xref: /freebsd/sys/fs/ext2fs/ext2_bmap.c (revision 7447ca0eb235974642312b9555caec00b57d8fc1)
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/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 static int ext4_bmapext(struct vnode *, int32_t, int64_t *, int *, int *);
57 
58 /*
59  * Bmap converts the logical block number of a file to its physical block
60  * number on the disk. The conversion is done by using the logical block
61  * number to index into the array of block pointers described by the dinode.
62  */
63 int
64 ext2_bmap(struct vop_bmap_args *ap)
65 {
66 	daddr_t blkno;
67 	int error;
68 
69 	/*
70 	 * Check for underlying vnode requests and ensure that logical
71 	 * to physical mapping is requested.
72 	 */
73 	if (ap->a_bop != NULL)
74 		*ap->a_bop = &VTOI(ap->a_vp)->i_devvp->v_bufobj;
75 	if (ap->a_bnp == NULL)
76 		return (0);
77 
78 	if (VTOI(ap->a_vp)->i_flag & IN_E4EXTENTS)
79 		error = ext4_bmapext(ap->a_vp, ap->a_bn, &blkno,
80 		    ap->a_runp, ap->a_runb);
81 	else
82 		error = ext2_bmaparray(ap->a_vp, ap->a_bn, &blkno,
83 		    ap->a_runp, ap->a_runb);
84 	*ap->a_bnp = blkno;
85 	return (error);
86 }
87 
88 /*
89  * Convert the logical block number of a file to its physical block number
90  * on the disk within ext4 extents.
91  */
92 static int
93 ext4_bmapext(struct vnode *vp, int32_t bn, int64_t *bnp, int *runp, int *runb)
94 {
95 	struct inode *ip;
96 	struct m_ext2fs *fs;
97 	struct ext4_extent *ep;
98 	struct ext4_extent_path path = { .ep_bp = NULL };
99 	daddr_t lbn;
100 	int error;
101 
102 	ip = VTOI(vp);
103 	fs = ip->i_e2fs;
104 	lbn = bn;
105 
106 	if (runp != NULL)
107 		*runp = 0;
108 	if (runb != NULL)
109 		*runb = 0;
110 	error = 0;
111 
112 	ext4_ext_find_extent(fs, ip, lbn, &path);
113 	if (path.ep_is_sparse) {
114 		*bnp = -1;
115 		if (runp != NULL)
116 			*runp = path.ep_sparse_ext.e_len -
117 			    (lbn - path.ep_sparse_ext.e_blk) - 1;
118 		if (runb != NULL)
119 			*runb = lbn - path.ep_sparse_ext.e_blk;
120 	} else {
121 		if (path.ep_ext == NULL) {
122 			error = EIO;
123 			goto out;
124 		}
125 		ep = path.ep_ext;
126 		*bnp = fsbtodb(fs, lbn - ep->e_blk +
127 		    (ep->e_start_lo | (daddr_t)ep->e_start_hi << 32));
128 
129 		if (*bnp == 0)
130 			*bnp = -1;
131 
132 		if (runp != NULL)
133 			*runp = ep->e_len - (lbn - ep->e_blk) - 1;
134 		if (runb != NULL)
135 			*runb = lbn - ep->e_blk;
136 	}
137 
138 out:
139 	if (path.ep_bp != NULL)
140 		brelse(path.ep_bp);
141 
142 	return (error);
143 }
144 
145 /*
146  * Indirect blocks are now on the vnode for the file.  They are given negative
147  * logical block numbers.  Indirect blocks are addressed by the negative
148  * address of the first data block to which they point.  Double indirect blocks
149  * are addressed by one less than the address of the first indirect block to
150  * which they point.  Triple indirect blocks are addressed by one less than
151  * the address of the first double indirect block to which they point.
152  *
153  * ext2_bmaparray does the bmap conversion, and if requested returns the
154  * array of logical blocks which must be traversed to get to a block.
155  * Each entry contains the offset into that block that gets you to the
156  * next block and the disk address of the block (if it is assigned).
157  */
158 
159 int
160 ext2_bmaparray(struct vnode *vp, daddr_t bn, daddr_t *bnp, int *runp, int *runb)
161 {
162 	struct inode *ip;
163 	struct buf *bp;
164 	struct ext2mount *ump;
165 	struct mount *mp;
166 	struct indir a[NIADDR+1], *ap;
167 	daddr_t daddr;
168 	e2fs_lbn_t metalbn;
169 	int error, num, maxrun = 0, bsize;
170 	int *nump;
171 
172 	ap = NULL;
173 	ip = VTOI(vp);
174 	mp = vp->v_mount;
175 	ump = VFSTOEXT2(mp);
176 
177 	bsize = EXT2_BLOCK_SIZE(ump->um_e2fs);
178 
179 	if (runp) {
180 		maxrun = mp->mnt_iosize_max / bsize - 1;
181 		*runp = 0;
182 	}
183 
184 	if (runb) {
185 		*runb = 0;
186 	}
187 
188 
189 	ap = a;
190 	nump = &num;
191 	error = ext2_getlbns(vp, bn, ap, nump);
192 	if (error)
193 		return (error);
194 
195 	num = *nump;
196 	if (num == 0) {
197 		*bnp = blkptrtodb(ump, ip->i_db[bn]);
198 		if (*bnp == 0) {
199 			*bnp = -1;
200 		} else if (runp) {
201 			daddr_t bnb = bn;
202 			for (++bn; bn < NDADDR && *runp < maxrun &&
203 			    is_sequential(ump, ip->i_db[bn - 1], ip->i_db[bn]);
204 			    ++bn, ++*runp);
205 			bn = bnb;
206 			if (runb && (bn > 0)) {
207 				for (--bn; (bn >= 0) && (*runb < maxrun) &&
208 					is_sequential(ump, ip->i_db[bn],
209 						ip->i_db[bn + 1]);
210 						--bn, ++*runb);
211 			}
212 		}
213 		return (0);
214 	}
215 
216 
217 	/* Get disk address out of indirect block array */
218 	daddr = ip->i_ib[ap->in_off];
219 
220 	for (bp = NULL, ++ap; --num; ++ap) {
221 		/*
222 		 * Exit the loop if there is no disk address assigned yet and
223 		 * the indirect block isn't in the cache, or if we were
224 		 * looking for an indirect block and we've found it.
225 		 */
226 
227 		metalbn = ap->in_lbn;
228 		if ((daddr == 0 && !incore(&vp->v_bufobj, metalbn)) || metalbn == bn)
229 			break;
230 		/*
231 		 * If we get here, we've either got the block in the cache
232 		 * or we have a disk address for it, go fetch it.
233 		 */
234 		if (bp)
235 			bqrelse(bp);
236 
237 		bp = getblk(vp, metalbn, bsize, 0, 0, 0);
238 		if ((bp->b_flags & B_CACHE) == 0) {
239 #ifdef INVARIANTS
240 			if (!daddr)
241 				panic("ext2_bmaparray: indirect block not in cache");
242 #endif
243 			bp->b_blkno = blkptrtodb(ump, daddr);
244 			bp->b_iocmd = BIO_READ;
245 			bp->b_flags &= ~B_INVAL;
246 			bp->b_ioflags &= ~BIO_ERROR;
247 			vfs_busy_pages(bp, 0);
248 			bp->b_iooffset = dbtob(bp->b_blkno);
249 			bstrategy(bp);
250 #ifdef RACCT
251 			if (racct_enable) {
252 				PROC_LOCK(curproc);
253 				racct_add_buf(curproc, bp, 0);
254 				PROC_UNLOCK(curproc);
255 			}
256 #endif
257 			curthread->td_ru.ru_inblock++;
258 			error = bufwait(bp);
259 			if (error) {
260 				brelse(bp);
261 				return (error);
262 			}
263 		}
264 
265 		daddr = ((e2fs_daddr_t *)bp->b_data)[ap->in_off];
266 		if (num == 1 && daddr && runp) {
267 			for (bn = ap->in_off + 1;
268 			    bn < MNINDIR(ump) && *runp < maxrun &&
269 			    is_sequential(ump,
270 			    ((e2fs_daddr_t *)bp->b_data)[bn - 1],
271 			    ((e2fs_daddr_t *)bp->b_data)[bn]);
272 			    ++bn, ++*runp);
273 			bn = ap->in_off;
274 			if (runb && bn) {
275 				for (--bn; bn >= 0 && *runb < maxrun &&
276 					is_sequential(ump,
277 					((e2fs_daddr_t *)bp->b_data)[bn],
278 					((e2fs_daddr_t *)bp->b_data)[bn + 1]);
279 					--bn, ++*runb);
280 			}
281 		}
282 	}
283 	if (bp)
284 		bqrelse(bp);
285 
286 	/*
287 	 * Since this is FFS independent code, we are out of scope for the
288 	 * definitions of BLK_NOCOPY and BLK_SNAP, but we do know that they
289 	 * will fall in the range 1..um_seqinc, so we use that test and
290 	 * return a request for a zeroed out buffer if attempts are made
291 	 * to read a BLK_NOCOPY or BLK_SNAP block.
292 	 */
293 	if ((ip->i_flags & SF_SNAPSHOT) && daddr > 0 && daddr < ump->um_seqinc){
294 		*bnp = -1;
295 		return (0);
296 	}
297 	*bnp = blkptrtodb(ump, daddr);
298 	if (*bnp == 0) {
299 		*bnp = -1;
300 	}
301 	return (0);
302 }
303 
304 /*
305  * Create an array of logical block number/offset pairs which represent the
306  * path of indirect blocks required to access a data block.  The first "pair"
307  * contains the logical block number of the appropriate single, double or
308  * triple indirect block and the offset into the inode indirect block array.
309  * Note, the logical block number of the inode single/double/triple indirect
310  * block appears twice in the array, once with the offset into the i_ib and
311  * once with the offset into the page itself.
312  */
313 int
314 ext2_getlbns(struct vnode *vp, daddr_t bn, struct indir *ap, int *nump)
315 {
316 	long blockcnt;
317 	e2fs_lbn_t metalbn, realbn;
318 	struct ext2mount *ump;
319 	int i, numlevels, off;
320 	int64_t qblockcnt;
321 
322 	ump = VFSTOEXT2(vp->v_mount);
323 	if (nump)
324 		*nump = 0;
325 	numlevels = 0;
326 	realbn = bn;
327 	if ((long)bn < 0)
328 		bn = -(long)bn;
329 
330 	/* The first NDADDR blocks are direct blocks. */
331 	if (bn < NDADDR)
332 		return (0);
333 
334 	/*
335 	 * Determine the number of levels of indirection.  After this loop
336 	 * is done, blockcnt indicates the number of data blocks possible
337 	 * at the previous level of indirection, and NIADDR - i is the number
338 	 * of levels of indirection needed to locate the requested block.
339 	 */
340 	for (blockcnt = 1, i = NIADDR, bn -= NDADDR;; i--, bn -= blockcnt) {
341 		if (i == 0)
342 			return (EFBIG);
343 		/*
344 		 * Use int64_t's here to avoid overflow for triple indirect
345 		 * blocks when longs have 32 bits and the block size is more
346 		 * than 4K.
347 		 */
348 		qblockcnt = (int64_t)blockcnt * MNINDIR(ump);
349 		if (bn < qblockcnt)
350 			break;
351 		blockcnt = qblockcnt;
352 	}
353 
354 	/* Calculate the address of the first meta-block. */
355 	if (realbn >= 0)
356 		metalbn = -(realbn - bn + NIADDR - i);
357 	else
358 		metalbn = -(-realbn - bn + NIADDR - i);
359 
360 	/*
361 	 * At each iteration, off is the offset into the bap array which is
362 	 * an array of disk addresses at the current level of indirection.
363 	 * The logical block number and the offset in that block are stored
364 	 * into the argument array.
365 	 */
366 	ap->in_lbn = metalbn;
367 	ap->in_off = off = NIADDR - i;
368 	ap++;
369 	for (++numlevels; i <= NIADDR; i++) {
370 		/* If searching for a meta-data block, quit when found. */
371 		if (metalbn == realbn)
372 			break;
373 
374 		off = (bn / blockcnt) % MNINDIR(ump);
375 
376 		++numlevels;
377 		ap->in_lbn = metalbn;
378 		ap->in_off = off;
379 		++ap;
380 
381 		metalbn -= -1 + off * blockcnt;
382 		blockcnt /= MNINDIR(ump);
383 	}
384 	if (nump)
385 		*nump = numlevels;
386 	return (0);
387 }
388