Commit | Line | Data |
---|---|---|
d67c8bdf | 1 | /* |
70d794dc AT |
2 | Copyright (C) Andrew Tridgell 1996 |
3 | Copyright (C) Paul Mackerras 1996 | |
d67c8bdf | 4 | |
70d794dc AT |
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. | |
d67c8bdf | 9 | |
70d794dc AT |
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. | |
d67c8bdf | 14 | |
70d794dc AT |
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" | |
5914bf15 | 21 | #include "zlib/zlib.h" |
70d794dc AT |
22 | |
23 | extern int do_compression; | |
d67c8bdf | 24 | extern int module_id; |
e8a8167a | 25 | extern int def_compress_level; |
d67c8bdf | 26 | |
e8a8167a | 27 | static int compression_level; |
70d794dc | 28 | |
83fff1aa AT |
29 | /* determine the compression level based on a wildcard filename list */ |
30 | void set_compression(char *fname) | |
31 | { | |
83fff1aa AT |
32 | char *dont; |
33 | char *tok; | |
34 | ||
d67c8bdf WD |
35 | if (!do_compression) |
36 | return; | |
83fff1aa | 37 | |
e8a8167a | 38 | compression_level = def_compress_level; |
83fff1aa AT |
39 | dont = lp_dont_compress(module_id); |
40 | ||
d67c8bdf WD |
41 | if (!dont || !*dont) |
42 | return; | |
83fff1aa | 43 | |
d67c8bdf | 44 | if (dont[0] == '*' && !dont[1]) { |
63f0774f DD |
45 | /* an optimization to skip the rest of this routine */ |
46 | compression_level = 0; | |
47 | return; | |
48 | } | |
49 | ||
0fe987e2 WD |
50 | if ((tok = strrchr(fname, '/')) != NULL) |
51 | fname = tok + 1; | |
52 | ||
83fff1aa AT |
53 | dont = strdup(dont); |
54 | fname = strdup(fname); | |
d67c8bdf WD |
55 | if (!dont || !fname) |
56 | return; | |
83fff1aa AT |
57 | |
58 | strlower(dont); | |
59 | strlower(fname); | |
60 | ||
d67c8bdf | 61 | for (tok = strtok(dont, " "); tok; tok = strtok(NULL, " ")) { |
fe332038 | 62 | if (wildmatch(tok, fname)) { |
83fff1aa AT |
63 | compression_level = 0; |
64 | break; | |
65 | } | |
66 | } | |
67 | free(dont); | |
68 | free(fname); | |
69 | } | |
70d794dc AT |
70 | |
71 | /* non-compressing recv token */ | |
7fcbf9e4 | 72 | static int32 simple_recv_token(int f, char **data) |
70d794dc | 73 | { |
acc461c7 | 74 | static int32 residue; |
c5eb3650 | 75 | static char *buf; |
7fcbf9e4 | 76 | int32 n; |
70d794dc | 77 | |
c5eb3650 | 78 | if (!buf) { |
58cadc86 | 79 | buf = new_array(char, CHUNK_SIZE); |
d67c8bdf WD |
80 | if (!buf) |
81 | out_of_memory("simple_recv_token"); | |
c5eb3650 | 82 | } |
70d794dc | 83 | |
c5eb3650 | 84 | if (residue == 0) { |
7fcbf9e4 | 85 | int32 i = read_int(f); |
d67c8bdf WD |
86 | if (i <= 0) |
87 | return i; | |
c5eb3650 AT |
88 | residue = i; |
89 | } | |
70d794dc | 90 | |
c5eb3650 AT |
91 | *data = buf; |
92 | n = MIN(CHUNK_SIZE,residue); | |
93 | residue -= n; | |
94 | read_buf(f,buf,n); | |
95 | return n; | |
70d794dc AT |
96 | } |
97 | ||
98 | ||
99 | /* non-compressing send token */ | |
acc461c7 | 100 | static void simple_send_token(int f, int32 token, struct map_struct *buf, |
7fcbf9e4 | 101 | OFF_T offset, int32 n) |
70d794dc | 102 | { |
c5eb3650 | 103 | if (n > 0) { |
7fcbf9e4 WD |
104 | int32 len = 0; |
105 | while (len < n) { | |
106 | int32 n1 = MIN(CHUNK_SIZE, n-len); | |
107 | write_int(f, n1); | |
108 | write_buf(f, map_ptr(buf, offset+len, n1), n1); | |
109 | len += n1; | |
c5eb3650 AT |
110 | } |
111 | } | |
45f133b9 | 112 | /* a -2 token means to send data only and no token */ |
7fcbf9e4 WD |
113 | if (token != -2) |
114 | write_int(f, -(token+1)); | |
70d794dc AT |
115 | } |
116 | ||
117 | ||
861c20b4 PM |
118 | /* Flag bytes in compressed stream are encoded as follows: */ |
119 | #define END_FLAG 0 /* that's all folks */ | |
120 | #define TOKEN_LONG 0x20 /* followed by 32-bit token number */ | |
121 | #define TOKENRUN_LONG 0x21 /* ditto with 16-bit run count */ | |
122 | #define DEFLATED_DATA 0x40 /* + 6-bit high len, then low len byte */ | |
123 | #define TOKEN_REL 0x80 /* + 6-bit relative token number */ | |
124 | #define TOKENRUN_REL 0xc0 /* ditto with 16-bit run count */ | |
125 | ||
126 | #define MAX_DATA_COUNT 16383 /* fit 14 bit count into 2 bytes with flags */ | |
127 | ||
24733919 WD |
128 | /* zlib.h says that if we want to be able to compress something in a single |
129 | * call, avail_out must be at least 0.1% larger than avail_in plus 12 bytes. | |
130 | * We'll add in 0.1%+16, just to be safe (and we'll avoid floating point, | |
131 | * to ensure that this is a compile-time value). */ | |
132 | #define AVAIL_OUT_SIZE(avail_in_size) ((avail_in_size)*1001/1000+16) | |
133 | ||
861c20b4 | 134 | /* For coding runs of tokens */ |
acc461c7 WD |
135 | static int32 last_token = -1; |
136 | static int32 run_start; | |
137 | static int32 last_run_end; | |
861c20b4 PM |
138 | |
139 | /* Deflation state */ | |
140 | static z_stream tx_strm; | |
141 | ||
142 | /* Output buffer */ | |
3a6a366f | 143 | static char *obuf; |
24733919 WD |
144 | |
145 | /* We want obuf to be able to hold both MAX_DATA_COUNT+2 bytes as well as | |
146 | * AVAIL_OUT_SIZE(CHUNK_SIZE) bytes, so make sure that it's large enough. */ | |
147 | #if MAX_DATA_COUNT+2 > AVAIL_OUT_SIZE(CHUNK_SIZE) | |
148 | #define OBUF_SIZE (MAX_DATA_COUNT+2) | |
149 | #else | |
150 | #define OBUF_SIZE AVAIL_OUT_SIZE(CHUNK_SIZE) | |
151 | #endif | |
861c20b4 PM |
152 | |
153 | /* Send a deflated token */ | |
154 | static void | |
acc461c7 | 155 | send_deflated_token(int f, int32 token, struct map_struct *buf, OFF_T offset, |
7fcbf9e4 | 156 | int32 nb, int32 toklen) |
861c20b4 | 157 | { |
7fcbf9e4 | 158 | int32 n, r; |
5914bf15 PM |
159 | static int init_done, flush_pending; |
160 | ||
161 | if (last_token == -1) { | |
162 | /* initialization */ | |
163 | if (!init_done) { | |
164 | tx_strm.next_in = NULL; | |
165 | tx_strm.zalloc = NULL; | |
166 | tx_strm.zfree = NULL; | |
83fff1aa | 167 | if (deflateInit2(&tx_strm, compression_level, |
5914bf15 PM |
168 | Z_DEFLATED, -15, 8, |
169 | Z_DEFAULT_STRATEGY) != Z_OK) { | |
170 | rprintf(FERROR, "compression init failed\n"); | |
65417579 | 171 | exit_cleanup(RERR_STREAMIO); |
5914bf15 | 172 | } |
58cadc86 | 173 | if ((obuf = new_array(char, OBUF_SIZE)) == NULL) |
5914bf15 PM |
174 | out_of_memory("send_deflated_token"); |
175 | init_done = 1; | |
176 | } else | |
177 | deflateReset(&tx_strm); | |
178 | last_run_end = 0; | |
179 | run_start = token; | |
180 | flush_pending = 0; | |
181 | ||
182 | } else if (last_token == -2) { | |
183 | run_start = token; | |
184 | ||
185 | } else if (nb != 0 || token != last_token + 1 | |
186 | || token >= run_start + 65536) { | |
187 | /* output previous run */ | |
188 | r = run_start - last_run_end; | |
189 | n = last_token - run_start; | |
190 | if (r >= 0 && r <= 63) { | |
191 | write_byte(f, (n==0? TOKEN_REL: TOKENRUN_REL) + r); | |
192 | } else { | |
193 | write_byte(f, (n==0? TOKEN_LONG: TOKENRUN_LONG)); | |
194 | write_int(f, run_start); | |
195 | } | |
196 | if (n != 0) { | |
197 | write_byte(f, n); | |
198 | write_byte(f, n >> 8); | |
199 | } | |
200 | last_run_end = last_token; | |
201 | run_start = token; | |
861c20b4 | 202 | } |
5914bf15 PM |
203 | |
204 | last_token = token; | |
205 | ||
206 | if (nb != 0 || flush_pending) { | |
207 | /* deflate the data starting at offset */ | |
208 | int flush = Z_NO_FLUSH; | |
209 | tx_strm.avail_in = 0; | |
210 | tx_strm.avail_out = 0; | |
211 | do { | |
212 | if (tx_strm.avail_in == 0 && nb != 0) { | |
213 | /* give it some more input */ | |
214 | n = MIN(nb, CHUNK_SIZE); | |
215 | tx_strm.next_in = (Bytef *) | |
216 | map_ptr(buf, offset, n); | |
217 | tx_strm.avail_in = n; | |
218 | nb -= n; | |
219 | offset += n; | |
220 | } | |
221 | if (tx_strm.avail_out == 0) { | |
222 | tx_strm.next_out = (Bytef *)(obuf + 2); | |
223 | tx_strm.avail_out = MAX_DATA_COUNT; | |
224 | if (flush != Z_NO_FLUSH) { | |
225 | /* | |
226 | * We left the last 4 bytes in the | |
227 | * buffer, in case they are the | |
228 | * last 4. Move them to the front. | |
229 | */ | |
230 | memcpy(tx_strm.next_out, | |
231 | obuf+MAX_DATA_COUNT-2, 4); | |
232 | tx_strm.next_out += 4; | |
233 | tx_strm.avail_out -= 4; | |
234 | } | |
235 | } | |
236 | if (nb == 0 && token != -2) | |
237 | flush = Z_SYNC_FLUSH; | |
238 | r = deflate(&tx_strm, flush); | |
239 | if (r != Z_OK) { | |
240 | rprintf(FERROR, "deflate returned %d\n", r); | |
65417579 | 241 | exit_cleanup(RERR_STREAMIO); |
5914bf15 PM |
242 | } |
243 | if (nb == 0 || tx_strm.avail_out == 0) { | |
244 | n = MAX_DATA_COUNT - tx_strm.avail_out; | |
245 | if (flush != Z_NO_FLUSH) { | |
246 | /* | |
247 | * We have to trim off the last 4 | |
248 | * bytes of output when flushing | |
249 | * (they are just 0, 0, ff, ff). | |
250 | */ | |
251 | n -= 4; | |
252 | } | |
253 | if (n > 0) { | |
254 | obuf[0] = DEFLATED_DATA + (n >> 8); | |
255 | obuf[1] = n; | |
256 | write_buf(f, obuf, n+2); | |
257 | } | |
258 | } | |
259 | } while (nb != 0 || tx_strm.avail_out == 0); | |
260 | flush_pending = token == -2; | |
861c20b4 | 261 | } |
861c20b4 | 262 | |
5914bf15 PM |
263 | if (token == -1) { |
264 | /* end of file - clean up */ | |
265 | write_byte(f, END_FLAG); | |
5914bf15 | 266 | } else if (token != -2) { |
acc461c7 WD |
267 | /* Add the data in the current block to the compressor's |
268 | * history and hash table. */ | |
01f439ec WD |
269 | do { |
270 | /* Break up long sections in the same way that | |
271 | * see_deflate_token() does. */ | |
272 | int32 n1 = toklen > 0xffff ? 0xffff : toklen; | |
273 | toklen -= n1; | |
274 | tx_strm.next_in = (Bytef *)map_ptr(buf, offset, n1); | |
275 | tx_strm.avail_in = n1; | |
276 | tx_strm.next_out = (Bytef *) obuf; | |
277 | tx_strm.avail_out = AVAIL_OUT_SIZE(CHUNK_SIZE); | |
278 | r = deflate(&tx_strm, Z_INSERT_ONLY); | |
279 | if (r != Z_OK || tx_strm.avail_in != 0) { | |
280 | rprintf(FERROR, "deflate on token returned %d (%d bytes left)\n", | |
281 | r, tx_strm.avail_in); | |
282 | exit_cleanup(RERR_STREAMIO); | |
283 | } | |
284 | } while (toklen > 0); | |
861c20b4 | 285 | } |
861c20b4 PM |
286 | } |
287 | ||
288 | ||
289 | /* tells us what the receiver is in the middle of doing */ | |
290 | static enum { r_init, r_idle, r_running, r_inflating, r_inflated } recv_state; | |
70d794dc | 291 | |
861c20b4 PM |
292 | /* for inflating stuff */ |
293 | static z_stream rx_strm; | |
294 | static char *cbuf; | |
295 | static char *dbuf; | |
296 | ||
297 | /* for decoding runs of tokens */ | |
7fcbf9e4 WD |
298 | static int32 rx_token; |
299 | static int32 rx_run; | |
861c20b4 PM |
300 | |
301 | /* Receive a deflated token and inflate it */ | |
7fcbf9e4 | 302 | static int32 recv_deflated_token(int f, char **data) |
861c20b4 | 303 | { |
5914bf15 | 304 | static int init_done; |
acc461c7 WD |
305 | static int32 saved_flag; |
306 | int32 n, flag; | |
307 | int r; | |
5914bf15 PM |
308 | |
309 | for (;;) { | |
310 | switch (recv_state) { | |
311 | case r_init: | |
312 | if (!init_done) { | |
313 | rx_strm.next_out = NULL; | |
314 | rx_strm.zalloc = NULL; | |
315 | rx_strm.zfree = NULL; | |
316 | if (inflateInit2(&rx_strm, -15) != Z_OK) { | |
317 | rprintf(FERROR, "inflate init failed\n"); | |
65417579 | 318 | exit_cleanup(RERR_STREAMIO); |
5914bf15 | 319 | } |
58cadc86 WD |
320 | if (!(cbuf = new_array(char, MAX_DATA_COUNT)) |
321 | || !(dbuf = new_array(char, AVAIL_OUT_SIZE(CHUNK_SIZE)))) | |
5914bf15 PM |
322 | out_of_memory("recv_deflated_token"); |
323 | init_done = 1; | |
324 | } else { | |
325 | inflateReset(&rx_strm); | |
326 | } | |
b8d4524b | 327 | recv_state = r_idle; |
5914bf15 PM |
328 | rx_token = 0; |
329 | break; | |
330 | ||
331 | case r_idle: | |
332 | case r_inflated: | |
333 | if (saved_flag) { | |
334 | flag = saved_flag & 0xff; | |
335 | saved_flag = 0; | |
336 | } else | |
337 | flag = read_byte(f); | |
338 | if ((flag & 0xC0) == DEFLATED_DATA) { | |
339 | n = ((flag & 0x3f) << 8) + read_byte(f); | |
340 | read_buf(f, cbuf, n); | |
341 | rx_strm.next_in = (Bytef *)cbuf; | |
342 | rx_strm.avail_in = n; | |
343 | recv_state = r_inflating; | |
344 | break; | |
345 | } | |
346 | if (recv_state == r_inflated) { | |
347 | /* check previous inflated stuff ended correctly */ | |
348 | rx_strm.avail_in = 0; | |
349 | rx_strm.next_out = (Bytef *)dbuf; | |
1dbb94ca | 350 | rx_strm.avail_out = AVAIL_OUT_SIZE(CHUNK_SIZE); |
5914bf15 | 351 | r = inflate(&rx_strm, Z_SYNC_FLUSH); |
1dbb94ca | 352 | n = AVAIL_OUT_SIZE(CHUNK_SIZE) - rx_strm.avail_out; |
5914bf15 PM |
353 | /* |
354 | * Z_BUF_ERROR just means no progress was | |
355 | * made, i.e. the decompressor didn't have | |
356 | * any pending output for us. | |
357 | */ | |
358 | if (r != Z_OK && r != Z_BUF_ERROR) { | |
359 | rprintf(FERROR, "inflate flush returned %d (%d bytes)\n", | |
360 | r, n); | |
65417579 | 361 | exit_cleanup(RERR_STREAMIO); |
5914bf15 PM |
362 | } |
363 | if (n != 0 && r != Z_BUF_ERROR) { | |
364 | /* have to return some more data and | |
365 | save the flag for later. */ | |
366 | saved_flag = flag + 0x10000; | |
367 | *data = dbuf; | |
368 | return n; | |
369 | } | |
370 | /* | |
371 | * At this point the decompressor should | |
372 | * be expecting to see the 0, 0, ff, ff bytes. | |
373 | */ | |
374 | if (!inflateSyncPoint(&rx_strm)) { | |
375 | rprintf(FERROR, "decompressor lost sync!\n"); | |
65417579 | 376 | exit_cleanup(RERR_STREAMIO); |
5914bf15 PM |
377 | } |
378 | rx_strm.avail_in = 4; | |
379 | rx_strm.next_in = (Bytef *)cbuf; | |
380 | cbuf[0] = cbuf[1] = 0; | |
381 | cbuf[2] = cbuf[3] = 0xff; | |
382 | inflate(&rx_strm, Z_SYNC_FLUSH); | |
383 | recv_state = r_idle; | |
384 | } | |
385 | if (flag == END_FLAG) { | |
386 | /* that's all folks */ | |
387 | recv_state = r_init; | |
388 | return 0; | |
389 | } | |
390 | ||
391 | /* here we have a token of some kind */ | |
392 | if (flag & TOKEN_REL) { | |
393 | rx_token += flag & 0x3f; | |
394 | flag >>= 6; | |
395 | } else | |
396 | rx_token = read_int(f); | |
397 | if (flag & 1) { | |
398 | rx_run = read_byte(f); | |
399 | rx_run += read_byte(f) << 8; | |
400 | recv_state = r_running; | |
401 | } | |
402 | return -1 - rx_token; | |
403 | ||
404 | case r_inflating: | |
405 | rx_strm.next_out = (Bytef *)dbuf; | |
1dbb94ca | 406 | rx_strm.avail_out = AVAIL_OUT_SIZE(CHUNK_SIZE); |
5914bf15 | 407 | r = inflate(&rx_strm, Z_NO_FLUSH); |
1dbb94ca | 408 | n = AVAIL_OUT_SIZE(CHUNK_SIZE) - rx_strm.avail_out; |
5914bf15 PM |
409 | if (r != Z_OK) { |
410 | rprintf(FERROR, "inflate returned %d (%d bytes)\n", r, n); | |
65417579 | 411 | exit_cleanup(RERR_STREAMIO); |
5914bf15 PM |
412 | } |
413 | if (rx_strm.avail_in == 0) | |
414 | recv_state = r_inflated; | |
415 | if (n != 0) { | |
416 | *data = dbuf; | |
417 | return n; | |
418 | } | |
419 | break; | |
420 | ||
421 | case r_running: | |
422 | ++rx_token; | |
423 | if (--rx_run == 0) | |
424 | recv_state = r_idle; | |
425 | return -1 - rx_token; | |
b8d4524b | 426 | } |
861c20b4 | 427 | } |
861c20b4 PM |
428 | } |
429 | ||
430 | /* | |
431 | * put the data corresponding to a token that we've just returned | |
432 | * from recv_deflated_token into the decompressor's history buffer. | |
433 | */ | |
7fcbf9e4 | 434 | static void see_deflate_token(char *buf, int32 len) |
861c20b4 | 435 | { |
acc461c7 WD |
436 | int r; |
437 | int32 blklen; | |
5914bf15 PM |
438 | unsigned char hdr[5]; |
439 | ||
440 | rx_strm.avail_in = 0; | |
441 | blklen = 0; | |
442 | hdr[0] = 0; | |
443 | do { | |
444 | if (rx_strm.avail_in == 0 && len != 0) { | |
445 | if (blklen == 0) { | |
446 | /* Give it a fake stored-block header. */ | |
447 | rx_strm.next_in = (Bytef *)hdr; | |
448 | rx_strm.avail_in = 5; | |
449 | blklen = len; | |
450 | if (blklen > 0xffff) | |
451 | blklen = 0xffff; | |
452 | hdr[1] = blklen; | |
453 | hdr[2] = blklen >> 8; | |
454 | hdr[3] = ~hdr[1]; | |
455 | hdr[4] = ~hdr[2]; | |
456 | } else { | |
457 | rx_strm.next_in = (Bytef *)buf; | |
458 | rx_strm.avail_in = blklen; | |
459 | len -= blklen; | |
460 | blklen = 0; | |
461 | } | |
462 | } | |
463 | rx_strm.next_out = (Bytef *)dbuf; | |
1dbb94ca | 464 | rx_strm.avail_out = AVAIL_OUT_SIZE(CHUNK_SIZE); |
5914bf15 PM |
465 | r = inflate(&rx_strm, Z_SYNC_FLUSH); |
466 | if (r != Z_OK) { | |
467 | rprintf(FERROR, "inflate (token) returned %d\n", r); | |
65417579 | 468 | exit_cleanup(RERR_STREAMIO); |
5914bf15 PM |
469 | } |
470 | } while (len || rx_strm.avail_out == 0); | |
861c20b4 | 471 | } |
70d794dc | 472 | |
79f671cc MP |
473 | /** |
474 | * Transmit a verbatim buffer of length @p n followed by a token. | |
d67c8bdf | 475 | * If token == -1 then we have reached EOF |
70d794dc | 476 | * If n == 0 then don't send a buffer |
70d794dc | 477 | */ |
acc461c7 | 478 | void send_token(int f, int32 token, struct map_struct *buf, OFF_T offset, |
7fcbf9e4 | 479 | int32 n, int32 toklen) |
70d794dc | 480 | { |
7fcbf9e4 WD |
481 | if (!do_compression) |
482 | simple_send_token(f, token, buf, offset, n); | |
483 | else | |
5914bf15 | 484 | send_deflated_token(f, token, buf, offset, n, toklen); |
70d794dc AT |
485 | } |
486 | ||
487 | ||
488 | /* | |
489 | * receive a token or buffer from the other end. If the reurn value is >0 then | |
490 | * it is a data buffer of that length, and *data will point at the data. | |
491 | * if the return value is -i then it represents token i-1 | |
492 | * if the return value is 0 then the end has been reached | |
493 | */ | |
7fcbf9e4 | 494 | int32 recv_token(int f, char **data) |
70d794dc | 495 | { |
5914bf15 PM |
496 | int tok; |
497 | ||
498 | if (!do_compression) { | |
499 | tok = simple_recv_token(f,data); | |
500 | } else { | |
501 | tok = recv_deflated_token(f, data); | |
502 | } | |
503 | return tok; | |
861c20b4 PM |
504 | } |
505 | ||
506 | /* | |
507 | * look at the data corresponding to a token, if necessary | |
508 | */ | |
7fcbf9e4 | 509 | void see_token(char *data, int32 toklen) |
861c20b4 | 510 | { |
5914bf15 PM |
511 | if (do_compression) |
512 | see_deflate_token(data, toklen); | |
70d794dc | 513 | } |