1 /* $NetBSD: ext2fs_balloc.c,v 1.16 2002/01/26 08:32:06 chs Exp $ */ 2 3 /* 4 * Copyright (c) 1997 Manuel Bouyer. 5 * Copyright (c) 1982, 1986, 1989, 1993 6 * The Regents of the University of California. All rights reserved. 7 * 8 * Redistribution and use in source and binary forms, with or without 9 * modification, are permitted provided that the following conditions 10 * are met: 11 * 1. Redistributions of source code must retain the above copyright 12 * notice, this list of conditions and the following disclaimer. 13 * 2. Redistributions in binary form must reproduce the above copyright 14 * notice, this list of conditions and the following disclaimer in the 15 * documentation and/or other materials provided with the distribution. 16 * 3. All advertising materials mentioning features or use of this software 17 * must display the following acknowledgement: 18 * This product includes software developed by the University of 19 * California, Berkeley and its contributors. 20 * 4. Neither the name of the University nor the names of its contributors 21 * may be used to endorse or promote products derived from this software 22 * without specific prior written permission. 23 * 24 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND 25 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE 26 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE 27 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE 28 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL 29 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS 30 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) 31 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT 32 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY 33 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF 34 * SUCH DAMAGE. 35 * 36 * @(#)ffs_balloc.c 8.4 (Berkeley) 9/23/93 37 * Modified for ext2fs by Manuel Bouyer. 38 */ 39 40 #include <sys/cdefs.h> 41 __KERNEL_RCSID(0, "$NetBSD: ext2fs_balloc.c,v 1.16 2002/01/26 08:32:06 chs Exp $"); 42 43 #if defined(_KERNEL_OPT) 44 #include "opt_uvmhist.h" 45 #endif 46 47 #include <sys/param.h> 48 #include <sys/systm.h> 49 #include <sys/buf.h> 50 #include <sys/proc.h> 51 #include <sys/file.h> 52 #include <sys/vnode.h> 53 #include <sys/mount.h> 54 55 #include <uvm/uvm.h> 56 57 #include <ufs/ufs/inode.h> 58 #include <ufs/ufs/ufs_extern.h> 59 60 #include <ufs/ext2fs/ext2fs.h> 61 #include <ufs/ext2fs/ext2fs_extern.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 ext2fs_balloc(ip, bn, size, cred, bpp, flags) 70 struct inode *ip; 71 ufs_daddr_t bn; 72 int size; 73 struct ucred *cred; 74 struct buf **bpp; 75 int flags; 76 { 77 struct m_ext2fs *fs; 78 ufs_daddr_t nb; 79 struct buf *bp, *nbp; 80 struct vnode *vp = ITOV(ip); 81 struct indir indirs[NIADDR + 2]; 82 ufs_daddr_t newb, lbn, *bap, pref; 83 int num, i, error; 84 u_int deallocated; 85 ufs_daddr_t *allocib, *blkp, *allocblk, allociblk[NIADDR + 1]; 86 int unwindidx = -1; 87 UVMHIST_FUNC("ext2fs_balloc"); UVMHIST_CALLED(ubchist); 88 89 UVMHIST_LOG(ubchist, "bn 0x%x", bn,0,0,0); 90 91 if (bpp != NULL) { 92 *bpp = NULL; 93 } 94 if (bn < 0) 95 return (EFBIG); 96 fs = ip->i_e2fs; 97 lbn = bn; 98 99 /* 100 * The first NDADDR blocks are direct blocks 101 */ 102 if (bn < NDADDR) { 103 nb = fs2h32(ip->i_e2fs_blocks[bn]); 104 if (nb != 0) { 105 106 /* 107 * the block is already allocated, just read it. 108 */ 109 110 if (bpp != NULL) { 111 error = bread(vp, bn, fs->e2fs_bsize, NOCRED, 112 &bp); 113 if (error) { 114 brelse(bp); 115 return (error); 116 } 117 *bpp = bp; 118 } 119 return (0); 120 } 121 122 /* 123 * allocate a new direct block. 124 */ 125 126 error = ext2fs_alloc(ip, bn, 127 ext2fs_blkpref(ip, bn, bn, &ip->i_e2fs_blocks[0]), 128 cred, &newb); 129 if (error) 130 return (error); 131 ip->i_e2fs_last_lblk = lbn; 132 ip->i_e2fs_last_blk = newb; 133 ip->i_e2fs_blocks[bn] = h2fs32(newb); 134 ip->i_flag |= IN_CHANGE | IN_UPDATE; 135 if (bpp != NULL) { 136 bp = getblk(vp, bn, fs->e2fs_bsize, 0, 0); 137 bp->b_blkno = fsbtodb(fs, newb); 138 if (flags & B_CLRBUF) 139 clrbuf(bp); 140 *bpp = bp; 141 } 142 return (0); 143 } 144 /* 145 * Determine the number of levels of indirection. 146 */ 147 pref = 0; 148 if ((error = ufs_getlbns(vp, bn, indirs, &num)) != 0) 149 return(error); 150 #ifdef DIAGNOSTIC 151 if (num < 1) 152 panic ("ext2fs_balloc: ufs_getlbns returned indirect block\n"); 153 #endif 154 /* 155 * Fetch the first indirect block allocating if necessary. 156 */ 157 --num; 158 nb = fs2h32(ip->i_e2fs_blocks[NDADDR + indirs[0].in_off]); 159 allocib = NULL; 160 allocblk = allociblk; 161 if (nb == 0) { 162 pref = ext2fs_blkpref(ip, lbn, 0, (ufs_daddr_t *)0); 163 error = ext2fs_alloc(ip, lbn, pref, cred, &newb); 164 if (error) 165 return (error); 166 nb = newb; 167 *allocblk++ = nb; 168 ip->i_e2fs_last_blk = newb; 169 bp = getblk(vp, indirs[1].in_lbn, fs->e2fs_bsize, 0, 0); 170 bp->b_blkno = fsbtodb(fs, newb); 171 clrbuf(bp); 172 /* 173 * Write synchronously so that indirect blocks 174 * never point at garbage. 175 */ 176 if ((error = bwrite(bp)) != 0) 177 goto fail; 178 unwindidx = 0; 179 allocib = &ip->i_e2fs_blocks[NDADDR + indirs[0].in_off]; 180 *allocib = h2fs32(newb); 181 ip->i_flag |= IN_CHANGE | IN_UPDATE; 182 } 183 /* 184 * Fetch through the indirect blocks, allocating as necessary. 185 */ 186 for (i = 1;;) { 187 error = bread(vp, 188 indirs[i].in_lbn, (int)fs->e2fs_bsize, NOCRED, &bp); 189 if (error) { 190 brelse(bp); 191 goto fail; 192 } 193 bap = (ufs_daddr_t *)bp->b_data; 194 nb = fs2h32(bap[indirs[i].in_off]); 195 if (i == num) 196 break; 197 i++; 198 if (nb != 0) { 199 brelse(bp); 200 continue; 201 } 202 pref = ext2fs_blkpref(ip, lbn, 0, (ufs_daddr_t *)0); 203 error = ext2fs_alloc(ip, lbn, pref, cred, &newb); 204 if (error) { 205 brelse(bp); 206 goto fail; 207 } 208 nb = newb; 209 *allocblk++ = nb; 210 ip->i_e2fs_last_blk = newb; 211 nbp = getblk(vp, indirs[i].in_lbn, fs->e2fs_bsize, 0, 0); 212 nbp->b_blkno = fsbtodb(fs, nb); 213 clrbuf(nbp); 214 /* 215 * Write synchronously so that indirect blocks 216 * never point at garbage. 217 */ 218 if ((error = bwrite(nbp)) != 0) { 219 brelse(bp); 220 goto fail; 221 } 222 if (unwindidx < 0) 223 unwindidx = i - 1; 224 bap[indirs[i - 1].in_off] = h2fs32(nb); 225 /* 226 * If required, write synchronously, otherwise use 227 * delayed write. 228 */ 229 if (flags & B_SYNC) { 230 bwrite(bp); 231 } else { 232 bdwrite(bp); 233 } 234 } 235 /* 236 * Get the data block, allocating if necessary. 237 */ 238 if (nb == 0) { 239 pref = ext2fs_blkpref(ip, lbn, indirs[num].in_off, &bap[0]); 240 error = ext2fs_alloc(ip, lbn, pref, cred, &newb); 241 if (error) { 242 brelse(bp); 243 goto fail; 244 } 245 nb = newb; 246 *allocblk++ = nb; 247 ip->i_e2fs_last_lblk = lbn; 248 ip->i_e2fs_last_blk = newb; 249 bap[indirs[num].in_off] = h2fs32(nb); 250 /* 251 * If required, write synchronously, otherwise use 252 * delayed write. 253 */ 254 if (flags & B_SYNC) { 255 bwrite(bp); 256 } else { 257 bdwrite(bp); 258 } 259 if (bpp != NULL) { 260 nbp = getblk(vp, lbn, fs->e2fs_bsize, 0, 0); 261 nbp->b_blkno = fsbtodb(fs, nb); 262 if (flags & B_CLRBUF) 263 clrbuf(nbp); 264 *bpp = nbp; 265 } 266 return (0); 267 } 268 brelse(bp); 269 if (bpp != NULL) { 270 if (flags & B_CLRBUF) { 271 error = bread(vp, lbn, (int)fs->e2fs_bsize, NOCRED, 272 &nbp); 273 if (error) { 274 brelse(nbp); 275 goto fail; 276 } 277 } else { 278 nbp = getblk(vp, lbn, fs->e2fs_bsize, 0, 0); 279 nbp->b_blkno = fsbtodb(fs, nb); 280 } 281 *bpp = nbp; 282 } 283 return (0); 284 fail: 285 /* 286 * If we have failed part way through block allocation, we 287 * have to deallocate any indirect blocks that we have allocated. 288 */ 289 for (deallocated = 0, blkp = allociblk; blkp < allocblk; blkp++) { 290 ext2fs_blkfree(ip, *blkp); 291 deallocated += fs->e2fs_bsize; 292 } 293 if (unwindidx >= 0) { 294 if (unwindidx == 0) { 295 *allocib = 0; 296 } else { 297 int r; 298 299 r = bread(vp, indirs[unwindidx].in_lbn, 300 (int)fs->e2fs_bsize, NOCRED, &bp); 301 if (r) { 302 panic("Could not unwind indirect block, error %d", r); 303 brelse(bp); 304 } else { 305 bap = (ufs_daddr_t *)bp->b_data; 306 bap[indirs[unwindidx].in_off] = 0; 307 if (flags & B_SYNC) 308 bwrite(bp); 309 else 310 bdwrite(bp); 311 } 312 } 313 for (i = unwindidx + 1; i <= num; i++) { 314 bp = getblk(vp, indirs[i].in_lbn, (int)fs->e2fs_bsize, 315 0, 0); 316 bp->b_flags |= B_INVAL; 317 brelse(bp); 318 } 319 } 320 if (deallocated) { 321 ip->i_e2fs_nblock -= btodb(deallocated); 322 ip->i_e2fs_flags |= IN_CHANGE | IN_UPDATE; 323 } 324 return error; 325 } 326 327 int 328 ext2fs_gop_alloc(struct vnode *vp, off_t off, off_t len, int flags, 329 struct ucred *cred) 330 { 331 struct inode *ip = VTOI(vp); 332 struct m_ext2fs *fs = ip->i_e2fs; 333 int error, delta, bshift, bsize; 334 UVMHIST_FUNC("ext2fs_gop_alloc"); UVMHIST_CALLED(ubchist); 335 336 bshift = fs->e2fs_bshift; 337 bsize = 1 << bshift; 338 339 delta = off & (bsize - 1); 340 off -= delta; 341 len += delta; 342 343 while (len > 0) { 344 bsize = min(bsize, len); 345 UVMHIST_LOG(ubchist, "off 0x%x len 0x%x bsize 0x%x", 346 off, len, bsize, 0); 347 348 error = ext2fs_balloc(ip, lblkno(fs, off), bsize, cred, 349 NULL, flags); 350 if (error) { 351 UVMHIST_LOG(ubchist, "error %d", error, 0,0,0); 352 return error; 353 } 354 355 /* 356 * increase file size now, VOP_BALLOC() requires that 357 * EOF be up-to-date before each call. 358 */ 359 360 if (ip->i_e2fs_size < off + bsize) { 361 UVMHIST_LOG(ubchist, "old 0x%x new 0x%x", 362 ip->i_e2fs_size, off + bsize,0,0); 363 ip->i_e2fs_size = off + bsize; 364 } 365 366 off += bsize; 367 len -= bsize; 368 } 369 return 0; 370 } 371 372 /* 373 * allocate a range of blocks in a file. 374 * after this function returns, any page entirely contained within the range 375 * will map to invalid data and thus must be overwritten before it is made 376 * accessible to others. 377 */ 378 379 int 380 ext2fs_balloc_range(vp, off, len, cred, flags) 381 struct vnode *vp; 382 off_t off, len; 383 struct ucred *cred; 384 int flags; 385 { 386 off_t oldeof, eof, pagestart; 387 struct genfs_node *gp = VTOG(vp); 388 int i, delta, error, npages; 389 int bshift = vp->v_mount->mnt_fs_bshift; 390 int bsize = 1 << bshift; 391 int ppb = max(bsize >> PAGE_SHIFT, 1); 392 struct vm_page *pgs[ppb]; 393 UVMHIST_FUNC("ext2fs_balloc_range"); UVMHIST_CALLED(ubchist); 394 UVMHIST_LOG(ubchist, "vp %p off 0x%x len 0x%x u_size 0x%x", 395 vp, off, len, vp->v_size); 396 397 oldeof = vp->v_size; 398 eof = MAX(oldeof, off + len); 399 UVMHIST_LOG(ubchist, "new eof 0x%x", eof,0,0,0); 400 pgs[0] = NULL; 401 402 /* 403 * cache the new range of the file. this will create zeroed pages 404 * where the new block will be and keep them locked until the 405 * new block is allocated, so there will be no window where 406 * the old contents of the new block is visible to racing threads. 407 */ 408 409 pagestart = trunc_page(off) & ~(bsize - 1); 410 npages = MIN(ppb, (round_page(eof) - pagestart) >> PAGE_SHIFT); 411 memset(pgs, 0, npages); 412 simple_lock(&vp->v_interlock); 413 error = VOP_GETPAGES(vp, pagestart, pgs, &npages, 0, 414 VM_PROT_READ, 0, PGO_SYNCIO | PGO_PASTEOF); 415 if (error) { 416 UVMHIST_LOG(ubchist, "getpages %d", error,0,0,0); 417 return error; 418 } 419 for (i = 0; i < npages; i++) { 420 UVMHIST_LOG(ubchist, "got pgs[%d] %p", i, pgs[i],0,0); 421 KASSERT((pgs[i]->flags & PG_RELEASED) == 0); 422 pgs[i]->flags &= ~PG_CLEAN; 423 uvm_pageactivate(pgs[i]); 424 } 425 426 /* 427 * adjust off to be block-aligned. 428 */ 429 430 delta = off & (bsize - 1); 431 off -= delta; 432 len += delta; 433 434 /* 435 * now allocate the range. 436 */ 437 438 lockmgr(&gp->g_glock, LK_EXCLUSIVE, NULL); 439 error = GOP_ALLOC(vp, off, len, flags, cred); 440 UVMHIST_LOG(ubchist, "alloc %d", error,0,0,0); 441 lockmgr(&gp->g_glock, LK_RELEASE, NULL); 442 443 /* 444 * clear PG_RDONLY on any pages we are holding 445 * (since they now have backing store) and unbusy them. 446 * if we got an error, free any pages we created past the old eob. 447 */ 448 449 simple_lock(&vp->v_interlock); 450 for (i = 0; i < npages; i++) { 451 pgs[i]->flags &= ~PG_RDONLY; 452 if (error) { 453 pgs[i]->flags |= PG_RELEASED; 454 } 455 } 456 if (error) { 457 uvm_lock_pageq(); 458 uvm_page_unbusy(pgs, npages); 459 uvm_unlock_pageq(); 460 } else { 461 uvm_page_unbusy(pgs, npages); 462 } 463 simple_unlock(&vp->v_interlock); 464 return (error); 465 } 466