| 1 | /* |
| 2 | Copyright (C) Andrew Tridgell 1996 |
| 3 | Copyright (C) Paul Mackerras 1996 |
| 4 | |
| 5 | This program is free software; you can redistribute it and/or modify |
| 6 | it under the terms of the GNU General Public License as published by |
| 7 | the Free Software Foundation; either version 2 of the License, or |
| 8 | (at your option) any later version. |
| 9 | |
| 10 | This program is distributed in the hope that it will be useful, |
| 11 | but WITHOUT ANY WARRANTY; without even the implied warranty of |
| 12 | MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the |
| 13 | GNU General Public License for more details. |
| 14 | |
| 15 | You should have received a copy of the GNU General Public License |
| 16 | along with this program; if not, write to the Free Software |
| 17 | Foundation, Inc., 675 Mass Ave, Cambridge, MA 02139, USA. |
| 18 | */ |
| 19 | |
| 20 | #include "rsync.h" |
| 21 | #include "zlib/zlib.h" |
| 22 | |
| 23 | extern int do_compression; |
| 24 | extern int module_id; |
| 25 | extern int write_batch; |
| 26 | |
| 27 | static int compression_level = Z_DEFAULT_COMPRESSION; |
| 28 | |
| 29 | /* determine the compression level based on a wildcard filename list */ |
| 30 | void set_compression(char *fname) |
| 31 | { |
| 32 | char *dont; |
| 33 | char *tok; |
| 34 | |
| 35 | if (!do_compression) |
| 36 | return; |
| 37 | |
| 38 | compression_level = Z_DEFAULT_COMPRESSION; |
| 39 | dont = lp_dont_compress(module_id); |
| 40 | |
| 41 | if (!dont || !*dont) |
| 42 | return; |
| 43 | |
| 44 | if (dont[0] == '*' && !dont[1]) { |
| 45 | /* an optimization to skip the rest of this routine */ |
| 46 | compression_level = 0; |
| 47 | return; |
| 48 | } |
| 49 | |
| 50 | dont = strdup(dont); |
| 51 | fname = strdup(fname); |
| 52 | if (!dont || !fname) |
| 53 | return; |
| 54 | |
| 55 | strlower(dont); |
| 56 | strlower(fname); |
| 57 | |
| 58 | for (tok = strtok(dont, " "); tok; tok = strtok(NULL, " ")) { |
| 59 | if (wildmatch(tok, fname)) { |
| 60 | compression_level = 0; |
| 61 | break; |
| 62 | } |
| 63 | } |
| 64 | free(dont); |
| 65 | free(fname); |
| 66 | } |
| 67 | |
| 68 | /* non-compressing recv token */ |
| 69 | static int simple_recv_token(int f,char **data) |
| 70 | { |
| 71 | static int residue; |
| 72 | static char *buf; |
| 73 | int n; |
| 74 | |
| 75 | if (!buf) { |
| 76 | buf = new_array(char, CHUNK_SIZE); |
| 77 | if (!buf) |
| 78 | out_of_memory("simple_recv_token"); |
| 79 | } |
| 80 | |
| 81 | if (residue == 0) { |
| 82 | int i = read_int(f); |
| 83 | if (i <= 0) |
| 84 | return i; |
| 85 | residue = i; |
| 86 | } |
| 87 | |
| 88 | *data = buf; |
| 89 | n = MIN(CHUNK_SIZE,residue); |
| 90 | residue -= n; |
| 91 | read_buf(f,buf,n); |
| 92 | return n; |
| 93 | } |
| 94 | |
| 95 | |
| 96 | /* non-compressing send token */ |
| 97 | static void simple_send_token(int f,int token, |
| 98 | struct map_struct *buf,OFF_T offset,int n) |
| 99 | { |
| 100 | int hold_int; |
| 101 | |
| 102 | if (n > 0) { |
| 103 | int l = 0; |
| 104 | while (l < n) { |
| 105 | int n1 = MIN(CHUNK_SIZE,n-l); |
| 106 | write_int(f,n1); |
| 107 | write_buf(f,map_ptr(buf,offset+l,n1),n1); |
| 108 | if (write_batch) { |
| 109 | write_batch_delta_file( (char *) &n1, sizeof(int) ); |
| 110 | write_batch_delta_file(map_ptr(buf,offset+l,n1),n1); |
| 111 | } |
| 112 | l += n1; |
| 113 | } |
| 114 | } |
| 115 | /* a -2 token means to send data only and no token */ |
| 116 | if (token != -2) { |
| 117 | write_int(f,-(token+1)); |
| 118 | if (write_batch) { |
| 119 | hold_int = -(token+1); |
| 120 | write_batch_delta_file( (char *) &hold_int, sizeof(int) ); |
| 121 | } |
| 122 | } |
| 123 | } |
| 124 | |
| 125 | |
| 126 | /* Flag bytes in compressed stream are encoded as follows: */ |
| 127 | #define END_FLAG 0 /* that's all folks */ |
| 128 | #define TOKEN_LONG 0x20 /* followed by 32-bit token number */ |
| 129 | #define TOKENRUN_LONG 0x21 /* ditto with 16-bit run count */ |
| 130 | #define DEFLATED_DATA 0x40 /* + 6-bit high len, then low len byte */ |
| 131 | #define TOKEN_REL 0x80 /* + 6-bit relative token number */ |
| 132 | #define TOKENRUN_REL 0xc0 /* ditto with 16-bit run count */ |
| 133 | |
| 134 | #define MAX_DATA_COUNT 16383 /* fit 14 bit count into 2 bytes with flags */ |
| 135 | |
| 136 | /* zlib.h says that if we want to be able to compress something in a single |
| 137 | * call, avail_out must be at least 0.1% larger than avail_in plus 12 bytes. |
| 138 | * We'll add in 0.1%+16, just to be safe (and we'll avoid floating point, |
| 139 | * to ensure that this is a compile-time value). */ |
| 140 | #define AVAIL_OUT_SIZE(avail_in_size) ((avail_in_size)*1001/1000+16) |
| 141 | |
| 142 | /* For coding runs of tokens */ |
| 143 | static int last_token = -1; |
| 144 | static int run_start; |
| 145 | static int last_run_end; |
| 146 | |
| 147 | /* Deflation state */ |
| 148 | static z_stream tx_strm; |
| 149 | |
| 150 | /* Output buffer */ |
| 151 | static char *obuf; |
| 152 | |
| 153 | /* We want obuf to be able to hold both MAX_DATA_COUNT+2 bytes as well as |
| 154 | * AVAIL_OUT_SIZE(CHUNK_SIZE) bytes, so make sure that it's large enough. */ |
| 155 | #if MAX_DATA_COUNT+2 > AVAIL_OUT_SIZE(CHUNK_SIZE) |
| 156 | #define OBUF_SIZE (MAX_DATA_COUNT+2) |
| 157 | #else |
| 158 | #define OBUF_SIZE AVAIL_OUT_SIZE(CHUNK_SIZE) |
| 159 | #endif |
| 160 | |
| 161 | /* Send a deflated token */ |
| 162 | static void |
| 163 | send_deflated_token(int f, int token, |
| 164 | struct map_struct *buf, OFF_T offset, int nb, int toklen) |
| 165 | { |
| 166 | int n, r; |
| 167 | static int init_done, flush_pending; |
| 168 | char temp_byte; |
| 169 | |
| 170 | if (last_token == -1) { |
| 171 | /* initialization */ |
| 172 | if (!init_done) { |
| 173 | tx_strm.next_in = NULL; |
| 174 | tx_strm.zalloc = NULL; |
| 175 | tx_strm.zfree = NULL; |
| 176 | if (deflateInit2(&tx_strm, compression_level, |
| 177 | Z_DEFLATED, -15, 8, |
| 178 | Z_DEFAULT_STRATEGY) != Z_OK) { |
| 179 | rprintf(FERROR, "compression init failed\n"); |
| 180 | exit_cleanup(RERR_STREAMIO); |
| 181 | } |
| 182 | if ((obuf = new_array(char, OBUF_SIZE)) == NULL) |
| 183 | out_of_memory("send_deflated_token"); |
| 184 | init_done = 1; |
| 185 | } else |
| 186 | deflateReset(&tx_strm); |
| 187 | last_run_end = 0; |
| 188 | run_start = token; |
| 189 | flush_pending = 0; |
| 190 | |
| 191 | } else if (last_token == -2) { |
| 192 | run_start = token; |
| 193 | |
| 194 | } else if (nb != 0 || token != last_token + 1 |
| 195 | || token >= run_start + 65536) { |
| 196 | /* output previous run */ |
| 197 | r = run_start - last_run_end; |
| 198 | n = last_token - run_start; |
| 199 | if (r >= 0 && r <= 63) { |
| 200 | write_byte(f, (n==0? TOKEN_REL: TOKENRUN_REL) + r); |
| 201 | if (write_batch) { |
| 202 | temp_byte = (char)( (n==0? TOKEN_REL: TOKENRUN_REL) + r); |
| 203 | write_batch_delta_file(&temp_byte,sizeof(char)); |
| 204 | } |
| 205 | } else { |
| 206 | write_byte(f, (n==0? TOKEN_LONG: TOKENRUN_LONG)); |
| 207 | write_int(f, run_start); |
| 208 | if (write_batch) { |
| 209 | temp_byte = (char)(n==0? TOKEN_LONG: TOKENRUN_LONG); |
| 210 | write_batch_delta_file(&temp_byte,sizeof(char)); |
| 211 | write_batch_delta_file((char *)&run_start,sizeof(run_start)); |
| 212 | } |
| 213 | } |
| 214 | if (n != 0) { |
| 215 | write_byte(f, n); |
| 216 | write_byte(f, n >> 8); |
| 217 | if (write_batch) { |
| 218 | temp_byte = (char)n; |
| 219 | write_batch_delta_file(&temp_byte,sizeof(char)); |
| 220 | temp_byte = (char)(n >> 8); |
| 221 | write_batch_delta_file(&temp_byte,sizeof(char)); |
| 222 | } |
| 223 | } |
| 224 | last_run_end = last_token; |
| 225 | run_start = token; |
| 226 | } |
| 227 | |
| 228 | last_token = token; |
| 229 | |
| 230 | if (nb != 0 || flush_pending) { |
| 231 | /* deflate the data starting at offset */ |
| 232 | int flush = Z_NO_FLUSH; |
| 233 | tx_strm.avail_in = 0; |
| 234 | tx_strm.avail_out = 0; |
| 235 | do { |
| 236 | if (tx_strm.avail_in == 0 && nb != 0) { |
| 237 | /* give it some more input */ |
| 238 | n = MIN(nb, CHUNK_SIZE); |
| 239 | tx_strm.next_in = (Bytef *) |
| 240 | map_ptr(buf, offset, n); |
| 241 | tx_strm.avail_in = n; |
| 242 | nb -= n; |
| 243 | offset += n; |
| 244 | } |
| 245 | if (tx_strm.avail_out == 0) { |
| 246 | tx_strm.next_out = (Bytef *)(obuf + 2); |
| 247 | tx_strm.avail_out = MAX_DATA_COUNT; |
| 248 | if (flush != Z_NO_FLUSH) { |
| 249 | /* |
| 250 | * We left the last 4 bytes in the |
| 251 | * buffer, in case they are the |
| 252 | * last 4. Move them to the front. |
| 253 | */ |
| 254 | memcpy(tx_strm.next_out, |
| 255 | obuf+MAX_DATA_COUNT-2, 4); |
| 256 | tx_strm.next_out += 4; |
| 257 | tx_strm.avail_out -= 4; |
| 258 | } |
| 259 | } |
| 260 | if (nb == 0 && token != -2) |
| 261 | flush = Z_SYNC_FLUSH; |
| 262 | r = deflate(&tx_strm, flush); |
| 263 | if (r != Z_OK) { |
| 264 | rprintf(FERROR, "deflate returned %d\n", r); |
| 265 | exit_cleanup(RERR_STREAMIO); |
| 266 | } |
| 267 | if (nb == 0 || tx_strm.avail_out == 0) { |
| 268 | n = MAX_DATA_COUNT - tx_strm.avail_out; |
| 269 | if (flush != Z_NO_FLUSH) { |
| 270 | /* |
| 271 | * We have to trim off the last 4 |
| 272 | * bytes of output when flushing |
| 273 | * (they are just 0, 0, ff, ff). |
| 274 | */ |
| 275 | n -= 4; |
| 276 | } |
| 277 | if (n > 0) { |
| 278 | obuf[0] = DEFLATED_DATA + (n >> 8); |
| 279 | obuf[1] = n; |
| 280 | write_buf(f, obuf, n+2); |
| 281 | if (write_batch) |
| 282 | write_batch_delta_file(obuf,n+2); |
| 283 | } |
| 284 | } |
| 285 | } while (nb != 0 || tx_strm.avail_out == 0); |
| 286 | flush_pending = token == -2; |
| 287 | } |
| 288 | |
| 289 | if (token == -1) { |
| 290 | /* end of file - clean up */ |
| 291 | write_byte(f, END_FLAG); |
| 292 | if (write_batch) { |
| 293 | temp_byte = END_FLAG; |
| 294 | write_batch_delta_file(&temp_byte,sizeof(char)); |
| 295 | } |
| 296 | |
| 297 | } else if (token != -2) { |
| 298 | /* add the data in the current block to the compressor's |
| 299 | history and hash table */ |
| 300 | tx_strm.next_in = (Bytef *) map_ptr(buf, offset, toklen); |
| 301 | tx_strm.avail_in = toklen; |
| 302 | tx_strm.next_out = (Bytef *) obuf; |
| 303 | tx_strm.avail_out = AVAIL_OUT_SIZE(CHUNK_SIZE); |
| 304 | r = deflate(&tx_strm, Z_INSERT_ONLY); |
| 305 | if (r != Z_OK || tx_strm.avail_in != 0) { |
| 306 | rprintf(FERROR, "deflate on token returned %d (%d bytes left)\n", |
| 307 | r, tx_strm.avail_in); |
| 308 | exit_cleanup(RERR_STREAMIO); |
| 309 | } |
| 310 | } |
| 311 | } |
| 312 | |
| 313 | |
| 314 | /* tells us what the receiver is in the middle of doing */ |
| 315 | static enum { r_init, r_idle, r_running, r_inflating, r_inflated } recv_state; |
| 316 | |
| 317 | /* for inflating stuff */ |
| 318 | static z_stream rx_strm; |
| 319 | static char *cbuf; |
| 320 | static char *dbuf; |
| 321 | |
| 322 | /* for decoding runs of tokens */ |
| 323 | static int rx_token; |
| 324 | static int rx_run; |
| 325 | |
| 326 | /* Receive a deflated token and inflate it */ |
| 327 | static int |
| 328 | recv_deflated_token(int f, char **data) |
| 329 | { |
| 330 | int n, r, flag; |
| 331 | static int init_done; |
| 332 | static int saved_flag; |
| 333 | |
| 334 | for (;;) { |
| 335 | switch (recv_state) { |
| 336 | case r_init: |
| 337 | if (!init_done) { |
| 338 | rx_strm.next_out = NULL; |
| 339 | rx_strm.zalloc = NULL; |
| 340 | rx_strm.zfree = NULL; |
| 341 | if (inflateInit2(&rx_strm, -15) != Z_OK) { |
| 342 | rprintf(FERROR, "inflate init failed\n"); |
| 343 | exit_cleanup(RERR_STREAMIO); |
| 344 | } |
| 345 | if (!(cbuf = new_array(char, MAX_DATA_COUNT)) |
| 346 | || !(dbuf = new_array(char, AVAIL_OUT_SIZE(CHUNK_SIZE)))) |
| 347 | out_of_memory("recv_deflated_token"); |
| 348 | init_done = 1; |
| 349 | } else { |
| 350 | inflateReset(&rx_strm); |
| 351 | } |
| 352 | recv_state = r_idle; |
| 353 | rx_token = 0; |
| 354 | break; |
| 355 | |
| 356 | case r_idle: |
| 357 | case r_inflated: |
| 358 | if (saved_flag) { |
| 359 | flag = saved_flag & 0xff; |
| 360 | saved_flag = 0; |
| 361 | } else |
| 362 | flag = read_byte(f); |
| 363 | if ((flag & 0xC0) == DEFLATED_DATA) { |
| 364 | n = ((flag & 0x3f) << 8) + read_byte(f); |
| 365 | read_buf(f, cbuf, n); |
| 366 | rx_strm.next_in = (Bytef *)cbuf; |
| 367 | rx_strm.avail_in = n; |
| 368 | recv_state = r_inflating; |
| 369 | break; |
| 370 | } |
| 371 | if (recv_state == r_inflated) { |
| 372 | /* check previous inflated stuff ended correctly */ |
| 373 | rx_strm.avail_in = 0; |
| 374 | rx_strm.next_out = (Bytef *)dbuf; |
| 375 | rx_strm.avail_out = AVAIL_OUT_SIZE(CHUNK_SIZE); |
| 376 | r = inflate(&rx_strm, Z_SYNC_FLUSH); |
| 377 | n = AVAIL_OUT_SIZE(CHUNK_SIZE) - rx_strm.avail_out; |
| 378 | /* |
| 379 | * Z_BUF_ERROR just means no progress was |
| 380 | * made, i.e. the decompressor didn't have |
| 381 | * any pending output for us. |
| 382 | */ |
| 383 | if (r != Z_OK && r != Z_BUF_ERROR) { |
| 384 | rprintf(FERROR, "inflate flush returned %d (%d bytes)\n", |
| 385 | r, n); |
| 386 | exit_cleanup(RERR_STREAMIO); |
| 387 | } |
| 388 | if (n != 0 && r != Z_BUF_ERROR) { |
| 389 | /* have to return some more data and |
| 390 | save the flag for later. */ |
| 391 | saved_flag = flag + 0x10000; |
| 392 | *data = dbuf; |
| 393 | return n; |
| 394 | } |
| 395 | /* |
| 396 | * At this point the decompressor should |
| 397 | * be expecting to see the 0, 0, ff, ff bytes. |
| 398 | */ |
| 399 | if (!inflateSyncPoint(&rx_strm)) { |
| 400 | rprintf(FERROR, "decompressor lost sync!\n"); |
| 401 | exit_cleanup(RERR_STREAMIO); |
| 402 | } |
| 403 | rx_strm.avail_in = 4; |
| 404 | rx_strm.next_in = (Bytef *)cbuf; |
| 405 | cbuf[0] = cbuf[1] = 0; |
| 406 | cbuf[2] = cbuf[3] = 0xff; |
| 407 | inflate(&rx_strm, Z_SYNC_FLUSH); |
| 408 | recv_state = r_idle; |
| 409 | } |
| 410 | if (flag == END_FLAG) { |
| 411 | /* that's all folks */ |
| 412 | recv_state = r_init; |
| 413 | return 0; |
| 414 | } |
| 415 | |
| 416 | /* here we have a token of some kind */ |
| 417 | if (flag & TOKEN_REL) { |
| 418 | rx_token += flag & 0x3f; |
| 419 | flag >>= 6; |
| 420 | } else |
| 421 | rx_token = read_int(f); |
| 422 | if (flag & 1) { |
| 423 | rx_run = read_byte(f); |
| 424 | rx_run += read_byte(f) << 8; |
| 425 | recv_state = r_running; |
| 426 | } |
| 427 | return -1 - rx_token; |
| 428 | |
| 429 | case r_inflating: |
| 430 | rx_strm.next_out = (Bytef *)dbuf; |
| 431 | rx_strm.avail_out = AVAIL_OUT_SIZE(CHUNK_SIZE); |
| 432 | r = inflate(&rx_strm, Z_NO_FLUSH); |
| 433 | n = AVAIL_OUT_SIZE(CHUNK_SIZE) - rx_strm.avail_out; |
| 434 | if (r != Z_OK) { |
| 435 | rprintf(FERROR, "inflate returned %d (%d bytes)\n", r, n); |
| 436 | exit_cleanup(RERR_STREAMIO); |
| 437 | } |
| 438 | if (rx_strm.avail_in == 0) |
| 439 | recv_state = r_inflated; |
| 440 | if (n != 0) { |
| 441 | *data = dbuf; |
| 442 | return n; |
| 443 | } |
| 444 | break; |
| 445 | |
| 446 | case r_running: |
| 447 | ++rx_token; |
| 448 | if (--rx_run == 0) |
| 449 | recv_state = r_idle; |
| 450 | return -1 - rx_token; |
| 451 | } |
| 452 | } |
| 453 | } |
| 454 | |
| 455 | /* |
| 456 | * put the data corresponding to a token that we've just returned |
| 457 | * from recv_deflated_token into the decompressor's history buffer. |
| 458 | */ |
| 459 | static void see_deflate_token(char *buf, int len) |
| 460 | { |
| 461 | int r, blklen; |
| 462 | unsigned char hdr[5]; |
| 463 | |
| 464 | rx_strm.avail_in = 0; |
| 465 | blklen = 0; |
| 466 | hdr[0] = 0; |
| 467 | do { |
| 468 | if (rx_strm.avail_in == 0 && len != 0) { |
| 469 | if (blklen == 0) { |
| 470 | /* Give it a fake stored-block header. */ |
| 471 | rx_strm.next_in = (Bytef *)hdr; |
| 472 | rx_strm.avail_in = 5; |
| 473 | blklen = len; |
| 474 | if (blklen > 0xffff) |
| 475 | blklen = 0xffff; |
| 476 | hdr[1] = blklen; |
| 477 | hdr[2] = blklen >> 8; |
| 478 | hdr[3] = ~hdr[1]; |
| 479 | hdr[4] = ~hdr[2]; |
| 480 | } else { |
| 481 | rx_strm.next_in = (Bytef *)buf; |
| 482 | rx_strm.avail_in = blklen; |
| 483 | len -= blklen; |
| 484 | blklen = 0; |
| 485 | } |
| 486 | } |
| 487 | rx_strm.next_out = (Bytef *)dbuf; |
| 488 | rx_strm.avail_out = AVAIL_OUT_SIZE(CHUNK_SIZE); |
| 489 | r = inflate(&rx_strm, Z_SYNC_FLUSH); |
| 490 | if (r != Z_OK) { |
| 491 | rprintf(FERROR, "inflate (token) returned %d\n", r); |
| 492 | exit_cleanup(RERR_STREAMIO); |
| 493 | } |
| 494 | } while (len || rx_strm.avail_out == 0); |
| 495 | } |
| 496 | |
| 497 | /** |
| 498 | * Transmit a verbatim buffer of length @p n followed by a token. |
| 499 | * If token == -1 then we have reached EOF |
| 500 | * If n == 0 then don't send a buffer |
| 501 | */ |
| 502 | void send_token(int f,int token,struct map_struct *buf,OFF_T offset, |
| 503 | int n,int toklen) |
| 504 | { |
| 505 | if (!do_compression) { |
| 506 | simple_send_token(f,token,buf,offset,n); |
| 507 | } else { |
| 508 | send_deflated_token(f, token, buf, offset, n, toklen); |
| 509 | } |
| 510 | } |
| 511 | |
| 512 | |
| 513 | /* |
| 514 | * receive a token or buffer from the other end. If the reurn value is >0 then |
| 515 | * it is a data buffer of that length, and *data will point at the data. |
| 516 | * if the return value is -i then it represents token i-1 |
| 517 | * if the return value is 0 then the end has been reached |
| 518 | */ |
| 519 | int recv_token(int f,char **data) |
| 520 | { |
| 521 | int tok; |
| 522 | |
| 523 | if (!do_compression) { |
| 524 | tok = simple_recv_token(f,data); |
| 525 | } else { |
| 526 | tok = recv_deflated_token(f, data); |
| 527 | } |
| 528 | return tok; |
| 529 | } |
| 530 | |
| 531 | /* |
| 532 | * look at the data corresponding to a token, if necessary |
| 533 | */ |
| 534 | void see_token(char *data, int toklen) |
| 535 | { |
| 536 | if (do_compression) |
| 537 | see_deflate_token(data, toklen); |
| 538 | } |