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>
103 #include <linux/randomize_kstack.h>
104 #include <net/net_namespace.h>
107 #include <asm/bugs.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 *);
119 extern void init_IRQ(void);
120 extern void radix_tree_init(void);
123 * Debug helper: via this flag we know that we are in 'early bootup code'
124 * where only the boot processor is running with IRQ disabled. This means
125 * two things - IRQ must not be enabled before the flag is cleared and some
126 * operations which are not allowed with IRQ disabled are allowed while the
129 bool early_boot_irqs_disabled __read_mostly
;
131 enum system_states system_state __read_mostly
;
132 EXPORT_SYMBOL(system_state
);
135 * Boot command-line arguments
137 #define MAX_INIT_ARGS CONFIG_INIT_ENV_ARG_LIMIT
138 #define MAX_INIT_ENVS CONFIG_INIT_ENV_ARG_LIMIT
140 extern void time_init(void);
141 /* Default late time init is NULL. archs can override this later. */
142 void (*__initdata late_time_init
)(void);
144 /* Untouched command line saved by arch-specific code. */
145 char __initdata boot_command_line
[COMMAND_LINE_SIZE
];
146 /* Untouched saved command line (eg. for /proc) */
147 char *saved_command_line
;
148 /* Command line for parameter parsing */
149 static char *static_command_line
;
150 /* Untouched extra command line */
151 static char *extra_command_line
;
152 /* Extra init arguments */
153 static char *extra_init_args
;
155 #ifdef CONFIG_BOOT_CONFIG
156 /* Is bootconfig on command line? */
157 static bool bootconfig_found
;
158 static size_t initargs_offs
;
160 # define bootconfig_found false
161 # define initargs_offs 0
164 static char *execute_command
;
165 static char *ramdisk_execute_command
= "/init";
168 * Used to generate warnings if static_key manipulation functions are used
169 * before jump_label_init is called.
171 bool static_key_initialized __read_mostly
;
172 EXPORT_SYMBOL_GPL(static_key_initialized
);
175 * If set, this is an indication to the drivers that reset the underlying
176 * device before going ahead with the initialization otherwise driver might
177 * rely on the BIOS and skip the reset operation.
179 * This is useful if kernel is booting in an unreliable environment.
180 * For ex. kdump situation where previous kernel has crashed, BIOS has been
181 * skipped and devices will be in unknown state.
183 unsigned int reset_devices
;
184 EXPORT_SYMBOL(reset_devices
);
186 static int __init
set_reset_devices(char *str
)
192 __setup("reset_devices", set_reset_devices
);
194 static const char *argv_init
[MAX_INIT_ARGS
+2] = { "init", NULL
, };
195 const char *envp_init
[MAX_INIT_ENVS
+2] = { "HOME=/", "TERM=linux", NULL
, };
196 static const char *panic_later
, *panic_param
;
198 extern const struct obs_kernel_param __setup_start
[], __setup_end
[];
200 static bool __init
obsolete_checksetup(char *line
)
202 const struct obs_kernel_param
*p
;
203 bool had_early_param
= false;
207 int n
= strlen(p
->str
);
208 if (parameqn(line
, p
->str
, n
)) {
210 /* Already done in parse_early_param?
211 * (Needs exact match on param part).
212 * Keep iterating, as we can have early
213 * params and __setups of same names 8( */
214 if (line
[n
] == '\0' || line
[n
] == '=')
215 had_early_param
= true;
216 } else if (!p
->setup_func
) {
217 pr_warn("Parameter %s is obsolete, ignored\n",
220 } else if (p
->setup_func(line
+ n
))
224 } while (p
< __setup_end
);
226 return had_early_param
;
230 * This should be approx 2 Bo*oMips to start (note initial shift), and will
231 * still work even if initially too large, it will just take slightly longer
233 unsigned long loops_per_jiffy
= (1<<12);
234 EXPORT_SYMBOL(loops_per_jiffy
);
236 static int __init
debug_kernel(char *str
)
238 console_loglevel
= CONSOLE_LOGLEVEL_DEBUG
;
242 static int __init
quiet_kernel(char *str
)
244 console_loglevel
= CONSOLE_LOGLEVEL_QUIET
;
248 early_param("debug", debug_kernel
);
249 early_param("quiet", quiet_kernel
);
251 static int __init
loglevel(char *str
)
256 * Only update loglevel value when a correct setting was passed,
257 * to prevent blind crashes (when loglevel being set to 0) that
258 * are quite hard to debug
260 if (get_option(&str
, &newlevel
)) {
261 console_loglevel
= newlevel
;
268 early_param("loglevel", loglevel
);
270 #ifdef CONFIG_BLK_DEV_INITRD
271 static void * __init
get_boot_config_from_initrd(u32
*_size
, u32
*_csum
)
281 data
= (char *)initrd_end
- BOOTCONFIG_MAGIC_LEN
;
283 * Since Grub may align the size of initrd to 4, we must
284 * check the preceding 3 bytes as well.
286 for (i
= 0; i
< 4; i
++) {
287 if (!memcmp(data
, BOOTCONFIG_MAGIC
, BOOTCONFIG_MAGIC_LEN
))
294 hdr
= (u32
*)(data
- 8);
295 size
= le32_to_cpu(hdr
[0]);
296 csum
= le32_to_cpu(hdr
[1]);
298 data
= ((void *)hdr
) - size
;
299 if ((unsigned long)data
< initrd_start
) {
300 pr_err("bootconfig size %d is greater than initrd size %ld\n",
301 size
, initrd_end
- initrd_start
);
305 /* Remove bootconfig from initramfs/initrd */
306 initrd_end
= (unsigned long)data
;
315 static void * __init
get_boot_config_from_initrd(u32
*_size
, u32
*_csum
)
321 #ifdef CONFIG_BOOT_CONFIG
323 static char xbc_namebuf
[XBC_KEYLEN_MAX
] __initdata
;
325 #define rest(dst, end) ((end) > (dst) ? (end) - (dst) : 0)
327 static int __init
xbc_snprint_cmdline(char *buf
, size_t size
,
328 struct xbc_node
*root
)
330 struct xbc_node
*knode
, *vnode
;
331 char *end
= buf
+ size
;
335 xbc_node_for_each_key_value(root
, knode
, val
) {
336 ret
= xbc_node_compose_key_after(root
, knode
,
337 xbc_namebuf
, XBC_KEYLEN_MAX
);
341 vnode
= xbc_node_get_child(knode
);
343 ret
= snprintf(buf
, rest(buf
, end
), "%s ", xbc_namebuf
);
349 xbc_array_for_each_value(vnode
, val
) {
350 ret
= snprintf(buf
, rest(buf
, end
), "%s=\"%s\" ",
358 return buf
- (end
- size
);
362 /* Make an extra command line under given key word */
363 static char * __init
xbc_make_cmdline(const char *key
)
365 struct xbc_node
*root
;
369 root
= xbc_find_node(key
);
373 /* Count required buffer size */
374 len
= xbc_snprint_cmdline(NULL
, 0, root
);
378 new_cmdline
= memblock_alloc(len
+ 1, SMP_CACHE_BYTES
);
380 pr_err("Failed to allocate memory for extra kernel cmdline.\n");
384 ret
= xbc_snprint_cmdline(new_cmdline
, len
+ 1, root
);
385 if (ret
< 0 || ret
> len
) {
386 pr_err("Failed to print extra kernel cmdline.\n");
387 memblock_free_ptr(new_cmdline
, len
+ 1);
394 static int __init
bootconfig_params(char *param
, char *val
,
395 const char *unused
, void *arg
)
397 if (strcmp(param
, "bootconfig") == 0) {
398 bootconfig_found
= true;
403 static int __init
warn_bootconfig(char *str
)
405 /* The 'bootconfig' has been handled by bootconfig_params(). */
409 static void __init
setup_boot_config(void)
411 static char tmp_cmdline
[COMMAND_LINE_SIZE
] __initdata
;
415 char *data
, *copy
, *err
;
418 /* Cut out the bootconfig data even if we have no bootconfig option */
419 data
= get_boot_config_from_initrd(&size
, &csum
);
421 strlcpy(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
)
428 /* parse_args() stops at the next param of '--' and returns an address */
430 initargs_offs
= err
- tmp_cmdline
;
433 pr_err("'bootconfig' found on command line, but no bootconfig found\n");
437 if (size
>= XBC_DATA_MAX
) {
438 pr_err("bootconfig size %d greater than max size %d\n",
443 if (xbc_calc_checksum(data
, size
) != csum
) {
444 pr_err("bootconfig checksum failed\n");
448 copy
= memblock_alloc(size
+ 1, SMP_CACHE_BYTES
);
450 pr_err("Failed to allocate memory for bootconfig\n");
454 memcpy(copy
, data
, size
);
457 ret
= xbc_init(copy
, &msg
, &pos
);
460 pr_err("Failed to init bootconfig: %s.\n", msg
);
462 pr_err("Failed to parse bootconfig: %s at %d.\n",
465 pr_info("Load bootconfig: %d bytes %d nodes\n", size
, ret
);
466 /* keys starting with "kernel." are passed via cmdline */
467 extra_command_line
= xbc_make_cmdline("kernel");
468 /* Also, "init." keys are init arguments */
469 extra_init_args
= xbc_make_cmdline("init");
474 static void __init
exit_boot_config(void)
479 #else /* !CONFIG_BOOT_CONFIG */
481 static void __init
setup_boot_config(void)
483 /* Remove bootconfig data from initrd */
484 get_boot_config_from_initrd(NULL
, NULL
);
487 static int __init
warn_bootconfig(char *str
)
489 pr_warn("WARNING: 'bootconfig' found on the kernel command line but CONFIG_BOOT_CONFIG is not set.\n");
493 #define exit_boot_config() do {} while (0)
495 #endif /* CONFIG_BOOT_CONFIG */
497 early_param("bootconfig", warn_bootconfig
);
499 /* Change NUL term back to "=", to make "param" the whole string. */
500 static void __init
repair_env_string(char *param
, char *val
)
503 /* param=val or param="val"? */
504 if (val
== param
+strlen(param
)+1)
506 else if (val
== param
+strlen(param
)+2) {
508 memmove(val
-1, val
, strlen(val
)+1);
514 /* Anything after -- gets handed straight to init. */
515 static int __init
set_init_arg(char *param
, char *val
,
516 const char *unused
, void *arg
)
523 repair_env_string(param
, val
);
525 for (i
= 0; argv_init
[i
]; i
++) {
526 if (i
== MAX_INIT_ARGS
) {
527 panic_later
= "init";
532 argv_init
[i
] = param
;
537 * Unknown boot options get handed to init, unless they look like
538 * unused parameters (modprobe will find them in /proc/cmdline).
540 static int __init
unknown_bootoption(char *param
, char *val
,
541 const char *unused
, void *arg
)
543 size_t len
= strlen(param
);
545 repair_env_string(param
, val
);
547 /* Handle obsolete-style parameters */
548 if (obsolete_checksetup(param
))
551 /* Unused module parameter. */
552 if (strnchr(param
, len
, '.'))
559 /* Environment option */
561 for (i
= 0; envp_init
[i
]; i
++) {
562 if (i
== MAX_INIT_ENVS
) {
566 if (!strncmp(param
, envp_init
[i
], len
+1))
569 envp_init
[i
] = param
;
571 /* Command line option */
573 for (i
= 0; argv_init
[i
]; i
++) {
574 if (i
== MAX_INIT_ARGS
) {
575 panic_later
= "init";
579 argv_init
[i
] = param
;
584 static int __init
init_setup(char *str
)
588 execute_command
= str
;
590 * In case LILO is going to boot us with default command line,
591 * it prepends "auto" before the whole cmdline which makes
592 * the shell think it should execute a script with such name.
593 * So we ignore all arguments entered _before_ init=... [MJ]
595 for (i
= 1; i
< MAX_INIT_ARGS
; i
++)
599 __setup("init=", init_setup
);
601 static int __init
rdinit_setup(char *str
)
605 ramdisk_execute_command
= str
;
606 /* See "auto" comment in init_setup */
607 for (i
= 1; i
< MAX_INIT_ARGS
; i
++)
611 __setup("rdinit=", rdinit_setup
);
614 static const unsigned int setup_max_cpus
= NR_CPUS
;
615 static inline void setup_nr_cpu_ids(void) { }
616 static inline void smp_prepare_cpus(unsigned int maxcpus
) { }
620 * We need to store the untouched command line for future reference.
621 * We also need to store the touched command line since the parameter
622 * parsing is performed in place, and we should allow a component to
623 * store reference of name/value for future reference.
625 static void __init
setup_command_line(char *command_line
)
627 size_t len
, xlen
= 0, ilen
= 0;
629 if (extra_command_line
)
630 xlen
= strlen(extra_command_line
);
632 ilen
= strlen(extra_init_args
) + 4; /* for " -- " */
634 len
= xlen
+ strlen(boot_command_line
) + 1;
636 saved_command_line
= memblock_alloc(len
+ ilen
, SMP_CACHE_BYTES
);
637 if (!saved_command_line
)
638 panic("%s: Failed to allocate %zu bytes\n", __func__
, len
+ ilen
);
640 static_command_line
= memblock_alloc(len
, SMP_CACHE_BYTES
);
641 if (!static_command_line
)
642 panic("%s: Failed to allocate %zu bytes\n", __func__
, len
);
646 * We have to put extra_command_line before boot command
647 * lines because there could be dashes (separator of init
648 * command line) in the command lines.
650 strcpy(saved_command_line
, extra_command_line
);
651 strcpy(static_command_line
, extra_command_line
);
653 strcpy(saved_command_line
+ xlen
, boot_command_line
);
654 strcpy(static_command_line
+ xlen
, command_line
);
658 * Append supplemental init boot args to saved_command_line
659 * so that user can check what command line options passed
661 * The order should always be
662 * " -- "[bootconfig init-param][cmdline init-param]
665 len
= xlen
+ initargs_offs
;
666 strcpy(saved_command_line
+ len
, extra_init_args
);
667 len
+= ilen
- 4; /* strlen(extra_init_args) */
668 strcpy(saved_command_line
+ len
,
669 boot_command_line
+ initargs_offs
- 1);
671 len
= strlen(saved_command_line
);
672 strcpy(saved_command_line
+ len
, " -- ");
674 strcpy(saved_command_line
+ len
, extra_init_args
);
680 * We need to finalize in a non-__init function or else race conditions
681 * between the root thread and the init thread may cause start_kernel to
682 * be reaped by free_initmem before the root thread has proceeded to
685 * gcc-3.4 accidentally inlines this function, so use noinline.
688 static __initdata
DECLARE_COMPLETION(kthreadd_done
);
690 noinline
void __ref
rest_init(void)
692 struct task_struct
*tsk
;
695 rcu_scheduler_starting();
697 * We need to spawn init first so that it obtains pid 1, however
698 * the init task will end up wanting to create kthreads, which, if
699 * we schedule it before we create kthreadd, will OOPS.
701 pid
= kernel_thread(kernel_init
, NULL
, CLONE_FS
);
703 * Pin init on the boot CPU. Task migration is not properly working
704 * until sched_init_smp() has been run. It will set the allowed
705 * CPUs for init to the non isolated CPUs.
708 tsk
= find_task_by_pid_ns(pid
, &init_pid_ns
);
709 tsk
->flags
|= PF_NO_SETAFFINITY
;
710 set_cpus_allowed_ptr(tsk
, cpumask_of(smp_processor_id()));
713 numa_default_policy();
714 pid
= kernel_thread(kthreadd
, NULL
, CLONE_FS
| CLONE_FILES
);
716 kthreadd_task
= find_task_by_pid_ns(pid
, &init_pid_ns
);
720 * Enable might_sleep() and smp_processor_id() checks.
721 * They cannot be enabled earlier because with CONFIG_PREEMPTION=y
722 * kernel_thread() would trigger might_sleep() splats. With
723 * CONFIG_PREEMPT_VOLUNTARY=y the init task might have scheduled
724 * already, but it's stuck on the kthreadd_done completion.
726 system_state
= SYSTEM_SCHEDULING
;
728 complete(&kthreadd_done
);
731 * The boot idle thread must execute schedule()
732 * at least once to get things moving:
734 schedule_preempt_disabled();
735 /* Call into cpu_idle with preempt disabled */
736 cpu_startup_entry(CPUHP_ONLINE
);
739 /* Check for early params. */
740 static int __init
do_early_param(char *param
, char *val
,
741 const char *unused
, void *arg
)
743 const struct obs_kernel_param
*p
;
745 for (p
= __setup_start
; p
< __setup_end
; p
++) {
746 if ((p
->early
&& parameq(param
, p
->str
)) ||
747 (strcmp(param
, "console") == 0 &&
748 strcmp(p
->str
, "earlycon") == 0)
750 if (p
->setup_func(val
) != 0)
751 pr_warn("Malformed early option '%s'\n", param
);
754 /* We accept everything at this stage. */
758 void __init
parse_early_options(char *cmdline
)
760 parse_args("early options", cmdline
, NULL
, 0, 0, 0, NULL
,
764 /* Arch code calls this early on, or if not, just before other parsing. */
765 void __init
parse_early_param(void)
767 static int done __initdata
;
768 static char tmp_cmdline
[COMMAND_LINE_SIZE
] __initdata
;
773 /* All fall through to do_early_param. */
774 strlcpy(tmp_cmdline
, boot_command_line
, COMMAND_LINE_SIZE
);
775 parse_early_options(tmp_cmdline
);
779 void __init __weak
arch_post_acpi_subsys_init(void) { }
781 void __init __weak
smp_setup_processor_id(void)
785 # if THREAD_SIZE >= PAGE_SIZE
786 void __init __weak
thread_stack_cache_init(void)
791 void __init __weak
mem_encrypt_init(void) { }
793 void __init __weak
poking_init(void) { }
795 void __init __weak
pgtable_cache_init(void) { }
797 void __init __weak
trap_init(void) { }
800 core_param(initcall_debug
, initcall_debug
, bool, 0644);
802 #ifdef TRACEPOINTS_ENABLED
803 static void __init
initcall_debug_enable(void);
805 static inline void initcall_debug_enable(void)
810 /* Report memory auto-initialization states for this boot. */
811 static void __init
report_meminit(void)
815 if (IS_ENABLED(CONFIG_INIT_STACK_ALL_PATTERN
))
816 stack
= "all(pattern)";
817 else if (IS_ENABLED(CONFIG_INIT_STACK_ALL_ZERO
))
819 else if (IS_ENABLED(CONFIG_GCC_PLUGIN_STRUCTLEAK_BYREF_ALL
))
820 stack
= "byref_all(zero)";
821 else if (IS_ENABLED(CONFIG_GCC_PLUGIN_STRUCTLEAK_BYREF
))
822 stack
= "byref(zero)";
823 else if (IS_ENABLED(CONFIG_GCC_PLUGIN_STRUCTLEAK_USER
))
824 stack
= "__user(zero)";
828 pr_info("mem auto-init: stack:%s, heap alloc:%s, heap free:%s\n",
829 stack
, want_init_on_alloc(GFP_KERNEL
) ? "on" : "off",
830 want_init_on_free() ? "on" : "off");
831 if (want_init_on_free())
832 pr_info("mem auto-init: clearing system memory may take some time...\n");
836 * Set up kernel memory allocators
838 static void __init
mm_init(void)
841 * page_ext requires contiguous pages,
842 * bigger than MAX_ORDER unless SPARSEMEM.
844 page_ext_init_flatmem();
845 init_mem_debugging_and_hardening();
850 mem_init_print_info();
851 /* page_owner must be initialized after buddy is ready */
852 page_ext_init_flatmem_late();
856 debug_objects_mem_init();
858 /* Should be run before the first non-init thread is created */
860 /* Should be run after espfix64 is set up. */
864 #ifdef CONFIG_HAVE_ARCH_RANDOMIZE_KSTACK_OFFSET
865 DEFINE_STATIC_KEY_MAYBE_RO(CONFIG_RANDOMIZE_KSTACK_OFFSET_DEFAULT
,
866 randomize_kstack_offset
);
867 DEFINE_PER_CPU(u32
, kstack_offset
);
869 static int __init
early_randomize_kstack_offset(char *buf
)
874 ret
= kstrtobool(buf
, &bool_result
);
879 static_branch_enable(&randomize_kstack_offset
);
881 static_branch_disable(&randomize_kstack_offset
);
884 early_param("randomize_kstack_offset", early_randomize_kstack_offset
);
887 void __init __weak
arch_call_rest_init(void)
892 static void __init
print_unknown_bootoptions(void)
894 char *unknown_options
;
896 const char *const *p
;
899 if (panic_later
|| (!argv_init
[1] && !envp_init
[2]))
903 * Determine how many options we have to print out, plus a space
906 len
= 1; /* null terminator */
907 for (p
= &argv_init
[1]; *p
; p
++) {
911 for (p
= &envp_init
[2]; *p
; p
++) {
916 unknown_options
= memblock_alloc(len
, SMP_CACHE_BYTES
);
917 if (!unknown_options
) {
918 pr_err("%s: Failed to allocate %zu bytes\n",
922 end
= unknown_options
;
924 for (p
= &argv_init
[1]; *p
; p
++)
925 end
+= sprintf(end
, " %s", *p
);
926 for (p
= &envp_init
[2]; *p
; p
++)
927 end
+= sprintf(end
, " %s", *p
);
929 /* Start at unknown_options[1] to skip the initial space */
930 pr_notice("Unknown kernel command line parameters \"%s\", will be passed to user space.\n",
931 &unknown_options
[1]);
932 memblock_free_ptr(unknown_options
, len
);
935 asmlinkage __visible
void __init __no_sanitize_address
start_kernel(void)
940 set_task_stack_end_magic(&init_task
);
941 smp_setup_processor_id();
942 debug_objects_early_init();
943 init_vmlinux_build_id();
948 early_boot_irqs_disabled
= true;
951 * Interrupts are still disabled. Do necessary setups, then
956 pr_notice("%s", linux_banner
);
957 early_security_init();
958 setup_arch(&command_line
);
960 setup_command_line(command_line
);
962 setup_per_cpu_areas();
963 smp_prepare_boot_cpu(); /* arch-specific boot-cpu hooks */
964 boot_cpu_hotplug_init();
966 build_all_zonelists(NULL
);
969 pr_notice("Kernel command line: %s\n", saved_command_line
);
970 /* parameters may set static keys */
973 after_dashes
= parse_args("Booting kernel",
974 static_command_line
, __start___param
,
975 __stop___param
- __start___param
,
976 -1, -1, NULL
, &unknown_bootoption
);
977 print_unknown_bootoptions();
978 if (!IS_ERR_OR_NULL(after_dashes
))
979 parse_args("Setting init args", after_dashes
, NULL
, 0, -1, -1,
982 parse_args("Setting extra init args", extra_init_args
,
983 NULL
, 0, -1, -1, NULL
, set_init_arg
);
986 * These use large bootmem allocations and must precede
990 vfs_caches_init_early();
997 /* trace_printk can be enabled here */
1001 * Set up the scheduler prior starting any interrupts (such as the
1002 * timer interrupt). Full topology setup happens at smp_init()
1003 * time - but meanwhile we still have a functioning scheduler.
1007 if (WARN(!irqs_disabled(),
1008 "Interrupts were enabled *very* early, fixing it\n"))
1009 local_irq_disable();
1013 * Set up housekeeping before setting up workqueues to allow the unbound
1014 * workqueue to take non-housekeeping into account.
1016 housekeeping_init();
1019 * Allow workqueue creation and work item queueing/cancelling
1020 * early. Work item execution depends on kthreads and starts after
1023 workqueue_init_early();
1027 /* Trace events are available after this */
1031 initcall_debug_enable();
1033 context_tracking_init();
1034 /* init some links before init_ISA_irqs() */
1048 * For best initial stack canary entropy, prepare it after:
1049 * - setup_arch() for any UEFI RNG entropy and boot cmdline access
1050 * - timekeeping_init() for ktime entropy used in random_init()
1051 * - time_init() for making random_get_entropy() work on some platforms
1052 * - random_init() to initialize the RNG from from early entropy sources
1054 random_init(command_line
);
1055 boot_init_stack_canary();
1059 call_function_init();
1060 WARN(!irqs_disabled(), "Interrupts were enabled early\n");
1062 early_boot_irqs_disabled
= false;
1065 kmem_cache_init_late();
1068 * HACK ALERT! This is early. We're enabling the console before
1069 * we've done PCI setups etc, and console_init() must be aware of
1070 * this. But we do want output early, in case something goes wrong.
1074 panic("Too many boot %s vars at `%s'", panic_later
,
1080 * Need to run this when irqs are enabled, because it wants
1081 * to self-test [hard/soft]-irqs on/off lock inversion bugs
1087 * This needs to be called before any devices perform DMA
1088 * operations that might use the SWIOTLB bounce buffers. It will
1089 * mark the bounce buffers as decrypted so that their usage will
1090 * not cause "plain-text" data to be decrypted when accessed.
1094 #ifdef CONFIG_BLK_DEV_INITRD
1095 if (initrd_start
&& !initrd_below_start_ok
&&
1096 page_to_pfn(virt_to_page((void *)initrd_start
)) < min_low_pfn
) {
1097 pr_crit("initrd overwritten (0x%08lx < 0x%08lx) - disabling it.\n",
1098 page_to_pfn(virt_to_page((void *)initrd_start
)),
1103 setup_per_cpu_pageset();
1113 if (efi_enabled(EFI_RUNTIME_SERVICES
))
1114 efi_enter_virtual_mode();
1116 thread_stack_cache_init();
1133 taskstats_init_early();
1139 acpi_subsystem_init();
1140 arch_post_acpi_subsys_init();
1143 /* Do the rest non-__init'ed, we're now alive */
1144 arch_call_rest_init();
1146 prevent_tail_call_optimization();
1149 /* Call all constructor functions linked into the kernel. */
1150 static void __init
do_ctors(void)
1153 * For UML, the constructors have already been called by the
1154 * normal setup code as it's just a normal ELF binary, so we
1155 * cannot do it again - but we do need CONFIG_CONSTRUCTORS
1156 * even on UML for modules.
1158 #if defined(CONFIG_CONSTRUCTORS) && !defined(CONFIG_UML)
1159 ctor_fn_t
*fn
= (ctor_fn_t
*) __ctors_start
;
1161 for (; fn
< (ctor_fn_t
*) __ctors_end
; fn
++)
1166 #ifdef CONFIG_KALLSYMS
1167 struct blacklist_entry
{
1168 struct list_head next
;
1172 static __initdata_or_module
LIST_HEAD(blacklisted_initcalls
);
1174 static int __init
initcall_blacklist(char *str
)
1177 struct blacklist_entry
*entry
;
1179 /* str argument is a comma-separated list of functions */
1181 str_entry
= strsep(&str
, ",");
1183 pr_debug("blacklisting initcall %s\n", str_entry
);
1184 entry
= memblock_alloc(sizeof(*entry
),
1187 panic("%s: Failed to allocate %zu bytes\n",
1188 __func__
, sizeof(*entry
));
1189 entry
->buf
= memblock_alloc(strlen(str_entry
) + 1,
1192 panic("%s: Failed to allocate %zu bytes\n",
1193 __func__
, strlen(str_entry
) + 1);
1194 strcpy(entry
->buf
, str_entry
);
1195 list_add(&entry
->next
, &blacklisted_initcalls
);
1197 } while (str_entry
);
1202 static bool __init_or_module
initcall_blacklisted(initcall_t fn
)
1204 struct blacklist_entry
*entry
;
1205 char fn_name
[KSYM_SYMBOL_LEN
];
1208 if (list_empty(&blacklisted_initcalls
))
1211 addr
= (unsigned long) dereference_function_descriptor(fn
);
1212 sprint_symbol_no_offset(fn_name
, addr
);
1215 * fn will be "function_name [module_name]" where [module_name] is not
1216 * displayed for built-in init functions. Strip off the [module_name].
1218 strreplace(fn_name
, ' ', '\0');
1220 list_for_each_entry(entry
, &blacklisted_initcalls
, next
) {
1221 if (!strcmp(fn_name
, entry
->buf
)) {
1222 pr_debug("initcall %s blacklisted\n", fn_name
);
1230 static int __init
initcall_blacklist(char *str
)
1232 pr_warn("initcall_blacklist requires CONFIG_KALLSYMS\n");
1236 static bool __init_or_module
initcall_blacklisted(initcall_t fn
)
1241 __setup("initcall_blacklist=", initcall_blacklist
);
1243 static __init_or_module
void
1244 trace_initcall_start_cb(void *data
, initcall_t fn
)
1246 unsigned long *calltime
= (unsigned long *)data
;
1248 printk(KERN_DEBUG
"calling %pS @ %i\n", fn
, task_pid_nr(current
));
1249 *calltime
= local_clock();
1252 static __init_or_module
void
1253 trace_initcall_finish_cb(void *data
, initcall_t fn
, int ret
)
1255 unsigned long *calltime
= (unsigned long *)data
;
1256 unsigned long delta
, rettime
;
1257 unsigned long long duration
;
1259 rettime
= local_clock();
1260 delta
= rettime
- *calltime
;
1261 duration
= delta
>> 10;
1262 printk(KERN_DEBUG
"initcall %pS returned %d after %lld usecs\n",
1266 static ktime_t initcall_calltime
;
1268 #ifdef TRACEPOINTS_ENABLED
1269 static void __init
initcall_debug_enable(void)
1273 ret
= register_trace_initcall_start(trace_initcall_start_cb
,
1274 &initcall_calltime
);
1275 ret
|= register_trace_initcall_finish(trace_initcall_finish_cb
,
1276 &initcall_calltime
);
1277 WARN(ret
, "Failed to register initcall tracepoints\n");
1279 # define do_trace_initcall_start trace_initcall_start
1280 # define do_trace_initcall_finish trace_initcall_finish
1282 static inline void do_trace_initcall_start(initcall_t fn
)
1284 if (!initcall_debug
)
1286 trace_initcall_start_cb(&initcall_calltime
, fn
);
1288 static inline void do_trace_initcall_finish(initcall_t fn
, int ret
)
1290 if (!initcall_debug
)
1292 trace_initcall_finish_cb(&initcall_calltime
, fn
, ret
);
1294 #endif /* !TRACEPOINTS_ENABLED */
1296 int __init_or_module
do_one_initcall(initcall_t fn
)
1298 int count
= preempt_count();
1302 if (initcall_blacklisted(fn
))
1305 do_trace_initcall_start(fn
);
1307 do_trace_initcall_finish(fn
, ret
);
1311 if (preempt_count() != count
) {
1312 sprintf(msgbuf
, "preemption imbalance ");
1313 preempt_count_set(count
);
1315 if (irqs_disabled()) {
1316 strlcat(msgbuf
, "disabled interrupts ", sizeof(msgbuf
));
1319 WARN(msgbuf
[0], "initcall %pS returned with %s\n", fn
, msgbuf
);
1321 add_latent_entropy();
1326 extern initcall_entry_t __initcall_start
[];
1327 extern initcall_entry_t __initcall0_start
[];
1328 extern initcall_entry_t __initcall1_start
[];
1329 extern initcall_entry_t __initcall2_start
[];
1330 extern initcall_entry_t __initcall3_start
[];
1331 extern initcall_entry_t __initcall4_start
[];
1332 extern initcall_entry_t __initcall5_start
[];
1333 extern initcall_entry_t __initcall6_start
[];
1334 extern initcall_entry_t __initcall7_start
[];
1335 extern initcall_entry_t __initcall_end
[];
1337 static initcall_entry_t
*initcall_levels
[] __initdata
= {
1349 /* Keep these in sync with initcalls in include/linux/init.h */
1350 static const char *initcall_level_names
[] __initdata
= {
1361 static int __init
ignore_unknown_bootoption(char *param
, char *val
,
1362 const char *unused
, void *arg
)
1367 static void __init
do_initcall_level(int level
, char *command_line
)
1369 initcall_entry_t
*fn
;
1371 parse_args(initcall_level_names
[level
],
1372 command_line
, __start___param
,
1373 __stop___param
- __start___param
,
1375 NULL
, ignore_unknown_bootoption
);
1377 trace_initcall_level(initcall_level_names
[level
]);
1378 for (fn
= initcall_levels
[level
]; fn
< initcall_levels
[level
+1]; fn
++)
1379 do_one_initcall(initcall_from_entry(fn
));
1382 static void __init
do_initcalls(void)
1385 size_t len
= strlen(saved_command_line
) + 1;
1388 command_line
= kzalloc(len
, GFP_KERNEL
);
1390 panic("%s: Failed to allocate %zu bytes\n", __func__
, len
);
1392 for (level
= 0; level
< ARRAY_SIZE(initcall_levels
) - 1; level
++) {
1393 /* Parser modifies command_line, restore it each time */
1394 strcpy(command_line
, saved_command_line
);
1395 do_initcall_level(level
, command_line
);
1398 kfree(command_line
);
1402 * Ok, the machine is now initialized. None of the devices
1403 * have been touched yet, but the CPU subsystem is up and
1404 * running, and memory and process management works.
1406 * Now we can finally start doing some real work..
1408 static void __init
do_basic_setup(void)
1417 static void __init
do_pre_smp_initcalls(void)
1419 initcall_entry_t
*fn
;
1421 trace_initcall_level("early");
1422 for (fn
= __initcall_start
; fn
< __initcall0_start
; fn
++)
1423 do_one_initcall(initcall_from_entry(fn
));
1426 static int run_init_process(const char *init_filename
)
1428 const char *const *p
;
1430 argv_init
[0] = init_filename
;
1431 pr_info("Run %s as init process\n", init_filename
);
1432 pr_debug(" with arguments:\n");
1433 for (p
= argv_init
; *p
; p
++)
1434 pr_debug(" %s\n", *p
);
1435 pr_debug(" with environment:\n");
1436 for (p
= envp_init
; *p
; p
++)
1437 pr_debug(" %s\n", *p
);
1438 return kernel_execve(init_filename
, argv_init
, envp_init
);
1441 static int try_to_run_init_process(const char *init_filename
)
1445 ret
= run_init_process(init_filename
);
1447 if (ret
&& ret
!= -ENOENT
) {
1448 pr_err("Starting init: %s exists but couldn't execute it (error %d)\n",
1449 init_filename
, ret
);
1455 static noinline
void __init
kernel_init_freeable(void);
1457 #if defined(CONFIG_STRICT_KERNEL_RWX) || defined(CONFIG_STRICT_MODULE_RWX)
1458 bool rodata_enabled __ro_after_init
= true;
1459 static int __init
set_debug_rodata(char *str
)
1461 if (strtobool(str
, &rodata_enabled
))
1462 pr_warn("Invalid option string for rodata: '%s'\n", str
);
1465 __setup("rodata=", set_debug_rodata
);
1468 #ifdef CONFIG_STRICT_KERNEL_RWX
1469 static void mark_readonly(void)
1471 if (rodata_enabled
) {
1473 * load_module() results in W+X mappings, which are cleaned
1474 * up with call_rcu(). Let's make sure that queued work is
1475 * flushed so that we don't hit false positives looking for
1476 * insecure pages which are W+X.
1482 pr_info("Kernel memory protection disabled.\n");
1484 #elif defined(CONFIG_ARCH_HAS_STRICT_KERNEL_RWX)
1485 static inline void mark_readonly(void)
1487 pr_warn("Kernel memory protection not selected by kernel config.\n");
1490 static inline void mark_readonly(void)
1492 pr_warn("This architecture does not have kernel memory protection.\n");
1496 void __weak
free_initmem(void)
1498 free_initmem_default(POISON_FREE_INITMEM
);
1501 static int __ref
kernel_init(void *unused
)
1506 * Wait until kthreadd is all set-up.
1508 wait_for_completion(&kthreadd_done
);
1510 kernel_init_freeable();
1511 /* need to finish all async __init code before freeing the memory */
1512 async_synchronize_full();
1513 kprobe_free_init_mem();
1514 ftrace_free_init_mem();
1515 kgdb_free_init_mem();
1521 * Kernel mappings are now finalized - update the userspace page-table
1526 system_state
= SYSTEM_RUNNING
;
1527 numa_default_policy();
1529 rcu_end_inkernel_boot();
1533 if (ramdisk_execute_command
) {
1534 ret
= run_init_process(ramdisk_execute_command
);
1537 pr_err("Failed to execute %s (error %d)\n",
1538 ramdisk_execute_command
, ret
);
1542 * We try each of these until one succeeds.
1544 * The Bourne shell can be used instead of init if we are
1545 * trying to recover a really broken machine.
1547 if (execute_command
) {
1548 ret
= run_init_process(execute_command
);
1551 panic("Requested init %s failed (error %d).",
1552 execute_command
, ret
);
1555 if (CONFIG_DEFAULT_INIT
[0] != '\0') {
1556 ret
= run_init_process(CONFIG_DEFAULT_INIT
);
1558 pr_err("Default init %s failed (error %d)\n",
1559 CONFIG_DEFAULT_INIT
, ret
);
1564 if (!try_to_run_init_process("/sbin/init") ||
1565 !try_to_run_init_process("/etc/init") ||
1566 !try_to_run_init_process("/bin/init") ||
1567 !try_to_run_init_process("/bin/sh"))
1570 panic("No working init found. Try passing init= option to kernel. "
1571 "See Linux Documentation/admin-guide/init.rst for guidance.");
1574 /* Open /dev/console, for stdin/stdout/stderr, this should never fail */
1575 void __init
console_on_rootfs(void)
1577 struct file
*file
= filp_open("/dev/console", O_RDWR
, 0);
1580 pr_err("Warning: unable to open an initial console.\n");
1589 static noinline
void __init
kernel_init_freeable(void)
1591 /* Now the scheduler is fully set up and can do blocking allocations */
1592 gfp_allowed_mask
= __GFP_BITS_MASK
;
1595 * init can allocate pages on any node
1597 set_mems_allowed(node_states
[N_MEMORY
]);
1599 cad_pid
= get_pid(task_pid(current
));
1601 smp_prepare_cpus(setup_max_cpus
);
1605 init_mm_internals();
1607 rcu_init_tasks_generic();
1608 do_pre_smp_initcalls();
1609 lockup_detector_init();
1615 page_alloc_init_late();
1616 /* Initialize page ext after all struct pages are initialized. */
1621 kunit_run_all_tests();
1623 wait_for_initramfs();
1624 console_on_rootfs();
1627 * check if there is an early userspace init. If yes, let it do all
1630 if (init_eaccess(ramdisk_execute_command
) != 0) {
1631 ramdisk_execute_command
= NULL
;
1632 prepare_namespace();
1636 * Ok, we have completed the initial bootup, and
1637 * we're essentially up and running. Get rid of the
1638 * initmem segments and start the user-mode stuff..
1640 * rootfs is available now, try loading the public keys
1641 * and default modules
1644 integrity_load_keys();