1/* $NetBSD: ufs_bmap.c,v 1.50 2013/01/22 09:39:18 dholland 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 1.50 2013/01/22 09:39:18 dholland 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#include <sys/fstrans.h>
51
52#include <miscfs/specfs/specdev.h>
53
54#include <ufs/ufs/inode.h>
55#include <ufs/ufs/ufsmount.h>
56#include <ufs/ufs/ufs_extern.h>
57#include <ufs/ufs/ufs_bswap.h>
58
59static bool
60ufs_issequential(const struct ufsmount *ump, daddr_t daddr0, daddr_t daddr1)
61{
62
63 /* for ufs, blocks in a hole is not 'contiguous'. */
64 if (daddr0 == 0)
65 return false;
66
67 return (daddr0 + ump->um_seqinc == daddr1);
68}
69
70/*
71 * Bmap converts the logical block number of a file to its physical block
72 * number on the disk. The conversion is done by using the logical block
73 * number to index into the array of block pointers described by the dinode.
74 */
75int
76ufs_bmap(void *v)
77{
78 struct vop_bmap_args /* {
79 struct vnode *a_vp;
80 daddr_t a_bn;
81 struct vnode **a_vpp;
82 daddr_t *a_bnp;
83 int *a_runp;
84 } */ *ap = v;
85 int error;
86
87 /*
88 * Check for underlying vnode requests and ensure that logical
89 * to physical mapping is requested.
90 */
91 if (ap->a_vpp != NULL)
92 *ap->a_vpp = VTOI(ap->a_vp)->i_devvp;
93 if (ap->a_bnp == NULL)
94 return (0);
95
96 fstrans_start(ap->a_vp->v_mount, FSTRANS_SHARED);
97 error = ufs_bmaparray(ap->a_vp, ap->a_bn, ap->a_bnp, NULL, NULL,
98 ap->a_runp, ufs_issequential);
99 fstrans_done(ap->a_vp->v_mount);
100 return error;
101}
102
103/*
104 * Indirect blocks are now on the vnode for the file. They are given negative
105 * logical block numbers. Indirect blocks are addressed by the negative
106 * address of the first data block to which they point. Double indirect blocks
107 * are addressed by one less than the address of the first indirect block to
108 * which they point. Triple indirect blocks are addressed by one less than
109 * the address of the first double indirect block to which they point.
110 *
111 * ufs_bmaparray does the bmap conversion, and if requested returns the
112 * array of logical blocks which must be traversed to get to a block.
113 * Each entry contains the offset into that block that gets you to the
114 * next block and the disk address of the block (if it is assigned).
115 */
116
117int
118ufs_bmaparray(struct vnode *vp, daddr_t bn, daddr_t *bnp, struct indir *ap,
119 int *nump, int *runp, ufs_issequential_callback_t is_sequential)
120{
121 struct inode *ip;
122 struct buf *bp, *cbp;
123 struct ufsmount *ump;
124 struct mount *mp;
125 struct indir a[UFS_NIADDR + 1], *xap;
126 daddr_t daddr;
127 daddr_t metalbn;
128 int error, maxrun = 0, num;
129
130 ip = VTOI(vp);
131 mp = vp->v_mount;
132 ump = ip->i_ump;
133#ifdef DIAGNOSTIC
134 if ((ap != NULL && nump == NULL) || (ap == NULL && nump != NULL))
135 panic("ufs_bmaparray: invalid arguments");
136#endif
137
138 if (runp) {
139 /*
140 * XXX
141 * If MAXBSIZE is the largest transfer the disks can handle,
142 * we probably want maxrun to be 1 block less so that we
143 * don't create a block larger than the device can handle.
144 */
145 *runp = 0;
146 maxrun = MAXPHYS / mp->mnt_stat.f_iosize - 1;
147 }
148
149 if (bn >= 0 && bn < UFS_NDADDR) {
150 if (nump != NULL)
151 *nump = 0;
152 if (ump->um_fstype == UFS1)
153 daddr = ufs_rw32(ip->i_ffs1_db[bn],
154 UFS_MPNEEDSWAP(ump));
155 else
156 daddr = ufs_rw64(ip->i_ffs2_db[bn],
157 UFS_MPNEEDSWAP(ump));
158 *bnp = blkptrtodb(ump, daddr);
159 /*
160 * Since this is FFS independent code, we are out of
161 * scope for the definitions of BLK_NOCOPY and
162 * BLK_SNAP, but we do know that they will fall in
163 * the range 1..um_seqinc, so we use that test and
164 * return a request for a zeroed out buffer if attempts
165 * are made to read a BLK_NOCOPY or BLK_SNAP block.
166 */
167 if ((ip->i_flags & (SF_SNAPSHOT | SF_SNAPINVAL)) == SF_SNAPSHOT
168 && daddr > 0 &&
169 daddr < ump->um_seqinc) {
170 *bnp = -1;
171 } else if (*bnp == 0) {
172 if ((ip->i_flags & (SF_SNAPSHOT | SF_SNAPINVAL))
173 == SF_SNAPSHOT) {
174 *bnp = blkptrtodb(ump, bn * ump->um_seqinc);
175 } else {
176 *bnp = -1;
177 }
178 } else if (runp) {
179 if (ump->um_fstype == UFS1) {
180 for (++bn; bn < UFS_NDADDR && *runp < maxrun &&
181 is_sequential(ump,
182 ufs_rw32(ip->i_ffs1_db[bn - 1],
183 UFS_MPNEEDSWAP(ump)),
184 ufs_rw32(ip->i_ffs1_db[bn],
185 UFS_MPNEEDSWAP(ump)));
186 ++bn, ++*runp);
187 } else {
188 for (++bn; bn < UFS_NDADDR && *runp < maxrun &&
189 is_sequential(ump,
190 ufs_rw64(ip->i_ffs2_db[bn - 1],
191 UFS_MPNEEDSWAP(ump)),
192 ufs_rw64(ip->i_ffs2_db[bn],
193 UFS_MPNEEDSWAP(ump)));
194 ++bn, ++*runp);
195 }
196 }
197 return (0);
198 }
199
200 xap = ap == NULL ? a : ap;
201 if (!nump)
202 nump = &num;
203 if ((error = ufs_getlbns(vp, bn, xap, nump)) != 0)
204 return (error);
205
206 num = *nump;
207
208 /* Get disk address out of indirect block array */
209 if (ump->um_fstype == UFS1)
210 daddr = ufs_rw32(ip->i_ffs1_ib[xap->in_off],
211 UFS_MPNEEDSWAP(ump));
212 else
213 daddr = ufs_rw64(ip->i_ffs2_ib[xap->in_off],
214 UFS_MPNEEDSWAP(ump));
215
216 for (bp = NULL, ++xap; --num; ++xap) {
217 /*
218 * Exit the loop if there is no disk address assigned yet and
219 * the indirect block isn't in the cache, or if we were
220 * looking for an indirect block and we've found it.
221 */
222
223 metalbn = xap->in_lbn;
224 if (metalbn == bn)
225 break;
226 if (daddr == 0) {
227 mutex_enter(&bufcache_lock);
228 cbp = incore(vp, metalbn);
229 mutex_exit(&bufcache_lock);
230 if (cbp == NULL)
231 break;
232 }
233
234 /*
235 * If we get here, we've either got the block in the cache
236 * or we have a disk address for it, go fetch it.
237 */
238 if (bp)
239 brelse(bp, 0);
240
241 xap->in_exists = 1;
242 bp = getblk(vp, metalbn, mp->mnt_stat.f_iosize, 0, 0);
243 if (bp == NULL) {
244
245 /*
246 * getblk() above returns NULL only iff we are
247 * pagedaemon. See the implementation of getblk
248 * for detail.
249 */
250
251 return (ENOMEM);
252 }
253 if (bp->b_oflags & (BO_DONE | BO_DELWRI)) {
254 trace(TR_BREADHIT, pack(vp, size), metalbn);
255 }
256#ifdef DIAGNOSTIC
257 else if (!daddr)
258 panic("ufs_bmaparray: indirect block not in cache");
259#endif
260 else {
261 trace(TR_BREADMISS, pack(vp, size), metalbn);
262 bp->b_blkno = blkptrtodb(ump, daddr);
263 bp->b_flags |= B_READ;
264 BIO_SETPRIO(bp, BPRIO_TIMECRITICAL);
265 VOP_STRATEGY(vp, bp);
266 curlwp->l_ru.ru_inblock++; /* XXX */
267 if ((error = biowait(bp)) != 0) {
268 brelse(bp, 0);
269 return (error);
270 }
271 }
272 if (ump->um_fstype == UFS1) {
273 daddr = ufs_rw32(((u_int32_t *)bp->b_data)[xap->in_off],
274 UFS_MPNEEDSWAP(ump));
275 if (num == 1 && daddr && runp) {
276 for (bn = xap->in_off + 1;
277 bn < MNINDIR(ump) && *runp < maxrun &&
278 is_sequential(ump,
279 ufs_rw32(((int32_t *)bp->b_data)[bn-1],
280 UFS_MPNEEDSWAP(ump)),
281 ufs_rw32(((int32_t *)bp->b_data)[bn],
282 UFS_MPNEEDSWAP(ump)));
283 ++bn, ++*runp);
284 }
285 } else {
286 daddr = ufs_rw64(((u_int64_t *)bp->b_data)[xap->in_off],
287 UFS_MPNEEDSWAP(ump));
288 if (num == 1 && daddr && runp) {
289 for (bn = xap->in_off + 1;
290 bn < MNINDIR(ump) && *runp < maxrun &&
291 is_sequential(ump,
292 ufs_rw64(((int64_t *)bp->b_data)[bn-1],
293 UFS_MPNEEDSWAP(ump)),
294 ufs_rw64(((int64_t *)bp->b_data)[bn],
295 UFS_MPNEEDSWAP(ump)));
296 ++bn, ++*runp);
297 }
298 }
299 }
300 if (bp)
301 brelse(bp, 0);
302
303 /*
304 * Since this is FFS independent code, we are out of scope for the
305 * definitions of BLK_NOCOPY and BLK_SNAP, but we do know that they
306 * will fall in the range 1..um_seqinc, so we use that test and
307 * return a request for a zeroed out buffer if attempts are made
308 * to read a BLK_NOCOPY or BLK_SNAP block.
309 */
310 if ((ip->i_flags & (SF_SNAPSHOT | SF_SNAPINVAL)) == SF_SNAPSHOT
311 && daddr > 0 && daddr < ump->um_seqinc) {
312 *bnp = -1;
313 return (0);
314 }
315 *bnp = blkptrtodb(ump, daddr);
316 if (*bnp == 0) {
317 if ((ip->i_flags & (SF_SNAPSHOT | SF_SNAPINVAL))
318 == SF_SNAPSHOT) {
319 *bnp = blkptrtodb(ump, bn * ump->um_seqinc);
320 } else {
321 *bnp = -1;
322 }
323 }
324 return (0);
325}
326
327/*
328 * Create an array of logical block number/offset pairs which represent the
329 * path of indirect blocks required to access a data block. The first "pair"
330 * contains the logical block number of the appropriate single, double or
331 * triple indirect block and the offset into the inode indirect block array.
332 * Note, the logical block number of the inode single/double/triple indirect
333 * block appears twice in the array, once with the offset into the i_ffs1_ib and
334 * once with the offset into the page itself.
335 */
336int
337ufs_getlbns(struct vnode *vp, daddr_t bn, struct indir *ap, int *nump)
338{
339 daddr_t metalbn, realbn;
340 struct ufsmount *ump;
341 int64_t blockcnt;
342 int lbc;
343 int i, numlevels, off;
344
345 ump = VFSTOUFS(vp->v_mount);
346 if (nump)
347 *nump = 0;
348 numlevels = 0;
349 realbn = bn;
350 if (bn < 0)
351 bn = -bn;
352 KASSERT(bn >= UFS_NDADDR);
353
354 /*
355 * Determine the number of levels of indirection. After this loop
356 * is done, blockcnt indicates the number of data blocks possible
357 * at the given level of indirection, and UFS_NIADDR - i is the number
358 * of levels of indirection needed to locate the requested block.
359 */
360
361 bn -= UFS_NDADDR;
362 for (lbc = 0, i = UFS_NIADDR;; i--, bn -= blockcnt) {
363 if (i == 0)
364 return (EFBIG);
365
366 lbc += ump->um_lognindir;
367 blockcnt = (int64_t)1 << lbc;
368
369 if (bn < blockcnt)
370 break;
371 }
372
373 /* Calculate the address of the first meta-block. */
374 metalbn = -((realbn >= 0 ? realbn : -realbn) - bn + UFS_NIADDR - i);
375
376 /*
377 * At each iteration, off is the offset into the bap array which is
378 * an array of disk addresses at the current level of indirection.
379 * The logical block number and the offset in that block are stored
380 * into the argument array.
381 */
382 ap->in_lbn = metalbn;
383 ap->in_off = off = UFS_NIADDR - i;
384 ap->in_exists = 0;
385 ap++;
386 for (++numlevels; i <= UFS_NIADDR; i++) {
387 /* If searching for a meta-data block, quit when found. */
388 if (metalbn == realbn)
389 break;
390
391 lbc -= ump->um_lognindir;
392 off = (bn >> lbc) & (MNINDIR(ump) - 1);
393
394 ++numlevels;
395 ap->in_lbn = metalbn;
396 ap->in_off = off;
397 ap->in_exists = 0;
398 ++ap;
399
400 metalbn -= -1 + ((int64_t)off << lbc);
401 }
402 if (nump)
403 *nump = numlevels;
404 return (0);
405}
406