From eb1c94078f35d1b6e48741122c8bb9426befc5eb Mon Sep 17 00:00:00 2001 From: Glenn L McGrath Date: Wed, 20 Jun 2001 07:48:00 +0000 Subject: Reorganise unarchiving functions, more code re-use, only does single pass(no more linked lists), basis for supporting a cpio (and cheaper untar) applet, but cpio applet isnt included in this. It effects ar, dpkg-deb applets only --- libbb/unarchive.c | 500 ++++++++++++++++++++++++++++++++++++++++++++++++++++++ 1 file changed, 500 insertions(+) create mode 100644 libbb/unarchive.c (limited to 'libbb/unarchive.c') diff --git a/libbb/unarchive.c b/libbb/unarchive.c new file mode 100644 index 000000000..a0cc28eca --- /dev/null +++ b/libbb/unarchive.c @@ -0,0 +1,500 @@ +/* + * Copyright (C) 2000 by Glenn McGrath + * Copyright (C) 2001 by Laurence Anderson + * + * Based on previous work by busybox developers and others. + * + * This program is free software; you can redistribute it and/or modify + * it under the terms of the GNU General Public License as published by + * the Free Software Foundation; either version 2 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 Library General Public License for more details. + * + * You should have received a copy of the GNU General Public License + * along with this program; if not, write to the Free Software + * Foundation, Inc., 59 Temple Place - Suite 330, Boston, MA 02111-1307, USA. + */ + +#include +#include +#include +#include +#include +#include +#include "libbb.h" + +typedef struct file_headers_s { + char *name; + char *link_name; + off_t size; + uid_t uid; + gid_t gid; + mode_t mode; + time_t mtime; + dev_t device; +} file_header_t; + +off_t archive_offset; + +void seek_sub_file(FILE *src_stream, const int count) +{ + int i; + /* Try to fseek as faster */ + archive_offset += count; + if (fseek(src_stream, count, SEEK_CUR) != 0 && errno == ESPIPE) { + for (i = 0; i < count; i++) { + fgetc(src_stream); + } + } + return; +} + + +/* Extract the data postioned at src_stream to either filesystem, stdout or + * buffer depending on the value of 'function' which is defined in libbb.h + * + * prefix doesnt have to be just a directory, it may prefix the filename as well. + * + * e.g. '/var/lib/dpkg/info/dpkg.' will extract all files to the base bath + * '/var/lib/dpkg/info/' and all files/dirs created in that dir will have + * 'dpkg.' as their prefix + * + * For this reason if prefix does point to a dir then it must end with a + * trailing '/' or else the last dir will be assumed to be the file prefix + */ +char *extract_archive(FILE *src_stream, FILE *out_stream, const file_header_t *file_entry, + const int function, const char *prefix) +{ + FILE *dst_stream = NULL; + char *full_name = NULL; + char *buffer = NULL; + struct utimbuf t; + + /* prefix doesnt have to be a proper path it may prepend + * the filename as well */ + if (prefix != NULL) { + /* strip leading '/' in filename to extract as prefix may not be dir */ + /* Cant use concat_path_file here as prefix might not be a directory */ + char *path = file_entry->name; + if (*path == '/') { + path++; + } + full_name = xmalloc(strlen(prefix) + strlen(path) + 1); + strcpy(full_name, prefix); + strcat(full_name, path); + } else { + full_name = file_entry->name; + } + + if (function & extract_to_stdout) { + if (S_ISREG(file_entry->mode)) { + copy_file_chunk(src_stream, out_stream, file_entry->size); + archive_offset += file_entry->size; + } + } + else if (function & extract_one_to_buffer) { + if (S_ISREG(file_entry->mode)) { + buffer = (char *) xmalloc(file_entry->size + 1); + fread(buffer, 1, file_entry->size, src_stream); + archive_offset += file_entry->size; + return(buffer); + } + } + else if (function & extract_all_to_fs) { +#if 0 + struct stat oldfile; + if ( (S_ISLNK(file_entry->mode) ? lstat (full_name, &oldfile) : stat (full_name, &oldfile)) == 0) { /* The file already exists */ + if (function & extract_unconditional || oldfile.st_mtime < file_entry->mtime) { + if (!S_ISDIR(oldfile.st_mode)) { + unlink(full_name); /* Directories might not be empty etc */ + } + } else { + error_msg("%s not created: newer or same age file exists", file_entry->name); + if (S_ISREG(file_entry->mode)) { + seek_sub_file(src_stream, file_entry->size); + } + return (NULL); + } + } +#endif + switch(file_entry->mode & S_IFMT) { + case S_IFREG: + if (file_entry->link_name) { /* Found a cpio hard link */ + if (link(file_entry->link_name, full_name) != 0) { + perror_msg("Cannot link from %s to '%s'", + file_entry->name, file_entry->link_name); + } + } else { + if ((dst_stream = wfopen(full_name, "w")) == NULL) { + seek_sub_file(src_stream, file_entry->size); + return NULL; + } + archive_offset += file_entry->size; + copy_file_chunk(src_stream, dst_stream, file_entry->size); + fclose(dst_stream); + } + break; + case S_IFDIR: + /* Use create_path instead of mkdir incase prefix path + * hasnt been created */ + if (function & extract_create_dirs) { + if (create_path(full_name, file_entry->mode) == FALSE) { + return NULL; + } + } + break; + case S_IFLNK: + if (symlink(file_entry->link_name, full_name) < 0) { + perror_msg("Cannot create symlink from %s to '%s'", file_entry->name, file_entry->link_name); + return NULL; + } + break; + case S_IFSOCK: + case S_IFBLK: + case S_IFCHR: + case S_IFIFO: + if (mknod(full_name, file_entry->mode, file_entry->device) == -1) { + perror_msg("Cannot create node %s", file_entry->name); + return NULL; + } + break; + } + if (function & extract_preserve_date) { + t.actime = file_entry->mtime; + t.modtime = file_entry->mtime; + utime(full_name, &t); + } + chmod(full_name, file_entry->mode); + lchown(full_name, file_entry->uid, file_entry->gid); + } else { + /* If we arent extracting data we have to skip it, + * if data size is 0 then then just do it anyway + * (saves testing for it) */ + seek_sub_file(src_stream, file_entry->size); + } + + /* extract_list and extract_verbose_list can be used in conjunction + * with one of the above four extraction functions, so do this seperately */ + if (function & extract_verbose_list) { + fprintf(out_stream, "%s %d/%d %8d %s ", mode_string(file_entry->mode), + file_entry->uid, file_entry->gid, + (int) file_entry->size, time_string(file_entry->mtime)); + } + if ((function & extract_list) || (function & extract_verbose_list)){ + /* fputs doesnt add a trailing \n, so use fprintf */ + fprintf(out_stream, "%s\n", file_entry->name); + } + + free(full_name); + + return(NULL); /* Maybe we should say if failed */ +} + +#if defined BB_AR || defined BB_CPIO || defined BB_UNTAR +char *unarchive(FILE *src_stream, void *(*get_headers)(FILE *), + const int extract_function, const char *prefix, char **extract_names) +{ + file_header_t *file_entry; + int found; + int i; + char *buffer = NULL; + + archive_offset = 0; + while ((file_entry = (file_header_t *) get_headers(src_stream)) != NULL) { + found = FALSE; + if (extract_names[0] != NULL) { + for(i = 0; extract_names[i] != 0; i++) { + if (strcmp(extract_names[i], file_entry->name) == 0) { + found = TRUE; + } + } + if (!found) { + /* seek past the data entry */ + if (!S_ISLNK(file_entry->mode) && file_entry->link_name && file_entry->size == 0) { + error_msg("You should extract %s as other files are hardlinked to it", file_entry->name); + } + seek_sub_file(src_stream, file_entry->size); + continue; + } + } + buffer = extract_archive(src_stream, stdout, file_entry, extract_function, prefix); + } + return(buffer); +} +#endif + +#if defined BB_AR || defined BB_DPKG_DEB || defined BB_DPKG +void *get_header_ar(FILE *src_stream) +{ + file_header_t *typed; + union { + char raw[60]; + struct { + char name[16]; + char date[12]; + char uid[6]; + char gid[6]; + char mode[8]; + char size[10]; + char magic[2]; + } formated; + } ar; + static char *ar_long_names; + + if (fread(ar.raw, 1, 60, src_stream) != 60) { + return(NULL); + } + archive_offset += 60; + /* align the headers based on the header magic */ + if ((ar.formated.magic[0] != '`') || (ar.formated.magic[1] != '\n')) { + /* some version of ar, have an extra '\n' after each data entry, + * this puts the next header out by 1 */ + if (ar.formated.magic[1] != '`') { + error_msg("Invalid magic"); + return(NULL); + } + /* read the next char out of what would be the data section, + * if its a '\n' then it is a valid header offset by 1*/ + archive_offset++; + if (fgetc(src_stream) != '\n') { + error_msg("Invalid magic"); + return(NULL); + } + /* fix up the header, we started reading 1 byte too early */ + /* raw_header[60] wont be '\n' as it should, but it doesnt matter */ + memmove(ar.raw, &ar.raw[1], 59); + } + + typed = (file_header_t *) xcalloc(1, sizeof(file_header_t)); + + typed->size = (size_t) atoi(ar.formated.size); + /* long filenames have '/' as the first character */ + if (ar.formated.name[0] == '/') { + if (ar.formated.name[1] == '/') { + /* If the second char is a '/' then this entries data section + * stores long filename for multiple entries, they are stored + * in static variable long_names for use in future entries */ + ar_long_names = (char *) xrealloc(ar_long_names, typed->size); + fread(ar_long_names, 1, typed->size, src_stream); + archive_offset += typed->size; + /* This ar entries data section only contained filenames for other records + * they are stored in the static ar_long_names for future reference */ + return(NULL); + } else { + /* The number after the '/' indicates the offset in the ar data section + (saved in variable long_name) that conatains the real filename */ + if (!ar_long_names) { + error_msg("Cannot resolve long file name"); + return (NULL); + } + typed->name = xstrdup(ar_long_names + atoi(&ar.formated.name[1])); + } + } else { + /* short filenames */ + typed->name = xcalloc(1, 16); + strncpy(typed->name, ar.formated.name, 16); + } + typed->name[strcspn(typed->name, " /")]='\0'; + + /* convert the rest of the now valid char header to its typed struct */ + parse_mode(ar.formated.mode, &typed->mode); + typed->mtime = atoi(ar.formated.date); + typed->uid = atoi(ar.formated.uid); + typed->gid = atoi(ar.formated.gid); + + return(typed); +} +#endif + +#if defined BB_CPIO +void *get_header_cpio(FILE *src_stream) +{ + file_header_t *cpio_entry = NULL; + char cpio_header[110]; + char dummy[14]; + int namesize; + int major, minor, nlink; + + /* There can be padding before archive header */ + seek_sub_file(src_stream, (4 - (archive_offset % 4)) % 4); + if (fread(cpio_header, 1, 110, src_stream) == 110) { + archive_offset += 110; + if (strncmp(cpio_header, "07070", 5) != 0) { + error_msg("Unsupported format or invalid magic"); + return(NULL); + } + switch (cpio_header[5]) { + case '2': /* "crc" header format */ + /* Doesnt do the crc check yet */ + case '1': /* "newc" header format */ + cpio_entry = (file_header_t *) xcalloc(1, sizeof(file_header_t)); + sscanf(cpio_header, "%14c%8x%8x%8x%8x%8lx%8lx%16c%8x%8x%8x%8c", + dummy, &cpio_entry->mode, &cpio_entry->uid, &cpio_entry->gid, + &nlink, &cpio_entry->mtime, &cpio_entry->size, + dummy, &major, &minor, &namesize, dummy); + + cpio_entry->name = (char *) xcalloc(1, namesize); + fread(cpio_entry->name, 1, namesize, src_stream); /* Read in filename */ + archive_offset += namesize; + /* Skip padding before file contents */ + seek_sub_file(src_stream, (4 - (archive_offset % 4)) % 4); + if (strcmp(cpio_entry->name, "TRAILER!!!") == 0) { + printf("%d blocks\n", (int) (archive_offset % 512 ? (archive_offset / 512) + 1 : archive_offset / 512)); /* Always round up */ + return(NULL); + } + + if (S_ISLNK(cpio_entry->mode)) { + cpio_entry->link_name = (char *) xcalloc(1, cpio_entry->size + 1); + fread(cpio_entry->link_name, 1, cpio_entry->size, src_stream); + archive_offset += cpio_entry->size; + } + if (nlink > 1 && !S_ISDIR(cpio_entry->mode) && cpio_entry->size == 0) { + error_msg("%s not extracted: Cannot handle hard links yet", cpio_entry->name); + return(get_header_cpio(src_stream)); /* Recurse to next file */ + } + cpio_entry->device = (major << 8) | minor; + break; + default: + error_msg("Unsupported format"); + return(NULL); + } + if (ferror(src_stream) || feof(src_stream)) { + perror_msg("Stream error"); + return(NULL); + } + } + return(cpio_entry); +} +#endif + +#if defined BB_UNTAR || defined BB_DPKG_DEB || defined BB_DPKG +void *get_header_tar(FILE *tar_stream) +{ + union { + unsigned char raw[512]; + struct { + char name[100]; /* 0-99 */ + char mode[8]; /* 100-107 */ + char uid[8]; /* 108-115 */ + char gid[8]; /* 116-123 */ + char size[12]; /* 124-135 */ + char mtime[12]; /* 136-147 */ + char chksum[8]; /* 148-155 */ + char typeflag; /* 156-156 */ + char linkname[100]; /* 157-256 */ + char magic[6]; /* 257-262 */ + char version[2]; /* 263-264 */ + char uname[32]; /* 265-296 */ + char gname[32]; /* 297-328 */ + char devmajor[8]; /* 329-336 */ + char devminor[8]; /* 337-344 */ + char prefix[155]; /* 345-499 */ + char padding[12]; /* 500-512 */ + } formated; + } tar; + file_header_t *tar_entry = NULL; + long i; + long sum = 0; + + if (archive_offset % 512 != 0) { + seek_sub_file(tar_stream, 512 - (archive_offset % 512)); + } + + if (fread(tar.raw, 1, 512, tar_stream) != 512) { + error_msg("Couldnt read header"); + return(NULL); + } + archive_offset += 512; + + /* Check header has valid magic, unfortunately some tar files + * have empty (0'ed) tar entries at the end, which will + * cause this to fail, so fail silently for now + */ + if (strncmp(tar.formated.magic, "ustar", 5) != 0) { + return(NULL); + } + + /* Do checksum on headers */ + for (i = 0; i < 148 ; i++) { + sum += tar.raw[i]; + } + sum += ' ' * 8; + for (i = 156; i < 512 ; i++) { + sum += tar.raw[i]; + } + if (sum != strtol(tar.formated.chksum, NULL, 8)) { + error_msg("Invalid tar header checksum"); + return(NULL); + } + + /* convert to type'ed variables */ + tar_entry = xcalloc(1, sizeof(file_header_t)); + tar_entry->name = xstrdup(tar.formated.name); + + parse_mode(tar.formated.mode, &tar_entry->mode); + tar_entry->uid = strtol(tar.formated.uid, NULL, 8); + tar_entry->gid = strtol(tar.formated.gid, NULL, 8); + tar_entry->size = strtol(tar.formated.size, NULL, 8); + tar_entry->mtime = strtol(tar.formated.mtime, NULL, 8); + tar_entry->link_name = strlen(tar.formated.linkname) ? xstrdup(tar.formated.linkname) : NULL; + tar_entry->device = (strtol(tar.formated.devmajor, NULL, 8) << 8) + + strtol(tar.formated.devminor, NULL, 8); + + return(tar_entry); +} +#endif + +#if defined BB_DPKG || defined BB_DPKG_DEB +char *deb_extract(const char *package_filename, FILE *out_stream, const int extract_function, + const char *prefix, const char *filename) +{ + FILE *deb_stream; + FILE *uncompressed_stream = NULL; + file_header_t *ar_header = NULL; + char *output_buffer = NULL; + char *ared_file = NULL; + char ar_magic[8]; + char **file_list; + int gunzip_pid; + + file_list = malloc(sizeof(char *)); + file_list[0] = xstrdup(filename); + file_list[1] = NULL; + + if (extract_function & extract_control_tar_gz) { + ared_file = xstrdup("control.tar.gz"); + } + else if (extract_function & extract_data_tar_gz) { + ared_file = xstrdup("data.tar.gz"); + } + + /* open the debian package to be worked on */ + deb_stream = wfopen(package_filename, "r"); + + /* check ar magic */ + fread(ar_magic, 1, 8, deb_stream); + if (strncmp(ar_magic,"!",7) != 0) { + error_msg_and_die("invalid magic"); + } + archive_offset = 8; + + while ((ar_header = get_header_ar(deb_stream)) != NULL) { + if (strcmp(ared_file, ar_header->name) == 0) { + /* open a stream of decompressed data */ + uncompressed_stream = gz_open(deb_stream, &gunzip_pid); + archive_offset = 0; + output_buffer = unarchive(uncompressed_stream, get_header_tar, extract_function, prefix, file_list); + } + seek_sub_file(deb_stream, ar_header->size); + } + gz_close(gunzip_pid); + fclose(deb_stream); + fclose(uncompressed_stream); + free(ared_file); + return(output_buffer); +} +#endif -- cgit v1.2.3