]>
Commit | Line | Data |
---|---|---|
63cccae4 KZ |
1 | /* |
2 | * mkcramfs - make a cramfs file system | |
3 | * | |
19922f22 | 4 | * Copyright (C) 1999-2002 Transmeta Corporation |
63cccae4 KZ |
5 | * |
6 | * This program is free software; you can redistribute it and/or modify | |
7 | * it under the terms of the GNU General Public License as published by | |
8 | * the Free Software Foundation; either version 2 of the License, or | |
9 | * (at your option) any later version. | |
10 | * | |
11 | * This program is distributed in the hope that it will be useful, | |
12 | * but WITHOUT ANY WARRANTY; without even the implied warranty of | |
13 | * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the | |
14 | * GNU General Public License for more details. | |
15 | * | |
16 | * You should have received a copy of the GNU General Public License | |
17 | * along with this program; if not, write to the Free Software | |
18 | * Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA 02111-1307 USA | |
19 | */ | |
20 | ||
95f1bdee KZ |
21 | /* |
22 | * Old version would die on largish filesystems. Change to mmap the | |
23 | * files one by one instaed of all simultaneously. - aeb, 2002-11-01 | |
24 | */ | |
25 | ||
63cccae4 KZ |
26 | #include <sys/types.h> |
27 | #include <stdio.h> | |
28 | #include <sys/stat.h> | |
29 | #include <unistd.h> | |
30 | #include <sys/mman.h> | |
d5c28ebc | 31 | #include <fcntl.h> |
63cccae4 | 32 | #include <dirent.h> |
2b760a59 | 33 | #include <stddef.h> |
63cccae4 KZ |
34 | #include <stdlib.h> |
35 | #include <errno.h> | |
36 | #include <string.h> | |
63cccae4 | 37 | #include <getopt.h> |
2b760a59 | 38 | #include <zconf.h> |
63cccae4 KZ |
39 | #include <zlib.h> |
40 | ||
5260748c | 41 | #include "c.h" |
63cccae4 | 42 | #include "cramfs.h" |
95f1bdee | 43 | #include "md5.h" |
63cccae4 | 44 | #include "nls.h" |
9fae6c27 | 45 | #include "mkfs.h" |
64ab88d7 | 46 | #include "strutils.h" |
ccaa5275 SK |
47 | #define XALLOC_EXIT_CODE MKFS_ERROR |
48 | #include "xalloc.h" | |
19922f22 KZ |
49 | |
50 | /* The kernel only supports PAD_SIZE of 0 and 512. */ | |
51 | #define PAD_SIZE 512 | |
63cccae4 | 52 | |
63cccae4 KZ |
53 | static int verbose = 0; |
54 | ||
19922f22 | 55 | static unsigned int blksize; /* settable via -b option */ |
63cccae4 KZ |
56 | static long total_blocks = 0, total_nodes = 1; /* pre-count the root node */ |
57 | static int image_length = 0; | |
fbaec83b | 58 | static int cramfs_is_big_endian = 0; /* target is big endian */ |
63cccae4 KZ |
59 | |
60 | /* | |
61 | * If opt_holes is set, then mkcramfs can create explicit holes in the | |
62 | * data, which saves 26 bytes per hole (which is a lot smaller a | |
95f1bdee | 63 | * saving than for most filesystems). |
63cccae4 KZ |
64 | * |
65 | * Note that kernels up to at least 2.3.39 don't support cramfs holes, | |
66 | * which is why this is turned off by default. | |
67 | */ | |
68 | static int opt_edition = 0; | |
69 | static int opt_errors = 0; | |
70 | static int opt_holes = 0; | |
71 | static int opt_pad = 0; | |
72 | static char *opt_image = NULL; | |
73 | static char *opt_name = NULL; | |
74 | ||
95f1bdee KZ |
75 | static int warn_dev = 0; |
76 | static int warn_gid = 0; | |
77 | static int warn_namelen = 0; | |
78 | static int warn_skip = 0; | |
79 | static int warn_size = 0; | |
80 | static int warn_uid = 0; | |
63cccae4 KZ |
81 | |
82 | #ifndef MIN | |
83 | # define MIN(_a,_b) ((_a) < (_b) ? (_a) : (_b)) | |
84 | #endif | |
85 | ||
94ed908c KZ |
86 | /* entry.flags */ |
87 | #define CRAMFS_EFLAG_MD5 1 | |
88 | #define CRAMFS_EFLAG_INVALID 2 | |
89 | ||
63cccae4 KZ |
90 | /* In-core version of inode / directory entry. */ |
91 | struct entry { | |
92 | /* stats */ | |
19922f22 | 93 | unsigned char *name; |
63cccae4 | 94 | unsigned int mode, size, uid, gid; |
e429d141 | 95 | unsigned char md5sum[MD5LENGTH]; |
94ed908c | 96 | unsigned char flags; /* CRAMFS_EFLAG_* */ |
63cccae4 KZ |
97 | |
98 | /* FS data */ | |
95f1bdee | 99 | char *path; |
19922f22 | 100 | int fd; /* temporarily open files while mmapped */ |
11bcb652 SK |
101 | struct entry *same; /* points to other identical file */ |
102 | unsigned int offset; /* pointer to compressed data in archive */ | |
95f1bdee | 103 | unsigned int dir_offset; /* offset of directory entry in archive */ |
63cccae4 KZ |
104 | |
105 | /* organization */ | |
95f1bdee | 106 | struct entry *child; /* NULL for non-directory and empty dir */ |
63cccae4 KZ |
107 | struct entry *next; |
108 | }; | |
109 | ||
110 | /* | |
111 | * Width of various bitfields in struct cramfs_inode. | |
112 | * Used only to generate warnings. | |
113 | */ | |
114 | #define CRAMFS_SIZE_WIDTH 24 | |
115 | #define CRAMFS_UID_WIDTH 16 | |
116 | #define CRAMFS_GID_WIDTH 8 | |
117 | #define CRAMFS_OFFSET_WIDTH 26 | |
118 | ||
95f1bdee KZ |
119 | /* Input status of 0 to print help and exit without an error. */ |
120 | static void | |
121 | usage(int status) { | |
122 | FILE *stream = status ? stderr : stdout; | |
123 | ||
124 | fprintf(stream, | |
fbaec83b | 125 | _("usage: %s [-h] [-v] [-b blksize] [-e edition] [-N endian] [-i file] " |
b22550fa | 126 | "[-n name] dirname outfile\n" |
95f1bdee KZ |
127 | " -h print this help\n" |
128 | " -v be verbose\n" | |
129 | " -E make all warnings errors " | |
130 | "(non-zero exit status)\n" | |
19922f22 | 131 | " -b blksize use this blocksize, must equal page size\n" |
95f1bdee | 132 | " -e edition set edition number (part of fsid)\n" |
fbaec83b | 133 | " -N endian set cramfs endianness (big|little|host), default host\n" |
95f1bdee KZ |
134 | " -i file insert a file image into the filesystem " |
135 | "(requires >= 2.4.0)\n" | |
136 | " -n name set name of cramfs filesystem\n" | |
137 | " -p pad by %d bytes for boot code\n" | |
138 | " -s sort directory entries (old option, ignored)\n" | |
139 | " -z make explicit holes (requires >= 2.3.39)\n" | |
140 | " dirname root of the filesystem to be compressed\n" | |
141 | " outfile output file\n"), | |
5260748c | 142 | program_invocation_short_name, PAD_SIZE); |
95f1bdee KZ |
143 | |
144 | exit(status); | |
145 | } | |
146 | ||
95f1bdee KZ |
147 | static char * |
148 | do_mmap(char *path, unsigned int size, unsigned int mode){ | |
149 | int fd; | |
150 | char *start; | |
151 | ||
152 | if (!size) | |
153 | return NULL; | |
154 | ||
155 | if (S_ISLNK(mode)) { | |
156 | start = xmalloc(size); | |
157 | if (readlink(path, start, size) < 0) { | |
158 | perror(path); | |
159 | warn_skip = 1; | |
160 | start = NULL; | |
161 | } | |
162 | return start; | |
163 | } | |
164 | ||
165 | fd = open(path, O_RDONLY); | |
166 | if (fd < 0) { | |
167 | perror(path); | |
168 | warn_skip = 1; | |
169 | return NULL; | |
170 | } | |
171 | ||
172 | start = mmap(NULL, size, PROT_READ, MAP_PRIVATE, fd, 0); | |
16154b1f SK |
173 | if (-1 == (int) (long) start) |
174 | err(MKFS_ERROR, "mmap"); | |
95f1bdee KZ |
175 | close(fd); |
176 | ||
177 | return start; | |
178 | } | |
179 | ||
180 | static void | |
181 | do_munmap(char *start, unsigned int size, unsigned int mode){ | |
182 | if (S_ISLNK(mode)) | |
183 | free(start); | |
184 | else | |
185 | munmap(start, size); | |
186 | } | |
187 | ||
188 | /* compute md5sums, so that we do not have to compare every pair of files */ | |
189 | static void | |
190 | mdfile(struct entry *e) { | |
191 | MD5_CTX ctx; | |
192 | char *start; | |
193 | ||
194 | start = do_mmap(e->path, e->size, e->mode); | |
195 | if (start == NULL) { | |
94ed908c | 196 | e->flags |= CRAMFS_EFLAG_INVALID; |
95f1bdee KZ |
197 | } else { |
198 | MD5Init(&ctx); | |
3c7c38e1 | 199 | MD5Update(&ctx, (unsigned char *) start, e->size); |
95f1bdee KZ |
200 | MD5Final(e->md5sum, &ctx); |
201 | ||
202 | do_munmap(start, e->size, e->mode); | |
203 | ||
94ed908c | 204 | e->flags |= CRAMFS_EFLAG_MD5; |
95f1bdee KZ |
205 | } |
206 | } | |
207 | ||
208 | /* md5 digests are equal; files are almost certainly the same, | |
209 | but just to be sure, do the comparison */ | |
210 | static int | |
211 | identical_file(struct entry *e1, struct entry *e2){ | |
212 | char *start1, *start2; | |
213 | int equal; | |
214 | ||
215 | start1 = do_mmap(e1->path, e1->size, e1->mode); | |
216 | if (!start1) | |
217 | return 0; | |
218 | start2 = do_mmap(e2->path, e2->size, e2->mode); | |
219 | if (!start2) | |
220 | return 0; | |
221 | equal = !memcmp(start1, start2, e1->size); | |
222 | do_munmap(start1, e1->size, e1->mode); | |
223 | do_munmap(start2, e2->size, e2->mode); | |
224 | return equal; | |
225 | } | |
226 | ||
63cccae4 KZ |
227 | /* |
228 | * The longest file name component to allow for in the input directory tree. | |
229 | * Ext2fs (and many others) allow up to 255 bytes. A couple of filesystems | |
230 | * allow longer (e.g. smbfs 1024), but there isn't much use in supporting | |
231 | * >255-byte names in the input directory tree given that such names get | |
232 | * truncated to 255 bytes when written to cramfs. | |
233 | */ | |
234 | #define MAX_INPUT_NAMELEN 255 | |
235 | ||
c17e59b4 | 236 | static int find_identical_file(struct entry *orig, struct entry *new, loff_t *fslen_ub) |
63cccae4 | 237 | { |
11bcb652 | 238 | if (orig == new) |
95f1bdee | 239 | return 1; |
11bcb652 | 240 | if (!orig) |
95f1bdee | 241 | return 0; |
11bcb652 | 242 | if (orig->size == new->size && orig->path) { |
95f1bdee KZ |
243 | if (!orig->flags) |
244 | mdfile(orig); | |
245 | if (!new->flags) | |
246 | mdfile(new); | |
247 | ||
94ed908c KZ |
248 | if ((orig->flags & CRAMFS_EFLAG_MD5) && |
249 | (new->flags & CRAMFS_EFLAG_MD5) && | |
e429d141 | 250 | !memcmp(orig->md5sum, new->md5sum, MD5LENGTH) && |
95f1bdee KZ |
251 | identical_file(orig, new)) { |
252 | new->same = orig; | |
c17e59b4 | 253 | *fslen_ub -= new->size; |
95f1bdee KZ |
254 | return 1; |
255 | } | |
11bcb652 SK |
256 | } |
257 | return find_identical_file(orig->child, new, fslen_ub) || | |
258 | find_identical_file(orig->next, new, fslen_ub); | |
63cccae4 KZ |
259 | } |
260 | ||
c17e59b4 | 261 | static void eliminate_doubles(struct entry *root, struct entry *orig, loff_t *fslen_ub) { |
11bcb652 SK |
262 | if (orig) { |
263 | if (orig->size && orig->path) | |
c17e59b4 | 264 | find_identical_file(root,orig, fslen_ub); |
11bcb652 SK |
265 | eliminate_doubles(root,orig->child, fslen_ub); |
266 | eliminate_doubles(root,orig->next, fslen_ub); | |
267 | } | |
63cccae4 KZ |
268 | } |
269 | ||
270 | /* | |
271 | * We define our own sorting function instead of using alphasort which | |
272 | * uses strcoll and changes ordering based on locale information. | |
273 | */ | |
6f68ab6a | 274 | static int cramsort (const struct dirent **a, const struct dirent **b) |
63cccae4 | 275 | { |
6f68ab6a | 276 | return strcmp((*a)->d_name, (*b)->d_name); |
63cccae4 KZ |
277 | } |
278 | ||
279 | static unsigned int parse_directory(struct entry *root_entry, const char *name, struct entry **prev, loff_t *fslen_ub) | |
280 | { | |
281 | struct dirent **dirlist; | |
282 | int totalsize = 0, dircount, dirindex; | |
283 | char *path, *endpath; | |
284 | size_t len = strlen(name); | |
285 | ||
286 | /* Set up the path. */ | |
287 | /* TODO: Reuse the parent's buffer to save memcpy'ing and duplication. */ | |
95f1bdee | 288 | path = xmalloc(len + 1 + MAX_INPUT_NAMELEN + 1); |
63cccae4 KZ |
289 | memcpy(path, name, len); |
290 | endpath = path + len; | |
291 | *endpath = '/'; | |
292 | endpath++; | |
293 | ||
11bcb652 SK |
294 | /* read in the directory and sort */ |
295 | dircount = scandir(name, &dirlist, 0, cramsort); | |
63cccae4 | 296 | |
16154b1f SK |
297 | if (dircount < 0) |
298 | err(MKFS_ERROR, _("could not read directory %s"), name); | |
63cccae4 KZ |
299 | |
300 | /* process directory */ | |
301 | for (dirindex = 0; dirindex < dircount; dirindex++) { | |
302 | struct dirent *dirent; | |
303 | struct entry *entry; | |
304 | struct stat st; | |
305 | int size; | |
306 | size_t namelen; | |
307 | ||
308 | dirent = dirlist[dirindex]; | |
309 | ||
310 | /* Ignore "." and ".." - we won't be adding them | |
311 | to the archive */ | |
312 | if (dirent->d_name[0] == '.') { | |
313 | if (dirent->d_name[1] == '\0') | |
314 | continue; | |
315 | if (dirent->d_name[1] == '.') { | |
316 | if (dirent->d_name[2] == '\0') | |
317 | continue; | |
318 | } | |
319 | } | |
320 | namelen = strlen(dirent->d_name); | |
16154b1f SK |
321 | if (namelen > MAX_INPUT_NAMELEN) |
322 | errx(MKFS_ERROR, | |
3c7c38e1 | 323 | _("Very long (%zu bytes) filename `%s' found.\n" |
63cccae4 | 324 | " Please increase MAX_INPUT_NAMELEN in " |
16154b1f | 325 | "mkcramfs.c and recompile. Exiting."), |
63cccae4 | 326 | namelen, dirent->d_name); |
63cccae4 KZ |
327 | memcpy(endpath, dirent->d_name, namelen + 1); |
328 | ||
329 | if (lstat(path, &st) < 0) { | |
330 | perror(endpath); | |
331 | warn_skip = 1; | |
332 | continue; | |
333 | } | |
ccaa5275 SK |
334 | entry = xcalloc(1, sizeof(struct entry)); |
335 | entry->name = (unsigned char *)xstrdup(dirent->d_name); | |
63cccae4 KZ |
336 | if (namelen > 255) { |
337 | /* Can't happen when reading from ext2fs. */ | |
338 | ||
339 | /* TODO: we ought to avoid chopping in half | |
340 | multi-byte UTF8 characters. */ | |
341 | entry->name[namelen = 255] = '\0'; | |
342 | warn_namelen = 1; | |
343 | } | |
344 | entry->mode = st.st_mode; | |
345 | entry->size = st.st_size; | |
346 | entry->uid = st.st_uid; | |
347 | if (entry->uid >= 1 << CRAMFS_UID_WIDTH) | |
348 | warn_uid = 1; | |
349 | entry->gid = st.st_gid; | |
350 | if (entry->gid >= 1 << CRAMFS_GID_WIDTH) | |
351 | /* TODO: We ought to replace with a default | |
11bcb652 SK |
352 | gid instead of truncating; otherwise there |
353 | are security problems. Maybe mode should | |
354 | be &= ~070. Same goes for uid once Linux | |
355 | supports >16-bit uids. */ | |
63cccae4 KZ |
356 | warn_gid = 1; |
357 | size = sizeof(struct cramfs_inode) + ((namelen + 3) & ~3); | |
358 | *fslen_ub += size; | |
359 | if (S_ISDIR(st.st_mode)) { | |
360 | entry->size = parse_directory(root_entry, path, &entry->child, fslen_ub); | |
361 | } else if (S_ISREG(st.st_mode)) { | |
ccaa5275 | 362 | entry->path = xstrdup(path); |
63cccae4 | 363 | if (entry->size) { |
95f1bdee | 364 | if (entry->size >= (1 << CRAMFS_SIZE_WIDTH)) { |
63cccae4 KZ |
365 | warn_size = 1; |
366 | entry->size = (1 << CRAMFS_SIZE_WIDTH) - 1; | |
367 | } | |
63cccae4 | 368 | } |
63cccae4 | 369 | } else if (S_ISLNK(st.st_mode)) { |
ccaa5275 | 370 | entry->path = xstrdup(path); |
63cccae4 KZ |
371 | } else if (S_ISFIFO(st.st_mode) || S_ISSOCK(st.st_mode)) { |
372 | /* maybe we should skip sockets */ | |
373 | entry->size = 0; | |
374 | } else { | |
375 | entry->size = st.st_rdev; | |
376 | if (entry->size & -(1<<CRAMFS_SIZE_WIDTH)) | |
377 | warn_dev = 1; | |
378 | } | |
379 | ||
380 | if (S_ISREG(st.st_mode) || S_ISLNK(st.st_mode)) { | |
381 | int blocks = ((entry->size - 1) / blksize + 1); | |
382 | ||
383 | /* block pointers & data expansion allowance + data */ | |
95f1bdee | 384 | if (entry->size) |
63cccae4 | 385 | *fslen_ub += (4+26)*blocks + entry->size + 3; |
11bcb652 | 386 | } |
63cccae4 KZ |
387 | |
388 | /* Link it into the list */ | |
389 | *prev = entry; | |
390 | prev = &entry->next; | |
391 | totalsize += size; | |
392 | } | |
393 | free(path); | |
394 | free(dirlist); /* allocated by scandir() with malloc() */ | |
395 | return totalsize; | |
396 | } | |
397 | ||
398 | /* Returns sizeof(struct cramfs_super), which includes the root inode. */ | |
399 | static unsigned int write_superblock(struct entry *root, char *base, int size) | |
400 | { | |
401 | struct cramfs_super *super = (struct cramfs_super *) base; | |
402 | unsigned int offset = sizeof(struct cramfs_super) + image_length; | |
403 | ||
404 | if (opt_pad) { | |
405 | offset += opt_pad; | |
406 | } | |
407 | ||
408 | super->magic = CRAMFS_MAGIC; | |
409 | super->flags = CRAMFS_FLAG_FSID_VERSION_2 | CRAMFS_FLAG_SORTED_DIRS; | |
410 | if (opt_holes) | |
411 | super->flags |= CRAMFS_FLAG_HOLES; | |
412 | if (image_length > 0) | |
413 | super->flags |= CRAMFS_FLAG_SHIFTED_ROOT_OFFSET; | |
414 | super->size = size; | |
415 | memcpy(super->signature, CRAMFS_SIGNATURE, sizeof(super->signature)); | |
416 | ||
417 | super->fsid.crc = crc32(0L, Z_NULL, 0); | |
418 | super->fsid.edition = opt_edition; | |
419 | super->fsid.blocks = total_blocks; | |
420 | super->fsid.files = total_nodes; | |
421 | ||
422 | memset(super->name, 0x00, sizeof(super->name)); | |
423 | if (opt_name) | |
3c7c38e1 | 424 | strncpy((char *)super->name, opt_name, sizeof(super->name)); |
63cccae4 | 425 | else |
3c7c38e1 | 426 | strncpy((char *)super->name, "Compressed", sizeof(super->name)); |
63cccae4 KZ |
427 | |
428 | super->root.mode = root->mode; | |
429 | super->root.uid = root->uid; | |
430 | super->root.gid = root->gid; | |
431 | super->root.size = root->size; | |
432 | super->root.offset = offset >> 2; | |
433 | ||
fbaec83b SRP |
434 | super_toggle_endianness(cramfs_is_big_endian, super); |
435 | inode_from_host(cramfs_is_big_endian, &super->root, &super->root); | |
436 | ||
63cccae4 KZ |
437 | return offset; |
438 | } | |
439 | ||
440 | static void set_data_offset(struct entry *entry, char *base, unsigned long offset) | |
441 | { | |
442 | struct cramfs_inode *inode = (struct cramfs_inode *) (base + entry->dir_offset); | |
fbaec83b | 443 | inode_to_host(cramfs_is_big_endian, inode, inode); |
16154b1f SK |
444 | if (offset >= (1 << (2 + CRAMFS_OFFSET_WIDTH))) |
445 | errx(MKFS_ERROR, _("filesystem too big. Exiting.")); | |
63cccae4 | 446 | inode->offset = (offset >> 2); |
fbaec83b | 447 | inode_from_host(cramfs_is_big_endian, inode, inode); |
63cccae4 KZ |
448 | } |
449 | ||
450 | ||
451 | /* | |
452 | * We do a width-first printout of the directory | |
453 | * entries, using a stack to remember the directories | |
454 | * we've seen. | |
455 | */ | |
63cccae4 KZ |
456 | static unsigned int write_directory_structure(struct entry *entry, char *base, unsigned int offset) |
457 | { | |
458 | int stack_entries = 0; | |
3e52b13e KZ |
459 | int stack_size = 64; |
460 | struct entry **entry_stack; | |
461 | ||
462 | entry_stack = xmalloc(stack_size * sizeof(struct entry *)); | |
63cccae4 KZ |
463 | |
464 | for (;;) { | |
465 | int dir_start = stack_entries; | |
466 | while (entry) { | |
467 | struct cramfs_inode *inode = | |
468 | (struct cramfs_inode *) (base + offset); | |
3c7c38e1 | 469 | size_t len = strlen((const char *)entry->name); |
63cccae4 KZ |
470 | |
471 | entry->dir_offset = offset; | |
472 | ||
473 | inode->mode = entry->mode; | |
474 | inode->uid = entry->uid; | |
475 | inode->gid = entry->gid; | |
476 | inode->size = entry->size; | |
477 | inode->offset = 0; | |
478 | /* Non-empty directories, regfiles and symlinks will | |
479 | write over inode->offset later. */ | |
480 | ||
481 | offset += sizeof(struct cramfs_inode); | |
482 | total_nodes++; /* another node */ | |
483 | memcpy(base + offset, entry->name, len); | |
484 | /* Pad up the name to a 4-byte boundary */ | |
485 | while (len & 3) { | |
486 | *(base + offset + len) = '\0'; | |
487 | len++; | |
488 | } | |
489 | inode->namelen = len >> 2; | |
490 | offset += len; | |
491 | ||
492 | if (verbose) | |
493 | printf(" %s\n", entry->name); | |
494 | if (entry->child) { | |
3e52b13e KZ |
495 | if (stack_entries >= stack_size) { |
496 | stack_size *= 2; | |
ccaa5275 | 497 | entry_stack = xrealloc(entry_stack, stack_size * sizeof(struct entry *)); |
63cccae4 KZ |
498 | } |
499 | entry_stack[stack_entries] = entry; | |
500 | stack_entries++; | |
501 | } | |
fbaec83b | 502 | inode_from_host(cramfs_is_big_endian, inode, inode); |
63cccae4 KZ |
503 | entry = entry->next; |
504 | } | |
505 | ||
506 | /* | |
507 | * Reverse the order the stack entries pushed during | |
11bcb652 SK |
508 | * this directory, for a small optimization of disk |
509 | * access in the created fs. This change makes things | |
510 | * `ls -UR' order. | |
63cccae4 KZ |
511 | */ |
512 | { | |
513 | struct entry **lo = entry_stack + dir_start; | |
514 | struct entry **hi = entry_stack + stack_entries; | |
515 | struct entry *tmp; | |
516 | ||
517 | while (lo < --hi) { | |
518 | tmp = *lo; | |
519 | *lo++ = *hi; | |
520 | *hi = tmp; | |
521 | } | |
522 | } | |
523 | ||
524 | /* Pop a subdirectory entry from the stack, and recurse. */ | |
525 | if (!stack_entries) | |
526 | break; | |
527 | stack_entries--; | |
528 | entry = entry_stack[stack_entries]; | |
529 | ||
530 | set_data_offset(entry, base, offset); | |
531 | if (verbose) | |
532 | printf("'%s':\n", entry->name); | |
533 | entry = entry->child; | |
534 | } | |
3e52b13e | 535 | free(entry_stack); |
63cccae4 KZ |
536 | return offset; |
537 | } | |
538 | ||
3c7c38e1 | 539 | static int is_zero(unsigned char const *begin, unsigned len) |
63cccae4 KZ |
540 | { |
541 | if (opt_holes) | |
542 | /* Returns non-zero iff the first LEN bytes from BEGIN are | |
543 | all NULs. */ | |
544 | return (len-- == 0 || | |
545 | (begin[0] == '\0' && | |
546 | (len-- == 0 || | |
547 | (begin[1] == '\0' && | |
548 | (len-- == 0 || | |
549 | (begin[2] == '\0' && | |
550 | (len-- == 0 || | |
551 | (begin[3] == '\0' && | |
552 | memcmp(begin, begin + 4, len) == 0)))))))); | |
553 | else | |
554 | /* Never create holes. */ | |
555 | return 0; | |
556 | } | |
557 | ||
558 | /* | |
559 | * One 4-byte pointer per block and then the actual blocked | |
560 | * output. The first block does not need an offset pointer, | |
561 | * as it will start immediately after the pointer block; | |
562 | * so the i'th pointer points to the end of the i'th block | |
563 | * (i.e. the start of the (i+1)'th block or past EOF). | |
564 | * | |
565 | * Note that size > 0, as a zero-sized file wouldn't ever | |
566 | * have gotten here in the first place. | |
567 | */ | |
95f1bdee | 568 | static unsigned int |
3c7c38e1 | 569 | do_compress(char *base, unsigned int offset, unsigned char const *name, |
95f1bdee | 570 | char *path, unsigned int size, unsigned int mode) |
63cccae4 | 571 | { |
95f1bdee | 572 | unsigned long original_size, original_offset, new_size, blocks, curr; |
3c7c38e1 RD |
573 | long change; |
574 | char *start; | |
575 | Bytef *p; | |
95f1bdee KZ |
576 | |
577 | /* get uncompressed data */ | |
578 | start = do_mmap(path, size, mode); | |
579 | if (start == NULL) | |
580 | return offset; | |
3c7c38e1 | 581 | p = (Bytef *) start; |
95f1bdee KZ |
582 | |
583 | original_size = size; | |
584 | original_offset = offset; | |
585 | blocks = (size - 1) / blksize + 1; | |
586 | curr = offset + 4 * blocks; | |
63cccae4 KZ |
587 | |
588 | total_blocks += blocks; | |
589 | ||
590 | do { | |
3c7c38e1 RD |
591 | uLongf len = 2 * blksize; |
592 | uLongf input = size; | |
63cccae4 KZ |
593 | if (input > blksize) |
594 | input = blksize; | |
595 | size -= input; | |
95f1bdee | 596 | if (!is_zero (p, input)) { |
3c7c38e1 | 597 | compress((Bytef *)(base + curr), &len, p, input); |
63cccae4 KZ |
598 | curr += len; |
599 | } | |
95f1bdee | 600 | p += input; |
63cccae4 KZ |
601 | |
602 | if (len > blksize*2) { | |
603 | /* (I don't think this can happen with zlib.) */ | |
604 | printf(_("AIEEE: block \"compressed\" to > " | |
605 | "2*blocklength (%ld)\n"), | |
606 | len); | |
89f56db1 | 607 | exit(MKFS_ERROR); |
63cccae4 KZ |
608 | } |
609 | ||
d51f37a3 | 610 | *(uint32_t *) (base + offset) = u32_toggle_endianness(cramfs_is_big_endian, curr); |
63cccae4 KZ |
611 | offset += 4; |
612 | } while (size); | |
613 | ||
95f1bdee KZ |
614 | do_munmap(start, original_size, mode); |
615 | ||
63cccae4 KZ |
616 | curr = (curr + 3) & ~3; |
617 | new_size = curr - original_offset; | |
618 | /* TODO: Arguably, original_size in these 2 lines should be | |
95f1bdee | 619 | st_blocks * 512. But if you say that, then perhaps |
63cccae4 KZ |
620 | administrative data should also be included in both. */ |
621 | change = new_size - original_size; | |
622 | if (verbose) | |
3c7c38e1 | 623 | printf(_("%6.2f%% (%+ld bytes)\t%s\n"), |
63cccae4 KZ |
624 | (change * 100) / (double) original_size, change, name); |
625 | ||
626 | return curr; | |
627 | } | |
628 | ||
629 | ||
630 | /* | |
631 | * Traverse the entry tree, writing data for every item that has | |
95f1bdee | 632 | * non-null entry->path (i.e. every symlink and non-empty |
63cccae4 KZ |
633 | * regfile). |
634 | */ | |
95f1bdee KZ |
635 | static unsigned int |
636 | write_data(struct entry *entry, char *base, unsigned int offset) { | |
637 | struct entry *e; | |
638 | ||
639 | for (e = entry; e; e = e->next) { | |
640 | if (e->path) { | |
11bcb652 SK |
641 | if (e->same) { |
642 | set_data_offset(e, base, e->same->offset); | |
643 | e->offset = e->same->offset; | |
644 | } else if (e->size) { | |
645 | set_data_offset(e, base, offset); | |
646 | e->offset = offset; | |
647 | offset = do_compress(base, offset, e->name, | |
95f1bdee | 648 | e->path, e->size,e->mode); |
11bcb652 | 649 | } |
95f1bdee KZ |
650 | } else if (e->child) |
651 | offset = write_data(e->child, base, offset); | |
652 | } | |
63cccae4 KZ |
653 | return offset; |
654 | } | |
655 | ||
656 | static unsigned int write_file(char *file, char *base, unsigned int offset) | |
657 | { | |
658 | int fd; | |
659 | char *buf; | |
660 | ||
661 | fd = open(file, O_RDONLY); | |
16154b1f SK |
662 | if (fd < 0) |
663 | err(MKFS_ERROR, _("cannot open file %s"), file); | |
63cccae4 KZ |
664 | buf = mmap(NULL, image_length, PROT_READ, MAP_PRIVATE, fd, 0); |
665 | memcpy(base + offset, buf, image_length); | |
666 | munmap(buf, image_length); | |
16154b1f SK |
667 | if (close (fd) < 0) |
668 | err(MKFS_ERROR, _("closing file %s"), file); | |
63cccae4 KZ |
669 | /* Pad up the image_length to a 4-byte boundary */ |
670 | while (image_length & 3) { | |
671 | *(base + offset + image_length) = '\0'; | |
672 | image_length++; | |
673 | } | |
674 | return (offset + image_length); | |
675 | } | |
676 | ||
677 | /* | |
678 | * Maximum size fs you can create is roughly 256MB. (The last file's | |
679 | * data must begin within 256MB boundary but can extend beyond that.) | |
680 | * | |
681 | * Note that if you want it to fit in a ROM then you're limited to what the | |
682 | * hardware and kernel can support (64MB?). | |
683 | */ | |
95f1bdee KZ |
684 | static unsigned int |
685 | maxfslen(void) { | |
686 | return (((1 << CRAMFS_OFFSET_WIDTH) - 1) << 2) /* offset */ | |
687 | + (1 << CRAMFS_SIZE_WIDTH) - 1 /* filesize */ | |
688 | + (1 << CRAMFS_SIZE_WIDTH) * 4 / blksize; /* block pointers */ | |
689 | } | |
63cccae4 KZ |
690 | |
691 | /* | |
692 | * Usage: | |
693 | * | |
694 | * mkcramfs directory-name outfile | |
695 | * | |
696 | * where "directory-name" is simply the root of the directory | |
697 | * tree that we want to generate a compressed filesystem out | |
698 | * of. | |
699 | */ | |
700 | int main(int argc, char **argv) | |
701 | { | |
702 | struct stat st; /* used twice... */ | |
703 | struct entry *root_entry; | |
704 | char *rom_image; | |
705 | ssize_t offset, written; | |
706 | int fd; | |
707 | /* initial guess (upper-bound) of required filesystem size */ | |
708 | loff_t fslen_ub = sizeof(struct cramfs_super); | |
95f1bdee | 709 | unsigned int fslen_max; |
63cccae4 | 710 | char const *dirname, *outfile; |
d51f37a3 | 711 | uint32_t crc = crc32(0L, Z_NULL, 0); |
95f1bdee | 712 | int c; |
fbaec83b | 713 | cramfs_is_big_endian = HOST_IS_BIG_ENDIAN; /* default is to use host order */ |
63cccae4 | 714 | |
d144727b | 715 | blksize = getpagesize(); |
63cccae4 KZ |
716 | total_blocks = 0; |
717 | ||
2d57c1f0 BS |
718 | setlocale(LC_ALL, ""); |
719 | bindtextdomain(PACKAGE, LOCALEDIR); | |
720 | textdomain(PACKAGE); | |
721 | ||
63cccae4 | 722 | /* command line options */ |
fbaec83b | 723 | while ((c = getopt(argc, argv, "hb:Ee:i:n:N:psVvz")) != EOF) { |
63cccae4 KZ |
724 | switch (c) { |
725 | case 'h': | |
9d9fb355 | 726 | usage(MKFS_OK); |
95f1bdee | 727 | case 'b': |
64ab88d7 | 728 | blksize = strtoll_or_err(optarg, _("failed to parse blocksize argument")); |
95f1bdee | 729 | if (blksize <= 0) |
9d9fb355 | 730 | usage(MKFS_USAGE); |
95f1bdee | 731 | break; |
63cccae4 KZ |
732 | case 'E': |
733 | opt_errors = 1; | |
734 | break; | |
735 | case 'e': | |
64ab88d7 | 736 | opt_edition = strtoll_or_err(optarg, _("edition number argument failed")); |
fbaec83b SRP |
737 | break; |
738 | case 'N': | |
739 | if (strcmp(optarg, "big") == 0) { | |
740 | cramfs_is_big_endian = 1; | |
741 | } | |
742 | else if (strcmp(optarg, "little") == 0) { | |
743 | cramfs_is_big_endian = 0; | |
744 | } | |
745 | else if (strcmp(optarg, "host") == 0); /* default */ | |
16154b1f SK |
746 | else |
747 | errx(MKFS_USAGE, _("invalid endianness given." | |
748 | " Must be 'big', 'little', or 'host'")); | |
fbaec83b | 749 | |
63cccae4 KZ |
750 | break; |
751 | case 'i': | |
752 | opt_image = optarg; | |
16154b1f SK |
753 | if (lstat(opt_image, &st) < 0) |
754 | err(MKFS_USAGE, _("cannot stat %s"), opt_image); | |
63cccae4 KZ |
755 | image_length = st.st_size; /* may be padded later */ |
756 | fslen_ub += (image_length + 3); /* 3 is for padding */ | |
757 | break; | |
758 | case 'n': | |
759 | opt_name = optarg; | |
760 | break; | |
761 | case 'p': | |
762 | opt_pad = PAD_SIZE; | |
763 | fslen_ub += PAD_SIZE; | |
764 | break; | |
765 | case 's': | |
766 | /* old option, ignored */ | |
767 | break; | |
768 | case 'V': | |
baf39af1 | 769 | printf(_("%s (%s)\n"), |
5260748c | 770 | program_invocation_short_name, PACKAGE_STRING); |
89f56db1 | 771 | exit(MKFS_OK); |
63cccae4 KZ |
772 | case 'v': |
773 | verbose = 1; | |
774 | break; | |
775 | case 'z': | |
776 | opt_holes = 1; | |
777 | break; | |
778 | } | |
779 | } | |
780 | ||
781 | if ((argc - optind) != 2) | |
9d9fb355 | 782 | usage(MKFS_USAGE); |
63cccae4 KZ |
783 | dirname = argv[optind]; |
784 | outfile = argv[optind + 1]; | |
785 | ||
16154b1f SK |
786 | if (stat(dirname, &st) < 0) |
787 | err(MKFS_USAGE, _("cannot stat %s"), dirname); | |
63cccae4 | 788 | fd = open(outfile, O_WRONLY | O_CREAT | O_TRUNC, 0666); |
16154b1f SK |
789 | if (fd < 0) |
790 | err(MKFS_USAGE, _("cannot open %s"), outfile); | |
63cccae4 | 791 | |
ccaa5275 | 792 | root_entry = xcalloc(1, sizeof(struct entry)); |
63cccae4 KZ |
793 | root_entry->mode = st.st_mode; |
794 | root_entry->uid = st.st_uid; | |
795 | root_entry->gid = st.st_gid; | |
796 | ||
797 | root_entry->size = parse_directory(root_entry, dirname, &root_entry->child, &fslen_ub); | |
798 | ||
799 | /* always allocate a multiple of blksize bytes because that's | |
11bcb652 | 800 | what we're going to write later on */ |
63cccae4 | 801 | fslen_ub = ((fslen_ub - 1) | (blksize - 1)) + 1; |
95f1bdee | 802 | fslen_max = maxfslen(); |
63cccae4 | 803 | |
95f1bdee | 804 | if (fslen_ub > fslen_max) { |
16154b1f | 805 | warnx( _("warning: guestimate of required size (upper bound) " |
3c7c38e1 | 806 | "is %lldMB, but maximum image size is %uMB. " |
16154b1f | 807 | "We might die prematurely."), |
3c7c38e1 | 808 | (long long)fslen_ub >> 20, |
95f1bdee KZ |
809 | fslen_max >> 20); |
810 | fslen_ub = fslen_max; | |
63cccae4 KZ |
811 | } |
812 | ||
11bcb652 SK |
813 | /* find duplicate files */ |
814 | eliminate_doubles(root_entry,root_entry, &fslen_ub); | |
63cccae4 KZ |
815 | |
816 | /* TODO: Why do we use a private/anonymous mapping here | |
11bcb652 SK |
817 | followed by a write below, instead of just a shared mapping |
818 | and a couple of ftruncate calls? Is it just to save us | |
819 | having to deal with removing the file afterwards? If we | |
820 | really need this huge anonymous mapping, we ought to mmap | |
821 | in smaller chunks, so that the user doesn't need nn MB of | |
822 | RAM free. If the reason is to be able to write to | |
823 | un-mmappable block devices, then we could try shared mmap | |
824 | and revert to anonymous mmap if the shared mmap fails. */ | |
63cccae4 KZ |
825 | rom_image = mmap(NULL, |
826 | fslen_ub?fslen_ub:1, | |
827 | PROT_READ | PROT_WRITE, | |
828 | MAP_PRIVATE | MAP_ANONYMOUS, | |
829 | -1, 0); | |
830 | ||
16154b1f SK |
831 | if (-1 == (int) (long) rom_image) |
832 | err(MKFS_ERROR, _("ROM image map")); | |
63cccae4 KZ |
833 | |
834 | /* Skip the first opt_pad bytes for boot loader code */ | |
835 | offset = opt_pad; | |
836 | memset(rom_image, 0x00, opt_pad); | |
837 | ||
838 | /* Skip the superblock and come back to write it later. */ | |
839 | offset += sizeof(struct cramfs_super); | |
840 | ||
841 | /* Insert a file image. */ | |
842 | if (opt_image) { | |
843 | if (verbose) | |
844 | printf(_("Including: %s\n"), opt_image); | |
845 | offset = write_file(opt_image, rom_image, offset); | |
846 | } | |
847 | ||
848 | offset = write_directory_structure(root_entry->child, rom_image, offset); | |
849 | if (verbose) | |
3c7c38e1 | 850 | printf(_("Directory data: %zd bytes\n"), offset); |
63cccae4 KZ |
851 | |
852 | offset = write_data(root_entry, rom_image, offset); | |
853 | ||
854 | /* We always write a multiple of blksize bytes, so that | |
11bcb652 | 855 | losetup works. */ |
63cccae4 KZ |
856 | offset = ((offset - 1) | (blksize - 1)) + 1; |
857 | if (verbose) | |
3c7c38e1 | 858 | printf(_("Everything: %zd kilobytes\n"), offset >> 10); |
63cccae4 KZ |
859 | |
860 | /* Write the superblock now that we can fill in all of the fields. */ | |
861 | write_superblock(root_entry, rom_image+opt_pad, offset); | |
862 | if (verbose) | |
3c7c38e1 | 863 | printf(_("Super block: %zd bytes\n"), |
63cccae4 KZ |
864 | sizeof(struct cramfs_super)); |
865 | ||
866 | /* Put the checksum in. */ | |
3c7c38e1 | 867 | crc = crc32(crc, (unsigned char *) (rom_image+opt_pad), (offset-opt_pad)); |
fbaec83b | 868 | ((struct cramfs_super *) (rom_image+opt_pad))->fsid.crc = u32_toggle_endianness(cramfs_is_big_endian, crc); |
63cccae4 KZ |
869 | if (verbose) |
870 | printf(_("CRC: %x\n"), crc); | |
871 | ||
872 | /* Check to make sure we allocated enough space. */ | |
16154b1f SK |
873 | if (fslen_ub < offset) |
874 | errx(MKFS_ERROR, | |
63cccae4 | 875 | _("not enough space allocated for ROM image " |
16154b1f | 876 | "(%lld allocated, %zu used)"), |
3c7c38e1 | 877 | (long long) fslen_ub, offset); |
63cccae4 KZ |
878 | |
879 | written = write(fd, rom_image, offset); | |
16154b1f SK |
880 | if (written < 0) |
881 | err(MKFS_ERROR, _("ROM image")); | |
882 | if (offset != written) | |
883 | errx(MKFS_ERROR, _("ROM image write failed (%zd %zd)"), | |
63cccae4 | 884 | written, offset); |
63cccae4 KZ |
885 | |
886 | /* (These warnings used to come at the start, but they scroll off the | |
11bcb652 | 887 | screen too quickly.) */ |
63cccae4 | 888 | if (warn_namelen) /* (can't happen when reading from ext2fs) */ |
16154b1f SK |
889 | warnx(/* bytes, not chars: think UTF8. */ |
890 | _("warning: filenames truncated to 255 bytes.")); | |
63cccae4 | 891 | if (warn_skip) |
16154b1f | 892 | warnx( _("warning: files were skipped due to errors.")); |
63cccae4 | 893 | if (warn_size) |
16154b1f SK |
894 | warnx( _("warning: file sizes truncated to %luMB " |
895 | "(minus 1 byte)."), | |
63cccae4 KZ |
896 | 1L << (CRAMFS_SIZE_WIDTH - 20)); |
897 | if (warn_uid) /* (not possible with current Linux versions) */ | |
16154b1f SK |
898 | warnx( _("warning: uids truncated to %u bits. " |
899 | "(This may be a security concern.)"), | |
63cccae4 KZ |
900 | CRAMFS_UID_WIDTH); |
901 | if (warn_gid) | |
16154b1f SK |
902 | warnx( _("warning: gids truncated to %u bits. " |
903 | "(This may be a security concern.)"), | |
63cccae4 KZ |
904 | CRAMFS_GID_WIDTH); |
905 | if (warn_dev) | |
16154b1f | 906 | warnx( _("WARNING: device numbers truncated to %u bits. " |
63cccae4 | 907 | "This almost certainly means\n" |
16154b1f | 908 | "that some device files will be wrong."), |
63cccae4 KZ |
909 | CRAMFS_OFFSET_WIDTH); |
910 | if (opt_errors && | |
911 | (warn_namelen|warn_skip|warn_size|warn_uid|warn_gid|warn_dev)) | |
89f56db1 | 912 | exit(MKFS_ERROR); |
16154b1f | 913 | return EXIT_SUCCESS; |
63cccae4 | 914 | } |