freebsd-dev/lib/libarchive/archive_read_support_compression_gzip.c
Tim Kientzle 693285bc87 Use 'skip' when ignoring data in tar archives. This dramatically
increases performance when extracting a single entry from a large
uncompressed archive, especially on slow devices such as USB hard
drives.

Requires a number of changes:
   * New archive_read_open2() supports a 'skip' client function
   * Old archive_read_open() is implemented as a wrapper now, to
     continue supporting the old API/ABI.
   * _read_open_fd and _read_open_file sprout new 'skip' functions.
   * compression layer gets a new 'skip' operation.
   * compression_none passes skip requests through to client.
   * compression_{gzip,bzip2,compress} simply ignore skip requests.

Thanks to: Benjamin Lutz, who designed and implemented the whole thing.
   I'm just committing it.  ;-)

TODO: Need to update the documentation a little bit.
2006-07-30 00:29:01 +00:00

533 lines
14 KiB
C

/*-
* Copyright (c) 2003-2004 Tim Kientzle
* All rights reserved.
*
* Redistribution and use in source and binary forms, with or without
* modification, are permitted provided that the following conditions
* are met:
* 1. Redistributions of source code must retain the above copyright
* notice, this list of conditions and the following disclaimer
* in this position and unchanged.
* 2. Redistributions in binary form must reproduce the above copyright
* notice, this list of conditions and the following disclaimer in the
* documentation and/or other materials provided with the distribution.
*
* THIS SOFTWARE IS PROVIDED BY THE AUTHOR(S) ``AS IS'' AND ANY EXPRESS OR
* IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES
* OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE DISCLAIMED.
* IN NO EVENT SHALL THE AUTHOR(S) BE LIABLE FOR ANY DIRECT, INDIRECT,
* INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT
* NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
* DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
* THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
* (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF
* THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
*/
#include "archive_platform.h"
__FBSDID("$FreeBSD$");
#include <errno.h>
#include <stdlib.h>
#include <string.h>
#include <unistd.h>
#ifdef HAVE_ZLIB_H
#include <zlib.h>
#endif
#include "archive.h"
#include "archive_private.h"
#ifdef HAVE_ZLIB_H
struct private_data {
z_stream stream;
unsigned char *uncompressed_buffer;
size_t uncompressed_buffer_size;
unsigned char *read_next;
int64_t total_out;
unsigned long crc;
char header_done;
};
static int finish(struct archive *);
static ssize_t read_ahead(struct archive *, const void **, size_t);
static ssize_t read_consume(struct archive *, size_t);
static int drive_decompressor(struct archive *a, struct private_data *);
#endif
/* These two functions are defined even if we lack zlib. See below. */
static int bid(const void *, size_t);
static int init(struct archive *, const void *, size_t);
int
archive_read_support_compression_gzip(struct archive *a)
{
return (__archive_read_register_compression(a, bid, init));
}
/*
* Test whether we can handle this data.
*
* This logic returns zero if any part of the signature fails. It
* also tries to Do The Right Thing if a very short buffer prevents us
* from verifying as much as we would like.
*/
static int
bid(const void *buff, size_t len)
{
const unsigned char *buffer;
int bits_checked;
if (len < 1)
return (0);
buffer = buff;
bits_checked = 0;
if (buffer[0] != 037) /* Verify first ID byte. */
return (0);
bits_checked += 8;
if (len < 2)
return (bits_checked);
if (buffer[1] != 0213) /* Verify second ID byte. */
return (0);
bits_checked += 8;
if (len < 3)
return (bits_checked);
if (buffer[2] != 8) /* Compression must be 'deflate'. */
return (0);
bits_checked += 8;
if (len < 4)
return (bits_checked);
if ((buffer[3] & 0xE0)!= 0) /* No reserved flags set. */
return (0);
bits_checked += 3;
if (len < 5)
return (bits_checked);
/*
* TODO: Verify more; in particular, gzip has an optional
* header CRC, which would give us 16 more verified bits. We
* may also be able to verify certain constraints on other
* fields.
*/
return (bits_checked);
}
#ifndef HAVE_ZLIB_H
/*
* If we don't have zlib on this system, we can't actually do the
* decompression. We can, however, still detect gzip-compressed
* archives and emit a useful message.
*/
static int
init(struct archive *a, const void *buff, size_t n)
{
(void)a; /* UNUSED */
(void)buff; /* UNUSED */
(void)n; /* UNUSED */
archive_set_error(a, -1,
"This version of libarchive was compiled without gzip support");
return (ARCHIVE_FATAL);
}
#else
/*
* Setup the callbacks.
*/
static int
init(struct archive *a, const void *buff, size_t n)
{
struct private_data *state;
int ret;
a->compression_code = ARCHIVE_COMPRESSION_GZIP;
a->compression_name = "gzip";
state = malloc(sizeof(*state));
if (state == NULL) {
archive_set_error(a, ENOMEM,
"Can't allocate data for %s decompression",
a->compression_name);
return (ARCHIVE_FATAL);
}
memset(state, 0, sizeof(*state));
state->crc = crc32(0L, NULL, 0);
state->header_done = 0; /* We've not yet begun to parse header... */
state->uncompressed_buffer_size = 64 * 1024;
state->uncompressed_buffer = malloc(state->uncompressed_buffer_size);
state->stream.next_out = state->uncompressed_buffer;
state->read_next = state->uncompressed_buffer;
state->stream.avail_out = state->uncompressed_buffer_size;
if (state->uncompressed_buffer == NULL) {
archive_set_error(a, ENOMEM,
"Can't allocate %s decompression buffers",
a->compression_name);
free(state);
return (ARCHIVE_FATAL);
}
/*
* A bug in zlib.h: stream.next_in should be marked 'const'
* but isn't (the library never alters data through the
* next_in pointer, only reads it). The result: this ugly
* cast to remove 'const'.
*/
state->stream.next_in = (void *)(uintptr_t)(const void *)buff;
state->stream.avail_in = n;
a->compression_read_ahead = read_ahead;
a->compression_read_consume = read_consume;
a->compression_skip = NULL; /* not supported */
a->compression_finish = finish;
/*
* TODO: Do I need to parse the gzip header before calling
* inflateInit2()? In particular, one of the header bytes
* marks "best compression" or "fastest", which may be
* appropriate for setting the second parameter here.
* However, I think the only penalty for not setting it
* correctly is wasted memory. If this is necessary, it
* should probably go into drive_decompressor() below.
*/
/* Initialize compression library. */
ret = inflateInit2(&(state->stream),
-15 /* Don't check for zlib header */);
if (ret == Z_OK) {
a->compression_data = state;
return (ARCHIVE_OK);
}
/* Library setup failed: Clean up. */
archive_set_error(a, ARCHIVE_ERRNO_MISC,
"Internal error initializing %s library", a->compression_name);
free(state->uncompressed_buffer);
free(state);
/* Override the error message if we know what really went wrong. */
switch (ret) {
case Z_STREAM_ERROR:
archive_set_error(a, ARCHIVE_ERRNO_MISC,
"Internal error initializing compression library: "
"invalid setup parameter");
break;
case Z_MEM_ERROR:
archive_set_error(a, ENOMEM,
"Internal error initializing compression library: "
"out of memory");
break;
case Z_VERSION_ERROR:
archive_set_error(a, ARCHIVE_ERRNO_MISC,
"Internal error initializing compression library: "
"invalid library version");
break;
}
return (ARCHIVE_FATAL);
}
/*
* Return a block of data from the decompression buffer. Decompress more
* as necessary.
*/
static ssize_t
read_ahead(struct archive *a, const void **p, size_t min)
{
struct private_data *state;
int read_avail, was_avail, ret;
state = a->compression_data;
was_avail = -1;
if (!a->client_reader) {
archive_set_error(a, ARCHIVE_ERRNO_PROGRAMMER,
"No read callback is registered? "
"This is probably an internal programming error.");
return (ARCHIVE_FATAL);
}
read_avail = state->stream.next_out - state->read_next;
if (read_avail + state->stream.avail_out < min) {
memmove(state->uncompressed_buffer, state->read_next,
read_avail);
state->read_next = state->uncompressed_buffer;
state->stream.next_out = state->read_next + read_avail;
state->stream.avail_out
= state->uncompressed_buffer_size - read_avail;
}
while (was_avail < read_avail && /* Made some progress. */
read_avail < (int)min && /* Haven't satisfied min. */
read_avail < (int)state->uncompressed_buffer_size) { /* !full */
if ((ret = drive_decompressor(a, state)) != ARCHIVE_OK)
return (ret);
was_avail = read_avail;
read_avail = state->stream.next_out - state->read_next;
}
*p = state->read_next;
return (read_avail);
}
/*
* Mark a previously-returned block of data as read.
*/
static ssize_t
read_consume(struct archive *a, size_t n)
{
struct private_data *state;
state = a->compression_data;
a->file_position += n;
state->read_next += n;
if (state->read_next > state->stream.next_out)
__archive_errx(1, "Request to consume too many "
"bytes from gzip decompressor");
return (n);
}
/*
* Clean up the decompressor.
*/
static int
finish(struct archive *a)
{
struct private_data *state;
int ret;
state = a->compression_data;
ret = ARCHIVE_OK;
switch (inflateEnd(&(state->stream))) {
case Z_OK:
break;
default:
archive_set_error(a, ARCHIVE_ERRNO_MISC,
"Failed to clean up %s compressor", a->compression_name);
ret = ARCHIVE_FATAL;
}
free(state->uncompressed_buffer);
free(state);
a->compression_data = NULL;
if (a->client_closer != NULL)
(a->client_closer)(a, a->client_data);
return (ret);
}
/*
* Utility function to pull data through decompressor, reading input
* blocks as necessary.
*/
static int
drive_decompressor(struct archive *a, struct private_data *state)
{
ssize_t ret;
int decompressed, total_decompressed;
int count, flags, header_state;
unsigned char *output;
unsigned char b;
flags = 0;
count = 0;
header_state = 0;
total_decompressed = 0;
for (;;) {
if (state->stream.avail_in == 0) {
ret = (a->client_reader)(a, a->client_data,
(const void **)&state->stream.next_in);
if (ret < 0) {
/*
* TODO: Find a better way to handle
* this read failure.
*/
goto fatal;
}
if (ret == 0 && total_decompressed == 0) {
archive_set_error(a, EIO,
"Premature end of %s compressed data",
a->compression_name);
return (ARCHIVE_FATAL);
}
a->raw_position += ret;
state->stream.avail_in = ret;
}
if (!state->header_done) {
/*
* If still parsing the header, interpret the
* next byte.
*/
b = *(state->stream.next_in++);
state->stream.avail_in--;
/*
* Yes, this is somewhat crude, but it works,
* GZip format isn't likely to change anytime
* in the near future, and header parsing is
* certainly not a performance issue, so
* there's little point in making this more
* elegant. Of course, if you see an easy way
* to make this more elegant, please let me
* know.. ;-)
*/
switch (header_state) {
case 0: /* First byte of signature. */
if (b != 037)
goto fatal;
header_state = 1;
break;
case 1: /* Second byte of signature. */
if (b != 0213)
goto fatal;
header_state = 2;
break;
case 2: /* Compression type must be 8. */
if (b != 8)
goto fatal;
header_state = 3;
break;
case 3: /* GZip flags. */
flags = b;
header_state = 4;
break;
case 4: case 5: case 6: case 7: /* Mod time. */
header_state++;
break;
case 8: /* Deflate flags. */
header_state = 9;
break;
case 9: /* OS. */
header_state = 10;
break;
case 10: /* Optional Extra: First byte of Length. */
if ((flags & 4)) {
count = 255 & (int)b;
header_state = 11;
break;
}
/*
* Fall through if there is no
* Optional Extra field.
*/
case 11: /* Optional Extra: Second byte of Length. */
if ((flags & 4)) {
count = (0xff00 & ((int)b << 8)) | count;
header_state = 12;
break;
}
/*
* Fall through if there is no
* Optional Extra field.
*/
case 12: /* Optional Extra Field: counted length. */
if ((flags & 4)) {
--count;
if (count == 0) header_state = 13;
else header_state = 12;
break;
}
/*
* Fall through if there is no
* Optional Extra field.
*/
case 13: /* Optional Original Filename. */
if ((flags & 8)) {
if (b == 0) header_state = 14;
else header_state = 13;
break;
}
/*
* Fall through if no Optional
* Original Filename.
*/
case 14: /* Optional Comment. */
if ((flags & 16)) {
if (b == 0) header_state = 15;
else header_state = 14;
break;
}
/* Fall through if no Optional Comment. */
case 15: /* Optional Header CRC: First byte. */
if ((flags & 2)) {
header_state = 16;
break;
}
/* Fall through if no Optional Header CRC. */
case 16: /* Optional Header CRC: Second byte. */
if ((flags & 2)) {
header_state = 17;
break;
}
/* Fall through if no Optional Header CRC. */
case 17: /* First byte of compressed data. */
state->header_done = 1; /* done with header */
state->stream.avail_in++;
state->stream.next_in--;
}
/*
* TODO: Consider moving the inflateInit2 call
* here so it can include the compression type
* from the header?
*/
} else {
output = state->stream.next_out;
/* Decompress some data. */
ret = inflate(&(state->stream), 0);
decompressed = state->stream.next_out - output;
/* Accumulate the CRC of the uncompressed data. */
state->crc = crc32(state->crc, output, decompressed);
/* Accumulate the total bytes of output. */
state->total_out += decompressed;
total_decompressed += decompressed;
switch (ret) {
case Z_OK: /* Decompressor made some progress. */
if (decompressed > 0)
return (ARCHIVE_OK);
break;
case Z_STREAM_END: /* Found end of stream. */
/*
* TODO: Verify gzip trailer
* (uncompressed length and CRC).
*/
return (ARCHIVE_OK);
default:
/* Any other return value is an error. */
archive_set_error(a, ARCHIVE_ERRNO_MISC,
"gzip decompression failed (%s)",
state->stream.msg);
goto fatal;
}
}
}
return (ARCHIVE_OK);
/* Return a fatal error. */
fatal:
archive_set_error(a, ARCHIVE_ERRNO_MISC, "%s decompression failed",
a->compression_name);
return (ARCHIVE_FATAL);
}
#endif /* HAVE_ZLIB_H */