Commit | Line | Data |
---|---|---|
1ef63481 KM |
1 | /* Copyright (c) 1981 Regents of the University of California */ |
2 | ||
6994bf5d | 3 | /* fs.h 1.13 %G% */ |
1ef63481 KM |
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 | * | |
aca50d72 KM |
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 | |
6994bf5d | 25 | * the ``cgbase(fs, cg)'' macro. |
b6407c9d | 26 | * |
aca50d72 | 27 | * The first boot and super blocks are given in absolute disk addresses. |
b6407c9d | 28 | */ |
80cc8328 KM |
29 | #define BBSIZE 1024 |
30 | #define SBSIZE 8192 | |
31 | #define BBLOCK ((daddr_t)(0)) | |
32 | #define SBLOCK ((daddr_t)(BBLOCK + BBSIZE / DEV_BSIZE)) | |
1ef63481 KM |
33 | |
34 | /* | |
b6407c9d KM |
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. | |
1ef63481 | 40 | * |
b6407c9d KM |
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. | |
1ef63481 KM |
48 | * |
49 | * The file system records space availability at the fragment level; | |
50 | * to determine block availability, aligned fragments are examined. | |
743f1ef7 | 51 | * |
80cc8328 KM |
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 | * | |
743f1ef7 KM |
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. | |
80cc8328 KM |
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. | |
743f1ef7 | 102 | */ |
80cc8328 KM |
103 | #define NRPOS 8 /* number distinct rotational positions */ |
104 | #define ROTDELAY 2 | |
1ef63481 KM |
105 | |
106 | /* | |
107 | * Each file system has a number of inodes statically allocated. | |
80cc8328 | 108 | * We allocate one inode slot per NBPI bytes, expecting this |
5ae9a796 | 109 | * to be far more than we will ever need. |
80cc8328 KM |
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). | |
1ef63481 | 116 | */ |
80cc8328 KM |
117 | #define NBPI 2048 |
118 | #define MAXIPG 2048 /* max number inodes/cyl group */ | |
1ef63481 | 119 | |
b6407c9d KM |
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 */ | |
80cc8328 | 135 | |
1ef63481 | 136 | /* |
80cc8328 KM |
137 | * Per cylinder group information; summarized in blocks allocated |
138 | * from first cylinder group data blocks. These blocks have to be | |
139 | * read in from fs_csaddr (size fs_cssize) in addition to the | |
140 | * super block. | |
1ef63481 | 141 | * |
80cc8328 KM |
142 | * N.B. sizeof(struct csum) must be a power of two in order for |
143 | * the ``fs_cs'' macro to work (see below). | |
1ef63481 | 144 | */ |
80cc8328 KM |
145 | struct csum { |
146 | long cs_ndir; /* number of directories */ | |
147 | long cs_nbfree; /* number of free blocks */ | |
148 | long cs_nifree; /* number of free inodes */ | |
149 | long cs_nffree; /* number of free frags */ | |
150 | }; | |
1ef63481 | 151 | |
80cc8328 KM |
152 | /* |
153 | * Super block for a file system. | |
154 | */ | |
1ef63481 KM |
155 | #define FS_MAGIC 0x110854 |
156 | struct fs | |
157 | { | |
158 | long fs_magic; /* magic number */ | |
aca50d72 KM |
159 | daddr_t fs_bblkno; /* abs addr of boot-block in filesys */ |
160 | daddr_t fs_sblkno; /* abs addr of super-block in filesys */ | |
161 | daddr_t fs_cblkno; /* offset of cyl-block in filesys */ | |
162 | daddr_t fs_iblkno; /* offset of inode-blocks in filesys */ | |
163 | daddr_t fs_dblkno; /* offset of data-blocks in filesys */ | |
1ef63481 | 164 | time_t fs_time; /* last time written */ |
003319d1 KM |
165 | long fs_size; /* number of blocks in fs */ |
166 | long fs_dsize; /* number of data blocks in fs */ | |
167 | long fs_ncg; /* number of cylinder groups */ | |
b6407c9d KM |
168 | long fs_bsize; /* size of basic blocks in fs */ |
169 | long fs_fsize; /* size of frag blocks in fs */ | |
aca50d72 | 170 | short fs_frag; /* number of frags in a block in fs */ |
b6407c9d KM |
171 | short fs_minfree; /* minimum percentage of free blocks */ |
172 | short fs_rotdelay; /* num of ms for optimal next block */ | |
aca50d72 | 173 | short fs_rps; /* disk revolutions per second */ |
1ef63481 | 174 | /* sizes determined by number of cylinder groups and their sizes */ |
b6407c9d | 175 | daddr_t fs_csaddr; /* blk addr of cyl grp summary area */ |
003319d1 KM |
176 | long fs_cssize; /* size of cyl grp summary area */ |
177 | long fs_cgsize; /* cylinder group size */ | |
1ef63481 KM |
178 | /* these fields should be derived from the hardware */ |
179 | short fs_ntrak; /* tracks per cylinder */ | |
180 | short fs_nsect; /* sectors per track */ | |
003319d1 | 181 | long fs_spc; /* sectors per cylinder */ |
1ef63481 | 182 | /* this comes from the disk driver partitioning */ |
003319d1 | 183 | long fs_ncyl; /* cylinders in file system */ |
1ef63481 KM |
184 | /* these fields can be computed from the others */ |
185 | short fs_cpg; /* cylinders per group */ | |
1ef63481 | 186 | short fs_ipg; /* inodes per group */ |
b6407c9d | 187 | long fs_fpg; /* blocks per group * fs_frag */ |
003319d1 KM |
188 | /* this data must be re-computed after crashes */ |
189 | struct csum fs_cstotal; /* cylinder summary information */ | |
1ef63481 KM |
190 | /* these fields are cleared at mount time */ |
191 | char fs_fmod; /* super block modified flag */ | |
192 | char fs_ronly; /* mounted read-only flag */ | |
003319d1 | 193 | char fs_fsmnt[34]; /* name mounted on */ |
743f1ef7 | 194 | /* these fields retain the current block allocation info */ |
003319d1 | 195 | long fs_cgrotor; /* last cg searched */ |
743f1ef7 | 196 | struct csum *fs_csp[NBUF]; /* list of fs_cs info buffers */ |
aca50d72 KM |
197 | short fs_cpc; /* cyl per cycle in postbl */ |
198 | short fs_postbl[MAXCPG][NRPOS];/* head of blocks for each rotation */ | |
199 | u_char fs_rotbl[1]; /* list of blocks for each rotation */ | |
743f1ef7 | 200 | /* actually longer */ |
1ef63481 | 201 | }; |
b6407c9d KM |
202 | |
203 | /* | |
204 | * convert cylinder group to base address of its global summary info. | |
80cc8328 | 205 | * |
b6407c9d KM |
206 | * N.B. This macro assumes that sizeof(struct csum) is a power of two. |
207 | */ | |
208 | #define fs_cs(fs, indx) \ | |
209 | fs_csp[(indx) / ((fs)->fs_bsize / sizeof(struct csum))] \ | |
210 | [(indx) % ((fs)->fs_bsize / sizeof(struct csum))] | |
1ef63481 KM |
211 | |
212 | /* | |
aca50d72 KM |
213 | * MAXBPC bounds the size of the rotational layout tables and |
214 | * is limited by the fact that the super block is of size SBSIZE. | |
215 | * The size of these tables is INVERSELY proportional to the block | |
216 | * size of the file system. It is aggravated by sector sizes that | |
217 | * are not powers of two, as this increases the number of cylinders | |
218 | * included before the rotational pattern repeats (fs_cpc). | |
219 | * Its size is derived from the number of bytes remaining in (struct fs) | |
1ef63481 | 220 | */ |
aca50d72 | 221 | #define MAXBPC (SBSIZE - sizeof (struct fs)) |
1ef63481 KM |
222 | |
223 | /* | |
80cc8328 | 224 | * Cylinder group block for a file system. |
1ef63481 | 225 | */ |
1ef63481 KM |
226 | #define CG_MAGIC 0x092752 |
227 | struct cg { | |
228 | long cg_magic; /* magic number */ | |
229 | time_t cg_time; /* time last written */ | |
003319d1 | 230 | long cg_cgx; /* we are the cgx'th cylinder group */ |
1ef63481 KM |
231 | short cg_ncyl; /* number of cyl's this cg */ |
232 | short cg_niblk; /* number of inode blocks this cg */ | |
003319d1 KM |
233 | long cg_ndblk; /* number of data blocks this cg */ |
234 | struct csum cg_cs; /* cylinder summary information */ | |
235 | long cg_rotor; /* position of last used block */ | |
236 | long cg_frotor; /* position of last used frag */ | |
237 | long cg_irotor; /* position of last used inode */ | |
b6407c9d | 238 | long cg_frsum[MAXFRAG]; /* counts of available frags */ |
d37938c8 | 239 | long cg_btot[MAXCPG]; /* block totals per cylinder */ |
1ef63481 KM |
240 | short cg_b[MAXCPG][NRPOS]; /* positions of free blocks */ |
241 | char cg_iused[MAXIPG/NBBY]; /* used inode map */ | |
242 | char cg_free[1]; /* free block map */ | |
243 | /* actually longer */ | |
244 | }; | |
80cc8328 KM |
245 | |
246 | /* | |
247 | * MAXBPG bounds the number of blocks of data per cylinder group, | |
248 | * and is limited by the fact that cylinder groups are at most one block. | |
249 | * Its size is derived from the size of blocks and the (struct cg) size, | |
250 | * by the number of remaining bits. | |
251 | */ | |
252 | #define MAXBPG(fs) \ | |
253 | (NBBY * ((fs)->fs_bsize - (sizeof (struct cg))) / (fs)->fs_frag) | |
254 | ||
255 | /* | |
256 | * Turn file system block numbers into disk block addresses. | |
257 | * This maps file system blocks to device size blocks. | |
258 | */ | |
259 | #define fsbtodb(fs, b) ((b) * ((fs)->fs_fsize / DEV_BSIZE)) | |
260 | #define dbtofsb(fs, b) ((b) / ((fs)->fs_fsize / DEV_BSIZE)) | |
261 | ||
262 | /* | |
263 | * Cylinder group macros to locate things in cylinder groups. | |
264 | * | |
aca50d72 KM |
265 | * cylinder group to disk block address of spare boot block |
266 | * and super block | |
267 | * Note that these are in absolute addresses, and can NOT | |
268 | * in general be expressable in terms of file system addresses. | |
80cc8328 | 269 | */ |
6994bf5d KM |
270 | #define cgbblock(fs, c) (fsbtodb(fs, cgbase(fs, c)) + (fs)->fs_bblkno) |
271 | #define cgsblock(fs, c) (fsbtodb(fs, cgbase(fs, c)) + (fs)->fs_sblkno) | |
80cc8328 KM |
272 | |
273 | /* | |
d37938c8 | 274 | * file system addresses of cylinder group data structures |
80cc8328 | 275 | */ |
6994bf5d KM |
276 | #define cgbase(fs, c) ((daddr_t)((fs)->fs_fpg * (c))) /* base addr */ |
277 | #define cgtod(fs, c) (cgbase(fs, c) + (fs)->fs_cblkno) /* cg block */ | |
278 | #define cgimin(fs, c) (cgbase(fs, c) + (fs)->fs_iblkno) /* inode blk */ | |
279 | #define cgdmin(fs, c) (cgbase(fs, c) + (fs)->fs_dblkno) /* 1st data */ | |
80cc8328 KM |
280 | |
281 | /* | |
d37938c8 KM |
282 | * macros for handling inode numbers |
283 | * inode number to file system block offset | |
284 | * inode number to cylinder group number | |
285 | * inode number to file system block address | |
80cc8328 | 286 | */ |
6994bf5d KM |
287 | #define itoo(fs, x) ((x) % INOPB(fs)) |
288 | #define itog(fs, x) ((x) / (fs)->fs_ipg) | |
289 | #define itod(fs, x) \ | |
290 | ((daddr_t)(cgimin(fs, itog(fs, x)) + \ | |
aca50d72 | 291 | (x) % (fs)->fs_ipg / INOPB(fs) * (fs)->fs_frag)) |
80cc8328 | 292 | |
80cc8328 KM |
293 | /* |
294 | * give cylinder group number for a file system block | |
80cc8328 KM |
295 | * give cylinder group block number for a file system block |
296 | */ | |
6994bf5d KM |
297 | #define dtog(fs, d) ((d) / (fs)->fs_fpg) |
298 | #define dtogd(fs, d) ((d) % (fs)->fs_fpg) | |
80cc8328 | 299 | |
aca50d72 KM |
300 | /* |
301 | * compute the cylinder and rotational position of a cyl block addr | |
302 | */ | |
303 | #define cbtocylno(fs, bno) \ | |
304 | ((bno) * NSPF(fs) / (fs)->fs_spc) | |
305 | #define cbtorpos(fs, bno) \ | |
306 | ((bno) * NSPF(fs) % (fs)->fs_nsect * NRPOS / (fs)->fs_nsect) | |
307 | ||
80cc8328 KM |
308 | /* |
309 | * determining the size of a file block in the file system | |
310 | */ | |
311 | #define blksize(fs, ip, lbn) \ | |
312 | (((lbn) >= NDADDR || (ip)->i_size >= ((lbn) + 1) * (fs)->fs_bsize) \ | |
313 | ? (fs)->fs_bsize \ | |
314 | : (roundup((ip)->i_size % (fs)->fs_bsize, (fs)->fs_fsize))) | |
315 | #define dblksize(fs, dip, lbn) \ | |
316 | (((lbn) >= NDADDR || (dip)->di_size >= ((lbn) + 1) * (fs)->fs_bsize) \ | |
317 | ? (fs)->fs_bsize \ | |
318 | : (roundup((dip)->di_size % (fs)->fs_bsize, (fs)->fs_fsize))) | |
319 | ||
320 | /* | |
321 | * number of disk sectors per block; assumes DEV_BSIZE byte sector size | |
322 | */ | |
323 | #define NSPB(fs) ((fs)->fs_bsize / DEV_BSIZE) | |
324 | #define NSPF(fs) ((fs)->fs_fsize / DEV_BSIZE) | |
325 | ||
326 | /* | |
327 | * INOPB is the number of inodes in a secondary storage block | |
328 | */ | |
329 | #define INOPB(fs) ((fs)->fs_bsize / sizeof (struct dinode)) | |
330 | #define INOPF(fs) ((fs)->fs_fsize / sizeof (struct dinode)) | |
331 | ||
332 | /* | |
333 | * NINDIR is the number of indirects in a file system block | |
334 | */ | |
335 | #define NINDIR(fs) ((fs)->fs_bsize / sizeof (daddr_t)) | |
1ef63481 KM |
336 | |
337 | #ifdef KERNEL | |
338 | struct fs *getfs(); | |
339 | #endif |