1 /*- 2 * Copyright (c) 2003-2007 Tim Kientzle 3 * All rights reserved. 4 * 5 * Redistribution and use in source and binary forms, with or without 6 * modification, are permitted provided that the following conditions 7 * are met: 8 * 1. Redistributions of source code must retain the above copyright 9 * notice, this list of conditions and the following disclaimer. 10 * 2. Redistributions in binary form must reproduce the above copyright 11 * notice, this list of conditions and the following disclaimer in the 12 * documentation and/or other materials provided with the distribution. 13 * 14 * THIS SOFTWARE IS PROVIDED BY THE AUTHOR(S) ``AS IS'' AND ANY EXPRESS OR 15 * IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES 16 * OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE DISCLAIMED. 17 * IN NO EVENT SHALL THE AUTHOR(S) BE LIABLE FOR ANY DIRECT, INDIRECT, 18 * INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT 19 * NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, 20 * DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY 21 * THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT 22 * (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF 23 * THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE. 24 */ 25 26 #include "archive_platform.h" 27 __FBSDID("$FreeBSD: src/lib/libarchive/archive_read_open_filename.c,v 1.21 2008/02/19 06:10:48 kientzle Exp $"); 28 29 #ifdef HAVE_SYS_STAT_H 30 #include <sys/stat.h> 31 #endif 32 #ifdef HAVE_ERRNO_H 33 #include <errno.h> 34 #endif 35 #ifdef HAVE_FCNTL_H 36 #include <fcntl.h> 37 #endif 38 #ifdef HAVE_STDLIB_H 39 #include <stdlib.h> 40 #endif 41 #ifdef HAVE_STRING_H 42 #include <string.h> 43 #endif 44 #ifdef HAVE_UNISTD_H 45 #include <unistd.h> 46 #endif 47 48 #include "archive.h" 49 50 #ifndef O_BINARY 51 #define O_BINARY 0 52 #endif 53 54 struct read_file_data { 55 int fd; 56 size_t block_size; 57 void *buffer; 58 mode_t st_mode; /* Mode bits for opened file. */ 59 char can_skip; /* This file supports skipping. */ 60 char filename[1]; /* Must be last! */ 61 }; 62 63 static int file_close(struct archive *, void *); 64 static ssize_t file_read(struct archive *, void *, const void **buff); 65 #if ARCHIVE_API_VERSION < 2 66 static ssize_t file_skip(struct archive *, void *, size_t request); 67 #else 68 static off_t file_skip(struct archive *, void *, off_t request); 69 #endif 70 71 int 72 archive_read_open_file(struct archive *a, const char *filename, 73 size_t block_size) 74 { 75 return (archive_read_open_filename(a, filename, block_size)); 76 } 77 78 int 79 archive_read_open_filename(struct archive *a, const char *filename, 80 size_t block_size) 81 { 82 struct stat st; 83 struct read_file_data *mine; 84 void *b; 85 int fd; 86 87 if (filename == NULL || filename[0] == '\0') { 88 /* We used to invoke archive_read_open_fd(a,0,block_size) 89 * here, but that doesn't (and shouldn't) handle the 90 * end-of-file flush when reading stdout from a pipe. 91 * Basically, read_open_fd() is intended for folks who 92 * are willing to handle such details themselves. This 93 * API is intended to be a little smarter for folks who 94 * want easy handling of the common case. 95 */ 96 filename = ""; /* Normalize NULL to "" */ 97 fd = 0; 98 } else { 99 fd = open(filename, O_RDONLY | O_BINARY); 100 if (fd < 0) { 101 archive_set_error(a, errno, 102 "Failed to open '%s'", filename); 103 return (ARCHIVE_FATAL); 104 } 105 } 106 if (fstat(fd, &st) != 0) { 107 archive_set_error(a, errno, "Can't stat '%s'", filename); 108 return (ARCHIVE_FATAL); 109 } 110 111 mine = (struct read_file_data *)calloc(1, 112 sizeof(*mine) + strlen(filename)); 113 b = malloc(block_size); 114 if (mine == NULL || b == NULL) { 115 archive_set_error(a, ENOMEM, "No memory"); 116 free(mine); 117 free(b); 118 return (ARCHIVE_FATAL); 119 } 120 strcpy(mine->filename, filename); 121 mine->block_size = block_size; 122 mine->buffer = b; 123 mine->fd = fd; 124 /* Remember mode so close can decide whether to flush. */ 125 mine->st_mode = st.st_mode; 126 /* If we're reading a file from disk, ensure that we don't 127 overwrite it with an extracted file. */ 128 if (S_ISREG(st.st_mode)) { 129 archive_read_extract_set_skip_file(a, st.st_dev, st.st_ino); 130 /* 131 * Enabling skip here is a performance optimization 132 * for anything that supports lseek(). On FreeBSD 133 * (and probably many other systems), only regular 134 * files and raw disk devices support lseek() (on 135 * other input types, lseek() returns success but 136 * doesn't actually change the file pointer, which 137 * just completely screws up the position-tracking 138 * logic). In addition, I've yet to find a portable 139 * way to determine if a device is a raw disk device. 140 * So I don't see a way to do much better than to only 141 * enable this optimization for regular files. 142 */ 143 mine->can_skip = 1; 144 } 145 return (archive_read_open2(a, mine, 146 NULL, file_read, file_skip, file_close)); 147 } 148 149 static ssize_t 150 file_read(struct archive *a, void *client_data, const void **buff) 151 { 152 struct read_file_data *mine = (struct read_file_data *)client_data; 153 ssize_t bytes_read; 154 155 *buff = mine->buffer; 156 bytes_read = read(mine->fd, mine->buffer, mine->block_size); 157 if (bytes_read < 0) { 158 if (mine->filename[0] == '\0') 159 archive_set_error(a, errno, "Error reading stdin"); 160 else 161 archive_set_error(a, errno, "Error reading '%s'", 162 mine->filename); 163 } 164 return (bytes_read); 165 } 166 167 #if ARCHIVE_API_VERSION < 2 168 static ssize_t 169 file_skip(struct archive *a, void *client_data, size_t request) 170 #else 171 static off_t 172 file_skip(struct archive *a, void *client_data, off_t request) 173 #endif 174 { 175 struct read_file_data *mine = (struct read_file_data *)client_data; 176 off_t old_offset, new_offset; 177 178 if (!mine->can_skip) /* We can't skip, so ... */ 179 return (0); /* ... skip zero bytes. */ 180 181 /* Reduce request to the next smallest multiple of block_size */ 182 request = (request / mine->block_size) * mine->block_size; 183 if (request == 0) 184 return (0); 185 186 /* 187 * Hurray for lazy evaluation: if the first lseek fails, the second 188 * one will not be executed. 189 */ 190 if (((old_offset = lseek(mine->fd, 0, SEEK_CUR)) < 0) || 191 ((new_offset = lseek(mine->fd, request, SEEK_CUR)) < 0)) 192 { 193 /* If skip failed once, it will probably fail again. */ 194 mine->can_skip = 0; 195 196 if (errno == ESPIPE) 197 { 198 /* 199 * Failure to lseek() can be caused by the file 200 * descriptor pointing to a pipe, socket or FIFO. 201 * Return 0 here, so the compression layer will use 202 * read()s instead to advance the file descriptor. 203 * It's slower of course, but works as well. 204 */ 205 return (0); 206 } 207 /* 208 * There's been an error other than ESPIPE. This is most 209 * likely caused by a programmer error (too large request) 210 * or a corrupted archive file. 211 */ 212 if (mine->filename[0] == '\0') 213 /* 214 * Should never get here, since lseek() on stdin ought 215 * to return an ESPIPE error. 216 */ 217 archive_set_error(a, errno, "Error seeking in stdin"); 218 else 219 archive_set_error(a, errno, "Error seeking in '%s'", 220 mine->filename); 221 return (-1); 222 } 223 return (new_offset - old_offset); 224 } 225 226 static int 227 file_close(struct archive *a, void *client_data) 228 { 229 struct read_file_data *mine = (struct read_file_data *)client_data; 230 231 (void)a; /* UNUSED */ 232 233 /* Only flush and close if open succeeded. */ 234 if (mine->fd >= 0) { 235 /* 236 * Sometimes, we should flush the input before closing. 237 * Regular files: faster to just close without flush. 238 * Devices: must not flush (user might need to 239 * read the "next" item on a non-rewind device). 240 * Pipes and sockets: must flush (otherwise, the 241 * program feeding the pipe or socket may complain). 242 * Here, I flush everything except for regular files and 243 * device nodes. 244 */ 245 if (!S_ISREG(mine->st_mode) 246 && !S_ISCHR(mine->st_mode) 247 && !S_ISBLK(mine->st_mode)) { 248 ssize_t bytesRead; 249 do { 250 bytesRead = read(mine->fd, mine->buffer, 251 mine->block_size); 252 } while (bytesRead > 0); 253 } 254 /* If a named file was opened, then it needs to be closed. */ 255 if (mine->filename[0] != '\0') 256 close(mine->fd); 257 } 258 free(mine->buffer); 259 free(mine); 260 return (ARCHIVE_OK); 261 } 262