Merge commit 'v2.6.30-rc6' into perfcounters/core
[linux-2.6.git] / kernel / sysctl.c
1 /*
2  * sysctl.c: General linux system control interface
3  *
4  * Begun 24 March 1995, Stephen Tweedie
5  * Added /proc support, Dec 1995
6  * Added bdflush entry and intvec min/max checking, 2/23/96, Tom Dyas.
7  * Added hooks for /proc/sys/net (minor, minor patch), 96/4/1, Mike Shaver.
8  * Added kernel/java-{interpreter,appletviewer}, 96/5/10, Mike Shaver.
9  * Dynamic registration fixes, Stephen Tweedie.
10  * Added kswapd-interval, ctrl-alt-del, printk stuff, 1/8/97, Chris Horn.
11  * Made sysctl support optional via CONFIG_SYSCTL, 1/10/97, Chris
12  *  Horn.
13  * Added proc_doulongvec_ms_jiffies_minmax, 09/08/99, Carlos H. Bauer.
14  * Added proc_doulongvec_minmax, 09/08/99, Carlos H. Bauer.
15  * Changed linked lists to use list.h instead of lists.h, 02/24/00, Bill
16  *  Wendling.
17  * The list_for_each() macro wasn't appropriate for the sysctl loop.
18  *  Removed it and replaced it with older style, 03/23/00, Bill Wendling
19  */
20
21 #include <linux/module.h>
22 #include <linux/mm.h>
23 #include <linux/swap.h>
24 #include <linux/slab.h>
25 #include <linux/sysctl.h>
26 #include <linux/proc_fs.h>
27 #include <linux/security.h>
28 #include <linux/ctype.h>
29 #include <linux/utsname.h>
30 #include <linux/smp_lock.h>
31 #include <linux/fs.h>
32 #include <linux/init.h>
33 #include <linux/kernel.h>
34 #include <linux/kobject.h>
35 #include <linux/net.h>
36 #include <linux/sysrq.h>
37 #include <linux/highuid.h>
38 #include <linux/writeback.h>
39 #include <linux/hugetlb.h>
40 #include <linux/initrd.h>
41 #include <linux/key.h>
42 #include <linux/times.h>
43 #include <linux/limits.h>
44 #include <linux/dcache.h>
45 #include <linux/syscalls.h>
46 #include <linux/vmstat.h>
47 #include <linux/nfs_fs.h>
48 #include <linux/acpi.h>
49 #include <linux/reboot.h>
50 #include <linux/ftrace.h>
51 #include <linux/slow-work.h>
52 #include <linux/perf_counter.h>
53
54 #include <asm/uaccess.h>
55 #include <asm/processor.h>
56
57 #ifdef CONFIG_X86
58 #include <asm/nmi.h>
59 #include <asm/stacktrace.h>
60 #include <asm/io.h>
61 #endif
62
63 static int deprecated_sysctl_warning(struct __sysctl_args *args);
64
65 #if defined(CONFIG_SYSCTL)
66
67 /* External variables not in a header file. */
68 extern int C_A_D;
69 extern int print_fatal_signals;
70 extern int sysctl_overcommit_memory;
71 extern int sysctl_overcommit_ratio;
72 extern int sysctl_panic_on_oom;
73 extern int sysctl_oom_kill_allocating_task;
74 extern int sysctl_oom_dump_tasks;
75 extern int max_threads;
76 extern int core_uses_pid;
77 extern int suid_dumpable;
78 extern char core_pattern[];
79 extern int pid_max;
80 extern int min_free_kbytes;
81 extern int pid_max_min, pid_max_max;
82 extern int sysctl_drop_caches;
83 extern int percpu_pagelist_fraction;
84 extern int compat_log;
85 extern int latencytop_enabled;
86 extern int sysctl_nr_open_min, sysctl_nr_open_max;
87 #ifndef CONFIG_MMU
88 extern int sysctl_nr_trim_pages;
89 #endif
90 #ifdef CONFIG_RCU_TORTURE_TEST
91 extern int rcutorture_runnable;
92 #endif /* #ifdef CONFIG_RCU_TORTURE_TEST */
93
94 /* Constants used for minimum and  maximum */
95 #ifdef CONFIG_DETECT_SOFTLOCKUP
96 static int sixty = 60;
97 static int neg_one = -1;
98 #endif
99
100 static int zero;
101 static int __maybe_unused one = 1;
102 static int __maybe_unused two = 2;
103 static unsigned long one_ul = 1;
104 static int one_hundred = 100;
105
106 /* this is needed for the proc_doulongvec_minmax of vm_dirty_bytes */
107 static unsigned long dirty_bytes_min = 2 * PAGE_SIZE;
108
109 /* this is needed for the proc_dointvec_minmax for [fs_]overflow UID and GID */
110 static int maxolduid = 65535;
111 static int minolduid;
112 static int min_percpu_pagelist_fract = 8;
113
114 static int ngroups_max = NGROUPS_MAX;
115
116 #ifdef CONFIG_MODULES
117 extern char modprobe_path[];
118 #endif
119 #ifdef CONFIG_CHR_DEV_SG
120 extern int sg_big_buff;
121 #endif
122
123 #ifdef CONFIG_SPARC
124 #include <asm/system.h>
125 #endif
126
127 #ifdef CONFIG_SPARC64
128 extern int sysctl_tsb_ratio;
129 #endif
130
131 #ifdef __hppa__
132 extern int pwrsw_enabled;
133 extern int unaligned_enabled;
134 #endif
135
136 #ifdef CONFIG_S390
137 #ifdef CONFIG_MATHEMU
138 extern int sysctl_ieee_emulation_warnings;
139 #endif
140 extern int sysctl_userprocess_debug;
141 extern int spin_retry;
142 #endif
143
144 #ifdef CONFIG_BSD_PROCESS_ACCT
145 extern int acct_parm[];
146 #endif
147
148 #ifdef CONFIG_IA64
149 extern int no_unaligned_warning;
150 extern int unaligned_dump_stack;
151 #endif
152
153 #ifdef CONFIG_RT_MUTEXES
154 extern int max_lock_depth;
155 #endif
156
157 #ifdef CONFIG_PROC_SYSCTL
158 static int proc_do_cad_pid(struct ctl_table *table, int write, struct file *filp,
159                   void __user *buffer, size_t *lenp, loff_t *ppos);
160 static int proc_taint(struct ctl_table *table, int write, struct file *filp,
161                                void __user *buffer, size_t *lenp, loff_t *ppos);
162 #endif
163
164 static struct ctl_table root_table[];
165 static struct ctl_table_root sysctl_table_root;
166 static struct ctl_table_header root_table_header = {
167         .count = 1,
168         .ctl_table = root_table,
169         .ctl_entry = LIST_HEAD_INIT(sysctl_table_root.default_set.list),
170         .root = &sysctl_table_root,
171         .set = &sysctl_table_root.default_set,
172 };
173 static struct ctl_table_root sysctl_table_root = {
174         .root_list = LIST_HEAD_INIT(sysctl_table_root.root_list),
175         .default_set.list = LIST_HEAD_INIT(root_table_header.ctl_entry),
176 };
177
178 static struct ctl_table kern_table[];
179 static struct ctl_table vm_table[];
180 static struct ctl_table fs_table[];
181 static struct ctl_table debug_table[];
182 static struct ctl_table dev_table[];
183 extern struct ctl_table random_table[];
184 #ifdef CONFIG_INOTIFY_USER
185 extern struct ctl_table inotify_table[];
186 #endif
187 #ifdef CONFIG_EPOLL
188 extern struct ctl_table epoll_table[];
189 #endif
190
191 #ifdef HAVE_ARCH_PICK_MMAP_LAYOUT
192 int sysctl_legacy_va_layout;
193 #endif
194
195 extern int prove_locking;
196 extern int lock_stat;
197
198 /* The default sysctl tables: */
199
200 static struct ctl_table root_table[] = {
201         {
202                 .ctl_name       = CTL_KERN,
203                 .procname       = "kernel",
204                 .mode           = 0555,
205                 .child          = kern_table,
206         },
207         {
208                 .ctl_name       = CTL_VM,
209                 .procname       = "vm",
210                 .mode           = 0555,
211                 .child          = vm_table,
212         },
213         {
214                 .ctl_name       = CTL_FS,
215                 .procname       = "fs",
216                 .mode           = 0555,
217                 .child          = fs_table,
218         },
219         {
220                 .ctl_name       = CTL_DEBUG,
221                 .procname       = "debug",
222                 .mode           = 0555,
223                 .child          = debug_table,
224         },
225         {
226                 .ctl_name       = CTL_DEV,
227                 .procname       = "dev",
228                 .mode           = 0555,
229                 .child          = dev_table,
230         },
231 /*
232  * NOTE: do not add new entries to this table unless you have read
233  * Documentation/sysctl/ctl_unnumbered.txt
234  */
235         { .ctl_name = 0 }
236 };
237
238 #ifdef CONFIG_SCHED_DEBUG
239 static int min_sched_granularity_ns = 100000;           /* 100 usecs */
240 static int max_sched_granularity_ns = NSEC_PER_SEC;     /* 1 second */
241 static int min_wakeup_granularity_ns;                   /* 0 usecs */
242 static int max_wakeup_granularity_ns = NSEC_PER_SEC;    /* 1 second */
243 #endif
244
245 static struct ctl_table kern_table[] = {
246 #ifdef CONFIG_SCHED_DEBUG
247         {
248                 .ctl_name       = CTL_UNNUMBERED,
249                 .procname       = "sched_min_granularity_ns",
250                 .data           = &sysctl_sched_min_granularity,
251                 .maxlen         = sizeof(unsigned int),
252                 .mode           = 0644,
253                 .proc_handler   = &sched_nr_latency_handler,
254                 .strategy       = &sysctl_intvec,
255                 .extra1         = &min_sched_granularity_ns,
256                 .extra2         = &max_sched_granularity_ns,
257         },
258         {
259                 .ctl_name       = CTL_UNNUMBERED,
260                 .procname       = "sched_latency_ns",
261                 .data           = &sysctl_sched_latency,
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_wakeup_granularity_ns",
272                 .data           = &sysctl_sched_wakeup_granularity,
273                 .maxlen         = sizeof(unsigned int),
274                 .mode           = 0644,
275                 .proc_handler   = &proc_dointvec_minmax,
276                 .strategy       = &sysctl_intvec,
277                 .extra1         = &min_wakeup_granularity_ns,
278                 .extra2         = &max_wakeup_granularity_ns,
279         },
280         {
281                 .ctl_name       = CTL_UNNUMBERED,
282                 .procname       = "sched_shares_ratelimit",
283                 .data           = &sysctl_sched_shares_ratelimit,
284                 .maxlen         = sizeof(unsigned int),
285                 .mode           = 0644,
286                 .proc_handler   = &proc_dointvec,
287         },
288         {
289                 .ctl_name       = CTL_UNNUMBERED,
290                 .procname       = "sched_shares_thresh",
291                 .data           = &sysctl_sched_shares_thresh,
292                 .maxlen         = sizeof(unsigned int),
293                 .mode           = 0644,
294                 .proc_handler   = &proc_dointvec_minmax,
295                 .strategy       = &sysctl_intvec,
296                 .extra1         = &zero,
297         },
298         {
299                 .ctl_name       = CTL_UNNUMBERED,
300                 .procname       = "sched_child_runs_first",
301                 .data           = &sysctl_sched_child_runs_first,
302                 .maxlen         = sizeof(unsigned int),
303                 .mode           = 0644,
304                 .proc_handler   = &proc_dointvec,
305         },
306         {
307                 .ctl_name       = CTL_UNNUMBERED,
308                 .procname       = "sched_features",
309                 .data           = &sysctl_sched_features,
310                 .maxlen         = sizeof(unsigned int),
311                 .mode           = 0644,
312                 .proc_handler   = &proc_dointvec,
313         },
314         {
315                 .ctl_name       = CTL_UNNUMBERED,
316                 .procname       = "sched_migration_cost",
317                 .data           = &sysctl_sched_migration_cost,
318                 .maxlen         = sizeof(unsigned int),
319                 .mode           = 0644,
320                 .proc_handler   = &proc_dointvec,
321         },
322         {
323                 .ctl_name       = CTL_UNNUMBERED,
324                 .procname       = "sched_nr_migrate",
325                 .data           = &sysctl_sched_nr_migrate,
326                 .maxlen         = sizeof(unsigned int),
327                 .mode           = 0644,
328                 .proc_handler   = &proc_dointvec,
329         },
330 #endif
331         {
332                 .ctl_name       = CTL_UNNUMBERED,
333                 .procname       = "sched_rt_period_us",
334                 .data           = &sysctl_sched_rt_period,
335                 .maxlen         = sizeof(unsigned int),
336                 .mode           = 0644,
337                 .proc_handler   = &sched_rt_handler,
338         },
339         {
340                 .ctl_name       = CTL_UNNUMBERED,
341                 .procname       = "sched_rt_runtime_us",
342                 .data           = &sysctl_sched_rt_runtime,
343                 .maxlen         = sizeof(int),
344                 .mode           = 0644,
345                 .proc_handler   = &sched_rt_handler,
346         },
347         {
348                 .ctl_name       = CTL_UNNUMBERED,
349                 .procname       = "sched_compat_yield",
350                 .data           = &sysctl_sched_compat_yield,
351                 .maxlen         = sizeof(unsigned int),
352                 .mode           = 0644,
353                 .proc_handler   = &proc_dointvec,
354         },
355 #ifdef CONFIG_PROVE_LOCKING
356         {
357                 .ctl_name       = CTL_UNNUMBERED,
358                 .procname       = "prove_locking",
359                 .data           = &prove_locking,
360                 .maxlen         = sizeof(int),
361                 .mode           = 0644,
362                 .proc_handler   = &proc_dointvec,
363         },
364 #endif
365 #ifdef CONFIG_LOCK_STAT
366         {
367                 .ctl_name       = CTL_UNNUMBERED,
368                 .procname       = "lock_stat",
369                 .data           = &lock_stat,
370                 .maxlen         = sizeof(int),
371                 .mode           = 0644,
372                 .proc_handler   = &proc_dointvec,
373         },
374 #endif
375         {
376                 .ctl_name       = KERN_PANIC,
377                 .procname       = "panic",
378                 .data           = &panic_timeout,
379                 .maxlen         = sizeof(int),
380                 .mode           = 0644,
381                 .proc_handler   = &proc_dointvec,
382         },
383         {
384                 .ctl_name       = KERN_CORE_USES_PID,
385                 .procname       = "core_uses_pid",
386                 .data           = &core_uses_pid,
387                 .maxlen         = sizeof(int),
388                 .mode           = 0644,
389                 .proc_handler   = &proc_dointvec,
390         },
391         {
392                 .ctl_name       = KERN_CORE_PATTERN,
393                 .procname       = "core_pattern",
394                 .data           = core_pattern,
395                 .maxlen         = CORENAME_MAX_SIZE,
396                 .mode           = 0644,
397                 .proc_handler   = &proc_dostring,
398                 .strategy       = &sysctl_string,
399         },
400 #ifdef CONFIG_PROC_SYSCTL
401         {
402                 .procname       = "tainted",
403                 .maxlen         = sizeof(long),
404                 .mode           = 0644,
405                 .proc_handler   = &proc_taint,
406         },
407 #endif
408 #ifdef CONFIG_LATENCYTOP
409         {
410                 .procname       = "latencytop",
411                 .data           = &latencytop_enabled,
412                 .maxlen         = sizeof(int),
413                 .mode           = 0644,
414                 .proc_handler   = &proc_dointvec,
415         },
416 #endif
417 #ifdef CONFIG_BLK_DEV_INITRD
418         {
419                 .ctl_name       = KERN_REALROOTDEV,
420                 .procname       = "real-root-dev",
421                 .data           = &real_root_dev,
422                 .maxlen         = sizeof(int),
423                 .mode           = 0644,
424                 .proc_handler   = &proc_dointvec,
425         },
426 #endif
427         {
428                 .ctl_name       = CTL_UNNUMBERED,
429                 .procname       = "print-fatal-signals",
430                 .data           = &print_fatal_signals,
431                 .maxlen         = sizeof(int),
432                 .mode           = 0644,
433                 .proc_handler   = &proc_dointvec,
434         },
435 #ifdef CONFIG_SPARC
436         {
437                 .ctl_name       = KERN_SPARC_REBOOT,
438                 .procname       = "reboot-cmd",
439                 .data           = reboot_command,
440                 .maxlen         = 256,
441                 .mode           = 0644,
442                 .proc_handler   = &proc_dostring,
443                 .strategy       = &sysctl_string,
444         },
445         {
446                 .ctl_name       = KERN_SPARC_STOP_A,
447                 .procname       = "stop-a",
448                 .data           = &stop_a_enabled,
449                 .maxlen         = sizeof (int),
450                 .mode           = 0644,
451                 .proc_handler   = &proc_dointvec,
452         },
453         {
454                 .ctl_name       = KERN_SPARC_SCONS_PWROFF,
455                 .procname       = "scons-poweroff",
456                 .data           = &scons_pwroff,
457                 .maxlen         = sizeof (int),
458                 .mode           = 0644,
459                 .proc_handler   = &proc_dointvec,
460         },
461 #endif
462 #ifdef CONFIG_SPARC64
463         {
464                 .ctl_name       = CTL_UNNUMBERED,
465                 .procname       = "tsb-ratio",
466                 .data           = &sysctl_tsb_ratio,
467                 .maxlen         = sizeof (int),
468                 .mode           = 0644,
469                 .proc_handler   = &proc_dointvec,
470         },
471 #endif
472 #ifdef __hppa__
473         {
474                 .ctl_name       = KERN_HPPA_PWRSW,
475                 .procname       = "soft-power",
476                 .data           = &pwrsw_enabled,
477                 .maxlen         = sizeof (int),
478                 .mode           = 0644,
479                 .proc_handler   = &proc_dointvec,
480         },
481         {
482                 .ctl_name       = KERN_HPPA_UNALIGNED,
483                 .procname       = "unaligned-trap",
484                 .data           = &unaligned_enabled,
485                 .maxlen         = sizeof (int),
486                 .mode           = 0644,
487                 .proc_handler   = &proc_dointvec,
488         },
489 #endif
490         {
491                 .ctl_name       = KERN_CTLALTDEL,
492                 .procname       = "ctrl-alt-del",
493                 .data           = &C_A_D,
494                 .maxlen         = sizeof(int),
495                 .mode           = 0644,
496                 .proc_handler   = &proc_dointvec,
497         },
498 #ifdef CONFIG_FUNCTION_TRACER
499         {
500                 .ctl_name       = CTL_UNNUMBERED,
501                 .procname       = "ftrace_enabled",
502                 .data           = &ftrace_enabled,
503                 .maxlen         = sizeof(int),
504                 .mode           = 0644,
505                 .proc_handler   = &ftrace_enable_sysctl,
506         },
507 #endif
508 #ifdef CONFIG_STACK_TRACER
509         {
510                 .ctl_name       = CTL_UNNUMBERED,
511                 .procname       = "stack_tracer_enabled",
512                 .data           = &stack_tracer_enabled,
513                 .maxlen         = sizeof(int),
514                 .mode           = 0644,
515                 .proc_handler   = &stack_trace_sysctl,
516         },
517 #endif
518 #ifdef CONFIG_TRACING
519         {
520                 .ctl_name       = CTL_UNNUMBERED,
521                 .procname       = "ftrace_dump_on_oops",
522                 .data           = &ftrace_dump_on_oops,
523                 .maxlen         = sizeof(int),
524                 .mode           = 0644,
525                 .proc_handler   = &proc_dointvec,
526         },
527 #endif
528 #ifdef CONFIG_MODULES
529         {
530                 .ctl_name       = KERN_MODPROBE,
531                 .procname       = "modprobe",
532                 .data           = &modprobe_path,
533                 .maxlen         = KMOD_PATH_LEN,
534                 .mode           = 0644,
535                 .proc_handler   = &proc_dostring,
536                 .strategy       = &sysctl_string,
537         },
538 #endif
539 #if defined(CONFIG_HOTPLUG) && defined(CONFIG_NET)
540         {
541                 .ctl_name       = KERN_HOTPLUG,
542                 .procname       = "hotplug",
543                 .data           = &uevent_helper,
544                 .maxlen         = UEVENT_HELPER_PATH_LEN,
545                 .mode           = 0644,
546                 .proc_handler   = &proc_dostring,
547                 .strategy       = &sysctl_string,
548         },
549 #endif
550 #ifdef CONFIG_CHR_DEV_SG
551         {
552                 .ctl_name       = KERN_SG_BIG_BUFF,
553                 .procname       = "sg-big-buff",
554                 .data           = &sg_big_buff,
555                 .maxlen         = sizeof (int),
556                 .mode           = 0444,
557                 .proc_handler   = &proc_dointvec,
558         },
559 #endif
560 #ifdef CONFIG_BSD_PROCESS_ACCT
561         {
562                 .ctl_name       = KERN_ACCT,
563                 .procname       = "acct",
564                 .data           = &acct_parm,
565                 .maxlen         = 3*sizeof(int),
566                 .mode           = 0644,
567                 .proc_handler   = &proc_dointvec,
568         },
569 #endif
570 #ifdef CONFIG_MAGIC_SYSRQ
571         {
572                 .ctl_name       = KERN_SYSRQ,
573                 .procname       = "sysrq",
574                 .data           = &__sysrq_enabled,
575                 .maxlen         = sizeof (int),
576                 .mode           = 0644,
577                 .proc_handler   = &proc_dointvec,
578         },
579 #endif
580 #ifdef CONFIG_PROC_SYSCTL
581         {
582                 .procname       = "cad_pid",
583                 .data           = NULL,
584                 .maxlen         = sizeof (int),
585                 .mode           = 0600,
586                 .proc_handler   = &proc_do_cad_pid,
587         },
588 #endif
589         {
590                 .ctl_name       = KERN_MAX_THREADS,
591                 .procname       = "threads-max",
592                 .data           = &max_threads,
593                 .maxlen         = sizeof(int),
594                 .mode           = 0644,
595                 .proc_handler   = &proc_dointvec,
596         },
597         {
598                 .ctl_name       = KERN_RANDOM,
599                 .procname       = "random",
600                 .mode           = 0555,
601                 .child          = random_table,
602         },
603         {
604                 .ctl_name       = KERN_OVERFLOWUID,
605                 .procname       = "overflowuid",
606                 .data           = &overflowuid,
607                 .maxlen         = sizeof(int),
608                 .mode           = 0644,
609                 .proc_handler   = &proc_dointvec_minmax,
610                 .strategy       = &sysctl_intvec,
611                 .extra1         = &minolduid,
612                 .extra2         = &maxolduid,
613         },
614         {
615                 .ctl_name       = KERN_OVERFLOWGID,
616                 .procname       = "overflowgid",
617                 .data           = &overflowgid,
618                 .maxlen         = sizeof(int),
619                 .mode           = 0644,
620                 .proc_handler   = &proc_dointvec_minmax,
621                 .strategy       = &sysctl_intvec,
622                 .extra1         = &minolduid,
623                 .extra2         = &maxolduid,
624         },
625 #ifdef CONFIG_S390
626 #ifdef CONFIG_MATHEMU
627         {
628                 .ctl_name       = KERN_IEEE_EMULATION_WARNINGS,
629                 .procname       = "ieee_emulation_warnings",
630                 .data           = &sysctl_ieee_emulation_warnings,
631                 .maxlen         = sizeof(int),
632                 .mode           = 0644,
633                 .proc_handler   = &proc_dointvec,
634         },
635 #endif
636         {
637                 .ctl_name       = KERN_S390_USER_DEBUG_LOGGING,
638                 .procname       = "userprocess_debug",
639                 .data           = &sysctl_userprocess_debug,
640                 .maxlen         = sizeof(int),
641                 .mode           = 0644,
642                 .proc_handler   = &proc_dointvec,
643         },
644 #endif
645         {
646                 .ctl_name       = KERN_PIDMAX,
647                 .procname       = "pid_max",
648                 .data           = &pid_max,
649                 .maxlen         = sizeof (int),
650                 .mode           = 0644,
651                 .proc_handler   = &proc_dointvec_minmax,
652                 .strategy       = sysctl_intvec,
653                 .extra1         = &pid_max_min,
654                 .extra2         = &pid_max_max,
655         },
656         {
657                 .ctl_name       = KERN_PANIC_ON_OOPS,
658                 .procname       = "panic_on_oops",
659                 .data           = &panic_on_oops,
660                 .maxlen         = sizeof(int),
661                 .mode           = 0644,
662                 .proc_handler   = &proc_dointvec,
663         },
664 #if defined CONFIG_PRINTK
665         {
666                 .ctl_name       = KERN_PRINTK,
667                 .procname       = "printk",
668                 .data           = &console_loglevel,
669                 .maxlen         = 4*sizeof(int),
670                 .mode           = 0644,
671                 .proc_handler   = &proc_dointvec,
672         },
673         {
674                 .ctl_name       = KERN_PRINTK_RATELIMIT,
675                 .procname       = "printk_ratelimit",
676                 .data           = &printk_ratelimit_state.interval,
677                 .maxlen         = sizeof(int),
678                 .mode           = 0644,
679                 .proc_handler   = &proc_dointvec_jiffies,
680                 .strategy       = &sysctl_jiffies,
681         },
682         {
683                 .ctl_name       = KERN_PRINTK_RATELIMIT_BURST,
684                 .procname       = "printk_ratelimit_burst",
685                 .data           = &printk_ratelimit_state.burst,
686                 .maxlen         = sizeof(int),
687                 .mode           = 0644,
688                 .proc_handler   = &proc_dointvec,
689         },
690 #endif
691         {
692                 .ctl_name       = KERN_NGROUPS_MAX,
693                 .procname       = "ngroups_max",
694                 .data           = &ngroups_max,
695                 .maxlen         = sizeof (int),
696                 .mode           = 0444,
697                 .proc_handler   = &proc_dointvec,
698         },
699 #if defined(CONFIG_X86_LOCAL_APIC) && defined(CONFIG_X86)
700         {
701                 .ctl_name       = KERN_UNKNOWN_NMI_PANIC,
702                 .procname       = "unknown_nmi_panic",
703                 .data           = &unknown_nmi_panic,
704                 .maxlen         = sizeof (int),
705                 .mode           = 0644,
706                 .proc_handler   = &proc_dointvec,
707         },
708         {
709                 .procname       = "nmi_watchdog",
710                 .data           = &nmi_watchdog_enabled,
711                 .maxlen         = sizeof (int),
712                 .mode           = 0644,
713                 .proc_handler   = &proc_nmi_enabled,
714         },
715 #endif
716 #if defined(CONFIG_X86)
717         {
718                 .ctl_name       = KERN_PANIC_ON_NMI,
719                 .procname       = "panic_on_unrecovered_nmi",
720                 .data           = &panic_on_unrecovered_nmi,
721                 .maxlen         = sizeof(int),
722                 .mode           = 0644,
723                 .proc_handler   = &proc_dointvec,
724         },
725         {
726                 .ctl_name       = KERN_BOOTLOADER_TYPE,
727                 .procname       = "bootloader_type",
728                 .data           = &bootloader_type,
729                 .maxlen         = sizeof (int),
730                 .mode           = 0444,
731                 .proc_handler   = &proc_dointvec,
732         },
733         {
734                 .ctl_name       = CTL_UNNUMBERED,
735                 .procname       = "kstack_depth_to_print",
736                 .data           = &kstack_depth_to_print,
737                 .maxlen         = sizeof(int),
738                 .mode           = 0644,
739                 .proc_handler   = &proc_dointvec,
740         },
741         {
742                 .ctl_name       = CTL_UNNUMBERED,
743                 .procname       = "io_delay_type",
744                 .data           = &io_delay_type,
745                 .maxlen         = sizeof(int),
746                 .mode           = 0644,
747                 .proc_handler   = &proc_dointvec,
748         },
749 #endif
750 #if defined(CONFIG_MMU)
751         {
752                 .ctl_name       = KERN_RANDOMIZE,
753                 .procname       = "randomize_va_space",
754                 .data           = &randomize_va_space,
755                 .maxlen         = sizeof(int),
756                 .mode           = 0644,
757                 .proc_handler   = &proc_dointvec,
758         },
759 #endif
760 #if defined(CONFIG_S390) && defined(CONFIG_SMP)
761         {
762                 .ctl_name       = KERN_SPIN_RETRY,
763                 .procname       = "spin_retry",
764                 .data           = &spin_retry,
765                 .maxlen         = sizeof (int),
766                 .mode           = 0644,
767                 .proc_handler   = &proc_dointvec,
768         },
769 #endif
770 #if     defined(CONFIG_ACPI_SLEEP) && defined(CONFIG_X86)
771         {
772                 .procname       = "acpi_video_flags",
773                 .data           = &acpi_realmode_flags,
774                 .maxlen         = sizeof (unsigned long),
775                 .mode           = 0644,
776                 .proc_handler   = &proc_doulongvec_minmax,
777         },
778 #endif
779 #ifdef CONFIG_IA64
780         {
781                 .ctl_name       = KERN_IA64_UNALIGNED,
782                 .procname       = "ignore-unaligned-usertrap",
783                 .data           = &no_unaligned_warning,
784                 .maxlen         = sizeof (int),
785                 .mode           = 0644,
786                 .proc_handler   = &proc_dointvec,
787         },
788         {
789                 .ctl_name       = CTL_UNNUMBERED,
790                 .procname       = "unaligned-dump-stack",
791                 .data           = &unaligned_dump_stack,
792                 .maxlen         = sizeof (int),
793                 .mode           = 0644,
794                 .proc_handler   = &proc_dointvec,
795         },
796 #endif
797 #ifdef CONFIG_DETECT_SOFTLOCKUP
798         {
799                 .ctl_name       = CTL_UNNUMBERED,
800                 .procname       = "softlockup_panic",
801                 .data           = &softlockup_panic,
802                 .maxlen         = sizeof(int),
803                 .mode           = 0644,
804                 .proc_handler   = &proc_dointvec_minmax,
805                 .strategy       = &sysctl_intvec,
806                 .extra1         = &zero,
807                 .extra2         = &one,
808         },
809         {
810                 .ctl_name       = CTL_UNNUMBERED,
811                 .procname       = "softlockup_thresh",
812                 .data           = &softlockup_thresh,
813                 .maxlen         = sizeof(int),
814                 .mode           = 0644,
815                 .proc_handler   = &proc_dosoftlockup_thresh,
816                 .strategy       = &sysctl_intvec,
817                 .extra1         = &neg_one,
818                 .extra2         = &sixty,
819         },
820 #endif
821 #ifdef CONFIG_DETECT_HUNG_TASK
822         {
823                 .ctl_name       = CTL_UNNUMBERED,
824                 .procname       = "hung_task_panic",
825                 .data           = &sysctl_hung_task_panic,
826                 .maxlen         = sizeof(int),
827                 .mode           = 0644,
828                 .proc_handler   = &proc_dointvec_minmax,
829                 .strategy       = &sysctl_intvec,
830                 .extra1         = &zero,
831                 .extra2         = &one,
832         },
833         {
834                 .ctl_name       = CTL_UNNUMBERED,
835                 .procname       = "hung_task_check_count",
836                 .data           = &sysctl_hung_task_check_count,
837                 .maxlen         = sizeof(unsigned long),
838                 .mode           = 0644,
839                 .proc_handler   = &proc_doulongvec_minmax,
840                 .strategy       = &sysctl_intvec,
841         },
842         {
843                 .ctl_name       = CTL_UNNUMBERED,
844                 .procname       = "hung_task_timeout_secs",
845                 .data           = &sysctl_hung_task_timeout_secs,
846                 .maxlen         = sizeof(unsigned long),
847                 .mode           = 0644,
848                 .proc_handler   = &proc_dohung_task_timeout_secs,
849                 .strategy       = &sysctl_intvec,
850         },
851         {
852                 .ctl_name       = CTL_UNNUMBERED,
853                 .procname       = "hung_task_warnings",
854                 .data           = &sysctl_hung_task_warnings,
855                 .maxlen         = sizeof(unsigned long),
856                 .mode           = 0644,
857                 .proc_handler   = &proc_doulongvec_minmax,
858                 .strategy       = &sysctl_intvec,
859         },
860 #endif
861 #ifdef CONFIG_COMPAT
862         {
863                 .ctl_name       = KERN_COMPAT_LOG,
864                 .procname       = "compat-log",
865                 .data           = &compat_log,
866                 .maxlen         = sizeof (int),
867                 .mode           = 0644,
868                 .proc_handler   = &proc_dointvec,
869         },
870 #endif
871 #ifdef CONFIG_RT_MUTEXES
872         {
873                 .ctl_name       = KERN_MAX_LOCK_DEPTH,
874                 .procname       = "max_lock_depth",
875                 .data           = &max_lock_depth,
876                 .maxlen         = sizeof(int),
877                 .mode           = 0644,
878                 .proc_handler   = &proc_dointvec,
879         },
880 #endif
881         {
882                 .ctl_name       = CTL_UNNUMBERED,
883                 .procname       = "poweroff_cmd",
884                 .data           = &poweroff_cmd,
885                 .maxlen         = POWEROFF_CMD_PATH_LEN,
886                 .mode           = 0644,
887                 .proc_handler   = &proc_dostring,
888                 .strategy       = &sysctl_string,
889         },
890 #ifdef CONFIG_KEYS
891         {
892                 .ctl_name       = CTL_UNNUMBERED,
893                 .procname       = "keys",
894                 .mode           = 0555,
895                 .child          = key_sysctls,
896         },
897 #endif
898 #ifdef CONFIG_RCU_TORTURE_TEST
899         {
900                 .ctl_name       = CTL_UNNUMBERED,
901                 .procname       = "rcutorture_runnable",
902                 .data           = &rcutorture_runnable,
903                 .maxlen         = sizeof(int),
904                 .mode           = 0644,
905                 .proc_handler   = &proc_dointvec,
906         },
907 #endif
908 #ifdef CONFIG_SLOW_WORK
909         {
910                 .ctl_name       = CTL_UNNUMBERED,
911                 .procname       = "slow-work",
912                 .mode           = 0555,
913                 .child          = slow_work_sysctls,
914         },
915 #endif
916 #ifdef CONFIG_PERF_COUNTERS
917         {
918                 .ctl_name       = CTL_UNNUMBERED,
919                 .procname       = "perf_counter_privileged",
920                 .data           = &sysctl_perf_counter_priv,
921                 .maxlen         = sizeof(sysctl_perf_counter_priv),
922                 .mode           = 0644,
923                 .proc_handler   = &proc_dointvec,
924         },
925         {
926                 .ctl_name       = CTL_UNNUMBERED,
927                 .procname       = "perf_counter_mlock_kb",
928                 .data           = &sysctl_perf_counter_mlock,
929                 .maxlen         = sizeof(sysctl_perf_counter_mlock),
930                 .mode           = 0644,
931                 .proc_handler   = &proc_dointvec,
932         },
933 #endif
934 /*
935  * NOTE: do not add new entries to this table unless you have read
936  * Documentation/sysctl/ctl_unnumbered.txt
937  */
938         { .ctl_name = 0 }
939 };
940
941 static struct ctl_table vm_table[] = {
942         {
943                 .ctl_name       = VM_OVERCOMMIT_MEMORY,
944                 .procname       = "overcommit_memory",
945                 .data           = &sysctl_overcommit_memory,
946                 .maxlen         = sizeof(sysctl_overcommit_memory),
947                 .mode           = 0644,
948                 .proc_handler   = &proc_dointvec,
949         },
950         {
951                 .ctl_name       = VM_PANIC_ON_OOM,
952                 .procname       = "panic_on_oom",
953                 .data           = &sysctl_panic_on_oom,
954                 .maxlen         = sizeof(sysctl_panic_on_oom),
955                 .mode           = 0644,
956                 .proc_handler   = &proc_dointvec,
957         },
958         {
959                 .ctl_name       = CTL_UNNUMBERED,
960                 .procname       = "oom_kill_allocating_task",
961                 .data           = &sysctl_oom_kill_allocating_task,
962                 .maxlen         = sizeof(sysctl_oom_kill_allocating_task),
963                 .mode           = 0644,
964                 .proc_handler   = &proc_dointvec,
965         },
966         {
967                 .ctl_name       = CTL_UNNUMBERED,
968                 .procname       = "oom_dump_tasks",
969                 .data           = &sysctl_oom_dump_tasks,
970                 .maxlen         = sizeof(sysctl_oom_dump_tasks),
971                 .mode           = 0644,
972                 .proc_handler   = &proc_dointvec,
973         },
974         {
975                 .ctl_name       = VM_OVERCOMMIT_RATIO,
976                 .procname       = "overcommit_ratio",
977                 .data           = &sysctl_overcommit_ratio,
978                 .maxlen         = sizeof(sysctl_overcommit_ratio),
979                 .mode           = 0644,
980                 .proc_handler   = &proc_dointvec,
981         },
982         {
983                 .ctl_name       = VM_PAGE_CLUSTER,
984                 .procname       = "page-cluster", 
985                 .data           = &page_cluster,
986                 .maxlen         = sizeof(int),
987                 .mode           = 0644,
988                 .proc_handler   = &proc_dointvec,
989         },
990         {
991                 .ctl_name       = VM_DIRTY_BACKGROUND,
992                 .procname       = "dirty_background_ratio",
993                 .data           = &dirty_background_ratio,
994                 .maxlen         = sizeof(dirty_background_ratio),
995                 .mode           = 0644,
996                 .proc_handler   = &dirty_background_ratio_handler,
997                 .strategy       = &sysctl_intvec,
998                 .extra1         = &zero,
999                 .extra2         = &one_hundred,
1000         },
1001         {
1002                 .ctl_name       = CTL_UNNUMBERED,
1003                 .procname       = "dirty_background_bytes",
1004                 .data           = &dirty_background_bytes,
1005                 .maxlen         = sizeof(dirty_background_bytes),
1006                 .mode           = 0644,
1007                 .proc_handler   = &dirty_background_bytes_handler,
1008                 .strategy       = &sysctl_intvec,
1009                 .extra1         = &one_ul,
1010         },
1011         {
1012                 .ctl_name       = VM_DIRTY_RATIO,
1013                 .procname       = "dirty_ratio",
1014                 .data           = &vm_dirty_ratio,
1015                 .maxlen         = sizeof(vm_dirty_ratio),
1016                 .mode           = 0644,
1017                 .proc_handler   = &dirty_ratio_handler,
1018                 .strategy       = &sysctl_intvec,
1019                 .extra1         = &zero,
1020                 .extra2         = &one_hundred,
1021         },
1022         {
1023                 .ctl_name       = CTL_UNNUMBERED,
1024                 .procname       = "dirty_bytes",
1025                 .data           = &vm_dirty_bytes,
1026                 .maxlen         = sizeof(vm_dirty_bytes),
1027                 .mode           = 0644,
1028                 .proc_handler   = &dirty_bytes_handler,
1029                 .strategy       = &sysctl_intvec,
1030                 .extra1         = &dirty_bytes_min,
1031         },
1032         {
1033                 .procname       = "dirty_writeback_centisecs",
1034                 .data           = &dirty_writeback_interval,
1035                 .maxlen         = sizeof(dirty_writeback_interval),
1036                 .mode           = 0644,
1037                 .proc_handler   = &dirty_writeback_centisecs_handler,
1038         },
1039         {
1040                 .procname       = "dirty_expire_centisecs",
1041                 .data           = &dirty_expire_interval,
1042                 .maxlen         = sizeof(dirty_expire_interval),
1043                 .mode           = 0644,
1044                 .proc_handler   = &proc_dointvec,
1045         },
1046         {
1047                 .ctl_name       = VM_NR_PDFLUSH_THREADS,
1048                 .procname       = "nr_pdflush_threads",
1049                 .data           = &nr_pdflush_threads,
1050                 .maxlen         = sizeof nr_pdflush_threads,
1051                 .mode           = 0444 /* read-only*/,
1052                 .proc_handler   = &proc_dointvec,
1053         },
1054         {
1055                 .ctl_name       = VM_SWAPPINESS,
1056                 .procname       = "swappiness",
1057                 .data           = &vm_swappiness,
1058                 .maxlen         = sizeof(vm_swappiness),
1059                 .mode           = 0644,
1060                 .proc_handler   = &proc_dointvec_minmax,
1061                 .strategy       = &sysctl_intvec,
1062                 .extra1         = &zero,
1063                 .extra2         = &one_hundred,
1064         },
1065 #ifdef CONFIG_HUGETLB_PAGE
1066          {
1067                 .procname       = "nr_hugepages",
1068                 .data           = NULL,
1069                 .maxlen         = sizeof(unsigned long),
1070                 .mode           = 0644,
1071                 .proc_handler   = &hugetlb_sysctl_handler,
1072                 .extra1         = (void *)&hugetlb_zero,
1073                 .extra2         = (void *)&hugetlb_infinity,
1074          },
1075          {
1076                 .ctl_name       = VM_HUGETLB_GROUP,
1077                 .procname       = "hugetlb_shm_group",
1078                 .data           = &sysctl_hugetlb_shm_group,
1079                 .maxlen         = sizeof(gid_t),
1080                 .mode           = 0644,
1081                 .proc_handler   = &proc_dointvec,
1082          },
1083          {
1084                 .ctl_name       = CTL_UNNUMBERED,
1085                 .procname       = "hugepages_treat_as_movable",
1086                 .data           = &hugepages_treat_as_movable,
1087                 .maxlen         = sizeof(int),
1088                 .mode           = 0644,
1089                 .proc_handler   = &hugetlb_treat_movable_handler,
1090         },
1091         {
1092                 .ctl_name       = CTL_UNNUMBERED,
1093                 .procname       = "nr_overcommit_hugepages",
1094                 .data           = NULL,
1095                 .maxlen         = sizeof(unsigned long),
1096                 .mode           = 0644,
1097                 .proc_handler   = &hugetlb_overcommit_handler,
1098                 .extra1         = (void *)&hugetlb_zero,
1099                 .extra2         = (void *)&hugetlb_infinity,
1100         },
1101 #endif
1102         {
1103                 .ctl_name       = VM_LOWMEM_RESERVE_RATIO,
1104                 .procname       = "lowmem_reserve_ratio",
1105                 .data           = &sysctl_lowmem_reserve_ratio,
1106                 .maxlen         = sizeof(sysctl_lowmem_reserve_ratio),
1107                 .mode           = 0644,
1108                 .proc_handler   = &lowmem_reserve_ratio_sysctl_handler,
1109                 .strategy       = &sysctl_intvec,
1110         },
1111         {
1112                 .ctl_name       = VM_DROP_PAGECACHE,
1113                 .procname       = "drop_caches",
1114                 .data           = &sysctl_drop_caches,
1115                 .maxlen         = sizeof(int),
1116                 .mode           = 0644,
1117                 .proc_handler   = drop_caches_sysctl_handler,
1118                 .strategy       = &sysctl_intvec,
1119         },
1120         {
1121                 .ctl_name       = VM_MIN_FREE_KBYTES,
1122                 .procname       = "min_free_kbytes",
1123                 .data           = &min_free_kbytes,
1124                 .maxlen         = sizeof(min_free_kbytes),
1125                 .mode           = 0644,
1126                 .proc_handler   = &min_free_kbytes_sysctl_handler,
1127                 .strategy       = &sysctl_intvec,
1128                 .extra1         = &zero,
1129         },
1130         {
1131                 .ctl_name       = VM_PERCPU_PAGELIST_FRACTION,
1132                 .procname       = "percpu_pagelist_fraction",
1133                 .data           = &percpu_pagelist_fraction,
1134                 .maxlen         = sizeof(percpu_pagelist_fraction),
1135                 .mode           = 0644,
1136                 .proc_handler   = &percpu_pagelist_fraction_sysctl_handler,
1137                 .strategy       = &sysctl_intvec,
1138                 .extra1         = &min_percpu_pagelist_fract,
1139         },
1140 #ifdef CONFIG_MMU
1141         {
1142                 .ctl_name       = VM_MAX_MAP_COUNT,
1143                 .procname       = "max_map_count",
1144                 .data           = &sysctl_max_map_count,
1145                 .maxlen         = sizeof(sysctl_max_map_count),
1146                 .mode           = 0644,
1147                 .proc_handler   = &proc_dointvec
1148         },
1149 #else
1150         {
1151                 .ctl_name       = CTL_UNNUMBERED,
1152                 .procname       = "nr_trim_pages",
1153                 .data           = &sysctl_nr_trim_pages,
1154                 .maxlen         = sizeof(sysctl_nr_trim_pages),
1155                 .mode           = 0644,
1156                 .proc_handler   = &proc_dointvec_minmax,
1157                 .strategy       = &sysctl_intvec,
1158                 .extra1         = &zero,
1159         },
1160 #endif
1161         {
1162                 .ctl_name       = VM_LAPTOP_MODE,
1163                 .procname       = "laptop_mode",
1164                 .data           = &laptop_mode,
1165                 .maxlen         = sizeof(laptop_mode),
1166                 .mode           = 0644,
1167                 .proc_handler   = &proc_dointvec_jiffies,
1168                 .strategy       = &sysctl_jiffies,
1169         },
1170         {
1171                 .ctl_name       = VM_BLOCK_DUMP,
1172                 .procname       = "block_dump",
1173                 .data           = &block_dump,
1174                 .maxlen         = sizeof(block_dump),
1175                 .mode           = 0644,
1176                 .proc_handler   = &proc_dointvec,
1177                 .strategy       = &sysctl_intvec,
1178                 .extra1         = &zero,
1179         },
1180         {
1181                 .ctl_name       = VM_VFS_CACHE_PRESSURE,
1182                 .procname       = "vfs_cache_pressure",
1183                 .data           = &sysctl_vfs_cache_pressure,
1184                 .maxlen         = sizeof(sysctl_vfs_cache_pressure),
1185                 .mode           = 0644,
1186                 .proc_handler   = &proc_dointvec,
1187                 .strategy       = &sysctl_intvec,
1188                 .extra1         = &zero,
1189         },
1190 #ifdef HAVE_ARCH_PICK_MMAP_LAYOUT
1191         {
1192                 .ctl_name       = VM_LEGACY_VA_LAYOUT,
1193                 .procname       = "legacy_va_layout",
1194                 .data           = &sysctl_legacy_va_layout,
1195                 .maxlen         = sizeof(sysctl_legacy_va_layout),
1196                 .mode           = 0644,
1197                 .proc_handler   = &proc_dointvec,
1198                 .strategy       = &sysctl_intvec,
1199                 .extra1         = &zero,
1200         },
1201 #endif
1202 #ifdef CONFIG_NUMA
1203         {
1204                 .ctl_name       = VM_ZONE_RECLAIM_MODE,
1205                 .procname       = "zone_reclaim_mode",
1206                 .data           = &zone_reclaim_mode,
1207                 .maxlen         = sizeof(zone_reclaim_mode),
1208                 .mode           = 0644,
1209                 .proc_handler   = &proc_dointvec,
1210                 .strategy       = &sysctl_intvec,
1211                 .extra1         = &zero,
1212         },
1213         {
1214                 .ctl_name       = VM_MIN_UNMAPPED,
1215                 .procname       = "min_unmapped_ratio",
1216                 .data           = &sysctl_min_unmapped_ratio,
1217                 .maxlen         = sizeof(sysctl_min_unmapped_ratio),
1218                 .mode           = 0644,
1219                 .proc_handler   = &sysctl_min_unmapped_ratio_sysctl_handler,
1220                 .strategy       = &sysctl_intvec,
1221                 .extra1         = &zero,
1222                 .extra2         = &one_hundred,
1223         },
1224         {
1225                 .ctl_name       = VM_MIN_SLAB,
1226                 .procname       = "min_slab_ratio",
1227                 .data           = &sysctl_min_slab_ratio,
1228                 .maxlen         = sizeof(sysctl_min_slab_ratio),
1229                 .mode           = 0644,
1230                 .proc_handler   = &sysctl_min_slab_ratio_sysctl_handler,
1231                 .strategy       = &sysctl_intvec,
1232                 .extra1         = &zero,
1233                 .extra2         = &one_hundred,
1234         },
1235 #endif
1236 #ifdef CONFIG_SMP
1237         {
1238                 .ctl_name       = CTL_UNNUMBERED,
1239                 .procname       = "stat_interval",
1240                 .data           = &sysctl_stat_interval,
1241                 .maxlen         = sizeof(sysctl_stat_interval),
1242                 .mode           = 0644,
1243                 .proc_handler   = &proc_dointvec_jiffies,
1244                 .strategy       = &sysctl_jiffies,
1245         },
1246 #endif
1247 #ifdef CONFIG_SECURITY
1248         {
1249                 .ctl_name       = CTL_UNNUMBERED,
1250                 .procname       = "mmap_min_addr",
1251                 .data           = &mmap_min_addr,
1252                 .maxlen         = sizeof(unsigned long),
1253                 .mode           = 0644,
1254                 .proc_handler   = &proc_doulongvec_minmax,
1255         },
1256 #endif
1257 #ifdef CONFIG_NUMA
1258         {
1259                 .ctl_name       = CTL_UNNUMBERED,
1260                 .procname       = "numa_zonelist_order",
1261                 .data           = &numa_zonelist_order,
1262                 .maxlen         = NUMA_ZONELIST_ORDER_LEN,
1263                 .mode           = 0644,
1264                 .proc_handler   = &numa_zonelist_order_handler,
1265                 .strategy       = &sysctl_string,
1266         },
1267 #endif
1268 #if (defined(CONFIG_X86_32) && !defined(CONFIG_UML))|| \
1269    (defined(CONFIG_SUPERH) && defined(CONFIG_VSYSCALL))
1270         {
1271                 .ctl_name       = VM_VDSO_ENABLED,
1272                 .procname       = "vdso_enabled",
1273                 .data           = &vdso_enabled,
1274                 .maxlen         = sizeof(vdso_enabled),
1275                 .mode           = 0644,
1276                 .proc_handler   = &proc_dointvec,
1277                 .strategy       = &sysctl_intvec,
1278                 .extra1         = &zero,
1279         },
1280 #endif
1281 #ifdef CONFIG_HIGHMEM
1282         {
1283                 .ctl_name       = CTL_UNNUMBERED,
1284                 .procname       = "highmem_is_dirtyable",
1285                 .data           = &vm_highmem_is_dirtyable,
1286                 .maxlen         = sizeof(vm_highmem_is_dirtyable),
1287                 .mode           = 0644,
1288                 .proc_handler   = &proc_dointvec_minmax,
1289                 .strategy       = &sysctl_intvec,
1290                 .extra1         = &zero,
1291                 .extra2         = &one,
1292         },
1293 #endif
1294 #ifdef CONFIG_UNEVICTABLE_LRU
1295         {
1296                 .ctl_name       = CTL_UNNUMBERED,
1297                 .procname       = "scan_unevictable_pages",
1298                 .data           = &scan_unevictable_pages,
1299                 .maxlen         = sizeof(scan_unevictable_pages),
1300                 .mode           = 0644,
1301                 .proc_handler   = &scan_unevictable_handler,
1302         },
1303 #endif
1304 /*
1305  * NOTE: do not add new entries to this table unless you have read
1306  * Documentation/sysctl/ctl_unnumbered.txt
1307  */
1308         { .ctl_name = 0 }
1309 };
1310
1311 #if defined(CONFIG_BINFMT_MISC) || defined(CONFIG_BINFMT_MISC_MODULE)
1312 static struct ctl_table binfmt_misc_table[] = {
1313         { .ctl_name = 0 }
1314 };
1315 #endif
1316
1317 static struct ctl_table fs_table[] = {
1318         {
1319                 .ctl_name       = FS_NRINODE,
1320                 .procname       = "inode-nr",
1321                 .data           = &inodes_stat,
1322                 .maxlen         = 2*sizeof(int),
1323                 .mode           = 0444,
1324                 .proc_handler   = &proc_dointvec,
1325         },
1326         {
1327                 .ctl_name       = FS_STATINODE,
1328                 .procname       = "inode-state",
1329                 .data           = &inodes_stat,
1330                 .maxlen         = 7*sizeof(int),
1331                 .mode           = 0444,
1332                 .proc_handler   = &proc_dointvec,
1333         },
1334         {
1335                 .procname       = "file-nr",
1336                 .data           = &files_stat,
1337                 .maxlen         = 3*sizeof(int),
1338                 .mode           = 0444,
1339                 .proc_handler   = &proc_nr_files,
1340         },
1341         {
1342                 .ctl_name       = FS_MAXFILE,
1343                 .procname       = "file-max",
1344                 .data           = &files_stat.max_files,
1345                 .maxlen         = sizeof(int),
1346                 .mode           = 0644,
1347                 .proc_handler   = &proc_dointvec,
1348         },
1349         {
1350                 .ctl_name       = CTL_UNNUMBERED,
1351                 .procname       = "nr_open",
1352                 .data           = &sysctl_nr_open,
1353                 .maxlen         = sizeof(int),
1354                 .mode           = 0644,
1355                 .proc_handler   = &proc_dointvec_minmax,
1356                 .extra1         = &sysctl_nr_open_min,
1357                 .extra2         = &sysctl_nr_open_max,
1358         },
1359         {
1360                 .ctl_name       = FS_DENTRY,
1361                 .procname       = "dentry-state",
1362                 .data           = &dentry_stat,
1363                 .maxlen         = 6*sizeof(int),
1364                 .mode           = 0444,
1365                 .proc_handler   = &proc_dointvec,
1366         },
1367         {
1368                 .ctl_name       = FS_OVERFLOWUID,
1369                 .procname       = "overflowuid",
1370                 .data           = &fs_overflowuid,
1371                 .maxlen         = sizeof(int),
1372                 .mode           = 0644,
1373                 .proc_handler   = &proc_dointvec_minmax,
1374                 .strategy       = &sysctl_intvec,
1375                 .extra1         = &minolduid,
1376                 .extra2         = &maxolduid,
1377         },
1378         {
1379                 .ctl_name       = FS_OVERFLOWGID,
1380                 .procname       = "overflowgid",
1381                 .data           = &fs_overflowgid,
1382                 .maxlen         = sizeof(int),
1383                 .mode           = 0644,
1384                 .proc_handler   = &proc_dointvec_minmax,
1385                 .strategy       = &sysctl_intvec,
1386                 .extra1         = &minolduid,
1387                 .extra2         = &maxolduid,
1388         },
1389 #ifdef CONFIG_FILE_LOCKING
1390         {
1391                 .ctl_name       = FS_LEASES,
1392                 .procname       = "leases-enable",
1393                 .data           = &leases_enable,
1394                 .maxlen         = sizeof(int),
1395                 .mode           = 0644,
1396                 .proc_handler   = &proc_dointvec,
1397         },
1398 #endif
1399 #ifdef CONFIG_DNOTIFY
1400         {
1401                 .ctl_name       = FS_DIR_NOTIFY,
1402                 .procname       = "dir-notify-enable",
1403                 .data           = &dir_notify_enable,
1404                 .maxlen         = sizeof(int),
1405                 .mode           = 0644,
1406                 .proc_handler   = &proc_dointvec,
1407         },
1408 #endif
1409 #ifdef CONFIG_MMU
1410 #ifdef CONFIG_FILE_LOCKING
1411         {
1412                 .ctl_name       = FS_LEASE_TIME,
1413                 .procname       = "lease-break-time",
1414                 .data           = &lease_break_time,
1415                 .maxlen         = sizeof(int),
1416                 .mode           = 0644,
1417                 .proc_handler   = &proc_dointvec,
1418         },
1419 #endif
1420 #ifdef CONFIG_AIO
1421         {
1422                 .procname       = "aio-nr",
1423                 .data           = &aio_nr,
1424                 .maxlen         = sizeof(aio_nr),
1425                 .mode           = 0444,
1426                 .proc_handler   = &proc_doulongvec_minmax,
1427         },
1428         {
1429                 .procname       = "aio-max-nr",
1430                 .data           = &aio_max_nr,
1431                 .maxlen         = sizeof(aio_max_nr),
1432                 .mode           = 0644,
1433                 .proc_handler   = &proc_doulongvec_minmax,
1434         },
1435 #endif /* CONFIG_AIO */
1436 #ifdef CONFIG_INOTIFY_USER
1437         {
1438                 .ctl_name       = FS_INOTIFY,
1439                 .procname       = "inotify",
1440                 .mode           = 0555,
1441                 .child          = inotify_table,
1442         },
1443 #endif  
1444 #ifdef CONFIG_EPOLL
1445         {
1446                 .procname       = "epoll",
1447                 .mode           = 0555,
1448                 .child          = epoll_table,
1449         },
1450 #endif
1451 #endif
1452         {
1453                 .ctl_name       = KERN_SETUID_DUMPABLE,
1454                 .procname       = "suid_dumpable",
1455                 .data           = &suid_dumpable,
1456                 .maxlen         = sizeof(int),
1457                 .mode           = 0644,
1458                 .proc_handler   = &proc_dointvec_minmax,
1459                 .strategy       = &sysctl_intvec,
1460                 .extra1         = &zero,
1461                 .extra2         = &two,
1462         },
1463 #if defined(CONFIG_BINFMT_MISC) || defined(CONFIG_BINFMT_MISC_MODULE)
1464         {
1465                 .ctl_name       = CTL_UNNUMBERED,
1466                 .procname       = "binfmt_misc",
1467                 .mode           = 0555,
1468                 .child          = binfmt_misc_table,
1469         },
1470 #endif
1471 /*
1472  * NOTE: do not add new entries to this table unless you have read
1473  * Documentation/sysctl/ctl_unnumbered.txt
1474  */
1475         { .ctl_name = 0 }
1476 };
1477
1478 static struct ctl_table debug_table[] = {
1479 #if defined(CONFIG_X86) || defined(CONFIG_PPC)
1480         {
1481                 .ctl_name       = CTL_UNNUMBERED,
1482                 .procname       = "exception-trace",
1483                 .data           = &show_unhandled_signals,
1484                 .maxlen         = sizeof(int),
1485                 .mode           = 0644,
1486                 .proc_handler   = proc_dointvec
1487         },
1488 #endif
1489         { .ctl_name = 0 }
1490 };
1491
1492 static struct ctl_table dev_table[] = {
1493         { .ctl_name = 0 }
1494 };
1495
1496 static DEFINE_SPINLOCK(sysctl_lock);
1497
1498 /* called under sysctl_lock */
1499 static int use_table(struct ctl_table_header *p)
1500 {
1501         if (unlikely(p->unregistering))
1502                 return 0;
1503         p->used++;
1504         return 1;
1505 }
1506
1507 /* called under sysctl_lock */
1508 static void unuse_table(struct ctl_table_header *p)
1509 {
1510         if (!--p->used)
1511                 if (unlikely(p->unregistering))
1512                         complete(p->unregistering);
1513 }
1514
1515 /* called under sysctl_lock, will reacquire if has to wait */
1516 static void start_unregistering(struct ctl_table_header *p)
1517 {
1518         /*
1519          * if p->used is 0, nobody will ever touch that entry again;
1520          * we'll eliminate all paths to it before dropping sysctl_lock
1521          */
1522         if (unlikely(p->used)) {
1523                 struct completion wait;
1524                 init_completion(&wait);
1525                 p->unregistering = &wait;
1526                 spin_unlock(&sysctl_lock);
1527                 wait_for_completion(&wait);
1528                 spin_lock(&sysctl_lock);
1529         } else {
1530                 /* anything non-NULL; we'll never dereference it */
1531                 p->unregistering = ERR_PTR(-EINVAL);
1532         }
1533         /*
1534          * do not remove from the list until nobody holds it; walking the
1535          * list in do_sysctl() relies on that.
1536          */
1537         list_del_init(&p->ctl_entry);
1538 }
1539
1540 void sysctl_head_get(struct ctl_table_header *head)
1541 {
1542         spin_lock(&sysctl_lock);
1543         head->count++;
1544         spin_unlock(&sysctl_lock);
1545 }
1546
1547 void sysctl_head_put(struct ctl_table_header *head)
1548 {
1549         spin_lock(&sysctl_lock);
1550         if (!--head->count)
1551                 kfree(head);
1552         spin_unlock(&sysctl_lock);
1553 }
1554
1555 struct ctl_table_header *sysctl_head_grab(struct ctl_table_header *head)
1556 {
1557         if (!head)
1558                 BUG();
1559         spin_lock(&sysctl_lock);
1560         if (!use_table(head))
1561                 head = ERR_PTR(-ENOENT);
1562         spin_unlock(&sysctl_lock);
1563         return head;
1564 }
1565
1566 void sysctl_head_finish(struct ctl_table_header *head)
1567 {
1568         if (!head)
1569                 return;
1570         spin_lock(&sysctl_lock);
1571         unuse_table(head);
1572         spin_unlock(&sysctl_lock);
1573 }
1574
1575 static struct ctl_table_set *
1576 lookup_header_set(struct ctl_table_root *root, struct nsproxy *namespaces)
1577 {
1578         struct ctl_table_set *set = &root->default_set;
1579         if (root->lookup)
1580                 set = root->lookup(root, namespaces);
1581         return set;
1582 }
1583
1584 static struct list_head *
1585 lookup_header_list(struct ctl_table_root *root, struct nsproxy *namespaces)
1586 {
1587         struct ctl_table_set *set = lookup_header_set(root, namespaces);
1588         return &set->list;
1589 }
1590
1591 struct ctl_table_header *__sysctl_head_next(struct nsproxy *namespaces,
1592                                             struct ctl_table_header *prev)
1593 {
1594         struct ctl_table_root *root;
1595         struct list_head *header_list;
1596         struct ctl_table_header *head;
1597         struct list_head *tmp;
1598
1599         spin_lock(&sysctl_lock);
1600         if (prev) {
1601                 head = prev;
1602                 tmp = &prev->ctl_entry;
1603                 unuse_table(prev);
1604                 goto next;
1605         }
1606         tmp = &root_table_header.ctl_entry;
1607         for (;;) {
1608                 head = list_entry(tmp, struct ctl_table_header, ctl_entry);
1609
1610                 if (!use_table(head))
1611                         goto next;
1612                 spin_unlock(&sysctl_lock);
1613                 return head;
1614         next:
1615                 root = head->root;
1616                 tmp = tmp->next;
1617                 header_list = lookup_header_list(root, namespaces);
1618                 if (tmp != header_list)
1619                         continue;
1620
1621                 do {
1622                         root = list_entry(root->root_list.next,
1623                                         struct ctl_table_root, root_list);
1624                         if (root == &sysctl_table_root)
1625                                 goto out;
1626                         header_list = lookup_header_list(root, namespaces);
1627                 } while (list_empty(header_list));
1628                 tmp = header_list->next;
1629         }
1630 out:
1631         spin_unlock(&sysctl_lock);
1632         return NULL;
1633 }
1634
1635 struct ctl_table_header *sysctl_head_next(struct ctl_table_header *prev)
1636 {
1637         return __sysctl_head_next(current->nsproxy, prev);
1638 }
1639
1640 void register_sysctl_root(struct ctl_table_root *root)
1641 {
1642         spin_lock(&sysctl_lock);
1643         list_add_tail(&root->root_list, &sysctl_table_root.root_list);
1644         spin_unlock(&sysctl_lock);
1645 }
1646
1647 #ifdef CONFIG_SYSCTL_SYSCALL
1648 /* Perform the actual read/write of a sysctl table entry. */
1649 static int do_sysctl_strategy(struct ctl_table_root *root,
1650                         struct ctl_table *table,
1651                         void __user *oldval, size_t __user *oldlenp,
1652                         void __user *newval, size_t newlen)
1653 {
1654         int op = 0, rc;
1655
1656         if (oldval)
1657                 op |= MAY_READ;
1658         if (newval)
1659                 op |= MAY_WRITE;
1660         if (sysctl_perm(root, table, op))
1661                 return -EPERM;
1662
1663         if (table->strategy) {
1664                 rc = table->strategy(table, oldval, oldlenp, newval, newlen);
1665                 if (rc < 0)
1666                         return rc;
1667                 if (rc > 0)
1668                         return 0;
1669         }
1670
1671         /* If there is no strategy routine, or if the strategy returns
1672          * zero, proceed with automatic r/w */
1673         if (table->data && table->maxlen) {
1674                 rc = sysctl_data(table, oldval, oldlenp, newval, newlen);
1675                 if (rc < 0)
1676                         return rc;
1677         }
1678         return 0;
1679 }
1680
1681 static int parse_table(int __user *name, int nlen,
1682                        void __user *oldval, size_t __user *oldlenp,
1683                        void __user *newval, size_t newlen,
1684                        struct ctl_table_root *root,
1685                        struct ctl_table *table)
1686 {
1687         int n;
1688 repeat:
1689         if (!nlen)
1690                 return -ENOTDIR;
1691         if (get_user(n, name))
1692                 return -EFAULT;
1693         for ( ; table->ctl_name || table->procname; table++) {
1694                 if (!table->ctl_name)
1695                         continue;
1696                 if (n == table->ctl_name) {
1697                         int error;
1698                         if (table->child) {
1699                                 if (sysctl_perm(root, table, MAY_EXEC))
1700                                         return -EPERM;
1701                                 name++;
1702                                 nlen--;
1703                                 table = table->child;
1704                                 goto repeat;
1705                         }
1706                         error = do_sysctl_strategy(root, table,
1707                                                    oldval, oldlenp,
1708                                                    newval, newlen);
1709                         return error;
1710                 }
1711         }
1712         return -ENOTDIR;
1713 }
1714
1715 int do_sysctl(int __user *name, int nlen, void __user *oldval, size_t __user *oldlenp,
1716                void __user *newval, size_t newlen)
1717 {
1718         struct ctl_table_header *head;
1719         int error = -ENOTDIR;
1720
1721         if (nlen <= 0 || nlen >= CTL_MAXNAME)
1722                 return -ENOTDIR;
1723         if (oldval) {
1724                 int old_len;
1725                 if (!oldlenp || get_user(old_len, oldlenp))
1726                         return -EFAULT;
1727         }
1728
1729         for (head = sysctl_head_next(NULL); head;
1730                         head = sysctl_head_next(head)) {
1731                 error = parse_table(name, nlen, oldval, oldlenp, 
1732                                         newval, newlen,
1733                                         head->root, head->ctl_table);
1734                 if (error != -ENOTDIR) {
1735                         sysctl_head_finish(head);
1736                         break;
1737                 }
1738         }
1739         return error;
1740 }
1741
1742 SYSCALL_DEFINE1(sysctl, struct __sysctl_args __user *, args)
1743 {
1744         struct __sysctl_args tmp;
1745         int error;
1746
1747         if (copy_from_user(&tmp, args, sizeof(tmp)))
1748                 return -EFAULT;
1749
1750         error = deprecated_sysctl_warning(&tmp);
1751         if (error)
1752                 goto out;
1753
1754         lock_kernel();
1755         error = do_sysctl(tmp.name, tmp.nlen, tmp.oldval, tmp.oldlenp,
1756                           tmp.newval, tmp.newlen);
1757         unlock_kernel();
1758 out:
1759         return error;
1760 }
1761 #endif /* CONFIG_SYSCTL_SYSCALL */
1762
1763 /*
1764  * sysctl_perm does NOT grant the superuser all rights automatically, because
1765  * some sysctl variables are readonly even to root.
1766  */
1767
1768 static int test_perm(int mode, int op)
1769 {
1770         if (!current_euid())
1771                 mode >>= 6;
1772         else if (in_egroup_p(0))
1773                 mode >>= 3;
1774         if ((op & ~mode & (MAY_READ|MAY_WRITE|MAY_EXEC)) == 0)
1775                 return 0;
1776         return -EACCES;
1777 }
1778
1779 int sysctl_perm(struct ctl_table_root *root, struct ctl_table *table, int op)
1780 {
1781         int error;
1782         int mode;
1783
1784         error = security_sysctl(table, op & (MAY_READ | MAY_WRITE | MAY_EXEC));
1785         if (error)
1786                 return error;
1787
1788         if (root->permissions)
1789                 mode = root->permissions(root, current->nsproxy, table);
1790         else
1791                 mode = table->mode;
1792
1793         return test_perm(mode, op);
1794 }
1795
1796 static void sysctl_set_parent(struct ctl_table *parent, struct ctl_table *table)
1797 {
1798         for (; table->ctl_name || table->procname; table++) {
1799                 table->parent = parent;
1800                 if (table->child)
1801                         sysctl_set_parent(table, table->child);
1802         }
1803 }
1804
1805 static __init int sysctl_init(void)
1806 {
1807         sysctl_set_parent(NULL, root_table);
1808 #ifdef CONFIG_SYSCTL_SYSCALL_CHECK
1809         {
1810                 int err;
1811                 err = sysctl_check_table(current->nsproxy, root_table);
1812         }
1813 #endif
1814         return 0;
1815 }
1816
1817 core_initcall(sysctl_init);
1818
1819 static struct ctl_table *is_branch_in(struct ctl_table *branch,
1820                                       struct ctl_table *table)
1821 {
1822         struct ctl_table *p;
1823         const char *s = branch->procname;
1824
1825         /* branch should have named subdirectory as its first element */
1826         if (!s || !branch->child)
1827                 return NULL;
1828
1829         /* ... and nothing else */
1830         if (branch[1].procname || branch[1].ctl_name)
1831                 return NULL;
1832
1833         /* table should contain subdirectory with the same name */
1834         for (p = table; p->procname || p->ctl_name; p++) {
1835                 if (!p->child)
1836                         continue;
1837                 if (p->procname && strcmp(p->procname, s) == 0)
1838                         return p;
1839         }
1840         return NULL;
1841 }
1842
1843 /* see if attaching q to p would be an improvement */
1844 static void try_attach(struct ctl_table_header *p, struct ctl_table_header *q)
1845 {
1846         struct ctl_table *to = p->ctl_table, *by = q->ctl_table;
1847         struct ctl_table *next;
1848         int is_better = 0;
1849         int not_in_parent = !p->attached_by;
1850
1851         while ((next = is_branch_in(by, to)) != NULL) {
1852                 if (by == q->attached_by)
1853                         is_better = 1;
1854                 if (to == p->attached_by)
1855                         not_in_parent = 1;
1856                 by = by->child;
1857                 to = next->child;
1858         }
1859
1860         if (is_better && not_in_parent) {
1861                 q->attached_by = by;
1862                 q->attached_to = to;
1863                 q->parent = p;
1864         }
1865 }
1866
1867 /**
1868  * __register_sysctl_paths - register a sysctl hierarchy
1869  * @root: List of sysctl headers to register on
1870  * @namespaces: Data to compute which lists of sysctl entries are visible
1871  * @path: The path to the directory the sysctl table is in.
1872  * @table: the top-level table structure
1873  *
1874  * Register a sysctl table hierarchy. @table should be a filled in ctl_table
1875  * array. A completely 0 filled entry terminates the table.
1876  *
1877  * The members of the &struct ctl_table structure are used as follows:
1878  *
1879  * ctl_name - This is the numeric sysctl value used by sysctl(2). The number
1880  *            must be unique within that level of sysctl
1881  *
1882  * procname - the name of the sysctl file under /proc/sys. Set to %NULL to not
1883  *            enter a sysctl file
1884  *
1885  * data - a pointer to data for use by proc_handler
1886  *
1887  * maxlen - the maximum size in bytes of the data
1888  *
1889  * mode - the file permissions for the /proc/sys file, and for sysctl(2)
1890  *
1891  * child - a pointer to the child sysctl table if this entry is a directory, or
1892  *         %NULL.
1893  *
1894  * proc_handler - the text handler routine (described below)
1895  *
1896  * strategy - the strategy routine (described below)
1897  *
1898  * de - for internal use by the sysctl routines
1899  *
1900  * extra1, extra2 - extra pointers usable by the proc handler routines
1901  *
1902  * Leaf nodes in the sysctl tree will be represented by a single file
1903  * under /proc; non-leaf nodes will be represented by directories.
1904  *
1905  * sysctl(2) can automatically manage read and write requests through
1906  * the sysctl table.  The data and maxlen fields of the ctl_table
1907  * struct enable minimal validation of the values being written to be
1908  * performed, and the mode field allows minimal authentication.
1909  *
1910  * More sophisticated management can be enabled by the provision of a
1911  * strategy routine with the table entry.  This will be called before
1912  * any automatic read or write of the data is performed.
1913  *
1914  * The strategy routine may return
1915  *
1916  * < 0 - Error occurred (error is passed to user process)
1917  *
1918  * 0   - OK - proceed with automatic read or write.
1919  *
1920  * > 0 - OK - read or write has been done by the strategy routine, so
1921  *       return immediately.
1922  *
1923  * There must be a proc_handler routine for any terminal nodes
1924  * mirrored under /proc/sys (non-terminals are handled by a built-in
1925  * directory handler).  Several default handlers are available to
1926  * cover common cases -
1927  *
1928  * proc_dostring(), proc_dointvec(), proc_dointvec_jiffies(),
1929  * proc_dointvec_userhz_jiffies(), proc_dointvec_minmax(), 
1930  * proc_doulongvec_ms_jiffies_minmax(), proc_doulongvec_minmax()
1931  *
1932  * It is the handler's job to read the input buffer from user memory
1933  * and process it. The handler should return 0 on success.
1934  *
1935  * This routine returns %NULL on a failure to register, and a pointer
1936  * to the table header on success.
1937  */
1938 struct ctl_table_header *__register_sysctl_paths(
1939         struct ctl_table_root *root,
1940         struct nsproxy *namespaces,
1941         const struct ctl_path *path, struct ctl_table *table)
1942 {
1943         struct ctl_table_header *header;
1944         struct ctl_table *new, **prevp;
1945         unsigned int n, npath;
1946         struct ctl_table_set *set;
1947
1948         /* Count the path components */
1949         for (npath = 0; path[npath].ctl_name || path[npath].procname; ++npath)
1950                 ;
1951
1952         /*
1953          * For each path component, allocate a 2-element ctl_table array.
1954          * The first array element will be filled with the sysctl entry
1955          * for this, the second will be the sentinel (ctl_name == 0).
1956          *
1957          * We allocate everything in one go so that we don't have to
1958          * worry about freeing additional memory in unregister_sysctl_table.
1959          */
1960         header = kzalloc(sizeof(struct ctl_table_header) +
1961                          (2 * npath * sizeof(struct ctl_table)), GFP_KERNEL);
1962         if (!header)
1963                 return NULL;
1964
1965         new = (struct ctl_table *) (header + 1);
1966
1967         /* Now connect the dots */
1968         prevp = &header->ctl_table;
1969         for (n = 0; n < npath; ++n, ++path) {
1970                 /* Copy the procname */
1971                 new->procname = path->procname;
1972                 new->ctl_name = path->ctl_name;
1973                 new->mode     = 0555;
1974
1975                 *prevp = new;
1976                 prevp = &new->child;
1977
1978                 new += 2;
1979         }
1980         *prevp = table;
1981         header->ctl_table_arg = table;
1982
1983         INIT_LIST_HEAD(&header->ctl_entry);
1984         header->used = 0;
1985         header->unregistering = NULL;
1986         header->root = root;
1987         sysctl_set_parent(NULL, header->ctl_table);
1988         header->count = 1;
1989 #ifdef CONFIG_SYSCTL_SYSCALL_CHECK
1990         if (sysctl_check_table(namespaces, header->ctl_table)) {
1991                 kfree(header);
1992                 return NULL;
1993         }
1994 #endif
1995         spin_lock(&sysctl_lock);
1996         header->set = lookup_header_set(root, namespaces);
1997         header->attached_by = header->ctl_table;
1998         header->attached_to = root_table;
1999         header->parent = &root_table_header;
2000         for (set = header->set; set; set = set->parent) {
2001                 struct ctl_table_header *p;
2002                 list_for_each_entry(p, &set->list, ctl_entry) {
2003                         if (p->unregistering)
2004                                 continue;
2005                         try_attach(p, header);
2006                 }
2007         }
2008         header->parent->count++;
2009         list_add_tail(&header->ctl_entry, &header->set->list);
2010         spin_unlock(&sysctl_lock);
2011
2012         return header;
2013 }
2014
2015 /**
2016  * register_sysctl_table_path - register a sysctl table hierarchy
2017  * @path: The path to the directory the sysctl table is in.
2018  * @table: the top-level table structure
2019  *
2020  * Register a sysctl table hierarchy. @table should be a filled in ctl_table
2021  * array. A completely 0 filled entry terminates the table.
2022  *
2023  * See __register_sysctl_paths for more details.
2024  */
2025 struct ctl_table_header *register_sysctl_paths(const struct ctl_path *path,
2026                                                 struct ctl_table *table)
2027 {
2028         return __register_sysctl_paths(&sysctl_table_root, current->nsproxy,
2029                                         path, table);
2030 }
2031
2032 /**
2033  * register_sysctl_table - register a sysctl table hierarchy
2034  * @table: the top-level table structure
2035  *
2036  * Register a sysctl table hierarchy. @table should be a filled in ctl_table
2037  * array. A completely 0 filled entry terminates the table.
2038  *
2039  * See register_sysctl_paths for more details.
2040  */
2041 struct ctl_table_header *register_sysctl_table(struct ctl_table *table)
2042 {
2043         static const struct ctl_path null_path[] = { {} };
2044
2045         return register_sysctl_paths(null_path, table);
2046 }
2047
2048 /**
2049  * unregister_sysctl_table - unregister a sysctl table hierarchy
2050  * @header: the header returned from register_sysctl_table
2051  *
2052  * Unregisters the sysctl table and all children. proc entries may not
2053  * actually be removed until they are no longer used by anyone.
2054  */
2055 void unregister_sysctl_table(struct ctl_table_header * header)
2056 {
2057         might_sleep();
2058
2059         if (header == NULL)
2060                 return;
2061
2062         spin_lock(&sysctl_lock);
2063         start_unregistering(header);
2064         if (!--header->parent->count) {
2065                 WARN_ON(1);
2066                 kfree(header->parent);
2067         }
2068         if (!--header->count)
2069                 kfree(header);
2070         spin_unlock(&sysctl_lock);
2071 }
2072
2073 int sysctl_is_seen(struct ctl_table_header *p)
2074 {
2075         struct ctl_table_set *set = p->set;
2076         int res;
2077         spin_lock(&sysctl_lock);
2078         if (p->unregistering)
2079                 res = 0;
2080         else if (!set->is_seen)
2081                 res = 1;
2082         else
2083                 res = set->is_seen(set);
2084         spin_unlock(&sysctl_lock);
2085         return res;
2086 }
2087
2088 void setup_sysctl_set(struct ctl_table_set *p,
2089         struct ctl_table_set *parent,
2090         int (*is_seen)(struct ctl_table_set *))
2091 {
2092         INIT_LIST_HEAD(&p->list);
2093         p->parent = parent ? parent : &sysctl_table_root.default_set;
2094         p->is_seen = is_seen;
2095 }
2096
2097 #else /* !CONFIG_SYSCTL */
2098 struct ctl_table_header *register_sysctl_table(struct ctl_table * table)
2099 {
2100         return NULL;
2101 }
2102
2103 struct ctl_table_header *register_sysctl_paths(const struct ctl_path *path,
2104                                                     struct ctl_table *table)
2105 {
2106         return NULL;
2107 }
2108
2109 void unregister_sysctl_table(struct ctl_table_header * table)
2110 {
2111 }
2112
2113 void setup_sysctl_set(struct ctl_table_set *p,
2114         struct ctl_table_set *parent,
2115         int (*is_seen)(struct ctl_table_set *))
2116 {
2117 }
2118
2119 void sysctl_head_put(struct ctl_table_header *head)
2120 {
2121 }
2122
2123 #endif /* CONFIG_SYSCTL */
2124
2125 /*
2126  * /proc/sys support
2127  */
2128
2129 #ifdef CONFIG_PROC_SYSCTL
2130
2131 static int _proc_do_string(void* data, int maxlen, int write,
2132                            struct file *filp, void __user *buffer,
2133                            size_t *lenp, loff_t *ppos)
2134 {
2135         size_t len;
2136         char __user *p;
2137         char c;
2138
2139         if (!data || !maxlen || !*lenp) {
2140                 *lenp = 0;
2141                 return 0;
2142         }
2143
2144         if (write) {
2145                 len = 0;
2146                 p = buffer;
2147                 while (len < *lenp) {
2148                         if (get_user(c, p++))
2149                                 return -EFAULT;
2150                         if (c == 0 || c == '\n')
2151                                 break;
2152                         len++;
2153                 }
2154                 if (len >= maxlen)
2155                         len = maxlen-1;
2156                 if(copy_from_user(data, buffer, len))
2157                         return -EFAULT;
2158                 ((char *) data)[len] = 0;
2159                 *ppos += *lenp;
2160         } else {
2161                 len = strlen(data);
2162                 if (len > maxlen)
2163                         len = maxlen;
2164
2165                 if (*ppos > len) {
2166                         *lenp = 0;
2167                         return 0;
2168                 }
2169
2170                 data += *ppos;
2171                 len  -= *ppos;
2172
2173                 if (len > *lenp)
2174                         len = *lenp;
2175                 if (len)
2176                         if(copy_to_user(buffer, data, len))
2177                                 return -EFAULT;
2178                 if (len < *lenp) {
2179                         if(put_user('\n', ((char __user *) buffer) + len))
2180                                 return -EFAULT;
2181                         len++;
2182                 }
2183                 *lenp = len;
2184                 *ppos += len;
2185         }
2186         return 0;
2187 }
2188
2189 /**
2190  * proc_dostring - read a string sysctl
2191  * @table: the sysctl table
2192  * @write: %TRUE if this is a write to the sysctl file
2193  * @filp: the file structure
2194  * @buffer: the user buffer
2195  * @lenp: the size of the user buffer
2196  * @ppos: file position
2197  *
2198  * Reads/writes a string from/to the user buffer. If the kernel
2199  * buffer provided is not large enough to hold the string, the
2200  * string is truncated. The copied string is %NULL-terminated.
2201  * If the string is being read by the user process, it is copied
2202  * and a newline '\n' is added. It is truncated if the buffer is
2203  * not large enough.
2204  *
2205  * Returns 0 on success.
2206  */
2207 int proc_dostring(struct ctl_table *table, int write, struct file *filp,
2208                   void __user *buffer, size_t *lenp, loff_t *ppos)
2209 {
2210         return _proc_do_string(table->data, table->maxlen, write, filp,
2211                                buffer, lenp, ppos);
2212 }
2213
2214
2215 static int do_proc_dointvec_conv(int *negp, unsigned long *lvalp,
2216                                  int *valp,
2217                                  int write, void *data)
2218 {
2219         if (write) {
2220                 *valp = *negp ? -*lvalp : *lvalp;
2221         } else {
2222                 int val = *valp;
2223                 if (val < 0) {
2224                         *negp = -1;
2225                         *lvalp = (unsigned long)-val;
2226                 } else {
2227                         *negp = 0;
2228                         *lvalp = (unsigned long)val;
2229                 }
2230         }
2231         return 0;
2232 }
2233
2234 static int __do_proc_dointvec(void *tbl_data, struct ctl_table *table,
2235                   int write, struct file *filp, void __user *buffer,
2236                   size_t *lenp, loff_t *ppos,
2237                   int (*conv)(int *negp, unsigned long *lvalp, int *valp,
2238                               int write, void *data),
2239                   void *data)
2240 {
2241 #define TMPBUFLEN 21
2242         int *i, vleft, first=1, neg, val;
2243         unsigned long lval;
2244         size_t left, len;
2245         
2246         char buf[TMPBUFLEN], *p;
2247         char __user *s = buffer;
2248         
2249         if (!tbl_data || !table->maxlen || !*lenp ||
2250             (*ppos && !write)) {
2251                 *lenp = 0;
2252                 return 0;
2253         }
2254         
2255         i = (int *) tbl_data;
2256         vleft = table->maxlen / sizeof(*i);
2257         left = *lenp;
2258
2259         if (!conv)
2260                 conv = do_proc_dointvec_conv;
2261
2262         for (; left && vleft--; i++, first=0) {
2263                 if (write) {
2264                         while (left) {
2265                                 char c;
2266                                 if (get_user(c, s))
2267                                         return -EFAULT;
2268                                 if (!isspace(c))
2269                                         break;
2270                                 left--;
2271                                 s++;
2272                         }
2273                         if (!left)
2274                                 break;
2275                         neg = 0;
2276                         len = left;
2277                         if (len > sizeof(buf) - 1)
2278                                 len = sizeof(buf) - 1;
2279                         if (copy_from_user(buf, s, len))
2280                                 return -EFAULT;
2281                         buf[len] = 0;
2282                         p = buf;
2283                         if (*p == '-' && left > 1) {
2284                                 neg = 1;
2285                                 p++;
2286                         }
2287                         if (*p < '0' || *p > '9')
2288                                 break;
2289
2290                         lval = simple_strtoul(p, &p, 0);
2291
2292                         len = p-buf;
2293                         if ((len < left) && *p && !isspace(*p))
2294                                 break;
2295                         if (neg)
2296                                 val = -val;
2297                         s += len;
2298                         left -= len;
2299
2300                         if (conv(&neg, &lval, i, 1, data))
2301                                 break;
2302                 } else {
2303                         p = buf;
2304                         if (!first)
2305                                 *p++ = '\t';
2306         
2307                         if (conv(&neg, &lval, i, 0, data))
2308                                 break;
2309
2310                         sprintf(p, "%s%lu", neg ? "-" : "", lval);
2311                         len = strlen(buf);
2312                         if (len > left)
2313                                 len = left;
2314                         if(copy_to_user(s, buf, len))
2315                                 return -EFAULT;
2316                         left -= len;
2317                         s += len;
2318                 }
2319         }
2320
2321         if (!write && !first && left) {
2322                 if(put_user('\n', s))
2323                         return -EFAULT;
2324                 left--, s++;
2325         }
2326         if (write) {
2327                 while (left) {
2328                         char c;
2329                         if (get_user(c, s++))
2330                                 return -EFAULT;
2331                         if (!isspace(c))
2332                                 break;
2333                         left--;
2334                 }
2335         }
2336         if (write && first)
2337                 return -EINVAL;
2338         *lenp -= left;
2339         *ppos += *lenp;
2340         return 0;
2341 #undef TMPBUFLEN
2342 }
2343
2344 static int do_proc_dointvec(struct ctl_table *table, int write, struct file *filp,
2345                   void __user *buffer, size_t *lenp, loff_t *ppos,
2346                   int (*conv)(int *negp, unsigned long *lvalp, int *valp,
2347                               int write, void *data),
2348                   void *data)
2349 {
2350         return __do_proc_dointvec(table->data, table, write, filp,
2351                         buffer, lenp, ppos, conv, data);
2352 }
2353
2354 /**
2355  * proc_dointvec - read a vector of integers
2356  * @table: the sysctl table
2357  * @write: %TRUE if this is a write to the sysctl file
2358  * @filp: the file structure
2359  * @buffer: the user buffer
2360  * @lenp: the size of the user buffer
2361  * @ppos: file position
2362  *
2363  * Reads/writes up to table->maxlen/sizeof(unsigned int) integer
2364  * values from/to the user buffer, treated as an ASCII string. 
2365  *
2366  * Returns 0 on success.
2367  */
2368 int proc_dointvec(struct ctl_table *table, int write, struct file *filp,
2369                      void __user *buffer, size_t *lenp, loff_t *ppos)
2370 {
2371     return do_proc_dointvec(table,write,filp,buffer,lenp,ppos,
2372                             NULL,NULL);
2373 }
2374
2375 /*
2376  * Taint values can only be increased
2377  * This means we can safely use a temporary.
2378  */
2379 static int proc_taint(struct ctl_table *table, int write, struct file *filp,
2380                                void __user *buffer, size_t *lenp, loff_t *ppos)
2381 {
2382         struct ctl_table t;
2383         unsigned long tmptaint = get_taint();
2384         int err;
2385
2386         if (write && !capable(CAP_SYS_ADMIN))
2387                 return -EPERM;
2388
2389         t = *table;
2390         t.data = &tmptaint;
2391         err = proc_doulongvec_minmax(&t, write, filp, buffer, lenp, ppos);
2392         if (err < 0)
2393                 return err;
2394
2395         if (write) {
2396                 /*
2397                  * Poor man's atomic or. Not worth adding a primitive
2398                  * to everyone's atomic.h for this
2399                  */
2400                 int i;
2401                 for (i = 0; i < BITS_PER_LONG && tmptaint >> i; i++) {
2402                         if ((tmptaint >> i) & 1)
2403                                 add_taint(i);
2404                 }
2405         }
2406
2407         return err;
2408 }
2409
2410 struct do_proc_dointvec_minmax_conv_param {
2411         int *min;
2412         int *max;
2413 };
2414
2415 static int do_proc_dointvec_minmax_conv(int *negp, unsigned long *lvalp, 
2416                                         int *valp, 
2417                                         int write, void *data)
2418 {
2419         struct do_proc_dointvec_minmax_conv_param *param = data;
2420         if (write) {
2421                 int val = *negp ? -*lvalp : *lvalp;
2422                 if ((param->min && *param->min > val) ||
2423                     (param->max && *param->max < val))
2424                         return -EINVAL;
2425                 *valp = val;
2426         } else {
2427                 int val = *valp;
2428                 if (val < 0) {
2429                         *negp = -1;
2430                         *lvalp = (unsigned long)-val;
2431                 } else {
2432                         *negp = 0;
2433                         *lvalp = (unsigned long)val;
2434                 }
2435         }
2436         return 0;
2437 }
2438
2439 /**
2440  * proc_dointvec_minmax - read a vector of integers with min/max values
2441  * @table: the sysctl table
2442  * @write: %TRUE if this is a write to the sysctl file
2443  * @filp: the file structure
2444  * @buffer: the user buffer
2445  * @lenp: the size of the user buffer
2446  * @ppos: file position
2447  *
2448  * Reads/writes up to table->maxlen/sizeof(unsigned int) integer
2449  * values from/to the user buffer, treated as an ASCII string.
2450  *
2451  * This routine will ensure the values are within the range specified by
2452  * table->extra1 (min) and table->extra2 (max).
2453  *
2454  * Returns 0 on success.
2455  */
2456 int proc_dointvec_minmax(struct ctl_table *table, int write, struct file *filp,
2457                   void __user *buffer, size_t *lenp, loff_t *ppos)
2458 {
2459         struct do_proc_dointvec_minmax_conv_param param = {
2460                 .min = (int *) table->extra1,
2461                 .max = (int *) table->extra2,
2462         };
2463         return do_proc_dointvec(table, write, filp, buffer, lenp, ppos,
2464                                 do_proc_dointvec_minmax_conv, &param);
2465 }
2466
2467 static int __do_proc_doulongvec_minmax(void *data, struct ctl_table *table, int write,
2468                                      struct file *filp,
2469                                      void __user *buffer,
2470                                      size_t *lenp, loff_t *ppos,
2471                                      unsigned long convmul,
2472                                      unsigned long convdiv)
2473 {
2474 #define TMPBUFLEN 21
2475         unsigned long *i, *min, *max, val;
2476         int vleft, first=1, neg;
2477         size_t len, left;
2478         char buf[TMPBUFLEN], *p;
2479         char __user *s = buffer;
2480         
2481         if (!data || !table->maxlen || !*lenp ||
2482             (*ppos && !write)) {
2483                 *lenp = 0;
2484                 return 0;
2485         }
2486         
2487         i = (unsigned long *) data;
2488         min = (unsigned long *) table->extra1;
2489         max = (unsigned long *) table->extra2;
2490         vleft = table->maxlen / sizeof(unsigned long);
2491         left = *lenp;
2492         
2493         for (; left && vleft--; i++, min++, max++, first=0) {
2494                 if (write) {
2495                         while (left) {
2496                                 char c;
2497                                 if (get_user(c, s))
2498                                         return -EFAULT;
2499                                 if (!isspace(c))
2500                                         break;
2501                                 left--;
2502                                 s++;
2503                         }
2504                         if (!left)
2505                                 break;
2506                         neg = 0;
2507                         len = left;
2508                         if (len > TMPBUFLEN-1)
2509                                 len = TMPBUFLEN-1;
2510                         if (copy_from_user(buf, s, len))
2511                                 return -EFAULT;
2512                         buf[len] = 0;
2513                         p = buf;
2514                         if (*p == '-' && left > 1) {
2515                                 neg = 1;
2516                                 p++;
2517                         }
2518                         if (*p < '0' || *p > '9')
2519                                 break;
2520                         val = simple_strtoul(p, &p, 0) * convmul / convdiv ;
2521                         len = p-buf;
2522                         if ((len < left) && *p && !isspace(*p))
2523                                 break;
2524                         if (neg)
2525                                 val = -val;
2526                         s += len;
2527                         left -= len;
2528
2529                         if(neg)
2530                                 continue;
2531                         if ((min && val < *min) || (max && val > *max))
2532                                 continue;
2533                         *i = val;
2534                 } else {
2535                         p = buf;
2536                         if (!first)
2537                                 *p++ = '\t';
2538                         sprintf(p, "%lu", convdiv * (*i) / convmul);
2539                         len = strlen(buf);
2540                         if (len > left)
2541                                 len = left;
2542                         if(copy_to_user(s, buf, len))
2543                                 return -EFAULT;
2544                         left -= len;
2545                         s += len;
2546                 }
2547         }
2548
2549         if (!write && !first && left) {
2550                 if(put_user('\n', s))
2551                         return -EFAULT;
2552                 left--, s++;
2553         }
2554         if (write) {
2555                 while (left) {
2556                         char c;
2557                         if (get_user(c, s++))
2558                                 return -EFAULT;
2559                         if (!isspace(c))
2560                                 break;
2561                         left--;
2562                 }
2563         }
2564         if (write && first)
2565                 return -EINVAL;
2566         *lenp -= left;
2567         *ppos += *lenp;
2568         return 0;
2569 #undef TMPBUFLEN
2570 }
2571
2572 static int do_proc_doulongvec_minmax(struct ctl_table *table, int write,
2573                                      struct file *filp,
2574                                      void __user *buffer,
2575                                      size_t *lenp, loff_t *ppos,
2576                                      unsigned long convmul,
2577                                      unsigned long convdiv)
2578 {
2579         return __do_proc_doulongvec_minmax(table->data, table, write,
2580                         filp, buffer, lenp, ppos, convmul, convdiv);
2581 }
2582
2583 /**
2584  * proc_doulongvec_minmax - read a vector of long integers with min/max values
2585  * @table: the sysctl table
2586  * @write: %TRUE if this is a write to the sysctl file
2587  * @filp: the file structure
2588  * @buffer: the user buffer
2589  * @lenp: the size of the user buffer
2590  * @ppos: file position
2591  *
2592  * Reads/writes up to table->maxlen/sizeof(unsigned long) unsigned long
2593  * values from/to the user buffer, treated as an ASCII string.
2594  *
2595  * This routine will ensure the values are within the range specified by
2596  * table->extra1 (min) and table->extra2 (max).
2597  *
2598  * Returns 0 on success.
2599  */
2600 int proc_doulongvec_minmax(struct ctl_table *table, int write, struct file *filp,
2601                            void __user *buffer, size_t *lenp, loff_t *ppos)
2602 {
2603     return do_proc_doulongvec_minmax(table, write, filp, buffer, lenp, ppos, 1l, 1l);
2604 }
2605
2606 /**
2607  * proc_doulongvec_ms_jiffies_minmax - read a vector of millisecond values with min/max values
2608  * @table: the sysctl table
2609  * @write: %TRUE if this is a write to the sysctl file
2610  * @filp: the file structure
2611  * @buffer: the user buffer
2612  * @lenp: the size of the user buffer
2613  * @ppos: file position
2614  *
2615  * Reads/writes up to table->maxlen/sizeof(unsigned long) unsigned long
2616  * values from/to the user buffer, treated as an ASCII string. The values
2617  * are treated as milliseconds, and converted to jiffies when they are stored.
2618  *
2619  * This routine will ensure the values are within the range specified by
2620  * table->extra1 (min) and table->extra2 (max).
2621  *
2622  * Returns 0 on success.
2623  */
2624 int proc_doulongvec_ms_jiffies_minmax(struct ctl_table *table, int write,
2625                                       struct file *filp,
2626                                       void __user *buffer,
2627                                       size_t *lenp, loff_t *ppos)
2628 {
2629     return do_proc_doulongvec_minmax(table, write, filp, buffer,
2630                                      lenp, ppos, HZ, 1000l);
2631 }
2632
2633
2634 static int do_proc_dointvec_jiffies_conv(int *negp, unsigned long *lvalp,
2635                                          int *valp,
2636                                          int write, void *data)
2637 {
2638         if (write) {
2639                 if (*lvalp > LONG_MAX / HZ)
2640                         return 1;
2641                 *valp = *negp ? -(*lvalp*HZ) : (*lvalp*HZ);
2642         } else {
2643                 int val = *valp;
2644                 unsigned long lval;
2645                 if (val < 0) {
2646                         *negp = -1;
2647                         lval = (unsigned long)-val;
2648                 } else {
2649                         *negp = 0;
2650                         lval = (unsigned long)val;
2651                 }
2652                 *lvalp = lval / HZ;
2653         }
2654         return 0;
2655 }
2656
2657 static int do_proc_dointvec_userhz_jiffies_conv(int *negp, unsigned long *lvalp,
2658                                                 int *valp,
2659                                                 int write, void *data)
2660 {
2661         if (write) {
2662                 if (USER_HZ < HZ && *lvalp > (LONG_MAX / HZ) * USER_HZ)
2663                         return 1;
2664                 *valp = clock_t_to_jiffies(*negp ? -*lvalp : *lvalp);
2665         } else {
2666                 int val = *valp;
2667                 unsigned long lval;
2668                 if (val < 0) {
2669                         *negp = -1;
2670                         lval = (unsigned long)-val;
2671                 } else {
2672                         *negp = 0;
2673                         lval = (unsigned long)val;
2674                 }
2675                 *lvalp = jiffies_to_clock_t(lval);
2676         }
2677         return 0;
2678 }
2679
2680 static int do_proc_dointvec_ms_jiffies_conv(int *negp, unsigned long *lvalp,
2681                                             int *valp,
2682                                             int write, void *data)
2683 {
2684         if (write) {
2685                 *valp = msecs_to_jiffies(*negp ? -*lvalp : *lvalp);
2686         } else {
2687                 int val = *valp;
2688                 unsigned long lval;
2689                 if (val < 0) {
2690                         *negp = -1;
2691                         lval = (unsigned long)-val;
2692                 } else {
2693                         *negp = 0;
2694                         lval = (unsigned long)val;
2695                 }
2696                 *lvalp = jiffies_to_msecs(lval);
2697         }
2698         return 0;
2699 }
2700
2701 /**
2702  * proc_dointvec_jiffies - read a vector of integers as seconds
2703  * @table: the sysctl table
2704  * @write: %TRUE if this is a write to the sysctl file
2705  * @filp: the file structure
2706  * @buffer: the user buffer
2707  * @lenp: the size of the user buffer
2708  * @ppos: file position
2709  *
2710  * Reads/writes up to table->maxlen/sizeof(unsigned int) integer
2711  * values from/to the user buffer, treated as an ASCII string. 
2712  * The values read are assumed to be in seconds, and are converted into
2713  * jiffies.
2714  *
2715  * Returns 0 on success.
2716  */
2717 int proc_dointvec_jiffies(struct ctl_table *table, int write, struct file *filp,
2718                           void __user *buffer, size_t *lenp, loff_t *ppos)
2719 {
2720     return do_proc_dointvec(table,write,filp,buffer,lenp,ppos,
2721                             do_proc_dointvec_jiffies_conv,NULL);
2722 }
2723
2724 /**
2725  * proc_dointvec_userhz_jiffies - read a vector of integers as 1/USER_HZ seconds
2726  * @table: the sysctl table
2727  * @write: %TRUE if this is a write to the sysctl file
2728  * @filp: the file structure
2729  * @buffer: the user buffer
2730  * @lenp: the size of the user buffer
2731  * @ppos: pointer to the file position
2732  *
2733  * Reads/writes up to table->maxlen/sizeof(unsigned int) integer
2734  * values from/to the user buffer, treated as an ASCII string. 
2735  * The values read are assumed to be in 1/USER_HZ seconds, and 
2736  * are converted into jiffies.
2737  *
2738  * Returns 0 on success.
2739  */
2740 int proc_dointvec_userhz_jiffies(struct ctl_table *table, int write, struct file *filp,
2741                                  void __user *buffer, size_t *lenp, loff_t *ppos)
2742 {
2743     return do_proc_dointvec(table,write,filp,buffer,lenp,ppos,
2744                             do_proc_dointvec_userhz_jiffies_conv,NULL);
2745 }
2746
2747 /**
2748  * proc_dointvec_ms_jiffies - read a vector of integers as 1 milliseconds
2749  * @table: the sysctl table
2750  * @write: %TRUE if this is a write to the sysctl file
2751  * @filp: the file structure
2752  * @buffer: the user buffer
2753  * @lenp: the size of the user buffer
2754  * @ppos: file position
2755  * @ppos: the current position in the file
2756  *
2757  * Reads/writes up to table->maxlen/sizeof(unsigned int) integer
2758  * values from/to the user buffer, treated as an ASCII string. 
2759  * The values read are assumed to be in 1/1000 seconds, and 
2760  * are converted into jiffies.
2761  *
2762  * Returns 0 on success.
2763  */
2764 int proc_dointvec_ms_jiffies(struct ctl_table *table, int write, struct file *filp,
2765                              void __user *buffer, size_t *lenp, loff_t *ppos)
2766 {
2767         return do_proc_dointvec(table, write, filp, buffer, lenp, ppos,
2768                                 do_proc_dointvec_ms_jiffies_conv, NULL);
2769 }
2770
2771 static int proc_do_cad_pid(struct ctl_table *table, int write, struct file *filp,
2772                            void __user *buffer, size_t *lenp, loff_t *ppos)
2773 {
2774         struct pid *new_pid;
2775         pid_t tmp;
2776         int r;
2777
2778         tmp = pid_vnr(cad_pid);
2779
2780         r = __do_proc_dointvec(&tmp, table, write, filp, buffer,
2781                                lenp, ppos, NULL, NULL);
2782         if (r || !write)
2783                 return r;
2784
2785         new_pid = find_get_pid(tmp);
2786         if (!new_pid)
2787                 return -ESRCH;
2788
2789         put_pid(xchg(&cad_pid, new_pid));
2790         return 0;
2791 }
2792
2793 #else /* CONFIG_PROC_FS */
2794
2795 int proc_dostring(struct ctl_table *table, int write, struct file *filp,
2796                   void __user *buffer, size_t *lenp, loff_t *ppos)
2797 {
2798         return -ENOSYS;
2799 }
2800
2801 int proc_dointvec(struct ctl_table *table, int write, struct file *filp,
2802                   void __user *buffer, size_t *lenp, loff_t *ppos)
2803 {
2804         return -ENOSYS;
2805 }
2806
2807 int proc_dointvec_minmax(struct ctl_table *table, int write, struct file *filp,
2808                     void __user *buffer, size_t *lenp, loff_t *ppos)
2809 {
2810         return -ENOSYS;
2811 }
2812
2813 int proc_dointvec_jiffies(struct ctl_table *table, int write, struct file *filp,
2814                     void __user *buffer, size_t *lenp, loff_t *ppos)
2815 {
2816         return -ENOSYS;
2817 }
2818
2819 int proc_dointvec_userhz_jiffies(struct ctl_table *table, int write, struct file *filp,
2820                     void __user *buffer, size_t *lenp, loff_t *ppos)
2821 {
2822         return -ENOSYS;
2823 }
2824
2825 int proc_dointvec_ms_jiffies(struct ctl_table *table, int write, struct file *filp,
2826                              void __user *buffer, size_t *lenp, loff_t *ppos)
2827 {
2828         return -ENOSYS;
2829 }
2830
2831 int proc_doulongvec_minmax(struct ctl_table *table, int write, struct file *filp,
2832                     void __user *buffer, size_t *lenp, loff_t *ppos)
2833 {
2834         return -ENOSYS;
2835 }
2836
2837 int proc_doulongvec_ms_jiffies_minmax(struct ctl_table *table, int write,
2838                                       struct file *filp,
2839                                       void __user *buffer,
2840                                       size_t *lenp, loff_t *ppos)
2841 {
2842     return -ENOSYS;
2843 }
2844
2845
2846 #endif /* CONFIG_PROC_FS */
2847
2848
2849 #ifdef CONFIG_SYSCTL_SYSCALL
2850 /*
2851  * General sysctl support routines 
2852  */
2853
2854 /* The generic sysctl data routine (used if no strategy routine supplied) */
2855 int sysctl_data(struct ctl_table *table,
2856                 void __user *oldval, size_t __user *oldlenp,
2857                 void __user *newval, size_t newlen)
2858 {
2859         size_t len;
2860
2861         /* Get out of I don't have a variable */
2862         if (!table->data || !table->maxlen)
2863                 return -ENOTDIR;
2864
2865         if (oldval && oldlenp) {
2866                 if (get_user(len, oldlenp))
2867                         return -EFAULT;
2868                 if (len) {
2869                         if (len > table->maxlen)
2870                                 len = table->maxlen;
2871                         if (copy_to_user(oldval, table->data, len))
2872                                 return -EFAULT;
2873                         if (put_user(len, oldlenp))
2874                                 return -EFAULT;
2875                 }
2876         }
2877
2878         if (newval && newlen) {
2879                 if (newlen > table->maxlen)
2880                         newlen = table->maxlen;
2881
2882                 if (copy_from_user(table->data, newval, newlen))
2883                         return -EFAULT;
2884         }
2885         return 1;
2886 }
2887
2888 /* The generic string strategy routine: */
2889 int sysctl_string(struct ctl_table *table,
2890                   void __user *oldval, size_t __user *oldlenp,
2891                   void __user *newval, size_t newlen)
2892 {
2893         if (!table->data || !table->maxlen) 
2894                 return -ENOTDIR;
2895         
2896         if (oldval && oldlenp) {
2897                 size_t bufsize;
2898                 if (get_user(bufsize, oldlenp))
2899                         return -EFAULT;
2900                 if (bufsize) {
2901                         size_t len = strlen(table->data), copied;
2902
2903                         /* This shouldn't trigger for a well-formed sysctl */
2904                         if (len > table->maxlen)
2905                                 len = table->maxlen;
2906
2907                         /* Copy up to a max of bufsize-1 bytes of the string */
2908                         copied = (len >= bufsize) ? bufsize - 1 : len;
2909
2910                         if (copy_to_user(oldval, table->data, copied) ||
2911                             put_user(0, (char __user *)(oldval + copied)))
2912                                 return -EFAULT;
2913                         if (put_user(len, oldlenp))
2914                                 return -EFAULT;
2915                 }
2916         }
2917         if (newval && newlen) {
2918                 size_t len = newlen;
2919                 if (len > table->maxlen)
2920                         len = table->maxlen;
2921                 if(copy_from_user(table->data, newval, len))
2922                         return -EFAULT;
2923                 if (len == table->maxlen)
2924                         len--;
2925                 ((char *) table->data)[len] = 0;
2926         }
2927         return 1;
2928 }
2929
2930 /*
2931  * This function makes sure that all of the integers in the vector
2932  * are between the minimum and maximum values given in the arrays
2933  * table->extra1 and table->extra2, respectively.
2934  */
2935 int sysctl_intvec(struct ctl_table *table,
2936                 void __user *oldval, size_t __user *oldlenp,
2937                 void __user *newval, size_t newlen)
2938 {
2939
2940         if (newval && newlen) {
2941                 int __user *vec = (int __user *) newval;
2942                 int *min = (int *) table->extra1;
2943                 int *max = (int *) table->extra2;
2944                 size_t length;
2945                 int i;
2946
2947                 if (newlen % sizeof(int) != 0)
2948                         return -EINVAL;
2949
2950                 if (!table->extra1 && !table->extra2)
2951                         return 0;
2952
2953                 if (newlen > table->maxlen)
2954                         newlen = table->maxlen;
2955                 length = newlen / sizeof(int);
2956
2957                 for (i = 0; i < length; i++) {
2958                         int value;
2959                         if (get_user(value, vec + i))
2960                                 return -EFAULT;
2961                         if (min && value < min[i])
2962                                 return -EINVAL;
2963                         if (max && value > max[i])
2964                                 return -EINVAL;
2965                 }
2966         }
2967         return 0;
2968 }
2969
2970 /* Strategy function to convert jiffies to seconds */ 
2971 int sysctl_jiffies(struct ctl_table *table,
2972                 void __user *oldval, size_t __user *oldlenp,
2973                 void __user *newval, size_t newlen)
2974 {
2975         if (oldval && oldlenp) {
2976                 size_t olen;
2977
2978                 if (get_user(olen, oldlenp))
2979                         return -EFAULT;
2980                 if (olen) {
2981                         int val;
2982
2983                         if (olen < sizeof(int))
2984                                 return -EINVAL;
2985
2986                         val = *(int *)(table->data) / HZ;
2987                         if (put_user(val, (int __user *)oldval))
2988                                 return -EFAULT;
2989                         if (put_user(sizeof(int), oldlenp))
2990                                 return -EFAULT;
2991                 }
2992         }
2993         if (newval && newlen) { 
2994                 int new;
2995                 if (newlen != sizeof(int))
2996                         return -EINVAL; 
2997                 if (get_user(new, (int __user *)newval))
2998                         return -EFAULT;
2999                 *(int *)(table->data) = new*HZ; 
3000         }
3001         return 1;
3002 }
3003
3004 /* Strategy function to convert jiffies to seconds */ 
3005 int sysctl_ms_jiffies(struct ctl_table *table,
3006                 void __user *oldval, size_t __user *oldlenp,
3007                 void __user *newval, size_t newlen)
3008 {
3009         if (oldval && oldlenp) {
3010                 size_t olen;
3011
3012                 if (get_user(olen, oldlenp))
3013                         return -EFAULT;
3014                 if (olen) {
3015                         int val;
3016
3017                         if (olen < sizeof(int))
3018                                 return -EINVAL;
3019
3020                         val = jiffies_to_msecs(*(int *)(table->data));
3021                         if (put_user(val, (int __user *)oldval))
3022                                 return -EFAULT;
3023                         if (put_user(sizeof(int), oldlenp))
3024                                 return -EFAULT;
3025                 }
3026         }
3027         if (newval && newlen) { 
3028                 int new;
3029                 if (newlen != sizeof(int))
3030                         return -EINVAL; 
3031                 if (get_user(new, (int __user *)newval))
3032                         return -EFAULT;
3033                 *(int *)(table->data) = msecs_to_jiffies(new);
3034         }
3035         return 1;
3036 }
3037
3038
3039
3040 #else /* CONFIG_SYSCTL_SYSCALL */
3041
3042
3043 SYSCALL_DEFINE1(sysctl, struct __sysctl_args __user *, args)
3044 {
3045         struct __sysctl_args tmp;
3046         int error;
3047
3048         if (copy_from_user(&tmp, args, sizeof(tmp)))
3049                 return -EFAULT;
3050
3051         error = deprecated_sysctl_warning(&tmp);
3052
3053         /* If no error reading the parameters then just -ENOSYS ... */
3054         if (!error)
3055                 error = -ENOSYS;
3056
3057         return error;
3058 }
3059
3060 int sysctl_data(struct ctl_table *table,
3061                   void __user *oldval, size_t __user *oldlenp,
3062                   void __user *newval, size_t newlen)
3063 {
3064         return -ENOSYS;
3065 }
3066
3067 int sysctl_string(struct ctl_table *table,
3068                   void __user *oldval, size_t __user *oldlenp,
3069                   void __user *newval, size_t newlen)
3070 {
3071         return -ENOSYS;
3072 }
3073
3074 int sysctl_intvec(struct ctl_table *table,
3075                 void __user *oldval, size_t __user *oldlenp,
3076                 void __user *newval, size_t newlen)
3077 {
3078         return -ENOSYS;
3079 }
3080
3081 int sysctl_jiffies(struct ctl_table *table,
3082                 void __user *oldval, size_t __user *oldlenp,
3083                 void __user *newval, size_t newlen)
3084 {
3085         return -ENOSYS;
3086 }
3087
3088 int sysctl_ms_jiffies(struct ctl_table *table,
3089                 void __user *oldval, size_t __user *oldlenp,
3090                 void __user *newval, size_t newlen)
3091 {
3092         return -ENOSYS;
3093 }
3094
3095 #endif /* CONFIG_SYSCTL_SYSCALL */
3096
3097 static int deprecated_sysctl_warning(struct __sysctl_args *args)
3098 {
3099         static int msg_count;
3100         int name[CTL_MAXNAME];
3101         int i;
3102
3103         /* Check args->nlen. */
3104         if (args->nlen < 0 || args->nlen > CTL_MAXNAME)
3105                 return -ENOTDIR;
3106
3107         /* Read in the sysctl name for better debug message logging */
3108         for (i = 0; i < args->nlen; i++)
3109                 if (get_user(name[i], args->name + i))
3110                         return -EFAULT;
3111
3112         /* Ignore accesses to kernel.version */
3113         if ((args->nlen == 2) && (name[0] == CTL_KERN) && (name[1] == KERN_VERSION))
3114                 return 0;
3115
3116         if (msg_count < 5) {
3117                 msg_count++;
3118                 printk(KERN_INFO
3119                         "warning: process `%s' used the deprecated sysctl "
3120                         "system call with ", current->comm);
3121                 for (i = 0; i < args->nlen; i++)
3122                         printk("%d.", name[i]);
3123                 printk("\n");
3124         }
3125         return 0;
3126 }
3127
3128 /*
3129  * No sense putting this after each symbol definition, twice,
3130  * exception granted :-)
3131  */
3132 EXPORT_SYMBOL(proc_dointvec);
3133 EXPORT_SYMBOL(proc_dointvec_jiffies);
3134 EXPORT_SYMBOL(proc_dointvec_minmax);
3135 EXPORT_SYMBOL(proc_dointvec_userhz_jiffies);
3136 EXPORT_SYMBOL(proc_dointvec_ms_jiffies);
3137 EXPORT_SYMBOL(proc_dostring);
3138 EXPORT_SYMBOL(proc_doulongvec_minmax);
3139 EXPORT_SYMBOL(proc_doulongvec_ms_jiffies_minmax);
3140 EXPORT_SYMBOL(register_sysctl_table);
3141 EXPORT_SYMBOL(register_sysctl_paths);
3142 EXPORT_SYMBOL(sysctl_intvec);
3143 EXPORT_SYMBOL(sysctl_jiffies);
3144 EXPORT_SYMBOL(sysctl_ms_jiffies);
3145 EXPORT_SYMBOL(sysctl_string);
3146 EXPORT_SYMBOL(sysctl_data);
3147 EXPORT_SYMBOL(unregister_sysctl_table);