xzlib.c

00001 /**
00002  * xzlib.c: front end for the transparent suport of lzma compression
00003  *          at the I/O layer, based on an example file from lzma project
00004  *
00005  * See Copyright for the status of this software.
00006  *
00007  * Anders F Bjorklund <afb@users.sourceforge.net>
00008  */
00009 #define IN_LIBXML
00010 #include "libxml.h"
00011 #ifdef LIBXML_LZMA_ENABLED
00012 
00013 #include <string.h>
00014 #ifdef HAVE_ERRNO_H
00015 #include <errno.h>
00016 #endif
00017 
00018 
00019 #ifdef HAVE_SYS_TYPES_H
00020 #include <sys/types.h>
00021 #endif
00022 #ifdef HAVE_SYS_STAT_H
00023 #include <sys/stat.h>
00024 #endif
00025 #ifdef HAVE_FCNTL_H
00026 #include <fcntl.h>
00027 #endif
00028 #ifdef HAVE_UNISTD_H
00029 #include <unistd.h>
00030 #endif
00031 #ifdef HAVE_STDLIB_H
00032 #include <stdlib.h>
00033 #endif
00034 #ifdef HAVE_ZLIB_H
00035 #include <zlib.h>
00036 #endif
00037 #ifdef HAVE_LZMA_H
00038 #include <lzma.h>
00039 #endif
00040 
00041 #include "xzlib.h"
00042 #include <libxml/xmlmemory.h>
00043 
00044 /* values for xz_state how */
00045 #define LOOK 0                  /* look for a gzip/lzma header */
00046 #define COPY 1                  /* copy input directly */
00047 #define GZIP 2                  /* decompress a gzip stream */
00048 #define LZMA 3                  /* decompress a lzma stream */
00049 
00050 /* internal lzma file state data structure */
00051 typedef struct {
00052     int mode;                   /* see lzma modes above */
00053     int fd;                     /* file descriptor */
00054     char *path;                 /* path or fd for error messages */
00055     uint64_t pos;               /* current position in uncompressed data */
00056     unsigned int size;          /* buffer size, zero if not allocated yet */
00057     unsigned int want;          /* requested buffer size, default is BUFSIZ */
00058     unsigned char *in;          /* input buffer */
00059     unsigned char *out;         /* output buffer (double-sized when reading) */
00060     unsigned char *next;        /* next output data to deliver or write */
00061     unsigned int have;          /* amount of output data unused at next */
00062     int eof;                    /* true if end of input file reached */
00063     uint64_t start;             /* where the lzma data started, for rewinding */
00064     uint64_t raw;               /* where the raw data started, for seeking */
00065     int how;                    /* 0: get header, 1: copy, 2: decompress */
00066     int direct;                 /* true if last read direct, false if lzma */
00067     /* seek request */
00068     uint64_t skip;              /* amount to skip (already rewound if backwards) */
00069     int seek;                   /* true if seek request pending */
00070     /* error information */
00071     int err;                    /* error code */
00072     char *msg;                  /* error message */
00073     /* lzma stream */
00074     int init;                   /* is the iniflate stream initialized */
00075     lzma_stream strm;           /* stream structure in-place (not a pointer) */
00076     char padding1[32];          /* padding allowing to cope with possible
00077                                    extensions of above structure without
00078                    too much side effect */
00079 #ifdef HAVE_ZLIB_H
00080     /* zlib inflate or deflate stream */
00081     z_stream zstrm;             /* stream structure in-place (not a pointer) */
00082 #endif
00083     char padding2[32];          /* padding allowing to cope with possible
00084                                    extensions of above structure without
00085                    too much side effect */
00086 } xz_state, *xz_statep;
00087 
00088 static void
00089 xz_error(xz_statep state, int err, const char *msg)
00090 {
00091     /* free previously allocated message and clear */
00092     if (state->msg != NULL) {
00093         if (state->err != LZMA_MEM_ERROR)
00094             xmlFree(state->msg);
00095         state->msg = NULL;
00096     }
00097 
00098     /* set error code, and if no message, then done */
00099     state->err = err;
00100     if (msg == NULL)
00101         return;
00102 
00103     /* for an out of memory error, save as static string */
00104     if (err == LZMA_MEM_ERROR) {
00105         state->msg = (char *) msg;
00106         return;
00107     }
00108 
00109     /* construct error message with path */
00110     if ((state->msg =
00111          xmlMalloc(strlen(state->path) + strlen(msg) + 3)) == NULL) {
00112         state->err = LZMA_MEM_ERROR;
00113         state->msg = (char *) "out of memory";
00114         return;
00115     }
00116     strcpy(state->msg, state->path);
00117     strcat(state->msg, ": ");
00118     strcat(state->msg, msg);
00119     return;
00120 }
00121 
00122 static void
00123 xz_reset(xz_statep state)
00124 {
00125     state->have = 0;            /* no output data available */
00126     state->eof = 0;             /* not at end of file */
00127     state->how = LOOK;          /* look for gzip header */
00128     state->direct = 1;          /* default for empty file */
00129     state->seek = 0;            /* no seek request pending */
00130     xz_error(state, LZMA_OK, NULL);     /* clear error */
00131     state->pos = 0;             /* no uncompressed data yet */
00132     state->strm.avail_in = 0;   /* no input data yet */
00133 #ifdef HAVE_ZLIB_H
00134     state->zstrm.avail_in = 0;  /* no input data yet */
00135 #endif
00136 }
00137 
00138 static xzFile
00139 xz_open(const char *path, int fd, const char *mode ATTRIBUTE_UNUSED)
00140 {
00141     xz_statep state;
00142 
00143     /* allocate xzFile structure to return */
00144     state = xmlMalloc(sizeof(xz_state));
00145     if (state == NULL)
00146         return NULL;
00147     state->size = 0;            /* no buffers allocated yet */
00148     state->want = BUFSIZ;       /* requested buffer size */
00149     state->msg = NULL;          /* no error message yet */
00150     state->init = 0;            /* initialization of zlib data */
00151 
00152     /* save the path name for error messages */
00153     state->path = xmlMalloc(strlen(path) + 1);
00154     if (state->path == NULL) {
00155         xmlFree(state);
00156         return NULL;
00157     }
00158     strcpy(state->path, path);
00159 
00160     /* open the file with the appropriate mode (or just use fd) */
00161     state->fd = fd != -1 ? fd : open(path,
00162 #ifdef O_LARGEFILE
00163                                      O_LARGEFILE |
00164 #endif
00165 #ifdef O_BINARY
00166                                      O_BINARY |
00167 #endif
00168                                      O_RDONLY, 0666);
00169     if (state->fd == -1) {
00170         xmlFree(state->path);
00171         xmlFree(state);
00172         return NULL;
00173     }
00174 
00175     /* save the current position for rewinding (only if reading) */
00176     state->start = lseek(state->fd, 0, SEEK_CUR);
00177     if (state->start == (uint64_t) - 1)
00178         state->start = 0;
00179 
00180     /* initialize stream */
00181     xz_reset(state);
00182 
00183     /* return stream */
00184     return (xzFile) state;
00185 }
00186 
00187 static int
00188 xz_compressed(xzFile f) {
00189     xz_statep state;
00190 
00191     if (f == NULL)
00192         return(-1);
00193     state = (xz_statep) f;
00194     if (state->init <= 0)
00195         return(-1);
00196 
00197     switch (state->how) {
00198         case COPY:
00199         return(0);
00200     case GZIP:
00201     case LZMA:
00202         return(1);
00203     }
00204     return(-1);
00205 }
00206 
00207 xzFile
00208 __libxml2_xzopen(const char *path, const char *mode)
00209 {
00210     return xz_open(path, -1, mode);
00211 }
00212 
00213 int
00214 __libxml2_xzcompressed(xzFile f) {
00215     return xz_compressed(f);
00216 }
00217 
00218 xzFile
00219 __libxml2_xzdopen(int fd, const char *mode)
00220 {
00221     char *path;                 /* identifier for error messages */
00222     xzFile xz;
00223 
00224     if (fd == -1 || (path = xmlMalloc(7 + 3 * sizeof(int))) == NULL)
00225         return NULL;
00226     sprintf(path, "<fd:%d>", fd);       /* for debugging */
00227     xz = xz_open(path, fd, mode);
00228     xmlFree(path);
00229     return xz;
00230 }
00231 
00232 static int
00233 xz_load(xz_statep state, unsigned char *buf, unsigned int len,
00234         unsigned int *have)
00235 {
00236     int ret;
00237 
00238     *have = 0;
00239     do {
00240         ret = read(state->fd, buf + *have, len - *have);
00241         if (ret <= 0)
00242             break;
00243         *have += ret;
00244     } while (*have < len);
00245     if (ret < 0) {
00246         xz_error(state, -1, strerror(errno));
00247         return -1;
00248     }
00249     if (ret == 0)
00250         state->eof = 1;
00251     return 0;
00252 }
00253 
00254 static int
00255 xz_avail(xz_statep state)
00256 {
00257     lzma_stream *strm = &(state->strm);
00258 
00259     if (state->err != LZMA_OK)
00260         return -1;
00261     if (state->eof == 0) {
00262         /* avail_in is size_t, which is not necessary sizeof(unsigned) */
00263         unsigned tmp = strm->avail_in;
00264 
00265         if (xz_load(state, state->in, state->size, &tmp) == -1) {
00266             strm->avail_in = tmp;
00267             return -1;
00268         }
00269         strm->avail_in = tmp;
00270         strm->next_in = state->in;
00271     }
00272     return 0;
00273 }
00274 
00275 #ifdef HAVE_ZLIB_H
00276 static int
00277 xz_avail_zstrm(xz_statep state)
00278 {
00279     int ret;
00280     state->strm.avail_in = state->zstrm.avail_in;
00281     state->strm.next_in = state->zstrm.next_in;
00282     ret = xz_avail(state);
00283     state->zstrm.avail_in = (uInt) state->strm.avail_in;
00284     state->zstrm.next_in = (Bytef *) state->strm.next_in;
00285     return ret;
00286 }
00287 #endif
00288 
00289 static int
00290 is_format_xz(xz_statep state)
00291 {
00292     lzma_stream *strm = &(state->strm);
00293 
00294     return strm->avail_in >= 6 && memcmp(state->in, "\3757zXZ", 6) == 0;
00295 }
00296 
00297 static int
00298 is_format_lzma(xz_statep state)
00299 {
00300     lzma_stream *strm = &(state->strm);
00301 
00302     lzma_filter filter;
00303     lzma_options_lzma *opt;
00304     uint32_t dict_size;
00305     uint64_t uncompressed_size;
00306     size_t i;
00307 
00308     if (strm->avail_in < 13)
00309         return 0;
00310 
00311     filter.id = LZMA_FILTER_LZMA1;
00312     if (lzma_properties_decode(&filter, NULL, state->in, 5) != LZMA_OK)
00313         return 0;
00314 
00315     opt = filter.options;
00316     dict_size = opt->dict_size;
00317     free(opt); /* we can't use xmlFree on a string returned by zlib */
00318 
00319     /* A hack to ditch tons of false positives: We allow only dictionary
00320      * sizes that are 2^n or 2^n + 2^(n-1) or UINT32_MAX. LZMA_Alone
00321      * created only files with 2^n, but accepts any dictionary size.
00322      * If someone complains, this will be reconsidered.
00323      */
00324     if (dict_size != UINT32_MAX) {
00325         uint32_t d = dict_size - 1;
00326 
00327         d |= d >> 2;
00328         d |= d >> 3;
00329         d |= d >> 4;
00330         d |= d >> 8;
00331         d |= d >> 16;
00332         ++d;
00333         if (d != dict_size || dict_size == 0)
00334             return 0;
00335     }
00336 
00337     /* Another hack to ditch false positives: Assume that if the
00338      * uncompressed size is known, it must be less than 256 GiB.
00339      * Again, if someone complains, this will be reconsidered.
00340      */
00341     uncompressed_size = 0;
00342     for (i = 0; i < 8; ++i)
00343         uncompressed_size |= (uint64_t) (state->in[5 + i]) << (i * 8);
00344 
00345     if (uncompressed_size != UINT64_MAX
00346         && uncompressed_size > (UINT64_C(1) << 38))
00347         return 0;
00348 
00349     return 1;
00350 }
00351 
00352 #ifdef HAVE_ZLIB_H
00353 
00354 /* Get next byte from input, or -1 if end or error. */
00355 #define NEXT() ((strm->avail_in == 0 && xz_avail(state) == -1) ? -1 : \
00356                 (strm->avail_in == 0 ? -1 : \
00357                  (strm->avail_in--, *(strm->next_in)++)))
00358 /* Same thing, but from zstrm */
00359 #define NEXTZ() ((strm->avail_in == 0 && xz_avail_zstrm(state) == -1) ? -1 : \
00360                 (strm->avail_in == 0 ? -1 : \
00361                  (strm->avail_in--, *(strm->next_in)++)))
00362 
00363 /* Get a four-byte little-endian integer and return 0 on success and the value
00364    in *ret.  Otherwise -1 is returned and *ret is not modified. */
00365 static int
00366 gz_next4(xz_statep state, unsigned long *ret)
00367 {
00368     int ch;
00369     unsigned long val;
00370     z_streamp strm = &(state->zstrm);
00371 
00372     val = NEXTZ();
00373     val += (unsigned) NEXTZ() << 8;
00374     val += (unsigned long) NEXTZ() << 16;
00375     ch = NEXTZ();
00376     if (ch == -1)
00377         return -1;
00378     val += (unsigned long) ch << 24;
00379     *ret = val;
00380     return 0;
00381 }
00382 #endif
00383 
00384 static int
00385 xz_head(xz_statep state)
00386 {
00387     lzma_stream *strm = &(state->strm);
00388     lzma_stream init = LZMA_STREAM_INIT;
00389     int flags;
00390     unsigned len;
00391 
00392     /* allocate read buffers and inflate memory */
00393     if (state->size == 0) {
00394         /* allocate buffers */
00395         state->in = xmlMalloc(state->want);
00396         state->out = xmlMalloc(state->want << 1);
00397         if (state->in == NULL || state->out == NULL) {
00398             if (state->out != NULL)
00399                 xmlFree(state->out);
00400             if (state->in != NULL)
00401                 xmlFree(state->in);
00402             xz_error(state, LZMA_MEM_ERROR, "out of memory");
00403             return -1;
00404         }
00405         state->size = state->want;
00406 
00407         /* allocate decoder memory */
00408         state->strm = init;
00409         state->strm.avail_in = 0;
00410         state->strm.next_in = NULL;
00411         if (lzma_auto_decoder(&state->strm, UINT64_MAX, 0) != LZMA_OK) {
00412             xmlFree(state->out);
00413             xmlFree(state->in);
00414             state->size = 0;
00415             xz_error(state, LZMA_MEM_ERROR, "out of memory");
00416             return -1;
00417         }
00418 #ifdef HAVE_ZLIB_H
00419         /* allocate inflate memory */
00420         state->zstrm.zalloc = Z_NULL;
00421         state->zstrm.zfree = Z_NULL;
00422         state->zstrm.opaque = Z_NULL;
00423         state->zstrm.avail_in = 0;
00424         state->zstrm.next_in = Z_NULL;
00425         if (state->init == 0) {
00426             if (inflateInit2(&(state->zstrm), -15) != Z_OK) {/* raw inflate */
00427                 xmlFree(state->out);
00428                 xmlFree(state->in);
00429                 state->size = 0;
00430                 xz_error(state, LZMA_MEM_ERROR, "out of memory");
00431                 return -1;
00432             }
00433             state->init = 1;
00434         }
00435 #endif
00436     }
00437 
00438     /* get some data in the input buffer */
00439     if (strm->avail_in == 0) {
00440         if (xz_avail(state) == -1)
00441             return -1;
00442         if (strm->avail_in == 0)
00443             return 0;
00444     }
00445 
00446     /* look for the xz magic header bytes */
00447     if (is_format_xz(state) || is_format_lzma(state)) {
00448         state->how = LZMA;
00449         state->direct = 0;
00450         return 0;
00451     }
00452 #ifdef HAVE_ZLIB_H
00453     /* look for the gzip magic header bytes 31 and 139 */
00454     if (strm->next_in[0] == 31) {
00455         strm->avail_in--;
00456         strm->next_in++;
00457         if (strm->avail_in == 0 && xz_avail(state) == -1)
00458             return -1;
00459         if (strm->avail_in && strm->next_in[0] == 139) {
00460             /* we have a gzip header, woo hoo! */
00461             strm->avail_in--;
00462             strm->next_in++;
00463 
00464             /* skip rest of header */
00465             if (NEXT() != 8) {  /* compression method */
00466                 xz_error(state, LZMA_DATA_ERROR,
00467                          "unknown compression method");
00468                 return -1;
00469             }
00470             flags = NEXT();
00471             if (flags & 0xe0) { /* reserved flag bits */
00472                 xz_error(state, LZMA_DATA_ERROR,
00473                          "unknown header flags set");
00474                 return -1;
00475             }
00476             NEXT();             /* modification time */
00477             NEXT();
00478             NEXT();
00479             NEXT();
00480             NEXT();             /* extra flags */
00481             NEXT();             /* operating system */
00482             if (flags & 4) {    /* extra field */
00483                 len = (unsigned) NEXT();
00484                 len += (unsigned) NEXT() << 8;
00485                 while (len--)
00486                     if (NEXT() < 0)
00487                         break;
00488             }
00489             if (flags & 8)      /* file name */
00490                 while (NEXT() > 0) ;
00491             if (flags & 16)     /* comment */
00492                 while (NEXT() > 0) ;
00493             if (flags & 2) {    /* header crc */
00494                 NEXT();
00495                 NEXT();
00496             }
00497             /* an unexpected end of file is not checked for here -- it will be
00498              * noticed on the first request for uncompressed data */
00499 
00500             /* set up for decompression */
00501             inflateReset(&state->zstrm);
00502             state->zstrm.adler = crc32(0L, Z_NULL, 0);
00503             state->how = GZIP;
00504             state->direct = 0;
00505             return 0;
00506         } else {
00507             /* not a gzip file -- save first byte (31) and fall to raw i/o */
00508             state->out[0] = 31;
00509             state->have = 1;
00510         }
00511     }
00512 #endif
00513 
00514     /* doing raw i/o, save start of raw data for seeking, copy any leftover
00515      * input to output -- this assumes that the output buffer is larger than
00516      * the input buffer, which also assures space for gzungetc() */
00517     state->raw = state->pos;
00518     state->next = state->out;
00519     if (strm->avail_in) {
00520         memcpy(state->next + state->have, strm->next_in, strm->avail_in);
00521         state->have += strm->avail_in;
00522         strm->avail_in = 0;
00523     }
00524     state->how = COPY;
00525     state->direct = 1;
00526     return 0;
00527 }
00528 
00529 static int
00530 xz_decomp(xz_statep state)
00531 {
00532     int ret;
00533     unsigned had;
00534     unsigned long crc, len;
00535     lzma_stream *strm = &(state->strm);
00536 
00537     lzma_action action = LZMA_RUN;
00538 
00539     /* fill output buffer up to end of deflate stream */
00540     had = strm->avail_out;
00541     do {
00542         /* get more input for inflate() */
00543         if (strm->avail_in == 0 && xz_avail(state) == -1)
00544             return -1;
00545         if (strm->avail_in == 0) {
00546             xz_error(state, LZMA_DATA_ERROR, "unexpected end of file");
00547             return -1;
00548         }
00549         if (state->eof)
00550             action = LZMA_FINISH;
00551 
00552         /* decompress and handle errors */
00553 #ifdef HAVE_ZLIB_H
00554         if (state->how == GZIP) {
00555             state->zstrm.avail_in = (uInt) state->strm.avail_in;
00556             state->zstrm.next_in = (Bytef *) state->strm.next_in;
00557             state->zstrm.avail_out = (uInt) state->strm.avail_out;
00558             state->zstrm.next_out = (Bytef *) state->strm.next_out;
00559             ret = inflate(&state->zstrm, Z_NO_FLUSH);
00560             if (ret == Z_STREAM_ERROR || ret == Z_NEED_DICT) {
00561                 xz_error(state, Z_STREAM_ERROR,
00562                          "internal error: inflate stream corrupt");
00563                 return -1;
00564             }
00565             if (ret == Z_MEM_ERROR)
00566                 ret = LZMA_MEM_ERROR;
00567             if (ret == Z_DATA_ERROR)
00568                 ret = LZMA_DATA_ERROR;
00569             if (ret == Z_STREAM_END)
00570                 ret = LZMA_STREAM_END;
00571             state->strm.avail_in = state->zstrm.avail_in;
00572             state->strm.next_in = state->zstrm.next_in;
00573             state->strm.avail_out = state->zstrm.avail_out;
00574             state->strm.next_out = state->zstrm.next_out;
00575         } else                  /* state->how == LZMA */
00576 #endif
00577             ret = lzma_code(strm, action);
00578         if (ret == LZMA_MEM_ERROR) {
00579             xz_error(state, LZMA_MEM_ERROR, "out of memory");
00580             return -1;
00581         }
00582         if (ret == LZMA_DATA_ERROR) {
00583             xz_error(state, LZMA_DATA_ERROR, "compressed data error");
00584             return -1;
00585         }
00586         if (ret == LZMA_PROG_ERROR) {
00587             xz_error(state, LZMA_PROG_ERROR, "compression error");
00588             return -1;
00589         }
00590     } while (strm->avail_out && ret != LZMA_STREAM_END);
00591 
00592     /* update available output and crc check value */
00593     state->have = had - strm->avail_out;
00594     state->next = strm->next_out - state->have;
00595 #ifdef HAVE_ZLIB_H
00596     state->zstrm.adler =
00597         crc32(state->zstrm.adler, state->next, state->have);
00598 #endif
00599 
00600     if (ret == LZMA_STREAM_END) {
00601 #ifdef HAVE_ZLIB_H
00602         if (state->how == GZIP) {
00603             if (gz_next4(state, &crc) == -1 || gz_next4(state, &len) == -1) {
00604                 xz_error(state, LZMA_DATA_ERROR, "unexpected end of file");
00605                 return -1;
00606             }
00607             if (crc != state->zstrm.adler) {
00608                 xz_error(state, LZMA_DATA_ERROR, "incorrect data check");
00609                 return -1;
00610             }
00611             if (len != (state->zstrm.total_out & 0xffffffffL)) {
00612                 xz_error(state, LZMA_DATA_ERROR, "incorrect length check");
00613                 return -1;
00614             }
00615             state->strm.avail_in = 0;
00616             state->strm.next_in = NULL;
00617             state->strm.avail_out = 0;
00618             state->strm.next_out = NULL;
00619         } else
00620 #endif
00621         if (strm->avail_in != 0 || !state->eof) {
00622             xz_error(state, LZMA_DATA_ERROR, "trailing garbage");
00623             return -1;
00624         }
00625         state->how = LOOK;      /* ready for next stream, once have is 0 (leave
00626                                  * state->direct unchanged to remember how) */
00627     }
00628 
00629     /* good decompression */
00630     return 0;
00631 }
00632 
00633 static int
00634 xz_make(xz_statep state)
00635 {
00636     lzma_stream *strm = &(state->strm);
00637 
00638     if (state->how == LOOK) {   /* look for lzma / gzip header */
00639         if (xz_head(state) == -1)
00640             return -1;
00641         if (state->have)        /* got some data from xz_head() */
00642             return 0;
00643     }
00644     if (state->how == COPY) {   /* straight copy */
00645         if (xz_load(state, state->out, state->size << 1, &(state->have)) ==
00646             -1)
00647             return -1;
00648         state->next = state->out;
00649     } else if (state->how == LZMA || state->how == GZIP) {      /* decompress */
00650         strm->avail_out = state->size << 1;
00651         strm->next_out = state->out;
00652         if (xz_decomp(state) == -1)
00653             return -1;
00654     }
00655     return 0;
00656 }
00657 
00658 static int
00659 xz_skip(xz_statep state, uint64_t len)
00660 {
00661     unsigned n;
00662 
00663     /* skip over len bytes or reach end-of-file, whichever comes first */
00664     while (len)
00665         /* skip over whatever is in output buffer */
00666         if (state->have) {
00667             n = (uint64_t) state->have > len ?
00668                 (unsigned) len : state->have;
00669             state->have -= n;
00670             state->next += n;
00671             state->pos += n;
00672             len -= n;
00673         }
00674 
00675     /* output buffer empty -- return if we're at the end of the input */
00676         else if (state->eof && state->strm.avail_in == 0)
00677             break;
00678 
00679     /* need more data to skip -- load up output buffer */
00680         else {
00681             /* get more output, looking for header if required */
00682             if (xz_make(state) == -1)
00683                 return -1;
00684         }
00685     return 0;
00686 }
00687 
00688 int
00689 __libxml2_xzread(xzFile file, void *buf, unsigned len)
00690 {
00691     unsigned got, n;
00692     xz_statep state;
00693     lzma_stream *strm;
00694 
00695     /* get internal structure */
00696     if (file == NULL)
00697         return -1;
00698     state = (xz_statep) file;
00699     strm = &(state->strm);
00700 
00701     /* check that we're reading and that there's no error */
00702     if (state->err != LZMA_OK)
00703         return -1;
00704 
00705     /* since an int is returned, make sure len fits in one, otherwise return
00706      * with an error (this avoids the flaw in the interface) */
00707     if ((int) len < 0) {
00708         xz_error(state, LZMA_BUF_ERROR,
00709                  "requested length does not fit in int");
00710         return -1;
00711     }
00712 
00713     /* if len is zero, avoid unnecessary operations */
00714     if (len == 0)
00715         return 0;
00716 
00717     /* process a skip request */
00718     if (state->seek) {
00719         state->seek = 0;
00720         if (xz_skip(state, state->skip) == -1)
00721             return -1;
00722     }
00723 
00724     /* get len bytes to buf, or less than len if at the end */
00725     got = 0;
00726     do {
00727         /* first just try copying data from the output buffer */
00728         if (state->have) {
00729             n = state->have > len ? len : state->have;
00730             memcpy(buf, state->next, n);
00731             state->next += n;
00732             state->have -= n;
00733         }
00734 
00735         /* output buffer empty -- return if we're at the end of the input */
00736         else if (state->eof && strm->avail_in == 0)
00737             break;
00738 
00739         /* need output data -- for small len or new stream load up our output
00740          * buffer */
00741         else if (state->how == LOOK || len < (state->size << 1)) {
00742             /* get more output, looking for header if required */
00743             if (xz_make(state) == -1)
00744                 return -1;
00745             continue;           /* no progress yet -- go back to memcpy() above */
00746             /* the copy above assures that we will leave with space in the
00747              * output buffer, allowing at least one gzungetc() to succeed */
00748         }
00749 
00750         /* large len -- read directly into user buffer */
00751         else if (state->how == COPY) {  /* read directly */
00752             if (xz_load(state, buf, len, &n) == -1)
00753                 return -1;
00754         }
00755 
00756         /* large len -- decompress directly into user buffer */
00757         else {                  /* state->how == LZMA */
00758             strm->avail_out = len;
00759             strm->next_out = buf;
00760             if (xz_decomp(state) == -1)
00761                 return -1;
00762             n = state->have;
00763             state->have = 0;
00764         }
00765 
00766         /* update progress */
00767         len -= n;
00768         buf = (char *) buf + n;
00769         got += n;
00770         state->pos += n;
00771     } while (len);
00772 
00773     /* return number of bytes read into user buffer (will fit in int) */
00774     return (int) got;
00775 }
00776 
00777 int
00778 __libxml2_xzclose(xzFile file)
00779 {
00780     int ret;
00781     xz_statep state;
00782 
00783     /* get internal structure */
00784     if (file == NULL)
00785         return LZMA_DATA_ERROR;
00786     state = (xz_statep) file;
00787 
00788     /* free memory and close file */
00789     if (state->size) {
00790         lzma_end(&(state->strm));
00791 #ifdef HAVE_ZLIB_H
00792         if (state->init == 1)
00793             inflateEnd(&(state->zstrm));
00794         state->init = 0;
00795 #endif
00796         xmlFree(state->out);
00797         xmlFree(state->in);
00798     }
00799     xmlFree(state->path);
00800     ret = close(state->fd);
00801     xmlFree(state);
00802     return ret ? ret : LZMA_OK;
00803 }
00804 #endif /* LIBXML_LZMA_ENABLED */
00805