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 size_t initargs_offs
;
158 # define bootconfig_found false
159 # define initargs_offs 0
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");
385 memblock_free_ptr(new_cmdline
, len
+ 1);
392 static int __init
bootconfig_params(char *param
, char *val
,
393 const char *unused
, void *arg
)
395 if (strcmp(param
, "bootconfig") == 0) {
396 bootconfig_found
= true;
401 static int __init
warn_bootconfig(char *str
)
403 /* The 'bootconfig' has been handled by bootconfig_params(). */
407 static void __init
setup_boot_config(void)
409 static char tmp_cmdline
[COMMAND_LINE_SIZE
] __initdata
;
413 char *data
, *copy
, *err
;
416 /* Cut out the bootconfig data even if we have no bootconfig option */
417 data
= get_boot_config_from_initrd(&size
, &csum
);
419 strlcpy(tmp_cmdline
, boot_command_line
, COMMAND_LINE_SIZE
);
420 err
= parse_args("bootconfig", tmp_cmdline
, NULL
, 0, 0, 0, NULL
,
423 if (IS_ERR(err
) || !bootconfig_found
)
426 /* parse_args() stops at the next param of '--' and returns an address */
428 initargs_offs
= err
- tmp_cmdline
;
431 pr_err("'bootconfig' found on command line, but no bootconfig found\n");
435 if (size
>= XBC_DATA_MAX
) {
436 pr_err("bootconfig size %d greater than max size %d\n",
441 if (xbc_calc_checksum(data
, size
) != csum
) {
442 pr_err("bootconfig checksum failed\n");
446 copy
= memblock_alloc(size
+ 1, SMP_CACHE_BYTES
);
448 pr_err("Failed to allocate memory for bootconfig\n");
452 memcpy(copy
, data
, size
);
455 ret
= xbc_init(copy
, &msg
, &pos
);
458 pr_err("Failed to init bootconfig: %s.\n", msg
);
460 pr_err("Failed to parse bootconfig: %s at %d.\n",
463 pr_info("Load bootconfig: %d bytes %d nodes\n", size
, ret
);
464 /* keys starting with "kernel." are passed via cmdline */
465 extra_command_line
= xbc_make_cmdline("kernel");
466 /* Also, "init." keys are init arguments */
467 extra_init_args
= xbc_make_cmdline("init");
472 static void __init
exit_boot_config(void)
477 #else /* !CONFIG_BOOT_CONFIG */
479 static void __init
setup_boot_config(void)
481 /* Remove bootconfig data from initrd */
482 get_boot_config_from_initrd(NULL
, NULL
);
485 static int __init
warn_bootconfig(char *str
)
487 pr_warn("WARNING: 'bootconfig' found on the kernel command line but CONFIG_BOOT_CONFIG is not set.\n");
491 #define exit_boot_config() do {} while (0)
493 #endif /* CONFIG_BOOT_CONFIG */
495 early_param("bootconfig", warn_bootconfig
);
497 /* Change NUL term back to "=", to make "param" the whole string. */
498 static void __init
repair_env_string(char *param
, char *val
)
501 /* param=val or param="val"? */
502 if (val
== param
+strlen(param
)+1)
504 else if (val
== param
+strlen(param
)+2) {
506 memmove(val
-1, val
, strlen(val
)+1);
512 /* Anything after -- gets handed straight to init. */
513 static int __init
set_init_arg(char *param
, char *val
,
514 const char *unused
, void *arg
)
521 repair_env_string(param
, val
);
523 for (i
= 0; argv_init
[i
]; i
++) {
524 if (i
== MAX_INIT_ARGS
) {
525 panic_later
= "init";
530 argv_init
[i
] = param
;
535 * Unknown boot options get handed to init, unless they look like
536 * unused parameters (modprobe will find them in /proc/cmdline).
538 static int __init
unknown_bootoption(char *param
, char *val
,
539 const char *unused
, void *arg
)
541 size_t len
= strlen(param
);
543 repair_env_string(param
, val
);
545 /* Handle obsolete-style parameters */
546 if (obsolete_checksetup(param
))
549 /* Unused module parameter. */
550 if (strnchr(param
, len
, '.'))
557 /* Environment option */
559 for (i
= 0; envp_init
[i
]; i
++) {
560 if (i
== MAX_INIT_ENVS
) {
564 if (!strncmp(param
, envp_init
[i
], len
+1))
567 envp_init
[i
] = param
;
569 /* Command line option */
571 for (i
= 0; argv_init
[i
]; i
++) {
572 if (i
== MAX_INIT_ARGS
) {
573 panic_later
= "init";
577 argv_init
[i
] = param
;
582 static int __init
init_setup(char *str
)
586 execute_command
= str
;
588 * In case LILO is going to boot us with default command line,
589 * it prepends "auto" before the whole cmdline which makes
590 * the shell think it should execute a script with such name.
591 * So we ignore all arguments entered _before_ init=... [MJ]
593 for (i
= 1; i
< MAX_INIT_ARGS
; i
++)
597 __setup("init=", init_setup
);
599 static int __init
rdinit_setup(char *str
)
603 ramdisk_execute_command
= str
;
604 /* See "auto" comment in init_setup */
605 for (i
= 1; i
< MAX_INIT_ARGS
; i
++)
609 __setup("rdinit=", rdinit_setup
);
612 static const unsigned int setup_max_cpus
= NR_CPUS
;
613 static inline void setup_nr_cpu_ids(void) { }
614 static inline void smp_prepare_cpus(unsigned int maxcpus
) { }
618 * We need to store the untouched command line for future reference.
619 * We also need to store the touched command line since the parameter
620 * parsing is performed in place, and we should allow a component to
621 * store reference of name/value for future reference.
623 static void __init
setup_command_line(char *command_line
)
625 size_t len
, xlen
= 0, ilen
= 0;
627 if (extra_command_line
)
628 xlen
= strlen(extra_command_line
);
630 ilen
= strlen(extra_init_args
) + 4; /* for " -- " */
632 len
= xlen
+ strlen(boot_command_line
) + 1;
634 saved_command_line
= memblock_alloc(len
+ ilen
, SMP_CACHE_BYTES
);
635 if (!saved_command_line
)
636 panic("%s: Failed to allocate %zu bytes\n", __func__
, len
+ ilen
);
638 static_command_line
= memblock_alloc(len
, SMP_CACHE_BYTES
);
639 if (!static_command_line
)
640 panic("%s: Failed to allocate %zu bytes\n", __func__
, len
);
644 * We have to put extra_command_line before boot command
645 * lines because there could be dashes (separator of init
646 * command line) in the command lines.
648 strcpy(saved_command_line
, extra_command_line
);
649 strcpy(static_command_line
, extra_command_line
);
651 strcpy(saved_command_line
+ xlen
, boot_command_line
);
652 strcpy(static_command_line
+ xlen
, command_line
);
656 * Append supplemental init boot args to saved_command_line
657 * so that user can check what command line options passed
659 * The order should always be
660 * " -- "[bootconfig init-param][cmdline init-param]
663 len
= xlen
+ initargs_offs
;
664 strcpy(saved_command_line
+ len
, extra_init_args
);
665 len
+= ilen
- 4; /* strlen(extra_init_args) */
666 strcpy(saved_command_line
+ len
,
667 boot_command_line
+ initargs_offs
- 1);
669 len
= strlen(saved_command_line
);
670 strcpy(saved_command_line
+ len
, " -- ");
672 strcpy(saved_command_line
+ len
, extra_init_args
);
678 * We need to finalize in a non-__init function or else race conditions
679 * between the root thread and the init thread may cause start_kernel to
680 * be reaped by free_initmem before the root thread has proceeded to
683 * gcc-3.4 accidentally inlines this function, so use noinline.
686 static __initdata
DECLARE_COMPLETION(kthreadd_done
);
688 noinline
void __ref
rest_init(void)
690 struct task_struct
*tsk
;
693 rcu_scheduler_starting();
695 * We need to spawn init first so that it obtains pid 1, however
696 * the init task will end up wanting to create kthreads, which, if
697 * we schedule it before we create kthreadd, will OOPS.
699 pid
= kernel_thread(kernel_init
, NULL
, CLONE_FS
);
701 * Pin init on the boot CPU. Task migration is not properly working
702 * until sched_init_smp() has been run. It will set the allowed
703 * CPUs for init to the non isolated CPUs.
706 tsk
= find_task_by_pid_ns(pid
, &init_pid_ns
);
707 tsk
->flags
|= PF_NO_SETAFFINITY
;
708 set_cpus_allowed_ptr(tsk
, cpumask_of(smp_processor_id()));
711 numa_default_policy();
712 pid
= kernel_thread(kthreadd
, NULL
, CLONE_FS
| CLONE_FILES
);
714 kthreadd_task
= find_task_by_pid_ns(pid
, &init_pid_ns
);
718 * Enable might_sleep() and smp_processor_id() checks.
719 * They cannot be enabled earlier because with CONFIG_PREEMPTION=y
720 * kernel_thread() would trigger might_sleep() splats. With
721 * CONFIG_PREEMPT_VOLUNTARY=y the init task might have scheduled
722 * already, but it's stuck on the kthreadd_done completion.
724 system_state
= SYSTEM_SCHEDULING
;
726 complete(&kthreadd_done
);
729 * The boot idle thread must execute schedule()
730 * at least once to get things moving:
732 schedule_preempt_disabled();
733 /* Call into cpu_idle with preempt disabled */
734 cpu_startup_entry(CPUHP_ONLINE
);
737 /* Check for early params. */
738 static int __init
do_early_param(char *param
, char *val
,
739 const char *unused
, void *arg
)
741 const struct obs_kernel_param
*p
;
743 for (p
= __setup_start
; p
< __setup_end
; p
++) {
744 if ((p
->early
&& parameq(param
, p
->str
)) ||
745 (strcmp(param
, "console") == 0 &&
746 strcmp(p
->str
, "earlycon") == 0)
748 if (p
->setup_func(val
) != 0)
749 pr_warn("Malformed early option '%s'\n", param
);
752 /* We accept everything at this stage. */
756 void __init
parse_early_options(char *cmdline
)
758 parse_args("early options", cmdline
, NULL
, 0, 0, 0, NULL
,
762 /* Arch code calls this early on, or if not, just before other parsing. */
763 void __init
parse_early_param(void)
765 static int done __initdata
;
766 static char tmp_cmdline
[COMMAND_LINE_SIZE
] __initdata
;
771 /* All fall through to do_early_param. */
772 strlcpy(tmp_cmdline
, boot_command_line
, COMMAND_LINE_SIZE
);
773 parse_early_options(tmp_cmdline
);
777 void __init __weak
arch_post_acpi_subsys_init(void) { }
779 void __init __weak
smp_setup_processor_id(void)
783 # if THREAD_SIZE >= PAGE_SIZE
784 void __init __weak
thread_stack_cache_init(void)
789 void __init __weak
mem_encrypt_init(void) { }
791 void __init __weak
poking_init(void) { }
793 void __init __weak
pgtable_cache_init(void) { }
795 void __init __weak
trap_init(void) { }
798 core_param(initcall_debug
, initcall_debug
, bool, 0644);
800 #ifdef TRACEPOINTS_ENABLED
801 static void __init
initcall_debug_enable(void);
803 static inline void initcall_debug_enable(void)
808 /* Report memory auto-initialization states for this boot. */
809 static void __init
report_meminit(void)
813 if (IS_ENABLED(CONFIG_INIT_STACK_ALL_PATTERN
))
814 stack
= "all(pattern)";
815 else if (IS_ENABLED(CONFIG_INIT_STACK_ALL_ZERO
))
817 else if (IS_ENABLED(CONFIG_GCC_PLUGIN_STRUCTLEAK_BYREF_ALL
))
818 stack
= "byref_all(zero)";
819 else if (IS_ENABLED(CONFIG_GCC_PLUGIN_STRUCTLEAK_BYREF
))
820 stack
= "byref(zero)";
821 else if (IS_ENABLED(CONFIG_GCC_PLUGIN_STRUCTLEAK_USER
))
822 stack
= "__user(zero)";
826 pr_info("mem auto-init: stack:%s, heap alloc:%s, heap free:%s\n",
827 stack
, want_init_on_alloc(GFP_KERNEL
) ? "on" : "off",
828 want_init_on_free() ? "on" : "off");
829 if (want_init_on_free())
830 pr_info("mem auto-init: clearing system memory may take some time...\n");
834 * Set up kernel memory allocators
836 static void __init
mm_init(void)
839 * page_ext requires contiguous pages,
840 * bigger than MAX_ORDER unless SPARSEMEM.
842 page_ext_init_flatmem();
843 init_mem_debugging_and_hardening();
848 mem_init_print_info();
849 /* page_owner must be initialized after buddy is ready */
850 page_ext_init_flatmem_late();
854 debug_objects_mem_init();
856 /* Should be run before the first non-init thread is created */
858 /* Should be run after espfix64 is set up. */
862 #ifdef CONFIG_HAVE_ARCH_RANDOMIZE_KSTACK_OFFSET
863 DEFINE_STATIC_KEY_MAYBE_RO(CONFIG_RANDOMIZE_KSTACK_OFFSET_DEFAULT
,
864 randomize_kstack_offset
);
865 DEFINE_PER_CPU(u32
, kstack_offset
);
867 static int __init
early_randomize_kstack_offset(char *buf
)
872 ret
= kstrtobool(buf
, &bool_result
);
877 static_branch_enable(&randomize_kstack_offset
);
879 static_branch_disable(&randomize_kstack_offset
);
882 early_param("randomize_kstack_offset", early_randomize_kstack_offset
);
885 void __init __weak
arch_call_rest_init(void)
890 static void __init
print_unknown_bootoptions(void)
892 char *unknown_options
;
894 const char *const *p
;
897 if (panic_later
|| (!argv_init
[1] && !envp_init
[2]))
901 * Determine how many options we have to print out, plus a space
904 len
= 1; /* null terminator */
905 for (p
= &argv_init
[1]; *p
; p
++) {
909 for (p
= &envp_init
[2]; *p
; p
++) {
914 unknown_options
= memblock_alloc(len
, SMP_CACHE_BYTES
);
915 if (!unknown_options
) {
916 pr_err("%s: Failed to allocate %zu bytes\n",
920 end
= unknown_options
;
922 for (p
= &argv_init
[1]; *p
; p
++)
923 end
+= sprintf(end
, " %s", *p
);
924 for (p
= &envp_init
[2]; *p
; p
++)
925 end
+= sprintf(end
, " %s", *p
);
927 /* Start at unknown_options[1] to skip the initial space */
928 pr_notice("Unknown kernel command line parameters \"%s\", will be passed to user space.\n",
929 &unknown_options
[1]);
930 memblock_free_ptr(unknown_options
, len
);
933 asmlinkage __visible
void __init __no_sanitize_address
start_kernel(void)
938 set_task_stack_end_magic(&init_task
);
939 smp_setup_processor_id();
940 debug_objects_early_init();
941 init_vmlinux_build_id();
946 early_boot_irqs_disabled
= true;
949 * Interrupts are still disabled. Do necessary setups, then
954 pr_notice("%s", linux_banner
);
955 early_security_init();
956 setup_arch(&command_line
);
958 setup_command_line(command_line
);
960 setup_per_cpu_areas();
961 smp_prepare_boot_cpu(); /* arch-specific boot-cpu hooks */
962 boot_cpu_hotplug_init();
964 build_all_zonelists(NULL
);
967 pr_notice("Kernel command line: %s\n", saved_command_line
);
968 /* parameters may set static keys */
971 after_dashes
= parse_args("Booting kernel",
972 static_command_line
, __start___param
,
973 __stop___param
- __start___param
,
974 -1, -1, NULL
, &unknown_bootoption
);
975 print_unknown_bootoptions();
976 if (!IS_ERR_OR_NULL(after_dashes
))
977 parse_args("Setting init args", after_dashes
, NULL
, 0, -1, -1,
980 parse_args("Setting extra init args", extra_init_args
,
981 NULL
, 0, -1, -1, NULL
, set_init_arg
);
984 * These use large bootmem allocations and must precede
988 vfs_caches_init_early();
995 /* trace_printk can be enabled here */
999 * Set up the scheduler prior starting any interrupts (such as the
1000 * timer interrupt). Full topology setup happens at smp_init()
1001 * time - but meanwhile we still have a functioning scheduler.
1005 if (WARN(!irqs_disabled(),
1006 "Interrupts were enabled *very* early, fixing it\n"))
1007 local_irq_disable();
1011 * Set up housekeeping before setting up workqueues to allow the unbound
1012 * workqueue to take non-housekeeping into account.
1014 housekeeping_init();
1017 * Allow workqueue creation and work item queueing/cancelling
1018 * early. Work item execution depends on kthreads and starts after
1021 workqueue_init_early();
1025 /* Trace events are available after this */
1029 initcall_debug_enable();
1031 context_tracking_init();
1032 /* init some links before init_ISA_irqs() */
1045 * For best initial stack canary entropy, prepare it after:
1046 * - setup_arch() for any UEFI RNG entropy and boot cmdline access
1047 * - timekeeping_init() for ktime entropy used in rand_initialize()
1048 * - rand_initialize() to get any arch-specific entropy like RDRAND
1049 * - add_latent_entropy() to get any latent entropy
1050 * - adding command line entropy
1053 add_latent_entropy();
1054 add_device_randomness(command_line
, strlen(command_line
));
1055 boot_init_stack_canary();
1060 call_function_init();
1061 WARN(!irqs_disabled(), "Interrupts were enabled early\n");
1063 early_boot_irqs_disabled
= false;
1066 kmem_cache_init_late();
1069 * HACK ALERT! This is early. We're enabling the console before
1070 * we've done PCI setups etc, and console_init() must be aware of
1071 * this. But we do want output early, in case something goes wrong.
1075 panic("Too many boot %s vars at `%s'", panic_later
,
1081 * Need to run this when irqs are enabled, because it wants
1082 * to self-test [hard/soft]-irqs on/off lock inversion bugs
1088 * This needs to be called before any devices perform DMA
1089 * operations that might use the SWIOTLB bounce buffers. It will
1090 * mark the bounce buffers as decrypted so that their usage will
1091 * not cause "plain-text" data to be decrypted when accessed.
1095 #ifdef CONFIG_BLK_DEV_INITRD
1096 if (initrd_start
&& !initrd_below_start_ok
&&
1097 page_to_pfn(virt_to_page((void *)initrd_start
)) < min_low_pfn
) {
1098 pr_crit("initrd overwritten (0x%08lx < 0x%08lx) - disabling it.\n",
1099 page_to_pfn(virt_to_page((void *)initrd_start
)),
1104 setup_per_cpu_pageset();
1114 if (efi_enabled(EFI_RUNTIME_SERVICES
))
1115 efi_enter_virtual_mode();
1117 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 return strtobool(str
, &rodata_enabled
);
1463 __setup("rodata=", set_debug_rodata
);
1466 #ifdef CONFIG_STRICT_KERNEL_RWX
1467 static void mark_readonly(void)
1469 if (rodata_enabled
) {
1471 * load_module() results in W+X mappings, which are cleaned
1472 * up with call_rcu(). Let's make sure that queued work is
1473 * flushed so that we don't hit false positives looking for
1474 * insecure pages which are W+X.
1480 pr_info("Kernel memory protection disabled.\n");
1482 #elif defined(CONFIG_ARCH_HAS_STRICT_KERNEL_RWX)
1483 static inline void mark_readonly(void)
1485 pr_warn("Kernel memory protection not selected by kernel config.\n");
1488 static inline void mark_readonly(void)
1490 pr_warn("This architecture does not have kernel memory protection.\n");
1494 void __weak
free_initmem(void)
1496 free_initmem_default(POISON_FREE_INITMEM
);
1499 static int __ref
kernel_init(void *unused
)
1504 * Wait until kthreadd is all set-up.
1506 wait_for_completion(&kthreadd_done
);
1508 kernel_init_freeable();
1509 /* need to finish all async __init code before freeing the memory */
1510 async_synchronize_full();
1511 kprobe_free_init_mem();
1512 ftrace_free_init_mem();
1513 kgdb_free_init_mem();
1519 * Kernel mappings are now finalized - update the userspace page-table
1524 system_state
= SYSTEM_RUNNING
;
1525 numa_default_policy();
1527 rcu_end_inkernel_boot();
1531 if (ramdisk_execute_command
) {
1532 ret
= run_init_process(ramdisk_execute_command
);
1535 pr_err("Failed to execute %s (error %d)\n",
1536 ramdisk_execute_command
, ret
);
1540 * We try each of these until one succeeds.
1542 * The Bourne shell can be used instead of init if we are
1543 * trying to recover a really broken machine.
1545 if (execute_command
) {
1546 ret
= run_init_process(execute_command
);
1549 panic("Requested init %s failed (error %d).",
1550 execute_command
, ret
);
1553 if (CONFIG_DEFAULT_INIT
[0] != '\0') {
1554 ret
= run_init_process(CONFIG_DEFAULT_INIT
);
1556 pr_err("Default init %s failed (error %d)\n",
1557 CONFIG_DEFAULT_INIT
, ret
);
1562 if (!try_to_run_init_process("/sbin/init") ||
1563 !try_to_run_init_process("/etc/init") ||
1564 !try_to_run_init_process("/bin/init") ||
1565 !try_to_run_init_process("/bin/sh"))
1568 panic("No working init found. Try passing init= option to kernel. "
1569 "See Linux Documentation/admin-guide/init.rst for guidance.");
1572 /* Open /dev/console, for stdin/stdout/stderr, this should never fail */
1573 void __init
console_on_rootfs(void)
1575 struct file
*file
= filp_open("/dev/console", O_RDWR
, 0);
1578 pr_err("Warning: unable to open an initial console.\n");
1587 static noinline
void __init
kernel_init_freeable(void)
1589 /* Now the scheduler is fully set up and can do blocking allocations */
1590 gfp_allowed_mask
= __GFP_BITS_MASK
;
1593 * init can allocate pages on any node
1595 set_mems_allowed(node_states
[N_MEMORY
]);
1597 cad_pid
= get_pid(task_pid(current
));
1599 smp_prepare_cpus(setup_max_cpus
);
1603 init_mm_internals();
1605 rcu_init_tasks_generic();
1606 do_pre_smp_initcalls();
1607 lockup_detector_init();
1613 page_alloc_init_late();
1614 /* Initialize page ext after all struct pages are initialized. */
1619 kunit_run_all_tests();
1621 wait_for_initramfs();
1622 console_on_rootfs();
1625 * check if there is an early userspace init. If yes, let it do all
1628 if (init_eaccess(ramdisk_execute_command
) != 0) {
1629 ramdisk_execute_command
= NULL
;
1630 prepare_namespace();
1634 * Ok, we have completed the initial bootup, and
1635 * we're essentially up and running. Get rid of the
1636 * initmem segments and start the user-mode stuff..
1638 * rootfs is available now, try loading the public keys
1639 * and default modules
1642 integrity_load_keys();