nuttx-apps/system/lzf/lzf_main.c
Xiang Xiao 134b8b538f Replace all strcat with strlcat
Signed-off-by: Xiang Xiao <xiaoxiang@xiaomi.com>
2023-05-08 21:35:16 +03:00

578 lines
13 KiB
C

/****************************************************************************
* apps/system/lzf/lzf_main.c
*
* Copyright (c) 2006 Stefan Traby <stefan@hello-penguin.com>
*
* 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.
*
* 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 ``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 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.
*
****************************************************************************/
/****************************************************************************
* Included Files
****************************************************************************/
#include <nuttx/config.h>
#include <sys/types.h>
#include <sys/stat.h>
#include <stdbool.h>
#include <stdio.h>
#include <string.h>
#include <stdlib.h>
#include <unistd.h>
#include <fcntl.h>
#include <pthread.h>
#include <getopt.h>
#include <errno.h>
#include <limits.h>
#include <lzf.h>
/****************************************************************************
* Pre-processor Definitions
****************************************************************************/
#define BLOCKSIZE ((1 << CONFIG_SYSTEM_LZF_BLOG) - 1)
#define MAX_BLOCKSIZE BLOCKSIZE
/****************************************************************************
* Private Data
****************************************************************************/
#if CONFIG_SYSTEM_LZF != CONFIG_m
static pthread_mutex_t g_exclmutex = PTHREAD_MUTEX_INITIALIZER;
#endif
static off_t g_nread;
static off_t g_nwritten;
static FAR const char *g_imagename;
static enum
{
COMPRESS, UNCOMPRESS
} g_mode;
static bool g_verbose;
static bool g_force;
static unsigned long g_blocksize;
static lzf_state_t g_htab;
static uint8_t g_buf1[MAX_BLOCKSIZE + LZF_MAX_HDR_SIZE + 16];
static uint8_t g_buf2[MAX_BLOCKSIZE + LZF_MAX_HDR_SIZE + 16];
/****************************************************************************
* Private Functions
****************************************************************************/
#if CONFIG_SYSTEM_LZF != CONFIG_m
static void lzf_exit(int exitcode) noreturn_function;
static void lzf_exit(int exitcode)
{
pthread_mutex_unlock(&g_exclmutex);
exit(exitcode);
}
#else
# define lzf_exit(c) exit(c)
#endif
static void usage(int ret)
{
fprintf(stderr, "\n"
"lzf, a very lightweight compression/decompression "
"utility written by Stefan Traby.\n"
"uses liblzf written by Marc Lehmann <schmorp@schmorp.de> "
" You can find more info at\n"
"http://liblzf.plan9.de/\n"
"\n"
"usage: lzf [-dufhvb] [file ...]\n\n"
"-c Compress\n"
"-d Decompress\n"
"-f Force overwrite of output file\n"
"-h Give this help\n"
"-v Verbose mode\n"
"-b # Set blocksize (max %lu)\n"
"\n", (unsigned long)MAX_BLOCKSIZE);
lzf_exit(ret);
}
static inline ssize_t rread(int fd, FAR void *buf, size_t len)
{
FAR char *p = buf;
ssize_t ret = 0;
ssize_t offset = 0;
while (len && (ret = read(fd, &p[offset], len)) > 0)
{
offset += ret;
len -= ret;
}
g_nread += offset;
if (ret < 0)
{
return ret;
}
return offset;
}
/* Returns 0 if all written else -1 */
static inline ssize_t wwrite(int fd, void *buf, size_t len)
{
FAR char *b = buf;
ssize_t ret;
size_t l = len;
while (l)
{
ret = write(fd, b, l);
if (ret < 0)
{
fprintf(stderr, "%s: write error: %d\n", g_imagename, errno);
return -1;
}
l -= ret;
b += ret;
}
g_nwritten += len;
return 0;
}
/* Anatomy: an lzf file consists of any number of blocks
* in the following format:
*
* \x00 EOF (optional)
* "ZV\0" 2-byte-usize <uncompressed data>
* "ZV\1" 2-byte-csize 2-byte-usize <compressed data>
* "ZV\2" 4-byte-crc32-0xdebb20e3 (NYI)
*/
static int compress_fd(int from, int to)
{
FAR struct lzf_header_s *header;
ssize_t us;
ssize_t len;
g_nread = g_nwritten = 0;
while ((us = rread(from, &g_buf1[LZF_MAX_HDR_SIZE], g_blocksize)) > 0)
{
len = lzf_compress(&g_buf1[LZF_MAX_HDR_SIZE], us,
&g_buf2[LZF_MAX_HDR_SIZE],
us > 4 ? us - 4 : us, g_htab, &header);
if (wwrite(to, header, len) == -1)
{
return -1;
}
}
return 0;
}
static int uncompress_fd(int from, int to)
{
uint8_t header[LZF_MAX_HDR_SIZE];
FAR uint8_t *p;
int l;
int rd;
ssize_t ret;
ssize_t cs;
ssize_t us;
ssize_t bytes;
ssize_t over = 0;
g_nread = g_nwritten = 0;
while (1)
{
ret = rread(from, header + over, LZF_MAX_HDR_SIZE - over);
if (ret < 0)
{
fprintf(stderr, "%s: read error: %d\n", g_imagename, errno);
return -1;
}
ret += over;
over = 0;
if (!ret || header[0] == 0)
{
return 0;
}
if (ret < LZF_MIN_HDR_SIZE || header[0] != 'Z' || header[1] != 'V')
{
fprintf(stderr, "%s: invalid data stream - "
"magic not found or short header\n",
g_imagename);
return -1;
}
switch (header[2])
{
case 0:
cs = -1;
us = (header[3] << 8) | header[4];
p = &header[LZF_TYPE0_HDR_SIZE];
break;
case 1:
if (ret < LZF_TYPE1_HDR_SIZE)
{
goto short_read;
}
cs = (header[3] << 8) | header[4];
us = (header[5] << 8) | header[6];
p = &header[LZF_TYPE1_HDR_SIZE];
break;
default:
fprintf(stderr, "%s: unknown blocktype\n", g_imagename);
return -1;
}
bytes = cs == -1 ? us : cs;
l = &header[ret] - p;
if (l > 0)
{
memcpy(g_buf1, p, l);
}
if (l > bytes)
{
over = l - bytes;
memmove(header, &p[bytes], over);
}
p = &g_buf1[l];
rd = bytes - l;
if (rd > 0)
{
if ((ret = rread(from, p, rd)) != rd)
{
goto short_read;
}
}
if (cs == -1)
{
if (wwrite (to, g_buf1, us))
{
return -1;
}
}
else
{
if (lzf_decompress(g_buf1, cs, g_buf2, us) != us)
{
fprintf(stderr, "%s: decompress: invalid stream - "
"data corrupted\n",
g_imagename);
return -1;
}
if (wwrite(to, g_buf2, us))
{
return -1;
}
}
}
return 0;
short_read:
fprintf(stderr, "%s: short data\n", g_imagename);
return -1;
}
static int open_out(FAR const char *name)
{
int m = O_EXCL;
if (g_force)
{
m = 0;
}
return open(name, O_CREAT | O_WRONLY | O_TRUNC | m, 0600);
}
static int compose_name(FAR const char *fname, FAR char *oname, int namelen)
{
FAR char *p;
if (g_mode == COMPRESS)
{
if (strlen(fname) > PATH_MAX - 4)
{
fprintf(stderr, "%s: %s.lzf: name too long", g_imagename, fname);
return -1;
}
strlcpy(oname, fname, namelen);
p = strchr(oname, '.');
if (p != NULL)
{
*p = '_'; /* _ for dot */
}
strlcat(oname, ".lzf", namelen);
}
else
{
if (strlen(fname) > PATH_MAX)
{
fprintf(stderr, "%s: %s: name too long\n", g_imagename, fname);
return -1;
}
strlcpy(oname, fname, namelen);
p = strstr(oname, ".lzf");
if (p == NULL)
{
fprintf(stderr, "%s: %s: unknown suffix\n", g_imagename, fname);
return -1;
}
*p = 0;
p = strchr(oname, '_');
if (p != NULL)
{
*p = '.';
}
}
return 0;
}
static int run_file(FAR const char *fname)
{
struct stat mystat;
char oname[PATH_MAX];
int fd;
int fd2;
int ret;
if (compose_name(fname, oname, sizeof(oname)))
{
return -1;
}
ret = stat(fname, &mystat);
fd = open(fname, O_RDONLY);
if (ret || fd == -1)
{
fprintf(stderr, "%s: %s: %d\n", g_imagename, fname, errno);
return -1;
}
if (!S_ISREG(mystat.st_mode))
{
fprintf(stderr, "%s: %s: not a regular file.\n", g_imagename, fname);
close(fd);
return -1;
}
fd2 = open_out(oname);
if (fd2 == -1)
{
fprintf(stderr, "%s: %s: %d\n", g_imagename, oname, errno);
close(fd);
return -1;
}
if (g_mode == COMPRESS)
{
ret = compress_fd(fd, fd2);
if (!ret && g_verbose)
{
fprintf(stderr, "%s: %5.1f%% -- replaced with %s\n",
fname, g_nread == 0 ? 0 :
100.0 - g_nwritten / ((double) g_nread / 100.0), oname);
}
}
else
{
ret = uncompress_fd(fd, fd2);
if (!ret && g_verbose)
{
fprintf(stderr, "%s: %5.1f%% -- replaced with %s\n",
fname, g_nwritten == 0 ? 0 :
100.0 - g_nread / ((double) g_nwritten / 100.0), oname);
}
}
close(fd);
close(fd2);
if (!ret)
{
unlink(fname);
}
return ret;
}
/****************************************************************************
* lzf_main
****************************************************************************/
int main(int argc, FAR char *argv[])
{
FAR char *p = argv[0];
int optc;
int ret = 0;
#if CONFIG_SYSTEM_LZF != CONFIG_m
/* Get exclusive access to the global variables. Global variables are
* used because the hash table and buffers are too large to allocate on
* the embedded stack. But the use of global variables has the downside
* or forcing serialization of this logic in order to work in a multi-
* tasking environment.
*
* REVISIT: An alternative would be to pack all of the globals into a
* structure and allocate a per-thread instance of that structure here.
*
* NOTE: This applies only in the FLAT and PROTECTED build modes. In the
* KERNEL build mode, this will be a separate process with its own private
* global variables.
*/
pthread_mutex_lock(&g_exclmutex);
#endif
/* Set defaults. */
g_mode = COMPRESS;
g_verbose = false;
g_force = 0;
g_blocksize = BLOCKSIZE;
#ifndef CONFIG_DISABLE_ENVIRON
/* Block size may be specified as an environment variable */
p = getenv("LZF_BLOCKSIZE");
if (p)
{
g_blocksize = strtoul(p, 0, 0);
if (g_blocksize == 0 || g_blocksize > MAX_BLOCKSIZE)
{
g_blocksize = BLOCKSIZE;
}
}
#endif
/* Get the program name sans path */
p = strrchr(argv[0], '/');
g_imagename = p ? ++p : argv[0];
/* Handle command line options */
while ((optc = getopt(argc, argv, "cdfhvb:")) != -1)
{
switch (optc)
{
case 'c':
g_mode = COMPRESS;
break;
case 'd':
g_mode = UNCOMPRESS;
break;
case 'f':
g_force = true;
break;
case 'h':
usage(0);
break;
case 'v':
g_verbose = true;
break;
case 'b':
g_blocksize = strtoul(optarg, 0, 0);
if (g_blocksize == 0 || g_blocksize > MAX_BLOCKSIZE)
{
g_blocksize = BLOCKSIZE;
}
break;
default:
usage(1);
break;
}
}
if (optind == argc)
{
/* stdin stdout */
if (!g_force)
{
if ((g_mode == UNCOMPRESS) && isatty(0))
{
fprintf(stderr,
"%s: compressed data not read from a terminal. "
"Use -f to force decompression.\n", g_imagename);
lzf_exit(1);
}
if (g_mode == COMPRESS && isatty(1))
{
fprintf(stderr,
"%s: compressed data not written to a terminal. "
"Use -f to force compression.\n", g_imagename);
lzf_exit(1);
}
}
if (g_mode == COMPRESS)
{
ret = compress_fd(0, 1);
}
else
{
ret = uncompress_fd(0, 1);
}
lzf_exit(ret ? 1 : 0);
}
while (optind < argc)
{
ret |= run_file(argv[optind++]);
}
lzf_exit(ret ? 1 : 0);
}