1 /*- 2 * Copyright (c) 2007 Joerg Sonnenberger 3 * Copyright (c) 2012 Michihiro NAKAJIMA 4 * All rights reserved. 5 * 6 * Redistribution and use in source and binary forms, with or without 7 * modification, are permitted provided that the following conditions 8 * are met: 9 * 1. Redistributions of source code must retain the above copyright 10 * notice, this list of conditions and the following disclaimer. 11 * 2. Redistributions in binary form must reproduce the above copyright 12 * notice, this list of conditions and the following disclaimer in the 13 * documentation and/or other materials provided with the distribution. 14 * 15 * THIS SOFTWARE IS PROVIDED BY THE AUTHOR(S) ``AS IS'' AND ANY EXPRESS OR 16 * IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES 17 * OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE DISCLAIMED. 18 * IN NO EVENT SHALL THE AUTHOR(S) BE LIABLE FOR ANY DIRECT, INDIRECT, 19 * INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT 20 * NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, 21 * DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY 22 * THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT 23 * (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF 24 * THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE. 25 */ 26 27 #include "archive_platform.h" 28 29 #ifdef HAVE_SYS_WAIT_H 30 # include <sys/wait.h> 31 #endif 32 #ifdef HAVE_ERRNO_H 33 # include <errno.h> 34 #endif 35 #ifdef HAVE_FCNTL_H 36 # include <fcntl.h> 37 #endif 38 #ifdef HAVE_LIMITS_H 39 # include <limits.h> 40 #endif 41 #ifdef HAVE_SIGNAL_H 42 # include <signal.h> 43 #endif 44 #ifdef HAVE_STDLIB_H 45 # include <stdlib.h> 46 #endif 47 #ifdef HAVE_STRING_H 48 # include <string.h> 49 #endif 50 #ifdef HAVE_UNISTD_H 51 # include <unistd.h> 52 #endif 53 54 #include "archive.h" 55 #include "archive_private.h" 56 #include "archive_string.h" 57 #include "archive_read_private.h" 58 #include "filter_fork.h" 59 60 61 #if ARCHIVE_VERSION_NUMBER < 4000000 62 /* Deprecated; remove in libarchive 4.0 */ 63 int 64 archive_read_support_compression_program(struct archive *a, const char *cmd) 65 { 66 return archive_read_support_filter_program(a, cmd); 67 } 68 69 int 70 archive_read_support_compression_program_signature(struct archive *a, 71 const char *cmd, const void *signature, size_t signature_len) 72 { 73 return archive_read_support_filter_program_signature(a, 74 cmd, signature, signature_len); 75 } 76 #endif 77 78 int 79 archive_read_support_filter_program(struct archive *a, const char *cmd) 80 { 81 return (archive_read_support_filter_program_signature(a, cmd, NULL, 0)); 82 } 83 84 /* 85 * The bidder object stores the command and the signature to watch for. 86 * The 'inhibit' entry here is used to ensure that unchecked filters never 87 * bid twice in the same pipeline. 88 */ 89 struct program_bidder { 90 char *description; 91 char *cmd; 92 void *signature; 93 size_t signature_len; 94 int inhibit; 95 }; 96 97 static int program_bidder_bid(struct archive_read_filter_bidder *, 98 struct archive_read_filter *upstream); 99 static int program_bidder_init(struct archive_read_filter *); 100 static void program_bidder_free(struct archive_read_filter_bidder *); 101 102 /* 103 * The actual filter needs to track input and output data. 104 */ 105 struct program_filter { 106 struct archive_string description; 107 #if defined(_WIN32) && !defined(__CYGWIN__) 108 HANDLE child; 109 #else 110 pid_t child; 111 #endif 112 int exit_status; 113 pid_t waitpid_return; 114 int child_stdin, child_stdout; 115 116 char *out_buf; 117 size_t out_buf_len; 118 }; 119 120 static ssize_t program_filter_read(struct archive_read_filter *, 121 const void **); 122 static int program_filter_close(struct archive_read_filter *); 123 static void free_state(struct program_bidder *); 124 125 static const struct archive_read_filter_bidder_vtable 126 program_bidder_vtable = { 127 .bid = program_bidder_bid, 128 .init = program_bidder_init, 129 .free = program_bidder_free, 130 }; 131 132 int 133 archive_read_support_filter_program_signature(struct archive *_a, 134 const char *cmd, const void *signature, size_t signature_len) 135 { 136 struct archive_read *a = (struct archive_read *)_a; 137 struct program_bidder *state; 138 139 /* 140 * Allocate our private state. 141 */ 142 state = calloc(1, sizeof (*state)); 143 if (state == NULL) 144 goto memerr; 145 state->cmd = strdup(cmd); 146 if (state->cmd == NULL) 147 goto memerr; 148 149 if (signature != NULL && signature_len > 0) { 150 state->signature_len = signature_len; 151 state->signature = malloc(signature_len); 152 if (state->signature == NULL) 153 goto memerr; 154 memcpy(state->signature, signature, signature_len); 155 } 156 157 if (__archive_read_register_bidder(a, state, NULL, 158 &program_bidder_vtable) != ARCHIVE_OK) { 159 free_state(state); 160 return (ARCHIVE_FATAL); 161 } 162 return (ARCHIVE_OK); 163 164 memerr: 165 free_state(state); 166 archive_set_error(_a, ENOMEM, "Can't allocate memory"); 167 return (ARCHIVE_FATAL); 168 } 169 170 static void 171 program_bidder_free(struct archive_read_filter_bidder *self) 172 { 173 struct program_bidder *state = (struct program_bidder *)self->data; 174 175 free_state(state); 176 } 177 178 static void 179 free_state(struct program_bidder *state) 180 { 181 182 if (state) { 183 free(state->cmd); 184 free(state->signature); 185 free(state); 186 } 187 } 188 189 /* 190 * If we do have a signature, bid only if that matches. 191 * 192 * If there's no signature, we bid INT_MAX the first time 193 * we're called, then never bid again. 194 */ 195 static int 196 program_bidder_bid(struct archive_read_filter_bidder *self, 197 struct archive_read_filter *upstream) 198 { 199 struct program_bidder *state = self->data; 200 const char *p; 201 202 /* If we have a signature, use that to match. */ 203 if (state->signature_len > 0) { 204 p = __archive_read_filter_ahead(upstream, 205 state->signature_len, NULL); 206 if (p == NULL) 207 return (0); 208 /* No match, so don't bid. */ 209 if (memcmp(p, state->signature, state->signature_len) != 0) 210 return (0); 211 return ((int)state->signature_len * 8); 212 } 213 214 /* Otherwise, bid once and then never bid again. */ 215 if (state->inhibit) 216 return (0); 217 state->inhibit = 1; 218 return (INT_MAX); 219 } 220 221 /* 222 * Shut down the child, return ARCHIVE_OK if it exited normally. 223 * 224 * Note that the return value is sticky; if we're called again, 225 * we won't reap the child again, but we will return the same status 226 * (including error message if the child came to a bad end). 227 */ 228 static int 229 child_stop(struct archive_read_filter *self, struct program_filter *state) 230 { 231 /* Close our side of the I/O with the child. */ 232 if (state->child_stdin != -1) { 233 close(state->child_stdin); 234 state->child_stdin = -1; 235 } 236 if (state->child_stdout != -1) { 237 close(state->child_stdout); 238 state->child_stdout = -1; 239 } 240 241 if (state->child != 0) { 242 /* Reap the child. */ 243 do { 244 state->waitpid_return 245 = waitpid(state->child, &state->exit_status, 0); 246 } while (state->waitpid_return == -1 && errno == EINTR); 247 state->child = 0; 248 } 249 250 if (state->waitpid_return < 0) { 251 /* waitpid() failed? This is ugly. */ 252 archive_set_error(&self->archive->archive, ARCHIVE_ERRNO_MISC, 253 "Error closing child process"); 254 return (ARCHIVE_WARN); 255 } 256 257 #if !defined(_WIN32) || defined(__CYGWIN__) 258 if (WIFSIGNALED(state->exit_status)) { 259 #ifdef SIGPIPE 260 /* If the child died because we stopped reading before 261 * it was done, that's okay. Some archive formats 262 * have padding at the end that we routinely ignore. */ 263 /* The alternative to this would be to add a step 264 * before close(child_stdout) above to read from the 265 * child until the child has no more to write. */ 266 if (WTERMSIG(state->exit_status) == SIGPIPE) 267 return (ARCHIVE_OK); 268 #endif 269 archive_set_error(&self->archive->archive, ARCHIVE_ERRNO_MISC, 270 "Child process exited with signal %d", 271 WTERMSIG(state->exit_status)); 272 return (ARCHIVE_WARN); 273 } 274 #endif /* !_WIN32 || __CYGWIN__ */ 275 276 if (WIFEXITED(state->exit_status)) { 277 if (WEXITSTATUS(state->exit_status) == 0) 278 return (ARCHIVE_OK); 279 280 archive_set_error(&self->archive->archive, 281 ARCHIVE_ERRNO_MISC, 282 "Child process exited with status %d", 283 WEXITSTATUS(state->exit_status)); 284 return (ARCHIVE_WARN); 285 } 286 287 return (ARCHIVE_WARN); 288 } 289 290 /* 291 * Use select() to decide whether the child is ready for read or write. 292 */ 293 static ssize_t 294 child_read(struct archive_read_filter *self, char *buf, size_t buf_len) 295 { 296 struct program_filter *state = self->data; 297 ssize_t ret, requested, avail; 298 const char *p; 299 #if defined(_WIN32) && !defined(__CYGWIN__) 300 HANDLE handle = (HANDLE)_get_osfhandle(state->child_stdout); 301 #endif 302 303 requested = buf_len > SSIZE_MAX ? SSIZE_MAX : buf_len; 304 305 for (;;) { 306 do { 307 #if defined(_WIN32) && !defined(__CYGWIN__) 308 /* Avoid infinity wait. 309 * Note: If there is no data in the pipe, ReadFile() 310 * called in read() never returns and so we won't 311 * write remaining encoded data to the pipe. 312 * Note: This way may cause performance problem. 313 * we are looking forward to great code to resolve 314 * this. */ 315 DWORD pipe_avail = -1; 316 int cnt = 2; 317 318 while (PeekNamedPipe(handle, NULL, 0, NULL, 319 &pipe_avail, NULL) != 0 && pipe_avail == 0 && 320 cnt--) 321 Sleep(5); 322 if (pipe_avail == 0) { 323 ret = -1; 324 errno = EAGAIN; 325 break; 326 } 327 #endif 328 ret = read(state->child_stdout, buf, requested); 329 } while (ret == -1 && errno == EINTR); 330 331 if (ret > 0) 332 return (ret); 333 if (ret == 0 || (ret == -1 && errno == EPIPE)) 334 /* Child has closed its output; reap the child 335 * and return the status. */ 336 return (child_stop(self, state)); 337 if (ret == -1 && errno != EAGAIN) 338 return (-1); 339 340 if (state->child_stdin == -1) { 341 /* Block until child has some I/O ready. */ 342 __archive_check_child(state->child_stdin, 343 state->child_stdout); 344 continue; 345 } 346 347 /* Get some more data from upstream. */ 348 p = __archive_read_filter_ahead(self->upstream, 1, &avail); 349 if (p == NULL) { 350 close(state->child_stdin); 351 state->child_stdin = -1; 352 fcntl(state->child_stdout, F_SETFL, 0); 353 if (avail < 0) 354 return (avail); 355 continue; 356 } 357 358 do { 359 ret = write(state->child_stdin, p, avail); 360 } while (ret == -1 && errno == EINTR); 361 362 if (ret > 0) { 363 /* Consume whatever we managed to write. */ 364 __archive_read_filter_consume(self->upstream, ret); 365 } else if (ret == -1 && errno == EAGAIN) { 366 /* Block until child has some I/O ready. */ 367 __archive_check_child(state->child_stdin, 368 state->child_stdout); 369 } else { 370 /* Write failed. */ 371 close(state->child_stdin); 372 state->child_stdin = -1; 373 fcntl(state->child_stdout, F_SETFL, 0); 374 /* If it was a bad error, we're done; otherwise 375 * it was EPIPE or EOF, and we can still read 376 * from the child. */ 377 if (ret == -1 && errno != EPIPE) 378 return (-1); 379 } 380 } 381 } 382 383 static const struct archive_read_filter_vtable 384 program_reader_vtable = { 385 .read = program_filter_read, 386 .close = program_filter_close, 387 }; 388 389 int 390 __archive_read_program(struct archive_read_filter *self, const char *cmd) 391 { 392 struct program_filter *state; 393 static const size_t out_buf_len = 65536; 394 char *out_buf; 395 const char *prefix = "Program: "; 396 int ret; 397 size_t l; 398 399 l = strlen(prefix) + strlen(cmd) + 1; 400 state = calloc(1, sizeof(*state)); 401 out_buf = malloc(out_buf_len); 402 if (state == NULL || out_buf == NULL || 403 archive_string_ensure(&state->description, l) == NULL) { 404 archive_set_error(&self->archive->archive, ENOMEM, 405 "Can't allocate input data"); 406 if (state != NULL) { 407 archive_string_free(&state->description); 408 free(state); 409 } 410 free(out_buf); 411 return (ARCHIVE_FATAL); 412 } 413 archive_strcpy(&state->description, prefix); 414 archive_strcat(&state->description, cmd); 415 416 self->code = ARCHIVE_FILTER_PROGRAM; 417 self->name = state->description.s; 418 419 state->out_buf = out_buf; 420 state->out_buf_len = out_buf_len; 421 422 ret = __archive_create_child(cmd, &state->child_stdin, 423 &state->child_stdout, &state->child); 424 if (ret != ARCHIVE_OK) { 425 free(state->out_buf); 426 archive_string_free(&state->description); 427 free(state); 428 archive_set_error(&self->archive->archive, EINVAL, 429 "Can't initialize filter; unable to run program \"%s\"", 430 cmd); 431 return (ARCHIVE_FATAL); 432 } 433 434 self->data = state; 435 self->vtable = &program_reader_vtable; 436 437 /* XXX Check that we can read at least one byte? */ 438 return (ARCHIVE_OK); 439 } 440 441 static int 442 program_bidder_init(struct archive_read_filter *self) 443 { 444 struct program_bidder *bidder_state; 445 446 bidder_state = (struct program_bidder *)self->bidder->data; 447 return (__archive_read_program(self, bidder_state->cmd)); 448 } 449 450 static ssize_t 451 program_filter_read(struct archive_read_filter *self, const void **buff) 452 { 453 struct program_filter *state; 454 ssize_t bytes; 455 size_t total; 456 char *p; 457 458 state = (struct program_filter *)self->data; 459 460 total = 0; 461 p = state->out_buf; 462 while (state->child_stdout != -1 && total < state->out_buf_len) { 463 bytes = child_read(self, p, state->out_buf_len - total); 464 if (bytes < 0) 465 /* No recovery is possible if we can no longer 466 * read from the child. */ 467 return (ARCHIVE_FATAL); 468 if (bytes == 0) 469 /* We got EOF from the child. */ 470 break; 471 total += bytes; 472 p += bytes; 473 } 474 475 *buff = state->out_buf; 476 return (total); 477 } 478 479 static int 480 program_filter_close(struct archive_read_filter *self) 481 { 482 struct program_filter *state; 483 int e; 484 485 state = (struct program_filter *)self->data; 486 e = child_stop(self, state); 487 488 /* Release our private data. */ 489 free(state->out_buf); 490 archive_string_free(&state->description); 491 free(state); 492 493 return (e); 494 } 495
Indexes created Sun Aug 16 00:25:19 UTC 2026