| 1 | /* $NetBSD: ext2fs_alloc.c,v 1.51 2016/08/20 19:53:43 jdolecek 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. Neither the name of the University nor the names of its contributors |
| 16 | * may be used to endorse or promote products derived from this software |
| 17 | * without specific prior written permission. |
| 18 | * |
| 19 | * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND |
| 20 | * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE |
| 21 | * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE |
| 22 | * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE |
| 23 | * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL |
| 24 | * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS |
| 25 | * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) |
| 26 | * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT |
| 27 | * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY |
| 28 | * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF |
| 29 | * SUCH DAMAGE. |
| 30 | * |
| 31 | * @(#)ffs_alloc.c 8.11 (Berkeley) 10/27/94 |
| 32 | * Modified for ext2fs by Manuel Bouyer. |
| 33 | */ |
| 34 | |
| 35 | /* |
| 36 | * Copyright (c) 1997 Manuel Bouyer. |
| 37 | * |
| 38 | * Redistribution and use in source and binary forms, with or without |
| 39 | * modification, are permitted provided that the following conditions |
| 40 | * are met: |
| 41 | * 1. Redistributions of source code must retain the above copyright |
| 42 | * notice, this list of conditions and the following disclaimer. |
| 43 | * 2. Redistributions in binary form must reproduce the above copyright |
| 44 | * notice, this list of conditions and the following disclaimer in the |
| 45 | * documentation and/or other materials provided with the distribution. |
| 46 | * |
| 47 | * THIS SOFTWARE IS PROVIDED BY THE AUTHOR ``AS IS'' AND ANY EXPRESS OR |
| 48 | * IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES |
| 49 | * OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE DISCLAIMED. |
| 50 | * IN NO EVENT SHALL THE AUTHOR BE LIABLE FOR ANY DIRECT, INDIRECT, |
| 51 | * INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT |
| 52 | * NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, |
| 53 | * DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY |
| 54 | * THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT |
| 55 | * (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF |
| 56 | * THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE. |
| 57 | * |
| 58 | * @(#)ffs_alloc.c 8.11 (Berkeley) 10/27/94 |
| 59 | * Modified for ext2fs by Manuel Bouyer. |
| 60 | */ |
| 61 | |
| 62 | #include <sys/cdefs.h> |
| 63 | __KERNEL_RCSID(0, "$NetBSD: ext2fs_alloc.c,v 1.51 2016/08/20 19:53:43 jdolecek Exp $" ); |
| 64 | |
| 65 | #include <sys/param.h> |
| 66 | #include <sys/systm.h> |
| 67 | #include <sys/buf.h> |
| 68 | #include <sys/proc.h> |
| 69 | #include <sys/vnode.h> |
| 70 | #include <sys/mount.h> |
| 71 | #include <sys/kernel.h> |
| 72 | #include <sys/syslog.h> |
| 73 | #include <sys/kauth.h> |
| 74 | |
| 75 | #include <lib/libkern/crc16.h> |
| 76 | |
| 77 | #include <ufs/ufs/inode.h> |
| 78 | #include <ufs/ufs/ufs_extern.h> |
| 79 | #include <ufs/ufs/ufsmount.h> |
| 80 | |
| 81 | #include <ufs/ext2fs/ext2fs.h> |
| 82 | #include <ufs/ext2fs/ext2fs_extern.h> |
| 83 | |
| 84 | u_long ext2gennumber; |
| 85 | |
| 86 | static daddr_t ext2fs_alloccg(struct inode *, int, daddr_t, int); |
| 87 | static u_long ext2fs_dirpref(struct m_ext2fs *); |
| 88 | static void ext2fs_fserr(struct m_ext2fs *, u_int, const char *); |
| 89 | static u_long ext2fs_hashalloc(struct inode *, int, long, int, |
| 90 | daddr_t (*)(struct inode *, int, daddr_t, int)); |
| 91 | static daddr_t ext2fs_nodealloccg(struct inode *, int, daddr_t, int); |
| 92 | static daddr_t ext2fs_mapsearch(struct m_ext2fs *, char *, daddr_t); |
| 93 | static __inline void ext2fs_cg_update(struct m_ext2fs *, int, struct ext2_gd *, int, int, int, daddr_t); |
| 94 | static uint16_t ext2fs_cg_get_csum(struct m_ext2fs *, int, struct ext2_gd *); |
| 95 | static void ext2fs_init_bb(struct m_ext2fs *, int, struct ext2_gd *, char *); |
| 96 | |
| 97 | /* |
| 98 | * Allocate a block in the file system. |
| 99 | * |
| 100 | * A preference may be optionally specified. If a preference is given |
| 101 | * the following hierarchy is used to allocate a block: |
| 102 | * 1) allocate the requested block. |
| 103 | * 2) allocate a rotationally optimal block in the same cylinder. |
| 104 | * 3) allocate a block in the same cylinder group. |
| 105 | * 4) quadradically rehash into other cylinder groups, until an |
| 106 | * available block is located. |
| 107 | * If no block preference is given the following hierarchy is used |
| 108 | * to allocate a block: |
| 109 | * 1) allocate a block in the cylinder group that contains the |
| 110 | * inode for the file. |
| 111 | * 2) quadradically rehash into other cylinder groups, until an |
| 112 | * available block is located. |
| 113 | */ |
| 114 | int |
| 115 | ext2fs_alloc(struct inode *ip, daddr_t lbn, daddr_t bpref, |
| 116 | kauth_cred_t cred, daddr_t *bnp) |
| 117 | { |
| 118 | struct m_ext2fs *fs; |
| 119 | daddr_t bno; |
| 120 | int cg; |
| 121 | |
| 122 | *bnp = 0; |
| 123 | fs = ip->i_e2fs; |
| 124 | #ifdef DIAGNOSTIC |
| 125 | if (cred == NOCRED) |
| 126 | panic("ext2fs_alloc: missing credential" ); |
| 127 | #endif /* DIAGNOSTIC */ |
| 128 | if (fs->e2fs.e2fs_fbcount == 0) |
| 129 | goto nospace; |
| 130 | if (kauth_authorize_system(cred, KAUTH_SYSTEM_FS_RESERVEDSPACE, 0, NULL, |
| 131 | NULL, NULL) != 0 && |
| 132 | freespace(fs) <= 0) |
| 133 | goto nospace; |
| 134 | if (bpref >= fs->e2fs.e2fs_bcount) |
| 135 | bpref = 0; |
| 136 | if (bpref == 0) |
| 137 | cg = ino_to_cg(fs, ip->i_number); |
| 138 | else |
| 139 | cg = dtog(fs, bpref); |
| 140 | bno = (daddr_t)ext2fs_hashalloc(ip, cg, bpref, fs->e2fs_bsize, |
| 141 | ext2fs_alloccg); |
| 142 | if (bno > 0) { |
| 143 | ext2fs_setnblock(ip, ext2fs_nblock(ip) + btodb(fs->e2fs_bsize)); |
| 144 | ip->i_flag |= IN_CHANGE | IN_UPDATE; |
| 145 | *bnp = bno; |
| 146 | return 0; |
| 147 | } |
| 148 | nospace: |
| 149 | ext2fs_fserr(fs, kauth_cred_geteuid(cred), "file system full" ); |
| 150 | uprintf("\n%s: write failed, file system is full\n" , fs->e2fs_fsmnt); |
| 151 | return ENOSPC; |
| 152 | } |
| 153 | |
| 154 | /* |
| 155 | * Allocate an inode in the file system. |
| 156 | * |
| 157 | * If allocating a directory, use ext2fs_dirpref to select the inode. |
| 158 | * If allocating in a directory, the following hierarchy is followed: |
| 159 | * 1) allocate the preferred inode. |
| 160 | * 2) allocate an inode in the same cylinder group. |
| 161 | * 3) quadradically rehash into other cylinder groups, until an |
| 162 | * available inode is located. |
| 163 | * If no inode preference is given the following hierarchy is used |
| 164 | * to allocate an inode: |
| 165 | * 1) allocate an inode in cylinder group 0. |
| 166 | * 2) quadradically rehash into other cylinder groups, until an |
| 167 | * available inode is located. |
| 168 | */ |
| 169 | int |
| 170 | ext2fs_valloc(struct vnode *pvp, int mode, kauth_cred_t cred, |
| 171 | struct vnode **vpp) |
| 172 | { |
| 173 | struct inode *pip; |
| 174 | struct m_ext2fs *fs; |
| 175 | struct inode *ip; |
| 176 | ino_t ino, ipref; |
| 177 | int cg, error; |
| 178 | |
| 179 | *vpp = NULL; |
| 180 | pip = VTOI(pvp); |
| 181 | fs = pip->i_e2fs; |
| 182 | if (fs->e2fs.e2fs_ficount == 0) |
| 183 | goto noinodes; |
| 184 | |
| 185 | if ((mode & IFMT) == IFDIR) |
| 186 | cg = ext2fs_dirpref(fs); |
| 187 | else |
| 188 | cg = ino_to_cg(fs, pip->i_number); |
| 189 | ipref = cg * fs->e2fs.e2fs_ipg + 1; |
| 190 | ino = (ino_t)ext2fs_hashalloc(pip, cg, (long)ipref, mode, ext2fs_nodealloccg); |
| 191 | if (ino == 0) |
| 192 | goto noinodes; |
| 193 | error = VFS_VGET(pvp->v_mount, ino, vpp); |
| 194 | if (error) { |
| 195 | ext2fs_vfree(pvp, ino, mode); |
| 196 | return error; |
| 197 | } |
| 198 | ip = VTOI(*vpp); |
| 199 | |
| 200 | KASSERT(!E2FS_HAS_GD_CSUM(fs) || (fs->e2fs_gd[ino_to_cg(fs, ino)].ext2bgd_flags & h2fs16(E2FS_BG_INODE_ZEROED)) != 0); |
| 201 | |
| 202 | /* check for already used inode; makes sense only for ZEROED itable */ |
| 203 | if (__predict_false(ip->i_e2fs_mode && ip->i_e2fs_nlink != 0)) { |
| 204 | printf("mode = 0%o, nlinks %d, inum = %llu, fs = %s\n" , |
| 205 | ip->i_e2fs_mode, ip->i_e2fs_nlink, |
| 206 | (unsigned long long)ip->i_number, fs->e2fs_fsmnt); |
| 207 | panic("ext2fs_valloc: dup alloc" ); |
| 208 | } |
| 209 | |
| 210 | memset(ip->i_din.e2fs_din, 0, EXT2_DINODE_SIZE(fs)); |
| 211 | |
| 212 | /* |
| 213 | * Set up a new generation number for this inode. |
| 214 | */ |
| 215 | if (++ext2gennumber < time_second) |
| 216 | ext2gennumber = time_second; |
| 217 | ip->i_e2fs_gen = ext2gennumber; |
| 218 | return 0; |
| 219 | noinodes: |
| 220 | ext2fs_fserr(fs, kauth_cred_geteuid(cred), "out of inodes" ); |
| 221 | uprintf("\n%s: create/symlink failed, no inodes free\n" , fs->e2fs_fsmnt); |
| 222 | return ENOSPC; |
| 223 | } |
| 224 | |
| 225 | /* |
| 226 | * Find a cylinder to place a directory. |
| 227 | * |
| 228 | * The policy implemented by this algorithm is to select from |
| 229 | * among those cylinder groups with above the average number of |
| 230 | * free inodes, the one with the smallest number of directories. |
| 231 | */ |
| 232 | static u_long |
| 233 | ext2fs_dirpref(struct m_ext2fs *fs) |
| 234 | { |
| 235 | int cg, maxspace, mincg, avgifree; |
| 236 | |
| 237 | avgifree = fs->e2fs.e2fs_ficount / fs->e2fs_ncg; |
| 238 | maxspace = 0; |
| 239 | mincg = -1; |
| 240 | for (cg = 0; cg < fs->e2fs_ncg; cg++) |
| 241 | if (fs2h16(fs->e2fs_gd[cg].ext2bgd_nifree) >= avgifree) { |
| 242 | if (mincg == -1 || fs2h16(fs->e2fs_gd[cg].ext2bgd_nbfree) > maxspace) { |
| 243 | mincg = cg; |
| 244 | maxspace = fs2h16(fs->e2fs_gd[cg].ext2bgd_nbfree); |
| 245 | } |
| 246 | } |
| 247 | return mincg; |
| 248 | } |
| 249 | |
| 250 | /* |
| 251 | * Select the desired position for the next block in a file. The file is |
| 252 | * logically divided into sections. The first section is composed of the |
| 253 | * direct blocks. Each additional section contains fs_maxbpg blocks. |
| 254 | * |
| 255 | * If no blocks have been allocated in the first section, the policy is to |
| 256 | * request a block in the same cylinder group as the inode that describes |
| 257 | * the file. Otherwise, the policy is to try to allocate the blocks |
| 258 | * contigously. The two fields of the ext2 inode extension (see |
| 259 | * ufs/ufs/inode.h) help this. |
| 260 | */ |
| 261 | daddr_t |
| 262 | ext2fs_blkpref(struct inode *ip, daddr_t lbn, int indx, |
| 263 | int32_t *bap /* XXX ondisk32 */) |
| 264 | { |
| 265 | struct m_ext2fs *fs; |
| 266 | int cg, i; |
| 267 | |
| 268 | fs = ip->i_e2fs; |
| 269 | /* |
| 270 | * if we are doing contigous lbn allocation, try to alloc blocks |
| 271 | * contigously on disk |
| 272 | */ |
| 273 | |
| 274 | if ( ip->i_e2fs_last_blk && lbn == ip->i_e2fs_last_lblk + 1) { |
| 275 | return ip->i_e2fs_last_blk + 1; |
| 276 | } |
| 277 | |
| 278 | /* |
| 279 | * bap, if provided, gives us a list of blocks to which we want to |
| 280 | * stay close |
| 281 | */ |
| 282 | |
| 283 | if (bap) { |
| 284 | for (i = indx; i >= 0 ; i--) { |
| 285 | if (bap[i]) { |
| 286 | return fs2h32(bap[i]) + 1; |
| 287 | } |
| 288 | } |
| 289 | } |
| 290 | |
| 291 | /* fall back to the first block of the cylinder containing the inode */ |
| 292 | |
| 293 | cg = ino_to_cg(fs, ip->i_number); |
| 294 | return fs->e2fs.e2fs_bpg * cg + fs->e2fs.e2fs_first_dblock + 1; |
| 295 | } |
| 296 | |
| 297 | /* |
| 298 | * Implement the cylinder overflow algorithm. |
| 299 | * |
| 300 | * The policy implemented by this algorithm is: |
| 301 | * 1) allocate the block in its requested cylinder group. |
| 302 | * 2) quadradically rehash on the cylinder group number. |
| 303 | * 3) brute force search for a free block. |
| 304 | */ |
| 305 | static u_long |
| 306 | ext2fs_hashalloc(struct inode *ip, int cg, long pref, int size, |
| 307 | daddr_t (*allocator)(struct inode *, int, daddr_t, int)) |
| 308 | { |
| 309 | struct m_ext2fs *fs; |
| 310 | long result; |
| 311 | int i, icg = cg; |
| 312 | |
| 313 | fs = ip->i_e2fs; |
| 314 | /* |
| 315 | * 1: preferred cylinder group |
| 316 | */ |
| 317 | result = (*allocator)(ip, cg, pref, size); |
| 318 | if (result) |
| 319 | return result; |
| 320 | /* |
| 321 | * 2: quadratic rehash |
| 322 | */ |
| 323 | for (i = 1; i < fs->e2fs_ncg; i *= 2) { |
| 324 | cg += i; |
| 325 | if (cg >= fs->e2fs_ncg) |
| 326 | cg -= fs->e2fs_ncg; |
| 327 | result = (*allocator)(ip, cg, 0, size); |
| 328 | if (result) |
| 329 | return result; |
| 330 | } |
| 331 | /* |
| 332 | * 3: brute force search |
| 333 | * Note that we start at i == 2, since 0 was checked initially, |
| 334 | * and 1 is always checked in the quadratic rehash. |
| 335 | */ |
| 336 | cg = (icg + 2) % fs->e2fs_ncg; |
| 337 | for (i = 2; i < fs->e2fs_ncg; i++) { |
| 338 | result = (*allocator)(ip, cg, 0, size); |
| 339 | if (result) |
| 340 | return result; |
| 341 | cg++; |
| 342 | if (cg == fs->e2fs_ncg) |
| 343 | cg = 0; |
| 344 | } |
| 345 | return 0; |
| 346 | } |
| 347 | |
| 348 | /* |
| 349 | * Determine whether a block can be allocated. |
| 350 | * |
| 351 | * Check to see if a block of the appropriate size is available, |
| 352 | * and if it is, allocate it. |
| 353 | */ |
| 354 | |
| 355 | static daddr_t |
| 356 | ext2fs_alloccg(struct inode *ip, int cg, daddr_t bpref, int size) |
| 357 | { |
| 358 | struct m_ext2fs *fs; |
| 359 | char *bbp; |
| 360 | struct buf *bp; |
| 361 | /* XXX ondisk32 */ |
| 362 | int error, bno, start, end, loc; |
| 363 | |
| 364 | fs = ip->i_e2fs; |
| 365 | if (fs->e2fs_gd[cg].ext2bgd_nbfree == 0) |
| 366 | return 0; |
| 367 | error = bread(ip->i_devvp, EXT2_FSBTODB(fs, |
| 368 | fs2h32(fs->e2fs_gd[cg].ext2bgd_b_bitmap)), |
| 369 | (int)fs->e2fs_bsize, B_MODIFY, &bp); |
| 370 | if (error) { |
| 371 | return 0; |
| 372 | } |
| 373 | bbp = (char *)bp->b_data; |
| 374 | |
| 375 | if (dtog(fs, bpref) != cg) |
| 376 | bpref = 0; |
| 377 | |
| 378 | /* initialize block bitmap now if uninit */ |
| 379 | if (__predict_false(E2FS_HAS_GD_CSUM(fs) && |
| 380 | (fs->e2fs_gd[cg].ext2bgd_flags & h2fs16(E2FS_BG_BLOCK_UNINIT)))) { |
| 381 | ext2fs_init_bb(fs, cg, &fs->e2fs_gd[cg], bbp); |
| 382 | fs->e2fs_gd[cg].ext2bgd_flags &= h2fs16(~E2FS_BG_BLOCK_UNINIT); |
| 383 | } |
| 384 | |
| 385 | if (bpref != 0) { |
| 386 | bpref = dtogd(fs, bpref); |
| 387 | /* |
| 388 | * if the requested block is available, use it |
| 389 | */ |
| 390 | if (isclr(bbp, bpref)) { |
| 391 | bno = bpref; |
| 392 | goto gotit; |
| 393 | } |
| 394 | } |
| 395 | /* |
| 396 | * no blocks in the requested cylinder, so take next |
| 397 | * available one in this cylinder group. |
| 398 | * first try to get 8 contigous blocks, then fall back to a single |
| 399 | * block. |
| 400 | */ |
| 401 | if (bpref) |
| 402 | start = dtogd(fs, bpref) / NBBY; |
| 403 | else |
| 404 | start = 0; |
| 405 | end = howmany(fs->e2fs.e2fs_fpg, NBBY) - start; |
| 406 | for (loc = start; loc < end; loc++) { |
| 407 | if (bbp[loc] == 0) { |
| 408 | bno = loc * NBBY; |
| 409 | goto gotit; |
| 410 | } |
| 411 | } |
| 412 | for (loc = 0; loc < start; loc++) { |
| 413 | if (bbp[loc] == 0) { |
| 414 | bno = loc * NBBY; |
| 415 | goto gotit; |
| 416 | } |
| 417 | } |
| 418 | |
| 419 | bno = ext2fs_mapsearch(fs, bbp, bpref); |
| 420 | #if 0 |
| 421 | /* |
| 422 | * XXX jdolecek mapsearch actually never fails, it panics instead. |
| 423 | * If re-enabling, make sure to brele() before returning. |
| 424 | */ |
| 425 | if (bno < 0) |
| 426 | return 0; |
| 427 | #endif |
| 428 | gotit: |
| 429 | #ifdef DIAGNOSTIC |
| 430 | if (isset(bbp, (daddr_t)bno)) { |
| 431 | printf("ext2fs_alloccgblk: cg=%d bno=%d fs=%s\n" , |
| 432 | cg, bno, fs->e2fs_fsmnt); |
| 433 | panic("ext2fs_alloccg: dup alloc" ); |
| 434 | } |
| 435 | #endif |
| 436 | setbit(bbp, (daddr_t)bno); |
| 437 | fs->e2fs.e2fs_fbcount--; |
| 438 | ext2fs_cg_update(fs, cg, &fs->e2fs_gd[cg], -1, 0, 0, 0); |
| 439 | fs->e2fs_fmod = 1; |
| 440 | bdwrite(bp); |
| 441 | return cg * fs->e2fs.e2fs_fpg + fs->e2fs.e2fs_first_dblock + bno; |
| 442 | } |
| 443 | |
| 444 | /* |
| 445 | * Determine whether an inode can be allocated. |
| 446 | * |
| 447 | * Check to see if an inode is available, and if it is, |
| 448 | * allocate it using the following policy: |
| 449 | * 1) allocate the requested inode. |
| 450 | * 2) allocate the next available inode after the requested |
| 451 | * inode in the specified cylinder group. |
| 452 | */ |
| 453 | static daddr_t |
| 454 | ext2fs_nodealloccg(struct inode *ip, int cg, daddr_t ipref, int mode) |
| 455 | { |
| 456 | struct m_ext2fs *fs; |
| 457 | char *ibp; |
| 458 | struct buf *bp; |
| 459 | int error, start, len, loc, map, i; |
| 460 | |
| 461 | ipref--; /* to avoid a lot of (ipref -1) */ |
| 462 | if (ipref == -1) |
| 463 | ipref = 0; |
| 464 | fs = ip->i_e2fs; |
| 465 | if (fs->e2fs_gd[cg].ext2bgd_nifree == 0) |
| 466 | return 0; |
| 467 | error = bread(ip->i_devvp, EXT2_FSBTODB(fs, |
| 468 | fs2h32(fs->e2fs_gd[cg].ext2bgd_i_bitmap)), |
| 469 | (int)fs->e2fs_bsize, B_MODIFY, &bp); |
| 470 | if (error) { |
| 471 | return 0; |
| 472 | } |
| 473 | ibp = (char *)bp->b_data; |
| 474 | |
| 475 | KASSERT(!E2FS_HAS_GD_CSUM(fs) || (fs->e2fs_gd[cg].ext2bgd_flags & h2fs16(E2FS_BG_INODE_ZEROED)) != 0); |
| 476 | |
| 477 | /* initialize inode bitmap now if uninit */ |
| 478 | if (__predict_false(E2FS_HAS_GD_CSUM(fs) && |
| 479 | (fs->e2fs_gd[cg].ext2bgd_flags & h2fs16(E2FS_BG_INODE_UNINIT)))) { |
| 480 | KASSERT(fs2h16(fs->e2fs_gd[cg].ext2bgd_nifree) == fs->e2fs.e2fs_ipg); |
| 481 | memset(ibp, 0, fs->e2fs_bsize); |
| 482 | fs->e2fs_gd[cg].ext2bgd_flags &= h2fs16(~E2FS_BG_INODE_UNINIT); |
| 483 | } |
| 484 | |
| 485 | if (ipref) { |
| 486 | ipref %= fs->e2fs.e2fs_ipg; |
| 487 | if (isclr(ibp, ipref)) |
| 488 | goto gotit; |
| 489 | } |
| 490 | start = ipref / NBBY; |
| 491 | len = howmany(fs->e2fs.e2fs_ipg - ipref, NBBY); |
| 492 | loc = skpc(0xff, len, &ibp[start]); |
| 493 | if (loc == 0) { |
| 494 | len = start + 1; |
| 495 | start = 0; |
| 496 | loc = skpc(0xff, len, &ibp[0]); |
| 497 | if (loc == 0) { |
| 498 | printf("cg = %d, ipref = %lld, fs = %s\n" , |
| 499 | cg, (long long)ipref, fs->e2fs_fsmnt); |
| 500 | panic("ext2fs_nodealloccg: map corrupted" ); |
| 501 | /* NOTREACHED */ |
| 502 | } |
| 503 | } |
| 504 | i = start + len - loc; |
| 505 | map = ibp[i] ^ 0xff; |
| 506 | if (map == 0) { |
| 507 | printf("fs = %s\n" , fs->e2fs_fsmnt); |
| 508 | panic("ext2fs_nodealloccg: inode not in map" ); |
| 509 | } |
| 510 | ipref = i * NBBY + ffs(map) - 1; |
| 511 | gotit: |
| 512 | setbit(ibp, ipref); |
| 513 | fs->e2fs.e2fs_ficount--; |
| 514 | ext2fs_cg_update(fs, cg, &fs->e2fs_gd[cg], |
| 515 | 0, -1, ((mode & IFMT) == IFDIR) ? 1 : 0, ipref); |
| 516 | fs->e2fs_fmod = 1; |
| 517 | bdwrite(bp); |
| 518 | return cg * fs->e2fs.e2fs_ipg + ipref + 1; |
| 519 | } |
| 520 | |
| 521 | /* |
| 522 | * Free a block. |
| 523 | * |
| 524 | * The specified block is placed back in the |
| 525 | * free map. |
| 526 | */ |
| 527 | void |
| 528 | ext2fs_blkfree(struct inode *ip, daddr_t bno) |
| 529 | { |
| 530 | struct m_ext2fs *fs; |
| 531 | char *bbp; |
| 532 | struct buf *bp; |
| 533 | int error, cg; |
| 534 | |
| 535 | fs = ip->i_e2fs; |
| 536 | cg = dtog(fs, bno); |
| 537 | |
| 538 | KASSERT(!E2FS_HAS_GD_CSUM(fs) || (fs->e2fs_gd[cg].ext2bgd_flags & h2fs16(E2FS_BG_BLOCK_UNINIT)) == 0); |
| 539 | |
| 540 | if ((u_int)bno >= fs->e2fs.e2fs_bcount) { |
| 541 | printf("bad block %lld, ino %llu\n" , (long long)bno, |
| 542 | (unsigned long long)ip->i_number); |
| 543 | ext2fs_fserr(fs, ip->i_uid, "bad block" ); |
| 544 | return; |
| 545 | } |
| 546 | error = bread(ip->i_devvp, |
| 547 | EXT2_FSBTODB(fs, fs2h32(fs->e2fs_gd[cg].ext2bgd_b_bitmap)), |
| 548 | (int)fs->e2fs_bsize, B_MODIFY, &bp); |
| 549 | if (error) { |
| 550 | return; |
| 551 | } |
| 552 | bbp = (char *)bp->b_data; |
| 553 | bno = dtogd(fs, bno); |
| 554 | if (isclr(bbp, bno)) { |
| 555 | printf("dev = 0x%llx, block = %lld, fs = %s\n" , |
| 556 | (unsigned long long)ip->i_dev, (long long)bno, |
| 557 | fs->e2fs_fsmnt); |
| 558 | panic("blkfree: freeing free block" ); |
| 559 | } |
| 560 | clrbit(bbp, bno); |
| 561 | fs->e2fs.e2fs_fbcount++; |
| 562 | ext2fs_cg_update(fs, cg, &fs->e2fs_gd[cg], 1, 0, 0, 0); |
| 563 | fs->e2fs_fmod = 1; |
| 564 | bdwrite(bp); |
| 565 | } |
| 566 | |
| 567 | /* |
| 568 | * Free an inode. |
| 569 | * |
| 570 | * The specified inode is placed back in the free map. |
| 571 | */ |
| 572 | int |
| 573 | ext2fs_vfree(struct vnode *pvp, ino_t ino, int mode) |
| 574 | { |
| 575 | struct m_ext2fs *fs; |
| 576 | char *ibp; |
| 577 | struct inode *pip; |
| 578 | struct buf *bp; |
| 579 | int error, cg; |
| 580 | |
| 581 | pip = VTOI(pvp); |
| 582 | fs = pip->i_e2fs; |
| 583 | |
| 584 | if ((u_int)ino > fs->e2fs.e2fs_icount || (u_int)ino < EXT2_FIRSTINO) |
| 585 | panic("ifree: range: dev = 0x%llx, ino = %llu, fs = %s" , |
| 586 | (unsigned long long)pip->i_dev, (unsigned long long)ino, |
| 587 | fs->e2fs_fsmnt); |
| 588 | |
| 589 | cg = ino_to_cg(fs, ino); |
| 590 | |
| 591 | KASSERT(!E2FS_HAS_GD_CSUM(fs) || (fs->e2fs_gd[cg].ext2bgd_flags & h2fs16(E2FS_BG_INODE_UNINIT)) == 0); |
| 592 | |
| 593 | error = bread(pip->i_devvp, |
| 594 | EXT2_FSBTODB(fs, fs2h32(fs->e2fs_gd[cg].ext2bgd_i_bitmap)), |
| 595 | (int)fs->e2fs_bsize, B_MODIFY, &bp); |
| 596 | if (error) { |
| 597 | return 0; |
| 598 | } |
| 599 | ibp = (char *)bp->b_data; |
| 600 | ino = (ino - 1) % fs->e2fs.e2fs_ipg; |
| 601 | if (isclr(ibp, ino)) { |
| 602 | printf("dev = 0x%llx, ino = %llu, fs = %s\n" , |
| 603 | (unsigned long long)pip->i_dev, |
| 604 | (unsigned long long)ino, fs->e2fs_fsmnt); |
| 605 | if (fs->e2fs_ronly == 0) |
| 606 | panic("ifree: freeing free inode" ); |
| 607 | } |
| 608 | clrbit(ibp, ino); |
| 609 | fs->e2fs.e2fs_ficount++; |
| 610 | ext2fs_cg_update(fs, cg, &fs->e2fs_gd[cg], |
| 611 | 0, 1, ((mode & IFMT) == IFDIR) ? -1 : 0, 0); |
| 612 | fs->e2fs_fmod = 1; |
| 613 | bdwrite(bp); |
| 614 | return 0; |
| 615 | } |
| 616 | |
| 617 | /* |
| 618 | * Find a block in the specified cylinder group. |
| 619 | * |
| 620 | * It is a panic if a request is made to find a block if none are |
| 621 | * available. |
| 622 | */ |
| 623 | |
| 624 | static daddr_t |
| 625 | ext2fs_mapsearch(struct m_ext2fs *fs, char *bbp, daddr_t bpref) |
| 626 | { |
| 627 | int start, len, loc, i, map; |
| 628 | |
| 629 | /* |
| 630 | * find the fragment by searching through the free block |
| 631 | * map for an appropriate bit pattern |
| 632 | */ |
| 633 | if (bpref) |
| 634 | start = dtogd(fs, bpref) / NBBY; |
| 635 | else |
| 636 | start = 0; |
| 637 | len = howmany(fs->e2fs.e2fs_fpg, NBBY) - start; |
| 638 | loc = skpc(0xff, len, &bbp[start]); |
| 639 | if (loc == 0) { |
| 640 | len = start + 1; |
| 641 | start = 0; |
| 642 | loc = skpc(0xff, len, &bbp[start]); |
| 643 | if (loc == 0) { |
| 644 | printf("start = %d, len = %d, fs = %s\n" , |
| 645 | start, len, fs->e2fs_fsmnt); |
| 646 | panic("ext2fs_alloccg: map corrupted" ); |
| 647 | /* NOTREACHED */ |
| 648 | } |
| 649 | } |
| 650 | i = start + len - loc; |
| 651 | map = bbp[i] ^ 0xff; |
| 652 | if (map == 0) { |
| 653 | printf("fs = %s\n" , fs->e2fs_fsmnt); |
| 654 | panic("ext2fs_mapsearch: block not in map" ); |
| 655 | } |
| 656 | return i * NBBY + ffs(map) - 1; |
| 657 | } |
| 658 | |
| 659 | /* |
| 660 | * Fserr prints the name of a file system with an error diagnostic. |
| 661 | * |
| 662 | * The form of the error message is: |
| 663 | * fs: error message |
| 664 | */ |
| 665 | static void |
| 666 | ext2fs_fserr(struct m_ext2fs *fs, u_int uid, const char *cp) |
| 667 | { |
| 668 | |
| 669 | log(LOG_ERR, "uid %d on %s: %s\n" , uid, fs->e2fs_fsmnt, cp); |
| 670 | } |
| 671 | |
| 672 | static __inline void |
| 673 | ext2fs_cg_update(struct m_ext2fs *fs, int cg, struct ext2_gd *gd, int nbfree, int nifree, int ndirs, daddr_t ioff) |
| 674 | { |
| 675 | /* XXX disk32 */ |
| 676 | if (nifree) { |
| 677 | gd->ext2bgd_nifree = h2fs16(fs2h16(gd->ext2bgd_nifree) + nifree); |
| 678 | /* |
| 679 | * If we allocated inode on bigger offset than what was |
| 680 | * ever used before, bump the itable_unused count. This |
| 681 | * member only ever grows, and is used only for initialization |
| 682 | * !INODE_ZEROED groups with used inodes. Of course, by the |
| 683 | * time we get here the itables are already zeroed, but |
| 684 | * e2fstools fsck.ext4 still checks this. |
| 685 | */ |
| 686 | if (E2FS_HAS_GD_CSUM(fs) && nifree < 0 && (ioff+1) >= (fs->e2fs.e2fs_ipg - fs2h16(gd->ext2bgd_itable_unused_lo))) { |
| 687 | gd->ext2bgd_itable_unused_lo = h2fs16(fs->e2fs.e2fs_ipg - (ioff + 1)); |
| 688 | } |
| 689 | |
| 690 | KASSERT(!E2FS_HAS_GD_CSUM(fs) || gd->ext2bgd_itable_unused_lo <= gd->ext2bgd_nifree); |
| 691 | } |
| 692 | |
| 693 | |
| 694 | if (nbfree) |
| 695 | gd->ext2bgd_nbfree = h2fs16(fs2h16(gd->ext2bgd_nbfree) + nbfree); |
| 696 | |
| 697 | if (ndirs) |
| 698 | gd->ext2bgd_ndirs = h2fs16(fs2h16(gd->ext2bgd_ndirs) + ndirs); |
| 699 | |
| 700 | if (E2FS_HAS_GD_CSUM(fs)) |
| 701 | gd->ext2bgd_checksum = ext2fs_cg_get_csum(fs, cg, gd); |
| 702 | } |
| 703 | |
| 704 | /* |
| 705 | * Compute group description csum. Structure data must be LE (not host). |
| 706 | * Returned as LE (disk encoding). |
| 707 | */ |
| 708 | static uint16_t |
| 709 | ext2fs_cg_get_csum(struct m_ext2fs *fs, int cg, struct ext2_gd *gd) |
| 710 | { |
| 711 | uint16_t crc; |
| 712 | uint32_t cg_bswapped = h2fs32((uint32_t)cg); |
| 713 | size_t off; |
| 714 | |
| 715 | if (!EXT2F_HAS_ROCOMPAT_FEATURE(fs, EXT2F_ROCOMPAT_GDT_CSUM)) |
| 716 | return 0; |
| 717 | |
| 718 | off = offsetof(struct ext2_gd, ext2bgd_checksum); |
| 719 | |
| 720 | crc = crc16(~0, (uint8_t *)fs->e2fs.e2fs_uuid, sizeof(fs->e2fs.e2fs_uuid)); |
| 721 | crc = crc16(crc, (uint8_t *)&cg_bswapped, sizeof(cg_bswapped)); |
| 722 | crc = crc16(crc, (uint8_t *)gd, off); |
| 723 | /* XXX ondisk32 */ |
| 724 | |
| 725 | return h2fs16(crc); |
| 726 | } |
| 727 | |
| 728 | static void |
| 729 | ext2fs_init_bb(struct m_ext2fs *fs, int cg, struct ext2_gd *gd, char *bbp) |
| 730 | { |
| 731 | int i; |
| 732 | |
| 733 | memset(bbp, 0, fs->e2fs_bsize); |
| 734 | |
| 735 | /* |
| 736 | * No block was ever allocated on this cg before, so the only used |
| 737 | * blocks are metadata blocks on start of the group. We could optimize |
| 738 | * this to set by bytes, but since this is done once per the group |
| 739 | * in lifetime of filesystem, it really is not worth it. |
| 740 | */ |
| 741 | for(i=0; i < fs->e2fs.e2fs_bpg - fs2h16(gd->ext2bgd_nbfree); i++) |
| 742 | setbit(bbp, i); |
| 743 | } |
| 744 | |
| 745 | /* |
| 746 | * Verify csum and initialize itable if not done already |
| 747 | */ |
| 748 | int |
| 749 | ext2fs_cg_verify_and_initialize(struct vnode *devvp, struct m_ext2fs *fs, int ronly) |
| 750 | { |
| 751 | /* XXX disk32 */ |
| 752 | struct ext2_gd *gd; |
| 753 | ino_t ioff; |
| 754 | size_t boff; |
| 755 | struct buf *bp; |
| 756 | int cg, i, error; |
| 757 | |
| 758 | if (!E2FS_HAS_GD_CSUM(fs)) |
| 759 | return 0; |
| 760 | |
| 761 | for(cg=0; cg < fs->e2fs_ncg; cg++) { |
| 762 | gd = &fs->e2fs_gd[cg]; |
| 763 | |
| 764 | /* Verify checksum */ |
| 765 | if (gd->ext2bgd_checksum != ext2fs_cg_get_csum(fs, cg, gd)) { |
| 766 | printf("ext2fs_cg_verify_and_initialize: group %d invalid csum\n" , cg); |
| 767 | return EINVAL; |
| 768 | } |
| 769 | |
| 770 | /* if mounting read-write, zero itable if not already done */ |
| 771 | if (ronly || (gd->ext2bgd_flags & h2fs16(E2FS_BG_INODE_ZEROED)) != 0) |
| 772 | continue; |
| 773 | |
| 774 | /* |
| 775 | * We are skipping already used inodes, zero rest of itable |
| 776 | * blocks. First block to zero could be only partial wipe, all |
| 777 | * others are wiped completely. This might take a while, |
| 778 | * there could be many inode table blocks. We use |
| 779 | * delayed writes, so this shouldn't block for very |
| 780 | * long. |
| 781 | */ |
| 782 | ioff = fs->e2fs.e2fs_ipg - fs2h16(gd->ext2bgd_itable_unused_lo); |
| 783 | boff = (ioff % fs->e2fs_ipb) * EXT2_DINODE_SIZE(fs); |
| 784 | |
| 785 | for(i = ioff / fs->e2fs_ipb; i < fs->e2fs_itpg; i++) { |
| 786 | if (boff) { |
| 787 | /* partial wipe, must read old data */ |
| 788 | error = bread(devvp, |
| 789 | EXT2_FSBTODB(fs, fs2h32(gd->ext2bgd_i_tables) + i), |
| 790 | (int)fs->e2fs_bsize, B_MODIFY, &bp); |
| 791 | if (error) { |
| 792 | printf("ext2fs_cg_verify_and_initialize: can't read itable block" ); |
| 793 | return error; |
| 794 | } |
| 795 | memset((char *)bp->b_data + boff, 0, fs->e2fs_bsize - boff); |
| 796 | boff = 0; |
| 797 | } else { |
| 798 | /* |
| 799 | * Complete wipe, don't need to read data. This |
| 800 | * assumes nothing else is changing the data. |
| 801 | */ |
| 802 | bp = getblk(devvp, |
| 803 | EXT2_FSBTODB(fs, fs2h32(gd->ext2bgd_i_tables) + i), |
| 804 | (int)fs->e2fs_bsize, 0, 0); |
| 805 | clrbuf(bp); |
| 806 | } |
| 807 | |
| 808 | bdwrite(bp); |
| 809 | } |
| 810 | |
| 811 | gd->ext2bgd_flags |= h2fs16(E2FS_BG_INODE_ZEROED); |
| 812 | gd->ext2bgd_checksum = ext2fs_cg_get_csum(fs, cg, gd); |
| 813 | fs->e2fs_fmod = 1; |
| 814 | } |
| 815 | |
| 816 | return 0; |
| 817 | } |
| 818 | |