xref: /freebsd-src/sys/fs/ext2fs/ext2_bmap.c (revision 1c05a6ea6b849ff95e539c31adea887c644a6a01)
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  * 3. 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/racct.h>
46 #include <sys/resourcevar.h>
47 #include <sys/stat.h>
48 
49 #include <fs/ext2fs/inode.h>
50 #include <fs/ext2fs/fs.h>
51 #include <fs/ext2fs/ext2fs.h>
52 #include <fs/ext2fs/ext2_dinode.h>
53 #include <fs/ext2fs/ext2_extern.h>
54 #include <fs/ext2fs/ext2_mount.h>
55 
56 /*
57  * Bmap converts the logical block number of a file to its physical block
58  * number on the disk. The conversion is done by using the logical block
59  * number to index into the array of block pointers described by the dinode.
60  */
61 int
62 ext2_bmap(struct vop_bmap_args *ap)
63 {
64 	daddr_t blkno;
65 	int error;
66 
67 	/*
68 	 * Check for underlying vnode requests and ensure that logical
69 	 * to physical mapping is requested.
70 	 */
71 	if (ap->a_bop != NULL)
72 		*ap->a_bop = &VTOI(ap->a_vp)->i_devvp->v_bufobj;
73 	if (ap->a_bnp == NULL)
74 		return (0);
75 
76 	if (VTOI(ap->a_vp)->i_flag & IN_E4EXTENTS)
77 		error = ext4_bmapext(ap->a_vp, ap->a_bn, &blkno,
78 		    ap->a_runp, ap->a_runb);
79 	else
80 		error = ext2_bmaparray(ap->a_vp, ap->a_bn, &blkno,
81 		    ap->a_runp, ap->a_runb);
82 	*ap->a_bnp = blkno;
83 	return (error);
84 }
85 
86 /*
87  * Convert the logical block number of a file to its physical block number
88  * on the disk within ext4 extents.
89  */
90 int
91 ext4_bmapext(struct vnode *vp, int32_t bn, int64_t *bnp, int *runp, int *runb)
92 {
93 	struct inode *ip;
94 	struct m_ext2fs *fs;
95 	struct ext4_extent_header *ehp;
96 	struct ext4_extent *ep;
97 	struct ext4_extent_path *path = NULL;
98 	daddr_t lbn;
99 	int error, depth;
100 
101 	ip = VTOI(vp);
102 	fs = ip->i_e2fs;
103 	lbn = bn;
104 	ehp = (struct ext4_extent_header *)ip->i_data;
105 	depth = ehp->eh_depth;
106 
107 	*bnp = -1;
108 	if (runp != NULL)
109 		*runp = 0;
110 	if (runb != NULL)
111 		*runb = 0;
112 
113 	error = ext4_ext_find_extent(ip, lbn, &path);
114 	if (error)
115 		return (error);
116 
117 	ep = path[depth].ep_ext;
118 	if(ep) {
119 		if (lbn < ep->e_blk) {
120 			if (runp != NULL)
121 				*runp = ep->e_blk - lbn - 1;
122 		} else if (ep->e_blk <= lbn && lbn < ep->e_blk + ep->e_len) {
123 			*bnp = fsbtodb(fs, lbn - ep->e_blk +
124 			    (ep->e_start_lo | (daddr_t)ep->e_start_hi << 32));
125 			if (runp != NULL)
126 				*runp = ep->e_len - (lbn - ep->e_blk) - 1;
127 			if (runb != NULL)
128 				*runb = lbn - ep->e_blk;
129 		} else {
130 			if (runb != NULL)
131 				*runb = ep->e_blk + lbn - ep->e_len;
132 		}
133 	}
134 
135 	ext4_ext_path_free(path);
136 
137 	return (error);
138 }
139 
140 /*
141  * Indirect blocks are now on the vnode for the file.  They are given negative
142  * logical block numbers.  Indirect blocks are addressed by the negative
143  * address of the first data block to which they point.  Double indirect blocks
144  * are addressed by one less than the address of the first indirect block to
145  * which they point.  Triple indirect blocks are addressed by one less than
146  * the address of the first double indirect block to which they point.
147  *
148  * ext2_bmaparray does the bmap conversion, and if requested returns the
149  * array of logical blocks which must be traversed to get to a block.
150  * Each entry contains the offset into that block that gets you to the
151  * next block and the disk address of the block (if it is assigned).
152  */
153 
154 int
155 ext2_bmaparray(struct vnode *vp, daddr_t bn, daddr_t *bnp, int *runp, int *runb)
156 {
157 	struct inode *ip;
158 	struct buf *bp;
159 	struct ext2mount *ump;
160 	struct mount *mp;
161 	struct indir a[EXT2_NIADDR + 1], *ap;
162 	daddr_t daddr;
163 	e2fs_lbn_t metalbn;
164 	int error, num, maxrun = 0, bsize;
165 	int *nump;
166 
167 	ap = NULL;
168 	ip = VTOI(vp);
169 	mp = vp->v_mount;
170 	ump = VFSTOEXT2(mp);
171 
172 	bsize = EXT2_BLOCK_SIZE(ump->um_e2fs);
173 
174 	if (runp) {
175 		maxrun = mp->mnt_iosize_max / bsize - 1;
176 		*runp = 0;
177 	}
178 	if (runb)
179 		*runb = 0;
180 
181 
182 	ap = a;
183 	nump = &num;
184 	error = ext2_getlbns(vp, bn, ap, nump);
185 	if (error)
186 		return (error);
187 
188 	num = *nump;
189 	if (num == 0) {
190 		*bnp = blkptrtodb(ump, ip->i_db[bn]);
191 		if (*bnp == 0) {
192 			*bnp = -1;
193 		} else if (runp) {
194 			daddr_t bnb = bn;
195 
196 			for (++bn; bn < EXT2_NDADDR && *runp < maxrun &&
197 			    is_sequential(ump, ip->i_db[bn - 1], ip->i_db[bn]);
198 			    ++bn, ++*runp);
199 			bn = bnb;
200 			if (runb && (bn > 0)) {
201 				for (--bn; (bn >= 0) && (*runb < maxrun) &&
202 					is_sequential(ump, ip->i_db[bn],
203 						ip->i_db[bn + 1]);
204 						--bn, ++*runb);
205 			}
206 		}
207 		return (0);
208 	}
209 
210 	/* Get disk address out of indirect block array */
211 	daddr = ip->i_ib[ap->in_off];
212 
213 	for (bp = NULL, ++ap; --num; ++ap) {
214 		/*
215 		 * Exit the loop if there is no disk address assigned yet and
216 		 * the indirect block isn't in the cache, or if we were
217 		 * looking for an indirect block and we've found it.
218 		 */
219 
220 		metalbn = ap->in_lbn;
221 		if ((daddr == 0 && !incore(&vp->v_bufobj, metalbn)) || metalbn == bn)
222 			break;
223 		/*
224 		 * If we get here, we've either got the block in the cache
225 		 * or we have a disk address for it, go fetch it.
226 		 */
227 		if (bp)
228 			bqrelse(bp);
229 
230 		bp = getblk(vp, metalbn, bsize, 0, 0, 0);
231 		if ((bp->b_flags & B_CACHE) == 0) {
232 #ifdef INVARIANTS
233 			if (!daddr)
234 				panic("ext2_bmaparray: indirect block not in cache");
235 #endif
236 			bp->b_blkno = blkptrtodb(ump, daddr);
237 			bp->b_iocmd = BIO_READ;
238 			bp->b_flags &= ~B_INVAL;
239 			bp->b_ioflags &= ~BIO_ERROR;
240 			vfs_busy_pages(bp, 0);
241 			bp->b_iooffset = dbtob(bp->b_blkno);
242 			bstrategy(bp);
243 #ifdef RACCT
244 			if (racct_enable) {
245 				PROC_LOCK(curproc);
246 				racct_add_buf(curproc, bp, 0);
247 				PROC_UNLOCK(curproc);
248 			}
249 #endif
250 			curthread->td_ru.ru_inblock++;
251 			error = bufwait(bp);
252 			if (error) {
253 				brelse(bp);
254 				return (error);
255 			}
256 		}
257 
258 		daddr = ((e2fs_daddr_t *)bp->b_data)[ap->in_off];
259 		if (num == 1 && daddr && runp) {
260 			for (bn = ap->in_off + 1;
261 			    bn < MNINDIR(ump) && *runp < maxrun &&
262 			    is_sequential(ump,
263 			    ((e2fs_daddr_t *)bp->b_data)[bn - 1],
264 			    ((e2fs_daddr_t *)bp->b_data)[bn]);
265 			    ++bn, ++*runp);
266 			bn = ap->in_off;
267 			if (runb && bn) {
268 				for (--bn; bn >= 0 && *runb < maxrun &&
269 					is_sequential(ump,
270 					((e2fs_daddr_t *)bp->b_data)[bn],
271 					((e2fs_daddr_t *)bp->b_data)[bn + 1]);
272 					--bn, ++*runb);
273 			}
274 		}
275 	}
276 	if (bp)
277 		bqrelse(bp);
278 
279 	/*
280 	 * Since this is FFS independent code, we are out of scope for the
281 	 * definitions of BLK_NOCOPY and BLK_SNAP, but we do know that they
282 	 * will fall in the range 1..um_seqinc, so we use that test and
283 	 * return a request for a zeroed out buffer if attempts are made
284 	 * to read a BLK_NOCOPY or BLK_SNAP block.
285 	 */
286 	if ((ip->i_flags & SF_SNAPSHOT) && daddr > 0 && daddr < ump->um_seqinc) {
287 		*bnp = -1;
288 		return (0);
289 	}
290 	*bnp = blkptrtodb(ump, daddr);
291 	if (*bnp == 0) {
292 		*bnp = -1;
293 	}
294 	return (0);
295 }
296 
297 /*
298  * Create an array of logical block number/offset pairs which represent the
299  * path of indirect blocks required to access a data block.  The first "pair"
300  * contains the logical block number of the appropriate single, double or
301  * triple indirect block and the offset into the inode indirect block array.
302  * Note, the logical block number of the inode single/double/triple indirect
303  * block appears twice in the array, once with the offset into the i_ib and
304  * once with the offset into the page itself.
305  */
306 int
307 ext2_getlbns(struct vnode *vp, daddr_t bn, struct indir *ap, int *nump)
308 {
309 	long blockcnt;
310 	e2fs_lbn_t metalbn, realbn;
311 	struct ext2mount *ump;
312 	int i, numlevels, off;
313 	int64_t qblockcnt;
314 
315 	ump = VFSTOEXT2(vp->v_mount);
316 	if (nump)
317 		*nump = 0;
318 	numlevels = 0;
319 	realbn = bn;
320 	if ((long)bn < 0)
321 		bn = -(long)bn;
322 
323 	/* The first EXT2_NDADDR blocks are direct blocks. */
324 	if (bn < EXT2_NDADDR)
325 		return (0);
326 
327 	/*
328 	 * Determine the number of levels of indirection.  After this loop
329 	 * is done, blockcnt indicates the number of data blocks possible
330 	 * at the previous level of indirection, and EXT2_NIADDR - i is the
331 	 * number of levels of indirection needed to locate the requested block.
332 	 */
333 	for (blockcnt = 1, i = EXT2_NIADDR, bn -= EXT2_NDADDR; ;
334 	    i--, bn -= blockcnt) {
335 		if (i == 0)
336 			return (EFBIG);
337 		/*
338 		 * Use int64_t's here to avoid overflow for triple indirect
339 		 * blocks when longs have 32 bits and the block size is more
340 		 * than 4K.
341 		 */
342 		qblockcnt = (int64_t)blockcnt * MNINDIR(ump);
343 		if (bn < qblockcnt)
344 			break;
345 		blockcnt = qblockcnt;
346 	}
347 
348 	/* Calculate the address of the first meta-block. */
349 	if (realbn >= 0)
350 		metalbn = -(realbn - bn + EXT2_NIADDR - i);
351 	else
352 		metalbn = -(-realbn - bn + EXT2_NIADDR - i);
353 
354 	/*
355 	 * At each iteration, off is the offset into the bap array which is
356 	 * an array of disk addresses at the current level of indirection.
357 	 * The logical block number and the offset in that block are stored
358 	 * into the argument array.
359 	 */
360 	ap->in_lbn = metalbn;
361 	ap->in_off = off = EXT2_NIADDR - i;
362 	ap++;
363 	for (++numlevels; i <= EXT2_NIADDR; i++) {
364 		/* If searching for a meta-data block, quit when found. */
365 		if (metalbn == realbn)
366 			break;
367 
368 		off = (bn / blockcnt) % MNINDIR(ump);
369 
370 		++numlevels;
371 		ap->in_lbn = metalbn;
372 		ap->in_off = off;
373 		++ap;
374 
375 		metalbn -= -1 + off * blockcnt;
376 		blockcnt /= MNINDIR(ump);
377 	}
378 	if (nump)
379 		*nump = numlevels;
380 	return (0);
381 }
382