1 /*-
2 * SPDX-License-Identifier: BSD-3-Clause
3 *
4 * Copyright (c) 1989, 1991, 1993
5 * The Regents of the University of California. All rights reserved.
6 * (c) UNIX System Laboratories, Inc.
7 * All or some portions of this file are derived from material licensed
8 * to the University of California by American Telephone and Telegraph
9 * Co. or Unix System Laboratories, Inc. and are reproduced herein with
10 * the permission of UNIX System Laboratories, Inc.
11 *
12 * Redistribution and use in source and binary forms, with or without
13 * modification, are permitted provided that the following conditions
14 * are met:
15 * 1. Redistributions of source code must retain the above copyright
16 * notice, this list of conditions and the following disclaimer.
17 * 2. Redistributions in binary form must reproduce the above copyright
18 * notice, this list of conditions and the following disclaimer in the
19 * documentation and/or other materials provided with the distribution.
20 * 3. Neither the name of the University nor the names of its contributors
21 * may be used to endorse or promote products derived from this software
22 * without specific prior written permission.
23 *
24 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
25 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
26 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
27 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
28 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
29 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
30 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
31 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
32 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
33 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
34 * SUCH DAMAGE.
35 *
36 * @(#)ufs_bmap.c 8.7 (Berkeley) 3/21/95
37 */
38
39 #include <sys/param.h>
40 #include <sys/systm.h>
41 #include <sys/bio.h>
42 #include <sys/buf.h>
43 #include <sys/endian.h>
44 #include <sys/proc.h>
45 #include <sys/vnode.h>
46 #include <sys/mount.h>
47 #include <sys/racct.h>
48 #include <sys/resourcevar.h>
49 #include <sys/stat.h>
50
51 #include <fs/ext2fs/fs.h>
52 #include <fs/ext2fs/inode.h>
53 #include <fs/ext2fs/ext2fs.h>
54 #include <fs/ext2fs/ext2_dinode.h>
55 #include <fs/ext2fs/ext2_extern.h>
56 #include <fs/ext2fs/ext2_mount.h>
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
ext2_bmap(struct vop_bmap_args * ap)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 int
ext4_bmapext(struct vnode * vp,int32_t bn,int64_t * bnp,int * runp,int * runb)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 mount *mp;
98 struct ext2mount *ump;
99 struct ext4_extent_header *ehp;
100 struct ext4_extent *ep;
101 struct ext4_extent_path *path = NULL;
102 daddr_t lbn;
103 int error, depth, maxrun = 0, bsize;
104
105 ip = VTOI(vp);
106 fs = ip->i_e2fs;
107 mp = vp->v_mount;
108 ump = VFSTOEXT2(mp);
109 lbn = bn;
110 ehp = (struct ext4_extent_header *)ip->i_data;
111 depth = le16toh(ehp->eh_depth);
112 bsize = EXT2_BLOCK_SIZE(ump->um_e2fs);
113
114 *bnp = -1;
115 if (runp != NULL) {
116 maxrun = mp->mnt_iosize_max / bsize - 1;
117 *runp = 0;
118 }
119 if (runb != NULL)
120 *runb = 0;
121
122 error = ext4_ext_find_extent(ip, lbn, &path);
123 if (error)
124 return (error);
125
126 ep = path[depth].ep_ext;
127 if(ep) {
128 if (lbn < le32toh(ep->e_blk)) {
129 if (runp != NULL) {
130 *runp = min(maxrun, le32toh(ep->e_blk) - lbn - 1);
131 }
132 } else if (le32toh(ep->e_blk) <= lbn &&
133 lbn < le32toh(ep->e_blk) + le16toh(ep->e_len)) {
134 *bnp = fsbtodb(fs, lbn - le32toh(ep->e_blk) +
135 (le32toh(ep->e_start_lo) |
136 (daddr_t)le16toh(ep->e_start_hi) << 32));
137 if (runp != NULL) {
138 *runp = min(maxrun,
139 le16toh(ep->e_len) -
140 (lbn - le32toh(ep->e_blk)) - 1);
141 }
142 if (runb != NULL)
143 *runb = min(maxrun, lbn - le32toh(ep->e_blk));
144 } else {
145 if (runb != NULL)
146 *runb = min(maxrun, le32toh(ep->e_blk) + lbn -
147 le16toh(ep->e_len));
148 }
149 }
150
151 ext4_ext_path_free(path);
152
153 return (error);
154 }
155
156 static int
readindir(struct vnode * vp,e2fs_lbn_t lbn,e2fs_daddr_t daddr,struct buf ** bpp)157 readindir(struct vnode *vp, e2fs_lbn_t lbn, e2fs_daddr_t daddr, struct buf **bpp)
158 {
159 struct buf *bp;
160 struct mount *mp;
161 struct ext2mount *ump;
162 int error;
163
164 mp = vp->v_mount;
165 ump = VFSTOEXT2(mp);
166
167 bp = getblk(vp, lbn, mp->mnt_stat.f_iosize, 0, 0, 0);
168 if ((bp->b_flags & B_CACHE) == 0) {
169 KASSERT(daddr != 0,
170 ("readindir: indirect block not in cache"));
171
172 bp->b_blkno = blkptrtodb(ump, daddr);
173 bp->b_iocmd = BIO_READ;
174 bp->b_flags &= ~B_INVAL;
175 bp->b_ioflags &= ~BIO_ERROR;
176 vfs_busy_pages(bp, 0);
177 bp->b_iooffset = dbtob(bp->b_blkno);
178 bstrategy(bp);
179 #ifdef RACCT
180 if (racct_enable) {
181 PROC_LOCK(curproc);
182 racct_add_buf(curproc, bp, 0);
183 PROC_UNLOCK(curproc);
184 }
185 #endif
186 curthread->td_ru.ru_inblock++;
187 error = bufwait(bp);
188 if (error != 0) {
189 brelse(bp);
190 return (error);
191 }
192 }
193 *bpp = bp;
194 return (0);
195 }
196
197 /*
198 * Indirect blocks are now on the vnode for the file. They are given negative
199 * logical block numbers. Indirect blocks are addressed by the negative
200 * address of the first data block to which they point. Double indirect blocks
201 * are addressed by one less than the address of the first indirect block to
202 * which they point. Triple indirect blocks are addressed by one less than
203 * the address of the first double indirect block to which they point.
204 *
205 * ext2_bmaparray does the bmap conversion, and if requested returns the
206 * array of logical blocks which must be traversed to get to a block.
207 * Each entry contains the offset into that block that gets you to the
208 * next block and the disk address of the block (if it is assigned).
209 */
210
211 int
ext2_bmaparray(struct vnode * vp,daddr_t bn,daddr_t * bnp,int * runp,int * runb)212 ext2_bmaparray(struct vnode *vp, daddr_t bn, daddr_t *bnp, int *runp, int *runb)
213 {
214 struct inode *ip;
215 struct buf *bp;
216 struct ext2mount *ump;
217 struct mount *mp;
218 struct indir a[EXT2_NIADDR + 1], *ap;
219 daddr_t daddr;
220 e2fs_lbn_t metalbn;
221 int error, num, maxrun = 0, bsize;
222 int *nump;
223
224 ap = NULL;
225 ip = VTOI(vp);
226 mp = vp->v_mount;
227 ump = VFSTOEXT2(mp);
228
229 bsize = EXT2_BLOCK_SIZE(ump->um_e2fs);
230
231 if (runp) {
232 maxrun = mp->mnt_iosize_max / bsize - 1;
233 *runp = 0;
234 }
235 if (runb)
236 *runb = 0;
237
238 ap = a;
239 nump = #
240 error = ext2_getlbns(vp, bn, ap, nump);
241 if (error)
242 return (error);
243
244 num = *nump;
245 if (num == 0) {
246 *bnp = blkptrtodb(ump, ip->i_db[bn]);
247 if (*bnp == 0) {
248 *bnp = -1;
249 } else if (runp) {
250 daddr_t bnb = bn;
251
252 for (++bn; bn < EXT2_NDADDR && *runp < maxrun &&
253 is_sequential(ump, ip->i_db[bn - 1], ip->i_db[bn]);
254 ++bn, ++*runp);
255 bn = bnb;
256 if (runb && (bn > 0)) {
257 for (--bn; (bn >= 0) && (*runb < maxrun) &&
258 is_sequential(ump, ip->i_db[bn],
259 ip->i_db[bn + 1]);
260 --bn, ++*runb);
261 }
262 }
263 return (0);
264 }
265
266 /* Get disk address out of indirect block array */
267 daddr = ip->i_ib[ap->in_off];
268
269 for (bp = NULL, ++ap; --num; ++ap) {
270 /*
271 * Exit the loop if there is no disk address assigned yet and
272 * the indirect block isn't in the cache, or if we were
273 * looking for an indirect block and we've found it.
274 */
275
276 metalbn = ap->in_lbn;
277 if ((daddr == 0 && !incore(&vp->v_bufobj, metalbn)) || metalbn == bn)
278 break;
279 /*
280 * If we get here, we've either got the block in the cache
281 * or we have a disk address for it, go fetch it.
282 */
283 if (bp)
284 bqrelse(bp);
285 error = readindir(vp, metalbn, daddr, &bp);
286 if (error != 0)
287 return (error);
288
289 daddr = le32toh(((e2fs_daddr_t *)bp->b_data)[ap->in_off]);
290 if (num == 1 && daddr && runp) {
291 for (bn = ap->in_off + 1;
292 bn < MNINDIR(ump) && *runp < maxrun &&
293 is_sequential(ump,
294 ((e2fs_daddr_t *)bp->b_data)[bn - 1],
295 ((e2fs_daddr_t *)bp->b_data)[bn]);
296 ++bn, ++*runp);
297 bn = ap->in_off;
298 if (runb && bn) {
299 for (--bn; bn >= 0 && *runb < maxrun &&
300 is_sequential(ump,
301 ((e2fs_daddr_t *)bp->b_data)[bn],
302 ((e2fs_daddr_t *)bp->b_data)[bn + 1]);
303 --bn, ++*runb);
304 }
305 }
306 }
307 if (bp)
308 bqrelse(bp);
309
310 *bnp = blkptrtodb(ump, daddr);
311 if (*bnp == 0) {
312 *bnp = -1;
313 }
314 return (0);
315 }
316
317 static e2fs_lbn_t
lbn_count(struct ext2mount * ump,int level)318 lbn_count(struct ext2mount *ump, int level)
319
320 {
321 e2fs_lbn_t blockcnt;
322
323 for (blockcnt = 1; level > 0; level--)
324 blockcnt *= MNINDIR(ump);
325 return (blockcnt);
326 }
327
328 int
ext2_bmap_seekdata(struct vnode * vp,off_t * offp)329 ext2_bmap_seekdata(struct vnode *vp, off_t *offp)
330 {
331 struct buf *bp;
332 struct indir a[EXT2_NIADDR + 1], *ap;
333 struct inode *ip;
334 struct mount *mp;
335 struct ext2mount *ump;
336 e2fs_daddr_t bn, daddr, nextbn;
337 uint64_t bsize;
338 off_t numblks;
339 int error, num, num1, off;
340
341 bp = NULL;
342 error = 0;
343 ip = VTOI(vp);
344 mp = vp->v_mount;
345 ump = VFSTOEXT2(mp);
346
347 if (vp->v_type != VREG)
348 return (EINVAL);
349 if (*offp < 0 || *offp >= ip->i_size)
350 return (ENXIO);
351
352 bsize = mp->mnt_stat.f_iosize;
353 for (bn = *offp / bsize, numblks = howmany(ip->i_size, bsize);
354 bn < numblks; bn = nextbn) {
355 if (bn < EXT2_NDADDR) {
356 daddr = ip->i_db[bn];
357 if (daddr != 0)
358 break;
359 nextbn = bn + 1;
360 continue;
361 }
362
363 ap = a;
364 error = ext2_getlbns(vp, bn, ap, &num);
365 if (error != 0)
366 break;
367 MPASS(num >= 2);
368 daddr = ip->i_ib[ap->in_off];
369 ap++, num--;
370 for (nextbn = EXT2_NDADDR, num1 = num - 1; num1 > 0; num1--)
371 nextbn += lbn_count(ump, num1);
372 if (daddr == 0) {
373 nextbn += lbn_count(ump, num);
374 continue;
375 }
376
377 for (; daddr != 0 && num > 0; ap++, num--) {
378 if (bp != NULL)
379 bqrelse(bp);
380 error = readindir(vp, ap->in_lbn, daddr, &bp);
381 if (error != 0)
382 return (error);
383
384 /*
385 * Scan the indirect block until we find a non-zero
386 * pointer.
387 */
388 off = ap->in_off;
389 do {
390 daddr = le32toh(((e2fs_daddr_t *)bp->b_data)[off]);
391 } while (daddr == 0 && ++off < MNINDIR(ump));
392 nextbn += off * lbn_count(ump, num - 1);
393
394 /*
395 * We need to recompute the LBNs of indirect
396 * blocks, so restart with the updated block offset.
397 */
398 if (off != ap->in_off)
399 break;
400 }
401 if (num == 0) {
402 /*
403 * We found a data block.
404 */
405 bn = nextbn;
406 break;
407 }
408 }
409 if (bp != NULL)
410 bqrelse(bp);
411 if (bn >= numblks)
412 error = ENXIO;
413 if (error == 0 && *offp < bn * bsize)
414 *offp = bn * bsize;
415 return (error);
416 }
417
418 /*
419 * Create an array of logical block number/offset pairs which represent the
420 * path of indirect blocks required to access a data block. The first "pair"
421 * contains the logical block number of the appropriate single, double or
422 * triple indirect block and the offset into the inode indirect block array.
423 * Note, the logical block number of the inode single/double/triple indirect
424 * block appears twice in the array, once with the offset into the i_ib and
425 * once with the offset into the page itself.
426 */
427 int
ext2_getlbns(struct vnode * vp,daddr_t bn,struct indir * ap,int * nump)428 ext2_getlbns(struct vnode *vp, daddr_t bn, struct indir *ap, int *nump)
429 {
430 long blockcnt;
431 e2fs_lbn_t metalbn, realbn;
432 struct ext2mount *ump;
433 int i, numlevels, off;
434 int64_t qblockcnt;
435
436 ump = VFSTOEXT2(vp->v_mount);
437 if (nump)
438 *nump = 0;
439 numlevels = 0;
440 realbn = bn;
441 if ((long)bn < 0)
442 bn = -(long)bn;
443
444 /* The first EXT2_NDADDR blocks are direct blocks. */
445 if (bn < EXT2_NDADDR)
446 return (0);
447
448 /*
449 * Determine the number of levels of indirection. After this loop
450 * is done, blockcnt indicates the number of data blocks possible
451 * at the previous level of indirection, and EXT2_NIADDR - i is the
452 * number of levels of indirection needed to locate the requested block.
453 */
454 for (blockcnt = 1, i = EXT2_NIADDR, bn -= EXT2_NDADDR; ;
455 i--, bn -= blockcnt) {
456 if (i == 0)
457 return (EFBIG);
458 /*
459 * Use int64_t's here to avoid overflow for triple indirect
460 * blocks when longs have 32 bits and the block size is more
461 * than 4K.
462 */
463 qblockcnt = (int64_t)blockcnt * MNINDIR(ump);
464 if (bn < qblockcnt)
465 break;
466 blockcnt = qblockcnt;
467 }
468
469 /* Calculate the address of the first meta-block. */
470 if (realbn >= 0)
471 metalbn = -(realbn - bn + EXT2_NIADDR - i);
472 else
473 metalbn = -(-realbn - bn + EXT2_NIADDR - i);
474
475 /*
476 * At each iteration, off is the offset into the bap array which is
477 * an array of disk addresses at the current level of indirection.
478 * The logical block number and the offset in that block are stored
479 * into the argument array.
480 */
481 ap->in_lbn = metalbn;
482 ap->in_off = off = EXT2_NIADDR - i;
483 ap++;
484 for (++numlevels; i <= EXT2_NIADDR; i++) {
485 /* If searching for a meta-data block, quit when found. */
486 if (metalbn == realbn)
487 break;
488
489 off = (bn / blockcnt) % MNINDIR(ump);
490
491 ++numlevels;
492 ap->in_lbn = metalbn;
493 ap->in_off = off;
494 ++ap;
495
496 metalbn -= -1 + off * blockcnt;
497 blockcnt /= MNINDIR(ump);
498 }
499 if (nump)
500 *nump = numlevels;
501 return (0);
502 }
503