1 // SPDX-License-Identifier: GPL-2.0-only
5 * Copyright (C) 1991, 1992 Linus Torvalds
7 * GK 2/5/95 - Changed to support mounting root fs via NFS
8 * Added initrd & change_root: Werner Almesberger & Hans Lermen, Feb '96
9 * Moan early if gcc is old, avoiding bogus kernels - Paul Gortmaker, May '96
10 * Simplified starting of init: Michael A. Griffith <grif@acm.org>
13 #define DEBUG /* Enable initcall_debug */
15 #include <linux/types.h>
16 #include <linux/extable.h>
17 #include <linux/module.h>
18 #include <linux/proc_fs.h>
19 #include <linux/binfmts.h>
20 #include <linux/kernel.h>
21 #include <linux/syscalls.h>
22 #include <linux/stackprotector.h>
23 #include <linux/string.h>
24 #include <linux/ctype.h>
25 #include <linux/delay.h>
26 #include <linux/ioport.h>
27 #include <linux/init.h>
28 #include <linux/initrd.h>
29 #include <linux/memblock.h>
30 #include <linux/acpi.h>
31 #include <linux/bootconfig.h>
32 #include <linux/console.h>
33 #include <linux/nmi.h>
34 #include <linux/percpu.h>
35 #include <linux/kmod.h>
36 #include <linux/kprobes.h>
37 #include <linux/vmalloc.h>
38 #include <linux/kernel_stat.h>
39 #include <linux/start_kernel.h>
40 #include <linux/security.h>
41 #include <linux/smp.h>
42 #include <linux/profile.h>
43 #include <linux/kfence.h>
44 #include <linux/rcupdate.h>
45 #include <linux/srcu.h>
46 #include <linux/moduleparam.h>
47 #include <linux/kallsyms.h>
48 #include <linux/buildid.h>
49 #include <linux/writeback.h>
50 #include <linux/cpu.h>
51 #include <linux/cpuset.h>
52 #include <linux/cgroup.h>
53 #include <linux/efi.h>
54 #include <linux/tick.h>
55 #include <linux/sched/isolation.h>
56 #include <linux/interrupt.h>
57 #include <linux/taskstats_kern.h>
58 #include <linux/delayacct.h>
59 #include <linux/unistd.h>
60 #include <linux/utsname.h>
61 #include <linux/rmap.h>
62 #include <linux/mempolicy.h>
63 #include <linux/key.h>
64 #include <linux/page_ext.h>
65 #include <linux/debug_locks.h>
66 #include <linux/debugobjects.h>
67 #include <linux/lockdep.h>
68 #include <linux/kmemleak.h>
69 #include <linux/padata.h>
70 #include <linux/pid_namespace.h>
71 #include <linux/device/driver.h>
72 #include <linux/kthread.h>
73 #include <linux/sched.h>
74 #include <linux/sched/init.h>
75 #include <linux/signal.h>
76 #include <linux/idr.h>
77 #include <linux/kgdb.h>
78 #include <linux/ftrace.h>
79 #include <linux/async.h>
80 #include <linux/shmem_fs.h>
81 #include <linux/slab.h>
82 #include <linux/perf_event.h>
83 #include <linux/ptrace.h>
84 #include <linux/pti.h>
85 #include <linux/blkdev.h>
86 #include <linux/elevator.h>
87 #include <linux/sched/clock.h>
88 #include <linux/sched/task.h>
89 #include <linux/sched/task_stack.h>
90 #include <linux/context_tracking.h>
91 #include <linux/random.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/mem_encrypt.h>
100 #include <linux/kcsan.h>
101 #include <linux/init_syscalls.h>
102 #include <linux/stackdepot.h>
105 #include <asm/bugs.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 *);
117 extern void init_IRQ(void);
118 extern void radix_tree_init(void);
121 * Debug helper: via this flag we know that we are in 'early bootup code'
122 * where only the boot processor is running with IRQ disabled. This means
123 * two things - IRQ must not be enabled before the flag is cleared and some
124 * operations which are not allowed with IRQ disabled are allowed while the
127 bool early_boot_irqs_disabled __read_mostly
;
129 enum system_states system_state __read_mostly
;
130 EXPORT_SYMBOL(system_state
);
133 * Boot command-line arguments
135 #define MAX_INIT_ARGS CONFIG_INIT_ENV_ARG_LIMIT
136 #define MAX_INIT_ENVS CONFIG_INIT_ENV_ARG_LIMIT
138 extern void time_init(void);
139 /* Default late time init is NULL. archs can override this later. */
140 void (*__initdata late_time_init
)(void);
142 /* Untouched command line saved by arch-specific code. */
143 char __initdata boot_command_line
[COMMAND_LINE_SIZE
];
144 /* Untouched saved command line (eg. for /proc) */
145 char *saved_command_line
;
146 /* Command line for parameter parsing */
147 static char *static_command_line
;
148 /* Untouched extra command line */
149 static char *extra_command_line
;
150 /* Extra init arguments */
151 static char *extra_init_args
;
153 #ifdef CONFIG_BOOT_CONFIG
154 /* Is bootconfig on command line? */
155 static bool bootconfig_found
;
156 static bool initargs_found
;
158 # define bootconfig_found false
159 # define initargs_found false
162 static char *execute_command
;
163 static char *ramdisk_execute_command
= "/init";
166 * Used to generate warnings if static_key manipulation functions are used
167 * before jump_label_init is called.
169 bool static_key_initialized __read_mostly
;
170 EXPORT_SYMBOL_GPL(static_key_initialized
);
173 * If set, this is an indication to the drivers that reset the underlying
174 * device before going ahead with the initialization otherwise driver might
175 * rely on the BIOS and skip the reset operation.
177 * This is useful if kernel is booting in an unreliable environment.
178 * For ex. kdump situation where previous kernel has crashed, BIOS has been
179 * skipped and devices will be in unknown state.
181 unsigned int reset_devices
;
182 EXPORT_SYMBOL(reset_devices
);
184 static int __init
set_reset_devices(char *str
)
190 __setup("reset_devices", set_reset_devices
);
192 static const char *argv_init
[MAX_INIT_ARGS
+2] = { "init", NULL
, };
193 const char *envp_init
[MAX_INIT_ENVS
+2] = { "HOME=/", "TERM=linux", NULL
, };
194 static const char *panic_later
, *panic_param
;
196 extern const struct obs_kernel_param __setup_start
[], __setup_end
[];
198 static bool __init
obsolete_checksetup(char *line
)
200 const struct obs_kernel_param
*p
;
201 bool had_early_param
= false;
205 int n
= strlen(p
->str
);
206 if (parameqn(line
, p
->str
, n
)) {
208 /* Already done in parse_early_param?
209 * (Needs exact match on param part).
210 * Keep iterating, as we can have early
211 * params and __setups of same names 8( */
212 if (line
[n
] == '\0' || line
[n
] == '=')
213 had_early_param
= true;
214 } else if (!p
->setup_func
) {
215 pr_warn("Parameter %s is obsolete, ignored\n",
218 } else if (p
->setup_func(line
+ n
))
222 } while (p
< __setup_end
);
224 return had_early_param
;
228 * This should be approx 2 Bo*oMips to start (note initial shift), and will
229 * still work even if initially too large, it will just take slightly longer
231 unsigned long loops_per_jiffy
= (1<<12);
232 EXPORT_SYMBOL(loops_per_jiffy
);
234 static int __init
debug_kernel(char *str
)
236 console_loglevel
= CONSOLE_LOGLEVEL_DEBUG
;
240 static int __init
quiet_kernel(char *str
)
242 console_loglevel
= CONSOLE_LOGLEVEL_QUIET
;
246 early_param("debug", debug_kernel
);
247 early_param("quiet", quiet_kernel
);
249 static int __init
loglevel(char *str
)
254 * Only update loglevel value when a correct setting was passed,
255 * to prevent blind crashes (when loglevel being set to 0) that
256 * are quite hard to debug
258 if (get_option(&str
, &newlevel
)) {
259 console_loglevel
= newlevel
;
266 early_param("loglevel", loglevel
);
268 #ifdef CONFIG_BLK_DEV_INITRD
269 static void * __init
get_boot_config_from_initrd(u32
*_size
, u32
*_csum
)
279 data
= (char *)initrd_end
- BOOTCONFIG_MAGIC_LEN
;
281 * Since Grub may align the size of initrd to 4, we must
282 * check the preceding 3 bytes as well.
284 for (i
= 0; i
< 4; i
++) {
285 if (!memcmp(data
, BOOTCONFIG_MAGIC
, BOOTCONFIG_MAGIC_LEN
))
292 hdr
= (u32
*)(data
- 8);
293 size
= le32_to_cpu(hdr
[0]);
294 csum
= le32_to_cpu(hdr
[1]);
296 data
= ((void *)hdr
) - size
;
297 if ((unsigned long)data
< initrd_start
) {
298 pr_err("bootconfig size %d is greater than initrd size %ld\n",
299 size
, initrd_end
- initrd_start
);
303 /* Remove bootconfig from initramfs/initrd */
304 initrd_end
= (unsigned long)data
;
313 static void * __init
get_boot_config_from_initrd(u32
*_size
, u32
*_csum
)
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");
391 static int __init
bootconfig_params(char *param
, char *val
,
392 const char *unused
, void *arg
)
394 if (strcmp(param
, "bootconfig") == 0) {
395 bootconfig_found
= true;
400 static void __init
setup_boot_config(void)
402 static char tmp_cmdline
[COMMAND_LINE_SIZE
] __initdata
;
406 char *data
, *copy
, *err
;
409 /* Cut out the bootconfig data even if we have no bootconfig option */
410 data
= get_boot_config_from_initrd(&size
, &csum
);
412 strlcpy(tmp_cmdline
, boot_command_line
, COMMAND_LINE_SIZE
);
413 err
= parse_args("bootconfig", tmp_cmdline
, NULL
, 0, 0, 0, NULL
,
416 if (IS_ERR(err
) || !bootconfig_found
)
419 /* parse_args() stops at '--' and returns an address */
421 initargs_found
= true;
424 pr_err("'bootconfig' found on command line, but no bootconfig found\n");
428 if (size
>= XBC_DATA_MAX
) {
429 pr_err("bootconfig size %d greater than max size %d\n",
434 if (xbc_calc_checksum(data
, size
) != csum
) {
435 pr_err("bootconfig checksum failed\n");
439 copy
= memblock_alloc(size
+ 1, SMP_CACHE_BYTES
);
441 pr_err("Failed to allocate memory for bootconfig\n");
445 memcpy(copy
, data
, size
);
448 ret
= xbc_init(copy
, &msg
, &pos
);
451 pr_err("Failed to init bootconfig: %s.\n", msg
);
453 pr_err("Failed to parse bootconfig: %s at %d.\n",
456 pr_info("Load bootconfig: %d bytes %d nodes\n", 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");
467 static void __init
setup_boot_config(void)
469 /* Remove bootconfig data from initrd */
470 get_boot_config_from_initrd(NULL
, NULL
);
473 static int __init
warn_bootconfig(char *str
)
475 pr_warn("WARNING: 'bootconfig' found on the kernel command line but CONFIG_BOOT_CONFIG is not set.\n");
478 early_param("bootconfig", warn_bootconfig
);
482 /* Change NUL term back to "=", to make "param" the whole string. */
483 static void __init
repair_env_string(char *param
, char *val
)
486 /* param=val or param="val"? */
487 if (val
== param
+strlen(param
)+1)
489 else if (val
== param
+strlen(param
)+2) {
491 memmove(val
-1, val
, strlen(val
)+1);
497 /* Anything after -- gets handed straight to init. */
498 static int __init
set_init_arg(char *param
, char *val
,
499 const char *unused
, void *arg
)
506 repair_env_string(param
, val
);
508 for (i
= 0; argv_init
[i
]; i
++) {
509 if (i
== MAX_INIT_ARGS
) {
510 panic_later
= "init";
515 argv_init
[i
] = param
;
520 * Unknown boot options get handed to init, unless they look like
521 * unused parameters (modprobe will find them in /proc/cmdline).
523 static int __init
unknown_bootoption(char *param
, char *val
,
524 const char *unused
, void *arg
)
526 size_t len
= strlen(param
);
528 repair_env_string(param
, val
);
530 /* Handle obsolete-style parameters */
531 if (obsolete_checksetup(param
))
534 /* Unused module parameter. */
535 if (strnchr(param
, len
, '.'))
542 /* Environment option */
544 for (i
= 0; envp_init
[i
]; i
++) {
545 if (i
== MAX_INIT_ENVS
) {
549 if (!strncmp(param
, envp_init
[i
], len
+1))
552 envp_init
[i
] = param
;
554 /* Command line option */
556 for (i
= 0; argv_init
[i
]; i
++) {
557 if (i
== MAX_INIT_ARGS
) {
558 panic_later
= "init";
562 argv_init
[i
] = param
;
567 static int __init
init_setup(char *str
)
571 execute_command
= str
;
573 * In case LILO is going to boot us with default command line,
574 * it prepends "auto" before the whole cmdline which makes
575 * the shell think it should execute a script with such name.
576 * So we ignore all arguments entered _before_ init=... [MJ]
578 for (i
= 1; i
< MAX_INIT_ARGS
; i
++)
582 __setup("init=", init_setup
);
584 static int __init
rdinit_setup(char *str
)
588 ramdisk_execute_command
= str
;
589 /* See "auto" comment in init_setup */
590 for (i
= 1; i
< MAX_INIT_ARGS
; i
++)
594 __setup("rdinit=", rdinit_setup
);
597 static const unsigned int setup_max_cpus
= NR_CPUS
;
598 static inline void setup_nr_cpu_ids(void) { }
599 static inline void smp_prepare_cpus(unsigned int maxcpus
) { }
603 * We need to store the untouched command line for future reference.
604 * We also need to store the touched command line since the parameter
605 * parsing is performed in place, and we should allow a component to
606 * store reference of name/value for future reference.
608 static void __init
setup_command_line(char *command_line
)
610 size_t len
, xlen
= 0, ilen
= 0;
612 if (extra_command_line
)
613 xlen
= strlen(extra_command_line
);
615 ilen
= strlen(extra_init_args
) + 4; /* for " -- " */
617 len
= xlen
+ strlen(boot_command_line
) + 1;
619 saved_command_line
= memblock_alloc(len
+ ilen
, SMP_CACHE_BYTES
);
620 if (!saved_command_line
)
621 panic("%s: Failed to allocate %zu bytes\n", __func__
, len
+ ilen
);
623 static_command_line
= memblock_alloc(len
, SMP_CACHE_BYTES
);
624 if (!static_command_line
)
625 panic("%s: Failed to allocate %zu bytes\n", __func__
, len
);
629 * We have to put extra_command_line before boot command
630 * lines because there could be dashes (separator of init
631 * command line) in the command lines.
633 strcpy(saved_command_line
, extra_command_line
);
634 strcpy(static_command_line
, extra_command_line
);
636 strcpy(saved_command_line
+ xlen
, boot_command_line
);
637 strcpy(static_command_line
+ xlen
, command_line
);
641 * Append supplemental init boot args to saved_command_line
642 * so that user can check what command line options passed
645 len
= strlen(saved_command_line
);
646 if (initargs_found
) {
647 saved_command_line
[len
++] = ' ';
649 strcpy(saved_command_line
+ len
, " -- ");
653 strcpy(saved_command_line
+ len
, extra_init_args
);
658 * We need to finalize in a non-__init function or else race conditions
659 * between the root thread and the init thread may cause start_kernel to
660 * be reaped by free_initmem before the root thread has proceeded to
663 * gcc-3.4 accidentally inlines this function, so use noinline.
666 static __initdata
DECLARE_COMPLETION(kthreadd_done
);
668 noinline
void __ref
rest_init(void)
670 struct task_struct
*tsk
;
673 rcu_scheduler_starting();
675 * We need to spawn init first so that it obtains pid 1, however
676 * the init task will end up wanting to create kthreads, which, if
677 * we schedule it before we create kthreadd, will OOPS.
679 pid
= kernel_thread(kernel_init
, NULL
, CLONE_FS
);
681 * Pin init on the boot CPU. Task migration is not properly working
682 * until sched_init_smp() has been run. It will set the allowed
683 * CPUs for init to the non isolated CPUs.
686 tsk
= find_task_by_pid_ns(pid
, &init_pid_ns
);
687 tsk
->flags
|= PF_NO_SETAFFINITY
;
688 set_cpus_allowed_ptr(tsk
, cpumask_of(smp_processor_id()));
691 numa_default_policy();
692 pid
= kernel_thread(kthreadd
, NULL
, CLONE_FS
| CLONE_FILES
);
694 kthreadd_task
= find_task_by_pid_ns(pid
, &init_pid_ns
);
698 * Enable might_sleep() and smp_processor_id() checks.
699 * They cannot be enabled earlier because with CONFIG_PREEMPTION=y
700 * kernel_thread() would trigger might_sleep() splats. With
701 * CONFIG_PREEMPT_VOLUNTARY=y the init task might have scheduled
702 * already, but it's stuck on the kthreadd_done completion.
704 system_state
= SYSTEM_SCHEDULING
;
706 complete(&kthreadd_done
);
709 * The boot idle thread must execute schedule()
710 * at least once to get things moving:
712 schedule_preempt_disabled();
713 /* Call into cpu_idle with preempt disabled */
714 cpu_startup_entry(CPUHP_ONLINE
);
717 /* Check for early params. */
718 static int __init
do_early_param(char *param
, char *val
,
719 const char *unused
, void *arg
)
721 const struct obs_kernel_param
*p
;
723 for (p
= __setup_start
; p
< __setup_end
; p
++) {
724 if ((p
->early
&& parameq(param
, p
->str
)) ||
725 (strcmp(param
, "console") == 0 &&
726 strcmp(p
->str
, "earlycon") == 0)
728 if (p
->setup_func(val
) != 0)
729 pr_warn("Malformed early option '%s'\n", param
);
732 /* We accept everything at this stage. */
736 void __init
parse_early_options(char *cmdline
)
738 parse_args("early options", cmdline
, NULL
, 0, 0, 0, NULL
,
742 /* Arch code calls this early on, or if not, just before other parsing. */
743 void __init
parse_early_param(void)
745 static int done __initdata
;
746 static char tmp_cmdline
[COMMAND_LINE_SIZE
] __initdata
;
751 /* All fall through to do_early_param. */
752 strlcpy(tmp_cmdline
, boot_command_line
, COMMAND_LINE_SIZE
);
753 parse_early_options(tmp_cmdline
);
757 void __init __weak
arch_post_acpi_subsys_init(void) { }
759 void __init __weak
smp_setup_processor_id(void)
763 # if THREAD_SIZE >= PAGE_SIZE
764 void __init __weak
thread_stack_cache_init(void)
769 void __init __weak
mem_encrypt_init(void) { }
771 void __init __weak
poking_init(void) { }
773 void __init __weak
pgtable_cache_init(void) { }
776 core_param(initcall_debug
, initcall_debug
, bool, 0644);
778 #ifdef TRACEPOINTS_ENABLED
779 static void __init
initcall_debug_enable(void);
781 static inline void initcall_debug_enable(void)
786 /* Report memory auto-initialization states for this boot. */
787 static void __init
report_meminit(void)
791 if (IS_ENABLED(CONFIG_INIT_STACK_ALL_PATTERN
))
792 stack
= "all(pattern)";
793 else if (IS_ENABLED(CONFIG_INIT_STACK_ALL_ZERO
))
795 else if (IS_ENABLED(CONFIG_GCC_PLUGIN_STRUCTLEAK_BYREF_ALL
))
796 stack
= "byref_all(zero)";
797 else if (IS_ENABLED(CONFIG_GCC_PLUGIN_STRUCTLEAK_BYREF
))
798 stack
= "byref(zero)";
799 else if (IS_ENABLED(CONFIG_GCC_PLUGIN_STRUCTLEAK_USER
))
800 stack
= "__user(zero)";
804 pr_info("mem auto-init: stack:%s, heap alloc:%s, heap free:%s\n",
805 stack
, want_init_on_alloc(GFP_KERNEL
) ? "on" : "off",
806 want_init_on_free() ? "on" : "off");
807 if (want_init_on_free())
808 pr_info("mem auto-init: clearing system memory may take some time...\n");
812 * Set up kernel memory allocators
814 static void __init
mm_init(void)
817 * page_ext requires contiguous pages,
818 * bigger than MAX_ORDER unless SPARSEMEM.
820 page_ext_init_flatmem();
821 init_mem_debugging_and_hardening();
826 mem_init_print_info();
827 /* page_owner must be initialized after buddy is ready */
828 page_ext_init_flatmem_late();
832 debug_objects_mem_init();
834 /* Should be run before the first non-init thread is created */
836 /* Should be run after espfix64 is set up. */
840 #ifdef CONFIG_HAVE_ARCH_RANDOMIZE_KSTACK_OFFSET
841 DEFINE_STATIC_KEY_MAYBE_RO(CONFIG_RANDOMIZE_KSTACK_OFFSET_DEFAULT
,
842 randomize_kstack_offset
);
843 DEFINE_PER_CPU(u32
, kstack_offset
);
845 static int __init
early_randomize_kstack_offset(char *buf
)
850 ret
= kstrtobool(buf
, &bool_result
);
855 static_branch_enable(&randomize_kstack_offset
);
857 static_branch_disable(&randomize_kstack_offset
);
860 early_param("randomize_kstack_offset", early_randomize_kstack_offset
);
863 void __init __weak
arch_call_rest_init(void)
868 static void __init
print_unknown_bootoptions(void)
870 char *unknown_options
;
872 const char *const *p
;
875 if (panic_later
|| (!argv_init
[1] && !envp_init
[2]))
879 * Determine how many options we have to print out, plus a space
882 len
= 1; /* null terminator */
883 for (p
= &argv_init
[1]; *p
; p
++) {
887 for (p
= &envp_init
[2]; *p
; p
++) {
892 unknown_options
= memblock_alloc(len
, SMP_CACHE_BYTES
);
893 if (!unknown_options
) {
894 pr_err("%s: Failed to allocate %zu bytes\n",
898 end
= unknown_options
;
900 for (p
= &argv_init
[1]; *p
; p
++)
901 end
+= sprintf(end
, " %s", *p
);
902 for (p
= &envp_init
[2]; *p
; p
++)
903 end
+= sprintf(end
, " %s", *p
);
905 pr_notice("Unknown command line parameters:%s\n", unknown_options
);
906 memblock_free(__pa(unknown_options
), len
);
909 asmlinkage __visible
void __init __no_sanitize_address
start_kernel(void)
914 set_task_stack_end_magic(&init_task
);
915 smp_setup_processor_id();
916 debug_objects_early_init();
917 init_vmlinux_build_id();
922 early_boot_irqs_disabled
= true;
925 * Interrupts are still disabled. Do necessary setups, then
930 pr_notice("%s", linux_banner
);
931 early_security_init();
932 setup_arch(&command_line
);
934 setup_command_line(command_line
);
936 setup_per_cpu_areas();
937 smp_prepare_boot_cpu(); /* arch-specific boot-cpu hooks */
938 boot_cpu_hotplug_init();
940 build_all_zonelists(NULL
);
943 pr_notice("Kernel command line: %s\n", saved_command_line
);
944 /* parameters may set static keys */
947 after_dashes
= parse_args("Booting kernel",
948 static_command_line
, __start___param
,
949 __stop___param
- __start___param
,
950 -1, -1, NULL
, &unknown_bootoption
);
951 print_unknown_bootoptions();
952 if (!IS_ERR_OR_NULL(after_dashes
))
953 parse_args("Setting init args", after_dashes
, NULL
, 0, -1, -1,
956 parse_args("Setting extra init args", extra_init_args
,
957 NULL
, 0, -1, -1, NULL
, set_init_arg
);
960 * These use large bootmem allocations and must precede
964 vfs_caches_init_early();
971 /* trace_printk can be enabled here */
975 * Set up the scheduler prior starting any interrupts (such as the
976 * timer interrupt). Full topology setup happens at smp_init()
977 * time - but meanwhile we still have a functioning scheduler.
981 if (WARN(!irqs_disabled(),
982 "Interrupts were enabled *very* early, fixing it\n"))
987 * Set up housekeeping before setting up workqueues to allow the unbound
988 * workqueue to take non-housekeeping into account.
993 * Allow workqueue creation and work item queueing/cancelling
994 * early. Work item execution depends on kthreads and starts after
997 workqueue_init_early();
1001 /* Trace events are available after this */
1005 initcall_debug_enable();
1007 context_tracking_init();
1008 /* init some links before init_ISA_irqs() */
1021 * For best initial stack canary entropy, prepare it after:
1022 * - setup_arch() for any UEFI RNG entropy and boot cmdline access
1023 * - timekeeping_init() for ktime entropy used in rand_initialize()
1024 * - rand_initialize() to get any arch-specific entropy like RDRAND
1025 * - add_latent_entropy() to get any latent entropy
1026 * - adding command line entropy
1029 add_latent_entropy();
1030 add_device_randomness(command_line
, strlen(command_line
));
1031 boot_init_stack_canary();
1036 call_function_init();
1037 WARN(!irqs_disabled(), "Interrupts were enabled early\n");
1039 early_boot_irqs_disabled
= false;
1042 kmem_cache_init_late();
1045 * HACK ALERT! This is early. We're enabling the console before
1046 * we've done PCI setups etc, and console_init() must be aware of
1047 * this. But we do want output early, in case something goes wrong.
1051 panic("Too many boot %s vars at `%s'", panic_later
,
1057 * Need to run this when irqs are enabled, because it wants
1058 * to self-test [hard/soft]-irqs on/off lock inversion bugs
1064 * This needs to be called before any devices perform DMA
1065 * operations that might use the SWIOTLB bounce buffers. It will
1066 * mark the bounce buffers as decrypted so that their usage will
1067 * not cause "plain-text" data to be decrypted when accessed.
1071 #ifdef CONFIG_BLK_DEV_INITRD
1072 if (initrd_start
&& !initrd_below_start_ok
&&
1073 page_to_pfn(virt_to_page((void *)initrd_start
)) < min_low_pfn
) {
1074 pr_crit("initrd overwritten (0x%08lx < 0x%08lx) - disabling it.\n",
1075 page_to_pfn(virt_to_page((void *)initrd_start
)),
1080 setup_per_cpu_pageset();
1090 if (efi_enabled(EFI_RUNTIME_SERVICES
))
1091 efi_enter_virtual_mode();
1093 thread_stack_cache_init();
1109 taskstats_init_early();
1115 acpi_subsystem_init();
1116 arch_post_acpi_subsys_init();
1119 /* Do the rest non-__init'ed, we're now alive */
1120 arch_call_rest_init();
1122 prevent_tail_call_optimization();
1125 /* Call all constructor functions linked into the kernel. */
1126 static void __init
do_ctors(void)
1129 * For UML, the constructors have already been called by the
1130 * normal setup code as it's just a normal ELF binary, so we
1131 * cannot do it again - but we do need CONFIG_CONSTRUCTORS
1132 * even on UML for modules.
1134 #if defined(CONFIG_CONSTRUCTORS) && !defined(CONFIG_UML)
1135 ctor_fn_t
*fn
= (ctor_fn_t
*) __ctors_start
;
1137 for (; fn
< (ctor_fn_t
*) __ctors_end
; fn
++)
1142 #ifdef CONFIG_KALLSYMS
1143 struct blacklist_entry
{
1144 struct list_head next
;
1148 static __initdata_or_module
LIST_HEAD(blacklisted_initcalls
);
1150 static int __init
initcall_blacklist(char *str
)
1153 struct blacklist_entry
*entry
;
1155 /* str argument is a comma-separated list of functions */
1157 str_entry
= strsep(&str
, ",");
1159 pr_debug("blacklisting initcall %s\n", str_entry
);
1160 entry
= memblock_alloc(sizeof(*entry
),
1163 panic("%s: Failed to allocate %zu bytes\n",
1164 __func__
, sizeof(*entry
));
1165 entry
->buf
= memblock_alloc(strlen(str_entry
) + 1,
1168 panic("%s: Failed to allocate %zu bytes\n",
1169 __func__
, strlen(str_entry
) + 1);
1170 strcpy(entry
->buf
, str_entry
);
1171 list_add(&entry
->next
, &blacklisted_initcalls
);
1173 } while (str_entry
);
1178 static bool __init_or_module
initcall_blacklisted(initcall_t fn
)
1180 struct blacklist_entry
*entry
;
1181 char fn_name
[KSYM_SYMBOL_LEN
];
1184 if (list_empty(&blacklisted_initcalls
))
1187 addr
= (unsigned long) dereference_function_descriptor(fn
);
1188 sprint_symbol_no_offset(fn_name
, addr
);
1191 * fn will be "function_name [module_name]" where [module_name] is not
1192 * displayed for built-in init functions. Strip off the [module_name].
1194 strreplace(fn_name
, ' ', '\0');
1196 list_for_each_entry(entry
, &blacklisted_initcalls
, next
) {
1197 if (!strcmp(fn_name
, entry
->buf
)) {
1198 pr_debug("initcall %s blacklisted\n", fn_name
);
1206 static int __init
initcall_blacklist(char *str
)
1208 pr_warn("initcall_blacklist requires CONFIG_KALLSYMS\n");
1212 static bool __init_or_module
initcall_blacklisted(initcall_t fn
)
1217 __setup("initcall_blacklist=", initcall_blacklist
);
1219 static __init_or_module
void
1220 trace_initcall_start_cb(void *data
, initcall_t fn
)
1222 ktime_t
*calltime
= (ktime_t
*)data
;
1224 printk(KERN_DEBUG
"calling %pS @ %i\n", fn
, task_pid_nr(current
));
1225 *calltime
= ktime_get();
1228 static __init_or_module
void
1229 trace_initcall_finish_cb(void *data
, initcall_t fn
, int ret
)
1231 ktime_t
*calltime
= (ktime_t
*)data
;
1232 ktime_t delta
, rettime
;
1233 unsigned long long duration
;
1235 rettime
= ktime_get();
1236 delta
= ktime_sub(rettime
, *calltime
);
1237 duration
= (unsigned long long) ktime_to_ns(delta
) >> 10;
1238 printk(KERN_DEBUG
"initcall %pS returned %d after %lld usecs\n",
1242 static ktime_t initcall_calltime
;
1244 #ifdef TRACEPOINTS_ENABLED
1245 static void __init
initcall_debug_enable(void)
1249 ret
= register_trace_initcall_start(trace_initcall_start_cb
,
1250 &initcall_calltime
);
1251 ret
|= register_trace_initcall_finish(trace_initcall_finish_cb
,
1252 &initcall_calltime
);
1253 WARN(ret
, "Failed to register initcall tracepoints\n");
1255 # define do_trace_initcall_start trace_initcall_start
1256 # define do_trace_initcall_finish trace_initcall_finish
1258 static inline void do_trace_initcall_start(initcall_t fn
)
1260 if (!initcall_debug
)
1262 trace_initcall_start_cb(&initcall_calltime
, fn
);
1264 static inline void do_trace_initcall_finish(initcall_t fn
, int ret
)
1266 if (!initcall_debug
)
1268 trace_initcall_finish_cb(&initcall_calltime
, fn
, ret
);
1270 #endif /* !TRACEPOINTS_ENABLED */
1272 int __init_or_module
do_one_initcall(initcall_t fn
)
1274 int count
= preempt_count();
1278 if (initcall_blacklisted(fn
))
1281 do_trace_initcall_start(fn
);
1283 do_trace_initcall_finish(fn
, ret
);
1287 if (preempt_count() != count
) {
1288 sprintf(msgbuf
, "preemption imbalance ");
1289 preempt_count_set(count
);
1291 if (irqs_disabled()) {
1292 strlcat(msgbuf
, "disabled interrupts ", sizeof(msgbuf
));
1295 WARN(msgbuf
[0], "initcall %pS returned with %s\n", fn
, msgbuf
);
1297 add_latent_entropy();
1302 extern initcall_entry_t __initcall_start
[];
1303 extern initcall_entry_t __initcall0_start
[];
1304 extern initcall_entry_t __initcall1_start
[];
1305 extern initcall_entry_t __initcall2_start
[];
1306 extern initcall_entry_t __initcall3_start
[];
1307 extern initcall_entry_t __initcall4_start
[];
1308 extern initcall_entry_t __initcall5_start
[];
1309 extern initcall_entry_t __initcall6_start
[];
1310 extern initcall_entry_t __initcall7_start
[];
1311 extern initcall_entry_t __initcall_end
[];
1313 static initcall_entry_t
*initcall_levels
[] __initdata
= {
1325 /* Keep these in sync with initcalls in include/linux/init.h */
1326 static const char *initcall_level_names
[] __initdata
= {
1337 static int __init
ignore_unknown_bootoption(char *param
, char *val
,
1338 const char *unused
, void *arg
)
1343 static void __init
do_initcall_level(int level
, char *command_line
)
1345 initcall_entry_t
*fn
;
1347 parse_args(initcall_level_names
[level
],
1348 command_line
, __start___param
,
1349 __stop___param
- __start___param
,
1351 NULL
, ignore_unknown_bootoption
);
1353 trace_initcall_level(initcall_level_names
[level
]);
1354 for (fn
= initcall_levels
[level
]; fn
< initcall_levels
[level
+1]; fn
++)
1355 do_one_initcall(initcall_from_entry(fn
));
1358 static void __init
do_initcalls(void)
1361 size_t len
= strlen(saved_command_line
) + 1;
1364 command_line
= kzalloc(len
, GFP_KERNEL
);
1366 panic("%s: Failed to allocate %zu bytes\n", __func__
, len
);
1368 for (level
= 0; level
< ARRAY_SIZE(initcall_levels
) - 1; level
++) {
1369 /* Parser modifies command_line, restore it each time */
1370 strcpy(command_line
, saved_command_line
);
1371 do_initcall_level(level
, command_line
);
1374 kfree(command_line
);
1378 * Ok, the machine is now initialized. None of the devices
1379 * have been touched yet, but the CPU subsystem is up and
1380 * running, and memory and process management works.
1382 * Now we can finally start doing some real work..
1384 static void __init
do_basic_setup(void)
1390 usermodehelper_enable();
1394 static void __init
do_pre_smp_initcalls(void)
1396 initcall_entry_t
*fn
;
1398 trace_initcall_level("early");
1399 for (fn
= __initcall_start
; fn
< __initcall0_start
; fn
++)
1400 do_one_initcall(initcall_from_entry(fn
));
1403 static int run_init_process(const char *init_filename
)
1405 const char *const *p
;
1407 argv_init
[0] = init_filename
;
1408 pr_info("Run %s as init process\n", init_filename
);
1409 pr_debug(" with arguments:\n");
1410 for (p
= argv_init
; *p
; p
++)
1411 pr_debug(" %s\n", *p
);
1412 pr_debug(" with environment:\n");
1413 for (p
= envp_init
; *p
; p
++)
1414 pr_debug(" %s\n", *p
);
1415 return kernel_execve(init_filename
, argv_init
, envp_init
);
1418 static int try_to_run_init_process(const char *init_filename
)
1422 ret
= run_init_process(init_filename
);
1424 if (ret
&& ret
!= -ENOENT
) {
1425 pr_err("Starting init: %s exists but couldn't execute it (error %d)\n",
1426 init_filename
, ret
);
1432 static noinline
void __init
kernel_init_freeable(void);
1434 #if defined(CONFIG_STRICT_KERNEL_RWX) || defined(CONFIG_STRICT_MODULE_RWX)
1435 bool rodata_enabled __ro_after_init
= true;
1436 static int __init
set_debug_rodata(char *str
)
1438 return strtobool(str
, &rodata_enabled
);
1440 __setup("rodata=", set_debug_rodata
);
1443 #ifdef CONFIG_STRICT_KERNEL_RWX
1444 static void mark_readonly(void)
1446 if (rodata_enabled
) {
1448 * load_module() results in W+X mappings, which are cleaned
1449 * up with call_rcu(). Let's make sure that queued work is
1450 * flushed so that we don't hit false positives looking for
1451 * insecure pages which are W+X.
1457 pr_info("Kernel memory protection disabled.\n");
1459 #elif defined(CONFIG_ARCH_HAS_STRICT_KERNEL_RWX)
1460 static inline void mark_readonly(void)
1462 pr_warn("Kernel memory protection not selected by kernel config.\n");
1465 static inline void mark_readonly(void)
1467 pr_warn("This architecture does not have kernel memory protection.\n");
1471 void __weak
free_initmem(void)
1473 free_initmem_default(POISON_FREE_INITMEM
);
1476 static int __ref
kernel_init(void *unused
)
1481 * Wait until kthreadd is all set-up.
1483 wait_for_completion(&kthreadd_done
);
1485 kernel_init_freeable();
1486 /* need to finish all async __init code before freeing the memory */
1487 async_synchronize_full();
1488 kprobe_free_init_mem();
1489 ftrace_free_init_mem();
1490 kgdb_free_init_mem();
1495 * Kernel mappings are now finalized - update the userspace page-table
1500 system_state
= SYSTEM_RUNNING
;
1501 numa_default_policy();
1503 rcu_end_inkernel_boot();
1507 if (ramdisk_execute_command
) {
1508 ret
= run_init_process(ramdisk_execute_command
);
1511 pr_err("Failed to execute %s (error %d)\n",
1512 ramdisk_execute_command
, ret
);
1516 * We try each of these until one succeeds.
1518 * The Bourne shell can be used instead of init if we are
1519 * trying to recover a really broken machine.
1521 if (execute_command
) {
1522 ret
= run_init_process(execute_command
);
1525 panic("Requested init %s failed (error %d).",
1526 execute_command
, ret
);
1529 if (CONFIG_DEFAULT_INIT
[0] != '\0') {
1530 ret
= run_init_process(CONFIG_DEFAULT_INIT
);
1532 pr_err("Default init %s failed (error %d)\n",
1533 CONFIG_DEFAULT_INIT
, ret
);
1538 if (!try_to_run_init_process("/sbin/init") ||
1539 !try_to_run_init_process("/etc/init") ||
1540 !try_to_run_init_process("/bin/init") ||
1541 !try_to_run_init_process("/bin/sh"))
1544 panic("No working init found. Try passing init= option to kernel. "
1545 "See Linux Documentation/admin-guide/init.rst for guidance.");
1548 /* Open /dev/console, for stdin/stdout/stderr, this should never fail */
1549 void __init
console_on_rootfs(void)
1551 struct file
*file
= filp_open("/dev/console", O_RDWR
, 0);
1554 pr_err("Warning: unable to open an initial console.\n");
1563 static noinline
void __init
kernel_init_freeable(void)
1565 /* Now the scheduler is fully set up and can do blocking allocations */
1566 gfp_allowed_mask
= __GFP_BITS_MASK
;
1569 * init can allocate pages on any node
1571 set_mems_allowed(node_states
[N_MEMORY
]);
1573 cad_pid
= get_pid(task_pid(current
));
1575 smp_prepare_cpus(setup_max_cpus
);
1579 init_mm_internals();
1581 rcu_init_tasks_generic();
1582 do_pre_smp_initcalls();
1583 lockup_detector_init();
1589 page_alloc_init_late();
1590 /* Initialize page ext after all struct pages are initialized. */
1595 kunit_run_all_tests();
1597 wait_for_initramfs();
1598 console_on_rootfs();
1601 * check if there is an early userspace init. If yes, let it do all
1604 if (init_eaccess(ramdisk_execute_command
) != 0) {
1605 ramdisk_execute_command
= NULL
;
1606 prepare_namespace();
1610 * Ok, we have completed the initial bootup, and
1611 * we're essentially up and running. Get rid of the
1612 * initmem segments and start the user-mode stuff..
1614 * rootfs is available now, try loading the public keys
1615 * and default modules
1618 integrity_load_keys();