Btrfs: simplfy calculation of stripe length for discard operation
[linux-2.6/linux-acpi-2.6/ibm-acpi-2.6.git] / kernel / sysctl.c
blob2d2ecdcc8cdbb070999d46ae79b158dbbb28c8a3
1 /*
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
12 * Horn.
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
16 * Wendling.
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>
22 #include <linux/mm.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>
32 #include <linux/fs.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>
59 #include <linux/kmod.h>
61 #include <asm/uaccess.h>
62 #include <asm/processor.h>
64 #ifdef CONFIG_X86
65 #include <asm/nmi.h>
66 #include <asm/stacktrace.h>
67 #include <asm/io.h>
68 #endif
69 #ifdef CONFIG_BSD_PROCESS_ACCT
70 #include <linux/acct.h>
71 #endif
72 #ifdef CONFIG_RT_MUTEXES
73 #include <linux/rtmutex.h>
74 #endif
75 #if defined(CONFIG_PROVE_LOCKING) || defined(CONFIG_LOCK_STAT)
76 #include <linux/lockdep.h>
77 #endif
78 #ifdef CONFIG_CHR_DEV_SG
79 #include <scsi/sg.h>
80 #endif
82 #ifdef CONFIG_LOCKUP_DETECTOR
83 #include <linux/nmi.h>
84 #endif
87 #if defined(CONFIG_SYSCTL)
89 /* External variables not in a header file. */
90 extern int sysctl_overcommit_memory;
91 extern int sysctl_overcommit_ratio;
92 extern int max_threads;
93 extern int core_uses_pid;
94 extern int suid_dumpable;
95 extern char core_pattern[];
96 extern unsigned int core_pipe_limit;
97 extern int pid_max;
98 extern int min_free_kbytes;
99 extern int pid_max_min, pid_max_max;
100 extern int sysctl_drop_caches;
101 extern int percpu_pagelist_fraction;
102 extern int compat_log;
103 extern int latencytop_enabled;
104 extern int sysctl_nr_open_min, sysctl_nr_open_max;
105 #ifndef CONFIG_MMU
106 extern int sysctl_nr_trim_pages;
107 #endif
108 #ifdef CONFIG_BLOCK
109 extern int blk_iopoll_enabled;
110 #endif
112 /* Constants used for minimum and maximum */
113 #ifdef CONFIG_LOCKUP_DETECTOR
114 static int sixty = 60;
115 static int neg_one = -1;
116 #endif
118 static int zero;
119 static int __maybe_unused one = 1;
120 static int __maybe_unused two = 2;
121 static int __maybe_unused three = 3;
122 static unsigned long one_ul = 1;
123 static int one_hundred = 100;
124 #ifdef CONFIG_PRINTK
125 static int ten_thousand = 10000;
126 #endif
128 /* this is needed for the proc_doulongvec_minmax of vm_dirty_bytes */
129 static unsigned long dirty_bytes_min = 2 * PAGE_SIZE;
131 /* this is needed for the proc_dointvec_minmax for [fs_]overflow UID and GID */
132 static int maxolduid = 65535;
133 static int minolduid;
134 static int min_percpu_pagelist_fract = 8;
136 static int ngroups_max = NGROUPS_MAX;
138 #ifdef CONFIG_INOTIFY_USER
139 #include <linux/inotify.h>
140 #endif
141 #ifdef CONFIG_SPARC
142 #include <asm/system.h>
143 #endif
145 #ifdef CONFIG_SPARC64
146 extern int sysctl_tsb_ratio;
147 #endif
149 #ifdef __hppa__
150 extern int pwrsw_enabled;
151 extern int unaligned_enabled;
152 #endif
154 #ifdef CONFIG_S390
155 #ifdef CONFIG_MATHEMU
156 extern int sysctl_ieee_emulation_warnings;
157 #endif
158 extern int sysctl_userprocess_debug;
159 extern int spin_retry;
160 #endif
162 #ifdef CONFIG_IA64
163 extern int no_unaligned_warning;
164 extern int unaligned_dump_stack;
165 #endif
167 #ifdef CONFIG_PROC_SYSCTL
168 static int proc_do_cad_pid(struct ctl_table *table, int write,
169 void __user *buffer, size_t *lenp, loff_t *ppos);
170 static int proc_taint(struct ctl_table *table, int write,
171 void __user *buffer, size_t *lenp, loff_t *ppos);
172 #endif
174 #ifdef CONFIG_PRINTK
175 static int proc_dmesg_restrict(struct ctl_table *table, int write,
176 void __user *buffer, size_t *lenp, loff_t *ppos);
177 #endif
179 #ifdef CONFIG_MAGIC_SYSRQ
180 /* Note: sysrq code uses it's own private copy */
181 static int __sysrq_enabled = SYSRQ_DEFAULT_ENABLE;
183 static int sysrq_sysctl_handler(ctl_table *table, int write,
184 void __user *buffer, size_t *lenp,
185 loff_t *ppos)
187 int error;
189 error = proc_dointvec(table, write, buffer, lenp, ppos);
190 if (error)
191 return error;
193 if (write)
194 sysrq_toggle_support(__sysrq_enabled);
196 return 0;
199 #endif
201 static struct ctl_table root_table[];
202 static struct ctl_table_root sysctl_table_root;
203 static struct ctl_table_header root_table_header = {
204 {{.count = 1,
205 .ctl_table = root_table,
206 .ctl_entry = LIST_HEAD_INIT(sysctl_table_root.default_set.list),}},
207 .root = &sysctl_table_root,
208 .set = &sysctl_table_root.default_set,
210 static struct ctl_table_root sysctl_table_root = {
211 .root_list = LIST_HEAD_INIT(sysctl_table_root.root_list),
212 .default_set.list = LIST_HEAD_INIT(root_table_header.ctl_entry),
215 static struct ctl_table kern_table[];
216 static struct ctl_table vm_table[];
217 static struct ctl_table fs_table[];
218 static struct ctl_table debug_table[];
219 static struct ctl_table dev_table[];
220 extern struct ctl_table random_table[];
221 #ifdef CONFIG_EPOLL
222 extern struct ctl_table epoll_table[];
223 #endif
225 #ifdef HAVE_ARCH_PICK_MMAP_LAYOUT
226 int sysctl_legacy_va_layout;
227 #endif
229 /* The default sysctl tables: */
231 static struct ctl_table root_table[] = {
233 .procname = "kernel",
234 .mode = 0555,
235 .child = kern_table,
238 .procname = "vm",
239 .mode = 0555,
240 .child = vm_table,
243 .procname = "fs",
244 .mode = 0555,
245 .child = fs_table,
248 .procname = "debug",
249 .mode = 0555,
250 .child = debug_table,
253 .procname = "dev",
254 .mode = 0555,
255 .child = dev_table,
260 #ifdef CONFIG_SCHED_DEBUG
261 static int min_sched_granularity_ns = 100000; /* 100 usecs */
262 static int max_sched_granularity_ns = NSEC_PER_SEC; /* 1 second */
263 static int min_wakeup_granularity_ns; /* 0 usecs */
264 static int max_wakeup_granularity_ns = NSEC_PER_SEC; /* 1 second */
265 static int min_sched_tunable_scaling = SCHED_TUNABLESCALING_NONE;
266 static int max_sched_tunable_scaling = SCHED_TUNABLESCALING_END-1;
267 #endif
269 #ifdef CONFIG_COMPACTION
270 static int min_extfrag_threshold;
271 static int max_extfrag_threshold = 1000;
272 #endif
274 static struct ctl_table kern_table[] = {
276 .procname = "sched_child_runs_first",
277 .data = &sysctl_sched_child_runs_first,
278 .maxlen = sizeof(unsigned int),
279 .mode = 0644,
280 .proc_handler = proc_dointvec,
282 #ifdef CONFIG_SCHED_DEBUG
284 .procname = "sched_min_granularity_ns",
285 .data = &sysctl_sched_min_granularity,
286 .maxlen = sizeof(unsigned int),
287 .mode = 0644,
288 .proc_handler = sched_proc_update_handler,
289 .extra1 = &min_sched_granularity_ns,
290 .extra2 = &max_sched_granularity_ns,
293 .procname = "sched_latency_ns",
294 .data = &sysctl_sched_latency,
295 .maxlen = sizeof(unsigned int),
296 .mode = 0644,
297 .proc_handler = sched_proc_update_handler,
298 .extra1 = &min_sched_granularity_ns,
299 .extra2 = &max_sched_granularity_ns,
302 .procname = "sched_wakeup_granularity_ns",
303 .data = &sysctl_sched_wakeup_granularity,
304 .maxlen = sizeof(unsigned int),
305 .mode = 0644,
306 .proc_handler = sched_proc_update_handler,
307 .extra1 = &min_wakeup_granularity_ns,
308 .extra2 = &max_wakeup_granularity_ns,
311 .procname = "sched_tunable_scaling",
312 .data = &sysctl_sched_tunable_scaling,
313 .maxlen = sizeof(enum sched_tunable_scaling),
314 .mode = 0644,
315 .proc_handler = sched_proc_update_handler,
316 .extra1 = &min_sched_tunable_scaling,
317 .extra2 = &max_sched_tunable_scaling,
320 .procname = "sched_migration_cost",
321 .data = &sysctl_sched_migration_cost,
322 .maxlen = sizeof(unsigned int),
323 .mode = 0644,
324 .proc_handler = proc_dointvec,
327 .procname = "sched_nr_migrate",
328 .data = &sysctl_sched_nr_migrate,
329 .maxlen = sizeof(unsigned int),
330 .mode = 0644,
331 .proc_handler = proc_dointvec,
334 .procname = "sched_time_avg",
335 .data = &sysctl_sched_time_avg,
336 .maxlen = sizeof(unsigned int),
337 .mode = 0644,
338 .proc_handler = proc_dointvec,
341 .procname = "sched_shares_window",
342 .data = &sysctl_sched_shares_window,
343 .maxlen = sizeof(unsigned int),
344 .mode = 0644,
345 .proc_handler = proc_dointvec,
348 .procname = "timer_migration",
349 .data = &sysctl_timer_migration,
350 .maxlen = sizeof(unsigned int),
351 .mode = 0644,
352 .proc_handler = proc_dointvec_minmax,
353 .extra1 = &zero,
354 .extra2 = &one,
356 #endif
358 .procname = "sched_rt_period_us",
359 .data = &sysctl_sched_rt_period,
360 .maxlen = sizeof(unsigned int),
361 .mode = 0644,
362 .proc_handler = sched_rt_handler,
365 .procname = "sched_rt_runtime_us",
366 .data = &sysctl_sched_rt_runtime,
367 .maxlen = sizeof(int),
368 .mode = 0644,
369 .proc_handler = sched_rt_handler,
371 #ifdef CONFIG_SCHED_AUTOGROUP
373 .procname = "sched_autogroup_enabled",
374 .data = &sysctl_sched_autogroup_enabled,
375 .maxlen = sizeof(unsigned int),
376 .mode = 0644,
377 .proc_handler = proc_dointvec_minmax,
378 .extra1 = &zero,
379 .extra2 = &one,
381 #endif
382 #ifdef CONFIG_CFS_BANDWIDTH
384 .procname = "sched_cfs_bandwidth_slice_us",
385 .data = &sysctl_sched_cfs_bandwidth_slice,
386 .maxlen = sizeof(unsigned int),
387 .mode = 0644,
388 .proc_handler = proc_dointvec_minmax,
389 .extra1 = &one,
391 #endif
392 #ifdef CONFIG_PROVE_LOCKING
394 .procname = "prove_locking",
395 .data = &prove_locking,
396 .maxlen = sizeof(int),
397 .mode = 0644,
398 .proc_handler = proc_dointvec,
400 #endif
401 #ifdef CONFIG_LOCK_STAT
403 .procname = "lock_stat",
404 .data = &lock_stat,
405 .maxlen = sizeof(int),
406 .mode = 0644,
407 .proc_handler = proc_dointvec,
409 #endif
411 .procname = "panic",
412 .data = &panic_timeout,
413 .maxlen = sizeof(int),
414 .mode = 0644,
415 .proc_handler = proc_dointvec,
418 .procname = "core_uses_pid",
419 .data = &core_uses_pid,
420 .maxlen = sizeof(int),
421 .mode = 0644,
422 .proc_handler = proc_dointvec,
425 .procname = "core_pattern",
426 .data = core_pattern,
427 .maxlen = CORENAME_MAX_SIZE,
428 .mode = 0644,
429 .proc_handler = proc_dostring,
432 .procname = "core_pipe_limit",
433 .data = &core_pipe_limit,
434 .maxlen = sizeof(unsigned int),
435 .mode = 0644,
436 .proc_handler = proc_dointvec,
438 #ifdef CONFIG_PROC_SYSCTL
440 .procname = "tainted",
441 .maxlen = sizeof(long),
442 .mode = 0644,
443 .proc_handler = proc_taint,
445 #endif
446 #ifdef CONFIG_LATENCYTOP
448 .procname = "latencytop",
449 .data = &latencytop_enabled,
450 .maxlen = sizeof(int),
451 .mode = 0644,
452 .proc_handler = proc_dointvec,
454 #endif
455 #ifdef CONFIG_BLK_DEV_INITRD
457 .procname = "real-root-dev",
458 .data = &real_root_dev,
459 .maxlen = sizeof(int),
460 .mode = 0644,
461 .proc_handler = proc_dointvec,
463 #endif
465 .procname = "print-fatal-signals",
466 .data = &print_fatal_signals,
467 .maxlen = sizeof(int),
468 .mode = 0644,
469 .proc_handler = proc_dointvec,
471 #ifdef CONFIG_SPARC
473 .procname = "reboot-cmd",
474 .data = reboot_command,
475 .maxlen = 256,
476 .mode = 0644,
477 .proc_handler = proc_dostring,
480 .procname = "stop-a",
481 .data = &stop_a_enabled,
482 .maxlen = sizeof (int),
483 .mode = 0644,
484 .proc_handler = proc_dointvec,
487 .procname = "scons-poweroff",
488 .data = &scons_pwroff,
489 .maxlen = sizeof (int),
490 .mode = 0644,
491 .proc_handler = proc_dointvec,
493 #endif
494 #ifdef CONFIG_SPARC64
496 .procname = "tsb-ratio",
497 .data = &sysctl_tsb_ratio,
498 .maxlen = sizeof (int),
499 .mode = 0644,
500 .proc_handler = proc_dointvec,
502 #endif
503 #ifdef __hppa__
505 .procname = "soft-power",
506 .data = &pwrsw_enabled,
507 .maxlen = sizeof (int),
508 .mode = 0644,
509 .proc_handler = proc_dointvec,
512 .procname = "unaligned-trap",
513 .data = &unaligned_enabled,
514 .maxlen = sizeof (int),
515 .mode = 0644,
516 .proc_handler = proc_dointvec,
518 #endif
520 .procname = "ctrl-alt-del",
521 .data = &C_A_D,
522 .maxlen = sizeof(int),
523 .mode = 0644,
524 .proc_handler = proc_dointvec,
526 #ifdef CONFIG_FUNCTION_TRACER
528 .procname = "ftrace_enabled",
529 .data = &ftrace_enabled,
530 .maxlen = sizeof(int),
531 .mode = 0644,
532 .proc_handler = ftrace_enable_sysctl,
534 #endif
535 #ifdef CONFIG_STACK_TRACER
537 .procname = "stack_tracer_enabled",
538 .data = &stack_tracer_enabled,
539 .maxlen = sizeof(int),
540 .mode = 0644,
541 .proc_handler = stack_trace_sysctl,
543 #endif
544 #ifdef CONFIG_TRACING
546 .procname = "ftrace_dump_on_oops",
547 .data = &ftrace_dump_on_oops,
548 .maxlen = sizeof(int),
549 .mode = 0644,
550 .proc_handler = proc_dointvec,
552 #endif
553 #ifdef CONFIG_MODULES
555 .procname = "modprobe",
556 .data = &modprobe_path,
557 .maxlen = KMOD_PATH_LEN,
558 .mode = 0644,
559 .proc_handler = proc_dostring,
562 .procname = "modules_disabled",
563 .data = &modules_disabled,
564 .maxlen = sizeof(int),
565 .mode = 0644,
566 /* only handle a transition from default "0" to "1" */
567 .proc_handler = proc_dointvec_minmax,
568 .extra1 = &one,
569 .extra2 = &one,
571 #endif
572 #ifdef CONFIG_HOTPLUG
574 .procname = "hotplug",
575 .data = &uevent_helper,
576 .maxlen = UEVENT_HELPER_PATH_LEN,
577 .mode = 0644,
578 .proc_handler = proc_dostring,
580 #endif
581 #ifdef CONFIG_CHR_DEV_SG
583 .procname = "sg-big-buff",
584 .data = &sg_big_buff,
585 .maxlen = sizeof (int),
586 .mode = 0444,
587 .proc_handler = proc_dointvec,
589 #endif
590 #ifdef CONFIG_BSD_PROCESS_ACCT
592 .procname = "acct",
593 .data = &acct_parm,
594 .maxlen = 3*sizeof(int),
595 .mode = 0644,
596 .proc_handler = proc_dointvec,
598 #endif
599 #ifdef CONFIG_MAGIC_SYSRQ
601 .procname = "sysrq",
602 .data = &__sysrq_enabled,
603 .maxlen = sizeof (int),
604 .mode = 0644,
605 .proc_handler = sysrq_sysctl_handler,
607 #endif
608 #ifdef CONFIG_PROC_SYSCTL
610 .procname = "cad_pid",
611 .data = NULL,
612 .maxlen = sizeof (int),
613 .mode = 0600,
614 .proc_handler = proc_do_cad_pid,
616 #endif
618 .procname = "threads-max",
619 .data = &max_threads,
620 .maxlen = sizeof(int),
621 .mode = 0644,
622 .proc_handler = proc_dointvec,
625 .procname = "random",
626 .mode = 0555,
627 .child = random_table,
630 .procname = "usermodehelper",
631 .mode = 0555,
632 .child = usermodehelper_table,
635 .procname = "overflowuid",
636 .data = &overflowuid,
637 .maxlen = sizeof(int),
638 .mode = 0644,
639 .proc_handler = proc_dointvec_minmax,
640 .extra1 = &minolduid,
641 .extra2 = &maxolduid,
644 .procname = "overflowgid",
645 .data = &overflowgid,
646 .maxlen = sizeof(int),
647 .mode = 0644,
648 .proc_handler = proc_dointvec_minmax,
649 .extra1 = &minolduid,
650 .extra2 = &maxolduid,
652 #ifdef CONFIG_S390
653 #ifdef CONFIG_MATHEMU
655 .procname = "ieee_emulation_warnings",
656 .data = &sysctl_ieee_emulation_warnings,
657 .maxlen = sizeof(int),
658 .mode = 0644,
659 .proc_handler = proc_dointvec,
661 #endif
663 .procname = "userprocess_debug",
664 .data = &show_unhandled_signals,
665 .maxlen = sizeof(int),
666 .mode = 0644,
667 .proc_handler = proc_dointvec,
669 #endif
671 .procname = "pid_max",
672 .data = &pid_max,
673 .maxlen = sizeof (int),
674 .mode = 0644,
675 .proc_handler = proc_dointvec_minmax,
676 .extra1 = &pid_max_min,
677 .extra2 = &pid_max_max,
680 .procname = "panic_on_oops",
681 .data = &panic_on_oops,
682 .maxlen = sizeof(int),
683 .mode = 0644,
684 .proc_handler = proc_dointvec,
686 #if defined CONFIG_PRINTK
688 .procname = "printk",
689 .data = &console_loglevel,
690 .maxlen = 4*sizeof(int),
691 .mode = 0644,
692 .proc_handler = proc_dointvec,
695 .procname = "printk_ratelimit",
696 .data = &printk_ratelimit_state.interval,
697 .maxlen = sizeof(int),
698 .mode = 0644,
699 .proc_handler = proc_dointvec_jiffies,
702 .procname = "printk_ratelimit_burst",
703 .data = &printk_ratelimit_state.burst,
704 .maxlen = sizeof(int),
705 .mode = 0644,
706 .proc_handler = proc_dointvec,
709 .procname = "printk_delay",
710 .data = &printk_delay_msec,
711 .maxlen = sizeof(int),
712 .mode = 0644,
713 .proc_handler = proc_dointvec_minmax,
714 .extra1 = &zero,
715 .extra2 = &ten_thousand,
718 .procname = "dmesg_restrict",
719 .data = &dmesg_restrict,
720 .maxlen = sizeof(int),
721 .mode = 0644,
722 .proc_handler = proc_dointvec_minmax,
723 .extra1 = &zero,
724 .extra2 = &one,
727 .procname = "kptr_restrict",
728 .data = &kptr_restrict,
729 .maxlen = sizeof(int),
730 .mode = 0644,
731 .proc_handler = proc_dmesg_restrict,
732 .extra1 = &zero,
733 .extra2 = &two,
735 #endif
737 .procname = "ngroups_max",
738 .data = &ngroups_max,
739 .maxlen = sizeof (int),
740 .mode = 0444,
741 .proc_handler = proc_dointvec,
743 #if defined(CONFIG_LOCKUP_DETECTOR)
745 .procname = "watchdog",
746 .data = &watchdog_enabled,
747 .maxlen = sizeof (int),
748 .mode = 0644,
749 .proc_handler = proc_dowatchdog,
750 .extra1 = &zero,
751 .extra2 = &one,
754 .procname = "watchdog_thresh",
755 .data = &watchdog_thresh,
756 .maxlen = sizeof(int),
757 .mode = 0644,
758 .proc_handler = proc_dowatchdog,
759 .extra1 = &neg_one,
760 .extra2 = &sixty,
763 .procname = "softlockup_panic",
764 .data = &softlockup_panic,
765 .maxlen = sizeof(int),
766 .mode = 0644,
767 .proc_handler = proc_dointvec_minmax,
768 .extra1 = &zero,
769 .extra2 = &one,
772 .procname = "nmi_watchdog",
773 .data = &watchdog_enabled,
774 .maxlen = sizeof (int),
775 .mode = 0644,
776 .proc_handler = proc_dowatchdog,
777 .extra1 = &zero,
778 .extra2 = &one,
780 #endif
781 #if defined(CONFIG_X86_LOCAL_APIC) && defined(CONFIG_X86)
783 .procname = "unknown_nmi_panic",
784 .data = &unknown_nmi_panic,
785 .maxlen = sizeof (int),
786 .mode = 0644,
787 .proc_handler = proc_dointvec,
789 #endif
790 #if defined(CONFIG_X86)
792 .procname = "panic_on_unrecovered_nmi",
793 .data = &panic_on_unrecovered_nmi,
794 .maxlen = sizeof(int),
795 .mode = 0644,
796 .proc_handler = proc_dointvec,
799 .procname = "panic_on_io_nmi",
800 .data = &panic_on_io_nmi,
801 .maxlen = sizeof(int),
802 .mode = 0644,
803 .proc_handler = proc_dointvec,
806 .procname = "bootloader_type",
807 .data = &bootloader_type,
808 .maxlen = sizeof (int),
809 .mode = 0444,
810 .proc_handler = proc_dointvec,
813 .procname = "bootloader_version",
814 .data = &bootloader_version,
815 .maxlen = sizeof (int),
816 .mode = 0444,
817 .proc_handler = proc_dointvec,
820 .procname = "kstack_depth_to_print",
821 .data = &kstack_depth_to_print,
822 .maxlen = sizeof(int),
823 .mode = 0644,
824 .proc_handler = proc_dointvec,
827 .procname = "io_delay_type",
828 .data = &io_delay_type,
829 .maxlen = sizeof(int),
830 .mode = 0644,
831 .proc_handler = proc_dointvec,
833 #endif
834 #if defined(CONFIG_MMU)
836 .procname = "randomize_va_space",
837 .data = &randomize_va_space,
838 .maxlen = sizeof(int),
839 .mode = 0644,
840 .proc_handler = proc_dointvec,
842 #endif
843 #if defined(CONFIG_S390) && defined(CONFIG_SMP)
845 .procname = "spin_retry",
846 .data = &spin_retry,
847 .maxlen = sizeof (int),
848 .mode = 0644,
849 .proc_handler = proc_dointvec,
851 #endif
852 #if defined(CONFIG_ACPI_SLEEP) && defined(CONFIG_X86)
854 .procname = "acpi_video_flags",
855 .data = &acpi_realmode_flags,
856 .maxlen = sizeof (unsigned long),
857 .mode = 0644,
858 .proc_handler = proc_doulongvec_minmax,
860 #endif
861 #ifdef CONFIG_IA64
863 .procname = "ignore-unaligned-usertrap",
864 .data = &no_unaligned_warning,
865 .maxlen = sizeof (int),
866 .mode = 0644,
867 .proc_handler = proc_dointvec,
870 .procname = "unaligned-dump-stack",
871 .data = &unaligned_dump_stack,
872 .maxlen = sizeof (int),
873 .mode = 0644,
874 .proc_handler = proc_dointvec,
876 #endif
877 #ifdef CONFIG_DETECT_HUNG_TASK
879 .procname = "hung_task_panic",
880 .data = &sysctl_hung_task_panic,
881 .maxlen = sizeof(int),
882 .mode = 0644,
883 .proc_handler = proc_dointvec_minmax,
884 .extra1 = &zero,
885 .extra2 = &one,
888 .procname = "hung_task_check_count",
889 .data = &sysctl_hung_task_check_count,
890 .maxlen = sizeof(unsigned long),
891 .mode = 0644,
892 .proc_handler = proc_doulongvec_minmax,
895 .procname = "hung_task_timeout_secs",
896 .data = &sysctl_hung_task_timeout_secs,
897 .maxlen = sizeof(unsigned long),
898 .mode = 0644,
899 .proc_handler = proc_dohung_task_timeout_secs,
902 .procname = "hung_task_warnings",
903 .data = &sysctl_hung_task_warnings,
904 .maxlen = sizeof(unsigned long),
905 .mode = 0644,
906 .proc_handler = proc_doulongvec_minmax,
908 #endif
909 #ifdef CONFIG_COMPAT
911 .procname = "compat-log",
912 .data = &compat_log,
913 .maxlen = sizeof (int),
914 .mode = 0644,
915 .proc_handler = proc_dointvec,
917 #endif
918 #ifdef CONFIG_RT_MUTEXES
920 .procname = "max_lock_depth",
921 .data = &max_lock_depth,
922 .maxlen = sizeof(int),
923 .mode = 0644,
924 .proc_handler = proc_dointvec,
926 #endif
928 .procname = "poweroff_cmd",
929 .data = &poweroff_cmd,
930 .maxlen = POWEROFF_CMD_PATH_LEN,
931 .mode = 0644,
932 .proc_handler = proc_dostring,
934 #ifdef CONFIG_KEYS
936 .procname = "keys",
937 .mode = 0555,
938 .child = key_sysctls,
940 #endif
941 #ifdef CONFIG_RCU_TORTURE_TEST
943 .procname = "rcutorture_runnable",
944 .data = &rcutorture_runnable,
945 .maxlen = sizeof(int),
946 .mode = 0644,
947 .proc_handler = proc_dointvec,
949 #endif
950 #ifdef CONFIG_PERF_EVENTS
952 * User-space scripts rely on the existence of this file
953 * as a feature check for perf_events being enabled.
955 * So it's an ABI, do not remove!
958 .procname = "perf_event_paranoid",
959 .data = &sysctl_perf_event_paranoid,
960 .maxlen = sizeof(sysctl_perf_event_paranoid),
961 .mode = 0644,
962 .proc_handler = proc_dointvec,
965 .procname = "perf_event_mlock_kb",
966 .data = &sysctl_perf_event_mlock,
967 .maxlen = sizeof(sysctl_perf_event_mlock),
968 .mode = 0644,
969 .proc_handler = proc_dointvec,
972 .procname = "perf_event_max_sample_rate",
973 .data = &sysctl_perf_event_sample_rate,
974 .maxlen = sizeof(sysctl_perf_event_sample_rate),
975 .mode = 0644,
976 .proc_handler = perf_proc_update_handler,
978 #endif
979 #ifdef CONFIG_KMEMCHECK
981 .procname = "kmemcheck",
982 .data = &kmemcheck_enabled,
983 .maxlen = sizeof(int),
984 .mode = 0644,
985 .proc_handler = proc_dointvec,
987 #endif
988 #ifdef CONFIG_BLOCK
990 .procname = "blk_iopoll",
991 .data = &blk_iopoll_enabled,
992 .maxlen = sizeof(int),
993 .mode = 0644,
994 .proc_handler = proc_dointvec,
996 #endif
1000 static struct ctl_table vm_table[] = {
1002 .procname = "overcommit_memory",
1003 .data = &sysctl_overcommit_memory,
1004 .maxlen = sizeof(sysctl_overcommit_memory),
1005 .mode = 0644,
1006 .proc_handler = proc_dointvec_minmax,
1007 .extra1 = &zero,
1008 .extra2 = &two,
1011 .procname = "panic_on_oom",
1012 .data = &sysctl_panic_on_oom,
1013 .maxlen = sizeof(sysctl_panic_on_oom),
1014 .mode = 0644,
1015 .proc_handler = proc_dointvec_minmax,
1016 .extra1 = &zero,
1017 .extra2 = &two,
1020 .procname = "oom_kill_allocating_task",
1021 .data = &sysctl_oom_kill_allocating_task,
1022 .maxlen = sizeof(sysctl_oom_kill_allocating_task),
1023 .mode = 0644,
1024 .proc_handler = proc_dointvec,
1027 .procname = "oom_dump_tasks",
1028 .data = &sysctl_oom_dump_tasks,
1029 .maxlen = sizeof(sysctl_oom_dump_tasks),
1030 .mode = 0644,
1031 .proc_handler = proc_dointvec,
1034 .procname = "overcommit_ratio",
1035 .data = &sysctl_overcommit_ratio,
1036 .maxlen = sizeof(sysctl_overcommit_ratio),
1037 .mode = 0644,
1038 .proc_handler = proc_dointvec,
1041 .procname = "page-cluster",
1042 .data = &page_cluster,
1043 .maxlen = sizeof(int),
1044 .mode = 0644,
1045 .proc_handler = proc_dointvec_minmax,
1046 .extra1 = &zero,
1049 .procname = "dirty_background_ratio",
1050 .data = &dirty_background_ratio,
1051 .maxlen = sizeof(dirty_background_ratio),
1052 .mode = 0644,
1053 .proc_handler = dirty_background_ratio_handler,
1054 .extra1 = &zero,
1055 .extra2 = &one_hundred,
1058 .procname = "dirty_background_bytes",
1059 .data = &dirty_background_bytes,
1060 .maxlen = sizeof(dirty_background_bytes),
1061 .mode = 0644,
1062 .proc_handler = dirty_background_bytes_handler,
1063 .extra1 = &one_ul,
1066 .procname = "dirty_ratio",
1067 .data = &vm_dirty_ratio,
1068 .maxlen = sizeof(vm_dirty_ratio),
1069 .mode = 0644,
1070 .proc_handler = dirty_ratio_handler,
1071 .extra1 = &zero,
1072 .extra2 = &one_hundred,
1075 .procname = "dirty_bytes",
1076 .data = &vm_dirty_bytes,
1077 .maxlen = sizeof(vm_dirty_bytes),
1078 .mode = 0644,
1079 .proc_handler = dirty_bytes_handler,
1080 .extra1 = &dirty_bytes_min,
1083 .procname = "dirty_writeback_centisecs",
1084 .data = &dirty_writeback_interval,
1085 .maxlen = sizeof(dirty_writeback_interval),
1086 .mode = 0644,
1087 .proc_handler = dirty_writeback_centisecs_handler,
1090 .procname = "dirty_expire_centisecs",
1091 .data = &dirty_expire_interval,
1092 .maxlen = sizeof(dirty_expire_interval),
1093 .mode = 0644,
1094 .proc_handler = proc_dointvec_minmax,
1095 .extra1 = &zero,
1098 .procname = "nr_pdflush_threads",
1099 .data = &nr_pdflush_threads,
1100 .maxlen = sizeof nr_pdflush_threads,
1101 .mode = 0444 /* read-only*/,
1102 .proc_handler = proc_dointvec,
1105 .procname = "swappiness",
1106 .data = &vm_swappiness,
1107 .maxlen = sizeof(vm_swappiness),
1108 .mode = 0644,
1109 .proc_handler = proc_dointvec_minmax,
1110 .extra1 = &zero,
1111 .extra2 = &one_hundred,
1113 #ifdef CONFIG_HUGETLB_PAGE
1115 .procname = "nr_hugepages",
1116 .data = NULL,
1117 .maxlen = sizeof(unsigned long),
1118 .mode = 0644,
1119 .proc_handler = hugetlb_sysctl_handler,
1120 .extra1 = (void *)&hugetlb_zero,
1121 .extra2 = (void *)&hugetlb_infinity,
1123 #ifdef CONFIG_NUMA
1125 .procname = "nr_hugepages_mempolicy",
1126 .data = NULL,
1127 .maxlen = sizeof(unsigned long),
1128 .mode = 0644,
1129 .proc_handler = &hugetlb_mempolicy_sysctl_handler,
1130 .extra1 = (void *)&hugetlb_zero,
1131 .extra2 = (void *)&hugetlb_infinity,
1133 #endif
1135 .procname = "hugetlb_shm_group",
1136 .data = &sysctl_hugetlb_shm_group,
1137 .maxlen = sizeof(gid_t),
1138 .mode = 0644,
1139 .proc_handler = proc_dointvec,
1142 .procname = "hugepages_treat_as_movable",
1143 .data = &hugepages_treat_as_movable,
1144 .maxlen = sizeof(int),
1145 .mode = 0644,
1146 .proc_handler = hugetlb_treat_movable_handler,
1149 .procname = "nr_overcommit_hugepages",
1150 .data = NULL,
1151 .maxlen = sizeof(unsigned long),
1152 .mode = 0644,
1153 .proc_handler = hugetlb_overcommit_handler,
1154 .extra1 = (void *)&hugetlb_zero,
1155 .extra2 = (void *)&hugetlb_infinity,
1157 #endif
1159 .procname = "lowmem_reserve_ratio",
1160 .data = &sysctl_lowmem_reserve_ratio,
1161 .maxlen = sizeof(sysctl_lowmem_reserve_ratio),
1162 .mode = 0644,
1163 .proc_handler = lowmem_reserve_ratio_sysctl_handler,
1166 .procname = "drop_caches",
1167 .data = &sysctl_drop_caches,
1168 .maxlen = sizeof(int),
1169 .mode = 0644,
1170 .proc_handler = drop_caches_sysctl_handler,
1171 .extra1 = &one,
1172 .extra2 = &three,
1174 #ifdef CONFIG_COMPACTION
1176 .procname = "compact_memory",
1177 .data = &sysctl_compact_memory,
1178 .maxlen = sizeof(int),
1179 .mode = 0200,
1180 .proc_handler = sysctl_compaction_handler,
1183 .procname = "extfrag_threshold",
1184 .data = &sysctl_extfrag_threshold,
1185 .maxlen = sizeof(int),
1186 .mode = 0644,
1187 .proc_handler = sysctl_extfrag_handler,
1188 .extra1 = &min_extfrag_threshold,
1189 .extra2 = &max_extfrag_threshold,
1192 #endif /* CONFIG_COMPACTION */
1194 .procname = "min_free_kbytes",
1195 .data = &min_free_kbytes,
1196 .maxlen = sizeof(min_free_kbytes),
1197 .mode = 0644,
1198 .proc_handler = min_free_kbytes_sysctl_handler,
1199 .extra1 = &zero,
1202 .procname = "percpu_pagelist_fraction",
1203 .data = &percpu_pagelist_fraction,
1204 .maxlen = sizeof(percpu_pagelist_fraction),
1205 .mode = 0644,
1206 .proc_handler = percpu_pagelist_fraction_sysctl_handler,
1207 .extra1 = &min_percpu_pagelist_fract,
1209 #ifdef CONFIG_MMU
1211 .procname = "max_map_count",
1212 .data = &sysctl_max_map_count,
1213 .maxlen = sizeof(sysctl_max_map_count),
1214 .mode = 0644,
1215 .proc_handler = proc_dointvec_minmax,
1216 .extra1 = &zero,
1218 #else
1220 .procname = "nr_trim_pages",
1221 .data = &sysctl_nr_trim_pages,
1222 .maxlen = sizeof(sysctl_nr_trim_pages),
1223 .mode = 0644,
1224 .proc_handler = proc_dointvec_minmax,
1225 .extra1 = &zero,
1227 #endif
1229 .procname = "laptop_mode",
1230 .data = &laptop_mode,
1231 .maxlen = sizeof(laptop_mode),
1232 .mode = 0644,
1233 .proc_handler = proc_dointvec_jiffies,
1236 .procname = "block_dump",
1237 .data = &block_dump,
1238 .maxlen = sizeof(block_dump),
1239 .mode = 0644,
1240 .proc_handler = proc_dointvec,
1241 .extra1 = &zero,
1244 .procname = "vfs_cache_pressure",
1245 .data = &sysctl_vfs_cache_pressure,
1246 .maxlen = sizeof(sysctl_vfs_cache_pressure),
1247 .mode = 0644,
1248 .proc_handler = proc_dointvec,
1249 .extra1 = &zero,
1251 #ifdef HAVE_ARCH_PICK_MMAP_LAYOUT
1253 .procname = "legacy_va_layout",
1254 .data = &sysctl_legacy_va_layout,
1255 .maxlen = sizeof(sysctl_legacy_va_layout),
1256 .mode = 0644,
1257 .proc_handler = proc_dointvec,
1258 .extra1 = &zero,
1260 #endif
1261 #ifdef CONFIG_NUMA
1263 .procname = "zone_reclaim_mode",
1264 .data = &zone_reclaim_mode,
1265 .maxlen = sizeof(zone_reclaim_mode),
1266 .mode = 0644,
1267 .proc_handler = proc_dointvec,
1268 .extra1 = &zero,
1271 .procname = "min_unmapped_ratio",
1272 .data = &sysctl_min_unmapped_ratio,
1273 .maxlen = sizeof(sysctl_min_unmapped_ratio),
1274 .mode = 0644,
1275 .proc_handler = sysctl_min_unmapped_ratio_sysctl_handler,
1276 .extra1 = &zero,
1277 .extra2 = &one_hundred,
1280 .procname = "min_slab_ratio",
1281 .data = &sysctl_min_slab_ratio,
1282 .maxlen = sizeof(sysctl_min_slab_ratio),
1283 .mode = 0644,
1284 .proc_handler = sysctl_min_slab_ratio_sysctl_handler,
1285 .extra1 = &zero,
1286 .extra2 = &one_hundred,
1288 #endif
1289 #ifdef CONFIG_SMP
1291 .procname = "stat_interval",
1292 .data = &sysctl_stat_interval,
1293 .maxlen = sizeof(sysctl_stat_interval),
1294 .mode = 0644,
1295 .proc_handler = proc_dointvec_jiffies,
1297 #endif
1298 #ifdef CONFIG_MMU
1300 .procname = "mmap_min_addr",
1301 .data = &dac_mmap_min_addr,
1302 .maxlen = sizeof(unsigned long),
1303 .mode = 0644,
1304 .proc_handler = mmap_min_addr_handler,
1306 #endif
1307 #ifdef CONFIG_NUMA
1309 .procname = "numa_zonelist_order",
1310 .data = &numa_zonelist_order,
1311 .maxlen = NUMA_ZONELIST_ORDER_LEN,
1312 .mode = 0644,
1313 .proc_handler = numa_zonelist_order_handler,
1315 #endif
1316 #if (defined(CONFIG_X86_32) && !defined(CONFIG_UML))|| \
1317 (defined(CONFIG_SUPERH) && defined(CONFIG_VSYSCALL))
1319 .procname = "vdso_enabled",
1320 .data = &vdso_enabled,
1321 .maxlen = sizeof(vdso_enabled),
1322 .mode = 0644,
1323 .proc_handler = proc_dointvec,
1324 .extra1 = &zero,
1326 #endif
1327 #ifdef CONFIG_HIGHMEM
1329 .procname = "highmem_is_dirtyable",
1330 .data = &vm_highmem_is_dirtyable,
1331 .maxlen = sizeof(vm_highmem_is_dirtyable),
1332 .mode = 0644,
1333 .proc_handler = proc_dointvec_minmax,
1334 .extra1 = &zero,
1335 .extra2 = &one,
1337 #endif
1339 .procname = "scan_unevictable_pages",
1340 .data = &scan_unevictable_pages,
1341 .maxlen = sizeof(scan_unevictable_pages),
1342 .mode = 0644,
1343 .proc_handler = scan_unevictable_handler,
1345 #ifdef CONFIG_MEMORY_FAILURE
1347 .procname = "memory_failure_early_kill",
1348 .data = &sysctl_memory_failure_early_kill,
1349 .maxlen = sizeof(sysctl_memory_failure_early_kill),
1350 .mode = 0644,
1351 .proc_handler = proc_dointvec_minmax,
1352 .extra1 = &zero,
1353 .extra2 = &one,
1356 .procname = "memory_failure_recovery",
1357 .data = &sysctl_memory_failure_recovery,
1358 .maxlen = sizeof(sysctl_memory_failure_recovery),
1359 .mode = 0644,
1360 .proc_handler = proc_dointvec_minmax,
1361 .extra1 = &zero,
1362 .extra2 = &one,
1364 #endif
1368 #if defined(CONFIG_BINFMT_MISC) || defined(CONFIG_BINFMT_MISC_MODULE)
1369 static struct ctl_table binfmt_misc_table[] = {
1372 #endif
1374 static struct ctl_table fs_table[] = {
1376 .procname = "inode-nr",
1377 .data = &inodes_stat,
1378 .maxlen = 2*sizeof(int),
1379 .mode = 0444,
1380 .proc_handler = proc_nr_inodes,
1383 .procname = "inode-state",
1384 .data = &inodes_stat,
1385 .maxlen = 7*sizeof(int),
1386 .mode = 0444,
1387 .proc_handler = proc_nr_inodes,
1390 .procname = "file-nr",
1391 .data = &files_stat,
1392 .maxlen = sizeof(files_stat),
1393 .mode = 0444,
1394 .proc_handler = proc_nr_files,
1397 .procname = "file-max",
1398 .data = &files_stat.max_files,
1399 .maxlen = sizeof(files_stat.max_files),
1400 .mode = 0644,
1401 .proc_handler = proc_doulongvec_minmax,
1404 .procname = "nr_open",
1405 .data = &sysctl_nr_open,
1406 .maxlen = sizeof(int),
1407 .mode = 0644,
1408 .proc_handler = proc_dointvec_minmax,
1409 .extra1 = &sysctl_nr_open_min,
1410 .extra2 = &sysctl_nr_open_max,
1413 .procname = "dentry-state",
1414 .data = &dentry_stat,
1415 .maxlen = 6*sizeof(int),
1416 .mode = 0444,
1417 .proc_handler = proc_nr_dentry,
1420 .procname = "overflowuid",
1421 .data = &fs_overflowuid,
1422 .maxlen = sizeof(int),
1423 .mode = 0644,
1424 .proc_handler = proc_dointvec_minmax,
1425 .extra1 = &minolduid,
1426 .extra2 = &maxolduid,
1429 .procname = "overflowgid",
1430 .data = &fs_overflowgid,
1431 .maxlen = sizeof(int),
1432 .mode = 0644,
1433 .proc_handler = proc_dointvec_minmax,
1434 .extra1 = &minolduid,
1435 .extra2 = &maxolduid,
1437 #ifdef CONFIG_FILE_LOCKING
1439 .procname = "leases-enable",
1440 .data = &leases_enable,
1441 .maxlen = sizeof(int),
1442 .mode = 0644,
1443 .proc_handler = proc_dointvec,
1445 #endif
1446 #ifdef CONFIG_DNOTIFY
1448 .procname = "dir-notify-enable",
1449 .data = &dir_notify_enable,
1450 .maxlen = sizeof(int),
1451 .mode = 0644,
1452 .proc_handler = proc_dointvec,
1454 #endif
1455 #ifdef CONFIG_MMU
1456 #ifdef CONFIG_FILE_LOCKING
1458 .procname = "lease-break-time",
1459 .data = &lease_break_time,
1460 .maxlen = sizeof(int),
1461 .mode = 0644,
1462 .proc_handler = proc_dointvec,
1464 #endif
1465 #ifdef CONFIG_AIO
1467 .procname = "aio-nr",
1468 .data = &aio_nr,
1469 .maxlen = sizeof(aio_nr),
1470 .mode = 0444,
1471 .proc_handler = proc_doulongvec_minmax,
1474 .procname = "aio-max-nr",
1475 .data = &aio_max_nr,
1476 .maxlen = sizeof(aio_max_nr),
1477 .mode = 0644,
1478 .proc_handler = proc_doulongvec_minmax,
1480 #endif /* CONFIG_AIO */
1481 #ifdef CONFIG_INOTIFY_USER
1483 .procname = "inotify",
1484 .mode = 0555,
1485 .child = inotify_table,
1487 #endif
1488 #ifdef CONFIG_EPOLL
1490 .procname = "epoll",
1491 .mode = 0555,
1492 .child = epoll_table,
1494 #endif
1495 #endif
1497 .procname = "suid_dumpable",
1498 .data = &suid_dumpable,
1499 .maxlen = sizeof(int),
1500 .mode = 0644,
1501 .proc_handler = proc_dointvec_minmax,
1502 .extra1 = &zero,
1503 .extra2 = &two,
1505 #if defined(CONFIG_BINFMT_MISC) || defined(CONFIG_BINFMT_MISC_MODULE)
1507 .procname = "binfmt_misc",
1508 .mode = 0555,
1509 .child = binfmt_misc_table,
1511 #endif
1513 .procname = "pipe-max-size",
1514 .data = &pipe_max_size,
1515 .maxlen = sizeof(int),
1516 .mode = 0644,
1517 .proc_handler = &pipe_proc_fn,
1518 .extra1 = &pipe_min_size,
1523 static struct ctl_table debug_table[] = {
1524 #if defined(CONFIG_X86) || defined(CONFIG_PPC) || defined(CONFIG_SPARC) || \
1525 defined(CONFIG_S390) || defined(CONFIG_TILE)
1527 .procname = "exception-trace",
1528 .data = &show_unhandled_signals,
1529 .maxlen = sizeof(int),
1530 .mode = 0644,
1531 .proc_handler = proc_dointvec
1533 #endif
1534 #if defined(CONFIG_OPTPROBES)
1536 .procname = "kprobes-optimization",
1537 .data = &sysctl_kprobes_optimization,
1538 .maxlen = sizeof(int),
1539 .mode = 0644,
1540 .proc_handler = proc_kprobes_optimization_handler,
1541 .extra1 = &zero,
1542 .extra2 = &one,
1544 #endif
1548 static struct ctl_table dev_table[] = {
1552 static DEFINE_SPINLOCK(sysctl_lock);
1554 /* called under sysctl_lock */
1555 static int use_table(struct ctl_table_header *p)
1557 if (unlikely(p->unregistering))
1558 return 0;
1559 p->used++;
1560 return 1;
1563 /* called under sysctl_lock */
1564 static void unuse_table(struct ctl_table_header *p)
1566 if (!--p->used)
1567 if (unlikely(p->unregistering))
1568 complete(p->unregistering);
1571 /* called under sysctl_lock, will reacquire if has to wait */
1572 static void start_unregistering(struct ctl_table_header *p)
1575 * if p->used is 0, nobody will ever touch that entry again;
1576 * we'll eliminate all paths to it before dropping sysctl_lock
1578 if (unlikely(p->used)) {
1579 struct completion wait;
1580 init_completion(&wait);
1581 p->unregistering = &wait;
1582 spin_unlock(&sysctl_lock);
1583 wait_for_completion(&wait);
1584 spin_lock(&sysctl_lock);
1585 } else {
1586 /* anything non-NULL; we'll never dereference it */
1587 p->unregistering = ERR_PTR(-EINVAL);
1590 * do not remove from the list until nobody holds it; walking the
1591 * list in do_sysctl() relies on that.
1593 list_del_init(&p->ctl_entry);
1596 void sysctl_head_get(struct ctl_table_header *head)
1598 spin_lock(&sysctl_lock);
1599 head->count++;
1600 spin_unlock(&sysctl_lock);
1603 void sysctl_head_put(struct ctl_table_header *head)
1605 spin_lock(&sysctl_lock);
1606 if (!--head->count)
1607 kfree_rcu(head, rcu);
1608 spin_unlock(&sysctl_lock);
1611 struct ctl_table_header *sysctl_head_grab(struct ctl_table_header *head)
1613 if (!head)
1614 BUG();
1615 spin_lock(&sysctl_lock);
1616 if (!use_table(head))
1617 head = ERR_PTR(-ENOENT);
1618 spin_unlock(&sysctl_lock);
1619 return head;
1622 void sysctl_head_finish(struct ctl_table_header *head)
1624 if (!head)
1625 return;
1626 spin_lock(&sysctl_lock);
1627 unuse_table(head);
1628 spin_unlock(&sysctl_lock);
1631 static struct ctl_table_set *
1632 lookup_header_set(struct ctl_table_root *root, struct nsproxy *namespaces)
1634 struct ctl_table_set *set = &root->default_set;
1635 if (root->lookup)
1636 set = root->lookup(root, namespaces);
1637 return set;
1640 static struct list_head *
1641 lookup_header_list(struct ctl_table_root *root, struct nsproxy *namespaces)
1643 struct ctl_table_set *set = lookup_header_set(root, namespaces);
1644 return &set->list;
1647 struct ctl_table_header *__sysctl_head_next(struct nsproxy *namespaces,
1648 struct ctl_table_header *prev)
1650 struct ctl_table_root *root;
1651 struct list_head *header_list;
1652 struct ctl_table_header *head;
1653 struct list_head *tmp;
1655 spin_lock(&sysctl_lock);
1656 if (prev) {
1657 head = prev;
1658 tmp = &prev->ctl_entry;
1659 unuse_table(prev);
1660 goto next;
1662 tmp = &root_table_header.ctl_entry;
1663 for (;;) {
1664 head = list_entry(tmp, struct ctl_table_header, ctl_entry);
1666 if (!use_table(head))
1667 goto next;
1668 spin_unlock(&sysctl_lock);
1669 return head;
1670 next:
1671 root = head->root;
1672 tmp = tmp->next;
1673 header_list = lookup_header_list(root, namespaces);
1674 if (tmp != header_list)
1675 continue;
1677 do {
1678 root = list_entry(root->root_list.next,
1679 struct ctl_table_root, root_list);
1680 if (root == &sysctl_table_root)
1681 goto out;
1682 header_list = lookup_header_list(root, namespaces);
1683 } while (list_empty(header_list));
1684 tmp = header_list->next;
1686 out:
1687 spin_unlock(&sysctl_lock);
1688 return NULL;
1691 struct ctl_table_header *sysctl_head_next(struct ctl_table_header *prev)
1693 return __sysctl_head_next(current->nsproxy, prev);
1696 void register_sysctl_root(struct ctl_table_root *root)
1698 spin_lock(&sysctl_lock);
1699 list_add_tail(&root->root_list, &sysctl_table_root.root_list);
1700 spin_unlock(&sysctl_lock);
1704 * sysctl_perm does NOT grant the superuser all rights automatically, because
1705 * some sysctl variables are readonly even to root.
1708 static int test_perm(int mode, int op)
1710 if (!current_euid())
1711 mode >>= 6;
1712 else if (in_egroup_p(0))
1713 mode >>= 3;
1714 if ((op & ~mode & (MAY_READ|MAY_WRITE|MAY_EXEC)) == 0)
1715 return 0;
1716 return -EACCES;
1719 int sysctl_perm(struct ctl_table_root *root, struct ctl_table *table, int op)
1721 int mode;
1723 if (root->permissions)
1724 mode = root->permissions(root, current->nsproxy, table);
1725 else
1726 mode = table->mode;
1728 return test_perm(mode, op);
1731 static void sysctl_set_parent(struct ctl_table *parent, struct ctl_table *table)
1733 for (; table->procname; table++) {
1734 table->parent = parent;
1735 if (table->child)
1736 sysctl_set_parent(table, table->child);
1740 static __init int sysctl_init(void)
1742 sysctl_set_parent(NULL, root_table);
1743 #ifdef CONFIG_SYSCTL_SYSCALL_CHECK
1744 sysctl_check_table(current->nsproxy, root_table);
1745 #endif
1746 return 0;
1749 core_initcall(sysctl_init);
1751 static struct ctl_table *is_branch_in(struct ctl_table *branch,
1752 struct ctl_table *table)
1754 struct ctl_table *p;
1755 const char *s = branch->procname;
1757 /* branch should have named subdirectory as its first element */
1758 if (!s || !branch->child)
1759 return NULL;
1761 /* ... and nothing else */
1762 if (branch[1].procname)
1763 return NULL;
1765 /* table should contain subdirectory with the same name */
1766 for (p = table; p->procname; p++) {
1767 if (!p->child)
1768 continue;
1769 if (p->procname && strcmp(p->procname, s) == 0)
1770 return p;
1772 return NULL;
1775 /* see if attaching q to p would be an improvement */
1776 static void try_attach(struct ctl_table_header *p, struct ctl_table_header *q)
1778 struct ctl_table *to = p->ctl_table, *by = q->ctl_table;
1779 struct ctl_table *next;
1780 int is_better = 0;
1781 int not_in_parent = !p->attached_by;
1783 while ((next = is_branch_in(by, to)) != NULL) {
1784 if (by == q->attached_by)
1785 is_better = 1;
1786 if (to == p->attached_by)
1787 not_in_parent = 1;
1788 by = by->child;
1789 to = next->child;
1792 if (is_better && not_in_parent) {
1793 q->attached_by = by;
1794 q->attached_to = to;
1795 q->parent = p;
1800 * __register_sysctl_paths - register a sysctl hierarchy
1801 * @root: List of sysctl headers to register on
1802 * @namespaces: Data to compute which lists of sysctl entries are visible
1803 * @path: The path to the directory the sysctl table is in.
1804 * @table: the top-level table structure
1806 * Register a sysctl table hierarchy. @table should be a filled in ctl_table
1807 * array. A completely 0 filled entry terminates the table.
1809 * The members of the &struct ctl_table structure are used as follows:
1811 * procname - the name of the sysctl file under /proc/sys. Set to %NULL to not
1812 * enter a sysctl file
1814 * data - a pointer to data for use by proc_handler
1816 * maxlen - the maximum size in bytes of the data
1818 * mode - the file permissions for the /proc/sys file, and for sysctl(2)
1820 * child - a pointer to the child sysctl table if this entry is a directory, or
1821 * %NULL.
1823 * proc_handler - the text handler routine (described below)
1825 * de - for internal use by the sysctl routines
1827 * extra1, extra2 - extra pointers usable by the proc handler routines
1829 * Leaf nodes in the sysctl tree will be represented by a single file
1830 * under /proc; non-leaf nodes will be represented by directories.
1832 * sysctl(2) can automatically manage read and write requests through
1833 * the sysctl table. The data and maxlen fields of the ctl_table
1834 * struct enable minimal validation of the values being written to be
1835 * performed, and the mode field allows minimal authentication.
1837 * There must be a proc_handler routine for any terminal nodes
1838 * mirrored under /proc/sys (non-terminals are handled by a built-in
1839 * directory handler). Several default handlers are available to
1840 * cover common cases -
1842 * proc_dostring(), proc_dointvec(), proc_dointvec_jiffies(),
1843 * proc_dointvec_userhz_jiffies(), proc_dointvec_minmax(),
1844 * proc_doulongvec_ms_jiffies_minmax(), proc_doulongvec_minmax()
1846 * It is the handler's job to read the input buffer from user memory
1847 * and process it. The handler should return 0 on success.
1849 * This routine returns %NULL on a failure to register, and a pointer
1850 * to the table header on success.
1852 struct ctl_table_header *__register_sysctl_paths(
1853 struct ctl_table_root *root,
1854 struct nsproxy *namespaces,
1855 const struct ctl_path *path, struct ctl_table *table)
1857 struct ctl_table_header *header;
1858 struct ctl_table *new, **prevp;
1859 unsigned int n, npath;
1860 struct ctl_table_set *set;
1862 /* Count the path components */
1863 for (npath = 0; path[npath].procname; ++npath)
1867 * For each path component, allocate a 2-element ctl_table array.
1868 * The first array element will be filled with the sysctl entry
1869 * for this, the second will be the sentinel (procname == 0).
1871 * We allocate everything in one go so that we don't have to
1872 * worry about freeing additional memory in unregister_sysctl_table.
1874 header = kzalloc(sizeof(struct ctl_table_header) +
1875 (2 * npath * sizeof(struct ctl_table)), GFP_KERNEL);
1876 if (!header)
1877 return NULL;
1879 new = (struct ctl_table *) (header + 1);
1881 /* Now connect the dots */
1882 prevp = &header->ctl_table;
1883 for (n = 0; n < npath; ++n, ++path) {
1884 /* Copy the procname */
1885 new->procname = path->procname;
1886 new->mode = 0555;
1888 *prevp = new;
1889 prevp = &new->child;
1891 new += 2;
1893 *prevp = table;
1894 header->ctl_table_arg = table;
1896 INIT_LIST_HEAD(&header->ctl_entry);
1897 header->used = 0;
1898 header->unregistering = NULL;
1899 header->root = root;
1900 sysctl_set_parent(NULL, header->ctl_table);
1901 header->count = 1;
1902 #ifdef CONFIG_SYSCTL_SYSCALL_CHECK
1903 if (sysctl_check_table(namespaces, header->ctl_table)) {
1904 kfree(header);
1905 return NULL;
1907 #endif
1908 spin_lock(&sysctl_lock);
1909 header->set = lookup_header_set(root, namespaces);
1910 header->attached_by = header->ctl_table;
1911 header->attached_to = root_table;
1912 header->parent = &root_table_header;
1913 for (set = header->set; set; set = set->parent) {
1914 struct ctl_table_header *p;
1915 list_for_each_entry(p, &set->list, ctl_entry) {
1916 if (p->unregistering)
1917 continue;
1918 try_attach(p, header);
1921 header->parent->count++;
1922 list_add_tail(&header->ctl_entry, &header->set->list);
1923 spin_unlock(&sysctl_lock);
1925 return header;
1929 * register_sysctl_table_path - register a sysctl table hierarchy
1930 * @path: The path to the directory the sysctl table is in.
1931 * @table: the top-level table structure
1933 * Register a sysctl table hierarchy. @table should be a filled in ctl_table
1934 * array. A completely 0 filled entry terminates the table.
1936 * See __register_sysctl_paths for more details.
1938 struct ctl_table_header *register_sysctl_paths(const struct ctl_path *path,
1939 struct ctl_table *table)
1941 return __register_sysctl_paths(&sysctl_table_root, current->nsproxy,
1942 path, table);
1946 * register_sysctl_table - register a sysctl table hierarchy
1947 * @table: the top-level table structure
1949 * Register a sysctl table hierarchy. @table should be a filled in ctl_table
1950 * array. A completely 0 filled entry terminates the table.
1952 * See register_sysctl_paths for more details.
1954 struct ctl_table_header *register_sysctl_table(struct ctl_table *table)
1956 static const struct ctl_path null_path[] = { {} };
1958 return register_sysctl_paths(null_path, table);
1962 * unregister_sysctl_table - unregister a sysctl table hierarchy
1963 * @header: the header returned from register_sysctl_table
1965 * Unregisters the sysctl table and all children. proc entries may not
1966 * actually be removed until they are no longer used by anyone.
1968 void unregister_sysctl_table(struct ctl_table_header * header)
1970 might_sleep();
1972 if (header == NULL)
1973 return;
1975 spin_lock(&sysctl_lock);
1976 start_unregistering(header);
1977 if (!--header->parent->count) {
1978 WARN_ON(1);
1979 kfree_rcu(header->parent, rcu);
1981 if (!--header->count)
1982 kfree_rcu(header, rcu);
1983 spin_unlock(&sysctl_lock);
1986 int sysctl_is_seen(struct ctl_table_header *p)
1988 struct ctl_table_set *set = p->set;
1989 int res;
1990 spin_lock(&sysctl_lock);
1991 if (p->unregistering)
1992 res = 0;
1993 else if (!set->is_seen)
1994 res = 1;
1995 else
1996 res = set->is_seen(set);
1997 spin_unlock(&sysctl_lock);
1998 return res;
2001 void setup_sysctl_set(struct ctl_table_set *p,
2002 struct ctl_table_set *parent,
2003 int (*is_seen)(struct ctl_table_set *))
2005 INIT_LIST_HEAD(&p->list);
2006 p->parent = parent ? parent : &sysctl_table_root.default_set;
2007 p->is_seen = is_seen;
2010 #else /* !CONFIG_SYSCTL */
2011 struct ctl_table_header *register_sysctl_table(struct ctl_table * table)
2013 return NULL;
2016 struct ctl_table_header *register_sysctl_paths(const struct ctl_path *path,
2017 struct ctl_table *table)
2019 return NULL;
2022 void unregister_sysctl_table(struct ctl_table_header * table)
2026 void setup_sysctl_set(struct ctl_table_set *p,
2027 struct ctl_table_set *parent,
2028 int (*is_seen)(struct ctl_table_set *))
2032 void sysctl_head_put(struct ctl_table_header *head)
2036 #endif /* CONFIG_SYSCTL */
2039 * /proc/sys support
2042 #ifdef CONFIG_PROC_SYSCTL
2044 static int _proc_do_string(void* data, int maxlen, int write,
2045 void __user *buffer,
2046 size_t *lenp, loff_t *ppos)
2048 size_t len;
2049 char __user *p;
2050 char c;
2052 if (!data || !maxlen || !*lenp) {
2053 *lenp = 0;
2054 return 0;
2057 if (write) {
2058 len = 0;
2059 p = buffer;
2060 while (len < *lenp) {
2061 if (get_user(c, p++))
2062 return -EFAULT;
2063 if (c == 0 || c == '\n')
2064 break;
2065 len++;
2067 if (len >= maxlen)
2068 len = maxlen-1;
2069 if(copy_from_user(data, buffer, len))
2070 return -EFAULT;
2071 ((char *) data)[len] = 0;
2072 *ppos += *lenp;
2073 } else {
2074 len = strlen(data);
2075 if (len > maxlen)
2076 len = maxlen;
2078 if (*ppos > len) {
2079 *lenp = 0;
2080 return 0;
2083 data += *ppos;
2084 len -= *ppos;
2086 if (len > *lenp)
2087 len = *lenp;
2088 if (len)
2089 if(copy_to_user(buffer, data, len))
2090 return -EFAULT;
2091 if (len < *lenp) {
2092 if(put_user('\n', ((char __user *) buffer) + len))
2093 return -EFAULT;
2094 len++;
2096 *lenp = len;
2097 *ppos += len;
2099 return 0;
2103 * proc_dostring - read a string sysctl
2104 * @table: the sysctl table
2105 * @write: %TRUE if this is a write to the sysctl file
2106 * @buffer: the user buffer
2107 * @lenp: the size of the user buffer
2108 * @ppos: file position
2110 * Reads/writes a string from/to the user buffer. If the kernel
2111 * buffer provided is not large enough to hold the string, the
2112 * string is truncated. The copied string is %NULL-terminated.
2113 * If the string is being read by the user process, it is copied
2114 * and a newline '\n' is added. It is truncated if the buffer is
2115 * not large enough.
2117 * Returns 0 on success.
2119 int proc_dostring(struct ctl_table *table, int write,
2120 void __user *buffer, size_t *lenp, loff_t *ppos)
2122 return _proc_do_string(table->data, table->maxlen, write,
2123 buffer, lenp, ppos);
2126 static size_t proc_skip_spaces(char **buf)
2128 size_t ret;
2129 char *tmp = skip_spaces(*buf);
2130 ret = tmp - *buf;
2131 *buf = tmp;
2132 return ret;
2135 static void proc_skip_char(char **buf, size_t *size, const char v)
2137 while (*size) {
2138 if (**buf != v)
2139 break;
2140 (*size)--;
2141 (*buf)++;
2145 #define TMPBUFLEN 22
2147 * proc_get_long - reads an ASCII formatted integer from a user buffer
2149 * @buf: a kernel buffer
2150 * @size: size of the kernel buffer
2151 * @val: this is where the number will be stored
2152 * @neg: set to %TRUE if number is negative
2153 * @perm_tr: a vector which contains the allowed trailers
2154 * @perm_tr_len: size of the perm_tr vector
2155 * @tr: pointer to store the trailer character
2157 * In case of success %0 is returned and @buf and @size are updated with
2158 * the amount of bytes read. If @tr is non-NULL and a trailing
2159 * character exists (size is non-zero after returning from this
2160 * function), @tr is updated with the trailing character.
2162 static int proc_get_long(char **buf, size_t *size,
2163 unsigned long *val, bool *neg,
2164 const char *perm_tr, unsigned perm_tr_len, char *tr)
2166 int len;
2167 char *p, tmp[TMPBUFLEN];
2169 if (!*size)
2170 return -EINVAL;
2172 len = *size;
2173 if (len > TMPBUFLEN - 1)
2174 len = TMPBUFLEN - 1;
2176 memcpy(tmp, *buf, len);
2178 tmp[len] = 0;
2179 p = tmp;
2180 if (*p == '-' && *size > 1) {
2181 *neg = true;
2182 p++;
2183 } else
2184 *neg = false;
2185 if (!isdigit(*p))
2186 return -EINVAL;
2188 *val = simple_strtoul(p, &p, 0);
2190 len = p - tmp;
2192 /* We don't know if the next char is whitespace thus we may accept
2193 * invalid integers (e.g. 1234...a) or two integers instead of one
2194 * (e.g. 123...1). So lets not allow such large numbers. */
2195 if (len == TMPBUFLEN - 1)
2196 return -EINVAL;
2198 if (len < *size && perm_tr_len && !memchr(perm_tr, *p, perm_tr_len))
2199 return -EINVAL;
2201 if (tr && (len < *size))
2202 *tr = *p;
2204 *buf += len;
2205 *size -= len;
2207 return 0;
2211 * proc_put_long - converts an integer to a decimal ASCII formatted string
2213 * @buf: the user buffer
2214 * @size: the size of the user buffer
2215 * @val: the integer to be converted
2216 * @neg: sign of the number, %TRUE for negative
2218 * In case of success %0 is returned and @buf and @size are updated with
2219 * the amount of bytes written.
2221 static int proc_put_long(void __user **buf, size_t *size, unsigned long val,
2222 bool neg)
2224 int len;
2225 char tmp[TMPBUFLEN], *p = tmp;
2227 sprintf(p, "%s%lu", neg ? "-" : "", val);
2228 len = strlen(tmp);
2229 if (len > *size)
2230 len = *size;
2231 if (copy_to_user(*buf, tmp, len))
2232 return -EFAULT;
2233 *size -= len;
2234 *buf += len;
2235 return 0;
2237 #undef TMPBUFLEN
2239 static int proc_put_char(void __user **buf, size_t *size, char c)
2241 if (*size) {
2242 char __user **buffer = (char __user **)buf;
2243 if (put_user(c, *buffer))
2244 return -EFAULT;
2245 (*size)--, (*buffer)++;
2246 *buf = *buffer;
2248 return 0;
2251 static int do_proc_dointvec_conv(bool *negp, unsigned long *lvalp,
2252 int *valp,
2253 int write, void *data)
2255 if (write) {
2256 *valp = *negp ? -*lvalp : *lvalp;
2257 } else {
2258 int val = *valp;
2259 if (val < 0) {
2260 *negp = true;
2261 *lvalp = (unsigned long)-val;
2262 } else {
2263 *negp = false;
2264 *lvalp = (unsigned long)val;
2267 return 0;
2270 static const char proc_wspace_sep[] = { ' ', '\t', '\n' };
2272 static int __do_proc_dointvec(void *tbl_data, struct ctl_table *table,
2273 int write, void __user *buffer,
2274 size_t *lenp, loff_t *ppos,
2275 int (*conv)(bool *negp, unsigned long *lvalp, int *valp,
2276 int write, void *data),
2277 void *data)
2279 int *i, vleft, first = 1, err = 0;
2280 unsigned long page = 0;
2281 size_t left;
2282 char *kbuf;
2284 if (!tbl_data || !table->maxlen || !*lenp || (*ppos && !write)) {
2285 *lenp = 0;
2286 return 0;
2289 i = (int *) tbl_data;
2290 vleft = table->maxlen / sizeof(*i);
2291 left = *lenp;
2293 if (!conv)
2294 conv = do_proc_dointvec_conv;
2296 if (write) {
2297 if (left > PAGE_SIZE - 1)
2298 left = PAGE_SIZE - 1;
2299 page = __get_free_page(GFP_TEMPORARY);
2300 kbuf = (char *) page;
2301 if (!kbuf)
2302 return -ENOMEM;
2303 if (copy_from_user(kbuf, buffer, left)) {
2304 err = -EFAULT;
2305 goto free;
2307 kbuf[left] = 0;
2310 for (; left && vleft--; i++, first=0) {
2311 unsigned long lval;
2312 bool neg;
2314 if (write) {
2315 left -= proc_skip_spaces(&kbuf);
2317 if (!left)
2318 break;
2319 err = proc_get_long(&kbuf, &left, &lval, &neg,
2320 proc_wspace_sep,
2321 sizeof(proc_wspace_sep), NULL);
2322 if (err)
2323 break;
2324 if (conv(&neg, &lval, i, 1, data)) {
2325 err = -EINVAL;
2326 break;
2328 } else {
2329 if (conv(&neg, &lval, i, 0, data)) {
2330 err = -EINVAL;
2331 break;
2333 if (!first)
2334 err = proc_put_char(&buffer, &left, '\t');
2335 if (err)
2336 break;
2337 err = proc_put_long(&buffer, &left, lval, neg);
2338 if (err)
2339 break;
2343 if (!write && !first && left && !err)
2344 err = proc_put_char(&buffer, &left, '\n');
2345 if (write && !err && left)
2346 left -= proc_skip_spaces(&kbuf);
2347 free:
2348 if (write) {
2349 free_page(page);
2350 if (first)
2351 return err ? : -EINVAL;
2353 *lenp -= left;
2354 *ppos += *lenp;
2355 return err;
2358 static int do_proc_dointvec(struct ctl_table *table, int write,
2359 void __user *buffer, size_t *lenp, loff_t *ppos,
2360 int (*conv)(bool *negp, unsigned long *lvalp, int *valp,
2361 int write, void *data),
2362 void *data)
2364 return __do_proc_dointvec(table->data, table, write,
2365 buffer, lenp, ppos, conv, data);
2369 * proc_dointvec - read a vector of integers
2370 * @table: the sysctl table
2371 * @write: %TRUE if this is a write to the sysctl file
2372 * @buffer: the user buffer
2373 * @lenp: the size of the user buffer
2374 * @ppos: file position
2376 * Reads/writes up to table->maxlen/sizeof(unsigned int) integer
2377 * values from/to the user buffer, treated as an ASCII string.
2379 * Returns 0 on success.
2381 int proc_dointvec(struct ctl_table *table, int write,
2382 void __user *buffer, size_t *lenp, loff_t *ppos)
2384 return do_proc_dointvec(table,write,buffer,lenp,ppos,
2385 NULL,NULL);
2389 * Taint values can only be increased
2390 * This means we can safely use a temporary.
2392 static int proc_taint(struct ctl_table *table, int write,
2393 void __user *buffer, size_t *lenp, loff_t *ppos)
2395 struct ctl_table t;
2396 unsigned long tmptaint = get_taint();
2397 int err;
2399 if (write && !capable(CAP_SYS_ADMIN))
2400 return -EPERM;
2402 t = *table;
2403 t.data = &tmptaint;
2404 err = proc_doulongvec_minmax(&t, write, buffer, lenp, ppos);
2405 if (err < 0)
2406 return err;
2408 if (write) {
2410 * Poor man's atomic or. Not worth adding a primitive
2411 * to everyone's atomic.h for this
2413 int i;
2414 for (i = 0; i < BITS_PER_LONG && tmptaint >> i; i++) {
2415 if ((tmptaint >> i) & 1)
2416 add_taint(i);
2420 return err;
2423 #ifdef CONFIG_PRINTK
2424 static int proc_dmesg_restrict(struct ctl_table *table, int write,
2425 void __user *buffer, size_t *lenp, loff_t *ppos)
2427 if (write && !capable(CAP_SYS_ADMIN))
2428 return -EPERM;
2430 return proc_dointvec_minmax(table, write, buffer, lenp, ppos);
2432 #endif
2434 struct do_proc_dointvec_minmax_conv_param {
2435 int *min;
2436 int *max;
2439 static int do_proc_dointvec_minmax_conv(bool *negp, unsigned long *lvalp,
2440 int *valp,
2441 int write, void *data)
2443 struct do_proc_dointvec_minmax_conv_param *param = data;
2444 if (write) {
2445 int val = *negp ? -*lvalp : *lvalp;
2446 if ((param->min && *param->min > val) ||
2447 (param->max && *param->max < val))
2448 return -EINVAL;
2449 *valp = val;
2450 } else {
2451 int val = *valp;
2452 if (val < 0) {
2453 *negp = true;
2454 *lvalp = (unsigned long)-val;
2455 } else {
2456 *negp = false;
2457 *lvalp = (unsigned long)val;
2460 return 0;
2464 * proc_dointvec_minmax - read a vector of integers with min/max values
2465 * @table: the sysctl table
2466 * @write: %TRUE if this is a write to the sysctl file
2467 * @buffer: the user buffer
2468 * @lenp: the size of the user buffer
2469 * @ppos: file position
2471 * Reads/writes up to table->maxlen/sizeof(unsigned int) integer
2472 * values from/to the user buffer, treated as an ASCII string.
2474 * This routine will ensure the values are within the range specified by
2475 * table->extra1 (min) and table->extra2 (max).
2477 * Returns 0 on success.
2479 int proc_dointvec_minmax(struct ctl_table *table, int write,
2480 void __user *buffer, size_t *lenp, loff_t *ppos)
2482 struct do_proc_dointvec_minmax_conv_param param = {
2483 .min = (int *) table->extra1,
2484 .max = (int *) table->extra2,
2486 return do_proc_dointvec(table, write, buffer, lenp, ppos,
2487 do_proc_dointvec_minmax_conv, &param);
2490 static int __do_proc_doulongvec_minmax(void *data, struct ctl_table *table, int write,
2491 void __user *buffer,
2492 size_t *lenp, loff_t *ppos,
2493 unsigned long convmul,
2494 unsigned long convdiv)
2496 unsigned long *i, *min, *max;
2497 int vleft, first = 1, err = 0;
2498 unsigned long page = 0;
2499 size_t left;
2500 char *kbuf;
2502 if (!data || !table->maxlen || !*lenp || (*ppos && !write)) {
2503 *lenp = 0;
2504 return 0;
2507 i = (unsigned long *) data;
2508 min = (unsigned long *) table->extra1;
2509 max = (unsigned long *) table->extra2;
2510 vleft = table->maxlen / sizeof(unsigned long);
2511 left = *lenp;
2513 if (write) {
2514 if (left > PAGE_SIZE - 1)
2515 left = PAGE_SIZE - 1;
2516 page = __get_free_page(GFP_TEMPORARY);
2517 kbuf = (char *) page;
2518 if (!kbuf)
2519 return -ENOMEM;
2520 if (copy_from_user(kbuf, buffer, left)) {
2521 err = -EFAULT;
2522 goto free;
2524 kbuf[left] = 0;
2527 for (; left && vleft--; i++, first = 0) {
2528 unsigned long val;
2530 if (write) {
2531 bool neg;
2533 left -= proc_skip_spaces(&kbuf);
2535 err = proc_get_long(&kbuf, &left, &val, &neg,
2536 proc_wspace_sep,
2537 sizeof(proc_wspace_sep), NULL);
2538 if (err)
2539 break;
2540 if (neg)
2541 continue;
2542 if ((min && val < *min) || (max && val > *max))
2543 continue;
2544 *i = val;
2545 } else {
2546 val = convdiv * (*i) / convmul;
2547 if (!first)
2548 err = proc_put_char(&buffer, &left, '\t');
2549 err = proc_put_long(&buffer, &left, val, false);
2550 if (err)
2551 break;
2555 if (!write && !first && left && !err)
2556 err = proc_put_char(&buffer, &left, '\n');
2557 if (write && !err)
2558 left -= proc_skip_spaces(&kbuf);
2559 free:
2560 if (write) {
2561 free_page(page);
2562 if (first)
2563 return err ? : -EINVAL;
2565 *lenp -= left;
2566 *ppos += *lenp;
2567 return err;
2570 static int do_proc_doulongvec_minmax(struct ctl_table *table, int write,
2571 void __user *buffer,
2572 size_t *lenp, loff_t *ppos,
2573 unsigned long convmul,
2574 unsigned long convdiv)
2576 return __do_proc_doulongvec_minmax(table->data, table, write,
2577 buffer, lenp, ppos, convmul, convdiv);
2581 * proc_doulongvec_minmax - read a vector of long integers with min/max values
2582 * @table: the sysctl table
2583 * @write: %TRUE if this is a write to the sysctl file
2584 * @buffer: the user buffer
2585 * @lenp: the size of the user buffer
2586 * @ppos: file position
2588 * Reads/writes up to table->maxlen/sizeof(unsigned long) unsigned long
2589 * values from/to the user buffer, treated as an ASCII string.
2591 * This routine will ensure the values are within the range specified by
2592 * table->extra1 (min) and table->extra2 (max).
2594 * Returns 0 on success.
2596 int proc_doulongvec_minmax(struct ctl_table *table, int write,
2597 void __user *buffer, size_t *lenp, loff_t *ppos)
2599 return do_proc_doulongvec_minmax(table, write, buffer, lenp, ppos, 1l, 1l);
2603 * proc_doulongvec_ms_jiffies_minmax - read a vector of millisecond values with min/max values
2604 * @table: the sysctl table
2605 * @write: %TRUE if this is a write to the sysctl file
2606 * @buffer: the user buffer
2607 * @lenp: the size of the user buffer
2608 * @ppos: file position
2610 * Reads/writes up to table->maxlen/sizeof(unsigned long) unsigned long
2611 * values from/to the user buffer, treated as an ASCII string. The values
2612 * are treated as milliseconds, and converted to jiffies when they are stored.
2614 * This routine will ensure the values are within the range specified by
2615 * table->extra1 (min) and table->extra2 (max).
2617 * Returns 0 on success.
2619 int proc_doulongvec_ms_jiffies_minmax(struct ctl_table *table, int write,
2620 void __user *buffer,
2621 size_t *lenp, loff_t *ppos)
2623 return do_proc_doulongvec_minmax(table, write, buffer,
2624 lenp, ppos, HZ, 1000l);
2628 static int do_proc_dointvec_jiffies_conv(bool *negp, unsigned long *lvalp,
2629 int *valp,
2630 int write, void *data)
2632 if (write) {
2633 if (*lvalp > LONG_MAX / HZ)
2634 return 1;
2635 *valp = *negp ? -(*lvalp*HZ) : (*lvalp*HZ);
2636 } else {
2637 int val = *valp;
2638 unsigned long lval;
2639 if (val < 0) {
2640 *negp = true;
2641 lval = (unsigned long)-val;
2642 } else {
2643 *negp = false;
2644 lval = (unsigned long)val;
2646 *lvalp = lval / HZ;
2648 return 0;
2651 static int do_proc_dointvec_userhz_jiffies_conv(bool *negp, unsigned long *lvalp,
2652 int *valp,
2653 int write, void *data)
2655 if (write) {
2656 if (USER_HZ < HZ && *lvalp > (LONG_MAX / HZ) * USER_HZ)
2657 return 1;
2658 *valp = clock_t_to_jiffies(*negp ? -*lvalp : *lvalp);
2659 } else {
2660 int val = *valp;
2661 unsigned long lval;
2662 if (val < 0) {
2663 *negp = true;
2664 lval = (unsigned long)-val;
2665 } else {
2666 *negp = false;
2667 lval = (unsigned long)val;
2669 *lvalp = jiffies_to_clock_t(lval);
2671 return 0;
2674 static int do_proc_dointvec_ms_jiffies_conv(bool *negp, unsigned long *lvalp,
2675 int *valp,
2676 int write, void *data)
2678 if (write) {
2679 *valp = msecs_to_jiffies(*negp ? -*lvalp : *lvalp);
2680 } else {
2681 int val = *valp;
2682 unsigned long lval;
2683 if (val < 0) {
2684 *negp = true;
2685 lval = (unsigned long)-val;
2686 } else {
2687 *negp = false;
2688 lval = (unsigned long)val;
2690 *lvalp = jiffies_to_msecs(lval);
2692 return 0;
2696 * proc_dointvec_jiffies - read a vector of integers as seconds
2697 * @table: the sysctl table
2698 * @write: %TRUE if this is a write to the sysctl file
2699 * @buffer: the user buffer
2700 * @lenp: the size of the user buffer
2701 * @ppos: file position
2703 * Reads/writes up to table->maxlen/sizeof(unsigned int) integer
2704 * values from/to the user buffer, treated as an ASCII string.
2705 * The values read are assumed to be in seconds, and are converted into
2706 * jiffies.
2708 * Returns 0 on success.
2710 int proc_dointvec_jiffies(struct ctl_table *table, int write,
2711 void __user *buffer, size_t *lenp, loff_t *ppos)
2713 return do_proc_dointvec(table,write,buffer,lenp,ppos,
2714 do_proc_dointvec_jiffies_conv,NULL);
2718 * proc_dointvec_userhz_jiffies - read a vector of integers as 1/USER_HZ seconds
2719 * @table: the sysctl table
2720 * @write: %TRUE if this is a write to the sysctl file
2721 * @buffer: the user buffer
2722 * @lenp: the size of the user buffer
2723 * @ppos: pointer to the file position
2725 * Reads/writes up to table->maxlen/sizeof(unsigned int) integer
2726 * values from/to the user buffer, treated as an ASCII string.
2727 * The values read are assumed to be in 1/USER_HZ seconds, and
2728 * are converted into jiffies.
2730 * Returns 0 on success.
2732 int proc_dointvec_userhz_jiffies(struct ctl_table *table, int write,
2733 void __user *buffer, size_t *lenp, loff_t *ppos)
2735 return do_proc_dointvec(table,write,buffer,lenp,ppos,
2736 do_proc_dointvec_userhz_jiffies_conv,NULL);
2740 * proc_dointvec_ms_jiffies - read a vector of integers as 1 milliseconds
2741 * @table: the sysctl table
2742 * @write: %TRUE if this is a write to the sysctl file
2743 * @buffer: the user buffer
2744 * @lenp: the size of the user buffer
2745 * @ppos: file position
2746 * @ppos: the current position in the file
2748 * Reads/writes up to table->maxlen/sizeof(unsigned int) integer
2749 * values from/to the user buffer, treated as an ASCII string.
2750 * The values read are assumed to be in 1/1000 seconds, and
2751 * are converted into jiffies.
2753 * Returns 0 on success.
2755 int proc_dointvec_ms_jiffies(struct ctl_table *table, int write,
2756 void __user *buffer, size_t *lenp, loff_t *ppos)
2758 return do_proc_dointvec(table, write, buffer, lenp, ppos,
2759 do_proc_dointvec_ms_jiffies_conv, NULL);
2762 static int proc_do_cad_pid(struct ctl_table *table, int write,
2763 void __user *buffer, size_t *lenp, loff_t *ppos)
2765 struct pid *new_pid;
2766 pid_t tmp;
2767 int r;
2769 tmp = pid_vnr(cad_pid);
2771 r = __do_proc_dointvec(&tmp, table, write, buffer,
2772 lenp, ppos, NULL, NULL);
2773 if (r || !write)
2774 return r;
2776 new_pid = find_get_pid(tmp);
2777 if (!new_pid)
2778 return -ESRCH;
2780 put_pid(xchg(&cad_pid, new_pid));
2781 return 0;
2785 * proc_do_large_bitmap - read/write from/to a large bitmap
2786 * @table: the sysctl table
2787 * @write: %TRUE if this is a write to the sysctl file
2788 * @buffer: the user buffer
2789 * @lenp: the size of the user buffer
2790 * @ppos: file position
2792 * The bitmap is stored at table->data and the bitmap length (in bits)
2793 * in table->maxlen.
2795 * We use a range comma separated format (e.g. 1,3-4,10-10) so that
2796 * large bitmaps may be represented in a compact manner. Writing into
2797 * the file will clear the bitmap then update it with the given input.
2799 * Returns 0 on success.
2801 int proc_do_large_bitmap(struct ctl_table *table, int write,
2802 void __user *buffer, size_t *lenp, loff_t *ppos)
2804 int err = 0;
2805 bool first = 1;
2806 size_t left = *lenp;
2807 unsigned long bitmap_len = table->maxlen;
2808 unsigned long *bitmap = (unsigned long *) table->data;
2809 unsigned long *tmp_bitmap = NULL;
2810 char tr_a[] = { '-', ',', '\n' }, tr_b[] = { ',', '\n', 0 }, c;
2812 if (!bitmap_len || !left || (*ppos && !write)) {
2813 *lenp = 0;
2814 return 0;
2817 if (write) {
2818 unsigned long page = 0;
2819 char *kbuf;
2821 if (left > PAGE_SIZE - 1)
2822 left = PAGE_SIZE - 1;
2824 page = __get_free_page(GFP_TEMPORARY);
2825 kbuf = (char *) page;
2826 if (!kbuf)
2827 return -ENOMEM;
2828 if (copy_from_user(kbuf, buffer, left)) {
2829 free_page(page);
2830 return -EFAULT;
2832 kbuf[left] = 0;
2834 tmp_bitmap = kzalloc(BITS_TO_LONGS(bitmap_len) * sizeof(unsigned long),
2835 GFP_KERNEL);
2836 if (!tmp_bitmap) {
2837 free_page(page);
2838 return -ENOMEM;
2840 proc_skip_char(&kbuf, &left, '\n');
2841 while (!err && left) {
2842 unsigned long val_a, val_b;
2843 bool neg;
2845 err = proc_get_long(&kbuf, &left, &val_a, &neg, tr_a,
2846 sizeof(tr_a), &c);
2847 if (err)
2848 break;
2849 if (val_a >= bitmap_len || neg) {
2850 err = -EINVAL;
2851 break;
2854 val_b = val_a;
2855 if (left) {
2856 kbuf++;
2857 left--;
2860 if (c == '-') {
2861 err = proc_get_long(&kbuf, &left, &val_b,
2862 &neg, tr_b, sizeof(tr_b),
2863 &c);
2864 if (err)
2865 break;
2866 if (val_b >= bitmap_len || neg ||
2867 val_a > val_b) {
2868 err = -EINVAL;
2869 break;
2871 if (left) {
2872 kbuf++;
2873 left--;
2877 while (val_a <= val_b)
2878 set_bit(val_a++, tmp_bitmap);
2880 first = 0;
2881 proc_skip_char(&kbuf, &left, '\n');
2883 free_page(page);
2884 } else {
2885 unsigned long bit_a, bit_b = 0;
2887 while (left) {
2888 bit_a = find_next_bit(bitmap, bitmap_len, bit_b);
2889 if (bit_a >= bitmap_len)
2890 break;
2891 bit_b = find_next_zero_bit(bitmap, bitmap_len,
2892 bit_a + 1) - 1;
2894 if (!first) {
2895 err = proc_put_char(&buffer, &left, ',');
2896 if (err)
2897 break;
2899 err = proc_put_long(&buffer, &left, bit_a, false);
2900 if (err)
2901 break;
2902 if (bit_a != bit_b) {
2903 err = proc_put_char(&buffer, &left, '-');
2904 if (err)
2905 break;
2906 err = proc_put_long(&buffer, &left, bit_b, false);
2907 if (err)
2908 break;
2911 first = 0; bit_b++;
2913 if (!err)
2914 err = proc_put_char(&buffer, &left, '\n');
2917 if (!err) {
2918 if (write) {
2919 if (*ppos)
2920 bitmap_or(bitmap, bitmap, tmp_bitmap, bitmap_len);
2921 else
2922 memcpy(bitmap, tmp_bitmap,
2923 BITS_TO_LONGS(bitmap_len) * sizeof(unsigned long));
2925 kfree(tmp_bitmap);
2926 *lenp -= left;
2927 *ppos += *lenp;
2928 return 0;
2929 } else {
2930 kfree(tmp_bitmap);
2931 return err;
2935 #else /* CONFIG_PROC_SYSCTL */
2937 int proc_dostring(struct ctl_table *table, int write,
2938 void __user *buffer, size_t *lenp, loff_t *ppos)
2940 return -ENOSYS;
2943 int proc_dointvec(struct ctl_table *table, int write,
2944 void __user *buffer, size_t *lenp, loff_t *ppos)
2946 return -ENOSYS;
2949 int proc_dointvec_minmax(struct ctl_table *table, int write,
2950 void __user *buffer, size_t *lenp, loff_t *ppos)
2952 return -ENOSYS;
2955 int proc_dointvec_jiffies(struct ctl_table *table, int write,
2956 void __user *buffer, size_t *lenp, loff_t *ppos)
2958 return -ENOSYS;
2961 int proc_dointvec_userhz_jiffies(struct ctl_table *table, int write,
2962 void __user *buffer, size_t *lenp, loff_t *ppos)
2964 return -ENOSYS;
2967 int proc_dointvec_ms_jiffies(struct ctl_table *table, int write,
2968 void __user *buffer, size_t *lenp, loff_t *ppos)
2970 return -ENOSYS;
2973 int proc_doulongvec_minmax(struct ctl_table *table, int write,
2974 void __user *buffer, size_t *lenp, loff_t *ppos)
2976 return -ENOSYS;
2979 int proc_doulongvec_ms_jiffies_minmax(struct ctl_table *table, int write,
2980 void __user *buffer,
2981 size_t *lenp, loff_t *ppos)
2983 return -ENOSYS;
2987 #endif /* CONFIG_PROC_SYSCTL */
2990 * No sense putting this after each symbol definition, twice,
2991 * exception granted :-)
2993 EXPORT_SYMBOL(proc_dointvec);
2994 EXPORT_SYMBOL(proc_dointvec_jiffies);
2995 EXPORT_SYMBOL(proc_dointvec_minmax);
2996 EXPORT_SYMBOL(proc_dointvec_userhz_jiffies);
2997 EXPORT_SYMBOL(proc_dointvec_ms_jiffies);
2998 EXPORT_SYMBOL(proc_dostring);
2999 EXPORT_SYMBOL(proc_doulongvec_minmax);
3000 EXPORT_SYMBOL(proc_doulongvec_ms_jiffies_minmax);
3001 EXPORT_SYMBOL(register_sysctl_table);
3002 EXPORT_SYMBOL(register_sysctl_paths);
3003 EXPORT_SYMBOL(unregister_sysctl_table);