FreeBSD/Linux Kernel Cross Reference
sys/ufs/ufs/ufs_bmap.c
1 /* $NetBSD: ufs_bmap.c,v 1.28.2.3 2005/04/06 11:38:21 tron Exp $ */
2
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 8.8 (Berkeley) 8/11/95
37 */
38
39 #include <sys/cdefs.h>
40 __KERNEL_RCSID(0, "$NetBSD: ufs_bmap.c,v 1.28.2.3 2005/04/06 11:38:21 tron Exp $");
41
42 #include <sys/param.h>
43 #include <sys/systm.h>
44 #include <sys/buf.h>
45 #include <sys/proc.h>
46 #include <sys/vnode.h>
47 #include <sys/mount.h>
48 #include <sys/resourcevar.h>
49 #include <sys/trace.h>
50
51 #include <miscfs/specfs/specdev.h>
52
53 #include <ufs/ufs/inode.h>
54 #include <ufs/ufs/ufsmount.h>
55 #include <ufs/ufs/ufs_extern.h>
56 #include <ufs/ufs/ufs_bswap.h>
57
58 static boolean_t ufs_issequential __P((const struct ufsmount *,
59 daddr_t, daddr_t));
60
61 static boolean_t
62 ufs_issequential(ump, daddr0, daddr1)
63 const struct ufsmount *ump;
64 daddr_t daddr0;
65 daddr_t daddr1;
66 {
67
68 /* for ufs, blocks in a hole is not 'contiguous'. */
69 if (daddr0 == 0)
70 return FALSE;
71
72 return (daddr0 + ump->um_seqinc == daddr1);
73 }
74
75 /*
76 * Bmap converts a the logical block number of a file to its physical block
77 * number on the disk. The conversion is done by using the logical block
78 * number to index into the array of block pointers described by the dinode.
79 */
80 int
81 ufs_bmap(v)
82 void *v;
83 {
84 struct vop_bmap_args /* {
85 struct vnode *a_vp;
86 daddr_t a_bn;
87 struct vnode **a_vpp;
88 daddr_t *a_bnp;
89 int *a_runp;
90 } */ *ap = v;
91 /*
92 * Check for underlying vnode requests and ensure that logical
93 * to physical mapping is requested.
94 */
95 if (ap->a_vpp != NULL)
96 *ap->a_vpp = VTOI(ap->a_vp)->i_devvp;
97 if (ap->a_bnp == NULL)
98 return (0);
99
100 return (ufs_bmaparray(ap->a_vp, ap->a_bn, ap->a_bnp, NULL, NULL,
101 ap->a_runp, ufs_issequential));
102 }
103
104 /*
105 * Indirect blocks are now on the vnode for the file. They are given negative
106 * logical block numbers. Indirect blocks are addressed by the negative
107 * address of the first data block to which they point. Double indirect blocks
108 * are addressed by one less than the address of the first indirect block to
109 * which they point. Triple indirect blocks are addressed by one less than
110 * the address of the first double indirect block to which they point.
111 *
112 * ufs_bmaparray does the bmap conversion, and if requested returns the
113 * array of logical blocks which must be traversed to get to a block.
114 * Each entry contains the offset into that block that gets you to the
115 * next block and the disk address of the block (if it is assigned).
116 */
117
118 int
119 ufs_bmaparray(vp, bn, bnp, ap, nump, runp, is_sequential)
120 struct vnode *vp;
121 daddr_t bn;
122 daddr_t *bnp;
123 struct indir *ap;
124 int *nump;
125 int *runp;
126 ufs_issequential_callback_t is_sequential;
127 {
128 struct inode *ip;
129 struct buf *bp;
130 struct ufsmount *ump;
131 struct mount *mp;
132 struct indir a[NIADDR + 1], *xap;
133 daddr_t daddr;
134 daddr_t metalbn;
135 int error, maxrun = 0, num;
136
137 ip = VTOI(vp);
138 mp = vp->v_mount;
139 ump = VFSTOUFS(mp);
140 #ifdef DIAGNOSTIC
141 if ((ap != NULL && nump == NULL) || (ap == NULL && nump != NULL))
142 panic("ufs_bmaparray: invalid arguments");
143 #endif
144
145 if (runp) {
146 /*
147 * XXX
148 * If MAXBSIZE is the largest transfer the disks can handle,
149 * we probably want maxrun to be 1 block less so that we
150 * don't create a block larger than the device can handle.
151 */
152 *runp = 0;
153 maxrun = MAXPHYS / mp->mnt_stat.f_iosize - 1;
154 }
155
156 if (bn >= 0 && bn < NDADDR) {
157 if (nump != NULL)
158 *nump = 0;
159 if (ip->i_ump->um_fstype == UFS1)
160 daddr = ufs_rw32(ip->i_ffs1_db[bn],
161 UFS_MPNEEDSWAP(vp->v_mount));
162 else
163 daddr = ufs_rw64(ip->i_ffs2_db[bn],
164 UFS_MPNEEDSWAP(vp->v_mount));
165 *bnp = blkptrtodb(ump, daddr);
166 if (*bnp == 0)
167 *bnp = -1;
168 else if (runp) {
169 if (ip->i_ump->um_fstype == UFS1) {
170 for (++bn; bn < NDADDR && *runp < maxrun &&
171 is_sequential(ump,
172 ufs_rw32(ip->i_ffs1_db[bn - 1],
173 UFS_MPNEEDSWAP(vp->v_mount)),
174 ufs_rw32(ip->i_ffs1_db[bn],
175 UFS_MPNEEDSWAP(vp->v_mount)));
176 ++bn, ++*runp);
177 } else {
178 for (++bn; bn < NDADDR && *runp < maxrun &&
179 is_sequential(ump,
180 ufs_rw64(ip->i_ffs2_db[bn - 1],
181 UFS_MPNEEDSWAP(vp->v_mount)),
182 ufs_rw64(ip->i_ffs2_db[bn],
183 UFS_MPNEEDSWAP(vp->v_mount)));
184 ++bn, ++*runp);
185 }
186 }
187 return (0);
188 }
189
190 xap = ap == NULL ? a : ap;
191 if (!nump)
192 nump = #
193 if ((error = ufs_getlbns(vp, bn, xap, nump)) != 0)
194 return (error);
195
196 num = *nump;
197
198 /* Get disk address out of indirect block array */
199 if (ip->i_ump->um_fstype == UFS1)
200 daddr = ufs_rw32(ip->i_ffs1_ib[xap->in_off],
201 UFS_MPNEEDSWAP(vp->v_mount));
202 else
203 daddr = ufs_rw64(ip->i_ffs2_ib[xap->in_off],
204 UFS_MPNEEDSWAP(vp->v_mount));
205
206 for (bp = NULL, ++xap; --num; ++xap) {
207 /*
208 * Exit the loop if there is no disk address assigned yet and
209 * the indirect block isn't in the cache, or if we were
210 * looking for an indirect block and we've found it.
211 */
212
213 metalbn = xap->in_lbn;
214 if ((daddr == 0 && !incore(vp, metalbn)) || metalbn == bn)
215 break;
216 /*
217 * If we get here, we've either got the block in the cache
218 * or we have a disk address for it, go fetch it.
219 */
220 if (bp)
221 brelse(bp);
222
223 xap->in_exists = 1;
224 bp = getblk(vp, metalbn, mp->mnt_stat.f_iosize, 0, 0);
225 if (bp == NULL) {
226
227 /*
228 * getblk() above returns NULL only iff we are
229 * pagedaemon. See the implementation of getblk
230 * for detail.
231 */
232
233 return (ENOMEM);
234 }
235 if (bp->b_flags & (B_DONE | B_DELWRI)) {
236 trace(TR_BREADHIT, pack(vp, size), metalbn);
237 }
238 #ifdef DIAGNOSTIC
239 else if (!daddr)
240 panic("ufs_bmaparry: indirect block not in cache");
241 #endif
242 else {
243 trace(TR_BREADMISS, pack(vp, size), metalbn);
244 bp->b_blkno = blkptrtodb(ump, daddr);
245 bp->b_flags |= B_READ;
246 BIO_SETPRIO(bp, BPRIO_TIMECRITICAL);
247 VOP_STRATEGY(vp, bp);
248 curproc->p_stats->p_ru.ru_inblock++; /* XXX */
249 if ((error = biowait(bp)) != 0) {
250 brelse(bp);
251 return (error);
252 }
253 }
254 if (ip->i_ump->um_fstype == UFS1) {
255 daddr = ufs_rw32(((int32_t *)bp->b_data)[xap->in_off],
256 UFS_MPNEEDSWAP(mp));
257 if (num == 1 && daddr && runp) {
258 for (bn = xap->in_off + 1;
259 bn < MNINDIR(ump) && *runp < maxrun &&
260 is_sequential(ump,
261 ufs_rw32(((int32_t *)bp->b_data)[bn-1],
262 UFS_MPNEEDSWAP(mp)),
263 ufs_rw32(((int32_t *)bp->b_data)[bn],
264 UFS_MPNEEDSWAP(mp)));
265 ++bn, ++*runp);
266 }
267 } else {
268 daddr = ufs_rw64(((int64_t *)bp->b_data)[xap->in_off],
269 UFS_MPNEEDSWAP(mp));
270 if (num == 1 && daddr && runp) {
271 for (bn = xap->in_off + 1;
272 bn < MNINDIR(ump) && *runp < maxrun &&
273 is_sequential(ump,
274 ufs_rw64(((int64_t *)bp->b_data)[bn-1],
275 UFS_MPNEEDSWAP(mp)),
276 ufs_rw64(((int64_t *)bp->b_data)[bn],
277 UFS_MPNEEDSWAP(mp)));
278 ++bn, ++*runp);
279 }
280 }
281 }
282 if (bp)
283 brelse(bp);
284
285 *bnp = blkptrtodb(ump, daddr);
286 if (*bnp == 0)
287 *bnp = -1;
288 return (0);
289 }
290
291 /*
292 * Create an array of logical block number/offset pairs which represent the
293 * path of indirect blocks required to access a data block. The first "pair"
294 * contains the logical block number of the appropriate single, double or
295 * triple indirect block and the offset into the inode indirect block array.
296 * Note, the logical block number of the inode single/double/triple indirect
297 * block appears twice in the array, once with the offset into the i_ffs1_ib and
298 * once with the offset into the page itself.
299 */
300 int
301 ufs_getlbns(vp, bn, ap, nump)
302 struct vnode *vp;
303 daddr_t bn;
304 struct indir *ap;
305 int *nump;
306 {
307 daddr_t metalbn, realbn;
308 struct ufsmount *ump;
309 int64_t blockcnt;
310 int lbc;
311 int i, numlevels, off;
312
313 ump = VFSTOUFS(vp->v_mount);
314 if (nump)
315 *nump = 0;
316 numlevels = 0;
317 realbn = bn;
318 if (bn < 0)
319 bn = -bn;
320 KASSERT(bn >= NDADDR);
321
322 /*
323 * Determine the number of levels of indirection. After this loop
324 * is done, blockcnt indicates the number of data blocks possible
325 * at the given level of indirection, and NIADDR - i is the number
326 * of levels of indirection needed to locate the requested block.
327 */
328
329 bn -= NDADDR;
330 for (lbc = 0, i = NIADDR;; i--, bn -= blockcnt) {
331 if (i == 0)
332 return (EFBIG);
333
334 lbc += ump->um_lognindir;
335 blockcnt = (int64_t)1 << lbc;
336
337 if (bn < blockcnt)
338 break;
339 }
340
341 /* Calculate the address of the first meta-block. */
342 metalbn = 0; /* XXX: gcc3 */
343 if (realbn >= 0)
344 metalbn = -(realbn - bn + NIADDR - i);
345 else
346 metalbn = -(-realbn - bn + NIADDR - i);
347
348 /*
349 * At each iteration, off is the offset into the bap array which is
350 * an array of disk addresses at the current level of indirection.
351 * The logical block number and the offset in that block are stored
352 * into the argument array.
353 */
354 ap->in_lbn = metalbn;
355 ap->in_off = off = NIADDR - i;
356 ap->in_exists = 0;
357 ap++;
358 for (++numlevels; i <= NIADDR; i++) {
359 /* If searching for a meta-data block, quit when found. */
360 if (metalbn == realbn)
361 break;
362
363 lbc -= ump->um_lognindir;
364 off = (bn >> lbc) & (MNINDIR(ump) - 1);
365
366 ++numlevels;
367 ap->in_lbn = metalbn;
368 ap->in_off = off;
369 ap->in_exists = 0;
370 ++ap;
371
372 metalbn -= -1 + ((int64_t)off << lbc);
373 }
374 if (nump)
375 *nump = numlevels;
376 return (0);
377 }
Cache object: bfc093cec4066f962b8a0bccc5dfe145
|