xref: /dragonfly/sys/vfs/ext2fs/ext2_balloc.c (revision 37de577a)
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/endian.h>
44 #include <sys/bio.h>
45 #include <sys/buf2.h>
46 #include <sys/limits.h>
47 #include <sys/lock.h>
48 #include <sys/mount.h>
49 #include <sys/vnode.h>
50 #include <sys/mutex2.h>
51 
52 #include <vfs/ext2fs/fs.h>
53 #include <vfs/ext2fs/inode.h>
54 #include <vfs/ext2fs/ext2fs.h>
55 #include <vfs/ext2fs/ext2_dinode.h>
56 #include <vfs/ext2fs/ext2_extern.h>
57 #include <vfs/ext2fs/ext2_mount.h>
58 
59 static int
60 ext2_ext_balloc(struct inode *ip, uint32_t lbn, int size,
61     struct ucred *cred, struct buf **bpp, int flags)
62 {
63 	return (EINVAL);
64 }
65 
66 /*
67  * Balloc defines the structure of filesystem storage
68  * by allocating the physical blocks on a device given
69  * the inode and the logical block number in a file.
70  */
71 int
72 ext2_balloc(struct inode *ip, e2fs_lbn_t lbn, int size, struct ucred *cred,
73     struct buf **bpp, int flags)
74 {
75 	struct m_ext2fs *fs;
76 	struct ext2mount *ump;
77 	struct buf *bp, *nbp;
78 	struct vnode *vp = ITOV(ip);
79 	struct indir indirs[EXT2_NIADDR + 2];
80 	e4fs_daddr_t nb, newb;
81 	e2fs_daddr_t *bap, pref;
82 	int num, i, error;
83 
84 	*bpp = NULL;
85 	if (lbn < 0)
86 		return (EFBIG);
87 	fs = ip->i_e2fs;
88 	ump = ip->i_ump;
89 
90 	/*
91 	 * check if this is a sequential block allocation.
92 	 * If so, increment next_alloc fields to allow ext2_blkpref
93 	 * to make a good guess
94 	 */
95 	if (lbn == ip->i_next_alloc_block + 1) {
96 		ip->i_next_alloc_block++;
97 		ip->i_next_alloc_goal++;
98 	}
99 
100 	if (ip->i_flag & IN_E4EXTENTS)
101 		return (ext2_ext_balloc(ip, lbn, size, cred, bpp, flags));
102 
103 	/*
104 	 * The first EXT2_NDADDR blocks are direct blocks
105 	 */
106 	if (lbn < EXT2_NDADDR) {
107 		nb = ip->i_db[lbn];
108 		/*
109 		 * no new block is to be allocated, and no need to expand
110 		 * the file
111 		 */
112 		if (nb != 0) {
113 			error = bread(vp, lblktodoff(fs, lbn), fs->e2fs_bsize,
114 			    &bp);
115 			if (error) {
116 				brelse(bp);
117 				return (error);
118 			}
119 			bp->b_bio2.bio_offset = fsbtodoff(fs, nb);
120 			if (ip->i_size >= (lbn + 1) * fs->e2fs_bsize) {
121 				*bpp = bp;
122 				return (0);
123 			}
124 		} else {
125 			EXT2_LOCK(ump);
126 			error = ext2_alloc(ip, lbn,
127 			    ext2_blkpref(ip, lbn, (int)lbn, &ip->i_db[0], 0),
128 			    fs->e2fs_bsize, cred, &newb);
129 			if (error)
130 				return (error);
131 			/*
132 			 * If the newly allocated block exceeds 32-bit limit,
133 			 * we can not use it in file block maps.
134 			 */
135 			if (newb > UINT_MAX)
136 				return (EFBIG);
137 			bp = getblk(vp, lblktodoff(fs, lbn), fs->e2fs_bsize, 0, 0);
138 			bp->b_bio2.bio_offset = fsbtodoff(fs, newb);
139 			if (flags & BA_CLRBUF)
140 				vfs_bio_clrbuf(bp);
141 		}
142 		ip->i_db[lbn] = dofftofsb(fs, bp->b_bio2.bio_offset);
143 		ip->i_flag |= IN_CHANGE | IN_UPDATE;
144 		*bpp = bp;
145 		return (0);
146 	}
147 	/*
148 	 * Determine the number of levels of indirection.
149 	 */
150 	pref = 0;
151 	if ((error = ext2_getlbns(vp, lbn, indirs, &num)) != 0)
152 		return (error);
153 #ifdef INVARIANTS
154 	if (num < 1)
155 		panic("ext2_balloc: ext2_getlbns returned indirect block");
156 #endif
157 	/*
158 	 * Fetch the first indirect block allocating if necessary.
159 	 */
160 	--num;
161 	nb = ip->i_ib[indirs[0].in_off];
162 	if (nb == 0) {
163 		EXT2_LOCK(ump);
164 		pref = ext2_blkpref(ip, lbn, indirs[0].in_off +
165 		    EXT2_NDIR_BLOCKS, &ip->i_db[0], 0);
166 		if ((error = ext2_alloc(ip, lbn, pref, fs->e2fs_bsize, cred,
167 		    &newb)))
168 			return (error);
169 		if (newb > UINT_MAX)
170 			return (EFBIG);
171 		nb = newb;
172 		bp = getblk(vp, lblktodoff(fs, indirs[1].in_lbn),
173 		    fs->e2fs_bsize, 0, 0);
174 		bp->b_bio2.bio_offset = fsbtodoff(fs, newb);
175 		vfs_bio_clrbuf(bp);
176 		/*
177 		 * Write synchronously so that indirect blocks
178 		 * never point at garbage.
179 		 */
180 		if ((error = bwrite(bp)) != 0) {
181 			ext2_blkfree(ip, nb, fs->e2fs_bsize);
182 			return (error);
183 		}
184 		ip->i_ib[indirs[0].in_off] = newb;
185 		ip->i_flag |= IN_CHANGE | IN_UPDATE;
186 	}
187 	/*
188 	 * Fetch through the indirect blocks, allocating as necessary.
189 	 */
190 	for (i = 1;;) {
191 		error = bread(vp, lblktodoff(fs, indirs[i].in_lbn),
192 		    (int)fs->e2fs_bsize, &bp);
193 		if (error) {
194 			brelse(bp);
195 			return (error);
196 		}
197 		bap = (e2fs_daddr_t *)bp->b_data;
198 		nb = le32toh(bap[indirs[i].in_off]);
199 		if (i == num)
200 			break;
201 		i += 1;
202 		if (nb != 0) {
203 			bqrelse(bp);
204 			continue;
205 		}
206 		EXT2_LOCK(ump);
207 		if (pref == 0)
208 			pref = ext2_blkpref(ip, lbn, indirs[i].in_off, bap,
209 			    lblkno(fs, bp->b_loffset));
210 		error = ext2_alloc(ip, lbn, pref, (int)fs->e2fs_bsize, cred, &newb);
211 		if (error) {
212 			brelse(bp);
213 			return (error);
214 		}
215 		if (newb > UINT_MAX) {
216 			brelse(bp);
217 			return (EFBIG);
218 		}
219 		nb = newb;
220 		nbp = getblk(vp, lblktodoff(fs, indirs[i].in_lbn),
221 		    fs->e2fs_bsize, 0, 0);
222 		nbp->b_bio2.bio_offset = fsbtodoff(fs, nb);
223 		vfs_bio_clrbuf(nbp);
224 		/*
225 		 * Write synchronously so that indirect blocks
226 		 * never point at garbage.
227 		 */
228 		if ((error = bwrite(nbp)) != 0) {
229 			ext2_blkfree(ip, nb, fs->e2fs_bsize);
230 			brelse(bp);
231 			return (error);
232 		}
233 		bap[indirs[i - 1].in_off] = htole32(nb);
234 		/*
235 		 * If required, write synchronously, otherwise use
236 		 * delayed write.
237 		 */
238 		if (flags & IO_SYNC) {
239 			bwrite(bp);
240 		} else {
241 			if (bp->b_bufsize == fs->e2fs_bsize)
242 				bp->b_flags |= B_CLUSTEROK;
243 			bdwrite(bp);
244 		}
245 	}
246 	/*
247 	 * Get the data block, allocating if necessary.
248 	 */
249 	if (nb == 0) {
250 		EXT2_LOCK(ump);
251 		pref = ext2_blkpref(ip, lbn, indirs[i].in_off, &bap[0],
252 		    lblkno(fs, bp->b_loffset));
253 		if ((error = ext2_alloc(ip,
254 		    lbn, pref, (int)fs->e2fs_bsize, cred, &newb)) != 0) {
255 			brelse(bp);
256 			return (error);
257 		}
258 		if (newb > UINT_MAX) {
259 			brelse(bp);
260 			return (EFBIG);
261 		}
262 		nb = newb;
263 		nbp = getblk(vp, lblktodoff(fs, lbn), fs->e2fs_bsize, 0, 0);
264 		nbp->b_bio2.bio_offset = fsbtodoff(fs, nb);
265 		if (flags & BA_CLRBUF)
266 			vfs_bio_clrbuf(nbp);
267 		bap[indirs[i].in_off] = htole32(nb);
268 		/*
269 		 * If required, write synchronously, otherwise use
270 		 * delayed write.
271 		 */
272 		if (flags & IO_SYNC) {
273 			bwrite(bp);
274 		} else {
275 			if (bp->b_bufsize == fs->e2fs_bsize)
276 				bp->b_flags |= B_CLUSTEROK;
277 			bdwrite(bp);
278 		}
279 		*bpp = nbp;
280 		return (0);
281 	}
282 	brelse(bp);
283 	if (flags & BA_CLRBUF) {
284 		int seqcount = (flags & BA_SEQMASK) >> BA_SEQSHIFT;
285 
286 		if (seqcount && (vp->v_mount->mnt_flag & MNT_NOCLUSTERR) == 0) {
287 			error = cluster_read(vp, ip->i_size,
288 			    lblktodoff(fs, lbn), (int)fs->e2fs_bsize,
289 			    MAXBSIZE, (size_t)seqcount, &nbp);
290 		} else {
291 			error = bread(vp, lblktodoff(fs, lbn),
292 			    (int)fs->e2fs_bsize, &nbp);
293 		}
294 		if (error) {
295 			brelse(nbp);
296 			return (error);
297 		}
298 	} else {
299 		nbp = getblk(vp, lblktodoff(fs, lbn), fs->e2fs_bsize, 0, 0);
300 		nbp->b_bio2.bio_offset = fsbtodoff(fs, nb);
301 	}
302 	*bpp = nbp;
303 	return (0);
304 }
305