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/vmalloc.h>
38 #include <linux/kernel_stat.h>
39 #include <linux/start_kernel.h>
40 #include <linux/security.h>
41 #include <linux/smp.h>
42 #include <linux/profile.h>
43 #include <linux/kfence.h>
44 #include <linux/rcupdate.h>
45 #include <linux/srcu.h>
46 #include <linux/moduleparam.h>
47 #include <linux/kallsyms.h>
48 #include <linux/buildid.h>
49 #include <linux/writeback.h>
50 #include <linux/cpu.h>
51 #include <linux/cpuset.h>
52 #include <linux/cgroup.h>
53 #include <linux/efi.h>
54 #include <linux/tick.h>
55 #include <linux/sched/isolation.h>
56 #include <linux/interrupt.h>
57 #include <linux/taskstats_kern.h>
58 #include <linux/delayacct.h>
59 #include <linux/unistd.h>
60 #include <linux/utsname.h>
61 #include <linux/rmap.h>
62 #include <linux/mempolicy.h>
63 #include <linux/key.h>
64 #include <linux/page_ext.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/elevator.h>
87 #include <linux/sched/clock.h>
88 #include <linux/sched/task.h>
89 #include <linux/sched/task_stack.h>
90 #include <linux/context_tracking.h>
91 #include <linux/random.h>
92 #include <linux/list.h>
93 #include <linux/integrity.h>
94 #include <linux/proc_ns.h>
96 #include <linux/cache.h>
97 #include <linux/rodata_test.h>
98 #include <linux/jump_label.h>
99 #include <linux/mem_encrypt.h>
100 #include <linux/kcsan.h>
101 #include <linux/init_syscalls.h>
102 #include <linux/stackdepot.h>
105 #include <asm/bugs.h>
106 #include <asm/setup.h>
107 #include <asm/sections.h>
108 #include <asm/cacheflush.h>
110 #define CREATE_TRACE_POINTS
111 #include <trace/events/initcall.h>
113 #include <kunit/test.h>
115 static int kernel_init(void *);
117 extern void init_IRQ(void);
118 extern void radix_tree_init(void);
121 * Debug helper: via this flag we know that we are in 'early bootup code'
122 * where only the boot processor is running with IRQ disabled. This means
123 * two things - IRQ must not be enabled before the flag is cleared and some
124 * operations which are not allowed with IRQ disabled are allowed while the
127 bool early_boot_irqs_disabled __read_mostly;
129 enum system_states system_state __read_mostly;
130 EXPORT_SYMBOL(system_state);
133 * Boot command-line arguments
135 #define MAX_INIT_ARGS CONFIG_INIT_ENV_ARG_LIMIT
136 #define MAX_INIT_ENVS CONFIG_INIT_ENV_ARG_LIMIT
138 extern void time_init(void);
139 /* Default late time init is NULL. archs can override this later. */
140 void (*__initdata late_time_init)(void);
142 /* Untouched command line saved by arch-specific code. */
143 char __initdata boot_command_line[COMMAND_LINE_SIZE];
144 /* Untouched saved command line (eg. for /proc) */
145 char *saved_command_line;
146 /* Command line for parameter parsing */
147 static char *static_command_line;
148 /* Untouched extra command line */
149 static char *extra_command_line;
150 /* Extra init arguments */
151 static char *extra_init_args;
153 #ifdef CONFIG_BOOT_CONFIG
154 /* Is bootconfig on command line? */
155 static bool bootconfig_found;
156 static size_t initargs_offs;
158 # define bootconfig_found false
159 # define initargs_offs 0
162 static char *execute_command;
163 static char *ramdisk_execute_command = "/init";
166 * Used to generate warnings if static_key manipulation functions are used
167 * before jump_label_init is called.
169 bool static_key_initialized __read_mostly;
170 EXPORT_SYMBOL_GPL(static_key_initialized);
173 * If set, this is an indication to the drivers that reset the underlying
174 * device before going ahead with the initialization otherwise driver might
175 * rely on the BIOS and skip the reset operation.
177 * This is useful if kernel is booting in an unreliable environment.
178 * For ex. kdump situation where previous kernel has crashed, BIOS has been
179 * skipped and devices will be in unknown state.
181 unsigned int reset_devices;
182 EXPORT_SYMBOL(reset_devices);
184 static int __init set_reset_devices(char *str)
190 __setup("reset_devices", set_reset_devices);
192 static const char *argv_init[MAX_INIT_ARGS+2] = { "init", NULL, };
193 const char *envp_init[MAX_INIT_ENVS+2] = { "HOME=/", "TERM=linux", NULL, };
194 static const char *panic_later, *panic_param;
196 extern const struct obs_kernel_param __setup_start[], __setup_end[];
198 static bool __init obsolete_checksetup(char *line)
200 const struct obs_kernel_param *p;
201 bool had_early_param = false;
205 int n = strlen(p->str);
206 if (parameqn(line, p->str, n)) {
208 /* Already done in parse_early_param?
209 * (Needs exact match on param part).
210 * Keep iterating, as we can have early
211 * params and __setups of same names 8( */
212 if (line[n] == '\0' || line[n] == '=')
213 had_early_param = true;
214 } else if (!p->setup_func) {
215 pr_warn("Parameter %s is obsolete, ignored\n",
218 } else if (p->setup_func(line + n))
222 } while (p < __setup_end);
224 return had_early_param;
228 * This should be approx 2 Bo*oMips to start (note initial shift), and will
229 * still work even if initially too large, it will just take slightly longer
231 unsigned long loops_per_jiffy = (1<<12);
232 EXPORT_SYMBOL(loops_per_jiffy);
234 static int __init debug_kernel(char *str)
236 console_loglevel = CONSOLE_LOGLEVEL_DEBUG;
240 static int __init quiet_kernel(char *str)
242 console_loglevel = CONSOLE_LOGLEVEL_QUIET;
246 early_param("debug", debug_kernel);
247 early_param("quiet", quiet_kernel);
249 static int __init loglevel(char *str)
254 * Only update loglevel value when a correct setting was passed,
255 * to prevent blind crashes (when loglevel being set to 0) that
256 * are quite hard to debug
258 if (get_option(&str, &newlevel)) {
259 console_loglevel = newlevel;
266 early_param("loglevel", loglevel);
268 #ifdef CONFIG_BLK_DEV_INITRD
269 static void * __init get_boot_config_from_initrd(u32 *_size, u32 *_csum)
279 data = (char *)initrd_end - BOOTCONFIG_MAGIC_LEN;
281 * Since Grub may align the size of initrd to 4, we must
282 * check the preceding 3 bytes as well.
284 for (i = 0; i < 4; i++) {
285 if (!memcmp(data, BOOTCONFIG_MAGIC, BOOTCONFIG_MAGIC_LEN))
292 hdr = (u32 *)(data - 8);
293 size = le32_to_cpu(hdr[0]);
294 csum = le32_to_cpu(hdr[1]);
296 data = ((void *)hdr) - size;
297 if ((unsigned long)data < initrd_start) {
298 pr_err("bootconfig size %d is greater than initrd size %ld\n",
299 size, initrd_end - initrd_start);
303 /* Remove bootconfig from initramfs/initrd */
304 initrd_end = (unsigned long)data;
313 static void * __init get_boot_config_from_initrd(u32 *_size, u32 *_csum)
319 #ifdef CONFIG_BOOT_CONFIG
321 static char xbc_namebuf[XBC_KEYLEN_MAX] __initdata;
323 #define rest(dst, end) ((end) > (dst) ? (end) - (dst) : 0)
325 static int __init xbc_snprint_cmdline(char *buf, size_t size,
326 struct xbc_node *root)
328 struct xbc_node *knode, *vnode;
329 char *end = buf + size;
333 xbc_node_for_each_key_value(root, knode, val) {
334 ret = xbc_node_compose_key_after(root, knode,
335 xbc_namebuf, XBC_KEYLEN_MAX);
339 vnode = xbc_node_get_child(knode);
341 ret = snprintf(buf, rest(buf, end), "%s ", xbc_namebuf);
347 xbc_array_for_each_value(vnode, val) {
348 ret = snprintf(buf, rest(buf, end), "%s=\"%s\" ",
356 return buf - (end - size);
360 /* Make an extra command line under given key word */
361 static char * __init xbc_make_cmdline(const char *key)
363 struct xbc_node *root;
367 root = xbc_find_node(key);
371 /* Count required buffer size */
372 len = xbc_snprint_cmdline(NULL, 0, root);
376 new_cmdline = memblock_alloc(len + 1, SMP_CACHE_BYTES);
378 pr_err("Failed to allocate memory for extra kernel cmdline.\n");
382 ret = xbc_snprint_cmdline(new_cmdline, len + 1, root);
383 if (ret < 0 || ret > len) {
384 pr_err("Failed to print extra kernel cmdline.\n");
391 static int __init bootconfig_params(char *param, char *val,
392 const char *unused, void *arg)
394 if (strcmp(param, "bootconfig") == 0) {
395 bootconfig_found = true;
400 static int __init warn_bootconfig(char *str)
402 /* The 'bootconfig' has been handled by bootconfig_params(). */
406 static void __init setup_boot_config(void)
408 static char tmp_cmdline[COMMAND_LINE_SIZE] __initdata;
412 char *data, *copy, *err;
415 /* Cut out the bootconfig data even if we have no bootconfig option */
416 data = get_boot_config_from_initrd(&size, &csum);
418 strlcpy(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)
425 /* parse_args() stops at the next param of '--' and returns an address */
427 initargs_offs = err - tmp_cmdline;
430 pr_err("'bootconfig' found on command line, but no bootconfig found\n");
434 if (size >= XBC_DATA_MAX) {
435 pr_err("bootconfig size %d greater than max size %d\n",
440 if (xbc_calc_checksum(data, size) != csum) {
441 pr_err("bootconfig checksum failed\n");
445 copy = memblock_alloc(size + 1, SMP_CACHE_BYTES);
447 pr_err("Failed to allocate memory for bootconfig\n");
451 memcpy(copy, data, size);
454 ret = xbc_init(copy, &msg, &pos);
457 pr_err("Failed to init bootconfig: %s.\n", msg);
459 pr_err("Failed to parse bootconfig: %s at %d.\n",
462 pr_info("Load bootconfig: %d bytes %d nodes\n", 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, 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);
677 * We need to finalize in a non-__init function or else race conditions
678 * between the root thread and the init thread may cause start_kernel to
679 * be reaped by free_initmem before the root thread has proceeded to
682 * gcc-3.4 accidentally inlines this function, so use noinline.
685 static __initdata DECLARE_COMPLETION(kthreadd_done);
687 noinline void __ref rest_init(void)
689 struct task_struct *tsk;
692 rcu_scheduler_starting();
694 * We need to spawn init first so that it obtains pid 1, however
695 * the init task will end up wanting to create kthreads, which, if
696 * we schedule it before we create kthreadd, will OOPS.
698 pid = kernel_thread(kernel_init, NULL, CLONE_FS);
700 * Pin init on the boot CPU. Task migration is not properly working
701 * until sched_init_smp() has been run. It will set the allowed
702 * CPUs for init to the non isolated CPUs.
705 tsk = find_task_by_pid_ns(pid, &init_pid_ns);
706 tsk->flags |= PF_NO_SETAFFINITY;
707 set_cpus_allowed_ptr(tsk, cpumask_of(smp_processor_id()));
710 numa_default_policy();
711 pid = kernel_thread(kthreadd, NULL, CLONE_FS | CLONE_FILES);
713 kthreadd_task = find_task_by_pid_ns(pid, &init_pid_ns);
717 * Enable might_sleep() and smp_processor_id() checks.
718 * They cannot be enabled earlier because with CONFIG_PREEMPTION=y
719 * kernel_thread() would trigger might_sleep() splats. With
720 * CONFIG_PREEMPT_VOLUNTARY=y the init task might have scheduled
721 * already, but it's stuck on the kthreadd_done completion.
723 system_state = SYSTEM_SCHEDULING;
725 complete(&kthreadd_done);
728 * The boot idle thread must execute schedule()
729 * at least once to get things moving:
731 schedule_preempt_disabled();
732 /* Call into cpu_idle with preempt disabled */
733 cpu_startup_entry(CPUHP_ONLINE);
736 /* Check for early params. */
737 static int __init do_early_param(char *param, char *val,
738 const char *unused, void *arg)
740 const struct obs_kernel_param *p;
742 for (p = __setup_start; p < __setup_end; p++) {
743 if ((p->early && parameq(param, p->str)) ||
744 (strcmp(param, "console") == 0 &&
745 strcmp(p->str, "earlycon") == 0)
747 if (p->setup_func(val) != 0)
748 pr_warn("Malformed early option '%s'\n", param);
751 /* We accept everything at this stage. */
755 void __init parse_early_options(char *cmdline)
757 parse_args("early options", cmdline, NULL, 0, 0, 0, NULL,
761 /* Arch code calls this early on, or if not, just before other parsing. */
762 void __init parse_early_param(void)
764 static int done __initdata;
765 static char tmp_cmdline[COMMAND_LINE_SIZE] __initdata;
770 /* All fall through to do_early_param. */
771 strlcpy(tmp_cmdline, boot_command_line, COMMAND_LINE_SIZE);
772 parse_early_options(tmp_cmdline);
776 void __init __weak arch_post_acpi_subsys_init(void) { }
778 void __init __weak smp_setup_processor_id(void)
782 # if THREAD_SIZE >= PAGE_SIZE
783 void __init __weak thread_stack_cache_init(void)
788 void __init __weak mem_encrypt_init(void) { }
790 void __init __weak poking_init(void) { }
792 void __init __weak pgtable_cache_init(void) { }
794 void __init __weak trap_init(void) { }
797 core_param(initcall_debug, initcall_debug, bool, 0644);
799 #ifdef TRACEPOINTS_ENABLED
800 static void __init initcall_debug_enable(void);
802 static inline void initcall_debug_enable(void)
807 /* Report memory auto-initialization states for this boot. */
808 static void __init report_meminit(void)
812 if (IS_ENABLED(CONFIG_INIT_STACK_ALL_PATTERN))
813 stack = "all(pattern)";
814 else if (IS_ENABLED(CONFIG_INIT_STACK_ALL_ZERO))
816 else if (IS_ENABLED(CONFIG_GCC_PLUGIN_STRUCTLEAK_BYREF_ALL))
817 stack = "byref_all(zero)";
818 else if (IS_ENABLED(CONFIG_GCC_PLUGIN_STRUCTLEAK_BYREF))
819 stack = "byref(zero)";
820 else if (IS_ENABLED(CONFIG_GCC_PLUGIN_STRUCTLEAK_USER))
821 stack = "__user(zero)";
825 pr_info("mem auto-init: stack:%s, heap alloc:%s, heap free:%s\n",
826 stack, want_init_on_alloc(GFP_KERNEL) ? "on" : "off",
827 want_init_on_free() ? "on" : "off");
828 if (want_init_on_free())
829 pr_info("mem auto-init: clearing system memory may take some time...\n");
833 * Set up kernel memory allocators
835 static void __init mm_init(void)
838 * page_ext requires contiguous pages,
839 * bigger than MAX_ORDER unless SPARSEMEM.
841 page_ext_init_flatmem();
842 init_mem_debugging_and_hardening();
847 mem_init_print_info();
848 /* page_owner must be initialized after buddy is ready */
849 page_ext_init_flatmem_late();
853 debug_objects_mem_init();
855 /* Should be run before the first non-init thread is created */
857 /* Should be run after espfix64 is set up. */
861 #ifdef CONFIG_HAVE_ARCH_RANDOMIZE_KSTACK_OFFSET
862 DEFINE_STATIC_KEY_MAYBE_RO(CONFIG_RANDOMIZE_KSTACK_OFFSET_DEFAULT,
863 randomize_kstack_offset);
864 DEFINE_PER_CPU(u32, kstack_offset);
866 static int __init early_randomize_kstack_offset(char *buf)
871 ret = kstrtobool(buf, &bool_result);
876 static_branch_enable(&randomize_kstack_offset);
878 static_branch_disable(&randomize_kstack_offset);
881 early_param("randomize_kstack_offset", early_randomize_kstack_offset);
884 void __init __weak arch_call_rest_init(void)
889 static void __init print_unknown_bootoptions(void)
891 char *unknown_options;
893 const char *const *p;
896 if (panic_later || (!argv_init[1] && !envp_init[2]))
900 * Determine how many options we have to print out, plus a space
903 len = 1; /* null terminator */
904 for (p = &argv_init[1]; *p; p++) {
908 for (p = &envp_init[2]; *p; p++) {
913 unknown_options = memblock_alloc(len, SMP_CACHE_BYTES);
914 if (!unknown_options) {
915 pr_err("%s: Failed to allocate %zu bytes\n",
919 end = unknown_options;
921 for (p = &argv_init[1]; *p; p++)
922 end += sprintf(end, " %s", *p);
923 for (p = &envp_init[2]; *p; p++)
924 end += sprintf(end, " %s", *p);
926 pr_notice("Unknown command line parameters:%s\n", unknown_options);
927 memblock_free_ptr(unknown_options, len);
930 asmlinkage __visible void __init __no_sanitize_address start_kernel(void)
935 set_task_stack_end_magic(&init_task);
936 smp_setup_processor_id();
937 debug_objects_early_init();
938 init_vmlinux_build_id();
943 early_boot_irqs_disabled = true;
946 * Interrupts are still disabled. Do necessary setups, then
951 pr_notice("%s", linux_banner);
952 early_security_init();
953 setup_arch(&command_line);
955 setup_command_line(command_line);
957 setup_per_cpu_areas();
958 smp_prepare_boot_cpu(); /* arch-specific boot-cpu hooks */
959 boot_cpu_hotplug_init();
961 build_all_zonelists(NULL);
964 pr_notice("Kernel command line: %s\n", saved_command_line);
965 /* parameters may set static keys */
968 after_dashes = parse_args("Booting kernel",
969 static_command_line, __start___param,
970 __stop___param - __start___param,
971 -1, -1, NULL, &unknown_bootoption);
972 print_unknown_bootoptions();
973 if (!IS_ERR_OR_NULL(after_dashes))
974 parse_args("Setting init args", after_dashes, NULL, 0, -1, -1,
977 parse_args("Setting extra init args", extra_init_args,
978 NULL, 0, -1, -1, NULL, set_init_arg);
981 * These use large bootmem allocations and must precede
985 vfs_caches_init_early();
992 /* trace_printk can be enabled here */
996 * Set up the scheduler prior starting any interrupts (such as the
997 * timer interrupt). Full topology setup happens at smp_init()
998 * time - but meanwhile we still have a functioning scheduler.
1002 if (WARN(!irqs_disabled(),
1003 "Interrupts were enabled *very* early, fixing it\n"))
1004 local_irq_disable();
1008 * Set up housekeeping before setting up workqueues to allow the unbound
1009 * workqueue to take non-housekeeping into account.
1011 housekeeping_init();
1014 * Allow workqueue creation and work item queueing/cancelling
1015 * early. Work item execution depends on kthreads and starts after
1018 workqueue_init_early();
1022 /* Trace events are available after this */
1026 initcall_debug_enable();
1028 context_tracking_init();
1029 /* init some links before init_ISA_irqs() */
1042 * For best initial stack canary entropy, prepare it after:
1043 * - setup_arch() for any UEFI RNG entropy and boot cmdline access
1044 * - timekeeping_init() for ktime entropy used in rand_initialize()
1045 * - rand_initialize() to get any arch-specific entropy like RDRAND
1046 * - add_latent_entropy() to get any latent entropy
1047 * - adding command line entropy
1050 add_latent_entropy();
1051 add_device_randomness(command_line, strlen(command_line));
1052 boot_init_stack_canary();
1057 call_function_init();
1058 WARN(!irqs_disabled(), "Interrupts were enabled early\n");
1060 early_boot_irqs_disabled = false;
1063 kmem_cache_init_late();
1066 * HACK ALERT! This is early. We're enabling the console before
1067 * we've done PCI setups etc, and console_init() must be aware of
1068 * this. But we do want output early, in case something goes wrong.
1072 panic("Too many boot %s vars at `%s'", panic_later,
1078 * Need to run this when irqs are enabled, because it wants
1079 * to self-test [hard/soft]-irqs on/off lock inversion bugs
1085 * This needs to be called before any devices perform DMA
1086 * operations that might use the SWIOTLB bounce buffers. It will
1087 * mark the bounce buffers as decrypted so that their usage will
1088 * not cause "plain-text" data to be decrypted when accessed.
1092 #ifdef CONFIG_BLK_DEV_INITRD
1093 if (initrd_start && !initrd_below_start_ok &&
1094 page_to_pfn(virt_to_page((void *)initrd_start)) < min_low_pfn) {
1095 pr_crit("initrd overwritten (0x%08lx < 0x%08lx) - disabling it.\n",
1096 page_to_pfn(virt_to_page((void *)initrd_start)),
1101 setup_per_cpu_pageset();
1111 if (efi_enabled(EFI_RUNTIME_SERVICES))
1112 efi_enter_virtual_mode();
1114 thread_stack_cache_init();
1130 taskstats_init_early();
1136 acpi_subsystem_init();
1137 arch_post_acpi_subsys_init();
1140 /* Do the rest non-__init'ed, we're now alive */
1141 arch_call_rest_init();
1143 prevent_tail_call_optimization();
1146 /* Call all constructor functions linked into the kernel. */
1147 static void __init do_ctors(void)
1150 * For UML, the constructors have already been called by the
1151 * normal setup code as it's just a normal ELF binary, so we
1152 * cannot do it again - but we do need CONFIG_CONSTRUCTORS
1153 * even on UML for modules.
1155 #if defined(CONFIG_CONSTRUCTORS) && !defined(CONFIG_UML)
1156 ctor_fn_t *fn = (ctor_fn_t *) __ctors_start;
1158 for (; fn < (ctor_fn_t *) __ctors_end; fn++)
1163 #ifdef CONFIG_KALLSYMS
1164 struct blacklist_entry {
1165 struct list_head next;
1169 static __initdata_or_module LIST_HEAD(blacklisted_initcalls);
1171 static int __init initcall_blacklist(char *str)
1174 struct blacklist_entry *entry;
1176 /* str argument is a comma-separated list of functions */
1178 str_entry = strsep(&str, ",");
1180 pr_debug("blacklisting initcall %s\n", str_entry);
1181 entry = memblock_alloc(sizeof(*entry),
1184 panic("%s: Failed to allocate %zu bytes\n",
1185 __func__, sizeof(*entry));
1186 entry->buf = memblock_alloc(strlen(str_entry) + 1,
1189 panic("%s: Failed to allocate %zu bytes\n",
1190 __func__, strlen(str_entry) + 1);
1191 strcpy(entry->buf, str_entry);
1192 list_add(&entry->next, &blacklisted_initcalls);
1194 } while (str_entry);
1199 static bool __init_or_module initcall_blacklisted(initcall_t fn)
1201 struct blacklist_entry *entry;
1202 char fn_name[KSYM_SYMBOL_LEN];
1205 if (list_empty(&blacklisted_initcalls))
1208 addr = (unsigned long) dereference_function_descriptor(fn);
1209 sprint_symbol_no_offset(fn_name, addr);
1212 * fn will be "function_name [module_name]" where [module_name] is not
1213 * displayed for built-in init functions. Strip off the [module_name].
1215 strreplace(fn_name, ' ', '\0');
1217 list_for_each_entry(entry, &blacklisted_initcalls, next) {
1218 if (!strcmp(fn_name, entry->buf)) {
1219 pr_debug("initcall %s blacklisted\n", fn_name);
1227 static int __init initcall_blacklist(char *str)
1229 pr_warn("initcall_blacklist requires CONFIG_KALLSYMS\n");
1233 static bool __init_or_module initcall_blacklisted(initcall_t fn)
1238 __setup("initcall_blacklist=", initcall_blacklist);
1240 static __init_or_module void
1241 trace_initcall_start_cb(void *data, initcall_t fn)
1243 ktime_t *calltime = (ktime_t *)data;
1245 printk(KERN_DEBUG "calling %pS @ %i\n", fn, task_pid_nr(current));
1246 *calltime = ktime_get();
1249 static __init_or_module void
1250 trace_initcall_finish_cb(void *data, initcall_t fn, int ret)
1252 ktime_t *calltime = (ktime_t *)data;
1253 ktime_t delta, rettime;
1254 unsigned long long duration;
1256 rettime = ktime_get();
1257 delta = ktime_sub(rettime, *calltime);
1258 duration = (unsigned long long) ktime_to_ns(delta) >> 10;
1259 printk(KERN_DEBUG "initcall %pS returned %d after %lld usecs\n",
1263 static ktime_t initcall_calltime;
1265 #ifdef TRACEPOINTS_ENABLED
1266 static void __init initcall_debug_enable(void)
1270 ret = register_trace_initcall_start(trace_initcall_start_cb,
1271 &initcall_calltime);
1272 ret |= register_trace_initcall_finish(trace_initcall_finish_cb,
1273 &initcall_calltime);
1274 WARN(ret, "Failed to register initcall tracepoints\n");
1276 # define do_trace_initcall_start trace_initcall_start
1277 # define do_trace_initcall_finish trace_initcall_finish
1279 static inline void do_trace_initcall_start(initcall_t fn)
1281 if (!initcall_debug)
1283 trace_initcall_start_cb(&initcall_calltime, fn);
1285 static inline void do_trace_initcall_finish(initcall_t fn, int ret)
1287 if (!initcall_debug)
1289 trace_initcall_finish_cb(&initcall_calltime, fn, ret);
1291 #endif /* !TRACEPOINTS_ENABLED */
1293 int __init_or_module do_one_initcall(initcall_t fn)
1295 int count = preempt_count();
1299 if (initcall_blacklisted(fn))
1302 do_trace_initcall_start(fn);
1304 do_trace_initcall_finish(fn, ret);
1308 if (preempt_count() != count) {
1309 sprintf(msgbuf, "preemption imbalance ");
1310 preempt_count_set(count);
1312 if (irqs_disabled()) {
1313 strlcat(msgbuf, "disabled interrupts ", sizeof(msgbuf));
1316 WARN(msgbuf[0], "initcall %pS returned with %s\n", fn, msgbuf);
1318 add_latent_entropy();
1323 extern initcall_entry_t __initcall_start[];
1324 extern initcall_entry_t __initcall0_start[];
1325 extern initcall_entry_t __initcall1_start[];
1326 extern initcall_entry_t __initcall2_start[];
1327 extern initcall_entry_t __initcall3_start[];
1328 extern initcall_entry_t __initcall4_start[];
1329 extern initcall_entry_t __initcall5_start[];
1330 extern initcall_entry_t __initcall6_start[];
1331 extern initcall_entry_t __initcall7_start[];
1332 extern initcall_entry_t __initcall_end[];
1334 static initcall_entry_t *initcall_levels[] __initdata = {
1346 /* Keep these in sync with initcalls in include/linux/init.h */
1347 static const char *initcall_level_names[] __initdata = {
1358 static int __init ignore_unknown_bootoption(char *param, char *val,
1359 const char *unused, void *arg)
1364 static void __init do_initcall_level(int level, char *command_line)
1366 initcall_entry_t *fn;
1368 parse_args(initcall_level_names[level],
1369 command_line, __start___param,
1370 __stop___param - __start___param,
1372 NULL, ignore_unknown_bootoption);
1374 trace_initcall_level(initcall_level_names[level]);
1375 for (fn = initcall_levels[level]; fn < initcall_levels[level+1]; fn++)
1376 do_one_initcall(initcall_from_entry(fn));
1379 static void __init do_initcalls(void)
1382 size_t len = strlen(saved_command_line) + 1;
1385 command_line = kzalloc(len, GFP_KERNEL);
1387 panic("%s: Failed to allocate %zu bytes\n", __func__, len);
1389 for (level = 0; level < ARRAY_SIZE(initcall_levels) - 1; level++) {
1390 /* Parser modifies command_line, restore it each time */
1391 strcpy(command_line, saved_command_line);
1392 do_initcall_level(level, command_line);
1395 kfree(command_line);
1399 * Ok, the machine is now initialized. None of the devices
1400 * have been touched yet, but the CPU subsystem is up and
1401 * running, and memory and process management works.
1403 * Now we can finally start doing some real work..
1405 static void __init do_basic_setup(void)
1414 static void __init do_pre_smp_initcalls(void)
1416 initcall_entry_t *fn;
1418 trace_initcall_level("early");
1419 for (fn = __initcall_start; fn < __initcall0_start; fn++)
1420 do_one_initcall(initcall_from_entry(fn));
1423 static int run_init_process(const char *init_filename)
1425 const char *const *p;
1427 argv_init[0] = init_filename;
1428 pr_info("Run %s as init process\n", init_filename);
1429 pr_debug(" with arguments:\n");
1430 for (p = argv_init; *p; p++)
1431 pr_debug(" %s\n", *p);
1432 pr_debug(" with environment:\n");
1433 for (p = envp_init; *p; p++)
1434 pr_debug(" %s\n", *p);
1435 return kernel_execve(init_filename, argv_init, envp_init);
1438 static int try_to_run_init_process(const char *init_filename)
1442 ret = run_init_process(init_filename);
1444 if (ret && ret != -ENOENT) {
1445 pr_err("Starting init: %s exists but couldn't execute it (error %d)\n",
1446 init_filename, ret);
1452 static noinline void __init kernel_init_freeable(void);
1454 #if defined(CONFIG_STRICT_KERNEL_RWX) || defined(CONFIG_STRICT_MODULE_RWX)
1455 bool rodata_enabled __ro_after_init = true;
1456 static int __init set_debug_rodata(char *str)
1458 return strtobool(str, &rodata_enabled);
1460 __setup("rodata=", set_debug_rodata);
1463 #ifdef CONFIG_STRICT_KERNEL_RWX
1464 static void mark_readonly(void)
1466 if (rodata_enabled) {
1468 * load_module() results in W+X mappings, which are cleaned
1469 * up with call_rcu(). Let's make sure that queued work is
1470 * flushed so that we don't hit false positives looking for
1471 * insecure pages which are W+X.
1477 pr_info("Kernel memory protection disabled.\n");
1479 #elif defined(CONFIG_ARCH_HAS_STRICT_KERNEL_RWX)
1480 static inline void mark_readonly(void)
1482 pr_warn("Kernel memory protection not selected by kernel config.\n");
1485 static inline void mark_readonly(void)
1487 pr_warn("This architecture does not have kernel memory protection.\n");
1491 void __weak free_initmem(void)
1493 free_initmem_default(POISON_FREE_INITMEM);
1496 static int __ref kernel_init(void *unused)
1501 * Wait until kthreadd is all set-up.
1503 wait_for_completion(&kthreadd_done);
1505 kernel_init_freeable();
1506 /* need to finish all async __init code before freeing the memory */
1507 async_synchronize_full();
1508 kprobe_free_init_mem();
1509 ftrace_free_init_mem();
1510 kgdb_free_init_mem();
1516 * Kernel mappings are now finalized - update the userspace page-table
1521 system_state = SYSTEM_RUNNING;
1522 numa_default_policy();
1524 rcu_end_inkernel_boot();
1528 if (ramdisk_execute_command) {
1529 ret = run_init_process(ramdisk_execute_command);
1532 pr_err("Failed to execute %s (error %d)\n",
1533 ramdisk_execute_command, ret);
1537 * We try each of these until one succeeds.
1539 * The Bourne shell can be used instead of init if we are
1540 * trying to recover a really broken machine.
1542 if (execute_command) {
1543 ret = run_init_process(execute_command);
1546 panic("Requested init %s failed (error %d).",
1547 execute_command, ret);
1550 if (CONFIG_DEFAULT_INIT[0] != '\0') {
1551 ret = run_init_process(CONFIG_DEFAULT_INIT);
1553 pr_err("Default init %s failed (error %d)\n",
1554 CONFIG_DEFAULT_INIT, ret);
1559 if (!try_to_run_init_process("/sbin/init") ||
1560 !try_to_run_init_process("/etc/init") ||
1561 !try_to_run_init_process("/bin/init") ||
1562 !try_to_run_init_process("/bin/sh"))
1565 panic("No working init found. Try passing init= option to kernel. "
1566 "See Linux Documentation/admin-guide/init.rst for guidance.");
1569 /* Open /dev/console, for stdin/stdout/stderr, this should never fail */
1570 void __init console_on_rootfs(void)
1572 struct file *file = filp_open("/dev/console", O_RDWR, 0);
1575 pr_err("Warning: unable to open an initial console.\n");
1584 static noinline void __init kernel_init_freeable(void)
1586 /* Now the scheduler is fully set up and can do blocking allocations */
1587 gfp_allowed_mask = __GFP_BITS_MASK;
1590 * init can allocate pages on any node
1592 set_mems_allowed(node_states[N_MEMORY]);
1594 cad_pid = get_pid(task_pid(current));
1596 smp_prepare_cpus(setup_max_cpus);
1600 init_mm_internals();
1602 rcu_init_tasks_generic();
1603 do_pre_smp_initcalls();
1604 lockup_detector_init();
1610 page_alloc_init_late();
1611 /* Initialize page ext after all struct pages are initialized. */
1616 kunit_run_all_tests();
1618 wait_for_initramfs();
1619 console_on_rootfs();
1622 * check if there is an early userspace init. If yes, let it do all
1625 if (init_eaccess(ramdisk_execute_command) != 0) {
1626 ramdisk_execute_command = NULL;
1627 prepare_namespace();
1631 * Ok, we have completed the initial bootup, and
1632 * we're essentially up and running. Get rid of the
1633 * initmem segments and start the user-mode stuff..
1635 * rootfs is available now, try loading the public keys
1636 * and default modules
1639 integrity_load_keys();