diff options
Diffstat (limited to 'src/lzma')
| -rw-r--r-- | src/lzma/Makefile.am | 63 | ||||
| -rw-r--r-- | src/lzma/alloc.c | 106 | ||||
| -rw-r--r-- | src/lzma/alloc.h | 42 | ||||
| -rw-r--r-- | src/lzma/args.c | 566 | ||||
| -rw-r--r-- | src/lzma/args.h | 64 | ||||
| -rw-r--r-- | src/lzma/error.c | 156 | ||||
| -rw-r--r-- | src/lzma/error.h | 67 | ||||
| -rw-r--r-- | src/lzma/hardware.c | 99 | ||||
| -rw-r--r-- | src/lzma/hardware.h | 31 | ||||
| -rw-r--r-- | src/lzma/help.c | 178 | ||||
| -rw-r--r-- | src/lzma/help.h | 32 | ||||
| -rw-r--r-- | src/lzma/io.c | 664 | ||||
| -rw-r--r-- | src/lzma/io.h | 60 | ||||
| -rw-r--r-- | src/lzma/list.c | 477 | ||||
| -rw-r--r-- | src/lzma/main.c | 254 | ||||
| -rw-r--r-- | src/lzma/options.c | 346 | ||||
| -rw-r--r-- | src/lzma/options.h | 46 | ||||
| -rw-r--r-- | src/lzma/private.h | 55 | ||||
| -rw-r--r-- | src/lzma/process.c | 458 | ||||
| -rw-r--r-- | src/lzma/process.h | 30 | ||||
| -rw-r--r-- | src/lzma/suffix.c | 145 | ||||
| -rw-r--r-- | src/lzma/suffix.h | 25 | ||||
| -rw-r--r-- | src/lzma/util.c | 182 | ||||
| -rw-r--r-- | src/lzma/util.h | 32 |
24 files changed, 4178 insertions, 0 deletions
diff --git a/src/lzma/Makefile.am b/src/lzma/Makefile.am new file mode 100644 index 00000000..5fbd3358 --- /dev/null +++ b/src/lzma/Makefile.am @@ -0,0 +1,63 @@ +## +## Copyright (C) 2007 Lasse Collin +## +## This program is free software; you can redistribute it and/or +## modify it under the terms of the GNU Lesser General Public +## License as published by the Free Software Foundation; either +## version 2.1 of the License, or (at your option) any later version. +## +## This program is distributed in the hope that it will be useful, +## but WITHOUT ANY WARRANTY; without even the implied warranty of +## MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU +## Lesser General Public License for more details. +## + +bin_PROGRAMS = lzma + +lzma_SOURCES = \ + alloc.c \ + alloc.h \ + args.c \ + args.h \ + error.c \ + error.h \ + hardware.c \ + hardware.h \ + help.c \ + help.h \ + io.c \ + io.h \ + main.c \ + options.c \ + options.h \ + private.h \ + process.c \ + process.h \ + suffix.c \ + suffix.h \ + util.c \ + util.h + +## It must be able to find sysdefs.h, lzma_adv.h, and possible +## replacement headers. +lzma_CPPFLAGS = \ + -DLOCALEDIR=\"$(localedir)\" \ + -I@top_srcdir@/src/common \ + -I@top_srcdir@/src/liblzma/api \ + -I@top_builddir@/lib \ + -I@top_srcdir@/lib + +lzma_CFLAGS = @PTHREAD_CFLAGS@ + +## Always link the command line tool statically against liblzma. It is +## faster on x86, because no need for PIC. We also have one dependency less, +## which allows users to more freely copy the lzma binary to other boxes. +lzma_LDFLAGS = -static +lzma_LDADD = \ + @top_builddir@/src/liblzma/liblzma.la \ + @LTLIBINTL@ \ + @PTHREAD_LIBS@ + +if COND_GNULIB +lzma_LDADD += @top_builddir@/lib/libgnu_nls.a +endif diff --git a/src/lzma/alloc.c b/src/lzma/alloc.c new file mode 100644 index 00000000..d0fee68b --- /dev/null +++ b/src/lzma/alloc.c @@ -0,0 +1,106 @@ +/////////////////////////////////////////////////////////////////////////////// +// +/// \file alloc.c +/// \brief Memory allocation functions +// +// Copyright (C) 2007 Lasse Collin +// +// This program is free software; you can redistribute it and/or +// modify it under the terms of the GNU Lesser General Public +// License as published by the Free Software Foundation; either +// version 2.1 of the License, or (at your option) any later version. +// +// This program is distributed in the hope that it will be useful, +// but WITHOUT ANY WARRANTY; without even the implied warranty of +// MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU +// Lesser General Public License for more details. +// +/////////////////////////////////////////////////////////////////////////////// + +#include "private.h" + + +/// Called when memory allocation fails. Prints and error message and +/// quits the application. +static void lzma_attribute((noreturn)) +xerror(void) +{ + errmsg(V_ERROR, "%s", strerror(errno)); + my_exit(ERROR); +} + + +extern void * +xmalloc(size_t size) +{ + if (size < 1) { + errno = EINVAL; + xerror(); + } + + void *p = malloc(size); + if (p == NULL) + xerror(); + + return p; +} + + +/* +extern void * +xrealloc(void *ptr, size_t size) +{ + if (size < 1) { + errno = EINVAL; + xerror(); + } + + ptr = realloc(ptr, size); + if (ptr == NULL) + xerror(); + + return ptr; +} +*/ + + +extern char * +xstrdup(const char *src) +{ + if (src == NULL) { + errno = EINVAL; + xerror(); + } + + const size_t size = strlen(src) + 1; + char *dest = malloc(size); + if (dest == NULL) + xerror(); + + memcpy(dest, src, size); + + return dest; +} + + +extern void +xstrcpy(char **dest, const char *src) +{ + size_t len = strlen(src) + 1; + + *dest = realloc(*dest, len); + if (*dest == NULL) + xerror(); + + memcpy(*dest, src, len + 1); + + return; +} + + +extern void * +allocator(void *opaque lzma_attribute((unused)), + size_t nmemb lzma_attribute((unused)), size_t size) +{ + return xmalloc(size); +} diff --git a/src/lzma/alloc.h b/src/lzma/alloc.h new file mode 100644 index 00000000..80317269 --- /dev/null +++ b/src/lzma/alloc.h @@ -0,0 +1,42 @@ +/////////////////////////////////////////////////////////////////////////////// +// +/// \file alloc.h +/// \brief Memory allocation functions +// +// Copyright (C) 2007 Lasse Collin +// +// This program is free software; you can redistribute it and/or +// modify it under the terms of the GNU Lesser General Public +// License as published by the Free Software Foundation; either +// version 2.1 of the License, or (at your option) any later version. +// +// This program is distributed in the hope that it will be useful, +// but WITHOUT ANY WARRANTY; without even the implied warranty of +// MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU +// Lesser General Public License for more details. +// +/////////////////////////////////////////////////////////////////////////////// + +#ifndef ALLOC_H +#define ALLOC_H + +#include "private.h" + + +/// Safe malloc() that never returns NULL. +extern void *xmalloc(size_t size); + +/// Safe realloc() that never returns NULL. +extern void *xrealloc(void *ptr, size_t size); + +/// Safe strdup() that never returns NULL. +extern char *xstrdup(const char *src); + +/// xrealloc()s *dest to the size needed by src, and copies src to *dest. +extern void xstrcpy(char **dest, const char *src); + +/// Function for lzma_allocator.alloc. This uses xmalloc(). +extern void *allocator(void *opaque lzma_attribute((unused)), + size_t nmemb lzma_attribute((unused)), size_t size); + +#endif diff --git a/src/lzma/args.c b/src/lzma/args.c new file mode 100644 index 00000000..d6163ae7 --- /dev/null +++ b/src/lzma/args.c @@ -0,0 +1,566 @@ +/////////////////////////////////////////////////////////////////////////////// +// +/// \file args.c +/// \brief Argument parsing +/// +/// \note Filter-specific options parsing is in options.c. +// +// Copyright (C) 2007 Lasse Collin +// +// This program is free software; you can redistribute it and/or +// modify it under the terms of the GNU Lesser General Public +// License as published by the Free Software Foundation; either +// version 2.1 of the License, or (at your option) any later version. +// +// This program is distributed in the hope that it will be useful, +// but WITHOUT ANY WARRANTY; without even the implied warranty of +// MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU +// Lesser General Public License for more details. +// +/////////////////////////////////////////////////////////////////////////////// + +#include "private.h" + +#include "getopt.h" +#include <ctype.h> + + +enum tool_mode opt_mode = MODE_COMPRESS; +enum header_type opt_header = HEADER_AUTO; + +char *opt_suffix = NULL; + +char *opt_files_name = NULL; +char opt_files_split = '\0'; +FILE *opt_files_file = NULL; + +bool opt_stdout = false; +bool opt_force = false; +bool opt_keep_original = false; +bool opt_preserve_name = false; + +lzma_check_type opt_check = LZMA_CHECK_CRC64; +lzma_options_filter opt_filters[8]; + +// We don't modify or free() this, but we need to assign it in some +// non-const pointers. +const char *stdin_filename = "(stdin)"; + +static size_t preset_number = 7 - 1; +static bool preset_default = true; +static size_t filter_count = 0; + + +enum { + OPT_COPY = INT_MIN, + OPT_SUBBLOCK, + OPT_X86, + OPT_POWERPC, + OPT_IA64, + OPT_ARM, + OPT_ARMTHUMB, + OPT_SPARC, + OPT_DELTA, + OPT_LZMA, + + OPT_FILES, + OPT_FILES0, +}; + + +static const char short_opts[] = "cC:dfFhlLkqrStT:vVz123456789"; + + +static const struct option long_opts[] = { + // gzip-like options + { "fast", no_argument, NULL, '1' }, + { "best", no_argument, NULL, '9' }, + { "memory", required_argument, NULL, 'M' }, + { "name", no_argument, NULL, 'N' }, + { "suffix", required_argument, NULL, 'S' }, + { "threads", required_argument, NULL, 'T' }, + { "version", no_argument, NULL, 'V' }, + { "stdout", no_argument, NULL, 'c' }, + { "to-stdout", no_argument, NULL, 'c' }, + { "decompress", no_argument, NULL, 'd' }, + { "uncompress", no_argument, NULL, 'd' }, + { "force", no_argument, NULL, 'f' }, + { "help", no_argument, NULL, 'h' }, + { "list", no_argument, NULL, 'l' }, + { "info", no_argument, NULL, 'l' }, + { "keep", no_argument, NULL, 'k' }, + { "no-name", no_argument, NULL, 'n' }, + { "quiet", no_argument, NULL, 'q' }, +// { "recursive", no_argument, NULL, 'r' }, // TODO + { "test", no_argument, NULL, 't' }, + { "verbose", no_argument, NULL, 'v' }, + { "compress", no_argument, NULL, 'z' }, + + // Filters + { "copy", no_argument, NULL, OPT_COPY }, + { "subblock", optional_argument, NULL, OPT_SUBBLOCK }, + { "x86", no_argument, NULL, OPT_X86 }, + { "bcj", no_argument, NULL, OPT_X86 }, + { "powerpc", no_argument, NULL, OPT_POWERPC }, + { "ppc", no_argument, NULL, OPT_POWERPC }, + { "ia64", no_argument, NULL, OPT_IA64 }, + { "itanium", no_argument, NULL, OPT_IA64 }, + { "arm", no_argument, NULL, OPT_ARM }, + { "armthumb", no_argument, NULL, OPT_ARMTHUMB }, + { "sparc", no_argument, NULL, OPT_SPARC }, + { "delta", optional_argument, NULL, OPT_DELTA }, + { "lzma", optional_argument, NULL, OPT_LZMA }, + + // Other + { "format", required_argument, NULL, 'F' }, + { "check", required_argument, NULL, 'C' }, + { "files", optional_argument, NULL, OPT_FILES }, + { "files0", optional_argument, NULL, OPT_FILES0 }, + + { NULL, 0, NULL, 0 } +}; + + +static void +add_filter(lzma_vli id, const char *opt_str) +{ + if (filter_count == 7) { + errmsg(V_ERROR, _("Maximum number of filters is seven")); + my_exit(ERROR); + } + + opt_filters[filter_count].id = id; + + switch (id) { + case LZMA_FILTER_SUBBLOCK: + opt_filters[filter_count].options + = parse_options_subblock(opt_str); + break; + + case LZMA_FILTER_DELTA: + opt_filters[filter_count].options + = parse_options_delta(opt_str); + break; + + case LZMA_FILTER_LZMA: + opt_filters[filter_count].options + = parse_options_lzma(opt_str); + break; + + default: + assert(opt_str == NULL); + opt_filters[filter_count].options = NULL; + break; + } + + ++filter_count; + preset_default = false; + return; +} + + +static void +parse_real(int argc, char **argv) +{ + int c; + + while ((c = getopt_long(argc, argv, short_opts, long_opts, NULL)) + != -1) { + switch (c) { + // gzip-like options + + case '1': case '2': case '3': case '4': + case '5': case '6': case '7': case '8': case '9': + preset_number = c - '1'; + preset_default = false; + break; + + // --memory + case 'M': + opt_memory = str_to_uint64("memory", optarg, + 1, SIZE_MAX); + break; + + case 'N': + opt_preserve_name = true; + break; + + // --suffix + case 'S': + // Empty suffix and suffixes having a slash are + // rejected. Such suffixes would break things later. + if (optarg[0] == '\0' || strchr(optarg, '/') != NULL) { + errmsg(V_ERROR, _("%s: Invalid filename " + "suffix"), optarg); + my_exit(ERROR); + } + + free(opt_suffix); + opt_suffix = xstrdup(optarg); + break; + + case 'T': + opt_threads = str_to_uint64("threads", optarg, + 1, SIZE_MAX); + break; + + // --version + case 'V': + // This doesn't return. + show_version(); + + // --stdout + case 'c': + opt_stdout = true; + break; + + // --decompress + case 'd': + opt_mode = MODE_DECOMPRESS; + break; + + // --force + case 'f': + opt_force = true; + break; + + // --help + case 'h': + // This doesn't return. + show_help(); + + // --list + case 'l': + opt_mode = MODE_LIST; + break; + + // --keep + case 'k': + opt_keep_original = true; + break; + + case 'n': + opt_preserve_name = false; + break; + + // --quiet + case 'q': + if (verbosity > V_SILENT) + --verbosity; + + break; + + case 't': + opt_mode = MODE_TEST; + break; + + // --verbose + case 'v': + if (verbosity < V_DEBUG) + ++verbosity; + + break; + + case 'z': + opt_mode = MODE_COMPRESS; + break; + + // Filter setup + + case OPT_COPY: + add_filter(LZMA_FILTER_COPY, NULL); + break; + + case OPT_SUBBLOCK: + add_filter(LZMA_FILTER_SUBBLOCK, optarg); + break; + + case OPT_X86: + add_filter(LZMA_FILTER_X86, NULL); + break; + + case OPT_POWERPC: + add_filter(LZMA_FILTER_POWERPC, NULL); + break; + + case OPT_IA64: + add_filter(LZMA_FILTER_IA64, NULL); + break; + + case OPT_ARM: + add_filter(LZMA_FILTER_ARM, NULL); + break; + + case OPT_ARMTHUMB: + add_filter(LZMA_FILTER_ARMTHUMB, NULL); + break; + + case OPT_SPARC: + add_filter(LZMA_FILTER_SPARC, NULL); + break; + + case OPT_DELTA: + add_filter(LZMA_FILTER_DELTA, optarg); + break; + + case OPT_LZMA: + add_filter(LZMA_FILTER_LZMA, optarg); + break; + + // Other + + // --format + case 'F': { + static const char *types[] = { + "auto", + "native", + "single", + "multi", + "alone", +// "gzip", + NULL + }; + + opt_header = 0; + while (strcmp(types[opt_header], optarg) != 0) { + if (types[++opt_header] == NULL) { + errmsg(V_ERROR, _("%s: Unknown file " + "format type"), + optarg); + my_exit(ERROR); + } + } + + break; + } + + // --check + case 'C': { + static const struct { + const char *str; + unsigned int value; + } types[] = { + { "none", LZMA_CHECK_NONE }, + { "crc32", LZMA_CHECK_CRC32 }, + { "crc64", LZMA_CHECK_CRC64 }, + { "sha256", LZMA_CHECK_SHA256 }, + { NULL, 0 } + }; + + size_t i = 0; + while (strcmp(types[i].str, optarg) != 0) { + if (types[++i].str == NULL) { + errmsg(V_ERROR, _("%s: Unknown " + "integrity check " + "type"), optarg); + my_exit(ERROR); + } + } + + opt_check = types[i].value; + break; + } + + case OPT_FILES: + opt_files_split = '\n'; + + // Fall through + + case OPT_FILES0: + if (opt_files_name != NULL) { + errmsg(V_ERROR, _("Only one file can be " + "specified with `--files'" + "or `--files0'.")); + my_exit(ERROR); + } + + if (optarg == NULL) { + opt_files_name = (char *)stdin_filename; + opt_files_file = stdin; + } else { + opt_files_name = optarg; + opt_files_file = fopen(optarg, + c == OPT_FILES ? "r" : "rb"); + if (opt_files_file == NULL) { + errmsg(V_ERROR, "%s: %s", optarg, + strerror(errno)); + my_exit(ERROR); + } + } + + break; + + default: + show_try_help(); + my_exit(ERROR); + } + } + + return; +} + + +static void +parse_environment(void) +{ + char *env = getenv("LZMA_OPT"); + if (env == NULL) + return; + + env = xstrdup(env); + + // Calculate the number of arguments in env. + unsigned int argc = 1; + bool prev_was_space = true; + for (size_t i = 0; env[i] != '\0'; ++i) { + if (isspace(env[i])) { + prev_was_space = true; + } else if (prev_was_space) { + prev_was_space = false; + if (++argc > (unsigned int)(INT_MAX)) { + errmsg(V_ERROR, _("The environment variable " + "LZMA_OPT contains too many " + "arguments")); + my_exit(ERROR); + } + } + } + + char **argv = xmalloc((argc + 1) * sizeof(char*)); + argv[0] = argv0; + argv[argc] = NULL; + + argc = 1; + prev_was_space = true; + for (size_t i = 0; env[i] != '\0'; ++i) { + if (isspace(env[i])) { + prev_was_space = true; + } else if (prev_was_space) { + prev_was_space = false; + argv[argc++] = env + i; + } + } + + parse_real((int)(argc), argv); + + free(env); + + return; +} + + +static void +set_compression_settings(void) +{ + if (filter_count == 0) { + opt_filters[0].id = LZMA_FILTER_LZMA; + opt_filters[0].options = (lzma_options_lzma *)( + lzma_preset_lzma + preset_number); + filter_count = 1; + } + + // Terminate the filter options array. + opt_filters[filter_count].id = LZMA_VLI_VALUE_UNKNOWN; + + // Optimize the filter chain a little by removing all + // Copy filters. + for (size_t i = 0; opt_filters[i].id != LZMA_VLI_VALUE_UNKNOWN; ++i) { + while (opt_filters[i].id == LZMA_FILTER_COPY) { + size_t j = i; + do { + opt_filters[j] = opt_filters[j + 1]; + } while (opt_filters[++j].id + != LZMA_VLI_VALUE_UNKNOWN); + } + } + + const uint32_t memory_limit = opt_memory / (1024 * 1024) + 1; + uint32_t memory_usage = lzma_memory_usage(opt_filters, true); + + // Don't go over the memory limits when the default + // setting is used. + if (preset_default) { + while (memory_usage > memory_limit) { + if (preset_number == 0) { + errmsg(V_ERROR, _("Memory usage limit is too " + "small for any internal " + "filter preset")); + my_exit(ERROR); + } + + --preset_number; + opt_filters[0].options = (lzma_options_lzma *)( + lzma_preset_lzma + + preset_number); + memory_usage = lzma_memory_usage(opt_filters, + true); + } + } else { + if (memory_usage > memory_limit) { + errmsg(V_ERROR, _("Memory usage limit is too small " + "for the given filter setup")); + my_exit(ERROR); + } + } + + // Limit the number of worked threads so that memory usage + // limit isn't exceeded. + // FIXME: Probably should use bytes instead of mebibytes for + // memory_usage and memory_limit. + if (memory_usage == 0) + memory_usage = 1; + + size_t thread_limit = memory_limit / memory_usage; + if (thread_limit == 0) + thread_limit = 1; + + if (opt_threads > thread_limit) + opt_threads = thread_limit; + + return; +} + + +extern char ** +parse_args(int argc, char **argv) +{ + // Check how we were called. + { + const char *name = str_filename(argv[0]); + if (name != NULL) { + if (strstr(name, "cat") != NULL) { + opt_mode = MODE_DECOMPRESS; + opt_stdout = true; + } else if (strstr(name, "un") != NULL) { + opt_mode = MODE_DECOMPRESS; + } + } + } + + // First the flags from environment + parse_environment(); + + // Then from the command line + optind = 1; + parse_real(argc, argv); + + // Never remove the source file when the destination is not on disk. + // In test mode the data is written nowhere, but setting opt_stdout + // will make the rest of the code behave well. + if (opt_stdout || opt_mode == MODE_TEST) { + opt_keep_original = true; + opt_stdout = true; + } + + if (opt_mode == MODE_COMPRESS) + set_compression_settings(); + + // If no filenames are given, use stdin. + if (argv[optind] == NULL && opt_files_name == NULL) { + // We don't modify or free() the "-" constant. + static char *argv_stdin[2] = { (char *)"-", NULL }; + return argv_stdin; + } + + return argv + optind; +} diff --git a/src/lzma/args.h b/src/lzma/args.h new file mode 100644 index 00000000..4f19a01e --- /dev/null +++ b/src/lzma/args.h @@ -0,0 +1,64 @@ +/////////////////////////////////////////////////////////////////////////////// +// +/// \file args.h +/// \brief Argument parsing +// +// Copyright (C) 2007 Lasse Collin +// +// This program is free software; you can redistribute it and/or +// modify it under the terms of the GNU Lesser General Public +// License as published by the Free Software Foundation; either +// version 2.1 of the License, or (at your option) any later version. +// +// This program is distributed in the hope that it will be useful, +// but WITHOUT ANY WARRANTY; without even the implied warranty of +// MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU +// Lesser General Public License for more details. +// +/////////////////////////////////////////////////////////////////////////////// + +#ifndef ARGS_H +#define ARGS_H + +#include "private.h" + + +enum tool_mode { + MODE_COMPRESS, + MODE_DECOMPRESS, + MODE_TEST, + MODE_LIST, +}; + +enum header_type { + HEADER_AUTO, + HEADER_NATIVE, + HEADER_SINGLE, + HEADER_MULTI, + HEADER_ALONE, + // HEADER_GZIP, +}; + + +extern char *opt_suffix; + +extern char *opt_files_name; +extern char opt_files_split; +extern FILE *opt_files_file; + +extern bool opt_stdout; +extern bool opt_force; +extern bool opt_keep_original; +extern bool opt_preserve_name; +// extern bool opt_recursive; +extern enum tool_mode opt_mode; +extern enum header_type opt_header; + +extern lzma_check_type opt_check; +extern lzma_options_filter opt_filters[8]; + +extern const char *stdin_filename; + +extern char **parse_args(int argc, char **argv); + +#endif diff --git a/src/lzma/error.c b/src/lzma/error.c new file mode 100644 index 00000000..a83de27a --- /dev/null +++ b/src/lzma/error.c @@ -0,0 +1,156 @@ +/////////////////////////////////////////////////////////////////////////////// +// +/// \file error.c +/// \brief Error message printing +// +// Copyright (C) 2007 Lasse Collin +// +// This program is free software; you can redistribute it and/or +// modify it under the terms of the GNU Lesser General Public +// License as published by the Free Software Foundation; either +// version 2.1 of the License, or (at your option) any later version. +// +// This program is distributed in the hope that it will be useful, +// but WITHOUT ANY WARRANTY; without even the implied warranty of +// MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU +// Lesser General Public License for more details. +// +/////////////////////////////////////////////////////////////////////////////// + +#include "private.h" +#include <stdarg.h> + + +exit_status_type exit_status = SUCCESS; +verbosity_type verbosity = V_WARNING; +char *argv0 = NULL; +volatile sig_atomic_t user_abort = 0; + + +extern const char * +str_strm_error(lzma_ret code) +{ + switch (code) { + case LZMA_OK: + return _("Operation successful"); + + case LZMA_STREAM_END: + return _("Operation finished successfully"); + + case LZMA_PROG_ERROR: + return _("Internal error (bug)"); + + case LZMA_DATA_ERROR: + return _("Compressed data is corrupt"); + + case LZMA_MEM_ERROR: + return strerror(ENOMEM); + + case LZMA_BUF_ERROR: + return _("Unexpected end of input"); + + case LZMA_HEADER_ERROR: + return _("Unsupported options"); + + case LZMA_UNSUPPORTED_CHECK: + return _("Unsupported integrity check type"); + + default: + return NULL; + } +} + + +extern void +set_exit_status(exit_status_type new_status) +{ + static pthread_mutex_t mutex = PTHREAD_MUTEX_INITIALIZER; + pthread_mutex_lock(&mutex); + + if (new_status != WARNING || exit_status == SUCCESS) + exit_status = new_status; + + pthread_mutex_unlock(&mutex); + return; +} + + +extern void lzma_attribute((noreturn)) +my_exit(int status) +{ + // Close stdout. If something goes wrong, print an error message + // to stderr. + { + const int ferror_err = ferror(stdout); + const int fclose_err = fclose(stdout); + if (fclose_err) { + errmsg(V_ERROR, _("Writing to standard output " + "failed: %s"), strerror(errno)); + status = ERROR; + } else if (ferror_err) { + // Some error has occurred but we have no clue about + // the reason since fclose() succeeded. + errmsg(V_ERROR, _("Writing to standard output " + "failed: %s"), "Unknown error"); + status = ERROR; + } + } + + // Close stderr. If something goes wrong, there's nothing where we + // could print an error message. Just set the exit status. + { + const int ferror_err = ferror(stderr); + const int fclose_err = fclose(stderr); + if (fclose_err || ferror_err) + status = ERROR; + } + + exit(status); +} + + +extern void lzma_attribute((format(printf, 2, 3))) +errmsg(verbosity_type v, const char *fmt, ...) +{ + va_list ap; + + if (v <= verbosity) { + va_start(ap, fmt); + + static pthread_mutex_t mutex = PTHREAD_MUTEX_INITIALIZER; + pthread_mutex_lock(&mutex); + + fprintf(stderr, "%s: ", argv0); + vfprintf(stderr, fmt, ap); + fprintf(stderr, "\n"); + + pthread_mutex_unlock(&mutex); + + va_end(ap); + } + + if (v == V_ERROR) + set_exit_status(ERROR); + else if (v == V_WARNING) + set_exit_status(WARNING); + + return; +} + + +extern void +out_of_memory(void) +{ + errmsg(V_ERROR, "%s", strerror(ENOMEM)); + user_abort = 1; + return; +} + + +extern void +internal_error(void) +{ + errmsg(V_ERROR, _("Internal error (bug)")); + user_abort = 1; + return; +} diff --git a/src/lzma/error.h b/src/lzma/error.h new file mode 100644 index 00000000..34ec30e1 --- /dev/null +++ b/src/lzma/error.h @@ -0,0 +1,67 @@ +/////////////////////////////////////////////////////////////////////////////// +// +/// \file error.c +/// \brief Error message printing +// +// Copyright (C) 2007 Lasse Collin +// +// This program is free software; you can redistribute it and/or +// modify it under the terms of the GNU Lesser General Public +// License as published by the Free Software Foundation; either +// version 2.1 of the License, or (at your option) any later version. +// +// This program is distributed in the hope that it will be useful, +// but WITHOUT ANY WARRANTY; without even the implied warranty of +// MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU +// Lesser General Public License for more details. +// +/////////////////////////////////////////////////////////////////////////////// + +#ifndef ERROR_H +#define ERROR_H + +#include "private.h" + + +typedef enum { + SUCCESS = 0, + ERROR = 1, + WARNING = 2, +} exit_status_type; + + +typedef enum { + V_SILENT, + V_ERROR, + V_WARNING, + V_VERBOSE, + V_DEBUG, +} verbosity_type; + + +extern exit_status_type exit_status; + +extern verbosity_type verbosity; + +/// Like GNU's program_invocation_name but portable +extern char *argv0; + +/// Once this is non-zero, all threads must shutdown and clean up incomplete +/// output files from the disk. +extern volatile sig_atomic_t user_abort; + + +extern const char * str_strm_error(lzma_ret code); + +extern void errmsg(verbosity_type v, const char *fmt, ...) + lzma_attribute((format(printf, 2, 3))); + +extern void set_exit_status(exit_status_type new_status); + +extern void my_exit(int status) lzma_attribute((noreturn)); + +extern void out_of_memory(void); + +extern void internal_error(void); + +#endif diff --git a/src/lzma/hardware.c b/src/lzma/hardware.c new file mode 100644 index 00000000..6cb3cdfc --- /dev/null +++ b/src/lzma/hardware.c @@ -0,0 +1,99 @@ +/////////////////////////////////////////////////////////////////////////////// +// +/// \file hardware.c +/// \brief Detection of available hardware resources +// +// Copyright (C) 2007 Lasse Collin +// +// This program is free software; you can redistribute it and/or +// modify it under the terms of the GNU Lesser General Public +// License as published by the Free Software Foundation; either +// version 2.1 of the License, or (at your option) any later version. +// +// This program is distributed in the hope that it will be useful, +// but WITHOUT ANY WARRANTY; without even the implied warranty of +// MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU +// Lesser General Public License for more details. +// +/////////////////////////////////////////////////////////////////////////////// + +#include "private.h" +#include "physmem.h" + + +/// Maximum number of free *coder* threads. This can be set with +/// the --threads=NUM command line option. +size_t opt_threads = 1; + + +/// Number of bytes of memory to use at maximum (only a rough limit). +/// This can be set with the --memory=NUM command line option. +/// If no better value can be determined, the default is 14 MiB, which +/// should be quite safe even for older systems while still allowing +/// reasonable compression ratio. +size_t opt_memory = 14 * 1024 * 1024; + + +/// Get the amount of physical memory, and set opt_memory to 1/3 of it. +/// User can then override this with --memory command line option. +static void +hardware_memory(void) +{ + uint64_t mem = physmem(); + if (mem != 0) { + mem /= 3; + +#if UINT64_MAX > SIZE_MAX + if (mem > SIZE_MAX) + mem = SIZE_MAX; +#endif + + opt_memory = mem; + } + + return; +} + + +/// Get the number of CPU cores, and set opt_threads to default to that value. +/// User can then override this with --threads command line option. +static void +hardware_cores(void) +{ +#if defined(HAVE_NUM_PROCESSORS_SYSCONF) + const long cpus = sysconf(_SC_NPROCESSORS_ONLN); + if (cpus > 0) + opt_threads = (size_t)(cpus); + +#elif defined(HAVE_NUM_PROCESSORS_SYSCTL) + int name[2] = { CTL_HW, HW_NCPU }; + int cpus; + size_t cpus_size = sizeof(cpus); + if (!sysctl(name, &cpus, &cpus_size, NULL, NULL) + && cpus_size == sizeof(cpus) && cpus > 0) + opt_threads = (size_t)(cpus); +#endif + + // Limit opt_threads so that maximum number of threads doesn't exceed. + +#if defined(_SC_THREAD_THREADS_MAX) + const long threads_max = sysconf(_SC_THREAD_THREADS_MAX); + if (threads_max > 0 && (size_t)(threads_max) < opt_threads) + opt_threads = (size_t)(threads_max); + +#elif defined(PTHREAD_THREADS_MAX) + if (opt_threads > PTHREAD_THREADS_MAX) + opt_threads = PTHREAD_THREADS_MAX; +#endif + + return; +} + + +extern void +hardware_init(void) +{ + hardware_memory(); + hardware_cores(); + return; +} diff --git a/src/lzma/hardware.h b/src/lzma/hardware.h new file mode 100644 index 00000000..d47bd29f --- /dev/null +++ b/src/lzma/hardware.h @@ -0,0 +1,31 @@ +/////////////////////////////////////////////////////////////////////////////// +// +/// \file hardware.c +/// \brief Detection of available hardware resources +// +// Copyright (C) 2007 Lasse Collin +// +// This program is free software; you can redistribute it and/or +// modify it under the terms of the GNU Lesser General Public +// License as published by the Free Software Foundation; either +// version 2.1 of the License, or (at your option) any later version. +// +// This program is distributed in the hope that it will be useful, +// but WITHOUT ANY WARRANTY; without even the implied warranty of +// MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU +// Lesser General Public License for more details. +// +/////////////////////////////////////////////////////////////////////////////// + +#ifndef HARDWARE_H +#define HARDWARE_H + +#include "private.h" + + +extern size_t opt_threads; +extern size_t opt_memory; + +extern void hardware_init(void); + +#endif diff --git a/src/lzma/help.c b/src/lzma/help.c new file mode 100644 index 00000000..ad7dd861 --- /dev/null +++ b/src/lzma/help.c @@ -0,0 +1,178 @@ +/////////////////////////////////////////////////////////////////////////////// +// +/// \file help.c +/// \brief Help messages +// +// Copyright (C) 2007 Lasse Collin +// +// This program is free software; you can redistribute it and/or +// modify it under the terms of the GNU Lesser General Public +// License as published by the Free Software Foundation; either +// version 2.1 of the License, or (at your option) any later version. +// +// This program is distributed in the hope that it will be useful, +// but WITHOUT ANY WARRANTY; without even the implied warranty of +// MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU +// Lesser General Public License for more details. +// +/////////////////////////////////////////////////////////////////////////////// + +#include "private.h" + + +extern void +show_try_help(void) +{ + // Print this with V_WARNING instead of V_ERROR to prevent it from + // showing up when --quiet has been specified. + errmsg(V_WARNING, _("Try `%s --help' for more information."), argv0); + return; +} + + +extern void lzma_attribute((noreturn)) +show_help(void) +{ + printf(_("Usage: %s [OPTION]... [FILE]...\n" + "Compress or decompress FILEs in the .lzma format.\n" + "\n"), argv0); + + puts(_("Mandatory arguments to long options are mandatory for " + "short options too.\n")); + + puts(_( +" Operation mode:\n" +"\n" +" -z, --compress force compression\n" +" -d, --decompress force decompression\n" +" -t, --test test compressed file integrity\n" +" -l, --list list block sizes, total sizes, and possible metadata\n" +)); + + puts(_( +" Operation modifiers:\n" +"\n" +" -k, --keep keep (don't delete) input files\n" +" -f, --force force overwrite of output file and (de)compress links\n" +" -c, --stdout write to standard output and don't delete input files\n" +" -S, --suffix=.SUF use suffix `.SUF' on compressed files instead of `.lzma'\n" +" -F, --format=FMT file format to encode or decode; possible values are\n" +" `auto', `native', `single', `multi', and `alone'\n" +" --files=[FILE] read filenames to process from FILE; if FILE is\n" +" omitted, filenames are read from the standard input;\n" +" filenames must be terminated with the newline character\n" +" --files0=[FILE] like --files but use the nul byte as terminator\n" +)); + + puts(_( +" Compression presets and basic compression options:\n" +"\n" +" -1 .. -2 fast compression\n" +" -3 .. -6 good compression\n" +" -7 .. -9 excellent compression, but needs a lot of memory;\n" +" default is -7 if memory limit allows\n" +"\n" +" -C, --check=CHECK integrity check type: `crc32', `crc64' (default),\n" +" or `sha256'\n" +)); + + puts(_( +" Custom filter chain for compression (alternative for using presets):\n" +"\n" +" --lzma=[OPTS] LZMA filter; OPTS is a comma-separated list of zero or\n" +" more of the following options (valid values; default):\n" +" dict=NUM dictionary size in bytes (1 - 1Gi; 8Mi)\n" +" lc=NUM number of literal context bits (0-8; 3)\n" +" lp=NUM number of literal position bits (0-4; 0)\n" +" pb=NUM number of position bits (0-4; 2)\n" +" mode=MODE compression mode (`fast' or `best'; `best')\n" +" fb=NUM number of fast bytes (5-273; 128)\n" +" mf=NAME match finder (hc3, hc4, bt2, bt3, bt4; bt4)\n" +" mfc=NUM match finder cycles; 0=automatic (default)\n" +"\n" +" --x86 x86 filter (sometimes called BCJ filter)\n" +" --powerpc PowerPC (big endian) filter\n" +" --ia64 IA64 (Itanium) filter\n" +" --arm ARM filter\n" +" --armthumb ARM-Thumb filter\n" +" --sparc SPARC filter\n" +"\n" +" --copy No filtering (useful only when specified alone)\n" +" --subblock=[OPTS] Subblock filter; valid OPTS (valid values; default):\n" +" size=NUM number of bytes of data per subblock\n" +" (1 - 256Mi; 4Ki)\n" +" rle=NUM run-length encoder chunk size (0-256; 0)\n" +)); + +/* +These aren't implemented yet. + + puts(_( +" Metadata options:\n" +"\n" +" -N, --name save or restore the original filename and time stamp\n" +" -n, --no-name do not save or restore filename and time stamp (default)\n" +" -S, --sign=KEY sign the data with GnuPG when compressing, or verify\n" +" the signature when decompressing\n")); +*/ + + puts(_( +" Resource usage options:\n" +"\n" +" -M, --memory=NUM use roughly NUM bytes of memory at maximum\n" +" -T, --threads=NUM use at maximum of NUM (de)compression threads\n" +// " --threading=STR threading style; possible values are `auto' (default),\n" +// " `files', and `stream' +)); + + puts(_( +" Other options:\n" +"\n" +" -q, --quiet suppress warnings; specify twice to suppress errors too\n" +" -v, --verbose be verbose; specify twice for even more verbose\n" +"\n" +" -h, --help display this help and exit\n" +" -V, --version display version and license information and exit\n")); + + puts(_("With no FILE, or when FILE is -, read standard input.\n")); + + size_t mem_limit = opt_memory / (1024 * 1024); + if (mem_limit == 0) + mem_limit = 1; + + puts(_("On this system and configuration, the tool will use")); + printf(_(" * roughly %zu MiB of memory at maximum; and\n"), + mem_limit); + printf(N_( + " * at maximum of one thread for (de)compression.\n\n", + " * at maximum of %zu threads for (de)compression.\n\n", + opt_threads), opt_threads); + + printf(_("Report bugs to <%s> (in English or Finnish).\n"), + PACKAGE_BUGREPORT); + + my_exit(SUCCESS); +} + + +extern void lzma_attribute((noreturn)) +show_version(void) +{ + printf( +"lzma (LZMA Utils) " PACKAGE_VERSION "\n" +"\n" +"Copyright (C) 1999-2006 Igor Pavlov\n" +"Copyright (C) 2007 Lasse Collin\n" +"\n" +"This program is free software; you can redistribute it and/or modify\n" +"it under the terms of the GNU General Public License as published by\n" +"the Free Software Foundation; either version 2 of the License, or\n" +"(at your option) any later version.\n" +"\n" +"This program is distributed in the hope that it will be useful,\n" +"but WITHOUT ANY WARRANTY; without even the implied warranty of\n" +"MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the\n" +"GNU General Public License for more details.\n" +"\n"); + my_exit(SUCCESS); +} diff --git a/src/lzma/help.h b/src/lzma/help.h new file mode 100644 index 00000000..659c66a0 --- /dev/null +++ b/src/lzma/help.h @@ -0,0 +1,32 @@ +/////////////////////////////////////////////////////////////////////////////// +// +/// \file help.h +/// \brief Help messages +// +// Copyright (C) 2007 Lasse Collin +// +// This program is free software; you can redistribute it and/or +// modify it under the terms of the GNU Lesser General Public +// License as published by the Free Software Foundation; either +// version 2.1 of the License, or (at your option) any later version. +// +// This program is distributed in the hope that it will be useful, +// but WITHOUT ANY WARRANTY; without even the implied warranty of +// MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU +// Lesser General Public License for more details. +// +/////////////////////////////////////////////////////////////////////////////// + +#ifndef HELP_H +#define HELP_H + +#include "private.h" + + +extern void show_try_help(void); + +extern void show_help(void) lzma_attribute((noreturn)); + +extern void show_version(void) lzma_attribute((noreturn)); + +#endif diff --git a/src/lzma/io.c b/src/lzma/io.c new file mode 100644 index 00000000..a7683fcc --- /dev/null +++ b/src/lzma/io.c @@ -0,0 +1,664 @@ +/////////////////////////////////////////////////////////////////////////////// +// +/// \file io.c +/// \brief File opening, unlinking, and closing +// +// Copyright (C) 2007 Lasse Collin +// +// This program is free software; you can redistribute it and/or +// modify it under the terms of the GNU Lesser General Public +// License as published by the Free Software Foundation; either +// version 2.1 of the License, or (at your option) any later version. +// +// This program is distributed in the hope that it will be useful, +// but WITHOUT ANY WARRANTY; without even the implied warranty of +// MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU +// Lesser General Public License for more details. +// +/////////////////////////////////////////////////////////////////////////////// + +#include "private.h" + +#if defined(HAVE_FUTIMES) || defined(HAVE_FUTIMESAT) +# include <sys/time.h> +#endif + +#ifndef O_SEARCH +# define O_SEARCH O_RDONLY +#endif + + +/// \brief Number of open file_pairs +/// +/// Once the main() function has requested processing of all files, +/// we wait that open_pairs drops back to zero. Then it is safe to +/// exit from the program. +static size_t open_pairs = 0; + + +/// \brief mutex for file system operations +/// +/// All file system operations are done via the functions in this file. +/// They use fchdir() to avoid some race conditions (more portable than +/// openat() & co.). +/// +/// Synchronizing all file system operations shouldn't affect speed notably, +/// since the actual reading from and writing to files is done in parallel. +static pthread_mutex_t mutex = PTHREAD_MUTEX_INITIALIZER; + + +/// This condition is invoked when a file is closed and the value of +/// the open_files variable has dropped to zero. The only listener for +/// this condition is io_finish() which is called from main(). +static pthread_cond_t io_cond = PTHREAD_COND_INITIALIZER; + + +/// True when stdout is being used by some thread +static bool stdout_in_use = false; + + +/// This condition is signalled when a thread releases stdout (no longer +/// writes data to it). +static pthread_cond_t stdout_cond = PTHREAD_COND_INITIALIZER; + + +/// \brief Directory where we were started +/// +/// This is needed when a new file, whose name was given on command line, +/// is opened. +static int start_dir; + + +static uid_t uid; +static gid_t gid; + + +extern void +io_init(void) +{ + start_dir = open(".", O_SEARCH | O_NOCTTY); + if (start_dir == -1) { + errmsg(V_ERROR, _("Cannot get file descriptor of the current " + "directory: %s"), strerror(errno)); + my_exit(ERROR); + } + + uid = getuid(); + gid = getgid(); + + return; +} + + +/// Waits until the number of open file_pairs has dropped to zero. +extern void +io_finish(void) +{ + pthread_mutex_lock(&mutex); + + while (open_pairs != 0) + pthread_cond_wait(&io_cond, &mutex); + + (void)close(start_dir); + + pthread_mutex_unlock(&mutex); + + return; +} + + +/// \brief Unlinks a file +/// +/// \param dir_fd File descriptor of the directory containing the file +/// \param name Name of the file with or without path +/// +/// \return Zero on success. On error, -1 is returned and errno set. +/// +static void +io_unlink(int dir_fd, const char *name, ino_t ino) +{ + const char *base = str_filename(name); + if (base == NULL) { + // This shouldn't happen. + errmsg(V_ERROR, _("%s: Invalid filename"), name); + return; + } + + pthread_mutex_lock(&mutex); + + if (fchdir(dir_fd)) { + errmsg(V_ERROR, _("Cannot change directory: %s"), + strerror(errno)); + } else { + struct stat st; + if (lstat(base, &st) || st.st_ino != ino) + errmsg(V_ERROR, _("%s: File seems to be moved, " + "not removing"), name); + + // There's a race condition between lstat() and unlink() + // but at least we have tried to avoid removing wrong file. + else if (unlink(base)) + errmsg(V_ERROR, _("%s: Cannot remove: %s"), + name, strerror(errno)); + } + + pthread_mutex_unlock(&mutex); + + return; +} + + +/// \brief Copies owner/group and permissions +/// +/// \todo ACL and EA support +/// +static void +io_copy_attrs(const file_pair *pair) +{ + // This function is more tricky than you may think at first. + // Blindly copying permissions may permit users to access the + // destination file who didn't have permission to access the + // source file. + + if (uid == 0 && fchown(pair->dest_fd, pair->src_st.st_uid, -1)) + errmsg(V_WARNING, _("%s: Cannot set the file owner: %s"), + pair->dest_name, strerror(errno)); + + mode_t mode; + + if (fchown(pair->dest_fd, -1, pair->src_st.st_gid)) { + errmsg(V_WARNING, _("%s: Cannot set the file group: %s"), + pair->dest_name, strerror(errno)); + // We can still safely copy some additional permissions: + // `group' must be at least as strict as `other' and + // also vice versa. + // + // NOTE: After this, the owner of the source file may + // get additional permissions. This shouldn't be too bad, + // because the owner would have had permission to chmod + // the original file anyway. + mode = ((pair->src_st.st_mode & 0070) >> 3) + & (pair->src_st.st_mode & 0007); + mode = (pair->src_st.st_mode & 0700) | (mode << 3) | mode; + } else { + // Drop the setuid, setgid, and sticky bits. + mode = pair->src_st.st_mode & 0777; + } + + if (fchmod(pair->dest_fd, mode)) + errmsg(V_WARNING, _("%s: Cannot set the file permissions: %s"), + pair->dest_name, strerror(errno)); + + // Copy the timestamps only if we have a secure function to do it. +#if defined(HAVE_FUTIMES) || defined(HAVE_FUTIMESAT) + struct timeval tv[2]; + tv[0].tv_sec = pair->src_st.st_atime; + tv[1].tv_sec = pair->src_st.st_mtime; + +# if defined(HAVE_STRUCT_STAT_ST_ATIM_TV_NSEC) + tv[0].tv_usec = pair->src_st.st_atim.tv_nsec / 1000; +# elif defined(HAVE_STRUCT_STAT_ST_ATIMESPEC_TV_NSEC) + tv[0].tv_usec = pair->src_st.st_atimespec.tv_nsec / 1000; +# else + tv[0].tv_usec = 0; +# endif + +# if defined(HAVE_STRUCT_STAT_ST_MTIM_TV_NSEC) + tv[1].tv_usec = pair->src_st.st_mtim.tv_nsec / 1000; +# elif defined(HAVE_STRUCT_STAT_ST_MTIMESPEC_TV_NSEC) + tv[1].tv_usec = pair->src_st.st_mtimespec.tv_nsec / 1000; +# else + tv[1].tv_usec = 0; +# endif + +# ifdef HAVE_FUTIMES + (void)futimes(pair->dest_fd, tv); +# else + (void)futimesat(pair->dest_fd, NULL, tv); +# endif +#endif + + return; +} + + +/// Opens and changes into the directory containing the source file. +static int +io_open_dir(file_pair *pair) +{ + if (pair->src_name == stdin_filename) + return 0; + + if (fchdir(start_dir)) { + errmsg(V_ERROR, _("Cannot change directory: %s"), + strerror(errno)); + return -1; + } + + const char *split = strrchr(pair->src_name, '/'); + if (split == NULL) { + pair->dir_fd = start_dir; + } else { + // Copy also the slash. It's needed to support filenames + // like "/foo" (dirname being "/"), and it never hurts anyway. + const size_t dirname_len = split - pair->src_name + 1; + char dirname[dirname_len + 1]; + memcpy(dirname, pair->src_name, dirname_len); + dirname[dirname_len] = '\0'; + + // Open the directory and change into it. + pair->dir_fd = open(dirname, O_SEARCH | O_NOCTTY); + if (pair->dir_fd == -1 || fchdir(pair->dir_fd)) { + errmsg(V_ERROR, _("%s: Cannot open the directory " + "containing the file: %s"), + pair->src_name, strerror(errno)); + (void)close(pair->dir_fd); + return -1; + } + } + + return 0; +} + + +static void +io_close_dir(file_pair *pair) +{ + if (pair->dir_fd != start_dir) + (void)close(pair->dir_fd); + + return; +} + + +/// Opens the source file. The file is opened using the plain filename without +/// path, thus the file must be in the current working directory. This is +/// ensured because io_open_dir() is always called before this function. +static int +io_open_src(file_pair *pair) +{ + if (pair->src_name == stdin_filename) { + pair->src_fd = STDIN_FILENO; + } else { + // Strip the pathname. Thanks to io_open_dir(), the file + // is now in the current working directory. + const char *filename = str_filename(pair->src_name); + if (filename == NULL) + return -1; + + // Symlinks are followed if --stdout or --force has been + // specified. + const bool follow_symlinks = opt_stdout || opt_force; + pair->src_fd = open(filename, O_RDONLY | O_NOCTTY + | (follow_symlinks ? 0 : O_NOFOLLOW)); + if (pair->src_fd == -1) { + // Give an understandable error message in if reason + // for failing was that the file was a symbolic link. + // - Linux, OpenBSD, Solaris: ELOOP + // - FreeBSD: EMLINK + // - Tru64: ENOTSUP + // It seems to be safe to check for all these, since + // those errno values aren't used for other purporses + // on any of the listed operating system *when* the + // above flags are used with open(). + if (!follow_symlinks + && (errno == ELOOP +#ifdef EMLINK + || errno == EMLINK +#endif +#ifdef ENOTSUP + || errno == ENOTSUP +#endif + )) { + errmsg(V_WARNING, _("%s: Is a symbolic link, " + "skipping"), pair->src_name); + } else { + errmsg(V_ERROR, "%s: %s", pair->src_name, + strerror(errno)); + } + + return -1; + } + + if (fstat(pair->src_fd, &pair->src_st)) { + errmsg(V_ERROR, "%s: %s", pair->src_name, + strerror(errno)); + goto error; + } + + if (S_ISDIR(pair->src_st.st_mode)) { + errmsg(V_WARNING, _("%s: Is a directory, skipping"), + pair->src_name); + goto error; + } + + if (!opt_stdout) { + if (!opt_force && !S_ISREG(pair->src_st.st_mode)) { + errmsg(V_WARNING, _("%s: Not a regular file, " + "skipping"), pair->src_name); + goto error; + } + + if (pair->src_st.st_mode & (S_ISUID | S_ISGID)) { + // Setuid and setgid files are rejected even + // with --force. This is good for security + // (hopefully) but it's a bit weird to reject + // file when --force was given. At least this + // matches gzip's behavior. + errmsg(V_WARNING, _("%s: File has setuid or " + "setgid bit set, skipping"), + pair->src_name); + goto error; + } + + if (!opt_force && (pair->src_st.st_mode & S_ISVTX)) { + errmsg(V_WARNING, _("%s: File has sticky bit " + "set, skipping"), + pair->src_name); + goto error; + } + + if (pair->src_st.st_nlink > 1) { + errmsg(V_WARNING, _("%s: Input file has more " + "than one hard link, " + "skipping"), pair->src_name); + goto error; + } + } + } + + return 0; + +error: + (void)close(pair->src_fd); + return -1; +} + + +/// \brief Closes source file of the file_pair structure +/// +/// \param pair File whose src_fd should be closed +/// \param success If true, the file will be removed from the disk if +/// closing succeeds and --keep hasn't been used. +static void +io_close_src(file_pair *pair, bool success) +{ + if (pair->src_fd == STDIN_FILENO || pair->src_fd == -1) + return; + + if (close(pair->src_fd)) { + errmsg(V_ERROR, _("%s: Closing the file failed: %s"), + pair->src_name, strerror(errno)); + } else if (success && !opt_keep_original) { + io_unlink(pair->dir_fd, pair->src_name, pair->src_st.st_ino); + } + + return; +} + + +static int +io_open_dest(file_pair *pair) +{ + if (opt_stdout || pair->src_fd == STDIN_FILENO) { + // We don't modify or free() this. + pair->dest_name = (char *)"(stdout)"; + pair->dest_fd = STDOUT_FILENO; + + // Synchronize the order in which files get written to stdout. + // Unlocking the mutex is safe, because opening the file_pair + // can no longer fail. + while (stdout_in_use) + pthread_cond_wait(&stdout_cond, &mutex); + + stdout_in_use = true; + + } else { + pair->dest_name = get_dest_name(pair->src_name); + if (pair->dest_name == NULL) + return -1; + + // This cannot fail, because get_dest_name() doesn't return + // invalid names. + const char *filename = str_filename(pair->dest_name); + assert(filename != NULL); + + pair->dest_fd = open(filename, O_WRONLY | O_NOCTTY | O_CREAT + | (opt_force ? O_TRUNC : O_EXCL), + S_IRUSR | S_IWUSR); + if (pair->dest_fd == -1) { + errmsg(V_ERROR, "%s: %s", pair->dest_name, + strerror(errno)); + free(pair->dest_name); + return -1; + } + + // If this really fails... well, we have a safe fallback. + struct stat st; + if (fstat(pair->dest_fd, &st)) + pair->dest_ino = 0; + else + pair->dest_ino = st.st_ino; + } + + return 0; +} + + +/// \brief Closes destination file of the file_pair structure +/// +/// \param pair File whose dest_fd should be closed +/// \param success If false, the file will be removed from the disk. +/// +/// \return Zero if closing succeeds. On error, -1 is returned and +/// error message printed. +static int +io_close_dest(file_pair *pair, bool success) +{ + if (pair->dest_fd == -1) + return 0; + + if (pair->dest_fd == STDOUT_FILENO) { + stdout_in_use = false; + pthread_cond_signal(&stdout_cond); + return 0; + } + + if (close(pair->dest_fd)) { + errmsg(V_ERROR, _("%s: Closing the file failed: %s"), + pair->dest_name, strerror(errno)); + + // Closing destination file failed, so we cannot trust its + // contents. Get rid of junk: + io_unlink(pair->dir_fd, pair->dest_name, pair->dest_ino); + free(pair->dest_name); + return -1; + } + + // If the operation using this file wasn't successful, we git rid + // of the junk file. + if (!success) + io_unlink(pair->dir_fd, pair->dest_name, pair->dest_ino); + + free(pair->dest_name); + + return 0; +} + + +extern file_pair * +io_open(const char *src_name) +{ + if (is_empty_filename(src_name)) + return NULL; + + file_pair *pair = malloc(sizeof(file_pair)); + if (pair == NULL) { + out_of_memory(); + return NULL; + } + + *pair = (file_pair){ + .src_name = src_name, + .dest_name = NULL, + .dir_fd = -1, + .src_fd = -1, + .dest_fd = -1, + .src_eof = false, + }; + + pthread_mutex_lock(&mutex); + + ++open_pairs; + + if (io_open_dir(pair)) + goto error_dir; + + if (io_open_src(pair)) + goto error_src; + + if (user_abort || io_open_dest(pair)) + goto error_dest; + + pthread_mutex_unlock(&mutex); + + return pair; + +error_dest: + io_close_src(pair, false); +error_src: + io_close_dir(pair); +error_dir: + --open_pairs; + pthread_mutex_unlock(&mutex); + free(pair); + return NULL; +} + + +/// \brief Closes the file descriptors and frees the structure +extern void +io_close(file_pair *pair, bool success) +{ + if (success && pair->dest_fd != STDOUT_FILENO) + io_copy_attrs(pair); + + // Close the destination first. If it fails, we must not remove + // the source file! + if (!io_close_dest(pair, success)) { + // Closing destination file succeeded. Remove the source file + // if the operation using this file pair was successful + // and we haven't been requested to keep the source file. + io_close_src(pair, success); + } else { + // We don't care if operation using this file pair was + // successful or not, since closing the destination file + // failed. Don't remove the original file. + io_close_src(pair, false); + } + + io_close_dir(pair); + + free(pair); + + pthread_mutex_lock(&mutex); + + if (--open_pairs == 0) + pthread_cond_signal(&io_cond); + + pthread_mutex_unlock(&mutex); + + return; +} + + +/// \brief Reads from a file to a buffer +/// +/// \param pair File pair having the sourcefile open for reading +/// \param buf Destination buffer to hold the read data +/// \param size Size of the buffer; assumed be smaller than SSIZE_MAX +/// +/// \return On success, number of bytes read is returned. On end of +/// file zero is returned and pair->src_eof set to true. +/// On error, SIZE_MAX is returned and error message printed. +/// +/// \note This does no locking, thus two threads must not read from +/// the same file. This no problem in this program. +extern size_t +io_read(file_pair *pair, uint8_t *buf, size_t size) +{ + // We use small buffers here. + assert(size < SSIZE_MAX); + + size_t left = size; + + while (left > 0) { + const ssize_t amount = read(pair->src_fd, buf, left); + + if (amount == 0) { + pair->src_eof = true; + break; + } + + if (amount == -1) { + if (errno == EINTR) { + if (user_abort) + return SIZE_MAX; + + continue; + } + + errmsg(V_ERROR, _("%s: Read error: %s"), + pair->src_name, strerror(errno)); + + // FIXME Is this needed? + pair->src_eof = true; + + return SIZE_MAX; + } + + buf += (size_t)(amount); + left -= (size_t)(amount); + } + + return size - left; +} + + +/// \brief Writes a buffer to a file +/// +/// \param pair File pair having the destination file open for writing +/// \param buf Buffer containing the data to be written +/// \param size Size of the buffer; assumed be smaller than SSIZE_MAX +/// +/// \return On success, zero is returned. On error, -1 is returned +/// and error message printed. +/// +/// \note This does no locking, thus two threads must not write to +/// the same file. This no problem in this program. +extern int +io_write(const file_pair *pair, const uint8_t *buf, size_t size) +{ + assert(size < SSIZE_MAX); + + while (size > 0) { + const ssize_t amount = write(pair->dest_fd, buf, size); + if (amount == -1) { + if (errno == EINTR) { + if (user_abort) + return -1; + + continue; + } + + errmsg(V_ERROR, _("%s: Write error: %s"), + pair->dest_name, strerror(errno)); + return -1; + } + + buf += (size_t)(amount); + size -= (size_t)(amount); + } + + return 0; +} diff --git a/src/lzma/io.h b/src/lzma/io.h new file mode 100644 index 00000000..d1aa17f4 --- /dev/null +++ b/src/lzma/io.h @@ -0,0 +1,60 @@ +/////////////////////////////////////////////////////////////////////////////// +// +/// \file io.h +/// \brief I/O types and functions +// +// Copyright (C) 2007 Lasse Collin +// +// This program is free software; you can redistribute it and/or +// modify it under the terms of the GNU Lesser General Public +// License as published by the Free Software Foundation; either +// version 2.1 of the License, or (at your option) any later version. +// +// This program is distributed in the hope that it will be useful, +// but WITHOUT ANY WARRANTY; without even the implied warranty of +// MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU +// Lesser General Public License for more details. +// +/////////////////////////////////////////////////////////////////////////////// + +#ifndef IO_H +#define IO_H + +#include "private.h" + +#if BUFSIZ <= 1024 +# define IO_BUFFER_SIZE 8192 +#else +# define IO_BUFFER_SIZE BUFSIZ +#endif + + +typedef struct { + const char *src_name; + char *dest_name; + + int dir_fd; + int src_fd; + int dest_fd; + + struct stat src_st; + ino_t dest_ino; + + bool src_eof; +} file_pair; + + +extern void io_init(void); + +extern void io_finish(void); + +extern file_pair *io_open(const char *src_name); + +extern void io_close(file_pair *pair, bool success); + +extern size_t io_read(file_pair *pair, uint8_t *buf, size_t size); + +extern int io_write(const file_pair *pair, const uint8_t *buf, size_t size); + + +#endif diff --git a/src/lzma/list.c b/src/lzma/list.c new file mode 100644 index 00000000..61eb5702 --- /dev/null +++ b/src/lzma/list.c @@ -0,0 +1,477 @@ +/////////////////////////////////////////////////////////////////////////////// +// +/// \file list.c +/// \brief Listing information about .lzma files +// +// Copyright (C) 2007 Lasse Collin +// +// This program is free software; you can redistribute it and/or +// modify it under the terms of the GNU Lesser General Public +// License as published by the Free Software Foundation; either +// version 2.1 of the License, or (at your option) any later version. +// +// This program is distributed in the hope that it will be useful, +// but WITHOUT ANY WARRANTY; without even the implied warranty of +// MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU +// Lesser General Public License for more details. +// +/////////////////////////////////////////////////////////////////////////////// + +#include "private.h" + + +/* + +1. Check the file type: native, alone, unknown + +Alone: +1. Show info about header. Don't look for concatenated parts. + +Native: +1. Check that Stream Header is valid. +2. Seek to the end of the file. +3. Skip padding. +4. Reverse decode Stream Footer. +5. Seek Backward Size bytes. +6. + +*/ + + +static void +unsupported_file(file_handle *handle) +{ + errmsg(V_ERROR, "%s: Unsupported file type", handle->name); + set_exit_status(ERROR); + (void)io_close(handle); + return; +} + + +/// Primitive escaping function, that escapes only ASCII control characters. +static void +print_escaped(const uint8_t *str) +{ + while (*str != '\0') { + if (*str <= 0x1F || *str == 0x7F) + printf("\\x%02X", *str); + else + putchar(*str); + + ++str; + } + + return; +} + + +static void +list_native(file_handle *handle) +{ + lzma_stream strm = LZMA_STREAM_INIT; + lzma_stream_flags flags; + lzma_ret ret = lzma_stream_header_decoder(&strm, &flags); + +} + + +static void +list_alone(const listing_handle *handle) +{ + if (handle->buffer[0] > (4 * 5 + 4) * 9 + 8) { + unsupported_file(handle); + return; + } + + const unsigned int pb = handle->buffer[0] / (9 * 5); + handle->buffer[0] -= pb * 9 * 5; + const unsigned int lp = handle->buffer[0] / 9; + const unsigned int lc = handle->buffer[0] - lp * 9; + + uint32_t dict = 0; + for (size_t i = 1; i < 5; ++i) { + dict <<= 8; + dict |= header[i]; + } + + if (dict > LZMA_DICTIONARY_SIZE_MAX) { + unsupported_file(handle); + return; + } + + uint64_t uncompressed_size = 0; + for (size_t i = 5; i < 13; ++i) { + uncompressed_size <<= 8; + uncompressed_size |= header[i]; + } + + // Reject files with uncompressed size of 256 GiB or more. It's + // an arbitrary limit trying to avoid at least some false positives. + if (uncompressed_size != UINT64_MAX + && uncompressed_size >= (UINT64_C(1) << 38)) { + unsupported_file(handle); + return; + } + + if (verbosity < V_WARNING) { + printf("name="); + print_escaped(handle->name); + printf("\nformat=alone\n"); + + if (uncompressed_size == UINT64_MAX) + printf("uncompressed_size=unknown\n"); + else + printf("uncompressed_size=%" PRIu64 "\n", + uncompressed_size); + + printf("dict=%" PRIu32 "\n", dict); + + printf("lc=%u\nlp=%u\npb=%u\n\n", lc, lp, pb); + + } else { + printf("File name: "); + print_escaped(handle->name); + printf("\nFile format: LZMA_Alone\n") + + printf("Uncompressed size: "); + if (uncompressed_size == UINT64_MAX) + printf("unknown\n"); + else + printf("%," PRIu64 " bytes (%" PRIu64 " MiB)\n", + uncompressed_size, + (uncompressed_size + 1024 * 512) + / (1024 * 1024)); + + printf("Dictionary size: %," PRIu32 " bytes " + "(%" PRIu32 " MiB)\n", + dict, (dict + 1024 * 512) / (1024 * 1024)); + + printf("Literal context bits (lc): %u\n", lc); + printf("Literal position bits (lc): %u\n", lp); + printf("Position bits (pb): %u\n", pb); + } + + return; +} + + + + +typedef struct { + const char *filename; + struct stat st; + int fd; + + lzma_stream strm; + lzma_stream_flags stream_flags; + lzma_info *info; + + lzma_vli backward_size; + lzma_vli uncompressed_size; + + size_t buffer_size; + uint8_t buffer[IO_BUFFER_SIZE]; +} listing_handle; + + +static bool +listing_pread(listing_handle *handle, uint64_t offset) +{ + if (offset >= (uint64_t)(handle->st.st_size)) { + errmsg(V_ERROR, "%s: Trying to read past the end of " + "the file.", handle->filename); + return true; + } + +#ifdef HAVE_PREAD + const ssize_t ret = pread(handle->fd, handle->buffer, IO_BUFFER_SIZE, + (off_t)(offset)); +#else + // Use lseek() + read() since we don't have pread(). We don't care + // to which offset the reading position is left. + if (lseek(handle->fd, (off_t)(offset), SEEK_SET) == -1) { + errmsg(V_ERROR, "%s: %s", handle->filename, strerror(errno)); + return true; + } + + const ssize_t ret = read(handle->fd, handle->buffer, IO_BUFFER_SIZE); +#endif + + if (ret == -1) { + errmsg(V_ERROR, "%s: %s", handle->filename, strerror(errno)); + return true; + } + + if (ret == 0) { + errmsg(V_ERROR, "%s: Trying to read past the end of " + "the file.", handle->filename); + return true; + } + + handle->buffer_size = (size_t)(ret); + return false; +} + + + +static bool +parse_stream_header(listing_handle *handle) +{ + if (listing_pread(handle, 0)) + return true; + + // TODO Got enough input? + + lzma_ret ret = lzma_stream_header_decoder( + &handle->strm, &handle->stream_flags); + if (ret != LZMA_OK) { + errmsg(V_ERROR, "%s: %s", handle->name, str_strm_error(ret)); + return true; + } + + handle->strm.next_in = handle->buffer; + handle->strm.avail_in = handle->buffer_size; + ret = lzma_code(&handle->strm, LZMA_RUN); + if (ret != LZMA_STREAM_END) { + assert(ret != LZMA_OK); + errmsg(V_ERROR, "%s: %s", handle->name, str_strm_error(ret)); + return true; + } + + return false; +} + + +static bool +parse_stream_tail(listing_handle *handle) +{ + uint64_t offset = (uint64_t)(handle->st.st_size); + + // Skip padding + do { + if (offset == 0) { + errmsg(V_ERROR, "%s: %s", handle->name, + str_strm_error(LZMA_DATA_ERROR)); + return true; + } + + if (offset < IO_BUFFER_SIZE) + offset = 0; + else + offset -= IO_BUFFER_SIZE; + + if (listing_pread(handle, offset)) + return true; + + while (handle->buffer_size > 0 + && handle->buffer[handle->buffer_size - 1] + == '\0') + --handle->buffer_size; + + } while (handle->buffer_size == 0); + + if (handle->buffer_size < LZMA_STREAM_TAIL_SIZE) { + // TODO + } + + lzma_stream_flags stream_flags; + lzma_ret ret = lzma_stream_tail_decoder(&handle->strm, &stream_flags); + if (ret != LZMA_OK) { + errmsg(V_ERROR, "%s: %s", handle->name, str_strm_error(ret)); + return true; + } + + handle->strm.next_in = handle->buffer + handle->buffer_size + - LZMA_STREAM_TAIL_SIZE; + handle->strm.avail_in = LZMA_STREAM_TAIL_SIZE; + handle->buffer_size -= LZMA_STREAM_TAIL_SIZE; + ret = lzma_code(&handle->strm, LZMA_RUN); + if (ret != LZMA_OK) { + assert(ret != LZMA_OK); + errmsg(V_ERROR, "%s: %s", handle->name, str_strm_error(ret)); + return true; + } + + if (!lzma_stream_flags_is_equal(handle->stream_flags, stream_flags)) { + // TODO + // Possibly corrupt, possibly concatenated file. + } + + handle->backward_size = 0; + ret = lzma_vli_reverse_decode(&handle->backward_size, handle->buffer, + &handle->buffer_size); + if (ret != LZMA_OK) { + // It may be LZMA_BUF_ERROR too, but it doesn't make sense + // as an error message displayed to the user. + errmsg(V_ERROR, "%s: %s", handle->name, + str_strm_error(LZMA_DATA_ERROR)); + return true; + } + + if (!stream_flags.is_multi) { + handle->uncompressed_size = 0; + size_t tmp = handle->buffer_size; + ret = lzma_vli_reverse_decode(&handle->uncompressed_size, + handle->buffer, &tmp); + if (ret != LZMA_OK) + handle->uncompressed_size = LZMA_VLI_VALUE_UNKNOWN; + } + + // Calculate the Header Metadata Block start offset. + + + return false; +} + + + +static void +list_native(listing_handle *handle) +{ + lzma_memory_limitter *limitter + = lzma_memory_limitter_create(opt_memory); + if (limitter == NULL) { + errmsg(V_ERROR, + } + lzma_info *info = + + + // Parse Stream Header + // + // Single-Block Stream: + // - Parse Block Header + // - Parse Stream Footer + // - If Backward Size doesn't match, error out + // + // Multi-Block Stream: + // - Parse Header Metadata Block, if any + // - Parse Footer Metadata Block + // - Parse Stream Footer + // - If Footer Metadata Block doesn't match the Stream, error out + // + // In other words, we don't support concatened files. + if (parse_stream_header(handle)) + return; + + if (parse_block_header(handle)) + return; + + if (handle->stream_flags.is_multi) { + if (handle->block_options.is_metadata) { + if (parse_metadata(handle) + return; + } + + if (my_seek(handle, + + } else { + if (handle->block_options.is_metadata) { + FILE_IS_CORRUPT(); + return; + } + + if (parse_stream_footer(handle)) + return; + + // If Uncompressed Size isn't present in Block Header, + // it must be present in Stream Footer. + if (handle->block_options.uncompressed_size + == LZMA_VLI_VALUE_UNKNOWN + && handle->stream_flags.uncompressed_size + == LZMA_VLI_VALUE_UNKNOWN) { + FILE_IS_CORRUPT(); + return; + } + + // Construct a single-Record Index. + lzma_index *index = malloc(sizeof(lzma_index)); + if (index == NULL) { + out_of_memory(); + return; + } + + // Pohdintaa: + // Jos Block coder hoitaisi Uncompressed ja Backward Sizet, + // voisi index->total_sizeksi laittaa suoraan Backward Sizen. + index->total_size = + + if () { + + } + } + + + if (handle->block_options.is_metadata) { + if (!handle->stream_flags.is_multi) { + FILE_IS_CORRUPT(); + return; + } + + if (parse_metadata(handle)) + return; + + } +} + + + +extern void +list(const char *filename) +{ + if (strcmp(filename, "-") == 0) { + errmsg(V_ERROR, "%s: --list does not support reading from " + "standard input", filename); + return; + } + + if (is_empty_filename(filename)) + return; + + listing_handle handle; + handle.filename = filename; + + handle.fd = open(filename, O_RDONLY | O_NOCTTY); + if (handle.fd == -1) { + errmsg(V_ERROR, "%s: %s", filename, strerror(errno)); + return; + } + + if (fstat(handle.fd, &handle.st)) { + errmsg(V_ERROR, "%s: %s", filename, strerror(errno)); + goto out; + } + + if (!S_ISREG(handle.st.st_mode)) { + errmsg(V_WARNING, _("%s: Not a regular file, skipping"), + filename); + goto out; + } + + if (handle.st.st_size <= 0) { + errmsg(V_ERROR, _("%s: File is empty"), filename); + goto out; + } + + if (listing_pread(&handle, 0)) + goto out; + + if (handle.buffer[0] == 0xFF) { + if (opt_header == HEADER_ALONE) { + errmsg(V_ERROR, "%s: FIXME", filename); // FIXME + goto out; + } + + list_native(&handle); + } else { + if (opt_header != HEADER_AUTO && opt_header != HEADER_ALONE) { + errmsg(V_ERROR, "%s: FIXME", filename); // FIXME + goto out; + } + + list_alone(&handle); + } + +out: + (void)close(fd); + return; +} diff --git a/src/lzma/main.c b/src/lzma/main.c new file mode 100644 index 00000000..26edc47e --- /dev/null +++ b/src/lzma/main.c @@ -0,0 +1,254 @@ +/////////////////////////////////////////////////////////////////////////////// +// +/// \file main.c +/// \brief main() +// +// Copyright (C) 2007 Lasse Collin +// +// This program is free software; you can redistribute it and/or +// modify it under the terms of the GNU Lesser General Public +// License as published by the Free Software Foundation; either +// version 2.1 of the License, or (at your option) any later version. +// +// This program is distributed in the hope that it will be useful, +// but WITHOUT ANY WARRANTY; without even the implied warranty of +// MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU +// Lesser General Public License for more details. +// +/////////////////////////////////////////////////////////////////////////////// + +#include "private.h" +#include "open_stdxxx.h" +#include <ctype.h> + +static sig_atomic_t exit_signal = 0; + + +static void +signal_handler(int sig) +{ + // FIXME Is this thread-safe together with main()? + exit_signal = sig; + + user_abort = 1; + return; +} + + +static void +establish_signal_handlers(void) +{ + struct sigaction sa; + sa.sa_handler = &signal_handler; + sigfillset(&sa.sa_mask); + sa.sa_flags = 0; + + static const int sigs[] = { + SIGHUP, + SIGINT, + SIGPIPE, + SIGTERM, + SIGXCPU, + SIGXFSZ, + }; + + for (size_t i = 0; i < sizeof(sigs) / sizeof(sigs[0]); ++i) { + if (sigaction(sigs[i], &sa, NULL)) { + errmsg(V_ERROR, _("Cannot establish signal handlers")); + my_exit(ERROR); + } + } + + /* + SIGINFO/SIGUSR1 for status reporting? + */ +} + + +static bool +is_tty_stdin(void) +{ + const bool ret = isatty(STDIN_FILENO); + if (ret) { + // FIXME: Other threads may print between these lines. + // Maybe that should be fixed. Not a big issue in practice. + errmsg(V_ERROR, _("Compressed data not read from " + "a terminal.")); + errmsg(V_ERROR, _("Use `--force' to force decompression.")); + show_try_help(); + } + + return ret; +} + + +static bool +is_tty_stdout(void) +{ + const bool ret = isatty(STDOUT_FILENO); + if (ret) { + errmsg(V_ERROR, _("Compressed data not written to " + "a terminal.")); + errmsg(V_ERROR, _("Use `--force' to force decompression.")); + show_try_help(); + } + + return ret; +} + + +static char * +read_name(void) +{ + size_t size = 256; + size_t pos = 0; + char *name = malloc(size); + if (name == NULL) { + out_of_memory(); + return NULL; + } + + while (true) { + const int c = fgetc(opt_files_file); + if (c == EOF) { + free(name); + + if (ferror(opt_files_file)) + errmsg(V_ERROR, _("%s: Error reading " + "filenames: %s"), + opt_files_name, + strerror(errno)); + else if (pos != 0) + errmsg(V_ERROR, _("%s: Unexpected end of " + "input when reading " + "filenames"), opt_files_name); + + return NULL; + } + + if (c == '\0' || c == opt_files_split) + break; + + name[pos++] = c; + + if (pos == size) { + size *= 2; + char *tmp = realloc(name, size); + if (tmp == NULL) { + free(name); + out_of_memory(); + return NULL; + } + + name = tmp; + } + } + + if (name != NULL) + name[pos] = '\0'; + + return name; +} + + +int +main(int argc, char **argv) +{ + // Make sure that stdin, stdout, and and stderr are connected to + // a valid file descriptor. Exit immediatelly with exit code ERROR + // if we cannot make the file descriptors valid. Maybe we should + // print an error message, but our stderr could be screwed anyway. + open_stdxxx(ERROR); + + // Set the program invocation name used in various messages. + argv0 = argv[0]; + + setlocale(LC_ALL, "en_US.UTF-8"); + bindtextdomain(PACKAGE, LOCALEDIR); + textdomain(PACKAGE); + + // Set hardware-dependent default values. These can be overriden + // on the command line, thus this must be done before parse_args(). + hardware_init(); + + char **files = parse_args(argc, argv); + + if (opt_mode == MODE_COMPRESS && opt_stdout && is_tty_stdout()) + return ERROR; + + if (opt_mode == MODE_COMPRESS) + lzma_init_encoder(); + else + lzma_init_decoder(); + + io_init(); + process_init(); + + if (opt_mode == MODE_LIST) { + errmsg(V_ERROR, "--list is not implemented yet."); + my_exit(ERROR); + } + + // Hook the signal handlers. We don't need these before we start + // the actual action, so this is done after parsing the command + // line arguments. + establish_signal_handlers(); + + while (*files != NULL && !user_abort) { + if (strcmp("-", *files) == 0) { + if (!opt_force) { + if (opt_mode == MODE_COMPRESS) { + if (is_tty_stdout()) { + ++files; + continue; + } + } else if (is_tty_stdin()) { + ++files; + continue; + } + } + + if (opt_files_name == stdin_filename) { + errmsg(V_ERROR, _("Cannot read data from " + "standard input when " + "reading filenames " + "from standard input")); + ++files; + continue; + } + + *files = (char *)stdin_filename; + } + + process_file(*files++); + } + + if (opt_files_name != NULL) { + while (true) { + char *name = read_name(); + if (name == NULL) + break; + + if (name[0] != '\0') + process_file(name); + + free(name); + } + + if (opt_files_name != stdin_filename) + (void)fclose(opt_files_file); + } + + io_finish(); + + if (exit_signal != 0) { + struct sigaction sa; + sa.sa_handler = SIG_DFL; + sigfillset(&sa.sa_mask); + sa.sa_flags = 0; + sigaction(exit_signal, &sa, NULL); + raise(exit_signal); + } + + my_exit(exit_status); +} diff --git a/src/lzma/options.c b/src/lzma/options.c new file mode 100644 index 00000000..2928aafc --- /dev/null +++ b/src/lzma/options.c @@ -0,0 +1,346 @@ +/////////////////////////////////////////////////////////////////////////////// +// +/// \file options.c +/// \brief Parser for filter-specific options +// +// Copyright (C) 2007 Lasse Collin +// +// This program is free software; you can redistribute it and/or +// modify it under the terms of the GNU Lesser General Public +// License as published by the Free Software Foundation; either +// version 2.1 of the License, or (at your option) any later version. +// +// This program is distributed in the hope that it will be useful, +// but WITHOUT ANY WARRANTY; without even the implied warranty of +// MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU +// Lesser General Public License for more details. +// +/////////////////////////////////////////////////////////////////////////////// + +#include "private.h" + + +/////////////////// +// Generic stuff // +/////////////////// + +typedef struct { + const char *name; + uint64_t id; +} name_id_map; + + +typedef struct { + const char *name; + const name_id_map *map; + uint64_t min; + uint64_t max; +} option_map; + + +/// Parses option=value pairs that are separated with colons, semicolons, +/// or commas: opt=val:opt=val;opt=val,opt=val +/// +/// Each option is a string, that is converted to an integer using the +/// index where the option string is in the array. +/// +/// Value can be either a number with minimum and maximum value limit, or +/// a string-id map mapping a list of possible string values to integers. +/// +/// When parsing both option and value succeed, a filter-specific function +/// is called, which should update the given value to filter-specific +/// options structure. +/// +/// \param str String containing the options from the command line +/// \param opts Filter-specific option map +/// \param set Filter-specific function to update filter_options +/// \param filter_options Pointer to filter-specific options structure +/// +/// \return Returns only if no errors occur. +/// +static void +parse_options(const char *str, const option_map *opts, + void (*set)(void *filter_options, + uint32_t key, uint64_t value), + void *filter_options) +{ + if (str == NULL || str[0] == '\0') + return; + + char *s = xstrdup(str); + char *name = s; + + while (true) { + char *split = strchr(name, ','); + if (split != NULL) + *split = '\0'; + + char *value = strchr(name, '='); + if (value != NULL) + *value++ = '\0'; + + if (value == NULL || value[0] == '\0') { + errmsg(V_ERROR, _("%s: Options must be `name=value' " + "pairs separated with commas"), + str); + my_exit(ERROR); + } + + // Look for the option name from the option map. + bool found = false; + for (size_t i = 0; opts[i].name != NULL; ++i) { + if (strcmp(name, opts[i].name) != 0) + continue; + + if (opts[i].map == NULL) { + // value is an integer. + const uint64_t v = str_to_uint64(name, value, + opts[i].min, opts[i].max); + set(filter_options, i, v); + } else { + // value is a string which we should map + // to an integer. + size_t j; + for (j = 0; opts[i].map[j].name != NULL; ++j) { + if (strcmp(opts[i].map[j].name, value) + == 0) + break; + } + + if (opts[i].map[j].name == NULL) { + errmsg(V_ERROR, _("%s: Invalid option " + "value"), value); + my_exit(ERROR); + } + + set(filter_options, i, j); + } + + found = true; + break; + } + + if (!found) { + errmsg(V_ERROR, _("%s: Invalid option name"), name); + my_exit(ERROR); + } + + if (split == NULL) + break; + + name = split + 1; + } + + free(s); + return; +} + + +////////////// +// Subblock // +////////////// + +enum { + OPT_SIZE, + OPT_RLE, + OPT_ALIGN, +}; + + +static void +set_subblock(void *options, uint32_t key, uint64_t value) +{ + lzma_options_subblock *opt = options; + + switch (key) { + case OPT_SIZE: + opt->subblock_data_size = value; + break; + + case OPT_RLE: + opt->rle = value; + break; + + case OPT_ALIGN: + opt->alignment = value; + break; + } +} + + +extern lzma_options_subblock * +parse_options_subblock(const char *str) +{ + static const option_map opts[] = { + { "size", NULL, LZMA_SUBBLOCK_DATA_SIZE_MIN, + LZMA_SUBBLOCK_DATA_SIZE_MAX }, + { "rle", NULL, LZMA_SUBBLOCK_RLE_OFF, + LZMA_SUBBLOCK_RLE_MAX }, + { "align",NULL, LZMA_SUBBLOCK_ALIGNMENT_MIN, + LZMA_SUBBLOCK_ALIGNMENT_MAX }, + { NULL, NULL, 0, 0 } + }; + + lzma_options_subblock *options + = xmalloc(sizeof(lzma_options_subblock)); + *options = (lzma_options_subblock){ + .allow_subfilters = false, + .alignment = LZMA_SUBBLOCK_ALIGNMENT_DEFAULT, + .subblock_data_size = LZMA_SUBBLOCK_DATA_SIZE_DEFAULT, + .rle = LZMA_SUBBLOCK_RLE_OFF, + }; + + parse_options(str, opts, &set_subblock, options); + + return options; +} + + +/////////// +// Delta // +/////////// + +enum { + OPT_DISTANCE, +}; + + +static void +set_delta(void *options, uint32_t key, uint64_t value) +{ + lzma_options_delta *opt = options; + switch (key) { + case OPT_DISTANCE: + opt->distance = value; + break; + } +} + + +extern lzma_options_delta * +parse_options_delta(const char *str) +{ + static const option_map opts[] = { + { "distance", NULL, LZMA_DELTA_DISTANCE_MIN, + LZMA_DELTA_DISTANCE_MAX }, + { NULL, NULL, 0, 0 } + }; + + lzma_options_delta *options = xmalloc(sizeof(lzma_options_subblock)); + *options = (lzma_options_delta){ + // It's hard to give a useful default for this. + .distance = LZMA_DELTA_DISTANCE_MIN, + }; + + parse_options(str, opts, &set_delta, options); + + return options; +} + + +////////// +// LZMA // +////////// + +enum { + OPT_DICT, + OPT_LC, + OPT_LP, + OPT_PB, + OPT_MODE, + OPT_FB, + OPT_MF, + OPT_MC +}; + + +static void +set_lzma(void *options, uint32_t key, uint64_t value) +{ + lzma_options_lzma *opt = options; + + switch (key) { + case OPT_DICT: + opt->dictionary_size = value; + break; + + case OPT_LC: + opt->literal_context_bits = value; + break; + + case OPT_LP: + opt->literal_pos_bits = value; + break; + + case OPT_PB: + opt->pos_bits = value; + break; + + case OPT_MODE: + opt->mode = value; + break; + + case OPT_FB: + opt->fast_bytes = value; + break; + + case OPT_MF: + opt->match_finder = value; + break; + + case OPT_MC: + opt->match_finder_cycles = value; + break; + } +} + + +extern lzma_options_lzma * +parse_options_lzma(const char *str) +{ + static const name_id_map modes[] = { + { "fast", LZMA_MODE_FAST }, + { "best", LZMA_MODE_BEST }, + { NULL, 0 } + }; + + static const name_id_map mfs[] = { + { "hc3", LZMA_MF_HC3 }, + { "hc4", LZMA_MF_HC4 }, + { "bt2", LZMA_MF_BT2 }, + { "bt3", LZMA_MF_BT3 }, + { "bt4", LZMA_MF_BT4 }, + { NULL, 0 } + }; + + static const option_map opts[] = { + { "dict", NULL, LZMA_DICTIONARY_SIZE_MIN, + LZMA_DICTIONARY_SIZE_MAX }, + { "lc", NULL, LZMA_LITERAL_CONTEXT_BITS_MIN, + LZMA_LITERAL_CONTEXT_BITS_MAX }, + { "lp", NULL, LZMA_LITERAL_POS_BITS_MIN, + LZMA_LITERAL_POS_BITS_MAX }, + { "pb", NULL, LZMA_POS_BITS_MIN, LZMA_POS_BITS_MAX }, + { "mode", modes, 0, 0 }, + { "fb", NULL, LZMA_FAST_BYTES_MIN, LZMA_FAST_BYTES_MAX }, + { "mf", mfs, 0, 0 }, + { "mc", NULL, 0, UINT32_MAX }, + { NULL, NULL, 0, 0 } + }; + + lzma_options_lzma *options = xmalloc(sizeof(lzma_options_lzma)); + *options = (lzma_options_lzma){ + .dictionary_size = LZMA_DICTIONARY_SIZE_DEFAULT, + .literal_context_bits = LZMA_LITERAL_CONTEXT_BITS_DEFAULT, + .literal_pos_bits = LZMA_LITERAL_POS_BITS_DEFAULT, + .pos_bits = LZMA_POS_BITS_DEFAULT, + .mode = LZMA_MODE_BEST, + .fast_bytes = LZMA_FAST_BYTES_DEFAULT, + .match_finder = LZMA_MF_BT4, + .match_finder_cycles = 0, + }; + + parse_options(str, opts, &set_lzma, options); + + return options; +} diff --git a/src/lzma/options.h b/src/lzma/options.h new file mode 100644 index 00000000..885c5969 --- /dev/null +++ b/src/lzma/options.h @@ -0,0 +1,46 @@ +/////////////////////////////////////////////////////////////////////////////// +// +/// \file options.h +/// \brief Parser for filter-specific options +// +// Copyright (C) 2007 Lasse Collin +// +// This program is free software; you can redistribute it and/or +// modify it under the terms of the GNU Lesser General Public +// License as published by the Free Software Foundation; either +// version 2.1 of the License, or (at your option) any later version. +// +// This program is distributed in the hope that it will be useful, +// but WITHOUT ANY WARRANTY; without even the implied warranty of +// MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU +// Lesser General Public License for more details. +// +/////////////////////////////////////////////////////////////////////////////// + +#ifndef OPTIONS_H +#define OPTIONS_H + +#include "private.h" + + +/// \brief Parser for Subblock options +/// +/// \return Pointer to allocated options structure. +/// Doesn't return on error. +extern lzma_options_subblock *parse_options_subblock(const char *str); + + +/// \brief Parser for Delta options +/// +/// \return Pointer to allocated options structure. +/// Doesn't return on error. +extern lzma_options_delta *parse_options_delta(const char *str); + + +/// \brief Parser for LZMA options +/// +/// \return Pointer to allocated options structure. +/// Doesn't return on error. +extern lzma_options_lzma *parse_options_lzma(const char *str); + +#endif diff --git a/src/lzma/private.h b/src/lzma/private.h new file mode 100644 index 00000000..89afac9b --- /dev/null +++ b/src/lzma/private.h @@ -0,0 +1,55 @@ +/////////////////////////////////////////////////////////////////////////////// +// +/// \file private.h +/// \brief Common includes, definions, and prototypes +// +// Copyright (C) 2007 Lasse Collin +// +// This program is free software; you can redistribute it and/or +// modify it under the terms of the GNU Lesser General Public +// License as published by the Free Software Foundation; either +// version 2.1 of the License, or (at your option) any later version. +// +// This program is distributed in the hope that it will be useful, +// but WITHOUT ANY WARRANTY; without even the implied warranty of +// MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU +// Lesser General Public License for more details. +// +/////////////////////////////////////////////////////////////////////////////// + +#ifndef PRIVATE_H +#define PRIVATE_H + +#include "sysdefs.h" + +#ifdef HAVE_ERRNO_H +# include <errno.h> +#else +extern int errno; +#endif + +#include <sys/stat.h> +#include <limits.h> +#include <signal.h> +#include <pthread.h> +#include <locale.h> +#include <stdio.h> +#include <fcntl.h> +#include <unistd.h> + +#include "gettext.h" +#define _(msgid) gettext(msgid) +#define N_(msgid1, msgid2, n) ngettext(msgid1, msgid2, n) + +#include "alloc.h" +#include "args.h" +#include "error.h" +#include "hardware.h" +#include "help.h" +#include "io.h" +#include "options.h" +#include "process.h" +#include "suffix.h" +#include "util.h" + +#endif diff --git a/src/lzma/process.c b/src/lzma/process.c new file mode 100644 index 00000000..10a76b74 --- /dev/null +++ b/src/lzma/process.c @@ -0,0 +1,458 @@ +/////////////////////////////////////////////////////////////////////////////// +// +/// \file process.c +/// \brief Compresses or uncompresses a file +// +// Copyright (C) 2007 Lasse Collin +// +// This program is free software; you can redistribute it and/or +// modify it under the terms of the GNU Lesser General Public +// License as published by the Free Software Foundation; either +// version 2.1 of the License, or (at your option) any later version. +// +// This program is distributed in the hope that it will be useful, +// but WITHOUT ANY WARRANTY; without even the implied warranty of +// MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU +// Lesser General Public License for more details. +// +/////////////////////////////////////////////////////////////////////////////// + +#include "private.h" + + +typedef struct { + lzma_stream strm; + void *options; + + file_pair *pair; + + /// We don't need this for *anything* but seems that at least with + /// glibc pthread_create() doesn't allow NULL. + pthread_t thread; + + bool in_use; + +} thread_data; + + +/// Number of available threads +static size_t free_threads; + +/// Thread-specific data +static thread_data *threads; + +static pthread_mutex_t mutex = PTHREAD_MUTEX_INITIALIZER; +static pthread_cond_t cond = PTHREAD_COND_INITIALIZER; + +/// Attributes of new coder threads. They are created in detached state. +/// Coder threads signal to the service thread themselves when they are done. +static pthread_attr_t thread_attr; + + +////////// +// Init // +////////// + +extern void +process_init(void) +{ + threads = malloc(sizeof(thread_data) * opt_threads); + if (threads == NULL) { + out_of_memory(); + my_exit(ERROR); + } + + for (size_t i = 0; i < opt_threads; ++i) + threads[i] = (thread_data){ + .strm = LZMA_STREAM_INIT_VAR, + .options = NULL, + .pair = NULL, + .in_use = false, + }; + + if (pthread_attr_init(&thread_attr) + || pthread_attr_setdetachstate( + &thread_attr, PTHREAD_CREATE_DETACHED)) { + out_of_memory(); + my_exit(ERROR); + } + + free_threads = opt_threads; + + return; +} + + +////////////////////////// +// Thread-specific data // +////////////////////////// + +static thread_data * +get_thread_data(void) +{ + pthread_mutex_lock(&mutex); + + while (free_threads == 0) { + pthread_cond_wait(&cond, &mutex); + + if (user_abort) { + pthread_cond_signal(&cond); + pthread_mutex_unlock(&mutex); + return NULL; + } + } + + thread_data *t = threads; + while (t->in_use) + ++t; + + t->in_use = true; + --free_threads; + + pthread_mutex_unlock(&mutex); + + return t; +} + + +static void +release_thread_data(thread_data *t) +{ + pthread_mutex_lock(&mutex); + + t->in_use = false; + ++free_threads; + + pthread_cond_signal(&cond); + pthread_mutex_unlock(&mutex); + + return; +} + + +static int +create_thread(void *(*func)(thread_data *t), thread_data *t) +{ + if (opt_threads == 1) { + func(t); + } else { + const int err = pthread_create(&t->thread, &thread_attr, + (void *(*)(void *))(func), t); + if (err) { + errmsg(V_ERROR, _("Cannot create a thread: %s"), + strerror(err)); + user_abort = 1; + return -1; + } + } + + return 0; +} + + +///////////////////////// +// One thread per file // +///////////////////////// + +static int +single_init(thread_data *t) +{ + lzma_ret ret; + + if (opt_mode == MODE_COMPRESS) { + const lzma_vli uncompressed_size + = t->pair->src_fd != STDIN_FILENO + ? (lzma_vli)(t->pair->src_st.st_size) + : LZMA_VLI_VALUE_UNKNOWN; + + // TODO Support Multi-Block Streams to store Extra. + if (opt_header == HEADER_ALONE) { + lzma_options_alone alone; + alone.uncompressed_size = uncompressed_size; + memcpy(&alone.lzma, opt_filters[0].options, + sizeof(alone.lzma)); + ret = lzma_alone_encoder(&t->strm, &alone); + } else { + lzma_options_stream stream = { + .check = opt_check, + .has_crc32 = true, + .uncompressed_size = uncompressed_size, + .alignment = 0, + }; + memcpy(stream.filters, opt_filters, + sizeof(stream.filters)); + ret = lzma_stream_encoder_single(&t->strm, &stream); + } + } else { + // TODO Restrict file format if requested on the command line. + ret = lzma_auto_decoder(&t->strm, NULL, NULL); + } + + if (ret != LZMA_OK) { + if (ret == LZMA_MEM_ERROR) + out_of_memory(); + else + internal_error(); + + return -1; + } + + return 0; +} + + +static lzma_ret +single_skip_padding(thread_data *t, uint8_t *in_buf) +{ + // Handle decoding of concatenated Streams. There can be arbitrary + // number of nul-byte padding between the Streams, which must be + // ignored. + // + // NOTE: Concatenating LZMA_Alone files works only if at least + // one of lc, lp, and pb is non-zero. Using the concatenation + // on LZMA_Alone files is strongly discouraged. + while (true) { + while (t->strm.avail_in > 0) { + if (*t->strm.next_in != '\0') + return LZMA_OK; + + ++t->strm.next_in; + --t->strm.avail_in; + } + + if (t->pair->src_eof) + return LZMA_STREAM_END; + + t->strm.next_in = in_buf; + t->strm.avail_in = io_read(t->pair, in_buf, BUFSIZ); + if (t->strm.avail_in == SIZE_MAX) + return LZMA_DATA_ERROR; + } +} + + +static void * +single(thread_data *t) +{ + if (single_init(t)) { + io_close(t->pair, false); + release_thread_data(t); + return NULL; + } + + uint8_t in_buf[BUFSIZ]; + uint8_t out_buf[BUFSIZ]; + lzma_action action = LZMA_RUN; + lzma_ret ret; + bool success = false; + + t->strm.avail_in = 0; + + while (!user_abort) { + if (t->strm.avail_in == 0 && !t->pair->src_eof) { + t->strm.next_in = in_buf; + t->strm.avail_in = io_read(t->pair, in_buf, BUFSIZ); + + if (t->strm.avail_in == SIZE_MAX) + break; + else if (t->pair->src_eof + && opt_mode == MODE_COMPRESS) + action = LZMA_FINISH; + } + + t->strm.next_out = out_buf; + t->strm.avail_out = BUFSIZ; + + ret = lzma_code(&t->strm, action); + + if (opt_mode != MODE_TEST) + if (io_write(t->pair, out_buf, + BUFSIZ - t->strm.avail_out)) + break; + + if (ret != LZMA_OK) { + if (ret == LZMA_STREAM_END) { + if (opt_mode == MODE_COMPRESS) { + success = true; + break; + } + + // Support decoding concatenated .lzma files. + ret = single_skip_padding(t, in_buf); + + if (ret == LZMA_STREAM_END) { + assert(t->pair->src_eof); + success = true; + break; + } + + if (ret == LZMA_OK && !single_init(t)) + continue; + + break; + + } else { + errmsg(V_ERROR, "%s: %s", t->pair->src_name, + str_strm_error(ret)); + break; + } + } + } + + io_close(t->pair, success); + release_thread_data(t); + + return NULL; +} + + +/////////////////////////////// +// Multiple threads per file // +/////////////////////////////// + +// TODO + +// I'm not sure what would the best way to implement this. Here's one +// possible way: +// - Reader thread would read the input data and control the coders threads. +// - Every coder thread is associated with input and output buffer pools. +// The input buffer pool is filled by reader thread, and the output buffer +// pool is emptied by the writer thread. +// - Writer thread writes the output data of the oldest living coder thread. +// +// The per-file thread started by the application's main thread is used as +// the reader thread. In the beginning, it starts the writer thread and the +// first coder thread. The coder thread would be left waiting for input from +// the reader thread, and the writer thread would be waiting for input from +// the coder thread. +// +// The reader thread reads the input data into a ring buffer, whose size +// depends on the value returned by lzma_chunk_size(). If the ring buffer +// gets full, the buffer is marked "to be finished", which indicates to +// the coder thread that no more input is coming. Then a new coder thread +// would be started. +// +// TODO + +/* +typedef struct { + /// Buffers + uint8_t (*buffers)[BUFSIZ]; + + /// Number of buffers + size_t buffer_count; + + /// buffers[read_pos] is the buffer currently being read. Once finish + /// is true and read_pos == write_pos, end of input has been reached. + size_t read_pos; + + /// buffers[write_pos] is the buffer into which data is currently + /// being written. + size_t write_pos; + + /// This variable matters only when read_pos == write_pos && finish. + /// In that case, this variable will contain the size of the + /// buffers[read_pos]. + size_t last_size; + + /// True once no more data is being written to the buffer. When this + /// is set, the last_size variable must have been set too. + bool finish; + + /// Mutex to protect access to the variables in this structure + pthread_mutex_t mutex; + + /// Condition to indicate when another thread can continue + pthread_cond_t cond; +} mem_pool; + + +static foo +multi_reader(thread_data *t) +{ + bool done = false; + + do { + const size_t size = io_read(t->pair, + m->buffers + m->write_pos, BUFSIZ); + if (size == SIZE_MAX) { + // TODO + } else if (t->pair->src_eof) { + m->last_size = size; + } + + pthread_mutex_lock(&m->mutex); + + if (++m->write_pos == m->buffer_count) + m->write_pos = 0; + + if (m->write_pos == m->read_pos || t->pair->src_eof) + m->finish = true; + + pthread_cond_signal(&m->cond); + pthread_mutex_unlock(&m->mutex); + + } while (!m->finish); + + return done ? 0 : -1; +} + + +static foo +multi_code() +{ + lzma_action = LZMA_RUN; + + while (true) { + pthread_mutex_lock(&m->mutex); + + while (m->read_pos == m->write_pos && !m->finish) + pthread_cond_wait(&m->cond, &m->mutex); + + pthread_mutex_unlock(&m->mutex); + + if (m->finish) { + t->strm.avail_in = m->last_size; + if (opt_mode == MODE_COMPRESS) + action = LZMA_FINISH; + } else { + t->strm.avail_in = BUFSIZ; + } + + t->strm.next_in = m->buffers + m->read_pos; + + const lzma_ret ret = lzma_code(&t->strm, action); + + } +} + +*/ + + +/////////////////////// +// Starting new file // +/////////////////////// + +extern void +process_file(const char *filename) +{ + thread_data *t = get_thread_data(); + if (t == NULL) + return; // User abort + + // If this fails, it shows appropriate error messages too. + t->pair = io_open(filename); + if (t->pair == NULL) { + release_thread_data(t); + return; + } + + // TODO Currently only one-thread-per-file mode is implemented. + + if (create_thread(&single, t)) { + io_close(t->pair, false); + release_thread_data(t); + } + + return; +} diff --git a/src/lzma/process.h b/src/lzma/process.h new file mode 100644 index 00000000..7fdfbce6 --- /dev/null +++ b/src/lzma/process.h @@ -0,0 +1,30 @@ +/////////////////////////////////////////////////////////////////////////////// +// +/// \file process.c +/// \brief Compresses or uncompresses a file +// +// Copyright (C) 2007 Lasse Collin +// +// This program is free software; you can redistribute it and/or +// modify it under the terms of the GNU Lesser General Public +// License as published by the Free Software Foundation; either +// version 2.1 of the License, or (at your option) any later version. +// +// This program is distributed in the hope that it will be useful, +// but WITHOUT ANY WARRANTY; without even the implied warranty of +// MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU +// Lesser General Public License for more details. +// +/////////////////////////////////////////////////////////////////////////////// + +#ifndef PROCESS_H +#define PROCESS_H + +#include "private.h" + + +extern void process_init(void); + +extern void process_file(const char *filename); + +#endif diff --git a/src/lzma/suffix.c b/src/lzma/suffix.c new file mode 100644 index 00000000..57afce82 --- /dev/null +++ b/src/lzma/suffix.c @@ -0,0 +1,145 @@ +/////////////////////////////////////////////////////////////////////////////// +// +/// \file suffix.c +/// \brief Checks filename suffix and creates the destination filename +// +// Copyright (C) 2007 Lasse Collin +// +// This program is free software; you can redistribute it and/or +// modify it under the terms of the GNU Lesser General Public +// License as published by the Free Software Foundation; either +// version 2.1 of the License, or (at your option) any later version. +// +// This program is distributed in the hope that it will be useful, +// but WITHOUT ANY WARRANTY; without even the implied warranty of +// MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU +// Lesser General Public License for more details. +// +/////////////////////////////////////////////////////////////////////////////// + +#include "private.h" + + +static const struct { + const char *compressed; + const char *uncompressed; +} suffixes[] = { + { ".lzma", "" }, + { ".tlz", ".tar" }, + { ".ylz", ".yar" }, + { NULL, NULL } +}; + + +/// \brief Checks if src_name has given compressed_suffix +/// +/// \param suffix Filename suffix to look for +/// \param src_name Input filename +/// \param src_len strlen(src_name) +/// +/// \return If src_name has the suffix, src_len - strlen(suffix) is +/// returned. It's always a positive integer. Otherwise zero +/// is returned. +static size_t +test_suffix(const char *suffix, const char *src_name, size_t src_len) +{ + const size_t suffix_len = strlen(suffix); + + // The filename must have at least one character in addition to + // the suffix. src_name may contain path to the filename, so we + // need to check for directory separator too. + if (src_len <= suffix_len || src_name[src_len - suffix_len - 1] == '/') + return 0; + + if (strcmp(suffix, src_name + src_len - suffix_len) == 0) + return src_len - suffix_len; + + return 0; +} + + +/// \brief Removes the filename suffix of the compressed file +/// +/// \return Name of the uncompressed file, or NULL if file has unknown +/// suffix. +static char * +uncompressed_name(const char *src_name) +{ + const char *new_suffix = ""; + const size_t src_len = strlen(src_name); + size_t new_len = 0; + + for (size_t i = 0; suffixes[i].compressed != NULL; ++i) { + new_len = test_suffix(suffixes[i].compressed, + src_name, src_len); + if (new_len != 0) { + new_suffix = suffixes[i].uncompressed; + break; + } + } + + if (new_len == 0 && opt_suffix != NULL) + new_len = test_suffix(opt_suffix, src_name, src_len); + + if (new_len == 0) { + errmsg(V_WARNING, _("%s: Filename has an unknown suffix, " + "skipping"), src_name); + return NULL; + } + + const size_t new_suffix_len = strlen(new_suffix); + char *dest_name = malloc(new_len + new_suffix_len + 1); + if (dest_name == NULL) { + out_of_memory(); + return NULL; + } + + memcpy(dest_name, src_name, new_len); + memcpy(dest_name + new_len, new_suffix, new_suffix_len); + dest_name[new_len + new_suffix_len] = '\0'; + + return dest_name; +} + + +/// \brief Appends suffix to src_name +static char * +compressed_name(const char *src_name) +{ + const size_t src_len = strlen(src_name); + + for (size_t i = 0; suffixes[i].compressed != NULL; ++i) { + if (test_suffix(suffixes[i].compressed, src_name, src_len) + != 0) { + errmsg(V_WARNING, _("%s: File already has `%s' " + "suffix, skipping"), src_name, + suffixes[i].compressed); + return NULL; + } + } + + const char *suffix = opt_suffix != NULL + ? opt_suffix : suffixes[0].compressed; + const size_t suffix_len = strlen(suffix); + + char *dest_name = malloc(src_len + suffix_len + 1); + if (dest_name == NULL) { + out_of_memory(); + return NULL; + } + + memcpy(dest_name, src_name, src_len); + memcpy(dest_name + src_len, suffix, suffix_len); + dest_name[src_len + suffix_len] = '\0'; + + return dest_name; +} + + +extern char * +get_dest_name(const char *src_name) +{ + return opt_mode == MODE_COMPRESS + ? compressed_name(src_name) + : uncompressed_name(src_name); +} diff --git a/src/lzma/suffix.h b/src/lzma/suffix.h new file mode 100644 index 00000000..08315659 --- /dev/null +++ b/src/lzma/suffix.h @@ -0,0 +1,25 @@ +/////////////////////////////////////////////////////////////////////////////// +// +/// \file suffix.h +/// \brief Checks filename suffix and creates the destination filename +// +// Copyright (C) 2007 Lasse Collin +// +// This program is free software; you can redistribute it and/or +// modify it under the terms of the GNU Lesser General Public +// License as published by the Free Software Foundation; either +// version 2.1 of the License, or (at your option) any later version. +// +// This program is distributed in the hope that it will be useful, +// but WITHOUT ANY WARRANTY; without even the implied warranty of +// MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU +// Lesser General Public License for more details. +// +/////////////////////////////////////////////////////////////////////////////// + +#ifndef SUFFIX_H +#define SUFFIX_H + +extern char *get_dest_name(const char *src_name); + +#endif diff --git a/src/lzma/util.c b/src/lzma/util.c new file mode 100644 index 00000000..6ef6eb0d --- /dev/null +++ b/src/lzma/util.c @@ -0,0 +1,182 @@ +/////////////////////////////////////////////////////////////////////////////// +// +/// \file util.c +/// \brief Miscellaneous utility functions +// +// Copyright (C) 2007 Lasse Collin +// +// This program is free software; you can redistribute it and/or +// modify it under the terms of the GNU Lesser General Public +// License as published by the Free Software Foundation; either +// version 2.1 of the License, or (at your option) any later version. +// +// This program is distributed in the hope that it will be useful, +// but WITHOUT ANY WARRANTY; without even the implied warranty of +// MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU +// Lesser General Public License for more details. +// +/////////////////////////////////////////////////////////////////////////////// + +#include "private.h" + + +/// \brief Fancy version of strtoull() +/// +/// \param name Name of the option to show in case of an error +/// \param value String containing the number to be parsed; may +/// contain suffixes "k", "M", "G", "Ki", "Mi", or "Gi" +/// \param min Minimum valid value +/// \param max Maximum valid value +/// +/// \return Parsed value that is in the range [min, max]. Does not return +/// if an error occurs. +/// +extern uint64_t +str_to_uint64(const char *name, const char *value, uint64_t min, uint64_t max) +{ + uint64_t result = 0; + + // Skip blanks. + while (*value == ' ' || *value == '\t') + ++value; + + if (*value < '0' || *value > '9') { + errmsg(V_ERROR, _("%s: Value is not a non-negative " + "decimal integer"), + value); + my_exit(ERROR); + } + + do { + // Don't overflow. + if (result > (UINT64_MAX - 9) / 10) + goto error; + + result *= 10; + result += *value - '0'; + ++value; + } while (*value >= '0' && *value <= '9'); + + if (*value != '\0') { + // Look for suffix. + static const struct { + const char *name; + uint64_t multiplier; + } suffixes[] = { + { "k", UINT64_C(1000) }, + { "M", UINT64_C(1000000) }, + { "G", UINT64_C(1000000000) }, + { "Ki", UINT64_C(1024) }, + { "Mi", UINT64_C(1048576) }, + { "Gi", UINT64_C(1073741824) }, + { NULL, 0 } + }; + + uint64_t multiplier = 0; + for (size_t i = 0; suffixes[i].name != NULL; ++i) { + if (strcmp(value, suffixes[i].name) == 0) { + multiplier = suffixes[i].multiplier; + break; + } + } + + if (multiplier == 0) { + errmsg(V_ERROR, _("%s: Invalid multiplier suffix. " + "Valid suffixes:"), value); + errmsg(V_ERROR, "`k' (10^3), `M' (10^6), `G' (10^9) " + "`Ki' (2^10), `Mi' (2^20), " + "`Gi' (2^30)"); + my_exit(ERROR); + } + + // Don't overflow here either. + if (result > UINT64_MAX / multiplier) + goto error; + + result *= multiplier; + } + + if (result < min || result > max) + goto error; + + return result; + +error: + errmsg(V_ERROR, _("Value of the option `%s' must be in the range " + "[%llu, %llu]"), name, + (unsigned long long)(min), + (unsigned long long)(max)); + my_exit(ERROR); +} + + +/// \brief Gets filename part from pathname+filename +/// +/// \return Pointer in the filename where the actual filename starts. +/// If the last character is a slash, NULL is returned. +/// +extern const char * +str_filename(const char *name) +{ + const char *base = strrchr(name, '/'); + + if (base == NULL) { + base = name; + } else if (*++base == '\0') { + base = NULL; + errmsg(V_ERROR, _("%s: Invalid filename"), name); + } + + return base; +} + + +/* +/// \brief Simple quoting to get rid of ASCII control characters +/// +/// This is not so cool and locale-dependent, but should be good enough +/// At least we don't print any control characters on the terminal. +/// +extern char * +str_quote(const char *str) +{ + size_t dest_len = 0; + bool has_ctrl = false; + + while (str[dest_len] != '\0') + if (*(unsigned char *)(str + dest_len++) < 0x20) + has_ctrl = true; + + char *dest = malloc(dest_len + 1); + if (dest != NULL) { + if (has_ctrl) { + for (size_t i = 0; i < dest_len; ++i) + if (*(unsigned char *)(str + i) < 0x20) + dest[i] = '?'; + else + dest[i] = str[i]; + + dest[dest_len] = '\0'; + + } else { + // Usually there are no control characters, + // so we can optimize. + memcpy(dest, str, dest_len + 1); + } + } + + return dest; +} +*/ + + +extern bool +is_empty_filename(const char *filename) +{ + if (filename[0] == '\0') { + errmsg(V_WARNING, _("Empty filename, skipping")); + return true; + } + + return false; +} diff --git a/src/lzma/util.h b/src/lzma/util.h new file mode 100644 index 00000000..91bd9ba3 --- /dev/null +++ b/src/lzma/util.h @@ -0,0 +1,32 @@ +/////////////////////////////////////////////////////////////////////////////// +// +/// \file util.h +/// \brief Miscellaneous utility functions +// +// Copyright (C) 2007 Lasse Collin +// +// This program is free software; you can redistribute it and/or +// modify it under the terms of the GNU Lesser General Public +// License as published by the Free Software Foundation; either +// version 2.1 of the License, or (at your option) any later version. +// +// This program is distributed in the hope that it will be useful, +// but WITHOUT ANY WARRANTY; without even the implied warranty of +// MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU +// Lesser General Public License for more details. +// +/////////////////////////////////////////////////////////////////////////////// + +#ifndef UTIL_H +#define UTIL_H + +#include "private.h" + +extern uint64_t str_to_uint64(const char *name, const char *value, + uint64_t min, uint64_t max); + +extern const char *str_filename(const char *filename); + +extern bool is_empty_filename(const char *filename); + +#endif |