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
13 #define DEBUG /* Enable initcall_debug */
15 #include <linux/types.h>
16 #include <linux/extable.h>
17 #include <linux/module.h>
18 #include <linux/proc_fs.h>
19 #include <linux/binfmts.h>
20 #include <linux/kernel.h>
21 #include <linux/syscalls.h>
22 #include <linux/stackprotector.h>
23 #include <linux/string.h>
24 #include <linux/ctype.h>
25 #include <linux/delay.h>
26 #include <linux/ioport.h>
27 #include <linux/init.h>
28 #include <linux/initrd.h>
29 #include <linux/memblock.h>
30 #include <linux/acpi.h>
31 #include <linux/bootconfig.h>
32 #include <linux/console.h>
33 #include <linux/nmi.h>
34 #include <linux/percpu.h>
35 #include <linux/kmod.h>
36 #include <linux/kprobes.h>
37 #include <linux/kmsan.h>
38 #include <linux/vmalloc.h>
39 #include <linux/kernel_stat.h>
40 #include <linux/start_kernel.h>
41 #include <linux/security.h>
42 #include <linux/smp.h>
43 #include <linux/profile.h>
44 #include <linux/kfence.h>
45 #include <linux/rcupdate.h>
46 #include <linux/srcu.h>
47 #include <linux/moduleparam.h>
48 #include <linux/kallsyms.h>
49 #include <linux/buildid.h>
50 #include <linux/writeback.h>
51 #include <linux/cpu.h>
52 #include <linux/cpuset.h>
53 #include <linux/cgroup.h>
54 #include <linux/efi.h>
55 #include <linux/tick.h>
56 #include <linux/sched/isolation.h>
57 #include <linux/interrupt.h>
58 #include <linux/taskstats_kern.h>
59 #include <linux/delayacct.h>
60 #include <linux/unistd.h>
61 #include <linux/utsname.h>
62 #include <linux/rmap.h>
63 #include <linux/mempolicy.h>
64 #include <linux/key.h>
65 #include <linux/debug_locks.h>
66 #include <linux/debugobjects.h>
67 #include <linux/lockdep.h>
68 #include <linux/kmemleak.h>
69 #include <linux/padata.h>
70 #include <linux/pid_namespace.h>
71 #include <linux/device/driver.h>
72 #include <linux/kthread.h>
73 #include <linux/sched.h>
74 #include <linux/sched/init.h>
75 #include <linux/signal.h>
76 #include <linux/idr.h>
77 #include <linux/kgdb.h>
78 #include <linux/ftrace.h>
79 #include <linux/async.h>
80 #include <linux/shmem_fs.h>
81 #include <linux/slab.h>
82 #include <linux/perf_event.h>
83 #include <linux/ptrace.h>
84 #include <linux/pti.h>
85 #include <linux/blkdev.h>
86 #include <linux/sched/clock.h>
87 #include <linux/sched/task.h>
88 #include <linux/sched/task_stack.h>
89 #include <linux/context_tracking.h>
90 #include <linux/random.h>
91 #include <linux/moduleloader.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/kcsan.h>
100 #include <linux/init_syscalls.h>
101 #include <linux/stackdepot.h>
102 #include <linux/randomize_kstack.h>
103 #include <linux/pidfs.h>
104 #include <linux/ptdump.h>
105 #include <net/net_namespace.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 *);
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 /* Default late time init is NULL. archs can override this later. */
138 void (*__initdata late_time_init)(void);
140 /* Untouched command line saved by arch-specific code. */
141 char __initdata boot_command_line[COMMAND_LINE_SIZE];
142 /* Untouched saved command line (eg. for /proc) */
143 char *saved_command_line __ro_after_init;
144 unsigned int saved_command_line_len __ro_after_init;
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 static bool __init obsolete_checksetup(char *line)
197 const struct obs_kernel_param *p;
198 bool had_early_param = false;
202 int n = strlen(p->str);
203 if (parameqn(line, p->str, n)) {
205 /* Already done in parse_early_param?
206 * (Needs exact match on param part).
207 * Keep iterating, as we can have early
208 * params and __setups of same names 8( */
209 if (line[n] == '\0' || line[n] == '=')
210 had_early_param = true;
211 } else if (!p->setup_func) {
212 pr_warn("Parameter %s is obsolete, ignored\n",
215 } else if (p->setup_func(line + n))
219 } while (p < __setup_end);
221 return had_early_param;
225 * This should be approx 2 Bo*oMips to start (note initial shift), and will
226 * still work even if initially too large, it will just take slightly longer
228 unsigned long loops_per_jiffy = (1<<12);
229 EXPORT_SYMBOL(loops_per_jiffy);
231 static int __init debug_kernel(char *str)
233 console_loglevel = CONSOLE_LOGLEVEL_DEBUG;
237 static int __init quiet_kernel(char *str)
239 console_loglevel = CONSOLE_LOGLEVEL_QUIET;
243 early_param("debug", debug_kernel);
244 early_param("quiet", quiet_kernel);
246 static int __init loglevel(char *str)
251 * Only update loglevel value when a correct setting was passed,
252 * to prevent blind crashes (when loglevel being set to 0) that
253 * are quite hard to debug
255 if (get_option(&str, &newlevel)) {
256 console_loglevel = newlevel;
263 early_param("loglevel", loglevel);
265 #ifdef CONFIG_BLK_DEV_INITRD
266 static void * __init get_boot_config_from_initrd(size_t *_size)
276 data = (char *)initrd_end - BOOTCONFIG_MAGIC_LEN;
278 * Since Grub may align the size of initrd to 4, we must
279 * check the preceding 3 bytes as well.
281 for (i = 0; i < 4; i++) {
282 if (!memcmp(data, BOOTCONFIG_MAGIC, BOOTCONFIG_MAGIC_LEN))
289 hdr = (u32 *)(data - 8);
290 size = le32_to_cpu(hdr[0]);
291 csum = le32_to_cpu(hdr[1]);
293 data = ((void *)hdr) - size;
294 if ((unsigned long)data < initrd_start) {
295 pr_err("bootconfig size %d is greater than initrd size %ld\n",
296 size, initrd_end - initrd_start);
300 if (xbc_calc_checksum(data, size) != csum) {
301 pr_err("bootconfig checksum failed\n");
305 /* Remove bootconfig from initramfs/initrd */
306 initrd_end = (unsigned long)data;
313 static void * __init get_boot_config_from_initrd(size_t *_size)
319 #ifdef CONFIG_BOOT_CONFIG
321 static char xbc_namebuf[XBC_KEYLEN_MAX] __initdata;
323 #define rest(dst, end) ((end) > (dst) ? (end) - (dst) : 0)
325 static int __init xbc_snprint_cmdline(char *buf, size_t size,
326 struct xbc_node *root)
328 struct xbc_node *knode, *vnode;
329 char *end = buf + size;
333 xbc_node_for_each_key_value(root, knode, val) {
334 ret = xbc_node_compose_key_after(root, knode,
335 xbc_namebuf, XBC_KEYLEN_MAX);
339 vnode = xbc_node_get_child(knode);
341 ret = snprintf(buf, rest(buf, end), "%s ", xbc_namebuf);
347 xbc_array_for_each_value(vnode, val) {
349 * For prettier and more readable /proc/cmdline, only
350 * quote the value when necessary, i.e. when it contains
353 q = strpbrk(val, " \t\r\n") ? "\"" : "";
354 ret = snprintf(buf, rest(buf, end), "%s=%s%s%s ",
355 xbc_namebuf, q, val, q);
362 return buf - (end - size);
366 /* Make an extra command line under given key word */
367 static char * __init xbc_make_cmdline(const char *key)
369 struct xbc_node *root;
373 root = xbc_find_node(key);
377 /* Count required buffer size */
378 len = xbc_snprint_cmdline(NULL, 0, root);
382 new_cmdline = memblock_alloc(len + 1, SMP_CACHE_BYTES);
384 pr_err("Failed to allocate memory for extra kernel cmdline.\n");
388 ret = xbc_snprint_cmdline(new_cmdline, len + 1, root);
389 if (ret < 0 || ret > len) {
390 pr_err("Failed to print extra kernel cmdline.\n");
391 memblock_free(new_cmdline, len + 1);
398 static int __init bootconfig_params(char *param, char *val,
399 const char *unused, void *arg)
401 if (strcmp(param, "bootconfig") == 0) {
402 bootconfig_found = true;
407 static int __init warn_bootconfig(char *str)
409 /* The 'bootconfig' has been handled by bootconfig_params(). */
413 static void __init setup_boot_config(void)
415 static char tmp_cmdline[COMMAND_LINE_SIZE] __initdata;
416 const char *msg, *data;
421 /* Cut out the bootconfig data even if we have no bootconfig option */
422 data = get_boot_config_from_initrd(&size);
423 /* If there is no bootconfig in initrd, try embedded one. */
425 data = xbc_get_embedded_bootconfig(&size);
427 strscpy(tmp_cmdline, boot_command_line, COMMAND_LINE_SIZE);
428 err = parse_args("bootconfig", tmp_cmdline, NULL, 0, 0, 0, NULL,
431 if (IS_ERR(err) || !(bootconfig_found || IS_ENABLED(CONFIG_BOOT_CONFIG_FORCE)))
434 /* parse_args() stops at the next param of '--' and returns an address */
436 initargs_offs = err - tmp_cmdline;
439 /* If user intended to use bootconfig, show an error level message */
440 if (bootconfig_found)
441 pr_err("'bootconfig' found on command line, but no bootconfig found\n");
443 pr_info("No bootconfig data provided, so skipping bootconfig");
447 if (size >= XBC_DATA_MAX) {
448 pr_err("bootconfig size %ld greater than max size %d\n",
449 (long)size, XBC_DATA_MAX);
453 ret = xbc_init(data, size, &msg, &pos);
456 pr_err("Failed to init bootconfig: %s.\n", msg);
458 pr_err("Failed to parse bootconfig: %s at %d.\n",
461 xbc_get_info(&ret, NULL);
462 pr_info("Load bootconfig: %ld bytes %d nodes\n", (long)size, ret);
463 /* keys starting with "kernel." are passed via cmdline */
464 extra_command_line = xbc_make_cmdline("kernel");
465 /* Also, "init." keys are init arguments */
466 extra_init_args = xbc_make_cmdline("init");
471 static void __init exit_boot_config(void)
476 #else /* !CONFIG_BOOT_CONFIG */
478 static void __init setup_boot_config(void)
480 /* Remove bootconfig data from initrd */
481 get_boot_config_from_initrd(NULL);
484 static int __init warn_bootconfig(char *str)
486 pr_warn("WARNING: 'bootconfig' found on the kernel command line but CONFIG_BOOT_CONFIG is not set.\n");
490 #define exit_boot_config() do {} while (0)
492 #endif /* CONFIG_BOOT_CONFIG */
494 early_param("bootconfig", warn_bootconfig);
496 bool __init cmdline_has_extra_options(void)
498 return extra_command_line || extra_init_args;
501 /* Change NUL term back to "=", to make "param" the whole string. */
502 static void __init repair_env_string(char *param, char *val)
505 /* param=val or param="val"? */
506 if (val == param+strlen(param)+1)
508 else if (val == param+strlen(param)+2) {
510 memmove(val-1, val, strlen(val)+1);
516 /* Anything after -- gets handed straight to init. */
517 static int __init set_init_arg(char *param, char *val,
518 const char *unused, void *arg)
525 repair_env_string(param, val);
527 for (i = 0; argv_init[i]; i++) {
528 if (i == MAX_INIT_ARGS) {
529 panic_later = "init";
534 argv_init[i] = param;
539 * Unknown boot options get handed to init, unless they look like
540 * unused parameters (modprobe will find them in /proc/cmdline).
542 static int __init unknown_bootoption(char *param, char *val,
543 const char *unused, void *arg)
545 size_t len = strlen(param);
547 /* Handle params aliased to sysctls */
548 if (sysctl_is_alias(param))
551 repair_env_string(param, val);
553 /* Handle obsolete-style parameters */
554 if (obsolete_checksetup(param))
557 /* Unused module parameter. */
558 if (strnchr(param, len, '.'))
565 /* Environment option */
567 for (i = 0; envp_init[i]; i++) {
568 if (i == MAX_INIT_ENVS) {
572 if (!strncmp(param, envp_init[i], len+1))
575 envp_init[i] = param;
577 /* Command line option */
579 for (i = 0; argv_init[i]; i++) {
580 if (i == MAX_INIT_ARGS) {
581 panic_later = "init";
585 argv_init[i] = param;
590 static int __init init_setup(char *str)
594 execute_command = str;
596 * In case LILO is going to boot us with default command line,
597 * it prepends "auto" before the whole cmdline which makes
598 * the shell think it should execute a script with such name.
599 * So we ignore all arguments entered _before_ init=... [MJ]
601 for (i = 1; i < MAX_INIT_ARGS; i++)
605 __setup("init=", init_setup);
607 static int __init rdinit_setup(char *str)
611 ramdisk_execute_command = str;
612 /* See "auto" comment in init_setup */
613 for (i = 1; i < MAX_INIT_ARGS; i++)
617 __setup("rdinit=", rdinit_setup);
620 static inline void setup_nr_cpu_ids(void) { }
621 static inline void smp_prepare_cpus(unsigned int maxcpus) { }
625 * We need to store the untouched command line for future reference.
626 * We also need to store the touched command line since the parameter
627 * parsing is performed in place, and we should allow a component to
628 * store reference of name/value for future reference.
630 static void __init setup_command_line(char *command_line)
632 size_t len, xlen = 0, ilen = 0;
634 if (extra_command_line)
635 xlen = strlen(extra_command_line);
636 if (extra_init_args) {
637 extra_init_args = strim(extra_init_args); /* remove trailing space */
638 ilen = strlen(extra_init_args) + 4; /* for " -- " */
641 len = xlen + strlen(boot_command_line) + ilen + 1;
643 saved_command_line = memblock_alloc(len, SMP_CACHE_BYTES);
644 if (!saved_command_line)
645 panic("%s: Failed to allocate %zu bytes\n", __func__, len);
647 len = xlen + strlen(command_line) + 1;
649 static_command_line = memblock_alloc(len, SMP_CACHE_BYTES);
650 if (!static_command_line)
651 panic("%s: Failed to allocate %zu bytes\n", __func__, len);
655 * We have to put extra_command_line before boot command
656 * lines because there could be dashes (separator of init
657 * command line) in the command lines.
659 strcpy(saved_command_line, extra_command_line);
660 strcpy(static_command_line, extra_command_line);
662 strcpy(saved_command_line + xlen, boot_command_line);
663 strcpy(static_command_line + xlen, command_line);
667 * Append supplemental init boot args to saved_command_line
668 * so that user can check what command line options passed
670 * The order should always be
671 * " -- "[bootconfig init-param][cmdline init-param]
674 len = xlen + initargs_offs;
675 strcpy(saved_command_line + len, extra_init_args);
676 len += ilen - 4; /* strlen(extra_init_args) */
677 strcpy(saved_command_line + len,
678 boot_command_line + initargs_offs - 1);
680 len = strlen(saved_command_line);
681 strcpy(saved_command_line + len, " -- ");
683 strcpy(saved_command_line + len, extra_init_args);
687 saved_command_line_len = strlen(saved_command_line);
691 * We need to finalize in a non-__init function or else race conditions
692 * between the root thread and the init thread may cause start_kernel to
693 * be reaped by free_initmem before the root thread has proceeded to
696 * gcc-3.4 accidentally inlines this function, so use noinline.
699 static __initdata DECLARE_COMPLETION(kthreadd_done);
701 static noinline void __ref __noreturn rest_init(void)
703 struct task_struct *tsk;
706 rcu_scheduler_starting();
708 * We need to spawn init first so that it obtains pid 1, however
709 * the init task will end up wanting to create kthreads, which, if
710 * we schedule it before we create kthreadd, will OOPS.
712 pid = user_mode_thread(kernel_init, NULL, CLONE_FS);
714 * Pin init on the boot CPU. Task migration is not properly working
715 * until sched_init_smp() has been run. It will set the allowed
716 * CPUs for init to the non isolated CPUs.
719 tsk = find_task_by_pid_ns(pid, &init_pid_ns);
720 tsk->flags |= PF_NO_SETAFFINITY;
721 set_cpus_allowed_ptr(tsk, cpumask_of(smp_processor_id()));
724 numa_default_policy();
725 pid = kernel_thread(kthreadd, NULL, NULL, CLONE_FS | CLONE_FILES);
727 kthreadd_task = find_task_by_pid_ns(pid, &init_pid_ns);
731 * Enable might_sleep() and smp_processor_id() checks.
732 * They cannot be enabled earlier because with CONFIG_PREEMPTION=y
733 * kernel_thread() would trigger might_sleep() splats. With
734 * CONFIG_PREEMPT_VOLUNTARY=y the init task might have scheduled
735 * already, but it's stuck on the kthreadd_done completion.
737 system_state = SYSTEM_SCHEDULING;
739 complete(&kthreadd_done);
742 * The boot idle thread must execute schedule()
743 * at least once to get things moving:
745 schedule_preempt_disabled();
746 /* Call into cpu_idle with preempt disabled */
747 cpu_startup_entry(CPUHP_ONLINE);
750 /* Check for early params. */
751 static int __init do_early_param(char *param, char *val,
752 const char *unused, void *arg)
754 const struct obs_kernel_param *p;
756 for (p = __setup_start; p < __setup_end; p++) {
757 if ((p->early && parameq(param, p->str)) ||
758 (strcmp(param, "console") == 0 &&
759 strcmp(p->str, "earlycon") == 0)
761 if (p->setup_func(val) != 0)
762 pr_warn("Malformed early option '%s'\n", param);
765 /* We accept everything at this stage. */
769 void __init parse_early_options(char *cmdline)
771 parse_args("early options", cmdline, NULL, 0, 0, 0, NULL,
775 /* Arch code calls this early on, or if not, just before other parsing. */
776 void __init parse_early_param(void)
778 static int done __initdata;
779 static char tmp_cmdline[COMMAND_LINE_SIZE] __initdata;
784 /* All fall through to do_early_param. */
785 strscpy(tmp_cmdline, boot_command_line, COMMAND_LINE_SIZE);
786 parse_early_options(tmp_cmdline);
790 void __init __weak arch_post_acpi_subsys_init(void) { }
792 void __init __weak smp_setup_processor_id(void)
796 void __init __weak smp_prepare_boot_cpu(void)
800 # if THREAD_SIZE >= PAGE_SIZE
801 void __init __weak thread_stack_cache_init(void)
806 void __init __weak poking_init(void) { }
808 void __init __weak pgtable_cache_init(void) { }
810 void __init __weak trap_init(void) { }
813 core_param(initcall_debug, initcall_debug, bool, 0644);
815 #ifdef TRACEPOINTS_ENABLED
816 static void __init initcall_debug_enable(void);
818 static inline void initcall_debug_enable(void)
823 #ifdef CONFIG_RANDOMIZE_KSTACK_OFFSET
824 DEFINE_STATIC_KEY_MAYBE_RO(CONFIG_RANDOMIZE_KSTACK_OFFSET_DEFAULT,
825 randomize_kstack_offset);
826 DEFINE_PER_CPU(u32, kstack_offset);
828 static int __init early_randomize_kstack_offset(char *buf)
833 ret = kstrtobool(buf, &bool_result);
838 static_branch_enable(&randomize_kstack_offset);
840 static_branch_disable(&randomize_kstack_offset);
843 early_param("randomize_kstack_offset", early_randomize_kstack_offset);
846 static void __init print_unknown_bootoptions(void)
848 char *unknown_options;
850 const char *const *p;
853 if (panic_later || (!argv_init[1] && !envp_init[2]))
857 * Determine how many options we have to print out, plus a space
860 len = 1; /* null terminator */
861 for (p = &argv_init[1]; *p; p++) {
865 for (p = &envp_init[2]; *p; p++) {
870 unknown_options = memblock_alloc(len, SMP_CACHE_BYTES);
871 if (!unknown_options) {
872 pr_err("%s: Failed to allocate %zu bytes\n",
876 end = unknown_options;
878 for (p = &argv_init[1]; *p; p++)
879 end += sprintf(end, " %s", *p);
880 for (p = &envp_init[2]; *p; p++)
881 end += sprintf(end, " %s", *p);
883 /* Start at unknown_options[1] to skip the initial space */
884 pr_notice("Unknown kernel command line parameters \"%s\", will be passed to user space.\n",
885 &unknown_options[1]);
886 memblock_free(unknown_options, len);
889 static void __init early_numa_node_init(void)
891 #ifdef CONFIG_USE_PERCPU_NUMA_NODE_ID
895 /* The early_cpu_to_node() should be ready here. */
896 for_each_possible_cpu(cpu)
897 set_cpu_numa_node(cpu, early_cpu_to_node(cpu));
902 asmlinkage __visible __init __no_sanitize_address __noreturn __no_stack_protector
903 void start_kernel(void)
908 set_task_stack_end_magic(&init_task);
909 smp_setup_processor_id();
910 debug_objects_early_init();
911 init_vmlinux_build_id();
916 early_boot_irqs_disabled = true;
919 * Interrupts are still disabled. Do necessary setups, then
924 pr_notice("%s", linux_banner);
925 early_security_init();
926 setup_arch(&command_line);
928 setup_command_line(command_line);
930 setup_per_cpu_areas();
931 smp_prepare_boot_cpu(); /* arch-specific boot-cpu hooks */
932 early_numa_node_init();
933 boot_cpu_hotplug_init();
935 pr_notice("Kernel command line: %s\n", saved_command_line);
936 /* parameters may set static keys */
939 after_dashes = parse_args("Booting kernel",
940 static_command_line, __start___param,
941 __stop___param - __start___param,
942 -1, -1, NULL, &unknown_bootoption);
943 print_unknown_bootoptions();
944 if (!IS_ERR_OR_NULL(after_dashes))
945 parse_args("Setting init args", after_dashes, NULL, 0, -1, -1,
948 parse_args("Setting extra init args", extra_init_args,
949 NULL, 0, -1, -1, NULL, set_init_arg);
951 /* Architectural and non-timekeeping rng init, before allocator init */
952 random_init_early(command_line);
955 * These use large bootmem allocations and must precede
956 * initalization of page allocator
959 vfs_caches_init_early();
966 /* trace_printk can be enabled here */
970 * Set up the scheduler prior starting any interrupts (such as the
971 * timer interrupt). Full topology setup happens at smp_init()
972 * time - but meanwhile we still have a functioning scheduler.
976 if (WARN(!irqs_disabled(),
977 "Interrupts were enabled *very* early, fixing it\n"))
983 * Set up housekeeping before setting up workqueues to allow the unbound
984 * workqueue to take non-housekeeping into account.
989 * Allow workqueue creation and work item queueing/cancelling
990 * early. Work item execution depends on kthreads and starts after
993 workqueue_init_early();
997 /* Trace events are available after this */
1001 initcall_debug_enable();
1003 context_tracking_init();
1004 /* init some links before init_ISA_irqs() */
1016 /* This must be after timekeeping is initialized */
1019 /* These make use of the fully initialized rng */
1021 boot_init_stack_canary();
1025 call_function_init();
1026 WARN(!irqs_disabled(), "Interrupts were enabled early\n");
1028 early_boot_irqs_disabled = false;
1031 kmem_cache_init_late();
1034 * HACK ALERT! This is early. We're enabling the console before
1035 * we've done PCI setups etc, and console_init() must be aware of
1036 * this. But we do want output early, in case something goes wrong.
1040 panic("Too many boot %s vars at `%s'", panic_later,
1046 * Need to run this when irqs are enabled, because it wants
1047 * to self-test [hard/soft]-irqs on/off lock inversion bugs
1052 #ifdef CONFIG_BLK_DEV_INITRD
1053 if (initrd_start && !initrd_below_start_ok &&
1054 page_to_pfn(virt_to_page((void *)initrd_start)) < min_low_pfn) {
1055 pr_crit("initrd overwritten (0x%08lx < 0x%08lx) - disabling it.\n",
1056 page_to_pfn(virt_to_page((void *)initrd_start)),
1061 setup_per_cpu_pageset();
1069 arch_cpu_finalize_init();
1074 if (efi_enabled(EFI_RUNTIME_SERVICES))
1075 efi_enter_virtual_mode();
1077 thread_stack_cache_init();
1095 taskstats_init_early();
1098 acpi_subsystem_init();
1099 arch_post_acpi_subsys_init();
1102 /* Do the rest non-__init'ed, we're now alive */
1106 * Avoid stack canaries in callers of boot_init_stack_canary for gcc-10
1109 #if !__has_attribute(__no_stack_protector__)
1110 prevent_tail_call_optimization();
1114 /* Call all constructor functions linked into the kernel. */
1115 static void __init do_ctors(void)
1118 * For UML, the constructors have already been called by the
1119 * normal setup code as it's just a normal ELF binary, so we
1120 * cannot do it again - but we do need CONFIG_CONSTRUCTORS
1121 * even on UML for modules.
1123 #if defined(CONFIG_CONSTRUCTORS) && !defined(CONFIG_UML)
1124 ctor_fn_t *fn = (ctor_fn_t *) __ctors_start;
1126 for (; fn < (ctor_fn_t *) __ctors_end; fn++)
1131 #ifdef CONFIG_KALLSYMS
1132 struct blacklist_entry {
1133 struct list_head next;
1137 static __initdata_or_module LIST_HEAD(blacklisted_initcalls);
1139 static int __init initcall_blacklist(char *str)
1142 struct blacklist_entry *entry;
1144 /* str argument is a comma-separated list of functions */
1146 str_entry = strsep(&str, ",");
1148 pr_debug("blacklisting initcall %s\n", str_entry);
1149 entry = memblock_alloc(sizeof(*entry),
1152 panic("%s: Failed to allocate %zu bytes\n",
1153 __func__, sizeof(*entry));
1154 entry->buf = memblock_alloc(strlen(str_entry) + 1,
1157 panic("%s: Failed to allocate %zu bytes\n",
1158 __func__, strlen(str_entry) + 1);
1159 strcpy(entry->buf, str_entry);
1160 list_add(&entry->next, &blacklisted_initcalls);
1162 } while (str_entry);
1167 static bool __init_or_module initcall_blacklisted(initcall_t fn)
1169 struct blacklist_entry *entry;
1170 char fn_name[KSYM_SYMBOL_LEN];
1173 if (list_empty(&blacklisted_initcalls))
1176 addr = (unsigned long) dereference_function_descriptor(fn);
1177 sprint_symbol_no_offset(fn_name, addr);
1180 * fn will be "function_name [module_name]" where [module_name] is not
1181 * displayed for built-in init functions. Strip off the [module_name].
1183 strreplace(fn_name, ' ', '\0');
1185 list_for_each_entry(entry, &blacklisted_initcalls, next) {
1186 if (!strcmp(fn_name, entry->buf)) {
1187 pr_debug("initcall %s blacklisted\n", fn_name);
1195 static int __init initcall_blacklist(char *str)
1197 pr_warn("initcall_blacklist requires CONFIG_KALLSYMS\n");
1201 static bool __init_or_module initcall_blacklisted(initcall_t fn)
1206 __setup("initcall_blacklist=", initcall_blacklist);
1208 static __init_or_module void
1209 trace_initcall_start_cb(void *data, initcall_t fn)
1211 ktime_t *calltime = data;
1213 printk(KERN_DEBUG "calling %pS @ %i\n", fn, task_pid_nr(current));
1214 *calltime = ktime_get();
1217 static __init_or_module void
1218 trace_initcall_finish_cb(void *data, initcall_t fn, int ret)
1220 ktime_t rettime, *calltime = data;
1222 rettime = ktime_get();
1223 printk(KERN_DEBUG "initcall %pS returned %d after %lld usecs\n",
1224 fn, ret, (unsigned long long)ktime_us_delta(rettime, *calltime));
1227 static ktime_t initcall_calltime;
1229 #ifdef TRACEPOINTS_ENABLED
1230 static void __init initcall_debug_enable(void)
1234 ret = register_trace_initcall_start(trace_initcall_start_cb,
1235 &initcall_calltime);
1236 ret |= register_trace_initcall_finish(trace_initcall_finish_cb,
1237 &initcall_calltime);
1238 WARN(ret, "Failed to register initcall tracepoints\n");
1240 # define do_trace_initcall_start trace_initcall_start
1241 # define do_trace_initcall_finish trace_initcall_finish
1243 static inline void do_trace_initcall_start(initcall_t fn)
1245 if (!initcall_debug)
1247 trace_initcall_start_cb(&initcall_calltime, fn);
1249 static inline void do_trace_initcall_finish(initcall_t fn, int ret)
1251 if (!initcall_debug)
1253 trace_initcall_finish_cb(&initcall_calltime, fn, ret);
1255 #endif /* !TRACEPOINTS_ENABLED */
1257 int __init_or_module do_one_initcall(initcall_t fn)
1259 int count = preempt_count();
1263 if (initcall_blacklisted(fn))
1266 do_trace_initcall_start(fn);
1268 do_trace_initcall_finish(fn, ret);
1272 if (preempt_count() != count) {
1273 sprintf(msgbuf, "preemption imbalance ");
1274 preempt_count_set(count);
1276 if (irqs_disabled()) {
1277 strlcat(msgbuf, "disabled interrupts ", sizeof(msgbuf));
1280 WARN(msgbuf[0], "initcall %pS returned with %s\n", fn, msgbuf);
1282 add_latent_entropy();
1287 static initcall_entry_t *initcall_levels[] __initdata = {
1299 /* Keep these in sync with initcalls in include/linux/init.h */
1300 static const char *initcall_level_names[] __initdata = {
1311 static int __init ignore_unknown_bootoption(char *param, char *val,
1312 const char *unused, void *arg)
1317 static void __init do_initcall_level(int level, char *command_line)
1319 initcall_entry_t *fn;
1321 parse_args(initcall_level_names[level],
1322 command_line, __start___param,
1323 __stop___param - __start___param,
1325 NULL, ignore_unknown_bootoption);
1327 trace_initcall_level(initcall_level_names[level]);
1328 for (fn = initcall_levels[level]; fn < initcall_levels[level+1]; fn++)
1329 do_one_initcall(initcall_from_entry(fn));
1332 static void __init do_initcalls(void)
1335 size_t len = saved_command_line_len + 1;
1338 command_line = kzalloc(len, GFP_KERNEL);
1340 panic("%s: Failed to allocate %zu bytes\n", __func__, len);
1342 for (level = 0; level < ARRAY_SIZE(initcall_levels) - 1; level++) {
1343 /* Parser modifies command_line, restore it each time */
1344 strcpy(command_line, saved_command_line);
1345 do_initcall_level(level, command_line);
1348 kfree(command_line);
1352 * Ok, the machine is now initialized. None of the devices
1353 * have been touched yet, but the CPU subsystem is up and
1354 * running, and memory and process management works.
1356 * Now we can finally start doing some real work..
1358 static void __init do_basic_setup(void)
1367 static void __init do_pre_smp_initcalls(void)
1369 initcall_entry_t *fn;
1371 trace_initcall_level("early");
1372 for (fn = __initcall_start; fn < __initcall0_start; fn++)
1373 do_one_initcall(initcall_from_entry(fn));
1376 static int run_init_process(const char *init_filename)
1378 const char *const *p;
1380 argv_init[0] = init_filename;
1381 pr_info("Run %s as init process\n", init_filename);
1382 pr_debug(" with arguments:\n");
1383 for (p = argv_init; *p; p++)
1384 pr_debug(" %s\n", *p);
1385 pr_debug(" with environment:\n");
1386 for (p = envp_init; *p; p++)
1387 pr_debug(" %s\n", *p);
1388 return kernel_execve(init_filename, argv_init, envp_init);
1391 static int try_to_run_init_process(const char *init_filename)
1395 ret = run_init_process(init_filename);
1397 if (ret && ret != -ENOENT) {
1398 pr_err("Starting init: %s exists but couldn't execute it (error %d)\n",
1399 init_filename, ret);
1405 static noinline void __init kernel_init_freeable(void);
1407 #if defined(CONFIG_STRICT_KERNEL_RWX) || defined(CONFIG_STRICT_MODULE_RWX)
1408 bool rodata_enabled __ro_after_init = true;
1410 #ifndef arch_parse_debug_rodata
1411 static inline bool arch_parse_debug_rodata(char *str) { return false; }
1414 static int __init set_debug_rodata(char *str)
1416 if (arch_parse_debug_rodata(str))
1419 if (str && !strcmp(str, "on"))
1420 rodata_enabled = true;
1421 else if (str && !strcmp(str, "off"))
1422 rodata_enabled = false;
1424 pr_warn("Invalid option string for rodata: '%s'\n", str);
1427 early_param("rodata", set_debug_rodata);
1430 static void mark_readonly(void)
1432 if (IS_ENABLED(CONFIG_STRICT_KERNEL_RWX) && rodata_enabled) {
1434 * load_module() results in W+X mappings, which are cleaned
1435 * up with init_free_wq. Let's make sure that queued work is
1436 * flushed so that we don't hit false positives looking for
1437 * insecure pages which are W+X.
1439 flush_module_init_free_work();
1440 jump_label_init_ro();
1444 } else if (IS_ENABLED(CONFIG_STRICT_KERNEL_RWX)) {
1445 pr_info("Kernel memory protection disabled.\n");
1446 } else if (IS_ENABLED(CONFIG_ARCH_HAS_STRICT_KERNEL_RWX)) {
1447 pr_warn("Kernel memory protection not selected by kernel config.\n");
1449 pr_warn("This architecture does not have kernel memory protection.\n");
1453 void __weak free_initmem(void)
1455 free_initmem_default(POISON_FREE_INITMEM);
1458 static int __ref kernel_init(void *unused)
1463 * Wait until kthreadd is all set-up.
1465 wait_for_completion(&kthreadd_done);
1467 kernel_init_freeable();
1468 /* need to finish all async __init code before freeing the memory */
1469 async_synchronize_full();
1471 system_state = SYSTEM_FREEING_INITMEM;
1472 kprobe_free_init_mem();
1473 ftrace_free_init_mem();
1474 kgdb_free_init_mem();
1480 * Kernel mappings are now finalized - update the userspace page-table
1485 system_state = SYSTEM_RUNNING;
1486 numa_default_policy();
1488 rcu_end_inkernel_boot();
1492 if (ramdisk_execute_command) {
1493 ret = run_init_process(ramdisk_execute_command);
1496 pr_err("Failed to execute %s (error %d)\n",
1497 ramdisk_execute_command, ret);
1501 * We try each of these until one succeeds.
1503 * The Bourne shell can be used instead of init if we are
1504 * trying to recover a really broken machine.
1506 if (execute_command) {
1507 ret = run_init_process(execute_command);
1510 panic("Requested init %s failed (error %d).",
1511 execute_command, ret);
1514 if (CONFIG_DEFAULT_INIT[0] != '\0') {
1515 ret = run_init_process(CONFIG_DEFAULT_INIT);
1517 pr_err("Default init %s failed (error %d)\n",
1518 CONFIG_DEFAULT_INIT, ret);
1523 if (!try_to_run_init_process("/sbin/init") ||
1524 !try_to_run_init_process("/etc/init") ||
1525 !try_to_run_init_process("/bin/init") ||
1526 !try_to_run_init_process("/bin/sh"))
1529 panic("No working init found. Try passing init= option to kernel. "
1530 "See Linux Documentation/admin-guide/init.rst for guidance.");
1533 /* Open /dev/console, for stdin/stdout/stderr, this should never fail */
1534 void __init console_on_rootfs(void)
1536 struct file *file = filp_open("/dev/console", O_RDWR, 0);
1539 pr_err("Warning: unable to open an initial console.\n");
1548 static noinline void __init kernel_init_freeable(void)
1550 /* Now the scheduler is fully set up and can do blocking allocations */
1551 gfp_allowed_mask = __GFP_BITS_MASK;
1554 * init can allocate pages on any node
1556 set_mems_allowed(node_states[N_MEMORY]);
1558 cad_pid = get_pid(task_pid(current));
1560 smp_prepare_cpus(setup_max_cpus);
1564 init_mm_internals();
1566 rcu_init_tasks_generic();
1567 do_pre_smp_initcalls();
1568 lockup_detector_init();
1573 workqueue_init_topology();
1576 page_alloc_init_late();
1580 kunit_run_all_tests();
1582 wait_for_initramfs();
1583 console_on_rootfs();
1586 * check if there is an early userspace init. If yes, let it do all
1589 if (init_eaccess(ramdisk_execute_command) != 0) {
1590 ramdisk_execute_command = NULL;
1591 prepare_namespace();
1595 * Ok, we have completed the initial bootup, and
1596 * we're essentially up and running. Get rid of the
1597 * initmem segments and start the user-mode stuff..
1599 * rootfs is available now, try loading the public keys
1600 * and default modules
1603 integrity_load_keys();