Annotation of embedaddon/rsync/token.c, revision 1.1.1.2
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
1.1.1.2 ! misho 6: * Copyright (C) 2003-2013 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:
! 26: #ifndef Z_INSERT_ONLY
! 27: #define Z_INSERT_ONLY Z_SYNC_FLUSH
! 28: #endif
1.1 misho 29:
30: extern int do_compression;
1.1.1.2 ! misho 31: extern int protocol_version;
1.1 misho 32: extern int module_id;
33: extern int def_compress_level;
34: extern char *skip_compress;
35:
36: static int compression_level, per_file_default_level;
37:
38: struct suffix_tree {
39: struct suffix_tree *sibling;
40: struct suffix_tree *child;
41: char letter, word_end;
42: };
43:
44: static char *match_list;
45: static struct suffix_tree *suftree;
46:
47: static void add_suffix(struct suffix_tree **prior, char ltr, const char *str)
48: {
49: struct suffix_tree *node, *newnode;
50:
51: if (ltr == '[') {
52: const char *after = strchr(str, ']');
53: /* Treat "[foo" and "[]" as having a literal '['. */
54: if (after && after++ != str+1) {
55: while ((ltr = *str++) != ']')
56: add_suffix(prior, ltr, after);
57: return;
58: }
59: }
60:
61: for (node = *prior; node; prior = &node->sibling, node = node->sibling) {
62: if (node->letter == ltr) {
63: if (*str)
64: add_suffix(&node->child, *str, str+1);
65: else
66: node->word_end = 1;
67: return;
68: }
69: if (node->letter > ltr)
70: break;
71: }
72: if (!(newnode = new(struct suffix_tree)))
73: out_of_memory("add_suffix");
74: newnode->sibling = node;
75: newnode->child = NULL;
76: newnode->letter = ltr;
77: *prior = newnode;
78: if (*str) {
79: add_suffix(&newnode->child, *str, str+1);
80: newnode->word_end = 0;
81: } else
82: newnode->word_end = 1;
83: }
84:
85: static void add_nocompress_suffixes(const char *str)
86: {
87: char *buf, *t;
88: const char *f = str;
89:
90: if (!(buf = new_array(char, strlen(f) + 1)))
91: out_of_memory("add_nocompress_suffixes");
92:
93: while (*f) {
94: if (*f == '/') {
95: f++;
96: continue;
97: }
98:
99: t = buf;
100: do {
101: if (isUpper(f))
102: *t++ = toLower(f);
103: else
104: *t++ = *f;
105: } while (*++f != '/' && *f);
106: *t++ = '\0';
107:
108: add_suffix(&suftree, *buf, buf+1);
109: }
110:
111: free(buf);
112: }
113:
114: static void init_set_compression(void)
115: {
116: const char *f;
117: char *t, *start;
118:
119: if (skip_compress)
120: add_nocompress_suffixes(skip_compress);
121:
122: /* A non-daemon transfer skips the default suffix list if the
123: * user specified --skip-compress. */
124: if (skip_compress && module_id < 0)
125: f = "";
126: else
127: f = lp_dont_compress(module_id);
128:
129: if (!(match_list = t = new_array(char, strlen(f) + 2)))
130: out_of_memory("set_compression");
131:
132: per_file_default_level = def_compress_level;
133:
134: while (*f) {
135: if (*f == ' ') {
136: f++;
137: continue;
138: }
139:
140: start = t;
141: do {
142: if (isUpper(f))
143: *t++ = toLower(f);
144: else
145: *t++ = *f;
146: } while (*++f != ' ' && *f);
147: *t++ = '\0';
148:
149: if (t - start == 1+1 && *start == '*') {
150: /* Optimize a match-string of "*". */
151: *match_list = '\0';
152: suftree = NULL;
153: per_file_default_level = 0;
154: break;
155: }
156:
157: /* Move *.foo items into the stuffix tree. */
158: if (*start == '*' && start[1] == '.' && start[2]
159: && !strpbrk(start+2, ".?*")) {
160: add_suffix(&suftree, start[2], start+3);
161: t = start;
162: }
163: }
164: *t++ = '\0';
165: }
166:
167: /* determine the compression level based on a wildcard filename list */
168: void set_compression(const char *fname)
169: {
170: const struct suffix_tree *node;
171: const char *s;
172: char ltr;
173:
174: if (!do_compression)
175: return;
176:
177: if (!match_list)
178: init_set_compression();
179:
180: compression_level = per_file_default_level;
181:
182: if (!*match_list && !suftree)
183: return;
184:
185: if ((s = strrchr(fname, '/')) != NULL)
186: fname = s + 1;
187:
188: for (s = match_list; *s; s += strlen(s) + 1) {
189: if (iwildmatch(s, fname)) {
190: compression_level = 0;
191: return;
192: }
193: }
194:
195: if (!(node = suftree) || !(s = strrchr(fname, '.'))
196: || s == fname || !(ltr = *++s))
197: return;
198:
199: while (1) {
200: if (isUpper(<r))
201: ltr = toLower(<r);
202: while (node->letter != ltr) {
203: if (node->letter > ltr)
204: return;
205: if (!(node = node->sibling))
206: return;
207: }
208: if ((ltr = *++s) == '\0') {
209: if (node->word_end)
210: compression_level = 0;
211: return;
212: }
213: if (!(node = node->child))
214: return;
215: }
216: }
217:
218: /* non-compressing recv token */
219: static int32 simple_recv_token(int f, char **data)
220: {
221: static int32 residue;
222: static char *buf;
223: int32 n;
224:
225: if (!buf) {
226: buf = new_array(char, CHUNK_SIZE);
227: if (!buf)
228: out_of_memory("simple_recv_token");
229: }
230:
231: if (residue == 0) {
232: int32 i = read_int(f);
233: if (i <= 0)
234: return i;
235: residue = i;
236: }
237:
238: *data = buf;
239: n = MIN(CHUNK_SIZE,residue);
240: residue -= n;
241: read_buf(f,buf,n);
242: return n;
243: }
244:
245: /* non-compressing send token */
246: static void simple_send_token(int f, int32 token, struct map_struct *buf,
247: OFF_T offset, int32 n)
248: {
249: if (n > 0) {
250: int32 len = 0;
251: while (len < n) {
252: int32 n1 = MIN(CHUNK_SIZE, n-len);
253: write_int(f, n1);
254: write_buf(f, map_ptr(buf, offset+len, n1), n1);
255: len += n1;
256: }
257: }
258: /* a -2 token means to send data only and no token */
259: if (token != -2)
260: write_int(f, -(token+1));
261: }
262:
263: /* Flag bytes in compressed stream are encoded as follows: */
264: #define END_FLAG 0 /* that's all folks */
265: #define TOKEN_LONG 0x20 /* followed by 32-bit token number */
266: #define TOKENRUN_LONG 0x21 /* ditto with 16-bit run count */
267: #define DEFLATED_DATA 0x40 /* + 6-bit high len, then low len byte */
268: #define TOKEN_REL 0x80 /* + 6-bit relative token number */
269: #define TOKENRUN_REL 0xc0 /* ditto with 16-bit run count */
270:
271: #define MAX_DATA_COUNT 16383 /* fit 14 bit count into 2 bytes with flags */
272:
273: /* zlib.h says that if we want to be able to compress something in a single
274: * call, avail_out must be at least 0.1% larger than avail_in plus 12 bytes.
275: * We'll add in 0.1%+16, just to be safe (and we'll avoid floating point,
276: * to ensure that this is a compile-time value). */
277: #define AVAIL_OUT_SIZE(avail_in_size) ((avail_in_size)*1001/1000+16)
278:
279: /* For coding runs of tokens */
280: static int32 last_token = -1;
281: static int32 run_start;
282: static int32 last_run_end;
283:
284: /* Deflation state */
285: static z_stream tx_strm;
286:
287: /* Output buffer */
288: static char *obuf;
289:
290: /* We want obuf to be able to hold both MAX_DATA_COUNT+2 bytes as well as
291: * AVAIL_OUT_SIZE(CHUNK_SIZE) bytes, so make sure that it's large enough. */
292: #if MAX_DATA_COUNT+2 > AVAIL_OUT_SIZE(CHUNK_SIZE)
293: #define OBUF_SIZE (MAX_DATA_COUNT+2)
294: #else
295: #define OBUF_SIZE AVAIL_OUT_SIZE(CHUNK_SIZE)
296: #endif
297:
298: /* Send a deflated token */
299: static void
300: send_deflated_token(int f, int32 token, struct map_struct *buf, OFF_T offset,
301: int32 nb, int32 toklen)
302: {
303: int32 n, r;
304: static int init_done, flush_pending;
305:
306: if (last_token == -1) {
307: /* initialization */
308: if (!init_done) {
309: tx_strm.next_in = NULL;
310: tx_strm.zalloc = NULL;
311: tx_strm.zfree = NULL;
312: if (deflateInit2(&tx_strm, compression_level,
313: Z_DEFLATED, -15, 8,
314: Z_DEFAULT_STRATEGY) != Z_OK) {
315: rprintf(FERROR, "compression init failed\n");
1.1.1.2 ! misho 316: exit_cleanup(RERR_PROTOCOL);
1.1 misho 317: }
318: if ((obuf = new_array(char, OBUF_SIZE)) == NULL)
319: out_of_memory("send_deflated_token");
320: init_done = 1;
321: } else
322: deflateReset(&tx_strm);
323: last_run_end = 0;
324: run_start = token;
325: flush_pending = 0;
326: } else if (last_token == -2) {
327: run_start = token;
328: } else if (nb != 0 || token != last_token + 1
329: || token >= run_start + 65536) {
330: /* output previous run */
331: r = run_start - last_run_end;
332: n = last_token - run_start;
333: if (r >= 0 && r <= 63) {
334: write_byte(f, (n==0? TOKEN_REL: TOKENRUN_REL) + r);
335: } else {
336: write_byte(f, (n==0? TOKEN_LONG: TOKENRUN_LONG));
337: write_int(f, run_start);
338: }
339: if (n != 0) {
340: write_byte(f, n);
341: write_byte(f, n >> 8);
342: }
343: last_run_end = last_token;
344: run_start = token;
345: }
346:
347: last_token = token;
348:
349: if (nb != 0 || flush_pending) {
350: /* deflate the data starting at offset */
351: int flush = Z_NO_FLUSH;
352: tx_strm.avail_in = 0;
353: tx_strm.avail_out = 0;
354: do {
355: if (tx_strm.avail_in == 0 && nb != 0) {
356: /* give it some more input */
357: n = MIN(nb, CHUNK_SIZE);
358: tx_strm.next_in = (Bytef *)
359: map_ptr(buf, offset, n);
360: tx_strm.avail_in = n;
361: nb -= n;
362: offset += n;
363: }
364: if (tx_strm.avail_out == 0) {
365: tx_strm.next_out = (Bytef *)(obuf + 2);
366: tx_strm.avail_out = MAX_DATA_COUNT;
367: if (flush != Z_NO_FLUSH) {
368: /*
369: * We left the last 4 bytes in the
370: * buffer, in case they are the
371: * last 4. Move them to the front.
372: */
373: memcpy(tx_strm.next_out,
374: obuf+MAX_DATA_COUNT-2, 4);
375: tx_strm.next_out += 4;
376: tx_strm.avail_out -= 4;
377: }
378: }
379: if (nb == 0 && token != -2)
380: flush = Z_SYNC_FLUSH;
381: r = deflate(&tx_strm, flush);
382: if (r != Z_OK) {
383: rprintf(FERROR, "deflate returned %d\n", r);
384: exit_cleanup(RERR_STREAMIO);
385: }
386: if (nb == 0 || tx_strm.avail_out == 0) {
387: n = MAX_DATA_COUNT - tx_strm.avail_out;
388: if (flush != Z_NO_FLUSH) {
389: /*
390: * We have to trim off the last 4
391: * bytes of output when flushing
392: * (they are just 0, 0, ff, ff).
393: */
394: n -= 4;
395: }
396: if (n > 0) {
397: obuf[0] = DEFLATED_DATA + (n >> 8);
398: obuf[1] = n;
399: write_buf(f, obuf, n+2);
400: }
401: }
402: } while (nb != 0 || tx_strm.avail_out == 0);
403: flush_pending = token == -2;
404: }
405:
406: if (token == -1) {
407: /* end of file - clean up */
408: write_byte(f, END_FLAG);
409: } else if (token != -2) {
410: /* Add the data in the current block to the compressor's
411: * history and hash table. */
412: do {
413: /* Break up long sections in the same way that
414: * see_deflate_token() does. */
415: int32 n1 = toklen > 0xffff ? 0xffff : toklen;
416: toklen -= n1;
417: tx_strm.next_in = (Bytef *)map_ptr(buf, offset, n1);
418: tx_strm.avail_in = n1;
1.1.1.2 ! misho 419: if (protocol_version >= 31) /* Newer protocols avoid a data-duplicating bug */
! 420: offset += n1;
! 421: do {
! 422: tx_strm.next_out = (Bytef *) obuf;
! 423: tx_strm.avail_out = AVAIL_OUT_SIZE(CHUNK_SIZE);
! 424: r = deflate(&tx_strm, Z_INSERT_ONLY);
! 425: if (r != Z_OK) {
! 426: rprintf(FERROR, "deflate on token returned %d (%d bytes left)\n",
! 427: r, tx_strm.avail_in);
! 428: exit_cleanup(RERR_STREAMIO);
! 429: }
! 430: } while (tx_strm.avail_in != 0);
1.1 misho 431: } while (toklen > 0);
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: {
582: int r;
583: int32 blklen;
584: unsigned char hdr[5];
585:
586: rx_strm.avail_in = 0;
587: blklen = 0;
588: hdr[0] = 0;
589: do {
590: if (rx_strm.avail_in == 0 && len != 0) {
591: if (blklen == 0) {
592: /* Give it a fake stored-block header. */
593: rx_strm.next_in = (Bytef *)hdr;
594: rx_strm.avail_in = 5;
595: blklen = len;
596: if (blklen > 0xffff)
597: blklen = 0xffff;
598: hdr[1] = blklen;
599: hdr[2] = blklen >> 8;
600: hdr[3] = ~hdr[1];
601: hdr[4] = ~hdr[2];
602: } else {
603: rx_strm.next_in = (Bytef *)buf;
604: rx_strm.avail_in = blklen;
1.1.1.2 ! misho 605: if (protocol_version >= 31) /* Newer protocols avoid a data-duplicating bug */
! 606: buf += blklen;
1.1 misho 607: len -= blklen;
608: blklen = 0;
609: }
610: }
611: rx_strm.next_out = (Bytef *)dbuf;
612: rx_strm.avail_out = AVAIL_OUT_SIZE(CHUNK_SIZE);
613: r = inflate(&rx_strm, Z_SYNC_FLUSH);
614: if (r != Z_OK && r != Z_BUF_ERROR) {
615: rprintf(FERROR, "inflate (token) returned %d\n", r);
616: exit_cleanup(RERR_STREAMIO);
617: }
618: } while (len || rx_strm.avail_out == 0);
619: }
620:
621: /**
622: * Transmit a verbatim buffer of length @p n followed by a token.
623: * If token == -1 then we have reached EOF
624: * If n == 0 then don't send a buffer
625: */
626: void send_token(int f, int32 token, struct map_struct *buf, OFF_T offset,
627: int32 n, int32 toklen)
628: {
629: if (!do_compression)
630: simple_send_token(f, token, buf, offset, n);
631: else
632: send_deflated_token(f, token, buf, offset, n, toklen);
633: }
634:
635: /*
636: * receive a token or buffer from the other end. If the reurn value is >0 then
637: * it is a data buffer of that length, and *data will point at the data.
638: * if the return value is -i then it represents token i-1
639: * if the return value is 0 then the end has been reached
640: */
641: int32 recv_token(int f, char **data)
642: {
643: int tok;
644:
645: if (!do_compression) {
646: tok = simple_recv_token(f,data);
647: } else {
648: tok = recv_deflated_token(f, data);
649: }
650: return tok;
651: }
652:
653: /*
654: * look at the data corresponding to a token, if necessary
655: */
656: void see_token(char *data, int32 toklen)
657: {
658: if (do_compression)
659: see_deflate_token(data, toklen);
660: }
FreeBSD-CVSweb <freebsd-cvsweb@FreeBSD.org>