sched: Make tunable scaling style configurable
[linux-2.6.git] / kernel / sysctl.c
1 /*
2  * sysctl.c: General linux system control interface
3  *
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
19  */
20
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/proc_fs.h>
27 #include <linux/security.h>
28 #include <linux/ctype.h>
29 #include <linux/kmemcheck.h>
30 #include <linux/smp_lock.h>
31 #include <linux/fs.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
55 #include <asm/uaccess.h>
56 #include <asm/processor.h>
57
58 #ifdef CONFIG_X86
59 #include <asm/nmi.h>
60 #include <asm/stacktrace.h>
61 #include <asm/io.h>
62 #endif
63
64 static int deprecated_sysctl_warning(struct __sysctl_args *args);
65
66 #if defined(CONFIG_SYSCTL)
67
68 /* External variables not in a header file. */
69 extern int C_A_D;
70 extern int print_fatal_signals;
71 extern int sysctl_overcommit_memory;
72 extern int sysctl_overcommit_ratio;
73 extern int sysctl_panic_on_oom;
74 extern int sysctl_oom_kill_allocating_task;
75 extern int sysctl_oom_dump_tasks;
76 extern int max_threads;
77 extern int core_uses_pid;
78 extern int suid_dumpable;
79 extern char core_pattern[];
80 extern unsigned int core_pipe_limit;
81 extern int pid_max;
82 extern int min_free_kbytes;
83 extern int pid_max_min, pid_max_max;
84 extern int sysctl_drop_caches;
85 extern int percpu_pagelist_fraction;
86 extern int compat_log;
87 extern int latencytop_enabled;
88 extern int sysctl_nr_open_min, sysctl_nr_open_max;
89 #ifndef CONFIG_MMU
90 extern int sysctl_nr_trim_pages;
91 #endif
92 #ifdef CONFIG_RCU_TORTURE_TEST
93 extern int rcutorture_runnable;
94 #endif /* #ifdef CONFIG_RCU_TORTURE_TEST */
95 #ifdef CONFIG_BLOCK
96 extern int blk_iopoll_enabled;
97 #endif
98
99 /* Constants used for minimum and  maximum */
100 #ifdef CONFIG_DETECT_SOFTLOCKUP
101 static int sixty = 60;
102 static int neg_one = -1;
103 #endif
104
105 static int zero;
106 static int __maybe_unused one = 1;
107 static int __maybe_unused two = 2;
108 static unsigned long one_ul = 1;
109 static int one_hundred = 100;
110 #ifdef CONFIG_PRINTK
111 static int ten_thousand = 10000;
112 #endif
113
114 /* this is needed for the proc_doulongvec_minmax of vm_dirty_bytes */
115 static unsigned long dirty_bytes_min = 2 * PAGE_SIZE;
116
117 /* this is needed for the proc_dointvec_minmax for [fs_]overflow UID and GID */
118 static int maxolduid = 65535;
119 static int minolduid;
120 static int min_percpu_pagelist_fract = 8;
121
122 static int ngroups_max = NGROUPS_MAX;
123
124 #ifdef CONFIG_MODULES
125 extern char modprobe_path[];
126 extern int modules_disabled;
127 #endif
128 #ifdef CONFIG_CHR_DEV_SG
129 extern int sg_big_buff;
130 #endif
131
132 #ifdef CONFIG_SPARC
133 #include <asm/system.h>
134 #endif
135
136 #ifdef CONFIG_SPARC64
137 extern int sysctl_tsb_ratio;
138 #endif
139
140 #ifdef __hppa__
141 extern int pwrsw_enabled;
142 extern int unaligned_enabled;
143 #endif
144
145 #ifdef CONFIG_S390
146 #ifdef CONFIG_MATHEMU
147 extern int sysctl_ieee_emulation_warnings;
148 #endif
149 extern int sysctl_userprocess_debug;
150 extern int spin_retry;
151 #endif
152
153 #ifdef CONFIG_BSD_PROCESS_ACCT
154 extern int acct_parm[];
155 #endif
156
157 #ifdef CONFIG_IA64
158 extern int no_unaligned_warning;
159 extern int unaligned_dump_stack;
160 #endif
161
162 extern struct ratelimit_state printk_ratelimit_state;
163
164 #ifdef CONFIG_RT_MUTEXES
165 extern int max_lock_depth;
166 #endif
167
168 #ifdef CONFIG_PROC_SYSCTL
169 static int proc_do_cad_pid(struct ctl_table *table, int write,
170                   void __user *buffer, size_t *lenp, loff_t *ppos);
171 static int proc_taint(struct ctl_table *table, int write,
172                                void __user *buffer, size_t *lenp, loff_t *ppos);
173 #endif
174
175 static struct ctl_table root_table[];
176 static struct ctl_table_root sysctl_table_root;
177 static struct ctl_table_header root_table_header = {
178         .count = 1,
179         .ctl_table = root_table,
180         .ctl_entry = LIST_HEAD_INIT(sysctl_table_root.default_set.list),
181         .root = &sysctl_table_root,
182         .set = &sysctl_table_root.default_set,
183 };
184 static struct ctl_table_root sysctl_table_root = {
185         .root_list = LIST_HEAD_INIT(sysctl_table_root.root_list),
186         .default_set.list = LIST_HEAD_INIT(root_table_header.ctl_entry),
187 };
188
189 static struct ctl_table kern_table[];
190 static struct ctl_table vm_table[];
191 static struct ctl_table fs_table[];
192 static struct ctl_table debug_table[];
193 static struct ctl_table dev_table[];
194 extern struct ctl_table random_table[];
195 #ifdef CONFIG_INOTIFY_USER
196 extern struct ctl_table inotify_table[];
197 #endif
198 #ifdef CONFIG_EPOLL
199 extern struct ctl_table epoll_table[];
200 #endif
201
202 #ifdef HAVE_ARCH_PICK_MMAP_LAYOUT
203 int sysctl_legacy_va_layout;
204 #endif
205
206 extern int prove_locking;
207 extern int lock_stat;
208
209 /* The default sysctl tables: */
210
211 static struct ctl_table root_table[] = {
212         {
213                 .ctl_name       = CTL_KERN,
214                 .procname       = "kernel",
215                 .mode           = 0555,
216                 .child          = kern_table,
217         },
218         {
219                 .ctl_name       = CTL_VM,
220                 .procname       = "vm",
221                 .mode           = 0555,
222                 .child          = vm_table,
223         },
224         {
225                 .ctl_name       = CTL_FS,
226                 .procname       = "fs",
227                 .mode           = 0555,
228                 .child          = fs_table,
229         },
230         {
231                 .ctl_name       = CTL_DEBUG,
232                 .procname       = "debug",
233                 .mode           = 0555,
234                 .child          = debug_table,
235         },
236         {
237                 .ctl_name       = CTL_DEV,
238                 .procname       = "dev",
239                 .mode           = 0555,
240                 .child          = dev_table,
241         },
242 /*
243  * NOTE: do not add new entries to this table unless you have read
244  * Documentation/sysctl/ctl_unnumbered.txt
245  */
246         { .ctl_name = 0 }
247 };
248
249 #ifdef CONFIG_SCHED_DEBUG
250 static int min_sched_granularity_ns = 100000;           /* 100 usecs */
251 static int max_sched_granularity_ns = NSEC_PER_SEC;     /* 1 second */
252 static int min_wakeup_granularity_ns;                   /* 0 usecs */
253 static int max_wakeup_granularity_ns = NSEC_PER_SEC;    /* 1 second */
254 static int min_sched_tunable_scaling = SCHED_TUNABLESCALING_NONE;
255 static int max_sched_tunable_scaling = SCHED_TUNABLESCALING_END-1;
256 #endif
257
258 static struct ctl_table kern_table[] = {
259         {
260                 .ctl_name       = CTL_UNNUMBERED,
261                 .procname       = "sched_child_runs_first",
262                 .data           = &sysctl_sched_child_runs_first,
263                 .maxlen         = sizeof(unsigned int),
264                 .mode           = 0644,
265                 .proc_handler   = &proc_dointvec,
266         },
267 #ifdef CONFIG_SCHED_DEBUG
268         {
269                 .ctl_name       = CTL_UNNUMBERED,
270                 .procname       = "sched_min_granularity_ns",
271                 .data           = &sysctl_sched_min_granularity,
272                 .maxlen         = sizeof(unsigned int),
273                 .mode           = 0644,
274                 .proc_handler   = &sched_nr_latency_handler,
275                 .strategy       = &sysctl_intvec,
276                 .extra1         = &min_sched_granularity_ns,
277                 .extra2         = &max_sched_granularity_ns,
278         },
279         {
280                 .ctl_name       = CTL_UNNUMBERED,
281                 .procname       = "sched_latency_ns",
282                 .data           = &sysctl_sched_latency,
283                 .maxlen         = sizeof(unsigned int),
284                 .mode           = 0644,
285                 .proc_handler   = &sched_nr_latency_handler,
286                 .strategy       = &sysctl_intvec,
287                 .extra1         = &min_sched_granularity_ns,
288                 .extra2         = &max_sched_granularity_ns,
289         },
290         {
291                 .ctl_name       = CTL_UNNUMBERED,
292                 .procname       = "sched_wakeup_granularity_ns",
293                 .data           = &sysctl_sched_wakeup_granularity,
294                 .maxlen         = sizeof(unsigned int),
295                 .mode           = 0644,
296                 .proc_handler   = &proc_dointvec_minmax,
297                 .strategy       = &sysctl_intvec,
298                 .extra1         = &min_wakeup_granularity_ns,
299                 .extra2         = &max_wakeup_granularity_ns,
300         },
301         {
302                 .ctl_name       = CTL_UNNUMBERED,
303                 .procname       = "sched_shares_ratelimit",
304                 .data           = &sysctl_sched_shares_ratelimit,
305                 .maxlen         = sizeof(unsigned int),
306                 .mode           = 0644,
307                 .proc_handler   = &proc_dointvec,
308         },
309         {
310                 .ctl_name       = CTL_UNNUMBERED,
311                 .procname       = "sched_tunable_scaling",
312                 .data           = &sysctl_sched_tunable_scaling,
313                 .maxlen         = sizeof(enum sched_tunable_scaling),
314                 .mode           = 0644,
315                 .proc_handler   = &proc_dointvec_minmax,
316                 .strategy       = &sysctl_intvec,
317                 .extra1         = &min_sched_tunable_scaling,
318                 .extra2         = &max_sched_tunable_scaling,
319         },
320
321         {
322                 .ctl_name       = CTL_UNNUMBERED,
323                 .procname       = "sched_shares_thresh",
324                 .data           = &sysctl_sched_shares_thresh,
325                 .maxlen         = sizeof(unsigned int),
326                 .mode           = 0644,
327                 .proc_handler   = &proc_dointvec_minmax,
328                 .strategy       = &sysctl_intvec,
329                 .extra1         = &zero,
330         },
331         {
332                 .ctl_name       = CTL_UNNUMBERED,
333                 .procname       = "sched_migration_cost",
334                 .data           = &sysctl_sched_migration_cost,
335                 .maxlen         = sizeof(unsigned int),
336                 .mode           = 0644,
337                 .proc_handler   = &proc_dointvec,
338         },
339         {
340                 .ctl_name       = CTL_UNNUMBERED,
341                 .procname       = "sched_nr_migrate",
342                 .data           = &sysctl_sched_nr_migrate,
343                 .maxlen         = sizeof(unsigned int),
344                 .mode           = 0644,
345                 .proc_handler   = &proc_dointvec,
346         },
347         {
348                 .ctl_name       = CTL_UNNUMBERED,
349                 .procname       = "sched_time_avg",
350                 .data           = &sysctl_sched_time_avg,
351                 .maxlen         = sizeof(unsigned int),
352                 .mode           = 0644,
353                 .proc_handler   = &proc_dointvec,
354         },
355         {
356                 .ctl_name       = CTL_UNNUMBERED,
357                 .procname       = "timer_migration",
358                 .data           = &sysctl_timer_migration,
359                 .maxlen         = sizeof(unsigned int),
360                 .mode           = 0644,
361                 .proc_handler   = &proc_dointvec_minmax,
362                 .strategy       = &sysctl_intvec,
363                 .extra1         = &zero,
364                 .extra2         = &one,
365         },
366 #endif
367         {
368                 .ctl_name       = CTL_UNNUMBERED,
369                 .procname       = "sched_rt_period_us",
370                 .data           = &sysctl_sched_rt_period,
371                 .maxlen         = sizeof(unsigned int),
372                 .mode           = 0644,
373                 .proc_handler   = &sched_rt_handler,
374         },
375         {
376                 .ctl_name       = CTL_UNNUMBERED,
377                 .procname       = "sched_rt_runtime_us",
378                 .data           = &sysctl_sched_rt_runtime,
379                 .maxlen         = sizeof(int),
380                 .mode           = 0644,
381                 .proc_handler   = &sched_rt_handler,
382         },
383         {
384                 .ctl_name       = CTL_UNNUMBERED,
385                 .procname       = "sched_compat_yield",
386                 .data           = &sysctl_sched_compat_yield,
387                 .maxlen         = sizeof(unsigned int),
388                 .mode           = 0644,
389                 .proc_handler   = &proc_dointvec,
390         },
391 #ifdef CONFIG_PROVE_LOCKING
392         {
393                 .ctl_name       = CTL_UNNUMBERED,
394                 .procname       = "prove_locking",
395                 .data           = &prove_locking,
396                 .maxlen         = sizeof(int),
397                 .mode           = 0644,
398                 .proc_handler   = &proc_dointvec,
399         },
400 #endif
401 #ifdef CONFIG_LOCK_STAT
402         {
403                 .ctl_name       = CTL_UNNUMBERED,
404                 .procname       = "lock_stat",
405                 .data           = &lock_stat,
406                 .maxlen         = sizeof(int),
407                 .mode           = 0644,
408                 .proc_handler   = &proc_dointvec,
409         },
410 #endif
411         {
412                 .ctl_name       = KERN_PANIC,
413                 .procname       = "panic",
414                 .data           = &panic_timeout,
415                 .maxlen         = sizeof(int),
416                 .mode           = 0644,
417                 .proc_handler   = &proc_dointvec,
418         },
419         {
420                 .ctl_name       = KERN_CORE_USES_PID,
421                 .procname       = "core_uses_pid",
422                 .data           = &core_uses_pid,
423                 .maxlen         = sizeof(int),
424                 .mode           = 0644,
425                 .proc_handler   = &proc_dointvec,
426         },
427         {
428                 .ctl_name       = KERN_CORE_PATTERN,
429                 .procname       = "core_pattern",
430                 .data           = core_pattern,
431                 .maxlen         = CORENAME_MAX_SIZE,
432                 .mode           = 0644,
433                 .proc_handler   = &proc_dostring,
434                 .strategy       = &sysctl_string,
435         },
436         {
437                 .ctl_name       = CTL_UNNUMBERED,
438                 .procname       = "core_pipe_limit",
439                 .data           = &core_pipe_limit,
440                 .maxlen         = sizeof(unsigned int),
441                 .mode           = 0644,
442                 .proc_handler   = &proc_dointvec,
443         },
444 #ifdef CONFIG_PROC_SYSCTL
445         {
446                 .procname       = "tainted",
447                 .maxlen         = sizeof(long),
448                 .mode           = 0644,
449                 .proc_handler   = &proc_taint,
450         },
451 #endif
452 #ifdef CONFIG_LATENCYTOP
453         {
454                 .procname       = "latencytop",
455                 .data           = &latencytop_enabled,
456                 .maxlen         = sizeof(int),
457                 .mode           = 0644,
458                 .proc_handler   = &proc_dointvec,
459         },
460 #endif
461 #ifdef CONFIG_BLK_DEV_INITRD
462         {
463                 .ctl_name       = KERN_REALROOTDEV,
464                 .procname       = "real-root-dev",
465                 .data           = &real_root_dev,
466                 .maxlen         = sizeof(int),
467                 .mode           = 0644,
468                 .proc_handler   = &proc_dointvec,
469         },
470 #endif
471         {
472                 .ctl_name       = CTL_UNNUMBERED,
473                 .procname       = "print-fatal-signals",
474                 .data           = &print_fatal_signals,
475                 .maxlen         = sizeof(int),
476                 .mode           = 0644,
477                 .proc_handler   = &proc_dointvec,
478         },
479 #ifdef CONFIG_SPARC
480         {
481                 .ctl_name       = KERN_SPARC_REBOOT,
482                 .procname       = "reboot-cmd",
483                 .data           = reboot_command,
484                 .maxlen         = 256,
485                 .mode           = 0644,
486                 .proc_handler   = &proc_dostring,
487                 .strategy       = &sysctl_string,
488         },
489         {
490                 .ctl_name       = KERN_SPARC_STOP_A,
491                 .procname       = "stop-a",
492                 .data           = &stop_a_enabled,
493                 .maxlen         = sizeof (int),
494                 .mode           = 0644,
495                 .proc_handler   = &proc_dointvec,
496         },
497         {
498                 .ctl_name       = KERN_SPARC_SCONS_PWROFF,
499                 .procname       = "scons-poweroff",
500                 .data           = &scons_pwroff,
501                 .maxlen         = sizeof (int),
502                 .mode           = 0644,
503                 .proc_handler   = &proc_dointvec,
504         },
505 #endif
506 #ifdef CONFIG_SPARC64
507         {
508                 .ctl_name       = CTL_UNNUMBERED,
509                 .procname       = "tsb-ratio",
510                 .data           = &sysctl_tsb_ratio,
511                 .maxlen         = sizeof (int),
512                 .mode           = 0644,
513                 .proc_handler   = &proc_dointvec,
514         },
515 #endif
516 #ifdef __hppa__
517         {
518                 .ctl_name       = KERN_HPPA_PWRSW,
519                 .procname       = "soft-power",
520                 .data           = &pwrsw_enabled,
521                 .maxlen         = sizeof (int),
522                 .mode           = 0644,
523                 .proc_handler   = &proc_dointvec,
524         },
525         {
526                 .ctl_name       = KERN_HPPA_UNALIGNED,
527                 .procname       = "unaligned-trap",
528                 .data           = &unaligned_enabled,
529                 .maxlen         = sizeof (int),
530                 .mode           = 0644,
531                 .proc_handler   = &proc_dointvec,
532         },
533 #endif
534         {
535                 .ctl_name       = KERN_CTLALTDEL,
536                 .procname       = "ctrl-alt-del",
537                 .data           = &C_A_D,
538                 .maxlen         = sizeof(int),
539                 .mode           = 0644,
540                 .proc_handler   = &proc_dointvec,
541         },
542 #ifdef CONFIG_FUNCTION_TRACER
543         {
544                 .ctl_name       = CTL_UNNUMBERED,
545                 .procname       = "ftrace_enabled",
546                 .data           = &ftrace_enabled,
547                 .maxlen         = sizeof(int),
548                 .mode           = 0644,
549                 .proc_handler   = &ftrace_enable_sysctl,
550         },
551 #endif
552 #ifdef CONFIG_STACK_TRACER
553         {
554                 .ctl_name       = CTL_UNNUMBERED,
555                 .procname       = "stack_tracer_enabled",
556                 .data           = &stack_tracer_enabled,
557                 .maxlen         = sizeof(int),
558                 .mode           = 0644,
559                 .proc_handler   = &stack_trace_sysctl,
560         },
561 #endif
562 #ifdef CONFIG_TRACING
563         {
564                 .ctl_name       = CTL_UNNUMBERED,
565                 .procname       = "ftrace_dump_on_oops",
566                 .data           = &ftrace_dump_on_oops,
567                 .maxlen         = sizeof(int),
568                 .mode           = 0644,
569                 .proc_handler   = &proc_dointvec,
570         },
571 #endif
572 #ifdef CONFIG_MODULES
573         {
574                 .ctl_name       = KERN_MODPROBE,
575                 .procname       = "modprobe",
576                 .data           = &modprobe_path,
577                 .maxlen         = KMOD_PATH_LEN,
578                 .mode           = 0644,
579                 .proc_handler   = &proc_dostring,
580                 .strategy       = &sysctl_string,
581         },
582         {
583                 .ctl_name       = CTL_UNNUMBERED,
584                 .procname       = "modules_disabled",
585                 .data           = &modules_disabled,
586                 .maxlen         = sizeof(int),
587                 .mode           = 0644,
588                 /* only handle a transition from default "0" to "1" */
589                 .proc_handler   = &proc_dointvec_minmax,
590                 .extra1         = &one,
591                 .extra2         = &one,
592         },
593 #endif
594 #if defined(CONFIG_HOTPLUG) && defined(CONFIG_NET)
595         {
596                 .ctl_name       = KERN_HOTPLUG,
597                 .procname       = "hotplug",
598                 .data           = &uevent_helper,
599                 .maxlen         = UEVENT_HELPER_PATH_LEN,
600                 .mode           = 0644,
601                 .proc_handler   = &proc_dostring,
602                 .strategy       = &sysctl_string,
603         },
604 #endif
605 #ifdef CONFIG_CHR_DEV_SG
606         {
607                 .ctl_name       = KERN_SG_BIG_BUFF,
608                 .procname       = "sg-big-buff",
609                 .data           = &sg_big_buff,
610                 .maxlen         = sizeof (int),
611                 .mode           = 0444,
612                 .proc_handler   = &proc_dointvec,
613         },
614 #endif
615 #ifdef CONFIG_BSD_PROCESS_ACCT
616         {
617                 .ctl_name       = KERN_ACCT,
618                 .procname       = "acct",
619                 .data           = &acct_parm,
620                 .maxlen         = 3*sizeof(int),
621                 .mode           = 0644,
622                 .proc_handler   = &proc_dointvec,
623         },
624 #endif
625 #ifdef CONFIG_MAGIC_SYSRQ
626         {
627                 .ctl_name       = KERN_SYSRQ,
628                 .procname       = "sysrq",
629                 .data           = &__sysrq_enabled,
630                 .maxlen         = sizeof (int),
631                 .mode           = 0644,
632                 .proc_handler   = &proc_dointvec,
633         },
634 #endif
635 #ifdef CONFIG_PROC_SYSCTL
636         {
637                 .procname       = "cad_pid",
638                 .data           = NULL,
639                 .maxlen         = sizeof (int),
640                 .mode           = 0600,
641                 .proc_handler   = &proc_do_cad_pid,
642         },
643 #endif
644         {
645                 .ctl_name       = KERN_MAX_THREADS,
646                 .procname       = "threads-max",
647                 .data           = &max_threads,
648                 .maxlen         = sizeof(int),
649                 .mode           = 0644,
650                 .proc_handler   = &proc_dointvec,
651         },
652         {
653                 .ctl_name       = KERN_RANDOM,
654                 .procname       = "random",
655                 .mode           = 0555,
656                 .child          = random_table,
657         },
658         {
659                 .ctl_name       = KERN_OVERFLOWUID,
660                 .procname       = "overflowuid",
661                 .data           = &overflowuid,
662                 .maxlen         = sizeof(int),
663                 .mode           = 0644,
664                 .proc_handler   = &proc_dointvec_minmax,
665                 .strategy       = &sysctl_intvec,
666                 .extra1         = &minolduid,
667                 .extra2         = &maxolduid,
668         },
669         {
670                 .ctl_name       = KERN_OVERFLOWGID,
671                 .procname       = "overflowgid",
672                 .data           = &overflowgid,
673                 .maxlen         = sizeof(int),
674                 .mode           = 0644,
675                 .proc_handler   = &proc_dointvec_minmax,
676                 .strategy       = &sysctl_intvec,
677                 .extra1         = &minolduid,
678                 .extra2         = &maxolduid,
679         },
680 #ifdef CONFIG_S390
681 #ifdef CONFIG_MATHEMU
682         {
683                 .ctl_name       = KERN_IEEE_EMULATION_WARNINGS,
684                 .procname       = "ieee_emulation_warnings",
685                 .data           = &sysctl_ieee_emulation_warnings,
686                 .maxlen         = sizeof(int),
687                 .mode           = 0644,
688                 .proc_handler   = &proc_dointvec,
689         },
690 #endif
691         {
692                 .ctl_name       = KERN_S390_USER_DEBUG_LOGGING,
693                 .procname       = "userprocess_debug",
694                 .data           = &sysctl_userprocess_debug,
695                 .maxlen         = sizeof(int),
696                 .mode           = 0644,
697                 .proc_handler   = &proc_dointvec,
698         },
699 #endif
700         {
701                 .ctl_name       = KERN_PIDMAX,
702                 .procname       = "pid_max",
703                 .data           = &pid_max,
704                 .maxlen         = sizeof (int),
705                 .mode           = 0644,
706                 .proc_handler   = &proc_dointvec_minmax,
707                 .strategy       = sysctl_intvec,
708                 .extra1         = &pid_max_min,
709                 .extra2         = &pid_max_max,
710         },
711         {
712                 .ctl_name       = KERN_PANIC_ON_OOPS,
713                 .procname       = "panic_on_oops",
714                 .data           = &panic_on_oops,
715                 .maxlen         = sizeof(int),
716                 .mode           = 0644,
717                 .proc_handler   = &proc_dointvec,
718         },
719 #if defined CONFIG_PRINTK
720         {
721                 .ctl_name       = KERN_PRINTK,
722                 .procname       = "printk",
723                 .data           = &console_loglevel,
724                 .maxlen         = 4*sizeof(int),
725                 .mode           = 0644,
726                 .proc_handler   = &proc_dointvec,
727         },
728         {
729                 .ctl_name       = KERN_PRINTK_RATELIMIT,
730                 .procname       = "printk_ratelimit",
731                 .data           = &printk_ratelimit_state.interval,
732                 .maxlen         = sizeof(int),
733                 .mode           = 0644,
734                 .proc_handler   = &proc_dointvec_jiffies,
735                 .strategy       = &sysctl_jiffies,
736         },
737         {
738                 .ctl_name       = KERN_PRINTK_RATELIMIT_BURST,
739                 .procname       = "printk_ratelimit_burst",
740                 .data           = &printk_ratelimit_state.burst,
741                 .maxlen         = sizeof(int),
742                 .mode           = 0644,
743                 .proc_handler   = &proc_dointvec,
744         },
745         {
746                 .ctl_name       = CTL_UNNUMBERED,
747                 .procname       = "printk_delay",
748                 .data           = &printk_delay_msec,
749                 .maxlen         = sizeof(int),
750                 .mode           = 0644,
751                 .proc_handler   = &proc_dointvec_minmax,
752                 .strategy       = &sysctl_intvec,
753                 .extra1         = &zero,
754                 .extra2         = &ten_thousand,
755         },
756 #endif
757         {
758                 .ctl_name       = KERN_NGROUPS_MAX,
759                 .procname       = "ngroups_max",
760                 .data           = &ngroups_max,
761                 .maxlen         = sizeof (int),
762                 .mode           = 0444,
763                 .proc_handler   = &proc_dointvec,
764         },
765 #if defined(CONFIG_X86_LOCAL_APIC) && defined(CONFIG_X86)
766         {
767                 .ctl_name       = KERN_UNKNOWN_NMI_PANIC,
768                 .procname       = "unknown_nmi_panic",
769                 .data           = &unknown_nmi_panic,
770                 .maxlen         = sizeof (int),
771                 .mode           = 0644,
772                 .proc_handler   = &proc_dointvec,
773         },
774         {
775                 .procname       = "nmi_watchdog",
776                 .data           = &nmi_watchdog_enabled,
777                 .maxlen         = sizeof (int),
778                 .mode           = 0644,
779                 .proc_handler   = &proc_nmi_enabled,
780         },
781 #endif
782 #if defined(CONFIG_X86)
783         {
784                 .ctl_name       = KERN_PANIC_ON_NMI,
785                 .procname       = "panic_on_unrecovered_nmi",
786                 .data           = &panic_on_unrecovered_nmi,
787                 .maxlen         = sizeof(int),
788                 .mode           = 0644,
789                 .proc_handler   = &proc_dointvec,
790         },
791         {
792                 .ctl_name       = CTL_UNNUMBERED,
793                 .procname       = "panic_on_io_nmi",
794                 .data           = &panic_on_io_nmi,
795                 .maxlen         = sizeof(int),
796                 .mode           = 0644,
797                 .proc_handler   = &proc_dointvec,
798         },
799         {
800                 .ctl_name       = KERN_BOOTLOADER_TYPE,
801                 .procname       = "bootloader_type",
802                 .data           = &bootloader_type,
803                 .maxlen         = sizeof (int),
804                 .mode           = 0444,
805                 .proc_handler   = &proc_dointvec,
806         },
807         {
808                 .ctl_name       = CTL_UNNUMBERED,
809                 .procname       = "bootloader_version",
810                 .data           = &bootloader_version,
811                 .maxlen         = sizeof (int),
812                 .mode           = 0444,
813                 .proc_handler   = &proc_dointvec,
814         },
815         {
816                 .ctl_name       = CTL_UNNUMBERED,
817                 .procname       = "kstack_depth_to_print",
818                 .data           = &kstack_depth_to_print,
819                 .maxlen         = sizeof(int),
820                 .mode           = 0644,
821                 .proc_handler   = &proc_dointvec,
822         },
823         {
824                 .ctl_name       = CTL_UNNUMBERED,
825                 .procname       = "io_delay_type",
826                 .data           = &io_delay_type,
827                 .maxlen         = sizeof(int),
828                 .mode           = 0644,
829                 .proc_handler   = &proc_dointvec,
830         },
831 #endif
832 #if defined(CONFIG_MMU)
833         {
834                 .ctl_name       = KERN_RANDOMIZE,
835                 .procname       = "randomize_va_space",
836                 .data           = &randomize_va_space,
837                 .maxlen         = sizeof(int),
838                 .mode           = 0644,
839                 .proc_handler   = &proc_dointvec,
840         },
841 #endif
842 #if defined(CONFIG_S390) && defined(CONFIG_SMP)
843         {
844                 .ctl_name       = KERN_SPIN_RETRY,
845                 .procname       = "spin_retry",
846                 .data           = &spin_retry,
847                 .maxlen         = sizeof (int),
848                 .mode           = 0644,
849                 .proc_handler   = &proc_dointvec,
850         },
851 #endif
852 #if     defined(CONFIG_ACPI_SLEEP) && defined(CONFIG_X86)
853         {
854                 .procname       = "acpi_video_flags",
855                 .data           = &acpi_realmode_flags,
856                 .maxlen         = sizeof (unsigned long),
857                 .mode           = 0644,
858                 .proc_handler   = &proc_doulongvec_minmax,
859         },
860 #endif
861 #ifdef CONFIG_IA64
862         {
863                 .ctl_name       = KERN_IA64_UNALIGNED,
864                 .procname       = "ignore-unaligned-usertrap",
865                 .data           = &no_unaligned_warning,
866                 .maxlen         = sizeof (int),
867                 .mode           = 0644,
868                 .proc_handler   = &proc_dointvec,
869         },
870         {
871                 .ctl_name       = CTL_UNNUMBERED,
872                 .procname       = "unaligned-dump-stack",
873                 .data           = &unaligned_dump_stack,
874                 .maxlen         = sizeof (int),
875                 .mode           = 0644,
876                 .proc_handler   = &proc_dointvec,
877         },
878 #endif
879 #ifdef CONFIG_DETECT_SOFTLOCKUP
880         {
881                 .ctl_name       = CTL_UNNUMBERED,
882                 .procname       = "softlockup_panic",
883                 .data           = &softlockup_panic,
884                 .maxlen         = sizeof(int),
885                 .mode           = 0644,
886                 .proc_handler   = &proc_dointvec_minmax,
887                 .strategy       = &sysctl_intvec,
888                 .extra1         = &zero,
889                 .extra2         = &one,
890         },
891         {
892                 .ctl_name       = CTL_UNNUMBERED,
893                 .procname       = "softlockup_thresh",
894                 .data           = &softlockup_thresh,
895                 .maxlen         = sizeof(int),
896                 .mode           = 0644,
897                 .proc_handler   = &proc_dosoftlockup_thresh,
898                 .strategy       = &sysctl_intvec,
899                 .extra1         = &neg_one,
900                 .extra2         = &sixty,
901         },
902 #endif
903 #ifdef CONFIG_DETECT_HUNG_TASK
904         {
905                 .ctl_name       = CTL_UNNUMBERED,
906                 .procname       = "hung_task_panic",
907                 .data           = &sysctl_hung_task_panic,
908                 .maxlen         = sizeof(int),
909                 .mode           = 0644,
910                 .proc_handler   = &proc_dointvec_minmax,
911                 .strategy       = &sysctl_intvec,
912                 .extra1         = &zero,
913                 .extra2         = &one,
914         },
915         {
916                 .ctl_name       = CTL_UNNUMBERED,
917                 .procname       = "hung_task_check_count",
918                 .data           = &sysctl_hung_task_check_count,
919                 .maxlen         = sizeof(unsigned long),
920                 .mode           = 0644,
921                 .proc_handler   = &proc_doulongvec_minmax,
922                 .strategy       = &sysctl_intvec,
923         },
924         {
925                 .ctl_name       = CTL_UNNUMBERED,
926                 .procname       = "hung_task_timeout_secs",
927                 .data           = &sysctl_hung_task_timeout_secs,
928                 .maxlen         = sizeof(unsigned long),
929                 .mode           = 0644,
930                 .proc_handler   = &proc_dohung_task_timeout_secs,
931                 .strategy       = &sysctl_intvec,
932         },
933         {
934                 .ctl_name       = CTL_UNNUMBERED,
935                 .procname       = "hung_task_warnings",
936                 .data           = &sysctl_hung_task_warnings,
937                 .maxlen         = sizeof(unsigned long),
938                 .mode           = 0644,
939                 .proc_handler   = &proc_doulongvec_minmax,
940                 .strategy       = &sysctl_intvec,
941         },
942 #endif
943 #ifdef CONFIG_COMPAT
944         {
945                 .ctl_name       = KERN_COMPAT_LOG,
946                 .procname       = "compat-log",
947                 .data           = &compat_log,
948                 .maxlen         = sizeof (int),
949                 .mode           = 0644,
950                 .proc_handler   = &proc_dointvec,
951         },
952 #endif
953 #ifdef CONFIG_RT_MUTEXES
954         {
955                 .ctl_name       = KERN_MAX_LOCK_DEPTH,
956                 .procname       = "max_lock_depth",
957                 .data           = &max_lock_depth,
958                 .maxlen         = sizeof(int),
959                 .mode           = 0644,
960                 .proc_handler   = &proc_dointvec,
961         },
962 #endif
963         {
964                 .ctl_name       = CTL_UNNUMBERED,
965                 .procname       = "poweroff_cmd",
966                 .data           = &poweroff_cmd,
967                 .maxlen         = POWEROFF_CMD_PATH_LEN,
968                 .mode           = 0644,
969                 .proc_handler   = &proc_dostring,
970                 .strategy       = &sysctl_string,
971         },
972 #ifdef CONFIG_KEYS
973         {
974                 .ctl_name       = CTL_UNNUMBERED,
975                 .procname       = "keys",
976                 .mode           = 0555,
977                 .child          = key_sysctls,
978         },
979 #endif
980 #ifdef CONFIG_RCU_TORTURE_TEST
981         {
982                 .ctl_name       = CTL_UNNUMBERED,
983                 .procname       = "rcutorture_runnable",
984                 .data           = &rcutorture_runnable,
985                 .maxlen         = sizeof(int),
986                 .mode           = 0644,
987                 .proc_handler   = &proc_dointvec,
988         },
989 #endif
990 #ifdef CONFIG_SLOW_WORK
991         {
992                 .ctl_name       = CTL_UNNUMBERED,
993                 .procname       = "slow-work",
994                 .mode           = 0555,
995                 .child          = slow_work_sysctls,
996         },
997 #endif
998 #ifdef CONFIG_PERF_EVENTS
999         {
1000                 .ctl_name       = CTL_UNNUMBERED,
1001                 .procname       = "perf_event_paranoid",
1002                 .data           = &sysctl_perf_event_paranoid,
1003                 .maxlen         = sizeof(sysctl_perf_event_paranoid),
1004                 .mode           = 0644,
1005                 .proc_handler   = &proc_dointvec,
1006         },
1007         {
1008                 .ctl_name       = CTL_UNNUMBERED,
1009                 .procname       = "perf_event_mlock_kb",
1010                 .data           = &sysctl_perf_event_mlock,
1011                 .maxlen         = sizeof(sysctl_perf_event_mlock),
1012                 .mode           = 0644,
1013                 .proc_handler   = &proc_dointvec,
1014         },
1015         {
1016                 .ctl_name       = CTL_UNNUMBERED,
1017                 .procname       = "perf_event_max_sample_rate",
1018                 .data           = &sysctl_perf_event_sample_rate,
1019                 .maxlen         = sizeof(sysctl_perf_event_sample_rate),
1020                 .mode           = 0644,
1021                 .proc_handler   = &proc_dointvec,
1022         },
1023 #endif
1024 #ifdef CONFIG_KMEMCHECK
1025         {
1026                 .ctl_name       = CTL_UNNUMBERED,
1027                 .procname       = "kmemcheck",
1028                 .data           = &kmemcheck_enabled,
1029                 .maxlen         = sizeof(int),
1030                 .mode           = 0644,
1031                 .proc_handler   = &proc_dointvec,
1032         },
1033 #endif
1034 #ifdef CONFIG_BLOCK
1035         {
1036                 .ctl_name       = CTL_UNNUMBERED,
1037                 .procname       = "blk_iopoll",
1038                 .data           = &blk_iopoll_enabled,
1039                 .maxlen         = sizeof(int),
1040                 .mode           = 0644,
1041                 .proc_handler   = &proc_dointvec,
1042         },
1043 #endif
1044 /*
1045  * NOTE: do not add new entries to this table unless you have read
1046  * Documentation/sysctl/ctl_unnumbered.txt
1047  */
1048         { .ctl_name = 0 }
1049 };
1050
1051 static struct ctl_table vm_table[] = {
1052         {
1053                 .ctl_name       = VM_OVERCOMMIT_MEMORY,
1054                 .procname       = "overcommit_memory",
1055                 .data           = &sysctl_overcommit_memory,
1056                 .maxlen         = sizeof(sysctl_overcommit_memory),
1057                 .mode           = 0644,
1058                 .proc_handler   = &proc_dointvec,
1059         },
1060         {
1061                 .ctl_name       = VM_PANIC_ON_OOM,
1062                 .procname       = "panic_on_oom",
1063                 .data           = &sysctl_panic_on_oom,
1064                 .maxlen         = sizeof(sysctl_panic_on_oom),
1065                 .mode           = 0644,
1066                 .proc_handler   = &proc_dointvec,
1067         },
1068         {
1069                 .ctl_name       = CTL_UNNUMBERED,
1070                 .procname       = "oom_kill_allocating_task",
1071                 .data           = &sysctl_oom_kill_allocating_task,
1072                 .maxlen         = sizeof(sysctl_oom_kill_allocating_task),
1073                 .mode           = 0644,
1074                 .proc_handler   = &proc_dointvec,
1075         },
1076         {
1077                 .ctl_name       = CTL_UNNUMBERED,
1078                 .procname       = "oom_dump_tasks",
1079                 .data           = &sysctl_oom_dump_tasks,
1080                 .maxlen         = sizeof(sysctl_oom_dump_tasks),
1081                 .mode           = 0644,
1082                 .proc_handler   = &proc_dointvec,
1083         },
1084         {
1085                 .ctl_name       = VM_OVERCOMMIT_RATIO,
1086                 .procname       = "overcommit_ratio",
1087                 .data           = &sysctl_overcommit_ratio,
1088                 .maxlen         = sizeof(sysctl_overcommit_ratio),
1089                 .mode           = 0644,
1090                 .proc_handler   = &proc_dointvec,
1091         },
1092         {
1093                 .ctl_name       = VM_PAGE_CLUSTER,
1094                 .procname       = "page-cluster", 
1095                 .data           = &page_cluster,
1096                 .maxlen         = sizeof(int),
1097                 .mode           = 0644,
1098                 .proc_handler   = &proc_dointvec,
1099         },
1100         {
1101                 .ctl_name       = VM_DIRTY_BACKGROUND,
1102                 .procname       = "dirty_background_ratio",
1103                 .data           = &dirty_background_ratio,
1104                 .maxlen         = sizeof(dirty_background_ratio),
1105                 .mode           = 0644,
1106                 .proc_handler   = &dirty_background_ratio_handler,
1107                 .strategy       = &sysctl_intvec,
1108                 .extra1         = &zero,
1109                 .extra2         = &one_hundred,
1110         },
1111         {
1112                 .ctl_name       = CTL_UNNUMBERED,
1113                 .procname       = "dirty_background_bytes",
1114                 .data           = &dirty_background_bytes,
1115                 .maxlen         = sizeof(dirty_background_bytes),
1116                 .mode           = 0644,
1117                 .proc_handler   = &dirty_background_bytes_handler,
1118                 .strategy       = &sysctl_intvec,
1119                 .extra1         = &one_ul,
1120         },
1121         {
1122                 .ctl_name       = VM_DIRTY_RATIO,
1123                 .procname       = "dirty_ratio",
1124                 .data           = &vm_dirty_ratio,
1125                 .maxlen         = sizeof(vm_dirty_ratio),
1126                 .mode           = 0644,
1127                 .proc_handler   = &dirty_ratio_handler,
1128                 .strategy       = &sysctl_intvec,
1129                 .extra1         = &zero,
1130                 .extra2         = &one_hundred,
1131         },
1132         {
1133                 .ctl_name       = CTL_UNNUMBERED,
1134                 .procname       = "dirty_bytes",
1135                 .data           = &vm_dirty_bytes,
1136                 .maxlen         = sizeof(vm_dirty_bytes),
1137                 .mode           = 0644,
1138                 .proc_handler   = &dirty_bytes_handler,
1139                 .strategy       = &sysctl_intvec,
1140                 .extra1         = &dirty_bytes_min,
1141         },
1142         {
1143                 .procname       = "dirty_writeback_centisecs",
1144                 .data           = &dirty_writeback_interval,
1145                 .maxlen         = sizeof(dirty_writeback_interval),
1146                 .mode           = 0644,
1147                 .proc_handler   = &dirty_writeback_centisecs_handler,
1148         },
1149         {
1150                 .procname       = "dirty_expire_centisecs",
1151                 .data           = &dirty_expire_interval,
1152                 .maxlen         = sizeof(dirty_expire_interval),
1153                 .mode           = 0644,
1154                 .proc_handler   = &proc_dointvec,
1155         },
1156         {
1157                 .ctl_name       = VM_NR_PDFLUSH_THREADS,
1158                 .procname       = "nr_pdflush_threads",
1159                 .data           = &nr_pdflush_threads,
1160                 .maxlen         = sizeof nr_pdflush_threads,
1161                 .mode           = 0444 /* read-only*/,
1162                 .proc_handler   = &proc_dointvec,
1163         },
1164         {
1165                 .ctl_name       = VM_SWAPPINESS,
1166                 .procname       = "swappiness",
1167                 .data           = &vm_swappiness,
1168                 .maxlen         = sizeof(vm_swappiness),
1169                 .mode           = 0644,
1170                 .proc_handler   = &proc_dointvec_minmax,
1171                 .strategy       = &sysctl_intvec,
1172                 .extra1         = &zero,
1173                 .extra2         = &one_hundred,
1174         },
1175 #ifdef CONFIG_HUGETLB_PAGE
1176          {
1177                 .procname       = "nr_hugepages",
1178                 .data           = NULL,
1179                 .maxlen         = sizeof(unsigned long),
1180                 .mode           = 0644,
1181                 .proc_handler   = &hugetlb_sysctl_handler,
1182                 .extra1         = (void *)&hugetlb_zero,
1183                 .extra2         = (void *)&hugetlb_infinity,
1184          },
1185          {
1186                 .ctl_name       = VM_HUGETLB_GROUP,
1187                 .procname       = "hugetlb_shm_group",
1188                 .data           = &sysctl_hugetlb_shm_group,
1189                 .maxlen         = sizeof(gid_t),
1190                 .mode           = 0644,
1191                 .proc_handler   = &proc_dointvec,
1192          },
1193          {
1194                 .ctl_name       = CTL_UNNUMBERED,
1195                 .procname       = "hugepages_treat_as_movable",
1196                 .data           = &hugepages_treat_as_movable,
1197                 .maxlen         = sizeof(int),
1198                 .mode           = 0644,
1199                 .proc_handler   = &hugetlb_treat_movable_handler,
1200         },
1201         {
1202                 .ctl_name       = CTL_UNNUMBERED,
1203                 .procname       = "nr_overcommit_hugepages",
1204                 .data           = NULL,
1205                 .maxlen         = sizeof(unsigned long),
1206                 .mode           = 0644,
1207                 .proc_handler   = &hugetlb_overcommit_handler,
1208                 .extra1         = (void *)&hugetlb_zero,
1209                 .extra2         = (void *)&hugetlb_infinity,
1210         },
1211 #endif
1212         {
1213                 .ctl_name       = VM_LOWMEM_RESERVE_RATIO,
1214                 .procname       = "lowmem_reserve_ratio",
1215                 .data           = &sysctl_lowmem_reserve_ratio,
1216                 .maxlen         = sizeof(sysctl_lowmem_reserve_ratio),
1217                 .mode           = 0644,
1218                 .proc_handler   = &lowmem_reserve_ratio_sysctl_handler,
1219                 .strategy       = &sysctl_intvec,
1220         },
1221         {
1222                 .ctl_name       = VM_DROP_PAGECACHE,
1223                 .procname       = "drop_caches",
1224                 .data           = &sysctl_drop_caches,
1225                 .maxlen         = sizeof(int),
1226                 .mode           = 0644,
1227                 .proc_handler   = drop_caches_sysctl_handler,
1228                 .strategy       = &sysctl_intvec,
1229         },
1230         {
1231                 .ctl_name       = VM_MIN_FREE_KBYTES,
1232                 .procname       = "min_free_kbytes",
1233                 .data           = &min_free_kbytes,
1234                 .maxlen         = sizeof(min_free_kbytes),
1235                 .mode           = 0644,
1236                 .proc_handler   = &min_free_kbytes_sysctl_handler,
1237                 .strategy       = &sysctl_intvec,
1238                 .extra1         = &zero,
1239         },
1240         {
1241                 .ctl_name       = VM_PERCPU_PAGELIST_FRACTION,
1242                 .procname       = "percpu_pagelist_fraction",
1243                 .data           = &percpu_pagelist_fraction,
1244                 .maxlen         = sizeof(percpu_pagelist_fraction),
1245                 .mode           = 0644,
1246                 .proc_handler   = &percpu_pagelist_fraction_sysctl_handler,
1247                 .strategy       = &sysctl_intvec,
1248                 .extra1         = &min_percpu_pagelist_fract,
1249         },
1250 #ifdef CONFIG_MMU
1251         {
1252                 .ctl_name       = VM_MAX_MAP_COUNT,
1253                 .procname       = "max_map_count",
1254                 .data           = &sysctl_max_map_count,
1255                 .maxlen         = sizeof(sysctl_max_map_count),
1256                 .mode           = 0644,
1257                 .proc_handler   = &proc_dointvec
1258         },
1259 #else
1260         {
1261                 .ctl_name       = CTL_UNNUMBERED,
1262                 .procname       = "nr_trim_pages",
1263                 .data           = &sysctl_nr_trim_pages,
1264                 .maxlen         = sizeof(sysctl_nr_trim_pages),
1265                 .mode           = 0644,
1266                 .proc_handler   = &proc_dointvec_minmax,
1267                 .strategy       = &sysctl_intvec,
1268                 .extra1         = &zero,
1269         },
1270 #endif
1271         {
1272                 .ctl_name       = VM_LAPTOP_MODE,
1273                 .procname       = "laptop_mode",
1274                 .data           = &laptop_mode,
1275                 .maxlen         = sizeof(laptop_mode),
1276                 .mode           = 0644,
1277                 .proc_handler   = &proc_dointvec_jiffies,
1278                 .strategy       = &sysctl_jiffies,
1279         },
1280         {
1281                 .ctl_name       = VM_BLOCK_DUMP,
1282                 .procname       = "block_dump",
1283                 .data           = &block_dump,
1284                 .maxlen         = sizeof(block_dump),
1285                 .mode           = 0644,
1286                 .proc_handler   = &proc_dointvec,
1287                 .strategy       = &sysctl_intvec,
1288                 .extra1         = &zero,
1289         },
1290         {
1291                 .ctl_name       = VM_VFS_CACHE_PRESSURE,
1292                 .procname       = "vfs_cache_pressure",
1293                 .data           = &sysctl_vfs_cache_pressure,
1294                 .maxlen         = sizeof(sysctl_vfs_cache_pressure),
1295                 .mode           = 0644,
1296                 .proc_handler   = &proc_dointvec,
1297                 .strategy       = &sysctl_intvec,
1298                 .extra1         = &zero,
1299         },
1300 #ifdef HAVE_ARCH_PICK_MMAP_LAYOUT
1301         {
1302                 .ctl_name       = VM_LEGACY_VA_LAYOUT,
1303                 .procname       = "legacy_va_layout",
1304                 .data           = &sysctl_legacy_va_layout,
1305                 .maxlen         = sizeof(sysctl_legacy_va_layout),
1306                 .mode           = 0644,
1307                 .proc_handler   = &proc_dointvec,
1308                 .strategy       = &sysctl_intvec,
1309                 .extra1         = &zero,
1310         },
1311 #endif
1312 #ifdef CONFIG_NUMA
1313         {
1314                 .ctl_name       = VM_ZONE_RECLAIM_MODE,
1315                 .procname       = "zone_reclaim_mode",
1316                 .data           = &zone_reclaim_mode,
1317                 .maxlen         = sizeof(zone_reclaim_mode),
1318                 .mode           = 0644,
1319                 .proc_handler   = &proc_dointvec,
1320                 .strategy       = &sysctl_intvec,
1321                 .extra1         = &zero,
1322         },
1323         {
1324                 .ctl_name       = VM_MIN_UNMAPPED,
1325                 .procname       = "min_unmapped_ratio",
1326                 .data           = &sysctl_min_unmapped_ratio,
1327                 .maxlen         = sizeof(sysctl_min_unmapped_ratio),
1328                 .mode           = 0644,
1329                 .proc_handler   = &sysctl_min_unmapped_ratio_sysctl_handler,
1330                 .strategy       = &sysctl_intvec,
1331                 .extra1         = &zero,
1332                 .extra2         = &one_hundred,
1333         },
1334         {
1335                 .ctl_name       = VM_MIN_SLAB,
1336                 .procname       = "min_slab_ratio",
1337                 .data           = &sysctl_min_slab_ratio,
1338                 .maxlen         = sizeof(sysctl_min_slab_ratio),
1339                 .mode           = 0644,
1340                 .proc_handler   = &sysctl_min_slab_ratio_sysctl_handler,
1341                 .strategy       = &sysctl_intvec,
1342                 .extra1         = &zero,
1343                 .extra2         = &one_hundred,
1344         },
1345 #endif
1346 #ifdef CONFIG_SMP
1347         {
1348                 .ctl_name       = CTL_UNNUMBERED,
1349                 .procname       = "stat_interval",
1350                 .data           = &sysctl_stat_interval,
1351                 .maxlen         = sizeof(sysctl_stat_interval),
1352                 .mode           = 0644,
1353                 .proc_handler   = &proc_dointvec_jiffies,
1354                 .strategy       = &sysctl_jiffies,
1355         },
1356 #endif
1357         {
1358                 .ctl_name       = CTL_UNNUMBERED,
1359                 .procname       = "mmap_min_addr",
1360                 .data           = &dac_mmap_min_addr,
1361                 .maxlen         = sizeof(unsigned long),
1362                 .mode           = 0644,
1363                 .proc_handler   = &mmap_min_addr_handler,
1364         },
1365 #ifdef CONFIG_NUMA
1366         {
1367                 .ctl_name       = CTL_UNNUMBERED,
1368                 .procname       = "numa_zonelist_order",
1369                 .data           = &numa_zonelist_order,
1370                 .maxlen         = NUMA_ZONELIST_ORDER_LEN,
1371                 .mode           = 0644,
1372                 .proc_handler   = &numa_zonelist_order_handler,
1373                 .strategy       = &sysctl_string,
1374         },
1375 #endif
1376 #if (defined(CONFIG_X86_32) && !defined(CONFIG_UML))|| \
1377    (defined(CONFIG_SUPERH) && defined(CONFIG_VSYSCALL))
1378         {
1379                 .ctl_name       = VM_VDSO_ENABLED,
1380                 .procname       = "vdso_enabled",
1381                 .data           = &vdso_enabled,
1382                 .maxlen         = sizeof(vdso_enabled),
1383                 .mode           = 0644,
1384                 .proc_handler   = &proc_dointvec,
1385                 .strategy       = &sysctl_intvec,
1386                 .extra1         = &zero,
1387         },
1388 #endif
1389 #ifdef CONFIG_HIGHMEM
1390         {
1391                 .ctl_name       = CTL_UNNUMBERED,
1392                 .procname       = "highmem_is_dirtyable",
1393                 .data           = &vm_highmem_is_dirtyable,
1394                 .maxlen         = sizeof(vm_highmem_is_dirtyable),
1395                 .mode           = 0644,
1396                 .proc_handler   = &proc_dointvec_minmax,
1397                 .strategy       = &sysctl_intvec,
1398                 .extra1         = &zero,
1399                 .extra2         = &one,
1400         },
1401 #endif
1402         {
1403                 .ctl_name       = CTL_UNNUMBERED,
1404                 .procname       = "scan_unevictable_pages",
1405                 .data           = &scan_unevictable_pages,
1406                 .maxlen         = sizeof(scan_unevictable_pages),
1407                 .mode           = 0644,
1408                 .proc_handler   = &scan_unevictable_handler,
1409         },
1410 #ifdef CONFIG_MEMORY_FAILURE
1411         {
1412                 .ctl_name       = CTL_UNNUMBERED,
1413                 .procname       = "memory_failure_early_kill",
1414                 .data           = &sysctl_memory_failure_early_kill,
1415                 .maxlen         = sizeof(sysctl_memory_failure_early_kill),
1416                 .mode           = 0644,
1417                 .proc_handler   = &proc_dointvec_minmax,
1418                 .strategy       = &sysctl_intvec,
1419                 .extra1         = &zero,
1420                 .extra2         = &one,
1421         },
1422         {
1423                 .ctl_name       = CTL_UNNUMBERED,
1424                 .procname       = "memory_failure_recovery",
1425                 .data           = &sysctl_memory_failure_recovery,
1426                 .maxlen         = sizeof(sysctl_memory_failure_recovery),
1427                 .mode           = 0644,
1428                 .proc_handler   = &proc_dointvec_minmax,
1429                 .strategy       = &sysctl_intvec,
1430                 .extra1         = &zero,
1431                 .extra2         = &one,
1432         },
1433 #endif
1434
1435 /*
1436  * NOTE: do not add new entries to this table unless you have read
1437  * Documentation/sysctl/ctl_unnumbered.txt
1438  */
1439         { .ctl_name = 0 }
1440 };
1441
1442 #if defined(CONFIG_BINFMT_MISC) || defined(CONFIG_BINFMT_MISC_MODULE)
1443 static struct ctl_table binfmt_misc_table[] = {
1444         { .ctl_name = 0 }
1445 };
1446 #endif
1447
1448 static struct ctl_table fs_table[] = {
1449         {
1450                 .ctl_name       = FS_NRINODE,
1451                 .procname       = "inode-nr",
1452                 .data           = &inodes_stat,
1453                 .maxlen         = 2*sizeof(int),
1454                 .mode           = 0444,
1455                 .proc_handler   = &proc_dointvec,
1456         },
1457         {
1458                 .ctl_name       = FS_STATINODE,
1459                 .procname       = "inode-state",
1460                 .data           = &inodes_stat,
1461                 .maxlen         = 7*sizeof(int),
1462                 .mode           = 0444,
1463                 .proc_handler   = &proc_dointvec,
1464         },
1465         {
1466                 .procname       = "file-nr",
1467                 .data           = &files_stat,
1468                 .maxlen         = 3*sizeof(int),
1469                 .mode           = 0444,
1470                 .proc_handler   = &proc_nr_files,
1471         },
1472         {
1473                 .ctl_name       = FS_MAXFILE,
1474                 .procname       = "file-max",
1475                 .data           = &files_stat.max_files,
1476                 .maxlen         = sizeof(int),
1477                 .mode           = 0644,
1478                 .proc_handler   = &proc_dointvec,
1479         },
1480         {
1481                 .ctl_name       = CTL_UNNUMBERED,
1482                 .procname       = "nr_open",
1483                 .data           = &sysctl_nr_open,
1484                 .maxlen         = sizeof(int),
1485                 .mode           = 0644,
1486                 .proc_handler   = &proc_dointvec_minmax,
1487                 .extra1         = &sysctl_nr_open_min,
1488                 .extra2         = &sysctl_nr_open_max,
1489         },
1490         {
1491                 .ctl_name       = FS_DENTRY,
1492                 .procname       = "dentry-state",
1493                 .data           = &dentry_stat,
1494                 .maxlen         = 6*sizeof(int),
1495                 .mode           = 0444,
1496                 .proc_handler   = &proc_dointvec,
1497         },
1498         {
1499                 .ctl_name       = FS_OVERFLOWUID,
1500                 .procname       = "overflowuid",
1501                 .data           = &fs_overflowuid,
1502                 .maxlen         = sizeof(int),
1503                 .mode           = 0644,
1504                 .proc_handler   = &proc_dointvec_minmax,
1505                 .strategy       = &sysctl_intvec,
1506                 .extra1         = &minolduid,
1507                 .extra2         = &maxolduid,
1508         },
1509         {
1510                 .ctl_name       = FS_OVERFLOWGID,
1511                 .procname       = "overflowgid",
1512                 .data           = &fs_overflowgid,
1513                 .maxlen         = sizeof(int),
1514                 .mode           = 0644,
1515                 .proc_handler   = &proc_dointvec_minmax,
1516                 .strategy       = &sysctl_intvec,
1517                 .extra1         = &minolduid,
1518                 .extra2         = &maxolduid,
1519         },
1520 #ifdef CONFIG_FILE_LOCKING
1521         {
1522                 .ctl_name       = FS_LEASES,
1523                 .procname       = "leases-enable",
1524                 .data           = &leases_enable,
1525                 .maxlen         = sizeof(int),
1526                 .mode           = 0644,
1527                 .proc_handler   = &proc_dointvec,
1528         },
1529 #endif
1530 #ifdef CONFIG_DNOTIFY
1531         {
1532                 .ctl_name       = FS_DIR_NOTIFY,
1533                 .procname       = "dir-notify-enable",
1534                 .data           = &dir_notify_enable,
1535                 .maxlen         = sizeof(int),
1536                 .mode           = 0644,
1537                 .proc_handler   = &proc_dointvec,
1538         },
1539 #endif
1540 #ifdef CONFIG_MMU
1541 #ifdef CONFIG_FILE_LOCKING
1542         {
1543                 .ctl_name       = FS_LEASE_TIME,
1544                 .procname       = "lease-break-time",
1545                 .data           = &lease_break_time,
1546                 .maxlen         = sizeof(int),
1547                 .mode           = 0644,
1548                 .proc_handler   = &proc_dointvec,
1549         },
1550 #endif
1551 #ifdef CONFIG_AIO
1552         {
1553                 .procname       = "aio-nr",
1554                 .data           = &aio_nr,
1555                 .maxlen         = sizeof(aio_nr),
1556                 .mode           = 0444,
1557                 .proc_handler   = &proc_doulongvec_minmax,
1558         },
1559         {
1560                 .procname       = "aio-max-nr",
1561                 .data           = &aio_max_nr,
1562                 .maxlen         = sizeof(aio_max_nr),
1563                 .mode           = 0644,
1564                 .proc_handler   = &proc_doulongvec_minmax,
1565         },
1566 #endif /* CONFIG_AIO */
1567 #ifdef CONFIG_INOTIFY_USER
1568         {
1569                 .ctl_name       = FS_INOTIFY,
1570                 .procname       = "inotify",
1571                 .mode           = 0555,
1572                 .child          = inotify_table,
1573         },
1574 #endif  
1575 #ifdef CONFIG_EPOLL
1576         {
1577                 .procname       = "epoll",
1578                 .mode           = 0555,
1579                 .child          = epoll_table,
1580         },
1581 #endif
1582 #endif
1583         {
1584                 .ctl_name       = KERN_SETUID_DUMPABLE,
1585                 .procname       = "suid_dumpable",
1586                 .data           = &suid_dumpable,
1587                 .maxlen         = sizeof(int),
1588                 .mode           = 0644,
1589                 .proc_handler   = &proc_dointvec_minmax,
1590                 .strategy       = &sysctl_intvec,
1591                 .extra1         = &zero,
1592                 .extra2         = &two,
1593         },
1594 #if defined(CONFIG_BINFMT_MISC) || defined(CONFIG_BINFMT_MISC_MODULE)
1595         {
1596                 .ctl_name       = CTL_UNNUMBERED,
1597                 .procname       = "binfmt_misc",
1598                 .mode           = 0555,
1599                 .child          = binfmt_misc_table,
1600         },
1601 #endif
1602 /*
1603  * NOTE: do not add new entries to this table unless you have read
1604  * Documentation/sysctl/ctl_unnumbered.txt
1605  */
1606         { .ctl_name = 0 }
1607 };
1608
1609 static struct ctl_table debug_table[] = {
1610 #if defined(CONFIG_X86) || defined(CONFIG_PPC)
1611         {
1612                 .ctl_name       = CTL_UNNUMBERED,
1613                 .procname       = "exception-trace",
1614                 .data           = &show_unhandled_signals,
1615                 .maxlen         = sizeof(int),
1616                 .mode           = 0644,
1617                 .proc_handler   = proc_dointvec
1618         },
1619 #endif
1620         { .ctl_name = 0 }
1621 };
1622
1623 static struct ctl_table dev_table[] = {
1624         { .ctl_name = 0 }
1625 };
1626
1627 static DEFINE_SPINLOCK(sysctl_lock);
1628
1629 /* called under sysctl_lock */
1630 static int use_table(struct ctl_table_header *p)
1631 {
1632         if (unlikely(p->unregistering))
1633                 return 0;
1634         p->used++;
1635         return 1;
1636 }
1637
1638 /* called under sysctl_lock */
1639 static void unuse_table(struct ctl_table_header *p)
1640 {
1641         if (!--p->used)
1642                 if (unlikely(p->unregistering))
1643                         complete(p->unregistering);
1644 }
1645
1646 /* called under sysctl_lock, will reacquire if has to wait */
1647 static void start_unregistering(struct ctl_table_header *p)
1648 {
1649         /*
1650          * if p->used is 0, nobody will ever touch that entry again;
1651          * we'll eliminate all paths to it before dropping sysctl_lock
1652          */
1653         if (unlikely(p->used)) {
1654                 struct completion wait;
1655                 init_completion(&wait);
1656                 p->unregistering = &wait;
1657                 spin_unlock(&sysctl_lock);
1658                 wait_for_completion(&wait);
1659                 spin_lock(&sysctl_lock);
1660         } else {
1661                 /* anything non-NULL; we'll never dereference it */
1662                 p->unregistering = ERR_PTR(-EINVAL);
1663         }
1664         /*
1665          * do not remove from the list until nobody holds it; walking the
1666          * list in do_sysctl() relies on that.
1667          */
1668         list_del_init(&p->ctl_entry);
1669 }
1670
1671 void sysctl_head_get(struct ctl_table_header *head)
1672 {
1673         spin_lock(&sysctl_lock);
1674         head->count++;
1675         spin_unlock(&sysctl_lock);
1676 }
1677
1678 void sysctl_head_put(struct ctl_table_header *head)
1679 {
1680         spin_lock(&sysctl_lock);
1681         if (!--head->count)
1682                 kfree(head);
1683         spin_unlock(&sysctl_lock);
1684 }
1685
1686 struct ctl_table_header *sysctl_head_grab(struct ctl_table_header *head)
1687 {
1688         if (!head)
1689                 BUG();
1690         spin_lock(&sysctl_lock);
1691         if (!use_table(head))
1692                 head = ERR_PTR(-ENOENT);
1693         spin_unlock(&sysctl_lock);
1694         return head;
1695 }
1696
1697 void sysctl_head_finish(struct ctl_table_header *head)
1698 {
1699         if (!head)
1700                 return;
1701         spin_lock(&sysctl_lock);
1702         unuse_table(head);
1703         spin_unlock(&sysctl_lock);
1704 }
1705
1706 static struct ctl_table_set *
1707 lookup_header_set(struct ctl_table_root *root, struct nsproxy *namespaces)
1708 {
1709         struct ctl_table_set *set = &root->default_set;
1710         if (root->lookup)
1711                 set = root->lookup(root, namespaces);
1712         return set;
1713 }
1714
1715 static struct list_head *
1716 lookup_header_list(struct ctl_table_root *root, struct nsproxy *namespaces)
1717 {
1718         struct ctl_table_set *set = lookup_header_set(root, namespaces);
1719         return &set->list;
1720 }
1721
1722 struct ctl_table_header *__sysctl_head_next(struct nsproxy *namespaces,
1723                                             struct ctl_table_header *prev)
1724 {
1725         struct ctl_table_root *root;
1726         struct list_head *header_list;
1727         struct ctl_table_header *head;
1728         struct list_head *tmp;
1729
1730         spin_lock(&sysctl_lock);
1731         if (prev) {
1732                 head = prev;
1733                 tmp = &prev->ctl_entry;
1734                 unuse_table(prev);
1735                 goto next;
1736         }
1737         tmp = &root_table_header.ctl_entry;
1738         for (;;) {
1739                 head = list_entry(tmp, struct ctl_table_header, ctl_entry);
1740
1741                 if (!use_table(head))
1742                         goto next;
1743                 spin_unlock(&sysctl_lock);
1744                 return head;
1745         next:
1746                 root = head->root;
1747                 tmp = tmp->next;
1748                 header_list = lookup_header_list(root, namespaces);
1749                 if (tmp != header_list)
1750                         continue;
1751
1752                 do {
1753                         root = list_entry(root->root_list.next,
1754                                         struct ctl_table_root, root_list);
1755                         if (root == &sysctl_table_root)
1756                                 goto out;
1757                         header_list = lookup_header_list(root, namespaces);
1758                 } while (list_empty(header_list));
1759                 tmp = header_list->next;
1760         }
1761 out:
1762         spin_unlock(&sysctl_lock);
1763         return NULL;
1764 }
1765
1766 struct ctl_table_header *sysctl_head_next(struct ctl_table_header *prev)
1767 {
1768         return __sysctl_head_next(current->nsproxy, prev);
1769 }
1770
1771 void register_sysctl_root(struct ctl_table_root *root)
1772 {
1773         spin_lock(&sysctl_lock);
1774         list_add_tail(&root->root_list, &sysctl_table_root.root_list);
1775         spin_unlock(&sysctl_lock);
1776 }
1777
1778 #ifdef CONFIG_SYSCTL_SYSCALL
1779 /* Perform the actual read/write of a sysctl table entry. */
1780 static int do_sysctl_strategy(struct ctl_table_root *root,
1781                         struct ctl_table *table,
1782                         void __user *oldval, size_t __user *oldlenp,
1783                         void __user *newval, size_t newlen)
1784 {
1785         int op = 0, rc;
1786
1787         if (oldval)
1788                 op |= MAY_READ;
1789         if (newval)
1790                 op |= MAY_WRITE;
1791         if (sysctl_perm(root, table, op))
1792                 return -EPERM;
1793
1794         if (table->strategy) {
1795                 rc = table->strategy(table, oldval, oldlenp, newval, newlen);
1796                 if (rc < 0)
1797                         return rc;
1798                 if (rc > 0)
1799                         return 0;
1800         }
1801
1802         /* If there is no strategy routine, or if the strategy returns
1803          * zero, proceed with automatic r/w */
1804         if (table->data && table->maxlen) {
1805                 rc = sysctl_data(table, oldval, oldlenp, newval, newlen);
1806                 if (rc < 0)
1807                         return rc;
1808         }
1809         return 0;
1810 }
1811
1812 static int parse_table(int __user *name, int nlen,
1813                        void __user *oldval, size_t __user *oldlenp,
1814                        void __user *newval, size_t newlen,
1815                        struct ctl_table_root *root,
1816                        struct ctl_table *table)
1817 {
1818         int n;
1819 repeat:
1820         if (!nlen)
1821                 return -ENOTDIR;
1822         if (get_user(n, name))
1823                 return -EFAULT;
1824         for ( ; table->ctl_name || table->procname; table++) {
1825                 if (!table->ctl_name)
1826                         continue;
1827                 if (n == table->ctl_name) {
1828                         int error;
1829                         if (table->child) {
1830                                 if (sysctl_perm(root, table, MAY_EXEC))
1831                                         return -EPERM;
1832                                 name++;
1833                                 nlen--;
1834                                 table = table->child;
1835                                 goto repeat;
1836                         }
1837                         error = do_sysctl_strategy(root, table,
1838                                                    oldval, oldlenp,
1839                                                    newval, newlen);
1840                         return error;
1841                 }
1842         }
1843         return -ENOTDIR;
1844 }
1845
1846 int do_sysctl(int __user *name, int nlen, void __user *oldval, size_t __user *oldlenp,
1847                void __user *newval, size_t newlen)
1848 {
1849         struct ctl_table_header *head;
1850         int error = -ENOTDIR;
1851
1852         if (nlen <= 0 || nlen >= CTL_MAXNAME)
1853                 return -ENOTDIR;
1854         if (oldval) {
1855                 int old_len;
1856                 if (!oldlenp || get_user(old_len, oldlenp))
1857                         return -EFAULT;
1858         }
1859
1860         for (head = sysctl_head_next(NULL); head;
1861                         head = sysctl_head_next(head)) {
1862                 error = parse_table(name, nlen, oldval, oldlenp, 
1863                                         newval, newlen,
1864                                         head->root, head->ctl_table);
1865                 if (error != -ENOTDIR) {
1866                         sysctl_head_finish(head);
1867                         break;
1868                 }
1869         }
1870         return error;
1871 }
1872
1873 SYSCALL_DEFINE1(sysctl, struct __sysctl_args __user *, args)
1874 {
1875         struct __sysctl_args tmp;
1876         int error;
1877
1878         if (copy_from_user(&tmp, args, sizeof(tmp)))
1879                 return -EFAULT;
1880
1881         error = deprecated_sysctl_warning(&tmp);
1882         if (error)
1883                 goto out;
1884
1885         lock_kernel();
1886         error = do_sysctl(tmp.name, tmp.nlen, tmp.oldval, tmp.oldlenp,
1887                           tmp.newval, tmp.newlen);
1888         unlock_kernel();
1889 out:
1890         return error;
1891 }
1892 #endif /* CONFIG_SYSCTL_SYSCALL */
1893
1894 /*
1895  * sysctl_perm does NOT grant the superuser all rights automatically, because
1896  * some sysctl variables are readonly even to root.
1897  */
1898
1899 static int test_perm(int mode, int op)
1900 {
1901         if (!current_euid())
1902                 mode >>= 6;
1903         else if (in_egroup_p(0))
1904                 mode >>= 3;
1905         if ((op & ~mode & (MAY_READ|MAY_WRITE|MAY_EXEC)) == 0)
1906                 return 0;
1907         return -EACCES;
1908 }
1909
1910 int sysctl_perm(struct ctl_table_root *root, struct ctl_table *table, int op)
1911 {
1912         int error;
1913         int mode;
1914
1915         error = security_sysctl(table, op & (MAY_READ | MAY_WRITE | MAY_EXEC));
1916         if (error)
1917                 return error;
1918
1919         if (root->permissions)
1920                 mode = root->permissions(root, current->nsproxy, table);
1921         else
1922                 mode = table->mode;
1923
1924         return test_perm(mode, op);
1925 }
1926
1927 static void sysctl_set_parent(struct ctl_table *parent, struct ctl_table *table)
1928 {
1929         for (; table->ctl_name || table->procname; table++) {
1930                 table->parent = parent;
1931                 if (table->child)
1932                         sysctl_set_parent(table, table->child);
1933         }
1934 }
1935
1936 static __init int sysctl_init(void)
1937 {
1938         sysctl_set_parent(NULL, root_table);
1939 #ifdef CONFIG_SYSCTL_SYSCALL_CHECK
1940         {
1941                 int err;
1942                 err = sysctl_check_table(current->nsproxy, root_table);
1943         }
1944 #endif
1945         return 0;
1946 }
1947
1948 core_initcall(sysctl_init);
1949
1950 static struct ctl_table *is_branch_in(struct ctl_table *branch,
1951                                       struct ctl_table *table)
1952 {
1953         struct ctl_table *p;
1954         const char *s = branch->procname;
1955
1956         /* branch should have named subdirectory as its first element */
1957         if (!s || !branch->child)
1958                 return NULL;
1959
1960         /* ... and nothing else */
1961         if (branch[1].procname || branch[1].ctl_name)
1962                 return NULL;
1963
1964         /* table should contain subdirectory with the same name */
1965         for (p = table; p->procname || p->ctl_name; p++) {
1966                 if (!p->child)
1967                         continue;
1968                 if (p->procname && strcmp(p->procname, s) == 0)
1969                         return p;
1970         }
1971         return NULL;
1972 }
1973
1974 /* see if attaching q to p would be an improvement */
1975 static void try_attach(struct ctl_table_header *p, struct ctl_table_header *q)
1976 {
1977         struct ctl_table *to = p->ctl_table, *by = q->ctl_table;
1978         struct ctl_table *next;
1979         int is_better = 0;
1980         int not_in_parent = !p->attached_by;
1981
1982         while ((next = is_branch_in(by, to)) != NULL) {
1983                 if (by == q->attached_by)
1984                         is_better = 1;
1985                 if (to == p->attached_by)
1986                         not_in_parent = 1;
1987                 by = by->child;
1988                 to = next->child;
1989         }
1990
1991         if (is_better && not_in_parent) {
1992                 q->attached_by = by;
1993                 q->attached_to = to;
1994                 q->parent = p;
1995         }
1996 }
1997
1998 /**
1999  * __register_sysctl_paths - register a sysctl hierarchy
2000  * @root: List of sysctl headers to register on
2001  * @namespaces: Data to compute which lists of sysctl entries are visible
2002  * @path: The path to the directory the sysctl table is in.
2003  * @table: the top-level table structure
2004  *
2005  * Register a sysctl table hierarchy. @table should be a filled in ctl_table
2006  * array. A completely 0 filled entry terminates the table.
2007  *
2008  * The members of the &struct ctl_table structure are used as follows:
2009  *
2010  * ctl_name - This is the numeric sysctl value used by sysctl(2). The number
2011  *            must be unique within that level of sysctl
2012  *
2013  * procname - the name of the sysctl file under /proc/sys. Set to %NULL to not
2014  *            enter a sysctl file
2015  *
2016  * data - a pointer to data for use by proc_handler
2017  *
2018  * maxlen - the maximum size in bytes of the data
2019  *
2020  * mode - the file permissions for the /proc/sys file, and for sysctl(2)
2021  *
2022  * child - a pointer to the child sysctl table if this entry is a directory, or
2023  *         %NULL.
2024  *
2025  * proc_handler - the text handler routine (described below)
2026  *
2027  * strategy - the strategy routine (described below)
2028  *
2029  * de - for internal use by the sysctl routines
2030  *
2031  * extra1, extra2 - extra pointers usable by the proc handler routines
2032  *
2033  * Leaf nodes in the sysctl tree will be represented by a single file
2034  * under /proc; non-leaf nodes will be represented by directories.
2035  *
2036  * sysctl(2) can automatically manage read and write requests through
2037  * the sysctl table.  The data and maxlen fields of the ctl_table
2038  * struct enable minimal validation of the values being written to be
2039  * performed, and the mode field allows minimal authentication.
2040  *
2041  * More sophisticated management can be enabled by the provision of a
2042  * strategy routine with the table entry.  This will be called before
2043  * any automatic read or write of the data is performed.
2044  *
2045  * The strategy routine may return
2046  *
2047  * < 0 - Error occurred (error is passed to user process)
2048  *
2049  * 0   - OK - proceed with automatic read or write.
2050  *
2051  * > 0 - OK - read or write has been done by the strategy routine, so
2052  *       return immediately.
2053  *
2054  * There must be a proc_handler routine for any terminal nodes
2055  * mirrored under /proc/sys (non-terminals are handled by a built-in
2056  * directory handler).  Several default handlers are available to
2057  * cover common cases -
2058  *
2059  * proc_dostring(), proc_dointvec(), proc_dointvec_jiffies(),
2060  * proc_dointvec_userhz_jiffies(), proc_dointvec_minmax(), 
2061  * proc_doulongvec_ms_jiffies_minmax(), proc_doulongvec_minmax()
2062  *
2063  * It is the handler's job to read the input buffer from user memory
2064  * and process it. The handler should return 0 on success.
2065  *
2066  * This routine returns %NULL on a failure to register, and a pointer
2067  * to the table header on success.
2068  */
2069 struct ctl_table_header *__register_sysctl_paths(
2070         struct ctl_table_root *root,
2071         struct nsproxy *namespaces,
2072         const struct ctl_path *path, struct ctl_table *table)
2073 {
2074         struct ctl_table_header *header;
2075         struct ctl_table *new, **prevp;
2076         unsigned int n, npath;
2077         struct ctl_table_set *set;
2078
2079         /* Count the path components */
2080         for (npath = 0; path[npath].ctl_name || path[npath].procname; ++npath)
2081                 ;
2082
2083         /*
2084          * For each path component, allocate a 2-element ctl_table array.
2085          * The first array element will be filled with the sysctl entry
2086          * for this, the second will be the sentinel (ctl_name == 0).
2087          *
2088          * We allocate everything in one go so that we don't have to
2089          * worry about freeing additional memory in unregister_sysctl_table.
2090          */
2091         header = kzalloc(sizeof(struct ctl_table_header) +
2092                          (2 * npath * sizeof(struct ctl_table)), GFP_KERNEL);
2093         if (!header)
2094                 return NULL;
2095
2096         new = (struct ctl_table *) (header + 1);
2097
2098         /* Now connect the dots */
2099         prevp = &header->ctl_table;
2100         for (n = 0; n < npath; ++n, ++path) {
2101                 /* Copy the procname */
2102                 new->procname = path->procname;
2103                 new->ctl_name = path->ctl_name;
2104                 new->mode     = 0555;
2105
2106                 *prevp = new;
2107                 prevp = &new->child;
2108
2109                 new += 2;
2110         }
2111         *prevp = table;
2112         header->ctl_table_arg = table;
2113
2114         INIT_LIST_HEAD(&header->ctl_entry);
2115         header->used = 0;
2116         header->unregistering = NULL;
2117         header->root = root;
2118         sysctl_set_parent(NULL, header->ctl_table);
2119         header->count = 1;
2120 #ifdef CONFIG_SYSCTL_SYSCALL_CHECK
2121         if (sysctl_check_table(namespaces, header->ctl_table)) {
2122                 kfree(header);
2123                 return NULL;
2124         }
2125 #endif
2126         spin_lock(&sysctl_lock);
2127         header->set = lookup_header_set(root, namespaces);
2128         header->attached_by = header->ctl_table;
2129         header->attached_to = root_table;
2130         header->parent = &root_table_header;
2131         for (set = header->set; set; set = set->parent) {
2132                 struct ctl_table_header *p;
2133                 list_for_each_entry(p, &set->list, ctl_entry) {
2134                         if (p->unregistering)
2135                                 continue;
2136                         try_attach(p, header);
2137                 }
2138         }
2139         header->parent->count++;
2140         list_add_tail(&header->ctl_entry, &header->set->list);
2141         spin_unlock(&sysctl_lock);
2142
2143         return header;
2144 }
2145
2146 /**
2147  * register_sysctl_table_path - register a sysctl table hierarchy
2148  * @path: The path to the directory the sysctl table is in.
2149  * @table: the top-level table structure
2150  *
2151  * Register a sysctl table hierarchy. @table should be a filled in ctl_table
2152  * array. A completely 0 filled entry terminates the table.
2153  *
2154  * See __register_sysctl_paths for more details.
2155  */
2156 struct ctl_table_header *register_sysctl_paths(const struct ctl_path *path,
2157                                                 struct ctl_table *table)
2158 {
2159         return __register_sysctl_paths(&sysctl_table_root, current->nsproxy,
2160                                         path, table);
2161 }
2162
2163 /**
2164  * register_sysctl_table - register a sysctl table hierarchy
2165  * @table: the top-level table structure
2166  *
2167  * Register a sysctl table hierarchy. @table should be a filled in ctl_table
2168  * array. A completely 0 filled entry terminates the table.
2169  *
2170  * See register_sysctl_paths for more details.
2171  */
2172 struct ctl_table_header *register_sysctl_table(struct ctl_table *table)
2173 {
2174         static const struct ctl_path null_path[] = { {} };
2175
2176         return register_sysctl_paths(null_path, table);
2177 }
2178
2179 /**
2180  * unregister_sysctl_table - unregister a sysctl table hierarchy
2181  * @header: the header returned from register_sysctl_table
2182  *
2183  * Unregisters the sysctl table and all children. proc entries may not
2184  * actually be removed until they are no longer used by anyone.
2185  */
2186 void unregister_sysctl_table(struct ctl_table_header * header)
2187 {
2188         might_sleep();
2189
2190         if (header == NULL)
2191                 return;
2192
2193         spin_lock(&sysctl_lock);
2194         start_unregistering(header);
2195         if (!--header->parent->count) {
2196                 WARN_ON(1);
2197                 kfree(header->parent);
2198         }
2199         if (!--header->count)
2200                 kfree(header);
2201         spin_unlock(&sysctl_lock);
2202 }
2203
2204 int sysctl_is_seen(struct ctl_table_header *p)
2205 {
2206         struct ctl_table_set *set = p->set;
2207         int res;
2208         spin_lock(&sysctl_lock);
2209         if (p->unregistering)
2210                 res = 0;
2211         else if (!set->is_seen)
2212                 res = 1;
2213         else
2214                 res = set->is_seen(set);
2215         spin_unlock(&sysctl_lock);
2216         return res;
2217 }
2218
2219 void setup_sysctl_set(struct ctl_table_set *p,
2220         struct ctl_table_set *parent,
2221         int (*is_seen)(struct ctl_table_set *))
2222 {
2223         INIT_LIST_HEAD(&p->list);
2224         p->parent = parent ? parent : &sysctl_table_root.default_set;
2225         p->is_seen = is_seen;
2226 }
2227
2228 #else /* !CONFIG_SYSCTL */
2229 struct ctl_table_header *register_sysctl_table(struct ctl_table * table)
2230 {
2231         return NULL;
2232 }
2233
2234 struct ctl_table_header *register_sysctl_paths(const struct ctl_path *path,
2235                                                     struct ctl_table *table)
2236 {
2237         return NULL;
2238 }
2239
2240 void unregister_sysctl_table(struct ctl_table_header * table)
2241 {
2242 }
2243
2244 void setup_sysctl_set(struct ctl_table_set *p,
2245         struct ctl_table_set *parent,
2246         int (*is_seen)(struct ctl_table_set *))
2247 {
2248 }
2249
2250 void sysctl_head_put(struct ctl_table_header *head)
2251 {
2252 }
2253
2254 #endif /* CONFIG_SYSCTL */
2255
2256 /*
2257  * /proc/sys support
2258  */
2259
2260 #ifdef CONFIG_PROC_SYSCTL
2261
2262 static int _proc_do_string(void* data, int maxlen, int write,
2263                            void __user *buffer,
2264                            size_t *lenp, loff_t *ppos)
2265 {
2266         size_t len;
2267         char __user *p;
2268         char c;
2269
2270         if (!data || !maxlen || !*lenp) {
2271                 *lenp = 0;
2272                 return 0;
2273         }
2274
2275         if (write) {
2276                 len = 0;
2277                 p = buffer;
2278                 while (len < *lenp) {
2279                         if (get_user(c, p++))
2280                                 return -EFAULT;
2281                         if (c == 0 || c == '\n')
2282                                 break;
2283                         len++;
2284                 }
2285                 if (len >= maxlen)
2286                         len = maxlen-1;
2287                 if(copy_from_user(data, buffer, len))
2288                         return -EFAULT;
2289                 ((char *) data)[len] = 0;
2290                 *ppos += *lenp;
2291         } else {
2292                 len = strlen(data);
2293                 if (len > maxlen)
2294                         len = maxlen;
2295
2296                 if (*ppos > len) {
2297                         *lenp = 0;
2298                         return 0;
2299                 }
2300
2301                 data += *ppos;
2302                 len  -= *ppos;
2303
2304                 if (len > *lenp)
2305                         len = *lenp;
2306                 if (len)
2307                         if(copy_to_user(buffer, data, len))
2308                                 return -EFAULT;
2309                 if (len < *lenp) {
2310                         if(put_user('\n', ((char __user *) buffer) + len))
2311                                 return -EFAULT;
2312                         len++;
2313                 }
2314                 *lenp = len;
2315                 *ppos += len;
2316         }
2317         return 0;
2318 }
2319
2320 /**
2321  * proc_dostring - read a string sysctl
2322  * @table: the sysctl table
2323  * @write: %TRUE if this is a write to the sysctl file
2324  * @buffer: the user buffer
2325  * @lenp: the size of the user buffer
2326  * @ppos: file position
2327  *
2328  * Reads/writes a string from/to the user buffer. If the kernel
2329  * buffer provided is not large enough to hold the string, the
2330  * string is truncated. The copied string is %NULL-terminated.
2331  * If the string is being read by the user process, it is copied
2332  * and a newline '\n' is added. It is truncated if the buffer is
2333  * not large enough.
2334  *
2335  * Returns 0 on success.
2336  */
2337 int proc_dostring(struct ctl_table *table, int write,
2338                   void __user *buffer, size_t *lenp, loff_t *ppos)
2339 {
2340         return _proc_do_string(table->data, table->maxlen, write,
2341                                buffer, lenp, ppos);
2342 }
2343
2344
2345 static int do_proc_dointvec_conv(int *negp, unsigned long *lvalp,
2346                                  int *valp,
2347                                  int write, void *data)
2348 {
2349         if (write) {
2350                 *valp = *negp ? -*lvalp : *lvalp;
2351         } else {
2352                 int val = *valp;
2353                 if (val < 0) {
2354                         *negp = -1;
2355                         *lvalp = (unsigned long)-val;
2356                 } else {
2357                         *negp = 0;
2358                         *lvalp = (unsigned long)val;
2359                 }
2360         }
2361         return 0;
2362 }
2363
2364 static int __do_proc_dointvec(void *tbl_data, struct ctl_table *table,
2365                   int write, void __user *buffer,
2366                   size_t *lenp, loff_t *ppos,
2367                   int (*conv)(int *negp, unsigned long *lvalp, int *valp,
2368                               int write, void *data),
2369                   void *data)
2370 {
2371 #define TMPBUFLEN 21
2372         int *i, vleft, first = 1, neg;
2373         unsigned long lval;
2374         size_t left, len;
2375         
2376         char buf[TMPBUFLEN], *p;
2377         char __user *s = buffer;
2378         
2379         if (!tbl_data || !table->maxlen || !*lenp ||
2380             (*ppos && !write)) {
2381                 *lenp = 0;
2382                 return 0;
2383         }
2384         
2385         i = (int *) tbl_data;
2386         vleft = table->maxlen / sizeof(*i);
2387         left = *lenp;
2388
2389         if (!conv)
2390                 conv = do_proc_dointvec_conv;
2391
2392         for (; left && vleft--; i++, first=0) {
2393                 if (write) {
2394                         while (left) {
2395                                 char c;
2396                                 if (get_user(c, s))
2397                                         return -EFAULT;
2398                                 if (!isspace(c))
2399                                         break;
2400                                 left--;
2401                                 s++;
2402                         }
2403                         if (!left)
2404                                 break;
2405                         neg = 0;
2406                         len = left;
2407                         if (len > sizeof(buf) - 1)
2408                                 len = sizeof(buf) - 1;
2409                         if (copy_from_user(buf, s, len))
2410                                 return -EFAULT;
2411                         buf[len] = 0;
2412                         p = buf;
2413                         if (*p == '-' && left > 1) {
2414                                 neg = 1;
2415                                 p++;
2416                         }
2417                         if (*p < '0' || *p > '9')
2418                                 break;
2419
2420                         lval = simple_strtoul(p, &p, 0);
2421
2422                         len = p-buf;
2423                         if ((len < left) && *p && !isspace(*p))
2424                                 break;
2425                         s += len;
2426                         left -= len;
2427
2428                         if (conv(&neg, &lval, i, 1, data))
2429                                 break;
2430                 } else {
2431                         p = buf;
2432                         if (!first)
2433                                 *p++ = '\t';
2434         
2435                         if (conv(&neg, &lval, i, 0, data))
2436                                 break;
2437
2438                         sprintf(p, "%s%lu", neg ? "-" : "", lval);
2439                         len = strlen(buf);
2440                         if (len > left)
2441                                 len = left;
2442                         if(copy_to_user(s, buf, len))
2443                                 return -EFAULT;
2444                         left -= len;
2445                         s += len;
2446                 }
2447         }
2448
2449         if (!write && !first && left) {
2450                 if(put_user('\n', s))
2451                         return -EFAULT;
2452                 left--, s++;
2453         }
2454         if (write) {
2455                 while (left) {
2456                         char c;
2457                         if (get_user(c, s++))
2458                                 return -EFAULT;
2459                         if (!isspace(c))
2460                                 break;
2461                         left--;
2462                 }
2463         }
2464         if (write && first)
2465                 return -EINVAL;
2466         *lenp -= left;
2467         *ppos += *lenp;
2468         return 0;
2469 #undef TMPBUFLEN
2470 }
2471
2472 static int do_proc_dointvec(struct ctl_table *table, int write,
2473                   void __user *buffer, size_t *lenp, loff_t *ppos,
2474                   int (*conv)(int *negp, unsigned long *lvalp, int *valp,
2475                               int write, void *data),
2476                   void *data)
2477 {
2478         return __do_proc_dointvec(table->data, table, write,
2479                         buffer, lenp, ppos, conv, data);
2480 }
2481
2482 /**
2483  * proc_dointvec - read a vector of integers
2484  * @table: the sysctl table
2485  * @write: %TRUE if this is a write to the sysctl file
2486  * @buffer: the user buffer
2487  * @lenp: the size of the user buffer
2488  * @ppos: file position
2489  *
2490  * Reads/writes up to table->maxlen/sizeof(unsigned int) integer
2491  * values from/to the user buffer, treated as an ASCII string. 
2492  *
2493  * Returns 0 on success.
2494  */
2495 int proc_dointvec(struct ctl_table *table, int write,
2496                      void __user *buffer, size_t *lenp, loff_t *ppos)
2497 {
2498     return do_proc_dointvec(table,write,buffer,lenp,ppos,
2499                             NULL,NULL);
2500 }
2501
2502 /*
2503  * Taint values can only be increased
2504  * This means we can safely use a temporary.
2505  */
2506 static int proc_taint(struct ctl_table *table, int write,
2507                                void __user *buffer, size_t *lenp, loff_t *ppos)
2508 {
2509         struct ctl_table t;
2510         unsigned long tmptaint = get_taint();
2511         int err;
2512
2513         if (write && !capable(CAP_SYS_ADMIN))
2514                 return -EPERM;
2515
2516         t = *table;
2517         t.data = &tmptaint;
2518         err = proc_doulongvec_minmax(&t, write, buffer, lenp, ppos);
2519         if (err < 0)
2520                 return err;
2521
2522         if (write) {
2523                 /*
2524                  * Poor man's atomic or. Not worth adding a primitive
2525                  * to everyone's atomic.h for this
2526                  */
2527                 int i;
2528                 for (i = 0; i < BITS_PER_LONG && tmptaint >> i; i++) {
2529                         if ((tmptaint >> i) & 1)
2530                                 add_taint(i);
2531                 }
2532         }
2533
2534         return err;
2535 }
2536
2537 struct do_proc_dointvec_minmax_conv_param {
2538         int *min;
2539         int *max;
2540 };
2541
2542 static int do_proc_dointvec_minmax_conv(int *negp, unsigned long *lvalp, 
2543                                         int *valp, 
2544                                         int write, void *data)
2545 {
2546         struct do_proc_dointvec_minmax_conv_param *param = data;
2547         if (write) {
2548                 int val = *negp ? -*lvalp : *lvalp;
2549                 if ((param->min && *param->min > val) ||
2550                     (param->max && *param->max < val))
2551                         return -EINVAL;
2552                 *valp = val;
2553         } else {
2554                 int val = *valp;
2555                 if (val < 0) {
2556                         *negp = -1;
2557                         *lvalp = (unsigned long)-val;
2558                 } else {
2559                         *negp = 0;
2560                         *lvalp = (unsigned long)val;
2561                 }
2562         }
2563         return 0;
2564 }
2565
2566 /**
2567  * proc_dointvec_minmax - read a vector of integers with min/max values
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  *
2574  * Reads/writes up to table->maxlen/sizeof(unsigned int) integer
2575  * values from/to the user buffer, treated as an ASCII string.
2576  *
2577  * This routine will ensure the values are within the range specified by
2578  * table->extra1 (min) and table->extra2 (max).
2579  *
2580  * Returns 0 on success.
2581  */
2582 int proc_dointvec_minmax(struct ctl_table *table, int write,
2583                   void __user *buffer, size_t *lenp, loff_t *ppos)
2584 {
2585         struct do_proc_dointvec_minmax_conv_param param = {
2586                 .min = (int *) table->extra1,
2587                 .max = (int *) table->extra2,
2588         };
2589         return do_proc_dointvec(table, write, buffer, lenp, ppos,
2590                                 do_proc_dointvec_minmax_conv, &param);
2591 }
2592
2593 static int __do_proc_doulongvec_minmax(void *data, struct ctl_table *table, int write,
2594                                      void __user *buffer,
2595                                      size_t *lenp, loff_t *ppos,
2596                                      unsigned long convmul,
2597                                      unsigned long convdiv)
2598 {
2599 #define TMPBUFLEN 21
2600         unsigned long *i, *min, *max, val;
2601         int vleft, first=1, neg;
2602         size_t len, left;
2603         char buf[TMPBUFLEN], *p;
2604         char __user *s = buffer;
2605         
2606         if (!data || !table->maxlen || !*lenp ||
2607             (*ppos && !write)) {
2608                 *lenp = 0;
2609                 return 0;
2610         }
2611         
2612         i = (unsigned long *) data;
2613         min = (unsigned long *) table->extra1;
2614         max = (unsigned long *) table->extra2;
2615         vleft = table->maxlen / sizeof(unsigned long);
2616         left = *lenp;
2617         
2618         for (; left && vleft--; i++, min++, max++, first=0) {
2619                 if (write) {
2620                         while (left) {
2621                                 char c;
2622                                 if (get_user(c, s))
2623                                         return -EFAULT;
2624                                 if (!isspace(c))
2625                                         break;
2626                                 left--;
2627                                 s++;
2628                         }
2629                         if (!left)
2630                                 break;
2631                         neg = 0;
2632                         len = left;
2633                         if (len > TMPBUFLEN-1)
2634                                 len = TMPBUFLEN-1;
2635                         if (copy_from_user(buf, s, len))
2636                                 return -EFAULT;
2637                         buf[len] = 0;
2638                         p = buf;
2639                         if (*p == '-' && left > 1) {
2640                                 neg = 1;
2641                                 p++;
2642                         }
2643                         if (*p < '0' || *p > '9')
2644                                 break;
2645                         val = simple_strtoul(p, &p, 0) * convmul / convdiv ;
2646                         len = p-buf;
2647                         if ((len < left) && *p && !isspace(*p))
2648                                 break;
2649                         if (neg)
2650                                 val = -val;
2651                         s += len;
2652                         left -= len;
2653
2654                         if(neg)
2655                                 continue;
2656                         if ((min && val < *min) || (max && val > *max))
2657                                 continue;
2658                         *i = val;
2659                 } else {
2660                         p = buf;
2661                         if (!first)
2662                                 *p++ = '\t';
2663                         sprintf(p, "%lu", convdiv * (*i) / convmul);
2664                         len = strlen(buf);
2665                         if (len > left)
2666                                 len = left;
2667                         if(copy_to_user(s, buf, len))
2668                                 return -EFAULT;
2669                         left -= len;
2670                         s += len;
2671                 }
2672         }
2673
2674         if (!write && !first && left) {
2675                 if(put_user('\n', s))
2676                         return -EFAULT;
2677                 left--, s++;
2678         }
2679         if (write) {
2680                 while (left) {
2681                         char c;
2682                         if (get_user(c, s++))
2683                                 return -EFAULT;
2684                         if (!isspace(c))
2685                                 break;
2686                         left--;
2687                 }
2688         }
2689         if (write && first)
2690                 return -EINVAL;
2691         *lenp -= left;
2692         *ppos += *lenp;
2693         return 0;
2694 #undef TMPBUFLEN
2695 }
2696
2697 static int do_proc_doulongvec_minmax(struct ctl_table *table, int write,
2698                                      void __user *buffer,
2699                                      size_t *lenp, loff_t *ppos,
2700                                      unsigned long convmul,
2701                                      unsigned long convdiv)
2702 {
2703         return __do_proc_doulongvec_minmax(table->data, table, write,
2704                         buffer, lenp, ppos, convmul, convdiv);
2705 }
2706
2707 /**
2708  * proc_doulongvec_minmax - read a vector of long integers with min/max values
2709  * @table: the sysctl table
2710  * @write: %TRUE if this is a write to the sysctl file
2711  * @buffer: the user buffer
2712  * @lenp: the size of the user buffer
2713  * @ppos: file position
2714  *
2715  * Reads/writes up to table->maxlen/sizeof(unsigned long) unsigned long
2716  * values from/to the user buffer, treated as an ASCII string.
2717  *
2718  * This routine will ensure the values are within the range specified by
2719  * table->extra1 (min) and table->extra2 (max).
2720  *
2721  * Returns 0 on success.
2722  */
2723 int proc_doulongvec_minmax(struct ctl_table *table, int write,
2724                            void __user *buffer, size_t *lenp, loff_t *ppos)
2725 {
2726     return do_proc_doulongvec_minmax(table, write, buffer, lenp, ppos, 1l, 1l);
2727 }
2728
2729 /**
2730  * proc_doulongvec_ms_jiffies_minmax - read a vector of millisecond values with min/max values
2731  * @table: the sysctl table
2732  * @write: %TRUE if this is a write to the sysctl file
2733  * @buffer: the user buffer
2734  * @lenp: the size of the user buffer
2735  * @ppos: file position
2736  *
2737  * Reads/writes up to table->maxlen/sizeof(unsigned long) unsigned long
2738  * values from/to the user buffer, treated as an ASCII string. The values
2739  * are treated as milliseconds, and converted to jiffies when they are stored.
2740  *
2741  * This routine will ensure the values are within the range specified by
2742  * table->extra1 (min) and table->extra2 (max).
2743  *
2744  * Returns 0 on success.
2745  */
2746 int proc_doulongvec_ms_jiffies_minmax(struct ctl_table *table, int write,
2747                                       void __user *buffer,
2748                                       size_t *lenp, loff_t *ppos)
2749 {
2750     return do_proc_doulongvec_minmax(table, write, buffer,
2751                                      lenp, ppos, HZ, 1000l);
2752 }
2753
2754
2755 static int do_proc_dointvec_jiffies_conv(int *negp, unsigned long *lvalp,
2756                                          int *valp,
2757                                          int write, void *data)
2758 {
2759         if (write) {
2760                 if (*lvalp > LONG_MAX / HZ)
2761                         return 1;
2762                 *valp = *negp ? -(*lvalp*HZ) : (*lvalp*HZ);
2763         } else {
2764                 int val = *valp;
2765                 unsigned long lval;
2766                 if (val < 0) {
2767                         *negp = -1;
2768                         lval = (unsigned long)-val;
2769                 } else {
2770                         *negp = 0;
2771                         lval = (unsigned long)val;
2772                 }
2773                 *lvalp = lval / HZ;
2774         }
2775         return 0;
2776 }
2777
2778 static int do_proc_dointvec_userhz_jiffies_conv(int *negp, unsigned long *lvalp,
2779                                                 int *valp,
2780                                                 int write, void *data)
2781 {
2782         if (write) {
2783                 if (USER_HZ < HZ && *lvalp > (LONG_MAX / HZ) * USER_HZ)
2784                         return 1;
2785                 *valp = clock_t_to_jiffies(*negp ? -*lvalp : *lvalp);
2786         } else {
2787                 int val = *valp;
2788                 unsigned long lval;
2789                 if (val < 0) {
2790                         *negp = -1;
2791                         lval = (unsigned long)-val;
2792                 } else {
2793                         *negp = 0;
2794                         lval = (unsigned long)val;
2795                 }
2796                 *lvalp = jiffies_to_clock_t(lval);
2797         }
2798         return 0;
2799 }
2800
2801 static int do_proc_dointvec_ms_jiffies_conv(int *negp, unsigned long *lvalp,
2802                                             int *valp,
2803                                             int write, void *data)
2804 {
2805         if (write) {
2806                 *valp = msecs_to_jiffies(*negp ? -*lvalp : *lvalp);
2807         } else {
2808                 int val = *valp;
2809                 unsigned long lval;
2810                 if (val < 0) {
2811                         *negp = -1;
2812                         lval = (unsigned long)-val;
2813                 } else {
2814                         *negp = 0;
2815                         lval = (unsigned long)val;
2816                 }
2817                 *lvalp = jiffies_to_msecs(lval);
2818         }
2819         return 0;
2820 }
2821
2822 /**
2823  * proc_dointvec_jiffies - read a vector of integers as seconds
2824  * @table: the sysctl table
2825  * @write: %TRUE if this is a write to the sysctl file
2826  * @buffer: the user buffer
2827  * @lenp: the size of the user buffer
2828  * @ppos: file position
2829  *
2830  * Reads/writes up to table->maxlen/sizeof(unsigned int) integer
2831  * values from/to the user buffer, treated as an ASCII string. 
2832  * The values read are assumed to be in seconds, and are converted into
2833  * jiffies.
2834  *
2835  * Returns 0 on success.
2836  */
2837 int proc_dointvec_jiffies(struct ctl_table *table, int write,
2838                           void __user *buffer, size_t *lenp, loff_t *ppos)
2839 {
2840     return do_proc_dointvec(table,write,buffer,lenp,ppos,
2841                             do_proc_dointvec_jiffies_conv,NULL);
2842 }
2843
2844 /**
2845  * proc_dointvec_userhz_jiffies - read a vector of integers as 1/USER_HZ seconds
2846  * @table: the sysctl table
2847  * @write: %TRUE if this is a write to the sysctl file
2848  * @buffer: the user buffer
2849  * @lenp: the size of the user buffer
2850  * @ppos: pointer to the file position
2851  *
2852  * Reads/writes up to table->maxlen/sizeof(unsigned int) integer
2853  * values from/to the user buffer, treated as an ASCII string. 
2854  * The values read are assumed to be in 1/USER_HZ seconds, and 
2855  * are converted into jiffies.
2856  *
2857  * Returns 0 on success.
2858  */
2859 int proc_dointvec_userhz_jiffies(struct ctl_table *table, int write,
2860                                  void __user *buffer, size_t *lenp, loff_t *ppos)
2861 {
2862     return do_proc_dointvec(table,write,buffer,lenp,ppos,
2863                             do_proc_dointvec_userhz_jiffies_conv,NULL);
2864 }
2865
2866 /**
2867  * proc_dointvec_ms_jiffies - read a vector of integers as 1 milliseconds
2868  * @table: the sysctl table
2869  * @write: %TRUE if this is a write to the sysctl file
2870  * @buffer: the user buffer
2871  * @lenp: the size of the user buffer
2872  * @ppos: file position
2873  * @ppos: the current position in the file
2874  *
2875  * Reads/writes up to table->maxlen/sizeof(unsigned int) integer
2876  * values from/to the user buffer, treated as an ASCII string. 
2877  * The values read are assumed to be in 1/1000 seconds, and 
2878  * are converted into jiffies.
2879  *
2880  * Returns 0 on success.
2881  */
2882 int proc_dointvec_ms_jiffies(struct ctl_table *table, int write,
2883                              void __user *buffer, size_t *lenp, loff_t *ppos)
2884 {
2885         return do_proc_dointvec(table, write, buffer, lenp, ppos,
2886                                 do_proc_dointvec_ms_jiffies_conv, NULL);
2887 }
2888
2889 static int proc_do_cad_pid(struct ctl_table *table, int write,
2890                            void __user *buffer, size_t *lenp, loff_t *ppos)
2891 {
2892         struct pid *new_pid;
2893         pid_t tmp;
2894         int r;
2895
2896         tmp = pid_vnr(cad_pid);
2897
2898         r = __do_proc_dointvec(&tmp, table, write, buffer,
2899                                lenp, ppos, NULL, NULL);
2900         if (r || !write)
2901                 return r;
2902
2903         new_pid = find_get_pid(tmp);
2904         if (!new_pid)
2905                 return -ESRCH;
2906
2907         put_pid(xchg(&cad_pid, new_pid));
2908         return 0;
2909 }
2910
2911 #else /* CONFIG_PROC_FS */
2912
2913 int proc_dostring(struct ctl_table *table, int write,
2914                   void __user *buffer, size_t *lenp, loff_t *ppos)
2915 {
2916         return -ENOSYS;
2917 }
2918
2919 int proc_dointvec(struct ctl_table *table, int write,
2920                   void __user *buffer, size_t *lenp, loff_t *ppos)
2921 {
2922         return -ENOSYS;
2923 }
2924
2925 int proc_dointvec_minmax(struct ctl_table *table, int write,
2926                     void __user *buffer, size_t *lenp, loff_t *ppos)
2927 {
2928         return -ENOSYS;
2929 }
2930
2931 int proc_dointvec_jiffies(struct ctl_table *table, int write,
2932                     void __user *buffer, size_t *lenp, loff_t *ppos)
2933 {
2934         return -ENOSYS;
2935 }
2936
2937 int proc_dointvec_userhz_jiffies(struct ctl_table *table, int write,
2938                     void __user *buffer, size_t *lenp, loff_t *ppos)
2939 {
2940         return -ENOSYS;
2941 }
2942
2943 int proc_dointvec_ms_jiffies(struct ctl_table *table, int write,
2944                              void __user *buffer, size_t *lenp, loff_t *ppos)
2945 {
2946         return -ENOSYS;
2947 }
2948
2949 int proc_doulongvec_minmax(struct ctl_table *table, int write,
2950                     void __user *buffer, size_t *lenp, loff_t *ppos)
2951 {
2952         return -ENOSYS;
2953 }
2954
2955 int proc_doulongvec_ms_jiffies_minmax(struct ctl_table *table, int write,
2956                                       void __user *buffer,
2957                                       size_t *lenp, loff_t *ppos)
2958 {
2959     return -ENOSYS;
2960 }
2961
2962
2963 #endif /* CONFIG_PROC_FS */
2964
2965
2966 #ifdef CONFIG_SYSCTL_SYSCALL
2967 /*
2968  * General sysctl support routines 
2969  */
2970
2971 /* The generic sysctl data routine (used if no strategy routine supplied) */
2972 int sysctl_data(struct ctl_table *table,
2973                 void __user *oldval, size_t __user *oldlenp,
2974                 void __user *newval, size_t newlen)
2975 {
2976         size_t len;
2977
2978         /* Get out of I don't have a variable */
2979         if (!table->data || !table->maxlen)
2980                 return -ENOTDIR;
2981
2982         if (oldval && oldlenp) {
2983                 if (get_user(len, oldlenp))
2984                         return -EFAULT;
2985                 if (len) {
2986                         if (len > table->maxlen)
2987                                 len = table->maxlen;
2988                         if (copy_to_user(oldval, table->data, len))
2989                                 return -EFAULT;
2990                         if (put_user(len, oldlenp))
2991                                 return -EFAULT;
2992                 }
2993         }
2994
2995         if (newval && newlen) {
2996                 if (newlen > table->maxlen)
2997                         newlen = table->maxlen;
2998
2999                 if (copy_from_user(table->data, newval, newlen))
3000                         return -EFAULT;
3001         }
3002         return 1;
3003 }
3004
3005 /* The generic string strategy routine: */
3006 int sysctl_string(struct ctl_table *table,
3007                   void __user *oldval, size_t __user *oldlenp,
3008                   void __user *newval, size_t newlen)
3009 {
3010         if (!table->data || !table->maxlen) 
3011                 return -ENOTDIR;
3012         
3013         if (oldval && oldlenp) {
3014                 size_t bufsize;
3015                 if (get_user(bufsize, oldlenp))
3016                         return -EFAULT;
3017                 if (bufsize) {
3018                         size_t len = strlen(table->data), copied;
3019
3020                         /* This shouldn't trigger for a well-formed sysctl */
3021                         if (len > table->maxlen)
3022                                 len = table->maxlen;
3023
3024                         /* Copy up to a max of bufsize-1 bytes of the string */
3025                         copied = (len >= bufsize) ? bufsize - 1 : len;
3026
3027                         if (copy_to_user(oldval, table->data, copied) ||
3028                             put_user(0, (char __user *)(oldval + copied)))
3029                                 return -EFAULT;
3030                         if (put_user(len, oldlenp))
3031                                 return -EFAULT;
3032                 }
3033         }
3034         if (newval && newlen) {
3035                 size_t len = newlen;
3036                 if (len > table->maxlen)
3037                         len = table->maxlen;
3038                 if(copy_from_user(table->data, newval, len))
3039                         return -EFAULT;
3040                 if (len == table->maxlen)
3041                         len--;
3042                 ((char *) table->data)[len] = 0;
3043         }
3044         return 1;
3045 }
3046
3047 /*
3048  * This function makes sure that all of the integers in the vector
3049  * are between the minimum and maximum values given in the arrays
3050  * table->extra1 and table->extra2, respectively.
3051  */
3052 int sysctl_intvec(struct ctl_table *table,
3053                 void __user *oldval, size_t __user *oldlenp,
3054                 void __user *newval, size_t newlen)
3055 {
3056
3057         if (newval && newlen) {
3058                 int __user *vec = (int __user *) newval;
3059                 int *min = (int *) table->extra1;
3060                 int *max = (int *) table->extra2;
3061                 size_t length;
3062                 int i;
3063
3064                 if (newlen % sizeof(int) != 0)
3065                         return -EINVAL;
3066
3067                 if (!table->extra1 && !table->extra2)
3068                         return 0;
3069
3070                 if (newlen > table->maxlen)
3071                         newlen = table->maxlen;
3072                 length = newlen / sizeof(int);
3073
3074                 for (i = 0; i < length; i++) {
3075                         int value;
3076                         if (get_user(value, vec + i))
3077                                 return -EFAULT;
3078                         if (min && value < min[i])
3079                                 return -EINVAL;
3080                         if (max && value > max[i])
3081                                 return -EINVAL;
3082                 }
3083         }
3084         return 0;
3085 }
3086
3087 /* Strategy function to convert jiffies to seconds */ 
3088 int sysctl_jiffies(struct ctl_table *table,
3089                 void __user *oldval, size_t __user *oldlenp,
3090                 void __user *newval, size_t newlen)
3091 {
3092         if (oldval && oldlenp) {
3093                 size_t olen;
3094
3095                 if (get_user(olen, oldlenp))
3096                         return -EFAULT;
3097                 if (olen) {
3098                         int val;
3099
3100                         if (olen < sizeof(int))
3101                                 return -EINVAL;
3102
3103                         val = *(int *)(table->data) / HZ;
3104                         if (put_user(val, (int __user *)oldval))
3105                                 return -EFAULT;
3106                         if (put_user(sizeof(int), oldlenp))
3107                                 return -EFAULT;
3108                 }
3109         }
3110         if (newval && newlen) { 
3111                 int new;
3112                 if (newlen != sizeof(int))
3113                         return -EINVAL; 
3114                 if (get_user(new, (int __user *)newval))
3115                         return -EFAULT;
3116                 *(int *)(table->data) = new*HZ; 
3117         }
3118         return 1;
3119 }
3120
3121 /* Strategy function to convert jiffies to seconds */ 
3122 int sysctl_ms_jiffies(struct ctl_table *table,
3123                 void __user *oldval, size_t __user *oldlenp,
3124                 void __user *newval, size_t newlen)
3125 {
3126         if (oldval && oldlenp) {
3127                 size_t olen;
3128
3129                 if (get_user(olen, oldlenp))
3130                         return -EFAULT;
3131                 if (olen) {
3132                         int val;
3133
3134                         if (olen < sizeof(int))
3135                                 return -EINVAL;
3136
3137                         val = jiffies_to_msecs(*(int *)(table->data));
3138                         if (put_user(val, (int __user *)oldval))
3139                                 return -EFAULT;
3140                         if (put_user(sizeof(int), oldlenp))
3141                                 return -EFAULT;
3142                 }
3143         }
3144         if (newval && newlen) { 
3145                 int new;
3146                 if (newlen != sizeof(int))
3147                         return -EINVAL; 
3148                 if (get_user(new, (int __user *)newval))
3149                         return -EFAULT;
3150                 *(int *)(table->data) = msecs_to_jiffies(new);
3151         }
3152         return 1;
3153 }
3154
3155
3156
3157 #else /* CONFIG_SYSCTL_SYSCALL */
3158
3159
3160 SYSCALL_DEFINE1(sysctl, struct __sysctl_args __user *, args)
3161 {
3162         struct __sysctl_args tmp;
3163         int error;
3164
3165         if (copy_from_user(&tmp, args, sizeof(tmp)))
3166                 return -EFAULT;
3167
3168         error = deprecated_sysctl_warning(&tmp);
3169
3170         /* If no error reading the parameters then just -ENOSYS ... */
3171         if (!error)
3172                 error = -ENOSYS;
3173
3174         return error;
3175 }
3176
3177 int sysctl_data(struct ctl_table *table,
3178                   void __user *oldval, size_t __user *oldlenp,
3179                   void __user *newval, size_t newlen)
3180 {
3181         return -ENOSYS;
3182 }
3183
3184 int sysctl_string(struct ctl_table *table,
3185                   void __user *oldval, size_t __user *oldlenp,
3186                   void __user *newval, size_t newlen)
3187 {
3188         return -ENOSYS;
3189 }
3190
3191 int sysctl_intvec(struct ctl_table *table,
3192                 void __user *oldval, size_t __user *oldlenp,
3193                 void __user *newval, size_t newlen)
3194 {
3195         return -ENOSYS;
3196 }
3197
3198 int sysctl_jiffies(struct ctl_table *table,
3199                 void __user *oldval, size_t __user *oldlenp,
3200                 void __user *newval, size_t newlen)
3201 {
3202         return -ENOSYS;
3203 }
3204
3205 int sysctl_ms_jiffies(struct ctl_table *table,
3206                 void __user *oldval, size_t __user *oldlenp,
3207                 void __user *newval, size_t newlen)
3208 {
3209         return -ENOSYS;
3210 }
3211
3212 #endif /* CONFIG_SYSCTL_SYSCALL */
3213
3214 static int deprecated_sysctl_warning(struct __sysctl_args *args)
3215 {
3216         static int msg_count;
3217         int name[CTL_MAXNAME];
3218         int i;
3219
3220         /* Check args->nlen. */
3221         if (args->nlen < 0 || args->nlen > CTL_MAXNAME)
3222                 return -ENOTDIR;
3223
3224         /* Read in the sysctl name for better debug message logging */
3225         for (i = 0; i < args->nlen; i++)
3226                 if (get_user(name[i], args->name + i))
3227                         return -EFAULT;
3228
3229         /* Ignore accesses to kernel.version */
3230         if ((args->nlen == 2) && (name[0] == CTL_KERN) && (name[1] == KERN_VERSION))
3231                 return 0;
3232
3233         if (msg_count < 5) {
3234                 msg_count++;
3235                 printk(KERN_INFO
3236                         "warning: process `%s' used the deprecated sysctl "
3237                         "system call with ", current->comm);
3238                 for (i = 0; i < args->nlen; i++)
3239                         printk("%d.", name[i]);
3240                 printk("\n");
3241         }
3242         return 0;
3243 }
3244
3245 /*
3246  * No sense putting this after each symbol definition, twice,
3247  * exception granted :-)
3248  */
3249 EXPORT_SYMBOL(proc_dointvec);
3250 EXPORT_SYMBOL(proc_dointvec_jiffies);
3251 EXPORT_SYMBOL(proc_dointvec_minmax);
3252 EXPORT_SYMBOL(proc_dointvec_userhz_jiffies);
3253 EXPORT_SYMBOL(proc_dointvec_ms_jiffies);
3254 EXPORT_SYMBOL(proc_dostring);
3255 EXPORT_SYMBOL(proc_doulongvec_minmax);
3256 EXPORT_SYMBOL(proc_doulongvec_ms_jiffies_minmax);
3257 EXPORT_SYMBOL(register_sysctl_table);
3258 EXPORT_SYMBOL(register_sysctl_paths);
3259 EXPORT_SYMBOL(sysctl_intvec);
3260 EXPORT_SYMBOL(sysctl_jiffies);
3261 EXPORT_SYMBOL(sysctl_ms_jiffies);
3262 EXPORT_SYMBOL(sysctl_string);
3263 EXPORT_SYMBOL(sysctl_data);
3264 EXPORT_SYMBOL(unregister_sysctl_table);