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/console.h>
32 #include <linux/nmi.h>
33 #include <linux/percpu.h>
34 #include <linux/kmod.h>
35 #include <linux/vmalloc.h>
36 #include <linux/kernel_stat.h>
37 #include <linux/start_kernel.h>
38 #include <linux/security.h>
39 #include <linux/smp.h>
40 #include <linux/profile.h>
41 #include <linux/rcupdate.h>
42 #include <linux/moduleparam.h>
43 #include <linux/kallsyms.h>
44 #include <linux/writeback.h>
45 #include <linux/cpu.h>
46 #include <linux/cpuset.h>
47 #include <linux/cgroup.h>
48 #include <linux/efi.h>
49 #include <linux/tick.h>
50 #include <linux/sched/isolation.h>
51 #include <linux/interrupt.h>
52 #include <linux/taskstats_kern.h>
53 #include <linux/delayacct.h>
54 #include <linux/unistd.h>
55 #include <linux/utsname.h>
56 #include <linux/rmap.h>
57 #include <linux/mempolicy.h>
58 #include <linux/key.h>
59 #include <linux/buffer_head.h>
60 #include <linux/page_ext.h>
61 #include <linux/debug_locks.h>
62 #include <linux/debugobjects.h>
63 #include <linux/lockdep.h>
64 #include <linux/kmemleak.h>
65 #include <linux/pid_namespace.h>
66 #include <linux/device.h>
67 #include <linux/kthread.h>
68 #include <linux/sched.h>
69 #include <linux/sched/init.h>
70 #include <linux/signal.h>
71 #include <linux/idr.h>
72 #include <linux/kgdb.h>
73 #include <linux/ftrace.h>
74 #include <linux/async.h>
75 #include <linux/sfi.h>
76 #include <linux/shmem_fs.h>
77 #include <linux/slab.h>
78 #include <linux/perf_event.h>
79 #include <linux/ptrace.h>
80 #include <linux/pti.h>
81 #include <linux/blkdev.h>
82 #include <linux/elevator.h>
83 #include <linux/sched/clock.h>
84 #include <linux/sched/task.h>
85 #include <linux/sched/task_stack.h>
86 #include <linux/context_tracking.h>
87 #include <linux/random.h>
88 #include <linux/list.h>
89 #include <linux/integrity.h>
90 #include <linux/proc_ns.h>
92 #include <linux/cache.h>
93 #include <linux/rodata_test.h>
94 #include <linux/jump_label.h>
95 #include <linux/mem_encrypt.h>
99 #include <asm/setup.h>
100 #include <asm/sections.h>
101 #include <asm/cacheflush.h>
103 #define CREATE_TRACE_POINTS
104 #include <trace/events/initcall.h>
106 static int kernel_init(void *);
108 extern void init_IRQ(void);
109 extern void radix_tree_init(void);
112 * Debug helper: via this flag we know that we are in 'early bootup code'
113 * where only the boot processor is running with IRQ disabled. This means
114 * two things - IRQ must not be enabled before the flag is cleared and some
115 * operations which are not allowed with IRQ disabled are allowed while the
118 bool early_boot_irqs_disabled __read_mostly;
120 enum system_states system_state __read_mostly;
121 EXPORT_SYMBOL(system_state);
124 * Boot command-line arguments
126 #define MAX_INIT_ARGS CONFIG_INIT_ENV_ARG_LIMIT
127 #define MAX_INIT_ENVS CONFIG_INIT_ENV_ARG_LIMIT
129 extern void time_init(void);
130 /* Default late time init is NULL. archs can override this later. */
131 void (*__initdata late_time_init)(void);
133 /* Untouched command line saved by arch-specific code. */
134 char __initdata boot_command_line[COMMAND_LINE_SIZE];
135 /* Untouched saved command line (eg. for /proc) */
136 char *saved_command_line;
137 /* Command line for parameter parsing */
138 static char *static_command_line;
139 /* Command line for per-initcall parameter parsing */
140 static char *initcall_command_line;
142 static char *execute_command;
143 static char *ramdisk_execute_command;
146 * Used to generate warnings if static_key manipulation functions are used
147 * before jump_label_init is called.
149 bool static_key_initialized __read_mostly;
150 EXPORT_SYMBOL_GPL(static_key_initialized);
153 * If set, this is an indication to the drivers that reset the underlying
154 * device before going ahead with the initialization otherwise driver might
155 * rely on the BIOS and skip the reset operation.
157 * This is useful if kernel is booting in an unreliable environment.
158 * For ex. kdump situation where previous kernel has crashed, BIOS has been
159 * skipped and devices will be in unknown state.
161 unsigned int reset_devices;
162 EXPORT_SYMBOL(reset_devices);
164 static int __init set_reset_devices(char *str)
170 __setup("reset_devices", set_reset_devices);
172 static const char *argv_init[MAX_INIT_ARGS+2] = { "init", NULL, };
173 const char *envp_init[MAX_INIT_ENVS+2] = { "HOME=/", "TERM=linux", NULL, };
174 static const char *panic_later, *panic_param;
176 extern const struct obs_kernel_param __setup_start[], __setup_end[];
178 static bool __init obsolete_checksetup(char *line)
180 const struct obs_kernel_param *p;
181 bool had_early_param = false;
185 int n = strlen(p->str);
186 if (parameqn(line, p->str, n)) {
188 /* Already done in parse_early_param?
189 * (Needs exact match on param part).
190 * Keep iterating, as we can have early
191 * params and __setups of same names 8( */
192 if (line[n] == '\0' || line[n] == '=')
193 had_early_param = true;
194 } else if (!p->setup_func) {
195 pr_warn("Parameter %s is obsolete, ignored\n",
198 } else if (p->setup_func(line + n))
202 } while (p < __setup_end);
204 return had_early_param;
208 * This should be approx 2 Bo*oMips to start (note initial shift), and will
209 * still work even if initially too large, it will just take slightly longer
211 unsigned long loops_per_jiffy = (1<<12);
212 EXPORT_SYMBOL(loops_per_jiffy);
214 static int __init debug_kernel(char *str)
216 console_loglevel = CONSOLE_LOGLEVEL_DEBUG;
220 static int __init quiet_kernel(char *str)
222 console_loglevel = CONSOLE_LOGLEVEL_QUIET;
226 early_param("debug", debug_kernel);
227 early_param("quiet", quiet_kernel);
229 static int __init loglevel(char *str)
234 * Only update loglevel value when a correct setting was passed,
235 * to prevent blind crashes (when loglevel being set to 0) that
236 * are quite hard to debug
238 if (get_option(&str, &newlevel)) {
239 console_loglevel = newlevel;
246 early_param("loglevel", loglevel);
248 /* Change NUL term back to "=", to make "param" the whole string. */
249 static int __init repair_env_string(char *param, char *val,
250 const char *unused, void *arg)
253 /* param=val or param="val"? */
254 if (val == param+strlen(param)+1)
256 else if (val == param+strlen(param)+2) {
258 memmove(val-1, val, strlen(val)+1);
266 /* Anything after -- gets handed straight to init. */
267 static int __init set_init_arg(char *param, char *val,
268 const char *unused, void *arg)
275 repair_env_string(param, val, unused, NULL);
277 for (i = 0; argv_init[i]; i++) {
278 if (i == MAX_INIT_ARGS) {
279 panic_later = "init";
284 argv_init[i] = param;
289 * Unknown boot options get handed to init, unless they look like
290 * unused parameters (modprobe will find them in /proc/cmdline).
292 static int __init unknown_bootoption(char *param, char *val,
293 const char *unused, void *arg)
295 repair_env_string(param, val, unused, NULL);
297 /* Handle obsolete-style parameters */
298 if (obsolete_checksetup(param))
301 /* Unused module parameter. */
302 if (strchr(param, '.') && (!val || strchr(param, '.') < val))
309 /* Environment option */
311 for (i = 0; envp_init[i]; i++) {
312 if (i == MAX_INIT_ENVS) {
316 if (!strncmp(param, envp_init[i], val - param))
319 envp_init[i] = param;
321 /* Command line option */
323 for (i = 0; argv_init[i]; i++) {
324 if (i == MAX_INIT_ARGS) {
325 panic_later = "init";
329 argv_init[i] = param;
334 static int __init init_setup(char *str)
338 execute_command = str;
340 * In case LILO is going to boot us with default command line,
341 * it prepends "auto" before the whole cmdline which makes
342 * the shell think it should execute a script with such name.
343 * So we ignore all arguments entered _before_ init=... [MJ]
345 for (i = 1; i < MAX_INIT_ARGS; i++)
349 __setup("init=", init_setup);
351 static int __init rdinit_setup(char *str)
355 ramdisk_execute_command = str;
356 /* See "auto" comment in init_setup */
357 for (i = 1; i < MAX_INIT_ARGS; i++)
361 __setup("rdinit=", rdinit_setup);
364 static const unsigned int setup_max_cpus = NR_CPUS;
365 static inline void setup_nr_cpu_ids(void) { }
366 static inline void smp_prepare_cpus(unsigned int maxcpus) { }
370 * We need to store the untouched command line for future reference.
371 * We also need to store the touched command line since the parameter
372 * parsing is performed in place, and we should allow a component to
373 * store reference of name/value for future reference.
375 static void __init setup_command_line(char *command_line)
377 size_t len = strlen(boot_command_line) + 1;
379 saved_command_line = memblock_alloc(len, SMP_CACHE_BYTES);
380 if (!saved_command_line)
381 panic("%s: Failed to allocate %zu bytes\n", __func__, len);
383 initcall_command_line = memblock_alloc(len, SMP_CACHE_BYTES);
384 if (!initcall_command_line)
385 panic("%s: Failed to allocate %zu bytes\n", __func__, len);
387 static_command_line = memblock_alloc(len, SMP_CACHE_BYTES);
388 if (!static_command_line)
389 panic("%s: Failed to allocate %zu bytes\n", __func__, len);
391 strcpy(saved_command_line, boot_command_line);
392 strcpy(static_command_line, command_line);
396 * We need to finalize in a non-__init function or else race conditions
397 * between the root thread and the init thread may cause start_kernel to
398 * be reaped by free_initmem before the root thread has proceeded to
401 * gcc-3.4 accidentally inlines this function, so use noinline.
404 static __initdata DECLARE_COMPLETION(kthreadd_done);
406 noinline void __ref rest_init(void)
408 struct task_struct *tsk;
411 rcu_scheduler_starting();
413 * We need to spawn init first so that it obtains pid 1, however
414 * the init task will end up wanting to create kthreads, which, if
415 * we schedule it before we create kthreadd, will OOPS.
417 pid = kernel_thread(kernel_init, NULL, CLONE_FS);
419 * Pin init on the boot CPU. Task migration is not properly working
420 * until sched_init_smp() has been run. It will set the allowed
421 * CPUs for init to the non isolated CPUs.
424 tsk = find_task_by_pid_ns(pid, &init_pid_ns);
425 set_cpus_allowed_ptr(tsk, cpumask_of(smp_processor_id()));
428 numa_default_policy();
429 pid = kernel_thread(kthreadd, NULL, CLONE_FS | CLONE_FILES);
431 kthreadd_task = find_task_by_pid_ns(pid, &init_pid_ns);
435 * Enable might_sleep() and smp_processor_id() checks.
436 * They cannot be enabled earlier because with CONFIG_PREEMPT=y
437 * kernel_thread() would trigger might_sleep() splats. With
438 * CONFIG_PREEMPT_VOLUNTARY=y the init task might have scheduled
439 * already, but it's stuck on the kthreadd_done completion.
441 system_state = SYSTEM_SCHEDULING;
443 complete(&kthreadd_done);
446 * The boot idle thread must execute schedule()
447 * at least once to get things moving:
449 schedule_preempt_disabled();
450 /* Call into cpu_idle with preempt disabled */
451 cpu_startup_entry(CPUHP_ONLINE);
454 /* Check for early params. */
455 static int __init do_early_param(char *param, char *val,
456 const char *unused, void *arg)
458 const struct obs_kernel_param *p;
460 for (p = __setup_start; p < __setup_end; p++) {
461 if ((p->early && parameq(param, p->str)) ||
462 (strcmp(param, "console") == 0 &&
463 strcmp(p->str, "earlycon") == 0)
465 if (p->setup_func(val) != 0)
466 pr_warn("Malformed early option '%s'\n", param);
469 /* We accept everything at this stage. */
473 void __init parse_early_options(char *cmdline)
475 parse_args("early options", cmdline, NULL, 0, 0, 0, NULL,
479 /* Arch code calls this early on, or if not, just before other parsing. */
480 void __init parse_early_param(void)
482 static int done __initdata;
483 static char tmp_cmdline[COMMAND_LINE_SIZE] __initdata;
488 /* All fall through to do_early_param. */
489 strlcpy(tmp_cmdline, boot_command_line, COMMAND_LINE_SIZE);
490 parse_early_options(tmp_cmdline);
494 void __init __weak arch_post_acpi_subsys_init(void) { }
496 void __init __weak smp_setup_processor_id(void)
500 # if THREAD_SIZE >= PAGE_SIZE
501 void __init __weak thread_stack_cache_init(void)
506 void __init __weak mem_encrypt_init(void) { }
508 void __init __weak poking_init(void) { }
510 void __init __weak pgd_cache_init(void) { }
513 core_param(initcall_debug, initcall_debug, bool, 0644);
515 #ifdef TRACEPOINTS_ENABLED
516 static void __init initcall_debug_enable(void);
518 static inline void initcall_debug_enable(void)
524 * Set up kernel memory allocators
526 static void __init mm_init(void)
529 * page_ext requires contiguous pages,
530 * bigger than MAX_ORDER unless SPARSEMEM.
532 page_ext_init_flatmem();
536 debug_objects_mem_init();
539 /* Should be run before the first non-init thread is created */
541 /* Should be run after espfix64 is set up. */
546 void __init __weak arch_call_rest_init(void)
551 asmlinkage __visible void __init start_kernel(void)
556 set_task_stack_end_magic(&init_task);
557 smp_setup_processor_id();
558 debug_objects_early_init();
563 early_boot_irqs_disabled = true;
566 * Interrupts are still disabled. Do necessary setups, then
571 pr_notice("%s", linux_banner);
572 setup_arch(&command_line);
573 mm_init_cpumask(&init_mm);
574 setup_command_line(command_line);
576 setup_per_cpu_areas();
577 smp_prepare_boot_cpu(); /* arch-specific boot-cpu hooks */
578 boot_cpu_hotplug_init();
580 build_all_zonelists(NULL);
583 pr_notice("Kernel command line: %s\n", boot_command_line);
584 /* parameters may set static keys */
587 after_dashes = parse_args("Booting kernel",
588 static_command_line, __start___param,
589 __stop___param - __start___param,
590 -1, -1, NULL, &unknown_bootoption);
591 if (!IS_ERR_OR_NULL(after_dashes))
592 parse_args("Setting init args", after_dashes, NULL, 0, -1, -1,
596 * These use large bootmem allocations and must precede
600 vfs_caches_init_early();
607 /* trace_printk can be enabled here */
611 * Set up the scheduler prior starting any interrupts (such as the
612 * timer interrupt). Full topology setup happens at smp_init()
613 * time - but meanwhile we still have a functioning scheduler.
617 * Disable preemption - early bootup scheduling is extremely
618 * fragile until we cpu_idle() for the first time.
621 if (WARN(!irqs_disabled(),
622 "Interrupts were enabled *very* early, fixing it\n"))
627 * Set up housekeeping before setting up workqueues to allow the unbound
628 * workqueue to take non-housekeeping into account.
633 * Allow workqueue creation and work item queueing/cancelling
634 * early. Work item execution depends on kthreads and starts after
637 workqueue_init_early();
641 /* Trace events are available after this */
645 initcall_debug_enable();
647 context_tracking_init();
648 /* init some links before init_ISA_irqs() */
659 * For best initial stack canary entropy, prepare it after:
660 * - setup_arch() for any UEFI RNG entropy and boot cmdline access
661 * - timekeeping_init() for ktime entropy used in rand_initialize()
662 * - rand_initialize() to get any arch-specific entropy like RDRAND
663 * - add_latent_entropy() to get any latent entropy
664 * - adding command line entropy
667 add_latent_entropy();
668 add_device_randomness(command_line, strlen(command_line));
669 boot_init_stack_canary();
675 call_function_init();
676 WARN(!irqs_disabled(), "Interrupts were enabled early\n");
678 early_boot_irqs_disabled = false;
681 kmem_cache_init_late();
684 * HACK ALERT! This is early. We're enabling the console before
685 * we've done PCI setups etc, and console_init() must be aware of
686 * this. But we do want output early, in case something goes wrong.
690 panic("Too many boot %s vars at `%s'", panic_later,
696 * Need to run this when irqs are enabled, because it wants
697 * to self-test [hard/soft]-irqs on/off lock inversion bugs
703 * This needs to be called before any devices perform DMA
704 * operations that might use the SWIOTLB bounce buffers. It will
705 * mark the bounce buffers as decrypted so that their usage will
706 * not cause "plain-text" data to be decrypted when accessed.
710 #ifdef CONFIG_BLK_DEV_INITRD
711 if (initrd_start && !initrd_below_start_ok &&
712 page_to_pfn(virt_to_page((void *)initrd_start)) < min_low_pfn) {
713 pr_crit("initrd overwritten (0x%08lx < 0x%08lx) - disabling it.\n",
714 page_to_pfn(virt_to_page((void *)initrd_start)),
720 setup_per_cpu_pageset();
730 if (efi_enabled(EFI_RUNTIME_SERVICES))
731 efi_enter_virtual_mode();
733 thread_stack_cache_init();
750 taskstats_init_early();
756 acpi_subsystem_init();
757 arch_post_acpi_subsys_init();
760 /* Do the rest non-__init'ed, we're now alive */
761 arch_call_rest_init();
764 /* Call all constructor functions linked into the kernel. */
765 static void __init do_ctors(void)
767 #ifdef CONFIG_CONSTRUCTORS
768 ctor_fn_t *fn = (ctor_fn_t *) __ctors_start;
770 for (; fn < (ctor_fn_t *) __ctors_end; fn++)
775 #ifdef CONFIG_KALLSYMS
776 struct blacklist_entry {
777 struct list_head next;
781 static __initdata_or_module LIST_HEAD(blacklisted_initcalls);
783 static int __init initcall_blacklist(char *str)
786 struct blacklist_entry *entry;
788 /* str argument is a comma-separated list of functions */
790 str_entry = strsep(&str, ",");
792 pr_debug("blacklisting initcall %s\n", str_entry);
793 entry = memblock_alloc(sizeof(*entry),
796 panic("%s: Failed to allocate %zu bytes\n",
797 __func__, sizeof(*entry));
798 entry->buf = memblock_alloc(strlen(str_entry) + 1,
801 panic("%s: Failed to allocate %zu bytes\n",
802 __func__, strlen(str_entry) + 1);
803 strcpy(entry->buf, str_entry);
804 list_add(&entry->next, &blacklisted_initcalls);
811 static bool __init_or_module initcall_blacklisted(initcall_t fn)
813 struct blacklist_entry *entry;
814 char fn_name[KSYM_SYMBOL_LEN];
817 if (list_empty(&blacklisted_initcalls))
820 addr = (unsigned long) dereference_function_descriptor(fn);
821 sprint_symbol_no_offset(fn_name, addr);
824 * fn will be "function_name [module_name]" where [module_name] is not
825 * displayed for built-in init functions. Strip off the [module_name].
827 strreplace(fn_name, ' ', '\0');
829 list_for_each_entry(entry, &blacklisted_initcalls, next) {
830 if (!strcmp(fn_name, entry->buf)) {
831 pr_debug("initcall %s blacklisted\n", fn_name);
839 static int __init initcall_blacklist(char *str)
841 pr_warn("initcall_blacklist requires CONFIG_KALLSYMS\n");
845 static bool __init_or_module initcall_blacklisted(initcall_t fn)
850 __setup("initcall_blacklist=", initcall_blacklist);
852 static __init_or_module void
853 trace_initcall_start_cb(void *data, initcall_t fn)
855 ktime_t *calltime = (ktime_t *)data;
857 printk(KERN_DEBUG "calling %pS @ %i\n", fn, task_pid_nr(current));
858 *calltime = ktime_get();
861 static __init_or_module void
862 trace_initcall_finish_cb(void *data, initcall_t fn, int ret)
864 ktime_t *calltime = (ktime_t *)data;
865 ktime_t delta, rettime;
866 unsigned long long duration;
868 rettime = ktime_get();
869 delta = ktime_sub(rettime, *calltime);
870 duration = (unsigned long long) ktime_to_ns(delta) >> 10;
871 printk(KERN_DEBUG "initcall %pS returned %d after %lld usecs\n",
875 static ktime_t initcall_calltime;
877 #ifdef TRACEPOINTS_ENABLED
878 static void __init initcall_debug_enable(void)
882 ret = register_trace_initcall_start(trace_initcall_start_cb,
884 ret |= register_trace_initcall_finish(trace_initcall_finish_cb,
886 WARN(ret, "Failed to register initcall tracepoints\n");
888 # define do_trace_initcall_start trace_initcall_start
889 # define do_trace_initcall_finish trace_initcall_finish
891 static inline void do_trace_initcall_start(initcall_t fn)
895 trace_initcall_start_cb(&initcall_calltime, fn);
897 static inline void do_trace_initcall_finish(initcall_t fn, int ret)
901 trace_initcall_finish_cb(&initcall_calltime, fn, ret);
903 #endif /* !TRACEPOINTS_ENABLED */
905 int __init_or_module do_one_initcall(initcall_t fn)
907 int count = preempt_count();
911 if (initcall_blacklisted(fn))
914 do_trace_initcall_start(fn);
916 do_trace_initcall_finish(fn, ret);
920 if (preempt_count() != count) {
921 sprintf(msgbuf, "preemption imbalance ");
922 preempt_count_set(count);
924 if (irqs_disabled()) {
925 strlcat(msgbuf, "disabled interrupts ", sizeof(msgbuf));
928 WARN(msgbuf[0], "initcall %pS returned with %s\n", fn, msgbuf);
930 add_latent_entropy();
935 extern initcall_entry_t __initcall_start[];
936 extern initcall_entry_t __initcall0_start[];
937 extern initcall_entry_t __initcall1_start[];
938 extern initcall_entry_t __initcall2_start[];
939 extern initcall_entry_t __initcall3_start[];
940 extern initcall_entry_t __initcall4_start[];
941 extern initcall_entry_t __initcall5_start[];
942 extern initcall_entry_t __initcall6_start[];
943 extern initcall_entry_t __initcall7_start[];
944 extern initcall_entry_t __initcall_end[];
946 static initcall_entry_t *initcall_levels[] __initdata = {
958 /* Keep these in sync with initcalls in include/linux/init.h */
959 static const char *initcall_level_names[] __initdata = {
970 static void __init do_initcall_level(int level)
972 initcall_entry_t *fn;
974 strcpy(initcall_command_line, saved_command_line);
975 parse_args(initcall_level_names[level],
976 initcall_command_line, __start___param,
977 __stop___param - __start___param,
979 NULL, &repair_env_string);
981 trace_initcall_level(initcall_level_names[level]);
982 for (fn = initcall_levels[level]; fn < initcall_levels[level+1]; fn++)
983 do_one_initcall(initcall_from_entry(fn));
986 static void __init do_initcalls(void)
990 for (level = 0; level < ARRAY_SIZE(initcall_levels) - 1; level++)
991 do_initcall_level(level);
995 * Ok, the machine is now initialized. None of the devices
996 * have been touched yet, but the CPU subsystem is up and
997 * running, and memory and process management works.
999 * Now we can finally start doing some real work..
1001 static void __init do_basic_setup(void)
1008 usermodehelper_enable();
1012 static void __init do_pre_smp_initcalls(void)
1014 initcall_entry_t *fn;
1016 trace_initcall_level("early");
1017 for (fn = __initcall_start; fn < __initcall0_start; fn++)
1018 do_one_initcall(initcall_from_entry(fn));
1021 static int run_init_process(const char *init_filename)
1023 argv_init[0] = init_filename;
1024 pr_info("Run %s as init process\n", init_filename);
1025 return do_execve(getname_kernel(init_filename),
1026 (const char __user *const __user *)argv_init,
1027 (const char __user *const __user *)envp_init);
1030 static int try_to_run_init_process(const char *init_filename)
1034 ret = run_init_process(init_filename);
1036 if (ret && ret != -ENOENT) {
1037 pr_err("Starting init: %s exists but couldn't execute it (error %d)\n",
1038 init_filename, ret);
1044 static noinline void __init kernel_init_freeable(void);
1046 #if defined(CONFIG_STRICT_KERNEL_RWX) || defined(CONFIG_STRICT_MODULE_RWX)
1047 bool rodata_enabled __ro_after_init = true;
1048 static int __init set_debug_rodata(char *str)
1050 return strtobool(str, &rodata_enabled);
1052 __setup("rodata=", set_debug_rodata);
1055 #ifdef CONFIG_STRICT_KERNEL_RWX
1056 static void mark_readonly(void)
1058 if (rodata_enabled) {
1060 * load_module() results in W+X mappings, which are cleaned
1061 * up with call_rcu(). Let's make sure that queued work is
1062 * flushed so that we don't hit false positives looking for
1063 * insecure pages which are W+X.
1069 pr_info("Kernel memory protection disabled.\n");
1072 static inline void mark_readonly(void)
1074 pr_warn("This architecture does not have kernel memory protection.\n");
1078 void __weak free_initmem(void)
1080 free_initmem_default(POISON_FREE_INITMEM);
1083 static int __ref kernel_init(void *unused)
1087 kernel_init_freeable();
1088 /* need to finish all async __init code before freeing the memory */
1089 async_synchronize_full();
1090 ftrace_free_init_mem();
1095 * Kernel mappings are now finalized - update the userspace page-table
1100 system_state = SYSTEM_RUNNING;
1101 numa_default_policy();
1103 rcu_end_inkernel_boot();
1105 if (ramdisk_execute_command) {
1106 ret = run_init_process(ramdisk_execute_command);
1109 pr_err("Failed to execute %s (error %d)\n",
1110 ramdisk_execute_command, ret);
1114 * We try each of these until one succeeds.
1116 * The Bourne shell can be used instead of init if we are
1117 * trying to recover a really broken machine.
1119 if (execute_command) {
1120 ret = run_init_process(execute_command);
1123 panic("Requested init %s failed (error %d).",
1124 execute_command, ret);
1126 if (!try_to_run_init_process("/sbin/init") ||
1127 !try_to_run_init_process("/etc/init") ||
1128 !try_to_run_init_process("/bin/init") ||
1129 !try_to_run_init_process("/bin/sh"))
1132 panic("No working init found. Try passing init= option to kernel. "
1133 "See Linux Documentation/admin-guide/init.rst for guidance.");
1136 static noinline void __init kernel_init_freeable(void)
1139 * Wait until kthreadd is all set-up.
1141 wait_for_completion(&kthreadd_done);
1143 /* Now the scheduler is fully set up and can do blocking allocations */
1144 gfp_allowed_mask = __GFP_BITS_MASK;
1147 * init can allocate pages on any node
1149 set_mems_allowed(node_states[N_MEMORY]);
1151 cad_pid = task_pid(current);
1153 smp_prepare_cpus(setup_max_cpus);
1157 init_mm_internals();
1159 do_pre_smp_initcalls();
1160 lockup_detector_init();
1165 page_alloc_init_late();
1166 /* Initialize page ext after all struct pages are initialized. */
1171 /* Open the /dev/console on the rootfs, this should never fail */
1172 if (ksys_open((const char __user *) "/dev/console", O_RDWR, 0) < 0)
1173 pr_err("Warning: unable to open an initial console.\n");
1178 * check if there is an early userspace init. If yes, let it do all
1182 if (!ramdisk_execute_command)
1183 ramdisk_execute_command = "/init";
1185 if (ksys_access((const char __user *)
1186 ramdisk_execute_command, 0) != 0) {
1187 ramdisk_execute_command = NULL;
1188 prepare_namespace();
1192 * Ok, we have completed the initial bootup, and
1193 * we're essentially up and running. Get rid of the
1194 * initmem segments and start the user-mode stuff..
1196 * rootfs is available now, try loading the public keys
1197 * and default modules
1200 integrity_load_keys();