766 lines · c
1// SPDX-License-Identifier: GPL-2.02#include <linux/init.h>3#include <linux/initramfs.h>4#include <linux/async.h>5#include <linux/fs.h>6#include <linux/slab.h>7#include <linux/types.h>8#include <linux/fcntl.h>9#include <linux/delay.h>10#include <linux/string.h>11#include <linux/dirent.h>12#include <linux/syscalls.h>13#include <linux/utime.h>14#include <linux/file.h>15#include <linux/kstrtox.h>16#include <linux/memblock.h>17#include <linux/mm.h>18#include <linux/namei.h>19#include <linux/init_syscalls.h>20#include <linux/umh.h>21#include <linux/security.h>22 23#include "do_mounts.h"24 25static __initdata bool csum_present;26static __initdata u32 io_csum;27 28static ssize_t __init xwrite(struct file *file, const unsigned char *p,29 size_t count, loff_t *pos)30{31 ssize_t out = 0;32 33 /* sys_write only can write MAX_RW_COUNT aka 2G-4K bytes at most */34 while (count) {35 ssize_t rv = kernel_write(file, p, count, pos);36 37 if (rv < 0) {38 if (rv == -EINTR || rv == -EAGAIN)39 continue;40 return out ? out : rv;41 } else if (rv == 0)42 break;43 44 if (csum_present) {45 ssize_t i;46 47 for (i = 0; i < rv; i++)48 io_csum += p[i];49 }50 51 p += rv;52 out += rv;53 count -= rv;54 }55 56 return out;57}58 59static __initdata char *message;60static void __init error(char *x)61{62 if (!message)63 message = x;64}65 66#define panic_show_mem(fmt, ...) \67 ({ show_mem(); panic(fmt, ##__VA_ARGS__); })68 69/* link hash */70 71#define N_ALIGN(len) ((((len) + 1) & ~3) + 2)72 73static __initdata struct hash {74 int ino, minor, major;75 umode_t mode;76 struct hash *next;77 char name[N_ALIGN(PATH_MAX)];78} *head[32];79 80static inline int hash(int major, int minor, int ino)81{82 unsigned long tmp = ino + minor + (major << 3);83 tmp += tmp >> 5;84 return tmp & 31;85}86 87static char __init *find_link(int major, int minor, int ino,88 umode_t mode, char *name)89{90 struct hash **p, *q;91 for (p = head + hash(major, minor, ino); *p; p = &(*p)->next) {92 if ((*p)->ino != ino)93 continue;94 if ((*p)->minor != minor)95 continue;96 if ((*p)->major != major)97 continue;98 if (((*p)->mode ^ mode) & S_IFMT)99 continue;100 return (*p)->name;101 }102 q = kmalloc(sizeof(struct hash), GFP_KERNEL);103 if (!q)104 panic_show_mem("can't allocate link hash entry");105 q->major = major;106 q->minor = minor;107 q->ino = ino;108 q->mode = mode;109 strcpy(q->name, name);110 q->next = NULL;111 *p = q;112 return NULL;113}114 115static void __init free_hash(void)116{117 struct hash **p, *q;118 for (p = head; p < head + 32; p++) {119 while (*p) {120 q = *p;121 *p = q->next;122 kfree(q);123 }124 }125}126 127#ifdef CONFIG_INITRAMFS_PRESERVE_MTIME128static void __init do_utime(char *filename, time64_t mtime)129{130 struct timespec64 t[2] = { { .tv_sec = mtime }, { .tv_sec = mtime } };131 init_utimes(filename, t);132}133 134static void __init do_utime_path(const struct path *path, time64_t mtime)135{136 struct timespec64 t[2] = { { .tv_sec = mtime }, { .tv_sec = mtime } };137 vfs_utimes(path, t);138}139 140static __initdata LIST_HEAD(dir_list);141struct dir_entry {142 struct list_head list;143 time64_t mtime;144 char name[];145};146 147static void __init dir_add(const char *name, time64_t mtime)148{149 size_t nlen = strlen(name) + 1;150 struct dir_entry *de;151 152 de = kmalloc(sizeof(struct dir_entry) + nlen, GFP_KERNEL);153 if (!de)154 panic_show_mem("can't allocate dir_entry buffer");155 INIT_LIST_HEAD(&de->list);156 strscpy(de->name, name, nlen);157 de->mtime = mtime;158 list_add(&de->list, &dir_list);159}160 161static void __init dir_utime(void)162{163 struct dir_entry *de, *tmp;164 list_for_each_entry_safe(de, tmp, &dir_list, list) {165 list_del(&de->list);166 do_utime(de->name, de->mtime);167 kfree(de);168 }169}170#else171static void __init do_utime(char *filename, time64_t mtime) {}172static void __init do_utime_path(const struct path *path, time64_t mtime) {}173static void __init dir_add(const char *name, time64_t mtime) {}174static void __init dir_utime(void) {}175#endif176 177static __initdata time64_t mtime;178 179/* cpio header parsing */180 181static __initdata unsigned long ino, major, minor, nlink;182static __initdata umode_t mode;183static __initdata unsigned long body_len, name_len;184static __initdata uid_t uid;185static __initdata gid_t gid;186static __initdata unsigned rdev;187static __initdata u32 hdr_csum;188 189static void __init parse_header(char *s)190{191 unsigned long parsed[13];192 char buf[9];193 int i;194 195 buf[8] = '\0';196 for (i = 0, s += 6; i < 13; i++, s += 8) {197 memcpy(buf, s, 8);198 parsed[i] = simple_strtoul(buf, NULL, 16);199 }200 ino = parsed[0];201 mode = parsed[1];202 uid = parsed[2];203 gid = parsed[3];204 nlink = parsed[4];205 mtime = parsed[5]; /* breaks in y2106 */206 body_len = parsed[6];207 major = parsed[7];208 minor = parsed[8];209 rdev = new_encode_dev(MKDEV(parsed[9], parsed[10]));210 name_len = parsed[11];211 hdr_csum = parsed[12];212}213 214/* FSM */215 216static __initdata enum state {217 Start,218 Collect,219 GotHeader,220 SkipIt,221 GotName,222 CopyFile,223 GotSymlink,224 Reset225} state, next_state;226 227static __initdata char *victim;228static unsigned long byte_count __initdata;229static __initdata loff_t this_header, next_header;230 231static inline void __init eat(unsigned n)232{233 victim += n;234 this_header += n;235 byte_count -= n;236}237 238static __initdata char *collected;239static long remains __initdata;240static __initdata char *collect;241 242static void __init read_into(char *buf, unsigned size, enum state next)243{244 if (byte_count >= size) {245 collected = victim;246 eat(size);247 state = next;248 } else {249 collect = collected = buf;250 remains = size;251 next_state = next;252 state = Collect;253 }254}255 256static __initdata char *header_buf, *symlink_buf, *name_buf;257 258static int __init do_start(void)259{260 read_into(header_buf, 110, GotHeader);261 return 0;262}263 264static int __init do_collect(void)265{266 unsigned long n = remains;267 if (byte_count < n)268 n = byte_count;269 memcpy(collect, victim, n);270 eat(n);271 collect += n;272 if ((remains -= n) != 0)273 return 1;274 state = next_state;275 return 0;276}277 278static int __init do_header(void)279{280 if (!memcmp(collected, "070701", 6)) {281 csum_present = false;282 } else if (!memcmp(collected, "070702", 6)) {283 csum_present = true;284 } else {285 if (memcmp(collected, "070707", 6) == 0)286 error("incorrect cpio method used: use -H newc option");287 else288 error("no cpio magic");289 return 1;290 }291 parse_header(collected);292 next_header = this_header + N_ALIGN(name_len) + body_len;293 next_header = (next_header + 3) & ~3;294 state = SkipIt;295 if (name_len <= 0 || name_len > PATH_MAX)296 return 0;297 if (S_ISLNK(mode)) {298 if (body_len > PATH_MAX)299 return 0;300 collect = collected = symlink_buf;301 remains = N_ALIGN(name_len) + body_len;302 next_state = GotSymlink;303 state = Collect;304 return 0;305 }306 if (S_ISREG(mode) || !body_len)307 read_into(name_buf, N_ALIGN(name_len), GotName);308 return 0;309}310 311static int __init do_skip(void)312{313 if (this_header + byte_count < next_header) {314 eat(byte_count);315 return 1;316 } else {317 eat(next_header - this_header);318 state = next_state;319 return 0;320 }321}322 323static int __init do_reset(void)324{325 while (byte_count && *victim == '\0')326 eat(1);327 if (byte_count && (this_header & 3))328 error("broken padding");329 return 1;330}331 332static void __init clean_path(char *path, umode_t fmode)333{334 struct kstat st;335 336 if (!init_stat(path, &st, AT_SYMLINK_NOFOLLOW) &&337 (st.mode ^ fmode) & S_IFMT) {338 if (S_ISDIR(st.mode))339 init_rmdir(path);340 else341 init_unlink(path);342 }343}344 345static int __init maybe_link(void)346{347 if (nlink >= 2) {348 char *old = find_link(major, minor, ino, mode, collected);349 if (old) {350 clean_path(collected, 0);351 return (init_link(old, collected) < 0) ? -1 : 1;352 }353 }354 return 0;355}356 357static __initdata struct file *wfile;358static __initdata loff_t wfile_pos;359 360static int __init do_name(void)361{362 state = SkipIt;363 next_state = Reset;364 if (strcmp(collected, "TRAILER!!!") == 0) {365 free_hash();366 return 0;367 }368 clean_path(collected, mode);369 if (S_ISREG(mode)) {370 int ml = maybe_link();371 if (ml >= 0) {372 int openflags = O_WRONLY|O_CREAT|O_LARGEFILE;373 if (ml != 1)374 openflags |= O_TRUNC;375 wfile = filp_open(collected, openflags, mode);376 if (IS_ERR(wfile))377 return 0;378 wfile_pos = 0;379 io_csum = 0;380 381 vfs_fchown(wfile, uid, gid);382 vfs_fchmod(wfile, mode);383 if (body_len)384 vfs_truncate(&wfile->f_path, body_len);385 state = CopyFile;386 }387 } else if (S_ISDIR(mode)) {388 init_mkdir(collected, mode);389 init_chown(collected, uid, gid, 0);390 init_chmod(collected, mode);391 dir_add(collected, mtime);392 } else if (S_ISBLK(mode) || S_ISCHR(mode) ||393 S_ISFIFO(mode) || S_ISSOCK(mode)) {394 if (maybe_link() == 0) {395 init_mknod(collected, mode, rdev);396 init_chown(collected, uid, gid, 0);397 init_chmod(collected, mode);398 do_utime(collected, mtime);399 }400 }401 return 0;402}403 404static int __init do_copy(void)405{406 if (byte_count >= body_len) {407 if (xwrite(wfile, victim, body_len, &wfile_pos) != body_len)408 error("write error");409 410 do_utime_path(&wfile->f_path, mtime);411 fput(wfile);412 if (csum_present && io_csum != hdr_csum)413 error("bad data checksum");414 eat(body_len);415 state = SkipIt;416 return 0;417 } else {418 if (xwrite(wfile, victim, byte_count, &wfile_pos) != byte_count)419 error("write error");420 body_len -= byte_count;421 eat(byte_count);422 return 1;423 }424}425 426static int __init do_symlink(void)427{428 collected[N_ALIGN(name_len) + body_len] = '\0';429 clean_path(collected, 0);430 init_symlink(collected + N_ALIGN(name_len), collected);431 init_chown(collected, uid, gid, AT_SYMLINK_NOFOLLOW);432 do_utime(collected, mtime);433 state = SkipIt;434 next_state = Reset;435 return 0;436}437 438static __initdata int (*actions[])(void) = {439 [Start] = do_start,440 [Collect] = do_collect,441 [GotHeader] = do_header,442 [SkipIt] = do_skip,443 [GotName] = do_name,444 [CopyFile] = do_copy,445 [GotSymlink] = do_symlink,446 [Reset] = do_reset,447};448 449static HWJS_SUSPENDS long __init write_buffer(char *buf, unsigned long len)450{451 byte_count = len;452 victim = buf;453 454 while (!actions[state]())455 ;456 return len - byte_count;457}458 459static HWJS_SUSPENDS long __init flush_buffer(void *bufv, unsigned long len)460{461 char *buf = bufv;462 long written;463 long origLen = len;464 if (message)465 return -1;466 while ((written = write_buffer(buf, len)) < len && !message) {467 char c = buf[written];468 if (c == '0') {469 buf += written;470 len -= written;471 state = Start;472 } else if (c == 0) {473 buf += written;474 len -= written;475 state = Reset;476 } else477 error("junk within compressed archive");478 }479 return origLen;480}481 482static unsigned long my_inptr __initdata; /* index of next byte to be processed in inbuf */483 484#include <linux/decompress/generic.h>485 486char * __init unpack_to_rootfs(char *buf, unsigned long len)487{488 long written;489 decompress_fn decompress;490 const char *compress_name;491 static __initdata char msg_buf[64];492 493 header_buf = kmalloc(110, GFP_KERNEL);494 symlink_buf = kmalloc(PATH_MAX + N_ALIGN(PATH_MAX) + 1, GFP_KERNEL);495 name_buf = kmalloc(N_ALIGN(PATH_MAX), GFP_KERNEL);496 497 if (!header_buf || !symlink_buf || !name_buf)498 panic_show_mem("can't allocate buffers");499 500 state = Start;501 this_header = 0;502 message = NULL;503 while (!message && len) {504 loff_t saved_offset = this_header;505 if (*buf == '0' && !(this_header & 3)) {506 state = Start;507 written = write_buffer(buf, len);508 buf += written;509 len -= written;510 continue;511 }512 if (!*buf) {513 buf++;514 len--;515 this_header++;516 continue;517 }518 this_header = 0;519 decompress = decompress_method(buf, len, &compress_name);520 pr_debug("Detected %s compressed data\n", compress_name);521 if (decompress) {522 int res = decompress(buf, len, NULL, flush_buffer, NULL,523 &my_inptr, error);524 if (res)525 error("decompressor failed");526 } else if (compress_name) {527 if (!message) {528 snprintf(msg_buf, sizeof msg_buf,529 "compression method %s not configured",530 compress_name);531 message = msg_buf;532 }533 } else534 error("invalid magic at start of compressed archive");535 if (state != Reset)536 error("junk at the end of compressed archive");537 this_header = saved_offset + my_inptr;538 buf += my_inptr;539 len -= my_inptr;540 }541 dir_utime();542 kfree(name_buf);543 kfree(symlink_buf);544 kfree(header_buf);545 return message;546}547 548static int __initdata do_retain_initrd;549 550static int __init retain_initrd_param(char *str)551{552 if (*str)553 return 0;554 do_retain_initrd = 1;555 return 1;556}557__setup("retain_initrd", retain_initrd_param);558 559#ifdef CONFIG_ARCH_HAS_KEEPINITRD560static int __init keepinitrd_setup(char *__unused)561{562 do_retain_initrd = 1;563 return 1;564}565__setup("keepinitrd", keepinitrd_setup);566#endif567 568static bool __initdata initramfs_async = true;569static int __init initramfs_async_setup(char *str)570{571 return kstrtobool(str, &initramfs_async) == 0;572}573__setup("initramfs_async=", initramfs_async_setup);574 575extern char __initramfs_start[];576extern unsigned long __initramfs_size;577#include <linux/initrd.h>578#include <linux/kexec.h>579 580static BIN_ATTR(initrd, 0440, sysfs_bin_attr_simple_read, NULL, 0);581 582void __init reserve_initrd_mem(void)583{584 phys_addr_t start;585 unsigned long size;586 587 /* Ignore the virtul address computed during device tree parsing */588 initrd_start = initrd_end = 0;589 590 if (!phys_initrd_size)591 return;592 /*593 * Round the memory region to page boundaries as per free_initrd_mem()594 * This allows us to detect whether the pages overlapping the initrd595 * are in use, but more importantly, reserves the entire set of pages596 * as we don't want these pages allocated for other purposes.597 */598 start = round_down(phys_initrd_start, PAGE_SIZE);599 size = phys_initrd_size + (phys_initrd_start - start);600 size = round_up(size, PAGE_SIZE);601 602 if (!memblock_is_region_memory(start, size)) {603 pr_err("INITRD: 0x%08llx+0x%08lx is not a memory region",604 (u64)start, size);605 goto disable;606 }607 608 if (memblock_is_region_reserved(start, size)) {609 pr_err("INITRD: 0x%08llx+0x%08lx overlaps in-use memory region\n",610 (u64)start, size);611 goto disable;612 }613 614 memblock_reserve(start, size);615 /* Now convert initrd to virtual addresses */616 initrd_start = (unsigned long)__va(phys_initrd_start);617 initrd_end = initrd_start + phys_initrd_size;618 initrd_below_start_ok = 1;619 620 return;621disable:622 pr_cont(" - disabling initrd\n");623 initrd_start = 0;624 initrd_end = 0;625}626 627void __weak __init free_initrd_mem(unsigned long start, unsigned long end)628{629#ifdef CONFIG_ARCH_KEEP_MEMBLOCK630 unsigned long aligned_start = ALIGN_DOWN(start, PAGE_SIZE);631 unsigned long aligned_end = ALIGN(end, PAGE_SIZE);632 633 memblock_free((void *)aligned_start, aligned_end - aligned_start);634#endif635 636 free_reserved_area((void *)start, (void *)end, POISON_FREE_INITMEM,637 "initrd");638}639 640#ifdef CONFIG_CRASH_RESERVE641static bool __init kexec_free_initrd(void)642{643 unsigned long crashk_start = (unsigned long)__va(crashk_res.start);644 unsigned long crashk_end = (unsigned long)__va(crashk_res.end);645 646 /*647 * If the initrd region is overlapped with crashkernel reserved region,648 * free only memory that is not part of crashkernel region.649 */650 if (initrd_start >= crashk_end || initrd_end <= crashk_start)651 return false;652 653 /*654 * Initialize initrd memory region since the kexec boot does not do.655 */656 memset((void *)initrd_start, 0, initrd_end - initrd_start);657 if (initrd_start < crashk_start)658 free_initrd_mem(initrd_start, crashk_start);659 if (initrd_end > crashk_end)660 free_initrd_mem(crashk_end, initrd_end);661 return true;662}663#else664static inline bool kexec_free_initrd(void)665{666 return false;667}668#endif /* CONFIG_KEXEC_CORE */669 670#ifdef CONFIG_BLK_DEV_RAM671static void __init populate_initrd_image(char *err)672{673 ssize_t written;674 struct file *file;675 loff_t pos = 0;676 677 printk(KERN_INFO "rootfs image is not initramfs (%s); looks like an initrd\n",678 err);679 file = filp_open("/initrd.image", O_WRONLY|O_CREAT|O_LARGEFILE, 0700);680 if (IS_ERR(file))681 return;682 683 written = xwrite(file, (char *)initrd_start, initrd_end - initrd_start,684 &pos);685 if (written != initrd_end - initrd_start)686 pr_err("/initrd.image: incomplete write (%zd != %ld)\n",687 written, initrd_end - initrd_start);688 fput(file);689}690#endif /* CONFIG_BLK_DEV_RAM */691 692static void __init do_populate_rootfs(void *unused, async_cookie_t cookie)693{694 /* Load the built in initramfs */695 char *err = unpack_to_rootfs(__initramfs_start, __initramfs_size);696 if (err)697 panic_show_mem("%s", err); /* Failed to decompress INTERNAL initramfs */698 699 if (!initrd_start || IS_ENABLED(CONFIG_INITRAMFS_FORCE))700 goto done;701 702 if (IS_ENABLED(CONFIG_BLK_DEV_RAM))703 printk(KERN_INFO "Trying to unpack rootfs image as initramfs...\n");704 else705 printk(KERN_INFO "Unpacking initramfs...\n");706 707 err = unpack_to_rootfs((char *)initrd_start, initrd_end - initrd_start);708 if (err) {709#ifdef CONFIG_BLK_DEV_RAM710 populate_initrd_image(err);711#else712 printk(KERN_EMERG "Initramfs unpacking failed: %s\n", err);713#endif714 }715 716done:717 security_initramfs_populated();718 719 /*720 * If the initrd region is overlapped with crashkernel reserved region,721 * free only memory that is not part of crashkernel region.722 */723 if (!do_retain_initrd && initrd_start && !kexec_free_initrd()) {724 free_initrd_mem(initrd_start, initrd_end);725 } else if (do_retain_initrd && initrd_start) {726 bin_attr_initrd.size = initrd_end - initrd_start;727 bin_attr_initrd.private = (void *)initrd_start;728 if (sysfs_create_bin_file(firmware_kobj, &bin_attr_initrd))729 pr_err("Failed to create initrd sysfs file");730 }731 initrd_start = 0;732 initrd_end = 0;733 734 init_flush_fput();735}736 737static ASYNC_DOMAIN_EXCLUSIVE(initramfs_domain);738static async_cookie_t initramfs_cookie;739 740void wait_for_initramfs(void)741{742 if (!initramfs_cookie) {743 /*744 * Something before rootfs_initcall wants to access745 * the filesystem/initramfs. Probably a bug. Make a746 * note, avoid deadlocking the machine, and let the747 * caller's access fail as it used to.748 */749 pr_warn_once("wait_for_initramfs() called before rootfs_initcalls\n");750 return;751 }752 async_synchronize_cookie_domain(initramfs_cookie + 1, &initramfs_domain);753}754EXPORT_SYMBOL_GPL(wait_for_initramfs);755 756static HWJS_SUSPENDS int __init populate_rootfs(void)757{758 initramfs_cookie = async_schedule_domain(do_populate_rootfs, NULL,759 &initramfs_domain);760 usermodehelper_enable();761 if (!initramfs_async)762 wait_for_initramfs();763 return 0;764}765rootfs_initcall(populate_rootfs);766