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) {
348 ret = snprintf(buf, rest(buf, end), "%s=\"%s\" ",
356 return buf - (end - size);
360 /* Make an extra command line under given key word */
361 static char * __init xbc_make_cmdline(const char *key)
363 struct xbc_node *root;
367 root = xbc_find_node(key);
371 /* Count required buffer size */
372 len = xbc_snprint_cmdline(NULL, 0, root);
376 new_cmdline = memblock_alloc(len + 1, SMP_CACHE_BYTES);
378 pr_err("Failed to allocate memory for extra kernel cmdline.\n");
382 ret = xbc_snprint_cmdline(new_cmdline, len + 1, root);
383 if (ret < 0 || ret > len) {
384 pr_err("Failed to print extra kernel cmdline.\n");
385 memblock_free(new_cmdline, len + 1);
392 static int __init bootconfig_params(char *param, char *val,
393 const char *unused, void *arg)
395 if (strcmp(param, "bootconfig") == 0) {
396 bootconfig_found = true;
401 static int __init warn_bootconfig(char *str)
403 /* The 'bootconfig' has been handled by bootconfig_params(). */
407 static void __init setup_boot_config(void)
409 static char tmp_cmdline[COMMAND_LINE_SIZE] __initdata;
410 const char *msg, *data;
415 /* Cut out the bootconfig data even if we have no bootconfig option */
416 data = get_boot_config_from_initrd(&size);
417 /* If there is no bootconfig in initrd, try embedded one. */
419 data = xbc_get_embedded_bootconfig(&size);
421 strscpy(tmp_cmdline, boot_command_line, COMMAND_LINE_SIZE);
422 err = parse_args("bootconfig", tmp_cmdline, NULL, 0, 0, 0, NULL,
425 if (IS_ERR(err) || !(bootconfig_found || IS_ENABLED(CONFIG_BOOT_CONFIG_FORCE)))
428 /* parse_args() stops at the next param of '--' and returns an address */
430 initargs_offs = err - tmp_cmdline;
433 /* If user intended to use bootconfig, show an error level message */
434 if (bootconfig_found)
435 pr_err("'bootconfig' found on command line, but no bootconfig found\n");
437 pr_info("No bootconfig data provided, so skipping bootconfig");
441 if (size >= XBC_DATA_MAX) {
442 pr_err("bootconfig size %ld greater than max size %d\n",
443 (long)size, XBC_DATA_MAX);
447 ret = xbc_init(data, size, &msg, &pos);
450 pr_err("Failed to init bootconfig: %s.\n", msg);
452 pr_err("Failed to parse bootconfig: %s at %d.\n",
455 xbc_get_info(&ret, NULL);
456 pr_info("Load bootconfig: %ld bytes %d nodes\n", (long)size, ret);
457 /* keys starting with "kernel." are passed via cmdline */
458 extra_command_line = xbc_make_cmdline("kernel");
459 /* Also, "init." keys are init arguments */
460 extra_init_args = xbc_make_cmdline("init");
465 static void __init exit_boot_config(void)
470 #else /* !CONFIG_BOOT_CONFIG */
472 static void __init setup_boot_config(void)
474 /* Remove bootconfig data from initrd */
475 get_boot_config_from_initrd(NULL);
478 static int __init warn_bootconfig(char *str)
480 pr_warn("WARNING: 'bootconfig' found on the kernel command line but CONFIG_BOOT_CONFIG is not set.\n");
484 #define exit_boot_config() do {} while (0)
486 #endif /* CONFIG_BOOT_CONFIG */
488 early_param("bootconfig", warn_bootconfig);
490 /* Change NUL term back to "=", to make "param" the whole string. */
491 static void __init repair_env_string(char *param, char *val)
494 /* param=val or param="val"? */
495 if (val == param+strlen(param)+1)
497 else if (val == param+strlen(param)+2) {
499 memmove(val-1, val, strlen(val)+1);
505 /* Anything after -- gets handed straight to init. */
506 static int __init set_init_arg(char *param, char *val,
507 const char *unused, void *arg)
514 repair_env_string(param, val);
516 for (i = 0; argv_init[i]; i++) {
517 if (i == MAX_INIT_ARGS) {
518 panic_later = "init";
523 argv_init[i] = param;
528 * Unknown boot options get handed to init, unless they look like
529 * unused parameters (modprobe will find them in /proc/cmdline).
531 static int __init unknown_bootoption(char *param, char *val,
532 const char *unused, void *arg)
534 size_t len = strlen(param);
536 /* Handle params aliased to sysctls */
537 if (sysctl_is_alias(param))
540 repair_env_string(param, val);
542 /* Handle obsolete-style parameters */
543 if (obsolete_checksetup(param))
546 /* Unused module parameter. */
547 if (strnchr(param, len, '.'))
554 /* Environment option */
556 for (i = 0; envp_init[i]; i++) {
557 if (i == MAX_INIT_ENVS) {
561 if (!strncmp(param, envp_init[i], len+1))
564 envp_init[i] = param;
566 /* Command line option */
568 for (i = 0; argv_init[i]; i++) {
569 if (i == MAX_INIT_ARGS) {
570 panic_later = "init";
574 argv_init[i] = param;
579 static int __init init_setup(char *str)
583 execute_command = str;
585 * In case LILO is going to boot us with default command line,
586 * it prepends "auto" before the whole cmdline which makes
587 * the shell think it should execute a script with such name.
588 * So we ignore all arguments entered _before_ init=... [MJ]
590 for (i = 1; i < MAX_INIT_ARGS; i++)
594 __setup("init=", init_setup);
596 static int __init rdinit_setup(char *str)
600 ramdisk_execute_command = str;
601 /* See "auto" comment in init_setup */
602 for (i = 1; i < MAX_INIT_ARGS; i++)
606 __setup("rdinit=", rdinit_setup);
609 static inline void setup_nr_cpu_ids(void) { }
610 static inline void smp_prepare_cpus(unsigned int maxcpus) { }
614 * We need to store the untouched command line for future reference.
615 * We also need to store the touched command line since the parameter
616 * parsing is performed in place, and we should allow a component to
617 * store reference of name/value for future reference.
619 static void __init setup_command_line(char *command_line)
621 size_t len, xlen = 0, ilen = 0;
623 if (extra_command_line)
624 xlen = strlen(extra_command_line);
626 ilen = strlen(extra_init_args) + 4; /* for " -- " */
628 len = xlen + strlen(boot_command_line) + 1;
630 saved_command_line = memblock_alloc(len + ilen, SMP_CACHE_BYTES);
631 if (!saved_command_line)
632 panic("%s: Failed to allocate %zu bytes\n", __func__, len + ilen);
634 static_command_line = memblock_alloc(len, SMP_CACHE_BYTES);
635 if (!static_command_line)
636 panic("%s: Failed to allocate %zu bytes\n", __func__, len);
640 * We have to put extra_command_line before boot command
641 * lines because there could be dashes (separator of init
642 * command line) in the command lines.
644 strcpy(saved_command_line, extra_command_line);
645 strcpy(static_command_line, extra_command_line);
647 strcpy(saved_command_line + xlen, boot_command_line);
648 strcpy(static_command_line + xlen, command_line);
652 * Append supplemental init boot args to saved_command_line
653 * so that user can check what command line options passed
655 * The order should always be
656 * " -- "[bootconfig init-param][cmdline init-param]
659 len = xlen + initargs_offs;
660 strcpy(saved_command_line + len, extra_init_args);
661 len += ilen - 4; /* strlen(extra_init_args) */
662 strcpy(saved_command_line + len,
663 boot_command_line + initargs_offs - 1);
665 len = strlen(saved_command_line);
666 strcpy(saved_command_line + len, " -- ");
668 strcpy(saved_command_line + len, extra_init_args);
672 saved_command_line_len = strlen(saved_command_line);
676 * We need to finalize in a non-__init function or else race conditions
677 * between the root thread and the init thread may cause start_kernel to
678 * be reaped by free_initmem before the root thread has proceeded to
681 * gcc-3.4 accidentally inlines this function, so use noinline.
684 static __initdata DECLARE_COMPLETION(kthreadd_done);
686 static noinline void __ref __noreturn rest_init(void)
688 struct task_struct *tsk;
691 rcu_scheduler_starting();
693 * We need to spawn init first so that it obtains pid 1, however
694 * the init task will end up wanting to create kthreads, which, if
695 * we schedule it before we create kthreadd, will OOPS.
697 pid = user_mode_thread(kernel_init, NULL, CLONE_FS);
699 * Pin init on the boot CPU. Task migration is not properly working
700 * until sched_init_smp() has been run. It will set the allowed
701 * CPUs for init to the non isolated CPUs.
704 tsk = find_task_by_pid_ns(pid, &init_pid_ns);
705 tsk->flags |= PF_NO_SETAFFINITY;
706 set_cpus_allowed_ptr(tsk, cpumask_of(smp_processor_id()));
709 numa_default_policy();
710 pid = kernel_thread(kthreadd, NULL, NULL, CLONE_FS | CLONE_FILES);
712 kthreadd_task = find_task_by_pid_ns(pid, &init_pid_ns);
716 * Enable might_sleep() and smp_processor_id() checks.
717 * They cannot be enabled earlier because with CONFIG_PREEMPTION=y
718 * kernel_thread() would trigger might_sleep() splats. With
719 * CONFIG_PREEMPT_VOLUNTARY=y the init task might have scheduled
720 * already, but it's stuck on the kthreadd_done completion.
722 system_state = SYSTEM_SCHEDULING;
724 complete(&kthreadd_done);
727 * The boot idle thread must execute schedule()
728 * at least once to get things moving:
730 schedule_preempt_disabled();
731 /* Call into cpu_idle with preempt disabled */
732 cpu_startup_entry(CPUHP_ONLINE);
735 /* Check for early params. */
736 static int __init do_early_param(char *param, char *val,
737 const char *unused, void *arg)
739 const struct obs_kernel_param *p;
741 for (p = __setup_start; p < __setup_end; p++) {
742 if ((p->early && parameq(param, p->str)) ||
743 (strcmp(param, "console") == 0 &&
744 strcmp(p->str, "earlycon") == 0)
746 if (p->setup_func(val) != 0)
747 pr_warn("Malformed early option '%s'\n", param);
750 /* We accept everything at this stage. */
754 void __init parse_early_options(char *cmdline)
756 parse_args("early options", cmdline, NULL, 0, 0, 0, NULL,
760 /* Arch code calls this early on, or if not, just before other parsing. */
761 void __init parse_early_param(void)
763 static int done __initdata;
764 static char tmp_cmdline[COMMAND_LINE_SIZE] __initdata;
769 /* All fall through to do_early_param. */
770 strscpy(tmp_cmdline, boot_command_line, COMMAND_LINE_SIZE);
771 parse_early_options(tmp_cmdline);
775 void __init __weak arch_post_acpi_subsys_init(void) { }
777 void __init __weak smp_setup_processor_id(void)
781 void __init __weak smp_prepare_boot_cpu(void)
785 # if THREAD_SIZE >= PAGE_SIZE
786 void __init __weak thread_stack_cache_init(void)
791 void __init __weak poking_init(void) { }
793 void __init __weak pgtable_cache_init(void) { }
795 void __init __weak trap_init(void) { }
798 core_param(initcall_debug, initcall_debug, bool, 0644);
800 #ifdef TRACEPOINTS_ENABLED
801 static void __init initcall_debug_enable(void);
803 static inline void initcall_debug_enable(void)
808 #ifdef CONFIG_RANDOMIZE_KSTACK_OFFSET
809 DEFINE_STATIC_KEY_MAYBE_RO(CONFIG_RANDOMIZE_KSTACK_OFFSET_DEFAULT,
810 randomize_kstack_offset);
811 DEFINE_PER_CPU(u32, kstack_offset);
813 static int __init early_randomize_kstack_offset(char *buf)
818 ret = kstrtobool(buf, &bool_result);
823 static_branch_enable(&randomize_kstack_offset);
825 static_branch_disable(&randomize_kstack_offset);
828 early_param("randomize_kstack_offset", early_randomize_kstack_offset);
831 static void __init print_unknown_bootoptions(void)
833 char *unknown_options;
835 const char *const *p;
838 if (panic_later || (!argv_init[1] && !envp_init[2]))
842 * Determine how many options we have to print out, plus a space
845 len = 1; /* null terminator */
846 for (p = &argv_init[1]; *p; p++) {
850 for (p = &envp_init[2]; *p; p++) {
855 unknown_options = memblock_alloc(len, SMP_CACHE_BYTES);
856 if (!unknown_options) {
857 pr_err("%s: Failed to allocate %zu bytes\n",
861 end = unknown_options;
863 for (p = &argv_init[1]; *p; p++)
864 end += sprintf(end, " %s", *p);
865 for (p = &envp_init[2]; *p; p++)
866 end += sprintf(end, " %s", *p);
868 /* Start at unknown_options[1] to skip the initial space */
869 pr_notice("Unknown kernel command line parameters \"%s\", will be passed to user space.\n",
870 &unknown_options[1]);
871 memblock_free(unknown_options, len);
874 asmlinkage __visible __init __no_sanitize_address __noreturn __no_stack_protector
875 void start_kernel(void)
880 set_task_stack_end_magic(&init_task);
881 smp_setup_processor_id();
882 debug_objects_early_init();
883 init_vmlinux_build_id();
888 early_boot_irqs_disabled = true;
891 * Interrupts are still disabled. Do necessary setups, then
896 pr_notice("%s", linux_banner);
897 early_security_init();
898 setup_arch(&command_line);
900 setup_command_line(command_line);
902 setup_per_cpu_areas();
903 smp_prepare_boot_cpu(); /* arch-specific boot-cpu hooks */
904 boot_cpu_hotplug_init();
906 pr_notice("Kernel command line: %s\n", saved_command_line);
907 /* parameters may set static keys */
910 after_dashes = parse_args("Booting kernel",
911 static_command_line, __start___param,
912 __stop___param - __start___param,
913 -1, -1, NULL, &unknown_bootoption);
914 print_unknown_bootoptions();
915 if (!IS_ERR_OR_NULL(after_dashes))
916 parse_args("Setting init args", after_dashes, NULL, 0, -1, -1,
919 parse_args("Setting extra init args", extra_init_args,
920 NULL, 0, -1, -1, NULL, set_init_arg);
922 /* Architectural and non-timekeeping rng init, before allocator init */
923 random_init_early(command_line);
926 * These use large bootmem allocations and must precede
927 * initalization of page allocator
930 vfs_caches_init_early();
937 /* trace_printk can be enabled here */
941 * Set up the scheduler prior starting any interrupts (such as the
942 * timer interrupt). Full topology setup happens at smp_init()
943 * time - but meanwhile we still have a functioning scheduler.
947 if (WARN(!irqs_disabled(),
948 "Interrupts were enabled *very* early, fixing it\n"))
954 * Set up housekeeping before setting up workqueues to allow the unbound
955 * workqueue to take non-housekeeping into account.
960 * Allow workqueue creation and work item queueing/cancelling
961 * early. Work item execution depends on kthreads and starts after
964 workqueue_init_early();
968 /* Trace events are available after this */
972 initcall_debug_enable();
974 context_tracking_init();
975 /* init some links before init_ISA_irqs() */
987 /* This must be after timekeeping is initialized */
990 /* These make use of the fully initialized rng */
992 boot_init_stack_canary();
996 call_function_init();
997 WARN(!irqs_disabled(), "Interrupts were enabled early\n");
999 early_boot_irqs_disabled = false;
1002 kmem_cache_init_late();
1005 * HACK ALERT! This is early. We're enabling the console before
1006 * we've done PCI setups etc, and console_init() must be aware of
1007 * this. But we do want output early, in case something goes wrong.
1011 panic("Too many boot %s vars at `%s'", panic_later,
1017 * Need to run this when irqs are enabled, because it wants
1018 * to self-test [hard/soft]-irqs on/off lock inversion bugs
1023 #ifdef CONFIG_BLK_DEV_INITRD
1024 if (initrd_start && !initrd_below_start_ok &&
1025 page_to_pfn(virt_to_page((void *)initrd_start)) < min_low_pfn) {
1026 pr_crit("initrd overwritten (0x%08lx < 0x%08lx) - disabling it.\n",
1027 page_to_pfn(virt_to_page((void *)initrd_start)),
1032 setup_per_cpu_pageset();
1040 arch_cpu_finalize_init();
1045 if (efi_enabled(EFI_RUNTIME_SERVICES))
1046 efi_enter_virtual_mode();
1048 thread_stack_cache_init();
1066 taskstats_init_early();
1069 acpi_subsystem_init();
1070 arch_post_acpi_subsys_init();
1073 /* Do the rest non-__init'ed, we're now alive */
1077 * Avoid stack canaries in callers of boot_init_stack_canary for gcc-10
1080 #if !__has_attribute(__no_stack_protector__)
1081 prevent_tail_call_optimization();
1085 /* Call all constructor functions linked into the kernel. */
1086 static void __init do_ctors(void)
1089 * For UML, the constructors have already been called by the
1090 * normal setup code as it's just a normal ELF binary, so we
1091 * cannot do it again - but we do need CONFIG_CONSTRUCTORS
1092 * even on UML for modules.
1094 #if defined(CONFIG_CONSTRUCTORS) && !defined(CONFIG_UML)
1095 ctor_fn_t *fn = (ctor_fn_t *) __ctors_start;
1097 for (; fn < (ctor_fn_t *) __ctors_end; fn++)
1102 #ifdef CONFIG_KALLSYMS
1103 struct blacklist_entry {
1104 struct list_head next;
1108 static __initdata_or_module LIST_HEAD(blacklisted_initcalls);
1110 static int __init initcall_blacklist(char *str)
1113 struct blacklist_entry *entry;
1115 /* str argument is a comma-separated list of functions */
1117 str_entry = strsep(&str, ",");
1119 pr_debug("blacklisting initcall %s\n", str_entry);
1120 entry = memblock_alloc(sizeof(*entry),
1123 panic("%s: Failed to allocate %zu bytes\n",
1124 __func__, sizeof(*entry));
1125 entry->buf = memblock_alloc(strlen(str_entry) + 1,
1128 panic("%s: Failed to allocate %zu bytes\n",
1129 __func__, strlen(str_entry) + 1);
1130 strcpy(entry->buf, str_entry);
1131 list_add(&entry->next, &blacklisted_initcalls);
1133 } while (str_entry);
1138 static bool __init_or_module initcall_blacklisted(initcall_t fn)
1140 struct blacklist_entry *entry;
1141 char fn_name[KSYM_SYMBOL_LEN];
1144 if (list_empty(&blacklisted_initcalls))
1147 addr = (unsigned long) dereference_function_descriptor(fn);
1148 sprint_symbol_no_offset(fn_name, addr);
1151 * fn will be "function_name [module_name]" where [module_name] is not
1152 * displayed for built-in init functions. Strip off the [module_name].
1154 strreplace(fn_name, ' ', '\0');
1156 list_for_each_entry(entry, &blacklisted_initcalls, next) {
1157 if (!strcmp(fn_name, entry->buf)) {
1158 pr_debug("initcall %s blacklisted\n", fn_name);
1166 static int __init initcall_blacklist(char *str)
1168 pr_warn("initcall_blacklist requires CONFIG_KALLSYMS\n");
1172 static bool __init_or_module initcall_blacklisted(initcall_t fn)
1177 __setup("initcall_blacklist=", initcall_blacklist);
1179 static __init_or_module void
1180 trace_initcall_start_cb(void *data, initcall_t fn)
1182 ktime_t *calltime = data;
1184 printk(KERN_DEBUG "calling %pS @ %i\n", fn, task_pid_nr(current));
1185 *calltime = ktime_get();
1188 static __init_or_module void
1189 trace_initcall_finish_cb(void *data, initcall_t fn, int ret)
1191 ktime_t rettime, *calltime = data;
1193 rettime = ktime_get();
1194 printk(KERN_DEBUG "initcall %pS returned %d after %lld usecs\n",
1195 fn, ret, (unsigned long long)ktime_us_delta(rettime, *calltime));
1198 static ktime_t initcall_calltime;
1200 #ifdef TRACEPOINTS_ENABLED
1201 static void __init initcall_debug_enable(void)
1205 ret = register_trace_initcall_start(trace_initcall_start_cb,
1206 &initcall_calltime);
1207 ret |= register_trace_initcall_finish(trace_initcall_finish_cb,
1208 &initcall_calltime);
1209 WARN(ret, "Failed to register initcall tracepoints\n");
1211 # define do_trace_initcall_start trace_initcall_start
1212 # define do_trace_initcall_finish trace_initcall_finish
1214 static inline void do_trace_initcall_start(initcall_t fn)
1216 if (!initcall_debug)
1218 trace_initcall_start_cb(&initcall_calltime, fn);
1220 static inline void do_trace_initcall_finish(initcall_t fn, int ret)
1222 if (!initcall_debug)
1224 trace_initcall_finish_cb(&initcall_calltime, fn, ret);
1226 #endif /* !TRACEPOINTS_ENABLED */
1228 int __init_or_module do_one_initcall(initcall_t fn)
1230 int count = preempt_count();
1234 if (initcall_blacklisted(fn))
1237 do_trace_initcall_start(fn);
1239 do_trace_initcall_finish(fn, ret);
1243 if (preempt_count() != count) {
1244 sprintf(msgbuf, "preemption imbalance ");
1245 preempt_count_set(count);
1247 if (irqs_disabled()) {
1248 strlcat(msgbuf, "disabled interrupts ", sizeof(msgbuf));
1251 WARN(msgbuf[0], "initcall %pS returned with %s\n", fn, msgbuf);
1253 add_latent_entropy();
1258 static initcall_entry_t *initcall_levels[] __initdata = {
1270 /* Keep these in sync with initcalls in include/linux/init.h */
1271 static const char *initcall_level_names[] __initdata = {
1282 static int __init ignore_unknown_bootoption(char *param, char *val,
1283 const char *unused, void *arg)
1288 static void __init do_initcall_level(int level, char *command_line)
1290 initcall_entry_t *fn;
1292 parse_args(initcall_level_names[level],
1293 command_line, __start___param,
1294 __stop___param - __start___param,
1296 NULL, ignore_unknown_bootoption);
1298 trace_initcall_level(initcall_level_names[level]);
1299 for (fn = initcall_levels[level]; fn < initcall_levels[level+1]; fn++)
1300 do_one_initcall(initcall_from_entry(fn));
1303 static void __init do_initcalls(void)
1306 size_t len = saved_command_line_len + 1;
1309 command_line = kzalloc(len, GFP_KERNEL);
1311 panic("%s: Failed to allocate %zu bytes\n", __func__, len);
1313 for (level = 0; level < ARRAY_SIZE(initcall_levels) - 1; level++) {
1314 /* Parser modifies command_line, restore it each time */
1315 strcpy(command_line, saved_command_line);
1316 do_initcall_level(level, command_line);
1319 kfree(command_line);
1323 * Ok, the machine is now initialized. None of the devices
1324 * have been touched yet, but the CPU subsystem is up and
1325 * running, and memory and process management works.
1327 * Now we can finally start doing some real work..
1329 static void __init do_basic_setup(void)
1338 static void __init do_pre_smp_initcalls(void)
1340 initcall_entry_t *fn;
1342 trace_initcall_level("early");
1343 for (fn = __initcall_start; fn < __initcall0_start; fn++)
1344 do_one_initcall(initcall_from_entry(fn));
1347 static int run_init_process(const char *init_filename)
1349 const char *const *p;
1351 argv_init[0] = init_filename;
1352 pr_info("Run %s as init process\n", init_filename);
1353 pr_debug(" with arguments:\n");
1354 for (p = argv_init; *p; p++)
1355 pr_debug(" %s\n", *p);
1356 pr_debug(" with environment:\n");
1357 for (p = envp_init; *p; p++)
1358 pr_debug(" %s\n", *p);
1359 return kernel_execve(init_filename, argv_init, envp_init);
1362 static int try_to_run_init_process(const char *init_filename)
1366 ret = run_init_process(init_filename);
1368 if (ret && ret != -ENOENT) {
1369 pr_err("Starting init: %s exists but couldn't execute it (error %d)\n",
1370 init_filename, ret);
1376 static noinline void __init kernel_init_freeable(void);
1378 #if defined(CONFIG_STRICT_KERNEL_RWX) || defined(CONFIG_STRICT_MODULE_RWX)
1379 bool rodata_enabled __ro_after_init = true;
1381 #ifndef arch_parse_debug_rodata
1382 static inline bool arch_parse_debug_rodata(char *str) { return false; }
1385 static int __init set_debug_rodata(char *str)
1387 if (arch_parse_debug_rodata(str))
1390 if (str && !strcmp(str, "on"))
1391 rodata_enabled = true;
1392 else if (str && !strcmp(str, "off"))
1393 rodata_enabled = false;
1395 pr_warn("Invalid option string for rodata: '%s'\n", str);
1398 early_param("rodata", set_debug_rodata);
1401 static void mark_readonly(void)
1403 if (IS_ENABLED(CONFIG_STRICT_KERNEL_RWX) && rodata_enabled) {
1405 * load_module() results in W+X mappings, which are cleaned
1406 * up with init_free_wq. Let's make sure that queued work is
1407 * flushed so that we don't hit false positives looking for
1408 * insecure pages which are W+X.
1410 flush_module_init_free_work();
1414 } else if (IS_ENABLED(CONFIG_STRICT_KERNEL_RWX)) {
1415 pr_info("Kernel memory protection disabled.\n");
1416 } else if (IS_ENABLED(CONFIG_ARCH_HAS_STRICT_KERNEL_RWX)) {
1417 pr_warn("Kernel memory protection not selected by kernel config.\n");
1419 pr_warn("This architecture does not have kernel memory protection.\n");
1423 void __weak free_initmem(void)
1425 free_initmem_default(POISON_FREE_INITMEM);
1428 static int __ref kernel_init(void *unused)
1433 * Wait until kthreadd is all set-up.
1435 wait_for_completion(&kthreadd_done);
1437 kernel_init_freeable();
1438 /* need to finish all async __init code before freeing the memory */
1439 async_synchronize_full();
1441 system_state = SYSTEM_FREEING_INITMEM;
1442 kprobe_free_init_mem();
1443 ftrace_free_init_mem();
1444 kgdb_free_init_mem();
1450 * Kernel mappings are now finalized - update the userspace page-table
1455 system_state = SYSTEM_RUNNING;
1456 numa_default_policy();
1458 rcu_end_inkernel_boot();
1462 if (ramdisk_execute_command) {
1463 ret = run_init_process(ramdisk_execute_command);
1466 pr_err("Failed to execute %s (error %d)\n",
1467 ramdisk_execute_command, ret);
1471 * We try each of these until one succeeds.
1473 * The Bourne shell can be used instead of init if we are
1474 * trying to recover a really broken machine.
1476 if (execute_command) {
1477 ret = run_init_process(execute_command);
1480 panic("Requested init %s failed (error %d).",
1481 execute_command, ret);
1484 if (CONFIG_DEFAULT_INIT[0] != '\0') {
1485 ret = run_init_process(CONFIG_DEFAULT_INIT);
1487 pr_err("Default init %s failed (error %d)\n",
1488 CONFIG_DEFAULT_INIT, ret);
1493 if (!try_to_run_init_process("/sbin/init") ||
1494 !try_to_run_init_process("/etc/init") ||
1495 !try_to_run_init_process("/bin/init") ||
1496 !try_to_run_init_process("/bin/sh"))
1499 panic("No working init found. Try passing init= option to kernel. "
1500 "See Linux Documentation/admin-guide/init.rst for guidance.");
1503 /* Open /dev/console, for stdin/stdout/stderr, this should never fail */
1504 void __init console_on_rootfs(void)
1506 struct file *file = filp_open("/dev/console", O_RDWR, 0);
1509 pr_err("Warning: unable to open an initial console.\n");
1518 static noinline void __init kernel_init_freeable(void)
1520 /* Now the scheduler is fully set up and can do blocking allocations */
1521 gfp_allowed_mask = __GFP_BITS_MASK;
1524 * init can allocate pages on any node
1526 set_mems_allowed(node_states[N_MEMORY]);
1528 cad_pid = get_pid(task_pid(current));
1530 smp_prepare_cpus(setup_max_cpus);
1534 init_mm_internals();
1536 rcu_init_tasks_generic();
1537 do_pre_smp_initcalls();
1538 lockup_detector_init();
1543 workqueue_init_topology();
1546 page_alloc_init_late();
1550 kunit_run_all_tests();
1552 wait_for_initramfs();
1553 console_on_rootfs();
1556 * check if there is an early userspace init. If yes, let it do all
1559 if (init_eaccess(ramdisk_execute_command) != 0) {
1560 ramdisk_execute_command = NULL;
1561 prepare_namespace();
1565 * Ok, we have completed the initial bootup, and
1566 * we're essentially up and running. Get rid of the
1567 * initmem segments and start the user-mode stuff..
1569 * rootfs is available now, try loading the public keys
1570 * and default modules
1573 integrity_load_keys();