ffs_balloc.c revision 1.23.2.5
1 /* $NetBSD: ffs_balloc.c,v 1.23.2.5 2001/10/08 20:11:51 nathanw 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/lwp.h> 46 #include <sys/proc.h> 47 #include <sys/file.h> 48 #include <sys/mount.h> 49 #include <sys/vnode.h> 50 #include <sys/mount.h> 51 52 #include <ufs/ufs/quota.h> 53 #include <ufs/ufs/ufsmount.h> 54 #include <ufs/ufs/inode.h> 55 #include <ufs/ufs/ufs_extern.h> 56 #include <ufs/ufs/ufs_bswap.h> 57 58 #include <ufs/ffs/fs.h> 59 #include <ufs/ffs/ffs_extern.h> 60 61 #include <uvm/uvm.h> 62 63 /* 64 * Balloc defines the structure of file system storage 65 * by allocating the physical blocks on a device given 66 * the inode and the logical block number in a file. 67 */ 68 int 69 ffs_balloc(v) 70 void *v; 71 { 72 struct vop_balloc_args /* { 73 struct vnode *a_vp; 74 off_t a_startoffset; 75 int a_size; 76 struct ucred *a_cred; 77 int a_flags; 78 struct buf **a_bpp; 79 } */ *ap = v; 80 ufs_daddr_t lbn; 81 int size; 82 struct ucred *cred; 83 int flags; 84 ufs_daddr_t nb; 85 struct buf *bp, *nbp; 86 struct vnode *vp = ap->a_vp; 87 struct inode *ip = VTOI(vp); 88 struct fs *fs = ip->i_fs; 89 struct indir indirs[NIADDR + 2]; 90 ufs_daddr_t newb, *bap, pref; 91 int deallocated, osize, nsize, num, i, error; 92 ufs_daddr_t *allocib, *blkp, *allocblk, allociblk[NIADDR + 1]; 93 int unwindidx = -1; 94 struct buf **bpp = ap->a_bpp; 95 off_t off; 96 #ifdef FFS_EI 97 const int needswap = UFS_FSNEEDSWAP(fs); 98 #endif 99 UVMHIST_FUNC("ffs_balloc"); UVMHIST_CALLED(ubchist); 100 101 lbn = lblkno(fs, ap->a_startoffset); 102 size = blkoff(fs, ap->a_startoffset) + ap->a_size; 103 if (size > fs->fs_bsize) 104 panic("ffs_balloc: blk too big"); 105 if (bpp != NULL) { 106 *bpp = NULL; 107 } 108 UVMHIST_LOG(ubchist, "vp %p lbn 0x%x size 0x%x", vp, lbn, size,0); 109 110 KASSERT(size <= fs->fs_bsize); 111 if (lbn < 0) 112 return (EFBIG); 113 cred = ap->a_cred; 114 flags = ap->a_flags; 115 116 /* 117 * If the next write will extend the file into a new block, 118 * and the file is currently composed of a fragment 119 * this fragment has to be extended to be a full block. 120 */ 121 122 nb = lblkno(fs, ip->i_ffs_size); 123 if (nb < NDADDR && nb < lbn) { 124 osize = blksize(fs, ip, nb); 125 if (osize < fs->fs_bsize && osize > 0) { 126 error = ffs_realloccg(ip, nb, 127 ffs_blkpref(ip, nb, (int)nb, &ip->i_ffs_db[0]), 128 osize, (int)fs->fs_bsize, cred, bpp, &newb); 129 if (error) 130 return (error); 131 if (DOINGSOFTDEP(vp)) 132 softdep_setup_allocdirect(ip, nb, newb, 133 ufs_rw32(ip->i_ffs_db[nb], needswap), 134 fs->fs_bsize, osize, bpp ? *bpp : NULL); 135 ip->i_ffs_size = lblktosize(fs, nb + 1); 136 uvm_vnp_setsize(vp, ip->i_ffs_size); 137 ip->i_ffs_db[nb] = ufs_rw32(newb, needswap); 138 ip->i_flag |= IN_CHANGE | IN_UPDATE; 139 if (bpp) { 140 if (flags & B_SYNC) 141 bwrite(*bpp); 142 else 143 bawrite(*bpp); 144 } 145 } 146 } 147 148 /* 149 * The first NDADDR blocks are direct blocks 150 */ 151 152 if (lbn < NDADDR) { 153 nb = ufs_rw32(ip->i_ffs_db[lbn], needswap); 154 if (nb != 0 && ip->i_ffs_size >= lblktosize(fs, lbn + 1)) { 155 156 /* 157 * The block is an already-allocated direct block 158 * and the file already extends past this block, 159 * thus this must be a whole block. 160 * Just read the block (if requested). 161 */ 162 163 if (bpp != NULL) { 164 error = bread(vp, lbn, fs->fs_bsize, NOCRED, 165 bpp); 166 if (error) { 167 brelse(*bpp); 168 return (error); 169 } 170 } 171 return (0); 172 } 173 if (nb != 0) { 174 175 /* 176 * Consider need to reallocate a fragment. 177 */ 178 179 osize = fragroundup(fs, blkoff(fs, ip->i_ffs_size)); 180 nsize = fragroundup(fs, size); 181 if (nsize <= osize) { 182 183 /* 184 * The existing block is already 185 * at least as big as we want. 186 * Just read the block (if requested). 187 */ 188 189 if (bpp != NULL) { 190 error = bread(vp, lbn, osize, NOCRED, 191 bpp); 192 if (error) { 193 brelse(*bpp); 194 return (error); 195 } 196 } 197 return 0; 198 } else { 199 200 /* 201 * The existing block is smaller than we want, 202 * grow it. 203 */ 204 205 error = ffs_realloccg(ip, lbn, 206 ffs_blkpref(ip, lbn, (int)lbn, 207 &ip->i_ffs_db[0]), osize, nsize, cred, 208 bpp, &newb); 209 if (error) 210 return (error); 211 if (DOINGSOFTDEP(vp)) 212 softdep_setup_allocdirect(ip, lbn, 213 newb, nb, nsize, osize, 214 bpp ? *bpp : NULL); 215 } 216 } else { 217 218 /* 219 * the block was not previously allocated, 220 * allocate a new block or fragment. 221 */ 222 223 if (ip->i_ffs_size < lblktosize(fs, lbn + 1)) 224 nsize = fragroundup(fs, size); 225 else 226 nsize = fs->fs_bsize; 227 error = ffs_alloc(ip, lbn, 228 ffs_blkpref(ip, lbn, (int)lbn, &ip->i_ffs_db[0]), 229 nsize, cred, &newb); 230 if (error) 231 return (error); 232 if (bpp != NULL) { 233 bp = getblk(vp, lbn, nsize, 0, 0); 234 bp->b_blkno = fsbtodb(fs, newb); 235 if (flags & B_CLRBUF) 236 clrbuf(bp); 237 *bpp = bp; 238 } 239 if (DOINGSOFTDEP(vp)) { 240 softdep_setup_allocdirect(ip, lbn, newb, 0, 241 nsize, 0, bpp ? *bpp : NULL); 242 } 243 } 244 ip->i_ffs_db[lbn] = ufs_rw32(newb, needswap); 245 ip->i_flag |= IN_CHANGE | IN_UPDATE; 246 return (0); 247 } 248 /* 249 * Determine the number of levels of indirection. 250 */ 251 pref = 0; 252 if ((error = ufs_getlbns(vp, lbn, indirs, &num)) != 0) 253 return(error); 254 255 #ifdef DIAGNOSTIC 256 if (num < 1) 257 panic ("ffs_balloc: ufs_bmaparray returned indirect block\n"); 258 #endif 259 /* 260 * Fetch the first indirect block allocating if necessary. 261 */ 262 --num; 263 nb = ufs_rw32(ip->i_ffs_ib[indirs[0].in_off], needswap); 264 allocib = NULL; 265 allocblk = allociblk; 266 if (nb == 0) { 267 pref = ffs_blkpref(ip, lbn, 0, (ufs_daddr_t *)0); 268 error = ffs_alloc(ip, lbn, pref, (int)fs->fs_bsize, cred, 269 &newb); 270 if (error) 271 goto fail; 272 nb = newb; 273 *allocblk++ = nb; 274 bp = getblk(vp, indirs[1].in_lbn, fs->fs_bsize, 0, 0); 275 bp->b_blkno = fsbtodb(fs, nb); 276 clrbuf(bp); 277 if (DOINGSOFTDEP(vp)) { 278 softdep_setup_allocdirect(ip, NDADDR + indirs[0].in_off, 279 newb, 0, fs->fs_bsize, 0, bp); 280 bdwrite(bp); 281 } else { 282 /* 283 * Write synchronously so that indirect blocks 284 * never point at garbage. 285 */ 286 if ((error = bwrite(bp)) != 0) 287 goto fail; 288 } 289 unwindidx = 0; 290 allocib = &ip->i_ffs_ib[indirs[0].in_off]; 291 *allocib = ufs_rw32(nb, needswap); 292 ip->i_flag |= IN_CHANGE | IN_UPDATE; 293 } 294 /* 295 * Fetch through the indirect blocks, allocating as necessary. 296 */ 297 for (i = 1;;) { 298 error = bread(vp, 299 indirs[i].in_lbn, (int)fs->fs_bsize, NOCRED, &bp); 300 if (error) { 301 brelse(bp); 302 goto fail; 303 } 304 bap = (ufs_daddr_t *)bp->b_data; 305 nb = ufs_rw32(bap[indirs[i].in_off], needswap); 306 if (i == num) 307 break; 308 i++; 309 if (nb != 0) { 310 brelse(bp); 311 continue; 312 } 313 if (pref == 0) 314 pref = ffs_blkpref(ip, lbn, 0, (ufs_daddr_t *)0); 315 error = ffs_alloc(ip, lbn, pref, (int)fs->fs_bsize, cred, 316 &newb); 317 if (error) { 318 brelse(bp); 319 goto fail; 320 } 321 nb = newb; 322 *allocblk++ = nb; 323 nbp = getblk(vp, indirs[i].in_lbn, fs->fs_bsize, 0, 0); 324 nbp->b_blkno = fsbtodb(fs, nb); 325 clrbuf(nbp); 326 if (DOINGSOFTDEP(vp)) { 327 softdep_setup_allocindir_meta(nbp, ip, bp, 328 indirs[i - 1].in_off, nb); 329 bdwrite(nbp); 330 } else { 331 /* 332 * Write synchronously so that indirect blocks 333 * never point at garbage. 334 */ 335 if ((error = bwrite(nbp)) != 0) { 336 brelse(bp); 337 goto fail; 338 } 339 } 340 if (unwindidx < 0) 341 unwindidx = i - 1; 342 bap[indirs[i - 1].in_off] = ufs_rw32(nb, needswap); 343 /* 344 * If required, write synchronously, otherwise use 345 * delayed write. 346 */ 347 if (flags & B_SYNC) { 348 bwrite(bp); 349 } else { 350 bdwrite(bp); 351 } 352 } 353 /* 354 * Get the data block, allocating if necessary. 355 */ 356 if (nb == 0) { 357 pref = ffs_blkpref(ip, lbn, indirs[num].in_off, &bap[0]); 358 error = ffs_alloc(ip, lbn, pref, (int)fs->fs_bsize, cred, 359 &newb); 360 if (error) { 361 brelse(bp); 362 goto fail; 363 } 364 nb = newb; 365 *allocblk++ = nb; 366 if (bpp != NULL) { 367 nbp = getblk(vp, lbn, fs->fs_bsize, 0, 0); 368 nbp->b_blkno = fsbtodb(fs, nb); 369 if (flags & B_CLRBUF) 370 clrbuf(nbp); 371 *bpp = nbp; 372 } 373 if (DOINGSOFTDEP(vp)) 374 softdep_setup_allocindir_page(ip, lbn, bp, 375 indirs[num].in_off, nb, 0, bpp ? *bpp : NULL); 376 bap[indirs[num].in_off] = ufs_rw32(nb, needswap); 377 if (allocib == NULL && unwindidx < 0) { 378 unwindidx = i - 1; 379 } 380 /* 381 * If required, write synchronously, otherwise use 382 * delayed write. 383 */ 384 if (flags & B_SYNC) { 385 bwrite(bp); 386 } else { 387 bdwrite(bp); 388 } 389 return (0); 390 } 391 brelse(bp); 392 if (bpp != NULL) { 393 if (flags & B_CLRBUF) { 394 error = bread(vp, lbn, (int)fs->fs_bsize, NOCRED, &nbp); 395 if (error) { 396 brelse(nbp); 397 goto fail; 398 } 399 } else { 400 nbp = getblk(vp, lbn, fs->fs_bsize, 0, 0); 401 nbp->b_blkno = fsbtodb(fs, nb); 402 clrbuf(nbp); 403 } 404 *bpp = nbp; 405 } 406 return (0); 407 408 fail: 409 410 /* 411 * Restore the UVM state to what the rest of the FFS code is 412 * expecting. Unbusy any pages that we allocated and left busy up in 413 * ufs_balloc_range(). the following VOP_FSYNC() will try to busy 414 * those pages again, which would deadlock if they are still busy 415 * from before. After this we're back to a state where we can undo 416 * any partial allocation. 417 */ 418 419 simple_lock(&vp->v_uobj.vmobjlock); 420 for (off = ap->a_startoffset; off < ap->a_startoffset + fs->fs_bsize; 421 off += PAGE_SIZE) { 422 struct vm_page *pg; 423 424 pg = uvm_pagelookup(&vp->v_uobj, off); 425 if (pg == NULL) { 426 break; 427 } 428 uvm_pageactivate(pg); 429 KASSERT((pg->flags & PG_FAKE) == 0); 430 pg->flags &= ~(PG_BUSY); 431 UVM_PAGE_OWN(pg, NULL); 432 } 433 simple_unlock(&vp->v_uobj.vmobjlock); 434 435 /* 436 * If we have failed part way through block allocation, we 437 * have to deallocate any indirect blocks that we have allocated. 438 * We have to fsync the file before we start to get rid of all 439 * of its dependencies so that we do not leave them dangling. 440 * We have to sync it at the end so that the soft updates code 441 * does not find any untracked changes. Although this is really 442 * slow, running out of disk space is not expected to be a common 443 * occurence. The error return from fsync is ignored as we already 444 * have an error to return to the user. 445 */ 446 447 (void) VOP_FSYNC(vp, cred, FSYNC_WAIT, 0, 0, curproc->l_proc); 448 for (deallocated = 0, blkp = allociblk; blkp < allocblk; blkp++) { 449 ffs_blkfree(ip, *blkp, fs->fs_bsize); 450 deallocated += fs->fs_bsize; 451 } 452 if (unwindidx >= 0) { 453 if (unwindidx == 0) { 454 *allocib = 0; 455 } else { 456 int r; 457 458 r = bread(vp, indirs[unwindidx].in_lbn, 459 (int)fs->fs_bsize, NOCRED, &bp); 460 if (r) { 461 panic("Could not unwind indirect block, error %d", r); 462 brelse(bp); 463 } else { 464 bap = (ufs_daddr_t *)bp->b_data; 465 bap[indirs[unwindidx].in_off] = 0; 466 if (flags & B_SYNC) 467 bwrite(bp); 468 else 469 bdwrite(bp); 470 } 471 } 472 for (i = unwindidx + 1; i <= num; i++) { 473 bp = getblk(vp, indirs[i].in_lbn, (int)fs->fs_bsize, 0, 474 0); 475 bp->b_flags |= B_INVAL; 476 brelse(bp); 477 } 478 } 479 if (deallocated) { 480 #ifdef QUOTA 481 /* 482 * Restore user's disk quota because allocation failed. 483 */ 484 (void)chkdq(ip, (long)-btodb(deallocated), cred, FORCE); 485 #endif 486 ip->i_ffs_blocks -= btodb(deallocated); 487 ip->i_flag |= IN_CHANGE | IN_UPDATE; 488 } 489 (void) VOP_FSYNC(vp, cred, FSYNC_WAIT, 0, 0, curproc->l_proc); 490 return (error); 491 } 492 493 494 int 495 ffs_gop_alloc(struct vnode *vp, off_t off, off_t len, int flags, 496 struct ucred *cred) 497 { 498 struct inode *ip = VTOI(vp); 499 struct fs *fs = ip->i_fs; 500 int error, delta, bshift, bsize; 501 UVMHIST_FUNC("ffs_gop_alloc"); UVMHIST_CALLED(ubchist); 502 503 error = 0; 504 bshift = fs->fs_bshift; 505 bsize = 1 << bshift; 506 507 delta = off & (bsize - 1); 508 off -= delta; 509 len += delta; 510 511 while (len > 0) { 512 bsize = MIN(bsize, len); 513 514 error = VOP_BALLOC(vp, off, bsize, cred, flags, NULL); 515 if (error) { 516 goto out; 517 } 518 519 /* 520 * increase file size now, VOP_BALLOC() requires that 521 * EOF be up-to-date before each call. 522 */ 523 524 if (ip->i_ffs_size < off + bsize) { 525 UVMHIST_LOG(ubchist, "vp %p old 0x%x new 0x%x", 526 vp, ip->i_ffs_size, off + bsize, 0); 527 ip->i_ffs_size = off + bsize; 528 } 529 530 off += bsize; 531 len -= bsize; 532 } 533 534 out: 535 return error; 536 } 537
Indexes created Tue Oct 14 15:09:51 GMT 2025