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/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>
101 #include <linux/stackdepot.h>
102 #include <linux/randomize_kstack.h>
103 #include <net/net_namespace.h>
106 #include <asm/bugs.h>
107 #include <asm/setup.h>
108 #include <asm/sections.h>
109 #include <asm/cacheflush.h>
111 #define CREATE_TRACE_POINTS
112 #include <trace/events/initcall.h>
114 #include <kunit/test.h>
116 static int kernel_init(void *);
118 extern void init_IRQ(void);
119 extern void radix_tree_init(void);
122 * Debug helper: via this flag we know that we are in 'early bootup code'
123 * where only the boot processor is running with IRQ disabled. This means
124 * two things - IRQ must not be enabled before the flag is cleared and some
125 * operations which are not allowed with IRQ disabled are allowed while the
128 bool early_boot_irqs_disabled __read_mostly
;
130 enum system_states system_state __read_mostly
;
131 EXPORT_SYMBOL(system_state
);
134 * Boot command-line arguments
136 #define MAX_INIT_ARGS CONFIG_INIT_ENV_ARG_LIMIT
137 #define MAX_INIT_ENVS CONFIG_INIT_ENV_ARG_LIMIT
139 extern void time_init(void);
140 /* Default late time init is NULL. archs can override this later. */
141 void (*__initdata late_time_init
)(void);
143 /* Untouched command line saved by arch-specific code. */
144 char __initdata boot_command_line
[COMMAND_LINE_SIZE
];
145 /* Untouched saved command line (eg. for /proc) */
146 char *saved_command_line
;
147 /* Command line for parameter parsing */
148 static char *static_command_line
;
149 /* Untouched extra command line */
150 static char *extra_command_line
;
151 /* Extra init arguments */
152 static char *extra_init_args
;
154 #ifdef CONFIG_BOOT_CONFIG
155 /* Is bootconfig on command line? */
156 static bool bootconfig_found
;
157 static size_t initargs_offs
;
159 # define bootconfig_found false
160 # define initargs_offs 0
163 static char *execute_command
;
164 static char *ramdisk_execute_command
= "/init";
167 * Used to generate warnings if static_key manipulation functions are used
168 * before jump_label_init is called.
170 bool static_key_initialized __read_mostly
;
171 EXPORT_SYMBOL_GPL(static_key_initialized
);
174 * If set, this is an indication to the drivers that reset the underlying
175 * device before going ahead with the initialization otherwise driver might
176 * rely on the BIOS and skip the reset operation.
178 * This is useful if kernel is booting in an unreliable environment.
179 * For ex. kdump situation where previous kernel has crashed, BIOS has been
180 * skipped and devices will be in unknown state.
182 unsigned int reset_devices
;
183 EXPORT_SYMBOL(reset_devices
);
185 static int __init
set_reset_devices(char *str
)
191 __setup("reset_devices", set_reset_devices
);
193 static const char *argv_init
[MAX_INIT_ARGS
+2] = { "init", NULL
, };
194 const char *envp_init
[MAX_INIT_ENVS
+2] = { "HOME=/", "TERM=linux", NULL
, };
195 static const char *panic_later
, *panic_param
;
197 extern const struct obs_kernel_param __setup_start
[], __setup_end
[];
199 static bool __init
obsolete_checksetup(char *line
)
201 const struct obs_kernel_param
*p
;
202 bool had_early_param
= false;
206 int n
= strlen(p
->str
);
207 if (parameqn(line
, p
->str
, n
)) {
209 /* Already done in parse_early_param?
210 * (Needs exact match on param part).
211 * Keep iterating, as we can have early
212 * params and __setups of same names 8( */
213 if (line
[n
] == '\0' || line
[n
] == '=')
214 had_early_param
= true;
215 } else if (!p
->setup_func
) {
216 pr_warn("Parameter %s is obsolete, ignored\n",
219 } else if (p
->setup_func(line
+ n
))
223 } while (p
< __setup_end
);
225 return had_early_param
;
229 * This should be approx 2 Bo*oMips to start (note initial shift), and will
230 * still work even if initially too large, it will just take slightly longer
232 unsigned long loops_per_jiffy
= (1<<12);
233 EXPORT_SYMBOL(loops_per_jiffy
);
235 static int __init
debug_kernel(char *str
)
237 console_loglevel
= CONSOLE_LOGLEVEL_DEBUG
;
241 static int __init
quiet_kernel(char *str
)
243 console_loglevel
= CONSOLE_LOGLEVEL_QUIET
;
247 early_param("debug", debug_kernel
);
248 early_param("quiet", quiet_kernel
);
250 static int __init
loglevel(char *str
)
255 * Only update loglevel value when a correct setting was passed,
256 * to prevent blind crashes (when loglevel being set to 0) that
257 * are quite hard to debug
259 if (get_option(&str
, &newlevel
)) {
260 console_loglevel
= newlevel
;
267 early_param("loglevel", loglevel
);
269 #ifdef CONFIG_BLK_DEV_INITRD
270 static void * __init
get_boot_config_from_initrd(size_t *_size
)
280 data
= (char *)initrd_end
- BOOTCONFIG_MAGIC_LEN
;
282 * Since Grub may align the size of initrd to 4, we must
283 * check the preceding 3 bytes as well.
285 for (i
= 0; i
< 4; i
++) {
286 if (!memcmp(data
, BOOTCONFIG_MAGIC
, BOOTCONFIG_MAGIC_LEN
))
293 hdr
= (u32
*)(data
- 8);
294 size
= le32_to_cpu(hdr
[0]);
295 csum
= le32_to_cpu(hdr
[1]);
297 data
= ((void *)hdr
) - size
;
298 if ((unsigned long)data
< initrd_start
) {
299 pr_err("bootconfig size %d is greater than initrd size %ld\n",
300 size
, initrd_end
- initrd_start
);
304 if (xbc_calc_checksum(data
, size
) != csum
) {
305 pr_err("bootconfig checksum failed\n");
309 /* Remove bootconfig from initramfs/initrd */
310 initrd_end
= (unsigned long)data
;
317 static void * __init
get_boot_config_from_initrd(size_t *_size
)
323 #ifdef CONFIG_BOOT_CONFIG
325 static char xbc_namebuf
[XBC_KEYLEN_MAX
] __initdata
;
327 #define rest(dst, end) ((end) > (dst) ? (end) - (dst) : 0)
329 static int __init
xbc_snprint_cmdline(char *buf
, size_t size
,
330 struct xbc_node
*root
)
332 struct xbc_node
*knode
, *vnode
;
333 char *end
= buf
+ size
;
337 xbc_node_for_each_key_value(root
, knode
, val
) {
338 ret
= xbc_node_compose_key_after(root
, knode
,
339 xbc_namebuf
, XBC_KEYLEN_MAX
);
343 vnode
= xbc_node_get_child(knode
);
345 ret
= snprintf(buf
, rest(buf
, end
), "%s ", xbc_namebuf
);
351 xbc_array_for_each_value(vnode
, val
) {
352 ret
= snprintf(buf
, rest(buf
, end
), "%s=\"%s\" ",
360 return buf
- (end
- size
);
364 /* Make an extra command line under given key word */
365 static char * __init
xbc_make_cmdline(const char *key
)
367 struct xbc_node
*root
;
371 root
= xbc_find_node(key
);
375 /* Count required buffer size */
376 len
= xbc_snprint_cmdline(NULL
, 0, root
);
380 new_cmdline
= memblock_alloc(len
+ 1, SMP_CACHE_BYTES
);
382 pr_err("Failed to allocate memory for extra kernel cmdline.\n");
386 ret
= xbc_snprint_cmdline(new_cmdline
, len
+ 1, root
);
387 if (ret
< 0 || ret
> len
) {
388 pr_err("Failed to print extra kernel cmdline.\n");
389 memblock_free(new_cmdline
, len
+ 1);
396 static int __init
bootconfig_params(char *param
, char *val
,
397 const char *unused
, void *arg
)
399 if (strcmp(param
, "bootconfig") == 0) {
400 bootconfig_found
= true;
405 static int __init
warn_bootconfig(char *str
)
407 /* The 'bootconfig' has been handled by bootconfig_params(). */
411 static void __init
setup_boot_config(void)
413 static char tmp_cmdline
[COMMAND_LINE_SIZE
] __initdata
;
414 const char *msg
, *data
;
419 /* Cut out the bootconfig data even if we have no bootconfig option */
420 data
= get_boot_config_from_initrd(&size
);
421 /* If there is no bootconfig in initrd, try embedded one. */
423 data
= xbc_get_embedded_bootconfig(&size
);
425 strlcpy(tmp_cmdline
, boot_command_line
, COMMAND_LINE_SIZE
);
426 err
= parse_args("bootconfig", tmp_cmdline
, NULL
, 0, 0, 0, NULL
,
429 if (IS_ERR(err
) || !bootconfig_found
)
432 /* parse_args() stops at the next param of '--' and returns an address */
434 initargs_offs
= err
- tmp_cmdline
;
437 pr_err("'bootconfig' found on command line, but no bootconfig found\n");
441 if (size
>= XBC_DATA_MAX
) {
442 pr_err("bootconfig size %ld greater than max size %d\n",
443 (long)size
, XBC_DATA_MAX
);
447 ret
= xbc_init(data
, size
, &msg
, &pos
);
450 pr_err("Failed to init bootconfig: %s.\n", msg
);
452 pr_err("Failed to parse bootconfig: %s at %d.\n",
455 xbc_get_info(&ret
, NULL
);
456 pr_info("Load bootconfig: %ld bytes %d nodes\n", (long)size
, ret
);
457 /* keys starting with "kernel." are passed via cmdline */
458 extra_command_line
= xbc_make_cmdline("kernel");
459 /* Also, "init." keys are init arguments */
460 extra_init_args
= xbc_make_cmdline("init");
465 static void __init
exit_boot_config(void)
470 #else /* !CONFIG_BOOT_CONFIG */
472 static void __init
setup_boot_config(void)
474 /* Remove bootconfig data from initrd */
475 get_boot_config_from_initrd(NULL
);
478 static int __init
warn_bootconfig(char *str
)
480 pr_warn("WARNING: 'bootconfig' found on the kernel command line but CONFIG_BOOT_CONFIG is not set.\n");
484 #define exit_boot_config() do {} while (0)
486 #endif /* CONFIG_BOOT_CONFIG */
488 early_param("bootconfig", warn_bootconfig
);
490 /* Change NUL term back to "=", to make "param" the whole string. */
491 static void __init
repair_env_string(char *param
, char *val
)
494 /* param=val or param="val"? */
495 if (val
== param
+strlen(param
)+1)
497 else if (val
== param
+strlen(param
)+2) {
499 memmove(val
-1, val
, strlen(val
)+1);
505 /* Anything after -- gets handed straight to init. */
506 static int __init
set_init_arg(char *param
, char *val
,
507 const char *unused
, void *arg
)
514 repair_env_string(param
, val
);
516 for (i
= 0; argv_init
[i
]; i
++) {
517 if (i
== MAX_INIT_ARGS
) {
518 panic_later
= "init";
523 argv_init
[i
] = param
;
528 * Unknown boot options get handed to init, unless they look like
529 * unused parameters (modprobe will find them in /proc/cmdline).
531 static int __init
unknown_bootoption(char *param
, char *val
,
532 const char *unused
, void *arg
)
534 size_t len
= strlen(param
);
536 repair_env_string(param
, val
);
538 /* Handle obsolete-style parameters */
539 if (obsolete_checksetup(param
))
542 /* Unused module parameter. */
543 if (strnchr(param
, len
, '.'))
550 /* Environment option */
552 for (i
= 0; envp_init
[i
]; i
++) {
553 if (i
== MAX_INIT_ENVS
) {
557 if (!strncmp(param
, envp_init
[i
], len
+1))
560 envp_init
[i
] = param
;
562 /* Command line option */
564 for (i
= 0; argv_init
[i
]; i
++) {
565 if (i
== MAX_INIT_ARGS
) {
566 panic_later
= "init";
570 argv_init
[i
] = param
;
575 static int __init
init_setup(char *str
)
579 execute_command
= str
;
581 * In case LILO is going to boot us with default command line,
582 * it prepends "auto" before the whole cmdline which makes
583 * the shell think it should execute a script with such name.
584 * So we ignore all arguments entered _before_ init=... [MJ]
586 for (i
= 1; i
< MAX_INIT_ARGS
; i
++)
590 __setup("init=", init_setup
);
592 static int __init
rdinit_setup(char *str
)
596 ramdisk_execute_command
= str
;
597 /* See "auto" comment in init_setup */
598 for (i
= 1; i
< MAX_INIT_ARGS
; i
++)
602 __setup("rdinit=", rdinit_setup
);
605 static const unsigned int setup_max_cpus
= NR_CPUS
;
606 static inline void setup_nr_cpu_ids(void) { }
607 static inline void smp_prepare_cpus(unsigned int maxcpus
) { }
611 * We need to store the untouched command line for future reference.
612 * We also need to store the touched command line since the parameter
613 * parsing is performed in place, and we should allow a component to
614 * store reference of name/value for future reference.
616 static void __init
setup_command_line(char *command_line
)
618 size_t len
, xlen
= 0, ilen
= 0;
620 if (extra_command_line
)
621 xlen
= strlen(extra_command_line
);
623 ilen
= strlen(extra_init_args
) + 4; /* for " -- " */
625 len
= xlen
+ strlen(boot_command_line
) + 1;
627 saved_command_line
= memblock_alloc(len
+ ilen
, SMP_CACHE_BYTES
);
628 if (!saved_command_line
)
629 panic("%s: Failed to allocate %zu bytes\n", __func__
, len
+ ilen
);
631 static_command_line
= memblock_alloc(len
, SMP_CACHE_BYTES
);
632 if (!static_command_line
)
633 panic("%s: Failed to allocate %zu bytes\n", __func__
, len
);
637 * We have to put extra_command_line before boot command
638 * lines because there could be dashes (separator of init
639 * command line) in the command lines.
641 strcpy(saved_command_line
, extra_command_line
);
642 strcpy(static_command_line
, extra_command_line
);
644 strcpy(saved_command_line
+ xlen
, boot_command_line
);
645 strcpy(static_command_line
+ xlen
, command_line
);
649 * Append supplemental init boot args to saved_command_line
650 * so that user can check what command line options passed
652 * The order should always be
653 * " -- "[bootconfig init-param][cmdline init-param]
656 len
= xlen
+ initargs_offs
;
657 strcpy(saved_command_line
+ len
, extra_init_args
);
658 len
+= ilen
- 4; /* strlen(extra_init_args) */
659 strcpy(saved_command_line
+ len
,
660 boot_command_line
+ initargs_offs
- 1);
662 len
= strlen(saved_command_line
);
663 strcpy(saved_command_line
+ len
, " -- ");
665 strcpy(saved_command_line
+ len
, extra_init_args
);
671 * We need to finalize in a non-__init function or else race conditions
672 * between the root thread and the init thread may cause start_kernel to
673 * be reaped by free_initmem before the root thread has proceeded to
676 * gcc-3.4 accidentally inlines this function, so use noinline.
679 static __initdata
DECLARE_COMPLETION(kthreadd_done
);
681 noinline
void __ref
rest_init(void)
683 struct task_struct
*tsk
;
686 rcu_scheduler_starting();
688 * We need to spawn init first so that it obtains pid 1, however
689 * the init task will end up wanting to create kthreads, which, if
690 * we schedule it before we create kthreadd, will OOPS.
692 pid
= user_mode_thread(kernel_init
, NULL
, CLONE_FS
);
694 * Pin init on the boot CPU. Task migration is not properly working
695 * until sched_init_smp() has been run. It will set the allowed
696 * CPUs for init to the non isolated CPUs.
699 tsk
= find_task_by_pid_ns(pid
, &init_pid_ns
);
700 tsk
->flags
|= PF_NO_SETAFFINITY
;
701 set_cpus_allowed_ptr(tsk
, cpumask_of(smp_processor_id()));
704 numa_default_policy();
705 pid
= kernel_thread(kthreadd
, NULL
, CLONE_FS
| CLONE_FILES
);
707 kthreadd_task
= find_task_by_pid_ns(pid
, &init_pid_ns
);
711 * Enable might_sleep() and smp_processor_id() checks.
712 * They cannot be enabled earlier because with CONFIG_PREEMPTION=y
713 * kernel_thread() would trigger might_sleep() splats. With
714 * CONFIG_PREEMPT_VOLUNTARY=y the init task might have scheduled
715 * already, but it's stuck on the kthreadd_done completion.
717 system_state
= SYSTEM_SCHEDULING
;
719 complete(&kthreadd_done
);
722 * The boot idle thread must execute schedule()
723 * at least once to get things moving:
725 schedule_preempt_disabled();
726 /* Call into cpu_idle with preempt disabled */
727 cpu_startup_entry(CPUHP_ONLINE
);
730 /* Check for early params. */
731 static int __init
do_early_param(char *param
, char *val
,
732 const char *unused
, void *arg
)
734 const struct obs_kernel_param
*p
;
736 for (p
= __setup_start
; p
< __setup_end
; p
++) {
737 if ((p
->early
&& parameq(param
, p
->str
)) ||
738 (strcmp(param
, "console") == 0 &&
739 strcmp(p
->str
, "earlycon") == 0)
741 if (p
->setup_func(val
) != 0)
742 pr_warn("Malformed early option '%s'\n", param
);
745 /* We accept everything at this stage. */
749 void __init
parse_early_options(char *cmdline
)
751 parse_args("early options", cmdline
, NULL
, 0, 0, 0, NULL
,
755 /* Arch code calls this early on, or if not, just before other parsing. */
756 void __init
parse_early_param(void)
758 static int done __initdata
;
759 static char tmp_cmdline
[COMMAND_LINE_SIZE
] __initdata
;
764 /* All fall through to do_early_param. */
765 strlcpy(tmp_cmdline
, boot_command_line
, COMMAND_LINE_SIZE
);
766 parse_early_options(tmp_cmdline
);
770 void __init __weak
arch_post_acpi_subsys_init(void) { }
772 void __init __weak
smp_setup_processor_id(void)
776 # if THREAD_SIZE >= PAGE_SIZE
777 void __init __weak
thread_stack_cache_init(void)
782 void __init __weak
mem_encrypt_init(void) { }
784 void __init __weak
poking_init(void) { }
786 void __init __weak
pgtable_cache_init(void) { }
788 void __init __weak
trap_init(void) { }
791 core_param(initcall_debug
, initcall_debug
, bool, 0644);
793 #ifdef TRACEPOINTS_ENABLED
794 static void __init
initcall_debug_enable(void);
796 static inline void initcall_debug_enable(void)
801 /* Report memory auto-initialization states for this boot. */
802 static void __init
report_meminit(void)
806 if (IS_ENABLED(CONFIG_INIT_STACK_ALL_PATTERN
))
807 stack
= "all(pattern)";
808 else if (IS_ENABLED(CONFIG_INIT_STACK_ALL_ZERO
))
810 else if (IS_ENABLED(CONFIG_GCC_PLUGIN_STRUCTLEAK_BYREF_ALL
))
811 stack
= "byref_all(zero)";
812 else if (IS_ENABLED(CONFIG_GCC_PLUGIN_STRUCTLEAK_BYREF
))
813 stack
= "byref(zero)";
814 else if (IS_ENABLED(CONFIG_GCC_PLUGIN_STRUCTLEAK_USER
))
815 stack
= "__user(zero)";
819 pr_info("mem auto-init: stack:%s, heap alloc:%s, heap free:%s\n",
820 stack
, want_init_on_alloc(GFP_KERNEL
) ? "on" : "off",
821 want_init_on_free() ? "on" : "off");
822 if (want_init_on_free())
823 pr_info("mem auto-init: clearing system memory may take some time...\n");
827 * Set up kernel memory allocators
829 static void __init
mm_init(void)
832 * page_ext requires contiguous pages,
833 * bigger than MAX_ORDER unless SPARSEMEM.
835 page_ext_init_flatmem();
836 init_mem_debugging_and_hardening();
839 stack_depot_early_init();
841 mem_init_print_info();
844 * page_owner must be initialized after buddy is ready, and also after
845 * slab is ready so that stack_depot_init() works properly
847 page_ext_init_flatmem_late();
850 debug_objects_mem_init();
852 /* Should be run before the first non-init thread is created */
854 /* Should be run after espfix64 is set up. */
858 #ifdef CONFIG_RANDOMIZE_KSTACK_OFFSET
859 DEFINE_STATIC_KEY_MAYBE_RO(CONFIG_RANDOMIZE_KSTACK_OFFSET_DEFAULT
,
860 randomize_kstack_offset
);
861 DEFINE_PER_CPU(u32
, kstack_offset
);
863 static int __init
early_randomize_kstack_offset(char *buf
)
868 ret
= kstrtobool(buf
, &bool_result
);
873 static_branch_enable(&randomize_kstack_offset
);
875 static_branch_disable(&randomize_kstack_offset
);
878 early_param("randomize_kstack_offset", early_randomize_kstack_offset
);
881 void __init __weak
arch_call_rest_init(void)
886 static void __init
print_unknown_bootoptions(void)
888 char *unknown_options
;
890 const char *const *p
;
893 if (panic_later
|| (!argv_init
[1] && !envp_init
[2]))
897 * Determine how many options we have to print out, plus a space
900 len
= 1; /* null terminator */
901 for (p
= &argv_init
[1]; *p
; p
++) {
905 for (p
= &envp_init
[2]; *p
; p
++) {
910 unknown_options
= memblock_alloc(len
, SMP_CACHE_BYTES
);
911 if (!unknown_options
) {
912 pr_err("%s: Failed to allocate %zu bytes\n",
916 end
= unknown_options
;
918 for (p
= &argv_init
[1]; *p
; p
++)
919 end
+= sprintf(end
, " %s", *p
);
920 for (p
= &envp_init
[2]; *p
; p
++)
921 end
+= sprintf(end
, " %s", *p
);
923 /* Start at unknown_options[1] to skip the initial space */
924 pr_notice("Unknown kernel command line parameters \"%s\", will be passed to user space.\n",
925 &unknown_options
[1]);
926 memblock_free(unknown_options
, len
);
929 asmlinkage __visible
void __init __no_sanitize_address
start_kernel(void)
934 set_task_stack_end_magic(&init_task
);
935 smp_setup_processor_id();
936 debug_objects_early_init();
937 init_vmlinux_build_id();
942 early_boot_irqs_disabled
= true;
945 * Interrupts are still disabled. Do necessary setups, then
950 pr_notice("%s", linux_banner
);
951 early_security_init();
952 setup_arch(&command_line
);
954 setup_command_line(command_line
);
956 setup_per_cpu_areas();
957 smp_prepare_boot_cpu(); /* arch-specific boot-cpu hooks */
958 boot_cpu_hotplug_init();
960 build_all_zonelists(NULL
);
963 pr_notice("Kernel command line: %s\n", saved_command_line
);
964 /* parameters may set static keys */
967 after_dashes
= parse_args("Booting kernel",
968 static_command_line
, __start___param
,
969 __stop___param
- __start___param
,
970 -1, -1, NULL
, &unknown_bootoption
);
971 print_unknown_bootoptions();
972 if (!IS_ERR_OR_NULL(after_dashes
))
973 parse_args("Setting init args", after_dashes
, NULL
, 0, -1, -1,
976 parse_args("Setting extra init args", extra_init_args
,
977 NULL
, 0, -1, -1, NULL
, set_init_arg
);
980 * These use large bootmem allocations and must precede
984 vfs_caches_init_early();
991 /* trace_printk can be enabled here */
995 * Set up the scheduler prior starting any interrupts (such as the
996 * timer interrupt). Full topology setup happens at smp_init()
997 * time - but meanwhile we still have a functioning scheduler.
1001 if (WARN(!irqs_disabled(),
1002 "Interrupts were enabled *very* early, fixing it\n"))
1003 local_irq_disable();
1007 * Set up housekeeping before setting up workqueues to allow the unbound
1008 * workqueue to take non-housekeeping into account.
1010 housekeeping_init();
1013 * Allow workqueue creation and work item queueing/cancelling
1014 * early. Work item execution depends on kthreads and starts after
1017 workqueue_init_early();
1021 /* Trace events are available after this */
1025 initcall_debug_enable();
1027 context_tracking_init();
1028 /* init some links before init_ISA_irqs() */
1042 * For best initial stack canary entropy, prepare it after:
1043 * - setup_arch() for any UEFI RNG entropy and boot cmdline access
1044 * - timekeeping_init() for ktime entropy used in random_init()
1045 * - time_init() for making random_get_entropy() work on some platforms
1046 * - random_init() to initialize the RNG from from early entropy sources
1048 random_init(command_line
);
1049 boot_init_stack_canary();
1053 call_function_init();
1054 WARN(!irqs_disabled(), "Interrupts were enabled early\n");
1056 early_boot_irqs_disabled
= false;
1059 kmem_cache_init_late();
1062 * HACK ALERT! This is early. We're enabling the console before
1063 * we've done PCI setups etc, and console_init() must be aware of
1064 * this. But we do want output early, in case something goes wrong.
1068 panic("Too many boot %s vars at `%s'", panic_later
,
1074 * Need to run this when irqs are enabled, because it wants
1075 * to self-test [hard/soft]-irqs on/off lock inversion bugs
1081 * This needs to be called before any devices perform DMA
1082 * operations that might use the SWIOTLB bounce buffers. It will
1083 * mark the bounce buffers as decrypted so that their usage will
1084 * not cause "plain-text" data to be decrypted when accessed.
1088 #ifdef CONFIG_BLK_DEV_INITRD
1089 if (initrd_start
&& !initrd_below_start_ok
&&
1090 page_to_pfn(virt_to_page((void *)initrd_start
)) < min_low_pfn
) {
1091 pr_crit("initrd overwritten (0x%08lx < 0x%08lx) - disabling it.\n",
1092 page_to_pfn(virt_to_page((void *)initrd_start
)),
1097 setup_per_cpu_pageset();
1107 if (efi_enabled(EFI_RUNTIME_SERVICES
))
1108 efi_enter_virtual_mode();
1110 thread_stack_cache_init();
1127 taskstats_init_early();
1133 acpi_subsystem_init();
1134 arch_post_acpi_subsys_init();
1137 /* Do the rest non-__init'ed, we're now alive */
1138 arch_call_rest_init();
1140 prevent_tail_call_optimization();
1143 /* Call all constructor functions linked into the kernel. */
1144 static void __init
do_ctors(void)
1147 * For UML, the constructors have already been called by the
1148 * normal setup code as it's just a normal ELF binary, so we
1149 * cannot do it again - but we do need CONFIG_CONSTRUCTORS
1150 * even on UML for modules.
1152 #if defined(CONFIG_CONSTRUCTORS) && !defined(CONFIG_UML)
1153 ctor_fn_t
*fn
= (ctor_fn_t
*) __ctors_start
;
1155 for (; fn
< (ctor_fn_t
*) __ctors_end
; fn
++)
1160 #ifdef CONFIG_KALLSYMS
1161 struct blacklist_entry
{
1162 struct list_head next
;
1166 static __initdata_or_module
LIST_HEAD(blacklisted_initcalls
);
1168 static int __init
initcall_blacklist(char *str
)
1171 struct blacklist_entry
*entry
;
1173 /* str argument is a comma-separated list of functions */
1175 str_entry
= strsep(&str
, ",");
1177 pr_debug("blacklisting initcall %s\n", str_entry
);
1178 entry
= memblock_alloc(sizeof(*entry
),
1181 panic("%s: Failed to allocate %zu bytes\n",
1182 __func__
, sizeof(*entry
));
1183 entry
->buf
= memblock_alloc(strlen(str_entry
) + 1,
1186 panic("%s: Failed to allocate %zu bytes\n",
1187 __func__
, strlen(str_entry
) + 1);
1188 strcpy(entry
->buf
, str_entry
);
1189 list_add(&entry
->next
, &blacklisted_initcalls
);
1191 } while (str_entry
);
1196 static bool __init_or_module
initcall_blacklisted(initcall_t fn
)
1198 struct blacklist_entry
*entry
;
1199 char fn_name
[KSYM_SYMBOL_LEN
];
1202 if (list_empty(&blacklisted_initcalls
))
1205 addr
= (unsigned long) dereference_function_descriptor(fn
);
1206 sprint_symbol_no_offset(fn_name
, addr
);
1209 * fn will be "function_name [module_name]" where [module_name] is not
1210 * displayed for built-in init functions. Strip off the [module_name].
1212 strreplace(fn_name
, ' ', '\0');
1214 list_for_each_entry(entry
, &blacklisted_initcalls
, next
) {
1215 if (!strcmp(fn_name
, entry
->buf
)) {
1216 pr_debug("initcall %s blacklisted\n", fn_name
);
1224 static int __init
initcall_blacklist(char *str
)
1226 pr_warn("initcall_blacklist requires CONFIG_KALLSYMS\n");
1230 static bool __init_or_module
initcall_blacklisted(initcall_t fn
)
1235 __setup("initcall_blacklist=", initcall_blacklist
);
1237 static __init_or_module
void
1238 trace_initcall_start_cb(void *data
, initcall_t fn
)
1240 ktime_t
*calltime
= (ktime_t
*)data
;
1242 printk(KERN_DEBUG
"calling %pS @ %i\n", fn
, task_pid_nr(current
));
1243 *calltime
= ktime_get();
1246 static __init_or_module
void
1247 trace_initcall_finish_cb(void *data
, initcall_t fn
, int ret
)
1249 ktime_t rettime
, *calltime
= (ktime_t
*)data
;
1251 rettime
= ktime_get();
1252 printk(KERN_DEBUG
"initcall %pS returned %d after %lld usecs\n",
1253 fn
, ret
, (unsigned long long)ktime_us_delta(rettime
, *calltime
));
1256 static ktime_t initcall_calltime
;
1258 #ifdef TRACEPOINTS_ENABLED
1259 static void __init
initcall_debug_enable(void)
1263 ret
= register_trace_initcall_start(trace_initcall_start_cb
,
1264 &initcall_calltime
);
1265 ret
|= register_trace_initcall_finish(trace_initcall_finish_cb
,
1266 &initcall_calltime
);
1267 WARN(ret
, "Failed to register initcall tracepoints\n");
1269 # define do_trace_initcall_start trace_initcall_start
1270 # define do_trace_initcall_finish trace_initcall_finish
1272 static inline void do_trace_initcall_start(initcall_t fn
)
1274 if (!initcall_debug
)
1276 trace_initcall_start_cb(&initcall_calltime
, fn
);
1278 static inline void do_trace_initcall_finish(initcall_t fn
, int ret
)
1280 if (!initcall_debug
)
1282 trace_initcall_finish_cb(&initcall_calltime
, fn
, ret
);
1284 #endif /* !TRACEPOINTS_ENABLED */
1286 int __init_or_module
do_one_initcall(initcall_t fn
)
1288 int count
= preempt_count();
1292 if (initcall_blacklisted(fn
))
1295 do_trace_initcall_start(fn
);
1297 do_trace_initcall_finish(fn
, ret
);
1301 if (preempt_count() != count
) {
1302 sprintf(msgbuf
, "preemption imbalance ");
1303 preempt_count_set(count
);
1305 if (irqs_disabled()) {
1306 strlcat(msgbuf
, "disabled interrupts ", sizeof(msgbuf
));
1309 WARN(msgbuf
[0], "initcall %pS returned with %s\n", fn
, msgbuf
);
1311 add_latent_entropy();
1316 extern initcall_entry_t __initcall_start
[];
1317 extern initcall_entry_t __initcall0_start
[];
1318 extern initcall_entry_t __initcall1_start
[];
1319 extern initcall_entry_t __initcall2_start
[];
1320 extern initcall_entry_t __initcall3_start
[];
1321 extern initcall_entry_t __initcall4_start
[];
1322 extern initcall_entry_t __initcall5_start
[];
1323 extern initcall_entry_t __initcall6_start
[];
1324 extern initcall_entry_t __initcall7_start
[];
1325 extern initcall_entry_t __initcall_end
[];
1327 static initcall_entry_t
*initcall_levels
[] __initdata
= {
1339 /* Keep these in sync with initcalls in include/linux/init.h */
1340 static const char *initcall_level_names
[] __initdata
= {
1351 static int __init
ignore_unknown_bootoption(char *param
, char *val
,
1352 const char *unused
, void *arg
)
1357 static void __init
do_initcall_level(int level
, char *command_line
)
1359 initcall_entry_t
*fn
;
1361 parse_args(initcall_level_names
[level
],
1362 command_line
, __start___param
,
1363 __stop___param
- __start___param
,
1365 NULL
, ignore_unknown_bootoption
);
1367 trace_initcall_level(initcall_level_names
[level
]);
1368 for (fn
= initcall_levels
[level
]; fn
< initcall_levels
[level
+1]; fn
++)
1369 do_one_initcall(initcall_from_entry(fn
));
1372 static void __init
do_initcalls(void)
1375 size_t len
= strlen(saved_command_line
) + 1;
1378 command_line
= kzalloc(len
, GFP_KERNEL
);
1380 panic("%s: Failed to allocate %zu bytes\n", __func__
, len
);
1382 for (level
= 0; level
< ARRAY_SIZE(initcall_levels
) - 1; level
++) {
1383 /* Parser modifies command_line, restore it each time */
1384 strcpy(command_line
, saved_command_line
);
1385 do_initcall_level(level
, command_line
);
1388 kfree(command_line
);
1392 * Ok, the machine is now initialized. None of the devices
1393 * have been touched yet, but the CPU subsystem is up and
1394 * running, and memory and process management works.
1396 * Now we can finally start doing some real work..
1398 static void __init
do_basic_setup(void)
1407 static void __init
do_pre_smp_initcalls(void)
1409 initcall_entry_t
*fn
;
1411 trace_initcall_level("early");
1412 for (fn
= __initcall_start
; fn
< __initcall0_start
; fn
++)
1413 do_one_initcall(initcall_from_entry(fn
));
1416 static int run_init_process(const char *init_filename
)
1418 const char *const *p
;
1420 argv_init
[0] = init_filename
;
1421 pr_info("Run %s as init process\n", init_filename
);
1422 pr_debug(" with arguments:\n");
1423 for (p
= argv_init
; *p
; p
++)
1424 pr_debug(" %s\n", *p
);
1425 pr_debug(" with environment:\n");
1426 for (p
= envp_init
; *p
; p
++)
1427 pr_debug(" %s\n", *p
);
1428 return kernel_execve(init_filename
, argv_init
, envp_init
);
1431 static int try_to_run_init_process(const char *init_filename
)
1435 ret
= run_init_process(init_filename
);
1437 if (ret
&& ret
!= -ENOENT
) {
1438 pr_err("Starting init: %s exists but couldn't execute it (error %d)\n",
1439 init_filename
, ret
);
1445 static noinline
void __init
kernel_init_freeable(void);
1447 #if defined(CONFIG_STRICT_KERNEL_RWX) || defined(CONFIG_STRICT_MODULE_RWX)
1448 bool rodata_enabled __ro_after_init
= true;
1449 static int __init
set_debug_rodata(char *str
)
1451 if (strtobool(str
, &rodata_enabled
))
1452 pr_warn("Invalid option string for rodata: '%s'\n", str
);
1455 __setup("rodata=", set_debug_rodata
);
1458 #ifdef CONFIG_STRICT_KERNEL_RWX
1459 static void mark_readonly(void)
1461 if (rodata_enabled
) {
1463 * load_module() results in W+X mappings, which are cleaned
1464 * up with call_rcu(). Let's make sure that queued work is
1465 * flushed so that we don't hit false positives looking for
1466 * insecure pages which are W+X.
1472 pr_info("Kernel memory protection disabled.\n");
1474 #elif defined(CONFIG_ARCH_HAS_STRICT_KERNEL_RWX)
1475 static inline void mark_readonly(void)
1477 pr_warn("Kernel memory protection not selected by kernel config.\n");
1480 static inline void mark_readonly(void)
1482 pr_warn("This architecture does not have kernel memory protection.\n");
1486 void __weak
free_initmem(void)
1488 free_initmem_default(POISON_FREE_INITMEM
);
1491 static int __ref
kernel_init(void *unused
)
1496 * Wait until kthreadd is all set-up.
1498 wait_for_completion(&kthreadd_done
);
1500 kernel_init_freeable();
1501 /* need to finish all async __init code before freeing the memory */
1502 async_synchronize_full();
1504 system_state
= SYSTEM_FREEING_INITMEM
;
1505 kprobe_free_init_mem();
1506 ftrace_free_init_mem();
1507 kgdb_free_init_mem();
1513 * Kernel mappings are now finalized - update the userspace page-table
1518 system_state
= SYSTEM_RUNNING
;
1519 numa_default_policy();
1521 rcu_end_inkernel_boot();
1525 if (ramdisk_execute_command
) {
1526 ret
= run_init_process(ramdisk_execute_command
);
1529 pr_err("Failed to execute %s (error %d)\n",
1530 ramdisk_execute_command
, ret
);
1534 * We try each of these until one succeeds.
1536 * The Bourne shell can be used instead of init if we are
1537 * trying to recover a really broken machine.
1539 if (execute_command
) {
1540 ret
= run_init_process(execute_command
);
1543 panic("Requested init %s failed (error %d).",
1544 execute_command
, ret
);
1547 if (CONFIG_DEFAULT_INIT
[0] != '\0') {
1548 ret
= run_init_process(CONFIG_DEFAULT_INIT
);
1550 pr_err("Default init %s failed (error %d)\n",
1551 CONFIG_DEFAULT_INIT
, ret
);
1556 if (!try_to_run_init_process("/sbin/init") ||
1557 !try_to_run_init_process("/etc/init") ||
1558 !try_to_run_init_process("/bin/init") ||
1559 !try_to_run_init_process("/bin/sh"))
1562 panic("No working init found. Try passing init= option to kernel. "
1563 "See Linux Documentation/admin-guide/init.rst for guidance.");
1566 /* Open /dev/console, for stdin/stdout/stderr, this should never fail */
1567 void __init
console_on_rootfs(void)
1569 struct file
*file
= filp_open("/dev/console", O_RDWR
, 0);
1572 pr_err("Warning: unable to open an initial console.\n");
1581 static noinline
void __init
kernel_init_freeable(void)
1583 /* Now the scheduler is fully set up and can do blocking allocations */
1584 gfp_allowed_mask
= __GFP_BITS_MASK
;
1587 * init can allocate pages on any node
1589 set_mems_allowed(node_states
[N_MEMORY
]);
1591 cad_pid
= get_pid(task_pid(current
));
1593 smp_prepare_cpus(setup_max_cpus
);
1597 init_mm_internals();
1599 rcu_init_tasks_generic();
1600 do_pre_smp_initcalls();
1601 lockup_detector_init();
1607 page_alloc_init_late();
1608 /* Initialize page ext after all struct pages are initialized. */
1613 kunit_run_all_tests();
1615 wait_for_initramfs();
1616 console_on_rootfs();
1619 * check if there is an early userspace init. If yes, let it do all
1622 if (init_eaccess(ramdisk_execute_command
) != 0) {
1623 ramdisk_execute_command
= NULL
;
1624 prepare_namespace();
1628 * Ok, we have completed the initial bootup, and
1629 * we're essentially up and running. Get rid of the
1630 * initmem segments and start the user-mode stuff..
1632 * rootfs is available now, try loading the public keys
1633 * and default modules
1636 integrity_load_keys();