ffs_balloc.c revision 1.25.2.1
1 /* $NetBSD: ffs_balloc.c,v 1.25.2.1 2001/10/01 12:48:21 fvdl Exp $ */ 2 3 /* 4 * Copyright (c) 1982, 1986, 1989, 1993 5 * The Regents of the University of California. All rights reserved. 6 * 7 * Redistribution and use in source and binary forms, with or without 8 * modification, are permitted provided that the following conditions 9 * are met: 10 * 1. Redistributions of source code must retain the above copyright 11 * notice, this list of conditions and the following disclaimer. 12 * 2. Redistributions in binary form must reproduce the above copyright 13 * notice, this list of conditions and the following disclaimer in the 14 * documentation and/or other materials provided with the distribution. 15 * 3. All advertising materials mentioning features or use of this software 16 * must display the following acknowledgement: 17 * This product includes software developed by the University of 18 * California, Berkeley and its contributors. 19 * 4. Neither the name of the University nor the names of its contributors 20 * may be used to endorse or promote products derived from this software 21 * without specific prior written permission. 22 * 23 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND 24 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE 25 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE 26 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE 27 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL 28 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS 29 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) 30 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT 31 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY 32 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF 33 * SUCH DAMAGE. 34 * 35 * @(#)ffs_balloc.c 8.8 (Berkeley) 6/16/95 36 */ 37 38 #if defined(_KERNEL_OPT) 39 #include "opt_quota.h" 40 #endif 41 42 #include <sys/param.h> 43 #include <sys/systm.h> 44 #include <sys/buf.h> 45 #include <sys/proc.h> 46 #include <sys/file.h> 47 #include <sys/mount.h> 48 #include <sys/vnode.h> 49 #include <sys/mount.h> 50 51 #include <ufs/ufs/quota.h> 52 #include <ufs/ufs/ufsmount.h> 53 #include <ufs/ufs/inode.h> 54 #include <ufs/ufs/ufs_extern.h> 55 #include <ufs/ufs/ufs_bswap.h> 56 57 #include <ufs/ffs/fs.h> 58 #include <ufs/ffs/ffs_extern.h> 59 60 #include <uvm/uvm.h> 61 62 /* 63 * Balloc defines the structure of file system storage 64 * by allocating the physical blocks on a device given 65 * the inode and the logical block number in a file. 66 */ 67 int 68 ffs_balloc(v) 69 void *v; 70 { 71 struct vop_balloc_args /* { 72 struct vnode *a_vp; 73 off_t a_startoffset; 74 int a_size; 75 struct ucred *a_cred; 76 int a_flags; 77 struct buf **a_bpp; 78 } */ *ap = v; 79 ufs_daddr_t lbn; 80 int size; 81 struct ucred *cred; 82 int flags; 83 ufs_daddr_t nb; 84 struct buf *bp, *nbp; 85 struct vnode *vp = ap->a_vp; 86 struct inode *ip = VTOI(vp); 87 struct fs *fs = ip->i_fs; 88 struct indir indirs[NIADDR + 2]; 89 ufs_daddr_t newb, *bap, pref; 90 int deallocated, osize, nsize, num, i, error; 91 ufs_daddr_t *allocib, *blkp, *allocblk, allociblk[NIADDR + 1]; 92 int unwindidx = -1; 93 struct buf **bpp = ap->a_bpp; 94 off_t off; 95 #ifdef FFS_EI 96 const int needswap = UFS_FSNEEDSWAP(fs); 97 #endif 98 UVMHIST_FUNC("ffs_balloc"); UVMHIST_CALLED(ubchist); 99 100 lbn = lblkno(fs, ap->a_startoffset); 101 size = blkoff(fs, ap->a_startoffset) + ap->a_size; 102 if (size > fs->fs_bsize) 103 panic("ffs_balloc: blk too big"); 104 if (bpp != NULL) { 105 *bpp = NULL; 106 } 107 UVMHIST_LOG(ubchist, "vp %p lbn 0x%x size 0x%x", vp, lbn, size,0); 108 109 KASSERT(size <= fs->fs_bsize); 110 if (lbn < 0) 111 return (EFBIG); 112 cred = ap->a_cred; 113 flags = ap->a_flags; 114 115 /* 116 * If the next write will extend the file into a new block, 117 * and the file is currently composed of a fragment 118 * this fragment has to be extended to be a full block. 119 */ 120 121 nb = lblkno(fs, ip->i_ffs_size); 122 if (nb < NDADDR && nb < lbn) { 123 osize = blksize(fs, ip, nb); 124 if (osize < fs->fs_bsize && osize > 0) { 125 error = ffs_realloccg(ip, nb, 126 ffs_blkpref(ip, nb, (int)nb, &ip->i_ffs_db[0]), 127 osize, (int)fs->fs_bsize, cred, bpp, &newb); 128 if (error) 129 return (error); 130 if (DOINGSOFTDEP(vp)) 131 softdep_setup_allocdirect(ip, nb, newb, 132 ufs_rw32(ip->i_ffs_db[nb], needswap), 133 fs->fs_bsize, osize, bpp ? *bpp : NULL); 134 ip->i_ffs_size = lblktosize(fs, nb + 1); 135 uvm_vnp_setsize(vp, ip->i_ffs_size); 136 ip->i_ffs_db[nb] = ufs_rw32(newb, needswap); 137 ip->i_flag |= IN_CHANGE | IN_UPDATE; 138 if (bpp) { 139 if (flags & B_SYNC) 140 bwrite(*bpp); 141 else 142 bawrite(*bpp); 143 } 144 } 145 } 146 147 /* 148 * The first NDADDR blocks are direct blocks 149 */ 150 151 if (lbn < NDADDR) { 152 nb = ufs_rw32(ip->i_ffs_db[lbn], needswap); 153 if (nb != 0 && ip->i_ffs_size >= lblktosize(fs, lbn + 1)) { 154 155 /* 156 * The block is an already-allocated direct block 157 * and the file already extends past this block, 158 * thus this must be a whole block. 159 * Just read the block (if requested). 160 */ 161 162 if (bpp != NULL) { 163 error = bread(vp, lbn, fs->fs_bsize, NOCRED, 164 bpp); 165 if (error) { 166 brelse(*bpp); 167 return (error); 168 } 169 } 170 return (0); 171 } 172 if (nb != 0) { 173 174 /* 175 * Consider need to reallocate a fragment. 176 */ 177 178 osize = fragroundup(fs, blkoff(fs, ip->i_ffs_size)); 179 nsize = fragroundup(fs, size); 180 if (nsize <= osize) { 181 182 /* 183 * The existing block is already 184 * at least as big as we want. 185 * Just read the block (if requested). 186 */ 187 188 if (bpp != NULL) { 189 error = bread(vp, lbn, osize, NOCRED, 190 bpp); 191 if (error) { 192 brelse(*bpp); 193 return (error); 194 } 195 } 196 return 0; 197 } else { 198 199 /* 200 * The existing block is smaller than we want, 201 * grow it. 202 */ 203 204 error = ffs_realloccg(ip, lbn, 205 ffs_blkpref(ip, lbn, (int)lbn, 206 &ip->i_ffs_db[0]), osize, nsize, cred, 207 bpp, &newb); 208 if (error) 209 return (error); 210 if (DOINGSOFTDEP(vp)) 211 softdep_setup_allocdirect(ip, lbn, 212 newb, nb, nsize, osize, 213 bpp ? *bpp : NULL); 214 } 215 } else { 216 217 /* 218 * the block was not previously allocated, 219 * allocate a new block or fragment. 220 */ 221 222 if (ip->i_ffs_size < lblktosize(fs, lbn + 1)) 223 nsize = fragroundup(fs, size); 224 else 225 nsize = fs->fs_bsize; 226 error = ffs_alloc(ip, lbn, 227 ffs_blkpref(ip, lbn, (int)lbn, &ip->i_ffs_db[0]), 228 nsize, cred, &newb); 229 if (error) 230 return (error); 231 if (bpp != NULL) { 232 bp = getblk(vp, lbn, nsize, 0, 0); 233 bp->b_blkno = fsbtodb(fs, newb); 234 if (flags & B_CLRBUF) 235 clrbuf(bp); 236 *bpp = bp; 237 } 238 if (DOINGSOFTDEP(vp)) { 239 softdep_setup_allocdirect(ip, lbn, newb, 0, 240 nsize, 0, bpp ? *bpp : NULL); 241 } 242 } 243 ip->i_ffs_db[lbn] = ufs_rw32(newb, needswap); 244 ip->i_flag |= IN_CHANGE | IN_UPDATE; 245 return (0); 246 } 247 /* 248 * Determine the number of levels of indirection. 249 */ 250 pref = 0; 251 if ((error = ufs_getlbns(vp, lbn, indirs, &num)) != 0) 252 return(error); 253 254 #ifdef DIAGNOSTIC 255 if (num < 1) 256 panic ("ffs_balloc: ufs_bmaparray returned indirect block\n"); 257 #endif 258 /* 259 * Fetch the first indirect block allocating if necessary. 260 */ 261 --num; 262 nb = ufs_rw32(ip->i_ffs_ib[indirs[0].in_off], needswap); 263 allocib = NULL; 264 allocblk = allociblk; 265 if (nb == 0) { 266 pref = ffs_blkpref(ip, lbn, 0, (ufs_daddr_t *)0); 267 error = ffs_alloc(ip, lbn, pref, (int)fs->fs_bsize, cred, 268 &newb); 269 if (error) 270 goto fail; 271 nb = newb; 272 *allocblk++ = nb; 273 bp = getblk(vp, indirs[1].in_lbn, fs->fs_bsize, 0, 0); 274 bp->b_blkno = fsbtodb(fs, nb); 275 clrbuf(bp); 276 if (DOINGSOFTDEP(vp)) { 277 softdep_setup_allocdirect(ip, NDADDR + indirs[0].in_off, 278 newb, 0, fs->fs_bsize, 0, bp); 279 bdwrite(bp); 280 } else { 281 /* 282 * Write synchronously so that indirect blocks 283 * never point at garbage. 284 */ 285 if ((error = bwrite(bp)) != 0) 286 goto fail; 287 } 288 unwindidx = 0; 289 allocib = &ip->i_ffs_ib[indirs[0].in_off]; 290 *allocib = ufs_rw32(nb, needswap); 291 ip->i_flag |= IN_CHANGE | IN_UPDATE; 292 } 293 /* 294 * Fetch through the indirect blocks, allocating as necessary. 295 */ 296 for (i = 1;;) { 297 error = bread(vp, 298 indirs[i].in_lbn, (int)fs->fs_bsize, NOCRED, &bp); 299 if (error) { 300 brelse(bp); 301 goto fail; 302 } 303 bap = (ufs_daddr_t *)bp->b_data; 304 nb = ufs_rw32(bap[indirs[i].in_off], needswap); 305 if (i == num) 306 break; 307 i++; 308 if (nb != 0) { 309 brelse(bp); 310 continue; 311 } 312 if (pref == 0) 313 pref = ffs_blkpref(ip, lbn, 0, (ufs_daddr_t *)0); 314 error = ffs_alloc(ip, lbn, pref, (int)fs->fs_bsize, cred, 315 &newb); 316 if (error) { 317 brelse(bp); 318 goto fail; 319 } 320 nb = newb; 321 *allocblk++ = nb; 322 nbp = getblk(vp, indirs[i].in_lbn, fs->fs_bsize, 0, 0); 323 nbp->b_blkno = fsbtodb(fs, nb); 324 clrbuf(nbp); 325 if (DOINGSOFTDEP(vp)) { 326 softdep_setup_allocindir_meta(nbp, ip, bp, 327 indirs[i - 1].in_off, nb); 328 bdwrite(nbp); 329 } else { 330 /* 331 * Write synchronously so that indirect blocks 332 * never point at garbage. 333 */ 334 if ((error = bwrite(nbp)) != 0) { 335 brelse(bp); 336 goto fail; 337 } 338 } 339 if (unwindidx < 0) 340 unwindidx = i - 1; 341 bap[indirs[i - 1].in_off] = ufs_rw32(nb, needswap); 342 /* 343 * If required, write synchronously, otherwise use 344 * delayed write. 345 */ 346 if (flags & B_SYNC) { 347 bwrite(bp); 348 } else { 349 bdwrite(bp); 350 } 351 } 352 /* 353 * Get the data block, allocating if necessary. 354 */ 355 if (nb == 0) { 356 pref = ffs_blkpref(ip, lbn, indirs[num].in_off, &bap[0]); 357 error = ffs_alloc(ip, lbn, pref, (int)fs->fs_bsize, cred, 358 &newb); 359 if (error) { 360 brelse(bp); 361 goto fail; 362 } 363 nb = newb; 364 *allocblk++ = nb; 365 if (bpp != NULL) { 366 nbp = getblk(vp, lbn, fs->fs_bsize, 0, 0); 367 nbp->b_blkno = fsbtodb(fs, nb); 368 if (flags & B_CLRBUF) 369 clrbuf(nbp); 370 *bpp = nbp; 371 } 372 if (DOINGSOFTDEP(vp)) 373 softdep_setup_allocindir_page(ip, lbn, bp, 374 indirs[num].in_off, nb, 0, bpp ? *bpp : NULL); 375 bap[indirs[num].in_off] = ufs_rw32(nb, needswap); 376 if (allocib == NULL && unwindidx < 0) { 377 unwindidx = i - 1; 378 } 379 /* 380 * If required, write synchronously, otherwise use 381 * delayed write. 382 */ 383 if (flags & B_SYNC) { 384 bwrite(bp); 385 } else { 386 bdwrite(bp); 387 } 388 return (0); 389 } 390 brelse(bp); 391 if (bpp != NULL) { 392 if (flags & B_CLRBUF) { 393 error = bread(vp, lbn, (int)fs->fs_bsize, NOCRED, &nbp); 394 if (error) { 395 brelse(nbp); 396 goto fail; 397 } 398 } else { 399 nbp = getblk(vp, lbn, fs->fs_bsize, 0, 0); 400 nbp->b_blkno = fsbtodb(fs, nb); 401 clrbuf(nbp); 402 } 403 *bpp = nbp; 404 } 405 return (0); 406 407 fail: 408 409 /* 410 * Restore the UVM state to what the rest of the FFS code is 411 * expecting. Unbusy any pages that we allocated and left busy up in 412 * ufs_balloc_range(). the following VOP_FSYNC() will try to busy 413 * those pages again, which would deadlock if they are still busy 414 * from before. After this we're back to a state where we can undo 415 * any partial allocation. 416 */ 417 418 simple_lock(&vp->v_uobj.vmobjlock); 419 for (off = ap->a_startoffset; off < ap->a_startoffset + fs->fs_bsize; 420 off += PAGE_SIZE) { 421 struct vm_page *pg; 422 423 pg = uvm_pagelookup(&vp->v_uobj, off); 424 if (pg == NULL) { 425 break; 426 } 427 uvm_pageactivate(pg); 428 KASSERT((pg->flags & PG_FAKE) == 0); 429 pg->flags &= ~(PG_BUSY); 430 UVM_PAGE_OWN(pg, NULL); 431 } 432 simple_unlock(&vp->v_uobj.vmobjlock); 433 434 /* 435 * If we have failed part way through block allocation, we 436 * have to deallocate any indirect blocks that we have allocated. 437 * We have to fsync the file before we start to get rid of all 438 * of its dependencies so that we do not leave them dangling. 439 * We have to sync it at the end so that the soft updates code 440 * does not find any untracked changes. Although this is really 441 * slow, running out of disk space is not expected to be a common 442 * occurence. The error return from fsync is ignored as we already 443 * have an error to return to the user. 444 */ 445 446 (void) VOP_FSYNC(vp, cred, FSYNC_WAIT, 0, 0, curproc); 447 for (deallocated = 0, blkp = allociblk; blkp < allocblk; blkp++) { 448 ffs_blkfree(ip, *blkp, fs->fs_bsize); 449 deallocated += fs->fs_bsize; 450 } 451 if (unwindidx >= 0) { 452 if (unwindidx == 0) { 453 *allocib = 0; 454 } else { 455 int r; 456 457 r = bread(vp, indirs[unwindidx].in_lbn, 458 (int)fs->fs_bsize, NOCRED, &bp); 459 if (r) { 460 panic("Could not unwind indirect block, error %d", r); 461 brelse(bp); 462 } else { 463 bap = (ufs_daddr_t *)bp->b_data; 464 bap[indirs[unwindidx].in_off] = 0; 465 if (flags & B_SYNC) 466 bwrite(bp); 467 else 468 bdwrite(bp); 469 } 470 } 471 for (i = unwindidx + 1; i <= num; i++) { 472 bp = getblk(vp, indirs[i].in_lbn, (int)fs->fs_bsize, 0, 473 0); 474 bp->b_flags |= B_INVAL; 475 brelse(bp); 476 } 477 } 478 if (deallocated) { 479 #ifdef QUOTA 480 /* 481 * Restore user's disk quota because allocation failed. 482 */ 483 (void)chkdq(ip, (long)-btodb(deallocated), cred, FORCE); 484 #endif 485 ip->i_ffs_blocks -= btodb(deallocated); 486 ip->i_flag |= IN_CHANGE | IN_UPDATE; 487 } 488 (void) VOP_FSYNC(vp, cred, FSYNC_WAIT, 0, 0, curproc); 489 return (error); 490 } 491 492 493 int 494 ffs_gop_alloc(struct vnode *vp, off_t off, off_t len, int flags, 495 struct ucred *cred) 496 { 497 struct inode *ip = VTOI(vp); 498 struct fs *fs = ip->i_fs; 499 int error, delta, bshift, bsize; 500 UVMHIST_FUNC("ffs_gop_alloc"); UVMHIST_CALLED(ubchist); 501 502 error = 0; 503 bshift = fs->fs_bshift; 504 bsize = 1 << bshift; 505 506 delta = off & (bsize - 1); 507 off -= delta; 508 len += delta; 509 510 while (len > 0) { 511 bsize = MIN(bsize, len); 512 513 error = VOP_BALLOC(vp, off, bsize, cred, flags, NULL); 514 if (error) { 515 goto out; 516 } 517 518 /* 519 * increase file size now, VOP_BALLOC() requires that 520 * EOF be up-to-date before each call. 521 */ 522 523 if (ip->i_ffs_size < off + bsize) { 524 UVMHIST_LOG(ubchist, "vp %p old 0x%x new 0x%x", 525 vp, ip->i_ffs_size, off + bsize, 0); 526 ip->i_ffs_size = off + bsize; 527 } 528 529 off += bsize; 530 len -= bsize; 531 } 532 533 out: 534 return error; 535 } 536
Indexes created Sun Oct 12 22:09:51 GMT 2025