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/proc_fs.h>
28 #include <linux/security.h>
29 #include <linux/ctype.h>
30 #include <linux/kmemcheck.h>
32 #include <linux/init.h>
33 #include <linux/kernel.h>
34 #include <linux/kobject.h>
35 #include <linux/net.h>
36 #include <linux/sysrq.h>
37 #include <linux/highuid.h>
38 #include <linux/writeback.h>
39 #include <linux/ratelimit.h>
40 #include <linux/hugetlb.h>
41 #include <linux/initrd.h>
42 #include <linux/key.h>
43 #include <linux/times.h>
44 #include <linux/limits.h>
45 #include <linux/dcache.h>
46 #include <linux/syscalls.h>
47 #include <linux/vmstat.h>
48 #include <linux/nfs_fs.h>
49 #include <linux/acpi.h>
50 #include <linux/reboot.h>
51 #include <linux/ftrace.h>
52 #include <linux/slow-work.h>
53 #include <linux/perf_event.h>
54 #include <linux/kprobes.h>
56 #include <asm/uaccess.h>
57 #include <asm/processor.h>
61 #include <asm/stacktrace.h>
64 #ifdef CONFIG_BSD_PROCESS_ACCT
65 #include <linux/acct.h>
67 #ifdef CONFIG_CHR_DEV_SG
72 #if defined(CONFIG_SYSCTL)
74 /* External variables not in a header file. */
75 extern int sysctl_overcommit_memory;
76 extern int sysctl_overcommit_ratio;
77 extern int sysctl_panic_on_oom;
78 extern int sysctl_oom_kill_allocating_task;
79 extern int sysctl_oom_dump_tasks;
80 extern int max_threads;
81 extern int core_uses_pid;
82 extern int suid_dumpable;
83 extern char core_pattern[];
84 extern unsigned int core_pipe_limit;
86 extern int min_free_kbytes;
87 extern int pid_max_min, pid_max_max;
88 extern int sysctl_drop_caches;
89 extern int percpu_pagelist_fraction;
90 extern int compat_log;
91 extern int latencytop_enabled;
92 extern int sysctl_nr_open_min, sysctl_nr_open_max;
94 extern int sysctl_nr_trim_pages;
97 extern int blk_iopoll_enabled;
100 /* Constants used for minimum and maximum */
101 #ifdef CONFIG_DETECT_SOFTLOCKUP
102 static int sixty = 60;
103 static int neg_one = -1;
107 static int __maybe_unused one = 1;
108 static int __maybe_unused two = 2;
109 static unsigned long one_ul = 1;
110 static int one_hundred = 100;
112 static int ten_thousand = 10000;
115 /* this is needed for the proc_doulongvec_minmax of vm_dirty_bytes */
116 static unsigned long dirty_bytes_min = 2 * PAGE_SIZE;
118 /* this is needed for the proc_dointvec_minmax for [fs_]overflow UID and GID */
119 static int maxolduid = 65535;
120 static int minolduid;
121 static int min_percpu_pagelist_fract = 8;
123 static int ngroups_max = NGROUPS_MAX;
126 #include <asm/system.h>
129 #ifdef CONFIG_SPARC64
130 extern int sysctl_tsb_ratio;
134 extern int pwrsw_enabled;
135 extern int unaligned_enabled;
139 #ifdef CONFIG_MATHEMU
140 extern int sysctl_ieee_emulation_warnings;
142 extern int sysctl_userprocess_debug;
143 extern int spin_retry;
147 extern int no_unaligned_warning;
148 extern int unaligned_dump_stack;
151 extern struct ratelimit_state printk_ratelimit_state;
153 #ifdef CONFIG_RT_MUTEXES
154 extern int max_lock_depth;
157 #ifdef CONFIG_PROC_SYSCTL
158 static int proc_do_cad_pid(struct ctl_table *table, int write,
159 void __user *buffer, size_t *lenp, loff_t *ppos);
160 static int proc_taint(struct ctl_table *table, int write,
161 void __user *buffer, size_t *lenp, loff_t *ppos);
164 static struct ctl_table root_table[];
165 static struct ctl_table_root sysctl_table_root;
166 static struct ctl_table_header root_table_header = {
168 .ctl_table = root_table,
169 .ctl_entry = LIST_HEAD_INIT(sysctl_table_root.default_set.list),
170 .root = &sysctl_table_root,
171 .set = &sysctl_table_root.default_set,
173 static struct ctl_table_root sysctl_table_root = {
174 .root_list = LIST_HEAD_INIT(sysctl_table_root.root_list),
175 .default_set.list = LIST_HEAD_INIT(root_table_header.ctl_entry),
178 static struct ctl_table kern_table[];
179 static struct ctl_table vm_table[];
180 static struct ctl_table fs_table[];
181 static struct ctl_table debug_table[];
182 static struct ctl_table dev_table[];
183 extern struct ctl_table random_table[];
184 #ifdef CONFIG_INOTIFY_USER
185 extern struct ctl_table inotify_table[];
188 extern struct ctl_table epoll_table[];
191 #ifdef HAVE_ARCH_PICK_MMAP_LAYOUT
192 int sysctl_legacy_va_layout;
195 extern int prove_locking;
196 extern int lock_stat;
198 /* The default sysctl tables: */
200 static struct ctl_table root_table[] = {
202 .procname = "kernel",
219 .child = debug_table,
227 * NOTE: do not add new entries to this table unless you have read
228 * Documentation/sysctl/ctl_unnumbered.txt
233 #ifdef CONFIG_SCHED_DEBUG
234 static int min_sched_granularity_ns = 100000; /* 100 usecs */
235 static int max_sched_granularity_ns = NSEC_PER_SEC; /* 1 second */
236 static int min_wakeup_granularity_ns; /* 0 usecs */
237 static int max_wakeup_granularity_ns = NSEC_PER_SEC; /* 1 second */
238 static int min_sched_tunable_scaling = SCHED_TUNABLESCALING_NONE;
239 static int max_sched_tunable_scaling = SCHED_TUNABLESCALING_END-1;
240 static int min_sched_shares_ratelimit = 100000; /* 100 usec */
241 static int max_sched_shares_ratelimit = NSEC_PER_SEC; /* 1 second */
244 static struct ctl_table kern_table[] = {
246 .procname = "sched_child_runs_first",
247 .data = &sysctl_sched_child_runs_first,
248 .maxlen = sizeof(unsigned int),
250 .proc_handler = proc_dointvec,
252 #ifdef CONFIG_SCHED_DEBUG
254 .procname = "sched_min_granularity_ns",
255 .data = &sysctl_sched_min_granularity,
256 .maxlen = sizeof(unsigned int),
258 .proc_handler = sched_proc_update_handler,
259 .extra1 = &min_sched_granularity_ns,
260 .extra2 = &max_sched_granularity_ns,
263 .procname = "sched_latency_ns",
264 .data = &sysctl_sched_latency,
265 .maxlen = sizeof(unsigned int),
267 .proc_handler = sched_proc_update_handler,
268 .extra1 = &min_sched_granularity_ns,
269 .extra2 = &max_sched_granularity_ns,
272 .procname = "sched_wakeup_granularity_ns",
273 .data = &sysctl_sched_wakeup_granularity,
274 .maxlen = sizeof(unsigned int),
276 .proc_handler = sched_proc_update_handler,
277 .extra1 = &min_wakeup_granularity_ns,
278 .extra2 = &max_wakeup_granularity_ns,
281 .procname = "sched_shares_ratelimit",
282 .data = &sysctl_sched_shares_ratelimit,
283 .maxlen = sizeof(unsigned int),
285 .proc_handler = sched_proc_update_handler,
286 .extra1 = &min_sched_shares_ratelimit,
287 .extra2 = &max_sched_shares_ratelimit,
290 .procname = "sched_tunable_scaling",
291 .data = &sysctl_sched_tunable_scaling,
292 .maxlen = sizeof(enum sched_tunable_scaling),
294 .proc_handler = sched_proc_update_handler,
295 .extra1 = &min_sched_tunable_scaling,
296 .extra2 = &max_sched_tunable_scaling,
299 .procname = "sched_shares_thresh",
300 .data = &sysctl_sched_shares_thresh,
301 .maxlen = sizeof(unsigned int),
303 .proc_handler = proc_dointvec_minmax,
307 .procname = "sched_migration_cost",
308 .data = &sysctl_sched_migration_cost,
309 .maxlen = sizeof(unsigned int),
311 .proc_handler = proc_dointvec,
314 .procname = "sched_nr_migrate",
315 .data = &sysctl_sched_nr_migrate,
316 .maxlen = sizeof(unsigned int),
318 .proc_handler = proc_dointvec,
321 .procname = "sched_time_avg",
322 .data = &sysctl_sched_time_avg,
323 .maxlen = sizeof(unsigned int),
325 .proc_handler = proc_dointvec,
328 .procname = "timer_migration",
329 .data = &sysctl_timer_migration,
330 .maxlen = sizeof(unsigned int),
332 .proc_handler = proc_dointvec_minmax,
338 .procname = "sched_rt_period_us",
339 .data = &sysctl_sched_rt_period,
340 .maxlen = sizeof(unsigned int),
342 .proc_handler = sched_rt_handler,
345 .procname = "sched_rt_runtime_us",
346 .data = &sysctl_sched_rt_runtime,
347 .maxlen = sizeof(int),
349 .proc_handler = sched_rt_handler,
352 .procname = "sched_compat_yield",
353 .data = &sysctl_sched_compat_yield,
354 .maxlen = sizeof(unsigned int),
356 .proc_handler = proc_dointvec,
358 #ifdef CONFIG_PROVE_LOCKING
360 .procname = "prove_locking",
361 .data = &prove_locking,
362 .maxlen = sizeof(int),
364 .proc_handler = proc_dointvec,
367 #ifdef CONFIG_LOCK_STAT
369 .procname = "lock_stat",
371 .maxlen = sizeof(int),
373 .proc_handler = proc_dointvec,
378 .data = &panic_timeout,
379 .maxlen = sizeof(int),
381 .proc_handler = proc_dointvec,
384 .procname = "core_uses_pid",
385 .data = &core_uses_pid,
386 .maxlen = sizeof(int),
388 .proc_handler = proc_dointvec,
391 .procname = "core_pattern",
392 .data = core_pattern,
393 .maxlen = CORENAME_MAX_SIZE,
395 .proc_handler = proc_dostring,
398 .procname = "core_pipe_limit",
399 .data = &core_pipe_limit,
400 .maxlen = sizeof(unsigned int),
402 .proc_handler = proc_dointvec,
404 #ifdef CONFIG_PROC_SYSCTL
406 .procname = "tainted",
407 .maxlen = sizeof(long),
409 .proc_handler = proc_taint,
412 #ifdef CONFIG_LATENCYTOP
414 .procname = "latencytop",
415 .data = &latencytop_enabled,
416 .maxlen = sizeof(int),
418 .proc_handler = proc_dointvec,
421 #ifdef CONFIG_BLK_DEV_INITRD
423 .procname = "real-root-dev",
424 .data = &real_root_dev,
425 .maxlen = sizeof(int),
427 .proc_handler = proc_dointvec,
431 .procname = "print-fatal-signals",
432 .data = &print_fatal_signals,
433 .maxlen = sizeof(int),
435 .proc_handler = proc_dointvec,
439 .procname = "reboot-cmd",
440 .data = reboot_command,
443 .proc_handler = proc_dostring,
446 .procname = "stop-a",
447 .data = &stop_a_enabled,
448 .maxlen = sizeof (int),
450 .proc_handler = proc_dointvec,
453 .procname = "scons-poweroff",
454 .data = &scons_pwroff,
455 .maxlen = sizeof (int),
457 .proc_handler = proc_dointvec,
460 #ifdef CONFIG_SPARC64
462 .procname = "tsb-ratio",
463 .data = &sysctl_tsb_ratio,
464 .maxlen = sizeof (int),
466 .proc_handler = proc_dointvec,
471 .procname = "soft-power",
472 .data = &pwrsw_enabled,
473 .maxlen = sizeof (int),
475 .proc_handler = proc_dointvec,
478 .procname = "unaligned-trap",
479 .data = &unaligned_enabled,
480 .maxlen = sizeof (int),
482 .proc_handler = proc_dointvec,
486 .procname = "ctrl-alt-del",
488 .maxlen = sizeof(int),
490 .proc_handler = proc_dointvec,
492 #ifdef CONFIG_FUNCTION_TRACER
494 .procname = "ftrace_enabled",
495 .data = &ftrace_enabled,
496 .maxlen = sizeof(int),
498 .proc_handler = ftrace_enable_sysctl,
501 #ifdef CONFIG_STACK_TRACER
503 .procname = "stack_tracer_enabled",
504 .data = &stack_tracer_enabled,
505 .maxlen = sizeof(int),
507 .proc_handler = stack_trace_sysctl,
510 #ifdef CONFIG_TRACING
512 .procname = "ftrace_dump_on_oops",
513 .data = &ftrace_dump_on_oops,
514 .maxlen = sizeof(int),
516 .proc_handler = proc_dointvec,
519 #ifdef CONFIG_MODULES
521 .procname = "modprobe",
522 .data = &modprobe_path,
523 .maxlen = KMOD_PATH_LEN,
525 .proc_handler = proc_dostring,
528 .procname = "modules_disabled",
529 .data = &modules_disabled,
530 .maxlen = sizeof(int),
532 /* only handle a transition from default "0" to "1" */
533 .proc_handler = proc_dointvec_minmax,
538 #if defined(CONFIG_HOTPLUG) && defined(CONFIG_NET)
540 .procname = "hotplug",
541 .data = &uevent_helper,
542 .maxlen = UEVENT_HELPER_PATH_LEN,
544 .proc_handler = proc_dostring,
547 #ifdef CONFIG_CHR_DEV_SG
549 .procname = "sg-big-buff",
550 .data = &sg_big_buff,
551 .maxlen = sizeof (int),
553 .proc_handler = proc_dointvec,
556 #ifdef CONFIG_BSD_PROCESS_ACCT
560 .maxlen = 3*sizeof(int),
562 .proc_handler = proc_dointvec,
565 #ifdef CONFIG_MAGIC_SYSRQ
568 .data = &__sysrq_enabled,
569 .maxlen = sizeof (int),
571 .proc_handler = proc_dointvec,
574 #ifdef CONFIG_PROC_SYSCTL
576 .procname = "cad_pid",
578 .maxlen = sizeof (int),
580 .proc_handler = proc_do_cad_pid,
584 .procname = "threads-max",
585 .data = &max_threads,
586 .maxlen = sizeof(int),
588 .proc_handler = proc_dointvec,
591 .procname = "random",
593 .child = random_table,
596 .procname = "overflowuid",
597 .data = &overflowuid,
598 .maxlen = sizeof(int),
600 .proc_handler = proc_dointvec_minmax,
601 .extra1 = &minolduid,
602 .extra2 = &maxolduid,
605 .procname = "overflowgid",
606 .data = &overflowgid,
607 .maxlen = sizeof(int),
609 .proc_handler = proc_dointvec_minmax,
610 .extra1 = &minolduid,
611 .extra2 = &maxolduid,
614 #ifdef CONFIG_MATHEMU
616 .procname = "ieee_emulation_warnings",
617 .data = &sysctl_ieee_emulation_warnings,
618 .maxlen = sizeof(int),
620 .proc_handler = proc_dointvec,
624 .procname = "userprocess_debug",
625 .data = &sysctl_userprocess_debug,
626 .maxlen = sizeof(int),
628 .proc_handler = proc_dointvec,
632 .procname = "pid_max",
634 .maxlen = sizeof (int),
636 .proc_handler = proc_dointvec_minmax,
637 .extra1 = &pid_max_min,
638 .extra2 = &pid_max_max,
641 .procname = "panic_on_oops",
642 .data = &panic_on_oops,
643 .maxlen = sizeof(int),
645 .proc_handler = proc_dointvec,
647 #if defined CONFIG_PRINTK
649 .procname = "printk",
650 .data = &console_loglevel,
651 .maxlen = 4*sizeof(int),
653 .proc_handler = proc_dointvec,
656 .procname = "printk_ratelimit",
657 .data = &printk_ratelimit_state.interval,
658 .maxlen = sizeof(int),
660 .proc_handler = proc_dointvec_jiffies,
663 .procname = "printk_ratelimit_burst",
664 .data = &printk_ratelimit_state.burst,
665 .maxlen = sizeof(int),
667 .proc_handler = proc_dointvec,
670 .procname = "printk_delay",
671 .data = &printk_delay_msec,
672 .maxlen = sizeof(int),
674 .proc_handler = proc_dointvec_minmax,
676 .extra2 = &ten_thousand,
680 .procname = "ngroups_max",
681 .data = &ngroups_max,
682 .maxlen = sizeof (int),
684 .proc_handler = proc_dointvec,
686 #if defined(CONFIG_X86_LOCAL_APIC) && defined(CONFIG_X86)
688 .procname = "unknown_nmi_panic",
689 .data = &unknown_nmi_panic,
690 .maxlen = sizeof (int),
692 .proc_handler = proc_dointvec,
695 .procname = "nmi_watchdog",
696 .data = &nmi_watchdog_enabled,
697 .maxlen = sizeof (int),
699 .proc_handler = proc_nmi_enabled,
702 #if defined(CONFIG_X86)
704 .procname = "panic_on_unrecovered_nmi",
705 .data = &panic_on_unrecovered_nmi,
706 .maxlen = sizeof(int),
708 .proc_handler = proc_dointvec,
711 .procname = "panic_on_io_nmi",
712 .data = &panic_on_io_nmi,
713 .maxlen = sizeof(int),
715 .proc_handler = proc_dointvec,
718 .procname = "bootloader_type",
719 .data = &bootloader_type,
720 .maxlen = sizeof (int),
722 .proc_handler = proc_dointvec,
725 .procname = "bootloader_version",
726 .data = &bootloader_version,
727 .maxlen = sizeof (int),
729 .proc_handler = proc_dointvec,
732 .procname = "kstack_depth_to_print",
733 .data = &kstack_depth_to_print,
734 .maxlen = sizeof(int),
736 .proc_handler = proc_dointvec,
739 .procname = "io_delay_type",
740 .data = &io_delay_type,
741 .maxlen = sizeof(int),
743 .proc_handler = proc_dointvec,
746 #if defined(CONFIG_MMU)
748 .procname = "randomize_va_space",
749 .data = &randomize_va_space,
750 .maxlen = sizeof(int),
752 .proc_handler = proc_dointvec,
755 #if defined(CONFIG_S390) && defined(CONFIG_SMP)
757 .procname = "spin_retry",
759 .maxlen = sizeof (int),
761 .proc_handler = proc_dointvec,
764 #if defined(CONFIG_ACPI_SLEEP) && defined(CONFIG_X86)
766 .procname = "acpi_video_flags",
767 .data = &acpi_realmode_flags,
768 .maxlen = sizeof (unsigned long),
770 .proc_handler = proc_doulongvec_minmax,
775 .procname = "ignore-unaligned-usertrap",
776 .data = &no_unaligned_warning,
777 .maxlen = sizeof (int),
779 .proc_handler = proc_dointvec,
782 .procname = "unaligned-dump-stack",
783 .data = &unaligned_dump_stack,
784 .maxlen = sizeof (int),
786 .proc_handler = proc_dointvec,
789 #ifdef CONFIG_DETECT_SOFTLOCKUP
791 .procname = "softlockup_panic",
792 .data = &softlockup_panic,
793 .maxlen = sizeof(int),
795 .proc_handler = proc_dointvec_minmax,
800 .procname = "softlockup_thresh",
801 .data = &softlockup_thresh,
802 .maxlen = sizeof(int),
804 .proc_handler = proc_dosoftlockup_thresh,
809 #ifdef CONFIG_DETECT_HUNG_TASK
811 .procname = "hung_task_panic",
812 .data = &sysctl_hung_task_panic,
813 .maxlen = sizeof(int),
815 .proc_handler = proc_dointvec_minmax,
820 .procname = "hung_task_check_count",
821 .data = &sysctl_hung_task_check_count,
822 .maxlen = sizeof(unsigned long),
824 .proc_handler = proc_doulongvec_minmax,
827 .procname = "hung_task_timeout_secs",
828 .data = &sysctl_hung_task_timeout_secs,
829 .maxlen = sizeof(unsigned long),
831 .proc_handler = proc_dohung_task_timeout_secs,
834 .procname = "hung_task_warnings",
835 .data = &sysctl_hung_task_warnings,
836 .maxlen = sizeof(unsigned long),
838 .proc_handler = proc_doulongvec_minmax,
843 .procname = "compat-log",
845 .maxlen = sizeof (int),
847 .proc_handler = proc_dointvec,
850 #ifdef CONFIG_RT_MUTEXES
852 .procname = "max_lock_depth",
853 .data = &max_lock_depth,
854 .maxlen = sizeof(int),
856 .proc_handler = proc_dointvec,
860 .procname = "poweroff_cmd",
861 .data = &poweroff_cmd,
862 .maxlen = POWEROFF_CMD_PATH_LEN,
864 .proc_handler = proc_dostring,
870 .child = key_sysctls,
873 #ifdef CONFIG_RCU_TORTURE_TEST
875 .procname = "rcutorture_runnable",
876 .data = &rcutorture_runnable,
877 .maxlen = sizeof(int),
879 .proc_handler = proc_dointvec,
882 #ifdef CONFIG_SLOW_WORK
884 .procname = "slow-work",
886 .child = slow_work_sysctls,
889 #ifdef CONFIG_PERF_EVENTS
891 .procname = "perf_event_paranoid",
892 .data = &sysctl_perf_event_paranoid,
893 .maxlen = sizeof(sysctl_perf_event_paranoid),
895 .proc_handler = proc_dointvec,
898 .procname = "perf_event_mlock_kb",
899 .data = &sysctl_perf_event_mlock,
900 .maxlen = sizeof(sysctl_perf_event_mlock),
902 .proc_handler = proc_dointvec,
905 .procname = "perf_event_max_sample_rate",
906 .data = &sysctl_perf_event_sample_rate,
907 .maxlen = sizeof(sysctl_perf_event_sample_rate),
909 .proc_handler = proc_dointvec,
912 #ifdef CONFIG_KMEMCHECK
914 .procname = "kmemcheck",
915 .data = &kmemcheck_enabled,
916 .maxlen = sizeof(int),
918 .proc_handler = proc_dointvec,
923 .procname = "blk_iopoll",
924 .data = &blk_iopoll_enabled,
925 .maxlen = sizeof(int),
927 .proc_handler = proc_dointvec,
931 * NOTE: do not add new entries to this table unless you have read
932 * Documentation/sysctl/ctl_unnumbered.txt
937 static struct ctl_table vm_table[] = {
939 .procname = "overcommit_memory",
940 .data = &sysctl_overcommit_memory,
941 .maxlen = sizeof(sysctl_overcommit_memory),
943 .proc_handler = proc_dointvec,
946 .procname = "panic_on_oom",
947 .data = &sysctl_panic_on_oom,
948 .maxlen = sizeof(sysctl_panic_on_oom),
950 .proc_handler = proc_dointvec,
953 .procname = "oom_kill_allocating_task",
954 .data = &sysctl_oom_kill_allocating_task,
955 .maxlen = sizeof(sysctl_oom_kill_allocating_task),
957 .proc_handler = proc_dointvec,
960 .procname = "oom_dump_tasks",
961 .data = &sysctl_oom_dump_tasks,
962 .maxlen = sizeof(sysctl_oom_dump_tasks),
964 .proc_handler = proc_dointvec,
967 .procname = "overcommit_ratio",
968 .data = &sysctl_overcommit_ratio,
969 .maxlen = sizeof(sysctl_overcommit_ratio),
971 .proc_handler = proc_dointvec,
974 .procname = "page-cluster",
975 .data = &page_cluster,
976 .maxlen = sizeof(int),
978 .proc_handler = proc_dointvec,
981 .procname = "dirty_background_ratio",
982 .data = &dirty_background_ratio,
983 .maxlen = sizeof(dirty_background_ratio),
985 .proc_handler = dirty_background_ratio_handler,
987 .extra2 = &one_hundred,
990 .procname = "dirty_background_bytes",
991 .data = &dirty_background_bytes,
992 .maxlen = sizeof(dirty_background_bytes),
994 .proc_handler = dirty_background_bytes_handler,
998 .procname = "dirty_ratio",
999 .data = &vm_dirty_ratio,
1000 .maxlen = sizeof(vm_dirty_ratio),
1002 .proc_handler = dirty_ratio_handler,
1004 .extra2 = &one_hundred,
1007 .procname = "dirty_bytes",
1008 .data = &vm_dirty_bytes,
1009 .maxlen = sizeof(vm_dirty_bytes),
1011 .proc_handler = dirty_bytes_handler,
1012 .extra1 = &dirty_bytes_min,
1015 .procname = "dirty_writeback_centisecs",
1016 .data = &dirty_writeback_interval,
1017 .maxlen = sizeof(dirty_writeback_interval),
1019 .proc_handler = dirty_writeback_centisecs_handler,
1022 .procname = "dirty_expire_centisecs",
1023 .data = &dirty_expire_interval,
1024 .maxlen = sizeof(dirty_expire_interval),
1026 .proc_handler = proc_dointvec,
1029 .procname = "nr_pdflush_threads",
1030 .data = &nr_pdflush_threads,
1031 .maxlen = sizeof nr_pdflush_threads,
1032 .mode = 0444 /* read-only*/,
1033 .proc_handler = proc_dointvec,
1036 .procname = "swappiness",
1037 .data = &vm_swappiness,
1038 .maxlen = sizeof(vm_swappiness),
1040 .proc_handler = proc_dointvec_minmax,
1042 .extra2 = &one_hundred,
1044 #ifdef CONFIG_HUGETLB_PAGE
1046 .procname = "nr_hugepages",
1048 .maxlen = sizeof(unsigned long),
1050 .proc_handler = hugetlb_sysctl_handler,
1051 .extra1 = (void *)&hugetlb_zero,
1052 .extra2 = (void *)&hugetlb_infinity,
1056 .procname = "nr_hugepages_mempolicy",
1058 .maxlen = sizeof(unsigned long),
1060 .proc_handler = &hugetlb_mempolicy_sysctl_handler,
1061 .extra1 = (void *)&hugetlb_zero,
1062 .extra2 = (void *)&hugetlb_infinity,
1066 .procname = "hugetlb_shm_group",
1067 .data = &sysctl_hugetlb_shm_group,
1068 .maxlen = sizeof(gid_t),
1070 .proc_handler = proc_dointvec,
1073 .procname = "hugepages_treat_as_movable",
1074 .data = &hugepages_treat_as_movable,
1075 .maxlen = sizeof(int),
1077 .proc_handler = hugetlb_treat_movable_handler,
1080 .procname = "nr_overcommit_hugepages",
1082 .maxlen = sizeof(unsigned long),
1084 .proc_handler = hugetlb_overcommit_handler,
1085 .extra1 = (void *)&hugetlb_zero,
1086 .extra2 = (void *)&hugetlb_infinity,
1090 .procname = "lowmem_reserve_ratio",
1091 .data = &sysctl_lowmem_reserve_ratio,
1092 .maxlen = sizeof(sysctl_lowmem_reserve_ratio),
1094 .proc_handler = lowmem_reserve_ratio_sysctl_handler,
1097 .procname = "drop_caches",
1098 .data = &sysctl_drop_caches,
1099 .maxlen = sizeof(int),
1101 .proc_handler = drop_caches_sysctl_handler,
1104 .procname = "min_free_kbytes",
1105 .data = &min_free_kbytes,
1106 .maxlen = sizeof(min_free_kbytes),
1108 .proc_handler = min_free_kbytes_sysctl_handler,
1112 .procname = "percpu_pagelist_fraction",
1113 .data = &percpu_pagelist_fraction,
1114 .maxlen = sizeof(percpu_pagelist_fraction),
1116 .proc_handler = percpu_pagelist_fraction_sysctl_handler,
1117 .extra1 = &min_percpu_pagelist_fract,
1121 .procname = "max_map_count",
1122 .data = &sysctl_max_map_count,
1123 .maxlen = sizeof(sysctl_max_map_count),
1125 .proc_handler = proc_dointvec_minmax,
1130 .procname = "nr_trim_pages",
1131 .data = &sysctl_nr_trim_pages,
1132 .maxlen = sizeof(sysctl_nr_trim_pages),
1134 .proc_handler = proc_dointvec_minmax,
1139 .procname = "laptop_mode",
1140 .data = &laptop_mode,
1141 .maxlen = sizeof(laptop_mode),
1143 .proc_handler = proc_dointvec_jiffies,
1146 .procname = "block_dump",
1147 .data = &block_dump,
1148 .maxlen = sizeof(block_dump),
1150 .proc_handler = proc_dointvec,
1154 .procname = "vfs_cache_pressure",
1155 .data = &sysctl_vfs_cache_pressure,
1156 .maxlen = sizeof(sysctl_vfs_cache_pressure),
1158 .proc_handler = proc_dointvec,
1161 #ifdef HAVE_ARCH_PICK_MMAP_LAYOUT
1163 .procname = "legacy_va_layout",
1164 .data = &sysctl_legacy_va_layout,
1165 .maxlen = sizeof(sysctl_legacy_va_layout),
1167 .proc_handler = proc_dointvec,
1173 .procname = "zone_reclaim_mode",
1174 .data = &zone_reclaim_mode,
1175 .maxlen = sizeof(zone_reclaim_mode),
1177 .proc_handler = proc_dointvec,
1181 .procname = "min_unmapped_ratio",
1182 .data = &sysctl_min_unmapped_ratio,
1183 .maxlen = sizeof(sysctl_min_unmapped_ratio),
1185 .proc_handler = sysctl_min_unmapped_ratio_sysctl_handler,
1187 .extra2 = &one_hundred,
1190 .procname = "min_slab_ratio",
1191 .data = &sysctl_min_slab_ratio,
1192 .maxlen = sizeof(sysctl_min_slab_ratio),
1194 .proc_handler = sysctl_min_slab_ratio_sysctl_handler,
1196 .extra2 = &one_hundred,
1201 .procname = "stat_interval",
1202 .data = &sysctl_stat_interval,
1203 .maxlen = sizeof(sysctl_stat_interval),
1205 .proc_handler = proc_dointvec_jiffies,
1210 .procname = "mmap_min_addr",
1211 .data = &dac_mmap_min_addr,
1212 .maxlen = sizeof(unsigned long),
1214 .proc_handler = mmap_min_addr_handler,
1219 .procname = "numa_zonelist_order",
1220 .data = &numa_zonelist_order,
1221 .maxlen = NUMA_ZONELIST_ORDER_LEN,
1223 .proc_handler = numa_zonelist_order_handler,
1226 #if (defined(CONFIG_X86_32) && !defined(CONFIG_UML))|| \
1227 (defined(CONFIG_SUPERH) && defined(CONFIG_VSYSCALL))
1229 .procname = "vdso_enabled",
1230 .data = &vdso_enabled,
1231 .maxlen = sizeof(vdso_enabled),
1233 .proc_handler = proc_dointvec,
1237 #ifdef CONFIG_HIGHMEM
1239 .procname = "highmem_is_dirtyable",
1240 .data = &vm_highmem_is_dirtyable,
1241 .maxlen = sizeof(vm_highmem_is_dirtyable),
1243 .proc_handler = proc_dointvec_minmax,
1249 .procname = "scan_unevictable_pages",
1250 .data = &scan_unevictable_pages,
1251 .maxlen = sizeof(scan_unevictable_pages),
1253 .proc_handler = scan_unevictable_handler,
1255 #ifdef CONFIG_MEMORY_FAILURE
1257 .procname = "memory_failure_early_kill",
1258 .data = &sysctl_memory_failure_early_kill,
1259 .maxlen = sizeof(sysctl_memory_failure_early_kill),
1261 .proc_handler = proc_dointvec_minmax,
1266 .procname = "memory_failure_recovery",
1267 .data = &sysctl_memory_failure_recovery,
1268 .maxlen = sizeof(sysctl_memory_failure_recovery),
1270 .proc_handler = proc_dointvec_minmax,
1277 * NOTE: do not add new entries to this table unless you have read
1278 * Documentation/sysctl/ctl_unnumbered.txt
1283 #if defined(CONFIG_BINFMT_MISC) || defined(CONFIG_BINFMT_MISC_MODULE)
1284 static struct ctl_table binfmt_misc_table[] = {
1289 static struct ctl_table fs_table[] = {
1291 .procname = "inode-nr",
1292 .data = &inodes_stat,
1293 .maxlen = 2*sizeof(int),
1295 .proc_handler = proc_dointvec,
1298 .procname = "inode-state",
1299 .data = &inodes_stat,
1300 .maxlen = 7*sizeof(int),
1302 .proc_handler = proc_dointvec,
1305 .procname = "file-nr",
1306 .data = &files_stat,
1307 .maxlen = 3*sizeof(int),
1309 .proc_handler = proc_nr_files,
1312 .procname = "file-max",
1313 .data = &files_stat.max_files,
1314 .maxlen = sizeof(int),
1316 .proc_handler = proc_dointvec,
1319 .procname = "nr_open",
1320 .data = &sysctl_nr_open,
1321 .maxlen = sizeof(int),
1323 .proc_handler = proc_dointvec_minmax,
1324 .extra1 = &sysctl_nr_open_min,
1325 .extra2 = &sysctl_nr_open_max,
1328 .procname = "dentry-state",
1329 .data = &dentry_stat,
1330 .maxlen = 6*sizeof(int),
1332 .proc_handler = proc_dointvec,
1335 .procname = "overflowuid",
1336 .data = &fs_overflowuid,
1337 .maxlen = sizeof(int),
1339 .proc_handler = proc_dointvec_minmax,
1340 .extra1 = &minolduid,
1341 .extra2 = &maxolduid,
1344 .procname = "overflowgid",
1345 .data = &fs_overflowgid,
1346 .maxlen = sizeof(int),
1348 .proc_handler = proc_dointvec_minmax,
1349 .extra1 = &minolduid,
1350 .extra2 = &maxolduid,
1352 #ifdef CONFIG_FILE_LOCKING
1354 .procname = "leases-enable",
1355 .data = &leases_enable,
1356 .maxlen = sizeof(int),
1358 .proc_handler = proc_dointvec,
1361 #ifdef CONFIG_DNOTIFY
1363 .procname = "dir-notify-enable",
1364 .data = &dir_notify_enable,
1365 .maxlen = sizeof(int),
1367 .proc_handler = proc_dointvec,
1371 #ifdef CONFIG_FILE_LOCKING
1373 .procname = "lease-break-time",
1374 .data = &lease_break_time,
1375 .maxlen = sizeof(int),
1377 .proc_handler = proc_dointvec,
1382 .procname = "aio-nr",
1384 .maxlen = sizeof(aio_nr),
1386 .proc_handler = proc_doulongvec_minmax,
1389 .procname = "aio-max-nr",
1390 .data = &aio_max_nr,
1391 .maxlen = sizeof(aio_max_nr),
1393 .proc_handler = proc_doulongvec_minmax,
1395 #endif /* CONFIG_AIO */
1396 #ifdef CONFIG_INOTIFY_USER
1398 .procname = "inotify",
1400 .child = inotify_table,
1405 .procname = "epoll",
1407 .child = epoll_table,
1412 .procname = "suid_dumpable",
1413 .data = &suid_dumpable,
1414 .maxlen = sizeof(int),
1416 .proc_handler = proc_dointvec_minmax,
1420 #if defined(CONFIG_BINFMT_MISC) || defined(CONFIG_BINFMT_MISC_MODULE)
1422 .procname = "binfmt_misc",
1424 .child = binfmt_misc_table,
1428 * NOTE: do not add new entries to this table unless you have read
1429 * Documentation/sysctl/ctl_unnumbered.txt
1434 static struct ctl_table debug_table[] = {
1435 #if defined(CONFIG_X86) || defined(CONFIG_PPC) || defined(CONFIG_SPARC)
1437 .procname = "exception-trace",
1438 .data = &show_unhandled_signals,
1439 .maxlen = sizeof(int),
1441 .proc_handler = proc_dointvec
1444 #if defined(CONFIG_OPTPROBES)
1446 .procname = "kprobes-optimization",
1447 .data = &sysctl_kprobes_optimization,
1448 .maxlen = sizeof(int),
1450 .proc_handler = proc_kprobes_optimization_handler,
1458 static struct ctl_table dev_table[] = {
1462 static DEFINE_SPINLOCK(sysctl_lock);
1464 /* called under sysctl_lock */
1465 static int use_table(struct ctl_table_header *p)
1467 if (unlikely(p->unregistering))
1473 /* called under sysctl_lock */
1474 static void unuse_table(struct ctl_table_header *p)
1477 if (unlikely(p->unregistering))
1478 complete(p->unregistering);
1481 /* called under sysctl_lock, will reacquire if has to wait */
1482 static void start_unregistering(struct ctl_table_header *p)
1485 * if p->used is 0, nobody will ever touch that entry again;
1486 * we'll eliminate all paths to it before dropping sysctl_lock
1488 if (unlikely(p->used)) {
1489 struct completion wait;
1490 init_completion(&wait);
1491 p->unregistering = &wait;
1492 spin_unlock(&sysctl_lock);
1493 wait_for_completion(&wait);
1494 spin_lock(&sysctl_lock);
1496 /* anything non-NULL; we'll never dereference it */
1497 p->unregistering = ERR_PTR(-EINVAL);
1500 * do not remove from the list until nobody holds it; walking the
1501 * list in do_sysctl() relies on that.
1503 list_del_init(&p->ctl_entry);
1506 void sysctl_head_get(struct ctl_table_header *head)
1508 spin_lock(&sysctl_lock);
1510 spin_unlock(&sysctl_lock);
1513 void sysctl_head_put(struct ctl_table_header *head)
1515 spin_lock(&sysctl_lock);
1518 spin_unlock(&sysctl_lock);
1521 struct ctl_table_header *sysctl_head_grab(struct ctl_table_header *head)
1525 spin_lock(&sysctl_lock);
1526 if (!use_table(head))
1527 head = ERR_PTR(-ENOENT);
1528 spin_unlock(&sysctl_lock);
1532 void sysctl_head_finish(struct ctl_table_header *head)
1536 spin_lock(&sysctl_lock);
1538 spin_unlock(&sysctl_lock);
1541 static struct ctl_table_set *
1542 lookup_header_set(struct ctl_table_root *root, struct nsproxy *namespaces)
1544 struct ctl_table_set *set = &root->default_set;
1546 set = root->lookup(root, namespaces);
1550 static struct list_head *
1551 lookup_header_list(struct ctl_table_root *root, struct nsproxy *namespaces)
1553 struct ctl_table_set *set = lookup_header_set(root, namespaces);
1557 struct ctl_table_header *__sysctl_head_next(struct nsproxy *namespaces,
1558 struct ctl_table_header *prev)
1560 struct ctl_table_root *root;
1561 struct list_head *header_list;
1562 struct ctl_table_header *head;
1563 struct list_head *tmp;
1565 spin_lock(&sysctl_lock);
1568 tmp = &prev->ctl_entry;
1572 tmp = &root_table_header.ctl_entry;
1574 head = list_entry(tmp, struct ctl_table_header, ctl_entry);
1576 if (!use_table(head))
1578 spin_unlock(&sysctl_lock);
1583 header_list = lookup_header_list(root, namespaces);
1584 if (tmp != header_list)
1588 root = list_entry(root->root_list.next,
1589 struct ctl_table_root, root_list);
1590 if (root == &sysctl_table_root)
1592 header_list = lookup_header_list(root, namespaces);
1593 } while (list_empty(header_list));
1594 tmp = header_list->next;
1597 spin_unlock(&sysctl_lock);
1601 struct ctl_table_header *sysctl_head_next(struct ctl_table_header *prev)
1603 return __sysctl_head_next(current->nsproxy, prev);
1606 void register_sysctl_root(struct ctl_table_root *root)
1608 spin_lock(&sysctl_lock);
1609 list_add_tail(&root->root_list, &sysctl_table_root.root_list);
1610 spin_unlock(&sysctl_lock);
1614 * sysctl_perm does NOT grant the superuser all rights automatically, because
1615 * some sysctl variables are readonly even to root.
1618 static int test_perm(int mode, int op)
1620 if (!current_euid())
1622 else if (in_egroup_p(0))
1624 if ((op & ~mode & (MAY_READ|MAY_WRITE|MAY_EXEC)) == 0)
1629 int sysctl_perm(struct ctl_table_root *root, struct ctl_table *table, int op)
1634 error = security_sysctl(table, op & (MAY_READ | MAY_WRITE | MAY_EXEC));
1638 if (root->permissions)
1639 mode = root->permissions(root, current->nsproxy, table);
1643 return test_perm(mode, op);
1646 static void sysctl_set_parent(struct ctl_table *parent, struct ctl_table *table)
1648 for (; table->procname; table++) {
1649 table->parent = parent;
1651 sysctl_set_parent(table, table->child);
1655 static __init int sysctl_init(void)
1657 sysctl_set_parent(NULL, root_table);
1658 #ifdef CONFIG_SYSCTL_SYSCALL_CHECK
1661 err = sysctl_check_table(current->nsproxy, root_table);
1667 core_initcall(sysctl_init);
1669 static struct ctl_table *is_branch_in(struct ctl_table *branch,
1670 struct ctl_table *table)
1672 struct ctl_table *p;
1673 const char *s = branch->procname;
1675 /* branch should have named subdirectory as its first element */
1676 if (!s || !branch->child)
1679 /* ... and nothing else */
1680 if (branch[1].procname)
1683 /* table should contain subdirectory with the same name */
1684 for (p = table; p->procname; p++) {
1687 if (p->procname && strcmp(p->procname, s) == 0)
1693 /* see if attaching q to p would be an improvement */
1694 static void try_attach(struct ctl_table_header *p, struct ctl_table_header *q)
1696 struct ctl_table *to = p->ctl_table, *by = q->ctl_table;
1697 struct ctl_table *next;
1699 int not_in_parent = !p->attached_by;
1701 while ((next = is_branch_in(by, to)) != NULL) {
1702 if (by == q->attached_by)
1704 if (to == p->attached_by)
1710 if (is_better && not_in_parent) {
1711 q->attached_by = by;
1712 q->attached_to = to;
1718 * __register_sysctl_paths - register a sysctl hierarchy
1719 * @root: List of sysctl headers to register on
1720 * @namespaces: Data to compute which lists of sysctl entries are visible
1721 * @path: The path to the directory the sysctl table is in.
1722 * @table: the top-level table structure
1724 * Register a sysctl table hierarchy. @table should be a filled in ctl_table
1725 * array. A completely 0 filled entry terminates the table.
1727 * The members of the &struct ctl_table structure are used as follows:
1729 * procname - the name of the sysctl file under /proc/sys. Set to %NULL to not
1730 * enter a sysctl file
1732 * data - a pointer to data for use by proc_handler
1734 * maxlen - the maximum size in bytes of the data
1736 * mode - the file permissions for the /proc/sys file, and for sysctl(2)
1738 * child - a pointer to the child sysctl table if this entry is a directory, or
1741 * proc_handler - the text handler routine (described below)
1743 * de - for internal use by the sysctl routines
1745 * extra1, extra2 - extra pointers usable by the proc handler routines
1747 * Leaf nodes in the sysctl tree will be represented by a single file
1748 * under /proc; non-leaf nodes will be represented by directories.
1750 * sysctl(2) can automatically manage read and write requests through
1751 * the sysctl table. The data and maxlen fields of the ctl_table
1752 * struct enable minimal validation of the values being written to be
1753 * performed, and the mode field allows minimal authentication.
1755 * There must be a proc_handler routine for any terminal nodes
1756 * mirrored under /proc/sys (non-terminals are handled by a built-in
1757 * directory handler). Several default handlers are available to
1758 * cover common cases -
1760 * proc_dostring(), proc_dointvec(), proc_dointvec_jiffies(),
1761 * proc_dointvec_userhz_jiffies(), proc_dointvec_minmax(),
1762 * proc_doulongvec_ms_jiffies_minmax(), proc_doulongvec_minmax()
1764 * It is the handler's job to read the input buffer from user memory
1765 * and process it. The handler should return 0 on success.
1767 * This routine returns %NULL on a failure to register, and a pointer
1768 * to the table header on success.
1770 struct ctl_table_header *__register_sysctl_paths(
1771 struct ctl_table_root *root,
1772 struct nsproxy *namespaces,
1773 const struct ctl_path *path, struct ctl_table *table)
1775 struct ctl_table_header *header;
1776 struct ctl_table *new, **prevp;
1777 unsigned int n, npath;
1778 struct ctl_table_set *set;
1780 /* Count the path components */
1781 for (npath = 0; path[npath].procname; ++npath)
1785 * For each path component, allocate a 2-element ctl_table array.
1786 * The first array element will be filled with the sysctl entry
1787 * for this, the second will be the sentinel (procname == 0).
1789 * We allocate everything in one go so that we don't have to
1790 * worry about freeing additional memory in unregister_sysctl_table.
1792 header = kzalloc(sizeof(struct ctl_table_header) +
1793 (2 * npath * sizeof(struct ctl_table)), GFP_KERNEL);
1797 new = (struct ctl_table *) (header + 1);
1799 /* Now connect the dots */
1800 prevp = &header->ctl_table;
1801 for (n = 0; n < npath; ++n, ++path) {
1802 /* Copy the procname */
1803 new->procname = path->procname;
1807 prevp = &new->child;
1812 header->ctl_table_arg = table;
1814 INIT_LIST_HEAD(&header->ctl_entry);
1816 header->unregistering = NULL;
1817 header->root = root;
1818 sysctl_set_parent(NULL, header->ctl_table);
1820 #ifdef CONFIG_SYSCTL_SYSCALL_CHECK
1821 if (sysctl_check_table(namespaces, header->ctl_table)) {
1826 spin_lock(&sysctl_lock);
1827 header->set = lookup_header_set(root, namespaces);
1828 header->attached_by = header->ctl_table;
1829 header->attached_to = root_table;
1830 header->parent = &root_table_header;
1831 for (set = header->set; set; set = set->parent) {
1832 struct ctl_table_header *p;
1833 list_for_each_entry(p, &set->list, ctl_entry) {
1834 if (p->unregistering)
1836 try_attach(p, header);
1839 header->parent->count++;
1840 list_add_tail(&header->ctl_entry, &header->set->list);
1841 spin_unlock(&sysctl_lock);
1847 * register_sysctl_table_path - register a sysctl table hierarchy
1848 * @path: The path to the directory the sysctl table is in.
1849 * @table: the top-level table structure
1851 * Register a sysctl table hierarchy. @table should be a filled in ctl_table
1852 * array. A completely 0 filled entry terminates the table.
1854 * See __register_sysctl_paths for more details.
1856 struct ctl_table_header *register_sysctl_paths(const struct ctl_path *path,
1857 struct ctl_table *table)
1859 return __register_sysctl_paths(&sysctl_table_root, current->nsproxy,
1864 * register_sysctl_table - register a sysctl table hierarchy
1865 * @table: the top-level table structure
1867 * Register a sysctl table hierarchy. @table should be a filled in ctl_table
1868 * array. A completely 0 filled entry terminates the table.
1870 * See register_sysctl_paths for more details.
1872 struct ctl_table_header *register_sysctl_table(struct ctl_table *table)
1874 static const struct ctl_path null_path[] = { {} };
1876 return register_sysctl_paths(null_path, table);
1880 * unregister_sysctl_table - unregister a sysctl table hierarchy
1881 * @header: the header returned from register_sysctl_table
1883 * Unregisters the sysctl table and all children. proc entries may not
1884 * actually be removed until they are no longer used by anyone.
1886 void unregister_sysctl_table(struct ctl_table_header * header)
1893 spin_lock(&sysctl_lock);
1894 start_unregistering(header);
1895 if (!--header->parent->count) {
1897 kfree(header->parent);
1899 if (!--header->count)
1901 spin_unlock(&sysctl_lock);
1904 int sysctl_is_seen(struct ctl_table_header *p)
1906 struct ctl_table_set *set = p->set;
1908 spin_lock(&sysctl_lock);
1909 if (p->unregistering)
1911 else if (!set->is_seen)
1914 res = set->is_seen(set);
1915 spin_unlock(&sysctl_lock);
1919 void setup_sysctl_set(struct ctl_table_set *p,
1920 struct ctl_table_set *parent,
1921 int (*is_seen)(struct ctl_table_set *))
1923 INIT_LIST_HEAD(&p->list);
1924 p->parent = parent ? parent : &sysctl_table_root.default_set;
1925 p->is_seen = is_seen;
1928 #else /* !CONFIG_SYSCTL */
1929 struct ctl_table_header *register_sysctl_table(struct ctl_table * table)
1934 struct ctl_table_header *register_sysctl_paths(const struct ctl_path *path,
1935 struct ctl_table *table)
1940 void unregister_sysctl_table(struct ctl_table_header * table)
1944 void setup_sysctl_set(struct ctl_table_set *p,
1945 struct ctl_table_set *parent,
1946 int (*is_seen)(struct ctl_table_set *))
1950 void sysctl_head_put(struct ctl_table_header *head)
1954 #endif /* CONFIG_SYSCTL */
1960 #ifdef CONFIG_PROC_SYSCTL
1962 static int _proc_do_string(void* data, int maxlen, int write,
1963 void __user *buffer,
1964 size_t *lenp, loff_t *ppos)
1970 if (!data || !maxlen || !*lenp) {
1978 while (len < *lenp) {
1979 if (get_user(c, p++))
1981 if (c == 0 || c == '\n')
1987 if(copy_from_user(data, buffer, len))
1989 ((char *) data)[len] = 0;
2007 if(copy_to_user(buffer, data, len))
2010 if(put_user('\n', ((char __user *) buffer) + len))
2021 * proc_dostring - read a string sysctl
2022 * @table: the sysctl table
2023 * @write: %TRUE if this is a write to the sysctl file
2024 * @buffer: the user buffer
2025 * @lenp: the size of the user buffer
2026 * @ppos: file position
2028 * Reads/writes a string from/to the user buffer. If the kernel
2029 * buffer provided is not large enough to hold the string, the
2030 * string is truncated. The copied string is %NULL-terminated.
2031 * If the string is being read by the user process, it is copied
2032 * and a newline '\n' is added. It is truncated if the buffer is
2035 * Returns 0 on success.
2037 int proc_dostring(struct ctl_table *table, int write,
2038 void __user *buffer, size_t *lenp, loff_t *ppos)
2040 return _proc_do_string(table->data, table->maxlen, write,
2041 buffer, lenp, ppos);
2045 static int do_proc_dointvec_conv(int *negp, unsigned long *lvalp,
2047 int write, void *data)
2050 *valp = *negp ? -*lvalp : *lvalp;
2055 *lvalp = (unsigned long)-val;
2058 *lvalp = (unsigned long)val;
2064 static int __do_proc_dointvec(void *tbl_data, struct ctl_table *table,
2065 int write, void __user *buffer,
2066 size_t *lenp, loff_t *ppos,
2067 int (*conv)(int *negp, unsigned long *lvalp, int *valp,
2068 int write, void *data),
2071 #define TMPBUFLEN 21
2072 int *i, vleft, first = 1, neg;
2076 char buf[TMPBUFLEN], *p;
2077 char __user *s = buffer;
2079 if (!tbl_data || !table->maxlen || !*lenp ||
2080 (*ppos && !write)) {
2085 i = (int *) tbl_data;
2086 vleft = table->maxlen / sizeof(*i);
2090 conv = do_proc_dointvec_conv;
2092 for (; left && vleft--; i++, first=0) {
2107 if (len > sizeof(buf) - 1)
2108 len = sizeof(buf) - 1;
2109 if (copy_from_user(buf, s, len))
2113 if (*p == '-' && left > 1) {
2117 if (*p < '0' || *p > '9')
2120 lval = simple_strtoul(p, &p, 0);
2123 if ((len < left) && *p && !isspace(*p))
2128 if (conv(&neg, &lval, i, 1, data))
2135 if (conv(&neg, &lval, i, 0, data))
2138 sprintf(p, "%s%lu", neg ? "-" : "", lval);
2142 if(copy_to_user(s, buf, len))
2149 if (!write && !first && left) {
2150 if(put_user('\n', s))
2157 if (get_user(c, s++))
2172 static int do_proc_dointvec(struct ctl_table *table, int write,
2173 void __user *buffer, size_t *lenp, loff_t *ppos,
2174 int (*conv)(int *negp, unsigned long *lvalp, int *valp,
2175 int write, void *data),
2178 return __do_proc_dointvec(table->data, table, write,
2179 buffer, lenp, ppos, conv, data);
2183 * proc_dointvec - read a vector of integers
2184 * @table: the sysctl table
2185 * @write: %TRUE if this is a write to the sysctl file
2186 * @buffer: the user buffer
2187 * @lenp: the size of the user buffer
2188 * @ppos: file position
2190 * Reads/writes up to table->maxlen/sizeof(unsigned int) integer
2191 * values from/to the user buffer, treated as an ASCII string.
2193 * Returns 0 on success.
2195 int proc_dointvec(struct ctl_table *table, int write,
2196 void __user *buffer, size_t *lenp, loff_t *ppos)
2198 return do_proc_dointvec(table,write,buffer,lenp,ppos,
2203 * Taint values can only be increased
2204 * This means we can safely use a temporary.
2206 static int proc_taint(struct ctl_table *table, int write,
2207 void __user *buffer, size_t *lenp, loff_t *ppos)
2210 unsigned long tmptaint = get_taint();
2213 if (write && !capable(CAP_SYS_ADMIN))
2218 err = proc_doulongvec_minmax(&t, write, buffer, lenp, ppos);
2224 * Poor man's atomic or. Not worth adding a primitive
2225 * to everyone's atomic.h for this
2228 for (i = 0; i < BITS_PER_LONG && tmptaint >> i; i++) {
2229 if ((tmptaint >> i) & 1)
2237 struct do_proc_dointvec_minmax_conv_param {
2242 static int do_proc_dointvec_minmax_conv(int *negp, unsigned long *lvalp,
2244 int write, void *data)
2246 struct do_proc_dointvec_minmax_conv_param *param = data;
2248 int val = *negp ? -*lvalp : *lvalp;
2249 if ((param->min && *param->min > val) ||
2250 (param->max && *param->max < val))
2257 *lvalp = (unsigned long)-val;
2260 *lvalp = (unsigned long)val;
2267 * proc_dointvec_minmax - read a vector of integers with min/max values
2268 * @table: the sysctl table
2269 * @write: %TRUE if this is a write to the sysctl file
2270 * @buffer: the user buffer
2271 * @lenp: the size of the user buffer
2272 * @ppos: file position
2274 * Reads/writes up to table->maxlen/sizeof(unsigned int) integer
2275 * values from/to the user buffer, treated as an ASCII string.
2277 * This routine will ensure the values are within the range specified by
2278 * table->extra1 (min) and table->extra2 (max).
2280 * Returns 0 on success.
2282 int proc_dointvec_minmax(struct ctl_table *table, int write,
2283 void __user *buffer, size_t *lenp, loff_t *ppos)
2285 struct do_proc_dointvec_minmax_conv_param param = {
2286 .min = (int *) table->extra1,
2287 .max = (int *) table->extra2,
2289 return do_proc_dointvec(table, write, buffer, lenp, ppos,
2290 do_proc_dointvec_minmax_conv, ¶m);
2293 static int __do_proc_doulongvec_minmax(void *data, struct ctl_table *table, int write,
2294 void __user *buffer,
2295 size_t *lenp, loff_t *ppos,
2296 unsigned long convmul,
2297 unsigned long convdiv)
2299 #define TMPBUFLEN 21
2300 unsigned long *i, *min, *max, val;
2301 int vleft, first=1, neg;
2303 char buf[TMPBUFLEN], *p;
2304 char __user *s = buffer;
2306 if (!data || !table->maxlen || !*lenp ||
2307 (*ppos && !write)) {
2312 i = (unsigned long *) data;
2313 min = (unsigned long *) table->extra1;
2314 max = (unsigned long *) table->extra2;
2315 vleft = table->maxlen / sizeof(unsigned long);
2318 for (; left && vleft--; i++, min++, max++, first=0) {
2333 if (len > TMPBUFLEN-1)
2335 if (copy_from_user(buf, s, len))
2339 if (*p == '-' && left > 1) {
2343 if (*p < '0' || *p > '9')
2345 val = simple_strtoul(p, &p, 0) * convmul / convdiv ;
2347 if ((len < left) && *p && !isspace(*p))
2356 if ((min && val < *min) || (max && val > *max))
2363 sprintf(p, "%lu", convdiv * (*i) / convmul);
2367 if(copy_to_user(s, buf, len))
2374 if (!write && !first && left) {
2375 if(put_user('\n', s))
2382 if (get_user(c, s++))
2397 static int do_proc_doulongvec_minmax(struct ctl_table *table, int write,
2398 void __user *buffer,
2399 size_t *lenp, loff_t *ppos,
2400 unsigned long convmul,
2401 unsigned long convdiv)
2403 return __do_proc_doulongvec_minmax(table->data, table, write,
2404 buffer, lenp, ppos, convmul, convdiv);
2408 * proc_doulongvec_minmax - read a vector of long integers with min/max values
2409 * @table: the sysctl table
2410 * @write: %TRUE if this is a write to the sysctl file
2411 * @buffer: the user buffer
2412 * @lenp: the size of the user buffer
2413 * @ppos: file position
2415 * Reads/writes up to table->maxlen/sizeof(unsigned long) unsigned long
2416 * values from/to the user buffer, treated as an ASCII string.
2418 * This routine will ensure the values are within the range specified by
2419 * table->extra1 (min) and table->extra2 (max).
2421 * Returns 0 on success.
2423 int proc_doulongvec_minmax(struct ctl_table *table, int write,
2424 void __user *buffer, size_t *lenp, loff_t *ppos)
2426 return do_proc_doulongvec_minmax(table, write, buffer, lenp, ppos, 1l, 1l);
2430 * proc_doulongvec_ms_jiffies_minmax - read a vector of millisecond values with min/max values
2431 * @table: the sysctl table
2432 * @write: %TRUE if this is a write to the sysctl file
2433 * @buffer: the user buffer
2434 * @lenp: the size of the user buffer
2435 * @ppos: file position
2437 * Reads/writes up to table->maxlen/sizeof(unsigned long) unsigned long
2438 * values from/to the user buffer, treated as an ASCII string. The values
2439 * are treated as milliseconds, and converted to jiffies when they are stored.
2441 * This routine will ensure the values are within the range specified by
2442 * table->extra1 (min) and table->extra2 (max).
2444 * Returns 0 on success.
2446 int proc_doulongvec_ms_jiffies_minmax(struct ctl_table *table, int write,
2447 void __user *buffer,
2448 size_t *lenp, loff_t *ppos)
2450 return do_proc_doulongvec_minmax(table, write, buffer,
2451 lenp, ppos, HZ, 1000l);
2455 static int do_proc_dointvec_jiffies_conv(int *negp, unsigned long *lvalp,
2457 int write, void *data)
2460 if (*lvalp > LONG_MAX / HZ)
2462 *valp = *negp ? -(*lvalp*HZ) : (*lvalp*HZ);
2468 lval = (unsigned long)-val;
2471 lval = (unsigned long)val;
2478 static int do_proc_dointvec_userhz_jiffies_conv(int *negp, unsigned long *lvalp,
2480 int write, void *data)
2483 if (USER_HZ < HZ && *lvalp > (LONG_MAX / HZ) * USER_HZ)
2485 *valp = clock_t_to_jiffies(*negp ? -*lvalp : *lvalp);
2491 lval = (unsigned long)-val;
2494 lval = (unsigned long)val;
2496 *lvalp = jiffies_to_clock_t(lval);
2501 static int do_proc_dointvec_ms_jiffies_conv(int *negp, unsigned long *lvalp,
2503 int write, void *data)
2506 *valp = msecs_to_jiffies(*negp ? -*lvalp : *lvalp);
2512 lval = (unsigned long)-val;
2515 lval = (unsigned long)val;
2517 *lvalp = jiffies_to_msecs(lval);
2523 * proc_dointvec_jiffies - read a vector of integers as seconds
2524 * @table: the sysctl table
2525 * @write: %TRUE if this is a write to the sysctl file
2526 * @buffer: the user buffer
2527 * @lenp: the size of the user buffer
2528 * @ppos: file position
2530 * Reads/writes up to table->maxlen/sizeof(unsigned int) integer
2531 * values from/to the user buffer, treated as an ASCII string.
2532 * The values read are assumed to be in seconds, and are converted into
2535 * Returns 0 on success.
2537 int proc_dointvec_jiffies(struct ctl_table *table, int write,
2538 void __user *buffer, size_t *lenp, loff_t *ppos)
2540 return do_proc_dointvec(table,write,buffer,lenp,ppos,
2541 do_proc_dointvec_jiffies_conv,NULL);
2545 * proc_dointvec_userhz_jiffies - read a vector of integers as 1/USER_HZ seconds
2546 * @table: the sysctl table
2547 * @write: %TRUE if this is a write to the sysctl file
2548 * @buffer: the user buffer
2549 * @lenp: the size of the user buffer
2550 * @ppos: pointer to the file position
2552 * Reads/writes up to table->maxlen/sizeof(unsigned int) integer
2553 * values from/to the user buffer, treated as an ASCII string.
2554 * The values read are assumed to be in 1/USER_HZ seconds, and
2555 * are converted into jiffies.
2557 * Returns 0 on success.
2559 int proc_dointvec_userhz_jiffies(struct ctl_table *table, int write,
2560 void __user *buffer, size_t *lenp, loff_t *ppos)
2562 return do_proc_dointvec(table,write,buffer,lenp,ppos,
2563 do_proc_dointvec_userhz_jiffies_conv,NULL);
2567 * proc_dointvec_ms_jiffies - read a vector of integers as 1 milliseconds
2568 * @table: the sysctl table
2569 * @write: %TRUE if this is a write to the sysctl file
2570 * @buffer: the user buffer
2571 * @lenp: the size of the user buffer
2572 * @ppos: file position
2573 * @ppos: the current position in the file
2575 * Reads/writes up to table->maxlen/sizeof(unsigned int) integer
2576 * values from/to the user buffer, treated as an ASCII string.
2577 * The values read are assumed to be in 1/1000 seconds, and
2578 * are converted into jiffies.
2580 * Returns 0 on success.
2582 int proc_dointvec_ms_jiffies(struct ctl_table *table, int write,
2583 void __user *buffer, size_t *lenp, loff_t *ppos)
2585 return do_proc_dointvec(table, write, buffer, lenp, ppos,
2586 do_proc_dointvec_ms_jiffies_conv, NULL);
2589 static int proc_do_cad_pid(struct ctl_table *table, int write,
2590 void __user *buffer, size_t *lenp, loff_t *ppos)
2592 struct pid *new_pid;
2596 tmp = pid_vnr(cad_pid);
2598 r = __do_proc_dointvec(&tmp, table, write, buffer,
2599 lenp, ppos, NULL, NULL);
2603 new_pid = find_get_pid(tmp);
2607 put_pid(xchg(&cad_pid, new_pid));
2611 #else /* CONFIG_PROC_FS */
2613 int proc_dostring(struct ctl_table *table, int write,
2614 void __user *buffer, size_t *lenp, loff_t *ppos)
2619 int proc_dointvec(struct ctl_table *table, int write,
2620 void __user *buffer, size_t *lenp, loff_t *ppos)
2625 int proc_dointvec_minmax(struct ctl_table *table, int write,
2626 void __user *buffer, size_t *lenp, loff_t *ppos)
2631 int proc_dointvec_jiffies(struct ctl_table *table, int write,
2632 void __user *buffer, size_t *lenp, loff_t *ppos)
2637 int proc_dointvec_userhz_jiffies(struct ctl_table *table, int write,
2638 void __user *buffer, size_t *lenp, loff_t *ppos)
2643 int proc_dointvec_ms_jiffies(struct ctl_table *table, int write,
2644 void __user *buffer, size_t *lenp, loff_t *ppos)
2649 int proc_doulongvec_minmax(struct ctl_table *table, int write,
2650 void __user *buffer, size_t *lenp, loff_t *ppos)
2655 int proc_doulongvec_ms_jiffies_minmax(struct ctl_table *table, int write,
2656 void __user *buffer,
2657 size_t *lenp, loff_t *ppos)
2663 #endif /* CONFIG_PROC_FS */
2666 * No sense putting this after each symbol definition, twice,
2667 * exception granted :-)
2669 EXPORT_SYMBOL(proc_dointvec);
2670 EXPORT_SYMBOL(proc_dointvec_jiffies);
2671 EXPORT_SYMBOL(proc_dointvec_minmax);
2672 EXPORT_SYMBOL(proc_dointvec_userhz_jiffies);
2673 EXPORT_SYMBOL(proc_dointvec_ms_jiffies);
2674 EXPORT_SYMBOL(proc_dostring);
2675 EXPORT_SYMBOL(proc_doulongvec_minmax);
2676 EXPORT_SYMBOL(proc_doulongvec_ms_jiffies_minmax);
2677 EXPORT_SYMBOL(register_sysctl_table);
2678 EXPORT_SYMBOL(register_sysctl_paths);
2679 EXPORT_SYMBOL(unregister_sysctl_table);