1 // SPDX-License-Identifier: GPL-2.0
2 #include <linux/init.h>
3 #include <linux/fs.h>
4 #include <linux/slab.h>
5 #include <linux/types.h>
6 #include <linux/fcntl.h>
7 #include <linux/delay.h>
8 #include <linux/string.h>
9 #include <linux/dirent.h>
10 #include <linux/syscalls.h>
11 #include <linux/utime.h>
12 #include <linux/file.h>
13 #include <linux/memblock.h>
14 #include <linux/namei.h>
15 #include <linux/init_syscalls.h>
16
xwrite(struct file * file,const char * p,size_t count,loff_t * pos)17 static ssize_t __init xwrite(struct file *file, const char *p, size_t count,
18 loff_t *pos)
19 {
20 ssize_t out = 0;
21
22 /* sys_write only can write MAX_RW_COUNT aka 2G-4K bytes at most */
23 while (count) {
24 ssize_t rv = kernel_write(file, p, count, pos);
25
26 if (rv < 0) {
27 if (rv == -EINTR || rv == -EAGAIN)
28 continue;
29 return out ? out : rv;
30 } else if (rv == 0)
31 break;
32
33 p += rv;
34 out += rv;
35 count -= rv;
36 }
37
38 return out;
39 }
40
41 static __initdata char *message;
error(char * x)42 static void __init error(char *x)
43 {
44 if (!message)
45 message = x;
46 }
47
48 /* link hash */
49
50 #define N_ALIGN(len) ((((len) + 1) & ~3) + 2)
51
52 static __initdata struct hash {
53 int ino, minor, major;
54 umode_t mode;
55 struct hash *next;
56 char name[N_ALIGN(PATH_MAX)];
57 } *head[32];
58
hash(int major,int minor,int ino)59 static inline int hash(int major, int minor, int ino)
60 {
61 unsigned long tmp = ino + minor + (major << 3);
62 tmp += tmp >> 5;
63 return tmp & 31;
64 }
65
find_link(int major,int minor,int ino,umode_t mode,char * name)66 static char __init *find_link(int major, int minor, int ino,
67 umode_t mode, char *name)
68 {
69 struct hash **p, *q;
70 for (p = head + hash(major, minor, ino); *p; p = &(*p)->next) {
71 if ((*p)->ino != ino)
72 continue;
73 if ((*p)->minor != minor)
74 continue;
75 if ((*p)->major != major)
76 continue;
77 if (((*p)->mode ^ mode) & S_IFMT)
78 continue;
79 return (*p)->name;
80 }
81 q = kmalloc(sizeof(struct hash), GFP_KERNEL);
82 if (!q)
83 panic("can't allocate link hash entry");
84 q->major = major;
85 q->minor = minor;
86 q->ino = ino;
87 q->mode = mode;
88 strcpy(q->name, name);
89 q->next = NULL;
90 *p = q;
91 return NULL;
92 }
93
free_hash(void)94 static void __init free_hash(void)
95 {
96 struct hash **p, *q;
97 for (p = head; p < head + 32; p++) {
98 while (*p) {
99 q = *p;
100 *p = q->next;
101 kfree(q);
102 }
103 }
104 }
105
do_utime(char * filename,time64_t mtime)106 static long __init do_utime(char *filename, time64_t mtime)
107 {
108 struct timespec64 t[2];
109
110 t[0].tv_sec = mtime;
111 t[0].tv_nsec = 0;
112 t[1].tv_sec = mtime;
113 t[1].tv_nsec = 0;
114 return init_utimes(filename, t);
115 }
116
117 static __initdata LIST_HEAD(dir_list);
118 struct dir_entry {
119 struct list_head list;
120 char *name;
121 time64_t mtime;
122 };
123
dir_add(const char * name,time64_t mtime)124 static void __init dir_add(const char *name, time64_t mtime)
125 {
126 struct dir_entry *de = kmalloc(sizeof(struct dir_entry), GFP_KERNEL);
127 if (!de)
128 panic("can't allocate dir_entry buffer");
129 INIT_LIST_HEAD(&de->list);
130 de->name = kstrdup(name, GFP_KERNEL);
131 de->mtime = mtime;
132 list_add(&de->list, &dir_list);
133 }
134
dir_utime(void)135 static void __init dir_utime(void)
136 {
137 struct dir_entry *de, *tmp;
138 list_for_each_entry_safe(de, tmp, &dir_list, list) {
139 list_del(&de->list);
140 do_utime(de->name, de->mtime);
141 kfree(de->name);
142 kfree(de);
143 }
144 }
145
146 static __initdata time64_t mtime;
147
148 /* cpio header parsing */
149
150 static __initdata unsigned long ino, major, minor, nlink;
151 static __initdata umode_t mode;
152 static __initdata unsigned long body_len, name_len;
153 static __initdata uid_t uid;
154 static __initdata gid_t gid;
155 static __initdata unsigned rdev;
156
parse_header(char * s)157 static void __init parse_header(char *s)
158 {
159 unsigned long parsed[12];
160 char buf[9];
161 int i;
162
163 buf[8] = '\0';
164 for (i = 0, s += 6; i < 12; i++, s += 8) {
165 memcpy(buf, s, 8);
166 parsed[i] = simple_strtoul(buf, NULL, 16);
167 }
168 ino = parsed[0];
169 mode = parsed[1];
170 uid = parsed[2];
171 gid = parsed[3];
172 nlink = parsed[4];
173 mtime = parsed[5]; /* breaks in y2106 */
174 body_len = parsed[6];
175 major = parsed[7];
176 minor = parsed[8];
177 rdev = new_encode_dev(MKDEV(parsed[9], parsed[10]));
178 name_len = parsed[11];
179 }
180
181 /* FSM */
182
183 static __initdata enum state {
184 Start,
185 Collect,
186 GotHeader,
187 SkipIt,
188 GotName,
189 CopyFile,
190 GotSymlink,
191 Reset
192 } state, next_state;
193
194 static __initdata char *victim;
195 static unsigned long byte_count __initdata;
196 static __initdata loff_t this_header, next_header;
197
eat(unsigned n)198 static inline void __init eat(unsigned n)
199 {
200 victim += n;
201 this_header += n;
202 byte_count -= n;
203 }
204
205 static __initdata char *collected;
206 static long remains __initdata;
207 static __initdata char *collect;
208
read_into(char * buf,unsigned size,enum state next)209 static void __init read_into(char *buf, unsigned size, enum state next)
210 {
211 if (byte_count >= size) {
212 collected = victim;
213 eat(size);
214 state = next;
215 } else {
216 collect = collected = buf;
217 remains = size;
218 next_state = next;
219 state = Collect;
220 }
221 }
222
223 static __initdata char *header_buf, *symlink_buf, *name_buf;
224
do_start(void)225 static int __init do_start(void)
226 {
227 read_into(header_buf, 110, GotHeader);
228 return 0;
229 }
230
do_collect(void)231 static int __init do_collect(void)
232 {
233 unsigned long n = remains;
234 if (byte_count < n)
235 n = byte_count;
236 memcpy(collect, victim, n);
237 eat(n);
238 collect += n;
239 if ((remains -= n) != 0)
240 return 1;
241 state = next_state;
242 return 0;
243 }
244
do_header(void)245 static int __init do_header(void)
246 {
247 if (memcmp(collected, "070707", 6)==0) {
248 error("incorrect cpio method used: use -H newc option");
249 return 1;
250 }
251 if (memcmp(collected, "070701", 6)) {
252 error("no cpio magic");
253 return 1;
254 }
255 parse_header(collected);
256 next_header = this_header + N_ALIGN(name_len) + body_len;
257 next_header = (next_header + 3) & ~3;
258 state = SkipIt;
259 if (name_len <= 0 || name_len > PATH_MAX)
260 return 0;
261 if (S_ISLNK(mode)) {
262 if (body_len > PATH_MAX)
263 return 0;
264 collect = collected = symlink_buf;
265 remains = N_ALIGN(name_len) + body_len;
266 next_state = GotSymlink;
267 state = Collect;
268 return 0;
269 }
270 if (S_ISREG(mode) || !body_len)
271 read_into(name_buf, N_ALIGN(name_len), GotName);
272 return 0;
273 }
274
do_skip(void)275 static int __init do_skip(void)
276 {
277 if (this_header + byte_count < next_header) {
278 eat(byte_count);
279 return 1;
280 } else {
281 eat(next_header - this_header);
282 state = next_state;
283 return 0;
284 }
285 }
286
do_reset(void)287 static int __init do_reset(void)
288 {
289 while (byte_count && *victim == '\0')
290 eat(1);
291 if (byte_count && (this_header & 3))
292 error("broken padding");
293 return 1;
294 }
295
clean_path(char * path,umode_t fmode)296 static void __init clean_path(char *path, umode_t fmode)
297 {
298 struct kstat st;
299
300 if (!init_stat(path, &st, AT_SYMLINK_NOFOLLOW) &&
301 (st.mode ^ fmode) & S_IFMT) {
302 if (S_ISDIR(st.mode))
303 init_rmdir(path);
304 else
305 init_unlink(path);
306 }
307 }
308
maybe_link(void)309 static int __init maybe_link(void)
310 {
311 if (nlink >= 2) {
312 char *old = find_link(major, minor, ino, mode, collected);
313 if (old) {
314 clean_path(collected, 0);
315 return (init_link(old, collected) < 0) ? -1 : 1;
316 }
317 }
318 return 0;
319 }
320
321 static __initdata struct file *wfile;
322 static __initdata loff_t wfile_pos;
323
do_name(void)324 static int __init do_name(void)
325 {
326 state = SkipIt;
327 next_state = Reset;
328
329 /* name_len > 0 && name_len <= PATH_MAX checked in do_header */
330 if (collected[name_len - 1] != '\0') {
331 pr_err("initramfs name without nulterm: %.*s\n",
332 (int)name_len, collected);
333 error("malformed archive");
334 return 1;
335 }
336
337 if (strcmp(collected, "TRAILER!!!") == 0) {
338 free_hash();
339 return 0;
340 }
341 clean_path(collected, mode);
342 if (S_ISREG(mode)) {
343 int ml = maybe_link();
344 if (ml >= 0) {
345 int openflags = O_WRONLY|O_CREAT;
346 if (ml != 1)
347 openflags |= O_TRUNC;
348 wfile = filp_open(collected, openflags, mode);
349 if (IS_ERR(wfile))
350 return 0;
351 wfile_pos = 0;
352
353 vfs_fchown(wfile, uid, gid);
354 vfs_fchmod(wfile, mode);
355 if (body_len)
356 vfs_truncate(&wfile->f_path, body_len);
357 state = CopyFile;
358 }
359 } else if (S_ISDIR(mode)) {
360 init_mkdir(collected, mode);
361 init_chown(collected, uid, gid, 0);
362 init_chmod(collected, mode);
363 dir_add(collected, mtime);
364 } else if (S_ISBLK(mode) || S_ISCHR(mode) ||
365 S_ISFIFO(mode) || S_ISSOCK(mode)) {
366 if (maybe_link() == 0) {
367 init_mknod(collected, mode, rdev);
368 init_chown(collected, uid, gid, 0);
369 init_chmod(collected, mode);
370 do_utime(collected, mtime);
371 }
372 }
373 return 0;
374 }
375
do_copy(void)376 static int __init do_copy(void)
377 {
378 if (byte_count >= body_len) {
379 struct timespec64 t[2] = { };
380 if (xwrite(wfile, victim, body_len, &wfile_pos) != body_len)
381 error("write error");
382
383 t[0].tv_sec = mtime;
384 t[1].tv_sec = mtime;
385 vfs_utimes(&wfile->f_path, t);
386
387 fput(wfile);
388 eat(body_len);
389 state = SkipIt;
390 return 0;
391 } else {
392 if (xwrite(wfile, victim, byte_count, &wfile_pos) != byte_count)
393 error("write error");
394 body_len -= byte_count;
395 eat(byte_count);
396 return 1;
397 }
398 }
399
do_symlink(void)400 static int __init do_symlink(void)
401 {
402 if (collected[name_len - 1] != '\0') {
403 pr_err("initramfs symlink without nulterm: %.*s\n",
404 (int)name_len, collected);
405 error("malformed archive");
406 return 1;
407 }
408 collected[N_ALIGN(name_len) + body_len] = '\0';
409 clean_path(collected, 0);
410 init_symlink(collected + N_ALIGN(name_len), collected);
411 init_chown(collected, uid, gid, AT_SYMLINK_NOFOLLOW);
412 do_utime(collected, mtime);
413 state = SkipIt;
414 next_state = Reset;
415 return 0;
416 }
417
418 static __initdata int (*actions[])(void) = {
419 [Start] = do_start,
420 [Collect] = do_collect,
421 [GotHeader] = do_header,
422 [SkipIt] = do_skip,
423 [GotName] = do_name,
424 [CopyFile] = do_copy,
425 [GotSymlink] = do_symlink,
426 [Reset] = do_reset,
427 };
428
write_buffer(char * buf,unsigned long len)429 static long __init write_buffer(char *buf, unsigned long len)
430 {
431 byte_count = len;
432 victim = buf;
433
434 while (!actions[state]())
435 ;
436 return len - byte_count;
437 }
438
flush_buffer(void * bufv,unsigned long len)439 static long __init flush_buffer(void *bufv, unsigned long len)
440 {
441 char *buf = (char *) bufv;
442 long written;
443 long origLen = len;
444 if (message)
445 return -1;
446 while ((written = write_buffer(buf, len)) < len && !message) {
447 char c = buf[written];
448 if (c == '0') {
449 buf += written;
450 len -= written;
451 state = Start;
452 } else if (c == 0) {
453 buf += written;
454 len -= written;
455 state = Reset;
456 } else
457 error("junk within compressed archive");
458 }
459 return origLen;
460 }
461
462 static unsigned long my_inptr; /* index of next byte to be processed in inbuf */
463
464 #include <linux/decompress/generic.h>
465
unpack_to_rootfs(char * buf,unsigned long len)466 static char * __init unpack_to_rootfs(char *buf, unsigned long len)
467 {
468 long written;
469 decompress_fn decompress;
470 const char *compress_name;
471 static __initdata char msg_buf[64];
472
473 header_buf = kmalloc(110, GFP_KERNEL);
474 symlink_buf = kmalloc(PATH_MAX + N_ALIGN(PATH_MAX) + 1, GFP_KERNEL);
475 name_buf = kmalloc(N_ALIGN(PATH_MAX), GFP_KERNEL);
476
477 if (!header_buf || !symlink_buf || !name_buf)
478 panic("can't allocate buffers");
479
480 state = Start;
481 this_header = 0;
482 message = NULL;
483 while (!message && len) {
484 loff_t saved_offset = this_header;
485 if (*buf == '0' && !(this_header & 3)) {
486 state = Start;
487 written = write_buffer(buf, len);
488 buf += written;
489 len -= written;
490 continue;
491 }
492 if (!*buf) {
493 buf++;
494 len--;
495 this_header++;
496 continue;
497 }
498 this_header = 0;
499 decompress = decompress_method(buf, len, &compress_name);
500 pr_debug("Detected %s compressed data\n", compress_name);
501 if (decompress) {
502 int res = decompress(buf, len, NULL, flush_buffer, NULL,
503 &my_inptr, error);
504 if (res)
505 error("decompressor failed");
506 } else if (compress_name) {
507 if (!message) {
508 snprintf(msg_buf, sizeof msg_buf,
509 "compression method %s not configured",
510 compress_name);
511 message = msg_buf;
512 }
513 } else
514 error("invalid magic at start of compressed archive");
515 if (state != Reset)
516 error("junk at the end of compressed archive");
517 this_header = saved_offset + my_inptr;
518 buf += my_inptr;
519 len -= my_inptr;
520 }
521 dir_utime();
522 kfree(name_buf);
523 kfree(symlink_buf);
524 kfree(header_buf);
525 return message;
526 }
527
528 static int __initdata do_retain_initrd;
529
retain_initrd_param(char * str)530 static int __init retain_initrd_param(char *str)
531 {
532 if (*str)
533 return 0;
534 do_retain_initrd = 1;
535 return 1;
536 }
537 __setup("retain_initrd", retain_initrd_param);
538
539 #ifdef CONFIG_ARCH_HAS_KEEPINITRD
keepinitrd_setup(char * __unused)540 static int __init keepinitrd_setup(char *__unused)
541 {
542 do_retain_initrd = 1;
543 return 1;
544 }
545 __setup("keepinitrd", keepinitrd_setup);
546 #endif
547
548 extern char __initramfs_start[];
549 extern unsigned long __initramfs_size;
550 #include <linux/initrd.h>
551 #include <linux/kexec.h>
552
reserve_initrd_mem(void)553 void __init reserve_initrd_mem(void)
554 {
555 phys_addr_t start;
556 unsigned long size;
557
558 /* Ignore the virtul address computed during device tree parsing */
559 initrd_start = initrd_end = 0;
560
561 if (!phys_initrd_size)
562 return;
563 /*
564 * Round the memory region to page boundaries as per free_initrd_mem()
565 * This allows us to detect whether the pages overlapping the initrd
566 * are in use, but more importantly, reserves the entire set of pages
567 * as we don't want these pages allocated for other purposes.
568 */
569 start = round_down(phys_initrd_start, PAGE_SIZE);
570 size = phys_initrd_size + (phys_initrd_start - start);
571 size = round_up(size, PAGE_SIZE);
572
573 if (!memblock_is_region_memory(start, size)) {
574 pr_err("INITRD: 0x%08llx+0x%08lx is not a memory region",
575 (u64)start, size);
576 goto disable;
577 }
578
579 if (memblock_is_region_reserved(start, size)) {
580 pr_err("INITRD: 0x%08llx+0x%08lx overlaps in-use memory region\n",
581 (u64)start, size);
582 goto disable;
583 }
584
585 memblock_reserve(start, size);
586 /* Now convert initrd to virtual addresses */
587 initrd_start = (unsigned long)__va(phys_initrd_start);
588 initrd_end = initrd_start + phys_initrd_size;
589 initrd_below_start_ok = 1;
590
591 return;
592 disable:
593 pr_cont(" - disabling initrd\n");
594 initrd_start = 0;
595 initrd_end = 0;
596 }
597
free_initrd_mem(unsigned long start,unsigned long end)598 void __weak __init free_initrd_mem(unsigned long start, unsigned long end)
599 {
600 #ifdef CONFIG_ARCH_KEEP_MEMBLOCK
601 unsigned long aligned_start = ALIGN_DOWN(start, PAGE_SIZE);
602 unsigned long aligned_end = ALIGN(end, PAGE_SIZE);
603
604 memblock_free(__pa(aligned_start), aligned_end - aligned_start);
605 #endif
606
607 free_reserved_area((void *)start, (void *)end, POISON_FREE_INITMEM,
608 "initrd");
609 }
610
611 #ifdef CONFIG_KEXEC_CORE
kexec_free_initrd(void)612 static bool __init kexec_free_initrd(void)
613 {
614 unsigned long crashk_start = (unsigned long)__va(crashk_res.start);
615 unsigned long crashk_end = (unsigned long)__va(crashk_res.end);
616
617 /*
618 * If the initrd region is overlapped with crashkernel reserved region,
619 * free only memory that is not part of crashkernel region.
620 */
621 if (initrd_start >= crashk_end || initrd_end <= crashk_start)
622 return false;
623
624 /*
625 * Initialize initrd memory region since the kexec boot does not do.
626 */
627 memset((void *)initrd_start, 0, initrd_end - initrd_start);
628 if (initrd_start < crashk_start)
629 free_initrd_mem(initrd_start, crashk_start);
630 if (initrd_end > crashk_end)
631 free_initrd_mem(crashk_end, initrd_end);
632 return true;
633 }
634 #else
kexec_free_initrd(void)635 static inline bool kexec_free_initrd(void)
636 {
637 return false;
638 }
639 #endif /* CONFIG_KEXEC_CORE */
640
641 #ifdef CONFIG_BLK_DEV_RAM
populate_initrd_image(char * err)642 static void __init populate_initrd_image(char *err)
643 {
644 ssize_t written;
645 struct file *file;
646 loff_t pos = 0;
647
648 unpack_to_rootfs(__initramfs_start, __initramfs_size);
649
650 printk(KERN_INFO "rootfs image is not initramfs (%s); looks like an initrd\n",
651 err);
652 file = filp_open("/initrd.image", O_WRONLY | O_CREAT, 0700);
653 if (IS_ERR(file))
654 return;
655
656 written = xwrite(file, (char *)initrd_start, initrd_end - initrd_start,
657 &pos);
658 if (written != initrd_end - initrd_start)
659 pr_err("/initrd.image: incomplete write (%zd != %ld)\n",
660 written, initrd_end - initrd_start);
661 fput(file);
662 }
663 #endif /* CONFIG_BLK_DEV_RAM */
664
populate_rootfs(void)665 static int __init populate_rootfs(void)
666 {
667 /* Load the built in initramfs */
668 char *err = unpack_to_rootfs(__initramfs_start, __initramfs_size);
669 if (err)
670 panic("%s", err); /* Failed to decompress INTERNAL initramfs */
671
672 if (!initrd_start || IS_ENABLED(CONFIG_INITRAMFS_FORCE))
673 goto done;
674
675 if (IS_ENABLED(CONFIG_BLK_DEV_RAM))
676 printk(KERN_INFO "Trying to unpack rootfs image as initramfs...\n");
677 else
678 printk(KERN_INFO "Unpacking initramfs...\n");
679
680 err = unpack_to_rootfs((char *)initrd_start, initrd_end - initrd_start);
681 if (err) {
682 #ifdef CONFIG_BLK_DEV_RAM
683 populate_initrd_image(err);
684 #else
685 printk(KERN_EMERG "Initramfs unpacking failed: %s\n", err);
686 #endif
687 }
688
689 done:
690 /*
691 * If the initrd region is overlapped with crashkernel reserved region,
692 * free only memory that is not part of crashkernel region.
693 */
694 if (!do_retain_initrd && initrd_start && !kexec_free_initrd())
695 free_initrd_mem(initrd_start, initrd_end);
696 initrd_start = 0;
697 initrd_end = 0;
698
699 flush_delayed_fput();
700 return 0;
701 }
702 rootfs_initcall(populate_rootfs);
703