4 * Copyright (C) 1991, 1992 Linus Torvalds
6 * GK 2/5/95 - Changed to support mounting root fs via NFS
7 * Added initrd & change_root: Werner Almesberger & Hans Lermen, Feb '96
8 * Moan early if gcc is old, avoiding bogus kernels - Paul Gortmaker, May '96
9 * Simplified starting of init: Michael A. Griffith <grif@acm.org>
12 #define DEBUG /* Enable initcall_debug */
14 #include <linux/types.h>
15 #include <linux/extable.h>
16 #include <linux/module.h>
17 #include <linux/proc_fs.h>
18 #include <linux/binfmts.h>
19 #include <linux/kernel.h>
20 #include <linux/syscalls.h>
21 #include <linux/stackprotector.h>
22 #include <linux/string.h>
23 #include <linux/ctype.h>
24 #include <linux/delay.h>
25 #include <linux/ioport.h>
26 #include <linux/init.h>
27 #include <linux/initrd.h>
28 #include <linux/memblock.h>
29 #include <linux/acpi.h>
30 #include <linux/console.h>
31 #include <linux/nmi.h>
32 #include <linux/percpu.h>
33 #include <linux/kmod.h>
34 #include <linux/vmalloc.h>
35 #include <linux/kernel_stat.h>
36 #include <linux/start_kernel.h>
37 #include <linux/security.h>
38 #include <linux/smp.h>
39 #include <linux/profile.h>
40 #include <linux/rcupdate.h>
41 #include <linux/moduleparam.h>
42 #include <linux/kallsyms.h>
43 #include <linux/writeback.h>
44 #include <linux/cpu.h>
45 #include <linux/cpuset.h>
46 #include <linux/cgroup.h>
47 #include <linux/efi.h>
48 #include <linux/tick.h>
49 #include <linux/sched/isolation.h>
50 #include <linux/interrupt.h>
51 #include <linux/taskstats_kern.h>
52 #include <linux/delayacct.h>
53 #include <linux/unistd.h>
54 #include <linux/utsname.h>
55 #include <linux/rmap.h>
56 #include <linux/mempolicy.h>
57 #include <linux/key.h>
58 #include <linux/buffer_head.h>
59 #include <linux/page_ext.h>
60 #include <linux/debug_locks.h>
61 #include <linux/debugobjects.h>
62 #include <linux/lockdep.h>
63 #include <linux/kmemleak.h>
64 #include <linux/pid_namespace.h>
65 #include <linux/device.h>
66 #include <linux/kthread.h>
67 #include <linux/sched.h>
68 #include <linux/sched/init.h>
69 #include <linux/signal.h>
70 #include <linux/idr.h>
71 #include <linux/kgdb.h>
72 #include <linux/ftrace.h>
73 #include <linux/async.h>
74 #include <linux/sfi.h>
75 #include <linux/shmem_fs.h>
76 #include <linux/slab.h>
77 #include <linux/perf_event.h>
78 #include <linux/ptrace.h>
79 #include <linux/pti.h>
80 #include <linux/blkdev.h>
81 #include <linux/elevator.h>
82 #include <linux/sched/clock.h>
83 #include <linux/sched/task.h>
84 #include <linux/sched/task_stack.h>
85 #include <linux/context_tracking.h>
86 #include <linux/random.h>
87 #include <linux/list.h>
88 #include <linux/integrity.h>
89 #include <linux/proc_ns.h>
91 #include <linux/cache.h>
92 #include <linux/rodata_test.h>
93 #include <linux/jump_label.h>
94 #include <linux/mem_encrypt.h>
98 #include <asm/setup.h>
99 #include <asm/sections.h>
100 #include <asm/cacheflush.h>
102 #define CREATE_TRACE_POINTS
103 #include <trace/events/initcall.h>
105 static int kernel_init(void *);
107 extern void init_IRQ(void);
108 extern void fork_init(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)
378 memblock_alloc(strlen(boot_command_line) + 1, SMP_CACHE_BYTES);
379 initcall_command_line =
380 memblock_alloc(strlen(boot_command_line) + 1, SMP_CACHE_BYTES);
381 static_command_line = memblock_alloc(strlen(command_line) + 1,
383 strcpy(saved_command_line, boot_command_line);
384 strcpy(static_command_line, command_line);
388 * We need to finalize in a non-__init function or else race conditions
389 * between the root thread and the init thread may cause start_kernel to
390 * be reaped by free_initmem before the root thread has proceeded to
393 * gcc-3.4 accidentally inlines this function, so use noinline.
396 static __initdata DECLARE_COMPLETION(kthreadd_done);
398 noinline void __ref rest_init(void)
400 struct task_struct *tsk;
403 rcu_scheduler_starting();
405 * We need to spawn init first so that it obtains pid 1, however
406 * the init task will end up wanting to create kthreads, which, if
407 * we schedule it before we create kthreadd, will OOPS.
409 pid = kernel_thread(kernel_init, NULL, CLONE_FS);
411 * Pin init on the boot CPU. Task migration is not properly working
412 * until sched_init_smp() has been run. It will set the allowed
413 * CPUs for init to the non isolated CPUs.
416 tsk = find_task_by_pid_ns(pid, &init_pid_ns);
417 set_cpus_allowed_ptr(tsk, cpumask_of(smp_processor_id()));
420 numa_default_policy();
421 pid = kernel_thread(kthreadd, NULL, CLONE_FS | CLONE_FILES);
423 kthreadd_task = find_task_by_pid_ns(pid, &init_pid_ns);
427 * Enable might_sleep() and smp_processor_id() checks.
428 * They cannot be enabled earlier because with CONFIG_PREEMPT=y
429 * kernel_thread() would trigger might_sleep() splats. With
430 * CONFIG_PREEMPT_VOLUNTARY=y the init task might have scheduled
431 * already, but it's stuck on the kthreadd_done completion.
433 system_state = SYSTEM_SCHEDULING;
435 complete(&kthreadd_done);
438 * The boot idle thread must execute schedule()
439 * at least once to get things moving:
441 schedule_preempt_disabled();
442 /* Call into cpu_idle with preempt disabled */
443 cpu_startup_entry(CPUHP_ONLINE);
446 /* Check for early params. */
447 static int __init do_early_param(char *param, char *val,
448 const char *unused, void *arg)
450 const struct obs_kernel_param *p;
452 for (p = __setup_start; p < __setup_end; p++) {
453 if ((p->early && parameq(param, p->str)) ||
454 (strcmp(param, "console") == 0 &&
455 strcmp(p->str, "earlycon") == 0)
457 if (p->setup_func(val) != 0)
458 pr_warn("Malformed early option '%s'\n", param);
461 /* We accept everything at this stage. */
465 void __init parse_early_options(char *cmdline)
467 parse_args("early options", cmdline, NULL, 0, 0, 0, NULL,
471 /* Arch code calls this early on, or if not, just before other parsing. */
472 void __init parse_early_param(void)
474 static int done __initdata;
475 static char tmp_cmdline[COMMAND_LINE_SIZE] __initdata;
480 /* All fall through to do_early_param. */
481 strlcpy(tmp_cmdline, boot_command_line, COMMAND_LINE_SIZE);
482 parse_early_options(tmp_cmdline);
486 void __init __weak arch_post_acpi_subsys_init(void) { }
488 void __init __weak smp_setup_processor_id(void)
492 # if THREAD_SIZE >= PAGE_SIZE
493 void __init __weak thread_stack_cache_init(void)
498 void __init __weak mem_encrypt_init(void) { }
501 core_param(initcall_debug, initcall_debug, bool, 0644);
503 #ifdef TRACEPOINTS_ENABLED
504 static void __init initcall_debug_enable(void);
506 static inline void initcall_debug_enable(void)
512 * Set up kernel memory allocators
514 static void __init mm_init(void)
517 * page_ext requires contiguous pages,
518 * bigger than MAX_ORDER unless SPARSEMEM.
520 page_ext_init_flatmem();
526 /* Should be run before the first non-init thread is created */
528 /* Should be run after espfix64 is set up. */
532 void __init __weak arch_call_rest_init(void)
537 asmlinkage __visible void __init start_kernel(void)
542 set_task_stack_end_magic(&init_task);
543 smp_setup_processor_id();
544 debug_objects_early_init();
549 early_boot_irqs_disabled = true;
552 * Interrupts are still disabled. Do necessary setups, then
557 pr_notice("%s", linux_banner);
558 setup_arch(&command_line);
560 * Set up the the initial canary and entropy after arch
561 * and after adding latent and command line entropy.
563 add_latent_entropy();
564 add_device_randomness(command_line, strlen(command_line));
565 boot_init_stack_canary();
566 mm_init_cpumask(&init_mm);
567 setup_command_line(command_line);
569 setup_per_cpu_areas();
570 smp_prepare_boot_cpu(); /* arch-specific boot-cpu hooks */
571 boot_cpu_hotplug_init();
573 build_all_zonelists(NULL);
576 pr_notice("Kernel command line: %s\n", boot_command_line);
578 after_dashes = parse_args("Booting kernel",
579 static_command_line, __start___param,
580 __stop___param - __start___param,
581 -1, -1, NULL, &unknown_bootoption);
582 if (!IS_ERR_OR_NULL(after_dashes))
583 parse_args("Setting init args", after_dashes, NULL, 0, -1, -1,
589 * These use large bootmem allocations and must precede
593 vfs_caches_init_early();
600 /* trace_printk can be enabled here */
604 * Set up the scheduler prior starting any interrupts (such as the
605 * timer interrupt). Full topology setup happens at smp_init()
606 * time - but meanwhile we still have a functioning scheduler.
610 * Disable preemption - early bootup scheduling is extremely
611 * fragile until we cpu_idle() for the first time.
614 if (WARN(!irqs_disabled(),
615 "Interrupts were enabled *very* early, fixing it\n"))
620 * Set up housekeeping before setting up workqueues to allow the unbound
621 * workqueue to take non-housekeeping into account.
626 * Allow workqueue creation and work item queueing/cancelling
627 * early. Work item execution depends on kthreads and starts after
630 workqueue_init_early();
634 /* Trace events are available after this */
638 initcall_debug_enable();
640 context_tracking_init();
641 /* init some links before init_ISA_irqs() */
654 call_function_init();
655 WARN(!irqs_disabled(), "Interrupts were enabled early\n");
657 early_boot_irqs_disabled = false;
660 kmem_cache_init_late();
663 * HACK ALERT! This is early. We're enabling the console before
664 * we've done PCI setups etc, and console_init() must be aware of
665 * this. But we do want output early, in case something goes wrong.
669 panic("Too many boot %s vars at `%s'", panic_later,
675 * Need to run this when irqs are enabled, because it wants
676 * to self-test [hard/soft]-irqs on/off lock inversion bugs
682 * This needs to be called before any devices perform DMA
683 * operations that might use the SWIOTLB bounce buffers. It will
684 * mark the bounce buffers as decrypted so that their usage will
685 * not cause "plain-text" data to be decrypted when accessed.
689 #ifdef CONFIG_BLK_DEV_INITRD
690 if (initrd_start && !initrd_below_start_ok &&
691 page_to_pfn(virt_to_page((void *)initrd_start)) < min_low_pfn) {
692 pr_crit("initrd overwritten (0x%08lx < 0x%08lx) - disabling it.\n",
693 page_to_pfn(virt_to_page((void *)initrd_start)),
700 debug_objects_mem_init();
701 setup_per_cpu_pageset();
711 if (efi_enabled(EFI_RUNTIME_SERVICES))
712 efi_enter_virtual_mode();
714 thread_stack_cache_init();
731 taskstats_init_early();
736 acpi_subsystem_init();
737 arch_post_acpi_subsys_init();
740 if (efi_enabled(EFI_RUNTIME_SERVICES)) {
741 efi_free_boot_services();
744 /* Do the rest non-__init'ed, we're now alive */
745 arch_call_rest_init();
748 /* Call all constructor functions linked into the kernel. */
749 static void __init do_ctors(void)
751 #ifdef CONFIG_CONSTRUCTORS
752 ctor_fn_t *fn = (ctor_fn_t *) __ctors_start;
754 for (; fn < (ctor_fn_t *) __ctors_end; fn++)
759 #ifdef CONFIG_KALLSYMS
760 struct blacklist_entry {
761 struct list_head next;
765 static __initdata_or_module LIST_HEAD(blacklisted_initcalls);
767 static int __init initcall_blacklist(char *str)
770 struct blacklist_entry *entry;
772 /* str argument is a comma-separated list of functions */
774 str_entry = strsep(&str, ",");
776 pr_debug("blacklisting initcall %s\n", str_entry);
777 entry = memblock_alloc(sizeof(*entry),
779 entry->buf = memblock_alloc(strlen(str_entry) + 1,
781 strcpy(entry->buf, str_entry);
782 list_add(&entry->next, &blacklisted_initcalls);
789 static bool __init_or_module initcall_blacklisted(initcall_t fn)
791 struct blacklist_entry *entry;
792 char fn_name[KSYM_SYMBOL_LEN];
795 if (list_empty(&blacklisted_initcalls))
798 addr = (unsigned long) dereference_function_descriptor(fn);
799 sprint_symbol_no_offset(fn_name, addr);
802 * fn will be "function_name [module_name]" where [module_name] is not
803 * displayed for built-in init functions. Strip off the [module_name].
805 strreplace(fn_name, ' ', '\0');
807 list_for_each_entry(entry, &blacklisted_initcalls, next) {
808 if (!strcmp(fn_name, entry->buf)) {
809 pr_debug("initcall %s blacklisted\n", fn_name);
817 static int __init initcall_blacklist(char *str)
819 pr_warn("initcall_blacklist requires CONFIG_KALLSYMS\n");
823 static bool __init_or_module initcall_blacklisted(initcall_t fn)
828 __setup("initcall_blacklist=", initcall_blacklist);
830 static __init_or_module void
831 trace_initcall_start_cb(void *data, initcall_t fn)
833 ktime_t *calltime = (ktime_t *)data;
835 printk(KERN_DEBUG "calling %pF @ %i\n", fn, task_pid_nr(current));
836 *calltime = ktime_get();
839 static __init_or_module void
840 trace_initcall_finish_cb(void *data, initcall_t fn, int ret)
842 ktime_t *calltime = (ktime_t *)data;
843 ktime_t delta, rettime;
844 unsigned long long duration;
846 rettime = ktime_get();
847 delta = ktime_sub(rettime, *calltime);
848 duration = (unsigned long long) ktime_to_ns(delta) >> 10;
849 printk(KERN_DEBUG "initcall %pF returned %d after %lld usecs\n",
853 static ktime_t initcall_calltime;
855 #ifdef TRACEPOINTS_ENABLED
856 static void __init initcall_debug_enable(void)
860 ret = register_trace_initcall_start(trace_initcall_start_cb,
862 ret |= register_trace_initcall_finish(trace_initcall_finish_cb,
864 WARN(ret, "Failed to register initcall tracepoints\n");
866 # define do_trace_initcall_start trace_initcall_start
867 # define do_trace_initcall_finish trace_initcall_finish
869 static inline void do_trace_initcall_start(initcall_t fn)
873 trace_initcall_start_cb(&initcall_calltime, fn);
875 static inline void do_trace_initcall_finish(initcall_t fn, int ret)
879 trace_initcall_finish_cb(&initcall_calltime, fn, ret);
881 #endif /* !TRACEPOINTS_ENABLED */
883 int __init_or_module do_one_initcall(initcall_t fn)
885 int count = preempt_count();
889 if (initcall_blacklisted(fn))
892 do_trace_initcall_start(fn);
894 do_trace_initcall_finish(fn, ret);
898 if (preempt_count() != count) {
899 sprintf(msgbuf, "preemption imbalance ");
900 preempt_count_set(count);
902 if (irqs_disabled()) {
903 strlcat(msgbuf, "disabled interrupts ", sizeof(msgbuf));
906 WARN(msgbuf[0], "initcall %pF returned with %s\n", fn, msgbuf);
908 add_latent_entropy();
913 extern initcall_entry_t __initcall_start[];
914 extern initcall_entry_t __initcall0_start[];
915 extern initcall_entry_t __initcall1_start[];
916 extern initcall_entry_t __initcall2_start[];
917 extern initcall_entry_t __initcall3_start[];
918 extern initcall_entry_t __initcall4_start[];
919 extern initcall_entry_t __initcall5_start[];
920 extern initcall_entry_t __initcall6_start[];
921 extern initcall_entry_t __initcall7_start[];
922 extern initcall_entry_t __initcall_end[];
924 static initcall_entry_t *initcall_levels[] __initdata = {
936 /* Keep these in sync with initcalls in include/linux/init.h */
937 static char *initcall_level_names[] __initdata = {
948 static void __init do_initcall_level(int level)
950 initcall_entry_t *fn;
952 strcpy(initcall_command_line, saved_command_line);
953 parse_args(initcall_level_names[level],
954 initcall_command_line, __start___param,
955 __stop___param - __start___param,
957 NULL, &repair_env_string);
959 trace_initcall_level(initcall_level_names[level]);
960 for (fn = initcall_levels[level]; fn < initcall_levels[level+1]; fn++)
961 do_one_initcall(initcall_from_entry(fn));
964 static void __init do_initcalls(void)
968 for (level = 0; level < ARRAY_SIZE(initcall_levels) - 1; level++)
969 do_initcall_level(level);
973 * Ok, the machine is now initialized. None of the devices
974 * have been touched yet, but the CPU subsystem is up and
975 * running, and memory and process management works.
977 * Now we can finally start doing some real work..
979 static void __init do_basic_setup(void)
986 usermodehelper_enable();
990 static void __init do_pre_smp_initcalls(void)
992 initcall_entry_t *fn;
994 trace_initcall_level("early");
995 for (fn = __initcall_start; fn < __initcall0_start; fn++)
996 do_one_initcall(initcall_from_entry(fn));
1000 * This function requests modules which should be loaded by default and is
1001 * called twice right after initrd is mounted and right before init is
1002 * exec'd. If such modules are on either initrd or rootfs, they will be
1003 * loaded before control is passed to userland.
1005 void __init load_default_modules(void)
1007 load_default_elevator_module();
1010 static int run_init_process(const char *init_filename)
1012 argv_init[0] = init_filename;
1013 pr_info("Run %s as init process\n", init_filename);
1014 return do_execve(getname_kernel(init_filename),
1015 (const char __user *const __user *)argv_init,
1016 (const char __user *const __user *)envp_init);
1019 static int try_to_run_init_process(const char *init_filename)
1023 ret = run_init_process(init_filename);
1025 if (ret && ret != -ENOENT) {
1026 pr_err("Starting init: %s exists but couldn't execute it (error %d)\n",
1027 init_filename, ret);
1033 static noinline void __init kernel_init_freeable(void);
1035 #if defined(CONFIG_STRICT_KERNEL_RWX) || defined(CONFIG_STRICT_MODULE_RWX)
1036 bool rodata_enabled __ro_after_init = true;
1037 static int __init set_debug_rodata(char *str)
1039 return strtobool(str, &rodata_enabled);
1041 __setup("rodata=", set_debug_rodata);
1044 #ifdef CONFIG_STRICT_KERNEL_RWX
1045 static void mark_readonly(void)
1047 if (rodata_enabled) {
1049 * load_module() results in W+X mappings, which are cleaned up
1050 * with call_rcu_sched(). Let's make sure that queued work is
1051 * flushed so that we don't hit false positives looking for
1052 * insecure pages which are W+X.
1054 rcu_barrier_sched();
1058 pr_info("Kernel memory protection disabled.\n");
1061 static inline void mark_readonly(void)
1063 pr_warn("This architecture does not have kernel memory protection.\n");
1067 static int __ref kernel_init(void *unused)
1071 kernel_init_freeable();
1072 /* need to finish all async __init code before freeing the memory */
1073 async_synchronize_full();
1074 ftrace_free_init_mem();
1079 * Kernel mappings are now finalized - update the userspace page-table
1084 system_state = SYSTEM_RUNNING;
1085 numa_default_policy();
1087 rcu_end_inkernel_boot();
1089 if (ramdisk_execute_command) {
1090 ret = run_init_process(ramdisk_execute_command);
1093 pr_err("Failed to execute %s (error %d)\n",
1094 ramdisk_execute_command, ret);
1098 * We try each of these until one succeeds.
1100 * The Bourne shell can be used instead of init if we are
1101 * trying to recover a really broken machine.
1103 if (execute_command) {
1104 ret = run_init_process(execute_command);
1107 panic("Requested init %s failed (error %d).",
1108 execute_command, ret);
1110 if (!try_to_run_init_process("/sbin/init") ||
1111 !try_to_run_init_process("/etc/init") ||
1112 !try_to_run_init_process("/bin/init") ||
1113 !try_to_run_init_process("/bin/sh"))
1116 panic("No working init found. Try passing init= option to kernel. "
1117 "See Linux Documentation/admin-guide/init.rst for guidance.");
1120 static noinline void __init kernel_init_freeable(void)
1123 * Wait until kthreadd is all set-up.
1125 wait_for_completion(&kthreadd_done);
1127 /* Now the scheduler is fully set up and can do blocking allocations */
1128 gfp_allowed_mask = __GFP_BITS_MASK;
1131 * init can allocate pages on any node
1133 set_mems_allowed(node_states[N_MEMORY]);
1135 cad_pid = task_pid(current);
1137 smp_prepare_cpus(setup_max_cpus);
1141 init_mm_internals();
1143 do_pre_smp_initcalls();
1144 lockup_detector_init();
1149 page_alloc_init_late();
1153 /* Open the /dev/console on the rootfs, this should never fail */
1154 if (ksys_open((const char __user *) "/dev/console", O_RDWR, 0) < 0)
1155 pr_err("Warning: unable to open an initial console.\n");
1160 * check if there is an early userspace init. If yes, let it do all
1164 if (!ramdisk_execute_command)
1165 ramdisk_execute_command = "/init";
1167 if (ksys_access((const char __user *)
1168 ramdisk_execute_command, 0) != 0) {
1169 ramdisk_execute_command = NULL;
1170 prepare_namespace();
1174 * Ok, we have completed the initial bootup, and
1175 * we're essentially up and running. Get rid of the
1176 * initmem segments and start the user-mode stuff..
1178 * rootfs is available now, try loading the public keys
1179 * and default modules
1182 integrity_load_keys();
1183 load_default_modules();