ffs_inode.c revision 1.124.14.3
1 /* $NetBSD: ffs_inode.c,v 1.124.14.3 2020/04/21 18:42:45 martin 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 1.124.14.3 2020/04/21 18:42:45 martin 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 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 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], oldblks[UFS_NDADDR + UFS_NIADDR]; 213 struct fs *fs; 214 int extblocks; 215 int offset, pgoffset, level; 216 int64_t blocksreleased = 0, datablocks; 217 int i, aflag, nblocks; 218 int error, allerror = 0; 219 off_t osize; 220 int sync; 221 struct ufsmount *ump = oip->i_ump; 222 void *dcookie; 223 224 UFS_WAPBL_JLOCK_ASSERT(ump->um_mountp); 225 226 if (ovp->v_type == VCHR || ovp->v_type == VBLK || 227 ovp->v_type == VFIFO || ovp->v_type == VSOCK) { 228 KASSERT(oip->i_size == 0); 229 return 0; 230 } 231 232 if (length < 0) 233 return (EINVAL); 234 235 fs = oip->i_fs; 236 #define i_din2 i_din.ffs2_din 237 extblocks = 0; 238 datablocks = DIP(oip, blocks); 239 if (fs->fs_magic == FS_UFS2_MAGIC && oip->i_din2->di_extsize > 0) { 240 extblocks = btodb(ffs_fragroundup(fs, oip->i_din2->di_extsize)); 241 datablocks -= extblocks; 242 } 243 if ((ioflag & IO_EXT) && extblocks > 0) { 244 if (length != 0) 245 panic("ffs_truncate: partial trunc of extdata"); 246 { 247 #ifdef QUOTA 248 (void) chkdq(oip, -extblocks, NOCRED, FORCE); 249 #endif 250 vinvalbuf(ovp, 0, cred, curlwp, 0, 0); 251 osize = oip->i_din2->di_extsize; 252 oip->i_din2->di_blocks -= extblocks; 253 oip->i_din2->di_extsize = 0; 254 for (i = 0; i < UFS_NXADDR; i++) { 255 oldblks[i] = oip->i_din2->di_extb[i]; 256 oip->i_din2->di_extb[i] = 0; 257 } 258 oip->i_flag |= IN_CHANGE; 259 if ((error = ffs_update(ovp, NULL, NULL, 0))) 260 return (error); 261 for (i = 0; i < UFS_NXADDR; i++) { 262 if (oldblks[i] == 0) 263 continue; 264 ffs_blkfree(fs, oip->i_devvp, oldblks[i], 265 ffs_sblksize(fs, osize, i), oip->i_number); 266 } 267 extblocks = 0; 268 } 269 } 270 if (ovp->v_type == VLNK && 271 (oip->i_size < ump->um_maxsymlinklen || 272 (ump->um_maxsymlinklen == 0 && datablocks == 0))) { 273 KDASSERT(length == 0); 274 memset(SHORTLINK(oip), 0, (size_t)oip->i_size); 275 oip->i_size = 0; 276 DIP_ASSIGN(oip, size, 0); 277 oip->i_flag |= IN_CHANGE | IN_UPDATE; 278 return (ffs_update(ovp, NULL, NULL, 0)); 279 } 280 if (oip->i_size == length) { 281 /* still do a uvm_vnp_setsize() as writesize may be larger */ 282 uvm_vnp_setsize(ovp, length); 283 oip->i_flag |= IN_CHANGE | IN_UPDATE; 284 return (ffs_update(ovp, NULL, NULL, 0)); 285 } 286 if (length > ump->um_maxfilesize) 287 return (EFBIG); 288 289 if ((oip->i_flags & SF_SNAPSHOT) != 0) 290 ffs_snapremove(ovp); 291 292 osize = oip->i_size; 293 aflag = ioflag & IO_SYNC ? B_SYNC : 0; 294 295 /* 296 * Lengthen the size of the file. We must ensure that the 297 * last byte of the file is allocated. Since the smallest 298 * value of osize is 0, length will be at least 1. 299 */ 300 301 if (osize < length) { 302 if (ffs_lblkno(fs, osize) < UFS_NDADDR && 303 ffs_lblkno(fs, osize) != ffs_lblkno(fs, length) && 304 ffs_blkroundup(fs, osize) != osize) { 305 off_t eob; 306 307 eob = ffs_blkroundup(fs, osize); 308 uvm_vnp_setwritesize(ovp, eob); 309 error = ufs_balloc_range(ovp, osize, eob - osize, 310 cred, aflag); 311 if (error) { 312 (void) ffs_truncate(ovp, osize, 313 ioflag & IO_SYNC, cred); 314 return error; 315 } 316 if (ioflag & IO_SYNC) { 317 rw_enter(ovp->v_uobj.vmobjlock, RW_WRITER); 318 VOP_PUTPAGES(ovp, 319 trunc_page(osize & fs->fs_bmask), 320 round_page(eob), PGO_CLEANIT | PGO_SYNCIO | 321 PGO_JOURNALLOCKED); 322 } 323 } 324 uvm_vnp_setwritesize(ovp, length); 325 error = ufs_balloc_range(ovp, length - 1, 1, cred, aflag); 326 if (error) { 327 (void) ffs_truncate(ovp, osize, ioflag & IO_SYNC, cred); 328 return (error); 329 } 330 uvm_vnp_setsize(ovp, length); 331 oip->i_flag |= IN_CHANGE | IN_UPDATE; 332 KASSERT(ovp->v_size == oip->i_size); 333 return (ffs_update(ovp, NULL, NULL, 0)); 334 } 335 336 /* 337 * When truncating a regular file down to a non-block-aligned size, 338 * we must zero the part of last block which is past the new EOF. 339 * We must synchronously flush the zeroed pages to disk 340 * since the new pages will be invalidated as soon as we 341 * inform the VM system of the new, smaller size. 342 * We must do this before acquiring the GLOCK, since fetching 343 * the pages will acquire the GLOCK internally. 344 * So there is a window where another thread could see a whole 345 * zeroed page past EOF, but that's life. 346 */ 347 348 offset = ffs_blkoff(fs, length); 349 pgoffset = length & PAGE_MASK; 350 if (ovp->v_type == VREG && (pgoffset != 0 || offset != 0) && 351 osize > length) { 352 daddr_t lbn; 353 voff_t eoz; 354 int size; 355 356 if (offset != 0) { 357 error = ufs_balloc_range(ovp, length - 1, 1, cred, 358 aflag); 359 if (error) 360 return error; 361 } 362 lbn = ffs_lblkno(fs, length); 363 size = ffs_blksize(fs, oip, lbn); 364 eoz = MIN(MAX(ffs_lblktosize(fs, lbn) + size, round_page(pgoffset)), 365 osize); 366 ubc_zerorange(&ovp->v_uobj, length, eoz - length, 367 UBC_UNMAP_FLAG(ovp)); 368 if (round_page(eoz) > round_page(length)) { 369 rw_enter(ovp->v_uobj.vmobjlock, RW_WRITER); 370 error = VOP_PUTPAGES(ovp, round_page(length), 371 round_page(eoz), 372 PGO_CLEANIT | PGO_DEACTIVATE | PGO_JOURNALLOCKED | 373 ((ioflag & IO_SYNC) ? PGO_SYNCIO : 0)); 374 if (error) 375 return error; 376 } 377 } 378 379 genfs_node_wrlock(ovp); 380 oip->i_size = length; 381 DIP_ASSIGN(oip, size, length); 382 uvm_vnp_setsize(ovp, length); 383 /* 384 * Calculate index into inode's block list of 385 * last direct and indirect blocks (if any) 386 * which we want to keep. Lastblock is -1 when 387 * the file is truncated to 0. 388 */ 389 lastblock = ffs_lblkno(fs, length + fs->fs_bsize - 1) - 1; 390 lastiblock[SINGLE] = lastblock - UFS_NDADDR; 391 lastiblock[DOUBLE] = lastiblock[SINGLE] - FFS_NINDIR(fs); 392 lastiblock[TRIPLE] = lastiblock[DOUBLE] - FFS_NINDIR(fs) * FFS_NINDIR(fs); 393 nblocks = btodb(fs->fs_bsize); 394 /* 395 * Update file and block pointers on disk before we start freeing 396 * blocks. If we crash before free'ing blocks below, the blocks 397 * will be returned to the free list. lastiblock values are also 398 * normalized to -1 for calls to ffs_indirtrunc below. 399 */ 400 sync = 0; 401 for (level = TRIPLE; level >= SINGLE; level--) { 402 blks[UFS_NDADDR + level] = DIP(oip, ib[level]); 403 if (lastiblock[level] < 0 && blks[UFS_NDADDR + level] != 0) { 404 sync = 1; 405 DIP_ASSIGN(oip, ib[level], 0); 406 lastiblock[level] = -1; 407 } 408 } 409 for (i = 0; i < UFS_NDADDR; i++) { 410 blks[i] = DIP(oip, db[i]); 411 if (i > lastblock && blks[i] != 0) { 412 sync = 1; 413 DIP_ASSIGN(oip, db[i], 0); 414 } 415 } 416 oip->i_flag |= IN_CHANGE | IN_UPDATE; 417 if (sync) { 418 error = ffs_update(ovp, NULL, NULL, UPDATE_WAIT); 419 if (error && !allerror) 420 allerror = error; 421 } 422 423 /* 424 * Having written the new inode to disk, save its new configuration 425 * and put back the old block pointers long enough to process them. 426 * Note that we save the new block configuration so we can check it 427 * when we are done. 428 */ 429 for (i = 0; i < UFS_NDADDR; i++) { 430 bn = DIP(oip, db[i]); 431 DIP_ASSIGN(oip, db[i], blks[i]); 432 blks[i] = bn; 433 } 434 for (i = 0; i < UFS_NIADDR; i++) { 435 bn = DIP(oip, ib[i]); 436 DIP_ASSIGN(oip, ib[i], blks[UFS_NDADDR + i]); 437 blks[UFS_NDADDR + i] = bn; 438 } 439 440 oip->i_size = osize; 441 DIP_ASSIGN(oip, size, osize); 442 error = vtruncbuf(ovp, lastblock + 1, 0, 0); 443 if (error && !allerror) 444 allerror = error; 445 446 /* 447 * Indirect blocks first. 448 */ 449 indir_lbn[SINGLE] = -UFS_NDADDR; 450 indir_lbn[DOUBLE] = indir_lbn[SINGLE] - FFS_NINDIR(fs) - 1; 451 indir_lbn[TRIPLE] = indir_lbn[DOUBLE] - FFS_NINDIR(fs) * FFS_NINDIR(fs) - 1; 452 for (level = TRIPLE; level >= SINGLE; level--) { 453 bn = ffs_getib(fs, oip, level); 454 if (bn != 0) { 455 if (lastiblock[level] < 0 && 456 oip->i_ump->um_mountp->mnt_wapbl) { 457 error = UFS_WAPBL_REGISTER_DEALLOCATION( 458 oip->i_ump->um_mountp, 459 FFS_FSBTODB(fs, bn), fs->fs_bsize, 460 &dcookie); 461 if (error) 462 goto out; 463 } else { 464 dcookie = NULL; 465 } 466 467 error = ffs_indirtrunc(oip, indir_lbn[level], 468 FFS_FSBTODB(fs, bn), lastiblock[level], level, 469 &blocksreleased); 470 if (error) { 471 if (dcookie) { 472 UFS_WAPBL_UNREGISTER_DEALLOCATION( 473 oip->i_ump->um_mountp, dcookie); 474 } 475 goto out; 476 } 477 478 if (lastiblock[level] < 0) { 479 if (!dcookie) 480 ffs_blkfree(fs, oip->i_devvp, bn, 481 fs->fs_bsize, oip->i_number); 482 DIP_ASSIGN(oip, ib[level], 0); 483 blocksreleased += nblocks; 484 } 485 } 486 if (lastiblock[level] >= 0) 487 goto done; 488 } 489 490 /* 491 * All whole direct blocks or frags. 492 */ 493 for (i = UFS_NDADDR - 1; i > lastblock; i--) { 494 long bsize; 495 496 bn = ffs_getdb(fs, oip, i); 497 if (bn == 0) 498 continue; 499 500 bsize = ffs_blksize(fs, oip, i); 501 if ((oip->i_ump->um_mountp->mnt_wapbl) && 502 (ovp->v_type != VREG)) { 503 error = UFS_WAPBL_REGISTER_DEALLOCATION( 504 oip->i_ump->um_mountp, 505 FFS_FSBTODB(fs, bn), bsize, NULL); 506 if (error) 507 goto out; 508 } else 509 ffs_blkfree(fs, oip->i_devvp, bn, bsize, oip->i_number); 510 DIP_ASSIGN(oip, db[i], 0); 511 blocksreleased += btodb(bsize); 512 } 513 if (lastblock < 0) 514 goto done; 515 516 /* 517 * Finally, look for a change in size of the 518 * last direct block; release any frags. 519 */ 520 bn = ffs_getdb(fs, oip, lastblock); 521 if (bn != 0) { 522 long oldspace, newspace; 523 524 /* 525 * Calculate amount of space we're giving 526 * back as old block size minus new block size. 527 */ 528 oldspace = ffs_blksize(fs, oip, lastblock); 529 oip->i_size = length; 530 DIP_ASSIGN(oip, size, length); 531 newspace = ffs_blksize(fs, oip, lastblock); 532 if (newspace == 0) 533 panic("itrunc: newspace"); 534 if (oldspace - newspace > 0) { 535 /* 536 * Block number of space to be free'd is 537 * the old block # plus the number of frags 538 * required for the storage we're keeping. 539 */ 540 bn += ffs_numfrags(fs, newspace); 541 if ((oip->i_ump->um_mountp->mnt_wapbl) && 542 (ovp->v_type != VREG)) { 543 error = UFS_WAPBL_REGISTER_DEALLOCATION( 544 oip->i_ump->um_mountp, FFS_FSBTODB(fs, bn), 545 oldspace - newspace, NULL); 546 if (error) 547 goto out; 548 } else 549 ffs_blkfree(fs, oip->i_devvp, bn, 550 oldspace - newspace, oip->i_number); 551 blocksreleased += btodb(oldspace - newspace); 552 } 553 } 554 555 done: 556 for (level = SINGLE; level <= TRIPLE; level++) 557 KASSERTMSG((blks[UFS_NDADDR + level] == DIP(oip, ib[level])), 558 "itrunc1 blk mismatch: %jx != %jx", 559 (uintmax_t)blks[UFS_NDADDR + level], 560 (uintmax_t)DIP(oip, ib[level])); 561 for (i = 0; i < UFS_NDADDR; i++) 562 KASSERTMSG((blks[i] == DIP(oip, db[i])), 563 "itrunc2 blk mismatch: %jx != %jx", 564 (uintmax_t)blks[i], (uintmax_t)DIP(oip, db[i])); 565 KASSERTMSG((length != 0 || extblocks || LIST_EMPTY(&ovp->v_cleanblkhd)), 566 "itrunc3: zero length and nonempty cleanblkhd"); 567 KASSERTMSG((length != 0 || extblocks || LIST_EMPTY(&ovp->v_dirtyblkhd)), 568 "itrunc3: zero length and nonempty dirtyblkhd"); 569 570 out: 571 /* 572 * Set length back to old size if deallocation failed. Some indirect 573 * blocks were deallocated creating a hole, but that is okay. 574 */ 575 if (error == EAGAIN) { 576 if (!allerror) 577 allerror = error; 578 length = osize; 579 uvm_vnp_setsize(ovp, length); 580 } 581 582 /* 583 * Put back the real size. 584 */ 585 oip->i_size = length; 586 DIP_ASSIGN(oip, size, length); 587 DIP_ADD(oip, blocks, -blocksreleased); 588 genfs_node_unlock(ovp); 589 oip->i_flag |= IN_CHANGE; 590 UFS_WAPBL_UPDATE(ovp, NULL, NULL, 0); 591 #if defined(QUOTA) || defined(QUOTA2) 592 (void) chkdq(oip, -blocksreleased, NOCRED, 0); 593 #endif 594 KASSERT(ovp->v_type != VREG || ovp->v_size == oip->i_size); 595 return (allerror); 596 } 597 598 /* 599 * Release blocks associated with the inode ip and stored in the indirect 600 * block bn. Blocks are free'd in LIFO order up to (but not including) 601 * lastbn. If level is greater than SINGLE, the block is an indirect block 602 * and recursive calls to indirtrunc must be used to cleanse other indirect 603 * blocks. 604 * 605 * NB: triple indirect blocks are untested. 606 */ 607 static int 608 ffs_indirtrunc(struct inode *ip, daddr_t lbn, daddr_t dbn, daddr_t lastbn, 609 int level, int64_t *countp) 610 { 611 int i; 612 struct buf *bp; 613 struct fs *fs = ip->i_fs; 614 int32_t *bap1 = NULL; 615 int64_t *bap2 = NULL; 616 struct vnode *vp; 617 daddr_t nb, nlbn, last; 618 char *copy = NULL; 619 int64_t factor; 620 int64_t nblocks; 621 int error = 0, allerror = 0; 622 const int needswap = UFS_FSNEEDSWAP(fs); 623 const int wapbl = (ip->i_ump->um_mountp->mnt_wapbl != NULL); 624 void *dcookie; 625 626 #define RBAP(ip, i) (((ip)->i_ump->um_fstype == UFS1) ? \ 627 ufs_rw32(bap1[i], needswap) : ufs_rw64(bap2[i], needswap)) 628 #define BAP_ASSIGN(ip, i, value) \ 629 do { \ 630 if ((ip)->i_ump->um_fstype == UFS1) \ 631 bap1[i] = (value); \ 632 else \ 633 bap2[i] = (value); \ 634 } while(0) 635 636 /* 637 * Calculate index in current block of last 638 * block to be kept. -1 indicates the entire 639 * block so we need not calculate the index. 640 */ 641 factor = 1; 642 for (i = SINGLE; i < level; i++) 643 factor *= FFS_NINDIR(fs); 644 last = lastbn; 645 if (lastbn > 0) 646 last /= factor; 647 nblocks = btodb(fs->fs_bsize); 648 /* 649 * Get buffer of block pointers, zero those entries corresponding 650 * to blocks to be free'd, and update on disk copy first. Since 651 * double(triple) indirect before single(double) indirect, calls 652 * to bmap on these blocks will fail. However, we already have 653 * the on disk address, so we have to set the b_blkno field 654 * explicitly instead of letting bread do everything for us. 655 */ 656 vp = ITOV(ip); 657 error = ffs_getblk(vp, lbn, FFS_NOBLK, fs->fs_bsize, false, &bp); 658 if (error) 659 return error; 660 661 if (bp->b_oflags & (BO_DONE | BO_DELWRI)) { 662 /* Braces must be here in case trace evaluates to nothing. */ 663 trace(TR_BREADHIT, pack(vp, fs->fs_bsize), lbn); 664 } else { 665 trace(TR_BREADMISS, pack(vp, fs->fs_bsize), lbn); 666 curlwp->l_ru.ru_inblock++; /* pay for read */ 667 bp->b_flags |= B_READ; 668 bp->b_flags &= ~B_COWDONE; /* we change blkno below */ 669 if (bp->b_bcount > bp->b_bufsize) 670 panic("ffs_indirtrunc: bad buffer size"); 671 bp->b_blkno = dbn; 672 BIO_SETPRIO(bp, BPRIO_TIMECRITICAL); 673 VOP_STRATEGY(vp, bp); 674 error = biowait(bp); 675 if (error == 0) 676 error = fscow_run(bp, true); 677 } 678 if (error) { 679 brelse(bp, 0); 680 return error; 681 } 682 683 /* 684 * Clear reference to blocks to be removed on disk, before actually 685 * reclaiming them, so that fsck is more likely to be able to recover 686 * the filesystem if system goes down during the truncate process. 687 * This assumes the truncate process would not fail, contrary 688 * to the wapbl case. 689 */ 690 if (ip->i_ump->um_fstype == UFS1) 691 bap1 = (int32_t *)bp->b_data; 692 else 693 bap2 = (int64_t *)bp->b_data; 694 if (lastbn >= 0 && !wapbl) { 695 copy = kmem_alloc(fs->fs_bsize, KM_SLEEP); 696 memcpy((void *)copy, bp->b_data, (u_int)fs->fs_bsize); 697 for (i = last + 1; i < FFS_NINDIR(fs); i++) 698 BAP_ASSIGN(ip, i, 0); 699 error = bwrite(bp); 700 if (error) 701 allerror = error; 702 703 if (ip->i_ump->um_fstype == UFS1) 704 bap1 = (int32_t *)copy; 705 else 706 bap2 = (int64_t *)copy; 707 } 708 709 /* 710 * Recursively free totally unused blocks. 711 */ 712 for (i = FFS_NINDIR(fs) - 1, nlbn = lbn + 1 - i * factor; i > last; 713 i--, nlbn += factor) { 714 nb = RBAP(ip, i); 715 if (nb == 0) 716 continue; 717 718 if ((ip->i_ump->um_mountp->mnt_wapbl) && 719 ((level > SINGLE) || (ITOV(ip)->v_type != VREG))) { 720 error = UFS_WAPBL_REGISTER_DEALLOCATION( 721 ip->i_ump->um_mountp, 722 FFS_FSBTODB(fs, nb), fs->fs_bsize, 723 &dcookie); 724 if (error) 725 goto out; 726 } else { 727 dcookie = NULL; 728 } 729 730 if (level > SINGLE) { 731 error = ffs_indirtrunc(ip, nlbn, FFS_FSBTODB(fs, nb), 732 (daddr_t)-1, level - 1, countp); 733 if (error) { 734 if (dcookie) { 735 UFS_WAPBL_UNREGISTER_DEALLOCATION( 736 ip->i_ump->um_mountp, dcookie); 737 } 738 739 goto out; 740 } 741 } 742 743 if (!dcookie) 744 ffs_blkfree(fs, ip->i_devvp, nb, fs->fs_bsize, 745 ip->i_number); 746 747 BAP_ASSIGN(ip, i, 0); 748 *countp += nblocks; 749 } 750 751 /* 752 * Recursively free blocks on the now last partial indirect block. 753 */ 754 if (level > SINGLE && lastbn >= 0) { 755 last = lastbn % factor; 756 nb = RBAP(ip, i); 757 if (nb != 0) { 758 error = ffs_indirtrunc(ip, nlbn, FFS_FSBTODB(fs, nb), 759 last, level - 1, countp); 760 if (error) 761 goto out; 762 } 763 } 764 765 out: 766 if (error && !allerror) 767 allerror = error; 768 769 if (copy != NULL) { 770 kmem_free(copy, fs->fs_bsize); 771 } else if (lastbn < 0 && error == 0) { 772 /* all freed, release without writing back */ 773 brelse(bp, BC_INVAL); 774 } else if (wapbl) { 775 /* only partially freed, write the updated block */ 776 error = bwrite(bp); 777 if (!allerror) 778 allerror = error; 779 } 780 781 return (allerror); 782 } 783 784 void 785 ffs_itimes(struct inode *ip, const struct timespec *acc, 786 const struct timespec *mod, const struct timespec *cre) 787 { 788 struct timespec now; 789 790 if (!(ip->i_flag & (IN_ACCESS | IN_CHANGE | IN_UPDATE | IN_MODIFY))) { 791 return; 792 } 793 794 vfs_timestamp(&now); 795 if (ip->i_flag & IN_ACCESS) { 796 if (acc == NULL) 797 acc = &now; 798 DIP_ASSIGN(ip, atime, acc->tv_sec); 799 DIP_ASSIGN(ip, atimensec, acc->tv_nsec); 800 } 801 if (ip->i_flag & (IN_UPDATE | IN_MODIFY)) { 802 if ((ip->i_flags & SF_SNAPSHOT) == 0) { 803 if (mod == NULL) 804 mod = &now; 805 DIP_ASSIGN(ip, mtime, mod->tv_sec); 806 DIP_ASSIGN(ip, mtimensec, mod->tv_nsec); 807 } 808 ip->i_modrev++; 809 } 810 if (ip->i_flag & (IN_CHANGE | IN_MODIFY)) { 811 if (cre == NULL) 812 cre = &now; 813 DIP_ASSIGN(ip, ctime, cre->tv_sec); 814 DIP_ASSIGN(ip, ctimensec, cre->tv_nsec); 815 } 816 if (ip->i_flag & (IN_ACCESS | IN_MODIFY)) 817 ip->i_flag |= IN_ACCESSED; 818 if (ip->i_flag & (IN_UPDATE | IN_CHANGE)) 819 ip->i_flag |= IN_MODIFIED; 820 ip->i_flag &= ~(IN_ACCESS | IN_CHANGE | IN_UPDATE | IN_MODIFY); 821 } 822
Indexes created Wed Oct 15 01:09:54 GMT 2025