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/page_ext.h>
66 #include <linux/debug_locks.h>
67 #include <linux/debugobjects.h>
68 #include <linux/lockdep.h>
69 #include <linux/kmemleak.h>
70 #include <linux/padata.h>
71 #include <linux/pid_namespace.h>
72 #include <linux/device/driver.h>
73 #include <linux/kthread.h>
74 #include <linux/sched.h>
75 #include <linux/sched/init.h>
76 #include <linux/signal.h>
77 #include <linux/idr.h>
78 #include <linux/kgdb.h>
79 #include <linux/ftrace.h>
80 #include <linux/async.h>
81 #include <linux/shmem_fs.h>
82 #include <linux/slab.h>
83 #include <linux/perf_event.h>
84 #include <linux/ptrace.h>
85 #include <linux/pti.h>
86 #include <linux/blkdev.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>
103 #include <linux/randomize_kstack.h>
104 #include <net/net_namespace.h>
107 #include <asm/bugs.h>
108 #include <asm/setup.h>
109 #include <asm/sections.h>
110 #include <asm/cacheflush.h>
112 #define CREATE_TRACE_POINTS
113 #include <trace/events/initcall.h>
115 #include <kunit/test.h>
117 static int kernel_init(void *);
119 extern void init_IRQ(void);
120 extern void radix_tree_init(void);
121 extern void maple_tree_init(void);
124 * Debug helper: via this flag we know that we are in 'early bootup code'
125 * where only the boot processor is running with IRQ disabled. This means
126 * two things - IRQ must not be enabled before the flag is cleared and some
127 * operations which are not allowed with IRQ disabled are allowed while the
130 bool early_boot_irqs_disabled __read_mostly;
132 enum system_states system_state __read_mostly;
133 EXPORT_SYMBOL(system_state);
136 * Boot command-line arguments
138 #define MAX_INIT_ARGS CONFIG_INIT_ENV_ARG_LIMIT
139 #define MAX_INIT_ENVS CONFIG_INIT_ENV_ARG_LIMIT
141 extern void time_init(void);
142 /* Default late time init is NULL. archs can override this later. */
143 void (*__initdata late_time_init)(void);
145 /* Untouched command line saved by arch-specific code. */
146 char __initdata boot_command_line[COMMAND_LINE_SIZE];
147 /* Untouched saved command line (eg. for /proc) */
148 char *saved_command_line __ro_after_init;
149 unsigned int saved_command_line_len __ro_after_init;
150 /* Command line for parameter parsing */
151 static char *static_command_line;
152 /* Untouched extra command line */
153 static char *extra_command_line;
154 /* Extra init arguments */
155 static char *extra_init_args;
157 #ifdef CONFIG_BOOT_CONFIG
158 /* Is bootconfig on command line? */
159 static bool bootconfig_found;
160 static size_t initargs_offs;
162 # define bootconfig_found false
163 # define initargs_offs 0
166 static char *execute_command;
167 static char *ramdisk_execute_command = "/init";
170 * Used to generate warnings if static_key manipulation functions are used
171 * before jump_label_init is called.
173 bool static_key_initialized __read_mostly;
174 EXPORT_SYMBOL_GPL(static_key_initialized);
177 * If set, this is an indication to the drivers that reset the underlying
178 * device before going ahead with the initialization otherwise driver might
179 * rely on the BIOS and skip the reset operation.
181 * This is useful if kernel is booting in an unreliable environment.
182 * For ex. kdump situation where previous kernel has crashed, BIOS has been
183 * skipped and devices will be in unknown state.
185 unsigned int reset_devices;
186 EXPORT_SYMBOL(reset_devices);
188 static int __init set_reset_devices(char *str)
194 __setup("reset_devices", set_reset_devices);
196 static const char *argv_init[MAX_INIT_ARGS+2] = { "init", NULL, };
197 const char *envp_init[MAX_INIT_ENVS+2] = { "HOME=/", "TERM=linux", NULL, };
198 static const char *panic_later, *panic_param;
200 extern const struct obs_kernel_param __setup_start[], __setup_end[];
202 static bool __init obsolete_checksetup(char *line)
204 const struct obs_kernel_param *p;
205 bool had_early_param = false;
209 int n = strlen(p->str);
210 if (parameqn(line, p->str, n)) {
212 /* Already done in parse_early_param?
213 * (Needs exact match on param part).
214 * Keep iterating, as we can have early
215 * params and __setups of same names 8( */
216 if (line[n] == '\0' || line[n] == '=')
217 had_early_param = true;
218 } else if (!p->setup_func) {
219 pr_warn("Parameter %s is obsolete, ignored\n",
222 } else if (p->setup_func(line + n))
226 } while (p < __setup_end);
228 return had_early_param;
232 * This should be approx 2 Bo*oMips to start (note initial shift), and will
233 * still work even if initially too large, it will just take slightly longer
235 unsigned long loops_per_jiffy = (1<<12);
236 EXPORT_SYMBOL(loops_per_jiffy);
238 static int __init debug_kernel(char *str)
240 console_loglevel = CONSOLE_LOGLEVEL_DEBUG;
244 static int __init quiet_kernel(char *str)
246 console_loglevel = CONSOLE_LOGLEVEL_QUIET;
250 early_param("debug", debug_kernel);
251 early_param("quiet", quiet_kernel);
253 static int __init loglevel(char *str)
258 * Only update loglevel value when a correct setting was passed,
259 * to prevent blind crashes (when loglevel being set to 0) that
260 * are quite hard to debug
262 if (get_option(&str, &newlevel)) {
263 console_loglevel = newlevel;
270 early_param("loglevel", loglevel);
272 #ifdef CONFIG_BLK_DEV_INITRD
273 static void * __init get_boot_config_from_initrd(size_t *_size)
283 data = (char *)initrd_end - BOOTCONFIG_MAGIC_LEN;
285 * Since Grub may align the size of initrd to 4, we must
286 * check the preceding 3 bytes as well.
288 for (i = 0; i < 4; i++) {
289 if (!memcmp(data, BOOTCONFIG_MAGIC, BOOTCONFIG_MAGIC_LEN))
296 hdr = (u32 *)(data - 8);
297 size = le32_to_cpu(hdr[0]);
298 csum = le32_to_cpu(hdr[1]);
300 data = ((void *)hdr) - size;
301 if ((unsigned long)data < initrd_start) {
302 pr_err("bootconfig size %d is greater than initrd size %ld\n",
303 size, initrd_end - initrd_start);
307 if (xbc_calc_checksum(data, size) != csum) {
308 pr_err("bootconfig checksum failed\n");
312 /* Remove bootconfig from initramfs/initrd */
313 initrd_end = (unsigned long)data;
320 static void * __init get_boot_config_from_initrd(size_t *_size)
326 #ifdef CONFIG_BOOT_CONFIG
328 static char xbc_namebuf[XBC_KEYLEN_MAX] __initdata;
330 #define rest(dst, end) ((end) > (dst) ? (end) - (dst) : 0)
332 static int __init xbc_snprint_cmdline(char *buf, size_t size,
333 struct xbc_node *root)
335 struct xbc_node *knode, *vnode;
336 char *end = buf + size;
340 xbc_node_for_each_key_value(root, knode, val) {
341 ret = xbc_node_compose_key_after(root, knode,
342 xbc_namebuf, XBC_KEYLEN_MAX);
346 vnode = xbc_node_get_child(knode);
348 ret = snprintf(buf, rest(buf, end), "%s ", xbc_namebuf);
354 xbc_array_for_each_value(vnode, val) {
355 ret = snprintf(buf, rest(buf, end), "%s=\"%s\" ",
363 return buf - (end - size);
367 /* Make an extra command line under given key word */
368 static char * __init xbc_make_cmdline(const char *key)
370 struct xbc_node *root;
374 root = xbc_find_node(key);
378 /* Count required buffer size */
379 len = xbc_snprint_cmdline(NULL, 0, root);
383 new_cmdline = memblock_alloc(len + 1, SMP_CACHE_BYTES);
385 pr_err("Failed to allocate memory for extra kernel cmdline.\n");
389 ret = xbc_snprint_cmdline(new_cmdline, len + 1, root);
390 if (ret < 0 || ret > len) {
391 pr_err("Failed to print extra kernel cmdline.\n");
392 memblock_free(new_cmdline, len + 1);
399 static int __init bootconfig_params(char *param, char *val,
400 const char *unused, void *arg)
402 if (strcmp(param, "bootconfig") == 0) {
403 bootconfig_found = true;
408 static int __init warn_bootconfig(char *str)
410 /* The 'bootconfig' has been handled by bootconfig_params(). */
414 static void __init setup_boot_config(void)
416 static char tmp_cmdline[COMMAND_LINE_SIZE] __initdata;
417 const char *msg, *data;
422 /* Cut out the bootconfig data even if we have no bootconfig option */
423 data = get_boot_config_from_initrd(&size);
424 /* If there is no bootconfig in initrd, try embedded one. */
426 data = xbc_get_embedded_bootconfig(&size);
428 strscpy(tmp_cmdline, boot_command_line, COMMAND_LINE_SIZE);
429 err = parse_args("bootconfig", tmp_cmdline, NULL, 0, 0, 0, NULL,
432 if (IS_ERR(err) || !(bootconfig_found || IS_ENABLED(CONFIG_BOOT_CONFIG_FORCE)))
435 /* parse_args() stops at the next param of '--' and returns an address */
437 initargs_offs = err - tmp_cmdline;
440 /* If user intended to use bootconfig, show an error level message */
441 if (bootconfig_found)
442 pr_err("'bootconfig' found on command line, but no bootconfig found\n");
444 pr_info("No bootconfig data provided, so skipping bootconfig");
448 if (size >= XBC_DATA_MAX) {
449 pr_err("bootconfig size %ld greater than max size %d\n",
450 (long)size, XBC_DATA_MAX);
454 ret = xbc_init(data, size, &msg, &pos);
457 pr_err("Failed to init bootconfig: %s.\n", msg);
459 pr_err("Failed to parse bootconfig: %s at %d.\n",
462 xbc_get_info(&ret, NULL);
463 pr_info("Load bootconfig: %ld bytes %d nodes\n", (long)size, ret);
464 /* keys starting with "kernel." are passed via cmdline */
465 extra_command_line = xbc_make_cmdline("kernel");
466 /* Also, "init." keys are init arguments */
467 extra_init_args = xbc_make_cmdline("init");
472 static void __init exit_boot_config(void)
477 #else /* !CONFIG_BOOT_CONFIG */
479 static void __init setup_boot_config(void)
481 /* Remove bootconfig data from initrd */
482 get_boot_config_from_initrd(NULL);
485 static int __init warn_bootconfig(char *str)
487 pr_warn("WARNING: 'bootconfig' found on the kernel command line but CONFIG_BOOT_CONFIG is not set.\n");
491 #define exit_boot_config() do {} while (0)
493 #endif /* CONFIG_BOOT_CONFIG */
495 early_param("bootconfig", warn_bootconfig);
497 /* Change NUL term back to "=", to make "param" the whole string. */
498 static void __init repair_env_string(char *param, char *val)
501 /* param=val or param="val"? */
502 if (val == param+strlen(param)+1)
504 else if (val == param+strlen(param)+2) {
506 memmove(val-1, val, strlen(val)+1);
512 /* Anything after -- gets handed straight to init. */
513 static int __init set_init_arg(char *param, char *val,
514 const char *unused, void *arg)
521 repair_env_string(param, val);
523 for (i = 0; argv_init[i]; i++) {
524 if (i == MAX_INIT_ARGS) {
525 panic_later = "init";
530 argv_init[i] = param;
535 * Unknown boot options get handed to init, unless they look like
536 * unused parameters (modprobe will find them in /proc/cmdline).
538 static int __init unknown_bootoption(char *param, char *val,
539 const char *unused, void *arg)
541 size_t len = strlen(param);
543 repair_env_string(param, val);
545 /* Handle obsolete-style parameters */
546 if (obsolete_checksetup(param))
549 /* Unused module parameter. */
550 if (strnchr(param, len, '.'))
557 /* Environment option */
559 for (i = 0; envp_init[i]; i++) {
560 if (i == MAX_INIT_ENVS) {
564 if (!strncmp(param, envp_init[i], len+1))
567 envp_init[i] = param;
569 /* Command line option */
571 for (i = 0; argv_init[i]; i++) {
572 if (i == MAX_INIT_ARGS) {
573 panic_later = "init";
577 argv_init[i] = param;
582 static int __init init_setup(char *str)
586 execute_command = str;
588 * In case LILO is going to boot us with default command line,
589 * it prepends "auto" before the whole cmdline which makes
590 * the shell think it should execute a script with such name.
591 * So we ignore all arguments entered _before_ init=... [MJ]
593 for (i = 1; i < MAX_INIT_ARGS; i++)
597 __setup("init=", init_setup);
599 static int __init rdinit_setup(char *str)
603 ramdisk_execute_command = str;
604 /* See "auto" comment in init_setup */
605 for (i = 1; i < MAX_INIT_ARGS; i++)
609 __setup("rdinit=", rdinit_setup);
612 static const unsigned int setup_max_cpus = NR_CPUS;
613 static inline void setup_nr_cpu_ids(void) { }
614 static inline void smp_prepare_cpus(unsigned int maxcpus) { }
618 * We need to store the untouched command line for future reference.
619 * We also need to store the touched command line since the parameter
620 * parsing is performed in place, and we should allow a component to
621 * store reference of name/value for future reference.
623 static void __init setup_command_line(char *command_line)
625 size_t len, xlen = 0, ilen = 0;
627 if (extra_command_line)
628 xlen = strlen(extra_command_line);
630 ilen = strlen(extra_init_args) + 4; /* for " -- " */
632 len = xlen + strlen(boot_command_line) + 1;
634 saved_command_line = memblock_alloc(len + ilen, SMP_CACHE_BYTES);
635 if (!saved_command_line)
636 panic("%s: Failed to allocate %zu bytes\n", __func__, len + ilen);
638 static_command_line = memblock_alloc(len, SMP_CACHE_BYTES);
639 if (!static_command_line)
640 panic("%s: Failed to allocate %zu bytes\n", __func__, len);
644 * We have to put extra_command_line before boot command
645 * lines because there could be dashes (separator of init
646 * command line) in the command lines.
648 strcpy(saved_command_line, extra_command_line);
649 strcpy(static_command_line, extra_command_line);
651 strcpy(saved_command_line + xlen, boot_command_line);
652 strcpy(static_command_line + xlen, command_line);
656 * Append supplemental init boot args to saved_command_line
657 * so that user can check what command line options passed
659 * The order should always be
660 * " -- "[bootconfig init-param][cmdline init-param]
663 len = xlen + initargs_offs;
664 strcpy(saved_command_line + len, extra_init_args);
665 len += ilen - 4; /* strlen(extra_init_args) */
666 strcpy(saved_command_line + len,
667 boot_command_line + initargs_offs - 1);
669 len = strlen(saved_command_line);
670 strcpy(saved_command_line + len, " -- ");
672 strcpy(saved_command_line + len, extra_init_args);
676 saved_command_line_len = strlen(saved_command_line);
680 * We need to finalize in a non-__init function or else race conditions
681 * between the root thread and the init thread may cause start_kernel to
682 * be reaped by free_initmem before the root thread has proceeded to
685 * gcc-3.4 accidentally inlines this function, so use noinline.
688 static __initdata DECLARE_COMPLETION(kthreadd_done);
690 noinline void __ref rest_init(void)
692 struct task_struct *tsk;
695 rcu_scheduler_starting();
697 * We need to spawn init first so that it obtains pid 1, however
698 * the init task will end up wanting to create kthreads, which, if
699 * we schedule it before we create kthreadd, will OOPS.
701 pid = user_mode_thread(kernel_init, NULL, CLONE_FS);
703 * Pin init on the boot CPU. Task migration is not properly working
704 * until sched_init_smp() has been run. It will set the allowed
705 * CPUs for init to the non isolated CPUs.
708 tsk = find_task_by_pid_ns(pid, &init_pid_ns);
709 tsk->flags |= PF_NO_SETAFFINITY;
710 set_cpus_allowed_ptr(tsk, cpumask_of(smp_processor_id()));
713 numa_default_policy();
714 pid = kernel_thread(kthreadd, NULL, NULL, CLONE_FS | CLONE_FILES);
716 kthreadd_task = find_task_by_pid_ns(pid, &init_pid_ns);
720 * Enable might_sleep() and smp_processor_id() checks.
721 * They cannot be enabled earlier because with CONFIG_PREEMPTION=y
722 * kernel_thread() would trigger might_sleep() splats. With
723 * CONFIG_PREEMPT_VOLUNTARY=y the init task might have scheduled
724 * already, but it's stuck on the kthreadd_done completion.
726 system_state = SYSTEM_SCHEDULING;
728 complete(&kthreadd_done);
731 * The boot idle thread must execute schedule()
732 * at least once to get things moving:
734 schedule_preempt_disabled();
735 /* Call into cpu_idle with preempt disabled */
736 cpu_startup_entry(CPUHP_ONLINE);
739 /* Check for early params. */
740 static int __init do_early_param(char *param, char *val,
741 const char *unused, void *arg)
743 const struct obs_kernel_param *p;
745 for (p = __setup_start; p < __setup_end; p++) {
746 if ((p->early && parameq(param, p->str)) ||
747 (strcmp(param, "console") == 0 &&
748 strcmp(p->str, "earlycon") == 0)
750 if (p->setup_func(val) != 0)
751 pr_warn("Malformed early option '%s'\n", param);
754 /* We accept everything at this stage. */
758 void __init parse_early_options(char *cmdline)
760 parse_args("early options", cmdline, NULL, 0, 0, 0, NULL,
764 /* Arch code calls this early on, or if not, just before other parsing. */
765 void __init parse_early_param(void)
767 static int done __initdata;
768 static char tmp_cmdline[COMMAND_LINE_SIZE] __initdata;
773 /* All fall through to do_early_param. */
774 strscpy(tmp_cmdline, boot_command_line, COMMAND_LINE_SIZE);
775 parse_early_options(tmp_cmdline);
779 void __init __weak arch_post_acpi_subsys_init(void) { }
781 void __init __weak smp_setup_processor_id(void)
785 # if THREAD_SIZE >= PAGE_SIZE
786 void __init __weak thread_stack_cache_init(void)
791 void __init __weak mem_encrypt_init(void) { }
793 void __init __weak poking_init(void) { }
795 void __init __weak pgtable_cache_init(void) { }
797 void __init __weak trap_init(void) { }
800 core_param(initcall_debug, initcall_debug, bool, 0644);
802 #ifdef TRACEPOINTS_ENABLED
803 static void __init initcall_debug_enable(void);
805 static inline void initcall_debug_enable(void)
810 /* Report memory auto-initialization states for this boot. */
811 static void __init report_meminit(void)
815 if (IS_ENABLED(CONFIG_INIT_STACK_ALL_PATTERN))
816 stack = "all(pattern)";
817 else if (IS_ENABLED(CONFIG_INIT_STACK_ALL_ZERO))
819 else if (IS_ENABLED(CONFIG_GCC_PLUGIN_STRUCTLEAK_BYREF_ALL))
820 stack = "byref_all(zero)";
821 else if (IS_ENABLED(CONFIG_GCC_PLUGIN_STRUCTLEAK_BYREF))
822 stack = "byref(zero)";
823 else if (IS_ENABLED(CONFIG_GCC_PLUGIN_STRUCTLEAK_USER))
824 stack = "__user(zero)";
828 pr_info("mem auto-init: stack:%s, heap alloc:%s, heap free:%s\n",
829 stack, want_init_on_alloc(GFP_KERNEL) ? "on" : "off",
830 want_init_on_free() ? "on" : "off");
831 if (want_init_on_free())
832 pr_info("mem auto-init: clearing system memory may take some time...\n");
836 * Set up kernel memory allocators
838 static void __init mm_init(void)
841 * page_ext requires contiguous pages,
842 * bigger than MAX_ORDER unless SPARSEMEM.
844 page_ext_init_flatmem();
845 init_mem_debugging_and_hardening();
849 stack_depot_early_init();
851 mem_init_print_info();
854 * page_owner must be initialized after buddy is ready, and also after
855 * slab is ready so that stack_depot_init() works properly
857 page_ext_init_flatmem_late();
860 debug_objects_mem_init();
862 /* If no deferred init page_ext now, as vmap is fully initialized */
863 if (!deferred_struct_pages)
865 /* Should be run before the first non-init thread is created */
867 /* Should be run after espfix64 is set up. */
869 kmsan_init_runtime();
873 #ifdef CONFIG_RANDOMIZE_KSTACK_OFFSET
874 DEFINE_STATIC_KEY_MAYBE_RO(CONFIG_RANDOMIZE_KSTACK_OFFSET_DEFAULT,
875 randomize_kstack_offset);
876 DEFINE_PER_CPU(u32, kstack_offset);
878 static int __init early_randomize_kstack_offset(char *buf)
883 ret = kstrtobool(buf, &bool_result);
888 static_branch_enable(&randomize_kstack_offset);
890 static_branch_disable(&randomize_kstack_offset);
893 early_param("randomize_kstack_offset", early_randomize_kstack_offset);
896 void __init __weak arch_call_rest_init(void)
901 static void __init print_unknown_bootoptions(void)
903 char *unknown_options;
905 const char *const *p;
908 if (panic_later || (!argv_init[1] && !envp_init[2]))
912 * Determine how many options we have to print out, plus a space
915 len = 1; /* null terminator */
916 for (p = &argv_init[1]; *p; p++) {
920 for (p = &envp_init[2]; *p; p++) {
925 unknown_options = memblock_alloc(len, SMP_CACHE_BYTES);
926 if (!unknown_options) {
927 pr_err("%s: Failed to allocate %zu bytes\n",
931 end = unknown_options;
933 for (p = &argv_init[1]; *p; p++)
934 end += sprintf(end, " %s", *p);
935 for (p = &envp_init[2]; *p; p++)
936 end += sprintf(end, " %s", *p);
938 /* Start at unknown_options[1] to skip the initial space */
939 pr_notice("Unknown kernel command line parameters \"%s\", will be passed to user space.\n",
940 &unknown_options[1]);
941 memblock_free(unknown_options, len);
944 asmlinkage __visible void __init __no_sanitize_address start_kernel(void)
949 set_task_stack_end_magic(&init_task);
950 smp_setup_processor_id();
951 debug_objects_early_init();
952 init_vmlinux_build_id();
957 early_boot_irqs_disabled = true;
960 * Interrupts are still disabled. Do necessary setups, then
965 pr_notice("%s", linux_banner);
966 early_security_init();
967 setup_arch(&command_line);
969 setup_command_line(command_line);
971 setup_per_cpu_areas();
972 smp_prepare_boot_cpu(); /* arch-specific boot-cpu hooks */
973 boot_cpu_hotplug_init();
975 build_all_zonelists(NULL);
978 pr_notice("Kernel command line: %s\n", saved_command_line);
979 /* parameters may set static keys */
982 after_dashes = parse_args("Booting kernel",
983 static_command_line, __start___param,
984 __stop___param - __start___param,
985 -1, -1, NULL, &unknown_bootoption);
986 print_unknown_bootoptions();
987 if (!IS_ERR_OR_NULL(after_dashes))
988 parse_args("Setting init args", after_dashes, NULL, 0, -1, -1,
991 parse_args("Setting extra init args", extra_init_args,
992 NULL, 0, -1, -1, NULL, set_init_arg);
994 /* Architectural and non-timekeeping rng init, before allocator init */
995 random_init_early(command_line);
998 * These use large bootmem allocations and must precede
1002 vfs_caches_init_early();
1003 sort_main_extable();
1009 /* trace_printk can be enabled here */
1013 * Set up the scheduler prior starting any interrupts (such as the
1014 * timer interrupt). Full topology setup happens at smp_init()
1015 * time - but meanwhile we still have a functioning scheduler.
1019 if (WARN(!irqs_disabled(),
1020 "Interrupts were enabled *very* early, fixing it\n"))
1021 local_irq_disable();
1026 * Set up housekeeping before setting up workqueues to allow the unbound
1027 * workqueue to take non-housekeeping into account.
1029 housekeeping_init();
1032 * Allow workqueue creation and work item queueing/cancelling
1033 * early. Work item execution depends on kthreads and starts after
1036 workqueue_init_early();
1040 /* Trace events are available after this */
1044 initcall_debug_enable();
1046 context_tracking_init();
1047 /* init some links before init_ISA_irqs() */
1059 /* This must be after timekeeping is initialized */
1062 /* These make use of the fully initialized rng */
1064 boot_init_stack_canary();
1068 call_function_init();
1069 WARN(!irqs_disabled(), "Interrupts were enabled early\n");
1071 early_boot_irqs_disabled = false;
1074 kmem_cache_init_late();
1077 * HACK ALERT! This is early. We're enabling the console before
1078 * we've done PCI setups etc, and console_init() must be aware of
1079 * this. But we do want output early, in case something goes wrong.
1083 panic("Too many boot %s vars at `%s'", panic_later,
1089 * Need to run this when irqs are enabled, because it wants
1090 * to self-test [hard/soft]-irqs on/off lock inversion bugs
1095 #ifdef CONFIG_BLK_DEV_INITRD
1096 if (initrd_start && !initrd_below_start_ok &&
1097 page_to_pfn(virt_to_page((void *)initrd_start)) < min_low_pfn) {
1098 pr_crit("initrd overwritten (0x%08lx < 0x%08lx) - disabling it.\n",
1099 page_to_pfn(virt_to_page((void *)initrd_start)),
1104 setup_per_cpu_pageset();
1113 * This needs to be called before any devices perform DMA
1114 * operations that might use the SWIOTLB bounce buffers. It will
1115 * mark the bounce buffers as decrypted so that their usage will
1116 * not cause "plain-text" data to be decrypted when accessed. It
1117 * must be called after late_time_init() so that Hyper-V x86/x64
1118 * hypercalls work when the SWIOTLB bounce buffers are decrypted.
1125 if (efi_enabled(EFI_RUNTIME_SERVICES))
1126 efi_enter_virtual_mode();
1128 thread_stack_cache_init();
1145 taskstats_init_early();
1150 acpi_subsystem_init();
1151 arch_post_acpi_subsys_init();
1154 /* Do the rest non-__init'ed, we're now alive */
1155 arch_call_rest_init();
1157 prevent_tail_call_optimization();
1160 /* Call all constructor functions linked into the kernel. */
1161 static void __init do_ctors(void)
1164 * For UML, the constructors have already been called by the
1165 * normal setup code as it's just a normal ELF binary, so we
1166 * cannot do it again - but we do need CONFIG_CONSTRUCTORS
1167 * even on UML for modules.
1169 #if defined(CONFIG_CONSTRUCTORS) && !defined(CONFIG_UML)
1170 ctor_fn_t *fn = (ctor_fn_t *) __ctors_start;
1172 for (; fn < (ctor_fn_t *) __ctors_end; fn++)
1177 #ifdef CONFIG_KALLSYMS
1178 struct blacklist_entry {
1179 struct list_head next;
1183 static __initdata_or_module LIST_HEAD(blacklisted_initcalls);
1185 static int __init initcall_blacklist(char *str)
1188 struct blacklist_entry *entry;
1190 /* str argument is a comma-separated list of functions */
1192 str_entry = strsep(&str, ",");
1194 pr_debug("blacklisting initcall %s\n", str_entry);
1195 entry = memblock_alloc(sizeof(*entry),
1198 panic("%s: Failed to allocate %zu bytes\n",
1199 __func__, sizeof(*entry));
1200 entry->buf = memblock_alloc(strlen(str_entry) + 1,
1203 panic("%s: Failed to allocate %zu bytes\n",
1204 __func__, strlen(str_entry) + 1);
1205 strcpy(entry->buf, str_entry);
1206 list_add(&entry->next, &blacklisted_initcalls);
1208 } while (str_entry);
1213 static bool __init_or_module initcall_blacklisted(initcall_t fn)
1215 struct blacklist_entry *entry;
1216 char fn_name[KSYM_SYMBOL_LEN];
1219 if (list_empty(&blacklisted_initcalls))
1222 addr = (unsigned long) dereference_function_descriptor(fn);
1223 sprint_symbol_no_offset(fn_name, addr);
1226 * fn will be "function_name [module_name]" where [module_name] is not
1227 * displayed for built-in init functions. Strip off the [module_name].
1229 strreplace(fn_name, ' ', '\0');
1231 list_for_each_entry(entry, &blacklisted_initcalls, next) {
1232 if (!strcmp(fn_name, entry->buf)) {
1233 pr_debug("initcall %s blacklisted\n", fn_name);
1241 static int __init initcall_blacklist(char *str)
1243 pr_warn("initcall_blacklist requires CONFIG_KALLSYMS\n");
1247 static bool __init_or_module initcall_blacklisted(initcall_t fn)
1252 __setup("initcall_blacklist=", initcall_blacklist);
1254 static __init_or_module void
1255 trace_initcall_start_cb(void *data, initcall_t fn)
1257 ktime_t *calltime = data;
1259 printk(KERN_DEBUG "calling %pS @ %i\n", fn, task_pid_nr(current));
1260 *calltime = ktime_get();
1263 static __init_or_module void
1264 trace_initcall_finish_cb(void *data, initcall_t fn, int ret)
1266 ktime_t rettime, *calltime = data;
1268 rettime = ktime_get();
1269 printk(KERN_DEBUG "initcall %pS returned %d after %lld usecs\n",
1270 fn, ret, (unsigned long long)ktime_us_delta(rettime, *calltime));
1273 static ktime_t initcall_calltime;
1275 #ifdef TRACEPOINTS_ENABLED
1276 static void __init initcall_debug_enable(void)
1280 ret = register_trace_initcall_start(trace_initcall_start_cb,
1281 &initcall_calltime);
1282 ret |= register_trace_initcall_finish(trace_initcall_finish_cb,
1283 &initcall_calltime);
1284 WARN(ret, "Failed to register initcall tracepoints\n");
1286 # define do_trace_initcall_start trace_initcall_start
1287 # define do_trace_initcall_finish trace_initcall_finish
1289 static inline void do_trace_initcall_start(initcall_t fn)
1291 if (!initcall_debug)
1293 trace_initcall_start_cb(&initcall_calltime, fn);
1295 static inline void do_trace_initcall_finish(initcall_t fn, int ret)
1297 if (!initcall_debug)
1299 trace_initcall_finish_cb(&initcall_calltime, fn, ret);
1301 #endif /* !TRACEPOINTS_ENABLED */
1303 int __init_or_module do_one_initcall(initcall_t fn)
1305 int count = preempt_count();
1309 if (initcall_blacklisted(fn))
1312 do_trace_initcall_start(fn);
1314 do_trace_initcall_finish(fn, ret);
1318 if (preempt_count() != count) {
1319 sprintf(msgbuf, "preemption imbalance ");
1320 preempt_count_set(count);
1322 if (irqs_disabled()) {
1323 strlcat(msgbuf, "disabled interrupts ", sizeof(msgbuf));
1326 WARN(msgbuf[0], "initcall %pS returned with %s\n", fn, msgbuf);
1328 add_latent_entropy();
1333 extern initcall_entry_t __initcall_start[];
1334 extern initcall_entry_t __initcall0_start[];
1335 extern initcall_entry_t __initcall1_start[];
1336 extern initcall_entry_t __initcall2_start[];
1337 extern initcall_entry_t __initcall3_start[];
1338 extern initcall_entry_t __initcall4_start[];
1339 extern initcall_entry_t __initcall5_start[];
1340 extern initcall_entry_t __initcall6_start[];
1341 extern initcall_entry_t __initcall7_start[];
1342 extern initcall_entry_t __initcall_end[];
1344 static initcall_entry_t *initcall_levels[] __initdata = {
1356 /* Keep these in sync with initcalls in include/linux/init.h */
1357 static const char *initcall_level_names[] __initdata = {
1368 static int __init ignore_unknown_bootoption(char *param, char *val,
1369 const char *unused, void *arg)
1374 static void __init do_initcall_level(int level, char *command_line)
1376 initcall_entry_t *fn;
1378 parse_args(initcall_level_names[level],
1379 command_line, __start___param,
1380 __stop___param - __start___param,
1382 NULL, ignore_unknown_bootoption);
1384 trace_initcall_level(initcall_level_names[level]);
1385 for (fn = initcall_levels[level]; fn < initcall_levels[level+1]; fn++)
1386 do_one_initcall(initcall_from_entry(fn));
1389 static void __init do_initcalls(void)
1392 size_t len = saved_command_line_len + 1;
1395 command_line = kzalloc(len, GFP_KERNEL);
1397 panic("%s: Failed to allocate %zu bytes\n", __func__, len);
1399 for (level = 0; level < ARRAY_SIZE(initcall_levels) - 1; level++) {
1400 /* Parser modifies command_line, restore it each time */
1401 strcpy(command_line, saved_command_line);
1402 do_initcall_level(level, command_line);
1405 kfree(command_line);
1409 * Ok, the machine is now initialized. None of the devices
1410 * have been touched yet, but the CPU subsystem is up and
1411 * running, and memory and process management works.
1413 * Now we can finally start doing some real work..
1415 static void __init do_basic_setup(void)
1424 static void __init do_pre_smp_initcalls(void)
1426 initcall_entry_t *fn;
1428 trace_initcall_level("early");
1429 for (fn = __initcall_start; fn < __initcall0_start; fn++)
1430 do_one_initcall(initcall_from_entry(fn));
1433 static int run_init_process(const char *init_filename)
1435 const char *const *p;
1437 argv_init[0] = init_filename;
1438 pr_info("Run %s as init process\n", init_filename);
1439 pr_debug(" with arguments:\n");
1440 for (p = argv_init; *p; p++)
1441 pr_debug(" %s\n", *p);
1442 pr_debug(" with environment:\n");
1443 for (p = envp_init; *p; p++)
1444 pr_debug(" %s\n", *p);
1445 return kernel_execve(init_filename, argv_init, envp_init);
1448 static int try_to_run_init_process(const char *init_filename)
1452 ret = run_init_process(init_filename);
1454 if (ret && ret != -ENOENT) {
1455 pr_err("Starting init: %s exists but couldn't execute it (error %d)\n",
1456 init_filename, ret);
1462 static noinline void __init kernel_init_freeable(void);
1464 #if defined(CONFIG_STRICT_KERNEL_RWX) || defined(CONFIG_STRICT_MODULE_RWX)
1465 bool rodata_enabled __ro_after_init = true;
1467 #ifndef arch_parse_debug_rodata
1468 static inline bool arch_parse_debug_rodata(char *str) { return false; }
1471 static int __init set_debug_rodata(char *str)
1473 if (arch_parse_debug_rodata(str))
1476 if (str && !strcmp(str, "on"))
1477 rodata_enabled = true;
1478 else if (str && !strcmp(str, "off"))
1479 rodata_enabled = false;
1481 pr_warn("Invalid option string for rodata: '%s'\n", str);
1484 early_param("rodata", set_debug_rodata);
1487 #ifdef CONFIG_STRICT_KERNEL_RWX
1488 static void mark_readonly(void)
1490 if (rodata_enabled) {
1492 * load_module() results in W+X mappings, which are cleaned
1493 * up with call_rcu(). Let's make sure that queued work is
1494 * flushed so that we don't hit false positives looking for
1495 * insecure pages which are W+X.
1501 pr_info("Kernel memory protection disabled.\n");
1503 #elif defined(CONFIG_ARCH_HAS_STRICT_KERNEL_RWX)
1504 static inline void mark_readonly(void)
1506 pr_warn("Kernel memory protection not selected by kernel config.\n");
1509 static inline void mark_readonly(void)
1511 pr_warn("This architecture does not have kernel memory protection.\n");
1515 void __weak free_initmem(void)
1517 free_initmem_default(POISON_FREE_INITMEM);
1520 static int __ref kernel_init(void *unused)
1525 * Wait until kthreadd is all set-up.
1527 wait_for_completion(&kthreadd_done);
1529 kernel_init_freeable();
1530 /* need to finish all async __init code before freeing the memory */
1531 async_synchronize_full();
1533 system_state = SYSTEM_FREEING_INITMEM;
1534 kprobe_free_init_mem();
1535 ftrace_free_init_mem();
1536 kgdb_free_init_mem();
1542 * Kernel mappings are now finalized - update the userspace page-table
1547 system_state = SYSTEM_RUNNING;
1548 numa_default_policy();
1550 rcu_end_inkernel_boot();
1554 if (ramdisk_execute_command) {
1555 ret = run_init_process(ramdisk_execute_command);
1558 pr_err("Failed to execute %s (error %d)\n",
1559 ramdisk_execute_command, ret);
1563 * We try each of these until one succeeds.
1565 * The Bourne shell can be used instead of init if we are
1566 * trying to recover a really broken machine.
1568 if (execute_command) {
1569 ret = run_init_process(execute_command);
1572 panic("Requested init %s failed (error %d).",
1573 execute_command, ret);
1576 if (CONFIG_DEFAULT_INIT[0] != '\0') {
1577 ret = run_init_process(CONFIG_DEFAULT_INIT);
1579 pr_err("Default init %s failed (error %d)\n",
1580 CONFIG_DEFAULT_INIT, ret);
1585 if (!try_to_run_init_process("/sbin/init") ||
1586 !try_to_run_init_process("/etc/init") ||
1587 !try_to_run_init_process("/bin/init") ||
1588 !try_to_run_init_process("/bin/sh"))
1591 panic("No working init found. Try passing init= option to kernel. "
1592 "See Linux Documentation/admin-guide/init.rst for guidance.");
1595 /* Open /dev/console, for stdin/stdout/stderr, this should never fail */
1596 void __init console_on_rootfs(void)
1598 struct file *file = filp_open("/dev/console", O_RDWR, 0);
1601 pr_err("Warning: unable to open an initial console.\n");
1610 static noinline void __init kernel_init_freeable(void)
1612 /* Now the scheduler is fully set up and can do blocking allocations */
1613 gfp_allowed_mask = __GFP_BITS_MASK;
1616 * init can allocate pages on any node
1618 set_mems_allowed(node_states[N_MEMORY]);
1620 cad_pid = get_pid(task_pid(current));
1622 smp_prepare_cpus(setup_max_cpus);
1626 init_mm_internals();
1628 rcu_init_tasks_generic();
1629 do_pre_smp_initcalls();
1630 lockup_detector_init();
1636 page_alloc_init_late();
1637 /* Initialize page ext after all struct pages are initialized. */
1638 if (deferred_struct_pages)
1643 kunit_run_all_tests();
1645 wait_for_initramfs();
1646 console_on_rootfs();
1649 * check if there is an early userspace init. If yes, let it do all
1652 if (init_eaccess(ramdisk_execute_command) != 0) {
1653 ramdisk_execute_command = NULL;
1654 prepare_namespace();
1658 * Ok, we have completed the initial bootup, and
1659 * we're essentially up and running. Get rid of the
1660 * initmem segments and start the user-mode stuff..
1662 * rootfs is available now, try loading the public keys
1663 * and default modules
1666 integrity_load_keys();