs(bbp[loc]) - 1;
890 /* Run was too short, start a new one. */
893 runstart = loc * NBBY + bit;
896 /* If the current run is long enough, use it. */
907 bno = ext2_mapsearch(fs, bbp, bpref);
915 if (isset(bbp, bno)) {
916 printf("ext2fs_alloccgblk: cg=%d bno=%jd fs=%s\n",
917 cg, (intmax_t)bno, fs->e2fs_fsmnt);
918 panic("ext2fs_alloccg: dup alloc");
923 ext2_clusteracct(fs, bbp, cg, bno, -1);
924 fs->e2fs->e2fs_fbcount--;
925 fs->e2fs_gd[cg].ext2bgd_nbfree--;
929 return (cg * fs->e2fs->e2fs_fpg + fs->e2fs->e2fs_first_dblock + bno);
933 * Determine whether a cluster can be allocated.
936 ext2_clusteralloc(struct inode *ip, int cg, daddr_t bpref, int len)
939 struct ext2mount *ump;
942 int bit, error, got, i, loc, run;
949 if (fs->e2fs_maxcluster[cg] < len)
953 error = bread(ip->i_devvp,
954 fsbtodb(fs, fs->e2fs_gd[cg].ext2bgd_b_bitmap),
955 (int)fs->e2fs_bsize, NOCRED, &bp);
959 bbp = (char *)bp->b_data;
962 * Check to see if a cluster of the needed size (or bigger) is
963 * available in this cylinder group.
965 lp = &fs->e2fs_clustersum[cg].cs_sum[len];
966 for (i = len; i <= fs->e2fs_contigsumsize; i++)
969 if (i > fs->e2fs_contigsumsize) {
971 * Update the cluster summary information to reflect
972 * the true maximum-sized cluster so that future cluster
973 * allocation requests can avoid reading the bitmap only
974 * to find no cluster.
976 lp = &fs->e2fs_clustersum[cg].cs_sum[len - 1];
977 for (i = len - 1; i > 0; i--)
980 fs->e2fs_maxcluster[cg] = i;
985 /* Search the bitmap to find a big enough cluster like in FFS. */
986 if (dtog(fs, bpref) != cg)
989 bpref = dtogd(fs, bpref);
991 bit = 1 << (bpref % NBBY);
992 for (run = 0, got = bpref; got < fs->e2fs->e2fs_fpg; got++) {
993 if ((bbp[loc] & bit) != 0)
1000 if ((got & (NBBY - 1)) != (NBBY - 1))
1008 if (got >= fs->e2fs->e2fs_fpg)
1011 /* Allocate the cluster that we found. */
1012 for (i = 1; i < len; i++)
1013 if (!isclr(bbp, got - run + i))
1014 panic("ext2_clusteralloc: map mismatch");
1016 bno = got - run + 1;
1017 if (bno >= fs->e2fs->e2fs_fpg)
1018 panic("ext2_clusteralloc: allocated out of group");
1021 for (i = 0; i < len; i += fs->e2fs_fpb) {
1022 setbit(bbp, bno + i);
1023 ext2_clusteracct(fs, bbp, cg, bno + i, -1);
1024 fs->e2fs->e2fs_fbcount--;
1025 fs->e2fs_gd[cg].ext2bgd_nbfree--;
1031 return (cg * fs->e2fs->e2fs_fpg + fs->e2fs->e2fs_first_dblock + bno);
1041 ext2_zero_inode_table(struct inode *ip, int cg)
1043 struct m_ext2fs *fs;
1045 int i, all_blks, used_blks;
1049 if (fs->e2fs_gd[cg].ext4bgd_flags & EXT2_BG_INODE_ZEROED)
1052 all_blks = fs->e2fs->e2fs_inode_size * fs->e2fs->e2fs_ipg /
1055 used_blks = howmany(fs->e2fs->e2fs_ipg -
1056 fs->e2fs_gd[cg].ext4bgd_i_unused,
1057 fs->e2fs_bsize / EXT2_INODE_SIZE(fs));
1059 for (i = 0; i < all_blks - used_blks; i++) {
1060 bp = getblk(ip->i_devvp, fsbtodb(fs,
1061 fs->e2fs_gd[cg].ext2bgd_i_tables + used_blks + i),
1062 fs->e2fs_bsize, 0, 0, 0);
1066 vfs_bio_bzero_buf(bp, 0, fs->e2fs_bsize);
1070 fs->e2fs_gd[cg].ext4bgd_flags |= EXT2_BG_INODE_ZEROED;
1076 * Determine whether an inode can be allocated.
1078 * Check to see if an inode is available, and if it is,
1079 * allocate it using tode in the specified cylinder group.
1082 ext2_nodealloccg(struct inode *ip, int cg, daddr_t ipref, int mode)
1084 struct m_ext2fs *fs;
1086 struct ext2mount *ump;
1087 int error, start, len;
1090 ipref--; /* to avoid a lot of (ipref -1) */
1095 if (fs->e2fs_gd[cg].ext2bgd_nifree == 0)
1098 error = bread(ip->i_devvp, fsbtodb(fs,
1099 fs->e2fs_gd[cg].ext2bgd_i_bitmap),
1100 (int)fs->e2fs_bsize, NOCRED, &bp);
1106 if (EXT2_HAS_RO_COMPAT_FEATURE(fs, EXT2F_ROCOMPAT_GDT_CSUM)) {
1107 if (fs->e2fs_gd[cg].ext4bgd_flags & EXT2_BG_INODE_UNINIT) {
1108 memset(bp->b_data, 0, fs->e2fs_bsize);
1109 fs->e2fs_gd[cg].ext4bgd_flags &= ~EXT2_BG_INODE_UNINIT;
1111 error = ext2_zero_inode_table(ip, cg);
1118 if (fs->e2fs_gd[cg].ext2bgd_nifree == 0) {
1120 * Another thread allocated the last i-node in this
1121 * group while we were waiting for the buffer.
1127 ibp = (char *)bp->b_data;
1129 ipref %= fs->e2fs->e2fs_ipg;
1130 if (isclr(ibp, ipref))
1133 start = ipref / NBBY;
1134 len = howmany(fs->e2fs->e2fs_ipg - ipref, NBBY);
1135 loc = memcchr(&ibp[start], 0xff, len);
1139 loc = memcchr(&ibp[start], 0xff, len);
1141 printf("cg = %d, ipref = %lld, fs = %s\n",
1142 cg, (long long)ipref, fs->e2fs_fsmnt);
1143 panic("ext2fs_nodealloccg: map corrupted");
1147 ipref = (loc - ibp) * NBBY + ffs(~*loc) - 1;
1151 fs->e2fs_gd[cg].ext2bgd_nifree--;
1152 if (EXT2_HAS_RO_COMPAT_FEATURE(fs, EXT2F_ROCOMPAT_GDT_CSUM))
1153 fs->e2fs_gd[cg].ext4bgd_i_unused--;
1154 fs->e2fs->e2fs_ficount--;
1156 if ((mode & IFMT) == IFDIR) {
1157 fs->e2fs_gd[cg].ext2bgd_ndirs++;
1158 fs->e2fs_total_dir++;
1162 return (cg * fs->e2fs->e2fs_ipg + ipref + 1);
1166 * Free a block or fragment.
1170 ext2_blkfree(struct inode *ip, e4fs_daddr_t bno, long size)
1172 struct m_ext2fs *fs;
1174 struct ext2mount *ump;
1181 if ((u_int)bno >= fs->e2fs->e2fs_bcount) {
1182 printf("bad block %lld, ino %ju\n", (long long)bno,
1183 (uintmax_t)ip->i_number);
1184 ext2_fserr(fs, ip->i_uid, "bad block");
1187 error = bread(ip->i_devvp,
1188 fsbtodb(fs, fs->e2fs_gd[cg].ext2bgd_b_bitmap),
1189 (int)fs->e2fs_bsize, NOCRED, &bp);
1194 bbp = (char *)bp->b_data;
1195 bno = dtogd(fs, bno);
1196 if (isclr(bbp, bno)) {
1197 printf("block = %lld, fs = %s\n",
1198 (long long)bno, fs->e2fs_fsmnt);
1199 panic("ext2_blkfree: freeing free block");
1203 ext2_clusteracct(fs, bbp, cg, bno, 1);
1204 fs->e2fs->e2fs_fbcount++;
1205 fs->e2fs_gd[cg].ext2bgd_nbfree++;
1216 ext2_vfree(struct vnode *pvp, ino_t ino, int mode)
1218 struct m_ext2fs *fs;
1221 struct ext2mount *ump;
1228 if ((u_int)ino > fs->e2fs_ipg * fs->e2fs_gcount)
1229 panic("ext2_vfree: range: devvp = %p, ino = %ju, fs = %s",
1230 pip->i_devvp, (uintmax_t)ino, fs->e2fs_fsmnt);
1232 cg = ino_to_cg(fs, ino);
1233 error = bread(pip->i_devvp,
1234 fsbtodb(fs, fs->e2fs_gd[cg].ext2bgd_i_bitmap),
1235 (int)fs->e2fs_bsize, NOCRED, &bp);
1240 ibp = (char *)bp->b_data;
1241 ino = (ino - 1) % fs->e2fs->e2fs_ipg;
1242 if (isclr(ibp, ino)) {
1243 printf("ino = %llu, fs = %s\n",
1244 (unsigned long long)ino, fs->e2fs_fsmnt);
1245 if (fs->e2fs_ronly == 0)
1246 panic("ext2_vfree: freeing free inode");
1250 fs->e2fs->e2fs_ficount++;
1251 fs->e2fs_gd[cg].ext2bgd_nifree++;
1252 if (EXT2_HAS_RO_COMPAT_FEATURE(fs, EXT2F_ROCOMPAT_GDT_CSUM))
1253 fs->e2fs_gd[cg].ext4bgd_i_unused++;
1254 if ((mode & IFMT) == IFDIR) {
1255 fs->e2fs_gd[cg].ext2bgd_ndirs--;
1256 fs->e2fs_total_dir--;
1265 * Find a block in the specified cylinder group.
1267 * It is a panic if a request is made to find a block if none are
1271 ext2_mapsearch(struct m_ext2fs *fs, char *bbp, daddr_t bpref)
1277 * find the fragment by searching through the free block
1278 * map for an appropriate bit pattern
1281 start = dtogd(fs, bpref) / NBBY;
1284 len = howmany(fs->e2fs->e2fs_fpg, NBBY) - start;
1285 loc = memcchr(&bbp[start], 0xff, len);
1289 loc = memcchr(&bbp[start], 0xff, len);
1291 printf("start = %d, len = %d, fs = %s\n",
1292 start, len, fs->e2fs_fsmnt);
1293 panic("ext2_mapsearch: map corrupted");
1297 return ((loc - bbp) * NBBY + ffs(~*loc) - 1);
1301 * Fserr prints the name of a filesystem with an error diagnostic.
1303 * The form of the error message is:
1307 ext2_fserr(struct m_ext2fs *fs, uid_t uid, char *cp)
1310 log(LOG_ERR, "uid %u on %s: %s\n", uid, fs->e2fs_fsmnt, cp);
1314 ext2_cg_has_sb(struct m_ext2fs *fs, int cg)
1321 if (EXT2_HAS_COMPAT_FEATURE(fs, EXT2F_COMPAT_SPARSESUPER2)) {
1322 if (cg == fs->e2fs->e4fs_backup_bgs[0] ||
1323 cg == fs->e2fs->e4fs_backup_bgs[1])
1329 !EXT2_HAS_RO_COMPAT_FEATURE(fs, EXT2F_ROCOMPAT_SPARSESUPER))
1335 for (a3 = 3, a5 = 5, a7 = 7;
1336 a3 <= cg || a5 <= cg || a7 <= cg;
1337 a3 *= 3, a5 *= 5, a7 *= 7)
1338 if (cg == a3 || cg == a5 || cg == a7)