Annotation of embedaddon/rsync/token.c, revision 1.1.1.3
1.1 misho 1: /*
2: * Routines used by the file-transfer code.
3: *
4: * Copyright (C) 1996 Andrew Tridgell
5: * Copyright (C) 1996 Paul Mackerras
1.1.1.3 ! misho 6: * Copyright (C) 2003-2015 Wayne Davison
1.1 misho 7: *
8: * This program is free software; you can redistribute it and/or modify
9: * it under the terms of the GNU General Public License as published by
10: * the Free Software Foundation; either version 3 of the License, or
11: * (at your option) any later version.
12: *
13: * This program is distributed in the hope that it will be useful,
14: * but WITHOUT ANY WARRANTY; without even the implied warranty of
15: * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
16: * GNU General Public License for more details.
17: *
18: * You should have received a copy of the GNU General Public License along
19: * with this program; if not, visit the http://fsf.org website.
20: */
21:
22: #include "rsync.h"
1.1.1.2 misho 23: #include "itypes.h"
24: #include <zlib.h>
25:
1.1 misho 26: extern int do_compression;
1.1.1.2 misho 27: extern int protocol_version;
1.1 misho 28: extern int module_id;
29: extern int def_compress_level;
30: extern char *skip_compress;
31:
32: static int compression_level, per_file_default_level;
33:
34: struct suffix_tree {
35: struct suffix_tree *sibling;
36: struct suffix_tree *child;
37: char letter, word_end;
38: };
39:
40: static char *match_list;
41: static struct suffix_tree *suftree;
42:
43: static void add_suffix(struct suffix_tree **prior, char ltr, const char *str)
44: {
45: struct suffix_tree *node, *newnode;
46:
47: if (ltr == '[') {
48: const char *after = strchr(str, ']');
49: /* Treat "[foo" and "[]" as having a literal '['. */
50: if (after && after++ != str+1) {
51: while ((ltr = *str++) != ']')
52: add_suffix(prior, ltr, after);
53: return;
54: }
55: }
56:
57: for (node = *prior; node; prior = &node->sibling, node = node->sibling) {
58: if (node->letter == ltr) {
59: if (*str)
60: add_suffix(&node->child, *str, str+1);
61: else
62: node->word_end = 1;
63: return;
64: }
65: if (node->letter > ltr)
66: break;
67: }
68: if (!(newnode = new(struct suffix_tree)))
69: out_of_memory("add_suffix");
70: newnode->sibling = node;
71: newnode->child = NULL;
72: newnode->letter = ltr;
73: *prior = newnode;
74: if (*str) {
75: add_suffix(&newnode->child, *str, str+1);
76: newnode->word_end = 0;
77: } else
78: newnode->word_end = 1;
79: }
80:
81: static void add_nocompress_suffixes(const char *str)
82: {
83: char *buf, *t;
84: const char *f = str;
85:
86: if (!(buf = new_array(char, strlen(f) + 1)))
87: out_of_memory("add_nocompress_suffixes");
88:
89: while (*f) {
90: if (*f == '/') {
91: f++;
92: continue;
93: }
94:
95: t = buf;
96: do {
97: if (isUpper(f))
98: *t++ = toLower(f);
99: else
100: *t++ = *f;
101: } while (*++f != '/' && *f);
102: *t++ = '\0';
103:
104: add_suffix(&suftree, *buf, buf+1);
105: }
106:
107: free(buf);
108: }
109:
110: static void init_set_compression(void)
111: {
112: const char *f;
113: char *t, *start;
114:
115: if (skip_compress)
116: add_nocompress_suffixes(skip_compress);
117:
118: /* A non-daemon transfer skips the default suffix list if the
119: * user specified --skip-compress. */
120: if (skip_compress && module_id < 0)
121: f = "";
122: else
123: f = lp_dont_compress(module_id);
124:
125: if (!(match_list = t = new_array(char, strlen(f) + 2)))
126: out_of_memory("set_compression");
127:
128: per_file_default_level = def_compress_level;
129:
130: while (*f) {
131: if (*f == ' ') {
132: f++;
133: continue;
134: }
135:
136: start = t;
137: do {
138: if (isUpper(f))
139: *t++ = toLower(f);
140: else
141: *t++ = *f;
142: } while (*++f != ' ' && *f);
143: *t++ = '\0';
144:
145: if (t - start == 1+1 && *start == '*') {
146: /* Optimize a match-string of "*". */
147: *match_list = '\0';
148: suftree = NULL;
149: per_file_default_level = 0;
150: break;
151: }
152:
153: /* Move *.foo items into the stuffix tree. */
154: if (*start == '*' && start[1] == '.' && start[2]
155: && !strpbrk(start+2, ".?*")) {
156: add_suffix(&suftree, start[2], start+3);
157: t = start;
158: }
159: }
160: *t++ = '\0';
161: }
162:
163: /* determine the compression level based on a wildcard filename list */
164: void set_compression(const char *fname)
165: {
166: const struct suffix_tree *node;
167: const char *s;
168: char ltr;
169:
170: if (!do_compression)
171: return;
172:
173: if (!match_list)
174: init_set_compression();
175:
176: compression_level = per_file_default_level;
177:
178: if (!*match_list && !suftree)
179: return;
180:
181: if ((s = strrchr(fname, '/')) != NULL)
182: fname = s + 1;
183:
184: for (s = match_list; *s; s += strlen(s) + 1) {
185: if (iwildmatch(s, fname)) {
186: compression_level = 0;
187: return;
188: }
189: }
190:
191: if (!(node = suftree) || !(s = strrchr(fname, '.'))
192: || s == fname || !(ltr = *++s))
193: return;
194:
195: while (1) {
196: if (isUpper(<r))
197: ltr = toLower(<r);
198: while (node->letter != ltr) {
199: if (node->letter > ltr)
200: return;
201: if (!(node = node->sibling))
202: return;
203: }
204: if ((ltr = *++s) == '\0') {
205: if (node->word_end)
206: compression_level = 0;
207: return;
208: }
209: if (!(node = node->child))
210: return;
211: }
212: }
213:
214: /* non-compressing recv token */
215: static int32 simple_recv_token(int f, char **data)
216: {
217: static int32 residue;
218: static char *buf;
219: int32 n;
220:
221: if (!buf) {
222: buf = new_array(char, CHUNK_SIZE);
223: if (!buf)
224: out_of_memory("simple_recv_token");
225: }
226:
227: if (residue == 0) {
228: int32 i = read_int(f);
229: if (i <= 0)
230: return i;
231: residue = i;
232: }
233:
234: *data = buf;
235: n = MIN(CHUNK_SIZE,residue);
236: residue -= n;
237: read_buf(f,buf,n);
238: return n;
239: }
240:
241: /* non-compressing send token */
242: static void simple_send_token(int f, int32 token, struct map_struct *buf,
243: OFF_T offset, int32 n)
244: {
245: if (n > 0) {
246: int32 len = 0;
247: while (len < n) {
248: int32 n1 = MIN(CHUNK_SIZE, n-len);
249: write_int(f, n1);
250: write_buf(f, map_ptr(buf, offset+len, n1), n1);
251: len += n1;
252: }
253: }
254: /* a -2 token means to send data only and no token */
255: if (token != -2)
256: write_int(f, -(token+1));
257: }
258:
259: /* Flag bytes in compressed stream are encoded as follows: */
260: #define END_FLAG 0 /* that's all folks */
261: #define TOKEN_LONG 0x20 /* followed by 32-bit token number */
262: #define TOKENRUN_LONG 0x21 /* ditto with 16-bit run count */
263: #define DEFLATED_DATA 0x40 /* + 6-bit high len, then low len byte */
264: #define TOKEN_REL 0x80 /* + 6-bit relative token number */
265: #define TOKENRUN_REL 0xc0 /* ditto with 16-bit run count */
266:
267: #define MAX_DATA_COUNT 16383 /* fit 14 bit count into 2 bytes with flags */
268:
269: /* zlib.h says that if we want to be able to compress something in a single
270: * call, avail_out must be at least 0.1% larger than avail_in plus 12 bytes.
271: * We'll add in 0.1%+16, just to be safe (and we'll avoid floating point,
272: * to ensure that this is a compile-time value). */
273: #define AVAIL_OUT_SIZE(avail_in_size) ((avail_in_size)*1001/1000+16)
274:
275: /* For coding runs of tokens */
276: static int32 last_token = -1;
277: static int32 run_start;
278: static int32 last_run_end;
279:
280: /* Deflation state */
281: static z_stream tx_strm;
282:
283: /* Output buffer */
284: static char *obuf;
285:
286: /* We want obuf to be able to hold both MAX_DATA_COUNT+2 bytes as well as
287: * AVAIL_OUT_SIZE(CHUNK_SIZE) bytes, so make sure that it's large enough. */
288: #if MAX_DATA_COUNT+2 > AVAIL_OUT_SIZE(CHUNK_SIZE)
289: #define OBUF_SIZE (MAX_DATA_COUNT+2)
290: #else
291: #define OBUF_SIZE AVAIL_OUT_SIZE(CHUNK_SIZE)
292: #endif
293:
294: /* Send a deflated token */
295: static void
296: send_deflated_token(int f, int32 token, struct map_struct *buf, OFF_T offset,
297: int32 nb, int32 toklen)
298: {
299: int32 n, r;
300: static int init_done, flush_pending;
301:
302: if (last_token == -1) {
303: /* initialization */
304: if (!init_done) {
305: tx_strm.next_in = NULL;
306: tx_strm.zalloc = NULL;
307: tx_strm.zfree = NULL;
308: if (deflateInit2(&tx_strm, compression_level,
309: Z_DEFLATED, -15, 8,
310: Z_DEFAULT_STRATEGY) != Z_OK) {
311: rprintf(FERROR, "compression init failed\n");
1.1.1.2 misho 312: exit_cleanup(RERR_PROTOCOL);
1.1 misho 313: }
314: if ((obuf = new_array(char, OBUF_SIZE)) == NULL)
315: out_of_memory("send_deflated_token");
316: init_done = 1;
317: } else
318: deflateReset(&tx_strm);
319: last_run_end = 0;
320: run_start = token;
321: flush_pending = 0;
322: } else if (last_token == -2) {
323: run_start = token;
324: } else if (nb != 0 || token != last_token + 1
325: || token >= run_start + 65536) {
326: /* output previous run */
327: r = run_start - last_run_end;
328: n = last_token - run_start;
329: if (r >= 0 && r <= 63) {
330: write_byte(f, (n==0? TOKEN_REL: TOKENRUN_REL) + r);
331: } else {
332: write_byte(f, (n==0? TOKEN_LONG: TOKENRUN_LONG));
333: write_int(f, run_start);
334: }
335: if (n != 0) {
336: write_byte(f, n);
337: write_byte(f, n >> 8);
338: }
339: last_run_end = last_token;
340: run_start = token;
341: }
342:
343: last_token = token;
344:
345: if (nb != 0 || flush_pending) {
346: /* deflate the data starting at offset */
347: int flush = Z_NO_FLUSH;
348: tx_strm.avail_in = 0;
349: tx_strm.avail_out = 0;
350: do {
351: if (tx_strm.avail_in == 0 && nb != 0) {
352: /* give it some more input */
353: n = MIN(nb, CHUNK_SIZE);
354: tx_strm.next_in = (Bytef *)
355: map_ptr(buf, offset, n);
356: tx_strm.avail_in = n;
357: nb -= n;
358: offset += n;
359: }
360: if (tx_strm.avail_out == 0) {
361: tx_strm.next_out = (Bytef *)(obuf + 2);
362: tx_strm.avail_out = MAX_DATA_COUNT;
363: if (flush != Z_NO_FLUSH) {
364: /*
365: * We left the last 4 bytes in the
366: * buffer, in case they are the
367: * last 4. Move them to the front.
368: */
369: memcpy(tx_strm.next_out,
370: obuf+MAX_DATA_COUNT-2, 4);
371: tx_strm.next_out += 4;
372: tx_strm.avail_out -= 4;
373: }
374: }
375: if (nb == 0 && token != -2)
376: flush = Z_SYNC_FLUSH;
377: r = deflate(&tx_strm, flush);
378: if (r != Z_OK) {
379: rprintf(FERROR, "deflate returned %d\n", r);
380: exit_cleanup(RERR_STREAMIO);
381: }
382: if (nb == 0 || tx_strm.avail_out == 0) {
383: n = MAX_DATA_COUNT - tx_strm.avail_out;
384: if (flush != Z_NO_FLUSH) {
385: /*
386: * We have to trim off the last 4
387: * bytes of output when flushing
388: * (they are just 0, 0, ff, ff).
389: */
390: n -= 4;
391: }
392: if (n > 0) {
393: obuf[0] = DEFLATED_DATA + (n >> 8);
394: obuf[1] = n;
395: write_buf(f, obuf, n+2);
396: }
397: }
398: } while (nb != 0 || tx_strm.avail_out == 0);
399: flush_pending = token == -2;
400: }
401:
402: if (token == -1) {
403: /* end of file - clean up */
404: write_byte(f, END_FLAG);
1.1.1.3 ! misho 405: } else if (token != -2 && do_compression == 1) {
1.1 misho 406: /* Add the data in the current block to the compressor's
407: * history and hash table. */
1.1.1.3 ! misho 408: #ifndef EXTERNAL_ZLIB
1.1 misho 409: do {
410: /* Break up long sections in the same way that
411: * see_deflate_token() does. */
412: int32 n1 = toklen > 0xffff ? 0xffff : toklen;
413: toklen -= n1;
414: tx_strm.next_in = (Bytef *)map_ptr(buf, offset, n1);
415: tx_strm.avail_in = n1;
1.1.1.2 misho 416: if (protocol_version >= 31) /* Newer protocols avoid a data-duplicating bug */
417: offset += n1;
1.1.1.3 ! misho 418: tx_strm.next_out = (Bytef *) obuf;
! 419: tx_strm.avail_out = AVAIL_OUT_SIZE(CHUNK_SIZE);
! 420: r = deflate(&tx_strm, Z_INSERT_ONLY);
! 421: if (r != Z_OK || tx_strm.avail_in != 0) {
! 422: rprintf(FERROR, "deflate on token returned %d (%d bytes left)\n",
! 423: r, tx_strm.avail_in);
! 424: exit_cleanup(RERR_STREAMIO);
! 425: }
1.1 misho 426: } while (toklen > 0);
1.1.1.3 ! misho 427: #else
! 428: toklen++;
! 429: rprintf(FERROR, "Impossible error in external-zlib code (1).\n");
! 430: exit_cleanup(RERR_STREAMIO);
! 431: #endif
1.1 misho 432: }
433: }
434:
435: /* tells us what the receiver is in the middle of doing */
436: static enum { r_init, r_idle, r_running, r_inflating, r_inflated } recv_state;
437:
438: /* for inflating stuff */
439: static z_stream rx_strm;
440: static char *cbuf;
441: static char *dbuf;
442:
443: /* for decoding runs of tokens */
444: static int32 rx_token;
445: static int32 rx_run;
446:
447: /* Receive a deflated token and inflate it */
448: static int32 recv_deflated_token(int f, char **data)
449: {
450: static int init_done;
451: static int32 saved_flag;
452: int32 n, flag;
453: int r;
454:
455: for (;;) {
456: switch (recv_state) {
457: case r_init:
458: if (!init_done) {
459: rx_strm.next_out = NULL;
460: rx_strm.zalloc = NULL;
461: rx_strm.zfree = NULL;
462: if (inflateInit2(&rx_strm, -15) != Z_OK) {
463: rprintf(FERROR, "inflate init failed\n");
1.1.1.2 misho 464: exit_cleanup(RERR_PROTOCOL);
1.1 misho 465: }
466: if (!(cbuf = new_array(char, MAX_DATA_COUNT))
467: || !(dbuf = new_array(char, AVAIL_OUT_SIZE(CHUNK_SIZE))))
468: out_of_memory("recv_deflated_token");
469: init_done = 1;
470: } else {
471: inflateReset(&rx_strm);
472: }
473: recv_state = r_idle;
474: rx_token = 0;
475: break;
476:
477: case r_idle:
478: case r_inflated:
479: if (saved_flag) {
480: flag = saved_flag & 0xff;
481: saved_flag = 0;
482: } else
483: flag = read_byte(f);
484: if ((flag & 0xC0) == DEFLATED_DATA) {
485: n = ((flag & 0x3f) << 8) + read_byte(f);
486: read_buf(f, cbuf, n);
487: rx_strm.next_in = (Bytef *)cbuf;
488: rx_strm.avail_in = n;
489: recv_state = r_inflating;
490: break;
491: }
492: if (recv_state == r_inflated) {
493: /* check previous inflated stuff ended correctly */
494: rx_strm.avail_in = 0;
495: rx_strm.next_out = (Bytef *)dbuf;
496: rx_strm.avail_out = AVAIL_OUT_SIZE(CHUNK_SIZE);
497: r = inflate(&rx_strm, Z_SYNC_FLUSH);
498: n = AVAIL_OUT_SIZE(CHUNK_SIZE) - rx_strm.avail_out;
499: /*
500: * Z_BUF_ERROR just means no progress was
501: * made, i.e. the decompressor didn't have
502: * any pending output for us.
503: */
504: if (r != Z_OK && r != Z_BUF_ERROR) {
505: rprintf(FERROR, "inflate flush returned %d (%d bytes)\n",
506: r, n);
507: exit_cleanup(RERR_STREAMIO);
508: }
509: if (n != 0 && r != Z_BUF_ERROR) {
510: /* have to return some more data and
511: save the flag for later. */
512: saved_flag = flag + 0x10000;
513: *data = dbuf;
514: return n;
515: }
516: /*
517: * At this point the decompressor should
518: * be expecting to see the 0, 0, ff, ff bytes.
519: */
520: if (!inflateSyncPoint(&rx_strm)) {
521: rprintf(FERROR, "decompressor lost sync!\n");
522: exit_cleanup(RERR_STREAMIO);
523: }
524: rx_strm.avail_in = 4;
525: rx_strm.next_in = (Bytef *)cbuf;
526: cbuf[0] = cbuf[1] = 0;
527: cbuf[2] = cbuf[3] = 0xff;
528: inflate(&rx_strm, Z_SYNC_FLUSH);
529: recv_state = r_idle;
530: }
531: if (flag == END_FLAG) {
532: /* that's all folks */
533: recv_state = r_init;
534: return 0;
535: }
536:
537: /* here we have a token of some kind */
538: if (flag & TOKEN_REL) {
539: rx_token += flag & 0x3f;
540: flag >>= 6;
541: } else
542: rx_token = read_int(f);
543: if (flag & 1) {
544: rx_run = read_byte(f);
545: rx_run += read_byte(f) << 8;
546: recv_state = r_running;
547: }
548: return -1 - rx_token;
549:
550: case r_inflating:
551: rx_strm.next_out = (Bytef *)dbuf;
552: rx_strm.avail_out = AVAIL_OUT_SIZE(CHUNK_SIZE);
553: r = inflate(&rx_strm, Z_NO_FLUSH);
554: n = AVAIL_OUT_SIZE(CHUNK_SIZE) - rx_strm.avail_out;
555: if (r != Z_OK) {
556: rprintf(FERROR, "inflate returned %d (%d bytes)\n", r, n);
557: exit_cleanup(RERR_STREAMIO);
558: }
559: if (rx_strm.avail_in == 0)
560: recv_state = r_inflated;
561: if (n != 0) {
562: *data = dbuf;
563: return n;
564: }
565: break;
566:
567: case r_running:
568: ++rx_token;
569: if (--rx_run == 0)
570: recv_state = r_idle;
571: return -1 - rx_token;
572: }
573: }
574: }
575:
576: /*
577: * put the data corresponding to a token that we've just returned
578: * from recv_deflated_token into the decompressor's history buffer.
579: */
580: static void see_deflate_token(char *buf, int32 len)
581: {
1.1.1.3 ! misho 582: #ifndef EXTERNAL_ZLIB
1.1 misho 583: int r;
584: int32 blklen;
585: unsigned char hdr[5];
586:
587: rx_strm.avail_in = 0;
588: blklen = 0;
589: hdr[0] = 0;
590: do {
591: if (rx_strm.avail_in == 0 && len != 0) {
592: if (blklen == 0) {
593: /* Give it a fake stored-block header. */
594: rx_strm.next_in = (Bytef *)hdr;
595: rx_strm.avail_in = 5;
596: blklen = len;
597: if (blklen > 0xffff)
598: blklen = 0xffff;
599: hdr[1] = blklen;
600: hdr[2] = blklen >> 8;
601: hdr[3] = ~hdr[1];
602: hdr[4] = ~hdr[2];
603: } else {
604: rx_strm.next_in = (Bytef *)buf;
605: rx_strm.avail_in = blklen;
1.1.1.2 misho 606: if (protocol_version >= 31) /* Newer protocols avoid a data-duplicating bug */
607: buf += blklen;
1.1 misho 608: len -= blklen;
609: blklen = 0;
610: }
611: }
612: rx_strm.next_out = (Bytef *)dbuf;
613: rx_strm.avail_out = AVAIL_OUT_SIZE(CHUNK_SIZE);
614: r = inflate(&rx_strm, Z_SYNC_FLUSH);
615: if (r != Z_OK && r != Z_BUF_ERROR) {
616: rprintf(FERROR, "inflate (token) returned %d\n", r);
617: exit_cleanup(RERR_STREAMIO);
618: }
619: } while (len || rx_strm.avail_out == 0);
1.1.1.3 ! misho 620: #else
! 621: buf++; len++;
! 622: rprintf(FERROR, "Impossible error in external-zlib code (2).\n");
! 623: exit_cleanup(RERR_STREAMIO);
! 624: #endif
1.1 misho 625: }
626:
627: /**
628: * Transmit a verbatim buffer of length @p n followed by a token.
629: * If token == -1 then we have reached EOF
630: * If n == 0 then don't send a buffer
631: */
632: void send_token(int f, int32 token, struct map_struct *buf, OFF_T offset,
633: int32 n, int32 toklen)
634: {
635: if (!do_compression)
636: simple_send_token(f, token, buf, offset, n);
637: else
638: send_deflated_token(f, token, buf, offset, n, toklen);
639: }
640:
641: /*
642: * receive a token or buffer from the other end. If the reurn value is >0 then
643: * it is a data buffer of that length, and *data will point at the data.
644: * if the return value is -i then it represents token i-1
645: * if the return value is 0 then the end has been reached
646: */
647: int32 recv_token(int f, char **data)
648: {
649: int tok;
650:
651: if (!do_compression) {
652: tok = simple_recv_token(f,data);
653: } else {
654: tok = recv_deflated_token(f, data);
655: }
656: return tok;
657: }
658:
659: /*
660: * look at the data corresponding to a token, if necessary
661: */
662: void see_token(char *data, int32 toklen)
663: {
1.1.1.3 ! misho 664: if (do_compression == 1)
1.1 misho 665: see_deflate_token(data, toklen);
666: }
FreeBSD-CVSweb <freebsd-cvsweb@FreeBSD.org>