xref: /freebsd/sys/fs/ext2fs/ext2_bmap.c (revision a98ff317388a00b992f1bf8404dee596f9383f5e)
1 /*-
2  * Copyright (c) 1989, 1991, 1993
3  *	The Regents of the University of California.  All rights reserved.
4  * (c) UNIX System Laboratories, Inc.
5  * All or some portions of this file are derived from material licensed
6  * to the University of California by American Telephone and Telegraph
7  * Co. or Unix System Laboratories, Inc. and are reproduced herein with
8  * the permission of UNIX System Laboratories, Inc.
9  *
10  * Redistribution and use in source and binary forms, with or without
11  * modification, are permitted provided that the following conditions
12  * are met:
13  * 1. Redistributions of source code must retain the above copyright
14  *    notice, this list of conditions and the following disclaimer.
15  * 2. Redistributions in binary form must reproduce the above copyright
16  *    notice, this list of conditions and the following disclaimer in the
17  *    documentation and/or other materials provided with the distribution.
18  * 4. Neither the name of the University nor the names of its contributors
19  *    may be used to endorse or promote products derived from this software
20  *    without specific prior written permission.
21  *
22  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
23  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
24  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
25  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
26  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
27  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
28  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
29  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
30  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
31  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
32  * SUCH DAMAGE.
33  *
34  *	@(#)ufs_bmap.c	8.7 (Berkeley) 3/21/95
35  * $FreeBSD$
36  */
37 
38 #include <sys/param.h>
39 #include <sys/systm.h>
40 #include <sys/bio.h>
41 #include <sys/buf.h>
42 #include <sys/proc.h>
43 #include <sys/vnode.h>
44 #include <sys/mount.h>
45 #include <sys/resourcevar.h>
46 #include <sys/stat.h>
47 
48 #include <fs/ext2fs/inode.h>
49 #include <fs/ext2fs/ext2fs.h>
50 #include <fs/ext2fs/ext2_extern.h>
51 #include <fs/ext2fs/ext2_mount.h>
52 
53 /*
54  * Bmap converts the logical block number of a file to its physical block
55  * number on the disk. The conversion is done by using the logical block
56  * number to index into the array of block pointers described by the dinode.
57  */
58 int
59 ext2_bmap(struct vop_bmap_args *ap)
60 {
61 	int32_t blkno;
62 	int error;
63 
64 	/*
65 	 * Check for underlying vnode requests and ensure that logical
66 	 * to physical mapping is requested.
67 	 */
68 	if (ap->a_bop != NULL)
69 		*ap->a_bop = &VTOI(ap->a_vp)->i_devvp->v_bufobj;
70 	if (ap->a_bnp == NULL)
71 		return (0);
72 
73 	error = ext2_bmaparray(ap->a_vp, ap->a_bn, &blkno,
74 	    ap->a_runp, ap->a_runb);
75 	*ap->a_bnp = blkno;
76 	return (error);
77 }
78 
79 /*
80  * Indirect blocks are now on the vnode for the file.  They are given negative
81  * logical block numbers.  Indirect blocks are addressed by the negative
82  * address of the first data block to which they point.  Double indirect blocks
83  * are addressed by one less than the address of the first indirect block to
84  * which they point.  Triple indirect blocks are addressed by one less than
85  * the address of the first double indirect block to which they point.
86  *
87  * ext2_bmaparray does the bmap conversion, and if requested returns the
88  * array of logical blocks which must be traversed to get to a block.
89  * Each entry contains the offset into that block that gets you to the
90  * next block and the disk address of the block (if it is assigned).
91  */
92 
93 int
94 ext2_bmaparray(struct vnode *vp, int32_t bn, int32_t *bnp, int *runp, int *runb)
95 {
96 	struct inode *ip;
97 	struct buf *bp;
98 	struct ext2mount *ump;
99 	struct mount *mp;
100 	struct vnode *devvp;
101 	struct indir a[NIADDR+1], *ap;
102 	daddr_t daddr;
103 	e2fs_lbn_t metalbn;
104 	int error, num, maxrun = 0, bsize;
105 	int *nump;
106 
107 	ap = NULL;
108 	ip = VTOI(vp);
109 	mp = vp->v_mount;
110 	ump = VFSTOEXT2(mp);
111 	devvp = ump->um_devvp;
112 
113 	bsize = EXT2_BLOCK_SIZE(ump->um_e2fs);
114 
115 	if (runp) {
116 		maxrun = mp->mnt_iosize_max / bsize - 1;
117 		*runp = 0;
118 	}
119 
120 	if (runb) {
121 		*runb = 0;
122 	}
123 
124 
125 	ap = a;
126 	nump = &num;
127 	error = ext2_getlbns(vp, bn, ap, nump);
128 	if (error)
129 		return (error);
130 
131 	num = *nump;
132 	if (num == 0) {
133 		*bnp = blkptrtodb(ump, ip->i_db[bn]);
134 		if (*bnp == 0) {
135 			*bnp = -1;
136 		} else if (runp) {
137 			int32_t bnb = bn;
138 			for (++bn; bn < NDADDR && *runp < maxrun &&
139 			    is_sequential(ump, ip->i_db[bn - 1], ip->i_db[bn]);
140 			    ++bn, ++*runp);
141 			bn = bnb;
142 			if (runb && (bn > 0)) {
143 				for (--bn; (bn >= 0) && (*runb < maxrun) &&
144 					is_sequential(ump, ip->i_db[bn],
145 						ip->i_db[bn+1]);
146 						--bn, ++*runb);
147 			}
148 		}
149 		return (0);
150 	}
151 
152 
153 	/* Get disk address out of indirect block array */
154 	daddr = ip->i_ib[ap->in_off];
155 
156 	for (bp = NULL, ++ap; --num; ++ap) {
157 		/*
158 		 * Exit the loop if there is no disk address assigned yet and
159 		 * the indirect block isn't in the cache, or if we were
160 		 * looking for an indirect block and we've found it.
161 		 */
162 
163 		metalbn = ap->in_lbn;
164 		if ((daddr == 0 && !incore(&vp->v_bufobj, metalbn)) || metalbn == bn)
165 			break;
166 		/*
167 		 * If we get here, we've either got the block in the cache
168 		 * or we have a disk address for it, go fetch it.
169 		 */
170 		if (bp)
171 			bqrelse(bp);
172 
173 		bp = getblk(vp, metalbn, bsize, 0, 0, 0);
174 		if ((bp->b_flags & B_CACHE) == 0) {
175 #ifdef INVARIANTS
176 			if (!daddr)
177 				panic("ext2_bmaparray: indirect block not in cache");
178 #endif
179 			bp->b_blkno = blkptrtodb(ump, daddr);
180 			bp->b_iocmd = BIO_READ;
181 			bp->b_flags &= ~B_INVAL;
182 			bp->b_ioflags &= ~BIO_ERROR;
183 			vfs_busy_pages(bp, 0);
184 			bp->b_iooffset = dbtob(bp->b_blkno);
185 			bstrategy(bp);
186 			curthread->td_ru.ru_inblock++;
187 			error = bufwait(bp);
188 			if (error) {
189 				brelse(bp);
190 				return (error);
191 			}
192 		}
193 
194 		daddr = ((int32_t *)bp->b_data)[ap->in_off];
195 		if (num == 1 && daddr && runp) {
196 			for (bn = ap->in_off + 1;
197 			    bn < MNINDIR(ump) && *runp < maxrun &&
198 			    is_sequential(ump,
199 			    ((int32_t *)bp->b_data)[bn - 1],
200 			    ((int32_t *)bp->b_data)[bn]);
201 			    ++bn, ++*runp);
202 			bn = ap->in_off;
203 			if (runb && bn) {
204 				for (--bn; bn >= 0 && *runb < maxrun &&
205 			    		is_sequential(ump, ((int32_t *)bp->b_data)[bn],
206 					    ((int32_t *)bp->b_data)[bn+1]);
207 			    		--bn, ++*runb);
208 			}
209 		}
210 	}
211 	if (bp)
212 		bqrelse(bp);
213 
214 	/*
215 	 * Since this is FFS independent code, we are out of scope for the
216 	 * definitions of BLK_NOCOPY and BLK_SNAP, but we do know that they
217 	 * will fall in the range 1..um_seqinc, so we use that test and
218 	 * return a request for a zeroed out buffer if attempts are made
219 	 * to read a BLK_NOCOPY or BLK_SNAP block.
220 	 */
221 	if ((ip->i_flags & SF_SNAPSHOT) && daddr > 0 && daddr < ump->um_seqinc){
222 		*bnp = -1;
223 		return (0);
224 	}
225 	*bnp = blkptrtodb(ump, daddr);
226 	if (*bnp == 0) {
227 		*bnp = -1;
228 	}
229 	return (0);
230 }
231 
232 /*
233  * Create an array of logical block number/offset pairs which represent the
234  * path of indirect blocks required to access a data block.  The first "pair"
235  * contains the logical block number of the appropriate single, double or
236  * triple indirect block and the offset into the inode indirect block array.
237  * Note, the logical block number of the inode single/double/triple indirect
238  * block appears twice in the array, once with the offset into the i_ib and
239  * once with the offset into the page itself.
240  */
241 int
242 ext2_getlbns(struct vnode *vp, int32_t bn, struct indir *ap, int *nump)
243 {
244 	long blockcnt;
245 	e2fs_lbn_t metalbn, realbn;
246 	struct ext2mount *ump;
247 	int i, numlevels, off;
248 	int64_t qblockcnt;
249 
250 	ump = VFSTOEXT2(vp->v_mount);
251 	if (nump)
252 		*nump = 0;
253 	numlevels = 0;
254 	realbn = bn;
255 	if ((long)bn < 0)
256 		bn = -(long)bn;
257 
258 	/* The first NDADDR blocks are direct blocks. */
259 	if (bn < NDADDR)
260 		return (0);
261 
262 	/*
263 	 * Determine the number of levels of indirection.  After this loop
264 	 * is done, blockcnt indicates the number of data blocks possible
265 	 * at the previous level of indirection, and NIADDR - i is the number
266 	 * of levels of indirection needed to locate the requested block.
267 	 */
268 	for (blockcnt = 1, i = NIADDR, bn -= NDADDR;; i--, bn -= blockcnt) {
269 		if (i == 0)
270 			return (EFBIG);
271 		/*
272 		 * Use int64_t's here to avoid overflow for triple indirect
273 		 * blocks when longs have 32 bits and the block size is more
274 		 * than 4K.
275 		 */
276 		qblockcnt = (int64_t)blockcnt * MNINDIR(ump);
277 		if (bn < qblockcnt)
278 			break;
279 		blockcnt = qblockcnt;
280 	}
281 
282 	/* Calculate the address of the first meta-block. */
283 	if (realbn >= 0)
284 		metalbn = -(realbn - bn + NIADDR - i);
285 	else
286 		metalbn = -(-realbn - bn + NIADDR - i);
287 
288 	/*
289 	 * At each iteration, off is the offset into the bap array which is
290 	 * an array of disk addresses at the current level of indirection.
291 	 * The logical block number and the offset in that block are stored
292 	 * into the argument array.
293 	 */
294 	ap->in_lbn = metalbn;
295 	ap->in_off = off = NIADDR - i;
296 	ap++;
297 	for (++numlevels; i <= NIADDR; i++) {
298 		/* If searching for a meta-data block, quit when found. */
299 		if (metalbn == realbn)
300 			break;
301 
302 		off = (bn / blockcnt) % MNINDIR(ump);
303 
304 		++numlevels;
305 		ap->in_lbn = metalbn;
306 		ap->in_off = off;
307 		++ap;
308 
309 		metalbn -= -1 + off * blockcnt;
310 		blockcnt /= MNINDIR(ump);
311 	}
312 	if (nump)
313 		*nump = numlevels;
314 	return (0);
315 }
316