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 <machine/inttypes.h>
59 #include "ext2mount.h"
62 #include "ext2_fs_sb.h"
64 #include "ext2_extern.h"
66 static void ext2_fserr (struct ext2_sb_info *, u_int, char *);
69 * Linux calls this functions at the following locations:
70 * (1) the inode is freed
71 * (2) a preallocation miss occurs
72 * (3) truncate is called
73 * (4) release_file is called and f_mode & 2
75 * I call it in ext2_inactive, ext2_truncate, ext2_vfree and in (2)
76 * the call in vfree might be redundant
79 ext2_discard_prealloc(struct inode *ip)
81 #ifdef EXT2_PREALLOCATE
82 if (ip->i_prealloc_count) {
83 int i = ip->i_prealloc_count;
84 ip->i_prealloc_count = 0;
85 ext2_free_blocks (ITOV(ip)->v_mount,
93 * Allocate a block in the file system.
95 * this takes the framework from ffs_alloc. To implement the
96 * actual allocation, it calls ext2_new_block, the ported version
97 * of the same Linux routine.
99 * we note that this is always called in connection with ext2_blkpref
101 * preallocation is done as Linux does it
104 ext2_alloc(struct inode *ip, daddr_t lbn, daddr_t bpref, int size,
105 struct ucred *cred, daddr_t *bnp)
107 struct ext2_sb_info *fs;
116 if ((u_int)size > fs->s_blocksize || blkoff(fs, size) != 0) {
117 kprintf("dev = %s, bsize = %lu, size = %d, fs = %s\n",
118 devtoname(ip->i_dev), fs->s_blocksize, size, fs->fs_fsmnt);
119 panic("ext2_alloc: bad size");
122 panic("ext2_alloc: missing credential");
123 #endif /* DIAGNOSTIC */
124 if (size == fs->s_blocksize && fs->s_es->s_free_blocks_count == 0)
126 if (cred->cr_uid != 0 &&
127 fs->s_es->s_free_blocks_count < fs->s_es->s_r_blocks_count)
130 if ((error = ext2_chkdq(ip, (long)btodb(size), cred, 0)) != 0)
133 if (bpref >= fs->s_es->s_blocks_count)
135 /* call the Linux code */
136 #ifdef EXT2_PREALLOCATE
137 /* To have a preallocation hit, we must
138 * - have at least one block preallocated
139 * - and our preferred block must have that block number or one below
141 if (ip->i_prealloc_count &&
142 (bpref == ip->i_prealloc_block ||
143 bpref + 1 == ip->i_prealloc_block))
145 bno = ip->i_prealloc_block++;
146 ip->i_prealloc_count--;
147 /* ext2_debug ("preallocation hit (%lu/%lu).\n",
148 ++alloc_hits, ++alloc_attempts); */
150 /* Linux gets, clears, and releases the buffer at this
151 point - we don't have to that; we leave it to the caller
154 ext2_discard_prealloc (ip);
155 /* ext2_debug ("preallocation miss (%lu/%lu).\n",
156 alloc_hits, ++alloc_attempts); */
157 if (S_ISREG(ip->i_mode))
159 (ITOV(ip)->v_mount, bpref,
160 &ip->i_prealloc_count,
161 &ip->i_prealloc_block);
163 bno = (daddr_t)ext2_new_block(ITOV(ip)->v_mount,
167 bno = (daddr_t)ext2_new_block(ITOV(ip)->v_mount, bpref, 0, 0);
171 /* set next_alloc fields as done in block_getblk */
172 ip->i_next_alloc_block = lbn;
173 ip->i_next_alloc_goal = bno;
175 ip->i_blocks += btodb(size);
176 ip->i_flag |= IN_CHANGE | IN_UPDATE;
182 * Restore user's disk quota because allocation failed.
184 ext2_chkdq(ip, (long)-btodb(size), cred, FORCE);
187 ext2_fserr(fs, cred->cr_uid, "file system full");
188 uprintf("\n%s: write failed, file system is full\n", fs->fs_fsmnt);
193 * Reallocate a sequence of blocks into a contiguous sequence of blocks.
195 * The vnode and an array of buffer pointers for a range of sequential
196 * logical blocks to be made contiguous is given. The allocator attempts
197 * to find a range of sequential blocks starting as close as possible to
198 * an fs_rotdelay offset from the end of the allocation for the logical
199 * block immediately preceeding the current range. If successful, the
200 * physical block numbers in the buffer pointers and in the inode are
201 * changed to reflect the new allocation. If unsuccessful, the allocation
202 * is left unchanged. The success in doing the reallocation is returned.
203 * Note that the error return is not reflected back to the user. Rather
204 * the previous block allocation will be used.
208 #include <sys/sysctl.h>
209 static int doasyncfree = 1;
211 SYSCTL_INT(_debug, 14, doasyncfree, CTLFLAG_RW, &doasyncfree, 0, "");
212 #endif /* OPT_DEBUG */
216 * ext2_reallocblks(struct vnode *a_vp, struct cluster_save *a_buflist)
219 ext2_reallocblks(struct vop_reallocblks_args *ap)
221 #ifndef FANCY_REALLOC
222 /* kprintf("ext2_reallocblks not implemented\n"); */
226 struct ext2_sb_info *fs;
229 struct buf *sbp, *ebp;
230 daddr_t *bap, *sbap, *ebap;
231 struct cluster_save *buflist;
232 daddr_t start_lbn, end_lbn, soff, eoff, newblk, blkno;
233 struct indir start_ap[NIADDR + 1], end_ap[NIADDR + 1], *idp;
234 int i, len, start_lvl, end_lvl, pref, ssize;
240 if (fs->fs_contigsumsize <= 0)
243 buflist = ap->a_buflist;
244 len = buflist->bs_nchildren;
245 start_lbn = lblkno(fs, buflist->bs_children[0]->b_loffset);
246 end_lbn = start_lbn + len - 1;
248 for (i = 1; i < len; i++) {
249 if (buflist->bs_children[i]->b_loffset != lblktodoff(fs, start_lbn) + lblktodoff(fs, i))
250 panic("ext2_reallocblks: non-cluster");
254 * If the latest allocation is in a new cylinder group, assume that
255 * the filesystem has decided to move and do not force it back to
256 * the previous cylinder group.
258 if (dtog(fs, dofftofsb(fs, buflist->bs_children[0]->b_bio2.bio_offset)) !=
259 dtog(fs, dofftofsb(fs, buflist->bs_children[len - 1]->b_bio2.bio_offset)))
261 if (ext2_getlbns(vp, start_lbn, start_ap, &start_lvl) ||
262 ext2_getlbns(vp, end_lbn, end_ap, &end_lvl))
265 * Get the starting offset and block map for the first block.
267 if (start_lvl == 0) {
271 idp = &start_ap[start_lvl - 1];
272 if (bread(vp, lblktodoff(fs, idp->in_lbn), (int)fs->s_blocksize, NOCRED, &sbp)) {
276 sbap = (daddr_t *)sbp->b_data;
280 * Find the preferred location for the cluster.
282 pref = ext2_blkpref(ip, start_lbn, soff, sbap);
284 * If the block range spans two block maps, get the second map.
286 if (end_lvl == 0 || (idp = &end_ap[end_lvl - 1])->in_off + 1 >= len) {
290 if (start_ap[start_lvl-1].in_lbn == idp->in_lbn)
291 panic("ext2_reallocblk: start == end");
293 ssize = len - (idp->in_off + 1);
294 if (bread(vp, lblktodoff(fs, idp->in_lbn), (int)fs->s_blocksize, NOCRED, &ebp))
296 ebap = (daddr_t *)ebp->b_data;
299 * Search the block map looking for an allocation of the desired size.
301 if ((newblk = (daddr_t)ext2_hashalloc(ip, dtog(fs, pref), (long)pref,
302 len, (u_long (*)())ext2_clusteralloc)) == 0)
305 * We have found a new contiguous block.
307 * First we have to replace the old block pointers with the new
308 * block pointers in the inode and indirect blocks associated
312 for (bap = &sbap[soff], i = 0; i < len; i++, blkno += fs->s_frags_per_block) {
316 if (buflist->bs_children[i]->b_bio2.bio_offset != fsbtodoff(fs, *bap))
317 panic("ext2_reallocblks: alloc mismatch");
322 * Next we must write out the modified inode and indirect blocks.
323 * For strict correctness, the writes should be synchronous since
324 * the old block values may have been written to disk. In practise
325 * they are almost never written, but if we are concerned about
326 * strict correctness, the `doasyncfree' flag should be set to zero.
328 * The test on `doasyncfree' should be changed to test a flag
329 * that shows whether the associated buffers and inodes have
330 * been written. The flag should be set when the cluster is
331 * started and cleared whenever the buffer or inode is flushed.
332 * We can then check below to see if it is set, and do the
333 * synchronous write only when it has been cleared.
335 if (sbap != &ip->i_db[0]) {
341 ip->i_flag |= IN_CHANGE | IN_UPDATE;
351 * Last, free the old blocks and assign the new blocks to the buffers.
353 for (blkno = newblk, i = 0; i < len; i++, blkno += fs->s_frags_per_block) {
354 ext2_blkfree(ip, dofftofsb(fs, buflist->bs_children[i]->b_bio2.bio_offset),
356 buflist->bs_children[i]->b_bio2.bio_offset = fsbtodoff(fs, blkno);
363 if (sbap != &ip->i_db[0])
367 #endif /* FANCY_REALLOC */
371 * Allocate an inode in the file system.
373 * we leave the actual allocation strategy to the (modified)
374 * ext2_new_inode(), to make sure we get the policies right
377 ext2_valloc(struct vnode *pvp, int mode, struct ucred *cred, struct vnode **vpp)
380 struct ext2_sb_info *fs;
388 if (fs->s_es->s_free_inodes_count == 0)
391 /* call the Linux routine - it returns the inode number only */
392 ino = ext2_new_inode(pip, mode);
396 error = VFS_VGET(pvp->v_mount, NULL, ino, vpp);
398 EXT2_VFREE(pvp, ino, mode);
404 the question is whether using VGET was such good idea at all -
405 Linux doesn't read the old inode in when it's allocating a
406 new one. I will set at least i_size & i_blocks the zero.
412 /* now we want to make sure that the block pointers are zeroed out */
413 for (i = 0; i < NDADDR; i++)
415 for (i = 0; i < NIADDR; i++)
419 * Set up a new generation number for this inode.
420 * XXX check if this makes sense in ext2
422 if (ip->i_gen == 0 || ++ip->i_gen == 0)
423 ip->i_gen = krandom() / 2 + 1;
425 kprintf("ext2_valloc: allocated inode %d\n", ino);
429 ext2_fserr(fs, cred->cr_uid, "out of inodes");
430 uprintf("\n%s: create/symlink failed, no inodes free\n", fs->fs_fsmnt);
435 * Select the desired position for the next block in a file.
437 * we try to mimic what Remy does in inode_getblk/block_getblk
439 * we note: blocknr == 0 means that we're about to allocate either
440 * a direct block or a pointer block at the first level of indirection
441 * (In other words, stuff that will go in i_db[] or i_ib[])
443 * blocknr != 0 means that we're allocating a block that is none
444 * of the above. Then, blocknr tells us the number of the block
445 * that will hold the pointer
448 ext2_blkpref(struct inode *ip, daddr_t lbn, int indx, daddr_t *bap,
454 * if the next block is actually what we thought it is,
455 * then set the goal to what we thought it should be
457 if (ip->i_next_alloc_block == lbn && ip->i_next_alloc_goal != 0)
458 return ip->i_next_alloc_goal;
460 /* now check whether we were provided with an array that basically
461 tells us previous blocks to which we want to stay closeby
464 for (tmp = indx - 1; tmp >= 0; tmp--)
469 * else let's fall back to the blocknr, or, if there is none,
470 * follow the rule that a block should be allocated near its inode
472 return blocknr ? blocknr :
473 (daddr_t)(ip->i_block_group *
474 EXT2_BLOCKS_PER_GROUP(ip->i_e2fs)) +
475 ip->i_e2fs->s_es->s_first_data_block;
479 * Free a block or fragment.
481 * pass on to the Linux code
484 ext2_blkfree(struct inode *ip, daddr_t bno, long size)
486 struct ext2_sb_info *fs;
490 * call Linux code with mount *, block number, count
492 ext2_free_blocks(ITOV(ip)->v_mount, bno, size / fs->s_frag_size);
498 * the maintenance of the actual bitmaps is again up to the linux code
501 ext2_vfree(struct vnode *pvp, ino_t ino, int mode)
503 struct ext2_sb_info *fs;
509 if ((u_int)ino > fs->s_inodes_per_group * fs->s_groups_count)
510 panic("ext2_vfree: range: dev = (%d, %d), ino = %"PRId64", fs = %s",
511 major(pip->i_dev), minor(pip->i_dev), ino, fs->fs_fsmnt);
513 /* ext2_debug("ext2_vfree (%d, %d) called\n", pip->i_number, mode);
515 ext2_discard_prealloc(pip);
517 /* we need to make sure that ext2_free_inode can adjust the
518 used_dir_counts in the group summary information - I'd
519 really like to know what the rationale behind this
520 'set i_mode to zero to denote an unused inode' is
522 save_i_mode = pip->i_mode;
524 ext2_free_inode(pip);
525 pip->i_mode = save_i_mode;
530 * Fserr prints the name of a file system with an error diagnostic.
532 * The form of the error message is:
536 ext2_fserr(struct ext2_sb_info *fs, u_int uid, char *cp)
538 log(LOG_ERR, "uid %d on %s: %s\n", uid, fs->fs_fsmnt, cp);