ffs_balloc.c revision 33108
1219820Sjeff/* 2219820Sjeff * Copyright (c) 1982, 1986, 1989, 1993 3219820Sjeff * The Regents of the University of California. All rights reserved. 4219820Sjeff * 5219820Sjeff * Redistribution and use in source and binary forms, with or without 6219820Sjeff * modification, are permitted provided that the following conditions 7219820Sjeff * are met: 8219820Sjeff * 1. Redistributions of source code must retain the above copyright 9219820Sjeff * notice, this list of conditions and the following disclaimer. 10219820Sjeff * 2. Redistributions in binary form must reproduce the above copyright 11219820Sjeff * notice, this list of conditions and the following disclaimer in the 12219820Sjeff * documentation and/or other materials provided with the distribution. 13219820Sjeff * 3. All advertising materials mentioning features or use of this software 14219820Sjeff * must display the following acknowledgement: 15219820Sjeff * This product includes software developed by the University of 16219820Sjeff * California, Berkeley and its contributors. 17219820Sjeff * 4. Neither the name of the University nor the names of its contributors 18219820Sjeff * may be used to endorse or promote products derived from this software 19219820Sjeff * without specific prior written permission. 20219820Sjeff * 21219820Sjeff * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND 22219820Sjeff * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE 23219820Sjeff * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE 24219820Sjeff * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE 25219820Sjeff * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL 26219820Sjeff * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS 27219820Sjeff * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) 28219820Sjeff * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT 29219820Sjeff * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY 30219820Sjeff * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF 31219820Sjeff * SUCH DAMAGE. 32219820Sjeff * 33219820Sjeff * @(#)ffs_balloc.c 8.8 (Berkeley) 6/16/95 34219820Sjeff * $Id: ffs_balloc.c,v 1.17 1998/01/06 05:23:33 dyson Exp $ 35219820Sjeff */ 36219820Sjeff 37219820Sjeff#include "opt_diagnostic.h" 38219820Sjeff 39219820Sjeff#include <sys/param.h> 40219820Sjeff#include <sys/systm.h> 41219820Sjeff#include <sys/buf.h> 42219820Sjeff#include <sys/lock.h> 43219820Sjeff#include <sys/vnode.h> 44219820Sjeff 45219820Sjeff#include <ufs/ufs/quota.h> 46219820Sjeff#include <ufs/ufs/inode.h> 47219820Sjeff#include <ufs/ufs/ufs_extern.h> 48219820Sjeff 49219820Sjeff#include <ufs/ffs/fs.h> 50219820Sjeff#include <ufs/ffs/ffs_extern.h> 51219820Sjeff 52219820Sjeff/* 53219820Sjeff * Balloc defines the structure of file system storage 54219820Sjeff * by allocating the physical blocks on a device given 55219820Sjeff * the inode and the logical block number in a file. 56219820Sjeff */ 57219820Sjeffint 58219820Sjeffffs_balloc(ip, lbn, size, cred, bpp, flags) 59219820Sjeff register struct inode *ip; 60219820Sjeff register ufs_daddr_t lbn; 61219820Sjeff int size; 62219820Sjeff struct ucred *cred; 63219820Sjeff struct buf **bpp; 64219820Sjeff int flags; 65219820Sjeff{ 66219820Sjeff register struct fs *fs; 67219820Sjeff register ufs_daddr_t nb; 68219820Sjeff struct buf *bp, *nbp; 69219820Sjeff struct vnode *vp = ITOV(ip); 70219820Sjeff struct indir indirs[NIADDR + 2]; 71219820Sjeff ufs_daddr_t newb, *bap, pref; 72219820Sjeff int deallocated, osize, nsize, num, i, error; 73219820Sjeff ufs_daddr_t *allocib, *blkp, *allocblk, allociblk[NIADDR + 1]; 74219820Sjeff 75219820Sjeff *bpp = NULL; 76219820Sjeff if (lbn < 0) 77219820Sjeff return (EFBIG); 78219820Sjeff fs = ip->i_fs; 79219820Sjeff 80219820Sjeff /* 81219820Sjeff * If the next write will extend the file into a new block, 82219820Sjeff * and the file is currently composed of a fragment 83219820Sjeff * this fragment has to be extended to be a full block. 84219820Sjeff */ 85219820Sjeff nb = lblkno(fs, ip->i_size); 86219820Sjeff if (nb < NDADDR && nb < lbn) { 87219820Sjeff osize = blksize(fs, ip, nb); 88219820Sjeff if (osize < fs->fs_bsize && osize > 0) { 89219820Sjeff error = ffs_realloccg(ip, nb, 90219820Sjeff ffs_blkpref(ip, nb, (int)nb, &ip->i_db[0]), 91219820Sjeff osize, (int)fs->fs_bsize, cred, &bp); 92219820Sjeff if (error) 93219820Sjeff return (error); 94219820Sjeff ip->i_size = smalllblktosize(fs, nb + 1); 95219820Sjeff ip->i_db[nb] = dbtofsb(fs, bp->b_blkno); 96219820Sjeff ip->i_flag |= IN_CHANGE | IN_UPDATE; 97219820Sjeff if (flags & B_SYNC) 98219820Sjeff bwrite(bp); 99219820Sjeff else 100219820Sjeff bawrite(bp); 101219820Sjeff } 102219820Sjeff } 103219820Sjeff /* 104219820Sjeff * The first NDADDR blocks are direct blocks 105219820Sjeff */ 106219820Sjeff if (lbn < NDADDR) { 107219820Sjeff nb = ip->i_db[lbn]; 108219820Sjeff if (nb != 0 && ip->i_size >= smalllblktosize(fs, lbn + 1)) { 109219820Sjeff error = bread(vp, lbn, fs->fs_bsize, NOCRED, &bp); 110219820Sjeff if (error) { 111219820Sjeff brelse(bp); 112219820Sjeff return (error); 113219820Sjeff } 114219820Sjeff bp->b_blkno = fsbtodb(fs, nb); 115219820Sjeff *bpp = bp; 116219820Sjeff return (0); 117219820Sjeff } 118219820Sjeff if (nb != 0) { 119219820Sjeff /* 120219820Sjeff * Consider need to reallocate a fragment. 121219820Sjeff */ 122219820Sjeff osize = fragroundup(fs, blkoff(fs, ip->i_size)); 123219820Sjeff nsize = fragroundup(fs, size); 124219820Sjeff if (nsize <= osize) { 125219820Sjeff error = bread(vp, lbn, osize, NOCRED, &bp); 126219820Sjeff if (error) { 127219820Sjeff brelse(bp); 128219820Sjeff return (error); 129219820Sjeff } 130219820Sjeff bp->b_blkno = fsbtodb(fs, nb); 131219820Sjeff } else { 132219820Sjeff error = ffs_realloccg(ip, lbn, 133219820Sjeff ffs_blkpref(ip, lbn, (int)lbn, 134219820Sjeff &ip->i_db[0]), osize, nsize, cred, &bp); 135219820Sjeff if (error) 136219820Sjeff return (error); 137219820Sjeff } 138219820Sjeff } else { 139219820Sjeff if (ip->i_size < smalllblktosize(fs, lbn + 1)) 140219820Sjeff nsize = fragroundup(fs, size); 141219820Sjeff else 142219820Sjeff nsize = fs->fs_bsize; 143219820Sjeff error = ffs_alloc(ip, lbn, 144219820Sjeff ffs_blkpref(ip, lbn, (int)lbn, &ip->i_db[0]), 145219820Sjeff nsize, cred, &newb); 146219820Sjeff if (error) 147219820Sjeff return (error); 148219820Sjeff bp = getblk(vp, lbn, nsize, 0, 0); 149219820Sjeff bp->b_blkno = fsbtodb(fs, newb); 150219820Sjeff if (flags & B_CLRBUF) 151219820Sjeff vfs_bio_clrbuf(bp); 152219820Sjeff } 153219820Sjeff ip->i_db[lbn] = dbtofsb(fs, bp->b_blkno); 154219820Sjeff ip->i_flag |= IN_CHANGE | IN_UPDATE; 155219820Sjeff *bpp = bp; 156219820Sjeff return (0); 157219820Sjeff } 158219820Sjeff /* 159219820Sjeff * Determine the number of levels of indirection. 160219820Sjeff */ 161219820Sjeff pref = 0; 162219820Sjeff if (error = ufs_getlbns(vp, lbn, indirs, &num)) 163219820Sjeff return(error); 164219820Sjeff#ifdef DIAGNOSTIC 165219820Sjeff if (num < 1) 166219820Sjeff panic ("ffs_balloc: ufs_bmaparray returned indirect block"); 167219820Sjeff#endif 168219820Sjeff /* 169219820Sjeff * Fetch the first indirect block allocating if necessary. 170219820Sjeff */ 171219820Sjeff --num; 172219820Sjeff nb = ip->i_ib[indirs[0].in_off]; 173219820Sjeff allocib = NULL; 174219820Sjeff allocblk = allociblk; 175219820Sjeff if (nb == 0) { 176219820Sjeff pref = ffs_blkpref(ip, lbn, 0, (ufs_daddr_t *)0); 177219820Sjeff if (error = ffs_alloc(ip, lbn, pref, (int)fs->fs_bsize, 178219820Sjeff cred, &newb)) 179219820Sjeff return (error); 180219820Sjeff nb = newb; 181219820Sjeff *allocblk++ = nb; 182219820Sjeff bp = getblk(vp, indirs[1].in_lbn, fs->fs_bsize, 0, 0); 183219820Sjeff bp->b_blkno = fsbtodb(fs, nb); 184219820Sjeff vfs_bio_clrbuf(bp); 185219820Sjeff /* 186219820Sjeff * Write synchronously so that indirect blocks 187219820Sjeff * never point at garbage. 188219820Sjeff */ 189219820Sjeff if (error = bwrite(bp)) 190219820Sjeff goto fail; 191219820Sjeff allocib = &ip->i_ib[indirs[0].in_off]; 192219820Sjeff *allocib = nb; 193219820Sjeff ip->i_flag |= IN_CHANGE | IN_UPDATE; 194219820Sjeff } 195219820Sjeff /* 196219820Sjeff * Fetch through the indirect blocks, allocating as necessary. 197219820Sjeff */ 198219820Sjeff for (i = 1;;) { 199219820Sjeff error = bread(vp, 200219820Sjeff indirs[i].in_lbn, (int)fs->fs_bsize, NOCRED, &bp); 201219820Sjeff if (error) { 202219820Sjeff brelse(bp); 203219820Sjeff goto fail; 204219820Sjeff } 205219820Sjeff bap = (ufs_daddr_t *)bp->b_data; 206219820Sjeff nb = bap[indirs[i].in_off]; 207219820Sjeff if (i == num) 208219820Sjeff break; 209219820Sjeff i += 1; 210219820Sjeff if (nb != 0) { 211219820Sjeff bqrelse(bp); 212219820Sjeff continue; 213219820Sjeff } 214219820Sjeff if (pref == 0) 215219820Sjeff pref = ffs_blkpref(ip, lbn, 0, (ufs_daddr_t *)0); 216219820Sjeff if (error = 217219820Sjeff ffs_alloc(ip, lbn, pref, (int)fs->fs_bsize, cred, &newb)) { 218219820Sjeff brelse(bp); 219219820Sjeff goto fail; 220219820Sjeff } 221219820Sjeff nb = newb; 222219820Sjeff *allocblk++ = nb; 223219820Sjeff nbp = getblk(vp, indirs[i].in_lbn, fs->fs_bsize, 0, 0); 224219820Sjeff nbp->b_blkno = fsbtodb(fs, nb); 225219820Sjeff vfs_bio_clrbuf(nbp); 226219820Sjeff /* 227219820Sjeff * Write synchronously so that indirect blocks 228219820Sjeff * never point at garbage. 229219820Sjeff */ 230219820Sjeff if (error = bwrite(nbp)) { 231 brelse(bp); 232 goto fail; 233 } 234 bap[indirs[i - 1].in_off] = nb; 235 /* 236 * If required, write synchronously, otherwise use 237 * delayed write. 238 */ 239 if (flags & B_SYNC) { 240 bwrite(bp); 241 } else { 242 if (bp->b_bufsize == fs->fs_bsize) 243 bp->b_flags |= B_CLUSTEROK; 244 bdwrite(bp); 245 } 246 } 247 /* 248 * Get the data block, allocating if necessary. 249 */ 250 if (nb == 0) { 251 pref = ffs_blkpref(ip, lbn, indirs[i].in_off, &bap[0]); 252 error = ffs_alloc(ip, 253 lbn, pref, (int)fs->fs_bsize, cred, &newb); 254 if (error) { 255 brelse(bp); 256 goto fail; 257 } 258 nb = newb; 259 *allocblk++ = nb; 260 nbp = getblk(vp, lbn, fs->fs_bsize, 0, 0); 261 nbp->b_blkno = fsbtodb(fs, nb); 262 if (flags & B_CLRBUF) 263 vfs_bio_clrbuf(nbp); 264 bap[indirs[i].in_off] = nb; 265 /* 266 * If required, write synchronously, otherwise use 267 * delayed write. 268 */ 269 if (flags & B_SYNC) { 270 bwrite(bp); 271 } else { 272 if (bp->b_bufsize == fs->fs_bsize) 273 bp->b_flags |= B_CLUSTEROK; 274 bdwrite(bp); 275 } 276 *bpp = nbp; 277 return (0); 278 } 279 brelse(bp); 280 if (flags & B_CLRBUF) { 281 error = bread(vp, lbn, (int)fs->fs_bsize, NOCRED, &nbp); 282 if (error) { 283 brelse(nbp); 284 goto fail; 285 } 286 } else { 287 nbp = getblk(vp, lbn, fs->fs_bsize, 0, 0); 288 nbp->b_blkno = fsbtodb(fs, nb); 289 } 290 *bpp = nbp; 291 return (0); 292fail: 293 /* 294 * If we have failed part way through block allocation, we 295 * have to deallocate any indirect blocks that we have allocated. 296 */ 297 for (deallocated = 0, blkp = allociblk; blkp < allocblk; blkp++) { 298 ffs_blkfree(ip, *blkp, fs->fs_bsize); 299 deallocated += fs->fs_bsize; 300 } 301 if (allocib != NULL) 302 *allocib = 0; 303 if (deallocated) { 304#ifdef QUOTA 305 /* 306 * Restore user's disk quota because allocation failed. 307 */ 308 (void) chkdq(ip, (long)-btodb(deallocated), cred, FORCE); 309#endif 310 ip->i_blocks -= btodb(deallocated); 311 ip->i_flag |= IN_CHANGE | IN_UPDATE; 312 } 313 return (error); 314} 315