ffs_inode.c revision 1.98 1 /* $NetBSD: ffs_inode.c,v 1.98 2008/07/31 05:38:06 simonb Exp $ */
2
3 /*-
4 * Copyright (c) 2008 The NetBSD Foundation, Inc.
5 * All rights reserved.
6 *
7 * This code is derived from software contributed to The NetBSD Foundation
8 * by Wasabi Systems, 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 *
19 * THIS SOFTWARE IS PROVIDED BY THE NETBSD FOUNDATION, INC. AND CONTRIBUTORS
20 * ``AS IS'' AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED
21 * TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR
22 * PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE FOUNDATION OR CONTRIBUTORS
23 * BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR
24 * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF
25 * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS
26 * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN
27 * CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE)
28 * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE
29 * POSSIBILITY OF SUCH DAMAGE.
30 */
31
32 /*
33 * Copyright (c) 1982, 1986, 1989, 1993
34 * The Regents of the University of California. All rights reserved.
35 *
36 * Redistribution and use in source and binary forms, with or without
37 * modification, are permitted provided that the following conditions
38 * are met:
39 * 1. Redistributions of source code must retain the above copyright
40 * notice, this list of conditions and the following disclaimer.
41 * 2. Redistributions in binary form must reproduce the above copyright
42 * notice, this list of conditions and the following disclaimer in the
43 * documentation and/or other materials provided with the distribution.
44 * 3. Neither the name of the University nor the names of its contributors
45 * may be used to endorse or promote products derived from this software
46 * without specific prior written permission.
47 *
48 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
49 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
50 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
51 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
52 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
53 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
54 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
55 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
56 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
57 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
58 * SUCH DAMAGE.
59 *
60 * @(#)ffs_inode.c 8.13 (Berkeley) 4/21/95
61 */
62
63 #include <sys/cdefs.h>
64 __KERNEL_RCSID(0, "$NetBSD: ffs_inode.c,v 1.98 2008/07/31 05:38:06 simonb Exp $");
65
66 #if defined(_KERNEL_OPT)
67 #include "opt_ffs.h"
68 #include "opt_quota.h"
69 #endif
70
71 #include <sys/param.h>
72 #include <sys/systm.h>
73 #include <sys/buf.h>
74 #include <sys/file.h>
75 #include <sys/fstrans.h>
76 #include <sys/kauth.h>
77 #include <sys/kernel.h>
78 #include <sys/malloc.h>
79 #include <sys/mount.h>
80 #include <sys/proc.h>
81 #include <sys/resourcevar.h>
82 #include <sys/trace.h>
83 #include <sys/vnode.h>
84 #include <sys/wapbl.h>
85
86 #include <ufs/ufs/quota.h>
87 #include <ufs/ufs/inode.h>
88 #include <ufs/ufs/ufsmount.h>
89 #include <ufs/ufs/ufs_extern.h>
90 #include <ufs/ufs/ufs_bswap.h>
91 #include <ufs/ufs/ufs_wapbl.h>
92
93 #include <ufs/ffs/fs.h>
94 #include <ufs/ffs/ffs_extern.h>
95
96 static int ffs_indirtrunc(struct inode *, daddr_t, daddr_t, daddr_t, int,
97 int64_t *);
98
99 /*
100 * Update the access, modified, and inode change times as specified
101 * by the IN_ACCESS, IN_UPDATE, and IN_CHANGE flags respectively.
102 * The IN_MODIFIED flag is used to specify that the inode needs to be
103 * updated but that the times have already been set. The access
104 * and modified times are taken from the second and third parameters;
105 * the inode change time is always taken from the current time. If
106 * UPDATE_WAIT flag is set, or UPDATE_DIROP is set and we are not doing
107 * softupdates, then wait for the disk write of the inode to complete.
108 */
109
110 int
111 ffs_update(struct vnode *vp, const struct timespec *acc,
112 const struct timespec *mod, int updflags)
113 {
114 struct fs *fs;
115 struct buf *bp;
116 struct inode *ip;
117 int error;
118 void *cp;
119 int waitfor, flags;
120
121 if (vp->v_mount->mnt_flag & MNT_RDONLY)
122 return (0);
123 ip = VTOI(vp);
124 FFS_ITIMES(ip, acc, mod, NULL);
125 if (updflags & UPDATE_CLOSE)
126 flags = ip->i_flag & (IN_MODIFIED | IN_ACCESSED);
127 else
128 flags = ip->i_flag & IN_MODIFIED;
129 if (flags == 0)
130 return (0);
131 fs = ip->i_fs;
132
133 if ((flags & IN_MODIFIED) != 0 &&
134 (vp->v_mount->mnt_flag & MNT_ASYNC) == 0) {
135 waitfor = updflags & UPDATE_WAIT;
136 if ((updflags & UPDATE_DIROP) && !DOINGSOFTDEP(vp))
137 waitfor |= UPDATE_WAIT;
138 } else
139 waitfor = 0;
140
141 /*
142 * Ensure that uid and gid are correct. This is a temporary
143 * fix until fsck has been changed to do the update.
144 */
145 if (fs->fs_magic == FS_UFS1_MAGIC && /* XXX */
146 fs->fs_old_inodefmt < FS_44INODEFMT) { /* XXX */
147 ip->i_ffs1_ouid = ip->i_uid; /* XXX */
148 ip->i_ffs1_ogid = ip->i_gid; /* XXX */
149 } /* XXX */
150 error = bread(ip->i_devvp,
151 fsbtodb(fs, ino_to_fsba(fs, ip->i_number)),
152 (int)fs->fs_bsize, NOCRED, B_MODIFY, &bp);
153 if (error) {
154 brelse(bp, 0);
155 return (error);
156 }
157 ip->i_flag &= ~(IN_MODIFIED | IN_ACCESSED);
158 if (DOINGSOFTDEP(vp)) {
159 softdep_update_inodeblock(ip, bp, waitfor);
160 } else if (ip->i_ffs_effnlink != ip->i_nlink)
161 panic("ffs_update: bad link cnt");
162 /* Keep unlinked inode list up to date */
163 KDASSERT(DIP(ip, nlink) == ip->i_nlink);
164 if (ip->i_mode) {
165 if (ip->i_nlink > 0) {
166 UFS_WAPBL_UNREGISTER_INODE(ip->i_ump->um_mountp,
167 ip->i_number, ip->i_mode);
168 } else {
169 UFS_WAPBL_REGISTER_INODE(ip->i_ump->um_mountp,
170 ip->i_number, ip->i_mode);
171 }
172 }
173 if (fs->fs_magic == FS_UFS1_MAGIC) {
174 cp = (char *)bp->b_data +
175 (ino_to_fsbo(fs, ip->i_number) * DINODE1_SIZE);
176 #ifdef FFS_EI
177 if (UFS_FSNEEDSWAP(fs))
178 ffs_dinode1_swap(ip->i_din.ffs1_din,
179 (struct ufs1_dinode *)cp);
180 else
181 #endif
182 memcpy(cp, ip->i_din.ffs1_din, DINODE1_SIZE);
183 } else {
184 cp = (char *)bp->b_data +
185 (ino_to_fsbo(fs, ip->i_number) * DINODE2_SIZE);
186 #ifdef FFS_EI
187 if (UFS_FSNEEDSWAP(fs))
188 ffs_dinode2_swap(ip->i_din.ffs2_din,
189 (struct ufs2_dinode *)cp);
190 else
191 #endif
192 memcpy(cp, ip->i_din.ffs2_din, DINODE2_SIZE);
193 }
194 if (waitfor) {
195 return (bwrite(bp));
196 } else {
197 bdwrite(bp);
198 return (0);
199 }
200 }
201
202 #define SINGLE 0 /* index of single indirect block */
203 #define DOUBLE 1 /* index of double indirect block */
204 #define TRIPLE 2 /* index of triple indirect block */
205 /*
206 * Truncate the inode oip to at most length size, freeing the
207 * disk blocks.
208 */
209 int
210 ffs_truncate(struct vnode *ovp, off_t length, int ioflag, kauth_cred_t cred)
211 {
212 daddr_t lastblock;
213 struct inode *oip = VTOI(ovp);
214 daddr_t bn, lastiblock[NIADDR], indir_lbn[NIADDR];
215 daddr_t blks[NDADDR + NIADDR];
216 struct fs *fs;
217 int offset, pgoffset, level;
218 int64_t count, blocksreleased = 0;
219 int i, aflag, nblocks;
220 int error, allerror = 0;
221 off_t osize;
222 int sync;
223 struct ufsmount *ump = oip->i_ump;
224
225 if (ovp->v_type == VCHR || ovp->v_type == VBLK ||
226 ovp->v_type == VFIFO || ovp->v_type == VSOCK) {
227 KASSERT(oip->i_size == 0);
228 return 0;
229 }
230
231 if (length < 0)
232 return (EINVAL);
233
234 if (ovp->v_type == VLNK &&
235 (oip->i_size < ump->um_maxsymlinklen ||
236 (ump->um_maxsymlinklen == 0 && DIP(oip, blocks) == 0))) {
237 KDASSERT(length == 0);
238 memset(SHORTLINK(oip), 0, (size_t)oip->i_size);
239 oip->i_size = 0;
240 DIP_ASSIGN(oip, size, 0);
241 oip->i_flag |= IN_CHANGE | IN_UPDATE;
242 return (ffs_update(ovp, NULL, NULL, 0));
243 }
244 if (oip->i_size == length) {
245 oip->i_flag |= IN_CHANGE | IN_UPDATE;
246 return (ffs_update(ovp, NULL, NULL, 0));
247 }
248 fs = oip->i_fs;
249 if (length > ump->um_maxfilesize)
250 return (EFBIG);
251
252 if ((oip->i_flags & SF_SNAPSHOT) != 0)
253 ffs_snapremove(ovp);
254
255 osize = oip->i_size;
256 aflag = ioflag & IO_SYNC ? B_SYNC : 0;
257
258 /*
259 * Lengthen the size of the file. We must ensure that the
260 * last byte of the file is allocated. Since the smallest
261 * value of osize is 0, length will be at least 1.
262 */
263
264 if (osize < length) {
265 if (lblkno(fs, osize) < NDADDR &&
266 lblkno(fs, osize) != lblkno(fs, length) &&
267 blkroundup(fs, osize) != osize) {
268 off_t eob;
269
270 eob = blkroundup(fs, osize);
271 uvm_vnp_setwritesize(ovp, eob);
272 error = ufs_balloc_range(ovp, osize, eob - osize,
273 cred, aflag);
274 if (error)
275 return error;
276 if (ioflag & IO_SYNC) {
277 mutex_enter(&ovp->v_interlock);
278 VOP_PUTPAGES(ovp,
279 trunc_page(osize & fs->fs_bmask),
280 round_page(eob), PGO_CLEANIT | PGO_SYNCIO);
281 }
282 }
283 uvm_vnp_setwritesize(ovp, length);
284 error = ufs_balloc_range(ovp, length - 1, 1, cred, aflag);
285 if (error) {
286 (void) ffs_truncate(ovp, osize, ioflag & IO_SYNC, cred);
287 return (error);
288 }
289 uvm_vnp_setsize(ovp, length);
290 oip->i_flag |= IN_CHANGE | IN_UPDATE;
291 KASSERT(ovp->v_size == oip->i_size);
292 return (ffs_update(ovp, NULL, NULL, 0));
293 }
294
295 /*
296 * When truncating a regular file down to a non-block-aligned size,
297 * we must zero the part of last block which is past the new EOF.
298 * We must synchronously flush the zeroed pages to disk
299 * since the new pages will be invalidated as soon as we
300 * inform the VM system of the new, smaller size.
301 * We must do this before acquiring the GLOCK, since fetching
302 * the pages will acquire the GLOCK internally.
303 * So there is a window where another thread could see a whole
304 * zeroed page past EOF, but that's life.
305 */
306
307 offset = blkoff(fs, length);
308 pgoffset = length & PAGE_MASK;
309 if (ovp->v_type == VREG && (pgoffset != 0 || offset != 0) &&
310 osize > length) {
311 daddr_t lbn;
312 voff_t eoz;
313 int size;
314
315 if (offset != 0) {
316 error = ufs_balloc_range(ovp, length - 1, 1, cred,
317 aflag);
318 if (error)
319 return error;
320 }
321 lbn = lblkno(fs, length);
322 size = blksize(fs, oip, lbn);
323 eoz = MIN(MAX(lblktosize(fs, lbn) + size, round_page(pgoffset)),
324 osize);
325 uvm_vnp_zerorange(ovp, length, eoz - length);
326 if (round_page(eoz) > round_page(length)) {
327 mutex_enter(&ovp->v_interlock);
328 error = VOP_PUTPAGES(ovp, round_page(length),
329 round_page(eoz),
330 PGO_CLEANIT | PGO_DEACTIVATE |
331 ((ioflag & IO_SYNC) ? PGO_SYNCIO : 0));
332 if (error)
333 return error;
334 }
335 }
336
337 genfs_node_wrlock(ovp);
338
339 if (DOINGSOFTDEP(ovp)) {
340 if (length > 0) {
341 /*
342 * If a file is only partially truncated, then
343 * we have to clean up the data structures
344 * describing the allocation past the truncation
345 * point. Finding and deallocating those structures
346 * is a lot of work. Since partial truncation occurs
347 * rarely, we solve the problem by syncing the file
348 * so that it will have no data structures left.
349 */
350 if ((error = VOP_FSYNC(ovp, cred, FSYNC_WAIT,
351 0, 0)) != 0) {
352 genfs_node_unlock(ovp);
353 return (error);
354 }
355 mutex_enter(&ump->um_lock);
356 if (oip->i_flag & IN_SPACECOUNTED)
357 fs->fs_pendingblocks -= DIP(oip, blocks);
358 mutex_exit(&ump->um_lock);
359 } else {
360 uvm_vnp_setsize(ovp, length);
361 #ifdef QUOTA
362 (void) chkdq(oip, -DIP(oip, blocks), NOCRED, 0);
363 #endif
364 softdep_setup_freeblocks(oip, length, 0);
365 (void) vinvalbuf(ovp, 0, cred, curlwp, 0, 0);
366 genfs_node_unlock(ovp);
367 oip->i_flag |= IN_CHANGE | IN_UPDATE;
368 return (ffs_update(ovp, NULL, NULL, 0));
369 }
370 }
371 oip->i_size = length;
372 DIP_ASSIGN(oip, size, length);
373 uvm_vnp_setsize(ovp, length);
374 /*
375 * Calculate index into inode's block list of
376 * last direct and indirect blocks (if any)
377 * which we want to keep. Lastblock is -1 when
378 * the file is truncated to 0.
379 */
380 lastblock = lblkno(fs, length + fs->fs_bsize - 1) - 1;
381 lastiblock[SINGLE] = lastblock - NDADDR;
382 lastiblock[DOUBLE] = lastiblock[SINGLE] - NINDIR(fs);
383 lastiblock[TRIPLE] = lastiblock[DOUBLE] - NINDIR(fs) * NINDIR(fs);
384 nblocks = btodb(fs->fs_bsize);
385 /*
386 * Update file and block pointers on disk before we start freeing
387 * blocks. If we crash before free'ing blocks below, the blocks
388 * will be returned to the free list. lastiblock values are also
389 * normalized to -1 for calls to ffs_indirtrunc below.
390 */
391 sync = 0;
392 for (level = TRIPLE; level >= SINGLE; level--) {
393 blks[NDADDR + level] = DIP(oip, ib[level]);
394 if (lastiblock[level] < 0 && blks[NDADDR + level] != 0) {
395 sync = 1;
396 DIP_ASSIGN(oip, ib[level], 0);
397 lastiblock[level] = -1;
398 }
399 }
400 for (i = 0; i < NDADDR; i++) {
401 blks[i] = DIP(oip, db[i]);
402 if (i > lastblock && blks[i] != 0) {
403 sync = 1;
404 DIP_ASSIGN(oip, db[i], 0);
405 }
406 }
407 oip->i_flag |= IN_CHANGE | IN_UPDATE;
408 if (sync) {
409 error = ffs_update(ovp, NULL, NULL, UPDATE_WAIT);
410 if (error && !allerror)
411 allerror = error;
412 }
413
414 /*
415 * Having written the new inode to disk, save its new configuration
416 * and put back the old block pointers long enough to process them.
417 * Note that we save the new block configuration so we can check it
418 * when we are done.
419 */
420 for (i = 0; i < NDADDR; i++) {
421 bn = DIP(oip, db[i]);
422 DIP_ASSIGN(oip, db[i], blks[i]);
423 blks[i] = bn;
424 }
425 for (i = 0; i < NIADDR; i++) {
426 bn = DIP(oip, ib[i]);
427 DIP_ASSIGN(oip, ib[i], blks[NDADDR + i]);
428 blks[NDADDR + i] = bn;
429 }
430
431 oip->i_size = osize;
432 DIP_ASSIGN(oip, size, osize);
433 error = vtruncbuf(ovp, lastblock + 1, 0, 0);
434 if (error && !allerror)
435 allerror = error;
436
437 /*
438 * Indirect blocks first.
439 */
440 indir_lbn[SINGLE] = -NDADDR;
441 indir_lbn[DOUBLE] = indir_lbn[SINGLE] - NINDIR(fs) - 1;
442 indir_lbn[TRIPLE] = indir_lbn[DOUBLE] - NINDIR(fs) * NINDIR(fs) - 1;
443 for (level = TRIPLE; level >= SINGLE; level--) {
444 if (oip->i_ump->um_fstype == UFS1)
445 bn = ufs_rw32(oip->i_ffs1_ib[level],UFS_FSNEEDSWAP(fs));
446 else
447 bn = ufs_rw64(oip->i_ffs2_ib[level],UFS_FSNEEDSWAP(fs));
448 if (bn != 0) {
449 error = ffs_indirtrunc(oip, indir_lbn[level],
450 fsbtodb(fs, bn), lastiblock[level], level, &count);
451 if (error)
452 allerror = error;
453 blocksreleased += count;
454 if (lastiblock[level] < 0) {
455 DIP_ASSIGN(oip, ib[level], 0);
456 if (oip->i_ump->um_mountp->mnt_wapbl) {
457 UFS_WAPBL_REGISTER_DEALLOCATION(
458 oip->i_ump->um_mountp,
459 fsbtodb(fs, bn), fs->fs_bsize);
460 } else
461 ffs_blkfree(fs, oip->i_devvp, bn,
462 fs->fs_bsize, oip->i_number);
463 blocksreleased += nblocks;
464 }
465 }
466 if (lastiblock[level] >= 0)
467 goto done;
468 }
469
470 /*
471 * All whole direct blocks or frags.
472 */
473 for (i = NDADDR - 1; i > lastblock; i--) {
474 long bsize;
475
476 if (oip->i_ump->um_fstype == UFS1)
477 bn = ufs_rw32(oip->i_ffs1_db[i], UFS_FSNEEDSWAP(fs));
478 else
479 bn = ufs_rw64(oip->i_ffs2_db[i], UFS_FSNEEDSWAP(fs));
480 if (bn == 0)
481 continue;
482 DIP_ASSIGN(oip, db[i], 0);
483 bsize = blksize(fs, oip, i);
484 if ((oip->i_ump->um_mountp->mnt_wapbl) &&
485 (ovp->v_type != VREG)) {
486 UFS_WAPBL_REGISTER_DEALLOCATION(oip->i_ump->um_mountp,
487 fsbtodb(fs, bn), bsize);
488 } else
489 ffs_blkfree(fs, oip->i_devvp, bn, bsize, oip->i_number);
490 blocksreleased += btodb(bsize);
491 }
492 if (lastblock < 0)
493 goto done;
494
495 /*
496 * Finally, look for a change in size of the
497 * last direct block; release any frags.
498 */
499 if (oip->i_ump->um_fstype == UFS1)
500 bn = ufs_rw32(oip->i_ffs1_db[lastblock], UFS_FSNEEDSWAP(fs));
501 else
502 bn = ufs_rw64(oip->i_ffs2_db[lastblock], UFS_FSNEEDSWAP(fs));
503 if (bn != 0) {
504 long oldspace, newspace;
505
506 /*
507 * Calculate amount of space we're giving
508 * back as old block size minus new block size.
509 */
510 oldspace = blksize(fs, oip, lastblock);
511 oip->i_size = length;
512 DIP_ASSIGN(oip, size, length);
513 newspace = blksize(fs, oip, lastblock);
514 if (newspace == 0)
515 panic("itrunc: newspace");
516 if (oldspace - newspace > 0) {
517 /*
518 * Block number of space to be free'd is
519 * the old block # plus the number of frags
520 * required for the storage we're keeping.
521 */
522 bn += numfrags(fs, newspace);
523 if ((oip->i_ump->um_mountp->mnt_wapbl) &&
524 (ovp->v_type != VREG)) {
525 UFS_WAPBL_REGISTER_DEALLOCATION(
526 oip->i_ump->um_mountp, fsbtodb(fs, bn),
527 oldspace - newspace);
528 } else
529 ffs_blkfree(fs, oip->i_devvp, bn,
530 oldspace - newspace, oip->i_number);
531 blocksreleased += btodb(oldspace - newspace);
532 }
533 }
534
535 done:
536 #ifdef DIAGNOSTIC
537 for (level = SINGLE; level <= TRIPLE; level++)
538 if (blks[NDADDR + level] != DIP(oip, ib[level]))
539 panic("itrunc1");
540 for (i = 0; i < NDADDR; i++)
541 if (blks[i] != DIP(oip, db[i]))
542 panic("itrunc2");
543 if (length == 0 &&
544 (!LIST_EMPTY(&ovp->v_cleanblkhd) || !LIST_EMPTY(&ovp->v_dirtyblkhd)))
545 panic("itrunc3");
546 #endif /* DIAGNOSTIC */
547 /*
548 * Put back the real size.
549 */
550 oip->i_size = length;
551 DIP_ASSIGN(oip, size, length);
552 DIP_ADD(oip, blocks, -blocksreleased);
553 genfs_node_unlock(ovp);
554 oip->i_flag |= IN_CHANGE;
555 UFS_WAPBL_UPDATE(ovp, NULL, NULL, 0);
556 #ifdef QUOTA
557 (void) chkdq(oip, -blocksreleased, NOCRED, 0);
558 #endif
559 KASSERT(ovp->v_type != VREG || ovp->v_size == oip->i_size);
560 return (allerror);
561 }
562
563 /*
564 * Release blocks associated with the inode ip and stored in the indirect
565 * block bn. Blocks are free'd in LIFO order up to (but not including)
566 * lastbn. If level is greater than SINGLE, the block is an indirect block
567 * and recursive calls to indirtrunc must be used to cleanse other indirect
568 * blocks.
569 *
570 * NB: triple indirect blocks are untested.
571 */
572 static int
573 ffs_indirtrunc(struct inode *ip, daddr_t lbn, daddr_t dbn, daddr_t lastbn,
574 int level, int64_t *countp)
575 {
576 int i;
577 struct buf *bp;
578 struct fs *fs = ip->i_fs;
579 int32_t *bap1 = NULL;
580 int64_t *bap2 = NULL;
581 struct vnode *vp;
582 daddr_t nb, nlbn, last;
583 char *copy = NULL;
584 int64_t blkcount, factor, blocksreleased = 0;
585 int nblocks;
586 int error = 0, allerror = 0;
587 #ifdef FFS_EI
588 const int needswap = UFS_FSNEEDSWAP(fs);
589 #endif
590 #define RBAP(ip, i) (((ip)->i_ump->um_fstype == UFS1) ? \
591 ufs_rw32(bap1[i], needswap) : ufs_rw64(bap2[i], needswap))
592 #define BAP_ASSIGN(ip, i, value) \
593 do { \
594 if ((ip)->i_ump->um_fstype == UFS1) \
595 bap1[i] = (value); \
596 else \
597 bap2[i] = (value); \
598 } while(0)
599
600 /*
601 * Calculate index in current block of last
602 * block to be kept. -1 indicates the entire
603 * block so we need not calculate the index.
604 */
605 factor = 1;
606 for (i = SINGLE; i < level; i++)
607 factor *= NINDIR(fs);
608 last = lastbn;
609 if (lastbn > 0)
610 last /= factor;
611 nblocks = btodb(fs->fs_bsize);
612 /*
613 * Get buffer of block pointers, zero those entries corresponding
614 * to blocks to be free'd, and update on disk copy first. Since
615 * double(triple) indirect before single(double) indirect, calls
616 * to bmap on these blocks will fail. However, we already have
617 * the on disk address, so we have to set the b_blkno field
618 * explicitly instead of letting bread do everything for us.
619 */
620 vp = ITOV(ip);
621 error = ffs_getblk(vp, lbn, FFS_NOBLK, fs->fs_bsize, false, &bp);
622 if (error) {
623 *countp = 0;
624 return error;
625 }
626 if (bp->b_oflags & (BO_DONE | BO_DELWRI)) {
627 /* Braces must be here in case trace evaluates to nothing. */
628 trace(TR_BREADHIT, pack(vp, fs->fs_bsize), lbn);
629 } else {
630 trace(TR_BREADMISS, pack(vp, fs->fs_bsize), lbn);
631 curlwp->l_ru.ru_inblock++; /* pay for read */
632 bp->b_flags |= B_READ;
633 bp->b_flags &= ~B_COWDONE; /* we change blkno below */
634 if (bp->b_bcount > bp->b_bufsize)
635 panic("ffs_indirtrunc: bad buffer size");
636 bp->b_blkno = dbn;
637 BIO_SETPRIO(bp, BPRIO_TIMECRITICAL);
638 VOP_STRATEGY(vp, bp);
639 error = biowait(bp);
640 if (error == 0)
641 error = fscow_run(bp, true);
642 }
643 if (error) {
644 brelse(bp, 0);
645 *countp = 0;
646 return (error);
647 }
648
649 if (ip->i_ump->um_fstype == UFS1)
650 bap1 = (int32_t *)bp->b_data;
651 else
652 bap2 = (int64_t *)bp->b_data;
653 if (lastbn >= 0) {
654 copy = malloc(fs->fs_bsize, M_TEMP, M_WAITOK);
655 memcpy((void *)copy, bp->b_data, (u_int)fs->fs_bsize);
656 for (i = last + 1; i < NINDIR(fs); i++)
657 BAP_ASSIGN(ip, i, 0);
658 error = bwrite(bp);
659 if (error)
660 allerror = error;
661 if (ip->i_ump->um_fstype == UFS1)
662 bap1 = (int32_t *)copy;
663 else
664 bap2 = (int64_t *)copy;
665 }
666
667 /*
668 * Recursively free totally unused blocks.
669 */
670 for (i = NINDIR(fs) - 1, nlbn = lbn + 1 - i * factor; i > last;
671 i--, nlbn += factor) {
672 nb = RBAP(ip, i);
673 if (nb == 0)
674 continue;
675 if (level > SINGLE) {
676 error = ffs_indirtrunc(ip, nlbn, fsbtodb(fs, nb),
677 (daddr_t)-1, level - 1,
678 &blkcount);
679 if (error)
680 allerror = error;
681 blocksreleased += blkcount;
682 }
683 if ((ip->i_ump->um_mountp->mnt_wapbl) &&
684 ((level > SINGLE) || (ITOV(ip)->v_type != VREG))) {
685 UFS_WAPBL_REGISTER_DEALLOCATION(ip->i_ump->um_mountp,
686 fsbtodb(fs, nb), fs->fs_bsize);
687 } else
688 ffs_blkfree(fs, ip->i_devvp, nb, fs->fs_bsize,
689 ip->i_number);
690 blocksreleased += nblocks;
691 }
692
693 /*
694 * Recursively free last partial block.
695 */
696 if (level > SINGLE && lastbn >= 0) {
697 last = lastbn % factor;
698 nb = RBAP(ip, i);
699 if (nb != 0) {
700 error = ffs_indirtrunc(ip, nlbn, fsbtodb(fs, nb),
701 last, level - 1, &blkcount);
702 if (error)
703 allerror = error;
704 blocksreleased += blkcount;
705 }
706 }
707
708 if (copy != NULL) {
709 FREE(copy, M_TEMP);
710 } else {
711 brelse(bp, BC_INVAL);
712 }
713
714 *countp = blocksreleased;
715 return (allerror);
716 }
717
718 void
719 ffs_itimes(struct inode *ip, const struct timespec *acc,
720 const struct timespec *mod, const struct timespec *cre)
721 {
722 struct timespec now;
723
724 if (!(ip->i_flag & (IN_ACCESS | IN_CHANGE | IN_UPDATE | IN_MODIFY))) {
725 return;
726 }
727
728 vfs_timestamp(&now);
729 if (ip->i_flag & IN_ACCESS) {
730 if (acc == NULL)
731 acc = &now;
732 DIP_ASSIGN(ip, atime, acc->tv_sec);
733 DIP_ASSIGN(ip, atimensec, acc->tv_nsec);
734 }
735 if (ip->i_flag & (IN_UPDATE | IN_MODIFY)) {
736 if ((ip->i_flags & SF_SNAPSHOT) == 0) {
737 if (mod == NULL)
738 mod = &now;
739 DIP_ASSIGN(ip, mtime, mod->tv_sec);
740 DIP_ASSIGN(ip, mtimensec, mod->tv_nsec);
741 }
742 ip->i_modrev++;
743 }
744 if (ip->i_flag & (IN_CHANGE | IN_MODIFY)) {
745 if (cre == NULL)
746 cre = &now;
747 DIP_ASSIGN(ip, ctime, cre->tv_sec);
748 DIP_ASSIGN(ip, ctimensec, cre->tv_nsec);
749 }
750 if (ip->i_flag & (IN_ACCESS | IN_MODIFY))
751 ip->i_flag |= IN_ACCESSED;
752 if (ip->i_flag & (IN_UPDATE | IN_CHANGE))
753 ip->i_flag |= IN_MODIFIED;
754 ip->i_flag &= ~(IN_ACCESS | IN_CHANGE | IN_UPDATE | IN_MODIFY);
755 }
756