1 | #include "filecache.hfa"
|
---|
2 |
|
---|
3 | #include <assert.h>
|
---|
4 | #include <string.h>
|
---|
5 |
|
---|
6 | #include <fstream.hfa>
|
---|
7 | #include <stdlib.hfa>
|
---|
8 |
|
---|
9 | #include <errno.h>
|
---|
10 | #include <unistd.h>
|
---|
11 | extern "C" {
|
---|
12 | #include <fcntl.h>
|
---|
13 | #include <ftw.h>
|
---|
14 | }
|
---|
15 |
|
---|
16 | #include "options.hfa"
|
---|
17 |
|
---|
18 | static inline uint32_t murmur_32_scramble(uint32_t k) {
|
---|
19 | k *= 0xcc9e2d51;
|
---|
20 | k = (k << 15) | (k >> 17);
|
---|
21 | k *= 0x1b873593;
|
---|
22 | return k;
|
---|
23 | }
|
---|
24 |
|
---|
25 | uint32_t murmur3_32(const uint8_t* key, size_t len, uint32_t seed)
|
---|
26 | {
|
---|
27 | uint32_t h = seed;
|
---|
28 | uint32_t k;
|
---|
29 | /* Read in groups of 4. */
|
---|
30 | for (size_t i = len >> 2; i; i--) {
|
---|
31 | // Here is a source of differing results across endiannesses.
|
---|
32 | // A swap here has no effects on hash properties though.
|
---|
33 | memcpy(&k, key, sizeof(uint32_t));
|
---|
34 | key += sizeof(uint32_t);
|
---|
35 | h ^= murmur_32_scramble(k);
|
---|
36 | h = (h << 13) | (h >> 19);
|
---|
37 | h = h * 5 + 0xe6546b64;
|
---|
38 | }
|
---|
39 | /* Read the rest. */
|
---|
40 | k = 0;
|
---|
41 | for (size_t i = len & 3; i; i--) {
|
---|
42 | k <<= 8;
|
---|
43 | k |= key[i - 1];
|
---|
44 | }
|
---|
45 | // A swap is *not* necessary here because the preceding loop already
|
---|
46 | // places the low bytes in the low places according to whatever endianness
|
---|
47 | // we use. Swaps only apply when the memory is copied in a chunk.
|
---|
48 | h ^= murmur_32_scramble(k);
|
---|
49 | /* Finalize. */
|
---|
50 | h ^= len;
|
---|
51 | h ^= h >> 16;
|
---|
52 | h *= 0x85ebca6b;
|
---|
53 | h ^= h >> 13;
|
---|
54 | h *= 0xc2b2ae35;
|
---|
55 | h ^= h >> 16;
|
---|
56 | return h;
|
---|
57 | }
|
---|
58 |
|
---|
59 | static inline [unsigned size, char unit] human_size( size_t size ) {
|
---|
60 | int idx = 0;
|
---|
61 | static char units [] = { ' ', 'K', 'M', 'G', 'T' };
|
---|
62 | while( size >= 1024 ) {
|
---|
63 | idx++;
|
---|
64 | size /= 1024;
|
---|
65 | if(idx >= 5) {
|
---|
66 | abort("File too large to print\n");
|
---|
67 | }
|
---|
68 | }
|
---|
69 |
|
---|
70 | return [size, units[idx]];
|
---|
71 | }
|
---|
72 |
|
---|
73 | struct {
|
---|
74 | cache_line * entries;
|
---|
75 | size_t size;
|
---|
76 | int * rawfds;
|
---|
77 | int nfds;
|
---|
78 | } file_cache;
|
---|
79 |
|
---|
80 | void ?{}( cache_line & this ) with( this ) {
|
---|
81 | file = 0p;
|
---|
82 | size = 0;
|
---|
83 | fd = 0;
|
---|
84 | }
|
---|
85 |
|
---|
86 | [int fd, size_t size] get_file( * const char file, size_t len ) {
|
---|
87 | uint32_t idx = murmur3_32( (const uint8_t *)file, len, options.file_cache.hash_seed ) % file_cache.size;
|
---|
88 |
|
---|
89 | for(int i = 0;; i++) {
|
---|
90 | assert( i < file_cache.size );
|
---|
91 | cache_line & entry = file_cache.entries[idx];
|
---|
92 | if( !entry.file ) return [-1, 0];
|
---|
93 | #if !defined(REJECT_CONFLICTS)
|
---|
94 | if( strncmp(entry.file, file, len) != 0 ) {
|
---|
95 | idx = (idx + 1) % file_cache.size;
|
---|
96 | continue;
|
---|
97 | }
|
---|
98 | #endif
|
---|
99 | return [entry.fd, entry.size];
|
---|
100 | }
|
---|
101 | }
|
---|
102 |
|
---|
103 | int put_file( cache_line & entry, int fd ) {
|
---|
104 | uint32_t idx = murmur3_32( (const uint8_t *)entry.file, strlen(entry.file), options.file_cache.hash_seed ) % file_cache.size;
|
---|
105 |
|
---|
106 | int i = 0;
|
---|
107 | for(;file_cache.entries[idx].file; i++ ) {
|
---|
108 | assert( i < file_cache.size );
|
---|
109 | idx = (idx + 1) % file_cache.size;
|
---|
110 | }
|
---|
111 |
|
---|
112 | file_cache.entries[idx] = entry;
|
---|
113 | file_cache.entries[idx].fd = fd;
|
---|
114 | return i > 0 ? 1 : 0;
|
---|
115 | }
|
---|
116 |
|
---|
117 | // int ftw(const char *dirpath, int (*fn) (const char *fpath, const struct stat *sb, int typeflag), int nopenfd)
|
---|
118 | void fill_cache( const char * path ) {
|
---|
119 | int ret;
|
---|
120 | ret = chdir(path);
|
---|
121 | if(ret < 0) {
|
---|
122 | abort( "chdir error: (%d) %s\n", (int)errno, strerror(errno) );
|
---|
123 | }
|
---|
124 |
|
---|
125 | size_t fcount = 0;
|
---|
126 | size_t fsize = 16;
|
---|
127 | cache_line * raw = alloc(fsize);
|
---|
128 | // Step 1 get a dense array of all files
|
---|
129 | int walk(const char *fpath, const struct stat *sb, int typeflag) {
|
---|
130 | if(typeflag != FTW_F) return 0;
|
---|
131 |
|
---|
132 | int idx = fcount;
|
---|
133 | fcount++;
|
---|
134 | if(fcount > fsize) {
|
---|
135 | fsize *= 2;
|
---|
136 | raw = alloc(fsize, raw`realloc);
|
---|
137 | }
|
---|
138 |
|
---|
139 | raw[idx].file = strdup(fpath+2);
|
---|
140 | raw[idx].size = sb->st_size;
|
---|
141 | if( !options.file_cache.list ) {
|
---|
142 | raw[idx].fd = open( fpath, options.file_cache.open_flags );
|
---|
143 | if(raw[idx].fd < 0) {
|
---|
144 | abort( "open file error: (%d) %s\n", (int)errno, strerror(errno) );
|
---|
145 | }
|
---|
146 | }
|
---|
147 | return 0;
|
---|
148 | }
|
---|
149 |
|
---|
150 | ret = ftw(".", walk, 10);
|
---|
151 | if(ret < 0) {
|
---|
152 | abort( "ftw error: (%d) %s\n", (int)errno, strerror(errno) );
|
---|
153 | }
|
---|
154 |
|
---|
155 | if(fcount == 0) {
|
---|
156 | abort("No file found in path %s\n", path);
|
---|
157 | }
|
---|
158 |
|
---|
159 | // Step 2 create the cache
|
---|
160 | file_cache.size = options.file_cache.size > 0 ? options.file_cache.size : fsize;
|
---|
161 | if( file_cache.size < fcount ) {
|
---|
162 | abort("File Cache too small\n");
|
---|
163 | }
|
---|
164 |
|
---|
165 | file_cache.entries = anew(file_cache.size);
|
---|
166 |
|
---|
167 | if(options.file_cache.fixed_fds) {
|
---|
168 | file_cache.nfds = fcount;
|
---|
169 | file_cache.rawfds = alloc(fcount);
|
---|
170 | }
|
---|
171 |
|
---|
172 | // Step 3 fill the cache
|
---|
173 | int conflicts = 0;
|
---|
174 | for(i; fcount) {
|
---|
175 | int fd;
|
---|
176 | if(options.file_cache.fixed_fds) {
|
---|
177 | file_cache.rawfds[i] = raw[i].fd;
|
---|
178 | fd = i;
|
---|
179 | }
|
---|
180 | else {
|
---|
181 | fd = raw[i].fd;
|
---|
182 | }
|
---|
183 | conflicts += put_file( raw[i], fd );
|
---|
184 | }
|
---|
185 | sout | "Filled cache from path \"" | path | "\" with" | fcount | "files";
|
---|
186 | if( conflicts > 0 ) {
|
---|
187 | sout | "Found" | conflicts | "conflicts (seed: " | options.file_cache.hash_seed | ")";
|
---|
188 | #if defined(REJECT_CONFLICTS)
|
---|
189 | abort("Conflicts found in the cache");
|
---|
190 | #endif
|
---|
191 | }
|
---|
192 |
|
---|
193 | if(options.file_cache.list) {
|
---|
194 | sout | "Listing files and exiting";
|
---|
195 | for(i; fcount) {
|
---|
196 | int s; char u;
|
---|
197 | [s, u] = human_size(raw[i].size);
|
---|
198 | sout | s | u | "-" | raw[i].file;
|
---|
199 | free(raw[i].file);
|
---|
200 | }
|
---|
201 | free(raw);
|
---|
202 | adelete( file_cache.entries );
|
---|
203 | exit(0);
|
---|
204 | }
|
---|
205 |
|
---|
206 | // Step 4 clean up
|
---|
207 | free( raw );
|
---|
208 | }
|
---|
209 |
|
---|
210 | [int *, int] filefds(int extra) {
|
---|
211 | if(!options.file_cache.path) {
|
---|
212 | int * data = alloc(extra);
|
---|
213 | return [data, 0];
|
---|
214 | }
|
---|
215 |
|
---|
216 | if(!file_cache.entries) {
|
---|
217 | abort("File cache not filled!\n");
|
---|
218 | }
|
---|
219 |
|
---|
220 | size_t s = file_cache.nfds + extra;
|
---|
221 | int * data = alloc(s, file_cache.rawfds`realloc);
|
---|
222 | return [data, file_cache.nfds];
|
---|
223 | }
|
---|
224 |
|
---|
225 |
|
---|
226 | void close_cache() {
|
---|
227 | for(idx; file_cache.size) {
|
---|
228 | cache_line & entry = file_cache.entries[idx];
|
---|
229 | if( !entry.file ) continue;
|
---|
230 |
|
---|
231 | int ret = close( entry.fd );
|
---|
232 | if(ret < 0) {
|
---|
233 | abort( "close file error: (%d) %s\n", (int)errno, strerror(errno) );
|
---|
234 | }
|
---|
235 | }
|
---|
236 |
|
---|
237 | delete( file_cache.entries );
|
---|
238 | }
|
---|