1/* $NetBSD: ffs_inode.c,v 1.123 2016/11/11 10:50:16 hannken 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 1.123 2016/11/11 10:50:16 hannken 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/kmem.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
96static 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 then wait for the
107 * disk write of the inode to complete.
108 */
109
110int
111ffs_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) != 0)
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 FFS_FSBTODB(fs, ino_to_fsba(fs, ip->i_number)),
152 (int)fs->fs_bsize, B_MODIFY, &bp);
153 if (error) {
154 return (error);
155 }
156 ip->i_flag &= ~(IN_MODIFIED | IN_ACCESSED);
157 /* Keep unlinked inode list up to date */
158 KDASSERTMSG(DIP(ip, nlink) == ip->i_nlink,
159 "DIP(ip, nlink) [%d] == ip->i_nlink [%d]",
160 DIP(ip, nlink), ip->i_nlink);
161 if (ip->i_mode) {
162 if (ip->i_nlink > 0) {
163 UFS_WAPBL_UNREGISTER_INODE(ip->i_ump->um_mountp,
164 ip->i_number, ip->i_mode);
165 } else {
166 UFS_WAPBL_REGISTER_INODE(ip->i_ump->um_mountp,
167 ip->i_number, ip->i_mode);
168 }
169 }
170 if (fs->fs_magic == FS_UFS1_MAGIC) {
171 cp = (char *)bp->b_data +
172 (ino_to_fsbo(fs, ip->i_number) * DINODE1_SIZE);
173#ifdef FFS_EI
174 if (UFS_FSNEEDSWAP(fs))
175 ffs_dinode1_swap(ip->i_din.ffs1_din,
176 (struct ufs1_dinode *)cp);
177 else
178#endif
179 memcpy(cp, ip->i_din.ffs1_din, DINODE1_SIZE);
180 } else {
181 cp = (char *)bp->b_data +
182 (ino_to_fsbo(fs, ip->i_number) * DINODE2_SIZE);
183#ifdef FFS_EI
184 if (UFS_FSNEEDSWAP(fs))
185 ffs_dinode2_swap(ip->i_din.ffs2_din,
186 (struct ufs2_dinode *)cp);
187 else
188#endif
189 memcpy(cp, ip->i_din.ffs2_din, DINODE2_SIZE);
190 }
191 if (waitfor) {
192 return (bwrite(bp));
193 } else {
194 bdwrite(bp);
195 return (0);
196 }
197}
198
199#define SINGLE 0 /* index of single indirect block */
200#define DOUBLE 1 /* index of double indirect block */
201#define TRIPLE 2 /* index of triple indirect block */
202/*
203 * Truncate the inode oip to at most length size, freeing the
204 * disk blocks.
205 */
206int
207ffs_truncate(struct vnode *ovp, off_t length, int ioflag, kauth_cred_t cred)
208{
209 daddr_t lastblock;
210 struct inode *oip = VTOI(ovp);
211 daddr_t bn, lastiblock[UFS_NIADDR], indir_lbn[UFS_NIADDR];
212 daddr_t blks[UFS_NDADDR + UFS_NIADDR];
213 struct fs *fs;
214 int offset, pgoffset, level;
215 int64_t blocksreleased = 0;
216 int i, aflag, nblocks;
217 int error, allerror = 0;
218 off_t osize;
219 int sync;
220 struct ufsmount *ump = oip->i_ump;
221 void *dcookie;
222
223 UFS_WAPBL_JLOCK_ASSERT(ip->i_ump->um_mountp);
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 /* still do a uvm_vnp_setsize() as writesize may be larger */
246 uvm_vnp_setsize(ovp, length);
247 oip->i_flag |= IN_CHANGE | IN_UPDATE;
248 return (ffs_update(ovp, NULL, NULL, 0));
249 }
250 fs = oip->i_fs;
251 if (length > ump->um_maxfilesize)
252 return (EFBIG);
253
254 if ((oip->i_flags & SF_SNAPSHOT) != 0)
255 ffs_snapremove(ovp);
256
257 osize = oip->i_size;
258 aflag = ioflag & IO_SYNC ? B_SYNC : 0;
259
260 /*
261 * Lengthen the size of the file. We must ensure that the
262 * last byte of the file is allocated. Since the smallest
263 * value of osize is 0, length will be at least 1.
264 */
265
266 if (osize < length) {
267 if (ffs_lblkno(fs, osize) < UFS_NDADDR &&
268 ffs_lblkno(fs, osize) != ffs_lblkno(fs, length) &&
269 ffs_blkroundup(fs, osize) != osize) {
270 off_t eob;
271
272 eob = ffs_blkroundup(fs, osize);
273 uvm_vnp_setwritesize(ovp, eob);
274 error = ufs_balloc_range(ovp, osize, eob - osize,
275 cred, aflag);
276 if (error) {
277 (void) ffs_truncate(ovp, osize,
278 ioflag & IO_SYNC, cred);
279 return error;
280 }
281 if (ioflag & IO_SYNC) {
282 mutex_enter(ovp->v_interlock);
283 VOP_PUTPAGES(ovp,
284 trunc_page(osize & fs->fs_bmask),
285 round_page(eob), PGO_CLEANIT | PGO_SYNCIO |
286 PGO_JOURNALLOCKED);
287 }
288 }
289 uvm_vnp_setwritesize(ovp, length);
290 error = ufs_balloc_range(ovp, length - 1, 1, cred, aflag);
291 if (error) {
292 (void) ffs_truncate(ovp, osize, ioflag & IO_SYNC, cred);
293 return (error);
294 }
295 uvm_vnp_setsize(ovp, length);
296 oip->i_flag |= IN_CHANGE | IN_UPDATE;
297 KASSERT(ovp->v_size == oip->i_size);
298 return (ffs_update(ovp, NULL, NULL, 0));
299 }
300
301 /*
302 * When truncating a regular file down to a non-block-aligned size,
303 * we must zero the part of last block which is past the new EOF.
304 * We must synchronously flush the zeroed pages to disk
305 * since the new pages will be invalidated as soon as we
306 * inform the VM system of the new, smaller size.
307 * We must do this before acquiring the GLOCK, since fetching
308 * the pages will acquire the GLOCK internally.
309 * So there is a window where another thread could see a whole
310 * zeroed page past EOF, but that's life.
311 */
312
313 offset = ffs_blkoff(fs, length);
314 pgoffset = length & PAGE_MASK;
315 if (ovp->v_type == VREG && (pgoffset != 0 || offset != 0) &&
316 osize > length) {
317 daddr_t lbn;
318 voff_t eoz;
319 int size;
320
321 if (offset != 0) {
322 error = ufs_balloc_range(ovp, length - 1, 1, cred,
323 aflag);
324 if (error)
325 return error;
326 }
327 lbn = ffs_lblkno(fs, length);
328 size = ffs_blksize(fs, oip, lbn);
329 eoz = MIN(MAX(ffs_lblktosize(fs, lbn) + size, round_page(pgoffset)),
330 osize);
331 ubc_zerorange(&ovp->v_uobj, length, eoz - length,
332 UBC_UNMAP_FLAG(ovp));
333 if (round_page(eoz) > round_page(length)) {
334 mutex_enter(ovp->v_interlock);
335 error = VOP_PUTPAGES(ovp, round_page(length),
336 round_page(eoz),
337 PGO_CLEANIT | PGO_DEACTIVATE | PGO_JOURNALLOCKED |
338 ((ioflag & IO_SYNC) ? PGO_SYNCIO : 0));
339 if (error)
340 return error;
341 }
342 }
343
344 genfs_node_wrlock(ovp);
345 oip->i_size = length;
346 DIP_ASSIGN(oip, size, length);
347 uvm_vnp_setsize(ovp, length);
348 /*
349 * Calculate index into inode's block list of
350 * last direct and indirect blocks (if any)
351 * which we want to keep. Lastblock is -1 when
352 * the file is truncated to 0.
353 */
354 lastblock = ffs_lblkno(fs, length + fs->fs_bsize - 1) - 1;
355 lastiblock[SINGLE] = lastblock - UFS_NDADDR;
356 lastiblock[DOUBLE] = lastiblock[SINGLE] - FFS_NINDIR(fs);
357 lastiblock[TRIPLE] = lastiblock[DOUBLE] - FFS_NINDIR(fs) * FFS_NINDIR(fs);
358 nblocks = btodb(fs->fs_bsize);
359 /*
360 * Update file and block pointers on disk before we start freeing
361 * blocks. If we crash before free'ing blocks below, the blocks
362 * will be returned to the free list. lastiblock values are also
363 * normalized to -1 for calls to ffs_indirtrunc below.
364 */
365 sync = 0;
366 for (level = TRIPLE; level >= SINGLE; level--) {
367 blks[UFS_NDADDR + level] = DIP(oip, ib[level]);
368 if (lastiblock[level] < 0 && blks[UFS_NDADDR + level] != 0) {
369 sync = 1;
370 DIP_ASSIGN(oip, ib[level], 0);
371 lastiblock[level] = -1;
372 }
373 }
374 for (i = 0; i < UFS_NDADDR; i++) {
375 blks[i] = DIP(oip, db[i]);
376 if (i > lastblock && blks[i] != 0) {
377 sync = 1;
378 DIP_ASSIGN(oip, db[i], 0);
379 }
380 }
381 oip->i_flag |= IN_CHANGE | IN_UPDATE;
382 if (sync) {
383 error = ffs_update(ovp, NULL, NULL, UPDATE_WAIT);
384 if (error && !allerror)
385 allerror = error;
386 }
387
388 /*
389 * Having written the new inode to disk, save its new configuration
390 * and put back the old block pointers long enough to process them.
391 * Note that we save the new block configuration so we can check it
392 * when we are done.
393 */
394 for (i = 0; i < UFS_NDADDR; i++) {
395 bn = DIP(oip, db[i]);
396 DIP_ASSIGN(oip, db[i], blks[i]);
397 blks[i] = bn;
398 }
399 for (i = 0; i < UFS_NIADDR; i++) {
400 bn = DIP(oip, ib[i]);
401 DIP_ASSIGN(oip, ib[i], blks[UFS_NDADDR + i]);
402 blks[UFS_NDADDR + i] = bn;
403 }
404
405 oip->i_size = osize;
406 DIP_ASSIGN(oip, size, osize);
407 error = vtruncbuf(ovp, lastblock + 1, 0, 0);
408 if (error && !allerror)
409 allerror = error;
410
411 /*
412 * Indirect blocks first.
413 */
414 indir_lbn[SINGLE] = -UFS_NDADDR;
415 indir_lbn[DOUBLE] = indir_lbn[SINGLE] - FFS_NINDIR(fs) - 1;
416 indir_lbn[TRIPLE] = indir_lbn[DOUBLE] - FFS_NINDIR(fs) * FFS_NINDIR(fs) - 1;
417 for (level = TRIPLE; level >= SINGLE; level--) {
418 if (oip->i_ump->um_fstype == UFS1)
419 bn = ufs_rw32(oip->i_ffs1_ib[level],UFS_FSNEEDSWAP(fs));
420 else
421 bn = ufs_rw64(oip->i_ffs2_ib[level],UFS_FSNEEDSWAP(fs));
422 if (bn != 0) {
423 if (lastiblock[level] < 0 &&
424 oip->i_ump->um_mountp->mnt_wapbl) {
425 error = UFS_WAPBL_REGISTER_DEALLOCATION(
426 oip->i_ump->um_mountp,
427 FFS_FSBTODB(fs, bn), fs->fs_bsize,
428 &dcookie);
429 if (error)
430 goto out;
431 } else {
432 dcookie = NULL;
433 }
434
435 error = ffs_indirtrunc(oip, indir_lbn[level],
436 FFS_FSBTODB(fs, bn), lastiblock[level], level,
437 &blocksreleased);
438 if (error) {
439 if (dcookie) {
440 UFS_WAPBL_UNREGISTER_DEALLOCATION(
441 oip->i_ump->um_mountp, dcookie);
442 }
443 goto out;
444 }
445
446 if (lastiblock[level] < 0) {
447 if (!dcookie)
448 ffs_blkfree(fs, oip->i_devvp, bn,
449 fs->fs_bsize, oip->i_number);
450 DIP_ASSIGN(oip, ib[level], 0);
451 blocksreleased += nblocks;
452 }
453 }
454 if (lastiblock[level] >= 0)
455 goto done;
456 }
457
458 /*
459 * All whole direct blocks or frags.
460 */
461 for (i = UFS_NDADDR - 1; i > lastblock; i--) {
462 long bsize;
463
464 if (oip->i_ump->um_fstype == UFS1)
465 bn = ufs_rw32(oip->i_ffs1_db[i], UFS_FSNEEDSWAP(fs));
466 else
467 bn = ufs_rw64(oip->i_ffs2_db[i], UFS_FSNEEDSWAP(fs));
468 if (bn == 0)
469 continue;
470
471 bsize = ffs_blksize(fs, oip, i);
472 if ((oip->i_ump->um_mountp->mnt_wapbl) &&
473 (ovp->v_type != VREG)) {
474 error = UFS_WAPBL_REGISTER_DEALLOCATION(
475 oip->i_ump->um_mountp,
476 FFS_FSBTODB(fs, bn), bsize, NULL);
477 if (error)
478 goto out;
479 } else
480 ffs_blkfree(fs, oip->i_devvp, bn, bsize, oip->i_number);
481 DIP_ASSIGN(oip, db[i], 0);
482 blocksreleased += btodb(bsize);
483 }
484 if (lastblock < 0)
485 goto done;
486
487 /*
488 * Finally, look for a change in size of the
489 * last direct block; release any frags.
490 */
491 if (oip->i_ump->um_fstype == UFS1)
492 bn = ufs_rw32(oip->i_ffs1_db[lastblock], UFS_FSNEEDSWAP(fs));
493 else
494 bn = ufs_rw64(oip->i_ffs2_db[lastblock], UFS_FSNEEDSWAP(fs));
495 if (bn != 0) {
496 long oldspace, newspace;
497
498 /*
499 * Calculate amount of space we're giving
500 * back as old block size minus new block size.
501 */
502 oldspace = ffs_blksize(fs, oip, lastblock);
503 oip->i_size = length;
504 DIP_ASSIGN(oip, size, length);
505 newspace = ffs_blksize(fs, oip, lastblock);
506 if (newspace == 0)
507 panic("itrunc: newspace");
508 if (oldspace - newspace > 0) {
509 /*
510 * Block number of space to be free'd is
511 * the old block # plus the number of frags
512 * required for the storage we're keeping.
513 */
514 bn += ffs_numfrags(fs, newspace);
515 if ((oip->i_ump->um_mountp->mnt_wapbl) &&
516 (ovp->v_type != VREG)) {
517 error = UFS_WAPBL_REGISTER_DEALLOCATION(
518 oip->i_ump->um_mountp, FFS_FSBTODB(fs, bn),
519 oldspace - newspace, NULL);
520 if (error)
521 goto out;
522 } else
523 ffs_blkfree(fs, oip->i_devvp, bn,
524 oldspace - newspace, oip->i_number);
525 blocksreleased += btodb(oldspace - newspace);
526 }
527 }
528
529done:
530#ifdef DIAGNOSTIC
531 for (level = SINGLE; level <= TRIPLE; level++)
532 if (blks[UFS_NDADDR + level] != DIP(oip, ib[level]))
533 panic("itrunc1");
534 for (i = 0; i < UFS_NDADDR; i++)
535 if (blks[i] != DIP(oip, db[i]))
536 panic("itrunc2");
537 if (length == 0 &&
538 (!LIST_EMPTY(&ovp->v_cleanblkhd) || !LIST_EMPTY(&ovp->v_dirtyblkhd)))
539 panic("itrunc3");
540#endif /* DIAGNOSTIC */
541
542out:
543 /*
544 * Set length back to old size if deallocation failed. Some indirect
545 * blocks were deallocated creating a hole, but that is okay.
546 */
547 if (error == EAGAIN) {
548 if (!allerror)
549 allerror = error;
550 length = osize;
551 uvm_vnp_setsize(ovp, length);
552 }
553
554 /*
555 * Put back the real size.
556 */
557 oip->i_size = length;
558 DIP_ASSIGN(oip, size, length);
559 DIP_ADD(oip, blocks, -blocksreleased);
560 genfs_node_unlock(ovp);
561 oip->i_flag |= IN_CHANGE;
562 UFS_WAPBL_UPDATE(ovp, NULL, NULL, 0);
563#if defined(QUOTA) || defined(QUOTA2)
564 (void) chkdq(oip, -blocksreleased, NOCRED, 0);
565#endif
566 KASSERT(ovp->v_type != VREG || ovp->v_size == oip->i_size);
567 return (allerror);
568}
569
570/*
571 * Release blocks associated with the inode ip and stored in the indirect
572 * block bn. Blocks are free'd in LIFO order up to (but not including)
573 * lastbn. If level is greater than SINGLE, the block is an indirect block
574 * and recursive calls to indirtrunc must be used to cleanse other indirect
575 * blocks.
576 *
577 * NB: triple indirect blocks are untested.
578 */
579static int
580ffs_indirtrunc(struct inode *ip, daddr_t lbn, daddr_t dbn, daddr_t lastbn,
581 int level, int64_t *countp)
582{
583 int i;
584 struct buf *bp;
585 struct fs *fs = ip->i_fs;
586 int32_t *bap1 = NULL;
587 int64_t *bap2 = NULL;
588 struct vnode *vp;
589 daddr_t nb, nlbn, last;
590 char *copy = NULL;
591 int64_t factor;
592 int64_t nblocks;
593 int error = 0, allerror = 0;
594 const int needswap = UFS_FSNEEDSWAP(fs);
595 const int wapbl = (ip->i_ump->um_mountp->mnt_wapbl != NULL);
596 void *dcookie;
597
598#define RBAP(ip, i) (((ip)->i_ump->um_fstype == UFS1) ? \
599 ufs_rw32(bap1[i], needswap) : ufs_rw64(bap2[i], needswap))
600#define BAP_ASSIGN(ip, i, value) \
601 do { \
602 if ((ip)->i_ump->um_fstype == UFS1) \
603 bap1[i] = (value); \
604 else \
605 bap2[i] = (value); \
606 } while(0)
607
608 /*
609 * Calculate index in current block of last
610 * block to be kept. -1 indicates the entire
611 * block so we need not calculate the index.
612 */
613 factor = 1;
614 for (i = SINGLE; i < level; i++)
615 factor *= FFS_NINDIR(fs);
616 last = lastbn;
617 if (lastbn > 0)
618 last /= factor;
619 nblocks = btodb(fs->fs_bsize);
620 /*
621 * Get buffer of block pointers, zero those entries corresponding
622 * to blocks to be free'd, and update on disk copy first. Since
623 * double(triple) indirect before single(double) indirect, calls
624 * to bmap on these blocks will fail. However, we already have
625 * the on disk address, so we have to set the b_blkno field
626 * explicitly instead of letting bread do everything for us.
627 */
628 vp = ITOV(ip);
629 error = ffs_getblk(vp, lbn, FFS_NOBLK, fs->fs_bsize, false, &bp);
630 if (error)
631 return error;
632
633 if (bp->b_oflags & (BO_DONE | BO_DELWRI)) {
634 /* Braces must be here in case trace evaluates to nothing. */
635 trace(TR_BREADHIT, pack(vp, fs->fs_bsize), lbn);
636 } else {
637 trace(TR_BREADMISS, pack(vp, fs->fs_bsize), lbn);
638 curlwp->l_ru.ru_inblock++; /* pay for read */
639 bp->b_flags |= B_READ;
640 bp->b_flags &= ~B_COWDONE; /* we change blkno below */
641 if (bp->b_bcount > bp->b_bufsize)
642 panic("ffs_indirtrunc: bad buffer size");
643 bp->b_blkno = dbn;
644 BIO_SETPRIO(bp, BPRIO_TIMECRITICAL);
645 VOP_STRATEGY(vp, bp);
646 error = biowait(bp);
647 if (error == 0)
648 error = fscow_run(bp, true);
649 }
650 if (error) {
651 brelse(bp, 0);
652 return error;
653 }
654
655 /*
656 * Clear reference to blocks to be removed on disk, before actually
657 * reclaiming them, so that fsck is more likely to be able to recover
658 * the filesystem if system goes down during the truncate process.
659 * This assumes the truncate process would not fail, contrary
660 * to the wapbl case.
661 */
662 if (ip->i_ump->um_fstype == UFS1)
663 bap1 = (int32_t *)bp->b_data;
664 else
665 bap2 = (int64_t *)bp->b_data;
666 if (lastbn >= 0 && !wapbl) {
667 copy = kmem_alloc(fs->fs_bsize, KM_SLEEP);
668 memcpy((void *)copy, bp->b_data, (u_int)fs->fs_bsize);
669 for (i = last + 1; i < FFS_NINDIR(fs); i++)
670 BAP_ASSIGN(ip, i, 0);
671 error = bwrite(bp);
672 if (error)
673 allerror = error;
674
675 if (ip->i_ump->um_fstype == UFS1)
676 bap1 = (int32_t *)copy;
677 else
678 bap2 = (int64_t *)copy;
679 }
680
681 /*
682 * Recursively free totally unused blocks.
683 */
684 for (i = FFS_NINDIR(fs) - 1, nlbn = lbn + 1 - i * factor; i > last;
685 i--, nlbn += factor) {
686 nb = RBAP(ip, i);
687 if (nb == 0)
688 continue;
689
690 if ((ip->i_ump->um_mountp->mnt_wapbl) &&
691 ((level > SINGLE) || (ITOV(ip)->v_type != VREG))) {
692 error = UFS_WAPBL_REGISTER_DEALLOCATION(
693 ip->i_ump->um_mountp,
694 FFS_FSBTODB(fs, nb), fs->fs_bsize,
695 &dcookie);
696 if (error)
697 goto out;
698 } else {
699 dcookie = NULL;
700 }
701
702 if (level > SINGLE) {
703 error = ffs_indirtrunc(ip, nlbn, FFS_FSBTODB(fs, nb),
704 (daddr_t)-1, level - 1, countp);
705 if (error) {
706 if (dcookie) {
707 UFS_WAPBL_UNREGISTER_DEALLOCATION(
708 ip->i_ump->um_mountp, dcookie);
709 }
710
711 goto out;
712 }
713 }
714
715 if (!dcookie)
716 ffs_blkfree(fs, ip->i_devvp, nb, fs->fs_bsize,
717 ip->i_number);
718
719 BAP_ASSIGN(ip, i, 0);
720 *countp += nblocks;
721 }
722
723 /*
724 * Recursively free blocks on the now last partial indirect block.
725 */
726 if (level > SINGLE && lastbn >= 0) {
727 last = lastbn % factor;
728 nb = RBAP(ip, i);
729 if (nb != 0) {
730 error = ffs_indirtrunc(ip, nlbn, FFS_FSBTODB(fs, nb),
731 last, level - 1, countp);
732 if (error)
733 goto out;
734 }
735 }
736
737out:
738 if (error && !allerror)
739 allerror = error;
740
741 if (copy != NULL) {
742 kmem_free(copy, fs->fs_bsize);
743 } else if (lastbn < 0 && error == 0) {
744 /* all freed, release without writing back */
745 brelse(bp, BC_INVAL);
746 } else if (wapbl) {
747 /* only partially freed, write the updated block */
748 error = bwrite(bp);
749 if (!allerror)
750 allerror = error;
751 }
752
753 return (allerror);
754}
755
756void
757ffs_itimes(struct inode *ip, const struct timespec *acc,
758 const struct timespec *mod, const struct timespec *cre)
759{
760 struct timespec now;
761
762 if (!(ip->i_flag & (IN_ACCESS | IN_CHANGE | IN_UPDATE | IN_MODIFY))) {
763 return;
764 }
765
766 vfs_timestamp(&now);
767 if (ip->i_flag & IN_ACCESS) {
768 if (acc == NULL)
769 acc = &now;
770 DIP_ASSIGN(ip, atime, acc->tv_sec);
771 DIP_ASSIGN(ip, atimensec, acc->tv_nsec);
772 }
773 if (ip->i_flag & (IN_UPDATE | IN_MODIFY)) {
774 if ((ip->i_flags & SF_SNAPSHOT) == 0) {
775 if (mod == NULL)
776 mod = &now;
777 DIP_ASSIGN(ip, mtime, mod->tv_sec);
778 DIP_ASSIGN(ip, mtimensec, mod->tv_nsec);
779 }
780 ip->i_modrev++;
781 }
782 if (ip->i_flag & (IN_CHANGE | IN_MODIFY)) {
783 if (cre == NULL)
784 cre = &now;
785 DIP_ASSIGN(ip, ctime, cre->tv_sec);
786 DIP_ASSIGN(ip, ctimensec, cre->tv_nsec);
787 }
788 if (ip->i_flag & (IN_ACCESS | IN_MODIFY))
789 ip->i_flag |= IN_ACCESSED;
790 if (ip->i_flag & (IN_UPDATE | IN_CHANGE))
791 ip->i_flag |= IN_MODIFIED;
792 ip->i_flag &= ~(IN_ACCESS | IN_CHANGE | IN_UPDATE | IN_MODIFY);
793}
794