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/writeback.h>
49 #include <linux/cpu.h>
50 #include <linux/cpuset.h>
51 #include <linux/cgroup.h>
52 #include <linux/efi.h>
53 #include <linux/tick.h>
54 #include <linux/sched/isolation.h>
55 #include <linux/interrupt.h>
56 #include <linux/taskstats_kern.h>
57 #include <linux/delayacct.h>
58 #include <linux/unistd.h>
59 #include <linux/utsname.h>
60 #include <linux/rmap.h>
61 #include <linux/mempolicy.h>
62 #include <linux/key.h>
63 #include <linux/page_ext.h>
64 #include <linux/debug_locks.h>
65 #include <linux/debugobjects.h>
66 #include <linux/lockdep.h>
67 #include <linux/kmemleak.h>
68 #include <linux/padata.h>
69 #include <linux/pid_namespace.h>
70 #include <linux/device/driver.h>
71 #include <linux/kthread.h>
72 #include <linux/sched.h>
73 #include <linux/sched/init.h>
74 #include <linux/signal.h>
75 #include <linux/idr.h>
76 #include <linux/kgdb.h>
77 #include <linux/ftrace.h>
78 #include <linux/async.h>
79 #include <linux/shmem_fs.h>
80 #include <linux/slab.h>
81 #include <linux/perf_event.h>
82 #include <linux/ptrace.h>
83 #include <linux/pti.h>
84 #include <linux/blkdev.h>
85 #include <linux/elevator.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>
104 #include <asm/bugs.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 *);
116 extern void init_IRQ(void);
117 extern void radix_tree_init(void);
120 * Debug helper: via this flag we know that we are in 'early bootup code'
121 * where only the boot processor is running with IRQ disabled. This means
122 * two things - IRQ must not be enabled before the flag is cleared and some
123 * operations which are not allowed with IRQ disabled are allowed while the
126 bool early_boot_irqs_disabled __read_mostly;
128 enum system_states system_state __read_mostly;
129 EXPORT_SYMBOL(system_state);
132 * Boot command-line arguments
134 #define MAX_INIT_ARGS CONFIG_INIT_ENV_ARG_LIMIT
135 #define MAX_INIT_ENVS CONFIG_INIT_ENV_ARG_LIMIT
137 extern void time_init(void);
138 /* Default late time init is NULL. archs can override this later. */
139 void (*__initdata late_time_init)(void);
141 /* Untouched command line saved by arch-specific code. */
142 char __initdata boot_command_line[COMMAND_LINE_SIZE];
143 /* Untouched saved command line (eg. for /proc) */
144 char *saved_command_line;
145 /* Command line for parameter parsing */
146 static char *static_command_line;
147 /* Untouched extra command line */
148 static char *extra_command_line;
149 /* Extra init arguments */
150 static char *extra_init_args;
152 #ifdef CONFIG_BOOT_CONFIG
153 /* Is bootconfig on command line? */
154 static bool bootconfig_found;
155 static bool initargs_found;
157 # define bootconfig_found false
158 # define initargs_found false
161 static char *execute_command;
162 static char *ramdisk_execute_command = "/init";
165 * Used to generate warnings if static_key manipulation functions are used
166 * before jump_label_init is called.
168 bool static_key_initialized __read_mostly;
169 EXPORT_SYMBOL_GPL(static_key_initialized);
172 * If set, this is an indication to the drivers that reset the underlying
173 * device before going ahead with the initialization otherwise driver might
174 * rely on the BIOS and skip the reset operation.
176 * This is useful if kernel is booting in an unreliable environment.
177 * For ex. kdump situation where previous kernel has crashed, BIOS has been
178 * skipped and devices will be in unknown state.
180 unsigned int reset_devices;
181 EXPORT_SYMBOL(reset_devices);
183 static int __init set_reset_devices(char *str)
189 __setup("reset_devices", set_reset_devices);
191 static const char *argv_init[MAX_INIT_ARGS+2] = { "init", NULL, };
192 const char *envp_init[MAX_INIT_ENVS+2] = { "HOME=/", "TERM=linux", NULL, };
193 static const char *panic_later, *panic_param;
195 extern const struct obs_kernel_param __setup_start[], __setup_end[];
197 static bool __init obsolete_checksetup(char *line)
199 const struct obs_kernel_param *p;
200 bool had_early_param = false;
204 int n = strlen(p->str);
205 if (parameqn(line, p->str, n)) {
207 /* Already done in parse_early_param?
208 * (Needs exact match on param part).
209 * Keep iterating, as we can have early
210 * params and __setups of same names 8( */
211 if (line[n] == '\0' || line[n] == '=')
212 had_early_param = true;
213 } else if (!p->setup_func) {
214 pr_warn("Parameter %s is obsolete, ignored\n",
217 } else if (p->setup_func(line + n))
221 } while (p < __setup_end);
223 return had_early_param;
227 * This should be approx 2 Bo*oMips to start (note initial shift), and will
228 * still work even if initially too large, it will just take slightly longer
230 unsigned long loops_per_jiffy = (1<<12);
231 EXPORT_SYMBOL(loops_per_jiffy);
233 static int __init debug_kernel(char *str)
235 console_loglevel = CONSOLE_LOGLEVEL_DEBUG;
239 static int __init quiet_kernel(char *str)
241 console_loglevel = CONSOLE_LOGLEVEL_QUIET;
245 early_param("debug", debug_kernel);
246 early_param("quiet", quiet_kernel);
248 static int __init loglevel(char *str)
253 * Only update loglevel value when a correct setting was passed,
254 * to prevent blind crashes (when loglevel being set to 0) that
255 * are quite hard to debug
257 if (get_option(&str, &newlevel)) {
258 console_loglevel = newlevel;
265 early_param("loglevel", loglevel);
267 #ifdef CONFIG_BLK_DEV_INITRD
268 static void * __init get_boot_config_from_initrd(u32 *_size, u32 *_csum)
278 data = (char *)initrd_end - BOOTCONFIG_MAGIC_LEN;
280 * Since Grub may align the size of initrd to 4, we must
281 * check the preceding 3 bytes as well.
283 for (i = 0; i < 4; i++) {
284 if (!memcmp(data, BOOTCONFIG_MAGIC, BOOTCONFIG_MAGIC_LEN))
291 hdr = (u32 *)(data - 8);
292 size = le32_to_cpu(hdr[0]);
293 csum = le32_to_cpu(hdr[1]);
295 data = ((void *)hdr) - size;
296 if ((unsigned long)data < initrd_start) {
297 pr_err("bootconfig size %d is greater than initrd size %ld\n",
298 size, initrd_end - initrd_start);
302 /* Remove bootconfig from initramfs/initrd */
303 initrd_end = (unsigned long)data;
312 static void * __init get_boot_config_from_initrd(u32 *_size, u32 *_csum)
318 #ifdef CONFIG_BOOT_CONFIG
320 static char xbc_namebuf[XBC_KEYLEN_MAX] __initdata;
322 #define rest(dst, end) ((end) > (dst) ? (end) - (dst) : 0)
324 static int __init xbc_snprint_cmdline(char *buf, size_t size,
325 struct xbc_node *root)
327 struct xbc_node *knode, *vnode;
328 char *end = buf + size;
332 xbc_node_for_each_key_value(root, knode, val) {
333 ret = xbc_node_compose_key_after(root, knode,
334 xbc_namebuf, XBC_KEYLEN_MAX);
338 vnode = xbc_node_get_child(knode);
340 ret = snprintf(buf, rest(buf, end), "%s ", xbc_namebuf);
346 xbc_array_for_each_value(vnode, val) {
347 ret = snprintf(buf, rest(buf, end), "%s=\"%s\" ",
355 return buf - (end - size);
359 /* Make an extra command line under given key word */
360 static char * __init xbc_make_cmdline(const char *key)
362 struct xbc_node *root;
366 root = xbc_find_node(key);
370 /* Count required buffer size */
371 len = xbc_snprint_cmdline(NULL, 0, root);
375 new_cmdline = memblock_alloc(len + 1, SMP_CACHE_BYTES);
377 pr_err("Failed to allocate memory for extra kernel cmdline.\n");
381 ret = xbc_snprint_cmdline(new_cmdline, len + 1, root);
382 if (ret < 0 || ret > len) {
383 pr_err("Failed to print extra kernel cmdline.\n");
390 static int __init bootconfig_params(char *param, char *val,
391 const char *unused, void *arg)
393 if (strcmp(param, "bootconfig") == 0) {
394 bootconfig_found = true;
399 static void __init setup_boot_config(void)
401 static char tmp_cmdline[COMMAND_LINE_SIZE] __initdata;
405 char *data, *copy, *err;
408 /* Cut out the bootconfig data even if we have no bootconfig option */
409 data = get_boot_config_from_initrd(&size, &csum);
411 strlcpy(tmp_cmdline, boot_command_line, COMMAND_LINE_SIZE);
412 err = parse_args("bootconfig", tmp_cmdline, NULL, 0, 0, 0, NULL,
415 if (IS_ERR(err) || !bootconfig_found)
418 /* parse_args() stops at '--' and returns an address */
420 initargs_found = true;
423 pr_err("'bootconfig' found on command line, but no bootconfig found\n");
427 if (size >= XBC_DATA_MAX) {
428 pr_err("bootconfig size %d greater than max size %d\n",
433 if (xbc_calc_checksum(data, size) != csum) {
434 pr_err("bootconfig checksum failed\n");
438 copy = memblock_alloc(size + 1, SMP_CACHE_BYTES);
440 pr_err("Failed to allocate memory for bootconfig\n");
444 memcpy(copy, data, size);
447 ret = xbc_init(copy, &msg, &pos);
450 pr_err("Failed to init bootconfig: %s.\n", msg);
452 pr_err("Failed to parse bootconfig: %s at %d.\n",
455 pr_info("Load bootconfig: %d bytes %d nodes\n", size, ret);
456 /* keys starting with "kernel." are passed via cmdline */
457 extra_command_line = xbc_make_cmdline("kernel");
458 /* Also, "init." keys are init arguments */
459 extra_init_args = xbc_make_cmdline("init");
466 static void __init setup_boot_config(void)
468 /* Remove bootconfig data from initrd */
469 get_boot_config_from_initrd(NULL, NULL);
472 static int __init warn_bootconfig(char *str)
474 pr_warn("WARNING: 'bootconfig' found on the kernel command line but CONFIG_BOOT_CONFIG is not set.\n");
477 early_param("bootconfig", warn_bootconfig);
481 /* Change NUL term back to "=", to make "param" the whole string. */
482 static void __init repair_env_string(char *param, char *val)
485 /* param=val or param="val"? */
486 if (val == param+strlen(param)+1)
488 else if (val == param+strlen(param)+2) {
490 memmove(val-1, val, strlen(val)+1);
496 /* Anything after -- gets handed straight to init. */
497 static int __init set_init_arg(char *param, char *val,
498 const char *unused, void *arg)
505 repair_env_string(param, val);
507 for (i = 0; argv_init[i]; i++) {
508 if (i == MAX_INIT_ARGS) {
509 panic_later = "init";
514 argv_init[i] = param;
519 * Unknown boot options get handed to init, unless they look like
520 * unused parameters (modprobe will find them in /proc/cmdline).
522 static int __init unknown_bootoption(char *param, char *val,
523 const char *unused, void *arg)
525 size_t len = strlen(param);
527 repair_env_string(param, val);
529 /* Handle obsolete-style parameters */
530 if (obsolete_checksetup(param))
533 /* Unused module parameter. */
534 if (strnchr(param, len, '.'))
541 /* Environment option */
543 for (i = 0; envp_init[i]; i++) {
544 if (i == MAX_INIT_ENVS) {
548 if (!strncmp(param, envp_init[i], len+1))
551 envp_init[i] = param;
553 /* Command line option */
555 for (i = 0; argv_init[i]; i++) {
556 if (i == MAX_INIT_ARGS) {
557 panic_later = "init";
561 argv_init[i] = param;
566 static int __init init_setup(char *str)
570 execute_command = str;
572 * In case LILO is going to boot us with default command line,
573 * it prepends "auto" before the whole cmdline which makes
574 * the shell think it should execute a script with such name.
575 * So we ignore all arguments entered _before_ init=... [MJ]
577 for (i = 1; i < MAX_INIT_ARGS; i++)
581 __setup("init=", init_setup);
583 static int __init rdinit_setup(char *str)
587 ramdisk_execute_command = str;
588 /* See "auto" comment in init_setup */
589 for (i = 1; i < MAX_INIT_ARGS; i++)
593 __setup("rdinit=", rdinit_setup);
596 static const unsigned int setup_max_cpus = NR_CPUS;
597 static inline void setup_nr_cpu_ids(void) { }
598 static inline void smp_prepare_cpus(unsigned int maxcpus) { }
602 * We need to store the untouched command line for future reference.
603 * We also need to store the touched command line since the parameter
604 * parsing is performed in place, and we should allow a component to
605 * store reference of name/value for future reference.
607 static void __init setup_command_line(char *command_line)
609 size_t len, xlen = 0, ilen = 0;
611 if (extra_command_line)
612 xlen = strlen(extra_command_line);
614 ilen = strlen(extra_init_args) + 4; /* for " -- " */
616 len = xlen + strlen(boot_command_line) + 1;
618 saved_command_line = memblock_alloc(len + ilen, SMP_CACHE_BYTES);
619 if (!saved_command_line)
620 panic("%s: Failed to allocate %zu bytes\n", __func__, len + ilen);
622 static_command_line = memblock_alloc(len, SMP_CACHE_BYTES);
623 if (!static_command_line)
624 panic("%s: Failed to allocate %zu bytes\n", __func__, len);
628 * We have to put extra_command_line before boot command
629 * lines because there could be dashes (separator of init
630 * command line) in the command lines.
632 strcpy(saved_command_line, extra_command_line);
633 strcpy(static_command_line, extra_command_line);
635 strcpy(saved_command_line + xlen, boot_command_line);
636 strcpy(static_command_line + xlen, command_line);
640 * Append supplemental init boot args to saved_command_line
641 * so that user can check what command line options passed
644 len = strlen(saved_command_line);
645 if (initargs_found) {
646 saved_command_line[len++] = ' ';
648 strcpy(saved_command_line + len, " -- ");
652 strcpy(saved_command_line + len, extra_init_args);
657 * We need to finalize in a non-__init function or else race conditions
658 * between the root thread and the init thread may cause start_kernel to
659 * be reaped by free_initmem before the root thread has proceeded to
662 * gcc-3.4 accidentally inlines this function, so use noinline.
665 static __initdata DECLARE_COMPLETION(kthreadd_done);
667 noinline void __ref rest_init(void)
669 struct task_struct *tsk;
672 rcu_scheduler_starting();
674 * We need to spawn init first so that it obtains pid 1, however
675 * the init task will end up wanting to create kthreads, which, if
676 * we schedule it before we create kthreadd, will OOPS.
678 pid = kernel_thread(kernel_init, NULL, CLONE_FS);
680 * Pin init on the boot CPU. Task migration is not properly working
681 * until sched_init_smp() has been run. It will set the allowed
682 * CPUs for init to the non isolated CPUs.
685 tsk = find_task_by_pid_ns(pid, &init_pid_ns);
686 tsk->flags |= PF_NO_SETAFFINITY;
687 set_cpus_allowed_ptr(tsk, cpumask_of(smp_processor_id()));
690 numa_default_policy();
691 pid = kernel_thread(kthreadd, NULL, CLONE_FS | CLONE_FILES);
693 kthreadd_task = find_task_by_pid_ns(pid, &init_pid_ns);
697 * Enable might_sleep() and smp_processor_id() checks.
698 * They cannot be enabled earlier because with CONFIG_PREEMPTION=y
699 * kernel_thread() would trigger might_sleep() splats. With
700 * CONFIG_PREEMPT_VOLUNTARY=y the init task might have scheduled
701 * already, but it's stuck on the kthreadd_done completion.
703 system_state = SYSTEM_SCHEDULING;
705 complete(&kthreadd_done);
708 * The boot idle thread must execute schedule()
709 * at least once to get things moving:
711 schedule_preempt_disabled();
712 /* Call into cpu_idle with preempt disabled */
713 cpu_startup_entry(CPUHP_ONLINE);
716 /* Check for early params. */
717 static int __init do_early_param(char *param, char *val,
718 const char *unused, void *arg)
720 const struct obs_kernel_param *p;
722 for (p = __setup_start; p < __setup_end; p++) {
723 if ((p->early && parameq(param, p->str)) ||
724 (strcmp(param, "console") == 0 &&
725 strcmp(p->str, "earlycon") == 0)
727 if (p->setup_func(val) != 0)
728 pr_warn("Malformed early option '%s'\n", param);
731 /* We accept everything at this stage. */
735 void __init parse_early_options(char *cmdline)
737 parse_args("early options", cmdline, NULL, 0, 0, 0, NULL,
741 /* Arch code calls this early on, or if not, just before other parsing. */
742 void __init parse_early_param(void)
744 static int done __initdata;
745 static char tmp_cmdline[COMMAND_LINE_SIZE] __initdata;
750 /* All fall through to do_early_param. */
751 strlcpy(tmp_cmdline, boot_command_line, COMMAND_LINE_SIZE);
752 parse_early_options(tmp_cmdline);
756 void __init __weak arch_post_acpi_subsys_init(void) { }
758 void __init __weak smp_setup_processor_id(void)
762 # if THREAD_SIZE >= PAGE_SIZE
763 void __init __weak thread_stack_cache_init(void)
768 void __init __weak mem_encrypt_init(void) { }
770 void __init __weak poking_init(void) { }
772 void __init __weak pgtable_cache_init(void) { }
775 core_param(initcall_debug, initcall_debug, bool, 0644);
777 #ifdef TRACEPOINTS_ENABLED
778 static void __init initcall_debug_enable(void);
780 static inline void initcall_debug_enable(void)
785 /* Report memory auto-initialization states for this boot. */
786 static void __init report_meminit(void)
790 if (IS_ENABLED(CONFIG_INIT_STACK_ALL_PATTERN))
791 stack = "all(pattern)";
792 else if (IS_ENABLED(CONFIG_INIT_STACK_ALL_ZERO))
794 else if (IS_ENABLED(CONFIG_GCC_PLUGIN_STRUCTLEAK_BYREF_ALL))
795 stack = "byref_all(zero)";
796 else if (IS_ENABLED(CONFIG_GCC_PLUGIN_STRUCTLEAK_BYREF))
797 stack = "byref(zero)";
798 else if (IS_ENABLED(CONFIG_GCC_PLUGIN_STRUCTLEAK_USER))
799 stack = "__user(zero)";
803 pr_info("mem auto-init: stack:%s, heap alloc:%s, heap free:%s\n",
804 stack, want_init_on_alloc(GFP_KERNEL) ? "on" : "off",
805 want_init_on_free() ? "on" : "off");
806 if (want_init_on_free())
807 pr_info("mem auto-init: clearing system memory may take some time...\n");
811 * Set up kernel memory allocators
813 static void __init mm_init(void)
816 * page_ext requires contiguous pages,
817 * bigger than MAX_ORDER unless SPARSEMEM.
819 page_ext_init_flatmem();
820 init_mem_debugging_and_hardening();
825 mem_init_print_info();
826 /* page_owner must be initialized after buddy is ready */
827 page_ext_init_flatmem_late();
831 debug_objects_mem_init();
833 /* Should be run before the first non-init thread is created */
835 /* Should be run after espfix64 is set up. */
839 #ifdef CONFIG_HAVE_ARCH_RANDOMIZE_KSTACK_OFFSET
840 DEFINE_STATIC_KEY_MAYBE_RO(CONFIG_RANDOMIZE_KSTACK_OFFSET_DEFAULT,
841 randomize_kstack_offset);
842 DEFINE_PER_CPU(u32, kstack_offset);
844 static int __init early_randomize_kstack_offset(char *buf)
849 ret = kstrtobool(buf, &bool_result);
854 static_branch_enable(&randomize_kstack_offset);
856 static_branch_disable(&randomize_kstack_offset);
859 early_param("randomize_kstack_offset", early_randomize_kstack_offset);
862 void __init __weak arch_call_rest_init(void)
867 static void __init print_unknown_bootoptions(void)
869 char *unknown_options;
871 const char *const *p;
874 if (panic_later || (!argv_init[1] && !envp_init[2]))
878 * Determine how many options we have to print out, plus a space
881 len = 1; /* null terminator */
882 for (p = &argv_init[1]; *p; p++) {
886 for (p = &envp_init[2]; *p; p++) {
891 unknown_options = memblock_alloc(len, SMP_CACHE_BYTES);
892 if (!unknown_options) {
893 pr_err("%s: Failed to allocate %zu bytes\n",
897 end = unknown_options;
899 for (p = &argv_init[1]; *p; p++)
900 end += sprintf(end, " %s", *p);
901 for (p = &envp_init[2]; *p; p++)
902 end += sprintf(end, " %s", *p);
904 pr_notice("Unknown command line parameters:%s\n", unknown_options);
905 memblock_free(__pa(unknown_options), len);
908 asmlinkage __visible void __init __no_sanitize_address start_kernel(void)
913 set_task_stack_end_magic(&init_task);
914 smp_setup_processor_id();
915 debug_objects_early_init();
920 early_boot_irqs_disabled = true;
923 * Interrupts are still disabled. Do necessary setups, then
928 pr_notice("%s", linux_banner);
929 early_security_init();
930 setup_arch(&command_line);
932 setup_command_line(command_line);
934 setup_per_cpu_areas();
935 smp_prepare_boot_cpu(); /* arch-specific boot-cpu hooks */
936 boot_cpu_hotplug_init();
938 build_all_zonelists(NULL);
941 pr_notice("Kernel command line: %s\n", saved_command_line);
942 /* parameters may set static keys */
945 after_dashes = parse_args("Booting kernel",
946 static_command_line, __start___param,
947 __stop___param - __start___param,
948 -1, -1, NULL, &unknown_bootoption);
949 print_unknown_bootoptions();
950 if (!IS_ERR_OR_NULL(after_dashes))
951 parse_args("Setting init args", after_dashes, NULL, 0, -1, -1,
954 parse_args("Setting extra init args", extra_init_args,
955 NULL, 0, -1, -1, NULL, set_init_arg);
958 * These use large bootmem allocations and must precede
962 vfs_caches_init_early();
969 /* trace_printk can be enabled here */
973 * Set up the scheduler prior starting any interrupts (such as the
974 * timer interrupt). Full topology setup happens at smp_init()
975 * time - but meanwhile we still have a functioning scheduler.
979 if (WARN(!irqs_disabled(),
980 "Interrupts were enabled *very* early, fixing it\n"))
985 * Set up housekeeping before setting up workqueues to allow the unbound
986 * workqueue to take non-housekeeping into account.
991 * Allow workqueue creation and work item queueing/cancelling
992 * early. Work item execution depends on kthreads and starts after
995 workqueue_init_early();
999 /* Trace events are available after this */
1003 initcall_debug_enable();
1005 context_tracking_init();
1006 /* init some links before init_ISA_irqs() */
1019 * For best initial stack canary entropy, prepare it after:
1020 * - setup_arch() for any UEFI RNG entropy and boot cmdline access
1021 * - timekeeping_init() for ktime entropy used in rand_initialize()
1022 * - rand_initialize() to get any arch-specific entropy like RDRAND
1023 * - add_latent_entropy() to get any latent entropy
1024 * - adding command line entropy
1027 add_latent_entropy();
1028 add_device_randomness(command_line, strlen(command_line));
1029 boot_init_stack_canary();
1034 call_function_init();
1035 WARN(!irqs_disabled(), "Interrupts were enabled early\n");
1037 early_boot_irqs_disabled = false;
1040 kmem_cache_init_late();
1043 * HACK ALERT! This is early. We're enabling the console before
1044 * we've done PCI setups etc, and console_init() must be aware of
1045 * this. But we do want output early, in case something goes wrong.
1049 panic("Too many boot %s vars at `%s'", panic_later,
1055 * Need to run this when irqs are enabled, because it wants
1056 * to self-test [hard/soft]-irqs on/off lock inversion bugs
1062 * This needs to be called before any devices perform DMA
1063 * operations that might use the SWIOTLB bounce buffers. It will
1064 * mark the bounce buffers as decrypted so that their usage will
1065 * not cause "plain-text" data to be decrypted when accessed.
1069 #ifdef CONFIG_BLK_DEV_INITRD
1070 if (initrd_start && !initrd_below_start_ok &&
1071 page_to_pfn(virt_to_page((void *)initrd_start)) < min_low_pfn) {
1072 pr_crit("initrd overwritten (0x%08lx < 0x%08lx) - disabling it.\n",
1073 page_to_pfn(virt_to_page((void *)initrd_start)),
1078 setup_per_cpu_pageset();
1088 if (efi_enabled(EFI_RUNTIME_SERVICES))
1089 efi_enter_virtual_mode();
1091 thread_stack_cache_init();
1107 taskstats_init_early();
1113 acpi_subsystem_init();
1114 arch_post_acpi_subsys_init();
1117 /* Do the rest non-__init'ed, we're now alive */
1118 arch_call_rest_init();
1120 prevent_tail_call_optimization();
1123 /* Call all constructor functions linked into the kernel. */
1124 static void __init do_ctors(void)
1127 * For UML, the constructors have already been called by the
1128 * normal setup code as it's just a normal ELF binary, so we
1129 * cannot do it again - but we do need CONFIG_CONSTRUCTORS
1130 * even on UML for modules.
1132 #if defined(CONFIG_CONSTRUCTORS) && !defined(CONFIG_UML)
1133 ctor_fn_t *fn = (ctor_fn_t *) __ctors_start;
1135 for (; fn < (ctor_fn_t *) __ctors_end; fn++)
1140 #ifdef CONFIG_KALLSYMS
1141 struct blacklist_entry {
1142 struct list_head next;
1146 static __initdata_or_module LIST_HEAD(blacklisted_initcalls);
1148 static int __init initcall_blacklist(char *str)
1151 struct blacklist_entry *entry;
1153 /* str argument is a comma-separated list of functions */
1155 str_entry = strsep(&str, ",");
1157 pr_debug("blacklisting initcall %s\n", str_entry);
1158 entry = memblock_alloc(sizeof(*entry),
1161 panic("%s: Failed to allocate %zu bytes\n",
1162 __func__, sizeof(*entry));
1163 entry->buf = memblock_alloc(strlen(str_entry) + 1,
1166 panic("%s: Failed to allocate %zu bytes\n",
1167 __func__, strlen(str_entry) + 1);
1168 strcpy(entry->buf, str_entry);
1169 list_add(&entry->next, &blacklisted_initcalls);
1171 } while (str_entry);
1176 static bool __init_or_module initcall_blacklisted(initcall_t fn)
1178 struct blacklist_entry *entry;
1179 char fn_name[KSYM_SYMBOL_LEN];
1182 if (list_empty(&blacklisted_initcalls))
1185 addr = (unsigned long) dereference_function_descriptor(fn);
1186 sprint_symbol_no_offset(fn_name, addr);
1189 * fn will be "function_name [module_name]" where [module_name] is not
1190 * displayed for built-in init functions. Strip off the [module_name].
1192 strreplace(fn_name, ' ', '\0');
1194 list_for_each_entry(entry, &blacklisted_initcalls, next) {
1195 if (!strcmp(fn_name, entry->buf)) {
1196 pr_debug("initcall %s blacklisted\n", fn_name);
1204 static int __init initcall_blacklist(char *str)
1206 pr_warn("initcall_blacklist requires CONFIG_KALLSYMS\n");
1210 static bool __init_or_module initcall_blacklisted(initcall_t fn)
1215 __setup("initcall_blacklist=", initcall_blacklist);
1217 static __init_or_module void
1218 trace_initcall_start_cb(void *data, initcall_t fn)
1220 ktime_t *calltime = (ktime_t *)data;
1222 printk(KERN_DEBUG "calling %pS @ %i\n", fn, task_pid_nr(current));
1223 *calltime = ktime_get();
1226 static __init_or_module void
1227 trace_initcall_finish_cb(void *data, initcall_t fn, int ret)
1229 ktime_t *calltime = (ktime_t *)data;
1230 ktime_t delta, rettime;
1231 unsigned long long duration;
1233 rettime = ktime_get();
1234 delta = ktime_sub(rettime, *calltime);
1235 duration = (unsigned long long) ktime_to_ns(delta) >> 10;
1236 printk(KERN_DEBUG "initcall %pS returned %d after %lld usecs\n",
1240 static ktime_t initcall_calltime;
1242 #ifdef TRACEPOINTS_ENABLED
1243 static void __init initcall_debug_enable(void)
1247 ret = register_trace_initcall_start(trace_initcall_start_cb,
1248 &initcall_calltime);
1249 ret |= register_trace_initcall_finish(trace_initcall_finish_cb,
1250 &initcall_calltime);
1251 WARN(ret, "Failed to register initcall tracepoints\n");
1253 # define do_trace_initcall_start trace_initcall_start
1254 # define do_trace_initcall_finish trace_initcall_finish
1256 static inline void do_trace_initcall_start(initcall_t fn)
1258 if (!initcall_debug)
1260 trace_initcall_start_cb(&initcall_calltime, fn);
1262 static inline void do_trace_initcall_finish(initcall_t fn, int ret)
1264 if (!initcall_debug)
1266 trace_initcall_finish_cb(&initcall_calltime, fn, ret);
1268 #endif /* !TRACEPOINTS_ENABLED */
1270 int __init_or_module do_one_initcall(initcall_t fn)
1272 int count = preempt_count();
1276 if (initcall_blacklisted(fn))
1279 do_trace_initcall_start(fn);
1281 do_trace_initcall_finish(fn, ret);
1285 if (preempt_count() != count) {
1286 sprintf(msgbuf, "preemption imbalance ");
1287 preempt_count_set(count);
1289 if (irqs_disabled()) {
1290 strlcat(msgbuf, "disabled interrupts ", sizeof(msgbuf));
1293 WARN(msgbuf[0], "initcall %pS returned with %s\n", fn, msgbuf);
1295 add_latent_entropy();
1300 extern initcall_entry_t __initcall_start[];
1301 extern initcall_entry_t __initcall0_start[];
1302 extern initcall_entry_t __initcall1_start[];
1303 extern initcall_entry_t __initcall2_start[];
1304 extern initcall_entry_t __initcall3_start[];
1305 extern initcall_entry_t __initcall4_start[];
1306 extern initcall_entry_t __initcall5_start[];
1307 extern initcall_entry_t __initcall6_start[];
1308 extern initcall_entry_t __initcall7_start[];
1309 extern initcall_entry_t __initcall_end[];
1311 static initcall_entry_t *initcall_levels[] __initdata = {
1323 /* Keep these in sync with initcalls in include/linux/init.h */
1324 static const char *initcall_level_names[] __initdata = {
1335 static int __init ignore_unknown_bootoption(char *param, char *val,
1336 const char *unused, void *arg)
1341 static void __init do_initcall_level(int level, char *command_line)
1343 initcall_entry_t *fn;
1345 parse_args(initcall_level_names[level],
1346 command_line, __start___param,
1347 __stop___param - __start___param,
1349 NULL, ignore_unknown_bootoption);
1351 trace_initcall_level(initcall_level_names[level]);
1352 for (fn = initcall_levels[level]; fn < initcall_levels[level+1]; fn++)
1353 do_one_initcall(initcall_from_entry(fn));
1356 static void __init do_initcalls(void)
1359 size_t len = strlen(saved_command_line) + 1;
1362 command_line = kzalloc(len, GFP_KERNEL);
1364 panic("%s: Failed to allocate %zu bytes\n", __func__, len);
1366 for (level = 0; level < ARRAY_SIZE(initcall_levels) - 1; level++) {
1367 /* Parser modifies command_line, restore it each time */
1368 strcpy(command_line, saved_command_line);
1369 do_initcall_level(level, command_line);
1372 kfree(command_line);
1376 * Ok, the machine is now initialized. None of the devices
1377 * have been touched yet, but the CPU subsystem is up and
1378 * running, and memory and process management works.
1380 * Now we can finally start doing some real work..
1382 static void __init do_basic_setup(void)
1388 usermodehelper_enable();
1392 static void __init do_pre_smp_initcalls(void)
1394 initcall_entry_t *fn;
1396 trace_initcall_level("early");
1397 for (fn = __initcall_start; fn < __initcall0_start; fn++)
1398 do_one_initcall(initcall_from_entry(fn));
1401 static int run_init_process(const char *init_filename)
1403 const char *const *p;
1405 argv_init[0] = init_filename;
1406 pr_info("Run %s as init process\n", init_filename);
1407 pr_debug(" with arguments:\n");
1408 for (p = argv_init; *p; p++)
1409 pr_debug(" %s\n", *p);
1410 pr_debug(" with environment:\n");
1411 for (p = envp_init; *p; p++)
1412 pr_debug(" %s\n", *p);
1413 return kernel_execve(init_filename, argv_init, envp_init);
1416 static int try_to_run_init_process(const char *init_filename)
1420 ret = run_init_process(init_filename);
1422 if (ret && ret != -ENOENT) {
1423 pr_err("Starting init: %s exists but couldn't execute it (error %d)\n",
1424 init_filename, ret);
1430 static noinline void __init kernel_init_freeable(void);
1432 #if defined(CONFIG_STRICT_KERNEL_RWX) || defined(CONFIG_STRICT_MODULE_RWX)
1433 bool rodata_enabled __ro_after_init = true;
1434 static int __init set_debug_rodata(char *str)
1436 return strtobool(str, &rodata_enabled);
1438 __setup("rodata=", set_debug_rodata);
1441 #ifdef CONFIG_STRICT_KERNEL_RWX
1442 static void mark_readonly(void)
1444 if (rodata_enabled) {
1446 * load_module() results in W+X mappings, which are cleaned
1447 * up with call_rcu(). Let's make sure that queued work is
1448 * flushed so that we don't hit false positives looking for
1449 * insecure pages which are W+X.
1455 pr_info("Kernel memory protection disabled.\n");
1457 #elif defined(CONFIG_ARCH_HAS_STRICT_KERNEL_RWX)
1458 static inline void mark_readonly(void)
1460 pr_warn("Kernel memory protection not selected by kernel config.\n");
1463 static inline void mark_readonly(void)
1465 pr_warn("This architecture does not have kernel memory protection.\n");
1469 void __weak free_initmem(void)
1471 free_initmem_default(POISON_FREE_INITMEM);
1474 static int __ref kernel_init(void *unused)
1479 * Wait until kthreadd is all set-up.
1481 wait_for_completion(&kthreadd_done);
1483 kernel_init_freeable();
1484 /* need to finish all async __init code before freeing the memory */
1485 async_synchronize_full();
1486 kprobe_free_init_mem();
1487 ftrace_free_init_mem();
1488 kgdb_free_init_mem();
1493 * Kernel mappings are now finalized - update the userspace page-table
1498 system_state = SYSTEM_RUNNING;
1499 numa_default_policy();
1501 rcu_end_inkernel_boot();
1505 if (ramdisk_execute_command) {
1506 ret = run_init_process(ramdisk_execute_command);
1509 pr_err("Failed to execute %s (error %d)\n",
1510 ramdisk_execute_command, ret);
1514 * We try each of these until one succeeds.
1516 * The Bourne shell can be used instead of init if we are
1517 * trying to recover a really broken machine.
1519 if (execute_command) {
1520 ret = run_init_process(execute_command);
1523 panic("Requested init %s failed (error %d).",
1524 execute_command, ret);
1527 if (CONFIG_DEFAULT_INIT[0] != '\0') {
1528 ret = run_init_process(CONFIG_DEFAULT_INIT);
1530 pr_err("Default init %s failed (error %d)\n",
1531 CONFIG_DEFAULT_INIT, ret);
1536 if (!try_to_run_init_process("/sbin/init") ||
1537 !try_to_run_init_process("/etc/init") ||
1538 !try_to_run_init_process("/bin/init") ||
1539 !try_to_run_init_process("/bin/sh"))
1542 panic("No working init found. Try passing init= option to kernel. "
1543 "See Linux Documentation/admin-guide/init.rst for guidance.");
1546 /* Open /dev/console, for stdin/stdout/stderr, this should never fail */
1547 void __init console_on_rootfs(void)
1549 struct file *file = filp_open("/dev/console", O_RDWR, 0);
1552 pr_err("Warning: unable to open an initial console.\n");
1561 static noinline void __init kernel_init_freeable(void)
1563 /* Now the scheduler is fully set up and can do blocking allocations */
1564 gfp_allowed_mask = __GFP_BITS_MASK;
1567 * init can allocate pages on any node
1569 set_mems_allowed(node_states[N_MEMORY]);
1571 cad_pid = get_pid(task_pid(current));
1573 smp_prepare_cpus(setup_max_cpus);
1577 init_mm_internals();
1579 rcu_init_tasks_generic();
1580 do_pre_smp_initcalls();
1581 lockup_detector_init();
1587 page_alloc_init_late();
1588 /* Initialize page ext after all struct pages are initialized. */
1593 kunit_run_all_tests();
1595 wait_for_initramfs();
1596 console_on_rootfs();
1599 * check if there is an early userspace init. If yes, let it do all
1602 if (init_eaccess(ramdisk_execute_command) != 0) {
1603 ramdisk_execute_command = NULL;
1604 prepare_namespace();
1608 * Ok, we have completed the initial bootup, and
1609 * we're essentially up and running. Get rid of the
1610 * initmem segments and start the user-mode stuff..
1612 * rootfs is available now, try loading the public keys
1613 * and default modules
1616 integrity_load_keys();