1 /* $NetBSD: ffs_inode.c,v 1.117 2015/03/28 19:24:04 maxv 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 1.117 2015/03/28 19:24:04 maxv 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 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 count, 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
222 if (ovp->v_type == VCHR || ovp->v_type == VBLK ||
223 ovp->v_type == VFIFO || ovp->v_type == VSOCK) {
224 KASSERT(oip->i_size == 0);
225 return 0;
226 }
227
228 if (length < 0)
229 return (EINVAL);
230
231 if (ovp->v_type == VLNK &&
232 (oip->i_size < ump->um_maxsymlinklen ||
233 (ump->um_maxsymlinklen == 0 && DIP(oip, blocks) == 0))) {
234 KDASSERT(length == 0);
235 memset(SHORTLINK(oip), 0, (size_t)oip->i_size);
236 oip->i_size = 0;
237 DIP_ASSIGN(oip, size, 0);
238 oip->i_flag |= IN_CHANGE | IN_UPDATE;
239 return (ffs_update(ovp, NULL, NULL, 0));
240 }
241 if (oip->i_size == length) {
242 /* still do a uvm_vnp_setsize() as writesize may be larger */
243 uvm_vnp_setsize(ovp, length);
244 oip->i_flag |= IN_CHANGE | IN_UPDATE;
245 return (ffs_update(ovp, NULL, NULL, 0));
246 }
247 fs = oip->i_fs;
248 if (length > ump->um_maxfilesize)
249 return (EFBIG);
250
251 if ((oip->i_flags & SF_SNAPSHOT) != 0)
252 ffs_snapremove(ovp);
253
254 osize = oip->i_size;
255 aflag = ioflag & IO_SYNC ? B_SYNC : 0;
256
257 /*
258 * Lengthen the size of the file. We must ensure that the
259 * last byte of the file is allocated. Since the smallest
260 * value of osize is 0, length will be at least 1.
261 */
262
263 if (osize < length) {
264 if (ffs_lblkno(fs, osize) < UFS_NDADDR &&
265 ffs_lblkno(fs, osize) != ffs_lblkno(fs, length) &&
266 ffs_blkroundup(fs, osize) != osize) {
267 off_t eob;
268
269 eob = ffs_blkroundup(fs, osize);
270 uvm_vnp_setwritesize(ovp, eob);
271 error = ufs_balloc_range(ovp, osize, eob - osize,
272 cred, aflag);
273 if (error) {
274 (void) ffs_truncate(ovp, osize,
275 ioflag & IO_SYNC, cred);
276 return error;
277 }
278 if (ioflag & IO_SYNC) {
279 mutex_enter(ovp->v_interlock);
280 VOP_PUTPAGES(ovp,
281 trunc_page(osize & fs->fs_bmask),
282 round_page(eob), PGO_CLEANIT | PGO_SYNCIO |
283 PGO_JOURNALLOCKED);
284 }
285 }
286 uvm_vnp_setwritesize(ovp, length);
287 error = ufs_balloc_range(ovp, length - 1, 1, cred, aflag);
288 if (error) {
289 (void) ffs_truncate(ovp, osize, ioflag & IO_SYNC, cred);
290 return (error);
291 }
292 uvm_vnp_setsize(ovp, length);
293 oip->i_flag |= IN_CHANGE | IN_UPDATE;
294 KASSERT(ovp->v_size == oip->i_size);
295 return (ffs_update(ovp, NULL, NULL, 0));
296 }
297
298 /*
299 * When truncating a regular file down to a non-block-aligned size,
300 * we must zero the part of last block which is past the new EOF.
301 * We must synchronously flush the zeroed pages to disk
302 * since the new pages will be invalidated as soon as we
303 * inform the VM system of the new, smaller size.
304 * We must do this before acquiring the GLOCK, since fetching
305 * the pages will acquire the GLOCK internally.
306 * So there is a window where another thread could see a whole
307 * zeroed page past EOF, but that's life.
308 */
309
310 offset = ffs_blkoff(fs, length);
311 pgoffset = length & PAGE_MASK;
312 if (ovp->v_type == VREG && (pgoffset != 0 || offset != 0) &&
313 osize > length) {
314 daddr_t lbn;
315 voff_t eoz;
316 int size;
317
318 if (offset != 0) {
319 error = ufs_balloc_range(ovp, length - 1, 1, cred,
320 aflag);
321 if (error)
322 return error;
323 }
324 lbn = ffs_lblkno(fs, length);
325 size = ffs_blksize(fs, oip, lbn);
326 eoz = MIN(MAX(ffs_lblktosize(fs, lbn) + size, round_page(pgoffset)),
327 osize);
328 ubc_zerorange(&ovp->v_uobj, length, eoz - length,
329 UBC_UNMAP_FLAG(ovp));
330 if (round_page(eoz) > round_page(length)) {
331 mutex_enter(ovp->v_interlock);
332 error = VOP_PUTPAGES(ovp, round_page(length),
333 round_page(eoz),
334 PGO_CLEANIT | PGO_DEACTIVATE | PGO_JOURNALLOCKED |
335 ((ioflag & IO_SYNC) ? PGO_SYNCIO : 0));
336 if (error)
337 return error;
338 }
339 }
340
341 genfs_node_wrlock(ovp);
342 oip->i_size = length;
343 DIP_ASSIGN(oip, size, length);
344 uvm_vnp_setsize(ovp, length);
345 /*
346 * Calculate index into inode's block list of
347 * last direct and indirect blocks (if any)
348 * which we want to keep. Lastblock is -1 when
349 * the file is truncated to 0.
350 */
351 lastblock = ffs_lblkno(fs, length + fs->fs_bsize - 1) - 1;
352 lastiblock[SINGLE] = lastblock - UFS_NDADDR;
353 lastiblock[DOUBLE] = lastiblock[SINGLE] - FFS_NINDIR(fs);
354 lastiblock[TRIPLE] = lastiblock[DOUBLE] - FFS_NINDIR(fs) * FFS_NINDIR(fs);
355 nblocks = btodb(fs->fs_bsize);
356 /*
357 * Update file and block pointers on disk before we start freeing
358 * blocks. If we crash before free'ing blocks below, the blocks
359 * will be returned to the free list. lastiblock values are also
360 * normalized to -1 for calls to ffs_indirtrunc below.
361 */
362 sync = 0;
363 for (level = TRIPLE; level >= SINGLE; level--) {
364 blks[UFS_NDADDR + level] = DIP(oip, ib[level]);
365 if (lastiblock[level] < 0 && blks[UFS_NDADDR + level] != 0) {
366 sync = 1;
367 DIP_ASSIGN(oip, ib[level], 0);
368 lastiblock[level] = -1;
369 }
370 }
371 for (i = 0; i < UFS_NDADDR; i++) {
372 blks[i] = DIP(oip, db[i]);
373 if (i > lastblock && blks[i] != 0) {
374 sync = 1;
375 DIP_ASSIGN(oip, db[i], 0);
376 }
377 }
378 oip->i_flag |= IN_CHANGE | IN_UPDATE;
379 if (sync) {
380 error = ffs_update(ovp, NULL, NULL, UPDATE_WAIT);
381 if (error && !allerror)
382 allerror = error;
383 }
384
385 /*
386 * Having written the new inode to disk, save its new configuration
387 * and put back the old block pointers long enough to process them.
388 * Note that we save the new block configuration so we can check it
389 * when we are done.
390 */
391 for (i = 0; i < UFS_NDADDR; i++) {
392 bn = DIP(oip, db[i]);
393 DIP_ASSIGN(oip, db[i], blks[i]);
394 blks[i] = bn;
395 }
396 for (i = 0; i < UFS_NIADDR; i++) {
397 bn = DIP(oip, ib[i]);
398 DIP_ASSIGN(oip, ib[i], blks[UFS_NDADDR + i]);
399 blks[UFS_NDADDR + i] = bn;
400 }
401
402 oip->i_size = osize;
403 DIP_ASSIGN(oip, size, osize);
404 error = vtruncbuf(ovp, lastblock + 1, 0, 0);
405 if (error && !allerror)
406 allerror = error;
407
408 /*
409 * Indirect blocks first.
410 */
411 indir_lbn[SINGLE] = -UFS_NDADDR;
412 indir_lbn[DOUBLE] = indir_lbn[SINGLE] - FFS_NINDIR(fs) - 1;
413 indir_lbn[TRIPLE] = indir_lbn[DOUBLE] - FFS_NINDIR(fs) * FFS_NINDIR(fs) - 1;
414 for (level = TRIPLE; level >= SINGLE; level--) {
415 if (oip->i_ump->um_fstype == UFS1)
416 bn = ufs_rw32(oip->i_ffs1_ib[level],UFS_FSNEEDSWAP(fs));
417 else
418 bn = ufs_rw64(oip->i_ffs2_ib[level],UFS_FSNEEDSWAP(fs));
419 if (bn != 0) {
420 error = ffs_indirtrunc(oip, indir_lbn[level],
421 FFS_FSBTODB(fs, bn), lastiblock[level], level, &count);
422 if (error)
423 allerror = error;
424 blocksreleased += count;
425 if (lastiblock[level] < 0) {
426 DIP_ASSIGN(oip, ib[level], 0);
427 if (oip->i_ump->um_mountp->mnt_wapbl) {
428 UFS_WAPBL_REGISTER_DEALLOCATION(
429 oip->i_ump->um_mountp,
430 FFS_FSBTODB(fs, bn), fs->fs_bsize);
431 } else
432 ffs_blkfree(fs, oip->i_devvp, bn,
433 fs->fs_bsize, oip->i_number);
434 blocksreleased += nblocks;
435 }
436 }
437 if (lastiblock[level] >= 0)
438 goto done;
439 }
440
441 /*
442 * All whole direct blocks or frags.
443 */
444 for (i = UFS_NDADDR - 1; i > lastblock; i--) {
445 long bsize;
446
447 if (oip->i_ump->um_fstype == UFS1)
448 bn = ufs_rw32(oip->i_ffs1_db[i], UFS_FSNEEDSWAP(fs));
449 else
450 bn = ufs_rw64(oip->i_ffs2_db[i], UFS_FSNEEDSWAP(fs));
451 if (bn == 0)
452 continue;
453 DIP_ASSIGN(oip, db[i], 0);
454 bsize = ffs_blksize(fs, oip, i);
455 if ((oip->i_ump->um_mountp->mnt_wapbl) &&
456 (ovp->v_type != VREG)) {
457 UFS_WAPBL_REGISTER_DEALLOCATION(oip->i_ump->um_mountp,
458 FFS_FSBTODB(fs, bn), bsize);
459 } else
460 ffs_blkfree(fs, oip->i_devvp, bn, bsize, oip->i_number);
461 blocksreleased += btodb(bsize);
462 }
463 if (lastblock < 0)
464 goto done;
465
466 /*
467 * Finally, look for a change in size of the
468 * last direct block; release any frags.
469 */
470 if (oip->i_ump->um_fstype == UFS1)
471 bn = ufs_rw32(oip->i_ffs1_db[lastblock], UFS_FSNEEDSWAP(fs));
472 else
473 bn = ufs_rw64(oip->i_ffs2_db[lastblock], UFS_FSNEEDSWAP(fs));
474 if (bn != 0) {
475 long oldspace, newspace;
476
477 /*
478 * Calculate amount of space we're giving
479 * back as old block size minus new block size.
480 */
481 oldspace = ffs_blksize(fs, oip, lastblock);
482 oip->i_size = length;
483 DIP_ASSIGN(oip, size, length);
484 newspace = ffs_blksize(fs, oip, lastblock);
485 if (newspace == 0)
486 panic("itrunc: newspace");
487 if (oldspace - newspace > 0) {
488 /*
489 * Block number of space to be free'd is
490 * the old block # plus the number of frags
491 * required for the storage we're keeping.
492 */
493 bn += ffs_numfrags(fs, newspace);
494 if ((oip->i_ump->um_mountp->mnt_wapbl) &&
495 (ovp->v_type != VREG)) {
496 UFS_WAPBL_REGISTER_DEALLOCATION(
497 oip->i_ump->um_mountp, FFS_FSBTODB(fs, bn),
498 oldspace - newspace);
499 } else
500 ffs_blkfree(fs, oip->i_devvp, bn,
501 oldspace - newspace, oip->i_number);
502 blocksreleased += btodb(oldspace - newspace);
503 }
504 }
505
506 done:
507 #ifdef DIAGNOSTIC
508 for (level = SINGLE; level <= TRIPLE; level++)
509 if (blks[UFS_NDADDR + level] != DIP(oip, ib[level]))
510 panic("itrunc1");
511 for (i = 0; i < UFS_NDADDR; i++)
512 if (blks[i] != DIP(oip, db[i]))
513 panic("itrunc2");
514 if (length == 0 &&
515 (!LIST_EMPTY(&ovp->v_cleanblkhd) || !LIST_EMPTY(&ovp->v_dirtyblkhd)))
516 panic("itrunc3");
517 #endif /* DIAGNOSTIC */
518 /*
519 * Put back the real size.
520 */
521 oip->i_size = length;
522 DIP_ASSIGN(oip, size, length);
523 DIP_ADD(oip, blocks, -blocksreleased);
524 genfs_node_unlock(ovp);
525 oip->i_flag |= IN_CHANGE;
526 UFS_WAPBL_UPDATE(ovp, NULL, NULL, 0);
527 #if defined(QUOTA) || defined(QUOTA2)
528 (void) chkdq(oip, -blocksreleased, NOCRED, 0);
529 #endif
530 KASSERT(ovp->v_type != VREG || ovp->v_size == oip->i_size);
531 return (allerror);
532 }
533
534 /*
535 * Release blocks associated with the inode ip and stored in the indirect
536 * block bn. Blocks are free'd in LIFO order up to (but not including)
537 * lastbn. If level is greater than SINGLE, the block is an indirect block
538 * and recursive calls to indirtrunc must be used to cleanse other indirect
539 * blocks.
540 *
541 * NB: triple indirect blocks are untested.
542 */
543 static int
ffs_indirtrunc(struct inode * ip,daddr_t lbn,daddr_t dbn,daddr_t lastbn,int level,int64_t * countp)544 ffs_indirtrunc(struct inode *ip, daddr_t lbn, daddr_t dbn, daddr_t lastbn,
545 int level, int64_t *countp)
546 {
547 int i;
548 struct buf *bp;
549 struct fs *fs = ip->i_fs;
550 int32_t *bap1 = NULL;
551 int64_t *bap2 = NULL;
552 struct vnode *vp;
553 daddr_t nb, nlbn, last;
554 char *copy = NULL;
555 int64_t blkcount, factor, blocksreleased = 0;
556 int nblocks;
557 int error = 0, allerror = 0;
558 const int needswap = UFS_FSNEEDSWAP(fs);
559 #define RBAP(ip, i) (((ip)->i_ump->um_fstype == UFS1) ? \
560 ufs_rw32(bap1[i], needswap) : ufs_rw64(bap2[i], needswap))
561 #define BAP_ASSIGN(ip, i, value) \
562 do { \
563 if ((ip)->i_ump->um_fstype == UFS1) \
564 bap1[i] = (value); \
565 else \
566 bap2[i] = (value); \
567 } while(0)
568
569 /*
570 * Calculate index in current block of last
571 * block to be kept. -1 indicates the entire
572 * block so we need not calculate the index.
573 */
574 factor = 1;
575 for (i = SINGLE; i < level; i++)
576 factor *= FFS_NINDIR(fs);
577 last = lastbn;
578 if (lastbn > 0)
579 last /= factor;
580 nblocks = btodb(fs->fs_bsize);
581 /*
582 * Get buffer of block pointers, zero those entries corresponding
583 * to blocks to be free'd, and update on disk copy first. Since
584 * double(triple) indirect before single(double) indirect, calls
585 * to bmap on these blocks will fail. However, we already have
586 * the on disk address, so we have to set the b_blkno field
587 * explicitly instead of letting bread do everything for us.
588 */
589 vp = ITOV(ip);
590 error = ffs_getblk(vp, lbn, FFS_NOBLK, fs->fs_bsize, false, &bp);
591 if (error) {
592 *countp = 0;
593 return error;
594 }
595 if (bp->b_oflags & (BO_DONE | BO_DELWRI)) {
596 /* Braces must be here in case trace evaluates to nothing. */
597 trace(TR_BREADHIT, pack(vp, fs->fs_bsize), lbn);
598 } else {
599 trace(TR_BREADMISS, pack(vp, fs->fs_bsize), lbn);
600 curlwp->l_ru.ru_inblock++; /* pay for read */
601 bp->b_flags |= B_READ;
602 bp->b_flags &= ~B_COWDONE; /* we change blkno below */
603 if (bp->b_bcount > bp->b_bufsize)
604 panic("ffs_indirtrunc: bad buffer size");
605 bp->b_blkno = dbn;
606 BIO_SETPRIO(bp, BPRIO_TIMECRITICAL);
607 VOP_STRATEGY(vp, bp);
608 error = biowait(bp);
609 if (error == 0)
610 error = fscow_run(bp, true);
611 }
612 if (error) {
613 brelse(bp, 0);
614 *countp = 0;
615 return (error);
616 }
617
618 if (ip->i_ump->um_fstype == UFS1)
619 bap1 = (int32_t *)bp->b_data;
620 else
621 bap2 = (int64_t *)bp->b_data;
622 if (lastbn >= 0) {
623 copy = kmem_alloc(fs->fs_bsize, KM_SLEEP);
624 memcpy((void *)copy, bp->b_data, (u_int)fs->fs_bsize);
625 for (i = last + 1; i < FFS_NINDIR(fs); i++)
626 BAP_ASSIGN(ip, i, 0);
627 error = bwrite(bp);
628 if (error)
629 allerror = error;
630 if (ip->i_ump->um_fstype == UFS1)
631 bap1 = (int32_t *)copy;
632 else
633 bap2 = (int64_t *)copy;
634 }
635
636 /*
637 * Recursively free totally unused blocks.
638 */
639 for (i = FFS_NINDIR(fs) - 1, nlbn = lbn + 1 - i * factor; i > last;
640 i--, nlbn += factor) {
641 nb = RBAP(ip, i);
642 if (nb == 0)
643 continue;
644 if (level > SINGLE) {
645 error = ffs_indirtrunc(ip, nlbn, FFS_FSBTODB(fs, nb),
646 (daddr_t)-1, level - 1,
647 &blkcount);
648 if (error)
649 allerror = error;
650 blocksreleased += blkcount;
651 }
652 if ((ip->i_ump->um_mountp->mnt_wapbl) &&
653 ((level > SINGLE) || (ITOV(ip)->v_type != VREG))) {
654 UFS_WAPBL_REGISTER_DEALLOCATION(ip->i_ump->um_mountp,
655 FFS_FSBTODB(fs, nb), fs->fs_bsize);
656 } else
657 ffs_blkfree(fs, ip->i_devvp, nb, fs->fs_bsize,
658 ip->i_number);
659 blocksreleased += nblocks;
660 }
661
662 /*
663 * Recursively free last partial block.
664 */
665 if (level > SINGLE && lastbn >= 0) {
666 last = lastbn % factor;
667 nb = RBAP(ip, i);
668 if (nb != 0) {
669 error = ffs_indirtrunc(ip, nlbn, FFS_FSBTODB(fs, nb),
670 last, level - 1, &blkcount);
671 if (error)
672 allerror = error;
673 blocksreleased += blkcount;
674 }
675 }
676
677 if (copy != NULL) {
678 kmem_free(copy, fs->fs_bsize);
679 } else {
680 brelse(bp, BC_INVAL);
681 }
682
683 *countp = blocksreleased;
684 return (allerror);
685 }
686
687 void
ffs_itimes(struct inode * ip,const struct timespec * acc,const struct timespec * mod,const struct timespec * cre)688 ffs_itimes(struct inode *ip, const struct timespec *acc,
689 const struct timespec *mod, const struct timespec *cre)
690 {
691 struct timespec now;
692
693 if (!(ip->i_flag & (IN_ACCESS | IN_CHANGE | IN_UPDATE | IN_MODIFY))) {
694 return;
695 }
696
697 vfs_timestamp(&now);
698 if (ip->i_flag & IN_ACCESS) {
699 if (acc == NULL)
700 acc = &now;
701 DIP_ASSIGN(ip, atime, acc->tv_sec);
702 DIP_ASSIGN(ip, atimensec, acc->tv_nsec);
703 }
704 if (ip->i_flag & (IN_UPDATE | IN_MODIFY)) {
705 if ((ip->i_flags & SF_SNAPSHOT) == 0) {
706 if (mod == NULL)
707 mod = &now;
708 DIP_ASSIGN(ip, mtime, mod->tv_sec);
709 DIP_ASSIGN(ip, mtimensec, mod->tv_nsec);
710 }
711 ip->i_modrev++;
712 }
713 if (ip->i_flag & (IN_CHANGE | IN_MODIFY)) {
714 if (cre == NULL)
715 cre = &now;
716 DIP_ASSIGN(ip, ctime, cre->tv_sec);
717 DIP_ASSIGN(ip, ctimensec, cre->tv_nsec);
718 }
719 if (ip->i_flag & (IN_ACCESS | IN_MODIFY))
720 ip->i_flag |= IN_ACCESSED;
721 if (ip->i_flag & (IN_UPDATE | IN_CHANGE))
722 ip->i_flag |= IN_MODIFIED;
723 ip->i_flag &= ~(IN_ACCESS | IN_CHANGE | IN_UPDATE | IN_MODIFY);
724 }
725