1 /* $NetBSD: ffs_inode.c,v 1.131 2020/07/31 04:07:30 chs 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 1.131 2020/07/31 04:07:30 chs 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
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 then wait for the
107 * disk write of the inode to complete.
108 */
109
110 int
ffs_update(struct vnode * vp,const struct timespec * acc,const struct timespec * mod,int updflags)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) != 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 */
206 int
ffs_truncate(struct vnode * ovp,off_t length,int ioflag,kauth_cred_t cred)207 ffs_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 struct mount *omp = ovp->v_mount;
212 daddr_t bn, lastiblock[UFS_NIADDR], indir_lbn[UFS_NIADDR];
213 daddr_t blks[UFS_NDADDR + UFS_NIADDR], oldblks[UFS_NDADDR + UFS_NIADDR];
214 struct fs *fs;
215 int extblocks;
216 int offset, pgoffset, level;
217 int64_t blocksreleased = 0, datablocks;
218 int i, aflag, nblocks;
219 int error, allerror = 0;
220 off_t osize;
221 int sync;
222 struct ufsmount *ump = oip->i_ump;
223 void *dcookie;
224 long bsize;
225 bool wapbl = omp->mnt_wapbl != NULL;
226
227 UFS_WAPBL_JLOCK_ASSERT(ump->um_mountp);
228
229 if (ovp->v_type == VCHR || ovp->v_type == VBLK ||
230 ovp->v_type == VFIFO || ovp->v_type == VSOCK) {
231 KASSERT(oip->i_size == 0);
232 return 0;
233 }
234
235 if (length < 0)
236 return (EINVAL);
237
238 /*
239 * Historically clients did not have to specify which data
240 * they were truncating. So, if not specified, we assume
241 * traditional behavior, e.g., just the normal data.
242 */
243 if ((ioflag & (IO_EXT | IO_NORMAL)) == 0)
244 ioflag |= IO_NORMAL;
245
246 fs = oip->i_fs;
247 #define i_din2 i_din.ffs2_din
248 extblocks = 0;
249 datablocks = DIP(oip, blocks);
250 if (fs->fs_magic == FS_UFS2_MAGIC && oip->i_din2->di_extsize > 0) {
251 extblocks = btodb(ffs_fragroundup(fs, oip->i_din2->di_extsize));
252 datablocks -= extblocks;
253 }
254 if ((ioflag & IO_EXT) && extblocks > 0) {
255 if (length != 0)
256 panic("ffs_truncate: partial trunc of extdata");
257 {
258 #ifdef QUOTA
259 (void) chkdq(oip, -extblocks, NOCRED, FORCE);
260 #endif
261 osize = oip->i_din2->di_extsize;
262 oip->i_din2->di_blocks -= extblocks;
263 oip->i_din2->di_extsize = 0;
264 for (i = 0; i < UFS_NXADDR; i++) {
265 binvalbuf(ovp, -1 - i);
266 oldblks[i] = oip->i_din2->di_extb[i];
267 oip->i_din2->di_extb[i] = 0;
268 }
269 oip->i_flag |= IN_CHANGE;
270 if ((error = ffs_update(ovp, NULL, NULL, 0)))
271 return (error);
272 for (i = 0; i < UFS_NXADDR; i++) {
273 if (oldblks[i] == 0)
274 continue;
275 bsize = ffs_sblksize(fs, osize, i);
276 if (wapbl) {
277 error = UFS_WAPBL_REGISTER_DEALLOCATION(omp,
278 FFS_FSBTODB(fs, oldblks[i]), bsize, NULL);
279 if (error)
280 return error;
281 } else
282 ffs_blkfree(fs, oip->i_devvp, oldblks[i],
283 bsize, oip->i_number);
284 }
285 extblocks = 0;
286 }
287 }
288 if ((ioflag & IO_NORMAL) == 0)
289 return (0);
290 if (ovp->v_type == VLNK &&
291 (oip->i_size < ump->um_maxsymlinklen ||
292 (ump->um_maxsymlinklen == 0 && datablocks == 0))) {
293 KDASSERT(length == 0);
294 memset(SHORTLINK(oip), 0, (size_t)oip->i_size);
295 oip->i_size = 0;
296 DIP_ASSIGN(oip, size, 0);
297 oip->i_flag |= IN_CHANGE | IN_UPDATE;
298 return (ffs_update(ovp, NULL, NULL, 0));
299 }
300 if (oip->i_size == length) {
301 /* still do a uvm_vnp_setsize() as writesize may be larger */
302 uvm_vnp_setsize(ovp, length);
303 oip->i_flag |= IN_CHANGE | IN_UPDATE;
304 return (ffs_update(ovp, NULL, NULL, 0));
305 }
306 if (length > ump->um_maxfilesize)
307 return (EFBIG);
308
309 if ((oip->i_flags & SF_SNAPSHOT) != 0)
310 ffs_snapremove(ovp);
311
312 osize = oip->i_size;
313 aflag = ioflag & IO_SYNC ? B_SYNC : 0;
314
315 /*
316 * Lengthen the size of the file. We must ensure that the
317 * last byte of the file is allocated. Since the smallest
318 * value of osize is 0, length will be at least 1.
319 */
320
321 if (osize < length) {
322 if (ffs_lblkno(fs, osize) < UFS_NDADDR &&
323 ffs_lblkno(fs, osize) != ffs_lblkno(fs, length) &&
324 ffs_blkroundup(fs, osize) != osize) {
325 off_t eob;
326
327 eob = ffs_blkroundup(fs, osize);
328 uvm_vnp_setwritesize(ovp, eob);
329 error = ufs_balloc_range(ovp, osize, eob - osize,
330 cred, aflag);
331 if (error) {
332 (void) ffs_truncate(ovp, osize,
333 ioflag & IO_SYNC, cred);
334 return error;
335 }
336 if (ioflag & IO_SYNC) {
337 rw_enter(ovp->v_uobj.vmobjlock, RW_WRITER);
338 VOP_PUTPAGES(ovp,
339 trunc_page(osize & fs->fs_bmask),
340 round_page(eob), PGO_CLEANIT | PGO_SYNCIO |
341 PGO_JOURNALLOCKED);
342 }
343 }
344 uvm_vnp_setwritesize(ovp, length);
345 error = ufs_balloc_range(ovp, length - 1, 1, cred, aflag);
346 if (error) {
347 (void) ffs_truncate(ovp, osize, ioflag & IO_SYNC, cred);
348 return (error);
349 }
350 uvm_vnp_setsize(ovp, length);
351 oip->i_flag |= IN_CHANGE | IN_UPDATE;
352 KASSERT(ovp->v_size == oip->i_size);
353 return (ffs_update(ovp, NULL, NULL, 0));
354 }
355
356 /*
357 * When truncating a regular file down to a non-block-aligned size,
358 * we must zero the part of last block which is past the new EOF.
359 * We must synchronously flush the zeroed pages to disk
360 * since the new pages will be invalidated as soon as we
361 * inform the VM system of the new, smaller size.
362 * We must do this before acquiring the GLOCK, since fetching
363 * the pages will acquire the GLOCK internally.
364 * So there is a window where another thread could see a whole
365 * zeroed page past EOF, but that's life.
366 */
367
368 offset = ffs_blkoff(fs, length);
369 pgoffset = length & PAGE_MASK;
370 if (ovp->v_type == VREG && (pgoffset != 0 || offset != 0) &&
371 osize > length) {
372 daddr_t lbn;
373 voff_t eoz;
374 int size;
375
376 if (offset != 0) {
377 error = ufs_balloc_range(ovp, length - 1, 1, cred,
378 aflag);
379 if (error)
380 return error;
381 }
382 lbn = ffs_lblkno(fs, length);
383 size = ffs_blksize(fs, oip, lbn);
384 eoz = MIN(MAX(ffs_lblktosize(fs, lbn) + size, round_page(pgoffset)),
385 osize);
386 ubc_zerorange(&ovp->v_uobj, length, eoz - length,
387 UBC_VNODE_FLAGS(ovp));
388 if (round_page(eoz) > round_page(length)) {
389 rw_enter(ovp->v_uobj.vmobjlock, RW_WRITER);
390 error = VOP_PUTPAGES(ovp, round_page(length),
391 round_page(eoz),
392 PGO_CLEANIT | PGO_DEACTIVATE | PGO_JOURNALLOCKED |
393 ((ioflag & IO_SYNC) ? PGO_SYNCIO : 0));
394 if (error)
395 return error;
396 }
397 }
398
399 genfs_node_wrlock(ovp);
400 oip->i_size = length;
401 DIP_ASSIGN(oip, size, length);
402 uvm_vnp_setsize(ovp, length);
403 /*
404 * Calculate index into inode's block list of
405 * last direct and indirect blocks (if any)
406 * which we want to keep. Lastblock is -1 when
407 * the file is truncated to 0.
408 */
409 lastblock = ffs_lblkno(fs, length + fs->fs_bsize - 1) - 1;
410 lastiblock[SINGLE] = lastblock - UFS_NDADDR;
411 lastiblock[DOUBLE] = lastiblock[SINGLE] - FFS_NINDIR(fs);
412 lastiblock[TRIPLE] = lastiblock[DOUBLE] - FFS_NINDIR(fs) * FFS_NINDIR(fs);
413 nblocks = btodb(fs->fs_bsize);
414 /*
415 * Update file and block pointers on disk before we start freeing
416 * blocks. If we crash before free'ing blocks below, the blocks
417 * will be returned to the free list. lastiblock values are also
418 * normalized to -1 for calls to ffs_indirtrunc below.
419 */
420 sync = 0;
421 for (level = TRIPLE; level >= SINGLE; level--) {
422 blks[UFS_NDADDR + level] = DIP(oip, ib[level]);
423 if (lastiblock[level] < 0 && blks[UFS_NDADDR + level] != 0) {
424 sync = 1;
425 DIP_ASSIGN(oip, ib[level], 0);
426 lastiblock[level] = -1;
427 }
428 }
429 for (i = 0; i < UFS_NDADDR; i++) {
430 blks[i] = DIP(oip, db[i]);
431 if (i > lastblock && blks[i] != 0) {
432 sync = 1;
433 DIP_ASSIGN(oip, db[i], 0);
434 }
435 }
436 oip->i_flag |= IN_CHANGE | IN_UPDATE;
437 if (sync) {
438 error = ffs_update(ovp, NULL, NULL, UPDATE_WAIT);
439 if (error && !allerror)
440 allerror = error;
441 }
442
443 /*
444 * Having written the new inode to disk, save its new configuration
445 * and put back the old block pointers long enough to process them.
446 * Note that we save the new block configuration so we can check it
447 * when we are done.
448 */
449 for (i = 0; i < UFS_NDADDR; i++) {
450 bn = DIP(oip, db[i]);
451 DIP_ASSIGN(oip, db[i], blks[i]);
452 blks[i] = bn;
453 }
454 for (i = 0; i < UFS_NIADDR; i++) {
455 bn = DIP(oip, ib[i]);
456 DIP_ASSIGN(oip, ib[i], blks[UFS_NDADDR + i]);
457 blks[UFS_NDADDR + i] = bn;
458 }
459
460 oip->i_size = osize;
461 DIP_ASSIGN(oip, size, osize);
462 error = vtruncbuf(ovp, lastblock + 1, 0, 0);
463 if (error && !allerror)
464 allerror = error;
465
466 /*
467 * Indirect blocks first.
468 */
469 indir_lbn[SINGLE] = -UFS_NDADDR;
470 indir_lbn[DOUBLE] = indir_lbn[SINGLE] - FFS_NINDIR(fs) - 1;
471 indir_lbn[TRIPLE] = indir_lbn[DOUBLE] - FFS_NINDIR(fs) * FFS_NINDIR(fs) - 1;
472 for (level = TRIPLE; level >= SINGLE; level--) {
473 bn = ffs_getib(fs, oip, level);
474 if (bn != 0) {
475 if (lastiblock[level] < 0 &&
476 oip->i_ump->um_mountp->mnt_wapbl) {
477 error = UFS_WAPBL_REGISTER_DEALLOCATION(
478 oip->i_ump->um_mountp,
479 FFS_FSBTODB(fs, bn), fs->fs_bsize,
480 &dcookie);
481 if (error)
482 goto out;
483 } else {
484 dcookie = NULL;
485 }
486
487 error = ffs_indirtrunc(oip, indir_lbn[level],
488 FFS_FSBTODB(fs, bn), lastiblock[level], level,
489 &blocksreleased);
490 if (error) {
491 if (dcookie) {
492 UFS_WAPBL_UNREGISTER_DEALLOCATION(
493 oip->i_ump->um_mountp, dcookie);
494 }
495 goto out;
496 }
497
498 if (lastiblock[level] < 0) {
499 if (!dcookie)
500 ffs_blkfree(fs, oip->i_devvp, bn,
501 fs->fs_bsize, oip->i_number);
502 DIP_ASSIGN(oip, ib[level], 0);
503 blocksreleased += nblocks;
504 }
505 }
506 if (lastiblock[level] >= 0)
507 goto done;
508 }
509
510 /*
511 * All whole direct blocks or frags.
512 */
513 for (i = UFS_NDADDR - 1; i > lastblock; i--) {
514 bn = ffs_getdb(fs, oip, i);
515 if (bn == 0)
516 continue;
517
518 bsize = ffs_blksize(fs, oip, i);
519 if ((oip->i_ump->um_mountp->mnt_wapbl) &&
520 (ovp->v_type != VREG)) {
521 error = UFS_WAPBL_REGISTER_DEALLOCATION(
522 oip->i_ump->um_mountp,
523 FFS_FSBTODB(fs, bn), bsize, NULL);
524 if (error)
525 goto out;
526 } else
527 ffs_blkfree(fs, oip->i_devvp, bn, bsize, oip->i_number);
528 DIP_ASSIGN(oip, db[i], 0);
529 blocksreleased += btodb(bsize);
530 }
531 if (lastblock < 0)
532 goto done;
533
534 /*
535 * Finally, look for a change in size of the
536 * last direct block; release any frags.
537 */
538 bn = ffs_getdb(fs, oip, lastblock);
539 if (bn != 0) {
540 long oldspace, newspace;
541
542 /*
543 * Calculate amount of space we're giving
544 * back as old block size minus new block size.
545 */
546 oldspace = ffs_blksize(fs, oip, lastblock);
547 oip->i_size = length;
548 DIP_ASSIGN(oip, size, length);
549 newspace = ffs_blksize(fs, oip, lastblock);
550 if (newspace == 0)
551 panic("itrunc: newspace");
552 if (oldspace - newspace > 0) {
553 /*
554 * Block number of space to be free'd is
555 * the old block # plus the number of frags
556 * required for the storage we're keeping.
557 */
558 bn += ffs_numfrags(fs, newspace);
559 if ((oip->i_ump->um_mountp->mnt_wapbl) &&
560 (ovp->v_type != VREG)) {
561 error = UFS_WAPBL_REGISTER_DEALLOCATION(
562 oip->i_ump->um_mountp, FFS_FSBTODB(fs, bn),
563 oldspace - newspace, NULL);
564 if (error)
565 goto out;
566 } else
567 ffs_blkfree(fs, oip->i_devvp, bn,
568 oldspace - newspace, oip->i_number);
569 blocksreleased += btodb(oldspace - newspace);
570 }
571 }
572
573 done:
574 for (level = SINGLE; level <= TRIPLE; level++)
575 KASSERTMSG((blks[UFS_NDADDR + level] == DIP(oip, ib[level])),
576 "itrunc1 blk mismatch: %jx != %jx",
577 (uintmax_t)blks[UFS_NDADDR + level],
578 (uintmax_t)DIP(oip, ib[level]));
579 for (i = 0; i < UFS_NDADDR; i++)
580 KASSERTMSG((blks[i] == DIP(oip, db[i])),
581 "itrunc2 blk mismatch: %jx != %jx",
582 (uintmax_t)blks[i], (uintmax_t)DIP(oip, db[i]));
583 KASSERTMSG((length != 0 || extblocks || LIST_EMPTY(&ovp->v_cleanblkhd)),
584 "itrunc3: zero length and nonempty cleanblkhd");
585 KASSERTMSG((length != 0 || extblocks || LIST_EMPTY(&ovp->v_dirtyblkhd)),
586 "itrunc3: zero length and nonempty dirtyblkhd");
587
588 out:
589 /*
590 * Set length back to old size if deallocation failed. Some indirect
591 * blocks were deallocated creating a hole, but that is okay.
592 */
593 if (error == EAGAIN) {
594 if (!allerror)
595 allerror = error;
596 length = osize;
597 uvm_vnp_setsize(ovp, length);
598 }
599
600 /*
601 * Put back the real size.
602 */
603 oip->i_size = length;
604 DIP_ASSIGN(oip, size, length);
605 DIP_ADD(oip, blocks, -blocksreleased);
606 genfs_node_unlock(ovp);
607 oip->i_flag |= IN_CHANGE;
608 UFS_WAPBL_UPDATE(ovp, NULL, NULL, 0);
609 #if defined(QUOTA) || defined(QUOTA2)
610 (void) chkdq(oip, -blocksreleased, NOCRED, 0);
611 #endif
612 KASSERT(ovp->v_type != VREG || ovp->v_size == oip->i_size);
613 return (allerror);
614 }
615
616 /*
617 * Release blocks associated with the inode ip and stored in the indirect
618 * block bn. Blocks are free'd in LIFO order up to (but not including)
619 * lastbn. If level is greater than SINGLE, the block is an indirect block
620 * and recursive calls to indirtrunc must be used to cleanse other indirect
621 * blocks.
622 *
623 * NB: triple indirect blocks are untested.
624 */
625 static int
ffs_indirtrunc(struct inode * ip,daddr_t lbn,daddr_t dbn,daddr_t lastbn,int level,int64_t * countp)626 ffs_indirtrunc(struct inode *ip, daddr_t lbn, daddr_t dbn, daddr_t lastbn,
627 int level, int64_t *countp)
628 {
629 int i;
630 struct buf *bp;
631 struct fs *fs = ip->i_fs;
632 int32_t *bap1 = NULL;
633 int64_t *bap2 = NULL;
634 struct vnode *vp;
635 daddr_t nb, nlbn, last;
636 char *copy = NULL;
637 int64_t factor;
638 int64_t nblocks;
639 int error = 0, allerror = 0;
640 const int needswap = UFS_FSNEEDSWAP(fs);
641 const int wapbl = (ip->i_ump->um_mountp->mnt_wapbl != NULL);
642 void *dcookie;
643
644 #define RBAP(ip, i) (((ip)->i_ump->um_fstype == UFS1) ? \
645 ufs_rw32(bap1[i], needswap) : ufs_rw64(bap2[i], needswap))
646 #define BAP_ASSIGN(ip, i, value) \
647 do { \
648 if ((ip)->i_ump->um_fstype == UFS1) \
649 bap1[i] = (value); \
650 else \
651 bap2[i] = (value); \
652 } while(0)
653
654 /*
655 * Calculate index in current block of last
656 * block to be kept. -1 indicates the entire
657 * block so we need not calculate the index.
658 */
659 factor = 1;
660 for (i = SINGLE; i < level; i++)
661 factor *= FFS_NINDIR(fs);
662 last = lastbn;
663 if (lastbn > 0)
664 last /= factor;
665 nblocks = btodb(fs->fs_bsize);
666 /*
667 * Get buffer of block pointers, zero those entries corresponding
668 * to blocks to be free'd, and update on disk copy first. Since
669 * double(triple) indirect before single(double) indirect, calls
670 * to bmap on these blocks will fail. However, we already have
671 * the on disk address, so we have to set the b_blkno field
672 * explicitly instead of letting bread do everything for us.
673 */
674 vp = ITOV(ip);
675 error = ffs_getblk(vp, lbn, FFS_NOBLK, fs->fs_bsize, false, &bp);
676 if (error)
677 return error;
678
679 if (bp->b_oflags & (BO_DONE | BO_DELWRI)) {
680 /* Braces must be here in case trace evaluates to nothing. */
681 trace(TR_BREADHIT, pack(vp, fs->fs_bsize), lbn);
682 } else {
683 trace(TR_BREADMISS, pack(vp, fs->fs_bsize), lbn);
684 curlwp->l_ru.ru_inblock++; /* pay for read */
685 bp->b_flags |= B_READ;
686 bp->b_flags &= ~B_COWDONE; /* we change blkno below */
687 if (bp->b_bcount > bp->b_bufsize)
688 panic("ffs_indirtrunc: bad buffer size");
689 bp->b_blkno = dbn;
690 BIO_SETPRIO(bp, BPRIO_TIMECRITICAL);
691 VOP_STRATEGY(vp, bp);
692 error = biowait(bp);
693 if (error == 0)
694 error = fscow_run(bp, true);
695 }
696 if (error) {
697 brelse(bp, 0);
698 return error;
699 }
700
701 /*
702 * Clear reference to blocks to be removed on disk, before actually
703 * reclaiming them, so that fsck is more likely to be able to recover
704 * the filesystem if system goes down during the truncate process.
705 * This assumes the truncate process would not fail, contrary
706 * to the wapbl case.
707 */
708 if (ip->i_ump->um_fstype == UFS1)
709 bap1 = (int32_t *)bp->b_data;
710 else
711 bap2 = (int64_t *)bp->b_data;
712 if (lastbn >= 0 && !wapbl) {
713 copy = kmem_alloc(fs->fs_bsize, KM_SLEEP);
714 memcpy((void *)copy, bp->b_data, (u_int)fs->fs_bsize);
715 for (i = last + 1; i < FFS_NINDIR(fs); i++)
716 BAP_ASSIGN(ip, i, 0);
717 error = bwrite(bp);
718 if (error)
719 allerror = error;
720
721 if (ip->i_ump->um_fstype == UFS1)
722 bap1 = (int32_t *)copy;
723 else
724 bap2 = (int64_t *)copy;
725 }
726
727 /*
728 * Recursively free totally unused blocks.
729 */
730 for (i = FFS_NINDIR(fs) - 1, nlbn = lbn + 1 - i * factor; i > last;
731 i--, nlbn += factor) {
732 nb = RBAP(ip, i);
733 if (nb == 0)
734 continue;
735
736 if ((ip->i_ump->um_mountp->mnt_wapbl) &&
737 ((level > SINGLE) || (ITOV(ip)->v_type != VREG))) {
738 error = UFS_WAPBL_REGISTER_DEALLOCATION(
739 ip->i_ump->um_mountp,
740 FFS_FSBTODB(fs, nb), fs->fs_bsize,
741 &dcookie);
742 if (error)
743 goto out;
744 } else {
745 dcookie = NULL;
746 }
747
748 if (level > SINGLE) {
749 error = ffs_indirtrunc(ip, nlbn, FFS_FSBTODB(fs, nb),
750 (daddr_t)-1, level - 1, countp);
751 if (error) {
752 if (dcookie) {
753 UFS_WAPBL_UNREGISTER_DEALLOCATION(
754 ip->i_ump->um_mountp, dcookie);
755 }
756
757 goto out;
758 }
759 }
760
761 if (!dcookie)
762 ffs_blkfree(fs, ip->i_devvp, nb, fs->fs_bsize,
763 ip->i_number);
764
765 BAP_ASSIGN(ip, i, 0);
766 *countp += nblocks;
767 }
768
769 /*
770 * Recursively free blocks on the now last partial indirect block.
771 */
772 if (level > SINGLE && lastbn >= 0) {
773 last = lastbn % factor;
774 nb = RBAP(ip, i);
775 if (nb != 0) {
776 error = ffs_indirtrunc(ip, nlbn, FFS_FSBTODB(fs, nb),
777 last, level - 1, countp);
778 if (error)
779 goto out;
780 }
781 }
782
783 out:
784 if (error && !allerror)
785 allerror = error;
786
787 if (copy != NULL) {
788 kmem_free(copy, fs->fs_bsize);
789 } else if (lastbn < 0 && error == 0) {
790 /* all freed, release without writing back */
791 brelse(bp, BC_INVAL);
792 } else if (wapbl) {
793 /* only partially freed, write the updated block */
794 error = bwrite(bp);
795 if (!allerror)
796 allerror = error;
797 }
798
799 return (allerror);
800 }
801
802 void
ffs_itimes(struct inode * ip,const struct timespec * acc,const struct timespec * mod,const struct timespec * cre)803 ffs_itimes(struct inode *ip, const struct timespec *acc,
804 const struct timespec *mod, const struct timespec *cre)
805 {
806 struct timespec now;
807
808 if (!(ip->i_flag & (IN_ACCESS | IN_CHANGE | IN_UPDATE | IN_MODIFY))) {
809 return;
810 }
811
812 vfs_timestamp(&now);
813 if (ip->i_flag & IN_ACCESS) {
814 if (acc == NULL)
815 acc = &now;
816 DIP_ASSIGN(ip, atime, acc->tv_sec);
817 DIP_ASSIGN(ip, atimensec, acc->tv_nsec);
818 }
819 if (ip->i_flag & (IN_UPDATE | IN_MODIFY)) {
820 if ((ip->i_flags & SF_SNAPSHOT) == 0) {
821 if (mod == NULL)
822 mod = &now;
823 DIP_ASSIGN(ip, mtime, mod->tv_sec);
824 DIP_ASSIGN(ip, mtimensec, mod->tv_nsec);
825 }
826 ip->i_modrev++;
827 }
828 if (ip->i_flag & (IN_CHANGE | IN_MODIFY)) {
829 if (cre == NULL)
830 cre = &now;
831 DIP_ASSIGN(ip, ctime, cre->tv_sec);
832 DIP_ASSIGN(ip, ctimensec, cre->tv_nsec);
833 }
834 if (ip->i_flag & (IN_ACCESS | IN_MODIFY))
835 ip->i_flag |= IN_ACCESSED;
836 if (ip->i_flag & (IN_UPDATE | IN_CHANGE))
837 ip->i_flag |= IN_MODIFIED;
838 ip->i_flag &= ~(IN_ACCESS | IN_CHANGE | IN_UPDATE | IN_MODIFY);
839 }
840