xref: /freebsd/sys/fs/ext2fs/ext2_balloc.c (revision 0957b409)
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_balloc.c	8.4 (Berkeley) 9/23/93
38  * $FreeBSD$
39  */
40 
41 #include <sys/param.h>
42 #include <sys/systm.h>
43 #include <sys/bio.h>
44 #include <sys/buf.h>
45 #include <sys/limits.h>
46 #include <sys/lock.h>
47 #include <sys/mount.h>
48 #include <sys/vnode.h>
49 
50 #include <fs/ext2fs/fs.h>
51 #include <fs/ext2fs/inode.h>
52 #include <fs/ext2fs/ext2fs.h>
53 #include <fs/ext2fs/ext2_dinode.h>
54 #include <fs/ext2fs/ext2_extern.h>
55 #include <fs/ext2fs/ext2_mount.h>
56 
57 static int
58 ext2_ext_balloc(struct inode *ip, uint32_t lbn, int size,
59     struct ucred *cred, struct buf **bpp, int flags)
60 {
61 	struct m_ext2fs *fs;
62 	struct buf *bp = NULL;
63 	struct vnode *vp = ITOV(ip);
64 	daddr_t newblk;
65 	int osize, nsize, blks, error, allocated;
66 
67 	fs = ip->i_e2fs;
68 	blks = howmany(size, fs->e2fs_bsize);
69 
70 	error = ext4_ext_get_blocks(ip, lbn, blks, cred, NULL, &allocated, &newblk);
71 	if (error)
72 		return (error);
73 
74 	if (allocated) {
75 		if (ip->i_size < (lbn + 1) * fs->e2fs_bsize)
76 			nsize = fragroundup(fs, size);
77 		else
78 			nsize = fs->e2fs_bsize;
79 
80 		bp = getblk(vp, lbn, nsize, 0, 0, 0);
81 		if(!bp)
82 			return (EIO);
83 
84 		bp->b_blkno = fsbtodb(fs, newblk);
85 		if (flags & BA_CLRBUF)
86 			vfs_bio_clrbuf(bp);
87 	} else {
88 		if (ip->i_size >= (lbn + 1) * fs->e2fs_bsize) {
89 
90 			error = bread(vp, lbn, fs->e2fs_bsize, NOCRED, &bp);
91 			if (error) {
92 				brelse(bp);
93 				return (error);
94 			}
95 			bp->b_blkno = fsbtodb(fs, newblk);
96 			*bpp = bp;
97 			return (0);
98 		}
99 
100 		/*
101 		 * Consider need to reallocate a fragment.
102 		 */
103 		osize = fragroundup(fs, blkoff(fs, ip->i_size));
104 		nsize = fragroundup(fs, size);
105 		if (nsize <= osize)
106 			error = bread(vp, lbn, osize, NOCRED, &bp);
107 		else
108 			error = bread(vp, lbn, fs->e2fs_bsize, NOCRED, &bp);
109 		if (error) {
110 			brelse(bp);
111 			return (error);
112 		}
113 		bp->b_blkno = fsbtodb(fs, newblk);
114 	}
115 
116 	*bpp = bp;
117 
118 	return (error);
119 }
120 
121 /*
122  * Balloc defines the structure of filesystem storage
123  * by allocating the physical blocks on a device given
124  * the inode and the logical block number in a file.
125  */
126 int
127 ext2_balloc(struct inode *ip, e2fs_lbn_t lbn, int size, struct ucred *cred,
128     struct buf **bpp, int flags)
129 {
130 	struct m_ext2fs *fs;
131 	struct ext2mount *ump;
132 	struct buf *bp, *nbp;
133 	struct vnode *vp = ITOV(ip);
134 	struct indir indirs[EXT2_NIADDR + 2];
135 	e4fs_daddr_t nb, newb;
136 	e2fs_daddr_t *bap, pref;
137 	int osize, nsize, num, i, error;
138 
139 	*bpp = NULL;
140 	if (lbn < 0)
141 		return (EFBIG);
142 	fs = ip->i_e2fs;
143 	ump = ip->i_ump;
144 
145 	/*
146 	 * check if this is a sequential block allocation.
147 	 * If so, increment next_alloc fields to allow ext2_blkpref
148 	 * to make a good guess
149 	 */
150 	if (lbn == ip->i_next_alloc_block + 1) {
151 		ip->i_next_alloc_block++;
152 		ip->i_next_alloc_goal++;
153 	}
154 
155 	if (ip->i_flag & IN_E4EXTENTS)
156 		return (ext2_ext_balloc(ip, lbn, size, cred, bpp, flags));
157 
158 	/*
159 	 * The first EXT2_NDADDR blocks are direct blocks
160 	 */
161 	if (lbn < EXT2_NDADDR) {
162 		nb = ip->i_db[lbn];
163 		/*
164 		 * no new block is to be allocated, and no need to expand
165 		 * the file
166 		 */
167 		if (nb != 0 && ip->i_size >= (lbn + 1) * fs->e2fs_bsize) {
168 			error = bread(vp, lbn, fs->e2fs_bsize, NOCRED, &bp);
169 			if (error) {
170 				brelse(bp);
171 				return (error);
172 			}
173 			bp->b_blkno = fsbtodb(fs, nb);
174 			*bpp = bp;
175 			return (0);
176 		}
177 		if (nb != 0) {
178 			/*
179 			 * Consider need to reallocate a fragment.
180 			 */
181 			osize = fragroundup(fs, blkoff(fs, ip->i_size));
182 			nsize = fragroundup(fs, size);
183 			if (nsize <= osize) {
184 				error = bread(vp, lbn, osize, NOCRED, &bp);
185 				if (error) {
186 					brelse(bp);
187 					return (error);
188 				}
189 				bp->b_blkno = fsbtodb(fs, nb);
190 			} else {
191 				/*
192 				 * Godmar thinks: this shouldn't happen w/o
193 				 * fragments
194 				 */
195 				printf("nsize %d(%d) > osize %d(%d) nb %d\n",
196 				    (int)nsize, (int)size, (int)osize,
197 				    (int)ip->i_size, (int)nb);
198 				panic(
199 				    "ext2_balloc: Something is terribly wrong");
200 /*
201  * please note there haven't been any changes from here on -
202  * FFS seems to work.
203  */
204 			}
205 		} else {
206 			if (ip->i_size < (lbn + 1) * fs->e2fs_bsize)
207 				nsize = fragroundup(fs, size);
208 			else
209 				nsize = fs->e2fs_bsize;
210 			EXT2_LOCK(ump);
211 			error = ext2_alloc(ip, lbn,
212 			    ext2_blkpref(ip, lbn, (int)lbn, &ip->i_db[0], 0),
213 			    nsize, cred, &newb);
214 			if (error)
215 				return (error);
216 			/*
217 			 * If the newly allocated block exceeds 32-bit limit,
218 			 * we can not use it in file block maps.
219 			 */
220 			if (newb > UINT_MAX)
221 				return (EFBIG);
222 			bp = getblk(vp, lbn, nsize, 0, 0, 0);
223 			bp->b_blkno = fsbtodb(fs, newb);
224 			if (flags & BA_CLRBUF)
225 				vfs_bio_clrbuf(bp);
226 		}
227 		ip->i_db[lbn] = dbtofsb(fs, bp->b_blkno);
228 		ip->i_flag |= IN_CHANGE | IN_UPDATE;
229 		*bpp = bp;
230 		return (0);
231 	}
232 	/*
233 	 * Determine the number of levels of indirection.
234 	 */
235 	pref = 0;
236 	if ((error = ext2_getlbns(vp, lbn, indirs, &num)) != 0)
237 		return (error);
238 #ifdef INVARIANTS
239 	if (num < 1)
240 		panic("ext2_balloc: ext2_getlbns returned indirect block");
241 #endif
242 	/*
243 	 * Fetch the first indirect block allocating if necessary.
244 	 */
245 	--num;
246 	nb = ip->i_ib[indirs[0].in_off];
247 	if (nb == 0) {
248 		EXT2_LOCK(ump);
249 		pref = ext2_blkpref(ip, lbn, indirs[0].in_off +
250 		    EXT2_NDIR_BLOCKS, &ip->i_db[0], 0);
251 		if ((error = ext2_alloc(ip, lbn, pref, fs->e2fs_bsize, cred,
252 		    &newb)))
253 			return (error);
254 		if (newb > UINT_MAX)
255 			return (EFBIG);
256 		nb = newb;
257 		bp = getblk(vp, indirs[1].in_lbn, fs->e2fs_bsize, 0, 0, 0);
258 		bp->b_blkno = fsbtodb(fs, newb);
259 		vfs_bio_clrbuf(bp);
260 		/*
261 		 * Write synchronously so that indirect blocks
262 		 * never point at garbage.
263 		 */
264 		if ((error = bwrite(bp)) != 0) {
265 			ext2_blkfree(ip, nb, fs->e2fs_bsize);
266 			return (error);
267 		}
268 		ip->i_ib[indirs[0].in_off] = newb;
269 		ip->i_flag |= IN_CHANGE | IN_UPDATE;
270 	}
271 	/*
272 	 * Fetch through the indirect blocks, allocating as necessary.
273 	 */
274 	for (i = 1;;) {
275 		error = bread(vp,
276 		    indirs[i].in_lbn, (int)fs->e2fs_bsize, NOCRED, &bp);
277 		if (error) {
278 			brelse(bp);
279 			return (error);
280 		}
281 		bap = (e2fs_daddr_t *)bp->b_data;
282 		nb = bap[indirs[i].in_off];
283 		if (i == num)
284 			break;
285 		i += 1;
286 		if (nb != 0) {
287 			bqrelse(bp);
288 			continue;
289 		}
290 		EXT2_LOCK(ump);
291 		if (pref == 0)
292 			pref = ext2_blkpref(ip, lbn, indirs[i].in_off, bap,
293 			    bp->b_lblkno);
294 		error = ext2_alloc(ip, lbn, pref, (int)fs->e2fs_bsize, cred, &newb);
295 		if (error) {
296 			brelse(bp);
297 			return (error);
298 		}
299 		if (newb > UINT_MAX)
300 			return (EFBIG);
301 		nb = newb;
302 		nbp = getblk(vp, indirs[i].in_lbn, fs->e2fs_bsize, 0, 0, 0);
303 		nbp->b_blkno = fsbtodb(fs, nb);
304 		vfs_bio_clrbuf(nbp);
305 		/*
306 		 * Write synchronously so that indirect blocks
307 		 * never point at garbage.
308 		 */
309 		if ((error = bwrite(nbp)) != 0) {
310 			ext2_blkfree(ip, nb, fs->e2fs_bsize);
311 			EXT2_UNLOCK(ump);
312 			brelse(bp);
313 			return (error);
314 		}
315 		bap[indirs[i - 1].in_off] = nb;
316 		/*
317 		 * If required, write synchronously, otherwise use
318 		 * delayed write.
319 		 */
320 		if (flags & IO_SYNC) {
321 			bwrite(bp);
322 		} else {
323 			if (bp->b_bufsize == fs->e2fs_bsize)
324 				bp->b_flags |= B_CLUSTEROK;
325 			bdwrite(bp);
326 		}
327 	}
328 	/*
329 	 * Get the data block, allocating if necessary.
330 	 */
331 	if (nb == 0) {
332 		EXT2_LOCK(ump);
333 		pref = ext2_blkpref(ip, lbn, indirs[i].in_off, &bap[0],
334 		    bp->b_lblkno);
335 		if ((error = ext2_alloc(ip,
336 		    lbn, pref, (int)fs->e2fs_bsize, cred, &newb)) != 0) {
337 			brelse(bp);
338 			return (error);
339 		}
340 		if (newb > UINT_MAX)
341 			return (EFBIG);
342 		nb = newb;
343 		nbp = getblk(vp, lbn, fs->e2fs_bsize, 0, 0, 0);
344 		nbp->b_blkno = fsbtodb(fs, nb);
345 		if (flags & BA_CLRBUF)
346 			vfs_bio_clrbuf(nbp);
347 		bap[indirs[i].in_off] = nb;
348 		/*
349 		 * If required, write synchronously, otherwise use
350 		 * delayed write.
351 		 */
352 		if (flags & IO_SYNC) {
353 			bwrite(bp);
354 		} else {
355 			if (bp->b_bufsize == fs->e2fs_bsize)
356 				bp->b_flags |= B_CLUSTEROK;
357 			bdwrite(bp);
358 		}
359 		*bpp = nbp;
360 		return (0);
361 	}
362 	brelse(bp);
363 	if (flags & BA_CLRBUF) {
364 		int seqcount = (flags & BA_SEQMASK) >> BA_SEQSHIFT;
365 
366 		if (seqcount && (vp->v_mount->mnt_flag & MNT_NOCLUSTERR) == 0) {
367 			error = cluster_read(vp, ip->i_size, lbn,
368 			    (int)fs->e2fs_bsize, NOCRED,
369 			    MAXBSIZE, seqcount, 0, &nbp);
370 		} else {
371 			error = bread(vp, lbn, (int)fs->e2fs_bsize, NOCRED, &nbp);
372 		}
373 		if (error) {
374 			brelse(nbp);
375 			return (error);
376 		}
377 	} else {
378 		nbp = getblk(vp, lbn, fs->e2fs_bsize, 0, 0, 0);
379 		nbp->b_blkno = fsbtodb(fs, nb);
380 	}
381 	*bpp = nbp;
382 	return (0);
383 }
384