xref: /freebsd/sys/fs/ext2fs/ext2_bmap.c (revision 8fc257994d0ce2396196d7a06d50d20c8015f4b7)
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_mount.h>
51 #include <fs/ext2fs/ext2_extern.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(ap)
60 	struct vop_bmap_args /* {
61 		struct vnode *a_vp;
62 		daddr_t a_bn;
63 		struct bufobj **a_bop;
64 		daddr_t *a_bnp;
65 		int *a_runp;
66 		int *a_runb;
67 	} */ *ap;
68 {
69 	int32_t blkno;
70 	int error;
71 
72 	/*
73 	 * Check for underlying vnode requests and ensure that logical
74 	 * to physical mapping is requested.
75 	 */
76 	if (ap->a_bop != NULL)
77 		*ap->a_bop = &VTOI(ap->a_vp)->i_devvp->v_bufobj;
78 	if (ap->a_bnp == NULL)
79 		return (0);
80 
81 	error = ext2_bmaparray(ap->a_vp, ap->a_bn, &blkno,
82 	    ap->a_runp, ap->a_runb);
83 	*ap->a_bnp = blkno;
84 	return (error);
85 }
86 
87 /*
88  * Indirect blocks are now on the vnode for the file.  They are given negative
89  * logical block numbers.  Indirect blocks are addressed by the negative
90  * address of the first data block to which they point.  Double indirect blocks
91  * are addressed by one less than the address of the first indirect block to
92  * which they point.  Triple indirect blocks are addressed by one less than
93  * the address of the first double indirect block to which they point.
94  *
95  * ext2_bmaparray does the bmap conversion, and if requested returns the
96  * array of logical blocks which must be traversed to get to a block.
97  * Each entry contains the offset into that block that gets you to the
98  * next block and the disk address of the block (if it is assigned).
99  */
100 
101 int
102 ext2_bmaparray(vp, bn, bnp, runp, runb)
103 	struct vnode *vp;
104 	int32_t bn;
105 	int32_t *bnp;
106 	int *runp;
107 	int *runb;
108 {
109 	struct inode *ip;
110 	struct buf *bp;
111 	struct ext2mount *ump;
112 	struct mount *mp;
113 	struct vnode *devvp;
114 	struct indir a[NIADDR+1], *ap;
115 	int32_t daddr;
116 	long metalbn;
117 	int error, num, maxrun = 0, bsize;
118 	int *nump;
119 
120 	ap = NULL;
121 	ip = VTOI(vp);
122 	mp = vp->v_mount;
123 	ump = VFSTOEXT2(mp);
124 	devvp = ump->um_devvp;
125 
126 	bsize = EXT2_BLOCK_SIZE(ump->um_e2fs);
127 
128 	if (runp) {
129 		maxrun = mp->mnt_iosize_max / bsize - 1;
130 		*runp = 0;
131 	}
132 
133 	if (runb) {
134 		*runb = 0;
135 	}
136 
137 
138 	ap = a;
139 	nump = &num;
140 	error = ext2_getlbns(vp, bn, ap, nump);
141 	if (error)
142 		return (error);
143 
144 	num = *nump;
145 	if (num == 0) {
146 		*bnp = blkptrtodb(ump, ip->i_db[bn]);
147 		if (*bnp == 0) {
148 			*bnp = -1;
149 		} else if (runp) {
150 			int32_t bnb = bn;
151 			for (++bn; bn < NDADDR && *runp < maxrun &&
152 			    is_sequential(ump, ip->i_db[bn - 1], ip->i_db[bn]);
153 			    ++bn, ++*runp);
154 			bn = bnb;
155 			if (runb && (bn > 0)) {
156 				for (--bn; (bn >= 0) && (*runb < maxrun) &&
157 					is_sequential(ump, ip->i_db[bn],
158 						ip->i_db[bn+1]);
159 						--bn, ++*runb);
160 			}
161 		}
162 		return (0);
163 	}
164 
165 
166 	/* Get disk address out of indirect block array */
167 	daddr = ip->i_ib[ap->in_off];
168 
169 	for (bp = NULL, ++ap; --num; ++ap) {
170 		/*
171 		 * Exit the loop if there is no disk address assigned yet and
172 		 * the indirect block isn't in the cache, or if we were
173 		 * looking for an indirect block and we've found it.
174 		 */
175 
176 		metalbn = ap->in_lbn;
177 		if ((daddr == 0 && !incore(&vp->v_bufobj, metalbn)) || metalbn == bn)
178 			break;
179 		/*
180 		 * If we get here, we've either got the block in the cache
181 		 * or we have a disk address for it, go fetch it.
182 		 */
183 		if (bp)
184 			bqrelse(bp);
185 
186 		ap->in_exists = 1;
187 		bp = getblk(vp, metalbn, bsize, 0, 0, 0);
188 		if ((bp->b_flags & B_CACHE) == 0) {
189 #ifdef DIAGNOSTIC
190 			if (!daddr)
191 				panic("ufs_bmaparray: indirect block not in cache");
192 #endif
193 			bp->b_blkno = blkptrtodb(ump, daddr);
194 			bp->b_iocmd = BIO_READ;
195 			bp->b_flags &= ~B_INVAL;
196 			bp->b_ioflags &= ~BIO_ERROR;
197 			vfs_busy_pages(bp, 0);
198 			bp->b_iooffset = dbtob(bp->b_blkno);
199 			bstrategy(bp);
200 			curthread->td_ru.ru_inblock++;
201 			error = bufwait(bp);
202 			if (error) {
203 				brelse(bp);
204 				return (error);
205 			}
206 		}
207 
208 		daddr = ((int32_t *)bp->b_data)[ap->in_off];
209 		if (num == 1 && daddr && runp) {
210 			for (bn = ap->in_off + 1;
211 			    bn < MNINDIR(ump) && *runp < maxrun &&
212 			    is_sequential(ump,
213 			    ((int32_t *)bp->b_data)[bn - 1],
214 			    ((int32_t *)bp->b_data)[bn]);
215 			    ++bn, ++*runp);
216 			bn = ap->in_off;
217 			if (runb && bn) {
218 				for(--bn; bn >= 0 && *runb < maxrun &&
219 			    		is_sequential(ump, ((int32_t *)bp->b_data)[bn],
220 					    ((int32_t *)bp->b_data)[bn+1]);
221 			    		--bn, ++*runb);
222 			}
223 		}
224 	}
225 	if (bp)
226 		bqrelse(bp);
227 
228 	/*
229 	 * Since this is FFS independent code, we are out of scope for the
230 	 * definitions of BLK_NOCOPY and BLK_SNAP, but we do know that they
231 	 * will fall in the range 1..um_seqinc, so we use that test and
232 	 * return a request for a zeroed out buffer if attempts are made
233 	 * to read a BLK_NOCOPY or BLK_SNAP block.
234 	 */
235 	if ((ip->i_flags & SF_SNAPSHOT) && daddr > 0 && daddr < ump->um_seqinc){
236 		*bnp = -1;
237 		return (0);
238 	}
239 	*bnp = blkptrtodb(ump, daddr);
240 	if (*bnp == 0) {
241 		*bnp = -1;
242 	}
243 	return (0);
244 }
245 
246 /*
247  * Create an array of logical block number/offset pairs which represent the
248  * path of indirect blocks required to access a data block.  The first "pair"
249  * contains the logical block number of the appropriate single, double or
250  * triple indirect block and the offset into the inode indirect block array.
251  * Note, the logical block number of the inode single/double/triple indirect
252  * block appears twice in the array, once with the offset into the i_ib and
253  * once with the offset into the page itself.
254  */
255 int
256 ext2_getlbns(vp, bn, ap, nump)
257 	struct vnode *vp;
258 	int32_t bn;
259 	struct indir *ap;
260 	int *nump;
261 {
262 	long blockcnt, metalbn, realbn;
263 	struct ext2mount *ump;
264 	int i, numlevels, off;
265 	int64_t qblockcnt;
266 
267 	ump = VFSTOEXT2(vp->v_mount);
268 	if (nump)
269 		*nump = 0;
270 	numlevels = 0;
271 	realbn = bn;
272 	if ((long)bn < 0)
273 		bn = -(long)bn;
274 
275 	/* The first NDADDR blocks are direct blocks. */
276 	if (bn < NDADDR)
277 		return (0);
278 
279 	/*
280 	 * Determine the number of levels of indirection.  After this loop
281 	 * is done, blockcnt indicates the number of data blocks possible
282 	 * at the previous level of indirection, and NIADDR - i is the number
283 	 * of levels of indirection needed to locate the requested block.
284 	 */
285 	for (blockcnt = 1, i = NIADDR, bn -= NDADDR;; i--, bn -= blockcnt) {
286 		if (i == 0)
287 			return (EFBIG);
288 		/*
289 		 * Use int64_t's here to avoid overflow for triple indirect
290 		 * blocks when longs have 32 bits and the block size is more
291 		 * than 4K.
292 		 */
293 		qblockcnt = (int64_t)blockcnt * MNINDIR(ump);
294 		if (bn < qblockcnt)
295 			break;
296 		blockcnt = qblockcnt;
297 	}
298 
299 	/* Calculate the address of the first meta-block. */
300 	if (realbn >= 0)
301 		metalbn = -(realbn - bn + NIADDR - i);
302 	else
303 		metalbn = -(-realbn - bn + NIADDR - i);
304 
305 	/*
306 	 * At each iteration, off is the offset into the bap array which is
307 	 * an array of disk addresses at the current level of indirection.
308 	 * The logical block number and the offset in that block are stored
309 	 * into the argument array.
310 	 */
311 	ap->in_lbn = metalbn;
312 	ap->in_off = off = NIADDR - i;
313 	ap->in_exists = 0;
314 	ap++;
315 	for (++numlevels; i <= NIADDR; i++) {
316 		/* If searching for a meta-data block, quit when found. */
317 		if (metalbn == realbn)
318 			break;
319 
320 		off = (bn / blockcnt) % MNINDIR(ump);
321 
322 		++numlevels;
323 		ap->in_lbn = metalbn;
324 		ap->in_off = off;
325 		ap->in_exists = 0;
326 		++ap;
327 
328 		metalbn -= -1 + off * blockcnt;
329 		blockcnt /= MNINDIR(ump);
330 	}
331 	if (nump)
332 		*nump = numlevels;
333 	return (0);
334 }
335