xref: /freebsd/sys/fs/ext2fs/ext2_bmap.c (revision 8d20be1e22095c27faf8fe8b2f0d089739cc742e)
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/fs.h>
50 #include <fs/ext2fs/ext2fs.h>
51 #include <fs/ext2fs/ext2_dinode.h>
52 #include <fs/ext2fs/ext2_extern.h>
53 #include <fs/ext2fs/ext2_mount.h>
54 
55 static int ext4_bmapext(struct vnode *, int32_t, int64_t *, int *, int *);
56 
57 /*
58  * Bmap converts the logical block number of a file to its physical block
59  * number on the disk. The conversion is done by using the logical block
60  * number to index into the array of block pointers described by the dinode.
61  */
62 int
63 ext2_bmap(struct vop_bmap_args *ap)
64 {
65 	daddr_t blkno;
66 	int error;
67 
68 	/*
69 	 * Check for underlying vnode requests and ensure that logical
70 	 * to physical mapping is requested.
71 	 */
72 	if (ap->a_bop != NULL)
73 		*ap->a_bop = &VTOI(ap->a_vp)->i_devvp->v_bufobj;
74 	if (ap->a_bnp == NULL)
75 		return (0);
76 
77 	if (VTOI(ap->a_vp)->i_flags & EXT4_EXTENTS)
78 		error = ext4_bmapext(ap->a_vp, ap->a_bn, &blkno,
79 		    ap->a_runp, ap->a_runb);
80 	else
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  * This function converts the logical block number of a file to
89  * its physical block number on the disk within ext4 extents.
90  */
91 static int
92 ext4_bmapext(struct vnode *vp, int32_t bn, int64_t *bnp, int *runp, int *runb)
93 {
94 	struct inode *ip;
95 	struct m_ext2fs *fs;
96 	struct ext4_extent *ep;
97 	struct ext4_extent_path path;
98 	daddr_t lbn;
99 
100 	ip = VTOI(vp);
101 	fs = ip->i_e2fs;
102 	lbn = bn;
103 
104 	/*
105 	 * TODO: need to implement read ahead to improve the performance.
106 	 */
107 	if (runp != NULL)
108 		*runp = 0;
109 
110 	if (runb != NULL)
111 		*runb = 0;
112 
113 	ext4_ext_find_extent(fs, ip, lbn, &path);
114 	ep = path.ep_ext;
115 	if (ep == NULL)
116 		return (EIO);
117 
118 	*bnp = fsbtodb(fs, lbn - ep->e_blk +
119 	    (ep->e_start_lo | (daddr_t)ep->e_start_hi << 32));
120 
121 	if (*bnp == 0)
122 		*bnp = -1;
123 
124 	return (0);
125 }
126 
127 /*
128  * Indirect blocks are now on the vnode for the file.  They are given negative
129  * logical block numbers.  Indirect blocks are addressed by the negative
130  * address of the first data block to which they point.  Double indirect blocks
131  * are addressed by one less than the address of the first indirect block to
132  * which they point.  Triple indirect blocks are addressed by one less than
133  * the address of the first double indirect block to which they point.
134  *
135  * ext2_bmaparray does the bmap conversion, and if requested returns the
136  * array of logical blocks which must be traversed to get to a block.
137  * Each entry contains the offset into that block that gets you to the
138  * next block and the disk address of the block (if it is assigned).
139  */
140 
141 int
142 ext2_bmaparray(struct vnode *vp, daddr_t bn, daddr_t *bnp, int *runp, int *runb)
143 {
144 	struct inode *ip;
145 	struct buf *bp;
146 	struct ext2mount *ump;
147 	struct mount *mp;
148 	struct vnode *devvp;
149 	struct indir a[NIADDR+1], *ap;
150 	daddr_t daddr;
151 	e2fs_lbn_t metalbn;
152 	int error, num, maxrun = 0, bsize;
153 	int *nump;
154 
155 	ap = NULL;
156 	ip = VTOI(vp);
157 	mp = vp->v_mount;
158 	ump = VFSTOEXT2(mp);
159 	devvp = ump->um_devvp;
160 
161 	bsize = EXT2_BLOCK_SIZE(ump->um_e2fs);
162 
163 	if (runp) {
164 		maxrun = mp->mnt_iosize_max / bsize - 1;
165 		*runp = 0;
166 	}
167 
168 	if (runb) {
169 		*runb = 0;
170 	}
171 
172 
173 	ap = a;
174 	nump = &num;
175 	error = ext2_getlbns(vp, bn, ap, nump);
176 	if (error)
177 		return (error);
178 
179 	num = *nump;
180 	if (num == 0) {
181 		*bnp = blkptrtodb(ump, ip->i_db[bn]);
182 		if (*bnp == 0) {
183 			*bnp = -1;
184 		} else if (runp) {
185 			daddr_t bnb = bn;
186 			for (++bn; bn < NDADDR && *runp < maxrun &&
187 			    is_sequential(ump, ip->i_db[bn - 1], ip->i_db[bn]);
188 			    ++bn, ++*runp);
189 			bn = bnb;
190 			if (runb && (bn > 0)) {
191 				for (--bn; (bn >= 0) && (*runb < maxrun) &&
192 					is_sequential(ump, ip->i_db[bn],
193 						ip->i_db[bn + 1]);
194 						--bn, ++*runb);
195 			}
196 		}
197 		return (0);
198 	}
199 
200 
201 	/* Get disk address out of indirect block array */
202 	daddr = ip->i_ib[ap->in_off];
203 
204 	for (bp = NULL, ++ap; --num; ++ap) {
205 		/*
206 		 * Exit the loop if there is no disk address assigned yet and
207 		 * the indirect block isn't in the cache, or if we were
208 		 * looking for an indirect block and we've found it.
209 		 */
210 
211 		metalbn = ap->in_lbn;
212 		if ((daddr == 0 && !incore(&vp->v_bufobj, metalbn)) || metalbn == bn)
213 			break;
214 		/*
215 		 * If we get here, we've either got the block in the cache
216 		 * or we have a disk address for it, go fetch it.
217 		 */
218 		if (bp)
219 			bqrelse(bp);
220 
221 		bp = getblk(vp, metalbn, bsize, 0, 0, 0);
222 		if ((bp->b_flags & B_CACHE) == 0) {
223 #ifdef INVARIANTS
224 			if (!daddr)
225 				panic("ext2_bmaparray: indirect block not in cache");
226 #endif
227 			bp->b_blkno = blkptrtodb(ump, daddr);
228 			bp->b_iocmd = BIO_READ;
229 			bp->b_flags &= ~B_INVAL;
230 			bp->b_ioflags &= ~BIO_ERROR;
231 			vfs_busy_pages(bp, 0);
232 			bp->b_iooffset = dbtob(bp->b_blkno);
233 			bstrategy(bp);
234 			curthread->td_ru.ru_inblock++;
235 			error = bufwait(bp);
236 			if (error) {
237 				brelse(bp);
238 				return (error);
239 			}
240 		}
241 
242 		daddr = ((e2fs_daddr_t *)bp->b_data)[ap->in_off];
243 		if (num == 1 && daddr && runp) {
244 			for (bn = ap->in_off + 1;
245 			    bn < MNINDIR(ump) && *runp < maxrun &&
246 			    is_sequential(ump,
247 			    ((e2fs_daddr_t *)bp->b_data)[bn - 1],
248 			    ((e2fs_daddr_t *)bp->b_data)[bn]);
249 			    ++bn, ++*runp);
250 			bn = ap->in_off;
251 			if (runb && bn) {
252 				for (--bn; bn >= 0 && *runb < maxrun &&
253 			    		is_sequential(ump,
254 					((e2fs_daddr_t *)bp->b_data)[bn],
255 					((e2fs_daddr_t *)bp->b_data)[bn + 1]);
256 			    		--bn, ++*runb);
257 			}
258 		}
259 	}
260 	if (bp)
261 		bqrelse(bp);
262 
263 	/*
264 	 * Since this is FFS independent code, we are out of scope for the
265 	 * definitions of BLK_NOCOPY and BLK_SNAP, but we do know that they
266 	 * will fall in the range 1..um_seqinc, so we use that test and
267 	 * return a request for a zeroed out buffer if attempts are made
268 	 * to read a BLK_NOCOPY or BLK_SNAP block.
269 	 */
270 	if ((ip->i_flags & SF_SNAPSHOT) && daddr > 0 && daddr < ump->um_seqinc){
271 		*bnp = -1;
272 		return (0);
273 	}
274 	*bnp = blkptrtodb(ump, daddr);
275 	if (*bnp == 0) {
276 		*bnp = -1;
277 	}
278 	return (0);
279 }
280 
281 /*
282  * Create an array of logical block number/offset pairs which represent the
283  * path of indirect blocks required to access a data block.  The first "pair"
284  * contains the logical block number of the appropriate single, double or
285  * triple indirect block and the offset into the inode indirect block array.
286  * Note, the logical block number of the inode single/double/triple indirect
287  * block appears twice in the array, once with the offset into the i_ib and
288  * once with the offset into the page itself.
289  */
290 int
291 ext2_getlbns(struct vnode *vp, daddr_t bn, struct indir *ap, int *nump)
292 {
293 	long blockcnt;
294 	e2fs_lbn_t metalbn, realbn;
295 	struct ext2mount *ump;
296 	int i, numlevels, off;
297 	int64_t qblockcnt;
298 
299 	ump = VFSTOEXT2(vp->v_mount);
300 	if (nump)
301 		*nump = 0;
302 	numlevels = 0;
303 	realbn = bn;
304 	if ((long)bn < 0)
305 		bn = -(long)bn;
306 
307 	/* The first NDADDR blocks are direct blocks. */
308 	if (bn < NDADDR)
309 		return (0);
310 
311 	/*
312 	 * Determine the number of levels of indirection.  After this loop
313 	 * is done, blockcnt indicates the number of data blocks possible
314 	 * at the previous level of indirection, and NIADDR - i is the number
315 	 * of levels of indirection needed to locate the requested block.
316 	 */
317 	for (blockcnt = 1, i = NIADDR, bn -= NDADDR;; i--, bn -= blockcnt) {
318 		if (i == 0)
319 			return (EFBIG);
320 		/*
321 		 * Use int64_t's here to avoid overflow for triple indirect
322 		 * blocks when longs have 32 bits and the block size is more
323 		 * than 4K.
324 		 */
325 		qblockcnt = (int64_t)blockcnt * MNINDIR(ump);
326 		if (bn < qblockcnt)
327 			break;
328 		blockcnt = qblockcnt;
329 	}
330 
331 	/* Calculate the address of the first meta-block. */
332 	if (realbn >= 0)
333 		metalbn = -(realbn - bn + NIADDR - i);
334 	else
335 		metalbn = -(-realbn - bn + NIADDR - i);
336 
337 	/*
338 	 * At each iteration, off is the offset into the bap array which is
339 	 * an array of disk addresses at the current level of indirection.
340 	 * The logical block number and the offset in that block are stored
341 	 * into the argument array.
342 	 */
343 	ap->in_lbn = metalbn;
344 	ap->in_off = off = NIADDR - i;
345 	ap++;
346 	for (++numlevels; i <= NIADDR; i++) {
347 		/* If searching for a meta-data block, quit when found. */
348 		if (metalbn == realbn)
349 			break;
350 
351 		off = (bn / blockcnt) % MNINDIR(ump);
352 
353 		++numlevels;
354 		ap->in_lbn = metalbn;
355 		ap->in_off = off;
356 		++ap;
357 
358 		metalbn -= -1 + off * blockcnt;
359 		blockcnt /= MNINDIR(ump);
360 	}
361 	if (nump)
362 		*nump = numlevels;
363 	return (0);
364 }
365