1 /* 2 * Copyright (c) 1982, 1986, 1989, 1993 3 * The Regents of the University of California. All rights reserved. 4 * 5 * Redistribution and use in source and binary forms, with or without 6 * modification, are permitted provided that the following conditions 7 * are met: 8 * 1. Redistributions of source code must retain the above copyright 9 * notice, this list of conditions and the following disclaimer. 10 * 2. Redistributions in binary form must reproduce the above copyright 11 * notice, this list of conditions and the following disclaimer in the 12 * documentation and/or other materials provided with the distribution. 13 * 3. All advertising materials mentioning features or use of this software 14 * must display the following acknowledgement: 15 * This product includes software developed by the University of 16 * California, Berkeley and its contributors. 17 * 4. Neither the name of the University nor the names of its contributors 18 * may be used to endorse or promote products derived from this software 19 * without specific prior written permission. 20 * 21 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND 22 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE 23 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE 24 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE 25 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL 26 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS 27 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) 28 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT 29 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY 30 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF 31 * SUCH DAMAGE. 32 * 33 * @(#)ffs_balloc.c 8.8 (Berkeley) 6/16/95 34 * $FreeBSD$ 35 */ 36 37 #include <sys/param.h> 38 #include <sys/systm.h> 39 #include <sys/bio.h> 40 #include <sys/buf.h> 41 #include <sys/lock.h> 42 #include <sys/mount.h> 43 #include <sys/vnode.h> 44 45 #include <ufs/ufs/quota.h> 46 #include <ufs/ufs/inode.h> 47 #include <ufs/ufs/ufs_extern.h> 48 49 #include <ufs/ffs/fs.h> 50 #include <ufs/ffs/ffs_extern.h> 51 52 /* 53 * Balloc defines the structure of file system storage 54 * by allocating the physical blocks on a device given 55 * the inode and the logical block number in a file. 56 */ 57 int 58 ffs_balloc(struct vnode *a_vp, off_t a_startoffset, int a_size, 59 struct ucred *a_cred, int a_flags, struct buf **a_bpp) 60 { 61 struct inode *ip; 62 ufs_daddr_t lbn; 63 int size; 64 struct ucred *cred; 65 int flags; 66 struct fs *fs; 67 ufs_daddr_t nb; 68 struct buf *bp, *nbp; 69 struct vnode *vp; 70 struct indir indirs[NIADDR + 2]; 71 ufs_daddr_t newb, *bap, pref; 72 int deallocated, osize, nsize, num, i, error; 73 ufs_daddr_t *allocib, *blkp, *allocblk, allociblk[NIADDR + 1]; 74 int unwindidx = -1; 75 struct thread *td = curthread; /* XXX */ 76 77 vp = a_vp; 78 ip = VTOI(vp); 79 fs = ip->i_fs; 80 lbn = lblkno(fs, a_startoffset); 81 size = blkoff(fs, a_startoffset) + a_size; 82 if (size > fs->fs_bsize) 83 panic("ffs_balloc: blk too big"); 84 *a_bpp = NULL; 85 if (lbn < 0) 86 return (EFBIG); 87 cred = a_cred; 88 flags = a_flags; 89 90 /* 91 * If the next write will extend the file into a new block, 92 * and the file is currently composed of a fragment 93 * this fragment has to be extended to be a full block. 94 */ 95 nb = lblkno(fs, ip->i_size); 96 if (nb < NDADDR && nb < lbn) { 97 osize = blksize(fs, ip, nb); 98 if (osize < fs->fs_bsize && osize > 0) { 99 error = ffs_realloccg(ip, nb, 100 ffs_blkpref(ip, nb, (int)nb, &ip->i_db[0]), 101 osize, (int)fs->fs_bsize, cred, &bp); 102 if (error) 103 return (error); 104 if (DOINGSOFTDEP(vp)) 105 softdep_setup_allocdirect(ip, nb, 106 dbtofsb(fs, bp->b_blkno), ip->i_db[nb], 107 fs->fs_bsize, osize, bp); 108 ip->i_size = smalllblktosize(fs, nb + 1); 109 ip->i_db[nb] = dbtofsb(fs, bp->b_blkno); 110 ip->i_flag |= IN_CHANGE | IN_UPDATE; 111 if (flags & B_SYNC) 112 bwrite(bp); 113 else 114 bawrite(bp); 115 } 116 } 117 /* 118 * The first NDADDR blocks are direct blocks 119 */ 120 if (lbn < NDADDR) { 121 if (flags & B_METAONLY) 122 panic("ffs_balloc: B_METAONLY for direct block"); 123 nb = ip->i_db[lbn]; 124 if (nb != 0 && ip->i_size >= smalllblktosize(fs, lbn + 1)) { 125 error = bread(vp, lbn, fs->fs_bsize, NOCRED, &bp); 126 if (error) { 127 brelse(bp); 128 return (error); 129 } 130 bp->b_blkno = fsbtodb(fs, nb); 131 *a_bpp = bp; 132 return (0); 133 } 134 if (nb != 0) { 135 /* 136 * Consider need to reallocate a fragment. 137 */ 138 osize = fragroundup(fs, blkoff(fs, ip->i_size)); 139 nsize = fragroundup(fs, size); 140 if (nsize <= osize) { 141 error = bread(vp, lbn, osize, NOCRED, &bp); 142 if (error) { 143 brelse(bp); 144 return (error); 145 } 146 bp->b_blkno = fsbtodb(fs, nb); 147 } else { 148 error = ffs_realloccg(ip, lbn, 149 ffs_blkpref(ip, lbn, (int)lbn, 150 &ip->i_db[0]), osize, nsize, cred, &bp); 151 if (error) 152 return (error); 153 if (DOINGSOFTDEP(vp)) 154 softdep_setup_allocdirect(ip, lbn, 155 dbtofsb(fs, bp->b_blkno), nb, 156 nsize, osize, bp); 157 } 158 } else { 159 if (ip->i_size < smalllblktosize(fs, lbn + 1)) 160 nsize = fragroundup(fs, size); 161 else 162 nsize = fs->fs_bsize; 163 error = ffs_alloc(ip, lbn, 164 ffs_blkpref(ip, lbn, (int)lbn, &ip->i_db[0]), 165 nsize, cred, &newb); 166 if (error) 167 return (error); 168 bp = getblk(vp, lbn, nsize, 0, 0); 169 bp->b_blkno = fsbtodb(fs, newb); 170 if (flags & B_CLRBUF) 171 vfs_bio_clrbuf(bp); 172 if (DOINGSOFTDEP(vp)) 173 softdep_setup_allocdirect(ip, lbn, newb, 0, 174 nsize, 0, bp); 175 } 176 ip->i_db[lbn] = dbtofsb(fs, bp->b_blkno); 177 ip->i_flag |= IN_CHANGE | IN_UPDATE; 178 *a_bpp = bp; 179 return (0); 180 } 181 /* 182 * Determine the number of levels of indirection. 183 */ 184 pref = 0; 185 if ((error = ufs_getlbns(vp, lbn, indirs, &num)) != 0) 186 return(error); 187 #ifdef DIAGNOSTIC 188 if (num < 1) 189 panic ("ffs_balloc: ufs_bmaparray returned indirect block"); 190 #endif 191 /* 192 * Fetch the first indirect block allocating if necessary. 193 */ 194 --num; 195 nb = ip->i_ib[indirs[0].in_off]; 196 allocib = NULL; 197 allocblk = allociblk; 198 if (nb == 0) { 199 pref = ffs_blkpref(ip, lbn, 0, (ufs_daddr_t *)0); 200 if ((error = ffs_alloc(ip, lbn, pref, (int)fs->fs_bsize, 201 cred, &newb)) != 0) 202 return (error); 203 nb = newb; 204 *allocblk++ = nb; 205 bp = getblk(vp, indirs[1].in_lbn, fs->fs_bsize, 0, 0); 206 bp->b_blkno = fsbtodb(fs, nb); 207 vfs_bio_clrbuf(bp); 208 if (DOINGSOFTDEP(vp)) { 209 softdep_setup_allocdirect(ip, NDADDR + indirs[0].in_off, 210 newb, 0, fs->fs_bsize, 0, bp); 211 bdwrite(bp); 212 } else { 213 /* 214 * Write synchronously so that indirect blocks 215 * never point at garbage. 216 */ 217 if (DOINGASYNC(vp)) 218 bdwrite(bp); 219 else if ((error = bwrite(bp)) != 0) 220 goto fail; 221 } 222 allocib = &ip->i_ib[indirs[0].in_off]; 223 *allocib = nb; 224 ip->i_flag |= IN_CHANGE | IN_UPDATE; 225 } 226 /* 227 * Fetch through the indirect blocks, allocating as necessary. 228 */ 229 for (i = 1;;) { 230 error = bread(vp, 231 indirs[i].in_lbn, (int)fs->fs_bsize, NOCRED, &bp); 232 if (error) { 233 brelse(bp); 234 goto fail; 235 } 236 bap = (ufs_daddr_t *)bp->b_data; 237 nb = bap[indirs[i].in_off]; 238 if (i == num) 239 break; 240 i += 1; 241 if (nb != 0) { 242 bqrelse(bp); 243 continue; 244 } 245 if (pref == 0) 246 pref = ffs_blkpref(ip, lbn, 0, (ufs_daddr_t *)0); 247 if ((error = 248 ffs_alloc(ip, lbn, pref, (int)fs->fs_bsize, cred, &newb)) != 0) { 249 brelse(bp); 250 goto fail; 251 } 252 nb = newb; 253 *allocblk++ = nb; 254 nbp = getblk(vp, indirs[i].in_lbn, fs->fs_bsize, 0, 0); 255 nbp->b_blkno = fsbtodb(fs, nb); 256 vfs_bio_clrbuf(nbp); 257 if (DOINGSOFTDEP(vp)) { 258 softdep_setup_allocindir_meta(nbp, ip, bp, 259 indirs[i - 1].in_off, nb); 260 bdwrite(nbp); 261 } else { 262 /* 263 * Write synchronously so that indirect blocks 264 * never point at garbage. 265 */ 266 if ((error = bwrite(nbp)) != 0) { 267 brelse(bp); 268 goto fail; 269 } 270 } 271 bap[indirs[i - 1].in_off] = nb; 272 if (allocib == NULL && unwindidx < 0) 273 unwindidx = i - 1; 274 /* 275 * If required, write synchronously, otherwise use 276 * delayed write. 277 */ 278 if (flags & B_SYNC) { 279 bwrite(bp); 280 } else { 281 if (bp->b_bufsize == fs->fs_bsize) 282 bp->b_flags |= B_CLUSTEROK; 283 bdwrite(bp); 284 } 285 } 286 /* 287 * If asked only for the indirect block, then return it. 288 */ 289 if (flags & B_METAONLY) { 290 *a_bpp = bp; 291 return (0); 292 } 293 /* 294 * Get the data block, allocating if necessary. 295 */ 296 if (nb == 0) { 297 pref = ffs_blkpref(ip, lbn, indirs[i].in_off, &bap[0]); 298 error = ffs_alloc(ip, 299 lbn, pref, (int)fs->fs_bsize, cred, &newb); 300 if (error) { 301 brelse(bp); 302 goto fail; 303 } 304 nb = newb; 305 *allocblk++ = nb; 306 nbp = getblk(vp, lbn, fs->fs_bsize, 0, 0); 307 nbp->b_blkno = fsbtodb(fs, nb); 308 if (flags & B_CLRBUF) 309 vfs_bio_clrbuf(nbp); 310 if (DOINGSOFTDEP(vp)) 311 softdep_setup_allocindir_page(ip, lbn, bp, 312 indirs[i].in_off, nb, 0, nbp); 313 bap[indirs[i].in_off] = nb; 314 /* 315 * If required, write synchronously, otherwise use 316 * delayed write. 317 */ 318 if (flags & B_SYNC) { 319 bwrite(bp); 320 } else { 321 if (bp->b_bufsize == fs->fs_bsize) 322 bp->b_flags |= B_CLUSTEROK; 323 bdwrite(bp); 324 } 325 *a_bpp = nbp; 326 return (0); 327 } 328 brelse(bp); 329 if (flags & B_CLRBUF) { 330 error = bread(vp, lbn, (int)fs->fs_bsize, NOCRED, &nbp); 331 if (error) { 332 brelse(nbp); 333 goto fail; 334 } 335 } else { 336 nbp = getblk(vp, lbn, fs->fs_bsize, 0, 0); 337 nbp->b_blkno = fsbtodb(fs, nb); 338 } 339 *a_bpp = nbp; 340 return (0); 341 fail: 342 /* 343 * If we have failed part way through block allocation, we 344 * have to deallocate any indirect blocks that we have allocated. 345 * We have to fsync the file before we start to get rid of all 346 * of its dependencies so that we do not leave them dangling. 347 * We have to sync it at the end so that the soft updates code 348 * does not find any untracked changes. Although this is really 349 * slow, running out of disk space is not expected to be a common 350 * occurence. The error return from fsync is ignored as we already 351 * have an error to return to the user. 352 */ 353 (void) VOP_FSYNC(vp, cred, MNT_WAIT, td); 354 for (deallocated = 0, blkp = allociblk; blkp < allocblk; blkp++) { 355 ffs_blkfree(fs, ip->i_devvp, *blkp, fs->fs_bsize, ip->i_number); 356 deallocated += fs->fs_bsize; 357 } 358 if (allocib != NULL) { 359 *allocib = 0; 360 } else if (unwindidx >= 0) { 361 int r; 362 363 r = bread(vp, indirs[unwindidx].in_lbn, 364 (int)fs->fs_bsize, NOCRED, &bp); 365 if (r) { 366 panic("Could not unwind indirect block, error %d", r); 367 brelse(bp); 368 } else { 369 bap = (ufs_daddr_t *)bp->b_data; 370 bap[indirs[unwindidx].in_off] = 0; 371 if (flags & B_SYNC) { 372 bwrite(bp); 373 } else { 374 if (bp->b_bufsize == fs->fs_bsize) 375 bp->b_flags |= B_CLUSTEROK; 376 bdwrite(bp); 377 } 378 } 379 } 380 if (deallocated) { 381 #ifdef QUOTA 382 /* 383 * Restore user's disk quota because allocation failed. 384 */ 385 (void) chkdq(ip, (long)-btodb(deallocated), cred, FORCE); 386 #endif 387 ip->i_blocks -= btodb(deallocated); 388 ip->i_flag |= IN_CHANGE | IN_UPDATE; 389 } 390 (void) VOP_FSYNC(vp, cred, MNT_WAIT, td); 391 return (error); 392 } 393