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/debug_locks.h>
66 #include <linux/debugobjects.h>
67 #include <linux/lockdep.h>
68 #include <linux/kmemleak.h>
69 #include <linux/padata.h>
70 #include <linux/pid_namespace.h>
71 #include <linux/device/driver.h>
72 #include <linux/kthread.h>
73 #include <linux/sched.h>
74 #include <linux/sched/init.h>
75 #include <linux/signal.h>
76 #include <linux/idr.h>
77 #include <linux/kgdb.h>
78 #include <linux/ftrace.h>
79 #include <linux/async.h>
80 #include <linux/shmem_fs.h>
81 #include <linux/slab.h>
82 #include <linux/perf_event.h>
83 #include <linux/ptrace.h>
84 #include <linux/pti.h>
85 #include <linux/blkdev.h>
86 #include <linux/sched/clock.h>
87 #include <linux/sched/task.h>
88 #include <linux/sched/task_stack.h>
89 #include <linux/context_tracking.h>
90 #include <linux/random.h>
91 #include <linux/list.h>
92 #include <linux/integrity.h>
93 #include <linux/proc_ns.h>
95 #include <linux/cache.h>
96 #include <linux/rodata_test.h>
97 #include <linux/jump_label.h>
98 #include <linux/kcsan.h>
99 #include <linux/init_syscalls.h>
100 #include <linux/stackdepot.h>
101 #include <linux/randomize_kstack.h>
102 #include <net/net_namespace.h>
105 #include <asm/setup.h>
106 #include <asm/sections.h>
107 #include <asm/cacheflush.h>
109 #define CREATE_TRACE_POINTS
110 #include <trace/events/initcall.h>
112 #include <kunit/test.h>
114 static int kernel_init(void *);
117 * Debug helper: via this flag we know that we are in 'early bootup code'
118 * where only the boot processor is running with IRQ disabled. This means
119 * two things - IRQ must not be enabled before the flag is cleared and some
120 * operations which are not allowed with IRQ disabled are allowed while the
123 bool early_boot_irqs_disabled __read_mostly;
125 enum system_states system_state __read_mostly;
126 EXPORT_SYMBOL(system_state);
129 * Boot command-line arguments
131 #define MAX_INIT_ARGS CONFIG_INIT_ENV_ARG_LIMIT
132 #define MAX_INIT_ENVS CONFIG_INIT_ENV_ARG_LIMIT
134 /* Default late time init is NULL. archs can override this later. */
135 void (*__initdata late_time_init)(void);
137 /* Untouched command line saved by arch-specific code. */
138 char __initdata boot_command_line[COMMAND_LINE_SIZE];
139 /* Untouched saved command line (eg. for /proc) */
140 char *saved_command_line __ro_after_init;
141 unsigned int saved_command_line_len __ro_after_init;
142 /* Command line for parameter parsing */
143 static char *static_command_line;
144 /* Untouched extra command line */
145 static char *extra_command_line;
146 /* Extra init arguments */
147 static char *extra_init_args;
149 #ifdef CONFIG_BOOT_CONFIG
150 /* Is bootconfig on command line? */
151 static bool bootconfig_found;
152 static size_t initargs_offs;
154 # define bootconfig_found false
155 # define initargs_offs 0
158 static char *execute_command;
159 static char *ramdisk_execute_command = "/init";
162 * Used to generate warnings if static_key manipulation functions are used
163 * before jump_label_init is called.
165 bool static_key_initialized __read_mostly;
166 EXPORT_SYMBOL_GPL(static_key_initialized);
169 * If set, this is an indication to the drivers that reset the underlying
170 * device before going ahead with the initialization otherwise driver might
171 * rely on the BIOS and skip the reset operation.
173 * This is useful if kernel is booting in an unreliable environment.
174 * For ex. kdump situation where previous kernel has crashed, BIOS has been
175 * skipped and devices will be in unknown state.
177 unsigned int reset_devices;
178 EXPORT_SYMBOL(reset_devices);
180 static int __init set_reset_devices(char *str)
186 __setup("reset_devices", set_reset_devices);
188 static const char *argv_init[MAX_INIT_ARGS+2] = { "init", NULL, };
189 const char *envp_init[MAX_INIT_ENVS+2] = { "HOME=/", "TERM=linux", NULL, };
190 static const char *panic_later, *panic_param;
192 static bool __init obsolete_checksetup(char *line)
194 const struct obs_kernel_param *p;
195 bool had_early_param = false;
199 int n = strlen(p->str);
200 if (parameqn(line, p->str, n)) {
202 /* Already done in parse_early_param?
203 * (Needs exact match on param part).
204 * Keep iterating, as we can have early
205 * params and __setups of same names 8( */
206 if (line[n] == '\0' || line[n] == '=')
207 had_early_param = true;
208 } else if (!p->setup_func) {
209 pr_warn("Parameter %s is obsolete, ignored\n",
212 } else if (p->setup_func(line + n))
216 } while (p < __setup_end);
218 return had_early_param;
222 * This should be approx 2 Bo*oMips to start (note initial shift), and will
223 * still work even if initially too large, it will just take slightly longer
225 unsigned long loops_per_jiffy = (1<<12);
226 EXPORT_SYMBOL(loops_per_jiffy);
228 static int __init debug_kernel(char *str)
230 console_loglevel = CONSOLE_LOGLEVEL_DEBUG;
234 static int __init quiet_kernel(char *str)
236 console_loglevel = CONSOLE_LOGLEVEL_QUIET;
240 early_param("debug", debug_kernel);
241 early_param("quiet", quiet_kernel);
243 static int __init loglevel(char *str)
248 * Only update loglevel value when a correct setting was passed,
249 * to prevent blind crashes (when loglevel being set to 0) that
250 * are quite hard to debug
252 if (get_option(&str, &newlevel)) {
253 console_loglevel = newlevel;
260 early_param("loglevel", loglevel);
262 #ifdef CONFIG_BLK_DEV_INITRD
263 static void * __init get_boot_config_from_initrd(size_t *_size)
273 data = (char *)initrd_end - BOOTCONFIG_MAGIC_LEN;
275 * Since Grub may align the size of initrd to 4, we must
276 * check the preceding 3 bytes as well.
278 for (i = 0; i < 4; i++) {
279 if (!memcmp(data, BOOTCONFIG_MAGIC, BOOTCONFIG_MAGIC_LEN))
286 hdr = (u32 *)(data - 8);
287 size = le32_to_cpu(hdr[0]);
288 csum = le32_to_cpu(hdr[1]);
290 data = ((void *)hdr) - size;
291 if ((unsigned long)data < initrd_start) {
292 pr_err("bootconfig size %d is greater than initrd size %ld\n",
293 size, initrd_end - initrd_start);
297 if (xbc_calc_checksum(data, size) != csum) {
298 pr_err("bootconfig checksum failed\n");
302 /* Remove bootconfig from initramfs/initrd */
303 initrd_end = (unsigned long)data;
310 static void * __init get_boot_config_from_initrd(size_t *_size)
316 #ifdef CONFIG_BOOT_CONFIG
318 static char xbc_namebuf[XBC_KEYLEN_MAX] __initdata;
320 #define rest(dst, end) ((end) > (dst) ? (end) - (dst) : 0)
322 static int __init xbc_snprint_cmdline(char *buf, size_t size,
323 struct xbc_node *root)
325 struct xbc_node *knode, *vnode;
326 char *end = buf + size;
330 xbc_node_for_each_key_value(root, knode, val) {
331 ret = xbc_node_compose_key_after(root, knode,
332 xbc_namebuf, XBC_KEYLEN_MAX);
336 vnode = xbc_node_get_child(knode);
338 ret = snprintf(buf, rest(buf, end), "%s ", xbc_namebuf);
344 xbc_array_for_each_value(vnode, val) {
345 ret = snprintf(buf, rest(buf, end), "%s=\"%s\" ",
353 return buf - (end - size);
357 /* Make an extra command line under given key word */
358 static char * __init xbc_make_cmdline(const char *key)
360 struct xbc_node *root;
364 root = xbc_find_node(key);
368 /* Count required buffer size */
369 len = xbc_snprint_cmdline(NULL, 0, root);
373 new_cmdline = memblock_alloc(len + 1, SMP_CACHE_BYTES);
375 pr_err("Failed to allocate memory for extra kernel cmdline.\n");
379 ret = xbc_snprint_cmdline(new_cmdline, len + 1, root);
380 if (ret < 0 || ret > len) {
381 pr_err("Failed to print extra kernel cmdline.\n");
382 memblock_free(new_cmdline, len + 1);
389 static int __init bootconfig_params(char *param, char *val,
390 const char *unused, void *arg)
392 if (strcmp(param, "bootconfig") == 0) {
393 bootconfig_found = true;
398 static int __init warn_bootconfig(char *str)
400 /* The 'bootconfig' has been handled by bootconfig_params(). */
404 static void __init setup_boot_config(void)
406 static char tmp_cmdline[COMMAND_LINE_SIZE] __initdata;
407 const char *msg, *data;
412 /* Cut out the bootconfig data even if we have no bootconfig option */
413 data = get_boot_config_from_initrd(&size);
414 /* If there is no bootconfig in initrd, try embedded one. */
416 data = xbc_get_embedded_bootconfig(&size);
418 strscpy(tmp_cmdline, boot_command_line, COMMAND_LINE_SIZE);
419 err = parse_args("bootconfig", tmp_cmdline, NULL, 0, 0, 0, NULL,
422 if (IS_ERR(err) || !(bootconfig_found || IS_ENABLED(CONFIG_BOOT_CONFIG_FORCE)))
425 /* parse_args() stops at the next param of '--' and returns an address */
427 initargs_offs = err - tmp_cmdline;
430 /* If user intended to use bootconfig, show an error level message */
431 if (bootconfig_found)
432 pr_err("'bootconfig' found on command line, but no bootconfig found\n");
434 pr_info("No bootconfig data provided, so skipping bootconfig");
438 if (size >= XBC_DATA_MAX) {
439 pr_err("bootconfig size %ld greater than max size %d\n",
440 (long)size, XBC_DATA_MAX);
444 ret = xbc_init(data, size, &msg, &pos);
447 pr_err("Failed to init bootconfig: %s.\n", msg);
449 pr_err("Failed to parse bootconfig: %s at %d.\n",
452 xbc_get_info(&ret, NULL);
453 pr_info("Load bootconfig: %ld bytes %d nodes\n", (long)size, ret);
454 /* keys starting with "kernel." are passed via cmdline */
455 extra_command_line = xbc_make_cmdline("kernel");
456 /* Also, "init." keys are init arguments */
457 extra_init_args = xbc_make_cmdline("init");
462 static void __init exit_boot_config(void)
467 #else /* !CONFIG_BOOT_CONFIG */
469 static void __init setup_boot_config(void)
471 /* Remove bootconfig data from initrd */
472 get_boot_config_from_initrd(NULL);
475 static int __init warn_bootconfig(char *str)
477 pr_warn("WARNING: 'bootconfig' found on the kernel command line but CONFIG_BOOT_CONFIG is not set.\n");
481 #define exit_boot_config() do {} while (0)
483 #endif /* CONFIG_BOOT_CONFIG */
485 early_param("bootconfig", warn_bootconfig);
487 /* Change NUL term back to "=", to make "param" the whole string. */
488 static void __init repair_env_string(char *param, char *val)
491 /* param=val or param="val"? */
492 if (val == param+strlen(param)+1)
494 else if (val == param+strlen(param)+2) {
496 memmove(val-1, val, strlen(val)+1);
502 /* Anything after -- gets handed straight to init. */
503 static int __init set_init_arg(char *param, char *val,
504 const char *unused, void *arg)
511 repair_env_string(param, val);
513 for (i = 0; argv_init[i]; i++) {
514 if (i == MAX_INIT_ARGS) {
515 panic_later = "init";
520 argv_init[i] = param;
525 * Unknown boot options get handed to init, unless they look like
526 * unused parameters (modprobe will find them in /proc/cmdline).
528 static int __init unknown_bootoption(char *param, char *val,
529 const char *unused, void *arg)
531 size_t len = strlen(param);
533 repair_env_string(param, val);
535 /* Handle obsolete-style parameters */
536 if (obsolete_checksetup(param))
539 /* Unused module parameter. */
540 if (strnchr(param, len, '.'))
547 /* Environment option */
549 for (i = 0; envp_init[i]; i++) {
550 if (i == MAX_INIT_ENVS) {
554 if (!strncmp(param, envp_init[i], len+1))
557 envp_init[i] = param;
559 /* Command line option */
561 for (i = 0; argv_init[i]; i++) {
562 if (i == MAX_INIT_ARGS) {
563 panic_later = "init";
567 argv_init[i] = param;
572 static int __init init_setup(char *str)
576 execute_command = str;
578 * In case LILO is going to boot us with default command line,
579 * it prepends "auto" before the whole cmdline which makes
580 * the shell think it should execute a script with such name.
581 * So we ignore all arguments entered _before_ init=... [MJ]
583 for (i = 1; i < MAX_INIT_ARGS; i++)
587 __setup("init=", init_setup);
589 static int __init rdinit_setup(char *str)
593 ramdisk_execute_command = str;
594 /* See "auto" comment in init_setup */
595 for (i = 1; i < MAX_INIT_ARGS; i++)
599 __setup("rdinit=", rdinit_setup);
602 static const unsigned int setup_max_cpus = NR_CPUS;
603 static inline void setup_nr_cpu_ids(void) { }
604 static inline void smp_prepare_cpus(unsigned int maxcpus) { }
608 * We need to store the untouched command line for future reference.
609 * We also need to store the touched command line since the parameter
610 * parsing is performed in place, and we should allow a component to
611 * store reference of name/value for future reference.
613 static void __init setup_command_line(char *command_line)
615 size_t len, xlen = 0, ilen = 0;
617 if (extra_command_line)
618 xlen = strlen(extra_command_line);
620 ilen = strlen(extra_init_args) + 4; /* for " -- " */
622 len = xlen + strlen(boot_command_line) + 1;
624 saved_command_line = memblock_alloc(len + ilen, SMP_CACHE_BYTES);
625 if (!saved_command_line)
626 panic("%s: Failed to allocate %zu bytes\n", __func__, len + ilen);
628 static_command_line = memblock_alloc(len, SMP_CACHE_BYTES);
629 if (!static_command_line)
630 panic("%s: Failed to allocate %zu bytes\n", __func__, len);
634 * We have to put extra_command_line before boot command
635 * lines because there could be dashes (separator of init
636 * command line) in the command lines.
638 strcpy(saved_command_line, extra_command_line);
639 strcpy(static_command_line, extra_command_line);
641 strcpy(saved_command_line + xlen, boot_command_line);
642 strcpy(static_command_line + xlen, command_line);
646 * Append supplemental init boot args to saved_command_line
647 * so that user can check what command line options passed
649 * The order should always be
650 * " -- "[bootconfig init-param][cmdline init-param]
653 len = xlen + initargs_offs;
654 strcpy(saved_command_line + len, extra_init_args);
655 len += ilen - 4; /* strlen(extra_init_args) */
656 strcpy(saved_command_line + len,
657 boot_command_line + initargs_offs - 1);
659 len = strlen(saved_command_line);
660 strcpy(saved_command_line + len, " -- ");
662 strcpy(saved_command_line + len, extra_init_args);
666 saved_command_line_len = strlen(saved_command_line);
670 * We need to finalize in a non-__init function or else race conditions
671 * between the root thread and the init thread may cause start_kernel to
672 * be reaped by free_initmem before the root thread has proceeded to
675 * gcc-3.4 accidentally inlines this function, so use noinline.
678 static __initdata DECLARE_COMPLETION(kthreadd_done);
680 noinline void __ref __noreturn rest_init(void)
682 struct task_struct *tsk;
685 rcu_scheduler_starting();
687 * We need to spawn init first so that it obtains pid 1, however
688 * the init task will end up wanting to create kthreads, which, if
689 * we schedule it before we create kthreadd, will OOPS.
691 pid = user_mode_thread(kernel_init, NULL, CLONE_FS);
693 * Pin init on the boot CPU. Task migration is not properly working
694 * until sched_init_smp() has been run. It will set the allowed
695 * CPUs for init to the non isolated CPUs.
698 tsk = find_task_by_pid_ns(pid, &init_pid_ns);
699 tsk->flags |= PF_NO_SETAFFINITY;
700 set_cpus_allowed_ptr(tsk, cpumask_of(smp_processor_id()));
703 numa_default_policy();
704 pid = kernel_thread(kthreadd, NULL, NULL, CLONE_FS | CLONE_FILES);
706 kthreadd_task = find_task_by_pid_ns(pid, &init_pid_ns);
710 * Enable might_sleep() and smp_processor_id() checks.
711 * They cannot be enabled earlier because with CONFIG_PREEMPTION=y
712 * kernel_thread() would trigger might_sleep() splats. With
713 * CONFIG_PREEMPT_VOLUNTARY=y the init task might have scheduled
714 * already, but it's stuck on the kthreadd_done completion.
716 system_state = SYSTEM_SCHEDULING;
718 complete(&kthreadd_done);
721 * The boot idle thread must execute schedule()
722 * at least once to get things moving:
724 schedule_preempt_disabled();
725 /* Call into cpu_idle with preempt disabled */
726 cpu_startup_entry(CPUHP_ONLINE);
729 /* Check for early params. */
730 static int __init do_early_param(char *param, char *val,
731 const char *unused, void *arg)
733 const struct obs_kernel_param *p;
735 for (p = __setup_start; p < __setup_end; p++) {
736 if ((p->early && parameq(param, p->str)) ||
737 (strcmp(param, "console") == 0 &&
738 strcmp(p->str, "earlycon") == 0)
740 if (p->setup_func(val) != 0)
741 pr_warn("Malformed early option '%s'\n", param);
744 /* We accept everything at this stage. */
748 void __init parse_early_options(char *cmdline)
750 parse_args("early options", cmdline, NULL, 0, 0, 0, NULL,
754 /* Arch code calls this early on, or if not, just before other parsing. */
755 void __init parse_early_param(void)
757 static int done __initdata;
758 static char tmp_cmdline[COMMAND_LINE_SIZE] __initdata;
763 /* All fall through to do_early_param. */
764 strscpy(tmp_cmdline, boot_command_line, COMMAND_LINE_SIZE);
765 parse_early_options(tmp_cmdline);
769 void __init __weak arch_post_acpi_subsys_init(void) { }
771 void __init __weak smp_setup_processor_id(void)
775 # if THREAD_SIZE >= PAGE_SIZE
776 void __init __weak thread_stack_cache_init(void)
781 void __init __weak poking_init(void) { }
783 void __init __weak pgtable_cache_init(void) { }
785 void __init __weak trap_init(void) { }
788 core_param(initcall_debug, initcall_debug, bool, 0644);
790 #ifdef TRACEPOINTS_ENABLED
791 static void __init initcall_debug_enable(void);
793 static inline void initcall_debug_enable(void)
798 #ifdef CONFIG_RANDOMIZE_KSTACK_OFFSET
799 DEFINE_STATIC_KEY_MAYBE_RO(CONFIG_RANDOMIZE_KSTACK_OFFSET_DEFAULT,
800 randomize_kstack_offset);
801 DEFINE_PER_CPU(u32, kstack_offset);
803 static int __init early_randomize_kstack_offset(char *buf)
808 ret = kstrtobool(buf, &bool_result);
813 static_branch_enable(&randomize_kstack_offset);
815 static_branch_disable(&randomize_kstack_offset);
818 early_param("randomize_kstack_offset", early_randomize_kstack_offset);
821 void __init __weak __noreturn arch_call_rest_init(void)
826 static void __init print_unknown_bootoptions(void)
828 char *unknown_options;
830 const char *const *p;
833 if (panic_later || (!argv_init[1] && !envp_init[2]))
837 * Determine how many options we have to print out, plus a space
840 len = 1; /* null terminator */
841 for (p = &argv_init[1]; *p; p++) {
845 for (p = &envp_init[2]; *p; p++) {
850 unknown_options = memblock_alloc(len, SMP_CACHE_BYTES);
851 if (!unknown_options) {
852 pr_err("%s: Failed to allocate %zu bytes\n",
856 end = unknown_options;
858 for (p = &argv_init[1]; *p; p++)
859 end += sprintf(end, " %s", *p);
860 for (p = &envp_init[2]; *p; p++)
861 end += sprintf(end, " %s", *p);
863 /* Start at unknown_options[1] to skip the initial space */
864 pr_notice("Unknown kernel command line parameters \"%s\", will be passed to user space.\n",
865 &unknown_options[1]);
866 memblock_free(unknown_options, len);
869 asmlinkage __visible __init __no_sanitize_address __noreturn __no_stack_protector
870 void start_kernel(void)
875 set_task_stack_end_magic(&init_task);
876 smp_setup_processor_id();
877 debug_objects_early_init();
878 init_vmlinux_build_id();
883 early_boot_irqs_disabled = true;
886 * Interrupts are still disabled. Do necessary setups, then
891 pr_notice("%s", linux_banner);
892 early_security_init();
893 setup_arch(&command_line);
895 setup_command_line(command_line);
897 setup_per_cpu_areas();
898 smp_prepare_boot_cpu(); /* arch-specific boot-cpu hooks */
899 boot_cpu_hotplug_init();
901 pr_notice("Kernel command line: %s\n", saved_command_line);
902 /* parameters may set static keys */
905 after_dashes = parse_args("Booting kernel",
906 static_command_line, __start___param,
907 __stop___param - __start___param,
908 -1, -1, NULL, &unknown_bootoption);
909 print_unknown_bootoptions();
910 if (!IS_ERR_OR_NULL(after_dashes))
911 parse_args("Setting init args", after_dashes, NULL, 0, -1, -1,
914 parse_args("Setting extra init args", extra_init_args,
915 NULL, 0, -1, -1, NULL, set_init_arg);
917 /* Architectural and non-timekeeping rng init, before allocator init */
918 random_init_early(command_line);
921 * These use large bootmem allocations and must precede
922 * initalization of page allocator
925 vfs_caches_init_early();
932 /* trace_printk can be enabled here */
936 * Set up the scheduler prior starting any interrupts (such as the
937 * timer interrupt). Full topology setup happens at smp_init()
938 * time - but meanwhile we still have a functioning scheduler.
942 if (WARN(!irqs_disabled(),
943 "Interrupts were enabled *very* early, fixing it\n"))
949 * Set up housekeeping before setting up workqueues to allow the unbound
950 * workqueue to take non-housekeeping into account.
955 * Allow workqueue creation and work item queueing/cancelling
956 * early. Work item execution depends on kthreads and starts after
959 workqueue_init_early();
963 /* Trace events are available after this */
967 initcall_debug_enable();
969 context_tracking_init();
970 /* init some links before init_ISA_irqs() */
982 /* This must be after timekeeping is initialized */
985 /* These make use of the fully initialized rng */
987 boot_init_stack_canary();
991 call_function_init();
992 WARN(!irqs_disabled(), "Interrupts were enabled early\n");
994 early_boot_irqs_disabled = false;
997 kmem_cache_init_late();
1000 * HACK ALERT! This is early. We're enabling the console before
1001 * we've done PCI setups etc, and console_init() must be aware of
1002 * this. But we do want output early, in case something goes wrong.
1006 panic("Too many boot %s vars at `%s'", panic_later,
1012 * Need to run this when irqs are enabled, because it wants
1013 * to self-test [hard/soft]-irqs on/off lock inversion bugs
1018 #ifdef CONFIG_BLK_DEV_INITRD
1019 if (initrd_start && !initrd_below_start_ok &&
1020 page_to_pfn(virt_to_page((void *)initrd_start)) < min_low_pfn) {
1021 pr_crit("initrd overwritten (0x%08lx < 0x%08lx) - disabling it.\n",
1022 page_to_pfn(virt_to_page((void *)initrd_start)),
1027 setup_per_cpu_pageset();
1035 arch_cpu_finalize_init();
1040 if (efi_enabled(EFI_RUNTIME_SERVICES))
1041 efi_enter_virtual_mode();
1043 thread_stack_cache_init();
1060 taskstats_init_early();
1063 acpi_subsystem_init();
1064 arch_post_acpi_subsys_init();
1067 /* Do the rest non-__init'ed, we're now alive */
1068 arch_call_rest_init();
1071 * Avoid stack canaries in callers of boot_init_stack_canary for gcc-10
1074 #if !__has_attribute(__no_stack_protector__)
1075 prevent_tail_call_optimization();
1079 /* Call all constructor functions linked into the kernel. */
1080 static void __init do_ctors(void)
1083 * For UML, the constructors have already been called by the
1084 * normal setup code as it's just a normal ELF binary, so we
1085 * cannot do it again - but we do need CONFIG_CONSTRUCTORS
1086 * even on UML for modules.
1088 #if defined(CONFIG_CONSTRUCTORS) && !defined(CONFIG_UML)
1089 ctor_fn_t *fn = (ctor_fn_t *) __ctors_start;
1091 for (; fn < (ctor_fn_t *) __ctors_end; fn++)
1096 #ifdef CONFIG_KALLSYMS
1097 struct blacklist_entry {
1098 struct list_head next;
1102 static __initdata_or_module LIST_HEAD(blacklisted_initcalls);
1104 static int __init initcall_blacklist(char *str)
1107 struct blacklist_entry *entry;
1109 /* str argument is a comma-separated list of functions */
1111 str_entry = strsep(&str, ",");
1113 pr_debug("blacklisting initcall %s\n", str_entry);
1114 entry = memblock_alloc(sizeof(*entry),
1117 panic("%s: Failed to allocate %zu bytes\n",
1118 __func__, sizeof(*entry));
1119 entry->buf = memblock_alloc(strlen(str_entry) + 1,
1122 panic("%s: Failed to allocate %zu bytes\n",
1123 __func__, strlen(str_entry) + 1);
1124 strcpy(entry->buf, str_entry);
1125 list_add(&entry->next, &blacklisted_initcalls);
1127 } while (str_entry);
1132 static bool __init_or_module initcall_blacklisted(initcall_t fn)
1134 struct blacklist_entry *entry;
1135 char fn_name[KSYM_SYMBOL_LEN];
1138 if (list_empty(&blacklisted_initcalls))
1141 addr = (unsigned long) dereference_function_descriptor(fn);
1142 sprint_symbol_no_offset(fn_name, addr);
1145 * fn will be "function_name [module_name]" where [module_name] is not
1146 * displayed for built-in init functions. Strip off the [module_name].
1148 strreplace(fn_name, ' ', '\0');
1150 list_for_each_entry(entry, &blacklisted_initcalls, next) {
1151 if (!strcmp(fn_name, entry->buf)) {
1152 pr_debug("initcall %s blacklisted\n", fn_name);
1160 static int __init initcall_blacklist(char *str)
1162 pr_warn("initcall_blacklist requires CONFIG_KALLSYMS\n");
1166 static bool __init_or_module initcall_blacklisted(initcall_t fn)
1171 __setup("initcall_blacklist=", initcall_blacklist);
1173 static __init_or_module void
1174 trace_initcall_start_cb(void *data, initcall_t fn)
1176 ktime_t *calltime = data;
1178 printk(KERN_DEBUG "calling %pS @ %i\n", fn, task_pid_nr(current));
1179 *calltime = ktime_get();
1182 static __init_or_module void
1183 trace_initcall_finish_cb(void *data, initcall_t fn, int ret)
1185 ktime_t rettime, *calltime = data;
1187 rettime = ktime_get();
1188 printk(KERN_DEBUG "initcall %pS returned %d after %lld usecs\n",
1189 fn, ret, (unsigned long long)ktime_us_delta(rettime, *calltime));
1192 static ktime_t initcall_calltime;
1194 #ifdef TRACEPOINTS_ENABLED
1195 static void __init initcall_debug_enable(void)
1199 ret = register_trace_initcall_start(trace_initcall_start_cb,
1200 &initcall_calltime);
1201 ret |= register_trace_initcall_finish(trace_initcall_finish_cb,
1202 &initcall_calltime);
1203 WARN(ret, "Failed to register initcall tracepoints\n");
1205 # define do_trace_initcall_start trace_initcall_start
1206 # define do_trace_initcall_finish trace_initcall_finish
1208 static inline void do_trace_initcall_start(initcall_t fn)
1210 if (!initcall_debug)
1212 trace_initcall_start_cb(&initcall_calltime, fn);
1214 static inline void do_trace_initcall_finish(initcall_t fn, int ret)
1216 if (!initcall_debug)
1218 trace_initcall_finish_cb(&initcall_calltime, fn, ret);
1220 #endif /* !TRACEPOINTS_ENABLED */
1222 int __init_or_module do_one_initcall(initcall_t fn)
1224 int count = preempt_count();
1228 if (initcall_blacklisted(fn))
1231 do_trace_initcall_start(fn);
1233 do_trace_initcall_finish(fn, ret);
1237 if (preempt_count() != count) {
1238 sprintf(msgbuf, "preemption imbalance ");
1239 preempt_count_set(count);
1241 if (irqs_disabled()) {
1242 strlcat(msgbuf, "disabled interrupts ", sizeof(msgbuf));
1245 WARN(msgbuf[0], "initcall %pS returned with %s\n", fn, msgbuf);
1247 add_latent_entropy();
1252 static initcall_entry_t *initcall_levels[] __initdata = {
1264 /* Keep these in sync with initcalls in include/linux/init.h */
1265 static const char *initcall_level_names[] __initdata = {
1276 static int __init ignore_unknown_bootoption(char *param, char *val,
1277 const char *unused, void *arg)
1282 static void __init do_initcall_level(int level, char *command_line)
1284 initcall_entry_t *fn;
1286 parse_args(initcall_level_names[level],
1287 command_line, __start___param,
1288 __stop___param - __start___param,
1290 NULL, ignore_unknown_bootoption);
1292 trace_initcall_level(initcall_level_names[level]);
1293 for (fn = initcall_levels[level]; fn < initcall_levels[level+1]; fn++)
1294 do_one_initcall(initcall_from_entry(fn));
1297 static void __init do_initcalls(void)
1300 size_t len = saved_command_line_len + 1;
1303 command_line = kzalloc(len, GFP_KERNEL);
1305 panic("%s: Failed to allocate %zu bytes\n", __func__, len);
1307 for (level = 0; level < ARRAY_SIZE(initcall_levels) - 1; level++) {
1308 /* Parser modifies command_line, restore it each time */
1309 strcpy(command_line, saved_command_line);
1310 do_initcall_level(level, command_line);
1313 kfree(command_line);
1317 * Ok, the machine is now initialized. None of the devices
1318 * have been touched yet, but the CPU subsystem is up and
1319 * running, and memory and process management works.
1321 * Now we can finally start doing some real work..
1323 static void __init do_basic_setup(void)
1332 static void __init do_pre_smp_initcalls(void)
1334 initcall_entry_t *fn;
1336 trace_initcall_level("early");
1337 for (fn = __initcall_start; fn < __initcall0_start; fn++)
1338 do_one_initcall(initcall_from_entry(fn));
1341 static int run_init_process(const char *init_filename)
1343 const char *const *p;
1345 argv_init[0] = init_filename;
1346 pr_info("Run %s as init process\n", init_filename);
1347 pr_debug(" with arguments:\n");
1348 for (p = argv_init; *p; p++)
1349 pr_debug(" %s\n", *p);
1350 pr_debug(" with environment:\n");
1351 for (p = envp_init; *p; p++)
1352 pr_debug(" %s\n", *p);
1353 return kernel_execve(init_filename, argv_init, envp_init);
1356 static int try_to_run_init_process(const char *init_filename)
1360 ret = run_init_process(init_filename);
1362 if (ret && ret != -ENOENT) {
1363 pr_err("Starting init: %s exists but couldn't execute it (error %d)\n",
1364 init_filename, ret);
1370 static noinline void __init kernel_init_freeable(void);
1372 #if defined(CONFIG_STRICT_KERNEL_RWX) || defined(CONFIG_STRICT_MODULE_RWX)
1373 bool rodata_enabled __ro_after_init = true;
1375 #ifndef arch_parse_debug_rodata
1376 static inline bool arch_parse_debug_rodata(char *str) { return false; }
1379 static int __init set_debug_rodata(char *str)
1381 if (arch_parse_debug_rodata(str))
1384 if (str && !strcmp(str, "on"))
1385 rodata_enabled = true;
1386 else if (str && !strcmp(str, "off"))
1387 rodata_enabled = false;
1389 pr_warn("Invalid option string for rodata: '%s'\n", str);
1392 early_param("rodata", set_debug_rodata);
1395 #ifdef CONFIG_STRICT_KERNEL_RWX
1396 static void mark_readonly(void)
1398 if (rodata_enabled) {
1400 * load_module() results in W+X mappings, which are cleaned
1401 * up with call_rcu(). Let's make sure that queued work is
1402 * flushed so that we don't hit false positives looking for
1403 * insecure pages which are W+X.
1409 pr_info("Kernel memory protection disabled.\n");
1411 #elif defined(CONFIG_ARCH_HAS_STRICT_KERNEL_RWX)
1412 static inline void mark_readonly(void)
1414 pr_warn("Kernel memory protection not selected by kernel config.\n");
1417 static inline void mark_readonly(void)
1419 pr_warn("This architecture does not have kernel memory protection.\n");
1423 void __weak free_initmem(void)
1425 free_initmem_default(POISON_FREE_INITMEM);
1428 static int __ref kernel_init(void *unused)
1433 * Wait until kthreadd is all set-up.
1435 wait_for_completion(&kthreadd_done);
1437 kernel_init_freeable();
1438 /* need to finish all async __init code before freeing the memory */
1439 async_synchronize_full();
1441 system_state = SYSTEM_FREEING_INITMEM;
1442 kprobe_free_init_mem();
1443 ftrace_free_init_mem();
1444 kgdb_free_init_mem();
1450 * Kernel mappings are now finalized - update the userspace page-table
1455 system_state = SYSTEM_RUNNING;
1456 numa_default_policy();
1458 rcu_end_inkernel_boot();
1462 if (ramdisk_execute_command) {
1463 ret = run_init_process(ramdisk_execute_command);
1466 pr_err("Failed to execute %s (error %d)\n",
1467 ramdisk_execute_command, ret);
1471 * We try each of these until one succeeds.
1473 * The Bourne shell can be used instead of init if we are
1474 * trying to recover a really broken machine.
1476 if (execute_command) {
1477 ret = run_init_process(execute_command);
1480 panic("Requested init %s failed (error %d).",
1481 execute_command, ret);
1484 if (CONFIG_DEFAULT_INIT[0] != '\0') {
1485 ret = run_init_process(CONFIG_DEFAULT_INIT);
1487 pr_err("Default init %s failed (error %d)\n",
1488 CONFIG_DEFAULT_INIT, ret);
1493 if (!try_to_run_init_process("/sbin/init") ||
1494 !try_to_run_init_process("/etc/init") ||
1495 !try_to_run_init_process("/bin/init") ||
1496 !try_to_run_init_process("/bin/sh"))
1499 panic("No working init found. Try passing init= option to kernel. "
1500 "See Linux Documentation/admin-guide/init.rst for guidance.");
1503 /* Open /dev/console, for stdin/stdout/stderr, this should never fail */
1504 void __init console_on_rootfs(void)
1506 struct file *file = filp_open("/dev/console", O_RDWR, 0);
1509 pr_err("Warning: unable to open an initial console.\n");
1518 static noinline void __init kernel_init_freeable(void)
1520 /* Now the scheduler is fully set up and can do blocking allocations */
1521 gfp_allowed_mask = __GFP_BITS_MASK;
1524 * init can allocate pages on any node
1526 set_mems_allowed(node_states[N_MEMORY]);
1528 cad_pid = get_pid(task_pid(current));
1530 smp_prepare_cpus(setup_max_cpus);
1534 init_mm_internals();
1536 rcu_init_tasks_generic();
1537 do_pre_smp_initcalls();
1538 lockup_detector_init();
1543 workqueue_init_topology();
1545 page_alloc_init_late();
1549 kunit_run_all_tests();
1551 wait_for_initramfs();
1552 console_on_rootfs();
1555 * check if there is an early userspace init. If yes, let it do all
1558 if (init_eaccess(ramdisk_execute_command) != 0) {
1559 ramdisk_execute_command = NULL;
1560 prepare_namespace();
1564 * Ok, we have completed the initial bootup, and
1565 * we're essentially up and running. Get rid of the
1566 * initmem segments and start the user-mode stuff..
1568 * rootfs is available now, try loading the public keys
1569 * and default modules
1572 integrity_load_keys();