1 | /* $NetBSD: ulfs_bmap.c,v 1.7 2015/09/01 06:08:37 dholland Exp $ */ |
2 | /* from NetBSD: ufs_bmap.c,v 1.50 2013/01/22 09:39:18 dholland Exp */ |
3 | |
4 | /* |
5 | * Copyright (c) 1989, 1991, 1993 |
6 | * The Regents of the University of California. All rights reserved. |
7 | * (c) UNIX System Laboratories, Inc. |
8 | * All or some portions of this file are derived from material licensed |
9 | * to the University of California by American Telephone and Telegraph |
10 | * Co. or Unix System Laboratories, Inc. and are reproduced herein with |
11 | * the permission of UNIX System Laboratories, Inc. |
12 | * |
13 | * Redistribution and use in source and binary forms, with or without |
14 | * modification, are permitted provided that the following conditions |
15 | * are met: |
16 | * 1. Redistributions of source code must retain the above copyright |
17 | * notice, this list of conditions and the following disclaimer. |
18 | * 2. Redistributions in binary form must reproduce the above copyright |
19 | * notice, this list of conditions and the following disclaimer in the |
20 | * documentation and/or other materials provided with the distribution. |
21 | * 3. Neither the name of the University nor the names of its contributors |
22 | * may be used to endorse or promote products derived from this software |
23 | * without specific prior written permission. |
24 | * |
25 | * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND |
26 | * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE |
27 | * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE |
28 | * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE |
29 | * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL |
30 | * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS |
31 | * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) |
32 | * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT |
33 | * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY |
34 | * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF |
35 | * SUCH DAMAGE. |
36 | * |
37 | * @(#)ufs_bmap.c 8.8 (Berkeley) 8/11/95 |
38 | */ |
39 | |
40 | #include <sys/cdefs.h> |
41 | __KERNEL_RCSID(0, "$NetBSD: ulfs_bmap.c,v 1.7 2015/09/01 06:08:37 dholland Exp $" ); |
42 | |
43 | #include <sys/param.h> |
44 | #include <sys/systm.h> |
45 | #include <sys/stat.h> |
46 | #include <sys/buf.h> |
47 | #include <sys/proc.h> |
48 | #include <sys/vnode.h> |
49 | #include <sys/mount.h> |
50 | #include <sys/resourcevar.h> |
51 | #include <sys/trace.h> |
52 | #include <sys/fstrans.h> |
53 | |
54 | #include <miscfs/specfs/specdev.h> |
55 | |
56 | #include <ufs/lfs/ulfs_inode.h> |
57 | #include <ufs/lfs/ulfsmount.h> |
58 | #include <ufs/lfs/ulfs_extern.h> |
59 | #include <ufs/lfs/ulfs_bswap.h> |
60 | |
61 | static bool |
62 | ulfs_issequential(const struct lfs *fs, daddr_t daddr0, daddr_t daddr1) |
63 | { |
64 | |
65 | /* for ulfs, blocks in a hole is not 'contiguous'. */ |
66 | if (daddr0 == 0) |
67 | return false; |
68 | |
69 | return (daddr0 + fs->um_seqinc == daddr1); |
70 | } |
71 | |
72 | /* |
73 | * This is used for block pointers in inodes and elsewhere, which can |
74 | * contain the magic value UNWRITTEN, which is -2. This is mishandled |
75 | * by u32 -> u64 promotion unless special-cased. |
76 | * |
77 | * XXX this should be rolled into better inode accessors and go away. |
78 | */ |
79 | static inline uint64_t |
80 | ulfs_fix_unwritten(uint32_t val) |
81 | { |
82 | if (val == (uint32_t)UNWRITTEN) { |
83 | return (uint64_t)(int64_t)UNWRITTEN; |
84 | } else { |
85 | return val; |
86 | } |
87 | } |
88 | |
89 | |
90 | /* |
91 | * Bmap converts the logical block number of a file to its physical block |
92 | * number on the disk. The conversion is done by using the logical block |
93 | * number to index into the array of block pointers described by the dinode. |
94 | */ |
95 | int |
96 | ulfs_bmap(void *v) |
97 | { |
98 | struct vop_bmap_args /* { |
99 | struct vnode *a_vp; |
100 | daddr_t a_bn; |
101 | struct vnode **a_vpp; |
102 | daddr_t *a_bnp; |
103 | int *a_runp; |
104 | } */ *ap = v; |
105 | int error; |
106 | |
107 | /* |
108 | * Check for underlying vnode requests and ensure that logical |
109 | * to physical mapping is requested. |
110 | */ |
111 | if (ap->a_vpp != NULL) |
112 | *ap->a_vpp = VTOI(ap->a_vp)->i_devvp; |
113 | if (ap->a_bnp == NULL) |
114 | return (0); |
115 | |
116 | fstrans_start(ap->a_vp->v_mount, FSTRANS_SHARED); |
117 | error = ulfs_bmaparray(ap->a_vp, ap->a_bn, ap->a_bnp, NULL, NULL, |
118 | ap->a_runp, ulfs_issequential); |
119 | fstrans_done(ap->a_vp->v_mount); |
120 | return error; |
121 | } |
122 | |
123 | /* |
124 | * Indirect blocks are now on the vnode for the file. They are given negative |
125 | * logical block numbers. Indirect blocks are addressed by the negative |
126 | * address of the first data block to which they point. Double indirect blocks |
127 | * are addressed by one less than the address of the first indirect block to |
128 | * which they point. Triple indirect blocks are addressed by one less than |
129 | * the address of the first double indirect block to which they point. |
130 | * |
131 | * ulfs_bmaparray does the bmap conversion, and if requested returns the |
132 | * array of logical blocks which must be traversed to get to a block. |
133 | * Each entry contains the offset into that block that gets you to the |
134 | * next block and the disk address of the block (if it is assigned). |
135 | */ |
136 | |
137 | int |
138 | ulfs_bmaparray(struct vnode *vp, daddr_t bn, daddr_t *bnp, struct indir *ap, |
139 | int *nump, int *runp, ulfs_issequential_callback_t is_sequential) |
140 | { |
141 | struct inode *ip; |
142 | struct buf *bp, *cbp; |
143 | struct ulfsmount *ump; |
144 | struct lfs *fs; |
145 | struct mount *mp; |
146 | struct indir a[ULFS_NIADDR + 1], *xap; |
147 | daddr_t daddr; |
148 | daddr_t metalbn; |
149 | int error, maxrun = 0, num; |
150 | |
151 | ip = VTOI(vp); |
152 | mp = vp->v_mount; |
153 | ump = ip->i_ump; |
154 | fs = ip->i_lfs; |
155 | #ifdef DIAGNOSTIC |
156 | if ((ap != NULL && nump == NULL) || (ap == NULL && nump != NULL)) |
157 | panic("ulfs_bmaparray: invalid arguments" ); |
158 | #endif |
159 | |
160 | if (runp) { |
161 | /* |
162 | * XXX |
163 | * If MAXBSIZE is the largest transfer the disks can handle, |
164 | * we probably want maxrun to be 1 block less so that we |
165 | * don't create a block larger than the device can handle. |
166 | */ |
167 | *runp = 0; |
168 | maxrun = MAXPHYS / mp->mnt_stat.f_iosize - 1; |
169 | } |
170 | |
171 | if (bn >= 0 && bn < ULFS_NDADDR) { |
172 | if (nump != NULL) |
173 | *nump = 0; |
174 | if (ump->um_fstype == ULFS1) |
175 | daddr = ulfs_fix_unwritten(ulfs_rw32(ip->i_din->u_32.di_db[bn], |
176 | ULFS_MPNEEDSWAP(fs))); |
177 | else |
178 | daddr = ulfs_rw64(ip->i_din->u_64.di_db[bn], |
179 | ULFS_MPNEEDSWAP(fs)); |
180 | *bnp = blkptrtodb(fs, daddr); |
181 | /* |
182 | * Since this is FFS independent code, we are out of |
183 | * scope for the definitions of BLK_NOCOPY and |
184 | * BLK_SNAP, but we do know that they will fall in |
185 | * the range 1..um_seqinc, so we use that test and |
186 | * return a request for a zeroed out buffer if attempts |
187 | * are made to read a BLK_NOCOPY or BLK_SNAP block. |
188 | */ |
189 | if ((ip->i_flags & (SF_SNAPSHOT | SF_SNAPINVAL)) == SF_SNAPSHOT |
190 | && daddr > 0 && |
191 | daddr < fs->um_seqinc) { |
192 | *bnp = -1; |
193 | } else if (*bnp == 0) { |
194 | if ((ip->i_flags & (SF_SNAPSHOT | SF_SNAPINVAL)) |
195 | == SF_SNAPSHOT) { |
196 | *bnp = blkptrtodb(fs, bn * fs->um_seqinc); |
197 | } else { |
198 | *bnp = -1; |
199 | } |
200 | } else if (runp) { |
201 | if (ump->um_fstype == ULFS1) { |
202 | for (++bn; bn < ULFS_NDADDR && *runp < maxrun && |
203 | is_sequential(fs, |
204 | ulfs_fix_unwritten(ulfs_rw32(ip->i_din->u_32.di_db[bn - 1], |
205 | ULFS_MPNEEDSWAP(fs))), |
206 | ulfs_fix_unwritten(ulfs_rw32(ip->i_din->u_32.di_db[bn], |
207 | ULFS_MPNEEDSWAP(fs)))); |
208 | ++bn, ++*runp); |
209 | } else { |
210 | for (++bn; bn < ULFS_NDADDR && *runp < maxrun && |
211 | is_sequential(fs, |
212 | ulfs_rw64(ip->i_din->u_64.di_db[bn - 1], |
213 | ULFS_MPNEEDSWAP(fs)), |
214 | ulfs_rw64(ip->i_din->u_64.di_db[bn], |
215 | ULFS_MPNEEDSWAP(fs))); |
216 | ++bn, ++*runp); |
217 | } |
218 | } |
219 | return (0); |
220 | } |
221 | |
222 | xap = ap == NULL ? a : ap; |
223 | if (!nump) |
224 | nump = # |
225 | if ((error = ulfs_getlbns(vp, bn, xap, nump)) != 0) |
226 | return (error); |
227 | |
228 | num = *nump; |
229 | |
230 | /* Get disk address out of indirect block array */ |
231 | // XXX clean this up |
232 | if (ump->um_fstype == ULFS1) |
233 | daddr = ulfs_fix_unwritten(ulfs_rw32(ip->i_din->u_32.di_ib[xap->in_off], |
234 | ULFS_MPNEEDSWAP(fs))); |
235 | else |
236 | daddr = ulfs_rw64(ip->i_din->u_64.di_ib[xap->in_off], |
237 | ULFS_MPNEEDSWAP(fs)); |
238 | |
239 | for (bp = NULL, ++xap; --num; ++xap) { |
240 | /* |
241 | * Exit the loop if there is no disk address assigned yet and |
242 | * the indirect block isn't in the cache, or if we were |
243 | * looking for an indirect block and we've found it. |
244 | */ |
245 | |
246 | metalbn = xap->in_lbn; |
247 | if (metalbn == bn) |
248 | break; |
249 | if (daddr == 0) { |
250 | mutex_enter(&bufcache_lock); |
251 | cbp = incore(vp, metalbn); |
252 | mutex_exit(&bufcache_lock); |
253 | if (cbp == NULL) |
254 | break; |
255 | } |
256 | |
257 | /* |
258 | * If we get here, we've either got the block in the cache |
259 | * or we have a disk address for it, go fetch it. |
260 | */ |
261 | if (bp) |
262 | brelse(bp, 0); |
263 | |
264 | xap->in_exists = 1; |
265 | bp = getblk(vp, metalbn, mp->mnt_stat.f_iosize, 0, 0); |
266 | if (bp == NULL) { |
267 | |
268 | /* |
269 | * getblk() above returns NULL only iff we are |
270 | * pagedaemon. See the implementation of getblk |
271 | * for detail. |
272 | */ |
273 | |
274 | return (ENOMEM); |
275 | } |
276 | if (bp->b_oflags & (BO_DONE | BO_DELWRI)) { |
277 | trace(TR_BREADHIT, pack(vp, size), metalbn); |
278 | } |
279 | #ifdef DIAGNOSTIC |
280 | else if (!daddr) |
281 | panic("ulfs_bmaparray: indirect block not in cache" ); |
282 | #endif |
283 | else { |
284 | trace(TR_BREADMISS, pack(vp, size), metalbn); |
285 | bp->b_blkno = blkptrtodb(fs, daddr); |
286 | bp->b_flags |= B_READ; |
287 | BIO_SETPRIO(bp, BPRIO_TIMECRITICAL); |
288 | VOP_STRATEGY(vp, bp); |
289 | curlwp->l_ru.ru_inblock++; /* XXX */ |
290 | if ((error = biowait(bp)) != 0) { |
291 | brelse(bp, 0); |
292 | return (error); |
293 | } |
294 | } |
295 | if (ump->um_fstype == ULFS1) { |
296 | daddr = ulfs_fix_unwritten(ulfs_rw32(((u_int32_t *)bp->b_data)[xap->in_off], |
297 | ULFS_MPNEEDSWAP(fs))); |
298 | if (num == 1 && daddr && runp) { |
299 | for (bn = xap->in_off + 1; |
300 | bn < MNINDIR(fs) && *runp < maxrun && |
301 | is_sequential(fs, |
302 | ulfs_fix_unwritten(ulfs_rw32(((int32_t *)bp->b_data)[bn-1], |
303 | ULFS_MPNEEDSWAP(fs))), |
304 | ulfs_fix_unwritten(ulfs_rw32(((int32_t *)bp->b_data)[bn], |
305 | ULFS_MPNEEDSWAP(fs)))); |
306 | ++bn, ++*runp); |
307 | } |
308 | } else { |
309 | daddr = ulfs_rw64(((u_int64_t *)bp->b_data)[xap->in_off], |
310 | ULFS_MPNEEDSWAP(fs)); |
311 | if (num == 1 && daddr && runp) { |
312 | for (bn = xap->in_off + 1; |
313 | bn < MNINDIR(fs) && *runp < maxrun && |
314 | is_sequential(fs, |
315 | ulfs_rw64(((int64_t *)bp->b_data)[bn-1], |
316 | ULFS_MPNEEDSWAP(fs)), |
317 | ulfs_rw64(((int64_t *)bp->b_data)[bn], |
318 | ULFS_MPNEEDSWAP(fs))); |
319 | ++bn, ++*runp); |
320 | } |
321 | } |
322 | } |
323 | if (bp) |
324 | brelse(bp, 0); |
325 | |
326 | /* |
327 | * Since this is FFS independent code, we are out of scope for the |
328 | * definitions of BLK_NOCOPY and BLK_SNAP, but we do know that they |
329 | * will fall in the range 1..um_seqinc, so we use that test and |
330 | * return a request for a zeroed out buffer if attempts are made |
331 | * to read a BLK_NOCOPY or BLK_SNAP block. |
332 | */ |
333 | if ((ip->i_flags & (SF_SNAPSHOT | SF_SNAPINVAL)) == SF_SNAPSHOT |
334 | && daddr > 0 && daddr < fs->um_seqinc) { |
335 | *bnp = -1; |
336 | return (0); |
337 | } |
338 | *bnp = blkptrtodb(fs, daddr); |
339 | if (*bnp == 0) { |
340 | if ((ip->i_flags & (SF_SNAPSHOT | SF_SNAPINVAL)) |
341 | == SF_SNAPSHOT) { |
342 | *bnp = blkptrtodb(fs, bn * fs->um_seqinc); |
343 | } else { |
344 | *bnp = -1; |
345 | } |
346 | } |
347 | return (0); |
348 | } |
349 | |
350 | /* |
351 | * Create an array of logical block number/offset pairs which represent the |
352 | * path of indirect blocks required to access a data block. The first "pair" |
353 | * contains the logical block number of the appropriate single, double or |
354 | * triple indirect block and the offset into the inode indirect block array. |
355 | * Note, the logical block number of the inode single/double/triple indirect |
356 | * block appears twice in the array, once with the offset into the i_ffs1_ib and |
357 | * once with the offset into the page itself. |
358 | */ |
359 | int |
360 | ulfs_getlbns(struct vnode *vp, daddr_t bn, struct indir *ap, int *nump) |
361 | { |
362 | daddr_t metalbn, realbn; |
363 | struct ulfsmount *ump; |
364 | struct lfs *fs; |
365 | int64_t blockcnt; |
366 | int lbc; |
367 | int i, numlevels, off; |
368 | |
369 | ump = VFSTOULFS(vp->v_mount); |
370 | fs = ump->um_lfs; |
371 | if (nump) |
372 | *nump = 0; |
373 | numlevels = 0; |
374 | realbn = bn; |
375 | if (bn < 0) |
376 | bn = -bn; |
377 | KASSERT(bn >= ULFS_NDADDR); |
378 | |
379 | /* |
380 | * Determine the number of levels of indirection. After this loop |
381 | * is done, blockcnt indicates the number of data blocks possible |
382 | * at the given level of indirection, and ULFS_NIADDR - i is the number |
383 | * of levels of indirection needed to locate the requested block. |
384 | */ |
385 | |
386 | bn -= ULFS_NDADDR; |
387 | for (lbc = 0, i = ULFS_NIADDR;; i--, bn -= blockcnt) { |
388 | if (i == 0) |
389 | return (EFBIG); |
390 | |
391 | lbc += fs->um_lognindir; |
392 | blockcnt = (int64_t)1 << lbc; |
393 | |
394 | if (bn < blockcnt) |
395 | break; |
396 | } |
397 | |
398 | /* Calculate the address of the first meta-block. */ |
399 | metalbn = -((realbn >= 0 ? realbn : -realbn) - bn + ULFS_NIADDR - i); |
400 | |
401 | /* |
402 | * At each iteration, off is the offset into the bap array which is |
403 | * an array of disk addresses at the current level of indirection. |
404 | * The logical block number and the offset in that block are stored |
405 | * into the argument array. |
406 | */ |
407 | ap->in_lbn = metalbn; |
408 | ap->in_off = off = ULFS_NIADDR - i; |
409 | ap->in_exists = 0; |
410 | ap++; |
411 | for (++numlevels; i <= ULFS_NIADDR; i++) { |
412 | /* If searching for a meta-data block, quit when found. */ |
413 | if (metalbn == realbn) |
414 | break; |
415 | |
416 | lbc -= fs->um_lognindir; |
417 | off = (bn >> lbc) & (MNINDIR(fs) - 1); |
418 | |
419 | ++numlevels; |
420 | ap->in_lbn = metalbn; |
421 | ap->in_off = off; |
422 | ap->in_exists = 0; |
423 | ++ap; |
424 | |
425 | metalbn -= -1 + ((int64_t)off << lbc); |
426 | } |
427 | if (nump) |
428 | *nump = numlevels; |
429 | return (0); |
430 | } |
431 | |