2 * sysctl.c: General linux system control interface
4 * Begun 24 March 1995, Stephen Tweedie
5 * Added /proc support, Dec 1995
6 * Added bdflush entry and intvec min/max checking, 2/23/96, Tom Dyas.
7 * Added hooks for /proc/sys/net (minor, minor patch), 96/4/1, Mike Shaver.
8 * Added kernel/java-{interpreter,appletviewer}, 96/5/10, Mike Shaver.
9 * Dynamic registration fixes, Stephen Tweedie.
10 * Added kswapd-interval, ctrl-alt-del, printk stuff, 1/8/97, Chris Horn.
11 * Made sysctl support optional via CONFIG_SYSCTL, 1/10/97, Chris
13 * Added proc_doulongvec_ms_jiffies_minmax, 09/08/99, Carlos H. Bauer.
14 * Added proc_doulongvec_minmax, 09/08/99, Carlos H. Bauer.
15 * Changed linked lists to use list.h instead of lists.h, 02/24/00, Bill
17 * The list_for_each() macro wasn't appropriate for the sysctl loop.
18 * Removed it and replaced it with older style, 03/23/00, Bill Wendling
21 #include <linux/module.h>
23 #include <linux/swap.h>
24 #include <linux/slab.h>
25 #include <linux/sysctl.h>
26 #include <linux/signal.h>
27 #include <linux/printk.h>
28 #include <linux/proc_fs.h>
29 #include <linux/security.h>
30 #include <linux/ctype.h>
31 #include <linux/kmemcheck.h>
33 #include <linux/init.h>
34 #include <linux/kernel.h>
35 #include <linux/kobject.h>
36 #include <linux/net.h>
37 #include <linux/sysrq.h>
38 #include <linux/highuid.h>
39 #include <linux/writeback.h>
40 #include <linux/ratelimit.h>
41 #include <linux/compaction.h>
42 #include <linux/hugetlb.h>
43 #include <linux/initrd.h>
44 #include <linux/key.h>
45 #include <linux/times.h>
46 #include <linux/limits.h>
47 #include <linux/dcache.h>
48 #include <linux/dnotify.h>
49 #include <linux/syscalls.h>
50 #include <linux/vmstat.h>
51 #include <linux/nfs_fs.h>
52 #include <linux/acpi.h>
53 #include <linux/reboot.h>
54 #include <linux/ftrace.h>
55 #include <linux/perf_event.h>
56 #include <linux/kprobes.h>
57 #include <linux/pipe_fs_i.h>
58 #include <linux/oom.h>
60 #include <asm/uaccess.h>
61 #include <asm/processor.h>
65 #include <asm/stacktrace.h>
68 #ifdef CONFIG_BSD_PROCESS_ACCT
69 #include <linux/acct.h>
71 #ifdef CONFIG_RT_MUTEXES
72 #include <linux/rtmutex.h>
74 #if defined(CONFIG_PROVE_LOCKING) || defined(CONFIG_LOCK_STAT)
75 #include <linux/lockdep.h>
77 #ifdef CONFIG_CHR_DEV_SG
81 #ifdef CONFIG_LOCKUP_DETECTOR
82 #include <linux/nmi.h>
86 #if defined(CONFIG_SYSCTL)
88 /* External variables not in a header file. */
89 extern int sysctl_overcommit_memory;
90 extern int sysctl_overcommit_ratio;
91 extern int max_threads;
92 extern int core_uses_pid;
93 extern int suid_dumpable;
94 extern char core_pattern[];
95 extern unsigned int core_pipe_limit;
97 extern int min_free_kbytes;
98 extern int pid_max_min, pid_max_max;
99 extern int sysctl_drop_caches;
100 extern int percpu_pagelist_fraction;
101 extern int compat_log;
102 extern int latencytop_enabled;
103 extern int sysctl_nr_open_min, sysctl_nr_open_max;
105 extern int sysctl_nr_trim_pages;
108 extern int blk_iopoll_enabled;
111 /* Constants used for minimum and maximum */
112 #ifdef CONFIG_LOCKUP_DETECTOR
113 static int sixty = 60;
114 static int neg_one = -1;
118 static int __maybe_unused one = 1;
119 static int __maybe_unused two = 2;
120 static unsigned long one_ul = 1;
121 static int one_hundred = 100;
123 static int ten_thousand = 10000;
126 /* this is needed for the proc_doulongvec_minmax of vm_dirty_bytes */
127 static unsigned long dirty_bytes_min = 2 * PAGE_SIZE;
129 /* this is needed for the proc_dointvec_minmax for [fs_]overflow UID and GID */
130 static int maxolduid = 65535;
131 static int minolduid;
132 static int min_percpu_pagelist_fract = 8;
134 static int ngroups_max = NGROUPS_MAX;
136 #ifdef CONFIG_INOTIFY_USER
137 #include <linux/inotify.h>
140 #include <asm/system.h>
143 #ifdef CONFIG_SPARC64
144 extern int sysctl_tsb_ratio;
148 extern int pwrsw_enabled;
149 extern int unaligned_enabled;
153 #ifdef CONFIG_MATHEMU
154 extern int sysctl_ieee_emulation_warnings;
156 extern int sysctl_userprocess_debug;
157 extern int spin_retry;
161 extern int no_unaligned_warning;
162 extern int unaligned_dump_stack;
165 #ifdef CONFIG_PROC_SYSCTL
166 static int proc_do_cad_pid(struct ctl_table *table, int write,
167 void __user *buffer, size_t *lenp, loff_t *ppos);
168 static int proc_taint(struct ctl_table *table, int write,
169 void __user *buffer, size_t *lenp, loff_t *ppos);
172 #ifdef CONFIG_MAGIC_SYSRQ
173 /* Note: sysrq code uses it's own private copy */
174 static int __sysrq_enabled = SYSRQ_DEFAULT_ENABLE;
176 static int sysrq_sysctl_handler(ctl_table *table, int write,
177 void __user *buffer, size_t *lenp,
182 error = proc_dointvec(table, write, buffer, lenp, ppos);
187 sysrq_toggle_support(__sysrq_enabled);
194 static struct ctl_table root_table[];
195 static struct ctl_table_root sysctl_table_root;
196 static struct ctl_table_header root_table_header = {
198 .ctl_table = root_table,
199 .ctl_entry = LIST_HEAD_INIT(sysctl_table_root.default_set.list),}},
200 .root = &sysctl_table_root,
201 .set = &sysctl_table_root.default_set,
203 static struct ctl_table_root sysctl_table_root = {
204 .root_list = LIST_HEAD_INIT(sysctl_table_root.root_list),
205 .default_set.list = LIST_HEAD_INIT(root_table_header.ctl_entry),
208 static struct ctl_table kern_table[];
209 static struct ctl_table vm_table[];
210 static struct ctl_table fs_table[];
211 static struct ctl_table debug_table[];
212 static struct ctl_table dev_table[];
213 extern struct ctl_table random_table[];
215 extern struct ctl_table epoll_table[];
218 #ifdef HAVE_ARCH_PICK_MMAP_LAYOUT
219 int sysctl_legacy_va_layout;
222 /* The default sysctl tables: */
224 static struct ctl_table root_table[] = {
226 .procname = "kernel",
243 .child = debug_table,
253 #ifdef CONFIG_SCHED_DEBUG
254 static int min_sched_granularity_ns = 100000; /* 100 usecs */
255 static int max_sched_granularity_ns = NSEC_PER_SEC; /* 1 second */
256 static int min_wakeup_granularity_ns; /* 0 usecs */
257 static int max_wakeup_granularity_ns = NSEC_PER_SEC; /* 1 second */
258 static int min_sched_tunable_scaling = SCHED_TUNABLESCALING_NONE;
259 static int max_sched_tunable_scaling = SCHED_TUNABLESCALING_END-1;
262 #ifdef CONFIG_COMPACTION
263 static int min_extfrag_threshold;
264 static int max_extfrag_threshold = 1000;
267 static struct ctl_table kern_table[] = {
269 .procname = "sched_child_runs_first",
270 .data = &sysctl_sched_child_runs_first,
271 .maxlen = sizeof(unsigned int),
273 .proc_handler = proc_dointvec,
275 #ifdef CONFIG_SCHED_DEBUG
277 .procname = "sched_min_granularity_ns",
278 .data = &sysctl_sched_min_granularity,
279 .maxlen = sizeof(unsigned int),
281 .proc_handler = sched_proc_update_handler,
282 .extra1 = &min_sched_granularity_ns,
283 .extra2 = &max_sched_granularity_ns,
286 .procname = "sched_latency_ns",
287 .data = &sysctl_sched_latency,
288 .maxlen = sizeof(unsigned int),
290 .proc_handler = sched_proc_update_handler,
291 .extra1 = &min_sched_granularity_ns,
292 .extra2 = &max_sched_granularity_ns,
295 .procname = "sched_wakeup_granularity_ns",
296 .data = &sysctl_sched_wakeup_granularity,
297 .maxlen = sizeof(unsigned int),
299 .proc_handler = sched_proc_update_handler,
300 .extra1 = &min_wakeup_granularity_ns,
301 .extra2 = &max_wakeup_granularity_ns,
304 .procname = "sched_tunable_scaling",
305 .data = &sysctl_sched_tunable_scaling,
306 .maxlen = sizeof(enum sched_tunable_scaling),
308 .proc_handler = sched_proc_update_handler,
309 .extra1 = &min_sched_tunable_scaling,
310 .extra2 = &max_sched_tunable_scaling,
313 .procname = "sched_migration_cost",
314 .data = &sysctl_sched_migration_cost,
315 .maxlen = sizeof(unsigned int),
317 .proc_handler = proc_dointvec,
320 .procname = "sched_nr_migrate",
321 .data = &sysctl_sched_nr_migrate,
322 .maxlen = sizeof(unsigned int),
324 .proc_handler = proc_dointvec,
327 .procname = "sched_time_avg",
328 .data = &sysctl_sched_time_avg,
329 .maxlen = sizeof(unsigned int),
331 .proc_handler = proc_dointvec,
334 .procname = "sched_shares_window",
335 .data = &sysctl_sched_shares_window,
336 .maxlen = sizeof(unsigned int),
338 .proc_handler = proc_dointvec,
341 .procname = "timer_migration",
342 .data = &sysctl_timer_migration,
343 .maxlen = sizeof(unsigned int),
345 .proc_handler = proc_dointvec_minmax,
351 .procname = "sched_rt_period_us",
352 .data = &sysctl_sched_rt_period,
353 .maxlen = sizeof(unsigned int),
355 .proc_handler = sched_rt_handler,
358 .procname = "sched_rt_runtime_us",
359 .data = &sysctl_sched_rt_runtime,
360 .maxlen = sizeof(int),
362 .proc_handler = sched_rt_handler,
364 #ifdef CONFIG_SCHED_AUTOGROUP
366 .procname = "sched_autogroup_enabled",
367 .data = &sysctl_sched_autogroup_enabled,
368 .maxlen = sizeof(unsigned int),
370 .proc_handler = proc_dointvec_minmax,
375 #ifdef CONFIG_PROVE_LOCKING
377 .procname = "prove_locking",
378 .data = &prove_locking,
379 .maxlen = sizeof(int),
381 .proc_handler = proc_dointvec,
384 #ifdef CONFIG_LOCK_STAT
386 .procname = "lock_stat",
388 .maxlen = sizeof(int),
390 .proc_handler = proc_dointvec,
395 .data = &panic_timeout,
396 .maxlen = sizeof(int),
398 .proc_handler = proc_dointvec,
401 .procname = "core_uses_pid",
402 .data = &core_uses_pid,
403 .maxlen = sizeof(int),
405 .proc_handler = proc_dointvec,
408 .procname = "core_pattern",
409 .data = core_pattern,
410 .maxlen = CORENAME_MAX_SIZE,
412 .proc_handler = proc_dostring,
415 .procname = "core_pipe_limit",
416 .data = &core_pipe_limit,
417 .maxlen = sizeof(unsigned int),
419 .proc_handler = proc_dointvec,
421 #ifdef CONFIG_PROC_SYSCTL
423 .procname = "tainted",
424 .maxlen = sizeof(long),
426 .proc_handler = proc_taint,
429 #ifdef CONFIG_LATENCYTOP
431 .procname = "latencytop",
432 .data = &latencytop_enabled,
433 .maxlen = sizeof(int),
435 .proc_handler = proc_dointvec,
438 #ifdef CONFIG_BLK_DEV_INITRD
440 .procname = "real-root-dev",
441 .data = &real_root_dev,
442 .maxlen = sizeof(int),
444 .proc_handler = proc_dointvec,
448 .procname = "print-fatal-signals",
449 .data = &print_fatal_signals,
450 .maxlen = sizeof(int),
452 .proc_handler = proc_dointvec,
456 .procname = "reboot-cmd",
457 .data = reboot_command,
460 .proc_handler = proc_dostring,
463 .procname = "stop-a",
464 .data = &stop_a_enabled,
465 .maxlen = sizeof (int),
467 .proc_handler = proc_dointvec,
470 .procname = "scons-poweroff",
471 .data = &scons_pwroff,
472 .maxlen = sizeof (int),
474 .proc_handler = proc_dointvec,
477 #ifdef CONFIG_SPARC64
479 .procname = "tsb-ratio",
480 .data = &sysctl_tsb_ratio,
481 .maxlen = sizeof (int),
483 .proc_handler = proc_dointvec,
488 .procname = "soft-power",
489 .data = &pwrsw_enabled,
490 .maxlen = sizeof (int),
492 .proc_handler = proc_dointvec,
495 .procname = "unaligned-trap",
496 .data = &unaligned_enabled,
497 .maxlen = sizeof (int),
499 .proc_handler = proc_dointvec,
503 .procname = "ctrl-alt-del",
505 .maxlen = sizeof(int),
507 .proc_handler = proc_dointvec,
509 #ifdef CONFIG_FUNCTION_TRACER
511 .procname = "ftrace_enabled",
512 .data = &ftrace_enabled,
513 .maxlen = sizeof(int),
515 .proc_handler = ftrace_enable_sysctl,
518 #ifdef CONFIG_STACK_TRACER
520 .procname = "stack_tracer_enabled",
521 .data = &stack_tracer_enabled,
522 .maxlen = sizeof(int),
524 .proc_handler = stack_trace_sysctl,
527 #ifdef CONFIG_TRACING
529 .procname = "ftrace_dump_on_oops",
530 .data = &ftrace_dump_on_oops,
531 .maxlen = sizeof(int),
533 .proc_handler = proc_dointvec,
536 #ifdef CONFIG_MODULES
538 .procname = "modprobe",
539 .data = &modprobe_path,
540 .maxlen = KMOD_PATH_LEN,
542 .proc_handler = proc_dostring,
545 .procname = "modules_disabled",
546 .data = &modules_disabled,
547 .maxlen = sizeof(int),
549 /* only handle a transition from default "0" to "1" */
550 .proc_handler = proc_dointvec_minmax,
555 #ifdef CONFIG_HOTPLUG
557 .procname = "hotplug",
558 .data = &uevent_helper,
559 .maxlen = UEVENT_HELPER_PATH_LEN,
561 .proc_handler = proc_dostring,
564 #ifdef CONFIG_CHR_DEV_SG
566 .procname = "sg-big-buff",
567 .data = &sg_big_buff,
568 .maxlen = sizeof (int),
570 .proc_handler = proc_dointvec,
573 #ifdef CONFIG_BSD_PROCESS_ACCT
577 .maxlen = 3*sizeof(int),
579 .proc_handler = proc_dointvec,
582 #ifdef CONFIG_MAGIC_SYSRQ
585 .data = &__sysrq_enabled,
586 .maxlen = sizeof (int),
588 .proc_handler = sysrq_sysctl_handler,
591 #ifdef CONFIG_PROC_SYSCTL
593 .procname = "cad_pid",
595 .maxlen = sizeof (int),
597 .proc_handler = proc_do_cad_pid,
601 .procname = "threads-max",
602 .data = &max_threads,
603 .maxlen = sizeof(int),
605 .proc_handler = proc_dointvec,
608 .procname = "random",
610 .child = random_table,
613 .procname = "overflowuid",
614 .data = &overflowuid,
615 .maxlen = sizeof(int),
617 .proc_handler = proc_dointvec_minmax,
618 .extra1 = &minolduid,
619 .extra2 = &maxolduid,
622 .procname = "overflowgid",
623 .data = &overflowgid,
624 .maxlen = sizeof(int),
626 .proc_handler = proc_dointvec_minmax,
627 .extra1 = &minolduid,
628 .extra2 = &maxolduid,
631 #ifdef CONFIG_MATHEMU
633 .procname = "ieee_emulation_warnings",
634 .data = &sysctl_ieee_emulation_warnings,
635 .maxlen = sizeof(int),
637 .proc_handler = proc_dointvec,
641 .procname = "userprocess_debug",
642 .data = &show_unhandled_signals,
643 .maxlen = sizeof(int),
645 .proc_handler = proc_dointvec,
649 .procname = "pid_max",
651 .maxlen = sizeof (int),
653 .proc_handler = proc_dointvec_minmax,
654 .extra1 = &pid_max_min,
655 .extra2 = &pid_max_max,
658 .procname = "panic_on_oops",
659 .data = &panic_on_oops,
660 .maxlen = sizeof(int),
662 .proc_handler = proc_dointvec,
664 #if defined CONFIG_PRINTK
666 .procname = "printk",
667 .data = &console_loglevel,
668 .maxlen = 4*sizeof(int),
670 .proc_handler = proc_dointvec,
673 .procname = "printk_ratelimit",
674 .data = &printk_ratelimit_state.interval,
675 .maxlen = sizeof(int),
677 .proc_handler = proc_dointvec_jiffies,
680 .procname = "printk_ratelimit_burst",
681 .data = &printk_ratelimit_state.burst,
682 .maxlen = sizeof(int),
684 .proc_handler = proc_dointvec,
687 .procname = "printk_delay",
688 .data = &printk_delay_msec,
689 .maxlen = sizeof(int),
691 .proc_handler = proc_dointvec_minmax,
693 .extra2 = &ten_thousand,
696 .procname = "dmesg_restrict",
697 .data = &dmesg_restrict,
698 .maxlen = sizeof(int),
700 .proc_handler = proc_dointvec_minmax,
705 .procname = "kptr_restrict",
706 .data = &kptr_restrict,
707 .maxlen = sizeof(int),
709 .proc_handler = proc_dointvec_minmax,
715 .procname = "ngroups_max",
716 .data = &ngroups_max,
717 .maxlen = sizeof (int),
719 .proc_handler = proc_dointvec,
721 #if defined(CONFIG_LOCKUP_DETECTOR)
723 .procname = "watchdog",
724 .data = &watchdog_enabled,
725 .maxlen = sizeof (int),
727 .proc_handler = proc_dowatchdog_enabled,
730 .procname = "watchdog_thresh",
731 .data = &softlockup_thresh,
732 .maxlen = sizeof(int),
734 .proc_handler = proc_dowatchdog_thresh,
739 .procname = "softlockup_panic",
740 .data = &softlockup_panic,
741 .maxlen = sizeof(int),
743 .proc_handler = proc_dointvec_minmax,
748 .procname = "nmi_watchdog",
749 .data = &watchdog_enabled,
750 .maxlen = sizeof (int),
752 .proc_handler = proc_dowatchdog_enabled,
755 #if defined(CONFIG_X86_LOCAL_APIC) && defined(CONFIG_X86)
757 .procname = "unknown_nmi_panic",
758 .data = &unknown_nmi_panic,
759 .maxlen = sizeof (int),
761 .proc_handler = proc_dointvec,
764 #if defined(CONFIG_X86)
766 .procname = "panic_on_unrecovered_nmi",
767 .data = &panic_on_unrecovered_nmi,
768 .maxlen = sizeof(int),
770 .proc_handler = proc_dointvec,
773 .procname = "panic_on_io_nmi",
774 .data = &panic_on_io_nmi,
775 .maxlen = sizeof(int),
777 .proc_handler = proc_dointvec,
780 .procname = "bootloader_type",
781 .data = &bootloader_type,
782 .maxlen = sizeof (int),
784 .proc_handler = proc_dointvec,
787 .procname = "bootloader_version",
788 .data = &bootloader_version,
789 .maxlen = sizeof (int),
791 .proc_handler = proc_dointvec,
794 .procname = "kstack_depth_to_print",
795 .data = &kstack_depth_to_print,
796 .maxlen = sizeof(int),
798 .proc_handler = proc_dointvec,
801 .procname = "io_delay_type",
802 .data = &io_delay_type,
803 .maxlen = sizeof(int),
805 .proc_handler = proc_dointvec,
808 #if defined(CONFIG_MMU)
810 .procname = "randomize_va_space",
811 .data = &randomize_va_space,
812 .maxlen = sizeof(int),
814 .proc_handler = proc_dointvec,
817 #if defined(CONFIG_S390) && defined(CONFIG_SMP)
819 .procname = "spin_retry",
821 .maxlen = sizeof (int),
823 .proc_handler = proc_dointvec,
826 #if defined(CONFIG_ACPI_SLEEP) && defined(CONFIG_X86)
828 .procname = "acpi_video_flags",
829 .data = &acpi_realmode_flags,
830 .maxlen = sizeof (unsigned long),
832 .proc_handler = proc_doulongvec_minmax,
837 .procname = "ignore-unaligned-usertrap",
838 .data = &no_unaligned_warning,
839 .maxlen = sizeof (int),
841 .proc_handler = proc_dointvec,
844 .procname = "unaligned-dump-stack",
845 .data = &unaligned_dump_stack,
846 .maxlen = sizeof (int),
848 .proc_handler = proc_dointvec,
851 #ifdef CONFIG_DETECT_HUNG_TASK
853 .procname = "hung_task_panic",
854 .data = &sysctl_hung_task_panic,
855 .maxlen = sizeof(int),
857 .proc_handler = proc_dointvec_minmax,
862 .procname = "hung_task_check_count",
863 .data = &sysctl_hung_task_check_count,
864 .maxlen = sizeof(unsigned long),
866 .proc_handler = proc_doulongvec_minmax,
869 .procname = "hung_task_timeout_secs",
870 .data = &sysctl_hung_task_timeout_secs,
871 .maxlen = sizeof(unsigned long),
873 .proc_handler = proc_dohung_task_timeout_secs,
876 .procname = "hung_task_warnings",
877 .data = &sysctl_hung_task_warnings,
878 .maxlen = sizeof(unsigned long),
880 .proc_handler = proc_doulongvec_minmax,
885 .procname = "compat-log",
887 .maxlen = sizeof (int),
889 .proc_handler = proc_dointvec,
892 #ifdef CONFIG_RT_MUTEXES
894 .procname = "max_lock_depth",
895 .data = &max_lock_depth,
896 .maxlen = sizeof(int),
898 .proc_handler = proc_dointvec,
902 .procname = "poweroff_cmd",
903 .data = &poweroff_cmd,
904 .maxlen = POWEROFF_CMD_PATH_LEN,
906 .proc_handler = proc_dostring,
912 .child = key_sysctls,
915 #ifdef CONFIG_RCU_TORTURE_TEST
917 .procname = "rcutorture_runnable",
918 .data = &rcutorture_runnable,
919 .maxlen = sizeof(int),
921 .proc_handler = proc_dointvec,
924 #ifdef CONFIG_PERF_EVENTS
926 .procname = "perf_event_paranoid",
927 .data = &sysctl_perf_event_paranoid,
928 .maxlen = sizeof(sysctl_perf_event_paranoid),
930 .proc_handler = proc_dointvec,
933 .procname = "perf_event_mlock_kb",
934 .data = &sysctl_perf_event_mlock,
935 .maxlen = sizeof(sysctl_perf_event_mlock),
937 .proc_handler = proc_dointvec,
940 .procname = "perf_event_max_sample_rate",
941 .data = &sysctl_perf_event_sample_rate,
942 .maxlen = sizeof(sysctl_perf_event_sample_rate),
944 .proc_handler = perf_proc_update_handler,
947 #ifdef CONFIG_KMEMCHECK
949 .procname = "kmemcheck",
950 .data = &kmemcheck_enabled,
951 .maxlen = sizeof(int),
953 .proc_handler = proc_dointvec,
958 .procname = "blk_iopoll",
959 .data = &blk_iopoll_enabled,
960 .maxlen = sizeof(int),
962 .proc_handler = proc_dointvec,
968 static struct ctl_table vm_table[] = {
970 .procname = "overcommit_memory",
971 .data = &sysctl_overcommit_memory,
972 .maxlen = sizeof(sysctl_overcommit_memory),
974 .proc_handler = proc_dointvec,
977 .procname = "panic_on_oom",
978 .data = &sysctl_panic_on_oom,
979 .maxlen = sizeof(sysctl_panic_on_oom),
981 .proc_handler = proc_dointvec,
984 .procname = "oom_kill_allocating_task",
985 .data = &sysctl_oom_kill_allocating_task,
986 .maxlen = sizeof(sysctl_oom_kill_allocating_task),
988 .proc_handler = proc_dointvec,
991 .procname = "oom_dump_tasks",
992 .data = &sysctl_oom_dump_tasks,
993 .maxlen = sizeof(sysctl_oom_dump_tasks),
995 .proc_handler = proc_dointvec,
998 .procname = "overcommit_ratio",
999 .data = &sysctl_overcommit_ratio,
1000 .maxlen = sizeof(sysctl_overcommit_ratio),
1002 .proc_handler = proc_dointvec,
1005 .procname = "page-cluster",
1006 .data = &page_cluster,
1007 .maxlen = sizeof(int),
1009 .proc_handler = proc_dointvec,
1012 .procname = "dirty_background_ratio",
1013 .data = &dirty_background_ratio,
1014 .maxlen = sizeof(dirty_background_ratio),
1016 .proc_handler = dirty_background_ratio_handler,
1018 .extra2 = &one_hundred,
1021 .procname = "dirty_background_bytes",
1022 .data = &dirty_background_bytes,
1023 .maxlen = sizeof(dirty_background_bytes),
1025 .proc_handler = dirty_background_bytes_handler,
1029 .procname = "dirty_ratio",
1030 .data = &vm_dirty_ratio,
1031 .maxlen = sizeof(vm_dirty_ratio),
1033 .proc_handler = dirty_ratio_handler,
1035 .extra2 = &one_hundred,
1038 .procname = "dirty_bytes",
1039 .data = &vm_dirty_bytes,
1040 .maxlen = sizeof(vm_dirty_bytes),
1042 .proc_handler = dirty_bytes_handler,
1043 .extra1 = &dirty_bytes_min,
1046 .procname = "dirty_writeback_centisecs",
1047 .data = &dirty_writeback_interval,
1048 .maxlen = sizeof(dirty_writeback_interval),
1050 .proc_handler = dirty_writeback_centisecs_handler,
1053 .procname = "dirty_expire_centisecs",
1054 .data = &dirty_expire_interval,
1055 .maxlen = sizeof(dirty_expire_interval),
1057 .proc_handler = proc_dointvec,
1060 .procname = "nr_pdflush_threads",
1061 .data = &nr_pdflush_threads,
1062 .maxlen = sizeof nr_pdflush_threads,
1063 .mode = 0444 /* read-only*/,
1064 .proc_handler = proc_dointvec,
1067 .procname = "swappiness",
1068 .data = &vm_swappiness,
1069 .maxlen = sizeof(vm_swappiness),
1071 .proc_handler = proc_dointvec_minmax,
1073 .extra2 = &one_hundred,
1075 #ifdef CONFIG_HUGETLB_PAGE
1077 .procname = "nr_hugepages",
1079 .maxlen = sizeof(unsigned long),
1081 .proc_handler = hugetlb_sysctl_handler,
1082 .extra1 = (void *)&hugetlb_zero,
1083 .extra2 = (void *)&hugetlb_infinity,
1087 .procname = "nr_hugepages_mempolicy",
1089 .maxlen = sizeof(unsigned long),
1091 .proc_handler = &hugetlb_mempolicy_sysctl_handler,
1092 .extra1 = (void *)&hugetlb_zero,
1093 .extra2 = (void *)&hugetlb_infinity,
1097 .procname = "hugetlb_shm_group",
1098 .data = &sysctl_hugetlb_shm_group,
1099 .maxlen = sizeof(gid_t),
1101 .proc_handler = proc_dointvec,
1104 .procname = "hugepages_treat_as_movable",
1105 .data = &hugepages_treat_as_movable,
1106 .maxlen = sizeof(int),
1108 .proc_handler = hugetlb_treat_movable_handler,
1111 .procname = "nr_overcommit_hugepages",
1113 .maxlen = sizeof(unsigned long),
1115 .proc_handler = hugetlb_overcommit_handler,
1116 .extra1 = (void *)&hugetlb_zero,
1117 .extra2 = (void *)&hugetlb_infinity,
1121 .procname = "lowmem_reserve_ratio",
1122 .data = &sysctl_lowmem_reserve_ratio,
1123 .maxlen = sizeof(sysctl_lowmem_reserve_ratio),
1125 .proc_handler = lowmem_reserve_ratio_sysctl_handler,
1128 .procname = "drop_caches",
1129 .data = &sysctl_drop_caches,
1130 .maxlen = sizeof(int),
1132 .proc_handler = drop_caches_sysctl_handler,
1134 #ifdef CONFIG_COMPACTION
1136 .procname = "compact_memory",
1137 .data = &sysctl_compact_memory,
1138 .maxlen = sizeof(int),
1140 .proc_handler = sysctl_compaction_handler,
1143 .procname = "extfrag_threshold",
1144 .data = &sysctl_extfrag_threshold,
1145 .maxlen = sizeof(int),
1147 .proc_handler = sysctl_extfrag_handler,
1148 .extra1 = &min_extfrag_threshold,
1149 .extra2 = &max_extfrag_threshold,
1152 #endif /* CONFIG_COMPACTION */
1154 .procname = "min_free_kbytes",
1155 .data = &min_free_kbytes,
1156 .maxlen = sizeof(min_free_kbytes),
1158 .proc_handler = min_free_kbytes_sysctl_handler,
1162 .procname = "percpu_pagelist_fraction",
1163 .data = &percpu_pagelist_fraction,
1164 .maxlen = sizeof(percpu_pagelist_fraction),
1166 .proc_handler = percpu_pagelist_fraction_sysctl_handler,
1167 .extra1 = &min_percpu_pagelist_fract,
1171 .procname = "max_map_count",
1172 .data = &sysctl_max_map_count,
1173 .maxlen = sizeof(sysctl_max_map_count),
1175 .proc_handler = proc_dointvec_minmax,
1180 .procname = "nr_trim_pages",
1181 .data = &sysctl_nr_trim_pages,
1182 .maxlen = sizeof(sysctl_nr_trim_pages),
1184 .proc_handler = proc_dointvec_minmax,
1189 .procname = "laptop_mode",
1190 .data = &laptop_mode,
1191 .maxlen = sizeof(laptop_mode),
1193 .proc_handler = proc_dointvec_jiffies,
1196 .procname = "block_dump",
1197 .data = &block_dump,
1198 .maxlen = sizeof(block_dump),
1200 .proc_handler = proc_dointvec,
1204 .procname = "vfs_cache_pressure",
1205 .data = &sysctl_vfs_cache_pressure,
1206 .maxlen = sizeof(sysctl_vfs_cache_pressure),
1208 .proc_handler = proc_dointvec,
1211 #ifdef HAVE_ARCH_PICK_MMAP_LAYOUT
1213 .procname = "legacy_va_layout",
1214 .data = &sysctl_legacy_va_layout,
1215 .maxlen = sizeof(sysctl_legacy_va_layout),
1217 .proc_handler = proc_dointvec,
1223 .procname = "zone_reclaim_mode",
1224 .data = &zone_reclaim_mode,
1225 .maxlen = sizeof(zone_reclaim_mode),
1227 .proc_handler = proc_dointvec,
1231 .procname = "min_unmapped_ratio",
1232 .data = &sysctl_min_unmapped_ratio,
1233 .maxlen = sizeof(sysctl_min_unmapped_ratio),
1235 .proc_handler = sysctl_min_unmapped_ratio_sysctl_handler,
1237 .extra2 = &one_hundred,
1240 .procname = "min_slab_ratio",
1241 .data = &sysctl_min_slab_ratio,
1242 .maxlen = sizeof(sysctl_min_slab_ratio),
1244 .proc_handler = sysctl_min_slab_ratio_sysctl_handler,
1246 .extra2 = &one_hundred,
1251 .procname = "stat_interval",
1252 .data = &sysctl_stat_interval,
1253 .maxlen = sizeof(sysctl_stat_interval),
1255 .proc_handler = proc_dointvec_jiffies,
1260 .procname = "mmap_min_addr",
1261 .data = &dac_mmap_min_addr,
1262 .maxlen = sizeof(unsigned long),
1264 .proc_handler = mmap_min_addr_handler,
1269 .procname = "numa_zonelist_order",
1270 .data = &numa_zonelist_order,
1271 .maxlen = NUMA_ZONELIST_ORDER_LEN,
1273 .proc_handler = numa_zonelist_order_handler,
1276 #if (defined(CONFIG_X86_32) && !defined(CONFIG_UML))|| \
1277 (defined(CONFIG_SUPERH) && defined(CONFIG_VSYSCALL))
1279 .procname = "vdso_enabled",
1280 .data = &vdso_enabled,
1281 .maxlen = sizeof(vdso_enabled),
1283 .proc_handler = proc_dointvec,
1287 #ifdef CONFIG_HIGHMEM
1289 .procname = "highmem_is_dirtyable",
1290 .data = &vm_highmem_is_dirtyable,
1291 .maxlen = sizeof(vm_highmem_is_dirtyable),
1293 .proc_handler = proc_dointvec_minmax,
1299 .procname = "scan_unevictable_pages",
1300 .data = &scan_unevictable_pages,
1301 .maxlen = sizeof(scan_unevictable_pages),
1303 .proc_handler = scan_unevictable_handler,
1305 #ifdef CONFIG_MEMORY_FAILURE
1307 .procname = "memory_failure_early_kill",
1308 .data = &sysctl_memory_failure_early_kill,
1309 .maxlen = sizeof(sysctl_memory_failure_early_kill),
1311 .proc_handler = proc_dointvec_minmax,
1316 .procname = "memory_failure_recovery",
1317 .data = &sysctl_memory_failure_recovery,
1318 .maxlen = sizeof(sysctl_memory_failure_recovery),
1320 .proc_handler = proc_dointvec_minmax,
1328 #if defined(CONFIG_BINFMT_MISC) || defined(CONFIG_BINFMT_MISC_MODULE)
1329 static struct ctl_table binfmt_misc_table[] = {
1334 static struct ctl_table fs_table[] = {
1336 .procname = "inode-nr",
1337 .data = &inodes_stat,
1338 .maxlen = 2*sizeof(int),
1340 .proc_handler = proc_nr_inodes,
1343 .procname = "inode-state",
1344 .data = &inodes_stat,
1345 .maxlen = 7*sizeof(int),
1347 .proc_handler = proc_nr_inodes,
1350 .procname = "file-nr",
1351 .data = &files_stat,
1352 .maxlen = sizeof(files_stat),
1354 .proc_handler = proc_nr_files,
1357 .procname = "file-max",
1358 .data = &files_stat.max_files,
1359 .maxlen = sizeof(files_stat.max_files),
1361 .proc_handler = proc_doulongvec_minmax,
1364 .procname = "nr_open",
1365 .data = &sysctl_nr_open,
1366 .maxlen = sizeof(int),
1368 .proc_handler = proc_dointvec_minmax,
1369 .extra1 = &sysctl_nr_open_min,
1370 .extra2 = &sysctl_nr_open_max,
1373 .procname = "dentry-state",
1374 .data = &dentry_stat,
1375 .maxlen = 6*sizeof(int),
1377 .proc_handler = proc_nr_dentry,
1380 .procname = "overflowuid",
1381 .data = &fs_overflowuid,
1382 .maxlen = sizeof(int),
1384 .proc_handler = proc_dointvec_minmax,
1385 .extra1 = &minolduid,
1386 .extra2 = &maxolduid,
1389 .procname = "overflowgid",
1390 .data = &fs_overflowgid,
1391 .maxlen = sizeof(int),
1393 .proc_handler = proc_dointvec_minmax,
1394 .extra1 = &minolduid,
1395 .extra2 = &maxolduid,
1397 #ifdef CONFIG_FILE_LOCKING
1399 .procname = "leases-enable",
1400 .data = &leases_enable,
1401 .maxlen = sizeof(int),
1403 .proc_handler = proc_dointvec,
1406 #ifdef CONFIG_DNOTIFY
1408 .procname = "dir-notify-enable",
1409 .data = &dir_notify_enable,
1410 .maxlen = sizeof(int),
1412 .proc_handler = proc_dointvec,
1416 #ifdef CONFIG_FILE_LOCKING
1418 .procname = "lease-break-time",
1419 .data = &lease_break_time,
1420 .maxlen = sizeof(int),
1422 .proc_handler = proc_dointvec,
1427 .procname = "aio-nr",
1429 .maxlen = sizeof(aio_nr),
1431 .proc_handler = proc_doulongvec_minmax,
1434 .procname = "aio-max-nr",
1435 .data = &aio_max_nr,
1436 .maxlen = sizeof(aio_max_nr),
1438 .proc_handler = proc_doulongvec_minmax,
1440 #endif /* CONFIG_AIO */
1441 #ifdef CONFIG_INOTIFY_USER
1443 .procname = "inotify",
1445 .child = inotify_table,
1450 .procname = "epoll",
1452 .child = epoll_table,
1457 .procname = "suid_dumpable",
1458 .data = &suid_dumpable,
1459 .maxlen = sizeof(int),
1461 .proc_handler = proc_dointvec_minmax,
1465 #if defined(CONFIG_BINFMT_MISC) || defined(CONFIG_BINFMT_MISC_MODULE)
1467 .procname = "binfmt_misc",
1469 .child = binfmt_misc_table,
1473 .procname = "pipe-max-size",
1474 .data = &pipe_max_size,
1475 .maxlen = sizeof(int),
1477 .proc_handler = &pipe_proc_fn,
1478 .extra1 = &pipe_min_size,
1483 static struct ctl_table debug_table[] = {
1484 #if defined(CONFIG_X86) || defined(CONFIG_PPC) || defined(CONFIG_SPARC) || \
1485 defined(CONFIG_S390)
1487 .procname = "exception-trace",
1488 .data = &show_unhandled_signals,
1489 .maxlen = sizeof(int),
1491 .proc_handler = proc_dointvec
1494 #if defined(CONFIG_OPTPROBES)
1496 .procname = "kprobes-optimization",
1497 .data = &sysctl_kprobes_optimization,
1498 .maxlen = sizeof(int),
1500 .proc_handler = proc_kprobes_optimization_handler,
1508 static struct ctl_table dev_table[] = {
1512 static DEFINE_SPINLOCK(sysctl_lock);
1514 /* called under sysctl_lock */
1515 static int use_table(struct ctl_table_header *p)
1517 if (unlikely(p->unregistering))
1523 /* called under sysctl_lock */
1524 static void unuse_table(struct ctl_table_header *p)
1527 if (unlikely(p->unregistering))
1528 complete(p->unregistering);
1531 /* called under sysctl_lock, will reacquire if has to wait */
1532 static void start_unregistering(struct ctl_table_header *p)
1535 * if p->used is 0, nobody will ever touch that entry again;
1536 * we'll eliminate all paths to it before dropping sysctl_lock
1538 if (unlikely(p->used)) {
1539 struct completion wait;
1540 init_completion(&wait);
1541 p->unregistering = &wait;
1542 spin_unlock(&sysctl_lock);
1543 wait_for_completion(&wait);
1544 spin_lock(&sysctl_lock);
1546 /* anything non-NULL; we'll never dereference it */
1547 p->unregistering = ERR_PTR(-EINVAL);
1550 * do not remove from the list until nobody holds it; walking the
1551 * list in do_sysctl() relies on that.
1553 list_del_init(&p->ctl_entry);
1556 void sysctl_head_get(struct ctl_table_header *head)
1558 spin_lock(&sysctl_lock);
1560 spin_unlock(&sysctl_lock);
1563 static void free_head(struct rcu_head *rcu)
1565 kfree(container_of(rcu, struct ctl_table_header, rcu));
1568 void sysctl_head_put(struct ctl_table_header *head)
1570 spin_lock(&sysctl_lock);
1572 call_rcu(&head->rcu, free_head);
1573 spin_unlock(&sysctl_lock);
1576 struct ctl_table_header *sysctl_head_grab(struct ctl_table_header *head)
1580 spin_lock(&sysctl_lock);
1581 if (!use_table(head))
1582 head = ERR_PTR(-ENOENT);
1583 spin_unlock(&sysctl_lock);
1587 void sysctl_head_finish(struct ctl_table_header *head)
1591 spin_lock(&sysctl_lock);
1593 spin_unlock(&sysctl_lock);
1596 static struct ctl_table_set *
1597 lookup_header_set(struct ctl_table_root *root, struct nsproxy *namespaces)
1599 struct ctl_table_set *set = &root->default_set;
1601 set = root->lookup(root, namespaces);
1605 static struct list_head *
1606 lookup_header_list(struct ctl_table_root *root, struct nsproxy *namespaces)
1608 struct ctl_table_set *set = lookup_header_set(root, namespaces);
1612 struct ctl_table_header *__sysctl_head_next(struct nsproxy *namespaces,
1613 struct ctl_table_header *prev)
1615 struct ctl_table_root *root;
1616 struct list_head *header_list;
1617 struct ctl_table_header *head;
1618 struct list_head *tmp;
1620 spin_lock(&sysctl_lock);
1623 tmp = &prev->ctl_entry;
1627 tmp = &root_table_header.ctl_entry;
1629 head = list_entry(tmp, struct ctl_table_header, ctl_entry);
1631 if (!use_table(head))
1633 spin_unlock(&sysctl_lock);
1638 header_list = lookup_header_list(root, namespaces);
1639 if (tmp != header_list)
1643 root = list_entry(root->root_list.next,
1644 struct ctl_table_root, root_list);
1645 if (root == &sysctl_table_root)
1647 header_list = lookup_header_list(root, namespaces);
1648 } while (list_empty(header_list));
1649 tmp = header_list->next;
1652 spin_unlock(&sysctl_lock);
1656 struct ctl_table_header *sysctl_head_next(struct ctl_table_header *prev)
1658 return __sysctl_head_next(current->nsproxy, prev);
1661 void register_sysctl_root(struct ctl_table_root *root)
1663 spin_lock(&sysctl_lock);
1664 list_add_tail(&root->root_list, &sysctl_table_root.root_list);
1665 spin_unlock(&sysctl_lock);
1669 * sysctl_perm does NOT grant the superuser all rights automatically, because
1670 * some sysctl variables are readonly even to root.
1673 static int test_perm(int mode, int op)
1675 if (!current_euid())
1677 else if (in_egroup_p(0))
1679 if ((op & ~mode & (MAY_READ|MAY_WRITE|MAY_EXEC)) == 0)
1684 int sysctl_perm(struct ctl_table_root *root, struct ctl_table *table, int op)
1688 if (root->permissions)
1689 mode = root->permissions(root, current->nsproxy, table);
1693 return test_perm(mode, op);
1696 static void sysctl_set_parent(struct ctl_table *parent, struct ctl_table *table)
1698 for (; table->procname; table++) {
1699 table->parent = parent;
1701 sysctl_set_parent(table, table->child);
1705 static __init int sysctl_init(void)
1707 sysctl_set_parent(NULL, root_table);
1708 #ifdef CONFIG_SYSCTL_SYSCALL_CHECK
1709 sysctl_check_table(current->nsproxy, root_table);
1714 core_initcall(sysctl_init);
1716 static struct ctl_table *is_branch_in(struct ctl_table *branch,
1717 struct ctl_table *table)
1719 struct ctl_table *p;
1720 const char *s = branch->procname;
1722 /* branch should have named subdirectory as its first element */
1723 if (!s || !branch->child)
1726 /* ... and nothing else */
1727 if (branch[1].procname)
1730 /* table should contain subdirectory with the same name */
1731 for (p = table; p->procname; p++) {
1734 if (p->procname && strcmp(p->procname, s) == 0)
1740 /* see if attaching q to p would be an improvement */
1741 static void try_attach(struct ctl_table_header *p, struct ctl_table_header *q)
1743 struct ctl_table *to = p->ctl_table, *by = q->ctl_table;
1744 struct ctl_table *next;
1746 int not_in_parent = !p->attached_by;
1748 while ((next = is_branch_in(by, to)) != NULL) {
1749 if (by == q->attached_by)
1751 if (to == p->attached_by)
1757 if (is_better && not_in_parent) {
1758 q->attached_by = by;
1759 q->attached_to = to;
1765 * __register_sysctl_paths - register a sysctl hierarchy
1766 * @root: List of sysctl headers to register on
1767 * @namespaces: Data to compute which lists of sysctl entries are visible
1768 * @path: The path to the directory the sysctl table is in.
1769 * @table: the top-level table structure
1771 * Register a sysctl table hierarchy. @table should be a filled in ctl_table
1772 * array. A completely 0 filled entry terminates the table.
1774 * The members of the &struct ctl_table structure are used as follows:
1776 * procname - the name of the sysctl file under /proc/sys. Set to %NULL to not
1777 * enter a sysctl file
1779 * data - a pointer to data for use by proc_handler
1781 * maxlen - the maximum size in bytes of the data
1783 * mode - the file permissions for the /proc/sys file, and for sysctl(2)
1785 * child - a pointer to the child sysctl table if this entry is a directory, or
1788 * proc_handler - the text handler routine (described below)
1790 * de - for internal use by the sysctl routines
1792 * extra1, extra2 - extra pointers usable by the proc handler routines
1794 * Leaf nodes in the sysctl tree will be represented by a single file
1795 * under /proc; non-leaf nodes will be represented by directories.
1797 * sysctl(2) can automatically manage read and write requests through
1798 * the sysctl table. The data and maxlen fields of the ctl_table
1799 * struct enable minimal validation of the values being written to be
1800 * performed, and the mode field allows minimal authentication.
1802 * There must be a proc_handler routine for any terminal nodes
1803 * mirrored under /proc/sys (non-terminals are handled by a built-in
1804 * directory handler). Several default handlers are available to
1805 * cover common cases -
1807 * proc_dostring(), proc_dointvec(), proc_dointvec_jiffies(),
1808 * proc_dointvec_userhz_jiffies(), proc_dointvec_minmax(),
1809 * proc_doulongvec_ms_jiffies_minmax(), proc_doulongvec_minmax()
1811 * It is the handler's job to read the input buffer from user memory
1812 * and process it. The handler should return 0 on success.
1814 * This routine returns %NULL on a failure to register, and a pointer
1815 * to the table header on success.
1817 struct ctl_table_header *__register_sysctl_paths(
1818 struct ctl_table_root *root,
1819 struct nsproxy *namespaces,
1820 const struct ctl_path *path, struct ctl_table *table)
1822 struct ctl_table_header *header;
1823 struct ctl_table *new, **prevp;
1824 unsigned int n, npath;
1825 struct ctl_table_set *set;
1827 /* Count the path components */
1828 for (npath = 0; path[npath].procname; ++npath)
1832 * For each path component, allocate a 2-element ctl_table array.
1833 * The first array element will be filled with the sysctl entry
1834 * for this, the second will be the sentinel (procname == 0).
1836 * We allocate everything in one go so that we don't have to
1837 * worry about freeing additional memory in unregister_sysctl_table.
1839 header = kzalloc(sizeof(struct ctl_table_header) +
1840 (2 * npath * sizeof(struct ctl_table)), GFP_KERNEL);
1844 new = (struct ctl_table *) (header + 1);
1846 /* Now connect the dots */
1847 prevp = &header->ctl_table;
1848 for (n = 0; n < npath; ++n, ++path) {
1849 /* Copy the procname */
1850 new->procname = path->procname;
1854 prevp = &new->child;
1859 header->ctl_table_arg = table;
1861 INIT_LIST_HEAD(&header->ctl_entry);
1863 header->unregistering = NULL;
1864 header->root = root;
1865 sysctl_set_parent(NULL, header->ctl_table);
1867 #ifdef CONFIG_SYSCTL_SYSCALL_CHECK
1868 if (sysctl_check_table(namespaces, header->ctl_table)) {
1873 spin_lock(&sysctl_lock);
1874 header->set = lookup_header_set(root, namespaces);
1875 header->attached_by = header->ctl_table;
1876 header->attached_to = root_table;
1877 header->parent = &root_table_header;
1878 for (set = header->set; set; set = set->parent) {
1879 struct ctl_table_header *p;
1880 list_for_each_entry(p, &set->list, ctl_entry) {
1881 if (p->unregistering)
1883 try_attach(p, header);
1886 header->parent->count++;
1887 list_add_tail(&header->ctl_entry, &header->set->list);
1888 spin_unlock(&sysctl_lock);
1894 * register_sysctl_table_path - register a sysctl table hierarchy
1895 * @path: The path to the directory the sysctl table is in.
1896 * @table: the top-level table structure
1898 * Register a sysctl table hierarchy. @table should be a filled in ctl_table
1899 * array. A completely 0 filled entry terminates the table.
1901 * See __register_sysctl_paths for more details.
1903 struct ctl_table_header *register_sysctl_paths(const struct ctl_path *path,
1904 struct ctl_table *table)
1906 return __register_sysctl_paths(&sysctl_table_root, current->nsproxy,
1911 * register_sysctl_table - register a sysctl table hierarchy
1912 * @table: the top-level table structure
1914 * Register a sysctl table hierarchy. @table should be a filled in ctl_table
1915 * array. A completely 0 filled entry terminates the table.
1917 * See register_sysctl_paths for more details.
1919 struct ctl_table_header *register_sysctl_table(struct ctl_table *table)
1921 static const struct ctl_path null_path[] = { {} };
1923 return register_sysctl_paths(null_path, table);
1927 * unregister_sysctl_table - unregister a sysctl table hierarchy
1928 * @header: the header returned from register_sysctl_table
1930 * Unregisters the sysctl table and all children. proc entries may not
1931 * actually be removed until they are no longer used by anyone.
1933 void unregister_sysctl_table(struct ctl_table_header * header)
1940 spin_lock(&sysctl_lock);
1941 start_unregistering(header);
1942 if (!--header->parent->count) {
1944 call_rcu(&header->parent->rcu, free_head);
1946 if (!--header->count)
1947 call_rcu(&header->rcu, free_head);
1948 spin_unlock(&sysctl_lock);
1951 int sysctl_is_seen(struct ctl_table_header *p)
1953 struct ctl_table_set *set = p->set;
1955 spin_lock(&sysctl_lock);
1956 if (p->unregistering)
1958 else if (!set->is_seen)
1961 res = set->is_seen(set);
1962 spin_unlock(&sysctl_lock);
1966 void setup_sysctl_set(struct ctl_table_set *p,
1967 struct ctl_table_set *parent,
1968 int (*is_seen)(struct ctl_table_set *))
1970 INIT_LIST_HEAD(&p->list);
1971 p->parent = parent ? parent : &sysctl_table_root.default_set;
1972 p->is_seen = is_seen;
1975 #else /* !CONFIG_SYSCTL */
1976 struct ctl_table_header *register_sysctl_table(struct ctl_table * table)
1981 struct ctl_table_header *register_sysctl_paths(const struct ctl_path *path,
1982 struct ctl_table *table)
1987 void unregister_sysctl_table(struct ctl_table_header * table)
1991 void setup_sysctl_set(struct ctl_table_set *p,
1992 struct ctl_table_set *parent,
1993 int (*is_seen)(struct ctl_table_set *))
1997 void sysctl_head_put(struct ctl_table_header *head)
2001 #endif /* CONFIG_SYSCTL */
2007 #ifdef CONFIG_PROC_SYSCTL
2009 static int _proc_do_string(void* data, int maxlen, int write,
2010 void __user *buffer,
2011 size_t *lenp, loff_t *ppos)
2017 if (!data || !maxlen || !*lenp) {
2025 while (len < *lenp) {
2026 if (get_user(c, p++))
2028 if (c == 0 || c == '\n')
2034 if(copy_from_user(data, buffer, len))
2036 ((char *) data)[len] = 0;
2054 if(copy_to_user(buffer, data, len))
2057 if(put_user('\n', ((char __user *) buffer) + len))
2068 * proc_dostring - read a string sysctl
2069 * @table: the sysctl table
2070 * @write: %TRUE if this is a write to the sysctl file
2071 * @buffer: the user buffer
2072 * @lenp: the size of the user buffer
2073 * @ppos: file position
2075 * Reads/writes a string from/to the user buffer. If the kernel
2076 * buffer provided is not large enough to hold the string, the
2077 * string is truncated. The copied string is %NULL-terminated.
2078 * If the string is being read by the user process, it is copied
2079 * and a newline '\n' is added. It is truncated if the buffer is
2082 * Returns 0 on success.
2084 int proc_dostring(struct ctl_table *table, int write,
2085 void __user *buffer, size_t *lenp, loff_t *ppos)
2087 return _proc_do_string(table->data, table->maxlen, write,
2088 buffer, lenp, ppos);
2091 static size_t proc_skip_spaces(char **buf)
2094 char *tmp = skip_spaces(*buf);
2100 static void proc_skip_char(char **buf, size_t *size, const char v)
2110 #define TMPBUFLEN 22
2112 * proc_get_long - reads an ASCII formatted integer from a user buffer
2114 * @buf: a kernel buffer
2115 * @size: size of the kernel buffer
2116 * @val: this is where the number will be stored
2117 * @neg: set to %TRUE if number is negative
2118 * @perm_tr: a vector which contains the allowed trailers
2119 * @perm_tr_len: size of the perm_tr vector
2120 * @tr: pointer to store the trailer character
2122 * In case of success %0 is returned and @buf and @size are updated with
2123 * the amount of bytes read. If @tr is non-NULL and a trailing
2124 * character exists (size is non-zero after returning from this
2125 * function), @tr is updated with the trailing character.
2127 static int proc_get_long(char **buf, size_t *size,
2128 unsigned long *val, bool *neg,
2129 const char *perm_tr, unsigned perm_tr_len, char *tr)
2132 char *p, tmp[TMPBUFLEN];
2138 if (len > TMPBUFLEN - 1)
2139 len = TMPBUFLEN - 1;
2141 memcpy(tmp, *buf, len);
2145 if (*p == '-' && *size > 1) {
2153 *val = simple_strtoul(p, &p, 0);
2157 /* We don't know if the next char is whitespace thus we may accept
2158 * invalid integers (e.g. 1234...a) or two integers instead of one
2159 * (e.g. 123...1). So lets not allow such large numbers. */
2160 if (len == TMPBUFLEN - 1)
2163 if (len < *size && perm_tr_len && !memchr(perm_tr, *p, perm_tr_len))
2166 if (tr && (len < *size))
2176 * proc_put_long - converts an integer to a decimal ASCII formatted string
2178 * @buf: the user buffer
2179 * @size: the size of the user buffer
2180 * @val: the integer to be converted
2181 * @neg: sign of the number, %TRUE for negative
2183 * In case of success %0 is returned and @buf and @size are updated with
2184 * the amount of bytes written.
2186 static int proc_put_long(void __user **buf, size_t *size, unsigned long val,
2190 char tmp[TMPBUFLEN], *p = tmp;
2192 sprintf(p, "%s%lu", neg ? "-" : "", val);
2196 if (copy_to_user(*buf, tmp, len))
2204 static int proc_put_char(void __user **buf, size_t *size, char c)
2207 char __user **buffer = (char __user **)buf;
2208 if (put_user(c, *buffer))
2210 (*size)--, (*buffer)++;
2216 static int do_proc_dointvec_conv(bool *negp, unsigned long *lvalp,
2218 int write, void *data)
2221 *valp = *negp ? -*lvalp : *lvalp;
2226 *lvalp = (unsigned long)-val;
2229 *lvalp = (unsigned long)val;
2235 static const char proc_wspace_sep[] = { ' ', '\t', '\n' };
2237 static int __do_proc_dointvec(void *tbl_data, struct ctl_table *table,
2238 int write, void __user *buffer,
2239 size_t *lenp, loff_t *ppos,
2240 int (*conv)(bool *negp, unsigned long *lvalp, int *valp,
2241 int write, void *data),
2244 int *i, vleft, first = 1, err = 0;
2245 unsigned long page = 0;
2249 if (!tbl_data || !table->maxlen || !*lenp || (*ppos && !write)) {
2254 i = (int *) tbl_data;
2255 vleft = table->maxlen / sizeof(*i);
2259 conv = do_proc_dointvec_conv;
2262 if (left > PAGE_SIZE - 1)
2263 left = PAGE_SIZE - 1;
2264 page = __get_free_page(GFP_TEMPORARY);
2265 kbuf = (char *) page;
2268 if (copy_from_user(kbuf, buffer, left)) {
2275 for (; left && vleft--; i++, first=0) {
2280 left -= proc_skip_spaces(&kbuf);
2284 err = proc_get_long(&kbuf, &left, &lval, &neg,
2286 sizeof(proc_wspace_sep), NULL);
2289 if (conv(&neg, &lval, i, 1, data)) {
2294 if (conv(&neg, &lval, i, 0, data)) {
2299 err = proc_put_char(&buffer, &left, '\t');
2302 err = proc_put_long(&buffer, &left, lval, neg);
2308 if (!write && !first && left && !err)
2309 err = proc_put_char(&buffer, &left, '\n');
2310 if (write && !err && left)
2311 left -= proc_skip_spaces(&kbuf);
2316 return err ? : -EINVAL;
2323 static int do_proc_dointvec(struct ctl_table *table, int write,
2324 void __user *buffer, size_t *lenp, loff_t *ppos,
2325 int (*conv)(bool *negp, unsigned long *lvalp, int *valp,
2326 int write, void *data),
2329 return __do_proc_dointvec(table->data, table, write,
2330 buffer, lenp, ppos, conv, data);
2334 * proc_dointvec - read a vector of integers
2335 * @table: the sysctl table
2336 * @write: %TRUE if this is a write to the sysctl file
2337 * @buffer: the user buffer
2338 * @lenp: the size of the user buffer
2339 * @ppos: file position
2341 * Reads/writes up to table->maxlen/sizeof(unsigned int) integer
2342 * values from/to the user buffer, treated as an ASCII string.
2344 * Returns 0 on success.
2346 int proc_dointvec(struct ctl_table *table, int write,
2347 void __user *buffer, size_t *lenp, loff_t *ppos)
2349 return do_proc_dointvec(table,write,buffer,lenp,ppos,
2354 * Taint values can only be increased
2355 * This means we can safely use a temporary.
2357 static int proc_taint(struct ctl_table *table, int write,
2358 void __user *buffer, size_t *lenp, loff_t *ppos)
2361 unsigned long tmptaint = get_taint();
2364 if (write && !capable(CAP_SYS_ADMIN))
2369 err = proc_doulongvec_minmax(&t, write, buffer, lenp, ppos);
2375 * Poor man's atomic or. Not worth adding a primitive
2376 * to everyone's atomic.h for this
2379 for (i = 0; i < BITS_PER_LONG && tmptaint >> i; i++) {
2380 if ((tmptaint >> i) & 1)
2388 struct do_proc_dointvec_minmax_conv_param {
2393 static int do_proc_dointvec_minmax_conv(bool *negp, unsigned long *lvalp,
2395 int write, void *data)
2397 struct do_proc_dointvec_minmax_conv_param *param = data;
2399 int val = *negp ? -*lvalp : *lvalp;
2400 if ((param->min && *param->min > val) ||
2401 (param->max && *param->max < val))
2408 *lvalp = (unsigned long)-val;
2411 *lvalp = (unsigned long)val;
2418 * proc_dointvec_minmax - read a vector of integers with min/max values
2419 * @table: the sysctl table
2420 * @write: %TRUE if this is a write to the sysctl file
2421 * @buffer: the user buffer
2422 * @lenp: the size of the user buffer
2423 * @ppos: file position
2425 * Reads/writes up to table->maxlen/sizeof(unsigned int) integer
2426 * values from/to the user buffer, treated as an ASCII string.
2428 * This routine will ensure the values are within the range specified by
2429 * table->extra1 (min) and table->extra2 (max).
2431 * Returns 0 on success.
2433 int proc_dointvec_minmax(struct ctl_table *table, int write,
2434 void __user *buffer, size_t *lenp, loff_t *ppos)
2436 struct do_proc_dointvec_minmax_conv_param param = {
2437 .min = (int *) table->extra1,
2438 .max = (int *) table->extra2,
2440 return do_proc_dointvec(table, write, buffer, lenp, ppos,
2441 do_proc_dointvec_minmax_conv, ¶m);
2444 static int __do_proc_doulongvec_minmax(void *data, struct ctl_table *table, int write,
2445 void __user *buffer,
2446 size_t *lenp, loff_t *ppos,
2447 unsigned long convmul,
2448 unsigned long convdiv)
2450 unsigned long *i, *min, *max;
2451 int vleft, first = 1, err = 0;
2452 unsigned long page = 0;
2456 if (!data || !table->maxlen || !*lenp || (*ppos && !write)) {
2461 i = (unsigned long *) data;
2462 min = (unsigned long *) table->extra1;
2463 max = (unsigned long *) table->extra2;
2464 vleft = table->maxlen / sizeof(unsigned long);
2468 if (left > PAGE_SIZE - 1)
2469 left = PAGE_SIZE - 1;
2470 page = __get_free_page(GFP_TEMPORARY);
2471 kbuf = (char *) page;
2474 if (copy_from_user(kbuf, buffer, left)) {
2481 for (; left && vleft--; i++, first = 0) {
2487 left -= proc_skip_spaces(&kbuf);
2489 err = proc_get_long(&kbuf, &left, &val, &neg,
2491 sizeof(proc_wspace_sep), NULL);
2496 if ((min && val < *min) || (max && val > *max))
2500 val = convdiv * (*i) / convmul;
2502 err = proc_put_char(&buffer, &left, '\t');
2503 err = proc_put_long(&buffer, &left, val, false);
2509 if (!write && !first && left && !err)
2510 err = proc_put_char(&buffer, &left, '\n');
2512 left -= proc_skip_spaces(&kbuf);
2517 return err ? : -EINVAL;
2524 static int do_proc_doulongvec_minmax(struct ctl_table *table, int write,
2525 void __user *buffer,
2526 size_t *lenp, loff_t *ppos,
2527 unsigned long convmul,
2528 unsigned long convdiv)
2530 return __do_proc_doulongvec_minmax(table->data, table, write,
2531 buffer, lenp, ppos, convmul, convdiv);
2535 * proc_doulongvec_minmax - read a vector of long integers with min/max values
2536 * @table: the sysctl table
2537 * @write: %TRUE if this is a write to the sysctl file
2538 * @buffer: the user buffer
2539 * @lenp: the size of the user buffer
2540 * @ppos: file position
2542 * Reads/writes up to table->maxlen/sizeof(unsigned long) unsigned long
2543 * values from/to the user buffer, treated as an ASCII string.
2545 * This routine will ensure the values are within the range specified by
2546 * table->extra1 (min) and table->extra2 (max).
2548 * Returns 0 on success.
2550 int proc_doulongvec_minmax(struct ctl_table *table, int write,
2551 void __user *buffer, size_t *lenp, loff_t *ppos)
2553 return do_proc_doulongvec_minmax(table, write, buffer, lenp, ppos, 1l, 1l);
2557 * proc_doulongvec_ms_jiffies_minmax - read a vector of millisecond values with min/max values
2558 * @table: the sysctl table
2559 * @write: %TRUE if this is a write to the sysctl file
2560 * @buffer: the user buffer
2561 * @lenp: the size of the user buffer
2562 * @ppos: file position
2564 * Reads/writes up to table->maxlen/sizeof(unsigned long) unsigned long
2565 * values from/to the user buffer, treated as an ASCII string. The values
2566 * are treated as milliseconds, and converted to jiffies when they are stored.
2568 * This routine will ensure the values are within the range specified by
2569 * table->extra1 (min) and table->extra2 (max).
2571 * Returns 0 on success.
2573 int proc_doulongvec_ms_jiffies_minmax(struct ctl_table *table, int write,
2574 void __user *buffer,
2575 size_t *lenp, loff_t *ppos)
2577 return do_proc_doulongvec_minmax(table, write, buffer,
2578 lenp, ppos, HZ, 1000l);
2582 static int do_proc_dointvec_jiffies_conv(bool *negp, unsigned long *lvalp,
2584 int write, void *data)
2587 if (*lvalp > LONG_MAX / HZ)
2589 *valp = *negp ? -(*lvalp*HZ) : (*lvalp*HZ);
2595 lval = (unsigned long)-val;
2598 lval = (unsigned long)val;
2605 static int do_proc_dointvec_userhz_jiffies_conv(bool *negp, unsigned long *lvalp,
2607 int write, void *data)
2610 if (USER_HZ < HZ && *lvalp > (LONG_MAX / HZ) * USER_HZ)
2612 *valp = clock_t_to_jiffies(*negp ? -*lvalp : *lvalp);
2618 lval = (unsigned long)-val;
2621 lval = (unsigned long)val;
2623 *lvalp = jiffies_to_clock_t(lval);
2628 static int do_proc_dointvec_ms_jiffies_conv(bool *negp, unsigned long *lvalp,
2630 int write, void *data)
2633 *valp = msecs_to_jiffies(*negp ? -*lvalp : *lvalp);
2639 lval = (unsigned long)-val;
2642 lval = (unsigned long)val;
2644 *lvalp = jiffies_to_msecs(lval);
2650 * proc_dointvec_jiffies - read a vector of integers as seconds
2651 * @table: the sysctl table
2652 * @write: %TRUE if this is a write to the sysctl file
2653 * @buffer: the user buffer
2654 * @lenp: the size of the user buffer
2655 * @ppos: file position
2657 * Reads/writes up to table->maxlen/sizeof(unsigned int) integer
2658 * values from/to the user buffer, treated as an ASCII string.
2659 * The values read are assumed to be in seconds, and are converted into
2662 * Returns 0 on success.
2664 int proc_dointvec_jiffies(struct ctl_table *table, int write,
2665 void __user *buffer, size_t *lenp, loff_t *ppos)
2667 return do_proc_dointvec(table,write,buffer,lenp,ppos,
2668 do_proc_dointvec_jiffies_conv,NULL);
2672 * proc_dointvec_userhz_jiffies - read a vector of integers as 1/USER_HZ seconds
2673 * @table: the sysctl table
2674 * @write: %TRUE if this is a write to the sysctl file
2675 * @buffer: the user buffer
2676 * @lenp: the size of the user buffer
2677 * @ppos: pointer to the file position
2679 * Reads/writes up to table->maxlen/sizeof(unsigned int) integer
2680 * values from/to the user buffer, treated as an ASCII string.
2681 * The values read are assumed to be in 1/USER_HZ seconds, and
2682 * are converted into jiffies.
2684 * Returns 0 on success.
2686 int proc_dointvec_userhz_jiffies(struct ctl_table *table, int write,
2687 void __user *buffer, size_t *lenp, loff_t *ppos)
2689 return do_proc_dointvec(table,write,buffer,lenp,ppos,
2690 do_proc_dointvec_userhz_jiffies_conv,NULL);
2694 * proc_dointvec_ms_jiffies - read a vector of integers as 1 milliseconds
2695 * @table: the sysctl table
2696 * @write: %TRUE if this is a write to the sysctl file
2697 * @buffer: the user buffer
2698 * @lenp: the size of the user buffer
2699 * @ppos: file position
2700 * @ppos: the current position in the file
2702 * Reads/writes up to table->maxlen/sizeof(unsigned int) integer
2703 * values from/to the user buffer, treated as an ASCII string.
2704 * The values read are assumed to be in 1/1000 seconds, and
2705 * are converted into jiffies.
2707 * Returns 0 on success.
2709 int proc_dointvec_ms_jiffies(struct ctl_table *table, int write,
2710 void __user *buffer, size_t *lenp, loff_t *ppos)
2712 return do_proc_dointvec(table, write, buffer, lenp, ppos,
2713 do_proc_dointvec_ms_jiffies_conv, NULL);
2716 static int proc_do_cad_pid(struct ctl_table *table, int write,
2717 void __user *buffer, size_t *lenp, loff_t *ppos)
2719 struct pid *new_pid;
2723 tmp = pid_vnr(cad_pid);
2725 r = __do_proc_dointvec(&tmp, table, write, buffer,
2726 lenp, ppos, NULL, NULL);
2730 new_pid = find_get_pid(tmp);
2734 put_pid(xchg(&cad_pid, new_pid));
2739 * proc_do_large_bitmap - read/write from/to a large bitmap
2740 * @table: the sysctl table
2741 * @write: %TRUE if this is a write to the sysctl file
2742 * @buffer: the user buffer
2743 * @lenp: the size of the user buffer
2744 * @ppos: file position
2746 * The bitmap is stored at table->data and the bitmap length (in bits)
2749 * We use a range comma separated format (e.g. 1,3-4,10-10) so that
2750 * large bitmaps may be represented in a compact manner. Writing into
2751 * the file will clear the bitmap then update it with the given input.
2753 * Returns 0 on success.
2755 int proc_do_large_bitmap(struct ctl_table *table, int write,
2756 void __user *buffer, size_t *lenp, loff_t *ppos)
2760 size_t left = *lenp;
2761 unsigned long bitmap_len = table->maxlen;
2762 unsigned long *bitmap = (unsigned long *) table->data;
2763 unsigned long *tmp_bitmap = NULL;
2764 char tr_a[] = { '-', ',', '\n' }, tr_b[] = { ',', '\n', 0 }, c;
2766 if (!bitmap_len || !left || (*ppos && !write)) {
2772 unsigned long page = 0;
2775 if (left > PAGE_SIZE - 1)
2776 left = PAGE_SIZE - 1;
2778 page = __get_free_page(GFP_TEMPORARY);
2779 kbuf = (char *) page;
2782 if (copy_from_user(kbuf, buffer, left)) {
2788 tmp_bitmap = kzalloc(BITS_TO_LONGS(bitmap_len) * sizeof(unsigned long),
2794 proc_skip_char(&kbuf, &left, '\n');
2795 while (!err && left) {
2796 unsigned long val_a, val_b;
2799 err = proc_get_long(&kbuf, &left, &val_a, &neg, tr_a,
2803 if (val_a >= bitmap_len || neg) {
2815 err = proc_get_long(&kbuf, &left, &val_b,
2816 &neg, tr_b, sizeof(tr_b),
2820 if (val_b >= bitmap_len || neg ||
2831 while (val_a <= val_b)
2832 set_bit(val_a++, tmp_bitmap);
2835 proc_skip_char(&kbuf, &left, '\n');
2839 unsigned long bit_a, bit_b = 0;
2842 bit_a = find_next_bit(bitmap, bitmap_len, bit_b);
2843 if (bit_a >= bitmap_len)
2845 bit_b = find_next_zero_bit(bitmap, bitmap_len,
2849 err = proc_put_char(&buffer, &left, ',');
2853 err = proc_put_long(&buffer, &left, bit_a, false);
2856 if (bit_a != bit_b) {
2857 err = proc_put_char(&buffer, &left, '-');
2860 err = proc_put_long(&buffer, &left, bit_b, false);
2868 err = proc_put_char(&buffer, &left, '\n');
2874 bitmap_or(bitmap, bitmap, tmp_bitmap, bitmap_len);
2876 memcpy(bitmap, tmp_bitmap,
2877 BITS_TO_LONGS(bitmap_len) * sizeof(unsigned long));
2889 #else /* CONFIG_PROC_SYSCTL */
2891 int proc_dostring(struct ctl_table *table, int write,
2892 void __user *buffer, size_t *lenp, loff_t *ppos)
2897 int proc_dointvec(struct ctl_table *table, int write,
2898 void __user *buffer, size_t *lenp, loff_t *ppos)
2903 int proc_dointvec_minmax(struct ctl_table *table, int write,
2904 void __user *buffer, size_t *lenp, loff_t *ppos)
2909 int proc_dointvec_jiffies(struct ctl_table *table, int write,
2910 void __user *buffer, size_t *lenp, loff_t *ppos)
2915 int proc_dointvec_userhz_jiffies(struct ctl_table *table, int write,
2916 void __user *buffer, size_t *lenp, loff_t *ppos)
2921 int proc_dointvec_ms_jiffies(struct ctl_table *table, int write,
2922 void __user *buffer, size_t *lenp, loff_t *ppos)
2927 int proc_doulongvec_minmax(struct ctl_table *table, int write,
2928 void __user *buffer, size_t *lenp, loff_t *ppos)
2933 int proc_doulongvec_ms_jiffies_minmax(struct ctl_table *table, int write,
2934 void __user *buffer,
2935 size_t *lenp, loff_t *ppos)
2941 #endif /* CONFIG_PROC_SYSCTL */
2944 * No sense putting this after each symbol definition, twice,
2945 * exception granted :-)
2947 EXPORT_SYMBOL(proc_dointvec);
2948 EXPORT_SYMBOL(proc_dointvec_jiffies);
2949 EXPORT_SYMBOL(proc_dointvec_minmax);
2950 EXPORT_SYMBOL(proc_dointvec_userhz_jiffies);
2951 EXPORT_SYMBOL(proc_dointvec_ms_jiffies);
2952 EXPORT_SYMBOL(proc_dostring);
2953 EXPORT_SYMBOL(proc_doulongvec_minmax);
2954 EXPORT_SYMBOL(proc_doulongvec_ms_jiffies_minmax);
2955 EXPORT_SYMBOL(register_sysctl_table);
2956 EXPORT_SYMBOL(register_sysctl_paths);
2957 EXPORT_SYMBOL(unregister_sysctl_table);