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/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 size_t initargs_offs;
157 # define bootconfig_found false
158 # define initargs_offs 0
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");
384 memblock_free(new_cmdline, len + 1);
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;
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 ret = xbc_init(data, size, &msg, &pos);
448 pr_err("Failed to init bootconfig: %s.\n", msg);
450 pr_err("Failed to parse bootconfig: %s at %d.\n",
453 xbc_get_info(&ret, NULL);
454 pr_info("Load bootconfig: %d bytes %d nodes\n", size, ret);
455 /* keys starting with "kernel." are passed via cmdline */
456 extra_command_line = xbc_make_cmdline("kernel");
457 /* Also, "init." keys are init arguments */
458 extra_init_args = xbc_make_cmdline("init");
463 static void __init exit_boot_config(void)
468 #else /* !CONFIG_BOOT_CONFIG */
470 static void __init setup_boot_config(void)
472 /* Remove bootconfig data from initrd */
473 get_boot_config_from_initrd(NULL, NULL);
476 static int __init warn_bootconfig(char *str)
478 pr_warn("WARNING: 'bootconfig' found on the kernel command line but CONFIG_BOOT_CONFIG is not set.\n");
482 #define exit_boot_config() do {} while (0)
484 #endif /* CONFIG_BOOT_CONFIG */
486 early_param("bootconfig", warn_bootconfig);
488 /* Change NUL term back to "=", to make "param" the whole string. */
489 static void __init repair_env_string(char *param, char *val)
492 /* param=val or param="val"? */
493 if (val == param+strlen(param)+1)
495 else if (val == param+strlen(param)+2) {
497 memmove(val-1, val, strlen(val)+1);
503 /* Anything after -- gets handed straight to init. */
504 static int __init set_init_arg(char *param, char *val,
505 const char *unused, void *arg)
512 repair_env_string(param, val);
514 for (i = 0; argv_init[i]; i++) {
515 if (i == MAX_INIT_ARGS) {
516 panic_later = "init";
521 argv_init[i] = param;
526 * Unknown boot options get handed to init, unless they look like
527 * unused parameters (modprobe will find them in /proc/cmdline).
529 static int __init unknown_bootoption(char *param, char *val,
530 const char *unused, void *arg)
532 size_t len = strlen(param);
534 repair_env_string(param, val);
536 /* Handle obsolete-style parameters */
537 if (obsolete_checksetup(param))
540 /* Unused module parameter. */
541 if (strnchr(param, len, '.'))
548 /* Environment option */
550 for (i = 0; envp_init[i]; i++) {
551 if (i == MAX_INIT_ENVS) {
555 if (!strncmp(param, envp_init[i], len+1))
558 envp_init[i] = param;
560 /* Command line option */
562 for (i = 0; argv_init[i]; i++) {
563 if (i == MAX_INIT_ARGS) {
564 panic_later = "init";
568 argv_init[i] = param;
573 static int __init init_setup(char *str)
577 execute_command = str;
579 * In case LILO is going to boot us with default command line,
580 * it prepends "auto" before the whole cmdline which makes
581 * the shell think it should execute a script with such name.
582 * So we ignore all arguments entered _before_ init=... [MJ]
584 for (i = 1; i < MAX_INIT_ARGS; i++)
588 __setup("init=", init_setup);
590 static int __init rdinit_setup(char *str)
594 ramdisk_execute_command = str;
595 /* See "auto" comment in init_setup */
596 for (i = 1; i < MAX_INIT_ARGS; i++)
600 __setup("rdinit=", rdinit_setup);
603 static const unsigned int setup_max_cpus = NR_CPUS;
604 static inline void setup_nr_cpu_ids(void) { }
605 static inline void smp_prepare_cpus(unsigned int maxcpus) { }
609 * We need to store the untouched command line for future reference.
610 * We also need to store the touched command line since the parameter
611 * parsing is performed in place, and we should allow a component to
612 * store reference of name/value for future reference.
614 static void __init setup_command_line(char *command_line)
616 size_t len, xlen = 0, ilen = 0;
618 if (extra_command_line)
619 xlen = strlen(extra_command_line);
621 ilen = strlen(extra_init_args) + 4; /* for " -- " */
623 len = xlen + strlen(boot_command_line) + 1;
625 saved_command_line = memblock_alloc(len + ilen, SMP_CACHE_BYTES);
626 if (!saved_command_line)
627 panic("%s: Failed to allocate %zu bytes\n", __func__, len + ilen);
629 static_command_line = memblock_alloc(len, SMP_CACHE_BYTES);
630 if (!static_command_line)
631 panic("%s: Failed to allocate %zu bytes\n", __func__, len);
635 * We have to put extra_command_line before boot command
636 * lines because there could be dashes (separator of init
637 * command line) in the command lines.
639 strcpy(saved_command_line, extra_command_line);
640 strcpy(static_command_line, extra_command_line);
642 strcpy(saved_command_line + xlen, boot_command_line);
643 strcpy(static_command_line + xlen, command_line);
647 * Append supplemental init boot args to saved_command_line
648 * so that user can check what command line options passed
650 * The order should always be
651 * " -- "[bootconfig init-param][cmdline init-param]
654 len = xlen + initargs_offs;
655 strcpy(saved_command_line + len, extra_init_args);
656 len += ilen - 4; /* strlen(extra_init_args) */
657 strcpy(saved_command_line + len,
658 boot_command_line + initargs_offs - 1);
660 len = strlen(saved_command_line);
661 strcpy(saved_command_line + len, " -- ");
663 strcpy(saved_command_line + len, extra_init_args);
669 * We need to finalize in a non-__init function or else race conditions
670 * between the root thread and the init thread may cause start_kernel to
671 * be reaped by free_initmem before the root thread has proceeded to
674 * gcc-3.4 accidentally inlines this function, so use noinline.
677 static __initdata DECLARE_COMPLETION(kthreadd_done);
679 noinline void __ref rest_init(void)
681 struct task_struct *tsk;
684 rcu_scheduler_starting();
686 * We need to spawn init first so that it obtains pid 1, however
687 * the init task will end up wanting to create kthreads, which, if
688 * we schedule it before we create kthreadd, will OOPS.
690 pid = kernel_thread(kernel_init, NULL, CLONE_FS);
692 * Pin init on the boot CPU. Task migration is not properly working
693 * until sched_init_smp() has been run. It will set the allowed
694 * CPUs for init to the non isolated CPUs.
697 tsk = find_task_by_pid_ns(pid, &init_pid_ns);
698 tsk->flags |= PF_NO_SETAFFINITY;
699 set_cpus_allowed_ptr(tsk, cpumask_of(smp_processor_id()));
702 numa_default_policy();
703 pid = kernel_thread(kthreadd, NULL, CLONE_FS | CLONE_FILES);
705 kthreadd_task = find_task_by_pid_ns(pid, &init_pid_ns);
709 * Enable might_sleep() and smp_processor_id() checks.
710 * They cannot be enabled earlier because with CONFIG_PREEMPTION=y
711 * kernel_thread() would trigger might_sleep() splats. With
712 * CONFIG_PREEMPT_VOLUNTARY=y the init task might have scheduled
713 * already, but it's stuck on the kthreadd_done completion.
715 system_state = SYSTEM_SCHEDULING;
717 complete(&kthreadd_done);
720 * The boot idle thread must execute schedule()
721 * at least once to get things moving:
723 schedule_preempt_disabled();
724 /* Call into cpu_idle with preempt disabled */
725 cpu_startup_entry(CPUHP_ONLINE);
728 /* Check for early params. */
729 static int __init do_early_param(char *param, char *val,
730 const char *unused, void *arg)
732 const struct obs_kernel_param *p;
734 for (p = __setup_start; p < __setup_end; p++) {
735 if ((p->early && parameq(param, p->str)) ||
736 (strcmp(param, "console") == 0 &&
737 strcmp(p->str, "earlycon") == 0)
739 if (p->setup_func(val) != 0)
740 pr_warn("Malformed early option '%s'\n", param);
743 /* We accept everything at this stage. */
747 void __init parse_early_options(char *cmdline)
749 parse_args("early options", cmdline, NULL, 0, 0, 0, NULL,
753 /* Arch code calls this early on, or if not, just before other parsing. */
754 void __init parse_early_param(void)
756 static int done __initdata;
757 static char tmp_cmdline[COMMAND_LINE_SIZE] __initdata;
762 /* All fall through to do_early_param. */
763 strlcpy(tmp_cmdline, boot_command_line, COMMAND_LINE_SIZE);
764 parse_early_options(tmp_cmdline);
768 void __init __weak arch_post_acpi_subsys_init(void) { }
770 void __init __weak smp_setup_processor_id(void)
774 # if THREAD_SIZE >= PAGE_SIZE
775 void __init __weak thread_stack_cache_init(void)
780 void __init __weak mem_encrypt_init(void) { }
782 void __init __weak poking_init(void) { }
784 void __init __weak pgtable_cache_init(void) { }
786 void __init __weak trap_init(void) { }
789 core_param(initcall_debug, initcall_debug, bool, 0644);
791 #ifdef TRACEPOINTS_ENABLED
792 static void __init initcall_debug_enable(void);
794 static inline void initcall_debug_enable(void)
799 /* Report memory auto-initialization states for this boot. */
800 static void __init report_meminit(void)
804 if (IS_ENABLED(CONFIG_INIT_STACK_ALL_PATTERN))
805 stack = "all(pattern)";
806 else if (IS_ENABLED(CONFIG_INIT_STACK_ALL_ZERO))
808 else if (IS_ENABLED(CONFIG_GCC_PLUGIN_STRUCTLEAK_BYREF_ALL))
809 stack = "byref_all(zero)";
810 else if (IS_ENABLED(CONFIG_GCC_PLUGIN_STRUCTLEAK_BYREF))
811 stack = "byref(zero)";
812 else if (IS_ENABLED(CONFIG_GCC_PLUGIN_STRUCTLEAK_USER))
813 stack = "__user(zero)";
817 pr_info("mem auto-init: stack:%s, heap alloc:%s, heap free:%s\n",
818 stack, want_init_on_alloc(GFP_KERNEL) ? "on" : "off",
819 want_init_on_free() ? "on" : "off");
820 if (want_init_on_free())
821 pr_info("mem auto-init: clearing system memory may take some time...\n");
825 * Set up kernel memory allocators
827 static void __init mm_init(void)
830 * page_ext requires contiguous pages,
831 * bigger than MAX_ORDER unless SPARSEMEM.
833 page_ext_init_flatmem();
834 init_mem_debugging_and_hardening();
839 mem_init_print_info();
840 /* page_owner must be initialized after buddy is ready */
841 page_ext_init_flatmem_late();
845 debug_objects_mem_init();
847 /* Should be run before the first non-init thread is created */
849 /* Should be run after espfix64 is set up. */
853 #ifdef CONFIG_HAVE_ARCH_RANDOMIZE_KSTACK_OFFSET
854 DEFINE_STATIC_KEY_MAYBE_RO(CONFIG_RANDOMIZE_KSTACK_OFFSET_DEFAULT,
855 randomize_kstack_offset);
856 DEFINE_PER_CPU(u32, kstack_offset);
858 static int __init early_randomize_kstack_offset(char *buf)
863 ret = kstrtobool(buf, &bool_result);
868 static_branch_enable(&randomize_kstack_offset);
870 static_branch_disable(&randomize_kstack_offset);
873 early_param("randomize_kstack_offset", early_randomize_kstack_offset);
876 void __init __weak arch_call_rest_init(void)
881 static void __init print_unknown_bootoptions(void)
883 char *unknown_options;
885 const char *const *p;
888 if (panic_later || (!argv_init[1] && !envp_init[2]))
892 * Determine how many options we have to print out, plus a space
895 len = 1; /* null terminator */
896 for (p = &argv_init[1]; *p; p++) {
900 for (p = &envp_init[2]; *p; p++) {
905 unknown_options = memblock_alloc(len, SMP_CACHE_BYTES);
906 if (!unknown_options) {
907 pr_err("%s: Failed to allocate %zu bytes\n",
911 end = unknown_options;
913 for (p = &argv_init[1]; *p; p++)
914 end += sprintf(end, " %s", *p);
915 for (p = &envp_init[2]; *p; p++)
916 end += sprintf(end, " %s", *p);
918 /* Start at unknown_options[1] to skip the initial space */
919 pr_notice("Unknown kernel command line parameters \"%s\", will be passed to user space.\n",
920 &unknown_options[1]);
921 memblock_free(unknown_options, len);
924 asmlinkage __visible void __init __no_sanitize_address start_kernel(void)
929 set_task_stack_end_magic(&init_task);
930 smp_setup_processor_id();
931 debug_objects_early_init();
932 init_vmlinux_build_id();
937 early_boot_irqs_disabled = true;
940 * Interrupts are still disabled. Do necessary setups, then
945 pr_notice("%s", linux_banner);
946 early_security_init();
947 setup_arch(&command_line);
949 setup_command_line(command_line);
951 setup_per_cpu_areas();
952 smp_prepare_boot_cpu(); /* arch-specific boot-cpu hooks */
953 boot_cpu_hotplug_init();
955 build_all_zonelists(NULL);
958 pr_notice("Kernel command line: %s\n", saved_command_line);
959 /* parameters may set static keys */
962 after_dashes = parse_args("Booting kernel",
963 static_command_line, __start___param,
964 __stop___param - __start___param,
965 -1, -1, NULL, &unknown_bootoption);
966 print_unknown_bootoptions();
967 if (!IS_ERR_OR_NULL(after_dashes))
968 parse_args("Setting init args", after_dashes, NULL, 0, -1, -1,
971 parse_args("Setting extra init args", extra_init_args,
972 NULL, 0, -1, -1, NULL, set_init_arg);
975 * These use large bootmem allocations and must precede
979 vfs_caches_init_early();
986 /* trace_printk can be enabled here */
990 * Set up the scheduler prior starting any interrupts (such as the
991 * timer interrupt). Full topology setup happens at smp_init()
992 * time - but meanwhile we still have a functioning scheduler.
996 if (WARN(!irqs_disabled(),
997 "Interrupts were enabled *very* early, fixing it\n"))
1002 * Set up housekeeping before setting up workqueues to allow the unbound
1003 * workqueue to take non-housekeeping into account.
1005 housekeeping_init();
1008 * Allow workqueue creation and work item queueing/cancelling
1009 * early. Work item execution depends on kthreads and starts after
1012 workqueue_init_early();
1016 /* Trace events are available after this */
1020 initcall_debug_enable();
1022 context_tracking_init();
1023 /* init some links before init_ISA_irqs() */
1036 * For best initial stack canary entropy, prepare it after:
1037 * - setup_arch() for any UEFI RNG entropy and boot cmdline access
1038 * - timekeeping_init() for ktime entropy used in rand_initialize()
1039 * - rand_initialize() to get any arch-specific entropy like RDRAND
1040 * - add_latent_entropy() to get any latent entropy
1041 * - adding command line entropy
1044 add_latent_entropy();
1045 add_device_randomness(command_line, strlen(command_line));
1046 boot_init_stack_canary();
1051 call_function_init();
1052 WARN(!irqs_disabled(), "Interrupts were enabled early\n");
1054 early_boot_irqs_disabled = false;
1057 kmem_cache_init_late();
1060 * HACK ALERT! This is early. We're enabling the console before
1061 * we've done PCI setups etc, and console_init() must be aware of
1062 * this. But we do want output early, in case something goes wrong.
1066 panic("Too many boot %s vars at `%s'", panic_later,
1072 * Need to run this when irqs are enabled, because it wants
1073 * to self-test [hard/soft]-irqs on/off lock inversion bugs
1079 * This needs to be called before any devices perform DMA
1080 * operations that might use the SWIOTLB bounce buffers. It will
1081 * mark the bounce buffers as decrypted so that their usage will
1082 * not cause "plain-text" data to be decrypted when accessed.
1086 #ifdef CONFIG_BLK_DEV_INITRD
1087 if (initrd_start && !initrd_below_start_ok &&
1088 page_to_pfn(virt_to_page((void *)initrd_start)) < min_low_pfn) {
1089 pr_crit("initrd overwritten (0x%08lx < 0x%08lx) - disabling it.\n",
1090 page_to_pfn(virt_to_page((void *)initrd_start)),
1095 setup_per_cpu_pageset();
1105 if (efi_enabled(EFI_RUNTIME_SERVICES))
1106 efi_enter_virtual_mode();
1108 thread_stack_cache_init();
1124 taskstats_init_early();
1130 acpi_subsystem_init();
1131 arch_post_acpi_subsys_init();
1134 /* Do the rest non-__init'ed, we're now alive */
1135 arch_call_rest_init();
1137 prevent_tail_call_optimization();
1140 /* Call all constructor functions linked into the kernel. */
1141 static void __init do_ctors(void)
1144 * For UML, the constructors have already been called by the
1145 * normal setup code as it's just a normal ELF binary, so we
1146 * cannot do it again - but we do need CONFIG_CONSTRUCTORS
1147 * even on UML for modules.
1149 #if defined(CONFIG_CONSTRUCTORS) && !defined(CONFIG_UML)
1150 ctor_fn_t *fn = (ctor_fn_t *) __ctors_start;
1152 for (; fn < (ctor_fn_t *) __ctors_end; fn++)
1157 #ifdef CONFIG_KALLSYMS
1158 struct blacklist_entry {
1159 struct list_head next;
1163 static __initdata_or_module LIST_HEAD(blacklisted_initcalls);
1165 static int __init initcall_blacklist(char *str)
1168 struct blacklist_entry *entry;
1170 /* str argument is a comma-separated list of functions */
1172 str_entry = strsep(&str, ",");
1174 pr_debug("blacklisting initcall %s\n", str_entry);
1175 entry = memblock_alloc(sizeof(*entry),
1178 panic("%s: Failed to allocate %zu bytes\n",
1179 __func__, sizeof(*entry));
1180 entry->buf = memblock_alloc(strlen(str_entry) + 1,
1183 panic("%s: Failed to allocate %zu bytes\n",
1184 __func__, strlen(str_entry) + 1);
1185 strcpy(entry->buf, str_entry);
1186 list_add(&entry->next, &blacklisted_initcalls);
1188 } while (str_entry);
1193 static bool __init_or_module initcall_blacklisted(initcall_t fn)
1195 struct blacklist_entry *entry;
1196 char fn_name[KSYM_SYMBOL_LEN];
1199 if (list_empty(&blacklisted_initcalls))
1202 addr = (unsigned long) dereference_function_descriptor(fn);
1203 sprint_symbol_no_offset(fn_name, addr);
1206 * fn will be "function_name [module_name]" where [module_name] is not
1207 * displayed for built-in init functions. Strip off the [module_name].
1209 strreplace(fn_name, ' ', '\0');
1211 list_for_each_entry(entry, &blacklisted_initcalls, next) {
1212 if (!strcmp(fn_name, entry->buf)) {
1213 pr_debug("initcall %s blacklisted\n", fn_name);
1221 static int __init initcall_blacklist(char *str)
1223 pr_warn("initcall_blacklist requires CONFIG_KALLSYMS\n");
1227 static bool __init_or_module initcall_blacklisted(initcall_t fn)
1232 __setup("initcall_blacklist=", initcall_blacklist);
1234 static __init_or_module void
1235 trace_initcall_start_cb(void *data, initcall_t fn)
1237 ktime_t *calltime = (ktime_t *)data;
1239 printk(KERN_DEBUG "calling %pS @ %i\n", fn, task_pid_nr(current));
1240 *calltime = ktime_get();
1243 static __init_or_module void
1244 trace_initcall_finish_cb(void *data, initcall_t fn, int ret)
1246 ktime_t *calltime = (ktime_t *)data;
1247 ktime_t delta, rettime;
1248 unsigned long long duration;
1250 rettime = ktime_get();
1251 delta = ktime_sub(rettime, *calltime);
1252 duration = (unsigned long long) ktime_to_ns(delta) >> 10;
1253 printk(KERN_DEBUG "initcall %pS returned %d after %lld usecs\n",
1257 static ktime_t initcall_calltime;
1259 #ifdef TRACEPOINTS_ENABLED
1260 static void __init initcall_debug_enable(void)
1264 ret = register_trace_initcall_start(trace_initcall_start_cb,
1265 &initcall_calltime);
1266 ret |= register_trace_initcall_finish(trace_initcall_finish_cb,
1267 &initcall_calltime);
1268 WARN(ret, "Failed to register initcall tracepoints\n");
1270 # define do_trace_initcall_start trace_initcall_start
1271 # define do_trace_initcall_finish trace_initcall_finish
1273 static inline void do_trace_initcall_start(initcall_t fn)
1275 if (!initcall_debug)
1277 trace_initcall_start_cb(&initcall_calltime, fn);
1279 static inline void do_trace_initcall_finish(initcall_t fn, int ret)
1281 if (!initcall_debug)
1283 trace_initcall_finish_cb(&initcall_calltime, fn, ret);
1285 #endif /* !TRACEPOINTS_ENABLED */
1287 int __init_or_module do_one_initcall(initcall_t fn)
1289 int count = preempt_count();
1293 if (initcall_blacklisted(fn))
1296 do_trace_initcall_start(fn);
1298 do_trace_initcall_finish(fn, ret);
1302 if (preempt_count() != count) {
1303 sprintf(msgbuf, "preemption imbalance ");
1304 preempt_count_set(count);
1306 if (irqs_disabled()) {
1307 strlcat(msgbuf, "disabled interrupts ", sizeof(msgbuf));
1310 WARN(msgbuf[0], "initcall %pS returned with %s\n", fn, msgbuf);
1312 add_latent_entropy();
1317 extern initcall_entry_t __initcall_start[];
1318 extern initcall_entry_t __initcall0_start[];
1319 extern initcall_entry_t __initcall1_start[];
1320 extern initcall_entry_t __initcall2_start[];
1321 extern initcall_entry_t __initcall3_start[];
1322 extern initcall_entry_t __initcall4_start[];
1323 extern initcall_entry_t __initcall5_start[];
1324 extern initcall_entry_t __initcall6_start[];
1325 extern initcall_entry_t __initcall7_start[];
1326 extern initcall_entry_t __initcall_end[];
1328 static initcall_entry_t *initcall_levels[] __initdata = {
1340 /* Keep these in sync with initcalls in include/linux/init.h */
1341 static const char *initcall_level_names[] __initdata = {
1352 static int __init ignore_unknown_bootoption(char *param, char *val,
1353 const char *unused, void *arg)
1358 static void __init do_initcall_level(int level, char *command_line)
1360 initcall_entry_t *fn;
1362 parse_args(initcall_level_names[level],
1363 command_line, __start___param,
1364 __stop___param - __start___param,
1366 NULL, ignore_unknown_bootoption);
1368 trace_initcall_level(initcall_level_names[level]);
1369 for (fn = initcall_levels[level]; fn < initcall_levels[level+1]; fn++)
1370 do_one_initcall(initcall_from_entry(fn));
1373 static void __init do_initcalls(void)
1376 size_t len = strlen(saved_command_line) + 1;
1379 command_line = kzalloc(len, GFP_KERNEL);
1381 panic("%s: Failed to allocate %zu bytes\n", __func__, len);
1383 for (level = 0; level < ARRAY_SIZE(initcall_levels) - 1; level++) {
1384 /* Parser modifies command_line, restore it each time */
1385 strcpy(command_line, saved_command_line);
1386 do_initcall_level(level, command_line);
1389 kfree(command_line);
1393 * Ok, the machine is now initialized. None of the devices
1394 * have been touched yet, but the CPU subsystem is up and
1395 * running, and memory and process management works.
1397 * Now we can finally start doing some real work..
1399 static void __init do_basic_setup(void)
1408 static void __init do_pre_smp_initcalls(void)
1410 initcall_entry_t *fn;
1412 trace_initcall_level("early");
1413 for (fn = __initcall_start; fn < __initcall0_start; fn++)
1414 do_one_initcall(initcall_from_entry(fn));
1417 static int run_init_process(const char *init_filename)
1419 const char *const *p;
1421 argv_init[0] = init_filename;
1422 pr_info("Run %s as init process\n", init_filename);
1423 pr_debug(" with arguments:\n");
1424 for (p = argv_init; *p; p++)
1425 pr_debug(" %s\n", *p);
1426 pr_debug(" with environment:\n");
1427 for (p = envp_init; *p; p++)
1428 pr_debug(" %s\n", *p);
1429 return kernel_execve(init_filename, argv_init, envp_init);
1432 static int try_to_run_init_process(const char *init_filename)
1436 ret = run_init_process(init_filename);
1438 if (ret && ret != -ENOENT) {
1439 pr_err("Starting init: %s exists but couldn't execute it (error %d)\n",
1440 init_filename, ret);
1446 static noinline void __init kernel_init_freeable(void);
1448 #if defined(CONFIG_STRICT_KERNEL_RWX) || defined(CONFIG_STRICT_MODULE_RWX)
1449 bool rodata_enabled __ro_after_init = true;
1450 static int __init set_debug_rodata(char *str)
1452 return strtobool(str, &rodata_enabled);
1454 __setup("rodata=", set_debug_rodata);
1457 #ifdef CONFIG_STRICT_KERNEL_RWX
1458 static void mark_readonly(void)
1460 if (rodata_enabled) {
1462 * load_module() results in W+X mappings, which are cleaned
1463 * up with call_rcu(). Let's make sure that queued work is
1464 * flushed so that we don't hit false positives looking for
1465 * insecure pages which are W+X.
1471 pr_info("Kernel memory protection disabled.\n");
1473 #elif defined(CONFIG_ARCH_HAS_STRICT_KERNEL_RWX)
1474 static inline void mark_readonly(void)
1476 pr_warn("Kernel memory protection not selected by kernel config.\n");
1479 static inline void mark_readonly(void)
1481 pr_warn("This architecture does not have kernel memory protection.\n");
1485 void __weak free_initmem(void)
1487 free_initmem_default(POISON_FREE_INITMEM);
1490 static int __ref kernel_init(void *unused)
1495 * Wait until kthreadd is all set-up.
1497 wait_for_completion(&kthreadd_done);
1499 kernel_init_freeable();
1500 /* need to finish all async __init code before freeing the memory */
1501 async_synchronize_full();
1503 system_state = SYSTEM_FREEING_INITMEM;
1504 kprobe_free_init_mem();
1505 ftrace_free_init_mem();
1506 kgdb_free_init_mem();
1512 * Kernel mappings are now finalized - update the userspace page-table
1517 system_state = SYSTEM_RUNNING;
1518 numa_default_policy();
1520 rcu_end_inkernel_boot();
1524 if (ramdisk_execute_command) {
1525 ret = run_init_process(ramdisk_execute_command);
1528 pr_err("Failed to execute %s (error %d)\n",
1529 ramdisk_execute_command, ret);
1533 * We try each of these until one succeeds.
1535 * The Bourne shell can be used instead of init if we are
1536 * trying to recover a really broken machine.
1538 if (execute_command) {
1539 ret = run_init_process(execute_command);
1542 panic("Requested init %s failed (error %d).",
1543 execute_command, ret);
1546 if (CONFIG_DEFAULT_INIT[0] != '\0') {
1547 ret = run_init_process(CONFIG_DEFAULT_INIT);
1549 pr_err("Default init %s failed (error %d)\n",
1550 CONFIG_DEFAULT_INIT, ret);
1555 if (!try_to_run_init_process("/sbin/init") ||
1556 !try_to_run_init_process("/etc/init") ||
1557 !try_to_run_init_process("/bin/init") ||
1558 !try_to_run_init_process("/bin/sh"))
1561 panic("No working init found. Try passing init= option to kernel. "
1562 "See Linux Documentation/admin-guide/init.rst for guidance.");
1565 /* Open /dev/console, for stdin/stdout/stderr, this should never fail */
1566 void __init console_on_rootfs(void)
1568 struct file *file = filp_open("/dev/console", O_RDWR, 0);
1571 pr_err("Warning: unable to open an initial console.\n");
1580 static noinline void __init kernel_init_freeable(void)
1582 /* Now the scheduler is fully set up and can do blocking allocations */
1583 gfp_allowed_mask = __GFP_BITS_MASK;
1586 * init can allocate pages on any node
1588 set_mems_allowed(node_states[N_MEMORY]);
1590 cad_pid = get_pid(task_pid(current));
1592 smp_prepare_cpus(setup_max_cpus);
1596 init_mm_internals();
1598 rcu_init_tasks_generic();
1599 do_pre_smp_initcalls();
1600 lockup_detector_init();
1606 page_alloc_init_late();
1607 /* Initialize page ext after all struct pages are initialized. */
1612 kunit_run_all_tests();
1614 wait_for_initramfs();
1615 console_on_rootfs();
1618 * check if there is an early userspace init. If yes, let it do all
1621 if (init_eaccess(ramdisk_execute_command) != 0) {
1622 ramdisk_execute_command = NULL;
1623 prepare_namespace();
1627 * Ok, we have completed the initial bootup, and
1628 * we're essentially up and running. Get rid of the
1629 * initmem segments and start the user-mode stuff..
1631 * rootfs is available now, try loading the public keys
1632 * and default modules
1635 integrity_load_keys();