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_flag & IN_E4EXTENTS)
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 = {.ep_bp = NULL};
98 daddr_t lbn;
99 int ret = 0;
100
101 ip = VTOI(vp);
102 fs = ip->i_e2fs;
103 lbn = bn;
104
105 /*
106 * TODO: need to implement read ahead to improve the performance.
107 */
108 if (runp != NULL)
109 *runp = 0;
110
111 if (runb != NULL)
112 *runb = 0;
113
114 ext4_ext_find_extent(fs, ip, lbn, &path);
115 ep = path.ep_ext;
116 if (ep == NULL)
117 ret = EIO;
118 else {
119 *bnp = fsbtodb(fs, lbn - ep->e_blk +
120 (ep->e_start_lo | (daddr_t)ep->e_start_hi << 32));
121
122 if (*bnp == 0)
123 *bnp = -1;
124 }
125
126 if (path.ep_bp != NULL) {
127 brelse(path.ep_bp);
128 path.ep_bp = NULL;
129 }
130
131 return (ret);
132 }
133
134 /*
135 * Indirect blocks are now on the vnode for the file. They are given negative
136 * logical block numbers. Indirect blocks are addressed by the negative
137 * address of the first data block to which they point. Double indirect blocks
138 * are addressed by one less than the address of the first indirect block to
139 * which they point. Triple indirect blocks are addressed by one less than
140 * the address of the first double indirect block to which they point.
141 *
142 * ext2_bmaparray does the bmap conversion, and if requested returns the
143 * array of logical blocks which must be traversed to get to a block.
144 * Each entry contains the offset into that block that gets you to the
145 * next block and the disk address of the block (if it is assigned).
146 */
147
148 int
149 ext2_bmaparray(struct vnode *vp, daddr_t bn, daddr_t *bnp, int *runp, int *runb)
150 {
151 struct inode *ip;
152 struct buf *bp;
153 struct ext2mount *ump;
154 struct mount *mp;
155 struct indir a[NIADDR + 1], *ap;
156 daddr_t daddr;
157 e2fs_lbn_t metalbn;
158 int error, num, maxrun = 0, bsize;
159 int *nump;
160
161 ap = NULL;
162 ip = VTOI(vp);
163 mp = vp->v_mount;
164 ump = VFSTOEXT2(mp);
165
166 bsize = EXT2_BLOCK_SIZE(ump->um_e2fs);
167
168 if (runp) {
169 maxrun = mp->mnt_iosize_max / bsize - 1;
170 *runp = 0;
171 }
172 if (runb)
173 *runb = 0;
174
175
176 ap = a;
177 nump = #
178 error = ext2_getlbns(vp, bn, ap, nump);
179 if (error)
180 return (error);
181
182 num = *nump;
183 if (num == 0) {
184 *bnp = blkptrtodb(ump, ip->i_db[bn]);
185 if (*bnp == 0) {
186 *bnp = -1;
187 } else if (runp) {
188 daddr_t bnb = bn;
189
190 for (++bn; bn < NDADDR && *runp < maxrun &&
191 is_sequential(ump, ip->i_db[bn - 1], ip->i_db[bn]);
192 ++bn, ++*runp);
193 bn = bnb;
194 if (runb && (bn > 0)) {
195 for (--bn; (bn >= 0) && (*runb < maxrun) &&
196 is_sequential(ump, ip->i_db[bn],
197 ip->i_db[bn + 1]);
198 --bn, ++*runb);
199 }
200 }
201 return (0);
202 }
203
204 /* Get disk address out of indirect block array */
205 daddr = ip->i_ib[ap->in_off];
206
207 for (bp = NULL, ++ap; --num; ++ap) {
208 /*
209 * Exit the loop if there is no disk address assigned yet and
210 * the indirect block isn't in the cache, or if we were
211 * looking for an indirect block and we've found it.
212 */
213
214 metalbn = ap->in_lbn;
215 if ((daddr == 0 && !incore(&vp->v_bufobj, metalbn)) || metalbn == bn)
216 break;
217 /*
218 * If we get here, we've either got the block in the cache
219 * or we have a disk address for it, go fetch it.
220 */
221 if (bp)
222 bqrelse(bp);
223
224 bp = getblk(vp, metalbn, bsize, 0, 0, 0);
225 if ((bp->b_flags & B_CACHE) == 0) {
226 #ifdef INVARIANTS
227 if (!daddr)
228 panic("ext2_bmaparray: indirect block not in cache");
229 #endif
230 bp->b_blkno = blkptrtodb(ump, daddr);
231 bp->b_iocmd = BIO_READ;
232 bp->b_flags &= ~B_INVAL;
233 bp->b_ioflags &= ~BIO_ERROR;
234 vfs_busy_pages(bp, 0);
235 bp->b_iooffset = dbtob(bp->b_blkno);
236 bstrategy(bp);
237 curthread->td_ru.ru_inblock++;
238 error = bufwait(bp);
239 if (error) {
240 brelse(bp);
241 return (error);
242 }
243 }
244
245 daddr = ((e2fs_daddr_t *)bp->b_data)[ap->in_off];
246 if (num == 1 && daddr && runp) {
247 for (bn = ap->in_off + 1;
248 bn < MNINDIR(ump) && *runp < maxrun &&
249 is_sequential(ump,
250 ((e2fs_daddr_t *)bp->b_data)[bn - 1],
251 ((e2fs_daddr_t *)bp->b_data)[bn]);
252 ++bn, ++*runp);
253 bn = ap->in_off;
254 if (runb && bn) {
255 for (--bn; bn >= 0 && *runb < maxrun &&
256 is_sequential(ump,
257 ((e2fs_daddr_t *)bp->b_data)[bn],
258 ((e2fs_daddr_t *)bp->b_data)[bn + 1]);
259 --bn, ++*runb);
260 }
261 }
262 }
263 if (bp)
264 bqrelse(bp);
265
266 /*
267 * Since this is FFS independent code, we are out of scope for the
268 * definitions of BLK_NOCOPY and BLK_SNAP, but we do know that they
269 * will fall in the range 1..um_seqinc, so we use that test and
270 * return a request for a zeroed out buffer if attempts are made
271 * to read a BLK_NOCOPY or BLK_SNAP block.
272 */
273 if ((ip->i_flags & SF_SNAPSHOT) && daddr > 0 && daddr < ump->um_seqinc) {
274 *bnp = -1;
275 return (0);
276 }
277 *bnp = blkptrtodb(ump, daddr);
278 if (*bnp == 0) {
279 *bnp = -1;
280 }
281 return (0);
282 }
283
284 /*
285 * Create an array of logical block number/offset pairs which represent the
286 * path of indirect blocks required to access a data block. The first "pair"
287 * contains the logical block number of the appropriate single, double or
288 * triple indirect block and the offset into the inode indirect block array.
289 * Note, the logical block number of the inode single/double/triple indirect
290 * block appears twice in the array, once with the offset into the i_ib and
291 * once with the offset into the page itself.
292 */
293 int
294 ext2_getlbns(struct vnode *vp, daddr_t bn, struct indir *ap, int *nump)
295 {
296 long blockcnt;
297 e2fs_lbn_t metalbn, realbn;
298 struct ext2mount *ump;
299 int i, numlevels, off;
300 int64_t qblockcnt;
301
302 ump = VFSTOEXT2(vp->v_mount);
303 if (nump)
304 *nump = 0;
305 numlevels = 0;
306 realbn = bn;
307 if ((long)bn < 0)
308 bn = -(long)bn;
309
310 /* The first NDADDR blocks are direct blocks. */
311 if (bn < NDADDR)
312 return (0);
313
314 /*
315 * Determine the number of levels of indirection. After this loop
316 * is done, blockcnt indicates the number of data blocks possible
317 * at the previous level of indirection, and NIADDR - i is the number
318 * of levels of indirection needed to locate the requested block.
319 */
320 for (blockcnt = 1, i = NIADDR, bn -= NDADDR;; i--, bn -= blockcnt) {
321 if (i == 0)
322 return (EFBIG);
323 /*
324 * Use int64_t's here to avoid overflow for triple indirect
325 * blocks when longs have 32 bits and the block size is more
326 * than 4K.
327 */
328 qblockcnt = (int64_t)blockcnt * MNINDIR(ump);
329 if (bn < qblockcnt)
330 break;
331 blockcnt = qblockcnt;
332 }
333
334 /* Calculate the address of the first meta-block. */
335 if (realbn >= 0)
336 metalbn = -(realbn - bn + NIADDR - i);
337 else
338 metalbn = -(-realbn - bn + NIADDR - i);
339
340 /*
341 * At each iteration, off is the offset into the bap array which is
342 * an array of disk addresses at the current level of indirection.
343 * The logical block number and the offset in that block are stored
344 * into the argument array.
345 */
346 ap->in_lbn = metalbn;
347 ap->in_off = off = NIADDR - i;
348 ap++;
349 for (++numlevels; i <= NIADDR; i++) {
350 /* If searching for a meta-data block, quit when found. */
351 if (metalbn == realbn)
352 break;
353
354 off = (bn / blockcnt) % MNINDIR(ump);
355
356 ++numlevels;
357 ap->in_lbn = metalbn;
358 ap->in_off = off;
359 ++ap;
360
361 metalbn -= -1 + off * blockcnt;
362 blockcnt /= MNINDIR(ump);
363 }
364 if (nump)
365 *nump = numlevels;
366 return (0);
367 }
Cache object: 14c0c7a8973ca0a369990704eff5e521
|