1 /* Copyright (c) 1981 Regents of the University of California */ 2 3 /* fs.h 2.1 03/25/82 */ 4 5 /* 6 * Each disk drive contains some number of file systems. 7 * A file system consists of a number of cylinder groups. 8 * Each cylinder group has inodes and data. 9 * 10 * A file system is described by its super-block, which in turn 11 * describes the cylinder groups. The super-block is critical 12 * data and is replicated in each cylinder group to protect against 13 * catastrophic loss. This is done at mkfs time and the critical 14 * super-block data does not change, so the copies need not be 15 * referenced further unless disaster strikes. 16 * 17 * For file system fs, the offsets of the various blocks of interest 18 * are given in the super block as: 19 * [fs->fs_bblkno] Boot sector 20 * [fs->fs_sblkno] Super-block 21 * [fs->fs_cblkno] Cylinder group block 22 * [fs->fs_iblkno] Inode blocks 23 * [fs->fs_dblkno] Data blocks 24 * The beginning of cylinder group cg in fs, is given by 25 * the ``cgbase(fs, cg)'' macro. 26 * 27 * The first boot and super blocks are given in absolute disk addresses. 28 */ 29 #define BBSIZE 8192 30 #define SBSIZE 8192 31 #define BBLOCK ((daddr_t)(0)) 32 #define SBLOCK ((daddr_t)(BBLOCK + BBSIZE / DEV_BSIZE)) 33 34 /* 35 * Addresses stored in inodes are capable of addressing fragments 36 * of `blocks'. File system blocks of at most size MAXBSIZE can 37 * be optionally broken into 2, 4, or 8 pieces, each of which is 38 * addressible; these pieces may be DEV_BSIZE, or some multiple of 39 * a DEV_BSIZE unit. 40 * 41 * Large files consist of exclusively large data blocks. To avoid 42 * undue wasted disk space, the last data block of a small file may be 43 * allocated as only as many fragments of a large block as are 44 * necessary. The file system format retains only a single pointer 45 * to such a fragment, which is a piece of a single large block that 46 * has been divided. The size of such a fragment is determinable from 47 * information in the inode, using the ``blksize(fs, ip, lbn)'' macro. 48 * 49 * The file system records space availability at the fragment level; 50 * to determine block availability, aligned fragments are examined. 51 * 52 * The root inode is the root of the file system. 53 * Inode 0 can't be used for normal purposes and 54 * historically bad blocks were linked to inode 1, 55 * thus the root inode is 2. (inode 1 is no longer used for 56 * this purpose, however numerous dump tapes make this 57 * assumption, so we are stuck with it) 58 * The lost+found directory is given the next available 59 * inode when it is created by ``mkfs''. 60 */ 61 #define ROOTINO ((ino_t)2) /* i number of all roots */ 62 #define LOSTFOUNDINO (ROOTINO + 1) 63 64 /* 65 * MINFREE gives the minimum acceptable percentage of file system 66 * blocks which may be free. If the freelist drops below this level 67 * only the superuser may continue to allocate blocks. This may 68 * be set to 0 if no reserve of free blocks is deemed necessary, 69 * however severe performance degredations will be observed if the 70 * file system is run at greater than 90% full; thus the default 71 * value of fs_minfree is 10%. 72 * 73 * Empirically the best trade-off between block fragmentation and 74 * overall disk utilization at a loading of 90% comes with a 75 * fragmentation of 4, thus the default fragment size is a fourth 76 * of the block size. 77 */ 78 #define MINFREE 10 79 #define DESFRAG 4 80 81 /* 82 * Under current technology, most 300MB disks have 32 sectors and 83 * 19 tracks, thus these are the defaults used for fs_nsect and 84 * fs_ntrak respectively. 85 */ 86 #define DFLNSECT 32 87 #define DFLNTRAK 19 88 89 /* 90 * Cylinder group related limits. 91 * 92 * For each cylinder we keep track of the availability of blocks at different 93 * rotational positions, so that we can lay out the data to be picked 94 * up with minimum rotational latency. NRPOS is the number of rotational 95 * positions which we distinguish. With NRPOS 8 the resolution of our 96 * summary information is 2ms for a typical 3600 rpm drive. 97 * 98 * ROTDELAY gives the minimum number of milliseconds to initiate 99 * another disk transfer on the same cylinder. It is used in 100 * determining the rotationally optimal layout for disk blocks 101 * within a file; the default of fs_rotdelay is 2ms. 102 */ 103 #define NRPOS 8 /* number distinct rotational positions */ 104 #define ROTDELAY 2 105 106 /* 107 * Each file system has a number of inodes statically allocated. 108 * We allocate one inode slot per NBPI bytes, expecting this 109 * to be far more than we will ever need. 110 * 111 * MAXIPG bounds the number of inodes per cylinder group, and 112 * is needed only to keep the structure simpler by having the 113 * only a single variable size element (the free bit map). 114 * 115 * N.B.: MAXIPG must be a multiple of INOPB(fs). 116 */ 117 #define NBPI 2048 118 #define MAXIPG 2048 /* max number inodes/cyl group */ 119 120 /* 121 * MINBSIZE is the smallest allowable block size. 122 * In order to insure that it is possible to create files of size 123 * 2^32 with only two levels of indirection, MINBSIZE is set to 4096. 124 * MINBSIZE must be big enough to hold a cylinder group block, 125 * thus changes to (struct cg) must keep its size within MINBSIZE. 126 * MAXCPG is limited only to dimension an array in (struct cg); 127 * it can be made larger as long as that structures size remains 128 * within the bounds dictated by MINBSIZE. 129 * Note that super blocks are always of size MAXBSIZE, 130 * and that MAXBSIZE must be >= MINBSIZE. 131 */ 132 #define MINBSIZE 4096 133 #define DESCPG 16 /* desired fs_cpg */ 134 #define MAXCPG 32 /* maximum fs_cpg */ 135 136 /* 137 * The path name on which the file system is mounted is maintained 138 * in fs_fsmnt. MAXMNTLEN defines the amount of space allocated in 139 * the super block for this name. 140 * The limit on the amount of summary information per file system 141 * is defined by MAXCSBUFS. It is currently parameterized for 1Meg 142 * cylinders maximum. 143 */ 144 #define MAXMNTLEN 34 145 #define MAXCSBUFS 16 146 147 /* 148 * Per cylinder group information; summarized in blocks allocated 149 * from first cylinder group data blocks. These blocks have to be 150 * read in from fs_csaddr (size fs_cssize) in addition to the 151 * super block. 152 * 153 * N.B. sizeof(struct csum) must be a power of two in order for 154 * the ``fs_cs'' macro to work (see below). 155 */ 156 struct csum { 157 long cs_ndir; /* number of directories */ 158 long cs_nbfree; /* number of free blocks */ 159 long cs_nifree; /* number of free inodes */ 160 long cs_nffree; /* number of free frags */ 161 }; 162 163 /* 164 * Super block for a file system. 165 */ 166 #define FS_MAGIC 0x110854 167 struct fs 168 { 169 long fs_magic; /* magic number */ 170 daddr_t fs_bblkno; /* abs addr of boot-block in filesys */ 171 daddr_t fs_sblkno; /* abs addr of super-block in filesys */ 172 daddr_t fs_cblkno; /* offset of cyl-block in filesys */ 173 daddr_t fs_iblkno; /* offset of inode-blocks in filesys */ 174 daddr_t fs_dblkno; /* offset of data-blocks in filesys */ 175 time_t fs_time; /* last time written */ 176 long fs_size; /* number of blocks in fs */ 177 long fs_dsize; /* number of data blocks in fs */ 178 long fs_ncg; /* number of cylinder groups */ 179 long fs_bsize; /* size of basic blocks in fs */ 180 long fs_fsize; /* size of frag blocks in fs */ 181 short fs_frag; /* number of frags in a block in fs */ 182 short fs_minfree; /* minimum percentage of free blocks */ 183 short fs_rotdelay; /* num of ms for optimal next block */ 184 short fs_rps; /* disk revolutions per second */ 185 long fs_bmask; /* ``blkoff'' calc of blk offsets */ 186 long fs_fmask; /* ``fragoff'' calc of frag offsets */ 187 short fs_bshift; /* ``lblkno'' calc of logical blkno */ 188 short fs_fshift; /* ``numfrags'' calc number of frags */ 189 /* sizes determined by number of cylinder groups and their sizes */ 190 daddr_t fs_csaddr; /* blk addr of cyl grp summary area */ 191 long fs_cssize; /* size of cyl grp summary area */ 192 long fs_cgsize; /* cylinder group size */ 193 /* these fields should be derived from the hardware */ 194 short fs_ntrak; /* tracks per cylinder */ 195 short fs_nsect; /* sectors per track */ 196 long fs_spc; /* sectors per cylinder */ 197 /* this comes from the disk driver partitioning */ 198 long fs_ncyl; /* cylinders in file system */ 199 /* these fields can be computed from the others */ 200 short fs_cpg; /* cylinders per group */ 201 short fs_ipg; /* inodes per group */ 202 long fs_fpg; /* blocks per group * fs_frag */ 203 /* this data must be re-computed after crashes */ 204 struct csum fs_cstotal; /* cylinder summary information */ 205 /* these fields are cleared at mount time */ 206 char fs_fmod; /* super block modified flag */ 207 char fs_ronly; /* mounted read-only flag */ 208 char fs_fsmnt[MAXMNTLEN]; /* name mounted on */ 209 /* these fields retain the current block allocation info */ 210 long fs_cgrotor; /* last cg searched */ 211 struct csum *fs_csp[MAXCSBUFS];/* list of fs_cs info buffers */ 212 short fs_cpc; /* cyl per cycle in postbl */ 213 short fs_postbl[MAXCPG][NRPOS];/* head of blocks for each rotation */ 214 u_char fs_rotbl[1]; /* list of blocks for each rotation */ 215 /* actually longer */ 216 }; 217 218 /* 219 * Convert cylinder group to base address of its global summary info. 220 * 221 * N.B. This macro assumes that sizeof(struct csum) is a power of two. 222 */ 223 #define fs_cs(fs, indx) \ 224 fs_csp[(indx) / ((fs)->fs_bsize / sizeof(struct csum))] \ 225 [(indx) % ((fs)->fs_bsize / sizeof(struct csum))] 226 227 /* 228 * MAXBPC bounds the size of the rotational layout tables and 229 * is limited by the fact that the super block is of size SBSIZE. 230 * The size of these tables is INVERSELY proportional to the block 231 * size of the file system. It is aggravated by sector sizes that 232 * are not powers of two, as this increases the number of cylinders 233 * included before the rotational pattern repeats (fs_cpc). 234 * Its size is derived from the number of bytes remaining in (struct fs) 235 */ 236 #define MAXBPC (SBSIZE - sizeof (struct fs)) 237 238 /* 239 * Cylinder group block for a file system. 240 */ 241 #define CG_MAGIC 0x092752 242 struct cg { 243 long cg_magic; /* magic number */ 244 time_t cg_time; /* time last written */ 245 long cg_cgx; /* we are the cgx'th cylinder group */ 246 short cg_ncyl; /* number of cyl's this cg */ 247 short cg_niblk; /* number of inode blocks this cg */ 248 long cg_ndblk; /* number of data blocks this cg */ 249 struct csum cg_cs; /* cylinder summary information */ 250 long cg_rotor; /* position of last used block */ 251 long cg_frotor; /* position of last used frag */ 252 long cg_irotor; /* position of last used inode */ 253 long cg_frsum[MAXFRAG]; /* counts of available frags */ 254 long cg_btot[MAXCPG]; /* block totals per cylinder */ 255 short cg_b[MAXCPG][NRPOS]; /* positions of free blocks */ 256 char cg_iused[MAXIPG/NBBY]; /* used inode map */ 257 char cg_free[1]; /* free block map */ 258 /* actually longer */ 259 }; 260 261 /* 262 * MAXBPG bounds the number of blocks of data per cylinder group, 263 * and is limited by the fact that cylinder groups are at most one block. 264 * Its size is derived from the size of blocks and the (struct cg) size, 265 * by the number of remaining bits. 266 */ 267 #define MAXBPG(fs) \ 268 (NBBY * ((fs)->fs_bsize - (sizeof (struct cg))) / (fs)->fs_frag) 269 270 /* 271 * Turn file system block numbers into disk block addresses. 272 * This maps file system blocks to device size blocks. 273 */ 274 #define fsbtodb(fs, b) ((b) * ((fs)->fs_fsize / DEV_BSIZE)) 275 #define dbtofsb(fs, b) ((b) / ((fs)->fs_fsize / DEV_BSIZE)) 276 277 /* 278 * Cylinder group macros to locate things in cylinder groups. 279 * 280 * Cylinder group to disk block address of spare boot block 281 * and super block. 282 * Note that these are in absolute addresses, and can NOT 283 * in general be expressable in terms of file system addresses. 284 */ 285 #define cgbblock(fs, c) (fsbtodb(fs, cgbase(fs, c)) + (fs)->fs_bblkno) 286 #define cgsblock(fs, c) (fsbtodb(fs, cgbase(fs, c)) + (fs)->fs_sblkno) 287 288 /* 289 * File system addresses of cylinder group data structures. 290 */ 291 #define cgbase(fs, c) ((daddr_t)((fs)->fs_fpg * (c))) /* base addr */ 292 #define cgtod(fs, c) (cgbase(fs, c) + (fs)->fs_cblkno) /* cg block */ 293 #define cgimin(fs, c) (cgbase(fs, c) + (fs)->fs_iblkno) /* inode blk */ 294 #define cgdmin(fs, c) (cgbase(fs, c) + (fs)->fs_dblkno) /* 1st data */ 295 296 /* 297 * Macros for handling inode numbers: 298 * inode number to file system block offset. 299 * inode number to cylinder group number. 300 * inode number to file system block address. 301 */ 302 #define itoo(fs, x) ((x) % INOPB(fs)) 303 #define itog(fs, x) ((x) / (fs)->fs_ipg) 304 #define itod(fs, x) \ 305 ((daddr_t)(cgimin(fs, itog(fs, x)) + \ 306 (x) % (fs)->fs_ipg / INOPB(fs) * (fs)->fs_frag)) 307 308 /* 309 * Give cylinder group number for a file system block. 310 * Give cylinder group block number for a file system block. 311 */ 312 #define dtog(fs, d) ((d) / (fs)->fs_fpg) 313 #define dtogd(fs, d) ((d) % (fs)->fs_fpg) 314 315 /* 316 * Extract the bits for a block from a map. 317 * Compute the cylinder and rotational position of a cyl block addr. 318 */ 319 #define blkmap(fs, map, loc) \ 320 (((map)[loc / NBBY] >> (loc % NBBY)) & (0xff >> (NBBY - (fs)->fs_frag))) 321 #define cbtocylno(fs, bno) \ 322 ((bno) * NSPF(fs) / (fs)->fs_spc) 323 #define cbtorpos(fs, bno) \ 324 ((bno) * NSPF(fs) % (fs)->fs_nsect * NRPOS / (fs)->fs_nsect) 325 326 /* 327 * The following macros optimize certain frequently calculated 328 * quantities by using shifts and masks in place of divisions 329 * modulos and multiplications. 330 */ 331 #define blkoff(fs, loc) /* calculates (loc % fs->fs_bsize) */ \ 332 ((loc) & ~(fs)->fs_bmask) 333 #define fragoff(fs, loc) /* calculates (loc % fs->fs_fsize) */ \ 334 ((loc) & ~(fs)->fs_fmask) 335 #define lblkno(fs, loc) /* calculates (loc / fs->fs_bsize) */ \ 336 ((loc) >> (fs)->fs_bshift) 337 #define numfrags(fs, loc) /* calculates (loc / fs->fs_fsize) */ \ 338 ((loc) >> (fs)->fs_fshift) 339 #define blkroundup(fs, size) /* calculates roundup(size, fs->fs_bsize) */ \ 340 (((size) + (fs)->fs_bsize - 1) & (fs)->fs_bmask) 341 #define fragroundup(fs, size) /* calculates roundup(size, fs->fs_fsize) */ \ 342 (((size) + (fs)->fs_fsize - 1) & (fs)->fs_fmask) 343 344 /* 345 * Determining the size of a file block in the file system. 346 */ 347 #define blksize(fs, ip, lbn) \ 348 (((lbn) >= NDADDR || (ip)->i_size >= ((lbn) + 1) * (fs)->fs_bsize) \ 349 ? (fs)->fs_bsize \ 350 : (fragroundup(fs, blkoff(fs, (ip)->i_size)))) 351 #define dblksize(fs, dip, lbn) \ 352 (((lbn) >= NDADDR || (dip)->di_size >= ((lbn) + 1) * (fs)->fs_bsize) \ 353 ? (fs)->fs_bsize \ 354 : (fragroundup(fs, blkoff(fs, (dip)->di_size)))) 355 356 /* 357 * Number of disk sectors per block; assumes DEV_BSIZE byte sector size. 358 */ 359 #define NSPB(fs) ((fs)->fs_bsize / DEV_BSIZE) 360 #define NSPF(fs) ((fs)->fs_fsize / DEV_BSIZE) 361 362 /* 363 * INOPB is the number of inodes in a secondary storage block. 364 */ 365 #define INOPB(fs) ((fs)->fs_bsize / sizeof (struct dinode)) 366 #define INOPF(fs) ((fs)->fs_fsize / sizeof (struct dinode)) 367 368 /* 369 * NINDIR is the number of indirects in a file system block. 370 */ 371 #define NINDIR(fs) ((fs)->fs_bsize / sizeof (daddr_t)) 372 373 #ifdef KERNEL 374 struct fs *getfs(); 375 #endif 376