Annotation of embedaddon/libxml2/xzlib.c, revision 1.1.1.1
1.1 misho 1: /**
2: * xzlib.c: front end for the transparent suport of lzma compression
3: * at the I/O layer, based on an example file from lzma project
4: *
5: * See Copyright for the status of this software.
6: *
7: * Anders F Bjorklund <afb@users.sourceforge.net>
8: */
9: #define IN_LIBXML
10: #include "libxml.h"
11: #ifdef HAVE_LZMA_H
12:
13: #include <string.h>
14: #ifdef HAVE_ERRNO_H
15: #include <errno.h>
16: #endif
17:
18:
19: #ifdef HAVE_SYS_TYPES_H
20: #include <sys/types.h>
21: #endif
22: #ifdef HAVE_SYS_STAT_H
23: #include <sys/stat.h>
24: #endif
25: #ifdef HAVE_FCNTL_H
26: #include <fcntl.h>
27: #endif
28: #ifdef HAVE_UNISTD_H
29: #include <unistd.h>
30: #endif
31: #ifdef HAVE_STDLIB_H
32: #include <stdlib.h>
33: #endif
34: #ifdef HAVE_ZLIB_H
35: #include <zlib.h>
36: #endif
37: #include <lzma.h>
38:
39: #include "xzlib.h"
40: #include <libxml/xmlmemory.h>
41:
42: /* values for xz_state how */
43: #define LOOK 0 /* look for a gzip/lzma header */
44: #define COPY 1 /* copy input directly */
45: #define GZIP 2 /* decompress a gzip stream */
46: #define LZMA 3 /* decompress a lzma stream */
47:
48: /* internal lzma file state data structure */
49: typedef struct {
50: int mode; /* see lzma modes above */
51: int fd; /* file descriptor */
52: char *path; /* path or fd for error messages */
53: uint64_t pos; /* current position in uncompressed data */
54: unsigned int size; /* buffer size, zero if not allocated yet */
55: unsigned int want; /* requested buffer size, default is BUFSIZ */
56: unsigned char *in; /* input buffer */
57: unsigned char *out; /* output buffer (double-sized when reading) */
58: unsigned char *next; /* next output data to deliver or write */
59: unsigned int have; /* amount of output data unused at next */
60: int eof; /* true if end of input file reached */
61: uint64_t start; /* where the lzma data started, for rewinding */
62: uint64_t raw; /* where the raw data started, for seeking */
63: int how; /* 0: get header, 1: copy, 2: decompress */
64: int direct; /* true if last read direct, false if lzma */
65: /* seek request */
66: uint64_t skip; /* amount to skip (already rewound if backwards) */
67: int seek; /* true if seek request pending */
68: /* error information */
69: int err; /* error code */
70: char *msg; /* error message */
71: /* lzma stream */
72: int init; /* is the iniflate stream initialized */
73: lzma_stream strm; /* stream structure in-place (not a pointer) */
74: char padding1[32]; /* padding allowing to cope with possible
75: extensions of above structure without
76: too much side effect */
77: #ifdef HAVE_ZLIB_H
78: /* zlib inflate or deflate stream */
79: z_stream zstrm; /* stream structure in-place (not a pointer) */
80: #endif
81: char padding2[32]; /* padding allowing to cope with possible
82: extensions of above structure without
83: too much side effect */
84: } xz_state, *xz_statep;
85:
86: static void
87: xz_error(xz_statep state, int err, const char *msg)
88: {
89: /* free previously allocated message and clear */
90: if (state->msg != NULL) {
91: if (state->err != LZMA_MEM_ERROR)
92: xmlFree(state->msg);
93: state->msg = NULL;
94: }
95:
96: /* set error code, and if no message, then done */
97: state->err = err;
98: if (msg == NULL)
99: return;
100:
101: /* for an out of memory error, save as static string */
102: if (err == LZMA_MEM_ERROR) {
103: state->msg = (char *) msg;
104: return;
105: }
106:
107: /* construct error message with path */
108: if ((state->msg =
109: xmlMalloc(strlen(state->path) + strlen(msg) + 3)) == NULL) {
110: state->err = LZMA_MEM_ERROR;
111: state->msg = (char *) "out of memory";
112: return;
113: }
114: strcpy(state->msg, state->path);
115: strcat(state->msg, ": ");
116: strcat(state->msg, msg);
117: return;
118: }
119:
120: static void
121: xz_reset(xz_statep state)
122: {
123: state->have = 0; /* no output data available */
124: state->eof = 0; /* not at end of file */
125: state->how = LOOK; /* look for gzip header */
126: state->direct = 1; /* default for empty file */
127: state->seek = 0; /* no seek request pending */
128: xz_error(state, LZMA_OK, NULL); /* clear error */
129: state->pos = 0; /* no uncompressed data yet */
130: state->strm.avail_in = 0; /* no input data yet */
131: #ifdef HAVE_ZLIB_H
132: state->zstrm.avail_in = 0; /* no input data yet */
133: #endif
134: }
135:
136: static xzFile
137: xz_open(const char *path, int fd, const char *mode ATTRIBUTE_UNUSED)
138: {
139: xz_statep state;
140:
141: /* allocate xzFile structure to return */
142: state = xmlMalloc(sizeof(xz_state));
143: if (state == NULL)
144: return NULL;
145: state->size = 0; /* no buffers allocated yet */
146: state->want = BUFSIZ; /* requested buffer size */
147: state->msg = NULL; /* no error message yet */
148: state->init = 0; /* initialization of zlib data */
149:
150: /* save the path name for error messages */
151: state->path = xmlMalloc(strlen(path) + 1);
152: if (state->path == NULL) {
153: xmlFree(state);
154: return NULL;
155: }
156: strcpy(state->path, path);
157:
158: /* open the file with the appropriate mode (or just use fd) */
159: state->fd = fd != -1 ? fd : open(path,
160: #ifdef O_LARGEFILE
161: O_LARGEFILE |
162: #endif
163: #ifdef O_BINARY
164: O_BINARY |
165: #endif
166: O_RDONLY, 0666);
167: if (state->fd == -1) {
168: xmlFree(state->path);
169: xmlFree(state);
170: return NULL;
171: }
172:
173: /* save the current position for rewinding (only if reading) */
174: state->start = lseek(state->fd, 0, SEEK_CUR);
175: if (state->start == (uint64_t) - 1)
176: state->start = 0;
177:
178: /* initialize stream */
179: xz_reset(state);
180:
181: /* return stream */
182: return (xzFile) state;
183: }
184:
185: xzFile
186: __libxml2_xzopen(const char *path, const char *mode)
187: {
188: return xz_open(path, -1, mode);
189: }
190:
191: xzFile
192: __libxml2_xzdopen(int fd, const char *mode)
193: {
194: char *path; /* identifier for error messages */
195: xzFile xz;
196:
197: if (fd == -1 || (path = xmlMalloc(7 + 3 * sizeof(int))) == NULL)
198: return NULL;
199: sprintf(path, "<fd:%d>", fd); /* for debugging */
200: xz = xz_open(path, fd, mode);
201: xmlFree(path);
202: return xz;
203: }
204:
205: static int
206: xz_load(xz_statep state, unsigned char *buf, unsigned int len,
207: unsigned int *have)
208: {
209: int ret;
210:
211: *have = 0;
212: do {
213: ret = read(state->fd, buf + *have, len - *have);
214: if (ret <= 0)
215: break;
216: *have += ret;
217: } while (*have < len);
218: if (ret < 0) {
219: xz_error(state, -1, strerror(errno));
220: return -1;
221: }
222: if (ret == 0)
223: state->eof = 1;
224: return 0;
225: }
226:
227: static int
228: xz_avail(xz_statep state)
229: {
230: lzma_stream *strm = &(state->strm);
231:
232: if (state->err != LZMA_OK)
233: return -1;
234: if (state->eof == 0) {
235: /* avail_in is size_t, which is not necessary sizeof(unsigned) */
236: unsigned tmp = strm->avail_in;
237:
238: if (xz_load(state, state->in, state->size, &tmp) == -1) {
239: strm->avail_in = tmp;
240: return -1;
241: }
242: strm->avail_in = tmp;
243: strm->next_in = state->in;
244: }
245: return 0;
246: }
247:
248: static int
249: is_format_xz(xz_statep state)
250: {
251: lzma_stream *strm = &(state->strm);
252:
253: return strm->avail_in >= 6 && memcmp(state->in, "\3757zXZ", 6) == 0;
254: }
255:
256: static int
257: is_format_lzma(xz_statep state)
258: {
259: lzma_stream *strm = &(state->strm);
260:
261: lzma_filter filter;
262: lzma_options_lzma *opt;
263: uint32_t dict_size;
264: uint64_t uncompressed_size;
265: size_t i;
266:
267: if (strm->avail_in < 13)
268: return 0;
269:
270: filter.id = LZMA_FILTER_LZMA1;
271: if (lzma_properties_decode(&filter, NULL, state->in, 5) != LZMA_OK)
272: return 0;
273:
274: opt = filter.options;
275: dict_size = opt->dict_size;
276: free(opt); /* we can't use xmlFree on a string returned by zlib */
277:
278: /* A hack to ditch tons of false positives: We allow only dictionary
279: * sizes that are 2^n or 2^n + 2^(n-1) or UINT32_MAX. LZMA_Alone
280: * created only files with 2^n, but accepts any dictionary size.
281: * If someone complains, this will be reconsidered.
282: */
283: if (dict_size != UINT32_MAX) {
284: uint32_t d = dict_size - 1;
285:
286: d |= d >> 2;
287: d |= d >> 3;
288: d |= d >> 4;
289: d |= d >> 8;
290: d |= d >> 16;
291: ++d;
292: if (d != dict_size || dict_size == 0)
293: return 0;
294: }
295:
296: /* Another hack to ditch false positives: Assume that if the
297: * uncompressed size is known, it must be less than 256 GiB.
298: * Again, if someone complains, this will be reconsidered.
299: */
300: uncompressed_size = 0;
301: for (i = 0; i < 8; ++i)
302: uncompressed_size |= (uint64_t) (state->in[5 + i]) << (i * 8);
303:
304: if (uncompressed_size != UINT64_MAX
305: && uncompressed_size > (UINT64_C(1) << 38))
306: return 0;
307:
308: return 1;
309: }
310:
311: #ifdef HAVE_ZLIB_H
312:
313: /* Get next byte from input, or -1 if end or error. */
314: #define NEXT() ((strm->avail_in == 0 && xz_avail(state) == -1) ? -1 : \
315: (strm->avail_in == 0 ? -1 : \
316: (strm->avail_in--, *(strm->next_in)++)))
317:
318: /* Get a four-byte little-endian integer and return 0 on success and the value
319: in *ret. Otherwise -1 is returned and *ret is not modified. */
320: static int
321: gz_next4(xz_statep state, unsigned long *ret)
322: {
323: int ch;
324: unsigned long val;
325: z_streamp strm = &(state->zstrm);
326:
327: val = NEXT();
328: val += (unsigned) NEXT() << 8;
329: val += (unsigned long) NEXT() << 16;
330: ch = NEXT();
331: if (ch == -1)
332: return -1;
333: val += (unsigned long) ch << 24;
334: *ret = val;
335: return 0;
336: }
337: #endif
338:
339: static int
340: xz_head(xz_statep state)
341: {
342: lzma_stream *strm = &(state->strm);
343: lzma_stream init = LZMA_STREAM_INIT;
344: int flags;
345: unsigned len;
346:
347: /* allocate read buffers and inflate memory */
348: if (state->size == 0) {
349: /* allocate buffers */
350: state->in = xmlMalloc(state->want);
351: state->out = xmlMalloc(state->want << 1);
352: if (state->in == NULL || state->out == NULL) {
353: if (state->out != NULL)
354: xmlFree(state->out);
355: if (state->in != NULL)
356: xmlFree(state->in);
357: xz_error(state, LZMA_MEM_ERROR, "out of memory");
358: return -1;
359: }
360: state->size = state->want;
361:
362: /* allocate decoder memory */
363: state->strm = init;
364: state->strm.avail_in = 0;
365: state->strm.next_in = NULL;
366: if (lzma_auto_decoder(&state->strm, UINT64_MAX, 0) != LZMA_OK) {
367: xmlFree(state->out);
368: xmlFree(state->in);
369: state->size = 0;
370: xz_error(state, LZMA_MEM_ERROR, "out of memory");
371: return -1;
372: }
373: #ifdef HAVE_ZLIB_H
374: /* allocate inflate memory */
375: state->zstrm.zalloc = Z_NULL;
376: state->zstrm.zfree = Z_NULL;
377: state->zstrm.opaque = Z_NULL;
378: state->zstrm.avail_in = 0;
379: state->zstrm.next_in = Z_NULL;
380: if (state->init == 0) {
381: if (inflateInit2(&(state->zstrm), -15) != Z_OK) {/* raw inflate */
382: xmlFree(state->out);
383: xmlFree(state->in);
384: state->size = 0;
385: xz_error(state, LZMA_MEM_ERROR, "out of memory");
386: return -1;
387: }
388: state->init = 1;
389: }
390: #endif
391: }
392:
393: /* get some data in the input buffer */
394: if (strm->avail_in == 0) {
395: if (xz_avail(state) == -1)
396: return -1;
397: if (strm->avail_in == 0)
398: return 0;
399: }
400:
401: /* look for the xz magic header bytes */
402: if (is_format_xz(state) || is_format_lzma(state)) {
403: state->how = LZMA;
404: state->direct = 0;
405: return 0;
406: }
407: #ifdef HAVE_ZLIB_H
408: /* look for the gzip magic header bytes 31 and 139 */
409: if (strm->next_in[0] == 31) {
410: strm->avail_in--;
411: strm->next_in++;
412: if (strm->avail_in == 0 && xz_avail(state) == -1)
413: return -1;
414: if (strm->avail_in && strm->next_in[0] == 139) {
415: /* we have a gzip header, woo hoo! */
416: strm->avail_in--;
417: strm->next_in++;
418:
419: /* skip rest of header */
420: if (NEXT() != 8) { /* compression method */
421: xz_error(state, LZMA_DATA_ERROR,
422: "unknown compression method");
423: return -1;
424: }
425: flags = NEXT();
426: if (flags & 0xe0) { /* reserved flag bits */
427: xz_error(state, LZMA_DATA_ERROR,
428: "unknown header flags set");
429: return -1;
430: }
431: NEXT(); /* modification time */
432: NEXT();
433: NEXT();
434: NEXT();
435: NEXT(); /* extra flags */
436: NEXT(); /* operating system */
437: if (flags & 4) { /* extra field */
438: len = (unsigned) NEXT();
439: len += (unsigned) NEXT() << 8;
440: while (len--)
441: if (NEXT() < 0)
442: break;
443: }
444: if (flags & 8) /* file name */
445: while (NEXT() > 0) ;
446: if (flags & 16) /* comment */
447: while (NEXT() > 0) ;
448: if (flags & 2) { /* header crc */
449: NEXT();
450: NEXT();
451: }
452: /* an unexpected end of file is not checked for here -- it will be
453: * noticed on the first request for uncompressed data */
454:
455: /* set up for decompression */
456: inflateReset(&state->zstrm);
457: state->zstrm.adler = crc32(0L, Z_NULL, 0);
458: state->how = GZIP;
459: state->direct = 0;
460: return 0;
461: } else {
462: /* not a gzip file -- save first byte (31) and fall to raw i/o */
463: state->out[0] = 31;
464: state->have = 1;
465: }
466: }
467: #endif
468:
469: /* doing raw i/o, save start of raw data for seeking, copy any leftover
470: * input to output -- this assumes that the output buffer is larger than
471: * the input buffer, which also assures space for gzungetc() */
472: state->raw = state->pos;
473: state->next = state->out;
474: if (strm->avail_in) {
475: memcpy(state->next + state->have, strm->next_in, strm->avail_in);
476: state->have += strm->avail_in;
477: strm->avail_in = 0;
478: }
479: state->how = COPY;
480: state->direct = 1;
481: return 0;
482: }
483:
484: static int
485: xz_decomp(xz_statep state)
486: {
487: int ret;
488: unsigned had;
489: unsigned long crc, len;
490: lzma_stream *strm = &(state->strm);
491:
492: lzma_action action = LZMA_RUN;
493:
494: /* fill output buffer up to end of deflate stream */
495: had = strm->avail_out;
496: do {
497: /* get more input for inflate() */
498: if (strm->avail_in == 0 && xz_avail(state) == -1)
499: return -1;
500: if (strm->avail_in == 0) {
501: xz_error(state, LZMA_DATA_ERROR, "unexpected end of file");
502: return -1;
503: }
504: if (state->eof)
505: action = LZMA_FINISH;
506:
507: /* decompress and handle errors */
508: #ifdef HAVE_ZLIB_H
509: if (state->how == GZIP) {
510: state->zstrm.avail_in = (uInt) state->strm.avail_in;
511: state->zstrm.next_in = (Bytef *) state->strm.next_in;
512: state->zstrm.avail_out = (uInt) state->strm.avail_out;
513: state->zstrm.next_out = (Bytef *) state->strm.next_out;
514: ret = inflate(&state->zstrm, Z_NO_FLUSH);
515: if (ret == Z_STREAM_ERROR || ret == Z_NEED_DICT) {
516: xz_error(state, Z_STREAM_ERROR,
517: "internal error: inflate stream corrupt");
518: return -1;
519: }
520: if (ret == Z_MEM_ERROR)
521: ret = LZMA_MEM_ERROR;
522: if (ret == Z_DATA_ERROR)
523: ret = LZMA_DATA_ERROR;
524: if (ret == Z_STREAM_END)
525: ret = LZMA_STREAM_END;
526: state->strm.avail_in = state->zstrm.avail_in;
527: state->strm.next_in = state->zstrm.next_in;
528: state->strm.avail_out = state->zstrm.avail_out;
529: state->strm.next_out = state->zstrm.next_out;
530: } else /* state->how == LZMA */
531: #endif
532: ret = lzma_code(strm, action);
533: if (ret == LZMA_MEM_ERROR) {
534: xz_error(state, LZMA_MEM_ERROR, "out of memory");
535: return -1;
536: }
537: if (ret == LZMA_DATA_ERROR) {
538: xz_error(state, LZMA_DATA_ERROR, "compressed data error");
539: return -1;
540: }
541: } while (strm->avail_out && ret != LZMA_STREAM_END);
542:
543: /* update available output and crc check value */
544: state->have = had - strm->avail_out;
545: state->next = strm->next_out - state->have;
546: #ifdef HAVE_ZLIB_H
547: state->zstrm.adler =
548: crc32(state->zstrm.adler, state->next, state->have);
549: #endif
550:
551: if (ret == LZMA_STREAM_END) {
552: #ifdef HAVE_ZLIB_H
553: if (state->how == GZIP) {
554: if (gz_next4(state, &crc) == -1 || gz_next4(state, &len) == -1) {
555: xz_error(state, LZMA_DATA_ERROR, "unexpected end of file");
556: return -1;
557: }
558: if (crc != state->zstrm.adler) {
559: xz_error(state, LZMA_DATA_ERROR, "incorrect data check");
560: return -1;
561: }
562: if (len != (state->zstrm.total_out & 0xffffffffL)) {
563: xz_error(state, LZMA_DATA_ERROR, "incorrect length check");
564: return -1;
565: }
566: state->strm.avail_in = 0;
567: state->strm.next_in = NULL;
568: state->strm.avail_out = 0;
569: state->strm.next_out = NULL;
570: } else
571: #endif
572: if (strm->avail_in != 0 || !state->eof) {
573: xz_error(state, LZMA_DATA_ERROR, "trailing garbage");
574: return -1;
575: }
576: state->how = LOOK; /* ready for next stream, once have is 0 (leave
577: * state->direct unchanged to remember how) */
578: }
579:
580: /* good decompression */
581: return 0;
582: }
583:
584: static int
585: xz_make(xz_statep state)
586: {
587: lzma_stream *strm = &(state->strm);
588:
589: if (state->how == LOOK) { /* look for lzma / gzip header */
590: if (xz_head(state) == -1)
591: return -1;
592: if (state->have) /* got some data from xz_head() */
593: return 0;
594: }
595: if (state->how == COPY) { /* straight copy */
596: if (xz_load(state, state->out, state->size << 1, &(state->have)) ==
597: -1)
598: return -1;
599: state->next = state->out;
600: } else if (state->how == LZMA || state->how == GZIP) { /* decompress */
601: strm->avail_out = state->size << 1;
602: strm->next_out = state->out;
603: if (xz_decomp(state) == -1)
604: return -1;
605: }
606: return 0;
607: }
608:
609: static int
610: xz_skip(xz_statep state, uint64_t len)
611: {
612: unsigned n;
613:
614: /* skip over len bytes or reach end-of-file, whichever comes first */
615: while (len)
616: /* skip over whatever is in output buffer */
617: if (state->have) {
618: n = (uint64_t) state->have > len ?
619: (unsigned) len : state->have;
620: state->have -= n;
621: state->next += n;
622: state->pos += n;
623: len -= n;
624: }
625:
626: /* output buffer empty -- return if we're at the end of the input */
627: else if (state->eof && state->strm.avail_in == 0)
628: break;
629:
630: /* need more data to skip -- load up output buffer */
631: else {
632: /* get more output, looking for header if required */
633: if (xz_make(state) == -1)
634: return -1;
635: }
636: return 0;
637: }
638:
639: int
640: __libxml2_xzread(xzFile file, void *buf, unsigned len)
641: {
642: unsigned got, n;
643: xz_statep state;
644: lzma_stream *strm;
645:
646: /* get internal structure */
647: if (file == NULL)
648: return -1;
649: state = (xz_statep) file;
650: strm = &(state->strm);
651:
652: /* check that we're reading and that there's no error */
653: if (state->err != LZMA_OK)
654: return -1;
655:
656: /* since an int is returned, make sure len fits in one, otherwise return
657: * with an error (this avoids the flaw in the interface) */
658: if ((int) len < 0) {
659: xz_error(state, LZMA_BUF_ERROR,
660: "requested length does not fit in int");
661: return -1;
662: }
663:
664: /* if len is zero, avoid unnecessary operations */
665: if (len == 0)
666: return 0;
667:
668: /* process a skip request */
669: if (state->seek) {
670: state->seek = 0;
671: if (xz_skip(state, state->skip) == -1)
672: return -1;
673: }
674:
675: /* get len bytes to buf, or less than len if at the end */
676: got = 0;
677: do {
678: /* first just try copying data from the output buffer */
679: if (state->have) {
680: n = state->have > len ? len : state->have;
681: memcpy(buf, state->next, n);
682: state->next += n;
683: state->have -= n;
684: }
685:
686: /* output buffer empty -- return if we're at the end of the input */
687: else if (state->eof && strm->avail_in == 0)
688: break;
689:
690: /* need output data -- for small len or new stream load up our output
691: * buffer */
692: else if (state->how == LOOK || len < (state->size << 1)) {
693: /* get more output, looking for header if required */
694: if (xz_make(state) == -1)
695: return -1;
696: continue; /* no progress yet -- go back to memcpy() above */
697: /* the copy above assures that we will leave with space in the
698: * output buffer, allowing at least one gzungetc() to succeed */
699: }
700:
701: /* large len -- read directly into user buffer */
702: else if (state->how == COPY) { /* read directly */
703: if (xz_load(state, buf, len, &n) == -1)
704: return -1;
705: }
706:
707: /* large len -- decompress directly into user buffer */
708: else { /* state->how == LZMA */
709: strm->avail_out = len;
710: strm->next_out = buf;
711: if (xz_decomp(state) == -1)
712: return -1;
713: n = state->have;
714: state->have = 0;
715: }
716:
717: /* update progress */
718: len -= n;
719: buf = (char *) buf + n;
720: got += n;
721: state->pos += n;
722: } while (len);
723:
724: /* return number of bytes read into user buffer (will fit in int) */
725: return (int) got;
726: }
727:
728: int
729: __libxml2_xzclose(xzFile file)
730: {
731: int ret;
732: xz_statep state;
733:
734: /* get internal structure */
735: if (file == NULL)
736: return LZMA_DATA_ERROR;
737: state = (xz_statep) file;
738:
739: /* free memory and close file */
740: if (state->size) {
741: lzma_end(&(state->strm));
742: #ifdef HAVE_ZLIB_H
743: if (state->init == 1)
744: inflateEnd(&(state->zstrm));
745: state->init = 0;
746: #endif
747: xmlFree(state->out);
748: xmlFree(state->in);
749: }
750: xmlFree(state->path);
751: ret = close(state->fd);
752: xmlFree(state);
753: return ret ? ret : LZMA_OK;
754: }
755: #endif /* HAVE_LZMA_H */
FreeBSD-CVSweb <freebsd-cvsweb@FreeBSD.org>