6bb59f707402c6ea456a9f58322e67f5b611f64d
[linux-3.10.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/utsname.h>
30 #include <linux/kmemcheck.h>
31 #include <linux/smp_lock.h>
32 #include <linux/fs.h>
33 #include <linux/init.h>
34 #include <linux/kernel.h>
35 #include <linux/kobject.h>
36 #include <linux/net.h>
37 #include <linux/sysrq.h>
38 #include <linux/highuid.h>
39 #include <linux/writeback.h>
40 #include <linux/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_counter.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 int pid_max;
81 extern int min_free_kbytes;
82 extern int pid_max_min, pid_max_max;
83 extern int sysctl_drop_caches;
84 extern int percpu_pagelist_fraction;
85 extern int compat_log;
86 extern int latencytop_enabled;
87 extern int sysctl_nr_open_min, sysctl_nr_open_max;
88 #ifndef CONFIG_MMU
89 extern int sysctl_nr_trim_pages;
90 #endif
91 #ifdef CONFIG_RCU_TORTURE_TEST
92 extern int rcutorture_runnable;
93 #endif /* #ifdef CONFIG_RCU_TORTURE_TEST */
94 extern int blk_iopoll_enabled;
95
96 /* Constants used for minimum and  maximum */
97 #ifdef CONFIG_DETECT_SOFTLOCKUP
98 static int sixty = 60;
99 static int neg_one = -1;
100 #endif
101
102 static int zero;
103 static int __maybe_unused one = 1;
104 static int __maybe_unused two = 2;
105 static unsigned long one_ul = 1;
106 static int one_hundred = 100;
107
108 /* this is needed for the proc_doulongvec_minmax of vm_dirty_bytes */
109 static unsigned long dirty_bytes_min = 2 * PAGE_SIZE;
110
111 /* this is needed for the proc_dointvec_minmax for [fs_]overflow UID and GID */
112 static int maxolduid = 65535;
113 static int minolduid;
114 static int min_percpu_pagelist_fract = 8;
115
116 static int ngroups_max = NGROUPS_MAX;
117
118 #ifdef CONFIG_MODULES
119 extern char modprobe_path[];
120 extern int modules_disabled;
121 #endif
122 #ifdef CONFIG_CHR_DEV_SG
123 extern int sg_big_buff;
124 #endif
125
126 #ifdef CONFIG_SPARC
127 #include <asm/system.h>
128 #endif
129
130 #ifdef CONFIG_SPARC64
131 extern int sysctl_tsb_ratio;
132 #endif
133
134 #ifdef __hppa__
135 extern int pwrsw_enabled;
136 extern int unaligned_enabled;
137 #endif
138
139 #ifdef CONFIG_S390
140 #ifdef CONFIG_MATHEMU
141 extern int sysctl_ieee_emulation_warnings;
142 #endif
143 extern int sysctl_userprocess_debug;
144 extern int spin_retry;
145 #endif
146
147 #ifdef CONFIG_BSD_PROCESS_ACCT
148 extern int acct_parm[];
149 #endif
150
151 #ifdef CONFIG_IA64
152 extern int no_unaligned_warning;
153 extern int unaligned_dump_stack;
154 #endif
155
156 #ifdef CONFIG_RT_MUTEXES
157 extern int max_lock_depth;
158 #endif
159
160 #ifdef CONFIG_PROC_SYSCTL
161 static int proc_do_cad_pid(struct ctl_table *table, int write, struct file *filp,
162                   void __user *buffer, size_t *lenp, loff_t *ppos);
163 static int proc_taint(struct ctl_table *table, int write, struct file *filp,
164                                void __user *buffer, size_t *lenp, loff_t *ppos);
165 #endif
166
167 static struct ctl_table root_table[];
168 static struct ctl_table_root sysctl_table_root;
169 static struct ctl_table_header root_table_header = {
170         .count = 1,
171         .ctl_table = root_table,
172         .ctl_entry = LIST_HEAD_INIT(sysctl_table_root.default_set.list),
173         .root = &sysctl_table_root,
174         .set = &sysctl_table_root.default_set,
175 };
176 static struct ctl_table_root sysctl_table_root = {
177         .root_list = LIST_HEAD_INIT(sysctl_table_root.root_list),
178         .default_set.list = LIST_HEAD_INIT(root_table_header.ctl_entry),
179 };
180
181 static struct ctl_table kern_table[];
182 static struct ctl_table vm_table[];
183 static struct ctl_table fs_table[];
184 static struct ctl_table debug_table[];
185 static struct ctl_table dev_table[];
186 extern struct ctl_table random_table[];
187 #ifdef CONFIG_INOTIFY_USER
188 extern struct ctl_table inotify_table[];
189 #endif
190 #ifdef CONFIG_EPOLL
191 extern struct ctl_table epoll_table[];
192 #endif
193
194 #ifdef HAVE_ARCH_PICK_MMAP_LAYOUT
195 int sysctl_legacy_va_layout;
196 #endif
197
198 extern int prove_locking;
199 extern int lock_stat;
200
201 /* The default sysctl tables: */
202
203 static struct ctl_table root_table[] = {
204         {
205                 .ctl_name       = CTL_KERN,
206                 .procname       = "kernel",
207                 .mode           = 0555,
208                 .child          = kern_table,
209         },
210         {
211                 .ctl_name       = CTL_VM,
212                 .procname       = "vm",
213                 .mode           = 0555,
214                 .child          = vm_table,
215         },
216         {
217                 .ctl_name       = CTL_FS,
218                 .procname       = "fs",
219                 .mode           = 0555,
220                 .child          = fs_table,
221         },
222         {
223                 .ctl_name       = CTL_DEBUG,
224                 .procname       = "debug",
225                 .mode           = 0555,
226                 .child          = debug_table,
227         },
228         {
229                 .ctl_name       = CTL_DEV,
230                 .procname       = "dev",
231                 .mode           = 0555,
232                 .child          = dev_table,
233         },
234 /*
235  * NOTE: do not add new entries to this table unless you have read
236  * Documentation/sysctl/ctl_unnumbered.txt
237  */
238         { .ctl_name = 0 }
239 };
240
241 #ifdef CONFIG_SCHED_DEBUG
242 static int min_sched_granularity_ns = 100000;           /* 100 usecs */
243 static int max_sched_granularity_ns = NSEC_PER_SEC;     /* 1 second */
244 static int min_wakeup_granularity_ns;                   /* 0 usecs */
245 static int max_wakeup_granularity_ns = NSEC_PER_SEC;    /* 1 second */
246 #endif
247
248 static struct ctl_table kern_table[] = {
249         {
250                 .ctl_name       = CTL_UNNUMBERED,
251                 .procname       = "sched_child_runs_first",
252                 .data           = &sysctl_sched_child_runs_first,
253                 .maxlen         = sizeof(unsigned int),
254                 .mode           = 0644,
255                 .proc_handler   = &proc_dointvec,
256         },
257 #ifdef CONFIG_SCHED_DEBUG
258         {
259                 .ctl_name       = CTL_UNNUMBERED,
260                 .procname       = "sched_min_granularity_ns",
261                 .data           = &sysctl_sched_min_granularity,
262                 .maxlen         = sizeof(unsigned int),
263                 .mode           = 0644,
264                 .proc_handler   = &sched_nr_latency_handler,
265                 .strategy       = &sysctl_intvec,
266                 .extra1         = &min_sched_granularity_ns,
267                 .extra2         = &max_sched_granularity_ns,
268         },
269         {
270                 .ctl_name       = CTL_UNNUMBERED,
271                 .procname       = "sched_latency_ns",
272                 .data           = &sysctl_sched_latency,
273                 .maxlen         = sizeof(unsigned int),
274                 .mode           = 0644,
275                 .proc_handler   = &sched_nr_latency_handler,
276                 .strategy       = &sysctl_intvec,
277                 .extra1         = &min_sched_granularity_ns,
278                 .extra2         = &max_sched_granularity_ns,
279         },
280         {
281                 .ctl_name       = CTL_UNNUMBERED,
282                 .procname       = "sched_wakeup_granularity_ns",
283                 .data           = &sysctl_sched_wakeup_granularity,
284                 .maxlen         = sizeof(unsigned int),
285                 .mode           = 0644,
286                 .proc_handler   = &proc_dointvec_minmax,
287                 .strategy       = &sysctl_intvec,
288                 .extra1         = &min_wakeup_granularity_ns,
289                 .extra2         = &max_wakeup_granularity_ns,
290         },
291         {
292                 .ctl_name       = CTL_UNNUMBERED,
293                 .procname       = "sched_shares_ratelimit",
294                 .data           = &sysctl_sched_shares_ratelimit,
295                 .maxlen         = sizeof(unsigned int),
296                 .mode           = 0644,
297                 .proc_handler   = &proc_dointvec,
298         },
299         {
300                 .ctl_name       = CTL_UNNUMBERED,
301                 .procname       = "sched_shares_thresh",
302                 .data           = &sysctl_sched_shares_thresh,
303                 .maxlen         = sizeof(unsigned int),
304                 .mode           = 0644,
305                 .proc_handler   = &proc_dointvec_minmax,
306                 .strategy       = &sysctl_intvec,
307                 .extra1         = &zero,
308         },
309         {
310                 .ctl_name       = CTL_UNNUMBERED,
311                 .procname       = "sched_features",
312                 .data           = &sysctl_sched_features,
313                 .maxlen         = sizeof(unsigned int),
314                 .mode           = 0644,
315                 .proc_handler   = &proc_dointvec,
316         },
317         {
318                 .ctl_name       = CTL_UNNUMBERED,
319                 .procname       = "sched_migration_cost",
320                 .data           = &sysctl_sched_migration_cost,
321                 .maxlen         = sizeof(unsigned int),
322                 .mode           = 0644,
323                 .proc_handler   = &proc_dointvec,
324         },
325         {
326                 .ctl_name       = CTL_UNNUMBERED,
327                 .procname       = "sched_nr_migrate",
328                 .data           = &sysctl_sched_nr_migrate,
329                 .maxlen         = sizeof(unsigned int),
330                 .mode           = 0644,
331                 .proc_handler   = &proc_dointvec,
332         },
333         {
334                 .ctl_name       = CTL_UNNUMBERED,
335                 .procname       = "sched_time_avg",
336                 .data           = &sysctl_sched_time_avg,
337                 .maxlen         = sizeof(unsigned int),
338                 .mode           = 0644,
339                 .proc_handler   = &proc_dointvec,
340         },
341         {
342                 .ctl_name       = CTL_UNNUMBERED,
343                 .procname       = "timer_migration",
344                 .data           = &sysctl_timer_migration,
345                 .maxlen         = sizeof(unsigned int),
346                 .mode           = 0644,
347                 .proc_handler   = &proc_dointvec_minmax,
348                 .strategy       = &sysctl_intvec,
349                 .extra1         = &zero,
350                 .extra2         = &one,
351         },
352 #endif
353         {
354                 .ctl_name       = CTL_UNNUMBERED,
355                 .procname       = "sched_rt_period_us",
356                 .data           = &sysctl_sched_rt_period,
357                 .maxlen         = sizeof(unsigned int),
358                 .mode           = 0644,
359                 .proc_handler   = &sched_rt_handler,
360         },
361         {
362                 .ctl_name       = CTL_UNNUMBERED,
363                 .procname       = "sched_rt_runtime_us",
364                 .data           = &sysctl_sched_rt_runtime,
365                 .maxlen         = sizeof(int),
366                 .mode           = 0644,
367                 .proc_handler   = &sched_rt_handler,
368         },
369         {
370                 .ctl_name       = CTL_UNNUMBERED,
371                 .procname       = "sched_compat_yield",
372                 .data           = &sysctl_sched_compat_yield,
373                 .maxlen         = sizeof(unsigned int),
374                 .mode           = 0644,
375                 .proc_handler   = &proc_dointvec,
376         },
377 #ifdef CONFIG_PROVE_LOCKING
378         {
379                 .ctl_name       = CTL_UNNUMBERED,
380                 .procname       = "prove_locking",
381                 .data           = &prove_locking,
382                 .maxlen         = sizeof(int),
383                 .mode           = 0644,
384                 .proc_handler   = &proc_dointvec,
385         },
386 #endif
387 #ifdef CONFIG_LOCK_STAT
388         {
389                 .ctl_name       = CTL_UNNUMBERED,
390                 .procname       = "lock_stat",
391                 .data           = &lock_stat,
392                 .maxlen         = sizeof(int),
393                 .mode           = 0644,
394                 .proc_handler   = &proc_dointvec,
395         },
396 #endif
397         {
398                 .ctl_name       = KERN_PANIC,
399                 .procname       = "panic",
400                 .data           = &panic_timeout,
401                 .maxlen         = sizeof(int),
402                 .mode           = 0644,
403                 .proc_handler   = &proc_dointvec,
404         },
405         {
406                 .ctl_name       = KERN_CORE_USES_PID,
407                 .procname       = "core_uses_pid",
408                 .data           = &core_uses_pid,
409                 .maxlen         = sizeof(int),
410                 .mode           = 0644,
411                 .proc_handler   = &proc_dointvec,
412         },
413         {
414                 .ctl_name       = KERN_CORE_PATTERN,
415                 .procname       = "core_pattern",
416                 .data           = core_pattern,
417                 .maxlen         = CORENAME_MAX_SIZE,
418                 .mode           = 0644,
419                 .proc_handler   = &proc_dostring,
420                 .strategy       = &sysctl_string,
421         },
422 #ifdef CONFIG_PROC_SYSCTL
423         {
424                 .procname       = "tainted",
425                 .maxlen         = sizeof(long),
426                 .mode           = 0644,
427                 .proc_handler   = &proc_taint,
428         },
429 #endif
430 #ifdef CONFIG_LATENCYTOP
431         {
432                 .procname       = "latencytop",
433                 .data           = &latencytop_enabled,
434                 .maxlen         = sizeof(int),
435                 .mode           = 0644,
436                 .proc_handler   = &proc_dointvec,
437         },
438 #endif
439 #ifdef CONFIG_BLK_DEV_INITRD
440         {
441                 .ctl_name       = KERN_REALROOTDEV,
442                 .procname       = "real-root-dev",
443                 .data           = &real_root_dev,
444                 .maxlen         = sizeof(int),
445                 .mode           = 0644,
446                 .proc_handler   = &proc_dointvec,
447         },
448 #endif
449         {
450                 .ctl_name       = CTL_UNNUMBERED,
451                 .procname       = "print-fatal-signals",
452                 .data           = &print_fatal_signals,
453                 .maxlen         = sizeof(int),
454                 .mode           = 0644,
455                 .proc_handler   = &proc_dointvec,
456         },
457 #ifdef CONFIG_SPARC
458         {
459                 .ctl_name       = KERN_SPARC_REBOOT,
460                 .procname       = "reboot-cmd",
461                 .data           = reboot_command,
462                 .maxlen         = 256,
463                 .mode           = 0644,
464                 .proc_handler   = &proc_dostring,
465                 .strategy       = &sysctl_string,
466         },
467         {
468                 .ctl_name       = KERN_SPARC_STOP_A,
469                 .procname       = "stop-a",
470                 .data           = &stop_a_enabled,
471                 .maxlen         = sizeof (int),
472                 .mode           = 0644,
473                 .proc_handler   = &proc_dointvec,
474         },
475         {
476                 .ctl_name       = KERN_SPARC_SCONS_PWROFF,
477                 .procname       = "scons-poweroff",
478                 .data           = &scons_pwroff,
479                 .maxlen         = sizeof (int),
480                 .mode           = 0644,
481                 .proc_handler   = &proc_dointvec,
482         },
483 #endif
484 #ifdef CONFIG_SPARC64
485         {
486                 .ctl_name       = CTL_UNNUMBERED,
487                 .procname       = "tsb-ratio",
488                 .data           = &sysctl_tsb_ratio,
489                 .maxlen         = sizeof (int),
490                 .mode           = 0644,
491                 .proc_handler   = &proc_dointvec,
492         },
493 #endif
494 #ifdef __hppa__
495         {
496                 .ctl_name       = KERN_HPPA_PWRSW,
497                 .procname       = "soft-power",
498                 .data           = &pwrsw_enabled,
499                 .maxlen         = sizeof (int),
500                 .mode           = 0644,
501                 .proc_handler   = &proc_dointvec,
502         },
503         {
504                 .ctl_name       = KERN_HPPA_UNALIGNED,
505                 .procname       = "unaligned-trap",
506                 .data           = &unaligned_enabled,
507                 .maxlen         = sizeof (int),
508                 .mode           = 0644,
509                 .proc_handler   = &proc_dointvec,
510         },
511 #endif
512         {
513                 .ctl_name       = KERN_CTLALTDEL,
514                 .procname       = "ctrl-alt-del",
515                 .data           = &C_A_D,
516                 .maxlen         = sizeof(int),
517                 .mode           = 0644,
518                 .proc_handler   = &proc_dointvec,
519         },
520 #ifdef CONFIG_FUNCTION_TRACER
521         {
522                 .ctl_name       = CTL_UNNUMBERED,
523                 .procname       = "ftrace_enabled",
524                 .data           = &ftrace_enabled,
525                 .maxlen         = sizeof(int),
526                 .mode           = 0644,
527                 .proc_handler   = &ftrace_enable_sysctl,
528         },
529 #endif
530 #ifdef CONFIG_STACK_TRACER
531         {
532                 .ctl_name       = CTL_UNNUMBERED,
533                 .procname       = "stack_tracer_enabled",
534                 .data           = &stack_tracer_enabled,
535                 .maxlen         = sizeof(int),
536                 .mode           = 0644,
537                 .proc_handler   = &stack_trace_sysctl,
538         },
539 #endif
540 #ifdef CONFIG_TRACING
541         {
542                 .ctl_name       = CTL_UNNUMBERED,
543                 .procname       = "ftrace_dump_on_oops",
544                 .data           = &ftrace_dump_on_oops,
545                 .maxlen         = sizeof(int),
546                 .mode           = 0644,
547                 .proc_handler   = &proc_dointvec,
548         },
549 #endif
550 #ifdef CONFIG_MODULES
551         {
552                 .ctl_name       = KERN_MODPROBE,
553                 .procname       = "modprobe",
554                 .data           = &modprobe_path,
555                 .maxlen         = KMOD_PATH_LEN,
556                 .mode           = 0644,
557                 .proc_handler   = &proc_dostring,
558                 .strategy       = &sysctl_string,
559         },
560         {
561                 .ctl_name       = CTL_UNNUMBERED,
562                 .procname       = "modules_disabled",
563                 .data           = &modules_disabled,
564                 .maxlen         = sizeof(int),
565                 .mode           = 0644,
566                 /* only handle a transition from default "0" to "1" */
567                 .proc_handler   = &proc_dointvec_minmax,
568                 .extra1         = &one,
569                 .extra2         = &one,
570         },
571 #endif
572 #if defined(CONFIG_HOTPLUG) && defined(CONFIG_NET)
573         {
574                 .ctl_name       = KERN_HOTPLUG,
575                 .procname       = "hotplug",
576                 .data           = &uevent_helper,
577                 .maxlen         = UEVENT_HELPER_PATH_LEN,
578                 .mode           = 0644,
579                 .proc_handler   = &proc_dostring,
580                 .strategy       = &sysctl_string,
581         },
582 #endif
583 #ifdef CONFIG_CHR_DEV_SG
584         {
585                 .ctl_name       = KERN_SG_BIG_BUFF,
586                 .procname       = "sg-big-buff",
587                 .data           = &sg_big_buff,
588                 .maxlen         = sizeof (int),
589                 .mode           = 0444,
590                 .proc_handler   = &proc_dointvec,
591         },
592 #endif
593 #ifdef CONFIG_BSD_PROCESS_ACCT
594         {
595                 .ctl_name       = KERN_ACCT,
596                 .procname       = "acct",
597                 .data           = &acct_parm,
598                 .maxlen         = 3*sizeof(int),
599                 .mode           = 0644,
600                 .proc_handler   = &proc_dointvec,
601         },
602 #endif
603 #ifdef CONFIG_MAGIC_SYSRQ
604         {
605                 .ctl_name       = KERN_SYSRQ,
606                 .procname       = "sysrq",
607                 .data           = &__sysrq_enabled,
608                 .maxlen         = sizeof (int),
609                 .mode           = 0644,
610                 .proc_handler   = &proc_dointvec,
611         },
612 #endif
613 #ifdef CONFIG_PROC_SYSCTL
614         {
615                 .procname       = "cad_pid",
616                 .data           = NULL,
617                 .maxlen         = sizeof (int),
618                 .mode           = 0600,
619                 .proc_handler   = &proc_do_cad_pid,
620         },
621 #endif
622         {
623                 .ctl_name       = KERN_MAX_THREADS,
624                 .procname       = "threads-max",
625                 .data           = &max_threads,
626                 .maxlen         = sizeof(int),
627                 .mode           = 0644,
628                 .proc_handler   = &proc_dointvec,
629         },
630         {
631                 .ctl_name       = KERN_RANDOM,
632                 .procname       = "random",
633                 .mode           = 0555,
634                 .child          = random_table,
635         },
636         {
637                 .ctl_name       = KERN_OVERFLOWUID,
638                 .procname       = "overflowuid",
639                 .data           = &overflowuid,
640                 .maxlen         = sizeof(int),
641                 .mode           = 0644,
642                 .proc_handler   = &proc_dointvec_minmax,
643                 .strategy       = &sysctl_intvec,
644                 .extra1         = &minolduid,
645                 .extra2         = &maxolduid,
646         },
647         {
648                 .ctl_name       = KERN_OVERFLOWGID,
649                 .procname       = "overflowgid",
650                 .data           = &overflowgid,
651                 .maxlen         = sizeof(int),
652                 .mode           = 0644,
653                 .proc_handler   = &proc_dointvec_minmax,
654                 .strategy       = &sysctl_intvec,
655                 .extra1         = &minolduid,
656                 .extra2         = &maxolduid,
657         },
658 #ifdef CONFIG_S390
659 #ifdef CONFIG_MATHEMU
660         {
661                 .ctl_name       = KERN_IEEE_EMULATION_WARNINGS,
662                 .procname       = "ieee_emulation_warnings",
663                 .data           = &sysctl_ieee_emulation_warnings,
664                 .maxlen         = sizeof(int),
665                 .mode           = 0644,
666                 .proc_handler   = &proc_dointvec,
667         },
668 #endif
669         {
670                 .ctl_name       = KERN_S390_USER_DEBUG_LOGGING,
671                 .procname       = "userprocess_debug",
672                 .data           = &sysctl_userprocess_debug,
673                 .maxlen         = sizeof(int),
674                 .mode           = 0644,
675                 .proc_handler   = &proc_dointvec,
676         },
677 #endif
678         {
679                 .ctl_name       = KERN_PIDMAX,
680                 .procname       = "pid_max",
681                 .data           = &pid_max,
682                 .maxlen         = sizeof (int),
683                 .mode           = 0644,
684                 .proc_handler   = &proc_dointvec_minmax,
685                 .strategy       = sysctl_intvec,
686                 .extra1         = &pid_max_min,
687                 .extra2         = &pid_max_max,
688         },
689         {
690                 .ctl_name       = KERN_PANIC_ON_OOPS,
691                 .procname       = "panic_on_oops",
692                 .data           = &panic_on_oops,
693                 .maxlen         = sizeof(int),
694                 .mode           = 0644,
695                 .proc_handler   = &proc_dointvec,
696         },
697 #if defined CONFIG_PRINTK
698         {
699                 .ctl_name       = KERN_PRINTK,
700                 .procname       = "printk",
701                 .data           = &console_loglevel,
702                 .maxlen         = 4*sizeof(int),
703                 .mode           = 0644,
704                 .proc_handler   = &proc_dointvec,
705         },
706         {
707                 .ctl_name       = KERN_PRINTK_RATELIMIT,
708                 .procname       = "printk_ratelimit",
709                 .data           = &printk_ratelimit_state.interval,
710                 .maxlen         = sizeof(int),
711                 .mode           = 0644,
712                 .proc_handler   = &proc_dointvec_jiffies,
713                 .strategy       = &sysctl_jiffies,
714         },
715         {
716                 .ctl_name       = KERN_PRINTK_RATELIMIT_BURST,
717                 .procname       = "printk_ratelimit_burst",
718                 .data           = &printk_ratelimit_state.burst,
719                 .maxlen         = sizeof(int),
720                 .mode           = 0644,
721                 .proc_handler   = &proc_dointvec,
722         },
723 #endif
724         {
725                 .ctl_name       = KERN_NGROUPS_MAX,
726                 .procname       = "ngroups_max",
727                 .data           = &ngroups_max,
728                 .maxlen         = sizeof (int),
729                 .mode           = 0444,
730                 .proc_handler   = &proc_dointvec,
731         },
732 #if defined(CONFIG_X86_LOCAL_APIC) && defined(CONFIG_X86)
733         {
734                 .ctl_name       = KERN_UNKNOWN_NMI_PANIC,
735                 .procname       = "unknown_nmi_panic",
736                 .data           = &unknown_nmi_panic,
737                 .maxlen         = sizeof (int),
738                 .mode           = 0644,
739                 .proc_handler   = &proc_dointvec,
740         },
741         {
742                 .procname       = "nmi_watchdog",
743                 .data           = &nmi_watchdog_enabled,
744                 .maxlen         = sizeof (int),
745                 .mode           = 0644,
746                 .proc_handler   = &proc_nmi_enabled,
747         },
748 #endif
749 #if defined(CONFIG_X86)
750         {
751                 .ctl_name       = KERN_PANIC_ON_NMI,
752                 .procname       = "panic_on_unrecovered_nmi",
753                 .data           = &panic_on_unrecovered_nmi,
754                 .maxlen         = sizeof(int),
755                 .mode           = 0644,
756                 .proc_handler   = &proc_dointvec,
757         },
758         {
759                 .ctl_name       = CTL_UNNUMBERED,
760                 .procname       = "panic_on_io_nmi",
761                 .data           = &panic_on_io_nmi,
762                 .maxlen         = sizeof(int),
763                 .mode           = 0644,
764                 .proc_handler   = &proc_dointvec,
765         },
766         {
767                 .ctl_name       = KERN_BOOTLOADER_TYPE,
768                 .procname       = "bootloader_type",
769                 .data           = &bootloader_type,
770                 .maxlen         = sizeof (int),
771                 .mode           = 0444,
772                 .proc_handler   = &proc_dointvec,
773         },
774         {
775                 .ctl_name       = CTL_UNNUMBERED,
776                 .procname       = "bootloader_version",
777                 .data           = &bootloader_version,
778                 .maxlen         = sizeof (int),
779                 .mode           = 0444,
780                 .proc_handler   = &proc_dointvec,
781         },
782         {
783                 .ctl_name       = CTL_UNNUMBERED,
784                 .procname       = "kstack_depth_to_print",
785                 .data           = &kstack_depth_to_print,
786                 .maxlen         = sizeof(int),
787                 .mode           = 0644,
788                 .proc_handler   = &proc_dointvec,
789         },
790         {
791                 .ctl_name       = CTL_UNNUMBERED,
792                 .procname       = "io_delay_type",
793                 .data           = &io_delay_type,
794                 .maxlen         = sizeof(int),
795                 .mode           = 0644,
796                 .proc_handler   = &proc_dointvec,
797         },
798 #endif
799 #if defined(CONFIG_MMU)
800         {
801                 .ctl_name       = KERN_RANDOMIZE,
802                 .procname       = "randomize_va_space",
803                 .data           = &randomize_va_space,
804                 .maxlen         = sizeof(int),
805                 .mode           = 0644,
806                 .proc_handler   = &proc_dointvec,
807         },
808 #endif
809 #if defined(CONFIG_S390) && defined(CONFIG_SMP)
810         {
811                 .ctl_name       = KERN_SPIN_RETRY,
812                 .procname       = "spin_retry",
813                 .data           = &spin_retry,
814                 .maxlen         = sizeof (int),
815                 .mode           = 0644,
816                 .proc_handler   = &proc_dointvec,
817         },
818 #endif
819 #if     defined(CONFIG_ACPI_SLEEP) && defined(CONFIG_X86)
820         {
821                 .procname       = "acpi_video_flags",
822                 .data           = &acpi_realmode_flags,
823                 .maxlen         = sizeof (unsigned long),
824                 .mode           = 0644,
825                 .proc_handler   = &proc_doulongvec_minmax,
826         },
827 #endif
828 #ifdef CONFIG_IA64
829         {
830                 .ctl_name       = KERN_IA64_UNALIGNED,
831                 .procname       = "ignore-unaligned-usertrap",
832                 .data           = &no_unaligned_warning,
833                 .maxlen         = sizeof (int),
834                 .mode           = 0644,
835                 .proc_handler   = &proc_dointvec,
836         },
837         {
838                 .ctl_name       = CTL_UNNUMBERED,
839                 .procname       = "unaligned-dump-stack",
840                 .data           = &unaligned_dump_stack,
841                 .maxlen         = sizeof (int),
842                 .mode           = 0644,
843                 .proc_handler   = &proc_dointvec,
844         },
845 #endif
846 #ifdef CONFIG_DETECT_SOFTLOCKUP
847         {
848                 .ctl_name       = CTL_UNNUMBERED,
849                 .procname       = "softlockup_panic",
850                 .data           = &softlockup_panic,
851                 .maxlen         = sizeof(int),
852                 .mode           = 0644,
853                 .proc_handler   = &proc_dointvec_minmax,
854                 .strategy       = &sysctl_intvec,
855                 .extra1         = &zero,
856                 .extra2         = &one,
857         },
858         {
859                 .ctl_name       = CTL_UNNUMBERED,
860                 .procname       = "softlockup_thresh",
861                 .data           = &softlockup_thresh,
862                 .maxlen         = sizeof(int),
863                 .mode           = 0644,
864                 .proc_handler   = &proc_dosoftlockup_thresh,
865                 .strategy       = &sysctl_intvec,
866                 .extra1         = &neg_one,
867                 .extra2         = &sixty,
868         },
869 #endif
870 #ifdef CONFIG_DETECT_HUNG_TASK
871         {
872                 .ctl_name       = CTL_UNNUMBERED,
873                 .procname       = "hung_task_panic",
874                 .data           = &sysctl_hung_task_panic,
875                 .maxlen         = sizeof(int),
876                 .mode           = 0644,
877                 .proc_handler   = &proc_dointvec_minmax,
878                 .strategy       = &sysctl_intvec,
879                 .extra1         = &zero,
880                 .extra2         = &one,
881         },
882         {
883                 .ctl_name       = CTL_UNNUMBERED,
884                 .procname       = "hung_task_check_count",
885                 .data           = &sysctl_hung_task_check_count,
886                 .maxlen         = sizeof(unsigned long),
887                 .mode           = 0644,
888                 .proc_handler   = &proc_doulongvec_minmax,
889                 .strategy       = &sysctl_intvec,
890         },
891         {
892                 .ctl_name       = CTL_UNNUMBERED,
893                 .procname       = "hung_task_timeout_secs",
894                 .data           = &sysctl_hung_task_timeout_secs,
895                 .maxlen         = sizeof(unsigned long),
896                 .mode           = 0644,
897                 .proc_handler   = &proc_dohung_task_timeout_secs,
898                 .strategy       = &sysctl_intvec,
899         },
900         {
901                 .ctl_name       = CTL_UNNUMBERED,
902                 .procname       = "hung_task_warnings",
903                 .data           = &sysctl_hung_task_warnings,
904                 .maxlen         = sizeof(unsigned long),
905                 .mode           = 0644,
906                 .proc_handler   = &proc_doulongvec_minmax,
907                 .strategy       = &sysctl_intvec,
908         },
909 #endif
910 #ifdef CONFIG_COMPAT
911         {
912                 .ctl_name       = KERN_COMPAT_LOG,
913                 .procname       = "compat-log",
914                 .data           = &compat_log,
915                 .maxlen         = sizeof (int),
916                 .mode           = 0644,
917                 .proc_handler   = &proc_dointvec,
918         },
919 #endif
920 #ifdef CONFIG_RT_MUTEXES
921         {
922                 .ctl_name       = KERN_MAX_LOCK_DEPTH,
923                 .procname       = "max_lock_depth",
924                 .data           = &max_lock_depth,
925                 .maxlen         = sizeof(int),
926                 .mode           = 0644,
927                 .proc_handler   = &proc_dointvec,
928         },
929 #endif
930         {
931                 .ctl_name       = CTL_UNNUMBERED,
932                 .procname       = "poweroff_cmd",
933                 .data           = &poweroff_cmd,
934                 .maxlen         = POWEROFF_CMD_PATH_LEN,
935                 .mode           = 0644,
936                 .proc_handler   = &proc_dostring,
937                 .strategy       = &sysctl_string,
938         },
939 #ifdef CONFIG_KEYS
940         {
941                 .ctl_name       = CTL_UNNUMBERED,
942                 .procname       = "keys",
943                 .mode           = 0555,
944                 .child          = key_sysctls,
945         },
946 #endif
947 #ifdef CONFIG_RCU_TORTURE_TEST
948         {
949                 .ctl_name       = CTL_UNNUMBERED,
950                 .procname       = "rcutorture_runnable",
951                 .data           = &rcutorture_runnable,
952                 .maxlen         = sizeof(int),
953                 .mode           = 0644,
954                 .proc_handler   = &proc_dointvec,
955         },
956 #endif
957 #ifdef CONFIG_SLOW_WORK
958         {
959                 .ctl_name       = CTL_UNNUMBERED,
960                 .procname       = "slow-work",
961                 .mode           = 0555,
962                 .child          = slow_work_sysctls,
963         },
964 #endif
965 #ifdef CONFIG_PERF_COUNTERS
966         {
967                 .ctl_name       = CTL_UNNUMBERED,
968                 .procname       = "perf_counter_paranoid",
969                 .data           = &sysctl_perf_counter_paranoid,
970                 .maxlen         = sizeof(sysctl_perf_counter_paranoid),
971                 .mode           = 0644,
972                 .proc_handler   = &proc_dointvec,
973         },
974         {
975                 .ctl_name       = CTL_UNNUMBERED,
976                 .procname       = "perf_counter_mlock_kb",
977                 .data           = &sysctl_perf_counter_mlock,
978                 .maxlen         = sizeof(sysctl_perf_counter_mlock),
979                 .mode           = 0644,
980                 .proc_handler   = &proc_dointvec,
981         },
982         {
983                 .ctl_name       = CTL_UNNUMBERED,
984                 .procname       = "perf_counter_max_sample_rate",
985                 .data           = &sysctl_perf_counter_sample_rate,
986                 .maxlen         = sizeof(sysctl_perf_counter_sample_rate),
987                 .mode           = 0644,
988                 .proc_handler   = &proc_dointvec,
989         },
990 #endif
991 #ifdef CONFIG_KMEMCHECK
992         {
993                 .ctl_name       = CTL_UNNUMBERED,
994                 .procname       = "kmemcheck",
995                 .data           = &kmemcheck_enabled,
996                 .maxlen         = sizeof(int),
997                 .mode           = 0644,
998                 .proc_handler   = &proc_dointvec,
999         },
1000 #endif
1001         {
1002                 .ctl_name       = CTL_UNNUMBERED,
1003                 .procname       = "blk_iopoll",
1004                 .data           = &blk_iopoll_enabled,
1005                 .maxlen         = sizeof(int),
1006                 .mode           = 0644,
1007                 .proc_handler   = &proc_dointvec,
1008         },
1009 /*
1010  * NOTE: do not add new entries to this table unless you have read
1011  * Documentation/sysctl/ctl_unnumbered.txt
1012  */
1013         { .ctl_name = 0 }
1014 };
1015
1016 static struct ctl_table vm_table[] = {
1017         {
1018                 .ctl_name       = VM_OVERCOMMIT_MEMORY,
1019                 .procname       = "overcommit_memory",
1020                 .data           = &sysctl_overcommit_memory,
1021                 .maxlen         = sizeof(sysctl_overcommit_memory),
1022                 .mode           = 0644,
1023                 .proc_handler   = &proc_dointvec,
1024         },
1025         {
1026                 .ctl_name       = VM_PANIC_ON_OOM,
1027                 .procname       = "panic_on_oom",
1028                 .data           = &sysctl_panic_on_oom,
1029                 .maxlen         = sizeof(sysctl_panic_on_oom),
1030                 .mode           = 0644,
1031                 .proc_handler   = &proc_dointvec,
1032         },
1033         {
1034                 .ctl_name       = CTL_UNNUMBERED,
1035                 .procname       = "oom_kill_allocating_task",
1036                 .data           = &sysctl_oom_kill_allocating_task,
1037                 .maxlen         = sizeof(sysctl_oom_kill_allocating_task),
1038                 .mode           = 0644,
1039                 .proc_handler   = &proc_dointvec,
1040         },
1041         {
1042                 .ctl_name       = CTL_UNNUMBERED,
1043                 .procname       = "oom_dump_tasks",
1044                 .data           = &sysctl_oom_dump_tasks,
1045                 .maxlen         = sizeof(sysctl_oom_dump_tasks),
1046                 .mode           = 0644,
1047                 .proc_handler   = &proc_dointvec,
1048         },
1049         {
1050                 .ctl_name       = VM_OVERCOMMIT_RATIO,
1051                 .procname       = "overcommit_ratio",
1052                 .data           = &sysctl_overcommit_ratio,
1053                 .maxlen         = sizeof(sysctl_overcommit_ratio),
1054                 .mode           = 0644,
1055                 .proc_handler   = &proc_dointvec,
1056         },
1057         {
1058                 .ctl_name       = VM_PAGE_CLUSTER,
1059                 .procname       = "page-cluster", 
1060                 .data           = &page_cluster,
1061                 .maxlen         = sizeof(int),
1062                 .mode           = 0644,
1063                 .proc_handler   = &proc_dointvec,
1064         },
1065         {
1066                 .ctl_name       = VM_DIRTY_BACKGROUND,
1067                 .procname       = "dirty_background_ratio",
1068                 .data           = &dirty_background_ratio,
1069                 .maxlen         = sizeof(dirty_background_ratio),
1070                 .mode           = 0644,
1071                 .proc_handler   = &dirty_background_ratio_handler,
1072                 .strategy       = &sysctl_intvec,
1073                 .extra1         = &zero,
1074                 .extra2         = &one_hundred,
1075         },
1076         {
1077                 .ctl_name       = CTL_UNNUMBERED,
1078                 .procname       = "dirty_background_bytes",
1079                 .data           = &dirty_background_bytes,
1080                 .maxlen         = sizeof(dirty_background_bytes),
1081                 .mode           = 0644,
1082                 .proc_handler   = &dirty_background_bytes_handler,
1083                 .strategy       = &sysctl_intvec,
1084                 .extra1         = &one_ul,
1085         },
1086         {
1087                 .ctl_name       = VM_DIRTY_RATIO,
1088                 .procname       = "dirty_ratio",
1089                 .data           = &vm_dirty_ratio,
1090                 .maxlen         = sizeof(vm_dirty_ratio),
1091                 .mode           = 0644,
1092                 .proc_handler   = &dirty_ratio_handler,
1093                 .strategy       = &sysctl_intvec,
1094                 .extra1         = &zero,
1095                 .extra2         = &one_hundred,
1096         },
1097         {
1098                 .ctl_name       = CTL_UNNUMBERED,
1099                 .procname       = "dirty_bytes",
1100                 .data           = &vm_dirty_bytes,
1101                 .maxlen         = sizeof(vm_dirty_bytes),
1102                 .mode           = 0644,
1103                 .proc_handler   = &dirty_bytes_handler,
1104                 .strategy       = &sysctl_intvec,
1105                 .extra1         = &dirty_bytes_min,
1106         },
1107         {
1108                 .procname       = "dirty_writeback_centisecs",
1109                 .data           = &dirty_writeback_interval,
1110                 .maxlen         = sizeof(dirty_writeback_interval),
1111                 .mode           = 0644,
1112                 .proc_handler   = &dirty_writeback_centisecs_handler,
1113         },
1114         {
1115                 .procname       = "dirty_expire_centisecs",
1116                 .data           = &dirty_expire_interval,
1117                 .maxlen         = sizeof(dirty_expire_interval),
1118                 .mode           = 0644,
1119                 .proc_handler   = &proc_dointvec,
1120         },
1121         {
1122                 .ctl_name       = VM_NR_PDFLUSH_THREADS,
1123                 .procname       = "nr_pdflush_threads",
1124                 .data           = &nr_pdflush_threads,
1125                 .maxlen         = sizeof nr_pdflush_threads,
1126                 .mode           = 0444 /* read-only*/,
1127                 .proc_handler   = &proc_dointvec,
1128         },
1129         {
1130                 .ctl_name       = VM_SWAPPINESS,
1131                 .procname       = "swappiness",
1132                 .data           = &vm_swappiness,
1133                 .maxlen         = sizeof(vm_swappiness),
1134                 .mode           = 0644,
1135                 .proc_handler   = &proc_dointvec_minmax,
1136                 .strategy       = &sysctl_intvec,
1137                 .extra1         = &zero,
1138                 .extra2         = &one_hundred,
1139         },
1140 #ifdef CONFIG_HUGETLB_PAGE
1141          {
1142                 .procname       = "nr_hugepages",
1143                 .data           = NULL,
1144                 .maxlen         = sizeof(unsigned long),
1145                 .mode           = 0644,
1146                 .proc_handler   = &hugetlb_sysctl_handler,
1147                 .extra1         = (void *)&hugetlb_zero,
1148                 .extra2         = (void *)&hugetlb_infinity,
1149          },
1150          {
1151                 .ctl_name       = VM_HUGETLB_GROUP,
1152                 .procname       = "hugetlb_shm_group",
1153                 .data           = &sysctl_hugetlb_shm_group,
1154                 .maxlen         = sizeof(gid_t),
1155                 .mode           = 0644,
1156                 .proc_handler   = &proc_dointvec,
1157          },
1158          {
1159                 .ctl_name       = CTL_UNNUMBERED,
1160                 .procname       = "hugepages_treat_as_movable",
1161                 .data           = &hugepages_treat_as_movable,
1162                 .maxlen         = sizeof(int),
1163                 .mode           = 0644,
1164                 .proc_handler   = &hugetlb_treat_movable_handler,
1165         },
1166         {
1167                 .ctl_name       = CTL_UNNUMBERED,
1168                 .procname       = "nr_overcommit_hugepages",
1169                 .data           = NULL,
1170                 .maxlen         = sizeof(unsigned long),
1171                 .mode           = 0644,
1172                 .proc_handler   = &hugetlb_overcommit_handler,
1173                 .extra1         = (void *)&hugetlb_zero,
1174                 .extra2         = (void *)&hugetlb_infinity,
1175         },
1176 #endif
1177         {
1178                 .ctl_name       = VM_LOWMEM_RESERVE_RATIO,
1179                 .procname       = "lowmem_reserve_ratio",
1180                 .data           = &sysctl_lowmem_reserve_ratio,
1181                 .maxlen         = sizeof(sysctl_lowmem_reserve_ratio),
1182                 .mode           = 0644,
1183                 .proc_handler   = &lowmem_reserve_ratio_sysctl_handler,
1184                 .strategy       = &sysctl_intvec,
1185         },
1186         {
1187                 .ctl_name       = VM_DROP_PAGECACHE,
1188                 .procname       = "drop_caches",
1189                 .data           = &sysctl_drop_caches,
1190                 .maxlen         = sizeof(int),
1191                 .mode           = 0644,
1192                 .proc_handler   = drop_caches_sysctl_handler,
1193                 .strategy       = &sysctl_intvec,
1194         },
1195         {
1196                 .ctl_name       = VM_MIN_FREE_KBYTES,
1197                 .procname       = "min_free_kbytes",
1198                 .data           = &min_free_kbytes,
1199                 .maxlen         = sizeof(min_free_kbytes),
1200                 .mode           = 0644,
1201                 .proc_handler   = &min_free_kbytes_sysctl_handler,
1202                 .strategy       = &sysctl_intvec,
1203                 .extra1         = &zero,
1204         },
1205         {
1206                 .ctl_name       = VM_PERCPU_PAGELIST_FRACTION,
1207                 .procname       = "percpu_pagelist_fraction",
1208                 .data           = &percpu_pagelist_fraction,
1209                 .maxlen         = sizeof(percpu_pagelist_fraction),
1210                 .mode           = 0644,
1211                 .proc_handler   = &percpu_pagelist_fraction_sysctl_handler,
1212                 .strategy       = &sysctl_intvec,
1213                 .extra1         = &min_percpu_pagelist_fract,
1214         },
1215 #ifdef CONFIG_MMU
1216         {
1217                 .ctl_name       = VM_MAX_MAP_COUNT,
1218                 .procname       = "max_map_count",
1219                 .data           = &sysctl_max_map_count,
1220                 .maxlen         = sizeof(sysctl_max_map_count),
1221                 .mode           = 0644,
1222                 .proc_handler   = &proc_dointvec
1223         },
1224 #else
1225         {
1226                 .ctl_name       = CTL_UNNUMBERED,
1227                 .procname       = "nr_trim_pages",
1228                 .data           = &sysctl_nr_trim_pages,
1229                 .maxlen         = sizeof(sysctl_nr_trim_pages),
1230                 .mode           = 0644,
1231                 .proc_handler   = &proc_dointvec_minmax,
1232                 .strategy       = &sysctl_intvec,
1233                 .extra1         = &zero,
1234         },
1235 #endif
1236         {
1237                 .ctl_name       = VM_LAPTOP_MODE,
1238                 .procname       = "laptop_mode",
1239                 .data           = &laptop_mode,
1240                 .maxlen         = sizeof(laptop_mode),
1241                 .mode           = 0644,
1242                 .proc_handler   = &proc_dointvec_jiffies,
1243                 .strategy       = &sysctl_jiffies,
1244         },
1245         {
1246                 .ctl_name       = VM_BLOCK_DUMP,
1247                 .procname       = "block_dump",
1248                 .data           = &block_dump,
1249                 .maxlen         = sizeof(block_dump),
1250                 .mode           = 0644,
1251                 .proc_handler   = &proc_dointvec,
1252                 .strategy       = &sysctl_intvec,
1253                 .extra1         = &zero,
1254         },
1255         {
1256                 .ctl_name       = VM_VFS_CACHE_PRESSURE,
1257                 .procname       = "vfs_cache_pressure",
1258                 .data           = &sysctl_vfs_cache_pressure,
1259                 .maxlen         = sizeof(sysctl_vfs_cache_pressure),
1260                 .mode           = 0644,
1261                 .proc_handler   = &proc_dointvec,
1262                 .strategy       = &sysctl_intvec,
1263                 .extra1         = &zero,
1264         },
1265 #ifdef HAVE_ARCH_PICK_MMAP_LAYOUT
1266         {
1267                 .ctl_name       = VM_LEGACY_VA_LAYOUT,
1268                 .procname       = "legacy_va_layout",
1269                 .data           = &sysctl_legacy_va_layout,
1270                 .maxlen         = sizeof(sysctl_legacy_va_layout),
1271                 .mode           = 0644,
1272                 .proc_handler   = &proc_dointvec,
1273                 .strategy       = &sysctl_intvec,
1274                 .extra1         = &zero,
1275         },
1276 #endif
1277 #ifdef CONFIG_NUMA
1278         {
1279                 .ctl_name       = VM_ZONE_RECLAIM_MODE,
1280                 .procname       = "zone_reclaim_mode",
1281                 .data           = &zone_reclaim_mode,
1282                 .maxlen         = sizeof(zone_reclaim_mode),
1283                 .mode           = 0644,
1284                 .proc_handler   = &proc_dointvec,
1285                 .strategy       = &sysctl_intvec,
1286                 .extra1         = &zero,
1287         },
1288         {
1289                 .ctl_name       = VM_MIN_UNMAPPED,
1290                 .procname       = "min_unmapped_ratio",
1291                 .data           = &sysctl_min_unmapped_ratio,
1292                 .maxlen         = sizeof(sysctl_min_unmapped_ratio),
1293                 .mode           = 0644,
1294                 .proc_handler   = &sysctl_min_unmapped_ratio_sysctl_handler,
1295                 .strategy       = &sysctl_intvec,
1296                 .extra1         = &zero,
1297                 .extra2         = &one_hundred,
1298         },
1299         {
1300                 .ctl_name       = VM_MIN_SLAB,
1301                 .procname       = "min_slab_ratio",
1302                 .data           = &sysctl_min_slab_ratio,
1303                 .maxlen         = sizeof(sysctl_min_slab_ratio),
1304                 .mode           = 0644,
1305                 .proc_handler   = &sysctl_min_slab_ratio_sysctl_handler,
1306                 .strategy       = &sysctl_intvec,
1307                 .extra1         = &zero,
1308                 .extra2         = &one_hundred,
1309         },
1310 #endif
1311 #ifdef CONFIG_SMP
1312         {
1313                 .ctl_name       = CTL_UNNUMBERED,
1314                 .procname       = "stat_interval",
1315                 .data           = &sysctl_stat_interval,
1316                 .maxlen         = sizeof(sysctl_stat_interval),
1317                 .mode           = 0644,
1318                 .proc_handler   = &proc_dointvec_jiffies,
1319                 .strategy       = &sysctl_jiffies,
1320         },
1321 #endif
1322         {
1323                 .ctl_name       = CTL_UNNUMBERED,
1324                 .procname       = "mmap_min_addr",
1325                 .data           = &dac_mmap_min_addr,
1326                 .maxlen         = sizeof(unsigned long),
1327                 .mode           = 0644,
1328                 .proc_handler   = &mmap_min_addr_handler,
1329         },
1330 #ifdef CONFIG_NUMA
1331         {
1332                 .ctl_name       = CTL_UNNUMBERED,
1333                 .procname       = "numa_zonelist_order",
1334                 .data           = &numa_zonelist_order,
1335                 .maxlen         = NUMA_ZONELIST_ORDER_LEN,
1336                 .mode           = 0644,
1337                 .proc_handler   = &numa_zonelist_order_handler,
1338                 .strategy       = &sysctl_string,
1339         },
1340 #endif
1341 #if (defined(CONFIG_X86_32) && !defined(CONFIG_UML))|| \
1342    (defined(CONFIG_SUPERH) && defined(CONFIG_VSYSCALL))
1343         {
1344                 .ctl_name       = VM_VDSO_ENABLED,
1345                 .procname       = "vdso_enabled",
1346                 .data           = &vdso_enabled,
1347                 .maxlen         = sizeof(vdso_enabled),
1348                 .mode           = 0644,
1349                 .proc_handler   = &proc_dointvec,
1350                 .strategy       = &sysctl_intvec,
1351                 .extra1         = &zero,
1352         },
1353 #endif
1354 #ifdef CONFIG_HIGHMEM
1355         {
1356                 .ctl_name       = CTL_UNNUMBERED,
1357                 .procname       = "highmem_is_dirtyable",
1358                 .data           = &vm_highmem_is_dirtyable,
1359                 .maxlen         = sizeof(vm_highmem_is_dirtyable),
1360                 .mode           = 0644,
1361                 .proc_handler   = &proc_dointvec_minmax,
1362                 .strategy       = &sysctl_intvec,
1363                 .extra1         = &zero,
1364                 .extra2         = &one,
1365         },
1366 #endif
1367         {
1368                 .ctl_name       = CTL_UNNUMBERED,
1369                 .procname       = "scan_unevictable_pages",
1370                 .data           = &scan_unevictable_pages,
1371                 .maxlen         = sizeof(scan_unevictable_pages),
1372                 .mode           = 0644,
1373                 .proc_handler   = &scan_unevictable_handler,
1374         },
1375 /*
1376  * NOTE: do not add new entries to this table unless you have read
1377  * Documentation/sysctl/ctl_unnumbered.txt
1378  */
1379         { .ctl_name = 0 }
1380 };
1381
1382 #if defined(CONFIG_BINFMT_MISC) || defined(CONFIG_BINFMT_MISC_MODULE)
1383 static struct ctl_table binfmt_misc_table[] = {
1384         { .ctl_name = 0 }
1385 };
1386 #endif
1387
1388 static struct ctl_table fs_table[] = {
1389         {
1390                 .ctl_name       = FS_NRINODE,
1391                 .procname       = "inode-nr",
1392                 .data           = &inodes_stat,
1393                 .maxlen         = 2*sizeof(int),
1394                 .mode           = 0444,
1395                 .proc_handler   = &proc_dointvec,
1396         },
1397         {
1398                 .ctl_name       = FS_STATINODE,
1399                 .procname       = "inode-state",
1400                 .data           = &inodes_stat,
1401                 .maxlen         = 7*sizeof(int),
1402                 .mode           = 0444,
1403                 .proc_handler   = &proc_dointvec,
1404         },
1405         {
1406                 .procname       = "file-nr",
1407                 .data           = &files_stat,
1408                 .maxlen         = 3*sizeof(int),
1409                 .mode           = 0444,
1410                 .proc_handler   = &proc_nr_files,
1411         },
1412         {
1413                 .ctl_name       = FS_MAXFILE,
1414                 .procname       = "file-max",
1415                 .data           = &files_stat.max_files,
1416                 .maxlen         = sizeof(int),
1417                 .mode           = 0644,
1418                 .proc_handler   = &proc_dointvec,
1419         },
1420         {
1421                 .ctl_name       = CTL_UNNUMBERED,
1422                 .procname       = "nr_open",
1423                 .data           = &sysctl_nr_open,
1424                 .maxlen         = sizeof(int),
1425                 .mode           = 0644,
1426                 .proc_handler   = &proc_dointvec_minmax,
1427                 .extra1         = &sysctl_nr_open_min,
1428                 .extra2         = &sysctl_nr_open_max,
1429         },
1430         {
1431                 .ctl_name       = FS_DENTRY,
1432                 .procname       = "dentry-state",
1433                 .data           = &dentry_stat,
1434                 .maxlen         = 6*sizeof(int),
1435                 .mode           = 0444,
1436                 .proc_handler   = &proc_dointvec,
1437         },
1438         {
1439                 .ctl_name       = FS_OVERFLOWUID,
1440                 .procname       = "overflowuid",
1441                 .data           = &fs_overflowuid,
1442                 .maxlen         = sizeof(int),
1443                 .mode           = 0644,
1444                 .proc_handler   = &proc_dointvec_minmax,
1445                 .strategy       = &sysctl_intvec,
1446                 .extra1         = &minolduid,
1447                 .extra2         = &maxolduid,
1448         },
1449         {
1450                 .ctl_name       = FS_OVERFLOWGID,
1451                 .procname       = "overflowgid",
1452                 .data           = &fs_overflowgid,
1453                 .maxlen         = sizeof(int),
1454                 .mode           = 0644,
1455                 .proc_handler   = &proc_dointvec_minmax,
1456                 .strategy       = &sysctl_intvec,
1457                 .extra1         = &minolduid,
1458                 .extra2         = &maxolduid,
1459         },
1460 #ifdef CONFIG_FILE_LOCKING
1461         {
1462                 .ctl_name       = FS_LEASES,
1463                 .procname       = "leases-enable",
1464                 .data           = &leases_enable,
1465                 .maxlen         = sizeof(int),
1466                 .mode           = 0644,
1467                 .proc_handler   = &proc_dointvec,
1468         },
1469 #endif
1470 #ifdef CONFIG_DNOTIFY
1471         {
1472                 .ctl_name       = FS_DIR_NOTIFY,
1473                 .procname       = "dir-notify-enable",
1474                 .data           = &dir_notify_enable,
1475                 .maxlen         = sizeof(int),
1476                 .mode           = 0644,
1477                 .proc_handler   = &proc_dointvec,
1478         },
1479 #endif
1480 #ifdef CONFIG_MMU
1481 #ifdef CONFIG_FILE_LOCKING
1482         {
1483                 .ctl_name       = FS_LEASE_TIME,
1484                 .procname       = "lease-break-time",
1485                 .data           = &lease_break_time,
1486                 .maxlen         = sizeof(int),
1487                 .mode           = 0644,
1488                 .proc_handler   = &proc_dointvec,
1489         },
1490 #endif
1491 #ifdef CONFIG_AIO
1492         {
1493                 .procname       = "aio-nr",
1494                 .data           = &aio_nr,
1495                 .maxlen         = sizeof(aio_nr),
1496                 .mode           = 0444,
1497                 .proc_handler   = &proc_doulongvec_minmax,
1498         },
1499         {
1500                 .procname       = "aio-max-nr",
1501                 .data           = &aio_max_nr,
1502                 .maxlen         = sizeof(aio_max_nr),
1503                 .mode           = 0644,
1504                 .proc_handler   = &proc_doulongvec_minmax,
1505         },
1506 #endif /* CONFIG_AIO */
1507 #ifdef CONFIG_INOTIFY_USER
1508         {
1509                 .ctl_name       = FS_INOTIFY,
1510                 .procname       = "inotify",
1511                 .mode           = 0555,
1512                 .child          = inotify_table,
1513         },
1514 #endif  
1515 #ifdef CONFIG_EPOLL
1516         {
1517                 .procname       = "epoll",
1518                 .mode           = 0555,
1519                 .child          = epoll_table,
1520         },
1521 #endif
1522 #endif
1523         {
1524                 .ctl_name       = KERN_SETUID_DUMPABLE,
1525                 .procname       = "suid_dumpable",
1526                 .data           = &suid_dumpable,
1527                 .maxlen         = sizeof(int),
1528                 .mode           = 0644,
1529                 .proc_handler   = &proc_dointvec_minmax,
1530                 .strategy       = &sysctl_intvec,
1531                 .extra1         = &zero,
1532                 .extra2         = &two,
1533         },
1534 #if defined(CONFIG_BINFMT_MISC) || defined(CONFIG_BINFMT_MISC_MODULE)
1535         {
1536                 .ctl_name       = CTL_UNNUMBERED,
1537                 .procname       = "binfmt_misc",
1538                 .mode           = 0555,
1539                 .child          = binfmt_misc_table,
1540         },
1541 #endif
1542 /*
1543  * NOTE: do not add new entries to this table unless you have read
1544  * Documentation/sysctl/ctl_unnumbered.txt
1545  */
1546         { .ctl_name = 0 }
1547 };
1548
1549 static struct ctl_table debug_table[] = {
1550 #if defined(CONFIG_X86) || defined(CONFIG_PPC)
1551         {
1552                 .ctl_name       = CTL_UNNUMBERED,
1553                 .procname       = "exception-trace",
1554                 .data           = &show_unhandled_signals,
1555                 .maxlen         = sizeof(int),
1556                 .mode           = 0644,
1557                 .proc_handler   = proc_dointvec
1558         },
1559 #endif
1560         { .ctl_name = 0 }
1561 };
1562
1563 static struct ctl_table dev_table[] = {
1564         { .ctl_name = 0 }
1565 };
1566
1567 static DEFINE_SPINLOCK(sysctl_lock);
1568
1569 /* called under sysctl_lock */
1570 static int use_table(struct ctl_table_header *p)
1571 {
1572         if (unlikely(p->unregistering))
1573                 return 0;
1574         p->used++;
1575         return 1;
1576 }
1577
1578 /* called under sysctl_lock */
1579 static void unuse_table(struct ctl_table_header *p)
1580 {
1581         if (!--p->used)
1582                 if (unlikely(p->unregistering))
1583                         complete(p->unregistering);
1584 }
1585
1586 /* called under sysctl_lock, will reacquire if has to wait */
1587 static void start_unregistering(struct ctl_table_header *p)
1588 {
1589         /*
1590          * if p->used is 0, nobody will ever touch that entry again;
1591          * we'll eliminate all paths to it before dropping sysctl_lock
1592          */
1593         if (unlikely(p->used)) {
1594                 struct completion wait;
1595                 init_completion(&wait);
1596                 p->unregistering = &wait;
1597                 spin_unlock(&sysctl_lock);
1598                 wait_for_completion(&wait);
1599                 spin_lock(&sysctl_lock);
1600         } else {
1601                 /* anything non-NULL; we'll never dereference it */
1602                 p->unregistering = ERR_PTR(-EINVAL);
1603         }
1604         /*
1605          * do not remove from the list until nobody holds it; walking the
1606          * list in do_sysctl() relies on that.
1607          */
1608         list_del_init(&p->ctl_entry);
1609 }
1610
1611 void sysctl_head_get(struct ctl_table_header *head)
1612 {
1613         spin_lock(&sysctl_lock);
1614         head->count++;
1615         spin_unlock(&sysctl_lock);
1616 }
1617
1618 void sysctl_head_put(struct ctl_table_header *head)
1619 {
1620         spin_lock(&sysctl_lock);
1621         if (!--head->count)
1622                 kfree(head);
1623         spin_unlock(&sysctl_lock);
1624 }
1625
1626 struct ctl_table_header *sysctl_head_grab(struct ctl_table_header *head)
1627 {
1628         if (!head)
1629                 BUG();
1630         spin_lock(&sysctl_lock);
1631         if (!use_table(head))
1632                 head = ERR_PTR(-ENOENT);
1633         spin_unlock(&sysctl_lock);
1634         return head;
1635 }
1636
1637 void sysctl_head_finish(struct ctl_table_header *head)
1638 {
1639         if (!head)
1640                 return;
1641         spin_lock(&sysctl_lock);
1642         unuse_table(head);
1643         spin_unlock(&sysctl_lock);
1644 }
1645
1646 static struct ctl_table_set *
1647 lookup_header_set(struct ctl_table_root *root, struct nsproxy *namespaces)
1648 {
1649         struct ctl_table_set *set = &root->default_set;
1650         if (root->lookup)
1651                 set = root->lookup(root, namespaces);
1652         return set;
1653 }
1654
1655 static struct list_head *
1656 lookup_header_list(struct ctl_table_root *root, struct nsproxy *namespaces)
1657 {
1658         struct ctl_table_set *set = lookup_header_set(root, namespaces);
1659         return &set->list;
1660 }
1661
1662 struct ctl_table_header *__sysctl_head_next(struct nsproxy *namespaces,
1663                                             struct ctl_table_header *prev)
1664 {
1665         struct ctl_table_root *root;
1666         struct list_head *header_list;
1667         struct ctl_table_header *head;
1668         struct list_head *tmp;
1669
1670         spin_lock(&sysctl_lock);
1671         if (prev) {
1672                 head = prev;
1673                 tmp = &prev->ctl_entry;
1674                 unuse_table(prev);
1675                 goto next;
1676         }
1677         tmp = &root_table_header.ctl_entry;
1678         for (;;) {
1679                 head = list_entry(tmp, struct ctl_table_header, ctl_entry);
1680
1681                 if (!use_table(head))
1682                         goto next;
1683                 spin_unlock(&sysctl_lock);
1684                 return head;
1685         next:
1686                 root = head->root;
1687                 tmp = tmp->next;
1688                 header_list = lookup_header_list(root, namespaces);
1689                 if (tmp != header_list)
1690                         continue;
1691
1692                 do {
1693                         root = list_entry(root->root_list.next,
1694                                         struct ctl_table_root, root_list);
1695                         if (root == &sysctl_table_root)
1696                                 goto out;
1697                         header_list = lookup_header_list(root, namespaces);
1698                 } while (list_empty(header_list));
1699                 tmp = header_list->next;
1700         }
1701 out:
1702         spin_unlock(&sysctl_lock);
1703         return NULL;
1704 }
1705
1706 struct ctl_table_header *sysctl_head_next(struct ctl_table_header *prev)
1707 {
1708         return __sysctl_head_next(current->nsproxy, prev);
1709 }
1710
1711 void register_sysctl_root(struct ctl_table_root *root)
1712 {
1713         spin_lock(&sysctl_lock);
1714         list_add_tail(&root->root_list, &sysctl_table_root.root_list);
1715         spin_unlock(&sysctl_lock);
1716 }
1717
1718 #ifdef CONFIG_SYSCTL_SYSCALL
1719 /* Perform the actual read/write of a sysctl table entry. */
1720 static int do_sysctl_strategy(struct ctl_table_root *root,
1721                         struct ctl_table *table,
1722                         void __user *oldval, size_t __user *oldlenp,
1723                         void __user *newval, size_t newlen)
1724 {
1725         int op = 0, rc;
1726
1727         if (oldval)
1728                 op |= MAY_READ;
1729         if (newval)
1730                 op |= MAY_WRITE;
1731         if (sysctl_perm(root, table, op))
1732                 return -EPERM;
1733
1734         if (table->strategy) {
1735                 rc = table->strategy(table, oldval, oldlenp, newval, newlen);
1736                 if (rc < 0)
1737                         return rc;
1738                 if (rc > 0)
1739                         return 0;
1740         }
1741
1742         /* If there is no strategy routine, or if the strategy returns
1743          * zero, proceed with automatic r/w */
1744         if (table->data && table->maxlen) {
1745                 rc = sysctl_data(table, oldval, oldlenp, newval, newlen);
1746                 if (rc < 0)
1747                         return rc;
1748         }
1749         return 0;
1750 }
1751
1752 static int parse_table(int __user *name, int nlen,
1753                        void __user *oldval, size_t __user *oldlenp,
1754                        void __user *newval, size_t newlen,
1755                        struct ctl_table_root *root,
1756                        struct ctl_table *table)
1757 {
1758         int n;
1759 repeat:
1760         if (!nlen)
1761                 return -ENOTDIR;
1762         if (get_user(n, name))
1763                 return -EFAULT;
1764         for ( ; table->ctl_name || table->procname; table++) {
1765                 if (!table->ctl_name)
1766                         continue;
1767                 if (n == table->ctl_name) {
1768                         int error;
1769                         if (table->child) {
1770                                 if (sysctl_perm(root, table, MAY_EXEC))
1771                                         return -EPERM;
1772                                 name++;
1773                                 nlen--;
1774                                 table = table->child;
1775                                 goto repeat;
1776                         }
1777                         error = do_sysctl_strategy(root, table,
1778                                                    oldval, oldlenp,
1779                                                    newval, newlen);
1780                         return error;
1781                 }
1782         }
1783         return -ENOTDIR;
1784 }
1785
1786 int do_sysctl(int __user *name, int nlen, void __user *oldval, size_t __user *oldlenp,
1787                void __user *newval, size_t newlen)
1788 {
1789         struct ctl_table_header *head;
1790         int error = -ENOTDIR;
1791
1792         if (nlen <= 0 || nlen >= CTL_MAXNAME)
1793                 return -ENOTDIR;
1794         if (oldval) {
1795                 int old_len;
1796                 if (!oldlenp || get_user(old_len, oldlenp))
1797                         return -EFAULT;
1798         }
1799
1800         for (head = sysctl_head_next(NULL); head;
1801                         head = sysctl_head_next(head)) {
1802                 error = parse_table(name, nlen, oldval, oldlenp, 
1803                                         newval, newlen,
1804                                         head->root, head->ctl_table);
1805                 if (error != -ENOTDIR) {
1806                         sysctl_head_finish(head);
1807                         break;
1808                 }
1809         }
1810         return error;
1811 }
1812
1813 SYSCALL_DEFINE1(sysctl, struct __sysctl_args __user *, args)
1814 {
1815         struct __sysctl_args tmp;
1816         int error;
1817
1818         if (copy_from_user(&tmp, args, sizeof(tmp)))
1819                 return -EFAULT;
1820
1821         error = deprecated_sysctl_warning(&tmp);
1822         if (error)
1823                 goto out;
1824
1825         lock_kernel();
1826         error = do_sysctl(tmp.name, tmp.nlen, tmp.oldval, tmp.oldlenp,
1827                           tmp.newval, tmp.newlen);
1828         unlock_kernel();
1829 out:
1830         return error;
1831 }
1832 #endif /* CONFIG_SYSCTL_SYSCALL */
1833
1834 /*
1835  * sysctl_perm does NOT grant the superuser all rights automatically, because
1836  * some sysctl variables are readonly even to root.
1837  */
1838
1839 static int test_perm(int mode, int op)
1840 {
1841         if (!current_euid())
1842                 mode >>= 6;
1843         else if (in_egroup_p(0))
1844                 mode >>= 3;
1845         if ((op & ~mode & (MAY_READ|MAY_WRITE|MAY_EXEC)) == 0)
1846                 return 0;
1847         return -EACCES;
1848 }
1849
1850 int sysctl_perm(struct ctl_table_root *root, struct ctl_table *table, int op)
1851 {
1852         int error;
1853         int mode;
1854
1855         error = security_sysctl(table, op & (MAY_READ | MAY_WRITE | MAY_EXEC));
1856         if (error)
1857                 return error;
1858
1859         if (root->permissions)
1860                 mode = root->permissions(root, current->nsproxy, table);
1861         else
1862                 mode = table->mode;
1863
1864         return test_perm(mode, op);
1865 }
1866
1867 static void sysctl_set_parent(struct ctl_table *parent, struct ctl_table *table)
1868 {
1869         for (; table->ctl_name || table->procname; table++) {
1870                 table->parent = parent;
1871                 if (table->child)
1872                         sysctl_set_parent(table, table->child);
1873         }
1874 }
1875
1876 static __init int sysctl_init(void)
1877 {
1878         sysctl_set_parent(NULL, root_table);
1879 #ifdef CONFIG_SYSCTL_SYSCALL_CHECK
1880         {
1881                 int err;
1882                 err = sysctl_check_table(current->nsproxy, root_table);
1883         }
1884 #endif
1885         return 0;
1886 }
1887
1888 core_initcall(sysctl_init);
1889
1890 static struct ctl_table *is_branch_in(struct ctl_table *branch,
1891                                       struct ctl_table *table)
1892 {
1893         struct ctl_table *p;
1894         const char *s = branch->procname;
1895
1896         /* branch should have named subdirectory as its first element */
1897         if (!s || !branch->child)
1898                 return NULL;
1899
1900         /* ... and nothing else */
1901         if (branch[1].procname || branch[1].ctl_name)
1902                 return NULL;
1903
1904         /* table should contain subdirectory with the same name */
1905         for (p = table; p->procname || p->ctl_name; p++) {
1906                 if (!p->child)
1907                         continue;
1908                 if (p->procname && strcmp(p->procname, s) == 0)
1909                         return p;
1910         }
1911         return NULL;
1912 }
1913
1914 /* see if attaching q to p would be an improvement */
1915 static void try_attach(struct ctl_table_header *p, struct ctl_table_header *q)
1916 {
1917         struct ctl_table *to = p->ctl_table, *by = q->ctl_table;
1918         struct ctl_table *next;
1919         int is_better = 0;
1920         int not_in_parent = !p->attached_by;
1921
1922         while ((next = is_branch_in(by, to)) != NULL) {
1923                 if (by == q->attached_by)
1924                         is_better = 1;
1925                 if (to == p->attached_by)
1926                         not_in_parent = 1;
1927                 by = by->child;
1928                 to = next->child;
1929         }
1930
1931         if (is_better && not_in_parent) {
1932                 q->attached_by = by;
1933                 q->attached_to = to;
1934                 q->parent = p;
1935         }
1936 }
1937
1938 /**
1939  * __register_sysctl_paths - register a sysctl hierarchy
1940  * @root: List of sysctl headers to register on
1941  * @namespaces: Data to compute which lists of sysctl entries are visible
1942  * @path: The path to the directory the sysctl table is in.
1943  * @table: the top-level table structure
1944  *
1945  * Register a sysctl table hierarchy. @table should be a filled in ctl_table
1946  * array. A completely 0 filled entry terminates the table.
1947  *
1948  * The members of the &struct ctl_table structure are used as follows:
1949  *
1950  * ctl_name - This is the numeric sysctl value used by sysctl(2). The number
1951  *            must be unique within that level of sysctl
1952  *
1953  * procname - the name of the sysctl file under /proc/sys. Set to %NULL to not
1954  *            enter a sysctl file
1955  *
1956  * data - a pointer to data for use by proc_handler
1957  *
1958  * maxlen - the maximum size in bytes of the data
1959  *
1960  * mode - the file permissions for the /proc/sys file, and for sysctl(2)
1961  *
1962  * child - a pointer to the child sysctl table if this entry is a directory, or
1963  *         %NULL.
1964  *
1965  * proc_handler - the text handler routine (described below)
1966  *
1967  * strategy - the strategy routine (described below)
1968  *
1969  * de - for internal use by the sysctl routines
1970  *
1971  * extra1, extra2 - extra pointers usable by the proc handler routines
1972  *
1973  * Leaf nodes in the sysctl tree will be represented by a single file
1974  * under /proc; non-leaf nodes will be represented by directories.
1975  *
1976  * sysctl(2) can automatically manage read and write requests through
1977  * the sysctl table.  The data and maxlen fields of the ctl_table
1978  * struct enable minimal validation of the values being written to be
1979  * performed, and the mode field allows minimal authentication.
1980  *
1981  * More sophisticated management can be enabled by the provision of a
1982  * strategy routine with the table entry.  This will be called before
1983  * any automatic read or write of the data is performed.
1984  *
1985  * The strategy routine may return
1986  *
1987  * < 0 - Error occurred (error is passed to user process)
1988  *
1989  * 0   - OK - proceed with automatic read or write.
1990  *
1991  * > 0 - OK - read or write has been done by the strategy routine, so
1992  *       return immediately.
1993  *
1994  * There must be a proc_handler routine for any terminal nodes
1995  * mirrored under /proc/sys (non-terminals are handled by a built-in
1996  * directory handler).  Several default handlers are available to
1997  * cover common cases -
1998  *
1999  * proc_dostring(), proc_dointvec(), proc_dointvec_jiffies(),
2000  * proc_dointvec_userhz_jiffies(), proc_dointvec_minmax(), 
2001  * proc_doulongvec_ms_jiffies_minmax(), proc_doulongvec_minmax()
2002  *
2003  * It is the handler's job to read the input buffer from user memory
2004  * and process it. The handler should return 0 on success.
2005  *
2006  * This routine returns %NULL on a failure to register, and a pointer
2007  * to the table header on success.
2008  */
2009 struct ctl_table_header *__register_sysctl_paths(
2010         struct ctl_table_root *root,
2011         struct nsproxy *namespaces,
2012         const struct ctl_path *path, struct ctl_table *table)
2013 {
2014         struct ctl_table_header *header;
2015         struct ctl_table *new, **prevp;
2016         unsigned int n, npath;
2017         struct ctl_table_set *set;
2018
2019         /* Count the path components */
2020         for (npath = 0; path[npath].ctl_name || path[npath].procname; ++npath)
2021                 ;
2022
2023         /*
2024          * For each path component, allocate a 2-element ctl_table array.
2025          * The first array element will be filled with the sysctl entry
2026          * for this, the second will be the sentinel (ctl_name == 0).
2027          *
2028          * We allocate everything in one go so that we don't have to
2029          * worry about freeing additional memory in unregister_sysctl_table.
2030          */
2031         header = kzalloc(sizeof(struct ctl_table_header) +
2032                          (2 * npath * sizeof(struct ctl_table)), GFP_KERNEL);
2033         if (!header)
2034                 return NULL;
2035
2036         new = (struct ctl_table *) (header + 1);
2037
2038         /* Now connect the dots */
2039         prevp = &header->ctl_table;
2040         for (n = 0; n < npath; ++n, ++path) {
2041                 /* Copy the procname */
2042                 new->procname = path->procname;
2043                 new->ctl_name = path->ctl_name;
2044                 new->mode     = 0555;
2045
2046                 *prevp = new;
2047                 prevp = &new->child;
2048
2049                 new += 2;
2050         }
2051         *prevp = table;
2052         header->ctl_table_arg = table;
2053
2054         INIT_LIST_HEAD(&header->ctl_entry);
2055         header->used = 0;
2056         header->unregistering = NULL;
2057         header->root = root;
2058         sysctl_set_parent(NULL, header->ctl_table);
2059         header->count = 1;
2060 #ifdef CONFIG_SYSCTL_SYSCALL_CHECK
2061         if (sysctl_check_table(namespaces, header->ctl_table)) {
2062                 kfree(header);
2063                 return NULL;
2064         }
2065 #endif
2066         spin_lock(&sysctl_lock);
2067         header->set = lookup_header_set(root, namespaces);
2068         header->attached_by = header->ctl_table;
2069         header->attached_to = root_table;
2070         header->parent = &root_table_header;
2071         for (set = header->set; set; set = set->parent) {
2072                 struct ctl_table_header *p;
2073                 list_for_each_entry(p, &set->list, ctl_entry) {
2074                         if (p->unregistering)
2075                                 continue;
2076                         try_attach(p, header);
2077                 }
2078         }
2079         header->parent->count++;
2080         list_add_tail(&header->ctl_entry, &header->set->list);
2081         spin_unlock(&sysctl_lock);
2082
2083         return header;
2084 }
2085
2086 /**
2087  * register_sysctl_table_path - register a sysctl table hierarchy
2088  * @path: The path to the directory the sysctl table is in.
2089  * @table: the top-level table structure
2090  *
2091  * Register a sysctl table hierarchy. @table should be a filled in ctl_table
2092  * array. A completely 0 filled entry terminates the table.
2093  *
2094  * See __register_sysctl_paths for more details.
2095  */
2096 struct ctl_table_header *register_sysctl_paths(const struct ctl_path *path,
2097                                                 struct ctl_table *table)
2098 {
2099         return __register_sysctl_paths(&sysctl_table_root, current->nsproxy,
2100                                         path, table);
2101 }
2102
2103 /**
2104  * register_sysctl_table - register a sysctl table hierarchy
2105  * @table: the top-level table structure
2106  *
2107  * Register a sysctl table hierarchy. @table should be a filled in ctl_table
2108  * array. A completely 0 filled entry terminates the table.
2109  *
2110  * See register_sysctl_paths for more details.
2111  */
2112 struct ctl_table_header *register_sysctl_table(struct ctl_table *table)
2113 {
2114         static const struct ctl_path null_path[] = { {} };
2115
2116         return register_sysctl_paths(null_path, table);
2117 }
2118
2119 /**
2120  * unregister_sysctl_table - unregister a sysctl table hierarchy
2121  * @header: the header returned from register_sysctl_table
2122  *
2123  * Unregisters the sysctl table and all children. proc entries may not
2124  * actually be removed until they are no longer used by anyone.
2125  */
2126 void unregister_sysctl_table(struct ctl_table_header * header)
2127 {
2128         might_sleep();
2129
2130         if (header == NULL)
2131                 return;
2132
2133         spin_lock(&sysctl_lock);
2134         start_unregistering(header);
2135         if (!--header->parent->count) {
2136                 WARN_ON(1);
2137                 kfree(header->parent);
2138         }
2139         if (!--header->count)
2140                 kfree(header);
2141         spin_unlock(&sysctl_lock);
2142 }
2143
2144 int sysctl_is_seen(struct ctl_table_header *p)
2145 {
2146         struct ctl_table_set *set = p->set;
2147         int res;
2148         spin_lock(&sysctl_lock);
2149         if (p->unregistering)
2150                 res = 0;
2151         else if (!set->is_seen)
2152                 res = 1;
2153         else
2154                 res = set->is_seen(set);
2155         spin_unlock(&sysctl_lock);
2156         return res;
2157 }
2158
2159 void setup_sysctl_set(struct ctl_table_set *p,
2160         struct ctl_table_set *parent,
2161         int (*is_seen)(struct ctl_table_set *))
2162 {
2163         INIT_LIST_HEAD(&p->list);
2164         p->parent = parent ? parent : &sysctl_table_root.default_set;
2165         p->is_seen = is_seen;
2166 }
2167
2168 #else /* !CONFIG_SYSCTL */
2169 struct ctl_table_header *register_sysctl_table(struct ctl_table * table)
2170 {
2171         return NULL;
2172 }
2173
2174 struct ctl_table_header *register_sysctl_paths(const struct ctl_path *path,
2175                                                     struct ctl_table *table)
2176 {
2177         return NULL;
2178 }
2179
2180 void unregister_sysctl_table(struct ctl_table_header * table)
2181 {
2182 }
2183
2184 void setup_sysctl_set(struct ctl_table_set *p,
2185         struct ctl_table_set *parent,
2186         int (*is_seen)(struct ctl_table_set *))
2187 {
2188 }
2189
2190 void sysctl_head_put(struct ctl_table_header *head)
2191 {
2192 }
2193
2194 #endif /* CONFIG_SYSCTL */
2195
2196 /*
2197  * /proc/sys support
2198  */
2199
2200 #ifdef CONFIG_PROC_SYSCTL
2201
2202 static int _proc_do_string(void* data, int maxlen, int write,
2203                            struct file *filp, void __user *buffer,
2204                            size_t *lenp, loff_t *ppos)
2205 {
2206         size_t len;
2207         char __user *p;
2208         char c;
2209
2210         if (!data || !maxlen || !*lenp) {
2211                 *lenp = 0;
2212                 return 0;
2213         }
2214
2215         if (write) {
2216                 len = 0;
2217                 p = buffer;
2218                 while (len < *lenp) {
2219                         if (get_user(c, p++))
2220                                 return -EFAULT;
2221                         if (c == 0 || c == '\n')
2222                                 break;
2223                         len++;
2224                 }
2225                 if (len >= maxlen)
2226                         len = maxlen-1;
2227                 if(copy_from_user(data, buffer, len))
2228                         return -EFAULT;
2229                 ((char *) data)[len] = 0;
2230                 *ppos += *lenp;
2231         } else {
2232                 len = strlen(data);
2233                 if (len > maxlen)
2234                         len = maxlen;
2235
2236                 if (*ppos > len) {
2237                         *lenp = 0;
2238                         return 0;
2239                 }
2240
2241                 data += *ppos;
2242                 len  -= *ppos;
2243
2244                 if (len > *lenp)
2245                         len = *lenp;
2246                 if (len)
2247                         if(copy_to_user(buffer, data, len))
2248                                 return -EFAULT;
2249                 if (len < *lenp) {
2250                         if(put_user('\n', ((char __user *) buffer) + len))
2251                                 return -EFAULT;
2252                         len++;
2253                 }
2254                 *lenp = len;
2255                 *ppos += len;
2256         }
2257         return 0;
2258 }
2259
2260 /**
2261  * proc_dostring - read a string sysctl
2262  * @table: the sysctl table
2263  * @write: %TRUE if this is a write to the sysctl file
2264  * @filp: the file structure
2265  * @buffer: the user buffer
2266  * @lenp: the size of the user buffer
2267  * @ppos: file position
2268  *
2269  * Reads/writes a string from/to the user buffer. If the kernel
2270  * buffer provided is not large enough to hold the string, the
2271  * string is truncated. The copied string is %NULL-terminated.
2272  * If the string is being read by the user process, it is copied
2273  * and a newline '\n' is added. It is truncated if the buffer is
2274  * not large enough.
2275  *
2276  * Returns 0 on success.
2277  */
2278 int proc_dostring(struct ctl_table *table, int write, struct file *filp,
2279                   void __user *buffer, size_t *lenp, loff_t *ppos)
2280 {
2281         return _proc_do_string(table->data, table->maxlen, write, filp,
2282                                buffer, lenp, ppos);
2283 }
2284
2285
2286 static int do_proc_dointvec_conv(int *negp, unsigned long *lvalp,
2287                                  int *valp,
2288                                  int write, void *data)
2289 {
2290         if (write) {
2291                 *valp = *negp ? -*lvalp : *lvalp;
2292         } else {
2293                 int val = *valp;
2294                 if (val < 0) {
2295                         *negp = -1;
2296                         *lvalp = (unsigned long)-val;
2297                 } else {
2298                         *negp = 0;
2299                         *lvalp = (unsigned long)val;
2300                 }
2301         }
2302         return 0;
2303 }
2304
2305 static int __do_proc_dointvec(void *tbl_data, struct ctl_table *table,
2306                   int write, struct file *filp, void __user *buffer,
2307                   size_t *lenp, loff_t *ppos,
2308                   int (*conv)(int *negp, unsigned long *lvalp, int *valp,
2309                               int write, void *data),
2310                   void *data)
2311 {
2312 #define TMPBUFLEN 21
2313         int *i, vleft, first = 1, neg;
2314         unsigned long lval;
2315         size_t left, len;
2316         
2317         char buf[TMPBUFLEN], *p;
2318         char __user *s = buffer;
2319         
2320         if (!tbl_data || !table->maxlen || !*lenp ||
2321             (*ppos && !write)) {
2322                 *lenp = 0;
2323                 return 0;
2324         }
2325         
2326         i = (int *) tbl_data;
2327         vleft = table->maxlen / sizeof(*i);
2328         left = *lenp;
2329
2330         if (!conv)
2331                 conv = do_proc_dointvec_conv;
2332
2333         for (; left && vleft--; i++, first=0) {
2334                 if (write) {
2335                         while (left) {
2336                                 char c;
2337                                 if (get_user(c, s))
2338                                         return -EFAULT;
2339                                 if (!isspace(c))
2340                                         break;
2341                                 left--;
2342                                 s++;
2343                         }
2344                         if (!left)
2345                                 break;
2346                         neg = 0;
2347                         len = left;
2348                         if (len > sizeof(buf) - 1)
2349                                 len = sizeof(buf) - 1;
2350                         if (copy_from_user(buf, s, len))
2351                                 return -EFAULT;
2352                         buf[len] = 0;
2353                         p = buf;
2354                         if (*p == '-' && left > 1) {
2355                                 neg = 1;
2356                                 p++;
2357                         }
2358                         if (*p < '0' || *p > '9')
2359                                 break;
2360
2361                         lval = simple_strtoul(p, &p, 0);
2362
2363                         len = p-buf;
2364                         if ((len < left) && *p && !isspace(*p))
2365                                 break;
2366                         s += len;
2367                         left -= len;
2368
2369                         if (conv(&neg, &lval, i, 1, data))
2370                                 break;
2371                 } else {
2372                         p = buf;
2373                         if (!first)
2374                                 *p++ = '\t';
2375         
2376                         if (conv(&neg, &lval, i, 0, data))
2377                                 break;
2378
2379                         sprintf(p, "%s%lu", neg ? "-" : "", lval);
2380                         len = strlen(buf);
2381                         if (len > left)
2382                                 len = left;
2383                         if(copy_to_user(s, buf, len))
2384                                 return -EFAULT;
2385                         left -= len;
2386                         s += len;
2387                 }
2388         }
2389
2390         if (!write && !first && left) {
2391                 if(put_user('\n', s))
2392                         return -EFAULT;
2393                 left--, s++;
2394         }
2395         if (write) {
2396                 while (left) {
2397                         char c;
2398                         if (get_user(c, s++))
2399                                 return -EFAULT;
2400                         if (!isspace(c))
2401                                 break;
2402                         left--;
2403                 }
2404         }
2405         if (write && first)
2406                 return -EINVAL;
2407         *lenp -= left;
2408         *ppos += *lenp;
2409         return 0;
2410 #undef TMPBUFLEN
2411 }
2412
2413 static int do_proc_dointvec(struct ctl_table *table, int write, struct file *filp,
2414                   void __user *buffer, size_t *lenp, loff_t *ppos,
2415                   int (*conv)(int *negp, unsigned long *lvalp, int *valp,
2416                               int write, void *data),
2417                   void *data)
2418 {
2419         return __do_proc_dointvec(table->data, table, write, filp,
2420                         buffer, lenp, ppos, conv, data);
2421 }
2422
2423 /**
2424  * proc_dointvec - read a vector of integers
2425  * @table: the sysctl table
2426  * @write: %TRUE if this is a write to the sysctl file
2427  * @filp: the file structure
2428  * @buffer: the user buffer
2429  * @lenp: the size of the user buffer
2430  * @ppos: file position
2431  *
2432  * Reads/writes up to table->maxlen/sizeof(unsigned int) integer
2433  * values from/to the user buffer, treated as an ASCII string. 
2434  *
2435  * Returns 0 on success.
2436  */
2437 int proc_dointvec(struct ctl_table *table, int write, struct file *filp,
2438                      void __user *buffer, size_t *lenp, loff_t *ppos)
2439 {
2440     return do_proc_dointvec(table,write,filp,buffer,lenp,ppos,
2441                             NULL,NULL);
2442 }
2443
2444 /*
2445  * Taint values can only be increased
2446  * This means we can safely use a temporary.
2447  */
2448 static int proc_taint(struct ctl_table *table, int write, struct file *filp,
2449                                void __user *buffer, size_t *lenp, loff_t *ppos)
2450 {
2451         struct ctl_table t;
2452         unsigned long tmptaint = get_taint();
2453         int err;
2454
2455         if (write && !capable(CAP_SYS_ADMIN))
2456                 return -EPERM;
2457
2458         t = *table;
2459         t.data = &tmptaint;
2460         err = proc_doulongvec_minmax(&t, write, filp, buffer, lenp, ppos);
2461         if (err < 0)
2462                 return err;
2463
2464         if (write) {
2465                 /*
2466                  * Poor man's atomic or. Not worth adding a primitive
2467                  * to everyone's atomic.h for this
2468                  */
2469                 int i;
2470                 for (i = 0; i < BITS_PER_LONG && tmptaint >> i; i++) {
2471                         if ((tmptaint >> i) & 1)
2472                                 add_taint(i);
2473                 }
2474         }
2475
2476         return err;
2477 }
2478
2479 struct do_proc_dointvec_minmax_conv_param {
2480         int *min;
2481         int *max;
2482 };
2483
2484 static int do_proc_dointvec_minmax_conv(int *negp, unsigned long *lvalp, 
2485                                         int *valp, 
2486                                         int write, void *data)
2487 {
2488         struct do_proc_dointvec_minmax_conv_param *param = data;
2489         if (write) {
2490                 int val = *negp ? -*lvalp : *lvalp;
2491                 if ((param->min && *param->min > val) ||
2492                     (param->max && *param->max < val))
2493                         return -EINVAL;
2494                 *valp = val;
2495         } else {
2496                 int val = *valp;
2497                 if (val < 0) {
2498                         *negp = -1;
2499                         *lvalp = (unsigned long)-val;
2500                 } else {
2501                         *negp = 0;
2502                         *lvalp = (unsigned long)val;
2503                 }
2504         }
2505         return 0;
2506 }
2507
2508 /**
2509  * proc_dointvec_minmax - read a vector of integers with min/max values
2510  * @table: the sysctl table
2511  * @write: %TRUE if this is a write to the sysctl file
2512  * @filp: the file structure
2513  * @buffer: the user buffer
2514  * @lenp: the size of the user buffer
2515  * @ppos: file position
2516  *
2517  * Reads/writes up to table->maxlen/sizeof(unsigned int) integer
2518  * values from/to the user buffer, treated as an ASCII string.
2519  *
2520  * This routine will ensure the values are within the range specified by
2521  * table->extra1 (min) and table->extra2 (max).
2522  *
2523  * Returns 0 on success.
2524  */
2525 int proc_dointvec_minmax(struct ctl_table *table, int write, struct file *filp,
2526                   void __user *buffer, size_t *lenp, loff_t *ppos)
2527 {
2528         struct do_proc_dointvec_minmax_conv_param param = {
2529                 .min = (int *) table->extra1,
2530                 .max = (int *) table->extra2,
2531         };
2532         return do_proc_dointvec(table, write, filp, buffer, lenp, ppos,
2533                                 do_proc_dointvec_minmax_conv, &param);
2534 }
2535
2536 static int __do_proc_doulongvec_minmax(void *data, struct ctl_table *table, int write,
2537                                      struct file *filp,
2538                                      void __user *buffer,
2539                                      size_t *lenp, loff_t *ppos,
2540                                      unsigned long convmul,
2541                                      unsigned long convdiv)
2542 {
2543 #define TMPBUFLEN 21
2544         unsigned long *i, *min, *max, val;
2545         int vleft, first=1, neg;
2546         size_t len, left;
2547         char buf[TMPBUFLEN], *p;
2548         char __user *s = buffer;
2549         
2550         if (!data || !table->maxlen || !*lenp ||
2551             (*ppos && !write)) {
2552                 *lenp = 0;
2553                 return 0;
2554         }
2555         
2556         i = (unsigned long *) data;
2557         min = (unsigned long *) table->extra1;
2558         max = (unsigned long *) table->extra2;
2559         vleft = table->maxlen / sizeof(unsigned long);
2560         left = *lenp;
2561         
2562         for (; left && vleft--; i++, min++, max++, first=0) {
2563                 if (write) {
2564                         while (left) {
2565                                 char c;
2566                                 if (get_user(c, s))
2567                                         return -EFAULT;
2568                                 if (!isspace(c))
2569                                         break;
2570                                 left--;
2571                                 s++;
2572                         }
2573                         if (!left)
2574                                 break;
2575                         neg = 0;
2576                         len = left;
2577                         if (len > TMPBUFLEN-1)
2578                                 len = TMPBUFLEN-1;
2579                         if (copy_from_user(buf, s, len))
2580                                 return -EFAULT;
2581                         buf[len] = 0;
2582                         p = buf;
2583                         if (*p == '-' && left > 1) {
2584                                 neg = 1;
2585                                 p++;
2586                         }
2587                         if (*p < '0' || *p > '9')
2588                                 break;
2589                         val = simple_strtoul(p, &p, 0) * convmul / convdiv ;
2590                         len = p-buf;
2591                         if ((len < left) && *p && !isspace(*p))
2592                                 break;
2593                         if (neg)
2594                                 val = -val;
2595                         s += len;
2596                         left -= len;
2597
2598                         if(neg)
2599                                 continue;
2600                         if ((min && val < *min) || (max && val > *max))
2601                                 continue;
2602                         *i = val;
2603                 } else {
2604                         p = buf;
2605                         if (!first)
2606                                 *p++ = '\t';
2607                         sprintf(p, "%lu", convdiv * (*i) / convmul);
2608                         len = strlen(buf);
2609                         if (len > left)
2610                                 len = left;
2611                         if(copy_to_user(s, buf, len))
2612                                 return -EFAULT;
2613                         left -= len;
2614                         s += len;
2615                 }
2616         }
2617
2618         if (!write && !first && left) {
2619                 if(put_user('\n', s))
2620                         return -EFAULT;
2621                 left--, s++;
2622         }
2623         if (write) {
2624                 while (left) {
2625                         char c;
2626                         if (get_user(c, s++))
2627                                 return -EFAULT;
2628                         if (!isspace(c))
2629                                 break;
2630                         left--;
2631                 }
2632         }
2633         if (write && first)
2634                 return -EINVAL;
2635         *lenp -= left;
2636         *ppos += *lenp;
2637         return 0;
2638 #undef TMPBUFLEN
2639 }
2640
2641 static int do_proc_doulongvec_minmax(struct ctl_table *table, int write,
2642                                      struct file *filp,
2643                                      void __user *buffer,
2644                                      size_t *lenp, loff_t *ppos,
2645                                      unsigned long convmul,
2646                                      unsigned long convdiv)
2647 {
2648         return __do_proc_doulongvec_minmax(table->data, table, write,
2649                         filp, buffer, lenp, ppos, convmul, convdiv);
2650 }
2651
2652 /**
2653  * proc_doulongvec_minmax - read a vector of long integers with min/max values
2654  * @table: the sysctl table
2655  * @write: %TRUE if this is a write to the sysctl file
2656  * @filp: the file structure
2657  * @buffer: the user buffer
2658  * @lenp: the size of the user buffer
2659  * @ppos: file position
2660  *
2661  * Reads/writes up to table->maxlen/sizeof(unsigned long) unsigned long
2662  * values from/to the user buffer, treated as an ASCII string.
2663  *
2664  * This routine will ensure the values are within the range specified by
2665  * table->extra1 (min) and table->extra2 (max).
2666  *
2667  * Returns 0 on success.
2668  */
2669 int proc_doulongvec_minmax(struct ctl_table *table, int write, struct file *filp,
2670                            void __user *buffer, size_t *lenp, loff_t *ppos)
2671 {
2672     return do_proc_doulongvec_minmax(table, write, filp, buffer, lenp, ppos, 1l, 1l);
2673 }
2674
2675 /**
2676  * proc_doulongvec_ms_jiffies_minmax - read a vector of millisecond values with min/max values
2677  * @table: the sysctl table
2678  * @write: %TRUE if this is a write to the sysctl file
2679  * @filp: the file structure
2680  * @buffer: the user buffer
2681  * @lenp: the size of the user buffer
2682  * @ppos: file position
2683  *
2684  * Reads/writes up to table->maxlen/sizeof(unsigned long) unsigned long
2685  * values from/to the user buffer, treated as an ASCII string. The values
2686  * are treated as milliseconds, and converted to jiffies when they are stored.
2687  *
2688  * This routine will ensure the values are within the range specified by
2689  * table->extra1 (min) and table->extra2 (max).
2690  *
2691  * Returns 0 on success.
2692  */
2693 int proc_doulongvec_ms_jiffies_minmax(struct ctl_table *table, int write,
2694                                       struct file *filp,
2695                                       void __user *buffer,
2696                                       size_t *lenp, loff_t *ppos)
2697 {
2698     return do_proc_doulongvec_minmax(table, write, filp, buffer,
2699                                      lenp, ppos, HZ, 1000l);
2700 }
2701
2702
2703 static int do_proc_dointvec_jiffies_conv(int *negp, unsigned long *lvalp,
2704                                          int *valp,
2705                                          int write, void *data)
2706 {
2707         if (write) {
2708                 if (*lvalp > LONG_MAX / HZ)
2709                         return 1;
2710                 *valp = *negp ? -(*lvalp*HZ) : (*lvalp*HZ);
2711         } else {
2712                 int val = *valp;
2713                 unsigned long lval;
2714                 if (val < 0) {
2715                         *negp = -1;
2716                         lval = (unsigned long)-val;
2717                 } else {
2718                         *negp = 0;
2719                         lval = (unsigned long)val;
2720                 }
2721                 *lvalp = lval / HZ;
2722         }
2723         return 0;
2724 }
2725
2726 static int do_proc_dointvec_userhz_jiffies_conv(int *negp, unsigned long *lvalp,
2727                                                 int *valp,
2728                                                 int write, void *data)
2729 {
2730         if (write) {
2731                 if (USER_HZ < HZ && *lvalp > (LONG_MAX / HZ) * USER_HZ)
2732                         return 1;
2733                 *valp = clock_t_to_jiffies(*negp ? -*lvalp : *lvalp);
2734         } else {
2735                 int val = *valp;
2736                 unsigned long lval;
2737                 if (val < 0) {
2738                         *negp = -1;
2739                         lval = (unsigned long)-val;
2740                 } else {
2741                         *negp = 0;
2742                         lval = (unsigned long)val;
2743                 }
2744                 *lvalp = jiffies_to_clock_t(lval);
2745         }
2746         return 0;
2747 }
2748
2749 static int do_proc_dointvec_ms_jiffies_conv(int *negp, unsigned long *lvalp,
2750                                             int *valp,
2751                                             int write, void *data)
2752 {
2753         if (write) {
2754                 *valp = msecs_to_jiffies(*negp ? -*lvalp : *lvalp);
2755         } else {
2756                 int val = *valp;
2757                 unsigned long lval;
2758                 if (val < 0) {
2759                         *negp = -1;
2760                         lval = (unsigned long)-val;
2761                 } else {
2762                         *negp = 0;
2763                         lval = (unsigned long)val;
2764                 }
2765                 *lvalp = jiffies_to_msecs(lval);
2766         }
2767         return 0;
2768 }
2769
2770 /**
2771  * proc_dointvec_jiffies - read a vector of integers as seconds
2772  * @table: the sysctl table
2773  * @write: %TRUE if this is a write to the sysctl file
2774  * @filp: the file structure
2775  * @buffer: the user buffer
2776  * @lenp: the size of the user buffer
2777  * @ppos: file position
2778  *
2779  * Reads/writes up to table->maxlen/sizeof(unsigned int) integer
2780  * values from/to the user buffer, treated as an ASCII string. 
2781  * The values read are assumed to be in seconds, and are converted into
2782  * jiffies.
2783  *
2784  * Returns 0 on success.
2785  */
2786 int proc_dointvec_jiffies(struct ctl_table *table, int write, struct file *filp,
2787                           void __user *buffer, size_t *lenp, loff_t *ppos)
2788 {
2789     return do_proc_dointvec(table,write,filp,buffer,lenp,ppos,
2790                             do_proc_dointvec_jiffies_conv,NULL);
2791 }
2792
2793 /**
2794  * proc_dointvec_userhz_jiffies - read a vector of integers as 1/USER_HZ seconds
2795  * @table: the sysctl table
2796  * @write: %TRUE if this is a write to the sysctl file
2797  * @filp: the file structure
2798  * @buffer: the user buffer
2799  * @lenp: the size of the user buffer
2800  * @ppos: pointer to the file position
2801  *
2802  * Reads/writes up to table->maxlen/sizeof(unsigned int) integer
2803  * values from/to the user buffer, treated as an ASCII string. 
2804  * The values read are assumed to be in 1/USER_HZ seconds, and 
2805  * are converted into jiffies.
2806  *
2807  * Returns 0 on success.
2808  */
2809 int proc_dointvec_userhz_jiffies(struct ctl_table *table, int write, struct file *filp,
2810                                  void __user *buffer, size_t *lenp, loff_t *ppos)
2811 {
2812     return do_proc_dointvec(table,write,filp,buffer,lenp,ppos,
2813                             do_proc_dointvec_userhz_jiffies_conv,NULL);
2814 }
2815
2816 /**
2817  * proc_dointvec_ms_jiffies - read a vector of integers as 1 milliseconds
2818  * @table: the sysctl table
2819  * @write: %TRUE if this is a write to the sysctl file
2820  * @filp: the file structure
2821  * @buffer: the user buffer
2822  * @lenp: the size of the user buffer
2823  * @ppos: file position
2824  * @ppos: the current position in the file
2825  *
2826  * Reads/writes up to table->maxlen/sizeof(unsigned int) integer
2827  * values from/to the user buffer, treated as an ASCII string. 
2828  * The values read are assumed to be in 1/1000 seconds, and 
2829  * are converted into jiffies.
2830  *
2831  * Returns 0 on success.
2832  */
2833 int proc_dointvec_ms_jiffies(struct ctl_table *table, int write, struct file *filp,
2834                              void __user *buffer, size_t *lenp, loff_t *ppos)
2835 {
2836         return do_proc_dointvec(table, write, filp, buffer, lenp, ppos,
2837                                 do_proc_dointvec_ms_jiffies_conv, NULL);
2838 }
2839
2840 static int proc_do_cad_pid(struct ctl_table *table, int write, struct file *filp,
2841                            void __user *buffer, size_t *lenp, loff_t *ppos)
2842 {
2843         struct pid *new_pid;
2844         pid_t tmp;
2845         int r;
2846
2847         tmp = pid_vnr(cad_pid);
2848
2849         r = __do_proc_dointvec(&tmp, table, write, filp, buffer,
2850                                lenp, ppos, NULL, NULL);
2851         if (r || !write)
2852                 return r;
2853
2854         new_pid = find_get_pid(tmp);
2855         if (!new_pid)
2856                 return -ESRCH;
2857
2858         put_pid(xchg(&cad_pid, new_pid));
2859         return 0;
2860 }
2861
2862 #else /* CONFIG_PROC_FS */
2863
2864 int proc_dostring(struct ctl_table *table, int write, struct file *filp,
2865                   void __user *buffer, size_t *lenp, loff_t *ppos)
2866 {
2867         return -ENOSYS;
2868 }
2869
2870 int proc_dointvec(struct ctl_table *table, int write, struct file *filp,
2871                   void __user *buffer, size_t *lenp, loff_t *ppos)
2872 {
2873         return -ENOSYS;
2874 }
2875
2876 int proc_dointvec_minmax(struct ctl_table *table, int write, struct file *filp,
2877                     void __user *buffer, size_t *lenp, loff_t *ppos)
2878 {
2879         return -ENOSYS;
2880 }
2881
2882 int proc_dointvec_jiffies(struct ctl_table *table, int write, struct file *filp,
2883                     void __user *buffer, size_t *lenp, loff_t *ppos)
2884 {
2885         return -ENOSYS;
2886 }
2887
2888 int proc_dointvec_userhz_jiffies(struct ctl_table *table, int write, struct file *filp,
2889                     void __user *buffer, size_t *lenp, loff_t *ppos)
2890 {
2891         return -ENOSYS;
2892 }
2893
2894 int proc_dointvec_ms_jiffies(struct ctl_table *table, int write, struct file *filp,
2895                              void __user *buffer, size_t *lenp, loff_t *ppos)
2896 {
2897         return -ENOSYS;
2898 }
2899
2900 int proc_doulongvec_minmax(struct ctl_table *table, int write, struct file *filp,
2901                     void __user *buffer, size_t *lenp, loff_t *ppos)
2902 {
2903         return -ENOSYS;
2904 }
2905
2906 int proc_doulongvec_ms_jiffies_minmax(struct ctl_table *table, int write,
2907                                       struct file *filp,
2908                                       void __user *buffer,
2909                                       size_t *lenp, loff_t *ppos)
2910 {
2911     return -ENOSYS;
2912 }
2913
2914
2915 #endif /* CONFIG_PROC_FS */
2916
2917
2918 #ifdef CONFIG_SYSCTL_SYSCALL
2919 /*
2920  * General sysctl support routines 
2921  */
2922
2923 /* The generic sysctl data routine (used if no strategy routine supplied) */
2924 int sysctl_data(struct ctl_table *table,
2925                 void __user *oldval, size_t __user *oldlenp,
2926                 void __user *newval, size_t newlen)
2927 {
2928         size_t len;
2929
2930         /* Get out of I don't have a variable */
2931         if (!table->data || !table->maxlen)
2932                 return -ENOTDIR;
2933
2934         if (oldval && oldlenp) {
2935                 if (get_user(len, oldlenp))
2936                         return -EFAULT;
2937                 if (len) {
2938                         if (len > table->maxlen)
2939                                 len = table->maxlen;
2940                         if (copy_to_user(oldval, table->data, len))
2941                                 return -EFAULT;
2942                         if (put_user(len, oldlenp))
2943                                 return -EFAULT;
2944                 }
2945         }
2946
2947         if (newval && newlen) {
2948                 if (newlen > table->maxlen)
2949                         newlen = table->maxlen;
2950
2951                 if (copy_from_user(table->data, newval, newlen))
2952                         return -EFAULT;
2953         }
2954         return 1;
2955 }
2956
2957 /* The generic string strategy routine: */
2958 int sysctl_string(struct ctl_table *table,
2959                   void __user *oldval, size_t __user *oldlenp,
2960                   void __user *newval, size_t newlen)
2961 {
2962         if (!table->data || !table->maxlen) 
2963                 return -ENOTDIR;
2964         
2965         if (oldval && oldlenp) {
2966                 size_t bufsize;
2967                 if (get_user(bufsize, oldlenp))
2968                         return -EFAULT;
2969                 if (bufsize) {
2970                         size_t len = strlen(table->data), copied;
2971
2972                         /* This shouldn't trigger for a well-formed sysctl */
2973                         if (len > table->maxlen)
2974                                 len = table->maxlen;
2975
2976                         /* Copy up to a max of bufsize-1 bytes of the string */
2977                         copied = (len >= bufsize) ? bufsize - 1 : len;
2978
2979                         if (copy_to_user(oldval, table->data, copied) ||
2980                             put_user(0, (char __user *)(oldval + copied)))
2981                                 return -EFAULT;
2982                         if (put_user(len, oldlenp))
2983                                 return -EFAULT;
2984                 }
2985         }
2986         if (newval && newlen) {
2987                 size_t len = newlen;
2988                 if (len > table->maxlen)
2989                         len = table->maxlen;
2990                 if(copy_from_user(table->data, newval, len))
2991                         return -EFAULT;
2992                 if (len == table->maxlen)
2993                         len--;
2994                 ((char *) table->data)[len] = 0;
2995         }
2996         return 1;
2997 }
2998
2999 /*
3000  * This function makes sure that all of the integers in the vector
3001  * are between the minimum and maximum values given in the arrays
3002  * table->extra1 and table->extra2, respectively.
3003  */
3004 int sysctl_intvec(struct ctl_table *table,
3005                 void __user *oldval, size_t __user *oldlenp,
3006                 void __user *newval, size_t newlen)
3007 {
3008
3009         if (newval && newlen) {
3010                 int __user *vec = (int __user *) newval;
3011                 int *min = (int *) table->extra1;
3012                 int *max = (int *) table->extra2;
3013                 size_t length;
3014                 int i;
3015
3016                 if (newlen % sizeof(int) != 0)
3017                         return -EINVAL;
3018
3019                 if (!table->extra1 && !table->extra2)
3020                         return 0;
3021
3022                 if (newlen > table->maxlen)
3023                         newlen = table->maxlen;
3024                 length = newlen / sizeof(int);
3025
3026                 for (i = 0; i < length; i++) {
3027                         int value;
3028                         if (get_user(value, vec + i))
3029                                 return -EFAULT;
3030                         if (min && value < min[i])
3031                                 return -EINVAL;
3032                         if (max && value > max[i])
3033                                 return -EINVAL;
3034                 }
3035         }
3036         return 0;
3037 }
3038
3039 /* Strategy function to convert jiffies to seconds */ 
3040 int sysctl_jiffies(struct ctl_table *table,
3041                 void __user *oldval, size_t __user *oldlenp,
3042                 void __user *newval, size_t newlen)
3043 {
3044         if (oldval && oldlenp) {
3045                 size_t olen;
3046
3047                 if (get_user(olen, oldlenp))
3048                         return -EFAULT;
3049                 if (olen) {
3050                         int val;
3051
3052                         if (olen < sizeof(int))
3053                                 return -EINVAL;
3054
3055                         val = *(int *)(table->data) / HZ;
3056                         if (put_user(val, (int __user *)oldval))
3057                                 return -EFAULT;
3058                         if (put_user(sizeof(int), oldlenp))
3059                                 return -EFAULT;
3060                 }
3061         }
3062         if (newval && newlen) { 
3063                 int new;
3064                 if (newlen != sizeof(int))
3065                         return -EINVAL; 
3066                 if (get_user(new, (int __user *)newval))
3067                         return -EFAULT;
3068                 *(int *)(table->data) = new*HZ; 
3069         }
3070         return 1;
3071 }
3072
3073 /* Strategy function to convert jiffies to seconds */ 
3074 int sysctl_ms_jiffies(struct ctl_table *table,
3075                 void __user *oldval, size_t __user *oldlenp,
3076                 void __user *newval, size_t newlen)
3077 {
3078         if (oldval && oldlenp) {
3079                 size_t olen;
3080
3081                 if (get_user(olen, oldlenp))
3082                         return -EFAULT;
3083                 if (olen) {
3084                         int val;
3085
3086                         if (olen < sizeof(int))
3087                                 return -EINVAL;
3088
3089                         val = jiffies_to_msecs(*(int *)(table->data));
3090                         if (put_user(val, (int __user *)oldval))
3091                                 return -EFAULT;
3092                         if (put_user(sizeof(int), oldlenp))
3093                                 return -EFAULT;
3094                 }
3095         }
3096         if (newval && newlen) { 
3097                 int new;
3098                 if (newlen != sizeof(int))
3099                         return -EINVAL; 
3100                 if (get_user(new, (int __user *)newval))
3101                         return -EFAULT;
3102                 *(int *)(table->data) = msecs_to_jiffies(new);
3103         }
3104         return 1;
3105 }
3106
3107
3108
3109 #else /* CONFIG_SYSCTL_SYSCALL */
3110
3111
3112 SYSCALL_DEFINE1(sysctl, struct __sysctl_args __user *, args)
3113 {
3114         struct __sysctl_args tmp;
3115         int error;
3116
3117         if (copy_from_user(&tmp, args, sizeof(tmp)))
3118                 return -EFAULT;
3119
3120         error = deprecated_sysctl_warning(&tmp);
3121
3122         /* If no error reading the parameters then just -ENOSYS ... */
3123         if (!error)
3124                 error = -ENOSYS;
3125
3126         return error;
3127 }
3128
3129 int sysctl_data(struct ctl_table *table,
3130                   void __user *oldval, size_t __user *oldlenp,
3131                   void __user *newval, size_t newlen)
3132 {
3133         return -ENOSYS;
3134 }
3135
3136 int sysctl_string(struct ctl_table *table,
3137                   void __user *oldval, size_t __user *oldlenp,
3138                   void __user *newval, size_t newlen)
3139 {
3140         return -ENOSYS;
3141 }
3142
3143 int sysctl_intvec(struct ctl_table *table,
3144                 void __user *oldval, size_t __user *oldlenp,
3145                 void __user *newval, size_t newlen)
3146 {
3147         return -ENOSYS;
3148 }
3149
3150 int sysctl_jiffies(struct ctl_table *table,
3151                 void __user *oldval, size_t __user *oldlenp,
3152                 void __user *newval, size_t newlen)
3153 {
3154         return -ENOSYS;
3155 }
3156
3157 int sysctl_ms_jiffies(struct ctl_table *table,
3158                 void __user *oldval, size_t __user *oldlenp,
3159                 void __user *newval, size_t newlen)
3160 {
3161         return -ENOSYS;
3162 }
3163
3164 #endif /* CONFIG_SYSCTL_SYSCALL */
3165
3166 static int deprecated_sysctl_warning(struct __sysctl_args *args)
3167 {
3168         static int msg_count;
3169         int name[CTL_MAXNAME];
3170         int i;
3171
3172         /* Check args->nlen. */
3173         if (args->nlen < 0 || args->nlen > CTL_MAXNAME)
3174                 return -ENOTDIR;
3175
3176         /* Read in the sysctl name for better debug message logging */
3177         for (i = 0; i < args->nlen; i++)
3178                 if (get_user(name[i], args->name + i))
3179                         return -EFAULT;
3180
3181         /* Ignore accesses to kernel.version */
3182         if ((args->nlen == 2) && (name[0] == CTL_KERN) && (name[1] == KERN_VERSION))
3183                 return 0;
3184
3185         if (msg_count < 5) {
3186                 msg_count++;
3187                 printk(KERN_INFO
3188                         "warning: process `%s' used the deprecated sysctl "
3189                         "system call with ", current->comm);
3190                 for (i = 0; i < args->nlen; i++)
3191                         printk("%d.", name[i]);
3192                 printk("\n");
3193         }
3194         return 0;
3195 }
3196
3197 /*
3198  * No sense putting this after each symbol definition, twice,
3199  * exception granted :-)
3200  */
3201 EXPORT_SYMBOL(proc_dointvec);
3202 EXPORT_SYMBOL(proc_dointvec_jiffies);
3203 EXPORT_SYMBOL(proc_dointvec_minmax);
3204 EXPORT_SYMBOL(proc_dointvec_userhz_jiffies);
3205 EXPORT_SYMBOL(proc_dointvec_ms_jiffies);
3206 EXPORT_SYMBOL(proc_dostring);
3207 EXPORT_SYMBOL(proc_doulongvec_minmax);
3208 EXPORT_SYMBOL(proc_doulongvec_ms_jiffies_minmax);
3209 EXPORT_SYMBOL(register_sysctl_table);
3210 EXPORT_SYMBOL(register_sysctl_paths);
3211 EXPORT_SYMBOL(sysctl_intvec);
3212 EXPORT_SYMBOL(sysctl_jiffies);
3213 EXPORT_SYMBOL(sysctl_ms_jiffies);
3214 EXPORT_SYMBOL(sysctl_string);
3215 EXPORT_SYMBOL(sysctl_data);
3216 EXPORT_SYMBOL(unregister_sysctl_table);