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/rcupdate.h>
44 #include <linux/moduleparam.h>
45 #include <linux/kallsyms.h>
46 #include <linux/writeback.h>
47 #include <linux/cpu.h>
48 #include <linux/cpuset.h>
49 #include <linux/cgroup.h>
50 #include <linux/efi.h>
51 #include <linux/tick.h>
52 #include <linux/sched/isolation.h>
53 #include <linux/interrupt.h>
54 #include <linux/taskstats_kern.h>
55 #include <linux/delayacct.h>
56 #include <linux/unistd.h>
57 #include <linux/utsname.h>
58 #include <linux/rmap.h>
59 #include <linux/mempolicy.h>
60 #include <linux/key.h>
61 #include <linux/buffer_head.h>
62 #include <linux/page_ext.h>
63 #include <linux/debug_locks.h>
64 #include <linux/debugobjects.h>
65 #include <linux/lockdep.h>
66 #include <linux/kmemleak.h>
67 #include <linux/padata.h>
68 #include <linux/pid_namespace.h>
69 #include <linux/device/driver.h>
70 #include <linux/kthread.h>
71 #include <linux/sched.h>
72 #include <linux/sched/init.h>
73 #include <linux/signal.h>
74 #include <linux/idr.h>
75 #include <linux/kgdb.h>
76 #include <linux/ftrace.h>
77 #include <linux/async.h>
78 #include <linux/sfi.h>
79 #include <linux/shmem_fs.h>
80 #include <linux/slab.h>
81 #include <linux/perf_event.h>
82 #include <linux/ptrace.h>
83 #include <linux/pti.h>
84 #include <linux/blkdev.h>
85 #include <linux/elevator.h>
86 #include <linux/sched/clock.h>
87 #include <linux/sched/task.h>
88 #include <linux/sched/task_stack.h>
89 #include <linux/context_tracking.h>
90 #include <linux/random.h>
91 #include <linux/list.h>
92 #include <linux/integrity.h>
93 #include <linux/proc_ns.h>
95 #include <linux/cache.h>
96 #include <linux/rodata_test.h>
97 #include <linux/jump_label.h>
98 #include <linux/mem_encrypt.h>
99 #include <linux/kcsan.h>
100 #include <linux/init_syscalls.h>
103 #include <asm/bugs.h>
104 #include <asm/setup.h>
105 #include <asm/sections.h>
106 #include <asm/cacheflush.h>
108 #define CREATE_TRACE_POINTS
109 #include <trace/events/initcall.h>
111 static int kernel_init(void *);
113 extern void init_IRQ(void);
114 extern void radix_tree_init(void);
117 * Debug helper: via this flag we know that we are in 'early bootup code'
118 * where only the boot processor is running with IRQ disabled. This means
119 * two things - IRQ must not be enabled before the flag is cleared and some
120 * operations which are not allowed with IRQ disabled are allowed while the
123 bool early_boot_irqs_disabled __read_mostly
;
125 enum system_states system_state __read_mostly
;
126 EXPORT_SYMBOL(system_state
);
129 * Boot command-line arguments
131 #define MAX_INIT_ARGS CONFIG_INIT_ENV_ARG_LIMIT
132 #define MAX_INIT_ENVS CONFIG_INIT_ENV_ARG_LIMIT
134 extern void time_init(void);
135 /* Default late time init is NULL. archs can override this later. */
136 void (*__initdata late_time_init
)(void);
138 /* Untouched command line saved by arch-specific code. */
139 char __initdata boot_command_line
[COMMAND_LINE_SIZE
];
140 /* Untouched saved command line (eg. for /proc) */
141 char *saved_command_line
;
142 /* Command line for parameter parsing */
143 static char *static_command_line
;
144 /* Untouched extra command line */
145 static char *extra_command_line
;
146 /* Extra init arguments */
147 static char *extra_init_args
;
149 #ifdef CONFIG_BOOT_CONFIG
150 /* Is bootconfig on command line? */
151 static bool bootconfig_found
;
152 static bool initargs_found
;
154 # define bootconfig_found false
155 # define initargs_found false
158 static char *execute_command
;
159 static char *ramdisk_execute_command
= "/init";
162 * Used to generate warnings if static_key manipulation functions are used
163 * before jump_label_init is called.
165 bool static_key_initialized __read_mostly
;
166 EXPORT_SYMBOL_GPL(static_key_initialized
);
169 * If set, this is an indication to the drivers that reset the underlying
170 * device before going ahead with the initialization otherwise driver might
171 * rely on the BIOS and skip the reset operation.
173 * This is useful if kernel is booting in an unreliable environment.
174 * For ex. kdump situation where previous kernel has crashed, BIOS has been
175 * skipped and devices will be in unknown state.
177 unsigned int reset_devices
;
178 EXPORT_SYMBOL(reset_devices
);
180 static int __init
set_reset_devices(char *str
)
186 __setup("reset_devices", set_reset_devices
);
188 static const char *argv_init
[MAX_INIT_ARGS
+2] = { "init", NULL
, };
189 const char *envp_init
[MAX_INIT_ENVS
+2] = { "HOME=/", "TERM=linux", NULL
, };
190 static const char *panic_later
, *panic_param
;
192 extern const struct obs_kernel_param __setup_start
[], __setup_end
[];
194 static bool __init
obsolete_checksetup(char *line
)
196 const struct obs_kernel_param
*p
;
197 bool had_early_param
= false;
201 int n
= strlen(p
->str
);
202 if (parameqn(line
, p
->str
, n
)) {
204 /* Already done in parse_early_param?
205 * (Needs exact match on param part).
206 * Keep iterating, as we can have early
207 * params and __setups of same names 8( */
208 if (line
[n
] == '\0' || line
[n
] == '=')
209 had_early_param
= true;
210 } else if (!p
->setup_func
) {
211 pr_warn("Parameter %s is obsolete, ignored\n",
214 } else if (p
->setup_func(line
+ n
))
218 } while (p
< __setup_end
);
220 return had_early_param
;
224 * This should be approx 2 Bo*oMips to start (note initial shift), and will
225 * still work even if initially too large, it will just take slightly longer
227 unsigned long loops_per_jiffy
= (1<<12);
228 EXPORT_SYMBOL(loops_per_jiffy
);
230 static int __init
debug_kernel(char *str
)
232 console_loglevel
= CONSOLE_LOGLEVEL_DEBUG
;
236 static int __init
quiet_kernel(char *str
)
238 console_loglevel
= CONSOLE_LOGLEVEL_QUIET
;
242 early_param("debug", debug_kernel
);
243 early_param("quiet", quiet_kernel
);
245 static int __init
loglevel(char *str
)
250 * Only update loglevel value when a correct setting was passed,
251 * to prevent blind crashes (when loglevel being set to 0) that
252 * are quite hard to debug
254 if (get_option(&str
, &newlevel
)) {
255 console_loglevel
= newlevel
;
262 early_param("loglevel", loglevel
);
264 #ifdef CONFIG_BLK_DEV_INITRD
265 static void * __init
get_boot_config_from_initrd(u32
*_size
, u32
*_csum
)
274 data
= (char *)initrd_end
- BOOTCONFIG_MAGIC_LEN
;
275 if (memcmp(data
, BOOTCONFIG_MAGIC
, BOOTCONFIG_MAGIC_LEN
))
278 hdr
= (u32
*)(data
- 8);
282 data
= ((void *)hdr
) - size
;
283 if ((unsigned long)data
< initrd_start
) {
284 pr_err("bootconfig size %d is greater than initrd size %ld\n",
285 size
, initrd_end
- initrd_start
);
289 /* Remove bootconfig from initramfs/initrd */
290 initrd_end
= (unsigned long)data
;
299 static void * __init
get_boot_config_from_initrd(u32
*_size
, u32
*_csum
)
305 #ifdef CONFIG_BOOT_CONFIG
307 static char xbc_namebuf
[XBC_KEYLEN_MAX
] __initdata
;
309 #define rest(dst, end) ((end) > (dst) ? (end) - (dst) : 0)
311 static int __init
xbc_snprint_cmdline(char *buf
, size_t size
,
312 struct xbc_node
*root
)
314 struct xbc_node
*knode
, *vnode
;
315 char *end
= buf
+ size
;
319 xbc_node_for_each_key_value(root
, knode
, val
) {
320 ret
= xbc_node_compose_key_after(root
, knode
,
321 xbc_namebuf
, XBC_KEYLEN_MAX
);
325 vnode
= xbc_node_get_child(knode
);
327 ret
= snprintf(buf
, rest(buf
, end
), "%s ", xbc_namebuf
);
333 xbc_array_for_each_value(vnode
, val
) {
334 ret
= snprintf(buf
, rest(buf
, end
), "%s=\"%s\" ",
342 return buf
- (end
- size
);
346 /* Make an extra command line under given key word */
347 static char * __init
xbc_make_cmdline(const char *key
)
349 struct xbc_node
*root
;
353 root
= xbc_find_node(key
);
357 /* Count required buffer size */
358 len
= xbc_snprint_cmdline(NULL
, 0, root
);
362 new_cmdline
= memblock_alloc(len
+ 1, SMP_CACHE_BYTES
);
364 pr_err("Failed to allocate memory for extra kernel cmdline.\n");
368 ret
= xbc_snprint_cmdline(new_cmdline
, len
+ 1, root
);
369 if (ret
< 0 || ret
> len
) {
370 pr_err("Failed to print extra kernel cmdline.\n");
377 static u32
boot_config_checksum(unsigned char *p
, u32 size
)
387 static int __init
bootconfig_params(char *param
, char *val
,
388 const char *unused
, void *arg
)
390 if (strcmp(param
, "bootconfig") == 0) {
391 bootconfig_found
= true;
396 static void __init
setup_boot_config(const char *cmdline
)
398 static char tmp_cmdline
[COMMAND_LINE_SIZE
] __initdata
;
402 char *data
, *copy
, *err
;
405 /* Cut out the bootconfig data even if we have no bootconfig option */
406 data
= get_boot_config_from_initrd(&size
, &csum
);
408 strlcpy(tmp_cmdline
, boot_command_line
, COMMAND_LINE_SIZE
);
409 err
= parse_args("bootconfig", tmp_cmdline
, NULL
, 0, 0, 0, NULL
,
412 if (IS_ERR(err
) || !bootconfig_found
)
415 /* parse_args() stops at '--' and returns an address */
417 initargs_found
= true;
420 pr_err("'bootconfig' found on command line, but no bootconfig found\n");
424 if (size
>= XBC_DATA_MAX
) {
425 pr_err("bootconfig size %d greater than max size %d\n",
430 if (boot_config_checksum((unsigned char *)data
, size
) != csum
) {
431 pr_err("bootconfig checksum failed\n");
435 copy
= memblock_alloc(size
+ 1, SMP_CACHE_BYTES
);
437 pr_err("Failed to allocate memory for bootconfig\n");
441 memcpy(copy
, data
, size
);
444 ret
= xbc_init(copy
, &msg
, &pos
);
447 pr_err("Failed to init bootconfig: %s.\n", msg
);
449 pr_err("Failed to parse bootconfig: %s at %d.\n",
452 pr_info("Load bootconfig: %d bytes %d nodes\n", size
, ret
);
453 /* keys starting with "kernel." are passed via cmdline */
454 extra_command_line
= xbc_make_cmdline("kernel");
455 /* Also, "init." keys are init arguments */
456 extra_init_args
= xbc_make_cmdline("init");
463 static void __init
setup_boot_config(const char *cmdline
)
465 /* Remove bootconfig data from initrd */
466 get_boot_config_from_initrd(NULL
, NULL
);
469 static int __init
warn_bootconfig(char *str
)
471 pr_warn("WARNING: 'bootconfig' found on the kernel command line but CONFIG_BOOTCONFIG is not set.\n");
474 early_param("bootconfig", warn_bootconfig
);
478 /* Change NUL term back to "=", to make "param" the whole string. */
479 static void __init
repair_env_string(char *param
, char *val
)
482 /* param=val or param="val"? */
483 if (val
== param
+strlen(param
)+1)
485 else if (val
== param
+strlen(param
)+2) {
487 memmove(val
-1, val
, strlen(val
)+1);
493 /* Anything after -- gets handed straight to init. */
494 static int __init
set_init_arg(char *param
, char *val
,
495 const char *unused
, void *arg
)
502 repair_env_string(param
, val
);
504 for (i
= 0; argv_init
[i
]; i
++) {
505 if (i
== MAX_INIT_ARGS
) {
506 panic_later
= "init";
511 argv_init
[i
] = param
;
516 * Unknown boot options get handed to init, unless they look like
517 * unused parameters (modprobe will find them in /proc/cmdline).
519 static int __init
unknown_bootoption(char *param
, char *val
,
520 const char *unused
, void *arg
)
522 size_t len
= strlen(param
);
524 repair_env_string(param
, val
);
526 /* Handle obsolete-style parameters */
527 if (obsolete_checksetup(param
))
530 /* Unused module parameter. */
531 if (strnchr(param
, len
, '.'))
538 /* Environment option */
540 for (i
= 0; envp_init
[i
]; i
++) {
541 if (i
== MAX_INIT_ENVS
) {
545 if (!strncmp(param
, envp_init
[i
], len
+1))
548 envp_init
[i
] = param
;
550 /* Command line option */
552 for (i
= 0; argv_init
[i
]; i
++) {
553 if (i
== MAX_INIT_ARGS
) {
554 panic_later
= "init";
558 argv_init
[i
] = param
;
563 static int __init
init_setup(char *str
)
567 execute_command
= str
;
569 * In case LILO is going to boot us with default command line,
570 * it prepends "auto" before the whole cmdline which makes
571 * the shell think it should execute a script with such name.
572 * So we ignore all arguments entered _before_ init=... [MJ]
574 for (i
= 1; i
< MAX_INIT_ARGS
; i
++)
578 __setup("init=", init_setup
);
580 static int __init
rdinit_setup(char *str
)
584 ramdisk_execute_command
= str
;
585 /* See "auto" comment in init_setup */
586 for (i
= 1; i
< MAX_INIT_ARGS
; i
++)
590 __setup("rdinit=", rdinit_setup
);
593 static const unsigned int setup_max_cpus
= NR_CPUS
;
594 static inline void setup_nr_cpu_ids(void) { }
595 static inline void smp_prepare_cpus(unsigned int maxcpus
) { }
599 * We need to store the untouched command line for future reference.
600 * We also need to store the touched command line since the parameter
601 * parsing is performed in place, and we should allow a component to
602 * store reference of name/value for future reference.
604 static void __init
setup_command_line(char *command_line
)
606 size_t len
, xlen
= 0, ilen
= 0;
608 if (extra_command_line
)
609 xlen
= strlen(extra_command_line
);
611 ilen
= strlen(extra_init_args
) + 4; /* for " -- " */
613 len
= xlen
+ strlen(boot_command_line
) + 1;
615 saved_command_line
= memblock_alloc(len
+ ilen
, SMP_CACHE_BYTES
);
616 if (!saved_command_line
)
617 panic("%s: Failed to allocate %zu bytes\n", __func__
, len
+ ilen
);
619 static_command_line
= memblock_alloc(len
, SMP_CACHE_BYTES
);
620 if (!static_command_line
)
621 panic("%s: Failed to allocate %zu bytes\n", __func__
, len
);
625 * We have to put extra_command_line before boot command
626 * lines because there could be dashes (separator of init
627 * command line) in the command lines.
629 strcpy(saved_command_line
, extra_command_line
);
630 strcpy(static_command_line
, extra_command_line
);
632 strcpy(saved_command_line
+ xlen
, boot_command_line
);
633 strcpy(static_command_line
+ xlen
, command_line
);
637 * Append supplemental init boot args to saved_command_line
638 * so that user can check what command line options passed
641 len
= strlen(saved_command_line
);
642 if (initargs_found
) {
643 saved_command_line
[len
++] = ' ';
645 strcpy(saved_command_line
+ len
, " -- ");
649 strcpy(saved_command_line
+ len
, extra_init_args
);
654 * We need to finalize in a non-__init function or else race conditions
655 * between the root thread and the init thread may cause start_kernel to
656 * be reaped by free_initmem before the root thread has proceeded to
659 * gcc-3.4 accidentally inlines this function, so use noinline.
662 static __initdata
DECLARE_COMPLETION(kthreadd_done
);
664 noinline
void __ref
rest_init(void)
666 struct task_struct
*tsk
;
669 rcu_scheduler_starting();
671 * We need to spawn init first so that it obtains pid 1, however
672 * the init task will end up wanting to create kthreads, which, if
673 * we schedule it before we create kthreadd, will OOPS.
675 pid
= kernel_thread(kernel_init
, NULL
, CLONE_FS
);
677 * Pin init on the boot CPU. Task migration is not properly working
678 * until sched_init_smp() has been run. It will set the allowed
679 * CPUs for init to the non isolated CPUs.
682 tsk
= find_task_by_pid_ns(pid
, &init_pid_ns
);
683 set_cpus_allowed_ptr(tsk
, cpumask_of(smp_processor_id()));
686 numa_default_policy();
687 pid
= kernel_thread(kthreadd
, NULL
, CLONE_FS
| CLONE_FILES
);
689 kthreadd_task
= find_task_by_pid_ns(pid
, &init_pid_ns
);
693 * Enable might_sleep() and smp_processor_id() checks.
694 * They cannot be enabled earlier because with CONFIG_PREEMPTION=y
695 * kernel_thread() would trigger might_sleep() splats. With
696 * CONFIG_PREEMPT_VOLUNTARY=y the init task might have scheduled
697 * already, but it's stuck on the kthreadd_done completion.
699 system_state
= SYSTEM_SCHEDULING
;
701 complete(&kthreadd_done
);
704 * The boot idle thread must execute schedule()
705 * at least once to get things moving:
707 schedule_preempt_disabled();
708 /* Call into cpu_idle with preempt disabled */
709 cpu_startup_entry(CPUHP_ONLINE
);
712 /* Check for early params. */
713 static int __init
do_early_param(char *param
, char *val
,
714 const char *unused
, void *arg
)
716 const struct obs_kernel_param
*p
;
718 for (p
= __setup_start
; p
< __setup_end
; p
++) {
719 if ((p
->early
&& parameq(param
, p
->str
)) ||
720 (strcmp(param
, "console") == 0 &&
721 strcmp(p
->str
, "earlycon") == 0)
723 if (p
->setup_func(val
) != 0)
724 pr_warn("Malformed early option '%s'\n", param
);
727 /* We accept everything at this stage. */
731 void __init
parse_early_options(char *cmdline
)
733 parse_args("early options", cmdline
, NULL
, 0, 0, 0, NULL
,
737 /* Arch code calls this early on, or if not, just before other parsing. */
738 void __init
parse_early_param(void)
740 static int done __initdata
;
741 static char tmp_cmdline
[COMMAND_LINE_SIZE
] __initdata
;
746 /* All fall through to do_early_param. */
747 strlcpy(tmp_cmdline
, boot_command_line
, COMMAND_LINE_SIZE
);
748 parse_early_options(tmp_cmdline
);
752 void __init __weak
arch_post_acpi_subsys_init(void) { }
754 void __init __weak
smp_setup_processor_id(void)
758 # if THREAD_SIZE >= PAGE_SIZE
759 void __init __weak
thread_stack_cache_init(void)
764 void __init __weak
mem_encrypt_init(void) { }
766 void __init __weak
poking_init(void) { }
768 void __init __weak
pgtable_cache_init(void) { }
771 core_param(initcall_debug
, initcall_debug
, bool, 0644);
773 #ifdef TRACEPOINTS_ENABLED
774 static void __init
initcall_debug_enable(void);
776 static inline void initcall_debug_enable(void)
781 /* Report memory auto-initialization states for this boot. */
782 static void __init
report_meminit(void)
786 if (IS_ENABLED(CONFIG_INIT_STACK_ALL_PATTERN
))
787 stack
= "all(pattern)";
788 else if (IS_ENABLED(CONFIG_INIT_STACK_ALL_ZERO
))
790 else if (IS_ENABLED(CONFIG_GCC_PLUGIN_STRUCTLEAK_BYREF_ALL
))
791 stack
= "byref_all(zero)";
792 else if (IS_ENABLED(CONFIG_GCC_PLUGIN_STRUCTLEAK_BYREF
))
793 stack
= "byref(zero)";
794 else if (IS_ENABLED(CONFIG_GCC_PLUGIN_STRUCTLEAK_USER
))
795 stack
= "__user(zero)";
799 pr_info("mem auto-init: stack:%s, heap alloc:%s, heap free:%s\n",
800 stack
, want_init_on_alloc(GFP_KERNEL
) ? "on" : "off",
801 want_init_on_free() ? "on" : "off");
802 if (want_init_on_free())
803 pr_info("mem auto-init: clearing system memory may take some time...\n");
807 * Set up kernel memory allocators
809 static void __init
mm_init(void)
812 * page_ext requires contiguous pages,
813 * bigger than MAX_ORDER unless SPARSEMEM.
815 page_ext_init_flatmem();
816 init_debug_pagealloc();
822 debug_objects_mem_init();
825 /* Should be run before the first non-init thread is created */
827 /* Should be run after espfix64 is set up. */
831 void __init __weak
arch_call_rest_init(void)
836 asmlinkage __visible
void __init __no_sanitize_address
start_kernel(void)
841 set_task_stack_end_magic(&init_task
);
842 smp_setup_processor_id();
843 debug_objects_early_init();
848 early_boot_irqs_disabled
= true;
851 * Interrupts are still disabled. Do necessary setups, then
856 pr_notice("%s", linux_banner
);
857 early_security_init();
858 setup_arch(&command_line
);
859 setup_boot_config(command_line
);
860 setup_command_line(command_line
);
862 setup_per_cpu_areas();
863 smp_prepare_boot_cpu(); /* arch-specific boot-cpu hooks */
864 boot_cpu_hotplug_init();
866 build_all_zonelists(NULL
);
869 pr_notice("Kernel command line: %s\n", saved_command_line
);
870 /* parameters may set static keys */
873 after_dashes
= parse_args("Booting kernel",
874 static_command_line
, __start___param
,
875 __stop___param
- __start___param
,
876 -1, -1, NULL
, &unknown_bootoption
);
877 if (!IS_ERR_OR_NULL(after_dashes
))
878 parse_args("Setting init args", after_dashes
, NULL
, 0, -1, -1,
881 parse_args("Setting extra init args", extra_init_args
,
882 NULL
, 0, -1, -1, NULL
, set_init_arg
);
885 * These use large bootmem allocations and must precede
889 vfs_caches_init_early();
896 /* trace_printk can be enabled here */
900 * Set up the scheduler prior starting any interrupts (such as the
901 * timer interrupt). Full topology setup happens at smp_init()
902 * time - but meanwhile we still have a functioning scheduler.
906 * Disable preemption - early bootup scheduling is extremely
907 * fragile until we cpu_idle() for the first time.
910 if (WARN(!irqs_disabled(),
911 "Interrupts were enabled *very* early, fixing it\n"))
916 * Set up housekeeping before setting up workqueues to allow the unbound
917 * workqueue to take non-housekeeping into account.
922 * Allow workqueue creation and work item queueing/cancelling
923 * early. Work item execution depends on kthreads and starts after
926 workqueue_init_early();
930 /* Trace events are available after this */
934 initcall_debug_enable();
936 context_tracking_init();
937 /* init some links before init_ISA_irqs() */
948 * For best initial stack canary entropy, prepare it after:
949 * - setup_arch() for any UEFI RNG entropy and boot cmdline access
950 * - timekeeping_init() for ktime entropy used in rand_initialize()
951 * - rand_initialize() to get any arch-specific entropy like RDRAND
952 * - add_latent_entropy() to get any latent entropy
953 * - adding command line entropy
956 add_latent_entropy();
957 add_device_randomness(command_line
, strlen(command_line
));
958 boot_init_stack_canary();
963 call_function_init();
964 WARN(!irqs_disabled(), "Interrupts were enabled early\n");
966 early_boot_irqs_disabled
= false;
969 kmem_cache_init_late();
972 * HACK ALERT! This is early. We're enabling the console before
973 * we've done PCI setups etc, and console_init() must be aware of
974 * this. But we do want output early, in case something goes wrong.
978 panic("Too many boot %s vars at `%s'", panic_later
,
984 * Need to run this when irqs are enabled, because it wants
985 * to self-test [hard/soft]-irqs on/off lock inversion bugs
991 * This needs to be called before any devices perform DMA
992 * operations that might use the SWIOTLB bounce buffers. It will
993 * mark the bounce buffers as decrypted so that their usage will
994 * not cause "plain-text" data to be decrypted when accessed.
998 #ifdef CONFIG_BLK_DEV_INITRD
999 if (initrd_start
&& !initrd_below_start_ok
&&
1000 page_to_pfn(virt_to_page((void *)initrd_start
)) < min_low_pfn
) {
1001 pr_crit("initrd overwritten (0x%08lx < 0x%08lx) - disabling it.\n",
1002 page_to_pfn(virt_to_page((void *)initrd_start
)),
1007 setup_per_cpu_pageset();
1017 if (efi_enabled(EFI_RUNTIME_SERVICES
))
1018 efi_enter_virtual_mode();
1020 thread_stack_cache_init();
1037 taskstats_init_early();
1043 acpi_subsystem_init();
1044 arch_post_acpi_subsys_init();
1048 /* Do the rest non-__init'ed, we're now alive */
1049 arch_call_rest_init();
1051 prevent_tail_call_optimization();
1054 /* Call all constructor functions linked into the kernel. */
1055 static void __init
do_ctors(void)
1057 #ifdef CONFIG_CONSTRUCTORS
1058 ctor_fn_t
*fn
= (ctor_fn_t
*) __ctors_start
;
1060 for (; fn
< (ctor_fn_t
*) __ctors_end
; fn
++)
1065 #ifdef CONFIG_KALLSYMS
1066 struct blacklist_entry
{
1067 struct list_head next
;
1071 static __initdata_or_module
LIST_HEAD(blacklisted_initcalls
);
1073 static int __init
initcall_blacklist(char *str
)
1076 struct blacklist_entry
*entry
;
1078 /* str argument is a comma-separated list of functions */
1080 str_entry
= strsep(&str
, ",");
1082 pr_debug("blacklisting initcall %s\n", str_entry
);
1083 entry
= memblock_alloc(sizeof(*entry
),
1086 panic("%s: Failed to allocate %zu bytes\n",
1087 __func__
, sizeof(*entry
));
1088 entry
->buf
= memblock_alloc(strlen(str_entry
) + 1,
1091 panic("%s: Failed to allocate %zu bytes\n",
1092 __func__
, strlen(str_entry
) + 1);
1093 strcpy(entry
->buf
, str_entry
);
1094 list_add(&entry
->next
, &blacklisted_initcalls
);
1096 } while (str_entry
);
1101 static bool __init_or_module
initcall_blacklisted(initcall_t fn
)
1103 struct blacklist_entry
*entry
;
1104 char fn_name
[KSYM_SYMBOL_LEN
];
1107 if (list_empty(&blacklisted_initcalls
))
1110 addr
= (unsigned long) dereference_function_descriptor(fn
);
1111 sprint_symbol_no_offset(fn_name
, addr
);
1114 * fn will be "function_name [module_name]" where [module_name] is not
1115 * displayed for built-in init functions. Strip off the [module_name].
1117 strreplace(fn_name
, ' ', '\0');
1119 list_for_each_entry(entry
, &blacklisted_initcalls
, next
) {
1120 if (!strcmp(fn_name
, entry
->buf
)) {
1121 pr_debug("initcall %s blacklisted\n", fn_name
);
1129 static int __init
initcall_blacklist(char *str
)
1131 pr_warn("initcall_blacklist requires CONFIG_KALLSYMS\n");
1135 static bool __init_or_module
initcall_blacklisted(initcall_t fn
)
1140 __setup("initcall_blacklist=", initcall_blacklist
);
1142 static __init_or_module
void
1143 trace_initcall_start_cb(void *data
, initcall_t fn
)
1145 ktime_t
*calltime
= (ktime_t
*)data
;
1147 printk(KERN_DEBUG
"calling %pS @ %i\n", fn
, task_pid_nr(current
));
1148 *calltime
= ktime_get();
1151 static __init_or_module
void
1152 trace_initcall_finish_cb(void *data
, initcall_t fn
, int ret
)
1154 ktime_t
*calltime
= (ktime_t
*)data
;
1155 ktime_t delta
, rettime
;
1156 unsigned long long duration
;
1158 rettime
= ktime_get();
1159 delta
= ktime_sub(rettime
, *calltime
);
1160 duration
= (unsigned long long) ktime_to_ns(delta
) >> 10;
1161 printk(KERN_DEBUG
"initcall %pS returned %d after %lld usecs\n",
1165 static ktime_t initcall_calltime
;
1167 #ifdef TRACEPOINTS_ENABLED
1168 static void __init
initcall_debug_enable(void)
1172 ret
= register_trace_initcall_start(trace_initcall_start_cb
,
1173 &initcall_calltime
);
1174 ret
|= register_trace_initcall_finish(trace_initcall_finish_cb
,
1175 &initcall_calltime
);
1176 WARN(ret
, "Failed to register initcall tracepoints\n");
1178 # define do_trace_initcall_start trace_initcall_start
1179 # define do_trace_initcall_finish trace_initcall_finish
1181 static inline void do_trace_initcall_start(initcall_t fn
)
1183 if (!initcall_debug
)
1185 trace_initcall_start_cb(&initcall_calltime
, fn
);
1187 static inline void do_trace_initcall_finish(initcall_t fn
, int ret
)
1189 if (!initcall_debug
)
1191 trace_initcall_finish_cb(&initcall_calltime
, fn
, ret
);
1193 #endif /* !TRACEPOINTS_ENABLED */
1195 int __init_or_module
do_one_initcall(initcall_t fn
)
1197 int count
= preempt_count();
1201 if (initcall_blacklisted(fn
))
1204 do_trace_initcall_start(fn
);
1206 do_trace_initcall_finish(fn
, ret
);
1210 if (preempt_count() != count
) {
1211 sprintf(msgbuf
, "preemption imbalance ");
1212 preempt_count_set(count
);
1214 if (irqs_disabled()) {
1215 strlcat(msgbuf
, "disabled interrupts ", sizeof(msgbuf
));
1218 WARN(msgbuf
[0], "initcall %pS returned with %s\n", fn
, msgbuf
);
1220 add_latent_entropy();
1225 extern initcall_entry_t __initcall_start
[];
1226 extern initcall_entry_t __initcall0_start
[];
1227 extern initcall_entry_t __initcall1_start
[];
1228 extern initcall_entry_t __initcall2_start
[];
1229 extern initcall_entry_t __initcall3_start
[];
1230 extern initcall_entry_t __initcall4_start
[];
1231 extern initcall_entry_t __initcall5_start
[];
1232 extern initcall_entry_t __initcall6_start
[];
1233 extern initcall_entry_t __initcall7_start
[];
1234 extern initcall_entry_t __initcall_end
[];
1236 static initcall_entry_t
*initcall_levels
[] __initdata
= {
1248 /* Keep these in sync with initcalls in include/linux/init.h */
1249 static const char *initcall_level_names
[] __initdata
= {
1260 static int __init
ignore_unknown_bootoption(char *param
, char *val
,
1261 const char *unused
, void *arg
)
1266 static void __init
do_initcall_level(int level
, char *command_line
)
1268 initcall_entry_t
*fn
;
1270 parse_args(initcall_level_names
[level
],
1271 command_line
, __start___param
,
1272 __stop___param
- __start___param
,
1274 NULL
, ignore_unknown_bootoption
);
1276 trace_initcall_level(initcall_level_names
[level
]);
1277 for (fn
= initcall_levels
[level
]; fn
< initcall_levels
[level
+1]; fn
++)
1278 do_one_initcall(initcall_from_entry(fn
));
1281 static void __init
do_initcalls(void)
1284 size_t len
= strlen(saved_command_line
) + 1;
1287 command_line
= kzalloc(len
, GFP_KERNEL
);
1289 panic("%s: Failed to allocate %zu bytes\n", __func__
, len
);
1291 for (level
= 0; level
< ARRAY_SIZE(initcall_levels
) - 1; level
++) {
1292 /* Parser modifies command_line, restore it each time */
1293 strcpy(command_line
, saved_command_line
);
1294 do_initcall_level(level
, command_line
);
1297 kfree(command_line
);
1301 * Ok, the machine is now initialized. None of the devices
1302 * have been touched yet, but the CPU subsystem is up and
1303 * running, and memory and process management works.
1305 * Now we can finally start doing some real work..
1307 static void __init
do_basic_setup(void)
1313 usermodehelper_enable();
1317 static void __init
do_pre_smp_initcalls(void)
1319 initcall_entry_t
*fn
;
1321 trace_initcall_level("early");
1322 for (fn
= __initcall_start
; fn
< __initcall0_start
; fn
++)
1323 do_one_initcall(initcall_from_entry(fn
));
1326 static int run_init_process(const char *init_filename
)
1328 const char *const *p
;
1330 argv_init
[0] = init_filename
;
1331 pr_info("Run %s as init process\n", init_filename
);
1332 pr_debug(" with arguments:\n");
1333 for (p
= argv_init
; *p
; p
++)
1334 pr_debug(" %s\n", *p
);
1335 pr_debug(" with environment:\n");
1336 for (p
= envp_init
; *p
; p
++)
1337 pr_debug(" %s\n", *p
);
1338 return kernel_execve(init_filename
, argv_init
, envp_init
);
1341 static int try_to_run_init_process(const char *init_filename
)
1345 ret
= run_init_process(init_filename
);
1347 if (ret
&& ret
!= -ENOENT
) {
1348 pr_err("Starting init: %s exists but couldn't execute it (error %d)\n",
1349 init_filename
, ret
);
1355 static noinline
void __init
kernel_init_freeable(void);
1357 #if defined(CONFIG_STRICT_KERNEL_RWX) || defined(CONFIG_STRICT_MODULE_RWX)
1358 bool rodata_enabled __ro_after_init
= true;
1359 static int __init
set_debug_rodata(char *str
)
1361 return strtobool(str
, &rodata_enabled
);
1363 __setup("rodata=", set_debug_rodata
);
1366 #ifdef CONFIG_STRICT_KERNEL_RWX
1367 static void mark_readonly(void)
1369 if (rodata_enabled
) {
1371 * load_module() results in W+X mappings, which are cleaned
1372 * up with call_rcu(). Let's make sure that queued work is
1373 * flushed so that we don't hit false positives looking for
1374 * insecure pages which are W+X.
1380 pr_info("Kernel memory protection disabled.\n");
1382 #elif defined(CONFIG_ARCH_HAS_STRICT_KERNEL_RWX)
1383 static inline void mark_readonly(void)
1385 pr_warn("Kernel memory protection not selected by kernel config.\n");
1388 static inline void mark_readonly(void)
1390 pr_warn("This architecture does not have kernel memory protection.\n");
1394 void __weak
free_initmem(void)
1396 free_initmem_default(POISON_FREE_INITMEM
);
1399 static int __ref
kernel_init(void *unused
)
1403 kernel_init_freeable();
1404 /* need to finish all async __init code before freeing the memory */
1405 async_synchronize_full();
1406 kprobe_free_init_mem();
1407 ftrace_free_init_mem();
1412 * Kernel mappings are now finalized - update the userspace page-table
1417 system_state
= SYSTEM_RUNNING
;
1418 numa_default_policy();
1420 rcu_end_inkernel_boot();
1424 if (ramdisk_execute_command
) {
1425 ret
= run_init_process(ramdisk_execute_command
);
1428 pr_err("Failed to execute %s (error %d)\n",
1429 ramdisk_execute_command
, ret
);
1433 * We try each of these until one succeeds.
1435 * The Bourne shell can be used instead of init if we are
1436 * trying to recover a really broken machine.
1438 if (execute_command
) {
1439 ret
= run_init_process(execute_command
);
1442 panic("Requested init %s failed (error %d).",
1443 execute_command
, ret
);
1446 if (CONFIG_DEFAULT_INIT
[0] != '\0') {
1447 ret
= run_init_process(CONFIG_DEFAULT_INIT
);
1449 pr_err("Default init %s failed (error %d)\n",
1450 CONFIG_DEFAULT_INIT
, ret
);
1455 if (!try_to_run_init_process("/sbin/init") ||
1456 !try_to_run_init_process("/etc/init") ||
1457 !try_to_run_init_process("/bin/init") ||
1458 !try_to_run_init_process("/bin/sh"))
1461 panic("No working init found. Try passing init= option to kernel. "
1462 "See Linux Documentation/admin-guide/init.rst for guidance.");
1465 /* Open /dev/console, for stdin/stdout/stderr, this should never fail */
1466 void __init
console_on_rootfs(void)
1468 struct file
*file
= filp_open("/dev/console", O_RDWR
, 0);
1471 pr_err("Warning: unable to open an initial console.\n");
1480 static noinline
void __init
kernel_init_freeable(void)
1483 * Wait until kthreadd is all set-up.
1485 wait_for_completion(&kthreadd_done
);
1487 /* Now the scheduler is fully set up and can do blocking allocations */
1488 gfp_allowed_mask
= __GFP_BITS_MASK
;
1491 * init can allocate pages on any node
1493 set_mems_allowed(node_states
[N_MEMORY
]);
1495 cad_pid
= task_pid(current
);
1497 smp_prepare_cpus(setup_max_cpus
);
1501 init_mm_internals();
1503 do_pre_smp_initcalls();
1504 lockup_detector_init();
1510 page_alloc_init_late();
1511 /* Initialize page ext after all struct pages are initialized. */
1516 console_on_rootfs();
1519 * check if there is an early userspace init. If yes, let it do all
1522 if (init_eaccess(ramdisk_execute_command
) != 0) {
1523 ramdisk_execute_command
= NULL
;
1524 prepare_namespace();
1528 * Ok, we have completed the initial bootup, and
1529 * we're essentially up and running. Get rid of the
1530 * initmem segments and start the user-mode stuff..
1532 * rootfs is available now, try loading the public keys
1533 * and default modules
1536 integrity_load_keys();