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/mem_encrypt.h>
99 #include <linux/kcsan.h>
100 #include <linux/init_syscalls.h>
101 #include <linux/stackdepot.h>
102 #include <linux/randomize_kstack.h>
103 #include <net/net_namespace.h>
106 #include <asm/bugs.h>
107 #include <asm/setup.h>
108 #include <asm/sections.h>
109 #include <asm/cacheflush.h>
111 #define CREATE_TRACE_POINTS
112 #include <trace/events/initcall.h>
114 #include <kunit/test.h>
116 static int kernel_init(void *);
118 extern void init_IRQ(void);
119 extern void radix_tree_init(void);
120 extern void maple_tree_init(void);
123 * Debug helper: via this flag we know that we are in 'early bootup code'
124 * where only the boot processor is running with IRQ disabled. This means
125 * two things - IRQ must not be enabled before the flag is cleared and some
126 * operations which are not allowed with IRQ disabled are allowed while the
129 bool early_boot_irqs_disabled __read_mostly;
131 enum system_states system_state __read_mostly;
132 EXPORT_SYMBOL(system_state);
135 * Boot command-line arguments
137 #define MAX_INIT_ARGS CONFIG_INIT_ENV_ARG_LIMIT
138 #define MAX_INIT_ENVS CONFIG_INIT_ENV_ARG_LIMIT
140 extern void time_init(void);
141 /* Default late time init is NULL. archs can override this later. */
142 void (*__initdata late_time_init)(void);
144 /* Untouched command line saved by arch-specific code. */
145 char __initdata boot_command_line[COMMAND_LINE_SIZE];
146 /* Untouched saved command line (eg. for /proc) */
147 char *saved_command_line __ro_after_init;
148 unsigned int saved_command_line_len __ro_after_init;
149 /* Command line for parameter parsing */
150 static char *static_command_line;
151 /* Untouched extra command line */
152 static char *extra_command_line;
153 /* Extra init arguments */
154 static char *extra_init_args;
156 #ifdef CONFIG_BOOT_CONFIG
157 /* Is bootconfig on command line? */
158 static bool bootconfig_found;
159 static size_t initargs_offs;
161 # define bootconfig_found false
162 # define initargs_offs 0
165 static char *execute_command;
166 static char *ramdisk_execute_command = "/init";
169 * Used to generate warnings if static_key manipulation functions are used
170 * before jump_label_init is called.
172 bool static_key_initialized __read_mostly;
173 EXPORT_SYMBOL_GPL(static_key_initialized);
176 * If set, this is an indication to the drivers that reset the underlying
177 * device before going ahead with the initialization otherwise driver might
178 * rely on the BIOS and skip the reset operation.
180 * This is useful if kernel is booting in an unreliable environment.
181 * For ex. kdump situation where previous kernel has crashed, BIOS has been
182 * skipped and devices will be in unknown state.
184 unsigned int reset_devices;
185 EXPORT_SYMBOL(reset_devices);
187 static int __init set_reset_devices(char *str)
193 __setup("reset_devices", set_reset_devices);
195 static const char *argv_init[MAX_INIT_ARGS+2] = { "init", NULL, };
196 const char *envp_init[MAX_INIT_ENVS+2] = { "HOME=/", "TERM=linux", NULL, };
197 static const char *panic_later, *panic_param;
199 extern const struct obs_kernel_param __setup_start[], __setup_end[];
201 static bool __init obsolete_checksetup(char *line)
203 const struct obs_kernel_param *p;
204 bool had_early_param = false;
208 int n = strlen(p->str);
209 if (parameqn(line, p->str, n)) {
211 /* Already done in parse_early_param?
212 * (Needs exact match on param part).
213 * Keep iterating, as we can have early
214 * params and __setups of same names 8( */
215 if (line[n] == '\0' || line[n] == '=')
216 had_early_param = true;
217 } else if (!p->setup_func) {
218 pr_warn("Parameter %s is obsolete, ignored\n",
221 } else if (p->setup_func(line + n))
225 } while (p < __setup_end);
227 return had_early_param;
231 * This should be approx 2 Bo*oMips to start (note initial shift), and will
232 * still work even if initially too large, it will just take slightly longer
234 unsigned long loops_per_jiffy = (1<<12);
235 EXPORT_SYMBOL(loops_per_jiffy);
237 static int __init debug_kernel(char *str)
239 console_loglevel = CONSOLE_LOGLEVEL_DEBUG;
243 static int __init quiet_kernel(char *str)
245 console_loglevel = CONSOLE_LOGLEVEL_QUIET;
249 early_param("debug", debug_kernel);
250 early_param("quiet", quiet_kernel);
252 static int __init loglevel(char *str)
257 * Only update loglevel value when a correct setting was passed,
258 * to prevent blind crashes (when loglevel being set to 0) that
259 * are quite hard to debug
261 if (get_option(&str, &newlevel)) {
262 console_loglevel = newlevel;
269 early_param("loglevel", loglevel);
271 #ifdef CONFIG_BLK_DEV_INITRD
272 static void * __init get_boot_config_from_initrd(size_t *_size)
282 data = (char *)initrd_end - BOOTCONFIG_MAGIC_LEN;
284 * Since Grub may align the size of initrd to 4, we must
285 * check the preceding 3 bytes as well.
287 for (i = 0; i < 4; i++) {
288 if (!memcmp(data, BOOTCONFIG_MAGIC, BOOTCONFIG_MAGIC_LEN))
295 hdr = (u32 *)(data - 8);
296 size = le32_to_cpu(hdr[0]);
297 csum = le32_to_cpu(hdr[1]);
299 data = ((void *)hdr) - size;
300 if ((unsigned long)data < initrd_start) {
301 pr_err("bootconfig size %d is greater than initrd size %ld\n",
302 size, initrd_end - initrd_start);
306 if (xbc_calc_checksum(data, size) != csum) {
307 pr_err("bootconfig checksum failed\n");
311 /* Remove bootconfig from initramfs/initrd */
312 initrd_end = (unsigned long)data;
319 static void * __init get_boot_config_from_initrd(size_t *_size)
325 #ifdef CONFIG_BOOT_CONFIG
327 static char xbc_namebuf[XBC_KEYLEN_MAX] __initdata;
329 #define rest(dst, end) ((end) > (dst) ? (end) - (dst) : 0)
331 static int __init xbc_snprint_cmdline(char *buf, size_t size,
332 struct xbc_node *root)
334 struct xbc_node *knode, *vnode;
335 char *end = buf + size;
339 xbc_node_for_each_key_value(root, knode, val) {
340 ret = xbc_node_compose_key_after(root, knode,
341 xbc_namebuf, XBC_KEYLEN_MAX);
345 vnode = xbc_node_get_child(knode);
347 ret = snprintf(buf, rest(buf, end), "%s ", xbc_namebuf);
353 xbc_array_for_each_value(vnode, val) {
354 ret = snprintf(buf, rest(buf, end), "%s=\"%s\" ",
362 return buf - (end - size);
366 /* Make an extra command line under given key word */
367 static char * __init xbc_make_cmdline(const char *key)
369 struct xbc_node *root;
373 root = xbc_find_node(key);
377 /* Count required buffer size */
378 len = xbc_snprint_cmdline(NULL, 0, root);
382 new_cmdline = memblock_alloc(len + 1, SMP_CACHE_BYTES);
384 pr_err("Failed to allocate memory for extra kernel cmdline.\n");
388 ret = xbc_snprint_cmdline(new_cmdline, len + 1, root);
389 if (ret < 0 || ret > len) {
390 pr_err("Failed to print extra kernel cmdline.\n");
391 memblock_free(new_cmdline, len + 1);
398 static int __init bootconfig_params(char *param, char *val,
399 const char *unused, void *arg)
401 if (strcmp(param, "bootconfig") == 0) {
402 bootconfig_found = true;
407 static int __init warn_bootconfig(char *str)
409 /* The 'bootconfig' has been handled by bootconfig_params(). */
413 static void __init setup_boot_config(void)
415 static char tmp_cmdline[COMMAND_LINE_SIZE] __initdata;
416 const char *msg, *data;
421 /* Cut out the bootconfig data even if we have no bootconfig option */
422 data = get_boot_config_from_initrd(&size);
423 /* If there is no bootconfig in initrd, try embedded one. */
425 data = xbc_get_embedded_bootconfig(&size);
427 strscpy(tmp_cmdline, boot_command_line, COMMAND_LINE_SIZE);
428 err = parse_args("bootconfig", tmp_cmdline, NULL, 0, 0, 0, NULL,
431 if (IS_ERR(err) || !(bootconfig_found || IS_ENABLED(CONFIG_BOOT_CONFIG_FORCE)))
434 /* parse_args() stops at the next param of '--' and returns an address */
436 initargs_offs = err - tmp_cmdline;
439 /* If user intended to use bootconfig, show an error level message */
440 if (bootconfig_found)
441 pr_err("'bootconfig' found on command line, but no bootconfig found\n");
443 pr_info("No bootconfig data provided, so skipping bootconfig");
447 if (size >= XBC_DATA_MAX) {
448 pr_err("bootconfig size %ld greater than max size %d\n",
449 (long)size, XBC_DATA_MAX);
453 ret = xbc_init(data, size, &msg, &pos);
456 pr_err("Failed to init bootconfig: %s.\n", msg);
458 pr_err("Failed to parse bootconfig: %s at %d.\n",
461 xbc_get_info(&ret, NULL);
462 pr_info("Load bootconfig: %ld bytes %d nodes\n", (long)size, ret);
463 /* keys starting with "kernel." are passed via cmdline */
464 extra_command_line = xbc_make_cmdline("kernel");
465 /* Also, "init." keys are init arguments */
466 extra_init_args = xbc_make_cmdline("init");
471 static void __init exit_boot_config(void)
476 #else /* !CONFIG_BOOT_CONFIG */
478 static void __init setup_boot_config(void)
480 /* Remove bootconfig data from initrd */
481 get_boot_config_from_initrd(NULL);
484 static int __init warn_bootconfig(char *str)
486 pr_warn("WARNING: 'bootconfig' found on the kernel command line but CONFIG_BOOT_CONFIG is not set.\n");
490 #define exit_boot_config() do {} while (0)
492 #endif /* CONFIG_BOOT_CONFIG */
494 early_param("bootconfig", warn_bootconfig);
496 /* Change NUL term back to "=", to make "param" the whole string. */
497 static void __init repair_env_string(char *param, char *val)
500 /* param=val or param="val"? */
501 if (val == param+strlen(param)+1)
503 else if (val == param+strlen(param)+2) {
505 memmove(val-1, val, strlen(val)+1);
511 /* Anything after -- gets handed straight to init. */
512 static int __init set_init_arg(char *param, char *val,
513 const char *unused, void *arg)
520 repair_env_string(param, val);
522 for (i = 0; argv_init[i]; i++) {
523 if (i == MAX_INIT_ARGS) {
524 panic_later = "init";
529 argv_init[i] = param;
534 * Unknown boot options get handed to init, unless they look like
535 * unused parameters (modprobe will find them in /proc/cmdline).
537 static int __init unknown_bootoption(char *param, char *val,
538 const char *unused, void *arg)
540 size_t len = strlen(param);
542 repair_env_string(param, val);
544 /* Handle obsolete-style parameters */
545 if (obsolete_checksetup(param))
548 /* Unused module parameter. */
549 if (strnchr(param, len, '.'))
556 /* Environment option */
558 for (i = 0; envp_init[i]; i++) {
559 if (i == MAX_INIT_ENVS) {
563 if (!strncmp(param, envp_init[i], len+1))
566 envp_init[i] = param;
568 /* Command line option */
570 for (i = 0; argv_init[i]; i++) {
571 if (i == MAX_INIT_ARGS) {
572 panic_later = "init";
576 argv_init[i] = param;
581 static int __init init_setup(char *str)
585 execute_command = str;
587 * In case LILO is going to boot us with default command line,
588 * it prepends "auto" before the whole cmdline which makes
589 * the shell think it should execute a script with such name.
590 * So we ignore all arguments entered _before_ init=... [MJ]
592 for (i = 1; i < MAX_INIT_ARGS; i++)
596 __setup("init=", init_setup);
598 static int __init rdinit_setup(char *str)
602 ramdisk_execute_command = str;
603 /* See "auto" comment in init_setup */
604 for (i = 1; i < MAX_INIT_ARGS; i++)
608 __setup("rdinit=", rdinit_setup);
611 static const unsigned int setup_max_cpus = NR_CPUS;
612 static inline void setup_nr_cpu_ids(void) { }
613 static inline void smp_prepare_cpus(unsigned int maxcpus) { }
617 * We need to store the untouched command line for future reference.
618 * We also need to store the touched command line since the parameter
619 * parsing is performed in place, and we should allow a component to
620 * store reference of name/value for future reference.
622 static void __init setup_command_line(char *command_line)
624 size_t len, xlen = 0, ilen = 0;
626 if (extra_command_line)
627 xlen = strlen(extra_command_line);
629 ilen = strlen(extra_init_args) + 4; /* for " -- " */
631 len = xlen + strlen(boot_command_line) + 1;
633 saved_command_line = memblock_alloc(len + ilen, SMP_CACHE_BYTES);
634 if (!saved_command_line)
635 panic("%s: Failed to allocate %zu bytes\n", __func__, len + ilen);
637 static_command_line = memblock_alloc(len, SMP_CACHE_BYTES);
638 if (!static_command_line)
639 panic("%s: Failed to allocate %zu bytes\n", __func__, len);
643 * We have to put extra_command_line before boot command
644 * lines because there could be dashes (separator of init
645 * command line) in the command lines.
647 strcpy(saved_command_line, extra_command_line);
648 strcpy(static_command_line, extra_command_line);
650 strcpy(saved_command_line + xlen, boot_command_line);
651 strcpy(static_command_line + xlen, command_line);
655 * Append supplemental init boot args to saved_command_line
656 * so that user can check what command line options passed
658 * The order should always be
659 * " -- "[bootconfig init-param][cmdline init-param]
662 len = xlen + initargs_offs;
663 strcpy(saved_command_line + len, extra_init_args);
664 len += ilen - 4; /* strlen(extra_init_args) */
665 strcpy(saved_command_line + len,
666 boot_command_line + initargs_offs - 1);
668 len = strlen(saved_command_line);
669 strcpy(saved_command_line + len, " -- ");
671 strcpy(saved_command_line + len, extra_init_args);
675 saved_command_line_len = strlen(saved_command_line);
679 * We need to finalize in a non-__init function or else race conditions
680 * between the root thread and the init thread may cause start_kernel to
681 * be reaped by free_initmem before the root thread has proceeded to
684 * gcc-3.4 accidentally inlines this function, so use noinline.
687 static __initdata DECLARE_COMPLETION(kthreadd_done);
689 noinline void __ref __noreturn rest_init(void)
691 struct task_struct *tsk;
694 rcu_scheduler_starting();
696 * We need to spawn init first so that it obtains pid 1, however
697 * the init task will end up wanting to create kthreads, which, if
698 * we schedule it before we create kthreadd, will OOPS.
700 pid = user_mode_thread(kernel_init, NULL, CLONE_FS);
702 * Pin init on the boot CPU. Task migration is not properly working
703 * until sched_init_smp() has been run. It will set the allowed
704 * CPUs for init to the non isolated CPUs.
707 tsk = find_task_by_pid_ns(pid, &init_pid_ns);
708 tsk->flags |= PF_NO_SETAFFINITY;
709 set_cpus_allowed_ptr(tsk, cpumask_of(smp_processor_id()));
712 numa_default_policy();
713 pid = kernel_thread(kthreadd, NULL, NULL, CLONE_FS | CLONE_FILES);
715 kthreadd_task = find_task_by_pid_ns(pid, &init_pid_ns);
719 * Enable might_sleep() and smp_processor_id() checks.
720 * They cannot be enabled earlier because with CONFIG_PREEMPTION=y
721 * kernel_thread() would trigger might_sleep() splats. With
722 * CONFIG_PREEMPT_VOLUNTARY=y the init task might have scheduled
723 * already, but it's stuck on the kthreadd_done completion.
725 system_state = SYSTEM_SCHEDULING;
727 complete(&kthreadd_done);
730 * The boot idle thread must execute schedule()
731 * at least once to get things moving:
733 schedule_preempt_disabled();
734 /* Call into cpu_idle with preempt disabled */
735 cpu_startup_entry(CPUHP_ONLINE);
738 /* Check for early params. */
739 static int __init do_early_param(char *param, char *val,
740 const char *unused, void *arg)
742 const struct obs_kernel_param *p;
744 for (p = __setup_start; p < __setup_end; p++) {
745 if ((p->early && parameq(param, p->str)) ||
746 (strcmp(param, "console") == 0 &&
747 strcmp(p->str, "earlycon") == 0)
749 if (p->setup_func(val) != 0)
750 pr_warn("Malformed early option '%s'\n", param);
753 /* We accept everything at this stage. */
757 void __init parse_early_options(char *cmdline)
759 parse_args("early options", cmdline, NULL, 0, 0, 0, NULL,
763 /* Arch code calls this early on, or if not, just before other parsing. */
764 void __init parse_early_param(void)
766 static int done __initdata;
767 static char tmp_cmdline[COMMAND_LINE_SIZE] __initdata;
772 /* All fall through to do_early_param. */
773 strscpy(tmp_cmdline, boot_command_line, COMMAND_LINE_SIZE);
774 parse_early_options(tmp_cmdline);
778 void __init __weak arch_post_acpi_subsys_init(void) { }
780 void __init __weak smp_setup_processor_id(void)
784 # if THREAD_SIZE >= PAGE_SIZE
785 void __init __weak thread_stack_cache_init(void)
790 void __init __weak mem_encrypt_init(void) { }
792 void __init __weak poking_init(void) { }
794 void __init __weak pgtable_cache_init(void) { }
796 void __init __weak trap_init(void) { }
799 core_param(initcall_debug, initcall_debug, bool, 0644);
801 #ifdef TRACEPOINTS_ENABLED
802 static void __init initcall_debug_enable(void);
804 static inline void initcall_debug_enable(void)
809 #ifdef CONFIG_RANDOMIZE_KSTACK_OFFSET
810 DEFINE_STATIC_KEY_MAYBE_RO(CONFIG_RANDOMIZE_KSTACK_OFFSET_DEFAULT,
811 randomize_kstack_offset);
812 DEFINE_PER_CPU(u32, kstack_offset);
814 static int __init early_randomize_kstack_offset(char *buf)
819 ret = kstrtobool(buf, &bool_result);
824 static_branch_enable(&randomize_kstack_offset);
826 static_branch_disable(&randomize_kstack_offset);
829 early_param("randomize_kstack_offset", early_randomize_kstack_offset);
832 void __init __weak __noreturn arch_call_rest_init(void)
837 static void __init print_unknown_bootoptions(void)
839 char *unknown_options;
841 const char *const *p;
844 if (panic_later || (!argv_init[1] && !envp_init[2]))
848 * Determine how many options we have to print out, plus a space
851 len = 1; /* null terminator */
852 for (p = &argv_init[1]; *p; p++) {
856 for (p = &envp_init[2]; *p; p++) {
861 unknown_options = memblock_alloc(len, SMP_CACHE_BYTES);
862 if (!unknown_options) {
863 pr_err("%s: Failed to allocate %zu bytes\n",
867 end = unknown_options;
869 for (p = &argv_init[1]; *p; p++)
870 end += sprintf(end, " %s", *p);
871 for (p = &envp_init[2]; *p; p++)
872 end += sprintf(end, " %s", *p);
874 /* Start at unknown_options[1] to skip the initial space */
875 pr_notice("Unknown kernel command line parameters \"%s\", will be passed to user space.\n",
876 &unknown_options[1]);
877 memblock_free(unknown_options, len);
880 asmlinkage __visible void __init __no_sanitize_address __noreturn start_kernel(void)
885 set_task_stack_end_magic(&init_task);
886 smp_setup_processor_id();
887 debug_objects_early_init();
888 init_vmlinux_build_id();
893 early_boot_irqs_disabled = true;
896 * Interrupts are still disabled. Do necessary setups, then
901 pr_notice("%s", linux_banner);
902 early_security_init();
903 setup_arch(&command_line);
905 setup_command_line(command_line);
907 setup_per_cpu_areas();
908 smp_prepare_boot_cpu(); /* arch-specific boot-cpu hooks */
909 boot_cpu_hotplug_init();
911 pr_notice("Kernel command line: %s\n", saved_command_line);
912 /* parameters may set static keys */
915 after_dashes = parse_args("Booting kernel",
916 static_command_line, __start___param,
917 __stop___param - __start___param,
918 -1, -1, NULL, &unknown_bootoption);
919 print_unknown_bootoptions();
920 if (!IS_ERR_OR_NULL(after_dashes))
921 parse_args("Setting init args", after_dashes, NULL, 0, -1, -1,
924 parse_args("Setting extra init args", extra_init_args,
925 NULL, 0, -1, -1, NULL, set_init_arg);
927 /* Architectural and non-timekeeping rng init, before allocator init */
928 random_init_early(command_line);
931 * These use large bootmem allocations and must precede
932 * initalization of page allocator
935 vfs_caches_init_early();
942 /* trace_printk can be enabled here */
946 * Set up the scheduler prior starting any interrupts (such as the
947 * timer interrupt). Full topology setup happens at smp_init()
948 * time - but meanwhile we still have a functioning scheduler.
952 if (WARN(!irqs_disabled(),
953 "Interrupts were enabled *very* early, fixing it\n"))
959 * Set up housekeeping before setting up workqueues to allow the unbound
960 * workqueue to take non-housekeeping into account.
965 * Allow workqueue creation and work item queueing/cancelling
966 * early. Work item execution depends on kthreads and starts after
969 workqueue_init_early();
973 /* Trace events are available after this */
977 initcall_debug_enable();
979 context_tracking_init();
980 /* init some links before init_ISA_irqs() */
992 /* This must be after timekeeping is initialized */
995 /* These make use of the fully initialized rng */
997 boot_init_stack_canary();
1001 call_function_init();
1002 WARN(!irqs_disabled(), "Interrupts were enabled early\n");
1004 early_boot_irqs_disabled = false;
1007 kmem_cache_init_late();
1010 * HACK ALERT! This is early. We're enabling the console before
1011 * we've done PCI setups etc, and console_init() must be aware of
1012 * this. But we do want output early, in case something goes wrong.
1016 panic("Too many boot %s vars at `%s'", panic_later,
1022 * Need to run this when irqs are enabled, because it wants
1023 * to self-test [hard/soft]-irqs on/off lock inversion bugs
1028 #ifdef CONFIG_BLK_DEV_INITRD
1029 if (initrd_start && !initrd_below_start_ok &&
1030 page_to_pfn(virt_to_page((void *)initrd_start)) < min_low_pfn) {
1031 pr_crit("initrd overwritten (0x%08lx < 0x%08lx) - disabling it.\n",
1032 page_to_pfn(virt_to_page((void *)initrd_start)),
1037 setup_per_cpu_pageset();
1046 * This needs to be called before any devices perform DMA
1047 * operations that might use the SWIOTLB bounce buffers. It will
1048 * mark the bounce buffers as decrypted so that their usage will
1049 * not cause "plain-text" data to be decrypted when accessed. It
1050 * must be called after late_time_init() so that Hyper-V x86/x64
1051 * hypercalls work when the SWIOTLB bounce buffers are decrypted.
1058 if (efi_enabled(EFI_RUNTIME_SERVICES))
1059 efi_enter_virtual_mode();
1061 thread_stack_cache_init();
1078 taskstats_init_early();
1083 acpi_subsystem_init();
1084 arch_post_acpi_subsys_init();
1087 /* Do the rest non-__init'ed, we're now alive */
1088 arch_call_rest_init();
1090 prevent_tail_call_optimization();
1093 /* Call all constructor functions linked into the kernel. */
1094 static void __init do_ctors(void)
1097 * For UML, the constructors have already been called by the
1098 * normal setup code as it's just a normal ELF binary, so we
1099 * cannot do it again - but we do need CONFIG_CONSTRUCTORS
1100 * even on UML for modules.
1102 #if defined(CONFIG_CONSTRUCTORS) && !defined(CONFIG_UML)
1103 ctor_fn_t *fn = (ctor_fn_t *) __ctors_start;
1105 for (; fn < (ctor_fn_t *) __ctors_end; fn++)
1110 #ifdef CONFIG_KALLSYMS
1111 struct blacklist_entry {
1112 struct list_head next;
1116 static __initdata_or_module LIST_HEAD(blacklisted_initcalls);
1118 static int __init initcall_blacklist(char *str)
1121 struct blacklist_entry *entry;
1123 /* str argument is a comma-separated list of functions */
1125 str_entry = strsep(&str, ",");
1127 pr_debug("blacklisting initcall %s\n", str_entry);
1128 entry = memblock_alloc(sizeof(*entry),
1131 panic("%s: Failed to allocate %zu bytes\n",
1132 __func__, sizeof(*entry));
1133 entry->buf = memblock_alloc(strlen(str_entry) + 1,
1136 panic("%s: Failed to allocate %zu bytes\n",
1137 __func__, strlen(str_entry) + 1);
1138 strcpy(entry->buf, str_entry);
1139 list_add(&entry->next, &blacklisted_initcalls);
1141 } while (str_entry);
1146 static bool __init_or_module initcall_blacklisted(initcall_t fn)
1148 struct blacklist_entry *entry;
1149 char fn_name[KSYM_SYMBOL_LEN];
1152 if (list_empty(&blacklisted_initcalls))
1155 addr = (unsigned long) dereference_function_descriptor(fn);
1156 sprint_symbol_no_offset(fn_name, addr);
1159 * fn will be "function_name [module_name]" where [module_name] is not
1160 * displayed for built-in init functions. Strip off the [module_name].
1162 strreplace(fn_name, ' ', '\0');
1164 list_for_each_entry(entry, &blacklisted_initcalls, next) {
1165 if (!strcmp(fn_name, entry->buf)) {
1166 pr_debug("initcall %s blacklisted\n", fn_name);
1174 static int __init initcall_blacklist(char *str)
1176 pr_warn("initcall_blacklist requires CONFIG_KALLSYMS\n");
1180 static bool __init_or_module initcall_blacklisted(initcall_t fn)
1185 __setup("initcall_blacklist=", initcall_blacklist);
1187 static __init_or_module void
1188 trace_initcall_start_cb(void *data, initcall_t fn)
1190 ktime_t *calltime = data;
1192 printk(KERN_DEBUG "calling %pS @ %i\n", fn, task_pid_nr(current));
1193 *calltime = ktime_get();
1196 static __init_or_module void
1197 trace_initcall_finish_cb(void *data, initcall_t fn, int ret)
1199 ktime_t rettime, *calltime = data;
1201 rettime = ktime_get();
1202 printk(KERN_DEBUG "initcall %pS returned %d after %lld usecs\n",
1203 fn, ret, (unsigned long long)ktime_us_delta(rettime, *calltime));
1206 static ktime_t initcall_calltime;
1208 #ifdef TRACEPOINTS_ENABLED
1209 static void __init initcall_debug_enable(void)
1213 ret = register_trace_initcall_start(trace_initcall_start_cb,
1214 &initcall_calltime);
1215 ret |= register_trace_initcall_finish(trace_initcall_finish_cb,
1216 &initcall_calltime);
1217 WARN(ret, "Failed to register initcall tracepoints\n");
1219 # define do_trace_initcall_start trace_initcall_start
1220 # define do_trace_initcall_finish trace_initcall_finish
1222 static inline void do_trace_initcall_start(initcall_t fn)
1224 if (!initcall_debug)
1226 trace_initcall_start_cb(&initcall_calltime, fn);
1228 static inline void do_trace_initcall_finish(initcall_t fn, int ret)
1230 if (!initcall_debug)
1232 trace_initcall_finish_cb(&initcall_calltime, fn, ret);
1234 #endif /* !TRACEPOINTS_ENABLED */
1236 int __init_or_module do_one_initcall(initcall_t fn)
1238 int count = preempt_count();
1242 if (initcall_blacklisted(fn))
1245 do_trace_initcall_start(fn);
1247 do_trace_initcall_finish(fn, ret);
1251 if (preempt_count() != count) {
1252 sprintf(msgbuf, "preemption imbalance ");
1253 preempt_count_set(count);
1255 if (irqs_disabled()) {
1256 strlcat(msgbuf, "disabled interrupts ", sizeof(msgbuf));
1259 WARN(msgbuf[0], "initcall %pS returned with %s\n", fn, msgbuf);
1261 add_latent_entropy();
1266 extern initcall_entry_t __initcall_start[];
1267 extern initcall_entry_t __initcall0_start[];
1268 extern initcall_entry_t __initcall1_start[];
1269 extern initcall_entry_t __initcall2_start[];
1270 extern initcall_entry_t __initcall3_start[];
1271 extern initcall_entry_t __initcall4_start[];
1272 extern initcall_entry_t __initcall5_start[];
1273 extern initcall_entry_t __initcall6_start[];
1274 extern initcall_entry_t __initcall7_start[];
1275 extern initcall_entry_t __initcall_end[];
1277 static initcall_entry_t *initcall_levels[] __initdata = {
1289 /* Keep these in sync with initcalls in include/linux/init.h */
1290 static const char *initcall_level_names[] __initdata = {
1301 static int __init ignore_unknown_bootoption(char *param, char *val,
1302 const char *unused, void *arg)
1307 static void __init do_initcall_level(int level, char *command_line)
1309 initcall_entry_t *fn;
1311 parse_args(initcall_level_names[level],
1312 command_line, __start___param,
1313 __stop___param - __start___param,
1315 NULL, ignore_unknown_bootoption);
1317 trace_initcall_level(initcall_level_names[level]);
1318 for (fn = initcall_levels[level]; fn < initcall_levels[level+1]; fn++)
1319 do_one_initcall(initcall_from_entry(fn));
1322 static void __init do_initcalls(void)
1325 size_t len = saved_command_line_len + 1;
1328 command_line = kzalloc(len, GFP_KERNEL);
1330 panic("%s: Failed to allocate %zu bytes\n", __func__, len);
1332 for (level = 0; level < ARRAY_SIZE(initcall_levels) - 1; level++) {
1333 /* Parser modifies command_line, restore it each time */
1334 strcpy(command_line, saved_command_line);
1335 do_initcall_level(level, command_line);
1338 kfree(command_line);
1342 * Ok, the machine is now initialized. None of the devices
1343 * have been touched yet, but the CPU subsystem is up and
1344 * running, and memory and process management works.
1346 * Now we can finally start doing some real work..
1348 static void __init do_basic_setup(void)
1357 static void __init do_pre_smp_initcalls(void)
1359 initcall_entry_t *fn;
1361 trace_initcall_level("early");
1362 for (fn = __initcall_start; fn < __initcall0_start; fn++)
1363 do_one_initcall(initcall_from_entry(fn));
1366 static int run_init_process(const char *init_filename)
1368 const char *const *p;
1370 argv_init[0] = init_filename;
1371 pr_info("Run %s as init process\n", init_filename);
1372 pr_debug(" with arguments:\n");
1373 for (p = argv_init; *p; p++)
1374 pr_debug(" %s\n", *p);
1375 pr_debug(" with environment:\n");
1376 for (p = envp_init; *p; p++)
1377 pr_debug(" %s\n", *p);
1378 return kernel_execve(init_filename, argv_init, envp_init);
1381 static int try_to_run_init_process(const char *init_filename)
1385 ret = run_init_process(init_filename);
1387 if (ret && ret != -ENOENT) {
1388 pr_err("Starting init: %s exists but couldn't execute it (error %d)\n",
1389 init_filename, ret);
1395 static noinline void __init kernel_init_freeable(void);
1397 #if defined(CONFIG_STRICT_KERNEL_RWX) || defined(CONFIG_STRICT_MODULE_RWX)
1398 bool rodata_enabled __ro_after_init = true;
1400 #ifndef arch_parse_debug_rodata
1401 static inline bool arch_parse_debug_rodata(char *str) { return false; }
1404 static int __init set_debug_rodata(char *str)
1406 if (arch_parse_debug_rodata(str))
1409 if (str && !strcmp(str, "on"))
1410 rodata_enabled = true;
1411 else if (str && !strcmp(str, "off"))
1412 rodata_enabled = false;
1414 pr_warn("Invalid option string for rodata: '%s'\n", str);
1417 early_param("rodata", set_debug_rodata);
1420 #ifdef CONFIG_STRICT_KERNEL_RWX
1421 static void mark_readonly(void)
1423 if (rodata_enabled) {
1425 * load_module() results in W+X mappings, which are cleaned
1426 * up with call_rcu(). Let's make sure that queued work is
1427 * flushed so that we don't hit false positives looking for
1428 * insecure pages which are W+X.
1434 pr_info("Kernel memory protection disabled.\n");
1436 #elif defined(CONFIG_ARCH_HAS_STRICT_KERNEL_RWX)
1437 static inline void mark_readonly(void)
1439 pr_warn("Kernel memory protection not selected by kernel config.\n");
1442 static inline void mark_readonly(void)
1444 pr_warn("This architecture does not have kernel memory protection.\n");
1448 void __weak free_initmem(void)
1450 free_initmem_default(POISON_FREE_INITMEM);
1453 static int __ref kernel_init(void *unused)
1458 * Wait until kthreadd is all set-up.
1460 wait_for_completion(&kthreadd_done);
1462 kernel_init_freeable();
1463 /* need to finish all async __init code before freeing the memory */
1464 async_synchronize_full();
1466 system_state = SYSTEM_FREEING_INITMEM;
1467 kprobe_free_init_mem();
1468 ftrace_free_init_mem();
1469 kgdb_free_init_mem();
1475 * Kernel mappings are now finalized - update the userspace page-table
1480 system_state = SYSTEM_RUNNING;
1481 numa_default_policy();
1483 rcu_end_inkernel_boot();
1487 if (ramdisk_execute_command) {
1488 ret = run_init_process(ramdisk_execute_command);
1491 pr_err("Failed to execute %s (error %d)\n",
1492 ramdisk_execute_command, ret);
1496 * We try each of these until one succeeds.
1498 * The Bourne shell can be used instead of init if we are
1499 * trying to recover a really broken machine.
1501 if (execute_command) {
1502 ret = run_init_process(execute_command);
1505 panic("Requested init %s failed (error %d).",
1506 execute_command, ret);
1509 if (CONFIG_DEFAULT_INIT[0] != '\0') {
1510 ret = run_init_process(CONFIG_DEFAULT_INIT);
1512 pr_err("Default init %s failed (error %d)\n",
1513 CONFIG_DEFAULT_INIT, ret);
1518 if (!try_to_run_init_process("/sbin/init") ||
1519 !try_to_run_init_process("/etc/init") ||
1520 !try_to_run_init_process("/bin/init") ||
1521 !try_to_run_init_process("/bin/sh"))
1524 panic("No working init found. Try passing init= option to kernel. "
1525 "See Linux Documentation/admin-guide/init.rst for guidance.");
1528 /* Open /dev/console, for stdin/stdout/stderr, this should never fail */
1529 void __init console_on_rootfs(void)
1531 struct file *file = filp_open("/dev/console", O_RDWR, 0);
1534 pr_err("Warning: unable to open an initial console.\n");
1543 static noinline void __init kernel_init_freeable(void)
1545 /* Now the scheduler is fully set up and can do blocking allocations */
1546 gfp_allowed_mask = __GFP_BITS_MASK;
1549 * init can allocate pages on any node
1551 set_mems_allowed(node_states[N_MEMORY]);
1553 cad_pid = get_pid(task_pid(current));
1555 smp_prepare_cpus(setup_max_cpus);
1559 init_mm_internals();
1561 rcu_init_tasks_generic();
1562 do_pre_smp_initcalls();
1563 lockup_detector_init();
1569 page_alloc_init_late();
1573 kunit_run_all_tests();
1575 wait_for_initramfs();
1576 console_on_rootfs();
1579 * check if there is an early userspace init. If yes, let it do all
1582 if (init_eaccess(ramdisk_execute_command) != 0) {
1583 ramdisk_execute_command = NULL;
1584 prepare_namespace();
1588 * Ok, we have completed the initial bootup, and
1589 * we're essentially up and running. Get rid of the
1590 * initmem segments and start the user-mode stuff..
1592 * rootfs is available now, try loading the public keys
1593 * and default modules
1596 integrity_load_keys();