| 1 | /* |
| 2 | * Main program for agedu. |
| 3 | */ |
| 4 | |
| 5 | #include "agedu.h" |
| 6 | |
| 7 | #include "du.h" |
| 8 | #include "trie.h" |
| 9 | #include "index.h" |
| 10 | #include "alloc.h" |
| 11 | #include "html.h" |
| 12 | #include "httpd.h" |
| 13 | #include "fgetline.h" |
| 14 | |
| 15 | /* |
| 16 | * Path separator. This global variable affects the behaviour of |
| 17 | * various parts of the code when they need to deal with path |
| 18 | * separators. The path separator appropriate to a particular data |
| 19 | * set is encoded in the index file storing that data set; data |
| 20 | * sets generated on Unix will of course have the default '/', but |
| 21 | * foreign data sets are conceivable and must be handled correctly. |
| 22 | */ |
| 23 | char pathsep = '/'; |
| 24 | |
| 25 | void fatal(const char *fmt, ...) |
| 26 | { |
| 27 | va_list ap; |
| 28 | fprintf(stderr, "%s: ", PNAME); |
| 29 | va_start(ap, fmt); |
| 30 | vfprintf(stderr, fmt, ap); |
| 31 | va_end(ap); |
| 32 | fprintf(stderr, "\n"); |
| 33 | exit(1); |
| 34 | } |
| 35 | |
| 36 | struct inclusion_exclusion { |
| 37 | int type; |
| 38 | const char *wildcard; |
| 39 | int path; |
| 40 | }; |
| 41 | |
| 42 | struct ctx { |
| 43 | triebuild *tb; |
| 44 | dev_t datafile_dev, filesystem_dev; |
| 45 | ino_t datafile_ino; |
| 46 | time_t last_output_update; |
| 47 | int progress, progwidth; |
| 48 | int straight_to_dump; |
| 49 | struct inclusion_exclusion *inex; |
| 50 | int ninex; |
| 51 | int crossfs; |
| 52 | int usemtime; |
| 53 | int fakeatimes; |
| 54 | }; |
| 55 | |
| 56 | static void dump_line(const char *pathname, const struct trie_file *tf) |
| 57 | { |
| 58 | const char *p; |
| 59 | printf("%llu %llu ", tf->size, tf->atime); |
| 60 | for (p = pathname; *p; p++) { |
| 61 | if (*p >= ' ' && *p < 127 && *p != '%') |
| 62 | putchar(*p); |
| 63 | else |
| 64 | printf("%%%02x", (unsigned char)*p); |
| 65 | } |
| 66 | putchar('\n'); |
| 67 | } |
| 68 | |
| 69 | static int gotdata(void *vctx, const char *pathname, const STRUCT_STAT *st) |
| 70 | { |
| 71 | struct ctx *ctx = (struct ctx *)vctx; |
| 72 | struct trie_file file; |
| 73 | time_t t; |
| 74 | int i, include; |
| 75 | const char *filename; |
| 76 | |
| 77 | /* |
| 78 | * Filter out our own data file. |
| 79 | */ |
| 80 | if (st->st_dev == ctx->datafile_dev && st->st_ino == ctx->datafile_ino) |
| 81 | return 0; |
| 82 | |
| 83 | /* |
| 84 | * Don't cross the streams^W^Wany file system boundary. |
| 85 | */ |
| 86 | if (!ctx->crossfs && st->st_dev != ctx->filesystem_dev) |
| 87 | return 0; |
| 88 | |
| 89 | file.size = (unsigned long long)512 * st->st_blocks; |
| 90 | if (ctx->usemtime || (ctx->fakeatimes && S_ISDIR(st->st_mode))) |
| 91 | file.atime = st->st_mtime; |
| 92 | else |
| 93 | file.atime = max(st->st_mtime, st->st_atime); |
| 94 | |
| 95 | /* |
| 96 | * Filter based on wildcards. |
| 97 | */ |
| 98 | include = 1; |
| 99 | filename = strrchr(pathname, pathsep); |
| 100 | if (!filename) |
| 101 | filename = pathname; |
| 102 | else |
| 103 | filename++; |
| 104 | for (i = 0; i < ctx->ninex; i++) { |
| 105 | if (fnmatch(ctx->inex[i].wildcard, |
| 106 | ctx->inex[i].path ? pathname : filename, 0) == 0) |
| 107 | include = ctx->inex[i].type; |
| 108 | } |
| 109 | if (include == -1) |
| 110 | return 0; /* ignore this entry and any subdirs */ |
| 111 | if (include == 0) { |
| 112 | /* |
| 113 | * Here we are supposed to be filtering an entry out, but |
| 114 | * still recursing into it if it's a directory. However, |
| 115 | * we can't actually leave out any directory whose |
| 116 | * subdirectories we then look at. So we cheat, in that |
| 117 | * case, by setting the size to zero. |
| 118 | */ |
| 119 | if (!S_ISDIR(st->st_mode)) |
| 120 | return 0; /* just ignore */ |
| 121 | else |
| 122 | file.size = 0; |
| 123 | } |
| 124 | |
| 125 | if (ctx->straight_to_dump) |
| 126 | dump_line(pathname, &file); |
| 127 | else |
| 128 | triebuild_add(ctx->tb, pathname, &file); |
| 129 | |
| 130 | if (ctx->progress) { |
| 131 | t = time(NULL); |
| 132 | if (t != ctx->last_output_update) { |
| 133 | fprintf(stderr, "%-*.*s\r", ctx->progwidth, ctx->progwidth, |
| 134 | pathname); |
| 135 | fflush(stderr); |
| 136 | ctx->last_output_update = t; |
| 137 | } |
| 138 | } |
| 139 | |
| 140 | return 1; |
| 141 | } |
| 142 | |
| 143 | static void scan_error(void *vctx, const char *fmt, ...) |
| 144 | { |
| 145 | struct ctx *ctx = (struct ctx *)vctx; |
| 146 | va_list ap; |
| 147 | |
| 148 | if (ctx->progress) { |
| 149 | fprintf(stderr, "%-*s\r", ctx->progwidth, ""); |
| 150 | fflush(stderr); |
| 151 | } |
| 152 | |
| 153 | fprintf(stderr, "%s: ", PNAME); |
| 154 | va_start(ap, fmt); |
| 155 | vfprintf(stderr, fmt, ap); |
| 156 | va_end(ap); |
| 157 | |
| 158 | ctx->last_output_update--; /* force a progress report next time */ |
| 159 | } |
| 160 | |
| 161 | static void text_query(const void *mappedfile, const char *querydir, |
| 162 | time_t t, int depth) |
| 163 | { |
| 164 | size_t maxpathlen; |
| 165 | char *pathbuf; |
| 166 | unsigned long xi1, xi2; |
| 167 | unsigned long long s1, s2; |
| 168 | |
| 169 | maxpathlen = trie_maxpathlen(mappedfile); |
| 170 | pathbuf = snewn(maxpathlen + 1, char); |
| 171 | |
| 172 | /* |
| 173 | * We want to query everything between the supplied filename |
| 174 | * (inclusive) and that filename with a ^A on the end |
| 175 | * (exclusive). So find the x indices for each. |
| 176 | */ |
| 177 | strcpy(pathbuf, querydir); |
| 178 | make_successor(pathbuf); |
| 179 | xi1 = trie_before(mappedfile, querydir); |
| 180 | xi2 = trie_before(mappedfile, pathbuf); |
| 181 | |
| 182 | if (xi2 - xi1 == 1) |
| 183 | return; /* file, or empty dir => no display */ |
| 184 | |
| 185 | /* |
| 186 | * Now do the lookups in the age index. |
| 187 | */ |
| 188 | s1 = index_query(mappedfile, xi1, t); |
| 189 | s2 = index_query(mappedfile, xi2, t); |
| 190 | |
| 191 | if (s1 == s2) |
| 192 | return; /* no space taken up => no display */ |
| 193 | |
| 194 | if (depth > 0) { |
| 195 | /* |
| 196 | * Now scan for first-level subdirectories and report |
| 197 | * those too. |
| 198 | */ |
| 199 | xi1++; |
| 200 | while (xi1 < xi2) { |
| 201 | trie_getpath(mappedfile, xi1, pathbuf); |
| 202 | text_query(mappedfile, pathbuf, t, depth-1); |
| 203 | make_successor(pathbuf); |
| 204 | xi1 = trie_before(mappedfile, pathbuf); |
| 205 | } |
| 206 | } |
| 207 | |
| 208 | /* Display in units of 1Kb */ |
| 209 | printf("%-11llu %s\n", (s2 - s1) / 1024, querydir); |
| 210 | } |
| 211 | |
| 212 | /* |
| 213 | * Largely frivolous way to define all my command-line options. I |
| 214 | * present here a parametric macro which declares a series of |
| 215 | * _logical_ option identifiers, and for each one declares zero or |
| 216 | * more short option characters and zero or more long option |
| 217 | * words. Then I repeatedly invoke that macro with its arguments |
| 218 | * defined to be various other macros, which allows me to |
| 219 | * variously: |
| 220 | * |
| 221 | * - define an enum allocating a distinct integer value to each |
| 222 | * logical option id |
| 223 | * - define a string consisting of precisely all the short option |
| 224 | * characters |
| 225 | * - define a string array consisting of all the long option |
| 226 | * strings |
| 227 | * - define (with help from auxiliary enums) integer arrays |
| 228 | * parallel to both of the above giving the logical option id |
| 229 | * for each physical short and long option |
| 230 | * - define an array indexed by logical option id indicating |
| 231 | * whether the option in question takes a value |
| 232 | * - define a function which prints out brief online help for all |
| 233 | * the options. |
| 234 | * |
| 235 | * It's not at all clear to me that this trickery is actually |
| 236 | * particularly _efficient_ - it still, after all, requires going |
| 237 | * linearly through the option list at run time and doing a |
| 238 | * strcmp, whereas in an ideal world I'd have liked the lists of |
| 239 | * long and short options to be pre-sorted so that a binary search |
| 240 | * or some other more efficient lookup was possible. (Not that |
| 241 | * asymptotic algorithmic complexity is remotely vital in option |
| 242 | * parsing, but if I were doing this in, say, Lisp or something |
| 243 | * with an equivalently powerful preprocessor then once I'd had |
| 244 | * the idea of preparing the option-parsing data structures at |
| 245 | * compile time I would probably have made the effort to prepare |
| 246 | * them _properly_. I could have Perl generate me a source file |
| 247 | * from some sort of description, I suppose, but that would seem |
| 248 | * like overkill. And in any case, it's more of a challenge to |
| 249 | * achieve as much as possible by cunning use of cpp and enum than |
| 250 | * to just write some sensible and logical code in a Turing- |
| 251 | * complete language. I said it was largely frivolous :-) |
| 252 | * |
| 253 | * This approach does have the virtue that it brings together the |
| 254 | * option ids, option spellings and help text into a single |
| 255 | * combined list and defines them all in exactly one place. If I |
| 256 | * want to add a new option, or a new spelling for an option, I |
| 257 | * only have to modify the main OPTHELP macro below and then add |
| 258 | * code to process the new logical id. |
| 259 | * |
| 260 | * (Though, really, even that isn't ideal, since it still involves |
| 261 | * modifying the source file in more than one place. In a |
| 262 | * _properly_ ideal world, I'd be able to interleave the option |
| 263 | * definitions with the code fragments that process them. And then |
| 264 | * not bother defining logical identifiers for them at all - those |
| 265 | * would be automatically generated, since I wouldn't have any |
| 266 | * need to specify them manually in another part of the code.) |
| 267 | * |
| 268 | * One other helpful consequence of the enum-based structure here |
| 269 | * is that it causes a compiler error if I accidentally try to |
| 270 | * define the same option (short or long) twice. |
| 271 | */ |
| 272 | |
| 273 | #define OPTHELP(NOVAL, VAL, SHORT, LONG, HELPPFX, HELPARG, HELPLINE, HELPOPT) \ |
| 274 | HELPPFX("usage") HELPLINE(PNAME " [options] action [action...]") \ |
| 275 | HELPPFX("actions") \ |
| 276 | VAL(SCAN) SHORT(s) LONG(scan) \ |
| 277 | HELPARG("directory") HELPOPT("scan and index a directory") \ |
| 278 | NOVAL(HTTPD) SHORT(w) LONG(web) LONG(server) LONG(httpd) \ |
| 279 | HELPOPT("serve HTML reports from a temporary web server") \ |
| 280 | VAL(TEXT) SHORT(t) LONG(text) \ |
| 281 | HELPARG("subdir") HELPOPT("print a plain text report on a subdirectory") \ |
| 282 | NOVAL(REMOVE) SHORT(R) LONG(remove) LONG(delete) LONG(unlink) \ |
| 283 | HELPOPT("remove the index file") \ |
| 284 | NOVAL(DUMP) SHORT(D) LONG(dump) HELPOPT("dump the index file on stdout") \ |
| 285 | NOVAL(LOAD) SHORT(L) LONG(load) \ |
| 286 | HELPOPT("load and index a dump file") \ |
| 287 | VAL(SCANDUMP) SHORT(S) LONG(scan_dump) \ |
| 288 | HELPARG("directory") HELPOPT("scan only, generating a dump") \ |
| 289 | VAL(HTML) SHORT(H) LONG(html) \ |
| 290 | HELPARG("subdir") HELPOPT("print an HTML report on a subdirectory") \ |
| 291 | HELPPFX("options") \ |
| 292 | VAL(DATAFILE) SHORT(f) LONG(file) \ |
| 293 | HELPARG("filename") HELPOPT("[most modes] specify index file") \ |
| 294 | NOVAL(CROSSFS) LONG(cross_fs) \ |
| 295 | HELPOPT("[--scan] cross filesystem boundaries") \ |
| 296 | NOVAL(NOCROSSFS) LONG(no_cross_fs) \ |
| 297 | HELPOPT("[--scan] stick to one filesystem") \ |
| 298 | VAL(PRUNE) LONG(prune) \ |
| 299 | HELPARG("wildcard") HELPOPT("[--scan] prune files matching pattern") \ |
| 300 | VAL(PRUNEPATH) LONG(prune_path) \ |
| 301 | HELPARG("wildcard") HELPOPT("[--scan] prune pathnames matching pattern") \ |
| 302 | VAL(EXCLUDE) LONG(exclude) \ |
| 303 | HELPARG("wildcard") HELPOPT("[--scan] exclude files matching pattern") \ |
| 304 | VAL(EXCLUDEPATH) LONG(exclude_path) \ |
| 305 | HELPARG("wildcard") HELPOPT("[--scan] exclude pathnames matching pattern") \ |
| 306 | VAL(INCLUDE) LONG(include) \ |
| 307 | HELPARG("wildcard") HELPOPT("[--scan] include files matching pattern") \ |
| 308 | VAL(INCLUDEPATH) LONG(include_path) \ |
| 309 | HELPARG("wildcard") HELPOPT("[--scan] include pathnames matching pattern") \ |
| 310 | NOVAL(PROGRESS) LONG(progress) LONG(scan_progress) \ |
| 311 | HELPOPT("[--scan] report progress on stderr") \ |
| 312 | NOVAL(NOPROGRESS) LONG(no_progress) LONG(no_scan_progress) \ |
| 313 | HELPOPT("[--scan] do not report progress") \ |
| 314 | NOVAL(TTYPROGRESS) LONG(tty_progress) LONG(tty_scan_progress) \ |
| 315 | LONG(progress_tty) LONG(scan_progress_tty) \ |
| 316 | HELPOPT("[--scan] report progress if stderr is a tty") \ |
| 317 | NOVAL(DIRATIME) LONG(dir_atime) LONG(dir_atimes) \ |
| 318 | HELPOPT("[--scan,--load] keep real atimes on directories") \ |
| 319 | NOVAL(NODIRATIME) LONG(no_dir_atime) LONG(no_dir_atimes) \ |
| 320 | HELPOPT("[--scan,--load] fake atimes on directories") \ |
| 321 | NOVAL(MTIME) LONG(mtime) \ |
| 322 | HELPOPT("[--scan] use mtime instead of atime") \ |
| 323 | VAL(AGERANGE) SHORT(r) LONG(age_range) LONG(range) LONG(ages) \ |
| 324 | HELPARG("age[-age]") HELPOPT("[--web,--html] set limits of colour coding") \ |
| 325 | VAL(SERVERADDR) LONG(address) LONG(addr) LONG(server_address) \ |
| 326 | LONG(server_addr) \ |
| 327 | HELPARG("addr[:port]") HELPOPT("[--web] specify HTTP server address") \ |
| 328 | VAL(AUTH) LONG(auth) LONG(http_auth) LONG(httpd_auth) \ |
| 329 | LONG(server_auth) LONG(web_auth) \ |
| 330 | HELPARG("type") HELPOPT("[--web] specify HTTP authentication method") \ |
| 331 | VAL(AUTHFILE) LONG(auth_file) \ |
| 332 | HELPARG("filename") HELPOPT("[--web] read HTTP Basic user/pass from file") \ |
| 333 | VAL(AUTHFD) LONG(auth_fd) \ |
| 334 | HELPARG("fd") HELPOPT("[--web] read HTTP Basic user/pass from fd") \ |
| 335 | VAL(TQDEPTH) SHORT(d) LONG(depth) LONG(max_depth) LONG(maximum_depth) \ |
| 336 | HELPARG("levels") HELPOPT("[--text] recurse to this many levels") \ |
| 337 | VAL(MINAGE) SHORT(a) LONG(age) LONG(min_age) LONG(minimum_age) \ |
| 338 | HELPARG("age") HELPOPT("[--text] include only files older than this") \ |
| 339 | HELPPFX("also") \ |
| 340 | NOVAL(HELP) SHORT(h) LONG(help) HELPOPT("display this help text") \ |
| 341 | NOVAL(VERSION) SHORT(V) LONG(version) HELPOPT("report version number") \ |
| 342 | NOVAL(LICENCE) LONG(licence) LONG(license) \ |
| 343 | HELPOPT("display (MIT) licence text") \ |
| 344 | |
| 345 | #define IGNORE(x) |
| 346 | #define DEFENUM(x) OPT_ ## x, |
| 347 | #define ZERO(x) 0, |
| 348 | #define ONE(x) 1, |
| 349 | #define STRING(x) #x , |
| 350 | #define STRINGNOCOMMA(x) #x |
| 351 | #define SHORTNEWOPT(x) SHORTtmp_ ## x = OPT_ ## x, |
| 352 | #define SHORTTHISOPT(x) SHORTtmp2_ ## x, SHORTVAL_ ## x = SHORTtmp2_ ## x - 1, |
| 353 | #define SHORTOPTVAL(x) SHORTVAL_ ## x, |
| 354 | #define SHORTTMP(x) SHORTtmp3_ ## x, |
| 355 | #define LONGNEWOPT(x) LONGtmp_ ## x = OPT_ ## x, |
| 356 | #define LONGTHISOPT(x) LONGtmp2_ ## x, LONGVAL_ ## x = LONGtmp2_ ## x - 1, |
| 357 | #define LONGOPTVAL(x) LONGVAL_ ## x, |
| 358 | #define LONGTMP(x) SHORTtmp3_ ## x, |
| 359 | |
| 360 | #define OPTIONS(NOVAL, VAL, SHORT, LONG) \ |
| 361 | OPTHELP(NOVAL, VAL, SHORT, LONG, IGNORE, IGNORE, IGNORE, IGNORE) |
| 362 | |
| 363 | enum { OPTIONS(DEFENUM,DEFENUM,IGNORE,IGNORE) NOPTIONS }; |
| 364 | enum { OPTIONS(IGNORE,IGNORE,SHORTTMP,IGNORE) NSHORTOPTS }; |
| 365 | enum { OPTIONS(IGNORE,IGNORE,IGNORE,LONGTMP) NLONGOPTS }; |
| 366 | static const int opthasval[NOPTIONS] = {OPTIONS(ZERO,ONE,IGNORE,IGNORE)}; |
| 367 | static const char shortopts[] = {OPTIONS(IGNORE,IGNORE,STRINGNOCOMMA,IGNORE)}; |
| 368 | static const char *const longopts[] = {OPTIONS(IGNORE,IGNORE,IGNORE,STRING)}; |
| 369 | enum { OPTIONS(SHORTNEWOPT,SHORTNEWOPT,SHORTTHISOPT,IGNORE) }; |
| 370 | enum { OPTIONS(LONGNEWOPT,LONGNEWOPT,IGNORE,LONGTHISOPT) }; |
| 371 | static const int shortvals[] = {OPTIONS(IGNORE,IGNORE,SHORTOPTVAL,IGNORE)}; |
| 372 | static const int longvals[] = {OPTIONS(IGNORE,IGNORE,IGNORE,LONGOPTVAL)}; |
| 373 | |
| 374 | static void usage(FILE *fp) |
| 375 | { |
| 376 | char longbuf[80]; |
| 377 | const char *prefix, *shortopt, *longopt, *optarg; |
| 378 | int i, optex; |
| 379 | |
| 380 | #define HELPRESET prefix = shortopt = longopt = optarg = NULL, optex = -1 |
| 381 | #define HELPNOVAL(s) optex = 0; |
| 382 | #define HELPVAL(s) optex = 1; |
| 383 | #define HELPSHORT(s) if (!shortopt) shortopt = "-" #s; |
| 384 | #define HELPLONG(s) if (!longopt) { \ |
| 385 | strcpy(longbuf, "--" #s); longopt = longbuf; \ |
| 386 | for (i = 0; longbuf[i]; i++) if (longbuf[i] == '_') longbuf[i] = '-'; } |
| 387 | #define HELPPFX(s) prefix = s; |
| 388 | #define HELPARG(s) optarg = s; |
| 389 | #define HELPLINE(s) assert(optex == -1); \ |
| 390 | fprintf(fp, "%7s%c %s\n", prefix?prefix:"", prefix?':':' ', s); \ |
| 391 | HELPRESET; |
| 392 | #define HELPOPT(s) assert((optex == 1 && optarg) || (optex == 0 && !optarg)); \ |
| 393 | assert(shortopt || longopt); \ |
| 394 | i = fprintf(fp, "%7s%c %s%s%s%s%s", prefix?prefix:"", prefix?':':' ', \ |
| 395 | shortopt?shortopt:"", shortopt&&longopt?", ":"", longopt?longopt:"", \ |
| 396 | optarg?" ":"", optarg?optarg:""); \ |
| 397 | fprintf(fp, "%*s %s\n", i<32?32-i:0,"",s); HELPRESET; |
| 398 | |
| 399 | HELPRESET; |
| 400 | OPTHELP(HELPNOVAL, HELPVAL, HELPSHORT, HELPLONG, |
| 401 | HELPPFX, HELPARG, HELPLINE, HELPOPT); |
| 402 | |
| 403 | #undef HELPRESET |
| 404 | #undef HELPNOVAL |
| 405 | #undef HELPVAL |
| 406 | #undef HELPSHORT |
| 407 | #undef HELPLONG |
| 408 | #undef HELPPFX |
| 409 | #undef HELPARG |
| 410 | #undef HELPLINE |
| 411 | #undef HELPOPT |
| 412 | } |
| 413 | |
| 414 | static time_t parse_age(time_t now, const char *agestr) |
| 415 | { |
| 416 | time_t t; |
| 417 | struct tm tm; |
| 418 | int nunits; |
| 419 | char unit[2]; |
| 420 | |
| 421 | t = now; |
| 422 | |
| 423 | if (2 != sscanf(agestr, "%d%1[DdWwMmYy]", &nunits, unit)) { |
| 424 | fprintf(stderr, "%s: age specification should be a number followed by" |
| 425 | " one of d,w,m,y\n", PNAME); |
| 426 | exit(1); |
| 427 | } |
| 428 | |
| 429 | if (unit[0] == 'd') { |
| 430 | t -= 86400 * nunits; |
| 431 | } else if (unit[0] == 'w') { |
| 432 | t -= 86400 * 7 * nunits; |
| 433 | } else { |
| 434 | int ym; |
| 435 | |
| 436 | tm = *localtime(&t); |
| 437 | ym = tm.tm_year * 12 + tm.tm_mon; |
| 438 | |
| 439 | if (unit[0] == 'm') |
| 440 | ym -= nunits; |
| 441 | else |
| 442 | ym -= 12 * nunits; |
| 443 | |
| 444 | tm.tm_year = ym / 12; |
| 445 | tm.tm_mon = ym % 12; |
| 446 | |
| 447 | t = mktime(&tm); |
| 448 | } |
| 449 | |
| 450 | return t; |
| 451 | } |
| 452 | |
| 453 | int main(int argc, char **argv) |
| 454 | { |
| 455 | int fd, count; |
| 456 | struct ctx actx, *ctx = &actx; |
| 457 | struct stat st; |
| 458 | off_t totalsize, realsize; |
| 459 | void *mappedfile; |
| 460 | triewalk *tw; |
| 461 | indexbuild *ib; |
| 462 | const struct trie_file *tf, *prevtf; |
| 463 | char *filename = PNAME ".dat"; |
| 464 | int doing_opts = 1; |
| 465 | enum { TEXT, HTML, SCAN, DUMP, SCANDUMP, LOAD, HTTPD, REMOVE }; |
| 466 | struct action { |
| 467 | int mode; |
| 468 | char *arg; |
| 469 | } *actions = NULL; |
| 470 | int nactions = 0, actionsize = 0, action; |
| 471 | time_t now = time(NULL); |
| 472 | time_t textcutoff = now, htmlnewest = now, htmloldest = now; |
| 473 | int htmlautoagerange = 1; |
| 474 | const char *httpserveraddr = NULL; |
| 475 | int httpserverport = 0; |
| 476 | const char *httpauthdata = NULL; |
| 477 | int auth = HTTPD_AUTH_MAGIC | HTTPD_AUTH_BASIC; |
| 478 | int progress = 1; |
| 479 | struct inclusion_exclusion *inex = NULL; |
| 480 | int ninex = 0, inexsize = 0; |
| 481 | int crossfs = 0; |
| 482 | int tqdepth = 1; |
| 483 | int fakediratimes = 1; |
| 484 | int mtime = 0; |
| 485 | |
| 486 | #ifdef DEBUG_MAD_OPTION_PARSING_MACROS |
| 487 | { |
| 488 | static const char *const optnames[NOPTIONS] = { |
| 489 | OPTIONS(STRING,STRING,IGNORE,IGNORE) |
| 490 | }; |
| 491 | int i; |
| 492 | for (i = 0; i < NSHORTOPTS; i++) |
| 493 | printf("-%c == %s [%s]\n", shortopts[i], optnames[shortvals[i]], |
| 494 | opthasval[shortvals[i]] ? "value" : "no value"); |
| 495 | for (i = 0; i < NLONGOPTS; i++) |
| 496 | printf("--%s == %s [%s]\n", longopts[i], optnames[longvals[i]], |
| 497 | opthasval[longvals[i]] ? "value" : "no value"); |
| 498 | } |
| 499 | #endif |
| 500 | |
| 501 | while (--argc > 0) { |
| 502 | char *p = *++argv; |
| 503 | |
| 504 | if (doing_opts && *p == '-') { |
| 505 | int wordstart = 1; |
| 506 | |
| 507 | if (!strcmp(p, "--")) { |
| 508 | doing_opts = 0; |
| 509 | continue; |
| 510 | } |
| 511 | |
| 512 | p++; |
| 513 | while (*p) { |
| 514 | int optid = -1; |
| 515 | int i; |
| 516 | char *optval; |
| 517 | |
| 518 | if (wordstart && *p == '-') { |
| 519 | /* |
| 520 | * GNU-style long option. |
| 521 | */ |
| 522 | p++; |
| 523 | optval = strchr(p, '='); |
| 524 | if (optval) |
| 525 | *optval++ = '\0'; |
| 526 | |
| 527 | for (i = 0; i < NLONGOPTS; i++) { |
| 528 | const char *opt = longopts[i], *s = p; |
| 529 | int match = 1; |
| 530 | /* |
| 531 | * The underscores in the option names |
| 532 | * defined above may be given by the user |
| 533 | * as underscores or dashes, or omitted |
| 534 | * entirely. |
| 535 | */ |
| 536 | while (*opt) { |
| 537 | if (*opt == '_') { |
| 538 | if (*s == '-' || *s == '_') |
| 539 | s++; |
| 540 | } else { |
| 541 | if (*opt != *s) { |
| 542 | match = 0; |
| 543 | break; |
| 544 | } |
| 545 | s++; |
| 546 | } |
| 547 | opt++; |
| 548 | } |
| 549 | if (match && !*s) { |
| 550 | optid = longvals[i]; |
| 551 | break; |
| 552 | } |
| 553 | } |
| 554 | |
| 555 | if (optid < 0) { |
| 556 | fprintf(stderr, "%s: unrecognised option '--%s'\n", |
| 557 | PNAME, p); |
| 558 | return 1; |
| 559 | } |
| 560 | |
| 561 | if (!opthasval[optid]) { |
| 562 | if (optval) { |
| 563 | fprintf(stderr, "%s: unexpected argument to option" |
| 564 | " '--%s'\n", PNAME, p); |
| 565 | return 1; |
| 566 | } |
| 567 | } else { |
| 568 | if (!optval) { |
| 569 | if (--argc > 0) { |
| 570 | optval = *++argv; |
| 571 | } else { |
| 572 | fprintf(stderr, "%s: option '--%s' expects" |
| 573 | " an argument\n", PNAME, p); |
| 574 | return 1; |
| 575 | } |
| 576 | } |
| 577 | } |
| 578 | |
| 579 | p += strlen(p); /* finished with this argument word */ |
| 580 | } else { |
| 581 | /* |
| 582 | * Short option. |
| 583 | */ |
| 584 | char c = *p++; |
| 585 | |
| 586 | for (i = 0; i < NSHORTOPTS; i++) |
| 587 | if (c == shortopts[i]) { |
| 588 | optid = shortvals[i]; |
| 589 | break; |
| 590 | } |
| 591 | |
| 592 | if (optid < 0) { |
| 593 | fprintf(stderr, "%s: unrecognised option '-%c'\n", |
| 594 | PNAME, c); |
| 595 | return 1; |
| 596 | } |
| 597 | |
| 598 | if (opthasval[optid]) { |
| 599 | if (*p) { |
| 600 | optval = p; |
| 601 | p += strlen(p); |
| 602 | } else if (--argc > 0) { |
| 603 | optval = *++argv; |
| 604 | } else { |
| 605 | fprintf(stderr, "%s: option '-%c' expects" |
| 606 | " an argument\n", PNAME, c); |
| 607 | return 1; |
| 608 | } |
| 609 | } else { |
| 610 | optval = NULL; |
| 611 | } |
| 612 | } |
| 613 | |
| 614 | wordstart = 0; |
| 615 | |
| 616 | /* |
| 617 | * Now actually process the option. |
| 618 | */ |
| 619 | switch (optid) { |
| 620 | case OPT_HELP: |
| 621 | usage(stdout); |
| 622 | return 0; |
| 623 | case OPT_VERSION: |
| 624 | #ifdef PACKAGE_VERSION |
| 625 | printf("%s, revision %s\n", PNAME, PACKAGE_VERSION); |
| 626 | #else |
| 627 | printf("%s: version number not available when not built" |
| 628 | " via automake\n", PNAME); |
| 629 | #endif |
| 630 | return 0; |
| 631 | case OPT_LICENCE: |
| 632 | { |
| 633 | extern const char *const licence[]; |
| 634 | int i; |
| 635 | |
| 636 | for (i = 0; licence[i]; i++) |
| 637 | fputs(licence[i], stdout); |
| 638 | |
| 639 | return 0; |
| 640 | } |
| 641 | return 0; |
| 642 | case OPT_SCAN: |
| 643 | if (nactions >= actionsize) { |
| 644 | actionsize = nactions * 3 / 2 + 16; |
| 645 | actions = sresize(actions, actionsize, struct action); |
| 646 | } |
| 647 | actions[nactions].mode = SCAN; |
| 648 | actions[nactions].arg = optval; |
| 649 | nactions++; |
| 650 | break; |
| 651 | case OPT_SCANDUMP: |
| 652 | if (nactions >= actionsize) { |
| 653 | actionsize = nactions * 3 / 2 + 16; |
| 654 | actions = sresize(actions, actionsize, struct action); |
| 655 | } |
| 656 | actions[nactions].mode = SCANDUMP; |
| 657 | actions[nactions].arg = optval; |
| 658 | nactions++; |
| 659 | break; |
| 660 | case OPT_DUMP: |
| 661 | if (nactions >= actionsize) { |
| 662 | actionsize = nactions * 3 / 2 + 16; |
| 663 | actions = sresize(actions, actionsize, struct action); |
| 664 | } |
| 665 | actions[nactions].mode = DUMP; |
| 666 | actions[nactions].arg = NULL; |
| 667 | nactions++; |
| 668 | break; |
| 669 | case OPT_LOAD: |
| 670 | if (nactions >= actionsize) { |
| 671 | actionsize = nactions * 3 / 2 + 16; |
| 672 | actions = sresize(actions, actionsize, struct action); |
| 673 | } |
| 674 | actions[nactions].mode = LOAD; |
| 675 | actions[nactions].arg = NULL; |
| 676 | nactions++; |
| 677 | break; |
| 678 | case OPT_TEXT: |
| 679 | if (nactions >= actionsize) { |
| 680 | actionsize = nactions * 3 / 2 + 16; |
| 681 | actions = sresize(actions, actionsize, struct action); |
| 682 | } |
| 683 | actions[nactions].mode = TEXT; |
| 684 | actions[nactions].arg = optval; |
| 685 | nactions++; |
| 686 | break; |
| 687 | case OPT_HTML: |
| 688 | if (nactions >= actionsize) { |
| 689 | actionsize = nactions * 3 / 2 + 16; |
| 690 | actions = sresize(actions, actionsize, struct action); |
| 691 | } |
| 692 | actions[nactions].mode = HTML; |
| 693 | actions[nactions].arg = optval; |
| 694 | nactions++; |
| 695 | break; |
| 696 | case OPT_HTTPD: |
| 697 | if (nactions >= actionsize) { |
| 698 | actionsize = nactions * 3 / 2 + 16; |
| 699 | actions = sresize(actions, actionsize, struct action); |
| 700 | } |
| 701 | actions[nactions].mode = HTTPD; |
| 702 | actions[nactions].arg = NULL; |
| 703 | nactions++; |
| 704 | break; |
| 705 | case OPT_REMOVE: |
| 706 | if (nactions >= actionsize) { |
| 707 | actionsize = nactions * 3 / 2 + 16; |
| 708 | actions = sresize(actions, actionsize, struct action); |
| 709 | } |
| 710 | actions[nactions].mode = REMOVE; |
| 711 | actions[nactions].arg = NULL; |
| 712 | nactions++; |
| 713 | break; |
| 714 | case OPT_PROGRESS: |
| 715 | progress = 2; |
| 716 | break; |
| 717 | case OPT_NOPROGRESS: |
| 718 | progress = 0; |
| 719 | break; |
| 720 | case OPT_TTYPROGRESS: |
| 721 | progress = 1; |
| 722 | break; |
| 723 | case OPT_CROSSFS: |
| 724 | crossfs = 1; |
| 725 | break; |
| 726 | case OPT_NOCROSSFS: |
| 727 | crossfs = 0; |
| 728 | break; |
| 729 | case OPT_DIRATIME: |
| 730 | fakediratimes = 0; |
| 731 | break; |
| 732 | case OPT_NODIRATIME: |
| 733 | fakediratimes = 1; |
| 734 | break; |
| 735 | case OPT_MTIME: |
| 736 | mtime = 1; |
| 737 | break; |
| 738 | case OPT_DATAFILE: |
| 739 | filename = optval; |
| 740 | break; |
| 741 | case OPT_TQDEPTH: |
| 742 | tqdepth = atoi(optval); |
| 743 | break; |
| 744 | case OPT_MINAGE: |
| 745 | textcutoff = parse_age(now, optval); |
| 746 | break; |
| 747 | case OPT_AGERANGE: |
| 748 | if (!strcmp(optval, "auto")) { |
| 749 | htmlautoagerange = 1; |
| 750 | } else { |
| 751 | char *q = optval + strcspn(optval, "-:"); |
| 752 | if (*q) |
| 753 | *q++ = '\0'; |
| 754 | htmloldest = parse_age(now, optval); |
| 755 | htmlnewest = *q ? parse_age(now, q) : now; |
| 756 | htmlautoagerange = 0; |
| 757 | } |
| 758 | break; |
| 759 | case OPT_SERVERADDR: |
| 760 | { |
| 761 | char *port; |
| 762 | if (optval[0] == '[' && |
| 763 | (port = strchr(optval, ']')) != NULL) |
| 764 | port++; |
| 765 | else |
| 766 | port = optval; |
| 767 | port += strcspn(port, ":"); |
| 768 | if (port) |
| 769 | *port++ = '\0'; |
| 770 | httpserveraddr = optval; |
| 771 | httpserverport = atoi(port); |
| 772 | } |
| 773 | break; |
| 774 | case OPT_AUTH: |
| 775 | if (!strcmp(optval, "magic")) |
| 776 | auth = HTTPD_AUTH_MAGIC; |
| 777 | else if (!strcmp(optval, "basic")) |
| 778 | auth = HTTPD_AUTH_BASIC; |
| 779 | else if (!strcmp(optval, "none")) |
| 780 | auth = HTTPD_AUTH_NONE; |
| 781 | else if (!strcmp(optval, "default")) |
| 782 | auth = HTTPD_AUTH_MAGIC | HTTPD_AUTH_BASIC; |
| 783 | else if (!strcmp(optval, "help") || |
| 784 | !strcmp(optval, "list")) { |
| 785 | printf(PNAME ": supported HTTP authentication types" |
| 786 | " are:\n" |
| 787 | " magic use Linux /proc/net/tcp to" |
| 788 | " determine owner of peer socket\n" |
| 789 | " basic HTTP Basic username and" |
| 790 | " password authentication\n" |
| 791 | " default use 'magic' if possible, " |
| 792 | " otherwise fall back to 'basic'\n" |
| 793 | " none unauthenticated HTTP (if" |
| 794 | " the data file is non-confidential)\n"); |
| 795 | return 0; |
| 796 | } else { |
| 797 | fprintf(stderr, "%s: unrecognised authentication" |
| 798 | " type '%s'\n%*s options are 'magic'," |
| 799 | " 'basic', 'none', 'default'\n", |
| 800 | PNAME, optval, (int)strlen(PNAME), ""); |
| 801 | return 1; |
| 802 | } |
| 803 | break; |
| 804 | case OPT_AUTHFILE: |
| 805 | case OPT_AUTHFD: |
| 806 | { |
| 807 | int fd; |
| 808 | char namebuf[40]; |
| 809 | const char *name; |
| 810 | char *authbuf; |
| 811 | int authlen, authsize; |
| 812 | int ret; |
| 813 | |
| 814 | if (optid == OPT_AUTHFILE) { |
| 815 | fd = open(optval, O_RDONLY); |
| 816 | if (fd < 0) { |
| 817 | fprintf(stderr, "%s: %s: open: %s\n", PNAME, |
| 818 | optval, strerror(errno)); |
| 819 | return 1; |
| 820 | } |
| 821 | name = optval; |
| 822 | } else { |
| 823 | fd = atoi(optval); |
| 824 | name = namebuf; |
| 825 | sprintf(namebuf, "fd %d", fd); |
| 826 | } |
| 827 | |
| 828 | authlen = 0; |
| 829 | authsize = 256; |
| 830 | authbuf = snewn(authsize, char); |
| 831 | while ((ret = read(fd, authbuf+authlen, |
| 832 | authsize-authlen)) > 0) { |
| 833 | authlen += ret; |
| 834 | if ((authsize - authlen) < (authsize / 16)) { |
| 835 | authsize = authlen * 3 / 2 + 4096; |
| 836 | authbuf = sresize(authbuf, authsize, char); |
| 837 | } |
| 838 | } |
| 839 | if (ret < 0) { |
| 840 | fprintf(stderr, "%s: %s: read: %s\n", PNAME, |
| 841 | name, strerror(errno)); |
| 842 | return 1; |
| 843 | } |
| 844 | if (optid == OPT_AUTHFILE) |
| 845 | close(fd); |
| 846 | httpauthdata = authbuf; |
| 847 | } |
| 848 | break; |
| 849 | case OPT_INCLUDE: |
| 850 | case OPT_INCLUDEPATH: |
| 851 | case OPT_EXCLUDE: |
| 852 | case OPT_EXCLUDEPATH: |
| 853 | case OPT_PRUNE: |
| 854 | case OPT_PRUNEPATH: |
| 855 | if (ninex >= inexsize) { |
| 856 | inexsize = ninex * 3 / 2 + 16; |
| 857 | inex = sresize(inex, inexsize, |
| 858 | struct inclusion_exclusion); |
| 859 | } |
| 860 | inex[ninex].path = (optid == OPT_INCLUDEPATH || |
| 861 | optid == OPT_EXCLUDEPATH || |
| 862 | optid == OPT_PRUNEPATH); |
| 863 | inex[ninex].type = (optid == OPT_INCLUDE ? 1 : |
| 864 | optid == OPT_INCLUDEPATH ? 1 : |
| 865 | optid == OPT_EXCLUDE ? 0 : |
| 866 | optid == OPT_EXCLUDEPATH ? 0 : |
| 867 | optid == OPT_PRUNE ? -1 : |
| 868 | /* optid == OPT_PRUNEPATH ? */ -1); |
| 869 | inex[ninex].wildcard = optval; |
| 870 | ninex++; |
| 871 | break; |
| 872 | } |
| 873 | } |
| 874 | } else { |
| 875 | fprintf(stderr, "%s: unexpected argument '%s'\n", PNAME, p); |
| 876 | return 1; |
| 877 | } |
| 878 | } |
| 879 | |
| 880 | if (nactions == 0) { |
| 881 | usage(stderr); |
| 882 | return 1; |
| 883 | } |
| 884 | |
| 885 | for (action = 0; action < nactions; action++) { |
| 886 | int mode = actions[action].mode; |
| 887 | |
| 888 | if (mode == SCAN || mode == SCANDUMP || mode == LOAD) { |
| 889 | const char *scandir = actions[action].arg; |
| 890 | |
| 891 | if (mode == LOAD) { |
| 892 | char *buf = fgetline(stdin); |
| 893 | unsigned newpathsep; |
| 894 | buf[strcspn(buf, "\r\n")] = '\0'; |
| 895 | if (1 != sscanf(buf, DUMPHDR "%x", |
| 896 | &newpathsep)) { |
| 897 | fprintf(stderr, "%s: header in dump file not recognised\n", |
| 898 | PNAME); |
| 899 | return 1; |
| 900 | } |
| 901 | pathsep = (char)newpathsep; |
| 902 | sfree(buf); |
| 903 | } |
| 904 | |
| 905 | if (mode == SCAN || mode == LOAD) { |
| 906 | /* |
| 907 | * Prepare to write out the index file. |
| 908 | */ |
| 909 | fd = open(filename, O_RDWR | O_TRUNC | O_CREAT, |
| 910 | S_IRUSR | S_IWUSR); |
| 911 | if (fd < 0) { |
| 912 | fprintf(stderr, "%s: %s: open: %s\n", PNAME, filename, |
| 913 | strerror(errno)); |
| 914 | return 1; |
| 915 | } |
| 916 | if (fstat(fd, &st) < 0) { |
| 917 | perror(PNAME ": fstat"); |
| 918 | return 1; |
| 919 | } |
| 920 | ctx->datafile_dev = st.st_dev; |
| 921 | ctx->datafile_ino = st.st_ino; |
| 922 | ctx->straight_to_dump = 0; |
| 923 | } else { |
| 924 | ctx->datafile_dev = -1; |
| 925 | ctx->datafile_ino = -1; |
| 926 | ctx->straight_to_dump = 1; |
| 927 | } |
| 928 | |
| 929 | if (mode == SCAN || mode == SCANDUMP) { |
| 930 | if (stat(scandir, &st) < 0) { |
| 931 | fprintf(stderr, "%s: %s: stat: %s\n", PNAME, scandir, |
| 932 | strerror(errno)); |
| 933 | return 1; |
| 934 | } |
| 935 | ctx->filesystem_dev = crossfs ? 0 : st.st_dev; |
| 936 | } |
| 937 | |
| 938 | ctx->inex = inex; |
| 939 | ctx->ninex = ninex; |
| 940 | ctx->crossfs = crossfs; |
| 941 | ctx->fakeatimes = fakediratimes; |
| 942 | ctx->usemtime = mtime; |
| 943 | |
| 944 | ctx->last_output_update = time(NULL); |
| 945 | |
| 946 | /* progress==1 means report progress only if stderr is a tty */ |
| 947 | if (progress == 1) |
| 948 | progress = isatty(2) ? 2 : 0; |
| 949 | ctx->progress = progress; |
| 950 | { |
| 951 | struct winsize ws; |
| 952 | if (progress && |
| 953 | ioctl(2, TIOCGWINSZ, &ws) == 0 && |
| 954 | ws.ws_col > 0) |
| 955 | ctx->progwidth = ws.ws_col - 1; |
| 956 | else |
| 957 | ctx->progwidth = 79; |
| 958 | } |
| 959 | |
| 960 | if (mode == SCANDUMP) |
| 961 | printf(DUMPHDR "%02x\n", (unsigned char)pathsep); |
| 962 | |
| 963 | /* |
| 964 | * Scan the directory tree, and write out the trie component |
| 965 | * of the data file. |
| 966 | */ |
| 967 | if (mode != SCANDUMP) { |
| 968 | ctx->tb = triebuild_new(fd); |
| 969 | } |
| 970 | if (mode == LOAD) { |
| 971 | char *buf; |
| 972 | int line = 2; |
| 973 | while ((buf = fgetline(stdin)) != NULL) { |
| 974 | struct trie_file tf; |
| 975 | char *p, *q; |
| 976 | |
| 977 | buf[strcspn(buf, "\r\n")] = '\0'; |
| 978 | |
| 979 | p = buf; |
| 980 | q = p; |
| 981 | while (*p && *p != ' ') p++; |
| 982 | if (!*p) { |
| 983 | fprintf(stderr, "%s: dump file line %d: expected at least" |
| 984 | " three fields\n", PNAME, line); |
| 985 | return 1; |
| 986 | } |
| 987 | *p++ = '\0'; |
| 988 | tf.size = strtoull(q, NULL, 10); |
| 989 | q = p; |
| 990 | while (*p && *p != ' ') p++; |
| 991 | if (!*p) { |
| 992 | fprintf(stderr, "%s: dump file line %d: expected at least" |
| 993 | " three fields\n", PNAME, line); |
| 994 | return 1; |
| 995 | } |
| 996 | *p++ = '\0'; |
| 997 | tf.atime = strtoull(q, NULL, 10); |
| 998 | q = buf; |
| 999 | while (*p) { |
| 1000 | int c = *p; |
| 1001 | if (*p == '%') { |
| 1002 | int i; |
| 1003 | p++; |
| 1004 | c = 0; |
| 1005 | for (i = 0; i < 2; i++) { |
| 1006 | c *= 16; |
| 1007 | if (*p >= '0' && *p <= '9') |
| 1008 | c += *p - '0'; |
| 1009 | else if (*p >= 'A' && *p <= 'F') |
| 1010 | c += *p - ('A' - 10); |
| 1011 | else if (*p >= 'a' && *p <= 'f') |
| 1012 | c += *p - ('a' - 10); |
| 1013 | else { |
| 1014 | fprintf(stderr, "%s: dump file line %d: unable" |
| 1015 | " to parse hex escape\n", PNAME, line); |
| 1016 | } |
| 1017 | p++; |
| 1018 | } |
| 1019 | } |
| 1020 | *q++ = c; |
| 1021 | p++; |
| 1022 | } |
| 1023 | *q = '\0'; |
| 1024 | triebuild_add(ctx->tb, buf, &tf); |
| 1025 | sfree(buf); |
| 1026 | line++; |
| 1027 | } |
| 1028 | } else { |
| 1029 | du(scandir, gotdata, scan_error, ctx); |
| 1030 | } |
| 1031 | if (mode != SCANDUMP) { |
| 1032 | size_t maxpathlen; |
| 1033 | size_t delta; |
| 1034 | char *buf, *prevbuf; |
| 1035 | |
| 1036 | count = triebuild_finish(ctx->tb); |
| 1037 | triebuild_free(ctx->tb); |
| 1038 | |
| 1039 | if (ctx->progress) { |
| 1040 | fprintf(stderr, "%-*s\r", ctx->progwidth, ""); |
| 1041 | fflush(stderr); |
| 1042 | } |
| 1043 | |
| 1044 | /* |
| 1045 | * Work out how much space the cumulative index trees |
| 1046 | * will take; enlarge the file, and memory-map it. |
| 1047 | */ |
| 1048 | if (fstat(fd, &st) < 0) { |
| 1049 | perror(PNAME ": fstat"); |
| 1050 | return 1; |
| 1051 | } |
| 1052 | |
| 1053 | printf("Built pathname index, %d entries," |
| 1054 | " %llu bytes of index\n", count, |
| 1055 | (unsigned long long)st.st_size); |
| 1056 | |
| 1057 | totalsize = index_initial_size(st.st_size, count); |
| 1058 | totalsize += totalsize / 10; |
| 1059 | |
| 1060 | if (lseek(fd, totalsize-1, SEEK_SET) < 0) { |
| 1061 | perror(PNAME ": lseek"); |
| 1062 | return 1; |
| 1063 | } |
| 1064 | if (write(fd, "\0", 1) < 1) { |
| 1065 | perror(PNAME ": write"); |
| 1066 | return 1; |
| 1067 | } |
| 1068 | |
| 1069 | mappedfile = mmap(NULL, totalsize, PROT_READ|PROT_WRITE,MAP_SHARED, fd, 0); |
| 1070 | if (!mappedfile) { |
| 1071 | perror(PNAME ": mmap"); |
| 1072 | return 1; |
| 1073 | } |
| 1074 | |
| 1075 | if (fakediratimes) { |
| 1076 | printf("Faking directory atimes\n"); |
| 1077 | trie_fake_dir_atimes(mappedfile); |
| 1078 | } |
| 1079 | |
| 1080 | printf("Building index\n"); |
| 1081 | ib = indexbuild_new(mappedfile, st.st_size, count, &delta); |
| 1082 | maxpathlen = trie_maxpathlen(mappedfile); |
| 1083 | buf = snewn(maxpathlen, char); |
| 1084 | prevbuf = snewn(maxpathlen, char); |
| 1085 | tw = triewalk_new(mappedfile); |
| 1086 | prevbuf[0] = '\0'; |
| 1087 | tf = triewalk_next(tw, buf); |
| 1088 | assert(tf); |
| 1089 | while (1) { |
| 1090 | int i; |
| 1091 | |
| 1092 | if (totalsize - indexbuild_realsize(ib) < delta) { |
| 1093 | const void *oldfile = mappedfile; |
| 1094 | ptrdiff_t diff; |
| 1095 | |
| 1096 | /* |
| 1097 | * Unmap the file, grow it, and remap it. |
| 1098 | */ |
| 1099 | munmap(mappedfile, totalsize); |
| 1100 | |
| 1101 | totalsize += delta; |
| 1102 | totalsize += totalsize / 10; |
| 1103 | |
| 1104 | if (lseek(fd, totalsize-1, SEEK_SET) < 0) { |
| 1105 | perror(PNAME ": lseek"); |
| 1106 | return 1; |
| 1107 | } |
| 1108 | if (write(fd, "\0", 1) < 1) { |
| 1109 | perror(PNAME ": write"); |
| 1110 | return 1; |
| 1111 | } |
| 1112 | |
| 1113 | mappedfile = mmap(NULL, totalsize, PROT_READ|PROT_WRITE,MAP_SHARED, fd, 0); |
| 1114 | if (!mappedfile) { |
| 1115 | perror(PNAME ": mmap"); |
| 1116 | return 1; |
| 1117 | } |
| 1118 | |
| 1119 | indexbuild_rebase(ib, mappedfile); |
| 1120 | triewalk_rebase(tw, mappedfile); |
| 1121 | diff = (const unsigned char *)mappedfile - |
| 1122 | (const unsigned char *)oldfile; |
| 1123 | if (prevtf) |
| 1124 | prevtf = (const struct trie_file *) |
| 1125 | (((const unsigned char *)prevtf) + diff); |
| 1126 | if (tf) |
| 1127 | tf = (const struct trie_file *) |
| 1128 | (((const unsigned char *)tf) + diff); |
| 1129 | } |
| 1130 | |
| 1131 | /* |
| 1132 | * Get the next file from the index. So we are |
| 1133 | * currently holding, and have not yet |
| 1134 | * indexed, prevtf (with pathname prevbuf) and |
| 1135 | * tf (with pathname buf). |
| 1136 | */ |
| 1137 | prevtf = tf; |
| 1138 | memcpy(prevbuf, buf, maxpathlen); |
| 1139 | tf = triewalk_next(tw, buf); |
| 1140 | |
| 1141 | if (!tf) |
| 1142 | buf[0] = '\0'; |
| 1143 | |
| 1144 | /* |
| 1145 | * Find the first differing character position |
| 1146 | * between our two pathnames. |
| 1147 | */ |
| 1148 | for (i = 0; prevbuf[i] && prevbuf[i] == buf[i]; i++); |
| 1149 | |
| 1150 | /* |
| 1151 | * If prevbuf was a directory name and buf is |
| 1152 | * something inside that directory, then |
| 1153 | * trie_before() will be called on prevbuf |
| 1154 | * itself. Hence we must drop a tag before it, |
| 1155 | * so that the resulting index is usable. |
| 1156 | */ |
| 1157 | if ((!prevbuf[i] && (buf[i] == pathsep || |
| 1158 | (i > 0 && buf[i-1] == pathsep)))) |
| 1159 | indexbuild_tag(ib); |
| 1160 | |
| 1161 | /* |
| 1162 | * Add prevtf to the index. |
| 1163 | */ |
| 1164 | indexbuild_add(ib, prevtf); |
| 1165 | |
| 1166 | if (!tf) { |
| 1167 | /* |
| 1168 | * Drop an unconditional final tag, and |
| 1169 | * get out of this loop. |
| 1170 | */ |
| 1171 | indexbuild_tag(ib); |
| 1172 | break; |
| 1173 | } |
| 1174 | |
| 1175 | /* |
| 1176 | * If prevbuf was a filename inside some |
| 1177 | * directory which buf is outside, then |
| 1178 | * trie_before() will be called on some |
| 1179 | * pathname either equal to buf or epsilon |
| 1180 | * less than it. Either way, we're going to |
| 1181 | * need to drop a tag after prevtf. |
| 1182 | */ |
| 1183 | if (strchr(prevbuf+i, pathsep) || !tf) |
| 1184 | indexbuild_tag(ib); |
| 1185 | } |
| 1186 | |
| 1187 | triewalk_free(tw); |
| 1188 | realsize = indexbuild_realsize(ib); |
| 1189 | indexbuild_free(ib); |
| 1190 | |
| 1191 | munmap(mappedfile, totalsize); |
| 1192 | ftruncate(fd, realsize); |
| 1193 | close(fd); |
| 1194 | printf("Final index file size = %llu bytes\n", |
| 1195 | (unsigned long long)realsize); |
| 1196 | } |
| 1197 | } else if (mode == TEXT) { |
| 1198 | char *querydir = actions[action].arg; |
| 1199 | size_t pathlen; |
| 1200 | |
| 1201 | fd = open(filename, O_RDONLY); |
| 1202 | if (fd < 0) { |
| 1203 | fprintf(stderr, "%s: %s: open: %s\n", PNAME, filename, |
| 1204 | strerror(errno)); |
| 1205 | return 1; |
| 1206 | } |
| 1207 | if (fstat(fd, &st) < 0) { |
| 1208 | perror(PNAME ": fstat"); |
| 1209 | return 1; |
| 1210 | } |
| 1211 | totalsize = st.st_size; |
| 1212 | mappedfile = mmap(NULL, totalsize, PROT_READ, MAP_SHARED, fd, 0); |
| 1213 | if (!mappedfile) { |
| 1214 | perror(PNAME ": mmap"); |
| 1215 | return 1; |
| 1216 | } |
| 1217 | pathsep = trie_pathsep(mappedfile); |
| 1218 | |
| 1219 | /* |
| 1220 | * Trim trailing slash, just in case. |
| 1221 | */ |
| 1222 | pathlen = strlen(querydir); |
| 1223 | if (pathlen > 0 && querydir[pathlen-1] == pathsep) |
| 1224 | querydir[--pathlen] = '\0'; |
| 1225 | |
| 1226 | text_query(mappedfile, querydir, textcutoff, tqdepth); |
| 1227 | |
| 1228 | munmap(mappedfile, totalsize); |
| 1229 | } else if (mode == HTML) { |
| 1230 | char *querydir = actions[action].arg; |
| 1231 | size_t pathlen, maxpathlen; |
| 1232 | char *pathbuf; |
| 1233 | struct html_config cfg; |
| 1234 | unsigned long xi; |
| 1235 | char *html; |
| 1236 | |
| 1237 | fd = open(filename, O_RDONLY); |
| 1238 | if (fd < 0) { |
| 1239 | fprintf(stderr, "%s: %s: open: %s\n", PNAME, filename, |
| 1240 | strerror(errno)); |
| 1241 | return 1; |
| 1242 | } |
| 1243 | if (fstat(fd, &st) < 0) { |
| 1244 | perror(PNAME ": fstat"); |
| 1245 | return 1; |
| 1246 | } |
| 1247 | totalsize = st.st_size; |
| 1248 | mappedfile = mmap(NULL, totalsize, PROT_READ, MAP_SHARED, fd, 0); |
| 1249 | if (!mappedfile) { |
| 1250 | perror(PNAME ": mmap"); |
| 1251 | return 1; |
| 1252 | } |
| 1253 | pathsep = trie_pathsep(mappedfile); |
| 1254 | |
| 1255 | maxpathlen = trie_maxpathlen(mappedfile); |
| 1256 | pathbuf = snewn(maxpathlen, char); |
| 1257 | |
| 1258 | /* |
| 1259 | * Trim trailing slash, just in case. |
| 1260 | */ |
| 1261 | pathlen = strlen(querydir); |
| 1262 | if (pathlen > 0 && querydir[pathlen-1] == pathsep) |
| 1263 | querydir[--pathlen] = '\0'; |
| 1264 | |
| 1265 | xi = trie_before(mappedfile, querydir); |
| 1266 | if (xi >= trie_count(mappedfile) || |
| 1267 | (trie_getpath(mappedfile, xi, pathbuf), |
| 1268 | strcmp(pathbuf, querydir))) { |
| 1269 | fprintf(stderr, "%s: pathname '%s' does not exist in index\n" |
| 1270 | "%*s(check it is spelled exactly as it is in the " |
| 1271 | "index, including\n%*sany leading './')\n", |
| 1272 | PNAME, querydir, |
| 1273 | (int)(1+sizeof(PNAME)), "", |
| 1274 | (int)(1+sizeof(PNAME)), ""); |
| 1275 | } else if (!index_has_root(mappedfile, xi)) { |
| 1276 | fprintf(stderr, "%s: pathname '%s' is" |
| 1277 | " a file, not a directory\n", PNAME, querydir); |
| 1278 | } else { |
| 1279 | cfg.format = NULL; |
| 1280 | cfg.autoage = htmlautoagerange; |
| 1281 | cfg.oldest = htmloldest; |
| 1282 | cfg.newest = htmlnewest; |
| 1283 | html = html_query(mappedfile, xi, &cfg); |
| 1284 | fputs(html, stdout); |
| 1285 | } |
| 1286 | |
| 1287 | munmap(mappedfile, totalsize); |
| 1288 | sfree(pathbuf); |
| 1289 | } else if (mode == DUMP) { |
| 1290 | size_t maxpathlen; |
| 1291 | char *buf; |
| 1292 | |
| 1293 | fd = open(filename, O_RDONLY); |
| 1294 | if (fd < 0) { |
| 1295 | fprintf(stderr, "%s: %s: open: %s\n", PNAME, filename, |
| 1296 | strerror(errno)); |
| 1297 | return 1; |
| 1298 | } |
| 1299 | if (fstat(fd, &st) < 0) { |
| 1300 | perror(PNAME ": fstat"); |
| 1301 | return 1; |
| 1302 | } |
| 1303 | totalsize = st.st_size; |
| 1304 | mappedfile = mmap(NULL, totalsize, PROT_READ, MAP_SHARED, fd, 0); |
| 1305 | if (!mappedfile) { |
| 1306 | perror(PNAME ": mmap"); |
| 1307 | return 1; |
| 1308 | } |
| 1309 | pathsep = trie_pathsep(mappedfile); |
| 1310 | |
| 1311 | maxpathlen = trie_maxpathlen(mappedfile); |
| 1312 | buf = snewn(maxpathlen, char); |
| 1313 | |
| 1314 | printf(DUMPHDR "%02x\n", (unsigned char)pathsep); |
| 1315 | tw = triewalk_new(mappedfile); |
| 1316 | while ((tf = triewalk_next(tw, buf)) != NULL) |
| 1317 | dump_line(buf, tf); |
| 1318 | triewalk_free(tw); |
| 1319 | |
| 1320 | munmap(mappedfile, totalsize); |
| 1321 | } else if (mode == HTTPD) { |
| 1322 | struct html_config pcfg; |
| 1323 | struct httpd_config dcfg; |
| 1324 | |
| 1325 | fd = open(filename, O_RDONLY); |
| 1326 | if (fd < 0) { |
| 1327 | fprintf(stderr, "%s: %s: open: %s\n", PNAME, filename, |
| 1328 | strerror(errno)); |
| 1329 | return 1; |
| 1330 | } |
| 1331 | if (fstat(fd, &st) < 0) { |
| 1332 | perror(PNAME ": fstat"); |
| 1333 | return 1; |
| 1334 | } |
| 1335 | totalsize = st.st_size; |
| 1336 | mappedfile = mmap(NULL, totalsize, PROT_READ, MAP_SHARED, fd, 0); |
| 1337 | if (!mappedfile) { |
| 1338 | perror(PNAME ": mmap"); |
| 1339 | return 1; |
| 1340 | } |
| 1341 | pathsep = trie_pathsep(mappedfile); |
| 1342 | |
| 1343 | dcfg.address = httpserveraddr; |
| 1344 | dcfg.port = httpserverport; |
| 1345 | dcfg.basicauthdata = httpauthdata; |
| 1346 | pcfg.format = NULL; |
| 1347 | pcfg.autoage = htmlautoagerange; |
| 1348 | pcfg.oldest = htmloldest; |
| 1349 | pcfg.newest = htmlnewest; |
| 1350 | run_httpd(mappedfile, auth, &dcfg, &pcfg); |
| 1351 | munmap(mappedfile, totalsize); |
| 1352 | } else if (mode == REMOVE) { |
| 1353 | if (remove(filename) < 0) { |
| 1354 | fprintf(stderr, "%s: %s: remove: %s\n", PNAME, filename, |
| 1355 | strerror(errno)); |
| 1356 | return 1; |
| 1357 | } |
| 1358 | } |
| 1359 | } |
| 1360 | |
| 1361 | return 0; |
| 1362 | } |