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/vmalloc.h>
37 #include <linux/kernel_stat.h>
38 #include <linux/start_kernel.h>
39 #include <linux/security.h>
40 #include <linux/smp.h>
41 #include <linux/profile.h>
42 #include <linux/rcupdate.h>
43 #include <linux/moduleparam.h>
44 #include <linux/kallsyms.h>
45 #include <linux/writeback.h>
46 #include <linux/cpu.h>
47 #include <linux/cpuset.h>
48 #include <linux/cgroup.h>
49 #include <linux/efi.h>
50 #include <linux/tick.h>
51 #include <linux/sched/isolation.h>
52 #include <linux/interrupt.h>
53 #include <linux/taskstats_kern.h>
54 #include <linux/delayacct.h>
55 #include <linux/unistd.h>
56 #include <linux/utsname.h>
57 #include <linux/rmap.h>
58 #include <linux/mempolicy.h>
59 #include <linux/key.h>
60 #include <linux/buffer_head.h>
61 #include <linux/page_ext.h>
62 #include <linux/debug_locks.h>
63 #include <linux/debugobjects.h>
64 #include <linux/lockdep.h>
65 #include <linux/kmemleak.h>
66 #include <linux/padata.h>
67 #include <linux/pid_namespace.h>
68 #include <linux/device/driver.h>
69 #include <linux/kthread.h>
70 #include <linux/sched.h>
71 #include <linux/sched/init.h>
72 #include <linux/signal.h>
73 #include <linux/idr.h>
74 #include <linux/kgdb.h>
75 #include <linux/ftrace.h>
76 #include <linux/async.h>
77 #include <linux/sfi.h>
78 #include <linux/shmem_fs.h>
79 #include <linux/slab.h>
80 #include <linux/perf_event.h>
81 #include <linux/ptrace.h>
82 #include <linux/pti.h>
83 #include <linux/blkdev.h>
84 #include <linux/elevator.h>
85 #include <linux/sched/clock.h>
86 #include <linux/sched/task.h>
87 #include <linux/sched/task_stack.h>
88 #include <linux/context_tracking.h>
89 #include <linux/random.h>
90 #include <linux/list.h>
91 #include <linux/integrity.h>
92 #include <linux/proc_ns.h>
94 #include <linux/cache.h>
95 #include <linux/rodata_test.h>
96 #include <linux/jump_label.h>
97 #include <linux/mem_encrypt.h>
98 #include <linux/kcsan.h>
101 #include <asm/bugs.h>
102 #include <asm/setup.h>
103 #include <asm/sections.h>
104 #include <asm/cacheflush.h>
106 #define CREATE_TRACE_POINTS
107 #include <trace/events/initcall.h>
109 static int kernel_init(void *);
111 extern void init_IRQ(void);
112 extern void radix_tree_init(void);
115 * Debug helper: via this flag we know that we are in 'early bootup code'
116 * where only the boot processor is running with IRQ disabled. This means
117 * two things - IRQ must not be enabled before the flag is cleared and some
118 * operations which are not allowed with IRQ disabled are allowed while the
121 bool early_boot_irqs_disabled __read_mostly
;
123 enum system_states system_state __read_mostly
;
124 EXPORT_SYMBOL(system_state
);
127 * Boot command-line arguments
129 #define MAX_INIT_ARGS CONFIG_INIT_ENV_ARG_LIMIT
130 #define MAX_INIT_ENVS CONFIG_INIT_ENV_ARG_LIMIT
132 extern void time_init(void);
133 /* Default late time init is NULL. archs can override this later. */
134 void (*__initdata late_time_init
)(void);
136 /* Untouched command line saved by arch-specific code. */
137 char __initdata boot_command_line
[COMMAND_LINE_SIZE
];
138 /* Untouched saved command line (eg. for /proc) */
139 char *saved_command_line
;
140 /* Command line for parameter parsing */
141 static char *static_command_line
;
142 /* Untouched extra command line */
143 static char *extra_command_line
;
144 /* Extra init arguments */
145 static char *extra_init_args
;
147 #ifdef CONFIG_BOOT_CONFIG
148 /* Is bootconfig on command line? */
149 static bool bootconfig_found
;
150 static bool initargs_found
;
152 # define bootconfig_found false
153 # define initargs_found false
156 static char *execute_command
;
157 static char *ramdisk_execute_command
;
160 * Used to generate warnings if static_key manipulation functions are used
161 * before jump_label_init is called.
163 bool static_key_initialized __read_mostly
;
164 EXPORT_SYMBOL_GPL(static_key_initialized
);
167 * If set, this is an indication to the drivers that reset the underlying
168 * device before going ahead with the initialization otherwise driver might
169 * rely on the BIOS and skip the reset operation.
171 * This is useful if kernel is booting in an unreliable environment.
172 * For ex. kdump situation where previous kernel has crashed, BIOS has been
173 * skipped and devices will be in unknown state.
175 unsigned int reset_devices
;
176 EXPORT_SYMBOL(reset_devices
);
178 static int __init
set_reset_devices(char *str
)
184 __setup("reset_devices", set_reset_devices
);
186 static const char *argv_init
[MAX_INIT_ARGS
+2] = { "init", NULL
, };
187 const char *envp_init
[MAX_INIT_ENVS
+2] = { "HOME=/", "TERM=linux", NULL
, };
188 static const char *panic_later
, *panic_param
;
190 extern const struct obs_kernel_param __setup_start
[], __setup_end
[];
192 static bool __init
obsolete_checksetup(char *line
)
194 const struct obs_kernel_param
*p
;
195 bool had_early_param
= false;
199 int n
= strlen(p
->str
);
200 if (parameqn(line
, p
->str
, n
)) {
202 /* Already done in parse_early_param?
203 * (Needs exact match on param part).
204 * Keep iterating, as we can have early
205 * params and __setups of same names 8( */
206 if (line
[n
] == '\0' || line
[n
] == '=')
207 had_early_param
= true;
208 } else if (!p
->setup_func
) {
209 pr_warn("Parameter %s is obsolete, ignored\n",
212 } else if (p
->setup_func(line
+ n
))
216 } while (p
< __setup_end
);
218 return had_early_param
;
222 * This should be approx 2 Bo*oMips to start (note initial shift), and will
223 * still work even if initially too large, it will just take slightly longer
225 unsigned long loops_per_jiffy
= (1<<12);
226 EXPORT_SYMBOL(loops_per_jiffy
);
228 static int __init
debug_kernel(char *str
)
230 console_loglevel
= CONSOLE_LOGLEVEL_DEBUG
;
234 static int __init
quiet_kernel(char *str
)
236 console_loglevel
= CONSOLE_LOGLEVEL_QUIET
;
240 early_param("debug", debug_kernel
);
241 early_param("quiet", quiet_kernel
);
243 static int __init
loglevel(char *str
)
248 * Only update loglevel value when a correct setting was passed,
249 * to prevent blind crashes (when loglevel being set to 0) that
250 * are quite hard to debug
252 if (get_option(&str
, &newlevel
)) {
253 console_loglevel
= newlevel
;
260 early_param("loglevel", loglevel
);
262 #ifdef CONFIG_BLK_DEV_INITRD
263 static void * __init
get_boot_config_from_initrd(u32
*_size
, u32
*_csum
)
272 data
= (char *)initrd_end
- BOOTCONFIG_MAGIC_LEN
;
273 if (memcmp(data
, BOOTCONFIG_MAGIC
, BOOTCONFIG_MAGIC_LEN
))
276 hdr
= (u32
*)(data
- 8);
280 data
= ((void *)hdr
) - size
;
281 if ((unsigned long)data
< initrd_start
) {
282 pr_err("bootconfig size %d is greater than initrd size %ld\n",
283 size
, initrd_end
- initrd_start
);
287 /* Remove bootconfig from initramfs/initrd */
288 initrd_end
= (unsigned long)data
;
297 static void * __init
get_boot_config_from_initrd(u32
*_size
, u32
*_csum
)
303 #ifdef CONFIG_BOOT_CONFIG
305 char xbc_namebuf
[XBC_KEYLEN_MAX
] __initdata
;
307 #define rest(dst, end) ((end) > (dst) ? (end) - (dst) : 0)
309 static int __init
xbc_snprint_cmdline(char *buf
, size_t size
,
310 struct xbc_node
*root
)
312 struct xbc_node
*knode
, *vnode
;
313 char *end
= buf
+ size
;
317 xbc_node_for_each_key_value(root
, knode
, val
) {
318 ret
= xbc_node_compose_key_after(root
, knode
,
319 xbc_namebuf
, XBC_KEYLEN_MAX
);
323 vnode
= xbc_node_get_child(knode
);
325 ret
= snprintf(buf
, rest(buf
, end
), "%s ", xbc_namebuf
);
331 xbc_array_for_each_value(vnode
, val
) {
332 ret
= snprintf(buf
, rest(buf
, end
), "%s=\"%s\" ",
340 return buf
- (end
- size
);
344 /* Make an extra command line under given key word */
345 static char * __init
xbc_make_cmdline(const char *key
)
347 struct xbc_node
*root
;
351 root
= xbc_find_node(key
);
355 /* Count required buffer size */
356 len
= xbc_snprint_cmdline(NULL
, 0, root
);
360 new_cmdline
= memblock_alloc(len
+ 1, SMP_CACHE_BYTES
);
362 pr_err("Failed to allocate memory for extra kernel cmdline.\n");
366 ret
= xbc_snprint_cmdline(new_cmdline
, len
+ 1, root
);
367 if (ret
< 0 || ret
> len
) {
368 pr_err("Failed to print extra kernel cmdline.\n");
375 static u32
boot_config_checksum(unsigned char *p
, u32 size
)
385 static int __init
bootconfig_params(char *param
, char *val
,
386 const char *unused
, void *arg
)
388 if (strcmp(param
, "bootconfig") == 0) {
389 bootconfig_found
= true;
390 } else if (strcmp(param
, "--") == 0) {
391 initargs_found
= true;
396 static void __init
setup_boot_config(const char *cmdline
)
398 static char tmp_cmdline
[COMMAND_LINE_SIZE
] __initdata
;
405 /* Cut out the bootconfig data even if we have no bootconfig option */
406 data
= get_boot_config_from_initrd(&size
, &csum
);
408 strlcpy(tmp_cmdline
, boot_command_line
, COMMAND_LINE_SIZE
);
409 parse_args("bootconfig", tmp_cmdline
, NULL
, 0, 0, 0, NULL
,
412 if (!bootconfig_found
)
416 pr_err("'bootconfig' found on command line, but no bootconfig found\n");
420 if (size
>= XBC_DATA_MAX
) {
421 pr_err("bootconfig size %d greater than max size %d\n",
426 if (boot_config_checksum((unsigned char *)data
, size
) != csum
) {
427 pr_err("bootconfig checksum failed\n");
431 copy
= memblock_alloc(size
+ 1, SMP_CACHE_BYTES
);
433 pr_err("Failed to allocate memory for bootconfig\n");
437 memcpy(copy
, data
, size
);
440 ret
= xbc_init(copy
, &msg
, &pos
);
443 pr_err("Failed to init bootconfig: %s.\n", msg
);
445 pr_err("Failed to parse bootconfig: %s at %d.\n",
448 pr_info("Load bootconfig: %d bytes %d nodes\n", size
, ret
);
449 /* keys starting with "kernel." are passed via cmdline */
450 extra_command_line
= xbc_make_cmdline("kernel");
451 /* Also, "init." keys are init arguments */
452 extra_init_args
= xbc_make_cmdline("init");
459 static void __init
setup_boot_config(const char *cmdline
)
461 /* Remove bootconfig data from initrd */
462 get_boot_config_from_initrd(NULL
, NULL
);
465 static int __init
warn_bootconfig(char *str
)
467 pr_warn("WARNING: 'bootconfig' found on the kernel command line but CONFIG_BOOTCONFIG is not set.\n");
470 early_param("bootconfig", warn_bootconfig
);
474 /* Change NUL term back to "=", to make "param" the whole string. */
475 static void __init
repair_env_string(char *param
, char *val
)
478 /* param=val or param="val"? */
479 if (val
== param
+strlen(param
)+1)
481 else if (val
== param
+strlen(param
)+2) {
483 memmove(val
-1, val
, strlen(val
)+1);
489 /* Anything after -- gets handed straight to init. */
490 static int __init
set_init_arg(char *param
, char *val
,
491 const char *unused
, void *arg
)
498 repair_env_string(param
, val
);
500 for (i
= 0; argv_init
[i
]; i
++) {
501 if (i
== MAX_INIT_ARGS
) {
502 panic_later
= "init";
507 argv_init
[i
] = param
;
512 * Unknown boot options get handed to init, unless they look like
513 * unused parameters (modprobe will find them in /proc/cmdline).
515 static int __init
unknown_bootoption(char *param
, char *val
,
516 const char *unused
, void *arg
)
518 size_t len
= strlen(param
);
520 repair_env_string(param
, val
);
522 /* Handle obsolete-style parameters */
523 if (obsolete_checksetup(param
))
526 /* Unused module parameter. */
527 if (strnchr(param
, len
, '.'))
534 /* Environment option */
536 for (i
= 0; envp_init
[i
]; i
++) {
537 if (i
== MAX_INIT_ENVS
) {
541 if (!strncmp(param
, envp_init
[i
], len
+1))
544 envp_init
[i
] = param
;
546 /* Command line option */
548 for (i
= 0; argv_init
[i
]; i
++) {
549 if (i
== MAX_INIT_ARGS
) {
550 panic_later
= "init";
554 argv_init
[i
] = param
;
559 static int __init
init_setup(char *str
)
563 execute_command
= str
;
565 * In case LILO is going to boot us with default command line,
566 * it prepends "auto" before the whole cmdline which makes
567 * the shell think it should execute a script with such name.
568 * So we ignore all arguments entered _before_ init=... [MJ]
570 for (i
= 1; i
< MAX_INIT_ARGS
; i
++)
574 __setup("init=", init_setup
);
576 static int __init
rdinit_setup(char *str
)
580 ramdisk_execute_command
= str
;
581 /* See "auto" comment in init_setup */
582 for (i
= 1; i
< MAX_INIT_ARGS
; i
++)
586 __setup("rdinit=", rdinit_setup
);
589 static const unsigned int setup_max_cpus
= NR_CPUS
;
590 static inline void setup_nr_cpu_ids(void) { }
591 static inline void smp_prepare_cpus(unsigned int maxcpus
) { }
595 * We need to store the untouched command line for future reference.
596 * We also need to store the touched command line since the parameter
597 * parsing is performed in place, and we should allow a component to
598 * store reference of name/value for future reference.
600 static void __init
setup_command_line(char *command_line
)
602 size_t len
, xlen
= 0, ilen
= 0;
604 if (extra_command_line
)
605 xlen
= strlen(extra_command_line
);
607 ilen
= strlen(extra_init_args
) + 4; /* for " -- " */
609 len
= xlen
+ strlen(boot_command_line
) + 1;
611 saved_command_line
= memblock_alloc(len
+ ilen
, SMP_CACHE_BYTES
);
612 if (!saved_command_line
)
613 panic("%s: Failed to allocate %zu bytes\n", __func__
, len
+ ilen
);
615 static_command_line
= memblock_alloc(len
, SMP_CACHE_BYTES
);
616 if (!static_command_line
)
617 panic("%s: Failed to allocate %zu bytes\n", __func__
, len
);
621 * We have to put extra_command_line before boot command
622 * lines because there could be dashes (separator of init
623 * command line) in the command lines.
625 strcpy(saved_command_line
, extra_command_line
);
626 strcpy(static_command_line
, extra_command_line
);
628 strcpy(saved_command_line
+ xlen
, boot_command_line
);
629 strcpy(static_command_line
+ xlen
, command_line
);
633 * Append supplemental init boot args to saved_command_line
634 * so that user can check what command line options passed
637 len
= strlen(saved_command_line
);
638 if (initargs_found
) {
639 saved_command_line
[len
++] = ' ';
641 strcpy(saved_command_line
+ len
, " -- ");
645 strcpy(saved_command_line
+ len
, extra_init_args
);
650 * We need to finalize in a non-__init function or else race conditions
651 * between the root thread and the init thread may cause start_kernel to
652 * be reaped by free_initmem before the root thread has proceeded to
655 * gcc-3.4 accidentally inlines this function, so use noinline.
658 static __initdata
DECLARE_COMPLETION(kthreadd_done
);
660 noinline
void __ref
rest_init(void)
662 struct task_struct
*tsk
;
665 rcu_scheduler_starting();
667 * We need to spawn init first so that it obtains pid 1, however
668 * the init task will end up wanting to create kthreads, which, if
669 * we schedule it before we create kthreadd, will OOPS.
671 pid
= kernel_thread(kernel_init
, NULL
, CLONE_FS
);
673 * Pin init on the boot CPU. Task migration is not properly working
674 * until sched_init_smp() has been run. It will set the allowed
675 * CPUs for init to the non isolated CPUs.
678 tsk
= find_task_by_pid_ns(pid
, &init_pid_ns
);
679 set_cpus_allowed_ptr(tsk
, cpumask_of(smp_processor_id()));
682 numa_default_policy();
683 pid
= kernel_thread(kthreadd
, NULL
, CLONE_FS
| CLONE_FILES
);
685 kthreadd_task
= find_task_by_pid_ns(pid
, &init_pid_ns
);
689 * Enable might_sleep() and smp_processor_id() checks.
690 * They cannot be enabled earlier because with CONFIG_PREEMPTION=y
691 * kernel_thread() would trigger might_sleep() splats. With
692 * CONFIG_PREEMPT_VOLUNTARY=y the init task might have scheduled
693 * already, but it's stuck on the kthreadd_done completion.
695 system_state
= SYSTEM_SCHEDULING
;
697 complete(&kthreadd_done
);
700 * The boot idle thread must execute schedule()
701 * at least once to get things moving:
703 schedule_preempt_disabled();
704 /* Call into cpu_idle with preempt disabled */
705 cpu_startup_entry(CPUHP_ONLINE
);
708 /* Check for early params. */
709 static int __init
do_early_param(char *param
, char *val
,
710 const char *unused
, void *arg
)
712 const struct obs_kernel_param
*p
;
714 for (p
= __setup_start
; p
< __setup_end
; p
++) {
715 if ((p
->early
&& parameq(param
, p
->str
)) ||
716 (strcmp(param
, "console") == 0 &&
717 strcmp(p
->str
, "earlycon") == 0)
719 if (p
->setup_func(val
) != 0)
720 pr_warn("Malformed early option '%s'\n", param
);
723 /* We accept everything at this stage. */
727 void __init
parse_early_options(char *cmdline
)
729 parse_args("early options", cmdline
, NULL
, 0, 0, 0, NULL
,
733 /* Arch code calls this early on, or if not, just before other parsing. */
734 void __init
parse_early_param(void)
736 static int done __initdata
;
737 static char tmp_cmdline
[COMMAND_LINE_SIZE
] __initdata
;
742 /* All fall through to do_early_param. */
743 strlcpy(tmp_cmdline
, boot_command_line
, COMMAND_LINE_SIZE
);
744 parse_early_options(tmp_cmdline
);
748 void __init __weak
arch_post_acpi_subsys_init(void) { }
750 void __init __weak
smp_setup_processor_id(void)
754 # if THREAD_SIZE >= PAGE_SIZE
755 void __init __weak
thread_stack_cache_init(void)
760 void __init __weak
mem_encrypt_init(void) { }
762 void __init __weak
poking_init(void) { }
764 void __init __weak
pgtable_cache_init(void) { }
767 core_param(initcall_debug
, initcall_debug
, bool, 0644);
769 #ifdef TRACEPOINTS_ENABLED
770 static void __init
initcall_debug_enable(void);
772 static inline void initcall_debug_enable(void)
777 /* Report memory auto-initialization states for this boot. */
778 static void __init
report_meminit(void)
782 if (IS_ENABLED(CONFIG_INIT_STACK_ALL
))
784 else if (IS_ENABLED(CONFIG_GCC_PLUGIN_STRUCTLEAK_BYREF_ALL
))
786 else if (IS_ENABLED(CONFIG_GCC_PLUGIN_STRUCTLEAK_BYREF
))
788 else if (IS_ENABLED(CONFIG_GCC_PLUGIN_STRUCTLEAK_USER
))
793 pr_info("mem auto-init: stack:%s, heap alloc:%s, heap free:%s\n",
794 stack
, want_init_on_alloc(GFP_KERNEL
) ? "on" : "off",
795 want_init_on_free() ? "on" : "off");
796 if (want_init_on_free())
797 pr_info("mem auto-init: clearing system memory may take some time...\n");
801 * Set up kernel memory allocators
803 static void __init
mm_init(void)
806 * page_ext requires contiguous pages,
807 * bigger than MAX_ORDER unless SPARSEMEM.
809 page_ext_init_flatmem();
810 init_debug_pagealloc();
816 debug_objects_mem_init();
819 /* Should be run before the first non-init thread is created */
821 /* Should be run after espfix64 is set up. */
825 void __init __weak
arch_call_rest_init(void)
830 asmlinkage __visible
void __init
start_kernel(void)
835 set_task_stack_end_magic(&init_task
);
836 smp_setup_processor_id();
837 debug_objects_early_init();
842 early_boot_irqs_disabled
= true;
845 * Interrupts are still disabled. Do necessary setups, then
850 pr_notice("%s", linux_banner
);
851 early_security_init();
852 setup_arch(&command_line
);
853 setup_boot_config(command_line
);
854 setup_command_line(command_line
);
856 setup_per_cpu_areas();
857 smp_prepare_boot_cpu(); /* arch-specific boot-cpu hooks */
858 boot_cpu_hotplug_init();
860 build_all_zonelists(NULL
);
863 pr_notice("Kernel command line: %s\n", saved_command_line
);
864 /* parameters may set static keys */
867 after_dashes
= parse_args("Booting kernel",
868 static_command_line
, __start___param
,
869 __stop___param
- __start___param
,
870 -1, -1, NULL
, &unknown_bootoption
);
871 if (!IS_ERR_OR_NULL(after_dashes
))
872 parse_args("Setting init args", after_dashes
, NULL
, 0, -1, -1,
875 parse_args("Setting extra init args", extra_init_args
,
876 NULL
, 0, -1, -1, NULL
, set_init_arg
);
879 * These use large bootmem allocations and must precede
883 vfs_caches_init_early();
890 /* trace_printk can be enabled here */
894 * Set up the scheduler prior starting any interrupts (such as the
895 * timer interrupt). Full topology setup happens at smp_init()
896 * time - but meanwhile we still have a functioning scheduler.
900 * Disable preemption - early bootup scheduling is extremely
901 * fragile until we cpu_idle() for the first time.
904 if (WARN(!irqs_disabled(),
905 "Interrupts were enabled *very* early, fixing it\n"))
910 * Set up housekeeping before setting up workqueues to allow the unbound
911 * workqueue to take non-housekeeping into account.
916 * Allow workqueue creation and work item queueing/cancelling
917 * early. Work item execution depends on kthreads and starts after
920 workqueue_init_early();
924 /* Trace events are available after this */
928 initcall_debug_enable();
930 context_tracking_init();
931 /* init some links before init_ISA_irqs() */
942 * For best initial stack canary entropy, prepare it after:
943 * - setup_arch() for any UEFI RNG entropy and boot cmdline access
944 * - timekeeping_init() for ktime entropy used in rand_initialize()
945 * - rand_initialize() to get any arch-specific entropy like RDRAND
946 * - add_latent_entropy() to get any latent entropy
947 * - adding command line entropy
950 add_latent_entropy();
951 add_device_randomness(command_line
, strlen(command_line
));
952 boot_init_stack_canary();
957 call_function_init();
958 WARN(!irqs_disabled(), "Interrupts were enabled early\n");
960 early_boot_irqs_disabled
= false;
963 kmem_cache_init_late();
966 * HACK ALERT! This is early. We're enabling the console before
967 * we've done PCI setups etc, and console_init() must be aware of
968 * this. But we do want output early, in case something goes wrong.
972 panic("Too many boot %s vars at `%s'", panic_later
,
978 * Need to run this when irqs are enabled, because it wants
979 * to self-test [hard/soft]-irqs on/off lock inversion bugs
985 * This needs to be called before any devices perform DMA
986 * operations that might use the SWIOTLB bounce buffers. It will
987 * mark the bounce buffers as decrypted so that their usage will
988 * not cause "plain-text" data to be decrypted when accessed.
992 #ifdef CONFIG_BLK_DEV_INITRD
993 if (initrd_start
&& !initrd_below_start_ok
&&
994 page_to_pfn(virt_to_page((void *)initrd_start
)) < min_low_pfn
) {
995 pr_crit("initrd overwritten (0x%08lx < 0x%08lx) - disabling it.\n",
996 page_to_pfn(virt_to_page((void *)initrd_start
)),
1001 setup_per_cpu_pageset();
1011 if (efi_enabled(EFI_RUNTIME_SERVICES
))
1012 efi_enter_virtual_mode();
1014 thread_stack_cache_init();
1031 taskstats_init_early();
1037 acpi_subsystem_init();
1038 arch_post_acpi_subsys_init();
1042 /* Do the rest non-__init'ed, we're now alive */
1043 arch_call_rest_init();
1045 prevent_tail_call_optimization();
1048 /* Call all constructor functions linked into the kernel. */
1049 static void __init
do_ctors(void)
1051 #ifdef CONFIG_CONSTRUCTORS
1052 ctor_fn_t
*fn
= (ctor_fn_t
*) __ctors_start
;
1054 for (; fn
< (ctor_fn_t
*) __ctors_end
; fn
++)
1059 #ifdef CONFIG_KALLSYMS
1060 struct blacklist_entry
{
1061 struct list_head next
;
1065 static __initdata_or_module
LIST_HEAD(blacklisted_initcalls
);
1067 static int __init
initcall_blacklist(char *str
)
1070 struct blacklist_entry
*entry
;
1072 /* str argument is a comma-separated list of functions */
1074 str_entry
= strsep(&str
, ",");
1076 pr_debug("blacklisting initcall %s\n", str_entry
);
1077 entry
= memblock_alloc(sizeof(*entry
),
1080 panic("%s: Failed to allocate %zu bytes\n",
1081 __func__
, sizeof(*entry
));
1082 entry
->buf
= memblock_alloc(strlen(str_entry
) + 1,
1085 panic("%s: Failed to allocate %zu bytes\n",
1086 __func__
, strlen(str_entry
) + 1);
1087 strcpy(entry
->buf
, str_entry
);
1088 list_add(&entry
->next
, &blacklisted_initcalls
);
1090 } while (str_entry
);
1095 static bool __init_or_module
initcall_blacklisted(initcall_t fn
)
1097 struct blacklist_entry
*entry
;
1098 char fn_name
[KSYM_SYMBOL_LEN
];
1101 if (list_empty(&blacklisted_initcalls
))
1104 addr
= (unsigned long) dereference_function_descriptor(fn
);
1105 sprint_symbol_no_offset(fn_name
, addr
);
1108 * fn will be "function_name [module_name]" where [module_name] is not
1109 * displayed for built-in init functions. Strip off the [module_name].
1111 strreplace(fn_name
, ' ', '\0');
1113 list_for_each_entry(entry
, &blacklisted_initcalls
, next
) {
1114 if (!strcmp(fn_name
, entry
->buf
)) {
1115 pr_debug("initcall %s blacklisted\n", fn_name
);
1123 static int __init
initcall_blacklist(char *str
)
1125 pr_warn("initcall_blacklist requires CONFIG_KALLSYMS\n");
1129 static bool __init_or_module
initcall_blacklisted(initcall_t fn
)
1134 __setup("initcall_blacklist=", initcall_blacklist
);
1136 static __init_or_module
void
1137 trace_initcall_start_cb(void *data
, initcall_t fn
)
1139 ktime_t
*calltime
= (ktime_t
*)data
;
1141 printk(KERN_DEBUG
"calling %pS @ %i\n", fn
, task_pid_nr(current
));
1142 *calltime
= ktime_get();
1145 static __init_or_module
void
1146 trace_initcall_finish_cb(void *data
, initcall_t fn
, int ret
)
1148 ktime_t
*calltime
= (ktime_t
*)data
;
1149 ktime_t delta
, rettime
;
1150 unsigned long long duration
;
1152 rettime
= ktime_get();
1153 delta
= ktime_sub(rettime
, *calltime
);
1154 duration
= (unsigned long long) ktime_to_ns(delta
) >> 10;
1155 printk(KERN_DEBUG
"initcall %pS returned %d after %lld usecs\n",
1159 static ktime_t initcall_calltime
;
1161 #ifdef TRACEPOINTS_ENABLED
1162 static void __init
initcall_debug_enable(void)
1166 ret
= register_trace_initcall_start(trace_initcall_start_cb
,
1167 &initcall_calltime
);
1168 ret
|= register_trace_initcall_finish(trace_initcall_finish_cb
,
1169 &initcall_calltime
);
1170 WARN(ret
, "Failed to register initcall tracepoints\n");
1172 # define do_trace_initcall_start trace_initcall_start
1173 # define do_trace_initcall_finish trace_initcall_finish
1175 static inline void do_trace_initcall_start(initcall_t fn
)
1177 if (!initcall_debug
)
1179 trace_initcall_start_cb(&initcall_calltime
, fn
);
1181 static inline void do_trace_initcall_finish(initcall_t fn
, int ret
)
1183 if (!initcall_debug
)
1185 trace_initcall_finish_cb(&initcall_calltime
, fn
, ret
);
1187 #endif /* !TRACEPOINTS_ENABLED */
1189 int __init_or_module
do_one_initcall(initcall_t fn
)
1191 int count
= preempt_count();
1195 if (initcall_blacklisted(fn
))
1198 do_trace_initcall_start(fn
);
1200 do_trace_initcall_finish(fn
, ret
);
1204 if (preempt_count() != count
) {
1205 sprintf(msgbuf
, "preemption imbalance ");
1206 preempt_count_set(count
);
1208 if (irqs_disabled()) {
1209 strlcat(msgbuf
, "disabled interrupts ", sizeof(msgbuf
));
1212 WARN(msgbuf
[0], "initcall %pS returned with %s\n", fn
, msgbuf
);
1214 add_latent_entropy();
1219 extern initcall_entry_t __initcall_start
[];
1220 extern initcall_entry_t __initcall0_start
[];
1221 extern initcall_entry_t __initcall1_start
[];
1222 extern initcall_entry_t __initcall2_start
[];
1223 extern initcall_entry_t __initcall3_start
[];
1224 extern initcall_entry_t __initcall4_start
[];
1225 extern initcall_entry_t __initcall5_start
[];
1226 extern initcall_entry_t __initcall6_start
[];
1227 extern initcall_entry_t __initcall7_start
[];
1228 extern initcall_entry_t __initcall_end
[];
1230 static initcall_entry_t
*initcall_levels
[] __initdata
= {
1242 /* Keep these in sync with initcalls in include/linux/init.h */
1243 static const char *initcall_level_names
[] __initdata
= {
1254 static int __init
ignore_unknown_bootoption(char *param
, char *val
,
1255 const char *unused
, void *arg
)
1260 static void __init
do_initcall_level(int level
, char *command_line
)
1262 initcall_entry_t
*fn
;
1264 parse_args(initcall_level_names
[level
],
1265 command_line
, __start___param
,
1266 __stop___param
- __start___param
,
1268 NULL
, ignore_unknown_bootoption
);
1270 trace_initcall_level(initcall_level_names
[level
]);
1271 for (fn
= initcall_levels
[level
]; fn
< initcall_levels
[level
+1]; fn
++)
1272 do_one_initcall(initcall_from_entry(fn
));
1275 static void __init
do_initcalls(void)
1278 size_t len
= strlen(saved_command_line
) + 1;
1281 command_line
= kzalloc(len
, GFP_KERNEL
);
1283 panic("%s: Failed to allocate %zu bytes\n", __func__
, len
);
1285 for (level
= 0; level
< ARRAY_SIZE(initcall_levels
) - 1; level
++) {
1286 /* Parser modifies command_line, restore it each time */
1287 strcpy(command_line
, saved_command_line
);
1288 do_initcall_level(level
, command_line
);
1291 kfree(command_line
);
1295 * Ok, the machine is now initialized. None of the devices
1296 * have been touched yet, but the CPU subsystem is up and
1297 * running, and memory and process management works.
1299 * Now we can finally start doing some real work..
1301 static void __init
do_basic_setup(void)
1307 usermodehelper_enable();
1311 static void __init
do_pre_smp_initcalls(void)
1313 initcall_entry_t
*fn
;
1315 trace_initcall_level("early");
1316 for (fn
= __initcall_start
; fn
< __initcall0_start
; fn
++)
1317 do_one_initcall(initcall_from_entry(fn
));
1320 static int run_init_process(const char *init_filename
)
1322 const char *const *p
;
1324 argv_init
[0] = init_filename
;
1325 pr_info("Run %s as init process\n", init_filename
);
1326 pr_debug(" with arguments:\n");
1327 for (p
= argv_init
; *p
; p
++)
1328 pr_debug(" %s\n", *p
);
1329 pr_debug(" with environment:\n");
1330 for (p
= envp_init
; *p
; p
++)
1331 pr_debug(" %s\n", *p
);
1332 return do_execve(getname_kernel(init_filename
),
1333 (const char __user
*const __user
*)argv_init
,
1334 (const char __user
*const __user
*)envp_init
);
1337 static int try_to_run_init_process(const char *init_filename
)
1341 ret
= run_init_process(init_filename
);
1343 if (ret
&& ret
!= -ENOENT
) {
1344 pr_err("Starting init: %s exists but couldn't execute it (error %d)\n",
1345 init_filename
, ret
);
1351 static noinline
void __init
kernel_init_freeable(void);
1353 #if defined(CONFIG_STRICT_KERNEL_RWX) || defined(CONFIG_STRICT_MODULE_RWX)
1354 bool rodata_enabled __ro_after_init
= true;
1355 static int __init
set_debug_rodata(char *str
)
1357 return strtobool(str
, &rodata_enabled
);
1359 __setup("rodata=", set_debug_rodata
);
1362 #ifdef CONFIG_STRICT_KERNEL_RWX
1363 static void mark_readonly(void)
1365 if (rodata_enabled
) {
1367 * load_module() results in W+X mappings, which are cleaned
1368 * up with call_rcu(). Let's make sure that queued work is
1369 * flushed so that we don't hit false positives looking for
1370 * insecure pages which are W+X.
1376 pr_info("Kernel memory protection disabled.\n");
1378 #elif defined(CONFIG_ARCH_HAS_STRICT_KERNEL_RWX)
1379 static inline void mark_readonly(void)
1381 pr_warn("Kernel memory protection not selected by kernel config.\n");
1384 static inline void mark_readonly(void)
1386 pr_warn("This architecture does not have kernel memory protection.\n");
1390 void __weak
free_initmem(void)
1392 free_initmem_default(POISON_FREE_INITMEM
);
1395 static int __ref
kernel_init(void *unused
)
1399 kernel_init_freeable();
1400 /* need to finish all async __init code before freeing the memory */
1401 async_synchronize_full();
1402 ftrace_free_init_mem();
1407 * Kernel mappings are now finalized - update the userspace page-table
1412 system_state
= SYSTEM_RUNNING
;
1413 numa_default_policy();
1415 rcu_end_inkernel_boot();
1419 if (ramdisk_execute_command
) {
1420 ret
= run_init_process(ramdisk_execute_command
);
1423 pr_err("Failed to execute %s (error %d)\n",
1424 ramdisk_execute_command
, ret
);
1428 * We try each of these until one succeeds.
1430 * The Bourne shell can be used instead of init if we are
1431 * trying to recover a really broken machine.
1433 if (execute_command
) {
1434 ret
= run_init_process(execute_command
);
1437 panic("Requested init %s failed (error %d).",
1438 execute_command
, ret
);
1441 if (CONFIG_DEFAULT_INIT
[0] != '\0') {
1442 ret
= run_init_process(CONFIG_DEFAULT_INIT
);
1444 pr_err("Default init %s failed (error %d)\n",
1445 CONFIG_DEFAULT_INIT
, ret
);
1450 if (!try_to_run_init_process("/sbin/init") ||
1451 !try_to_run_init_process("/etc/init") ||
1452 !try_to_run_init_process("/bin/init") ||
1453 !try_to_run_init_process("/bin/sh"))
1456 panic("No working init found. Try passing init= option to kernel. "
1457 "See Linux Documentation/admin-guide/init.rst for guidance.");
1460 void console_on_rootfs(void)
1462 /* Open the /dev/console as stdin, this should never fail */
1463 if (ksys_open((const char __user
*) "/dev/console", O_RDWR
, 0) < 0)
1464 pr_err("Warning: unable to open an initial console.\n");
1466 /* create stdout/stderr */
1471 static noinline
void __init
kernel_init_freeable(void)
1474 * Wait until kthreadd is all set-up.
1476 wait_for_completion(&kthreadd_done
);
1478 /* Now the scheduler is fully set up and can do blocking allocations */
1479 gfp_allowed_mask
= __GFP_BITS_MASK
;
1482 * init can allocate pages on any node
1484 set_mems_allowed(node_states
[N_MEMORY
]);
1486 cad_pid
= task_pid(current
);
1488 smp_prepare_cpus(setup_max_cpus
);
1492 init_mm_internals();
1494 do_pre_smp_initcalls();
1495 lockup_detector_init();
1501 page_alloc_init_late();
1502 /* Initialize page ext after all struct pages are initialized. */
1507 console_on_rootfs();
1510 * check if there is an early userspace init. If yes, let it do all
1514 if (!ramdisk_execute_command
)
1515 ramdisk_execute_command
= "/init";
1517 if (ksys_access((const char __user
*)
1518 ramdisk_execute_command
, 0) != 0) {
1519 ramdisk_execute_command
= NULL
;
1520 prepare_namespace();
1524 * Ok, we have completed the initial bootup, and
1525 * we're essentially up and running. Get rid of the
1526 * initmem segments and start the user-mode stuff..
1528 * rootfs is available now, try loading the public keys
1529 * and default modules
1532 integrity_load_keys();