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/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/kcsan.h>
99 #include <linux/init_syscalls.h>
100 #include <linux/stackdepot.h>
101 #include <linux/randomize_kstack.h>
102 #include <linux/pidfs.h>
103 #include <net/net_namespace.h>
106 #include <asm/setup.h>
107 #include <asm/sections.h>
108 #include <asm/cacheflush.h>
110 #define CREATE_TRACE_POINTS
111 #include <trace/events/initcall.h>
113 #include <kunit/test.h>
115 static int kernel_init(void *);
118 * Debug helper: via this flag we know that we are in 'early bootup code'
119 * where only the boot processor is running with IRQ disabled. This means
120 * two things - IRQ must not be enabled before the flag is cleared and some
121 * operations which are not allowed with IRQ disabled are allowed while the
124 bool early_boot_irqs_disabled __read_mostly;
126 enum system_states system_state __read_mostly;
127 EXPORT_SYMBOL(system_state);
130 * Boot command-line arguments
132 #define MAX_INIT_ARGS CONFIG_INIT_ENV_ARG_LIMIT
133 #define MAX_INIT_ENVS CONFIG_INIT_ENV_ARG_LIMIT
135 /* Default late time init is NULL. archs can override this later. */
136 void (*__initdata late_time_init)(void);
138 /* Untouched command line saved by arch-specific code. */
139 char __initdata boot_command_line[COMMAND_LINE_SIZE];
140 /* Untouched saved command line (eg. for /proc) */
141 char *saved_command_line __ro_after_init;
142 unsigned int saved_command_line_len __ro_after_init;
143 /* Command line for parameter parsing */
144 static char *static_command_line;
145 /* Untouched extra command line */
146 static char *extra_command_line;
147 /* Extra init arguments */
148 static char *extra_init_args;
150 #ifdef CONFIG_BOOT_CONFIG
151 /* Is bootconfig on command line? */
152 static bool bootconfig_found;
153 static size_t initargs_offs;
155 # define bootconfig_found false
156 # define initargs_offs 0
159 static char *execute_command;
160 static char *ramdisk_execute_command = "/init";
163 * Used to generate warnings if static_key manipulation functions are used
164 * before jump_label_init is called.
166 bool static_key_initialized __read_mostly;
167 EXPORT_SYMBOL_GPL(static_key_initialized);
170 * If set, this is an indication to the drivers that reset the underlying
171 * device before going ahead with the initialization otherwise driver might
172 * rely on the BIOS and skip the reset operation.
174 * This is useful if kernel is booting in an unreliable environment.
175 * For ex. kdump situation where previous kernel has crashed, BIOS has been
176 * skipped and devices will be in unknown state.
178 unsigned int reset_devices;
179 EXPORT_SYMBOL(reset_devices);
181 static int __init set_reset_devices(char *str)
187 __setup("reset_devices", set_reset_devices);
189 static const char *argv_init[MAX_INIT_ARGS+2] = { "init", NULL, };
190 const char *envp_init[MAX_INIT_ENVS+2] = { "HOME=/", "TERM=linux", NULL, };
191 static const char *panic_later, *panic_param;
193 static bool __init obsolete_checksetup(char *line)
195 const struct obs_kernel_param *p;
196 bool had_early_param = false;
200 int n = strlen(p->str);
201 if (parameqn(line, p->str, n)) {
203 /* Already done in parse_early_param?
204 * (Needs exact match on param part).
205 * Keep iterating, as we can have early
206 * params and __setups of same names 8( */
207 if (line[n] == '\0' || line[n] == '=')
208 had_early_param = true;
209 } else if (!p->setup_func) {
210 pr_warn("Parameter %s is obsolete, ignored\n",
213 } else if (p->setup_func(line + n))
217 } while (p < __setup_end);
219 return had_early_param;
223 * This should be approx 2 Bo*oMips to start (note initial shift), and will
224 * still work even if initially too large, it will just take slightly longer
226 unsigned long loops_per_jiffy = (1<<12);
227 EXPORT_SYMBOL(loops_per_jiffy);
229 static int __init debug_kernel(char *str)
231 console_loglevel = CONSOLE_LOGLEVEL_DEBUG;
235 static int __init quiet_kernel(char *str)
237 console_loglevel = CONSOLE_LOGLEVEL_QUIET;
241 early_param("debug", debug_kernel);
242 early_param("quiet", quiet_kernel);
244 static int __init loglevel(char *str)
249 * Only update loglevel value when a correct setting was passed,
250 * to prevent blind crashes (when loglevel being set to 0) that
251 * are quite hard to debug
253 if (get_option(&str, &newlevel)) {
254 console_loglevel = newlevel;
261 early_param("loglevel", loglevel);
263 #ifdef CONFIG_BLK_DEV_INITRD
264 static void * __init get_boot_config_from_initrd(size_t *_size)
274 data = (char *)initrd_end - BOOTCONFIG_MAGIC_LEN;
276 * Since Grub may align the size of initrd to 4, we must
277 * check the preceding 3 bytes as well.
279 for (i = 0; i < 4; i++) {
280 if (!memcmp(data, BOOTCONFIG_MAGIC, BOOTCONFIG_MAGIC_LEN))
287 hdr = (u32 *)(data - 8);
288 size = le32_to_cpu(hdr[0]);
289 csum = le32_to_cpu(hdr[1]);
291 data = ((void *)hdr) - size;
292 if ((unsigned long)data < initrd_start) {
293 pr_err("bootconfig size %d is greater than initrd size %ld\n",
294 size, initrd_end - initrd_start);
298 if (xbc_calc_checksum(data, size) != csum) {
299 pr_err("bootconfig checksum failed\n");
303 /* Remove bootconfig from initramfs/initrd */
304 initrd_end = (unsigned long)data;
311 static void * __init get_boot_config_from_initrd(size_t *_size)
317 #ifdef CONFIG_BOOT_CONFIG
319 static char xbc_namebuf[XBC_KEYLEN_MAX] __initdata;
321 #define rest(dst, end) ((end) > (dst) ? (end) - (dst) : 0)
323 static int __init xbc_snprint_cmdline(char *buf, size_t size,
324 struct xbc_node *root)
326 struct xbc_node *knode, *vnode;
327 char *end = buf + size;
331 xbc_node_for_each_key_value(root, knode, val) {
332 ret = xbc_node_compose_key_after(root, knode,
333 xbc_namebuf, XBC_KEYLEN_MAX);
337 vnode = xbc_node_get_child(knode);
339 ret = snprintf(buf, rest(buf, end), "%s ", xbc_namebuf);
345 xbc_array_for_each_value(vnode, val) {
346 ret = snprintf(buf, rest(buf, end), "%s=\"%s\" ",
354 return buf - (end - size);
358 /* Make an extra command line under given key word */
359 static char * __init xbc_make_cmdline(const char *key)
361 struct xbc_node *root;
365 root = xbc_find_node(key);
369 /* Count required buffer size */
370 len = xbc_snprint_cmdline(NULL, 0, root);
374 new_cmdline = memblock_alloc(len + 1, SMP_CACHE_BYTES);
376 pr_err("Failed to allocate memory for extra kernel cmdline.\n");
380 ret = xbc_snprint_cmdline(new_cmdline, len + 1, root);
381 if (ret < 0 || ret > len) {
382 pr_err("Failed to print extra kernel cmdline.\n");
383 memblock_free(new_cmdline, len + 1);
390 static int __init bootconfig_params(char *param, char *val,
391 const char *unused, void *arg)
393 if (strcmp(param, "bootconfig") == 0) {
394 bootconfig_found = true;
399 static int __init warn_bootconfig(char *str)
401 /* The 'bootconfig' has been handled by bootconfig_params(). */
405 static void __init setup_boot_config(void)
407 static char tmp_cmdline[COMMAND_LINE_SIZE] __initdata;
408 const char *msg, *data;
413 /* Cut out the bootconfig data even if we have no bootconfig option */
414 data = get_boot_config_from_initrd(&size);
415 /* If there is no bootconfig in initrd, try embedded one. */
417 data = xbc_get_embedded_bootconfig(&size);
419 strscpy(tmp_cmdline, boot_command_line, COMMAND_LINE_SIZE);
420 err = parse_args("bootconfig", tmp_cmdline, NULL, 0, 0, 0, NULL,
423 if (IS_ERR(err) || !(bootconfig_found || IS_ENABLED(CONFIG_BOOT_CONFIG_FORCE)))
426 /* parse_args() stops at the next param of '--' and returns an address */
428 initargs_offs = err - tmp_cmdline;
431 /* If user intended to use bootconfig, show an error level message */
432 if (bootconfig_found)
433 pr_err("'bootconfig' found on command line, but no bootconfig found\n");
435 pr_info("No bootconfig data provided, so skipping bootconfig");
439 if (size >= XBC_DATA_MAX) {
440 pr_err("bootconfig size %ld greater than max size %d\n",
441 (long)size, XBC_DATA_MAX);
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: %ld bytes %d nodes\n", (long)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);
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 /* Handle params aliased to sysctls */
535 if (sysctl_is_alias(param))
538 repair_env_string(param, val);
540 /* Handle obsolete-style parameters */
541 if (obsolete_checksetup(param))
544 /* Unused module parameter. */
545 if (strnchr(param, len, '.'))
552 /* Environment option */
554 for (i = 0; envp_init[i]; i++) {
555 if (i == MAX_INIT_ENVS) {
559 if (!strncmp(param, envp_init[i], len+1))
562 envp_init[i] = param;
564 /* Command line option */
566 for (i = 0; argv_init[i]; i++) {
567 if (i == MAX_INIT_ARGS) {
568 panic_later = "init";
572 argv_init[i] = param;
577 static int __init init_setup(char *str)
581 execute_command = str;
583 * In case LILO is going to boot us with default command line,
584 * it prepends "auto" before the whole cmdline which makes
585 * the shell think it should execute a script with such name.
586 * So we ignore all arguments entered _before_ init=... [MJ]
588 for (i = 1; i < MAX_INIT_ARGS; i++)
592 __setup("init=", init_setup);
594 static int __init rdinit_setup(char *str)
598 ramdisk_execute_command = str;
599 /* See "auto" comment in init_setup */
600 for (i = 1; i < MAX_INIT_ARGS; i++)
604 __setup("rdinit=", rdinit_setup);
607 static inline void setup_nr_cpu_ids(void) { }
608 static inline void smp_prepare_cpus(unsigned int maxcpus) { }
612 * We need to store the untouched command line for future reference.
613 * We also need to store the touched command line since the parameter
614 * parsing is performed in place, and we should allow a component to
615 * store reference of name/value for future reference.
617 static void __init setup_command_line(char *command_line)
619 size_t len, xlen = 0, ilen = 0;
621 if (extra_command_line)
622 xlen = strlen(extra_command_line);
624 ilen = strlen(extra_init_args) + 4; /* for " -- " */
626 len = xlen + strlen(boot_command_line) + 1;
628 saved_command_line = memblock_alloc(len + ilen, SMP_CACHE_BYTES);
629 if (!saved_command_line)
630 panic("%s: Failed to allocate %zu bytes\n", __func__, len + ilen);
632 static_command_line = memblock_alloc(len, SMP_CACHE_BYTES);
633 if (!static_command_line)
634 panic("%s: Failed to allocate %zu bytes\n", __func__, len);
638 * We have to put extra_command_line before boot command
639 * lines because there could be dashes (separator of init
640 * command line) in the command lines.
642 strcpy(saved_command_line, extra_command_line);
643 strcpy(static_command_line, extra_command_line);
645 strcpy(saved_command_line + xlen, boot_command_line);
646 strcpy(static_command_line + xlen, command_line);
650 * Append supplemental init boot args to saved_command_line
651 * so that user can check what command line options passed
653 * The order should always be
654 * " -- "[bootconfig init-param][cmdline init-param]
657 len = xlen + initargs_offs;
658 strcpy(saved_command_line + len, extra_init_args);
659 len += ilen - 4; /* strlen(extra_init_args) */
660 strcpy(saved_command_line + len,
661 boot_command_line + initargs_offs - 1);
663 len = strlen(saved_command_line);
664 strcpy(saved_command_line + len, " -- ");
666 strcpy(saved_command_line + len, extra_init_args);
670 saved_command_line_len = strlen(saved_command_line);
674 * We need to finalize in a non-__init function or else race conditions
675 * between the root thread and the init thread may cause start_kernel to
676 * be reaped by free_initmem before the root thread has proceeded to
679 * gcc-3.4 accidentally inlines this function, so use noinline.
682 static __initdata DECLARE_COMPLETION(kthreadd_done);
684 noinline void __ref __noreturn rest_init(void)
686 struct task_struct *tsk;
689 rcu_scheduler_starting();
691 * We need to spawn init first so that it obtains pid 1, however
692 * the init task will end up wanting to create kthreads, which, if
693 * we schedule it before we create kthreadd, will OOPS.
695 pid = user_mode_thread(kernel_init, NULL, CLONE_FS);
697 * Pin init on the boot CPU. Task migration is not properly working
698 * until sched_init_smp() has been run. It will set the allowed
699 * CPUs for init to the non isolated CPUs.
702 tsk = find_task_by_pid_ns(pid, &init_pid_ns);
703 tsk->flags |= PF_NO_SETAFFINITY;
704 set_cpus_allowed_ptr(tsk, cpumask_of(smp_processor_id()));
707 numa_default_policy();
708 pid = kernel_thread(kthreadd, NULL, NULL, CLONE_FS | CLONE_FILES);
710 kthreadd_task = find_task_by_pid_ns(pid, &init_pid_ns);
714 * Enable might_sleep() and smp_processor_id() checks.
715 * They cannot be enabled earlier because with CONFIG_PREEMPTION=y
716 * kernel_thread() would trigger might_sleep() splats. With
717 * CONFIG_PREEMPT_VOLUNTARY=y the init task might have scheduled
718 * already, but it's stuck on the kthreadd_done completion.
720 system_state = SYSTEM_SCHEDULING;
722 complete(&kthreadd_done);
725 * The boot idle thread must execute schedule()
726 * at least once to get things moving:
728 schedule_preempt_disabled();
729 /* Call into cpu_idle with preempt disabled */
730 cpu_startup_entry(CPUHP_ONLINE);
733 /* Check for early params. */
734 static int __init do_early_param(char *param, char *val,
735 const char *unused, void *arg)
737 const struct obs_kernel_param *p;
739 for (p = __setup_start; p < __setup_end; p++) {
740 if ((p->early && parameq(param, p->str)) ||
741 (strcmp(param, "console") == 0 &&
742 strcmp(p->str, "earlycon") == 0)
744 if (p->setup_func(val) != 0)
745 pr_warn("Malformed early option '%s'\n", param);
748 /* We accept everything at this stage. */
752 void __init parse_early_options(char *cmdline)
754 parse_args("early options", cmdline, NULL, 0, 0, 0, NULL,
758 /* Arch code calls this early on, or if not, just before other parsing. */
759 void __init parse_early_param(void)
761 static int done __initdata;
762 static char tmp_cmdline[COMMAND_LINE_SIZE] __initdata;
767 /* All fall through to do_early_param. */
768 strscpy(tmp_cmdline, boot_command_line, COMMAND_LINE_SIZE);
769 parse_early_options(tmp_cmdline);
773 void __init __weak arch_post_acpi_subsys_init(void) { }
775 void __init __weak smp_setup_processor_id(void)
779 void __init __weak smp_prepare_boot_cpu(void)
783 # if THREAD_SIZE >= PAGE_SIZE
784 void __init __weak thread_stack_cache_init(void)
789 void __init __weak poking_init(void) { }
791 void __init __weak pgtable_cache_init(void) { }
793 void __init __weak trap_init(void) { }
796 core_param(initcall_debug, initcall_debug, bool, 0644);
798 #ifdef TRACEPOINTS_ENABLED
799 static void __init initcall_debug_enable(void);
801 static inline void initcall_debug_enable(void)
806 #ifdef CONFIG_RANDOMIZE_KSTACK_OFFSET
807 DEFINE_STATIC_KEY_MAYBE_RO(CONFIG_RANDOMIZE_KSTACK_OFFSET_DEFAULT,
808 randomize_kstack_offset);
809 DEFINE_PER_CPU(u32, kstack_offset);
811 static int __init early_randomize_kstack_offset(char *buf)
816 ret = kstrtobool(buf, &bool_result);
821 static_branch_enable(&randomize_kstack_offset);
823 static_branch_disable(&randomize_kstack_offset);
826 early_param("randomize_kstack_offset", early_randomize_kstack_offset);
829 void __init __weak __noreturn arch_call_rest_init(void)
834 static void __init print_unknown_bootoptions(void)
836 char *unknown_options;
838 const char *const *p;
841 if (panic_later || (!argv_init[1] && !envp_init[2]))
845 * Determine how many options we have to print out, plus a space
848 len = 1; /* null terminator */
849 for (p = &argv_init[1]; *p; p++) {
853 for (p = &envp_init[2]; *p; p++) {
858 unknown_options = memblock_alloc(len, SMP_CACHE_BYTES);
859 if (!unknown_options) {
860 pr_err("%s: Failed to allocate %zu bytes\n",
864 end = unknown_options;
866 for (p = &argv_init[1]; *p; p++)
867 end += sprintf(end, " %s", *p);
868 for (p = &envp_init[2]; *p; p++)
869 end += sprintf(end, " %s", *p);
871 /* Start at unknown_options[1] to skip the initial space */
872 pr_notice("Unknown kernel command line parameters \"%s\", will be passed to user space.\n",
873 &unknown_options[1]);
874 memblock_free(unknown_options, len);
877 asmlinkage __visible __init __no_sanitize_address __noreturn __no_stack_protector
878 void start_kernel(void)
883 set_task_stack_end_magic(&init_task);
884 smp_setup_processor_id();
885 debug_objects_early_init();
886 init_vmlinux_build_id();
891 early_boot_irqs_disabled = true;
894 * Interrupts are still disabled. Do necessary setups, then
899 pr_notice("%s", linux_banner);
900 early_security_init();
901 setup_arch(&command_line);
903 setup_command_line(command_line);
905 setup_per_cpu_areas();
906 smp_prepare_boot_cpu(); /* arch-specific boot-cpu hooks */
907 boot_cpu_hotplug_init();
909 pr_notice("Kernel command line: %s\n", saved_command_line);
910 /* parameters may set static keys */
913 after_dashes = parse_args("Booting kernel",
914 static_command_line, __start___param,
915 __stop___param - __start___param,
916 -1, -1, NULL, &unknown_bootoption);
917 print_unknown_bootoptions();
918 if (!IS_ERR_OR_NULL(after_dashes))
919 parse_args("Setting init args", after_dashes, NULL, 0, -1, -1,
922 parse_args("Setting extra init args", extra_init_args,
923 NULL, 0, -1, -1, NULL, set_init_arg);
925 /* Architectural and non-timekeeping rng init, before allocator init */
926 random_init_early(command_line);
929 * These use large bootmem allocations and must precede
930 * initalization of page allocator
933 vfs_caches_init_early();
940 /* trace_printk can be enabled here */
944 * Set up the scheduler prior starting any interrupts (such as the
945 * timer interrupt). Full topology setup happens at smp_init()
946 * time - but meanwhile we still have a functioning scheduler.
950 if (WARN(!irqs_disabled(),
951 "Interrupts were enabled *very* early, fixing it\n"))
957 * Set up housekeeping before setting up workqueues to allow the unbound
958 * workqueue to take non-housekeeping into account.
963 * Allow workqueue creation and work item queueing/cancelling
964 * early. Work item execution depends on kthreads and starts after
967 workqueue_init_early();
971 /* Trace events are available after this */
975 initcall_debug_enable();
977 context_tracking_init();
978 /* init some links before init_ISA_irqs() */
990 /* This must be after timekeeping is initialized */
993 /* These make use of the fully initialized rng */
995 boot_init_stack_canary();
999 call_function_init();
1000 WARN(!irqs_disabled(), "Interrupts were enabled early\n");
1002 early_boot_irqs_disabled = false;
1005 kmem_cache_init_late();
1008 * HACK ALERT! This is early. We're enabling the console before
1009 * we've done PCI setups etc, and console_init() must be aware of
1010 * this. But we do want output early, in case something goes wrong.
1014 panic("Too many boot %s vars at `%s'", panic_later,
1020 * Need to run this when irqs are enabled, because it wants
1021 * to self-test [hard/soft]-irqs on/off lock inversion bugs
1026 #ifdef CONFIG_BLK_DEV_INITRD
1027 if (initrd_start && !initrd_below_start_ok &&
1028 page_to_pfn(virt_to_page((void *)initrd_start)) < min_low_pfn) {
1029 pr_crit("initrd overwritten (0x%08lx < 0x%08lx) - disabling it.\n",
1030 page_to_pfn(virt_to_page((void *)initrd_start)),
1035 setup_per_cpu_pageset();
1043 arch_cpu_finalize_init();
1048 if (efi_enabled(EFI_RUNTIME_SERVICES))
1049 efi_enter_virtual_mode();
1051 thread_stack_cache_init();
1069 taskstats_init_early();
1072 acpi_subsystem_init();
1073 arch_post_acpi_subsys_init();
1076 /* Do the rest non-__init'ed, we're now alive */
1077 arch_call_rest_init();
1080 * Avoid stack canaries in callers of boot_init_stack_canary for gcc-10
1083 #if !__has_attribute(__no_stack_protector__)
1084 prevent_tail_call_optimization();
1088 /* Call all constructor functions linked into the kernel. */
1089 static void __init do_ctors(void)
1092 * For UML, the constructors have already been called by the
1093 * normal setup code as it's just a normal ELF binary, so we
1094 * cannot do it again - but we do need CONFIG_CONSTRUCTORS
1095 * even on UML for modules.
1097 #if defined(CONFIG_CONSTRUCTORS) && !defined(CONFIG_UML)
1098 ctor_fn_t *fn = (ctor_fn_t *) __ctors_start;
1100 for (; fn < (ctor_fn_t *) __ctors_end; fn++)
1105 #ifdef CONFIG_KALLSYMS
1106 struct blacklist_entry {
1107 struct list_head next;
1111 static __initdata_or_module LIST_HEAD(blacklisted_initcalls);
1113 static int __init initcall_blacklist(char *str)
1116 struct blacklist_entry *entry;
1118 /* str argument is a comma-separated list of functions */
1120 str_entry = strsep(&str, ",");
1122 pr_debug("blacklisting initcall %s\n", str_entry);
1123 entry = memblock_alloc(sizeof(*entry),
1126 panic("%s: Failed to allocate %zu bytes\n",
1127 __func__, sizeof(*entry));
1128 entry->buf = memblock_alloc(strlen(str_entry) + 1,
1131 panic("%s: Failed to allocate %zu bytes\n",
1132 __func__, strlen(str_entry) + 1);
1133 strcpy(entry->buf, str_entry);
1134 list_add(&entry->next, &blacklisted_initcalls);
1136 } while (str_entry);
1141 static bool __init_or_module initcall_blacklisted(initcall_t fn)
1143 struct blacklist_entry *entry;
1144 char fn_name[KSYM_SYMBOL_LEN];
1147 if (list_empty(&blacklisted_initcalls))
1150 addr = (unsigned long) dereference_function_descriptor(fn);
1151 sprint_symbol_no_offset(fn_name, addr);
1154 * fn will be "function_name [module_name]" where [module_name] is not
1155 * displayed for built-in init functions. Strip off the [module_name].
1157 strreplace(fn_name, ' ', '\0');
1159 list_for_each_entry(entry, &blacklisted_initcalls, next) {
1160 if (!strcmp(fn_name, entry->buf)) {
1161 pr_debug("initcall %s blacklisted\n", fn_name);
1169 static int __init initcall_blacklist(char *str)
1171 pr_warn("initcall_blacklist requires CONFIG_KALLSYMS\n");
1175 static bool __init_or_module initcall_blacklisted(initcall_t fn)
1180 __setup("initcall_blacklist=", initcall_blacklist);
1182 static __init_or_module void
1183 trace_initcall_start_cb(void *data, initcall_t fn)
1185 ktime_t *calltime = data;
1187 printk(KERN_DEBUG "calling %pS @ %i\n", fn, task_pid_nr(current));
1188 *calltime = ktime_get();
1191 static __init_or_module void
1192 trace_initcall_finish_cb(void *data, initcall_t fn, int ret)
1194 ktime_t rettime, *calltime = data;
1196 rettime = ktime_get();
1197 printk(KERN_DEBUG "initcall %pS returned %d after %lld usecs\n",
1198 fn, ret, (unsigned long long)ktime_us_delta(rettime, *calltime));
1201 static ktime_t initcall_calltime;
1203 #ifdef TRACEPOINTS_ENABLED
1204 static void __init initcall_debug_enable(void)
1208 ret = register_trace_initcall_start(trace_initcall_start_cb,
1209 &initcall_calltime);
1210 ret |= register_trace_initcall_finish(trace_initcall_finish_cb,
1211 &initcall_calltime);
1212 WARN(ret, "Failed to register initcall tracepoints\n");
1214 # define do_trace_initcall_start trace_initcall_start
1215 # define do_trace_initcall_finish trace_initcall_finish
1217 static inline void do_trace_initcall_start(initcall_t fn)
1219 if (!initcall_debug)
1221 trace_initcall_start_cb(&initcall_calltime, fn);
1223 static inline void do_trace_initcall_finish(initcall_t fn, int ret)
1225 if (!initcall_debug)
1227 trace_initcall_finish_cb(&initcall_calltime, fn, ret);
1229 #endif /* !TRACEPOINTS_ENABLED */
1231 int __init_or_module do_one_initcall(initcall_t fn)
1233 int count = preempt_count();
1237 if (initcall_blacklisted(fn))
1240 do_trace_initcall_start(fn);
1242 do_trace_initcall_finish(fn, ret);
1246 if (preempt_count() != count) {
1247 sprintf(msgbuf, "preemption imbalance ");
1248 preempt_count_set(count);
1250 if (irqs_disabled()) {
1251 strlcat(msgbuf, "disabled interrupts ", sizeof(msgbuf));
1254 WARN(msgbuf[0], "initcall %pS returned with %s\n", fn, msgbuf);
1256 add_latent_entropy();
1261 static initcall_entry_t *initcall_levels[] __initdata = {
1273 /* Keep these in sync with initcalls in include/linux/init.h */
1274 static const char *initcall_level_names[] __initdata = {
1285 static int __init ignore_unknown_bootoption(char *param, char *val,
1286 const char *unused, void *arg)
1291 static void __init do_initcall_level(int level, char *command_line)
1293 initcall_entry_t *fn;
1295 parse_args(initcall_level_names[level],
1296 command_line, __start___param,
1297 __stop___param - __start___param,
1299 NULL, ignore_unknown_bootoption);
1301 trace_initcall_level(initcall_level_names[level]);
1302 for (fn = initcall_levels[level]; fn < initcall_levels[level+1]; fn++)
1303 do_one_initcall(initcall_from_entry(fn));
1306 static void __init do_initcalls(void)
1309 size_t len = saved_command_line_len + 1;
1312 command_line = kzalloc(len, GFP_KERNEL);
1314 panic("%s: Failed to allocate %zu bytes\n", __func__, len);
1316 for (level = 0; level < ARRAY_SIZE(initcall_levels) - 1; level++) {
1317 /* Parser modifies command_line, restore it each time */
1318 strcpy(command_line, saved_command_line);
1319 do_initcall_level(level, command_line);
1322 kfree(command_line);
1326 * Ok, the machine is now initialized. None of the devices
1327 * have been touched yet, but the CPU subsystem is up and
1328 * running, and memory and process management works.
1330 * Now we can finally start doing some real work..
1332 static void __init do_basic_setup(void)
1341 static void __init do_pre_smp_initcalls(void)
1343 initcall_entry_t *fn;
1345 trace_initcall_level("early");
1346 for (fn = __initcall_start; fn < __initcall0_start; fn++)
1347 do_one_initcall(initcall_from_entry(fn));
1350 static int run_init_process(const char *init_filename)
1352 const char *const *p;
1354 argv_init[0] = init_filename;
1355 pr_info("Run %s as init process\n", init_filename);
1356 pr_debug(" with arguments:\n");
1357 for (p = argv_init; *p; p++)
1358 pr_debug(" %s\n", *p);
1359 pr_debug(" with environment:\n");
1360 for (p = envp_init; *p; p++)
1361 pr_debug(" %s\n", *p);
1362 return kernel_execve(init_filename, argv_init, envp_init);
1365 static int try_to_run_init_process(const char *init_filename)
1369 ret = run_init_process(init_filename);
1371 if (ret && ret != -ENOENT) {
1372 pr_err("Starting init: %s exists but couldn't execute it (error %d)\n",
1373 init_filename, ret);
1379 static noinline void __init kernel_init_freeable(void);
1381 #if defined(CONFIG_STRICT_KERNEL_RWX) || defined(CONFIG_STRICT_MODULE_RWX)
1382 bool rodata_enabled __ro_after_init = true;
1384 #ifndef arch_parse_debug_rodata
1385 static inline bool arch_parse_debug_rodata(char *str) { return false; }
1388 static int __init set_debug_rodata(char *str)
1390 if (arch_parse_debug_rodata(str))
1393 if (str && !strcmp(str, "on"))
1394 rodata_enabled = true;
1395 else if (str && !strcmp(str, "off"))
1396 rodata_enabled = false;
1398 pr_warn("Invalid option string for rodata: '%s'\n", str);
1401 early_param("rodata", set_debug_rodata);
1404 static void mark_readonly(void)
1406 if (IS_ENABLED(CONFIG_STRICT_KERNEL_RWX) && rodata_enabled) {
1408 * load_module() results in W+X mappings, which are cleaned
1409 * up with call_rcu(). Let's make sure that queued work is
1410 * flushed so that we don't hit false positives looking for
1411 * insecure pages which are W+X.
1416 } else if (IS_ENABLED(CONFIG_STRICT_KERNEL_RWX)) {
1417 pr_info("Kernel memory protection disabled.\n");
1418 } else if (IS_ENABLED(CONFIG_ARCH_HAS_STRICT_KERNEL_RWX)) {
1419 pr_warn("Kernel memory protection not selected by kernel config.\n");
1421 pr_warn("This architecture does not have kernel memory protection.\n");
1425 void __weak free_initmem(void)
1427 free_initmem_default(POISON_FREE_INITMEM);
1430 static int __ref kernel_init(void *unused)
1435 * Wait until kthreadd is all set-up.
1437 wait_for_completion(&kthreadd_done);
1439 kernel_init_freeable();
1440 /* need to finish all async __init code before freeing the memory */
1441 async_synchronize_full();
1443 system_state = SYSTEM_FREEING_INITMEM;
1444 kprobe_free_init_mem();
1445 ftrace_free_init_mem();
1446 kgdb_free_init_mem();
1452 * Kernel mappings are now finalized - update the userspace page-table
1457 system_state = SYSTEM_RUNNING;
1458 numa_default_policy();
1460 rcu_end_inkernel_boot();
1464 if (ramdisk_execute_command) {
1465 ret = run_init_process(ramdisk_execute_command);
1468 pr_err("Failed to execute %s (error %d)\n",
1469 ramdisk_execute_command, ret);
1473 * We try each of these until one succeeds.
1475 * The Bourne shell can be used instead of init if we are
1476 * trying to recover a really broken machine.
1478 if (execute_command) {
1479 ret = run_init_process(execute_command);
1482 panic("Requested init %s failed (error %d).",
1483 execute_command, ret);
1486 if (CONFIG_DEFAULT_INIT[0] != '\0') {
1487 ret = run_init_process(CONFIG_DEFAULT_INIT);
1489 pr_err("Default init %s failed (error %d)\n",
1490 CONFIG_DEFAULT_INIT, ret);
1495 if (!try_to_run_init_process("/sbin/init") ||
1496 !try_to_run_init_process("/etc/init") ||
1497 !try_to_run_init_process("/bin/init") ||
1498 !try_to_run_init_process("/bin/sh"))
1501 panic("No working init found. Try passing init= option to kernel. "
1502 "See Linux Documentation/admin-guide/init.rst for guidance.");
1505 /* Open /dev/console, for stdin/stdout/stderr, this should never fail */
1506 void __init console_on_rootfs(void)
1508 struct file *file = filp_open("/dev/console", O_RDWR, 0);
1511 pr_err("Warning: unable to open an initial console.\n");
1520 static noinline void __init kernel_init_freeable(void)
1522 /* Now the scheduler is fully set up and can do blocking allocations */
1523 gfp_allowed_mask = __GFP_BITS_MASK;
1526 * init can allocate pages on any node
1528 set_mems_allowed(node_states[N_MEMORY]);
1530 cad_pid = get_pid(task_pid(current));
1532 smp_prepare_cpus(setup_max_cpus);
1536 init_mm_internals();
1538 rcu_init_tasks_generic();
1539 do_pre_smp_initcalls();
1540 lockup_detector_init();
1545 workqueue_init_topology();
1548 page_alloc_init_late();
1552 kunit_run_all_tests();
1554 wait_for_initramfs();
1555 console_on_rootfs();
1558 * check if there is an early userspace init. If yes, let it do all
1561 if (init_eaccess(ramdisk_execute_command) != 0) {
1562 ramdisk_execute_command = NULL;
1563 prepare_namespace();
1567 * Ok, we have completed the initial bootup, and
1568 * we're essentially up and running. Get rid of the
1569 * initmem segments and start the user-mode stuff..
1571 * rootfs is available now, try loading the public keys
1572 * and default modules
1575 integrity_load_keys();