]>
Commit | Line | Data |
---|---|---|
63cccae4 KZ |
1 | /* |
2 | * mkcramfs - make a cramfs file system | |
3 | * | |
4 | * Copyright (C) 1999-2001 Transmeta Corporation | |
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> | |
31 | #include <sys/fcntl.h> | |
32 | #include <dirent.h> | |
33 | #include <stdlib.h> | |
34 | #include <errno.h> | |
35 | #include <string.h> | |
36 | #include <assert.h> | |
37 | #include <getopt.h> | |
38 | #include <zlib.h> | |
39 | ||
40 | #include "cramfs.h" | |
95f1bdee | 41 | #include "md5.h" |
63cccae4 KZ |
42 | #include "nls.h" |
43 | ||
44 | #define PAD_SIZE 512 /* only 0 and 512 supported by kernel */ | |
45 | ||
46 | static const char *progname = "mkcramfs"; | |
47 | static int verbose = 0; | |
48 | ||
63cccae4 KZ |
49 | #define PAGE_CACHE_SIZE (4096) |
50 | /* The kernel assumes PAGE_CACHE_SIZE as block size. */ | |
95f1bdee | 51 | static unsigned int blksize = PAGE_CACHE_SIZE; /* settable via -b option */ |
63cccae4 KZ |
52 | static long total_blocks = 0, total_nodes = 1; /* pre-count the root node */ |
53 | static int image_length = 0; | |
54 | ||
55 | /* | |
56 | * If opt_holes is set, then mkcramfs can create explicit holes in the | |
57 | * data, which saves 26 bytes per hole (which is a lot smaller a | |
95f1bdee | 58 | * saving than for most filesystems). |
63cccae4 KZ |
59 | * |
60 | * Note that kernels up to at least 2.3.39 don't support cramfs holes, | |
61 | * which is why this is turned off by default. | |
62 | */ | |
63 | static int opt_edition = 0; | |
64 | static int opt_errors = 0; | |
65 | static int opt_holes = 0; | |
66 | static int opt_pad = 0; | |
67 | static char *opt_image = NULL; | |
68 | static char *opt_name = NULL; | |
69 | ||
95f1bdee KZ |
70 | static int warn_dev = 0; |
71 | static int warn_gid = 0; | |
72 | static int warn_namelen = 0; | |
73 | static int warn_skip = 0; | |
74 | static int warn_size = 0; | |
75 | static int warn_uid = 0; | |
63cccae4 KZ |
76 | |
77 | #ifndef MIN | |
78 | # define MIN(_a,_b) ((_a) < (_b) ? (_a) : (_b)) | |
79 | #endif | |
80 | ||
81 | /* In-core version of inode / directory entry. */ | |
82 | struct entry { | |
83 | /* stats */ | |
84 | char *name; | |
85 | unsigned int mode, size, uid, gid; | |
95f1bdee KZ |
86 | unsigned char md5sum[16]; |
87 | unsigned char flags; | |
88 | #define HAVE_MD5 1 | |
89 | #define INVALID 2 | |
63cccae4 KZ |
90 | |
91 | /* FS data */ | |
95f1bdee KZ |
92 | char *path; |
93 | struct entry *same; /* points to other identical file */ | |
94 | unsigned int offset; /* pointer to compressed data in archive */ | |
95 | unsigned int dir_offset; /* offset of directory entry in archive */ | |
63cccae4 KZ |
96 | |
97 | /* organization */ | |
95f1bdee | 98 | struct entry *child; /* NULL for non-directory and empty dir */ |
63cccae4 KZ |
99 | struct entry *next; |
100 | }; | |
101 | ||
102 | /* | |
103 | * Width of various bitfields in struct cramfs_inode. | |
104 | * Used only to generate warnings. | |
105 | */ | |
106 | #define CRAMFS_SIZE_WIDTH 24 | |
107 | #define CRAMFS_UID_WIDTH 16 | |
108 | #define CRAMFS_GID_WIDTH 8 | |
109 | #define CRAMFS_OFFSET_WIDTH 26 | |
110 | ||
95f1bdee KZ |
111 | /* Input status of 0 to print help and exit without an error. */ |
112 | static void | |
113 | usage(int status) { | |
114 | FILE *stream = status ? stderr : stdout; | |
115 | ||
116 | fprintf(stream, | |
117 | _("usage: %s [-v] [-b blksz] [-e edition] [-i file] [-n name] " | |
118 | "dirname outfile\n" | |
119 | " -h print this help\n" | |
120 | " -v be verbose\n" | |
121 | " -E make all warnings errors " | |
122 | "(non-zero exit status)\n" | |
123 | " -b blksz use this blocksize, must equal page size\n" | |
124 | " -e edition set edition number (part of fsid)\n" | |
125 | " -i file insert a file image into the filesystem " | |
126 | "(requires >= 2.4.0)\n" | |
127 | " -n name set name of cramfs filesystem\n" | |
128 | " -p pad by %d bytes for boot code\n" | |
129 | " -s sort directory entries (old option, ignored)\n" | |
130 | " -z make explicit holes (requires >= 2.3.39)\n" | |
131 | " dirname root of the filesystem to be compressed\n" | |
132 | " outfile output file\n"), | |
133 | progname, PAD_SIZE); | |
134 | ||
135 | exit(status); | |
136 | } | |
137 | ||
138 | /* malloc or die */ | |
139 | static void * | |
140 | xmalloc (size_t size) { | |
141 | void *t = malloc(size); | |
142 | if (t == NULL) { | |
143 | perror(NULL); | |
144 | exit(8); /* out of memory */ | |
145 | } | |
146 | return t; | |
147 | } | |
148 | ||
149 | static char * | |
150 | do_mmap(char *path, unsigned int size, unsigned int mode){ | |
151 | int fd; | |
152 | char *start; | |
153 | ||
154 | if (!size) | |
155 | return NULL; | |
156 | ||
157 | if (S_ISLNK(mode)) { | |
158 | start = xmalloc(size); | |
159 | if (readlink(path, start, size) < 0) { | |
160 | perror(path); | |
161 | warn_skip = 1; | |
162 | start = NULL; | |
163 | } | |
164 | return start; | |
165 | } | |
166 | ||
167 | fd = open(path, O_RDONLY); | |
168 | if (fd < 0) { | |
169 | perror(path); | |
170 | warn_skip = 1; | |
171 | return NULL; | |
172 | } | |
173 | ||
174 | start = mmap(NULL, size, PROT_READ, MAP_PRIVATE, fd, 0); | |
175 | if (-1 == (int) (long) start) { | |
176 | perror("mmap"); | |
177 | exit(8); | |
178 | } | |
179 | close(fd); | |
180 | ||
181 | return start; | |
182 | } | |
183 | ||
184 | static void | |
185 | do_munmap(char *start, unsigned int size, unsigned int mode){ | |
186 | if (S_ISLNK(mode)) | |
187 | free(start); | |
188 | else | |
189 | munmap(start, size); | |
190 | } | |
191 | ||
192 | /* compute md5sums, so that we do not have to compare every pair of files */ | |
193 | static void | |
194 | mdfile(struct entry *e) { | |
195 | MD5_CTX ctx; | |
196 | char *start; | |
197 | ||
198 | start = do_mmap(e->path, e->size, e->mode); | |
199 | if (start == NULL) { | |
200 | e->flags |= INVALID; | |
201 | } else { | |
202 | MD5Init(&ctx); | |
203 | MD5Update(&ctx, start, e->size); | |
204 | MD5Final(e->md5sum, &ctx); | |
205 | ||
206 | do_munmap(start, e->size, e->mode); | |
207 | ||
208 | e->flags |= HAVE_MD5; | |
209 | } | |
210 | } | |
211 | ||
212 | /* md5 digests are equal; files are almost certainly the same, | |
213 | but just to be sure, do the comparison */ | |
214 | static int | |
215 | identical_file(struct entry *e1, struct entry *e2){ | |
216 | char *start1, *start2; | |
217 | int equal; | |
218 | ||
219 | start1 = do_mmap(e1->path, e1->size, e1->mode); | |
220 | if (!start1) | |
221 | return 0; | |
222 | start2 = do_mmap(e2->path, e2->size, e2->mode); | |
223 | if (!start2) | |
224 | return 0; | |
225 | equal = !memcmp(start1, start2, e1->size); | |
226 | do_munmap(start1, e1->size, e1->mode); | |
227 | do_munmap(start2, e2->size, e2->mode); | |
228 | return equal; | |
229 | } | |
230 | ||
63cccae4 KZ |
231 | /* |
232 | * The longest file name component to allow for in the input directory tree. | |
233 | * Ext2fs (and many others) allow up to 255 bytes. A couple of filesystems | |
234 | * allow longer (e.g. smbfs 1024), but there isn't much use in supporting | |
235 | * >255-byte names in the input directory tree given that such names get | |
236 | * truncated to 255 bytes when written to cramfs. | |
237 | */ | |
238 | #define MAX_INPUT_NAMELEN 255 | |
239 | ||
95f1bdee | 240 | static int find_identical_file(struct entry *orig, struct entry *new) |
63cccae4 | 241 | { |
95f1bdee KZ |
242 | if (orig == new) |
243 | return 1; | |
244 | if (!orig) | |
245 | return 0; | |
246 | if (orig->size == new->size && orig->path) { | |
247 | if (!orig->flags) | |
248 | mdfile(orig); | |
249 | if (!new->flags) | |
250 | mdfile(new); | |
251 | ||
252 | if ((orig->flags & HAVE_MD5) && (new->flags & HAVE_MD5) && | |
253 | !memcmp(orig->md5sum, new->md5sum, 16) && | |
254 | identical_file(orig, new)) { | |
255 | new->same = orig; | |
256 | return 1; | |
257 | } | |
63cccae4 | 258 | } |
95f1bdee KZ |
259 | return find_identical_file(orig->child, new) || |
260 | find_identical_file(orig->next, new); | |
63cccae4 KZ |
261 | } |
262 | ||
95f1bdee KZ |
263 | static void eliminate_doubles(struct entry *root, struct entry *orig) { |
264 | if (orig) { | |
265 | if (orig->size && orig->path) | |
63cccae4 KZ |
266 | find_identical_file(root,orig); |
267 | eliminate_doubles(root,orig->child); | |
268 | eliminate_doubles(root,orig->next); | |
269 | } | |
270 | } | |
271 | ||
272 | /* | |
273 | * We define our own sorting function instead of using alphasort which | |
274 | * uses strcoll and changes ordering based on locale information. | |
275 | */ | |
276 | static int cramsort (const void *a, const void *b) | |
277 | { | |
278 | return strcmp ((*(const struct dirent **) a)->d_name, | |
279 | (*(const struct dirent **) b)->d_name); | |
280 | } | |
281 | ||
282 | static unsigned int parse_directory(struct entry *root_entry, const char *name, struct entry **prev, loff_t *fslen_ub) | |
283 | { | |
284 | struct dirent **dirlist; | |
285 | int totalsize = 0, dircount, dirindex; | |
286 | char *path, *endpath; | |
287 | size_t len = strlen(name); | |
288 | ||
289 | /* Set up the path. */ | |
290 | /* TODO: Reuse the parent's buffer to save memcpy'ing and duplication. */ | |
95f1bdee | 291 | path = xmalloc(len + 1 + MAX_INPUT_NAMELEN + 1); |
63cccae4 KZ |
292 | memcpy(path, name, len); |
293 | endpath = path + len; | |
294 | *endpath = '/'; | |
295 | endpath++; | |
296 | ||
297 | /* read in the directory and sort */ | |
298 | dircount = scandir(name, &dirlist, 0, cramsort); | |
299 | ||
300 | if (dircount < 0) { | |
301 | perror(name); | |
302 | exit(8); | |
303 | } | |
304 | ||
305 | /* process directory */ | |
306 | for (dirindex = 0; dirindex < dircount; dirindex++) { | |
307 | struct dirent *dirent; | |
308 | struct entry *entry; | |
309 | struct stat st; | |
310 | int size; | |
311 | size_t namelen; | |
312 | ||
313 | dirent = dirlist[dirindex]; | |
314 | ||
315 | /* Ignore "." and ".." - we won't be adding them | |
316 | to the archive */ | |
317 | if (dirent->d_name[0] == '.') { | |
318 | if (dirent->d_name[1] == '\0') | |
319 | continue; | |
320 | if (dirent->d_name[1] == '.') { | |
321 | if (dirent->d_name[2] == '\0') | |
322 | continue; | |
323 | } | |
324 | } | |
325 | namelen = strlen(dirent->d_name); | |
326 | if (namelen > MAX_INPUT_NAMELEN) { | |
327 | fprintf(stderr, | |
328 | _("Very long (%u bytes) filename `%s' found.\n" | |
329 | " Please increase MAX_INPUT_NAMELEN in " | |
330 | "mkcramfs.c and recompile. Exiting.\n"), | |
331 | namelen, dirent->d_name); | |
332 | exit(8); | |
333 | } | |
334 | memcpy(endpath, dirent->d_name, namelen + 1); | |
335 | ||
336 | if (lstat(path, &st) < 0) { | |
337 | perror(endpath); | |
338 | warn_skip = 1; | |
339 | continue; | |
340 | } | |
341 | entry = calloc(1, sizeof(struct entry)); | |
342 | if (!entry) { | |
343 | perror(NULL); | |
344 | exit(8); | |
345 | } | |
346 | entry->name = strdup(dirent->d_name); | |
347 | if (!entry->name) { | |
348 | perror(NULL); | |
349 | exit(8); | |
350 | } | |
351 | if (namelen > 255) { | |
352 | /* Can't happen when reading from ext2fs. */ | |
353 | ||
354 | /* TODO: we ought to avoid chopping in half | |
355 | multi-byte UTF8 characters. */ | |
356 | entry->name[namelen = 255] = '\0'; | |
357 | warn_namelen = 1; | |
358 | } | |
359 | entry->mode = st.st_mode; | |
360 | entry->size = st.st_size; | |
361 | entry->uid = st.st_uid; | |
362 | if (entry->uid >= 1 << CRAMFS_UID_WIDTH) | |
363 | warn_uid = 1; | |
364 | entry->gid = st.st_gid; | |
365 | if (entry->gid >= 1 << CRAMFS_GID_WIDTH) | |
366 | /* TODO: We ought to replace with a default | |
367 | gid instead of truncating; otherwise there | |
368 | are security problems. Maybe mode should | |
369 | be &= ~070. Same goes for uid once Linux | |
370 | supports >16-bit uids. */ | |
371 | warn_gid = 1; | |
372 | size = sizeof(struct cramfs_inode) + ((namelen + 3) & ~3); | |
373 | *fslen_ub += size; | |
374 | if (S_ISDIR(st.st_mode)) { | |
375 | entry->size = parse_directory(root_entry, path, &entry->child, fslen_ub); | |
376 | } else if (S_ISREG(st.st_mode)) { | |
95f1bdee | 377 | entry->path = strdup(path); |
63cccae4 | 378 | if (entry->size) { |
95f1bdee | 379 | if (entry->size >= (1 << CRAMFS_SIZE_WIDTH)) { |
63cccae4 KZ |
380 | warn_size = 1; |
381 | entry->size = (1 << CRAMFS_SIZE_WIDTH) - 1; | |
382 | } | |
63cccae4 | 383 | } |
63cccae4 | 384 | } else if (S_ISLNK(st.st_mode)) { |
95f1bdee | 385 | entry->path = strdup(path); |
63cccae4 KZ |
386 | } else if (S_ISFIFO(st.st_mode) || S_ISSOCK(st.st_mode)) { |
387 | /* maybe we should skip sockets */ | |
388 | entry->size = 0; | |
389 | } else { | |
390 | entry->size = st.st_rdev; | |
391 | if (entry->size & -(1<<CRAMFS_SIZE_WIDTH)) | |
392 | warn_dev = 1; | |
393 | } | |
394 | ||
395 | if (S_ISREG(st.st_mode) || S_ISLNK(st.st_mode)) { | |
396 | int blocks = ((entry->size - 1) / blksize + 1); | |
397 | ||
398 | /* block pointers & data expansion allowance + data */ | |
95f1bdee | 399 | if (entry->size) |
63cccae4 KZ |
400 | *fslen_ub += (4+26)*blocks + entry->size + 3; |
401 | } | |
402 | ||
403 | /* Link it into the list */ | |
404 | *prev = entry; | |
405 | prev = &entry->next; | |
406 | totalsize += size; | |
407 | } | |
408 | free(path); | |
409 | free(dirlist); /* allocated by scandir() with malloc() */ | |
410 | return totalsize; | |
411 | } | |
412 | ||
413 | /* Returns sizeof(struct cramfs_super), which includes the root inode. */ | |
414 | static unsigned int write_superblock(struct entry *root, char *base, int size) | |
415 | { | |
416 | struct cramfs_super *super = (struct cramfs_super *) base; | |
417 | unsigned int offset = sizeof(struct cramfs_super) + image_length; | |
418 | ||
419 | if (opt_pad) { | |
420 | offset += opt_pad; | |
421 | } | |
422 | ||
423 | super->magic = CRAMFS_MAGIC; | |
424 | super->flags = CRAMFS_FLAG_FSID_VERSION_2 | CRAMFS_FLAG_SORTED_DIRS; | |
425 | if (opt_holes) | |
426 | super->flags |= CRAMFS_FLAG_HOLES; | |
427 | if (image_length > 0) | |
428 | super->flags |= CRAMFS_FLAG_SHIFTED_ROOT_OFFSET; | |
429 | super->size = size; | |
430 | memcpy(super->signature, CRAMFS_SIGNATURE, sizeof(super->signature)); | |
431 | ||
432 | super->fsid.crc = crc32(0L, Z_NULL, 0); | |
433 | super->fsid.edition = opt_edition; | |
434 | super->fsid.blocks = total_blocks; | |
435 | super->fsid.files = total_nodes; | |
436 | ||
437 | memset(super->name, 0x00, sizeof(super->name)); | |
438 | if (opt_name) | |
439 | strncpy(super->name, opt_name, sizeof(super->name)); | |
440 | else | |
441 | strncpy(super->name, "Compressed", sizeof(super->name)); | |
442 | ||
443 | super->root.mode = root->mode; | |
444 | super->root.uid = root->uid; | |
445 | super->root.gid = root->gid; | |
446 | super->root.size = root->size; | |
447 | super->root.offset = offset >> 2; | |
448 | ||
449 | return offset; | |
450 | } | |
451 | ||
452 | static void set_data_offset(struct entry *entry, char *base, unsigned long offset) | |
453 | { | |
454 | struct cramfs_inode *inode = (struct cramfs_inode *) (base + entry->dir_offset); | |
455 | if (offset >= (1 << (2 + CRAMFS_OFFSET_WIDTH))) { | |
456 | fprintf(stderr, _("filesystem too big. Exiting.\n")); | |
457 | exit(8); | |
458 | } | |
459 | inode->offset = (offset >> 2); | |
460 | } | |
461 | ||
462 | ||
463 | /* | |
464 | * We do a width-first printout of the directory | |
465 | * entries, using a stack to remember the directories | |
466 | * we've seen. | |
467 | */ | |
468 | #define MAXENTRIES (100) | |
469 | static unsigned int write_directory_structure(struct entry *entry, char *base, unsigned int offset) | |
470 | { | |
471 | int stack_entries = 0; | |
472 | struct entry *entry_stack[MAXENTRIES]; | |
473 | ||
474 | for (;;) { | |
475 | int dir_start = stack_entries; | |
476 | while (entry) { | |
477 | struct cramfs_inode *inode = | |
478 | (struct cramfs_inode *) (base + offset); | |
479 | size_t len = strlen(entry->name); | |
480 | ||
481 | entry->dir_offset = offset; | |
482 | ||
483 | inode->mode = entry->mode; | |
484 | inode->uid = entry->uid; | |
485 | inode->gid = entry->gid; | |
486 | inode->size = entry->size; | |
487 | inode->offset = 0; | |
488 | /* Non-empty directories, regfiles and symlinks will | |
489 | write over inode->offset later. */ | |
490 | ||
491 | offset += sizeof(struct cramfs_inode); | |
492 | total_nodes++; /* another node */ | |
493 | memcpy(base + offset, entry->name, len); | |
494 | /* Pad up the name to a 4-byte boundary */ | |
495 | while (len & 3) { | |
496 | *(base + offset + len) = '\0'; | |
497 | len++; | |
498 | } | |
499 | inode->namelen = len >> 2; | |
500 | offset += len; | |
501 | ||
502 | if (verbose) | |
503 | printf(" %s\n", entry->name); | |
504 | if (entry->child) { | |
505 | if (stack_entries >= MAXENTRIES) { | |
506 | fprintf(stderr, | |
507 | _("Exceeded MAXENTRIES. Raise" | |
508 | " this value in mkcramfs.c " | |
509 | "and recompile. Exiting.\n") | |
510 | ); | |
511 | exit(8); | |
512 | } | |
513 | entry_stack[stack_entries] = entry; | |
514 | stack_entries++; | |
515 | } | |
516 | entry = entry->next; | |
517 | } | |
518 | ||
519 | /* | |
520 | * Reverse the order the stack entries pushed during | |
521 | * this directory, for a small optimization of disk | |
522 | * access in the created fs. This change makes things | |
523 | * `ls -UR' order. | |
524 | */ | |
525 | { | |
526 | struct entry **lo = entry_stack + dir_start; | |
527 | struct entry **hi = entry_stack + stack_entries; | |
528 | struct entry *tmp; | |
529 | ||
530 | while (lo < --hi) { | |
531 | tmp = *lo; | |
532 | *lo++ = *hi; | |
533 | *hi = tmp; | |
534 | } | |
535 | } | |
536 | ||
537 | /* Pop a subdirectory entry from the stack, and recurse. */ | |
538 | if (!stack_entries) | |
539 | break; | |
540 | stack_entries--; | |
541 | entry = entry_stack[stack_entries]; | |
542 | ||
543 | set_data_offset(entry, base, offset); | |
544 | if (verbose) | |
545 | printf("'%s':\n", entry->name); | |
546 | entry = entry->child; | |
547 | } | |
548 | return offset; | |
549 | } | |
550 | ||
551 | static int is_zero(char const *begin, unsigned len) | |
552 | { | |
553 | if (opt_holes) | |
554 | /* Returns non-zero iff the first LEN bytes from BEGIN are | |
555 | all NULs. */ | |
556 | return (len-- == 0 || | |
557 | (begin[0] == '\0' && | |
558 | (len-- == 0 || | |
559 | (begin[1] == '\0' && | |
560 | (len-- == 0 || | |
561 | (begin[2] == '\0' && | |
562 | (len-- == 0 || | |
563 | (begin[3] == '\0' && | |
564 | memcmp(begin, begin + 4, len) == 0)))))))); | |
565 | else | |
566 | /* Never create holes. */ | |
567 | return 0; | |
568 | } | |
569 | ||
570 | /* | |
571 | * One 4-byte pointer per block and then the actual blocked | |
572 | * output. The first block does not need an offset pointer, | |
573 | * as it will start immediately after the pointer block; | |
574 | * so the i'th pointer points to the end of the i'th block | |
575 | * (i.e. the start of the (i+1)'th block or past EOF). | |
576 | * | |
577 | * Note that size > 0, as a zero-sized file wouldn't ever | |
578 | * have gotten here in the first place. | |
579 | */ | |
95f1bdee KZ |
580 | static unsigned int |
581 | do_compress(char *base, unsigned int offset, char const *name, | |
582 | char *path, unsigned int size, unsigned int mode) | |
63cccae4 | 583 | { |
95f1bdee | 584 | unsigned long original_size, original_offset, new_size, blocks, curr; |
63cccae4 | 585 | int change; |
95f1bdee KZ |
586 | char *p, *start; |
587 | ||
588 | /* get uncompressed data */ | |
589 | start = do_mmap(path, size, mode); | |
590 | if (start == NULL) | |
591 | return offset; | |
592 | p = start; | |
593 | ||
594 | original_size = size; | |
595 | original_offset = offset; | |
596 | blocks = (size - 1) / blksize + 1; | |
597 | curr = offset + 4 * blocks; | |
63cccae4 KZ |
598 | |
599 | total_blocks += blocks; | |
600 | ||
601 | do { | |
602 | unsigned long len = 2 * blksize; | |
603 | unsigned int input = size; | |
604 | if (input > blksize) | |
605 | input = blksize; | |
606 | size -= input; | |
95f1bdee KZ |
607 | if (!is_zero (p, input)) { |
608 | compress(base + curr, &len, p, input); | |
63cccae4 KZ |
609 | curr += len; |
610 | } | |
95f1bdee | 611 | p += input; |
63cccae4 KZ |
612 | |
613 | if (len > blksize*2) { | |
614 | /* (I don't think this can happen with zlib.) */ | |
615 | printf(_("AIEEE: block \"compressed\" to > " | |
616 | "2*blocklength (%ld)\n"), | |
617 | len); | |
618 | exit(8); | |
619 | } | |
620 | ||
621 | *(u32 *) (base + offset) = curr; | |
622 | offset += 4; | |
623 | } while (size); | |
624 | ||
95f1bdee KZ |
625 | do_munmap(start, original_size, mode); |
626 | ||
63cccae4 KZ |
627 | curr = (curr + 3) & ~3; |
628 | new_size = curr - original_offset; | |
629 | /* TODO: Arguably, original_size in these 2 lines should be | |
95f1bdee | 630 | st_blocks * 512. But if you say that, then perhaps |
63cccae4 KZ |
631 | administrative data should also be included in both. */ |
632 | change = new_size - original_size; | |
633 | if (verbose) | |
634 | printf(_("%6.2f%% (%+d bytes)\t%s\n"), | |
635 | (change * 100) / (double) original_size, change, name); | |
636 | ||
637 | return curr; | |
638 | } | |
639 | ||
640 | ||
641 | /* | |
642 | * Traverse the entry tree, writing data for every item that has | |
95f1bdee | 643 | * non-null entry->path (i.e. every symlink and non-empty |
63cccae4 KZ |
644 | * regfile). |
645 | */ | |
95f1bdee KZ |
646 | static unsigned int |
647 | write_data(struct entry *entry, char *base, unsigned int offset) { | |
648 | struct entry *e; | |
649 | ||
650 | for (e = entry; e; e = e->next) { | |
651 | if (e->path) { | |
652 | if (e->same) { | |
653 | set_data_offset(e, base, e->same->offset); | |
654 | e->offset = e->same->offset; | |
63cccae4 | 655 | } else { |
95f1bdee KZ |
656 | set_data_offset(e, base, offset); |
657 | e->offset = offset; | |
658 | offset = do_compress(base, offset, e->name, | |
659 | e->path, e->size,e->mode); | |
63cccae4 | 660 | } |
95f1bdee KZ |
661 | } else if (e->child) |
662 | offset = write_data(e->child, base, offset); | |
663 | } | |
63cccae4 KZ |
664 | return offset; |
665 | } | |
666 | ||
667 | static unsigned int write_file(char *file, char *base, unsigned int offset) | |
668 | { | |
669 | int fd; | |
670 | char *buf; | |
671 | ||
672 | fd = open(file, O_RDONLY); | |
673 | if (fd < 0) { | |
674 | perror(file); | |
675 | exit(8); | |
676 | } | |
677 | buf = mmap(NULL, image_length, PROT_READ, MAP_PRIVATE, fd, 0); | |
678 | memcpy(base + offset, buf, image_length); | |
679 | munmap(buf, image_length); | |
680 | close (fd); | |
681 | /* Pad up the image_length to a 4-byte boundary */ | |
682 | while (image_length & 3) { | |
683 | *(base + offset + image_length) = '\0'; | |
684 | image_length++; | |
685 | } | |
686 | return (offset + image_length); | |
687 | } | |
688 | ||
689 | /* | |
690 | * Maximum size fs you can create is roughly 256MB. (The last file's | |
691 | * data must begin within 256MB boundary but can extend beyond that.) | |
692 | * | |
693 | * Note that if you want it to fit in a ROM then you're limited to what the | |
694 | * hardware and kernel can support (64MB?). | |
695 | */ | |
95f1bdee KZ |
696 | static unsigned int |
697 | maxfslen(void) { | |
698 | return (((1 << CRAMFS_OFFSET_WIDTH) - 1) << 2) /* offset */ | |
699 | + (1 << CRAMFS_SIZE_WIDTH) - 1 /* filesize */ | |
700 | + (1 << CRAMFS_SIZE_WIDTH) * 4 / blksize; /* block pointers */ | |
701 | } | |
63cccae4 KZ |
702 | |
703 | /* | |
704 | * Usage: | |
705 | * | |
706 | * mkcramfs directory-name outfile | |
707 | * | |
708 | * where "directory-name" is simply the root of the directory | |
709 | * tree that we want to generate a compressed filesystem out | |
710 | * of. | |
711 | */ | |
712 | int main(int argc, char **argv) | |
713 | { | |
714 | struct stat st; /* used twice... */ | |
715 | struct entry *root_entry; | |
716 | char *rom_image; | |
717 | ssize_t offset, written; | |
718 | int fd; | |
719 | /* initial guess (upper-bound) of required filesystem size */ | |
720 | loff_t fslen_ub = sizeof(struct cramfs_super); | |
95f1bdee | 721 | unsigned int fslen_max; |
63cccae4 KZ |
722 | char const *dirname, *outfile; |
723 | u32 crc = crc32(0L, Z_NULL, 0); | |
95f1bdee | 724 | int c; |
63cccae4 KZ |
725 | |
726 | total_blocks = 0; | |
727 | ||
728 | if (argc) { | |
729 | char *p; | |
730 | progname = argv[0]; | |
731 | if ((p = strrchr(progname, '/')) != NULL) | |
732 | progname = p+1; | |
733 | } | |
734 | ||
735 | /* command line options */ | |
95f1bdee | 736 | while ((c = getopt(argc, argv, "hb:Ee:i:n:psVvz")) != EOF) { |
63cccae4 KZ |
737 | switch (c) { |
738 | case 'h': | |
739 | usage(0); | |
95f1bdee KZ |
740 | case 'b': |
741 | blksize = atoi(optarg); | |
742 | if (blksize <= 0) | |
743 | usage(1); | |
744 | break; | |
63cccae4 KZ |
745 | case 'E': |
746 | opt_errors = 1; | |
747 | break; | |
748 | case 'e': | |
749 | opt_edition = atoi(optarg); | |
750 | break; | |
751 | case 'i': | |
752 | opt_image = optarg; | |
753 | if (lstat(opt_image, &st) < 0) { | |
754 | perror(opt_image); | |
755 | exit(16); | |
756 | } | |
757 | image_length = st.st_size; /* may be padded later */ | |
758 | fslen_ub += (image_length + 3); /* 3 is for padding */ | |
759 | break; | |
760 | case 'n': | |
761 | opt_name = optarg; | |
762 | break; | |
763 | case 'p': | |
764 | opt_pad = PAD_SIZE; | |
765 | fslen_ub += PAD_SIZE; | |
766 | break; | |
767 | case 's': | |
768 | /* old option, ignored */ | |
769 | break; | |
770 | case 'V': | |
771 | printf(_("%s from %s\n"), | |
772 | progname, util_linux_version); | |
773 | exit(0); | |
774 | case 'v': | |
775 | verbose = 1; | |
776 | break; | |
777 | case 'z': | |
778 | opt_holes = 1; | |
779 | break; | |
780 | } | |
781 | } | |
782 | ||
783 | if ((argc - optind) != 2) | |
784 | usage(16); | |
785 | dirname = argv[optind]; | |
786 | outfile = argv[optind + 1]; | |
787 | ||
788 | if (stat(dirname, &st) < 0) { | |
789 | perror(dirname); | |
790 | exit(16); | |
791 | } | |
792 | fd = open(outfile, O_WRONLY | O_CREAT | O_TRUNC, 0666); | |
793 | ||
794 | root_entry = calloc(1, sizeof(struct entry)); | |
795 | if (!root_entry) { | |
796 | perror(NULL); | |
797 | exit(8); | |
798 | } | |
799 | root_entry->mode = st.st_mode; | |
800 | root_entry->uid = st.st_uid; | |
801 | root_entry->gid = st.st_gid; | |
802 | ||
803 | root_entry->size = parse_directory(root_entry, dirname, &root_entry->child, &fslen_ub); | |
804 | ||
805 | /* always allocate a multiple of blksize bytes because that's | |
806 | what we're going to write later on */ | |
807 | fslen_ub = ((fslen_ub - 1) | (blksize - 1)) + 1; | |
95f1bdee | 808 | fslen_max = maxfslen(); |
63cccae4 | 809 | |
95f1bdee | 810 | if (fslen_ub > fslen_max) { |
63cccae4 KZ |
811 | fprintf(stderr, |
812 | _("warning: guestimate of required size (upper bound) " | |
813 | "is %LdMB, but maximum image size is %uMB. " | |
814 | "We might die prematurely.\n"), | |
815 | fslen_ub >> 20, | |
95f1bdee KZ |
816 | fslen_max >> 20); |
817 | fslen_ub = fslen_max; | |
63cccae4 KZ |
818 | } |
819 | ||
95f1bdee | 820 | /* find duplicate files */ |
63cccae4 KZ |
821 | eliminate_doubles(root_entry,root_entry); |
822 | ||
823 | /* TODO: Why do we use a private/anonymous mapping here | |
824 | followed by a write below, instead of just a shared mapping | |
825 | and a couple of ftruncate calls? Is it just to save us | |
826 | having to deal with removing the file afterwards? If we | |
827 | really need this huge anonymous mapping, we ought to mmap | |
828 | in smaller chunks, so that the user doesn't need nn MB of | |
829 | RAM free. If the reason is to be able to write to | |
830 | un-mmappable block devices, then we could try shared mmap | |
831 | and revert to anonymous mmap if the shared mmap fails. */ | |
832 | rom_image = mmap(NULL, | |
833 | fslen_ub?fslen_ub:1, | |
834 | PROT_READ | PROT_WRITE, | |
835 | MAP_PRIVATE | MAP_ANONYMOUS, | |
836 | -1, 0); | |
837 | ||
838 | if (-1 == (int) (long) rom_image) { | |
839 | perror("ROM image map"); | |
840 | exit(8); | |
841 | } | |
842 | ||
843 | /* Skip the first opt_pad bytes for boot loader code */ | |
844 | offset = opt_pad; | |
845 | memset(rom_image, 0x00, opt_pad); | |
846 | ||
847 | /* Skip the superblock and come back to write it later. */ | |
848 | offset += sizeof(struct cramfs_super); | |
849 | ||
850 | /* Insert a file image. */ | |
851 | if (opt_image) { | |
852 | if (verbose) | |
853 | printf(_("Including: %s\n"), opt_image); | |
854 | offset = write_file(opt_image, rom_image, offset); | |
855 | } | |
856 | ||
857 | offset = write_directory_structure(root_entry->child, rom_image, offset); | |
858 | if (verbose) | |
859 | printf(_("Directory data: %d bytes\n"), offset); | |
860 | ||
861 | offset = write_data(root_entry, rom_image, offset); | |
862 | ||
863 | /* We always write a multiple of blksize bytes, so that | |
864 | losetup works. */ | |
865 | offset = ((offset - 1) | (blksize - 1)) + 1; | |
866 | if (verbose) | |
867 | printf(_("Everything: %d kilobytes\n"), offset >> 10); | |
868 | ||
869 | /* Write the superblock now that we can fill in all of the fields. */ | |
870 | write_superblock(root_entry, rom_image+opt_pad, offset); | |
871 | if (verbose) | |
872 | printf(_("Super block: %d bytes\n"), | |
873 | sizeof(struct cramfs_super)); | |
874 | ||
875 | /* Put the checksum in. */ | |
876 | crc = crc32(crc, (rom_image+opt_pad), (offset-opt_pad)); | |
877 | ((struct cramfs_super *) (rom_image+opt_pad))->fsid.crc = crc; | |
878 | if (verbose) | |
879 | printf(_("CRC: %x\n"), crc); | |
880 | ||
881 | /* Check to make sure we allocated enough space. */ | |
882 | if (fslen_ub < offset) { | |
883 | fprintf(stderr, | |
884 | _("not enough space allocated for ROM image " | |
885 | "(%Ld allocated, %d used)\n"), | |
886 | fslen_ub, offset); | |
887 | exit(8); | |
888 | } | |
889 | ||
890 | written = write(fd, rom_image, offset); | |
891 | if (written < 0) { | |
892 | perror("ROM image"); | |
893 | exit(8); | |
894 | } | |
895 | if (offset != written) { | |
896 | fprintf(stderr, _("ROM image write failed (%d %d)\n"), | |
897 | written, offset); | |
898 | exit(8); | |
899 | } | |
900 | ||
901 | /* (These warnings used to come at the start, but they scroll off the | |
902 | screen too quickly.) */ | |
903 | if (warn_namelen) /* (can't happen when reading from ext2fs) */ | |
904 | fprintf(stderr, /* bytes, not chars: think UTF8. */ | |
905 | _("warning: filenames truncated to 255 bytes.\n")); | |
906 | if (warn_skip) | |
907 | fprintf(stderr, | |
908 | _("warning: files were skipped due to errors.\n")); | |
909 | if (warn_size) | |
910 | fprintf(stderr, | |
911 | _("warning: file sizes truncated to %luMB " | |
912 | "(minus 1 byte).\n"), | |
913 | 1L << (CRAMFS_SIZE_WIDTH - 20)); | |
914 | if (warn_uid) /* (not possible with current Linux versions) */ | |
915 | fprintf(stderr, | |
916 | _("warning: uids truncated to %u bits. " | |
917 | "(This may be a security concern.)\n"), | |
918 | CRAMFS_UID_WIDTH); | |
919 | if (warn_gid) | |
920 | fprintf(stderr, | |
921 | _("warning: gids truncated to %u bits. " | |
922 | "(This may be a security concern.)\n"), | |
923 | CRAMFS_GID_WIDTH); | |
924 | if (warn_dev) | |
925 | fprintf(stderr, | |
926 | _("WARNING: device numbers truncated to %u bits. " | |
927 | "This almost certainly means\n" | |
928 | "that some device files will be wrong.\n"), | |
929 | CRAMFS_OFFSET_WIDTH); | |
930 | if (opt_errors && | |
931 | (warn_namelen|warn_skip|warn_size|warn_uid|warn_gid|warn_dev)) | |
932 | exit(8); | |
933 | return 0; | |
934 | } |