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. All advertising materials mentioning features or use of this software
20 * must display the following acknowledgement:
21 * This product includes software developed by the University of
22 * California, Berkeley and its contributors.
23 * 4. Neither the name of the University nor the names of its contributors
24 * may be used to endorse or promote products derived from this software
25 * without specific prior written permission.
27 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
28 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
29 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
30 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
31 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
32 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
33 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
34 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
35 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
36 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
39 * @(#)ext2_alloc.c 8.8 (Berkeley) 2/21/94
40 * $FreeBSD: src/sys/gnu/ext2fs/ext2_alloc.c,v 1.28.2.2 2002/07/01 00:18:51 iedowse Exp $
43 #include "opt_quota.h"
45 #include <sys/param.h>
46 #include <sys/systm.h>
48 #include <sys/vnode.h>
50 #include <sys/mount.h>
51 #include <sys/syslog.h>
53 #include <ufs/ufs/quota.h>
54 #include <ufs/ufs/inode.h>
55 #include <ufs/ufs/ufsmount.h>
57 #include <gnu/ext2fs/ext2_fs.h>
58 #include <gnu/ext2fs/ext2_fs_sb.h>
59 #include <gnu/ext2fs/fs.h>
60 #include <gnu/ext2fs/ext2_extern.h>
62 static void ext2_fserr __P((struct ext2_sb_info *, u_int, char *));
65 * Linux calls this functions at the following locations:
66 * (1) the inode is freed
67 * (2) a preallocation miss occurs
68 * (3) truncate is called
69 * (4) release_file is called and f_mode & 2
71 * I call it in ext2_inactive, ext2_truncate, ext2_vfree and in (2)
72 * the call in vfree might be redundant
75 ext2_discard_prealloc(ip)
78 #ifdef EXT2_PREALLOCATE
79 if (ip->i_prealloc_count) {
80 int i = ip->i_prealloc_count;
81 ip->i_prealloc_count = 0;
82 ext2_free_blocks (ITOV(ip)->v_mount,
90 * Allocate a block in the file system.
92 * this takes the framework from ffs_alloc. To implement the
93 * actual allocation, it calls ext2_new_block, the ported version
94 * of the same Linux routine.
96 * we note that this is always called in connection with ext2_blkpref
98 * preallocation is done as Linux does it
101 ext2_alloc(ip, lbn, bpref, size, cred, bnp)
102 register struct inode *ip;
108 register struct ext2_sb_info *fs;
117 if ((u_int)size > fs->s_blocksize || blkoff(fs, size) != 0) {
118 printf("dev = %s, bsize = %lu, size = %d, fs = %s\n",
119 devtoname(ip->i_dev), fs->s_blocksize, size, fs->fs_fsmnt);
120 panic("ext2_alloc: bad size");
123 panic("ext2_alloc: missing credential");
124 #endif /* DIAGNOSTIC */
125 if (size == fs->s_blocksize && fs->s_es->s_free_blocks_count == 0)
127 if (cred->cr_uid != 0 &&
128 fs->s_es->s_free_blocks_count < fs->s_es->s_r_blocks_count)
131 if ((error = chkdq(ip, (long)btodb(size), cred, 0)) != 0)
134 if (bpref >= fs->s_es->s_blocks_count)
136 /* call the Linux code */
137 #ifdef EXT2_PREALLOCATE
138 /* To have a preallocation hit, we must
139 * - have at least one block preallocated
140 * - and our preferred block must have that block number or one below
142 if (ip->i_prealloc_count &&
143 (bpref == ip->i_prealloc_block ||
144 bpref + 1 == ip->i_prealloc_block))
146 bno = ip->i_prealloc_block++;
147 ip->i_prealloc_count--;
148 /* ext2_debug ("preallocation hit (%lu/%lu).\n",
149 ++alloc_hits, ++alloc_attempts); */
151 /* Linux gets, clears, and releases the buffer at this
152 point - we don't have to that; we leave it to the caller
155 ext2_discard_prealloc (ip);
156 /* ext2_debug ("preallocation miss (%lu/%lu).\n",
157 alloc_hits, ++alloc_attempts); */
158 if (S_ISREG(ip->i_mode))
160 (ITOV(ip)->v_mount, bpref,
161 &ip->i_prealloc_count,
162 &ip->i_prealloc_block);
164 bno = (daddr_t)ext2_new_block(ITOV(ip)->v_mount,
168 bno = (daddr_t)ext2_new_block(ITOV(ip)->v_mount, bpref, 0, 0);
172 /* set next_alloc fields as done in block_getblk */
173 ip->i_next_alloc_block = lbn;
174 ip->i_next_alloc_goal = bno;
176 ip->i_blocks += btodb(size);
177 ip->i_flag |= IN_CHANGE | IN_UPDATE;
183 * Restore user's disk quota because allocation failed.
185 (void) chkdq(ip, (long)-btodb(size), cred, FORCE);
188 ext2_fserr(fs, cred->cr_uid, "file system full");
189 uprintf("\n%s: write failed, file system is full\n", fs->fs_fsmnt);
194 * Reallocate a sequence of blocks into a contiguous sequence of blocks.
196 * The vnode and an array of buffer pointers for a range of sequential
197 * logical blocks to be made contiguous is given. The allocator attempts
198 * to find a range of sequential blocks starting as close as possible to
199 * an fs_rotdelay offset from the end of the allocation for the logical
200 * block immediately preceeding the current range. If successful, the
201 * physical block numbers in the buffer pointers and in the inode are
202 * changed to reflect the new allocation. If unsuccessful, the allocation
203 * is left unchanged. The success in doing the reallocation is returned.
204 * Note that the error return is not reflected back to the user. Rather
205 * the previous block allocation will be used.
209 #include <sys/sysctl.h>
210 static int doasyncfree = 1;
212 SYSCTL_INT(_debug, 14, doasyncfree, CTLFLAG_RW, &doasyncfree, 0, "");
213 #endif /* OPT_DEBUG */
218 struct vop_reallocblks_args /* {
220 struct cluster_save *a_buflist;
223 #ifndef FANCY_REALLOC
224 /* printf("ext2_reallocblks not implemented\n"); */
228 struct ext2_sb_info *fs;
231 struct buf *sbp, *ebp;
232 daddr_t *bap, *sbap, *ebap;
233 struct cluster_save *buflist;
234 daddr_t start_lbn, end_lbn, soff, eoff, newblk, blkno;
235 struct indir start_ap[NIADDR + 1], end_ap[NIADDR + 1], *idp;
236 int i, len, start_lvl, end_lvl, pref, ssize;
242 if (fs->fs_contigsumsize <= 0)
245 buflist = ap->a_buflist;
246 len = buflist->bs_nchildren;
247 start_lbn = buflist->bs_children[0]->b_lblkno;
248 end_lbn = start_lbn + len - 1;
250 for (i = 1; i < len; i++)
251 if (buflist->bs_children[i]->b_lblkno != start_lbn + i)
252 panic("ext2_reallocblks: non-cluster");
255 * If the latest allocation is in a new cylinder group, assume that
256 * the filesystem has decided to move and do not force it back to
257 * the previous cylinder group.
259 if (dtog(fs, dbtofsb(fs, buflist->bs_children[0]->b_blkno)) !=
260 dtog(fs, dbtofsb(fs, buflist->bs_children[len - 1]->b_blkno)))
262 if (ufs_getlbns(vp, start_lbn, start_ap, &start_lvl) ||
263 ufs_getlbns(vp, end_lbn, end_ap, &end_lvl))
266 * Get the starting offset and block map for the first block.
268 if (start_lvl == 0) {
272 idp = &start_ap[start_lvl - 1];
273 if (bread(vp, idp->in_lbn, (int)fs->s_blocksize, NOCRED, &sbp)) {
277 sbap = (daddr_t *)sbp->b_data;
281 * Find the preferred location for the cluster.
283 pref = ext2_blkpref(ip, start_lbn, soff, sbap);
285 * If the block range spans two block maps, get the second map.
287 if (end_lvl == 0 || (idp = &end_ap[end_lvl - 1])->in_off + 1 >= len) {
291 if (start_ap[start_lvl-1].in_lbn == idp->in_lbn)
292 panic("ext2_reallocblk: start == end");
294 ssize = len - (idp->in_off + 1);
295 if (bread(vp, idp->in_lbn, (int)fs->s_blocksize, NOCRED, &ebp))
297 ebap = (daddr_t *)ebp->b_data;
300 * Search the block map looking for an allocation of the desired size.
302 if ((newblk = (daddr_t)ext2_hashalloc(ip, dtog(fs, pref), (long)pref,
303 len, (u_long (*)())ext2_clusteralloc)) == 0)
306 * We have found a new contiguous block.
308 * First we have to replace the old block pointers with the new
309 * block pointers in the inode and indirect blocks associated
313 for (bap = &sbap[soff], i = 0; i < len; i++, blkno += fs->s_frags_per_block) {
317 if (buflist->bs_children[i]->b_blkno != fsbtodb(fs, *bap))
318 panic("ext2_reallocblks: alloc mismatch");
323 * Next we must write out the modified inode and indirect blocks.
324 * For strict correctness, the writes should be synchronous since
325 * the old block values may have been written to disk. In practise
326 * they are almost never written, but if we are concerned about
327 * strict correctness, the `doasyncfree' flag should be set to zero.
329 * The test on `doasyncfree' should be changed to test a flag
330 * that shows whether the associated buffers and inodes have
331 * been written. The flag should be set when the cluster is
332 * started and cleared whenever the buffer or inode is flushed.
333 * We can then check below to see if it is set, and do the
334 * synchronous write only when it has been cleared.
336 if (sbap != &ip->i_db[0]) {
342 ip->i_flag |= IN_CHANGE | IN_UPDATE;
352 * Last, free the old blocks and assign the new blocks to the buffers.
354 for (blkno = newblk, i = 0; i < len; i++, blkno += fs->s_frags_per_block) {
355 ext2_blkfree(ip, dbtofsb(fs, buflist->bs_children[i]->b_blkno),
357 buflist->bs_children[i]->b_blkno = fsbtodb(fs, blkno);
364 if (sbap != &ip->i_db[0])
368 #endif /* FANCY_REALLOC */
372 * Allocate an inode in the file system.
374 * we leave the actual allocation strategy to the (modified)
375 * ext2_new_inode(), to make sure we get the policies right
378 ext2_valloc(pvp, mode, cred, vpp)
384 register struct inode *pip;
385 register struct ext2_sb_info *fs;
386 register struct inode *ip;
393 if (fs->s_es->s_free_inodes_count == 0)
396 /* call the Linux routine - it returns the inode number only */
397 ino = ext2_new_inode(pip, mode);
401 error = VFS_VGET(pvp->v_mount, ino, vpp);
403 UFS_VFREE(pvp, ino, mode);
409 the question is whether using VGET was such good idea at all -
410 Linux doesn't read the old inode in when it's allocating a
411 new one. I will set at least i_size & i_blocks the zero.
417 /* now we want to make sure that the block pointers are zeroed out */
418 for (i = 0; i < NDADDR; i++)
420 for (i = 0; i < NIADDR; i++)
424 * Set up a new generation number for this inode.
425 * XXX check if this makes sense in ext2
427 if (ip->i_gen == 0 || ++ip->i_gen == 0)
428 ip->i_gen = random() / 2 + 1;
430 printf("ext2_valloc: allocated inode %d\n", ino);
434 ext2_fserr(fs, cred->cr_uid, "out of inodes");
435 uprintf("\n%s: create/symlink failed, no inodes free\n", fs->fs_fsmnt);
440 * Select the desired position for the next block in a file.
442 * we try to mimic what Remy does in inode_getblk/block_getblk
444 * we note: blocknr == 0 means that we're about to allocate either
445 * a direct block or a pointer block at the first level of indirection
446 * (In other words, stuff that will go in i_db[] or i_ib[])
448 * blocknr != 0 means that we're allocating a block that is none
449 * of the above. Then, blocknr tells us the number of the block
450 * that will hold the pointer
453 ext2_blkpref(ip, lbn, indx, bap, blocknr)
462 /* if the next block is actually what we thought it is,
463 then set the goal to what we thought it should be
465 if(ip->i_next_alloc_block == lbn)
466 return ip->i_next_alloc_goal;
468 /* now check whether we were provided with an array that basically
469 tells us previous blocks to which we want to stay closeby
472 for (tmp = indx - 1; tmp >= 0; tmp--)
476 /* else let's fall back to the blocknr, or, if there is none,
477 follow the rule that a block should be allocated near its inode
479 return blocknr ? blocknr :
480 (daddr_t)(ip->i_block_group *
481 EXT2_BLOCKS_PER_GROUP(ip->i_e2fs)) +
482 ip->i_e2fs->s_es->s_first_data_block;
486 * Free a block or fragment.
488 * pass on to the Linux code
491 ext2_blkfree(ip, bno, size)
492 register struct inode *ip;
496 register struct ext2_sb_info *fs;
500 * call Linux code with mount *, block number, count
502 ext2_free_blocks(ITOV(ip)->v_mount, bno, size / fs->s_frag_size);
508 * the maintenance of the actual bitmaps is again up to the linux code
511 ext2_vfree(pvp, ino, mode)
516 register struct ext2_sb_info *fs;
517 register struct inode *pip;
518 register mode_t save_i_mode;
522 if ((u_int)ino > fs->s_inodes_per_group * fs->s_groups_count)
523 panic("ext2_vfree: range: dev = (%d, %d), ino = %d, fs = %s",
524 major(pip->i_dev), minor(pip->i_dev), ino, fs->fs_fsmnt);
526 /* ext2_debug("ext2_vfree (%d, %d) called\n", pip->i_number, mode);
528 ext2_discard_prealloc(pip);
530 /* we need to make sure that ext2_free_inode can adjust the
531 used_dir_counts in the group summary information - I'd
532 really like to know what the rationale behind this
533 'set i_mode to zero to denote an unused inode' is
535 save_i_mode = pip->i_mode;
537 ext2_free_inode(pip);
538 pip->i_mode = save_i_mode;
543 * Fserr prints the name of a file system with an error diagnostic.
545 * The form of the error message is:
549 ext2_fserr(fs, uid, cp)
550 struct ext2_sb_info *fs;
555 log(LOG_ERR, "uid %d on %s: %s\n", uid, fs->fs_fsmnt, cp);