1 // SPDX-License-Identifier: GPL-2.0-only
5 * Copyright (C) 1991, 1992 Linus Torvalds
7 * GK 2/5/95 - Changed to support mounting root fs via NFS
8 * Added initrd & change_root: Werner Almesberger & Hans Lermen, Feb '96
9 * Moan early if gcc is old, avoiding bogus kernels - Paul Gortmaker, May '96
10 * Simplified starting of init: Michael A. Griffith <grif@acm.org>
13 #define DEBUG /* Enable initcall_debug */
15 #include <linux/types.h>
16 #include <linux/extable.h>
17 #include <linux/module.h>
18 #include <linux/proc_fs.h>
19 #include <linux/binfmts.h>
20 #include <linux/kernel.h>
21 #include <linux/syscalls.h>
22 #include <linux/stackprotector.h>
23 #include <linux/string.h>
24 #include <linux/ctype.h>
25 #include <linux/delay.h>
26 #include <linux/ioport.h>
27 #include <linux/init.h>
28 #include <linux/initrd.h>
29 #include <linux/memblock.h>
30 #include <linux/acpi.h>
31 #include <linux/bootconfig.h>
32 #include <linux/console.h>
33 #include <linux/nmi.h>
34 #include <linux/percpu.h>
35 #include <linux/kmod.h>
36 #include <linux/kprobes.h>
37 #include <linux/kmsan.h>
38 #include <linux/vmalloc.h>
39 #include <linux/kernel_stat.h>
40 #include <linux/start_kernel.h>
41 #include <linux/security.h>
42 #include <linux/smp.h>
43 #include <linux/profile.h>
44 #include <linux/kfence.h>
45 #include <linux/rcupdate.h>
46 #include <linux/srcu.h>
47 #include <linux/moduleparam.h>
48 #include <linux/kallsyms.h>
49 #include <linux/buildid.h>
50 #include <linux/writeback.h>
51 #include <linux/cpu.h>
52 #include <linux/cpuset.h>
53 #include <linux/cgroup.h>
54 #include <linux/efi.h>
55 #include <linux/tick.h>
56 #include <linux/sched/isolation.h>
57 #include <linux/interrupt.h>
58 #include <linux/taskstats_kern.h>
59 #include <linux/delayacct.h>
60 #include <linux/unistd.h>
61 #include <linux/utsname.h>
62 #include <linux/rmap.h>
63 #include <linux/mempolicy.h>
64 #include <linux/key.h>
65 #include <linux/page_ext.h>
66 #include <linux/debug_locks.h>
67 #include <linux/debugobjects.h>
68 #include <linux/lockdep.h>
69 #include <linux/kmemleak.h>
70 #include <linux/padata.h>
71 #include <linux/pid_namespace.h>
72 #include <linux/device/driver.h>
73 #include <linux/kthread.h>
74 #include <linux/sched.h>
75 #include <linux/sched/init.h>
76 #include <linux/signal.h>
77 #include <linux/idr.h>
78 #include <linux/kgdb.h>
79 #include <linux/ftrace.h>
80 #include <linux/async.h>
81 #include <linux/shmem_fs.h>
82 #include <linux/slab.h>
83 #include <linux/perf_event.h>
84 #include <linux/ptrace.h>
85 #include <linux/pti.h>
86 #include <linux/blkdev.h>
87 #include <linux/sched/clock.h>
88 #include <linux/sched/task.h>
89 #include <linux/sched/task_stack.h>
90 #include <linux/context_tracking.h>
91 #include <linux/random.h>
92 #include <linux/list.h>
93 #include <linux/integrity.h>
94 #include <linux/proc_ns.h>
96 #include <linux/cache.h>
97 #include <linux/rodata_test.h>
98 #include <linux/jump_label.h>
99 #include <linux/mem_encrypt.h>
100 #include <linux/kcsan.h>
101 #include <linux/init_syscalls.h>
102 #include <linux/stackdepot.h>
103 #include <linux/randomize_kstack.h>
104 #include <net/net_namespace.h>
107 #include <asm/bugs.h>
108 #include <asm/setup.h>
109 #include <asm/sections.h>
110 #include <asm/cacheflush.h>
112 #define CREATE_TRACE_POINTS
113 #include <trace/events/initcall.h>
115 #include <kunit/test.h>
117 static int kernel_init(void *);
119 extern void init_IRQ(void);
120 extern void radix_tree_init(void);
121 extern void maple_tree_init(void);
124 * Debug helper: via this flag we know that we are in 'early bootup code'
125 * where only the boot processor is running with IRQ disabled. This means
126 * two things - IRQ must not be enabled before the flag is cleared and some
127 * operations which are not allowed with IRQ disabled are allowed while the
130 bool early_boot_irqs_disabled __read_mostly;
132 enum system_states system_state __read_mostly;
133 EXPORT_SYMBOL(system_state);
136 * Boot command-line arguments
138 #define MAX_INIT_ARGS CONFIG_INIT_ENV_ARG_LIMIT
139 #define MAX_INIT_ENVS CONFIG_INIT_ENV_ARG_LIMIT
141 extern void time_init(void);
142 /* Default late time init is NULL. archs can override this later. */
143 void (*__initdata late_time_init)(void);
145 /* Untouched command line saved by arch-specific code. */
146 char __initdata boot_command_line[COMMAND_LINE_SIZE];
147 /* Untouched saved command line (eg. for /proc) */
148 char *saved_command_line __ro_after_init;
149 unsigned int saved_command_line_len __ro_after_init;
150 /* Command line for parameter parsing */
151 static char *static_command_line;
152 /* Untouched extra command line */
153 static char *extra_command_line;
154 /* Extra init arguments */
155 static char *extra_init_args;
157 #ifdef CONFIG_BOOT_CONFIG
158 /* Is bootconfig on command line? */
159 static bool bootconfig_found;
160 static size_t initargs_offs;
162 # define bootconfig_found false
163 # define initargs_offs 0
166 static char *execute_command;
167 static char *ramdisk_execute_command = "/init";
170 * Used to generate warnings if static_key manipulation functions are used
171 * before jump_label_init is called.
173 bool static_key_initialized __read_mostly;
174 EXPORT_SYMBOL_GPL(static_key_initialized);
177 * If set, this is an indication to the drivers that reset the underlying
178 * device before going ahead with the initialization otherwise driver might
179 * rely on the BIOS and skip the reset operation.
181 * This is useful if kernel is booting in an unreliable environment.
182 * For ex. kdump situation where previous kernel has crashed, BIOS has been
183 * skipped and devices will be in unknown state.
185 unsigned int reset_devices;
186 EXPORT_SYMBOL(reset_devices);
188 static int __init set_reset_devices(char *str)
194 __setup("reset_devices", set_reset_devices);
196 static const char *argv_init[MAX_INIT_ARGS+2] = { "init", NULL, };
197 const char *envp_init[MAX_INIT_ENVS+2] = { "HOME=/", "TERM=linux", NULL, };
198 static const char *panic_later, *panic_param;
200 extern const struct obs_kernel_param __setup_start[], __setup_end[];
202 static bool __init obsolete_checksetup(char *line)
204 const struct obs_kernel_param *p;
205 bool had_early_param = false;
209 int n = strlen(p->str);
210 if (parameqn(line, p->str, n)) {
212 /* Already done in parse_early_param?
213 * (Needs exact match on param part).
214 * Keep iterating, as we can have early
215 * params and __setups of same names 8( */
216 if (line[n] == '\0' || line[n] == '=')
217 had_early_param = true;
218 } else if (!p->setup_func) {
219 pr_warn("Parameter %s is obsolete, ignored\n",
222 } else if (p->setup_func(line + n))
226 } while (p < __setup_end);
228 return had_early_param;
232 * This should be approx 2 Bo*oMips to start (note initial shift), and will
233 * still work even if initially too large, it will just take slightly longer
235 unsigned long loops_per_jiffy = (1<<12);
236 EXPORT_SYMBOL(loops_per_jiffy);
238 static int __init debug_kernel(char *str)
240 console_loglevel = CONSOLE_LOGLEVEL_DEBUG;
244 static int __init quiet_kernel(char *str)
246 console_loglevel = CONSOLE_LOGLEVEL_QUIET;
250 early_param("debug", debug_kernel);
251 early_param("quiet", quiet_kernel);
253 static int __init loglevel(char *str)
258 * Only update loglevel value when a correct setting was passed,
259 * to prevent blind crashes (when loglevel being set to 0) that
260 * are quite hard to debug
262 if (get_option(&str, &newlevel)) {
263 console_loglevel = newlevel;
270 early_param("loglevel", loglevel);
272 #ifdef CONFIG_BLK_DEV_INITRD
273 static void * __init get_boot_config_from_initrd(size_t *_size)
283 data = (char *)initrd_end - BOOTCONFIG_MAGIC_LEN;
285 * Since Grub may align the size of initrd to 4, we must
286 * check the preceding 3 bytes as well.
288 for (i = 0; i < 4; i++) {
289 if (!memcmp(data, BOOTCONFIG_MAGIC, BOOTCONFIG_MAGIC_LEN))
296 hdr = (u32 *)(data - 8);
297 size = le32_to_cpu(hdr[0]);
298 csum = le32_to_cpu(hdr[1]);
300 data = ((void *)hdr) - size;
301 if ((unsigned long)data < initrd_start) {
302 pr_err("bootconfig size %d is greater than initrd size %ld\n",
303 size, initrd_end - initrd_start);
307 if (xbc_calc_checksum(data, size) != csum) {
308 pr_err("bootconfig checksum failed\n");
312 /* Remove bootconfig from initramfs/initrd */
313 initrd_end = (unsigned long)data;
320 static void * __init get_boot_config_from_initrd(size_t *_size)
326 #ifdef CONFIG_BOOT_CONFIG
328 static char xbc_namebuf[XBC_KEYLEN_MAX] __initdata;
330 #define rest(dst, end) ((end) > (dst) ? (end) - (dst) : 0)
332 static int __init xbc_snprint_cmdline(char *buf, size_t size,
333 struct xbc_node *root)
335 struct xbc_node *knode, *vnode;
336 char *end = buf + size;
340 xbc_node_for_each_key_value(root, knode, val) {
341 ret = xbc_node_compose_key_after(root, knode,
342 xbc_namebuf, XBC_KEYLEN_MAX);
346 vnode = xbc_node_get_child(knode);
348 ret = snprintf(buf, rest(buf, end), "%s ", xbc_namebuf);
354 xbc_array_for_each_value(vnode, val) {
355 ret = snprintf(buf, rest(buf, end), "%s=\"%s\" ",
363 return buf - (end - size);
367 /* Make an extra command line under given key word */
368 static char * __init xbc_make_cmdline(const char *key)
370 struct xbc_node *root;
374 root = xbc_find_node(key);
378 /* Count required buffer size */
379 len = xbc_snprint_cmdline(NULL, 0, root);
383 new_cmdline = memblock_alloc(len + 1, SMP_CACHE_BYTES);
385 pr_err("Failed to allocate memory for extra kernel cmdline.\n");
389 ret = xbc_snprint_cmdline(new_cmdline, len + 1, root);
390 if (ret < 0 || ret > len) {
391 pr_err("Failed to print extra kernel cmdline.\n");
392 memblock_free(new_cmdline, len + 1);
399 static int __init bootconfig_params(char *param, char *val,
400 const char *unused, void *arg)
402 if (strcmp(param, "bootconfig") == 0) {
403 bootconfig_found = true;
408 static int __init warn_bootconfig(char *str)
410 /* The 'bootconfig' has been handled by bootconfig_params(). */
414 static void __init setup_boot_config(void)
416 static char tmp_cmdline[COMMAND_LINE_SIZE] __initdata;
417 const char *msg, *data;
422 /* Cut out the bootconfig data even if we have no bootconfig option */
423 data = get_boot_config_from_initrd(&size);
424 /* If there is no bootconfig in initrd, try embedded one. */
426 data = xbc_get_embedded_bootconfig(&size);
428 strscpy(tmp_cmdline, boot_command_line, COMMAND_LINE_SIZE);
429 err = parse_args("bootconfig", tmp_cmdline, NULL, 0, 0, 0, NULL,
432 if (IS_ERR(err) || !bootconfig_found)
435 /* parse_args() stops at the next param of '--' and returns an address */
437 initargs_offs = err - tmp_cmdline;
440 pr_err("'bootconfig' found on command line, but no bootconfig found\n");
444 if (size >= XBC_DATA_MAX) {
445 pr_err("bootconfig size %ld greater than max size %d\n",
446 (long)size, XBC_DATA_MAX);
450 ret = xbc_init(data, size, &msg, &pos);
453 pr_err("Failed to init bootconfig: %s.\n", msg);
455 pr_err("Failed to parse bootconfig: %s at %d.\n",
458 xbc_get_info(&ret, NULL);
459 pr_info("Load bootconfig: %ld bytes %d nodes\n", (long)size, ret);
460 /* keys starting with "kernel." are passed via cmdline */
461 extra_command_line = xbc_make_cmdline("kernel");
462 /* Also, "init." keys are init arguments */
463 extra_init_args = xbc_make_cmdline("init");
468 static void __init exit_boot_config(void)
473 #else /* !CONFIG_BOOT_CONFIG */
475 static void __init setup_boot_config(void)
477 /* Remove bootconfig data from initrd */
478 get_boot_config_from_initrd(NULL);
481 static int __init warn_bootconfig(char *str)
483 pr_warn("WARNING: 'bootconfig' found on the kernel command line but CONFIG_BOOT_CONFIG is not set.\n");
487 #define exit_boot_config() do {} while (0)
489 #endif /* CONFIG_BOOT_CONFIG */
491 early_param("bootconfig", warn_bootconfig);
493 /* Change NUL term back to "=", to make "param" the whole string. */
494 static void __init repair_env_string(char *param, char *val)
497 /* param=val or param="val"? */
498 if (val == param+strlen(param)+1)
500 else if (val == param+strlen(param)+2) {
502 memmove(val-1, val, strlen(val)+1);
508 /* Anything after -- gets handed straight to init. */
509 static int __init set_init_arg(char *param, char *val,
510 const char *unused, void *arg)
517 repair_env_string(param, val);
519 for (i = 0; argv_init[i]; i++) {
520 if (i == MAX_INIT_ARGS) {
521 panic_later = "init";
526 argv_init[i] = param;
531 * Unknown boot options get handed to init, unless they look like
532 * unused parameters (modprobe will find them in /proc/cmdline).
534 static int __init unknown_bootoption(char *param, char *val,
535 const char *unused, void *arg)
537 size_t len = strlen(param);
539 repair_env_string(param, val);
541 /* Handle obsolete-style parameters */
542 if (obsolete_checksetup(param))
545 /* Unused module parameter. */
546 if (strnchr(param, len, '.'))
553 /* Environment option */
555 for (i = 0; envp_init[i]; i++) {
556 if (i == MAX_INIT_ENVS) {
560 if (!strncmp(param, envp_init[i], len+1))
563 envp_init[i] = param;
565 /* Command line option */
567 for (i = 0; argv_init[i]; i++) {
568 if (i == MAX_INIT_ARGS) {
569 panic_later = "init";
573 argv_init[i] = param;
578 static int __init init_setup(char *str)
582 execute_command = str;
584 * In case LILO is going to boot us with default command line,
585 * it prepends "auto" before the whole cmdline which makes
586 * the shell think it should execute a script with such name.
587 * So we ignore all arguments entered _before_ init=... [MJ]
589 for (i = 1; i < MAX_INIT_ARGS; i++)
593 __setup("init=", init_setup);
595 static int __init rdinit_setup(char *str)
599 ramdisk_execute_command = str;
600 /* See "auto" comment in init_setup */
601 for (i = 1; i < MAX_INIT_ARGS; i++)
605 __setup("rdinit=", rdinit_setup);
608 static const unsigned int setup_max_cpus = NR_CPUS;
609 static inline void setup_nr_cpu_ids(void) { }
610 static inline void smp_prepare_cpus(unsigned int maxcpus) { }
614 * We need to store the untouched command line for future reference.
615 * We also need to store the touched command line since the parameter
616 * parsing is performed in place, and we should allow a component to
617 * store reference of name/value for future reference.
619 static void __init setup_command_line(char *command_line)
621 size_t len, xlen = 0, ilen = 0;
623 if (extra_command_line)
624 xlen = strlen(extra_command_line);
626 ilen = strlen(extra_init_args) + 4; /* for " -- " */
628 len = xlen + strlen(boot_command_line) + 1;
630 saved_command_line = memblock_alloc(len + ilen, SMP_CACHE_BYTES);
631 if (!saved_command_line)
632 panic("%s: Failed to allocate %zu bytes\n", __func__, len + ilen);
634 static_command_line = memblock_alloc(len, SMP_CACHE_BYTES);
635 if (!static_command_line)
636 panic("%s: Failed to allocate %zu bytes\n", __func__, len);
640 * We have to put extra_command_line before boot command
641 * lines because there could be dashes (separator of init
642 * command line) in the command lines.
644 strcpy(saved_command_line, extra_command_line);
645 strcpy(static_command_line, extra_command_line);
647 strcpy(saved_command_line + xlen, boot_command_line);
648 strcpy(static_command_line + xlen, command_line);
652 * Append supplemental init boot args to saved_command_line
653 * so that user can check what command line options passed
655 * The order should always be
656 * " -- "[bootconfig init-param][cmdline init-param]
659 len = xlen + initargs_offs;
660 strcpy(saved_command_line + len, extra_init_args);
661 len += ilen - 4; /* strlen(extra_init_args) */
662 strcpy(saved_command_line + len,
663 boot_command_line + initargs_offs - 1);
665 len = strlen(saved_command_line);
666 strcpy(saved_command_line + len, " -- ");
668 strcpy(saved_command_line + len, extra_init_args);
672 saved_command_line_len = strlen(saved_command_line);
676 * We need to finalize in a non-__init function or else race conditions
677 * between the root thread and the init thread may cause start_kernel to
678 * be reaped by free_initmem before the root thread has proceeded to
681 * gcc-3.4 accidentally inlines this function, so use noinline.
684 static __initdata DECLARE_COMPLETION(kthreadd_done);
686 noinline void __ref rest_init(void)
688 struct task_struct *tsk;
691 rcu_scheduler_starting();
693 * We need to spawn init first so that it obtains pid 1, however
694 * the init task will end up wanting to create kthreads, which, if
695 * we schedule it before we create kthreadd, will OOPS.
697 pid = user_mode_thread(kernel_init, NULL, CLONE_FS);
699 * Pin init on the boot CPU. Task migration is not properly working
700 * until sched_init_smp() has been run. It will set the allowed
701 * CPUs for init to the non isolated CPUs.
704 tsk = find_task_by_pid_ns(pid, &init_pid_ns);
705 tsk->flags |= PF_NO_SETAFFINITY;
706 set_cpus_allowed_ptr(tsk, cpumask_of(smp_processor_id()));
709 numa_default_policy();
710 pid = kernel_thread(kthreadd, NULL, CLONE_FS | CLONE_FILES);
712 kthreadd_task = find_task_by_pid_ns(pid, &init_pid_ns);
716 * Enable might_sleep() and smp_processor_id() checks.
717 * They cannot be enabled earlier because with CONFIG_PREEMPTION=y
718 * kernel_thread() would trigger might_sleep() splats. With
719 * CONFIG_PREEMPT_VOLUNTARY=y the init task might have scheduled
720 * already, but it's stuck on the kthreadd_done completion.
722 system_state = SYSTEM_SCHEDULING;
724 complete(&kthreadd_done);
727 * The boot idle thread must execute schedule()
728 * at least once to get things moving:
730 schedule_preempt_disabled();
731 /* Call into cpu_idle with preempt disabled */
732 cpu_startup_entry(CPUHP_ONLINE);
735 /* Check for early params. */
736 static int __init do_early_param(char *param, char *val,
737 const char *unused, void *arg)
739 const struct obs_kernel_param *p;
741 for (p = __setup_start; p < __setup_end; p++) {
742 if ((p->early && parameq(param, p->str)) ||
743 (strcmp(param, "console") == 0 &&
744 strcmp(p->str, "earlycon") == 0)
746 if (p->setup_func(val) != 0)
747 pr_warn("Malformed early option '%s'\n", param);
750 /* We accept everything at this stage. */
754 void __init parse_early_options(char *cmdline)
756 parse_args("early options", cmdline, NULL, 0, 0, 0, NULL,
760 /* Arch code calls this early on, or if not, just before other parsing. */
761 void __init parse_early_param(void)
763 static int done __initdata;
764 static char tmp_cmdline[COMMAND_LINE_SIZE] __initdata;
769 /* All fall through to do_early_param. */
770 strscpy(tmp_cmdline, boot_command_line, COMMAND_LINE_SIZE);
771 parse_early_options(tmp_cmdline);
775 void __init __weak arch_post_acpi_subsys_init(void) { }
777 void __init __weak smp_setup_processor_id(void)
781 # if THREAD_SIZE >= PAGE_SIZE
782 void __init __weak thread_stack_cache_init(void)
787 void __init __weak mem_encrypt_init(void) { }
789 void __init __weak poking_init(void) { }
791 void __init __weak pgtable_cache_init(void) { }
793 void __init __weak trap_init(void) { }
796 core_param(initcall_debug, initcall_debug, bool, 0644);
798 #ifdef TRACEPOINTS_ENABLED
799 static void __init initcall_debug_enable(void);
801 static inline void initcall_debug_enable(void)
806 /* Report memory auto-initialization states for this boot. */
807 static void __init report_meminit(void)
811 if (IS_ENABLED(CONFIG_INIT_STACK_ALL_PATTERN))
812 stack = "all(pattern)";
813 else if (IS_ENABLED(CONFIG_INIT_STACK_ALL_ZERO))
815 else if (IS_ENABLED(CONFIG_GCC_PLUGIN_STRUCTLEAK_BYREF_ALL))
816 stack = "byref_all(zero)";
817 else if (IS_ENABLED(CONFIG_GCC_PLUGIN_STRUCTLEAK_BYREF))
818 stack = "byref(zero)";
819 else if (IS_ENABLED(CONFIG_GCC_PLUGIN_STRUCTLEAK_USER))
820 stack = "__user(zero)";
824 pr_info("mem auto-init: stack:%s, heap alloc:%s, heap free:%s\n",
825 stack, want_init_on_alloc(GFP_KERNEL) ? "on" : "off",
826 want_init_on_free() ? "on" : "off");
827 if (want_init_on_free())
828 pr_info("mem auto-init: clearing system memory may take some time...\n");
832 * Set up kernel memory allocators
834 static void __init mm_init(void)
837 * page_ext requires contiguous pages,
838 * bigger than MAX_ORDER unless SPARSEMEM.
840 page_ext_init_flatmem();
841 init_mem_debugging_and_hardening();
845 stack_depot_early_init();
847 mem_init_print_info();
850 * page_owner must be initialized after buddy is ready, and also after
851 * slab is ready so that stack_depot_init() works properly
853 page_ext_init_flatmem_late();
856 debug_objects_mem_init();
858 /* Should be run after vmap initialization */
859 if (early_page_ext_enabled())
861 /* Should be run before the first non-init thread is created */
863 /* Should be run after espfix64 is set up. */
865 kmsan_init_runtime();
869 #ifdef CONFIG_RANDOMIZE_KSTACK_OFFSET
870 DEFINE_STATIC_KEY_MAYBE_RO(CONFIG_RANDOMIZE_KSTACK_OFFSET_DEFAULT,
871 randomize_kstack_offset);
872 DEFINE_PER_CPU(u32, kstack_offset);
874 static int __init early_randomize_kstack_offset(char *buf)
879 ret = kstrtobool(buf, &bool_result);
884 static_branch_enable(&randomize_kstack_offset);
886 static_branch_disable(&randomize_kstack_offset);
889 early_param("randomize_kstack_offset", early_randomize_kstack_offset);
892 void __init __weak arch_call_rest_init(void)
897 static void __init print_unknown_bootoptions(void)
899 char *unknown_options;
901 const char *const *p;
904 if (panic_later || (!argv_init[1] && !envp_init[2]))
908 * Determine how many options we have to print out, plus a space
911 len = 1; /* null terminator */
912 for (p = &argv_init[1]; *p; p++) {
916 for (p = &envp_init[2]; *p; p++) {
921 unknown_options = memblock_alloc(len, SMP_CACHE_BYTES);
922 if (!unknown_options) {
923 pr_err("%s: Failed to allocate %zu bytes\n",
927 end = unknown_options;
929 for (p = &argv_init[1]; *p; p++)
930 end += sprintf(end, " %s", *p);
931 for (p = &envp_init[2]; *p; p++)
932 end += sprintf(end, " %s", *p);
934 /* Start at unknown_options[1] to skip the initial space */
935 pr_notice("Unknown kernel command line parameters \"%s\", will be passed to user space.\n",
936 &unknown_options[1]);
937 memblock_free(unknown_options, len);
940 asmlinkage __visible void __init __no_sanitize_address start_kernel(void)
945 set_task_stack_end_magic(&init_task);
946 smp_setup_processor_id();
947 debug_objects_early_init();
948 init_vmlinux_build_id();
953 early_boot_irqs_disabled = true;
956 * Interrupts are still disabled. Do necessary setups, then
961 pr_notice("%s", linux_banner);
962 early_security_init();
963 setup_arch(&command_line);
965 setup_command_line(command_line);
967 setup_per_cpu_areas();
968 smp_prepare_boot_cpu(); /* arch-specific boot-cpu hooks */
969 boot_cpu_hotplug_init();
971 build_all_zonelists(NULL);
974 pr_notice("Kernel command line: %s\n", saved_command_line);
975 /* parameters may set static keys */
978 after_dashes = parse_args("Booting kernel",
979 static_command_line, __start___param,
980 __stop___param - __start___param,
981 -1, -1, NULL, &unknown_bootoption);
982 print_unknown_bootoptions();
983 if (!IS_ERR_OR_NULL(after_dashes))
984 parse_args("Setting init args", after_dashes, NULL, 0, -1, -1,
987 parse_args("Setting extra init args", extra_init_args,
988 NULL, 0, -1, -1, NULL, set_init_arg);
990 /* Architectural and non-timekeeping rng init, before allocator init */
991 random_init_early(command_line);
994 * These use large bootmem allocations and must precede
998 vfs_caches_init_early();
1005 /* trace_printk can be enabled here */
1009 * Set up the scheduler prior starting any interrupts (such as the
1010 * timer interrupt). Full topology setup happens at smp_init()
1011 * time - but meanwhile we still have a functioning scheduler.
1015 if (WARN(!irqs_disabled(),
1016 "Interrupts were enabled *very* early, fixing it\n"))
1017 local_irq_disable();
1022 * Set up housekeeping before setting up workqueues to allow the unbound
1023 * workqueue to take non-housekeeping into account.
1025 housekeeping_init();
1028 * Allow workqueue creation and work item queueing/cancelling
1029 * early. Work item execution depends on kthreads and starts after
1032 workqueue_init_early();
1036 /* Trace events are available after this */
1040 initcall_debug_enable();
1042 context_tracking_init();
1043 /* init some links before init_ISA_irqs() */
1055 /* This must be after timekeeping is initialized */
1058 /* These make use of the fully initialized rng */
1060 boot_init_stack_canary();
1064 call_function_init();
1065 WARN(!irqs_disabled(), "Interrupts were enabled early\n");
1067 early_boot_irqs_disabled = false;
1070 kmem_cache_init_late();
1073 * HACK ALERT! This is early. We're enabling the console before
1074 * we've done PCI setups etc, and console_init() must be aware of
1075 * this. But we do want output early, in case something goes wrong.
1079 panic("Too many boot %s vars at `%s'", panic_later,
1085 * Need to run this when irqs are enabled, because it wants
1086 * to self-test [hard/soft]-irqs on/off lock inversion bugs
1092 * This needs to be called before any devices perform DMA
1093 * operations that might use the SWIOTLB bounce buffers. It will
1094 * mark the bounce buffers as decrypted so that their usage will
1095 * not cause "plain-text" data to be decrypted when accessed.
1099 #ifdef CONFIG_BLK_DEV_INITRD
1100 if (initrd_start && !initrd_below_start_ok &&
1101 page_to_pfn(virt_to_page((void *)initrd_start)) < min_low_pfn) {
1102 pr_crit("initrd overwritten (0x%08lx < 0x%08lx) - disabling it.\n",
1103 page_to_pfn(virt_to_page((void *)initrd_start)),
1108 setup_per_cpu_pageset();
1118 if (efi_enabled(EFI_RUNTIME_SERVICES))
1119 efi_enter_virtual_mode();
1121 thread_stack_cache_init();
1138 taskstats_init_early();
1143 acpi_subsystem_init();
1144 arch_post_acpi_subsys_init();
1147 /* Do the rest non-__init'ed, we're now alive */
1148 arch_call_rest_init();
1150 prevent_tail_call_optimization();
1153 /* Call all constructor functions linked into the kernel. */
1154 static void __init do_ctors(void)
1157 * For UML, the constructors have already been called by the
1158 * normal setup code as it's just a normal ELF binary, so we
1159 * cannot do it again - but we do need CONFIG_CONSTRUCTORS
1160 * even on UML for modules.
1162 #if defined(CONFIG_CONSTRUCTORS) && !defined(CONFIG_UML)
1163 ctor_fn_t *fn = (ctor_fn_t *) __ctors_start;
1165 for (; fn < (ctor_fn_t *) __ctors_end; fn++)
1170 #ifdef CONFIG_KALLSYMS
1171 struct blacklist_entry {
1172 struct list_head next;
1176 static __initdata_or_module LIST_HEAD(blacklisted_initcalls);
1178 static int __init initcall_blacklist(char *str)
1181 struct blacklist_entry *entry;
1183 /* str argument is a comma-separated list of functions */
1185 str_entry = strsep(&str, ",");
1187 pr_debug("blacklisting initcall %s\n", str_entry);
1188 entry = memblock_alloc(sizeof(*entry),
1191 panic("%s: Failed to allocate %zu bytes\n",
1192 __func__, sizeof(*entry));
1193 entry->buf = memblock_alloc(strlen(str_entry) + 1,
1196 panic("%s: Failed to allocate %zu bytes\n",
1197 __func__, strlen(str_entry) + 1);
1198 strcpy(entry->buf, str_entry);
1199 list_add(&entry->next, &blacklisted_initcalls);
1201 } while (str_entry);
1206 static bool __init_or_module initcall_blacklisted(initcall_t fn)
1208 struct blacklist_entry *entry;
1209 char fn_name[KSYM_SYMBOL_LEN];
1212 if (list_empty(&blacklisted_initcalls))
1215 addr = (unsigned long) dereference_function_descriptor(fn);
1216 sprint_symbol_no_offset(fn_name, addr);
1219 * fn will be "function_name [module_name]" where [module_name] is not
1220 * displayed for built-in init functions. Strip off the [module_name].
1222 strreplace(fn_name, ' ', '\0');
1224 list_for_each_entry(entry, &blacklisted_initcalls, next) {
1225 if (!strcmp(fn_name, entry->buf)) {
1226 pr_debug("initcall %s blacklisted\n", fn_name);
1234 static int __init initcall_blacklist(char *str)
1236 pr_warn("initcall_blacklist requires CONFIG_KALLSYMS\n");
1240 static bool __init_or_module initcall_blacklisted(initcall_t fn)
1245 __setup("initcall_blacklist=", initcall_blacklist);
1247 static __init_or_module void
1248 trace_initcall_start_cb(void *data, initcall_t fn)
1250 ktime_t *calltime = data;
1252 printk(KERN_DEBUG "calling %pS @ %i\n", fn, task_pid_nr(current));
1253 *calltime = ktime_get();
1256 static __init_or_module void
1257 trace_initcall_finish_cb(void *data, initcall_t fn, int ret)
1259 ktime_t rettime, *calltime = data;
1261 rettime = ktime_get();
1262 printk(KERN_DEBUG "initcall %pS returned %d after %lld usecs\n",
1263 fn, ret, (unsigned long long)ktime_us_delta(rettime, *calltime));
1266 static ktime_t initcall_calltime;
1268 #ifdef TRACEPOINTS_ENABLED
1269 static void __init initcall_debug_enable(void)
1273 ret = register_trace_initcall_start(trace_initcall_start_cb,
1274 &initcall_calltime);
1275 ret |= register_trace_initcall_finish(trace_initcall_finish_cb,
1276 &initcall_calltime);
1277 WARN(ret, "Failed to register initcall tracepoints\n");
1279 # define do_trace_initcall_start trace_initcall_start
1280 # define do_trace_initcall_finish trace_initcall_finish
1282 static inline void do_trace_initcall_start(initcall_t fn)
1284 if (!initcall_debug)
1286 trace_initcall_start_cb(&initcall_calltime, fn);
1288 static inline void do_trace_initcall_finish(initcall_t fn, int ret)
1290 if (!initcall_debug)
1292 trace_initcall_finish_cb(&initcall_calltime, fn, ret);
1294 #endif /* !TRACEPOINTS_ENABLED */
1296 int __init_or_module do_one_initcall(initcall_t fn)
1298 int count = preempt_count();
1302 if (initcall_blacklisted(fn))
1305 do_trace_initcall_start(fn);
1307 do_trace_initcall_finish(fn, ret);
1311 if (preempt_count() != count) {
1312 sprintf(msgbuf, "preemption imbalance ");
1313 preempt_count_set(count);
1315 if (irqs_disabled()) {
1316 strlcat(msgbuf, "disabled interrupts ", sizeof(msgbuf));
1319 WARN(msgbuf[0], "initcall %pS returned with %s\n", fn, msgbuf);
1321 add_latent_entropy();
1326 extern initcall_entry_t __initcall_start[];
1327 extern initcall_entry_t __initcall0_start[];
1328 extern initcall_entry_t __initcall1_start[];
1329 extern initcall_entry_t __initcall2_start[];
1330 extern initcall_entry_t __initcall3_start[];
1331 extern initcall_entry_t __initcall4_start[];
1332 extern initcall_entry_t __initcall5_start[];
1333 extern initcall_entry_t __initcall6_start[];
1334 extern initcall_entry_t __initcall7_start[];
1335 extern initcall_entry_t __initcall_end[];
1337 static initcall_entry_t *initcall_levels[] __initdata = {
1349 /* Keep these in sync with initcalls in include/linux/init.h */
1350 static const char *initcall_level_names[] __initdata = {
1361 static int __init ignore_unknown_bootoption(char *param, char *val,
1362 const char *unused, void *arg)
1367 static void __init do_initcall_level(int level, char *command_line)
1369 initcall_entry_t *fn;
1371 parse_args(initcall_level_names[level],
1372 command_line, __start___param,
1373 __stop___param - __start___param,
1375 NULL, ignore_unknown_bootoption);
1377 trace_initcall_level(initcall_level_names[level]);
1378 for (fn = initcall_levels[level]; fn < initcall_levels[level+1]; fn++)
1379 do_one_initcall(initcall_from_entry(fn));
1382 static void __init do_initcalls(void)
1385 size_t len = saved_command_line_len + 1;
1388 command_line = kzalloc(len, GFP_KERNEL);
1390 panic("%s: Failed to allocate %zu bytes\n", __func__, len);
1392 for (level = 0; level < ARRAY_SIZE(initcall_levels) - 1; level++) {
1393 /* Parser modifies command_line, restore it each time */
1394 strcpy(command_line, saved_command_line);
1395 do_initcall_level(level, command_line);
1398 kfree(command_line);
1402 * Ok, the machine is now initialized. None of the devices
1403 * have been touched yet, but the CPU subsystem is up and
1404 * running, and memory and process management works.
1406 * Now we can finally start doing some real work..
1408 static void __init do_basic_setup(void)
1417 static void __init do_pre_smp_initcalls(void)
1419 initcall_entry_t *fn;
1421 trace_initcall_level("early");
1422 for (fn = __initcall_start; fn < __initcall0_start; fn++)
1423 do_one_initcall(initcall_from_entry(fn));
1426 static int run_init_process(const char *init_filename)
1428 const char *const *p;
1430 argv_init[0] = init_filename;
1431 pr_info("Run %s as init process\n", init_filename);
1432 pr_debug(" with arguments:\n");
1433 for (p = argv_init; *p; p++)
1434 pr_debug(" %s\n", *p);
1435 pr_debug(" with environment:\n");
1436 for (p = envp_init; *p; p++)
1437 pr_debug(" %s\n", *p);
1438 return kernel_execve(init_filename, argv_init, envp_init);
1441 static int try_to_run_init_process(const char *init_filename)
1445 ret = run_init_process(init_filename);
1447 if (ret && ret != -ENOENT) {
1448 pr_err("Starting init: %s exists but couldn't execute it (error %d)\n",
1449 init_filename, ret);
1455 static noinline void __init kernel_init_freeable(void);
1457 #if defined(CONFIG_STRICT_KERNEL_RWX) || defined(CONFIG_STRICT_MODULE_RWX)
1458 bool rodata_enabled __ro_after_init = true;
1460 #ifndef arch_parse_debug_rodata
1461 static inline bool arch_parse_debug_rodata(char *str) { return false; }
1464 static int __init set_debug_rodata(char *str)
1466 if (arch_parse_debug_rodata(str))
1469 if (str && !strcmp(str, "on"))
1470 rodata_enabled = true;
1471 else if (str && !strcmp(str, "off"))
1472 rodata_enabled = false;
1474 pr_warn("Invalid option string for rodata: '%s'\n", str);
1477 early_param("rodata", set_debug_rodata);
1480 #ifdef CONFIG_STRICT_KERNEL_RWX
1481 static void mark_readonly(void)
1483 if (rodata_enabled) {
1485 * load_module() results in W+X mappings, which are cleaned
1486 * up with call_rcu(). Let's make sure that queued work is
1487 * flushed so that we don't hit false positives looking for
1488 * insecure pages which are W+X.
1494 pr_info("Kernel memory protection disabled.\n");
1496 #elif defined(CONFIG_ARCH_HAS_STRICT_KERNEL_RWX)
1497 static inline void mark_readonly(void)
1499 pr_warn("Kernel memory protection not selected by kernel config.\n");
1502 static inline void mark_readonly(void)
1504 pr_warn("This architecture does not have kernel memory protection.\n");
1508 void __weak free_initmem(void)
1510 free_initmem_default(POISON_FREE_INITMEM);
1513 static int __ref kernel_init(void *unused)
1518 * Wait until kthreadd is all set-up.
1520 wait_for_completion(&kthreadd_done);
1522 kernel_init_freeable();
1523 /* need to finish all async __init code before freeing the memory */
1524 async_synchronize_full();
1526 system_state = SYSTEM_FREEING_INITMEM;
1527 kprobe_free_init_mem();
1528 ftrace_free_init_mem();
1529 kgdb_free_init_mem();
1535 * Kernel mappings are now finalized - update the userspace page-table
1540 system_state = SYSTEM_RUNNING;
1541 numa_default_policy();
1543 rcu_end_inkernel_boot();
1547 if (ramdisk_execute_command) {
1548 ret = run_init_process(ramdisk_execute_command);
1551 pr_err("Failed to execute %s (error %d)\n",
1552 ramdisk_execute_command, ret);
1556 * We try each of these until one succeeds.
1558 * The Bourne shell can be used instead of init if we are
1559 * trying to recover a really broken machine.
1561 if (execute_command) {
1562 ret = run_init_process(execute_command);
1565 panic("Requested init %s failed (error %d).",
1566 execute_command, ret);
1569 if (CONFIG_DEFAULT_INIT[0] != '\0') {
1570 ret = run_init_process(CONFIG_DEFAULT_INIT);
1572 pr_err("Default init %s failed (error %d)\n",
1573 CONFIG_DEFAULT_INIT, ret);
1578 if (!try_to_run_init_process("/sbin/init") ||
1579 !try_to_run_init_process("/etc/init") ||
1580 !try_to_run_init_process("/bin/init") ||
1581 !try_to_run_init_process("/bin/sh"))
1584 panic("No working init found. Try passing init= option to kernel. "
1585 "See Linux Documentation/admin-guide/init.rst for guidance.");
1588 /* Open /dev/console, for stdin/stdout/stderr, this should never fail */
1589 void __init console_on_rootfs(void)
1591 struct file *file = filp_open("/dev/console", O_RDWR, 0);
1594 pr_err("Warning: unable to open an initial console.\n");
1603 static noinline void __init kernel_init_freeable(void)
1605 /* Now the scheduler is fully set up and can do blocking allocations */
1606 gfp_allowed_mask = __GFP_BITS_MASK;
1609 * init can allocate pages on any node
1611 set_mems_allowed(node_states[N_MEMORY]);
1613 cad_pid = get_pid(task_pid(current));
1615 smp_prepare_cpus(setup_max_cpus);
1619 init_mm_internals();
1621 rcu_init_tasks_generic();
1622 do_pre_smp_initcalls();
1623 lockup_detector_init();
1629 page_alloc_init_late();
1630 /* Initialize page ext after all struct pages are initialized. */
1631 if (!early_page_ext_enabled())
1636 kunit_run_all_tests();
1638 wait_for_initramfs();
1639 console_on_rootfs();
1642 * check if there is an early userspace init. If yes, let it do all
1645 if (init_eaccess(ramdisk_execute_command) != 0) {
1646 ramdisk_execute_command = NULL;
1647 prepare_namespace();
1651 * Ok, we have completed the initial bootup, and
1652 * we're essentially up and running. Get rid of the
1653 * initmem segments and start the user-mode stuff..
1655 * rootfs is available now, try loading the public keys
1656 * and default modules
1659 integrity_load_keys();