1 /* Copyright (c) 1981 Regents of the University of California */ 2 3 /* fs.h 4.1 82/04/19 */ 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_sblkno] Super-block 20 * [fs->fs_cblkno] Cylinder group block 21 * [fs->fs_iblkno] Inode blocks 22 * [fs->fs_dblkno] Data blocks 23 * The beginning of cylinder group cg in fs, is given by 24 * the ``cgbase(fs, cg)'' macro. 25 * 26 * The first boot and super blocks are given in absolute disk addresses. 27 */ 28 #define BBSIZE 8192 29 #define SBSIZE 8192 30 #define BBLOCK ((daddr_t)(0)) 31 #define SBLOCK ((daddr_t)(BBLOCK + BBSIZE / DEV_BSIZE)) 32 33 /* 34 * Addresses stored in inodes are capable of addressing fragments 35 * of `blocks'. File system blocks of at most size MAXBSIZE can 36 * be optionally broken into 2, 4, or 8 pieces, each of which is 37 * addressible; these pieces may be DEV_BSIZE, or some multiple of 38 * a DEV_BSIZE unit. 39 * 40 * Large files consist of exclusively large data blocks. To avoid 41 * undue wasted disk space, the last data block of a small file may be 42 * allocated as only as many fragments of a large block as are 43 * necessary. The file system format retains only a single pointer 44 * to such a fragment, which is a piece of a single large block that 45 * has been divided. The size of such a fragment is determinable from 46 * information in the inode, using the ``blksize(fs, ip, lbn)'' macro. 47 * 48 * The file system records space availability at the fragment level; 49 * to determine block availability, aligned fragments are examined. 50 * 51 * The root inode is the root of the file system. 52 * Inode 0 can't be used for normal purposes and 53 * historically bad blocks were linked to inode 1, 54 * thus the root inode is 2. (inode 1 is no longer used for 55 * this purpose, however numerous dump tapes make this 56 * assumption, so we are stuck with it) 57 * The lost+found directory is given the next available 58 * inode when it is created by ``mkfs''. 59 */ 60 #define ROOTINO ((ino_t)2) /* i number of all roots */ 61 #define LOSTFOUNDINO (ROOTINO + 1) 62 63 /* 64 * MINFREE gives the minimum acceptable percentage of file system 65 * blocks which may be free. If the freelist drops below this level 66 * only the superuser may continue to allocate blocks. This may 67 * be set to 0 if no reserve of free blocks is deemed necessary, 68 * however severe performance degredations will be observed if the 69 * file system is run at greater than 90% full; thus the default 70 * value of fs_minfree is 10%. 71 * 72 * Empirically the best trade-off between block fragmentation and 73 * overall disk utilization at a loading of 90% comes with a 74 * fragmentation of 4, thus the default fragment size is a fourth 75 * of the block size. 76 */ 77 #define MINFREE 10 78 #define DESFRAG 4 79 80 /* 81 * Under current technology, most 300MB disks have 32 sectors and 82 * 16 tracks, thus these are the defaults used for fs_nsect and 83 * fs_ntrak respectively. 84 */ 85 #define DFLNSECT 32 86 #define DFLNTRAK 16 87 88 /* 89 * Cylinder group related limits. 90 * 91 * For each cylinder we keep track of the availability of blocks at different 92 * rotational positions, so that we can lay out the data to be picked 93 * up with minimum rotational latency. NRPOS is the number of rotational 94 * positions which we distinguish. With NRPOS 8 the resolution of our 95 * summary information is 2ms for a typical 3600 rpm drive. 96 * 97 * ROTDELAY gives the minimum number of milliseconds to initiate 98 * another disk transfer on the same cylinder. It is used in 99 * determining the rotationally optimal layout for disk blocks 100 * within a file; the default of fs_rotdelay is 2ms. 101 */ 102 #define NRPOS 8 /* number distinct rotational positions */ 103 #define ROTDELAY 2 104 105 /* 106 * Each file system has a number of inodes statically allocated. 107 * We allocate one inode slot per NBPI bytes, expecting this 108 * to be far more than we will ever need. 109 * 110 * MAXIPG bounds the number of inodes per cylinder group, and 111 * is needed only to keep the structure simpler by having the 112 * only a single variable size element (the free bit map). 113 * 114 * N.B.: MAXIPG must be a multiple of INOPB(fs). 115 */ 116 #define NBPI 2048 117 #define MAXIPG 2048 /* max number inodes/cyl group */ 118 119 /* 120 * MINBSIZE is the smallest allowable block size. 121 * In order to insure that it is possible to create files of size 122 * 2^32 with only two levels of indirection, MINBSIZE is set to 4096. 123 * MINBSIZE must be big enough to hold a cylinder group block, 124 * thus changes to (struct cg) must keep its size within MINBSIZE. 125 * MAXCPG is limited only to dimension an array in (struct cg); 126 * it can be made larger as long as that structures size remains 127 * within the bounds dictated by MINBSIZE. 128 * Note that super blocks are always of size MAXBSIZE, 129 * and that MAXBSIZE must be >= MINBSIZE. 130 */ 131 #define MINBSIZE 4096 132 #define DESCPG 16 /* desired fs_cpg */ 133 #define MAXCPG 32 /* maximum fs_cpg */ 134 135 /* 136 * The path name on which the file system is mounted is maintained 137 * in fs_fsmnt. MAXMNTLEN defines the amount of space allocated in 138 * the super block for this name. 139 * The limit on the amount of summary information per file system 140 * is defined by MAXCSBUFS. It is currently parameterized for a 141 * maximum of two million cylinders. 142 */ 143 #define MAXMNTLEN 512 144 #define MAXCSBUFS 32 145 146 /* 147 * Per cylinder group information; summarized in blocks allocated 148 * from first cylinder group data blocks. These blocks have to be 149 * read in from fs_csaddr (size fs_cssize) in addition to the 150 * super block. 151 * 152 * N.B. sizeof(struct csum) must be a power of two in order for 153 * the ``fs_cs'' macro to work (see below). 154 */ 155 struct csum { 156 long cs_ndir; /* number of directories */ 157 long cs_nbfree; /* number of free blocks */ 158 long cs_nifree; /* number of free inodes */ 159 long cs_nffree; /* number of free frags */ 160 }; 161 162 /* 163 * Super block for a file system. 164 */ 165 #define FS_MAGIC 0x011954 166 struct fs 167 { 168 struct fs *fs_link; /* linked list of file systems */ 169 struct fs *fs_rlink; /* used for incore super blocks */ 170 daddr_t fs_sblkno; /* addr of super-block in filesys */ 171 daddr_t fs_cblkno; /* offset of cyl-block in filesys */ 172 daddr_t fs_iblkno; /* offset of inode-blocks in filesys */ 173 daddr_t fs_dblkno; /* offset of first data after cg */ 174 long fs_cgoffset; /* cylinder group offset in cylinder */ 175 long fs_cgmask; /* used to calc mod fs_ntrak */ 176 time_t fs_time; /* last time written */ 177 long fs_size; /* number of blocks in fs */ 178 long fs_dsize; /* number of data blocks in fs */ 179 long fs_ncg; /* number of cylinder groups */ 180 long fs_bsize; /* size of basic blocks in fs */ 181 long fs_fsize; /* size of frag blocks in fs */ 182 long fs_frag; /* number of frags in a block in fs */ 183 long fs_minfree; /* minimum percentage of free blocks */ 184 long fs_rotdelay; /* num of ms for optimal next block */ 185 long fs_rps; /* disk revolutions per second */ 186 long fs_bmask; /* ``blkoff'' calc of blk offsets */ 187 long fs_fmask; /* ``fragoff'' calc of frag offsets */ 188 long fs_bshift; /* ``lblkno'' calc of logical blkno */ 189 long fs_fshift; /* ``numfrags'' calc number of frags */ 190 long fs_sparecon[16]; /* reserved for future constants */ 191 /* sizes determined by number of cylinder groups and their sizes */ 192 daddr_t fs_csaddr; /* blk addr of cyl grp summary area */ 193 long fs_cssize; /* size of cyl grp summary area */ 194 long fs_cgsize; /* cylinder group size */ 195 /* these fields should be derived from the hardware */ 196 long fs_ntrak; /* tracks per cylinder */ 197 long fs_nsect; /* sectors per track */ 198 long fs_spc; /* sectors per cylinder */ 199 /* this comes from the disk driver partitioning */ 200 long fs_ncyl; /* cylinders in file system */ 201 /* these fields can be computed from the others */ 202 long fs_cpg; /* cylinders per group */ 203 long fs_ipg; /* inodes per group */ 204 long fs_fpg; /* blocks per group * fs_frag */ 205 /* this data must be re-computed after crashes */ 206 struct csum fs_cstotal; /* cylinder summary information */ 207 /* these fields are cleared at mount time */ 208 char fs_fmod; /* super block modified flag */ 209 char fs_clean; /* file system is clean flag */ 210 char fs_ronly; /* mounted read-only flag */ 211 char fs_flags; /* currently unused flag */ 212 char fs_fsmnt[MAXMNTLEN]; /* name mounted on */ 213 /* these fields retain the current block allocation info */ 214 long fs_cgrotor; /* last cg searched */ 215 struct csum *fs_csp[MAXCSBUFS];/* list of fs_cs info buffers */ 216 long fs_cpc; /* cyl per cycle in postbl */ 217 short fs_postbl[MAXCPG][NRPOS];/* head of blocks for each rotation */ 218 long fs_magic; /* magic number */ 219 u_char fs_rotbl[1]; /* list of blocks for each rotation */ 220 /* actually longer */ 221 }; 222 223 /* 224 * Convert cylinder group to base address of its global summary info. 225 * 226 * N.B. This macro assumes that sizeof(struct csum) is a power of two. 227 */ 228 #define fs_cs(fs, indx) \ 229 fs_csp[(indx) / ((fs)->fs_bsize / sizeof(struct csum))] \ 230 [(indx) % ((fs)->fs_bsize / sizeof(struct csum))] 231 232 /* 233 * MAXBPC bounds the size of the rotational layout tables and 234 * is limited by the fact that the super block is of size SBSIZE. 235 * The size of these tables is INVERSELY proportional to the block 236 * size of the file system. It is aggravated by sector sizes that 237 * are not powers of two, as this increases the number of cylinders 238 * included before the rotational pattern repeats (fs_cpc). 239 * Its size is derived from the number of bytes remaining in (struct fs) 240 */ 241 #define MAXBPC (SBSIZE - sizeof (struct fs)) 242 243 /* 244 * Cylinder group block for a file system. 245 */ 246 #define CG_MAGIC 0x090255 247 struct cg { 248 struct cg *cg_link; /* linked list of cyl groups */ 249 struct cg *cg_rlink; /* used for incore cyl groups */ 250 time_t cg_time; /* time last written */ 251 long cg_cgx; /* we are the cgx'th cylinder group */ 252 short cg_ncyl; /* number of cyl's this cg */ 253 short cg_niblk; /* number of inode blocks this cg */ 254 long cg_ndblk; /* number of data blocks this cg */ 255 struct csum cg_cs; /* cylinder summary information */ 256 long cg_rotor; /* position of last used block */ 257 long cg_frotor; /* position of last used frag */ 258 long cg_irotor; /* position of last used inode */ 259 long cg_frsum[MAXFRAG]; /* counts of available frags */ 260 long cg_btot[MAXCPG]; /* block totals per cylinder */ 261 short cg_b[MAXCPG][NRPOS]; /* positions of free blocks */ 262 char cg_iused[MAXIPG/NBBY]; /* used inode map */ 263 long cg_magic; /* magic number */ 264 char cg_free[1]; /* free block map */ 265 /* actually longer */ 266 }; 267 268 /* 269 * MAXBPG bounds the number of blocks of data per cylinder group, 270 * and is limited by the fact that cylinder groups are at most one block. 271 * Its size is derived from the size of blocks and the (struct cg) size, 272 * by the number of remaining bits. 273 */ 274 #define MAXBPG(fs) \ 275 (NBBY * ((fs)->fs_bsize - (sizeof (struct cg))) / (fs)->fs_frag) 276 277 /* 278 * Turn file system block numbers into disk block addresses. 279 * This maps file system blocks to device size blocks. 280 */ 281 #define fsbtodb(fs, b) ((b) * ((fs)->fs_fsize / DEV_BSIZE)) 282 #define dbtofsb(fs, b) ((b) / ((fs)->fs_fsize / DEV_BSIZE)) 283 284 /* 285 * Cylinder group macros to locate things in cylinder groups. 286 * They calc file system addresses of cylinder group data structures. 287 */ 288 #define cgbase(fs, c) ((daddr_t)((fs)->fs_fpg * (c))) 289 #define cgstart(fs, c) \ 290 (cgbase(fs, c) + (fs)->fs_cgoffset * ((c) & ~((fs)->fs_cgmask))) 291 #define cgsblock(fs, c) (cgstart(fs, c) + (fs)->fs_sblkno) /* super blk */ 292 #define cgtod(fs, c) (cgstart(fs, c) + (fs)->fs_cblkno) /* cg block */ 293 #define cgimin(fs, c) (cgstart(fs, c) + (fs)->fs_iblkno) /* inode blk */ 294 #define cgdmin(fs, c) (cgstart(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