sysctl: fix suid_dumpable and lease-break-time sysctls
[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
52 #include <asm/uaccess.h>
53 #include <asm/processor.h>
54
55 #ifdef CONFIG_X86
56 #include <asm/nmi.h>
57 #include <asm/stacktrace.h>
58 #include <asm/io.h>
59 #endif
60
61 static int deprecated_sysctl_warning(struct __sysctl_args *args);
62
63 #if defined(CONFIG_SYSCTL)
64
65 /* External variables not in a header file. */
66 extern int C_A_D;
67 extern int print_fatal_signals;
68 extern int sysctl_overcommit_memory;
69 extern int sysctl_overcommit_ratio;
70 extern int sysctl_panic_on_oom;
71 extern int sysctl_oom_kill_allocating_task;
72 extern int sysctl_oom_dump_tasks;
73 extern int max_threads;
74 extern int core_uses_pid;
75 extern int suid_dumpable;
76 extern char core_pattern[];
77 extern int pid_max;
78 extern int min_free_kbytes;
79 extern int pid_max_min, pid_max_max;
80 extern int sysctl_drop_caches;
81 extern int percpu_pagelist_fraction;
82 extern int compat_log;
83 extern int latencytop_enabled;
84 extern int sysctl_nr_open_min, sysctl_nr_open_max;
85 #ifndef CONFIG_MMU
86 extern int sysctl_nr_trim_pages;
87 #endif
88 #ifdef CONFIG_RCU_TORTURE_TEST
89 extern int rcutorture_runnable;
90 #endif /* #ifdef CONFIG_RCU_TORTURE_TEST */
91
92 /* Constants used for minimum and  maximum */
93 #ifdef CONFIG_DETECT_SOFTLOCKUP
94 static int sixty = 60;
95 static int neg_one = -1;
96 #endif
97
98 static int zero;
99 static int one = 1;
100 static int two = 2;
101 static unsigned long one_ul = 1;
102 static int one_hundred = 100;
103
104 /* this is needed for the proc_dointvec_minmax for [fs_]overflow UID and GID */
105 static int maxolduid = 65535;
106 static int minolduid;
107 static int min_percpu_pagelist_fract = 8;
108
109 static int ngroups_max = NGROUPS_MAX;
110
111 #ifdef CONFIG_MODULES
112 extern char modprobe_path[];
113 #endif
114 #ifdef CONFIG_CHR_DEV_SG
115 extern int sg_big_buff;
116 #endif
117
118 #ifdef CONFIG_SPARC
119 #include <asm/system.h>
120 #endif
121
122 #ifdef CONFIG_SPARC64
123 extern int sysctl_tsb_ratio;
124 #endif
125
126 #ifdef __hppa__
127 extern int pwrsw_enabled;
128 extern int unaligned_enabled;
129 #endif
130
131 #ifdef CONFIG_S390
132 #ifdef CONFIG_MATHEMU
133 extern int sysctl_ieee_emulation_warnings;
134 #endif
135 extern int sysctl_userprocess_debug;
136 extern int spin_retry;
137 #endif
138
139 #ifdef CONFIG_BSD_PROCESS_ACCT
140 extern int acct_parm[];
141 #endif
142
143 #ifdef CONFIG_IA64
144 extern int no_unaligned_warning;
145 extern int unaligned_dump_stack;
146 #endif
147
148 #ifdef CONFIG_RT_MUTEXES
149 extern int max_lock_depth;
150 #endif
151
152 #ifdef CONFIG_PROC_SYSCTL
153 static int proc_do_cad_pid(struct ctl_table *table, int write, struct file *filp,
154                   void __user *buffer, size_t *lenp, loff_t *ppos);
155 static int proc_taint(struct ctl_table *table, int write, struct file *filp,
156                                void __user *buffer, size_t *lenp, loff_t *ppos);
157 #endif
158
159 static struct ctl_table root_table[];
160 static struct ctl_table_root sysctl_table_root;
161 static struct ctl_table_header root_table_header = {
162         .count = 1,
163         .ctl_table = root_table,
164         .ctl_entry = LIST_HEAD_INIT(sysctl_table_root.default_set.list),
165         .root = &sysctl_table_root,
166         .set = &sysctl_table_root.default_set,
167 };
168 static struct ctl_table_root sysctl_table_root = {
169         .root_list = LIST_HEAD_INIT(sysctl_table_root.root_list),
170         .default_set.list = LIST_HEAD_INIT(root_table_header.ctl_entry),
171 };
172
173 static struct ctl_table kern_table[];
174 static struct ctl_table vm_table[];
175 static struct ctl_table fs_table[];
176 static struct ctl_table debug_table[];
177 static struct ctl_table dev_table[];
178 extern struct ctl_table random_table[];
179 #ifdef CONFIG_INOTIFY_USER
180 extern struct ctl_table inotify_table[];
181 #endif
182 #ifdef CONFIG_EPOLL
183 extern struct ctl_table epoll_table[];
184 #endif
185
186 #ifdef HAVE_ARCH_PICK_MMAP_LAYOUT
187 int sysctl_legacy_va_layout;
188 #endif
189
190 extern int prove_locking;
191 extern int lock_stat;
192
193 /* The default sysctl tables: */
194
195 static struct ctl_table root_table[] = {
196         {
197                 .ctl_name       = CTL_KERN,
198                 .procname       = "kernel",
199                 .mode           = 0555,
200                 .child          = kern_table,
201         },
202         {
203                 .ctl_name       = CTL_VM,
204                 .procname       = "vm",
205                 .mode           = 0555,
206                 .child          = vm_table,
207         },
208         {
209                 .ctl_name       = CTL_FS,
210                 .procname       = "fs",
211                 .mode           = 0555,
212                 .child          = fs_table,
213         },
214         {
215                 .ctl_name       = CTL_DEBUG,
216                 .procname       = "debug",
217                 .mode           = 0555,
218                 .child          = debug_table,
219         },
220         {
221                 .ctl_name       = CTL_DEV,
222                 .procname       = "dev",
223                 .mode           = 0555,
224                 .child          = dev_table,
225         },
226 /*
227  * NOTE: do not add new entries to this table unless you have read
228  * Documentation/sysctl/ctl_unnumbered.txt
229  */
230         { .ctl_name = 0 }
231 };
232
233 #ifdef CONFIG_SCHED_DEBUG
234 static int min_sched_granularity_ns = 100000;           /* 100 usecs */
235 static int max_sched_granularity_ns = NSEC_PER_SEC;     /* 1 second */
236 static int min_wakeup_granularity_ns;                   /* 0 usecs */
237 static int max_wakeup_granularity_ns = NSEC_PER_SEC;    /* 1 second */
238 #endif
239
240 static struct ctl_table kern_table[] = {
241 #ifdef CONFIG_SCHED_DEBUG
242         {
243                 .ctl_name       = CTL_UNNUMBERED,
244                 .procname       = "sched_min_granularity_ns",
245                 .data           = &sysctl_sched_min_granularity,
246                 .maxlen         = sizeof(unsigned int),
247                 .mode           = 0644,
248                 .proc_handler   = &sched_nr_latency_handler,
249                 .strategy       = &sysctl_intvec,
250                 .extra1         = &min_sched_granularity_ns,
251                 .extra2         = &max_sched_granularity_ns,
252         },
253         {
254                 .ctl_name       = CTL_UNNUMBERED,
255                 .procname       = "sched_latency_ns",
256                 .data           = &sysctl_sched_latency,
257                 .maxlen         = sizeof(unsigned int),
258                 .mode           = 0644,
259                 .proc_handler   = &sched_nr_latency_handler,
260                 .strategy       = &sysctl_intvec,
261                 .extra1         = &min_sched_granularity_ns,
262                 .extra2         = &max_sched_granularity_ns,
263         },
264         {
265                 .ctl_name       = CTL_UNNUMBERED,
266                 .procname       = "sched_wakeup_granularity_ns",
267                 .data           = &sysctl_sched_wakeup_granularity,
268                 .maxlen         = sizeof(unsigned int),
269                 .mode           = 0644,
270                 .proc_handler   = &proc_dointvec_minmax,
271                 .strategy       = &sysctl_intvec,
272                 .extra1         = &min_wakeup_granularity_ns,
273                 .extra2         = &max_wakeup_granularity_ns,
274         },
275         {
276                 .ctl_name       = CTL_UNNUMBERED,
277                 .procname       = "sched_shares_ratelimit",
278                 .data           = &sysctl_sched_shares_ratelimit,
279                 .maxlen         = sizeof(unsigned int),
280                 .mode           = 0644,
281                 .proc_handler   = &proc_dointvec,
282         },
283         {
284                 .ctl_name       = CTL_UNNUMBERED,
285                 .procname       = "sched_shares_thresh",
286                 .data           = &sysctl_sched_shares_thresh,
287                 .maxlen         = sizeof(unsigned int),
288                 .mode           = 0644,
289                 .proc_handler   = &proc_dointvec_minmax,
290                 .strategy       = &sysctl_intvec,
291                 .extra1         = &zero,
292         },
293         {
294                 .ctl_name       = CTL_UNNUMBERED,
295                 .procname       = "sched_child_runs_first",
296                 .data           = &sysctl_sched_child_runs_first,
297                 .maxlen         = sizeof(unsigned int),
298                 .mode           = 0644,
299                 .proc_handler   = &proc_dointvec,
300         },
301         {
302                 .ctl_name       = CTL_UNNUMBERED,
303                 .procname       = "sched_features",
304                 .data           = &sysctl_sched_features,
305                 .maxlen         = sizeof(unsigned int),
306                 .mode           = 0644,
307                 .proc_handler   = &proc_dointvec,
308         },
309         {
310                 .ctl_name       = CTL_UNNUMBERED,
311                 .procname       = "sched_migration_cost",
312                 .data           = &sysctl_sched_migration_cost,
313                 .maxlen         = sizeof(unsigned int),
314                 .mode           = 0644,
315                 .proc_handler   = &proc_dointvec,
316         },
317         {
318                 .ctl_name       = CTL_UNNUMBERED,
319                 .procname       = "sched_nr_migrate",
320                 .data           = &sysctl_sched_nr_migrate,
321                 .maxlen         = sizeof(unsigned int),
322                 .mode           = 0644,
323                 .proc_handler   = &proc_dointvec,
324         },
325 #endif
326         {
327                 .ctl_name       = CTL_UNNUMBERED,
328                 .procname       = "sched_rt_period_us",
329                 .data           = &sysctl_sched_rt_period,
330                 .maxlen         = sizeof(unsigned int),
331                 .mode           = 0644,
332                 .proc_handler   = &sched_rt_handler,
333         },
334         {
335                 .ctl_name       = CTL_UNNUMBERED,
336                 .procname       = "sched_rt_runtime_us",
337                 .data           = &sysctl_sched_rt_runtime,
338                 .maxlen         = sizeof(int),
339                 .mode           = 0644,
340                 .proc_handler   = &sched_rt_handler,
341         },
342         {
343                 .ctl_name       = CTL_UNNUMBERED,
344                 .procname       = "sched_compat_yield",
345                 .data           = &sysctl_sched_compat_yield,
346                 .maxlen         = sizeof(unsigned int),
347                 .mode           = 0644,
348                 .proc_handler   = &proc_dointvec,
349         },
350 #ifdef CONFIG_PROVE_LOCKING
351         {
352                 .ctl_name       = CTL_UNNUMBERED,
353                 .procname       = "prove_locking",
354                 .data           = &prove_locking,
355                 .maxlen         = sizeof(int),
356                 .mode           = 0644,
357                 .proc_handler   = &proc_dointvec,
358         },
359 #endif
360 #ifdef CONFIG_LOCK_STAT
361         {
362                 .ctl_name       = CTL_UNNUMBERED,
363                 .procname       = "lock_stat",
364                 .data           = &lock_stat,
365                 .maxlen         = sizeof(int),
366                 .mode           = 0644,
367                 .proc_handler   = &proc_dointvec,
368         },
369 #endif
370         {
371                 .ctl_name       = KERN_PANIC,
372                 .procname       = "panic",
373                 .data           = &panic_timeout,
374                 .maxlen         = sizeof(int),
375                 .mode           = 0644,
376                 .proc_handler   = &proc_dointvec,
377         },
378         {
379                 .ctl_name       = KERN_CORE_USES_PID,
380                 .procname       = "core_uses_pid",
381                 .data           = &core_uses_pid,
382                 .maxlen         = sizeof(int),
383                 .mode           = 0644,
384                 .proc_handler   = &proc_dointvec,
385         },
386         {
387                 .ctl_name       = KERN_CORE_PATTERN,
388                 .procname       = "core_pattern",
389                 .data           = core_pattern,
390                 .maxlen         = CORENAME_MAX_SIZE,
391                 .mode           = 0644,
392                 .proc_handler   = &proc_dostring,
393                 .strategy       = &sysctl_string,
394         },
395 #ifdef CONFIG_PROC_SYSCTL
396         {
397                 .procname       = "tainted",
398                 .maxlen         = sizeof(long),
399                 .mode           = 0644,
400                 .proc_handler   = &proc_taint,
401         },
402 #endif
403 #ifdef CONFIG_LATENCYTOP
404         {
405                 .procname       = "latencytop",
406                 .data           = &latencytop_enabled,
407                 .maxlen         = sizeof(int),
408                 .mode           = 0644,
409                 .proc_handler   = &proc_dointvec,
410         },
411 #endif
412 #ifdef CONFIG_BLK_DEV_INITRD
413         {
414                 .ctl_name       = KERN_REALROOTDEV,
415                 .procname       = "real-root-dev",
416                 .data           = &real_root_dev,
417                 .maxlen         = sizeof(int),
418                 .mode           = 0644,
419                 .proc_handler   = &proc_dointvec,
420         },
421 #endif
422         {
423                 .ctl_name       = CTL_UNNUMBERED,
424                 .procname       = "print-fatal-signals",
425                 .data           = &print_fatal_signals,
426                 .maxlen         = sizeof(int),
427                 .mode           = 0644,
428                 .proc_handler   = &proc_dointvec,
429         },
430 #ifdef CONFIG_SPARC
431         {
432                 .ctl_name       = KERN_SPARC_REBOOT,
433                 .procname       = "reboot-cmd",
434                 .data           = reboot_command,
435                 .maxlen         = 256,
436                 .mode           = 0644,
437                 .proc_handler   = &proc_dostring,
438                 .strategy       = &sysctl_string,
439         },
440         {
441                 .ctl_name       = KERN_SPARC_STOP_A,
442                 .procname       = "stop-a",
443                 .data           = &stop_a_enabled,
444                 .maxlen         = sizeof (int),
445                 .mode           = 0644,
446                 .proc_handler   = &proc_dointvec,
447         },
448         {
449                 .ctl_name       = KERN_SPARC_SCONS_PWROFF,
450                 .procname       = "scons-poweroff",
451                 .data           = &scons_pwroff,
452                 .maxlen         = sizeof (int),
453                 .mode           = 0644,
454                 .proc_handler   = &proc_dointvec,
455         },
456 #endif
457 #ifdef CONFIG_SPARC64
458         {
459                 .ctl_name       = CTL_UNNUMBERED,
460                 .procname       = "tsb-ratio",
461                 .data           = &sysctl_tsb_ratio,
462                 .maxlen         = sizeof (int),
463                 .mode           = 0644,
464                 .proc_handler   = &proc_dointvec,
465         },
466 #endif
467 #ifdef __hppa__
468         {
469                 .ctl_name       = KERN_HPPA_PWRSW,
470                 .procname       = "soft-power",
471                 .data           = &pwrsw_enabled,
472                 .maxlen         = sizeof (int),
473                 .mode           = 0644,
474                 .proc_handler   = &proc_dointvec,
475         },
476         {
477                 .ctl_name       = KERN_HPPA_UNALIGNED,
478                 .procname       = "unaligned-trap",
479                 .data           = &unaligned_enabled,
480                 .maxlen         = sizeof (int),
481                 .mode           = 0644,
482                 .proc_handler   = &proc_dointvec,
483         },
484 #endif
485         {
486                 .ctl_name       = KERN_CTLALTDEL,
487                 .procname       = "ctrl-alt-del",
488                 .data           = &C_A_D,
489                 .maxlen         = sizeof(int),
490                 .mode           = 0644,
491                 .proc_handler   = &proc_dointvec,
492         },
493 #ifdef CONFIG_FUNCTION_TRACER
494         {
495                 .ctl_name       = CTL_UNNUMBERED,
496                 .procname       = "ftrace_enabled",
497                 .data           = &ftrace_enabled,
498                 .maxlen         = sizeof(int),
499                 .mode           = 0644,
500                 .proc_handler   = &ftrace_enable_sysctl,
501         },
502 #endif
503 #ifdef CONFIG_STACK_TRACER
504         {
505                 .ctl_name       = CTL_UNNUMBERED,
506                 .procname       = "stack_tracer_enabled",
507                 .data           = &stack_tracer_enabled,
508                 .maxlen         = sizeof(int),
509                 .mode           = 0644,
510                 .proc_handler   = &stack_trace_sysctl,
511         },
512 #endif
513 #ifdef CONFIG_TRACING
514         {
515                 .ctl_name       = CTL_UNNUMBERED,
516                 .procname       = "ftrace_dump_on_oops",
517                 .data           = &ftrace_dump_on_oops,
518                 .maxlen         = sizeof(int),
519                 .mode           = 0644,
520                 .proc_handler   = &proc_dointvec,
521         },
522 #endif
523 #ifdef CONFIG_MODULES
524         {
525                 .ctl_name       = KERN_MODPROBE,
526                 .procname       = "modprobe",
527                 .data           = &modprobe_path,
528                 .maxlen         = KMOD_PATH_LEN,
529                 .mode           = 0644,
530                 .proc_handler   = &proc_dostring,
531                 .strategy       = &sysctl_string,
532         },
533 #endif
534 #if defined(CONFIG_HOTPLUG) && defined(CONFIG_NET)
535         {
536                 .ctl_name       = KERN_HOTPLUG,
537                 .procname       = "hotplug",
538                 .data           = &uevent_helper,
539                 .maxlen         = UEVENT_HELPER_PATH_LEN,
540                 .mode           = 0644,
541                 .proc_handler   = &proc_dostring,
542                 .strategy       = &sysctl_string,
543         },
544 #endif
545 #ifdef CONFIG_CHR_DEV_SG
546         {
547                 .ctl_name       = KERN_SG_BIG_BUFF,
548                 .procname       = "sg-big-buff",
549                 .data           = &sg_big_buff,
550                 .maxlen         = sizeof (int),
551                 .mode           = 0444,
552                 .proc_handler   = &proc_dointvec,
553         },
554 #endif
555 #ifdef CONFIG_BSD_PROCESS_ACCT
556         {
557                 .ctl_name       = KERN_ACCT,
558                 .procname       = "acct",
559                 .data           = &acct_parm,
560                 .maxlen         = 3*sizeof(int),
561                 .mode           = 0644,
562                 .proc_handler   = &proc_dointvec,
563         },
564 #endif
565 #ifdef CONFIG_MAGIC_SYSRQ
566         {
567                 .ctl_name       = KERN_SYSRQ,
568                 .procname       = "sysrq",
569                 .data           = &__sysrq_enabled,
570                 .maxlen         = sizeof (int),
571                 .mode           = 0644,
572                 .proc_handler   = &proc_dointvec,
573         },
574 #endif
575 #ifdef CONFIG_PROC_SYSCTL
576         {
577                 .procname       = "cad_pid",
578                 .data           = NULL,
579                 .maxlen         = sizeof (int),
580                 .mode           = 0600,
581                 .proc_handler   = &proc_do_cad_pid,
582         },
583 #endif
584         {
585                 .ctl_name       = KERN_MAX_THREADS,
586                 .procname       = "threads-max",
587                 .data           = &max_threads,
588                 .maxlen         = sizeof(int),
589                 .mode           = 0644,
590                 .proc_handler   = &proc_dointvec,
591         },
592         {
593                 .ctl_name       = KERN_RANDOM,
594                 .procname       = "random",
595                 .mode           = 0555,
596                 .child          = random_table,
597         },
598         {
599                 .ctl_name       = KERN_OVERFLOWUID,
600                 .procname       = "overflowuid",
601                 .data           = &overflowuid,
602                 .maxlen         = sizeof(int),
603                 .mode           = 0644,
604                 .proc_handler   = &proc_dointvec_minmax,
605                 .strategy       = &sysctl_intvec,
606                 .extra1         = &minolduid,
607                 .extra2         = &maxolduid,
608         },
609         {
610                 .ctl_name       = KERN_OVERFLOWGID,
611                 .procname       = "overflowgid",
612                 .data           = &overflowgid,
613                 .maxlen         = sizeof(int),
614                 .mode           = 0644,
615                 .proc_handler   = &proc_dointvec_minmax,
616                 .strategy       = &sysctl_intvec,
617                 .extra1         = &minolduid,
618                 .extra2         = &maxolduid,
619         },
620 #ifdef CONFIG_S390
621 #ifdef CONFIG_MATHEMU
622         {
623                 .ctl_name       = KERN_IEEE_EMULATION_WARNINGS,
624                 .procname       = "ieee_emulation_warnings",
625                 .data           = &sysctl_ieee_emulation_warnings,
626                 .maxlen         = sizeof(int),
627                 .mode           = 0644,
628                 .proc_handler   = &proc_dointvec,
629         },
630 #endif
631         {
632                 .ctl_name       = KERN_S390_USER_DEBUG_LOGGING,
633                 .procname       = "userprocess_debug",
634                 .data           = &sysctl_userprocess_debug,
635                 .maxlen         = sizeof(int),
636                 .mode           = 0644,
637                 .proc_handler   = &proc_dointvec,
638         },
639 #endif
640         {
641                 .ctl_name       = KERN_PIDMAX,
642                 .procname       = "pid_max",
643                 .data           = &pid_max,
644                 .maxlen         = sizeof (int),
645                 .mode           = 0644,
646                 .proc_handler   = &proc_dointvec_minmax,
647                 .strategy       = sysctl_intvec,
648                 .extra1         = &pid_max_min,
649                 .extra2         = &pid_max_max,
650         },
651         {
652                 .ctl_name       = KERN_PANIC_ON_OOPS,
653                 .procname       = "panic_on_oops",
654                 .data           = &panic_on_oops,
655                 .maxlen         = sizeof(int),
656                 .mode           = 0644,
657                 .proc_handler   = &proc_dointvec,
658         },
659 #if defined CONFIG_PRINTK
660         {
661                 .ctl_name       = KERN_PRINTK,
662                 .procname       = "printk",
663                 .data           = &console_loglevel,
664                 .maxlen         = 4*sizeof(int),
665                 .mode           = 0644,
666                 .proc_handler   = &proc_dointvec,
667         },
668         {
669                 .ctl_name       = KERN_PRINTK_RATELIMIT,
670                 .procname       = "printk_ratelimit",
671                 .data           = &printk_ratelimit_state.interval,
672                 .maxlen         = sizeof(int),
673                 .mode           = 0644,
674                 .proc_handler   = &proc_dointvec_jiffies,
675                 .strategy       = &sysctl_jiffies,
676         },
677         {
678                 .ctl_name       = KERN_PRINTK_RATELIMIT_BURST,
679                 .procname       = "printk_ratelimit_burst",
680                 .data           = &printk_ratelimit_state.burst,
681                 .maxlen         = sizeof(int),
682                 .mode           = 0644,
683                 .proc_handler   = &proc_dointvec,
684         },
685 #endif
686         {
687                 .ctl_name       = KERN_NGROUPS_MAX,
688                 .procname       = "ngroups_max",
689                 .data           = &ngroups_max,
690                 .maxlen         = sizeof (int),
691                 .mode           = 0444,
692                 .proc_handler   = &proc_dointvec,
693         },
694 #if defined(CONFIG_X86_LOCAL_APIC) && defined(CONFIG_X86)
695         {
696                 .ctl_name       = KERN_UNKNOWN_NMI_PANIC,
697                 .procname       = "unknown_nmi_panic",
698                 .data           = &unknown_nmi_panic,
699                 .maxlen         = sizeof (int),
700                 .mode           = 0644,
701                 .proc_handler   = &proc_dointvec,
702         },
703         {
704                 .procname       = "nmi_watchdog",
705                 .data           = &nmi_watchdog_enabled,
706                 .maxlen         = sizeof (int),
707                 .mode           = 0644,
708                 .proc_handler   = &proc_nmi_enabled,
709         },
710 #endif
711 #if defined(CONFIG_X86)
712         {
713                 .ctl_name       = KERN_PANIC_ON_NMI,
714                 .procname       = "panic_on_unrecovered_nmi",
715                 .data           = &panic_on_unrecovered_nmi,
716                 .maxlen         = sizeof(int),
717                 .mode           = 0644,
718                 .proc_handler   = &proc_dointvec,
719         },
720         {
721                 .ctl_name       = KERN_BOOTLOADER_TYPE,
722                 .procname       = "bootloader_type",
723                 .data           = &bootloader_type,
724                 .maxlen         = sizeof (int),
725                 .mode           = 0444,
726                 .proc_handler   = &proc_dointvec,
727         },
728         {
729                 .ctl_name       = CTL_UNNUMBERED,
730                 .procname       = "kstack_depth_to_print",
731                 .data           = &kstack_depth_to_print,
732                 .maxlen         = sizeof(int),
733                 .mode           = 0644,
734                 .proc_handler   = &proc_dointvec,
735         },
736         {
737                 .ctl_name       = CTL_UNNUMBERED,
738                 .procname       = "io_delay_type",
739                 .data           = &io_delay_type,
740                 .maxlen         = sizeof(int),
741                 .mode           = 0644,
742                 .proc_handler   = &proc_dointvec,
743         },
744 #endif
745 #if defined(CONFIG_MMU)
746         {
747                 .ctl_name       = KERN_RANDOMIZE,
748                 .procname       = "randomize_va_space",
749                 .data           = &randomize_va_space,
750                 .maxlen         = sizeof(int),
751                 .mode           = 0644,
752                 .proc_handler   = &proc_dointvec,
753         },
754 #endif
755 #if defined(CONFIG_S390) && defined(CONFIG_SMP)
756         {
757                 .ctl_name       = KERN_SPIN_RETRY,
758                 .procname       = "spin_retry",
759                 .data           = &spin_retry,
760                 .maxlen         = sizeof (int),
761                 .mode           = 0644,
762                 .proc_handler   = &proc_dointvec,
763         },
764 #endif
765 #if     defined(CONFIG_ACPI_SLEEP) && defined(CONFIG_X86)
766         {
767                 .procname       = "acpi_video_flags",
768                 .data           = &acpi_realmode_flags,
769                 .maxlen         = sizeof (unsigned long),
770                 .mode           = 0644,
771                 .proc_handler   = &proc_doulongvec_minmax,
772         },
773 #endif
774 #ifdef CONFIG_IA64
775         {
776                 .ctl_name       = KERN_IA64_UNALIGNED,
777                 .procname       = "ignore-unaligned-usertrap",
778                 .data           = &no_unaligned_warning,
779                 .maxlen         = sizeof (int),
780                 .mode           = 0644,
781                 .proc_handler   = &proc_dointvec,
782         },
783         {
784                 .ctl_name       = CTL_UNNUMBERED,
785                 .procname       = "unaligned-dump-stack",
786                 .data           = &unaligned_dump_stack,
787                 .maxlen         = sizeof (int),
788                 .mode           = 0644,
789                 .proc_handler   = &proc_dointvec,
790         },
791 #endif
792 #ifdef CONFIG_DETECT_SOFTLOCKUP
793         {
794                 .ctl_name       = CTL_UNNUMBERED,
795                 .procname       = "softlockup_panic",
796                 .data           = &softlockup_panic,
797                 .maxlen         = sizeof(int),
798                 .mode           = 0644,
799                 .proc_handler   = &proc_dointvec_minmax,
800                 .strategy       = &sysctl_intvec,
801                 .extra1         = &zero,
802                 .extra2         = &one,
803         },
804         {
805                 .ctl_name       = CTL_UNNUMBERED,
806                 .procname       = "softlockup_thresh",
807                 .data           = &softlockup_thresh,
808                 .maxlen         = sizeof(int),
809                 .mode           = 0644,
810                 .proc_handler   = &proc_dosoftlockup_thresh,
811                 .strategy       = &sysctl_intvec,
812                 .extra1         = &neg_one,
813                 .extra2         = &sixty,
814         },
815         {
816                 .ctl_name       = CTL_UNNUMBERED,
817                 .procname       = "hung_task_check_count",
818                 .data           = &sysctl_hung_task_check_count,
819                 .maxlen         = sizeof(unsigned long),
820                 .mode           = 0644,
821                 .proc_handler   = &proc_doulongvec_minmax,
822                 .strategy       = &sysctl_intvec,
823         },
824         {
825                 .ctl_name       = CTL_UNNUMBERED,
826                 .procname       = "hung_task_timeout_secs",
827                 .data           = &sysctl_hung_task_timeout_secs,
828                 .maxlen         = sizeof(unsigned long),
829                 .mode           = 0644,
830                 .proc_handler   = &proc_doulongvec_minmax,
831                 .strategy       = &sysctl_intvec,
832         },
833         {
834                 .ctl_name       = CTL_UNNUMBERED,
835                 .procname       = "hung_task_warnings",
836                 .data           = &sysctl_hung_task_warnings,
837                 .maxlen         = sizeof(unsigned long),
838                 .mode           = 0644,
839                 .proc_handler   = &proc_doulongvec_minmax,
840                 .strategy       = &sysctl_intvec,
841         },
842 #endif
843 #ifdef CONFIG_COMPAT
844         {
845                 .ctl_name       = KERN_COMPAT_LOG,
846                 .procname       = "compat-log",
847                 .data           = &compat_log,
848                 .maxlen         = sizeof (int),
849                 .mode           = 0644,
850                 .proc_handler   = &proc_dointvec,
851         },
852 #endif
853 #ifdef CONFIG_RT_MUTEXES
854         {
855                 .ctl_name       = KERN_MAX_LOCK_DEPTH,
856                 .procname       = "max_lock_depth",
857                 .data           = &max_lock_depth,
858                 .maxlen         = sizeof(int),
859                 .mode           = 0644,
860                 .proc_handler   = &proc_dointvec,
861         },
862 #endif
863         {
864                 .ctl_name       = CTL_UNNUMBERED,
865                 .procname       = "poweroff_cmd",
866                 .data           = &poweroff_cmd,
867                 .maxlen         = POWEROFF_CMD_PATH_LEN,
868                 .mode           = 0644,
869                 .proc_handler   = &proc_dostring,
870                 .strategy       = &sysctl_string,
871         },
872 #ifdef CONFIG_KEYS
873         {
874                 .ctl_name       = CTL_UNNUMBERED,
875                 .procname       = "keys",
876                 .mode           = 0555,
877                 .child          = key_sysctls,
878         },
879 #endif
880 #ifdef CONFIG_RCU_TORTURE_TEST
881         {
882                 .ctl_name       = CTL_UNNUMBERED,
883                 .procname       = "rcutorture_runnable",
884                 .data           = &rcutorture_runnable,
885                 .maxlen         = sizeof(int),
886                 .mode           = 0644,
887                 .proc_handler   = &proc_dointvec,
888         },
889 #endif
890 #ifdef CONFIG_UNEVICTABLE_LRU
891         {
892                 .ctl_name       = CTL_UNNUMBERED,
893                 .procname       = "scan_unevictable_pages",
894                 .data           = &scan_unevictable_pages,
895                 .maxlen         = sizeof(scan_unevictable_pages),
896                 .mode           = 0644,
897                 .proc_handler   = &scan_unevictable_handler,
898         },
899 #endif
900 /*
901  * NOTE: do not add new entries to this table unless you have read
902  * Documentation/sysctl/ctl_unnumbered.txt
903  */
904         { .ctl_name = 0 }
905 };
906
907 static struct ctl_table vm_table[] = {
908         {
909                 .ctl_name       = VM_OVERCOMMIT_MEMORY,
910                 .procname       = "overcommit_memory",
911                 .data           = &sysctl_overcommit_memory,
912                 .maxlen         = sizeof(sysctl_overcommit_memory),
913                 .mode           = 0644,
914                 .proc_handler   = &proc_dointvec,
915         },
916         {
917                 .ctl_name       = VM_PANIC_ON_OOM,
918                 .procname       = "panic_on_oom",
919                 .data           = &sysctl_panic_on_oom,
920                 .maxlen         = sizeof(sysctl_panic_on_oom),
921                 .mode           = 0644,
922                 .proc_handler   = &proc_dointvec,
923         },
924         {
925                 .ctl_name       = CTL_UNNUMBERED,
926                 .procname       = "oom_kill_allocating_task",
927                 .data           = &sysctl_oom_kill_allocating_task,
928                 .maxlen         = sizeof(sysctl_oom_kill_allocating_task),
929                 .mode           = 0644,
930                 .proc_handler   = &proc_dointvec,
931         },
932         {
933                 .ctl_name       = CTL_UNNUMBERED,
934                 .procname       = "oom_dump_tasks",
935                 .data           = &sysctl_oom_dump_tasks,
936                 .maxlen         = sizeof(sysctl_oom_dump_tasks),
937                 .mode           = 0644,
938                 .proc_handler   = &proc_dointvec,
939         },
940         {
941                 .ctl_name       = VM_OVERCOMMIT_RATIO,
942                 .procname       = "overcommit_ratio",
943                 .data           = &sysctl_overcommit_ratio,
944                 .maxlen         = sizeof(sysctl_overcommit_ratio),
945                 .mode           = 0644,
946                 .proc_handler   = &proc_dointvec,
947         },
948         {
949                 .ctl_name       = VM_PAGE_CLUSTER,
950                 .procname       = "page-cluster", 
951                 .data           = &page_cluster,
952                 .maxlen         = sizeof(int),
953                 .mode           = 0644,
954                 .proc_handler   = &proc_dointvec,
955         },
956         {
957                 .ctl_name       = VM_DIRTY_BACKGROUND,
958                 .procname       = "dirty_background_ratio",
959                 .data           = &dirty_background_ratio,
960                 .maxlen         = sizeof(dirty_background_ratio),
961                 .mode           = 0644,
962                 .proc_handler   = &dirty_background_ratio_handler,
963                 .strategy       = &sysctl_intvec,
964                 .extra1         = &zero,
965                 .extra2         = &one_hundred,
966         },
967         {
968                 .ctl_name       = CTL_UNNUMBERED,
969                 .procname       = "dirty_background_bytes",
970                 .data           = &dirty_background_bytes,
971                 .maxlen         = sizeof(dirty_background_bytes),
972                 .mode           = 0644,
973                 .proc_handler   = &dirty_background_bytes_handler,
974                 .strategy       = &sysctl_intvec,
975                 .extra1         = &one_ul,
976         },
977         {
978                 .ctl_name       = VM_DIRTY_RATIO,
979                 .procname       = "dirty_ratio",
980                 .data           = &vm_dirty_ratio,
981                 .maxlen         = sizeof(vm_dirty_ratio),
982                 .mode           = 0644,
983                 .proc_handler   = &dirty_ratio_handler,
984                 .strategy       = &sysctl_intvec,
985                 .extra1         = &zero,
986                 .extra2         = &one_hundred,
987         },
988         {
989                 .ctl_name       = CTL_UNNUMBERED,
990                 .procname       = "dirty_bytes",
991                 .data           = &vm_dirty_bytes,
992                 .maxlen         = sizeof(vm_dirty_bytes),
993                 .mode           = 0644,
994                 .proc_handler   = &dirty_bytes_handler,
995                 .strategy       = &sysctl_intvec,
996                 .extra1         = &one_ul,
997         },
998         {
999                 .procname       = "dirty_writeback_centisecs",
1000                 .data           = &dirty_writeback_interval,
1001                 .maxlen         = sizeof(dirty_writeback_interval),
1002                 .mode           = 0644,
1003                 .proc_handler   = &dirty_writeback_centisecs_handler,
1004         },
1005         {
1006                 .procname       = "dirty_expire_centisecs",
1007                 .data           = &dirty_expire_interval,
1008                 .maxlen         = sizeof(dirty_expire_interval),
1009                 .mode           = 0644,
1010                 .proc_handler   = &proc_dointvec,
1011         },
1012         {
1013                 .ctl_name       = VM_NR_PDFLUSH_THREADS,
1014                 .procname       = "nr_pdflush_threads",
1015                 .data           = &nr_pdflush_threads,
1016                 .maxlen         = sizeof nr_pdflush_threads,
1017                 .mode           = 0444 /* read-only*/,
1018                 .proc_handler   = &proc_dointvec,
1019         },
1020         {
1021                 .ctl_name       = VM_SWAPPINESS,
1022                 .procname       = "swappiness",
1023                 .data           = &vm_swappiness,
1024                 .maxlen         = sizeof(vm_swappiness),
1025                 .mode           = 0644,
1026                 .proc_handler   = &proc_dointvec_minmax,
1027                 .strategy       = &sysctl_intvec,
1028                 .extra1         = &zero,
1029                 .extra2         = &one_hundred,
1030         },
1031 #ifdef CONFIG_HUGETLB_PAGE
1032          {
1033                 .procname       = "nr_hugepages",
1034                 .data           = NULL,
1035                 .maxlen         = sizeof(unsigned long),
1036                 .mode           = 0644,
1037                 .proc_handler   = &hugetlb_sysctl_handler,
1038                 .extra1         = (void *)&hugetlb_zero,
1039                 .extra2         = (void *)&hugetlb_infinity,
1040          },
1041          {
1042                 .ctl_name       = VM_HUGETLB_GROUP,
1043                 .procname       = "hugetlb_shm_group",
1044                 .data           = &sysctl_hugetlb_shm_group,
1045                 .maxlen         = sizeof(gid_t),
1046                 .mode           = 0644,
1047                 .proc_handler   = &proc_dointvec,
1048          },
1049          {
1050                 .ctl_name       = CTL_UNNUMBERED,
1051                 .procname       = "hugepages_treat_as_movable",
1052                 .data           = &hugepages_treat_as_movable,
1053                 .maxlen         = sizeof(int),
1054                 .mode           = 0644,
1055                 .proc_handler   = &hugetlb_treat_movable_handler,
1056         },
1057         {
1058                 .ctl_name       = CTL_UNNUMBERED,
1059                 .procname       = "nr_overcommit_hugepages",
1060                 .data           = NULL,
1061                 .maxlen         = sizeof(unsigned long),
1062                 .mode           = 0644,
1063                 .proc_handler   = &hugetlb_overcommit_handler,
1064                 .extra1         = (void *)&hugetlb_zero,
1065                 .extra2         = (void *)&hugetlb_infinity,
1066         },
1067 #endif
1068         {
1069                 .ctl_name       = VM_LOWMEM_RESERVE_RATIO,
1070                 .procname       = "lowmem_reserve_ratio",
1071                 .data           = &sysctl_lowmem_reserve_ratio,
1072                 .maxlen         = sizeof(sysctl_lowmem_reserve_ratio),
1073                 .mode           = 0644,
1074                 .proc_handler   = &lowmem_reserve_ratio_sysctl_handler,
1075                 .strategy       = &sysctl_intvec,
1076         },
1077         {
1078                 .ctl_name       = VM_DROP_PAGECACHE,
1079                 .procname       = "drop_caches",
1080                 .data           = &sysctl_drop_caches,
1081                 .maxlen         = sizeof(int),
1082                 .mode           = 0644,
1083                 .proc_handler   = drop_caches_sysctl_handler,
1084                 .strategy       = &sysctl_intvec,
1085         },
1086         {
1087                 .ctl_name       = VM_MIN_FREE_KBYTES,
1088                 .procname       = "min_free_kbytes",
1089                 .data           = &min_free_kbytes,
1090                 .maxlen         = sizeof(min_free_kbytes),
1091                 .mode           = 0644,
1092                 .proc_handler   = &min_free_kbytes_sysctl_handler,
1093                 .strategy       = &sysctl_intvec,
1094                 .extra1         = &zero,
1095         },
1096         {
1097                 .ctl_name       = VM_PERCPU_PAGELIST_FRACTION,
1098                 .procname       = "percpu_pagelist_fraction",
1099                 .data           = &percpu_pagelist_fraction,
1100                 .maxlen         = sizeof(percpu_pagelist_fraction),
1101                 .mode           = 0644,
1102                 .proc_handler   = &percpu_pagelist_fraction_sysctl_handler,
1103                 .strategy       = &sysctl_intvec,
1104                 .extra1         = &min_percpu_pagelist_fract,
1105         },
1106 #ifdef CONFIG_MMU
1107         {
1108                 .ctl_name       = VM_MAX_MAP_COUNT,
1109                 .procname       = "max_map_count",
1110                 .data           = &sysctl_max_map_count,
1111                 .maxlen         = sizeof(sysctl_max_map_count),
1112                 .mode           = 0644,
1113                 .proc_handler   = &proc_dointvec
1114         },
1115 #else
1116         {
1117                 .ctl_name       = CTL_UNNUMBERED,
1118                 .procname       = "nr_trim_pages",
1119                 .data           = &sysctl_nr_trim_pages,
1120                 .maxlen         = sizeof(sysctl_nr_trim_pages),
1121                 .mode           = 0644,
1122                 .proc_handler   = &proc_dointvec_minmax,
1123                 .strategy       = &sysctl_intvec,
1124                 .extra1         = &zero,
1125         },
1126 #endif
1127         {
1128                 .ctl_name       = VM_LAPTOP_MODE,
1129                 .procname       = "laptop_mode",
1130                 .data           = &laptop_mode,
1131                 .maxlen         = sizeof(laptop_mode),
1132                 .mode           = 0644,
1133                 .proc_handler   = &proc_dointvec_jiffies,
1134                 .strategy       = &sysctl_jiffies,
1135         },
1136         {
1137                 .ctl_name       = VM_BLOCK_DUMP,
1138                 .procname       = "block_dump",
1139                 .data           = &block_dump,
1140                 .maxlen         = sizeof(block_dump),
1141                 .mode           = 0644,
1142                 .proc_handler   = &proc_dointvec,
1143                 .strategy       = &sysctl_intvec,
1144                 .extra1         = &zero,
1145         },
1146         {
1147                 .ctl_name       = VM_VFS_CACHE_PRESSURE,
1148                 .procname       = "vfs_cache_pressure",
1149                 .data           = &sysctl_vfs_cache_pressure,
1150                 .maxlen         = sizeof(sysctl_vfs_cache_pressure),
1151                 .mode           = 0644,
1152                 .proc_handler   = &proc_dointvec,
1153                 .strategy       = &sysctl_intvec,
1154                 .extra1         = &zero,
1155         },
1156 #ifdef HAVE_ARCH_PICK_MMAP_LAYOUT
1157         {
1158                 .ctl_name       = VM_LEGACY_VA_LAYOUT,
1159                 .procname       = "legacy_va_layout",
1160                 .data           = &sysctl_legacy_va_layout,
1161                 .maxlen         = sizeof(sysctl_legacy_va_layout),
1162                 .mode           = 0644,
1163                 .proc_handler   = &proc_dointvec,
1164                 .strategy       = &sysctl_intvec,
1165                 .extra1         = &zero,
1166         },
1167 #endif
1168 #ifdef CONFIG_NUMA
1169         {
1170                 .ctl_name       = VM_ZONE_RECLAIM_MODE,
1171                 .procname       = "zone_reclaim_mode",
1172                 .data           = &zone_reclaim_mode,
1173                 .maxlen         = sizeof(zone_reclaim_mode),
1174                 .mode           = 0644,
1175                 .proc_handler   = &proc_dointvec,
1176                 .strategy       = &sysctl_intvec,
1177                 .extra1         = &zero,
1178         },
1179         {
1180                 .ctl_name       = VM_MIN_UNMAPPED,
1181                 .procname       = "min_unmapped_ratio",
1182                 .data           = &sysctl_min_unmapped_ratio,
1183                 .maxlen         = sizeof(sysctl_min_unmapped_ratio),
1184                 .mode           = 0644,
1185                 .proc_handler   = &sysctl_min_unmapped_ratio_sysctl_handler,
1186                 .strategy       = &sysctl_intvec,
1187                 .extra1         = &zero,
1188                 .extra2         = &one_hundred,
1189         },
1190         {
1191                 .ctl_name       = VM_MIN_SLAB,
1192                 .procname       = "min_slab_ratio",
1193                 .data           = &sysctl_min_slab_ratio,
1194                 .maxlen         = sizeof(sysctl_min_slab_ratio),
1195                 .mode           = 0644,
1196                 .proc_handler   = &sysctl_min_slab_ratio_sysctl_handler,
1197                 .strategy       = &sysctl_intvec,
1198                 .extra1         = &zero,
1199                 .extra2         = &one_hundred,
1200         },
1201 #endif
1202 #ifdef CONFIG_SMP
1203         {
1204                 .ctl_name       = CTL_UNNUMBERED,
1205                 .procname       = "stat_interval",
1206                 .data           = &sysctl_stat_interval,
1207                 .maxlen         = sizeof(sysctl_stat_interval),
1208                 .mode           = 0644,
1209                 .proc_handler   = &proc_dointvec_jiffies,
1210                 .strategy       = &sysctl_jiffies,
1211         },
1212 #endif
1213 #ifdef CONFIG_SECURITY
1214         {
1215                 .ctl_name       = CTL_UNNUMBERED,
1216                 .procname       = "mmap_min_addr",
1217                 .data           = &mmap_min_addr,
1218                 .maxlen         = sizeof(unsigned long),
1219                 .mode           = 0644,
1220                 .proc_handler   = &proc_doulongvec_minmax,
1221         },
1222 #endif
1223 #ifdef CONFIG_NUMA
1224         {
1225                 .ctl_name       = CTL_UNNUMBERED,
1226                 .procname       = "numa_zonelist_order",
1227                 .data           = &numa_zonelist_order,
1228                 .maxlen         = NUMA_ZONELIST_ORDER_LEN,
1229                 .mode           = 0644,
1230                 .proc_handler   = &numa_zonelist_order_handler,
1231                 .strategy       = &sysctl_string,
1232         },
1233 #endif
1234 #if (defined(CONFIG_X86_32) && !defined(CONFIG_UML))|| \
1235    (defined(CONFIG_SUPERH) && defined(CONFIG_VSYSCALL))
1236         {
1237                 .ctl_name       = VM_VDSO_ENABLED,
1238                 .procname       = "vdso_enabled",
1239                 .data           = &vdso_enabled,
1240                 .maxlen         = sizeof(vdso_enabled),
1241                 .mode           = 0644,
1242                 .proc_handler   = &proc_dointvec,
1243                 .strategy       = &sysctl_intvec,
1244                 .extra1         = &zero,
1245         },
1246 #endif
1247 #ifdef CONFIG_HIGHMEM
1248         {
1249                 .ctl_name       = CTL_UNNUMBERED,
1250                 .procname       = "highmem_is_dirtyable",
1251                 .data           = &vm_highmem_is_dirtyable,
1252                 .maxlen         = sizeof(vm_highmem_is_dirtyable),
1253                 .mode           = 0644,
1254                 .proc_handler   = &proc_dointvec_minmax,
1255                 .strategy       = &sysctl_intvec,
1256                 .extra1         = &zero,
1257                 .extra2         = &one,
1258         },
1259 #endif
1260 /*
1261  * NOTE: do not add new entries to this table unless you have read
1262  * Documentation/sysctl/ctl_unnumbered.txt
1263  */
1264         { .ctl_name = 0 }
1265 };
1266
1267 #if defined(CONFIG_BINFMT_MISC) || defined(CONFIG_BINFMT_MISC_MODULE)
1268 static struct ctl_table binfmt_misc_table[] = {
1269         { .ctl_name = 0 }
1270 };
1271 #endif
1272
1273 static struct ctl_table fs_table[] = {
1274         {
1275                 .ctl_name       = FS_NRINODE,
1276                 .procname       = "inode-nr",
1277                 .data           = &inodes_stat,
1278                 .maxlen         = 2*sizeof(int),
1279                 .mode           = 0444,
1280                 .proc_handler   = &proc_dointvec,
1281         },
1282         {
1283                 .ctl_name       = FS_STATINODE,
1284                 .procname       = "inode-state",
1285                 .data           = &inodes_stat,
1286                 .maxlen         = 7*sizeof(int),
1287                 .mode           = 0444,
1288                 .proc_handler   = &proc_dointvec,
1289         },
1290         {
1291                 .procname       = "file-nr",
1292                 .data           = &files_stat,
1293                 .maxlen         = 3*sizeof(int),
1294                 .mode           = 0444,
1295                 .proc_handler   = &proc_nr_files,
1296         },
1297         {
1298                 .ctl_name       = FS_MAXFILE,
1299                 .procname       = "file-max",
1300                 .data           = &files_stat.max_files,
1301                 .maxlen         = sizeof(int),
1302                 .mode           = 0644,
1303                 .proc_handler   = &proc_dointvec,
1304         },
1305         {
1306                 .ctl_name       = CTL_UNNUMBERED,
1307                 .procname       = "nr_open",
1308                 .data           = &sysctl_nr_open,
1309                 .maxlen         = sizeof(int),
1310                 .mode           = 0644,
1311                 .proc_handler   = &proc_dointvec_minmax,
1312                 .extra1         = &sysctl_nr_open_min,
1313                 .extra2         = &sysctl_nr_open_max,
1314         },
1315         {
1316                 .ctl_name       = FS_DENTRY,
1317                 .procname       = "dentry-state",
1318                 .data           = &dentry_stat,
1319                 .maxlen         = 6*sizeof(int),
1320                 .mode           = 0444,
1321                 .proc_handler   = &proc_dointvec,
1322         },
1323         {
1324                 .ctl_name       = FS_OVERFLOWUID,
1325                 .procname       = "overflowuid",
1326                 .data           = &fs_overflowuid,
1327                 .maxlen         = sizeof(int),
1328                 .mode           = 0644,
1329                 .proc_handler   = &proc_dointvec_minmax,
1330                 .strategy       = &sysctl_intvec,
1331                 .extra1         = &minolduid,
1332                 .extra2         = &maxolduid,
1333         },
1334         {
1335                 .ctl_name       = FS_OVERFLOWGID,
1336                 .procname       = "overflowgid",
1337                 .data           = &fs_overflowgid,
1338                 .maxlen         = sizeof(int),
1339                 .mode           = 0644,
1340                 .proc_handler   = &proc_dointvec_minmax,
1341                 .strategy       = &sysctl_intvec,
1342                 .extra1         = &minolduid,
1343                 .extra2         = &maxolduid,
1344         },
1345 #ifdef CONFIG_FILE_LOCKING
1346         {
1347                 .ctl_name       = FS_LEASES,
1348                 .procname       = "leases-enable",
1349                 .data           = &leases_enable,
1350                 .maxlen         = sizeof(int),
1351                 .mode           = 0644,
1352                 .proc_handler   = &proc_dointvec,
1353         },
1354 #endif
1355 #ifdef CONFIG_DNOTIFY
1356         {
1357                 .ctl_name       = FS_DIR_NOTIFY,
1358                 .procname       = "dir-notify-enable",
1359                 .data           = &dir_notify_enable,
1360                 .maxlen         = sizeof(int),
1361                 .mode           = 0644,
1362                 .proc_handler   = &proc_dointvec,
1363         },
1364 #endif
1365 #ifdef CONFIG_MMU
1366 #ifdef CONFIG_FILE_LOCKING
1367         {
1368                 .ctl_name       = FS_LEASE_TIME,
1369                 .procname       = "lease-break-time",
1370                 .data           = &lease_break_time,
1371                 .maxlen         = sizeof(int),
1372                 .mode           = 0644,
1373                 .proc_handler   = &proc_dointvec,
1374         },
1375 #endif
1376 #ifdef CONFIG_AIO
1377         {
1378                 .procname       = "aio-nr",
1379                 .data           = &aio_nr,
1380                 .maxlen         = sizeof(aio_nr),
1381                 .mode           = 0444,
1382                 .proc_handler   = &proc_doulongvec_minmax,
1383         },
1384         {
1385                 .procname       = "aio-max-nr",
1386                 .data           = &aio_max_nr,
1387                 .maxlen         = sizeof(aio_max_nr),
1388                 .mode           = 0644,
1389                 .proc_handler   = &proc_doulongvec_minmax,
1390         },
1391 #endif /* CONFIG_AIO */
1392 #ifdef CONFIG_INOTIFY_USER
1393         {
1394                 .ctl_name       = FS_INOTIFY,
1395                 .procname       = "inotify",
1396                 .mode           = 0555,
1397                 .child          = inotify_table,
1398         },
1399 #endif  
1400 #ifdef CONFIG_EPOLL
1401         {
1402                 .procname       = "epoll",
1403                 .mode           = 0555,
1404                 .child          = epoll_table,
1405         },
1406 #endif
1407 #endif
1408         {
1409                 .ctl_name       = KERN_SETUID_DUMPABLE,
1410                 .procname       = "suid_dumpable",
1411                 .data           = &suid_dumpable,
1412                 .maxlen         = sizeof(int),
1413                 .mode           = 0644,
1414                 .proc_handler   = &proc_dointvec_minmax,
1415                 .strategy       = &sysctl_intvec,
1416                 .extra1         = &zero,
1417                 .extra2         = &two,
1418         },
1419 #if defined(CONFIG_BINFMT_MISC) || defined(CONFIG_BINFMT_MISC_MODULE)
1420         {
1421                 .ctl_name       = CTL_UNNUMBERED,
1422                 .procname       = "binfmt_misc",
1423                 .mode           = 0555,
1424                 .child          = binfmt_misc_table,
1425         },
1426 #endif
1427 /*
1428  * NOTE: do not add new entries to this table unless you have read
1429  * Documentation/sysctl/ctl_unnumbered.txt
1430  */
1431         { .ctl_name = 0 }
1432 };
1433
1434 static struct ctl_table debug_table[] = {
1435 #if defined(CONFIG_X86) || defined(CONFIG_PPC)
1436         {
1437                 .ctl_name       = CTL_UNNUMBERED,
1438                 .procname       = "exception-trace",
1439                 .data           = &show_unhandled_signals,
1440                 .maxlen         = sizeof(int),
1441                 .mode           = 0644,
1442                 .proc_handler   = proc_dointvec
1443         },
1444 #endif
1445         { .ctl_name = 0 }
1446 };
1447
1448 static struct ctl_table dev_table[] = {
1449         { .ctl_name = 0 }
1450 };
1451
1452 static DEFINE_SPINLOCK(sysctl_lock);
1453
1454 /* called under sysctl_lock */
1455 static int use_table(struct ctl_table_header *p)
1456 {
1457         if (unlikely(p->unregistering))
1458                 return 0;
1459         p->used++;
1460         return 1;
1461 }
1462
1463 /* called under sysctl_lock */
1464 static void unuse_table(struct ctl_table_header *p)
1465 {
1466         if (!--p->used)
1467                 if (unlikely(p->unregistering))
1468                         complete(p->unregistering);
1469 }
1470
1471 /* called under sysctl_lock, will reacquire if has to wait */
1472 static void start_unregistering(struct ctl_table_header *p)
1473 {
1474         /*
1475          * if p->used is 0, nobody will ever touch that entry again;
1476          * we'll eliminate all paths to it before dropping sysctl_lock
1477          */
1478         if (unlikely(p->used)) {
1479                 struct completion wait;
1480                 init_completion(&wait);
1481                 p->unregistering = &wait;
1482                 spin_unlock(&sysctl_lock);
1483                 wait_for_completion(&wait);
1484                 spin_lock(&sysctl_lock);
1485         } else {
1486                 /* anything non-NULL; we'll never dereference it */
1487                 p->unregistering = ERR_PTR(-EINVAL);
1488         }
1489         /*
1490          * do not remove from the list until nobody holds it; walking the
1491          * list in do_sysctl() relies on that.
1492          */
1493         list_del_init(&p->ctl_entry);
1494 }
1495
1496 void sysctl_head_get(struct ctl_table_header *head)
1497 {
1498         spin_lock(&sysctl_lock);
1499         head->count++;
1500         spin_unlock(&sysctl_lock);
1501 }
1502
1503 void sysctl_head_put(struct ctl_table_header *head)
1504 {
1505         spin_lock(&sysctl_lock);
1506         if (!--head->count)
1507                 kfree(head);
1508         spin_unlock(&sysctl_lock);
1509 }
1510
1511 struct ctl_table_header *sysctl_head_grab(struct ctl_table_header *head)
1512 {
1513         if (!head)
1514                 BUG();
1515         spin_lock(&sysctl_lock);
1516         if (!use_table(head))
1517                 head = ERR_PTR(-ENOENT);
1518         spin_unlock(&sysctl_lock);
1519         return head;
1520 }
1521
1522 void sysctl_head_finish(struct ctl_table_header *head)
1523 {
1524         if (!head)
1525                 return;
1526         spin_lock(&sysctl_lock);
1527         unuse_table(head);
1528         spin_unlock(&sysctl_lock);
1529 }
1530
1531 static struct ctl_table_set *
1532 lookup_header_set(struct ctl_table_root *root, struct nsproxy *namespaces)
1533 {
1534         struct ctl_table_set *set = &root->default_set;
1535         if (root->lookup)
1536                 set = root->lookup(root, namespaces);
1537         return set;
1538 }
1539
1540 static struct list_head *
1541 lookup_header_list(struct ctl_table_root *root, struct nsproxy *namespaces)
1542 {
1543         struct ctl_table_set *set = lookup_header_set(root, namespaces);
1544         return &set->list;
1545 }
1546
1547 struct ctl_table_header *__sysctl_head_next(struct nsproxy *namespaces,
1548                                             struct ctl_table_header *prev)
1549 {
1550         struct ctl_table_root *root;
1551         struct list_head *header_list;
1552         struct ctl_table_header *head;
1553         struct list_head *tmp;
1554
1555         spin_lock(&sysctl_lock);
1556         if (prev) {
1557                 head = prev;
1558                 tmp = &prev->ctl_entry;
1559                 unuse_table(prev);
1560                 goto next;
1561         }
1562         tmp = &root_table_header.ctl_entry;
1563         for (;;) {
1564                 head = list_entry(tmp, struct ctl_table_header, ctl_entry);
1565
1566                 if (!use_table(head))
1567                         goto next;
1568                 spin_unlock(&sysctl_lock);
1569                 return head;
1570         next:
1571                 root = head->root;
1572                 tmp = tmp->next;
1573                 header_list = lookup_header_list(root, namespaces);
1574                 if (tmp != header_list)
1575                         continue;
1576
1577                 do {
1578                         root = list_entry(root->root_list.next,
1579                                         struct ctl_table_root, root_list);
1580                         if (root == &sysctl_table_root)
1581                                 goto out;
1582                         header_list = lookup_header_list(root, namespaces);
1583                 } while (list_empty(header_list));
1584                 tmp = header_list->next;
1585         }
1586 out:
1587         spin_unlock(&sysctl_lock);
1588         return NULL;
1589 }
1590
1591 struct ctl_table_header *sysctl_head_next(struct ctl_table_header *prev)
1592 {
1593         return __sysctl_head_next(current->nsproxy, prev);
1594 }
1595
1596 void register_sysctl_root(struct ctl_table_root *root)
1597 {
1598         spin_lock(&sysctl_lock);
1599         list_add_tail(&root->root_list, &sysctl_table_root.root_list);
1600         spin_unlock(&sysctl_lock);
1601 }
1602
1603 #ifdef CONFIG_SYSCTL_SYSCALL
1604 /* Perform the actual read/write of a sysctl table entry. */
1605 static int do_sysctl_strategy(struct ctl_table_root *root,
1606                         struct ctl_table *table,
1607                         void __user *oldval, size_t __user *oldlenp,
1608                         void __user *newval, size_t newlen)
1609 {
1610         int op = 0, rc;
1611
1612         if (oldval)
1613                 op |= MAY_READ;
1614         if (newval)
1615                 op |= MAY_WRITE;
1616         if (sysctl_perm(root, table, op))
1617                 return -EPERM;
1618
1619         if (table->strategy) {
1620                 rc = table->strategy(table, oldval, oldlenp, newval, newlen);
1621                 if (rc < 0)
1622                         return rc;
1623                 if (rc > 0)
1624                         return 0;
1625         }
1626
1627         /* If there is no strategy routine, or if the strategy returns
1628          * zero, proceed with automatic r/w */
1629         if (table->data && table->maxlen) {
1630                 rc = sysctl_data(table, oldval, oldlenp, newval, newlen);
1631                 if (rc < 0)
1632                         return rc;
1633         }
1634         return 0;
1635 }
1636
1637 static int parse_table(int __user *name, int nlen,
1638                        void __user *oldval, size_t __user *oldlenp,
1639                        void __user *newval, size_t newlen,
1640                        struct ctl_table_root *root,
1641                        struct ctl_table *table)
1642 {
1643         int n;
1644 repeat:
1645         if (!nlen)
1646                 return -ENOTDIR;
1647         if (get_user(n, name))
1648                 return -EFAULT;
1649         for ( ; table->ctl_name || table->procname; table++) {
1650                 if (!table->ctl_name)
1651                         continue;
1652                 if (n == table->ctl_name) {
1653                         int error;
1654                         if (table->child) {
1655                                 if (sysctl_perm(root, table, MAY_EXEC))
1656                                         return -EPERM;
1657                                 name++;
1658                                 nlen--;
1659                                 table = table->child;
1660                                 goto repeat;
1661                         }
1662                         error = do_sysctl_strategy(root, table,
1663                                                    oldval, oldlenp,
1664                                                    newval, newlen);
1665                         return error;
1666                 }
1667         }
1668         return -ENOTDIR;
1669 }
1670
1671 int do_sysctl(int __user *name, int nlen, void __user *oldval, size_t __user *oldlenp,
1672                void __user *newval, size_t newlen)
1673 {
1674         struct ctl_table_header *head;
1675         int error = -ENOTDIR;
1676
1677         if (nlen <= 0 || nlen >= CTL_MAXNAME)
1678                 return -ENOTDIR;
1679         if (oldval) {
1680                 int old_len;
1681                 if (!oldlenp || get_user(old_len, oldlenp))
1682                         return -EFAULT;
1683         }
1684
1685         for (head = sysctl_head_next(NULL); head;
1686                         head = sysctl_head_next(head)) {
1687                 error = parse_table(name, nlen, oldval, oldlenp, 
1688                                         newval, newlen,
1689                                         head->root, head->ctl_table);
1690                 if (error != -ENOTDIR) {
1691                         sysctl_head_finish(head);
1692                         break;
1693                 }
1694         }
1695         return error;
1696 }
1697
1698 SYSCALL_DEFINE1(sysctl, struct __sysctl_args __user *, args)
1699 {
1700         struct __sysctl_args tmp;
1701         int error;
1702
1703         if (copy_from_user(&tmp, args, sizeof(tmp)))
1704                 return -EFAULT;
1705
1706         error = deprecated_sysctl_warning(&tmp);
1707         if (error)
1708                 goto out;
1709
1710         lock_kernel();
1711         error = do_sysctl(tmp.name, tmp.nlen, tmp.oldval, tmp.oldlenp,
1712                           tmp.newval, tmp.newlen);
1713         unlock_kernel();
1714 out:
1715         return error;
1716 }
1717 #endif /* CONFIG_SYSCTL_SYSCALL */
1718
1719 /*
1720  * sysctl_perm does NOT grant the superuser all rights automatically, because
1721  * some sysctl variables are readonly even to root.
1722  */
1723
1724 static int test_perm(int mode, int op)
1725 {
1726         if (!current_euid())
1727                 mode >>= 6;
1728         else if (in_egroup_p(0))
1729                 mode >>= 3;
1730         if ((op & ~mode & (MAY_READ|MAY_WRITE|MAY_EXEC)) == 0)
1731                 return 0;
1732         return -EACCES;
1733 }
1734
1735 int sysctl_perm(struct ctl_table_root *root, struct ctl_table *table, int op)
1736 {
1737         int error;
1738         int mode;
1739
1740         error = security_sysctl(table, op & (MAY_READ | MAY_WRITE | MAY_EXEC));
1741         if (error)
1742                 return error;
1743
1744         if (root->permissions)
1745                 mode = root->permissions(root, current->nsproxy, table);
1746         else
1747                 mode = table->mode;
1748
1749         return test_perm(mode, op);
1750 }
1751
1752 static void sysctl_set_parent(struct ctl_table *parent, struct ctl_table *table)
1753 {
1754         for (; table->ctl_name || table->procname; table++) {
1755                 table->parent = parent;
1756                 if (table->child)
1757                         sysctl_set_parent(table, table->child);
1758         }
1759 }
1760
1761 static __init int sysctl_init(void)
1762 {
1763         sysctl_set_parent(NULL, root_table);
1764 #ifdef CONFIG_SYSCTL_SYSCALL_CHECK
1765         {
1766                 int err;
1767                 err = sysctl_check_table(current->nsproxy, root_table);
1768         }
1769 #endif
1770         return 0;
1771 }
1772
1773 core_initcall(sysctl_init);
1774
1775 static struct ctl_table *is_branch_in(struct ctl_table *branch,
1776                                       struct ctl_table *table)
1777 {
1778         struct ctl_table *p;
1779         const char *s = branch->procname;
1780
1781         /* branch should have named subdirectory as its first element */
1782         if (!s || !branch->child)
1783                 return NULL;
1784
1785         /* ... and nothing else */
1786         if (branch[1].procname || branch[1].ctl_name)
1787                 return NULL;
1788
1789         /* table should contain subdirectory with the same name */
1790         for (p = table; p->procname || p->ctl_name; p++) {
1791                 if (!p->child)
1792                         continue;
1793                 if (p->procname && strcmp(p->procname, s) == 0)
1794                         return p;
1795         }
1796         return NULL;
1797 }
1798
1799 /* see if attaching q to p would be an improvement */
1800 static void try_attach(struct ctl_table_header *p, struct ctl_table_header *q)
1801 {
1802         struct ctl_table *to = p->ctl_table, *by = q->ctl_table;
1803         struct ctl_table *next;
1804         int is_better = 0;
1805         int not_in_parent = !p->attached_by;
1806
1807         while ((next = is_branch_in(by, to)) != NULL) {
1808                 if (by == q->attached_by)
1809                         is_better = 1;
1810                 if (to == p->attached_by)
1811                         not_in_parent = 1;
1812                 by = by->child;
1813                 to = next->child;
1814         }
1815
1816         if (is_better && not_in_parent) {
1817                 q->attached_by = by;
1818                 q->attached_to = to;
1819                 q->parent = p;
1820         }
1821 }
1822
1823 /**
1824  * __register_sysctl_paths - register a sysctl hierarchy
1825  * @root: List of sysctl headers to register on
1826  * @namespaces: Data to compute which lists of sysctl entries are visible
1827  * @path: The path to the directory the sysctl table is in.
1828  * @table: the top-level table structure
1829  *
1830  * Register a sysctl table hierarchy. @table should be a filled in ctl_table
1831  * array. A completely 0 filled entry terminates the table.
1832  *
1833  * The members of the &struct ctl_table structure are used as follows:
1834  *
1835  * ctl_name - This is the numeric sysctl value used by sysctl(2). The number
1836  *            must be unique within that level of sysctl
1837  *
1838  * procname - the name of the sysctl file under /proc/sys. Set to %NULL to not
1839  *            enter a sysctl file
1840  *
1841  * data - a pointer to data for use by proc_handler
1842  *
1843  * maxlen - the maximum size in bytes of the data
1844  *
1845  * mode - the file permissions for the /proc/sys file, and for sysctl(2)
1846  *
1847  * child - a pointer to the child sysctl table if this entry is a directory, or
1848  *         %NULL.
1849  *
1850  * proc_handler - the text handler routine (described below)
1851  *
1852  * strategy - the strategy routine (described below)
1853  *
1854  * de - for internal use by the sysctl routines
1855  *
1856  * extra1, extra2 - extra pointers usable by the proc handler routines
1857  *
1858  * Leaf nodes in the sysctl tree will be represented by a single file
1859  * under /proc; non-leaf nodes will be represented by directories.
1860  *
1861  * sysctl(2) can automatically manage read and write requests through
1862  * the sysctl table.  The data and maxlen fields of the ctl_table
1863  * struct enable minimal validation of the values being written to be
1864  * performed, and the mode field allows minimal authentication.
1865  *
1866  * More sophisticated management can be enabled by the provision of a
1867  * strategy routine with the table entry.  This will be called before
1868  * any automatic read or write of the data is performed.
1869  *
1870  * The strategy routine may return
1871  *
1872  * < 0 - Error occurred (error is passed to user process)
1873  *
1874  * 0   - OK - proceed with automatic read or write.
1875  *
1876  * > 0 - OK - read or write has been done by the strategy routine, so
1877  *       return immediately.
1878  *
1879  * There must be a proc_handler routine for any terminal nodes
1880  * mirrored under /proc/sys (non-terminals are handled by a built-in
1881  * directory handler).  Several default handlers are available to
1882  * cover common cases -
1883  *
1884  * proc_dostring(), proc_dointvec(), proc_dointvec_jiffies(),
1885  * proc_dointvec_userhz_jiffies(), proc_dointvec_minmax(), 
1886  * proc_doulongvec_ms_jiffies_minmax(), proc_doulongvec_minmax()
1887  *
1888  * It is the handler's job to read the input buffer from user memory
1889  * and process it. The handler should return 0 on success.
1890  *
1891  * This routine returns %NULL on a failure to register, and a pointer
1892  * to the table header on success.
1893  */
1894 struct ctl_table_header *__register_sysctl_paths(
1895         struct ctl_table_root *root,
1896         struct nsproxy *namespaces,
1897         const struct ctl_path *path, struct ctl_table *table)
1898 {
1899         struct ctl_table_header *header;
1900         struct ctl_table *new, **prevp;
1901         unsigned int n, npath;
1902         struct ctl_table_set *set;
1903
1904         /* Count the path components */
1905         for (npath = 0; path[npath].ctl_name || path[npath].procname; ++npath)
1906                 ;
1907
1908         /*
1909          * For each path component, allocate a 2-element ctl_table array.
1910          * The first array element will be filled with the sysctl entry
1911          * for this, the second will be the sentinel (ctl_name == 0).
1912          *
1913          * We allocate everything in one go so that we don't have to
1914          * worry about freeing additional memory in unregister_sysctl_table.
1915          */
1916         header = kzalloc(sizeof(struct ctl_table_header) +
1917                          (2 * npath * sizeof(struct ctl_table)), GFP_KERNEL);
1918         if (!header)
1919                 return NULL;
1920
1921         new = (struct ctl_table *) (header + 1);
1922
1923         /* Now connect the dots */
1924         prevp = &header->ctl_table;
1925         for (n = 0; n < npath; ++n, ++path) {
1926                 /* Copy the procname */
1927                 new->procname = path->procname;
1928                 new->ctl_name = path->ctl_name;
1929                 new->mode     = 0555;
1930
1931                 *prevp = new;
1932                 prevp = &new->child;
1933
1934                 new += 2;
1935         }
1936         *prevp = table;
1937         header->ctl_table_arg = table;
1938
1939         INIT_LIST_HEAD(&header->ctl_entry);
1940         header->used = 0;
1941         header->unregistering = NULL;
1942         header->root = root;
1943         sysctl_set_parent(NULL, header->ctl_table);
1944         header->count = 1;
1945 #ifdef CONFIG_SYSCTL_SYSCALL_CHECK
1946         if (sysctl_check_table(namespaces, header->ctl_table)) {
1947                 kfree(header);
1948                 return NULL;
1949         }
1950 #endif
1951         spin_lock(&sysctl_lock);
1952         header->set = lookup_header_set(root, namespaces);
1953         header->attached_by = header->ctl_table;
1954         header->attached_to = root_table;
1955         header->parent = &root_table_header;
1956         for (set = header->set; set; set = set->parent) {
1957                 struct ctl_table_header *p;
1958                 list_for_each_entry(p, &set->list, ctl_entry) {
1959                         if (p->unregistering)
1960                                 continue;
1961                         try_attach(p, header);
1962                 }
1963         }
1964         header->parent->count++;
1965         list_add_tail(&header->ctl_entry, &header->set->list);
1966         spin_unlock(&sysctl_lock);
1967
1968         return header;
1969 }
1970
1971 /**
1972  * register_sysctl_table_path - register a sysctl table hierarchy
1973  * @path: The path to the directory the sysctl table is in.
1974  * @table: the top-level table structure
1975  *
1976  * Register a sysctl table hierarchy. @table should be a filled in ctl_table
1977  * array. A completely 0 filled entry terminates the table.
1978  *
1979  * See __register_sysctl_paths for more details.
1980  */
1981 struct ctl_table_header *register_sysctl_paths(const struct ctl_path *path,
1982                                                 struct ctl_table *table)
1983 {
1984         return __register_sysctl_paths(&sysctl_table_root, current->nsproxy,
1985                                         path, table);
1986 }
1987
1988 /**
1989  * register_sysctl_table - register a sysctl table hierarchy
1990  * @table: the top-level table structure
1991  *
1992  * Register a sysctl table hierarchy. @table should be a filled in ctl_table
1993  * array. A completely 0 filled entry terminates the table.
1994  *
1995  * See register_sysctl_paths for more details.
1996  */
1997 struct ctl_table_header *register_sysctl_table(struct ctl_table *table)
1998 {
1999         static const struct ctl_path null_path[] = { {} };
2000
2001         return register_sysctl_paths(null_path, table);
2002 }
2003
2004 /**
2005  * unregister_sysctl_table - unregister a sysctl table hierarchy
2006  * @header: the header returned from register_sysctl_table
2007  *
2008  * Unregisters the sysctl table and all children. proc entries may not
2009  * actually be removed until they are no longer used by anyone.
2010  */
2011 void unregister_sysctl_table(struct ctl_table_header * header)
2012 {
2013         might_sleep();
2014
2015         if (header == NULL)
2016                 return;
2017
2018         spin_lock(&sysctl_lock);
2019         start_unregistering(header);
2020         if (!--header->parent->count) {
2021                 WARN_ON(1);
2022                 kfree(header->parent);
2023         }
2024         if (!--header->count)
2025                 kfree(header);
2026         spin_unlock(&sysctl_lock);
2027 }
2028
2029 int sysctl_is_seen(struct ctl_table_header *p)
2030 {
2031         struct ctl_table_set *set = p->set;
2032         int res;
2033         spin_lock(&sysctl_lock);
2034         if (p->unregistering)
2035                 res = 0;
2036         else if (!set->is_seen)
2037                 res = 1;
2038         else
2039                 res = set->is_seen(set);
2040         spin_unlock(&sysctl_lock);
2041         return res;
2042 }
2043
2044 void setup_sysctl_set(struct ctl_table_set *p,
2045         struct ctl_table_set *parent,
2046         int (*is_seen)(struct ctl_table_set *))
2047 {
2048         INIT_LIST_HEAD(&p->list);
2049         p->parent = parent ? parent : &sysctl_table_root.default_set;
2050         p->is_seen = is_seen;
2051 }
2052
2053 #else /* !CONFIG_SYSCTL */
2054 struct ctl_table_header *register_sysctl_table(struct ctl_table * table)
2055 {
2056         return NULL;
2057 }
2058
2059 struct ctl_table_header *register_sysctl_paths(const struct ctl_path *path,
2060                                                     struct ctl_table *table)
2061 {
2062         return NULL;
2063 }
2064
2065 void unregister_sysctl_table(struct ctl_table_header * table)
2066 {
2067 }
2068
2069 void setup_sysctl_set(struct ctl_table_set *p,
2070         struct ctl_table_set *parent,
2071         int (*is_seen)(struct ctl_table_set *))
2072 {
2073 }
2074
2075 void sysctl_head_put(struct ctl_table_header *head)
2076 {
2077 }
2078
2079 #endif /* CONFIG_SYSCTL */
2080
2081 /*
2082  * /proc/sys support
2083  */
2084
2085 #ifdef CONFIG_PROC_SYSCTL
2086
2087 static int _proc_do_string(void* data, int maxlen, int write,
2088                            struct file *filp, void __user *buffer,
2089                            size_t *lenp, loff_t *ppos)
2090 {
2091         size_t len;
2092         char __user *p;
2093         char c;
2094
2095         if (!data || !maxlen || !*lenp) {
2096                 *lenp = 0;
2097                 return 0;
2098         }
2099
2100         if (write) {
2101                 len = 0;
2102                 p = buffer;
2103                 while (len < *lenp) {
2104                         if (get_user(c, p++))
2105                                 return -EFAULT;
2106                         if (c == 0 || c == '\n')
2107                                 break;
2108                         len++;
2109                 }
2110                 if (len >= maxlen)
2111                         len = maxlen-1;
2112                 if(copy_from_user(data, buffer, len))
2113                         return -EFAULT;
2114                 ((char *) data)[len] = 0;
2115                 *ppos += *lenp;
2116         } else {
2117                 len = strlen(data);
2118                 if (len > maxlen)
2119                         len = maxlen;
2120
2121                 if (*ppos > len) {
2122                         *lenp = 0;
2123                         return 0;
2124                 }
2125
2126                 data += *ppos;
2127                 len  -= *ppos;
2128
2129                 if (len > *lenp)
2130                         len = *lenp;
2131                 if (len)
2132                         if(copy_to_user(buffer, data, len))
2133                                 return -EFAULT;
2134                 if (len < *lenp) {
2135                         if(put_user('\n', ((char __user *) buffer) + len))
2136                                 return -EFAULT;
2137                         len++;
2138                 }
2139                 *lenp = len;
2140                 *ppos += len;
2141         }
2142         return 0;
2143 }
2144
2145 /**
2146  * proc_dostring - read a string sysctl
2147  * @table: the sysctl table
2148  * @write: %TRUE if this is a write to the sysctl file
2149  * @filp: the file structure
2150  * @buffer: the user buffer
2151  * @lenp: the size of the user buffer
2152  * @ppos: file position
2153  *
2154  * Reads/writes a string from/to the user buffer. If the kernel
2155  * buffer provided is not large enough to hold the string, the
2156  * string is truncated. The copied string is %NULL-terminated.
2157  * If the string is being read by the user process, it is copied
2158  * and a newline '\n' is added. It is truncated if the buffer is
2159  * not large enough.
2160  *
2161  * Returns 0 on success.
2162  */
2163 int proc_dostring(struct ctl_table *table, int write, struct file *filp,
2164                   void __user *buffer, size_t *lenp, loff_t *ppos)
2165 {
2166         return _proc_do_string(table->data, table->maxlen, write, filp,
2167                                buffer, lenp, ppos);
2168 }
2169
2170
2171 static int do_proc_dointvec_conv(int *negp, unsigned long *lvalp,
2172                                  int *valp,
2173                                  int write, void *data)
2174 {
2175         if (write) {
2176                 *valp = *negp ? -*lvalp : *lvalp;
2177         } else {
2178                 int val = *valp;
2179                 if (val < 0) {
2180                         *negp = -1;
2181                         *lvalp = (unsigned long)-val;
2182                 } else {
2183                         *negp = 0;
2184                         *lvalp = (unsigned long)val;
2185                 }
2186         }
2187         return 0;
2188 }
2189
2190 static int __do_proc_dointvec(void *tbl_data, struct ctl_table *table,
2191                   int write, struct file *filp, void __user *buffer,
2192                   size_t *lenp, loff_t *ppos,
2193                   int (*conv)(int *negp, unsigned long *lvalp, int *valp,
2194                               int write, void *data),
2195                   void *data)
2196 {
2197 #define TMPBUFLEN 21
2198         int *i, vleft, first=1, neg, val;
2199         unsigned long lval;
2200         size_t left, len;
2201         
2202         char buf[TMPBUFLEN], *p;
2203         char __user *s = buffer;
2204         
2205         if (!tbl_data || !table->maxlen || !*lenp ||
2206             (*ppos && !write)) {
2207                 *lenp = 0;
2208                 return 0;
2209         }
2210         
2211         i = (int *) tbl_data;
2212         vleft = table->maxlen / sizeof(*i);
2213         left = *lenp;
2214
2215         if (!conv)
2216                 conv = do_proc_dointvec_conv;
2217
2218         for (; left && vleft--; i++, first=0) {
2219                 if (write) {
2220                         while (left) {
2221                                 char c;
2222                                 if (get_user(c, s))
2223                                         return -EFAULT;
2224                                 if (!isspace(c))
2225                                         break;
2226                                 left--;
2227                                 s++;
2228                         }
2229                         if (!left)
2230                                 break;
2231                         neg = 0;
2232                         len = left;
2233                         if (len > sizeof(buf) - 1)
2234                                 len = sizeof(buf) - 1;
2235                         if (copy_from_user(buf, s, len))
2236                                 return -EFAULT;
2237                         buf[len] = 0;
2238                         p = buf;
2239                         if (*p == '-' && left > 1) {
2240                                 neg = 1;
2241                                 p++;
2242                         }
2243                         if (*p < '0' || *p > '9')
2244                                 break;
2245
2246                         lval = simple_strtoul(p, &p, 0);
2247
2248                         len = p-buf;
2249                         if ((len < left) && *p && !isspace(*p))
2250                                 break;
2251                         if (neg)
2252                                 val = -val;
2253                         s += len;
2254                         left -= len;
2255
2256                         if (conv(&neg, &lval, i, 1, data))
2257                                 break;
2258                 } else {
2259                         p = buf;
2260                         if (!first)
2261                                 *p++ = '\t';
2262         
2263                         if (conv(&neg, &lval, i, 0, data))
2264                                 break;
2265
2266                         sprintf(p, "%s%lu", neg ? "-" : "", lval);
2267                         len = strlen(buf);
2268                         if (len > left)
2269                                 len = left;
2270                         if(copy_to_user(s, buf, len))
2271                                 return -EFAULT;
2272                         left -= len;
2273                         s += len;
2274                 }
2275         }
2276
2277         if (!write && !first && left) {
2278                 if(put_user('\n', s))
2279                         return -EFAULT;
2280                 left--, s++;
2281         }
2282         if (write) {
2283                 while (left) {
2284                         char c;
2285                         if (get_user(c, s++))
2286                                 return -EFAULT;
2287                         if (!isspace(c))
2288                                 break;
2289                         left--;
2290                 }
2291         }
2292         if (write && first)
2293                 return -EINVAL;
2294         *lenp -= left;
2295         *ppos += *lenp;
2296         return 0;
2297 #undef TMPBUFLEN
2298 }
2299
2300 static int do_proc_dointvec(struct ctl_table *table, int write, struct file *filp,
2301                   void __user *buffer, size_t *lenp, loff_t *ppos,
2302                   int (*conv)(int *negp, unsigned long *lvalp, int *valp,
2303                               int write, void *data),
2304                   void *data)
2305 {
2306         return __do_proc_dointvec(table->data, table, write, filp,
2307                         buffer, lenp, ppos, conv, data);
2308 }
2309
2310 /**
2311  * proc_dointvec - read a vector of integers
2312  * @table: the sysctl table
2313  * @write: %TRUE if this is a write to the sysctl file
2314  * @filp: the file structure
2315  * @buffer: the user buffer
2316  * @lenp: the size of the user buffer
2317  * @ppos: file position
2318  *
2319  * Reads/writes up to table->maxlen/sizeof(unsigned int) integer
2320  * values from/to the user buffer, treated as an ASCII string. 
2321  *
2322  * Returns 0 on success.
2323  */
2324 int proc_dointvec(struct ctl_table *table, int write, struct file *filp,
2325                      void __user *buffer, size_t *lenp, loff_t *ppos)
2326 {
2327     return do_proc_dointvec(table,write,filp,buffer,lenp,ppos,
2328                             NULL,NULL);
2329 }
2330
2331 /*
2332  * Taint values can only be increased
2333  * This means we can safely use a temporary.
2334  */
2335 static int proc_taint(struct ctl_table *table, int write, struct file *filp,
2336                                void __user *buffer, size_t *lenp, loff_t *ppos)
2337 {
2338         struct ctl_table t;
2339         unsigned long tmptaint = get_taint();
2340         int err;
2341
2342         if (write && !capable(CAP_SYS_ADMIN))
2343                 return -EPERM;
2344
2345         t = *table;
2346         t.data = &tmptaint;
2347         err = proc_doulongvec_minmax(&t, write, filp, buffer, lenp, ppos);
2348         if (err < 0)
2349                 return err;
2350
2351         if (write) {
2352                 /*
2353                  * Poor man's atomic or. Not worth adding a primitive
2354                  * to everyone's atomic.h for this
2355                  */
2356                 int i;
2357                 for (i = 0; i < BITS_PER_LONG && tmptaint >> i; i++) {
2358                         if ((tmptaint >> i) & 1)
2359                                 add_taint(i);
2360                 }
2361         }
2362
2363         return err;
2364 }
2365
2366 struct do_proc_dointvec_minmax_conv_param {
2367         int *min;
2368         int *max;
2369 };
2370
2371 static int do_proc_dointvec_minmax_conv(int *negp, unsigned long *lvalp, 
2372                                         int *valp, 
2373                                         int write, void *data)
2374 {
2375         struct do_proc_dointvec_minmax_conv_param *param = data;
2376         if (write) {
2377                 int val = *negp ? -*lvalp : *lvalp;
2378                 if ((param->min && *param->min > val) ||
2379                     (param->max && *param->max < val))
2380                         return -EINVAL;
2381                 *valp = val;
2382         } else {
2383                 int val = *valp;
2384                 if (val < 0) {
2385                         *negp = -1;
2386                         *lvalp = (unsigned long)-val;
2387                 } else {
2388                         *negp = 0;
2389                         *lvalp = (unsigned long)val;
2390                 }
2391         }
2392         return 0;
2393 }
2394
2395 /**
2396  * proc_dointvec_minmax - read a vector of integers with min/max values
2397  * @table: the sysctl table
2398  * @write: %TRUE if this is a write to the sysctl file
2399  * @filp: the file structure
2400  * @buffer: the user buffer
2401  * @lenp: the size of the user buffer
2402  * @ppos: file position
2403  *
2404  * Reads/writes up to table->maxlen/sizeof(unsigned int) integer
2405  * values from/to the user buffer, treated as an ASCII string.
2406  *
2407  * This routine will ensure the values are within the range specified by
2408  * table->extra1 (min) and table->extra2 (max).
2409  *
2410  * Returns 0 on success.
2411  */
2412 int proc_dointvec_minmax(struct ctl_table *table, int write, struct file *filp,
2413                   void __user *buffer, size_t *lenp, loff_t *ppos)
2414 {
2415         struct do_proc_dointvec_minmax_conv_param param = {
2416                 .min = (int *) table->extra1,
2417                 .max = (int *) table->extra2,
2418         };
2419         return do_proc_dointvec(table, write, filp, buffer, lenp, ppos,
2420                                 do_proc_dointvec_minmax_conv, &param);
2421 }
2422
2423 static int __do_proc_doulongvec_minmax(void *data, struct ctl_table *table, int write,
2424                                      struct file *filp,
2425                                      void __user *buffer,
2426                                      size_t *lenp, loff_t *ppos,
2427                                      unsigned long convmul,
2428                                      unsigned long convdiv)
2429 {
2430 #define TMPBUFLEN 21
2431         unsigned long *i, *min, *max, val;
2432         int vleft, first=1, neg;
2433         size_t len, left;
2434         char buf[TMPBUFLEN], *p;
2435         char __user *s = buffer;
2436         
2437         if (!data || !table->maxlen || !*lenp ||
2438             (*ppos && !write)) {
2439                 *lenp = 0;
2440                 return 0;
2441         }
2442         
2443         i = (unsigned long *) data;
2444         min = (unsigned long *) table->extra1;
2445         max = (unsigned long *) table->extra2;
2446         vleft = table->maxlen / sizeof(unsigned long);
2447         left = *lenp;
2448         
2449         for (; left && vleft--; i++, min++, max++, first=0) {
2450                 if (write) {
2451                         while (left) {
2452                                 char c;
2453                                 if (get_user(c, s))
2454                                         return -EFAULT;
2455                                 if (!isspace(c))
2456                                         break;
2457                                 left--;
2458                                 s++;
2459                         }
2460                         if (!left)
2461                                 break;
2462                         neg = 0;
2463                         len = left;
2464                         if (len > TMPBUFLEN-1)
2465                                 len = TMPBUFLEN-1;
2466                         if (copy_from_user(buf, s, len))
2467                                 return -EFAULT;
2468                         buf[len] = 0;
2469                         p = buf;
2470                         if (*p == '-' && left > 1) {
2471                                 neg = 1;
2472                                 p++;
2473                         }
2474                         if (*p < '0' || *p > '9')
2475                                 break;
2476                         val = simple_strtoul(p, &p, 0) * convmul / convdiv ;
2477                         len = p-buf;
2478                         if ((len < left) && *p && !isspace(*p))
2479                                 break;
2480                         if (neg)
2481                                 val = -val;
2482                         s += len;
2483                         left -= len;
2484
2485                         if(neg)
2486                                 continue;
2487                         if ((min && val < *min) || (max && val > *max))
2488                                 continue;
2489                         *i = val;
2490                 } else {
2491                         p = buf;
2492                         if (!first)
2493                                 *p++ = '\t';
2494                         sprintf(p, "%lu", convdiv * (*i) / convmul);
2495                         len = strlen(buf);
2496                         if (len > left)
2497                                 len = left;
2498                         if(copy_to_user(s, buf, len))
2499                                 return -EFAULT;
2500                         left -= len;
2501                         s += len;
2502                 }
2503         }
2504
2505         if (!write && !first && left) {
2506                 if(put_user('\n', s))
2507                         return -EFAULT;
2508                 left--, s++;
2509         }
2510         if (write) {
2511                 while (left) {
2512                         char c;
2513                         if (get_user(c, s++))
2514                                 return -EFAULT;
2515                         if (!isspace(c))
2516                                 break;
2517                         left--;
2518                 }
2519         }
2520         if (write && first)
2521                 return -EINVAL;
2522         *lenp -= left;
2523         *ppos += *lenp;
2524         return 0;
2525 #undef TMPBUFLEN
2526 }
2527
2528 static int do_proc_doulongvec_minmax(struct ctl_table *table, int write,
2529                                      struct file *filp,
2530                                      void __user *buffer,
2531                                      size_t *lenp, loff_t *ppos,
2532                                      unsigned long convmul,
2533                                      unsigned long convdiv)
2534 {
2535         return __do_proc_doulongvec_minmax(table->data, table, write,
2536                         filp, buffer, lenp, ppos, convmul, convdiv);
2537 }
2538
2539 /**
2540  * proc_doulongvec_minmax - read a vector of long integers with min/max values
2541  * @table: the sysctl table
2542  * @write: %TRUE if this is a write to the sysctl file
2543  * @filp: the file structure
2544  * @buffer: the user buffer
2545  * @lenp: the size of the user buffer
2546  * @ppos: file position
2547  *
2548  * Reads/writes up to table->maxlen/sizeof(unsigned long) unsigned long
2549  * values from/to the user buffer, treated as an ASCII string.
2550  *
2551  * This routine will ensure the values are within the range specified by
2552  * table->extra1 (min) and table->extra2 (max).
2553  *
2554  * Returns 0 on success.
2555  */
2556 int proc_doulongvec_minmax(struct ctl_table *table, int write, struct file *filp,
2557                            void __user *buffer, size_t *lenp, loff_t *ppos)
2558 {
2559     return do_proc_doulongvec_minmax(table, write, filp, buffer, lenp, ppos, 1l, 1l);
2560 }
2561
2562 /**
2563  * proc_doulongvec_ms_jiffies_minmax - read a vector of millisecond values with min/max values
2564  * @table: the sysctl table
2565  * @write: %TRUE if this is a write to the sysctl file
2566  * @filp: the file structure
2567  * @buffer: the user buffer
2568  * @lenp: the size of the user buffer
2569  * @ppos: file position
2570  *
2571  * Reads/writes up to table->maxlen/sizeof(unsigned long) unsigned long
2572  * values from/to the user buffer, treated as an ASCII string. The values
2573  * are treated as milliseconds, and converted to jiffies when they are stored.
2574  *
2575  * This routine will ensure the values are within the range specified by
2576  * table->extra1 (min) and table->extra2 (max).
2577  *
2578  * Returns 0 on success.
2579  */
2580 int proc_doulongvec_ms_jiffies_minmax(struct ctl_table *table, int write,
2581                                       struct file *filp,
2582                                       void __user *buffer,
2583                                       size_t *lenp, loff_t *ppos)
2584 {
2585     return do_proc_doulongvec_minmax(table, write, filp, buffer,
2586                                      lenp, ppos, HZ, 1000l);
2587 }
2588
2589
2590 static int do_proc_dointvec_jiffies_conv(int *negp, unsigned long *lvalp,
2591                                          int *valp,
2592                                          int write, void *data)
2593 {
2594         if (write) {
2595                 if (*lvalp > LONG_MAX / HZ)
2596                         return 1;
2597                 *valp = *negp ? -(*lvalp*HZ) : (*lvalp*HZ);
2598         } else {
2599                 int val = *valp;
2600                 unsigned long lval;
2601                 if (val < 0) {
2602                         *negp = -1;
2603                         lval = (unsigned long)-val;
2604                 } else {
2605                         *negp = 0;
2606                         lval = (unsigned long)val;
2607                 }
2608                 *lvalp = lval / HZ;
2609         }
2610         return 0;
2611 }
2612
2613 static int do_proc_dointvec_userhz_jiffies_conv(int *negp, unsigned long *lvalp,
2614                                                 int *valp,
2615                                                 int write, void *data)
2616 {
2617         if (write) {
2618                 if (USER_HZ < HZ && *lvalp > (LONG_MAX / HZ) * USER_HZ)
2619                         return 1;
2620                 *valp = clock_t_to_jiffies(*negp ? -*lvalp : *lvalp);
2621         } else {
2622                 int val = *valp;
2623                 unsigned long lval;
2624                 if (val < 0) {
2625                         *negp = -1;
2626                         lval = (unsigned long)-val;
2627                 } else {
2628                         *negp = 0;
2629                         lval = (unsigned long)val;
2630                 }
2631                 *lvalp = jiffies_to_clock_t(lval);
2632         }
2633         return 0;
2634 }
2635
2636 static int do_proc_dointvec_ms_jiffies_conv(int *negp, unsigned long *lvalp,
2637                                             int *valp,
2638                                             int write, void *data)
2639 {
2640         if (write) {
2641                 *valp = msecs_to_jiffies(*negp ? -*lvalp : *lvalp);
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 = jiffies_to_msecs(lval);
2653         }
2654         return 0;
2655 }
2656
2657 /**
2658  * proc_dointvec_jiffies - read a vector of integers as seconds
2659  * @table: the sysctl table
2660  * @write: %TRUE if this is a write to the sysctl file
2661  * @filp: the file structure
2662  * @buffer: the user buffer
2663  * @lenp: the size of the user buffer
2664  * @ppos: file position
2665  *
2666  * Reads/writes up to table->maxlen/sizeof(unsigned int) integer
2667  * values from/to the user buffer, treated as an ASCII string. 
2668  * The values read are assumed to be in seconds, and are converted into
2669  * jiffies.
2670  *
2671  * Returns 0 on success.
2672  */
2673 int proc_dointvec_jiffies(struct ctl_table *table, int write, struct file *filp,
2674                           void __user *buffer, size_t *lenp, loff_t *ppos)
2675 {
2676     return do_proc_dointvec(table,write,filp,buffer,lenp,ppos,
2677                             do_proc_dointvec_jiffies_conv,NULL);
2678 }
2679
2680 /**
2681  * proc_dointvec_userhz_jiffies - read a vector of integers as 1/USER_HZ seconds
2682  * @table: the sysctl table
2683  * @write: %TRUE if this is a write to the sysctl file
2684  * @filp: the file structure
2685  * @buffer: the user buffer
2686  * @lenp: the size of the user buffer
2687  * @ppos: pointer to the file position
2688  *
2689  * Reads/writes up to table->maxlen/sizeof(unsigned int) integer
2690  * values from/to the user buffer, treated as an ASCII string. 
2691  * The values read are assumed to be in 1/USER_HZ seconds, and 
2692  * are converted into jiffies.
2693  *
2694  * Returns 0 on success.
2695  */
2696 int proc_dointvec_userhz_jiffies(struct ctl_table *table, int write, struct file *filp,
2697                                  void __user *buffer, size_t *lenp, loff_t *ppos)
2698 {
2699     return do_proc_dointvec(table,write,filp,buffer,lenp,ppos,
2700                             do_proc_dointvec_userhz_jiffies_conv,NULL);
2701 }
2702
2703 /**
2704  * proc_dointvec_ms_jiffies - read a vector of integers as 1 milliseconds
2705  * @table: the sysctl table
2706  * @write: %TRUE if this is a write to the sysctl file
2707  * @filp: the file structure
2708  * @buffer: the user buffer
2709  * @lenp: the size of the user buffer
2710  * @ppos: file position
2711  * @ppos: the current position in the file
2712  *
2713  * Reads/writes up to table->maxlen/sizeof(unsigned int) integer
2714  * values from/to the user buffer, treated as an ASCII string. 
2715  * The values read are assumed to be in 1/1000 seconds, and 
2716  * are converted into jiffies.
2717  *
2718  * Returns 0 on success.
2719  */
2720 int proc_dointvec_ms_jiffies(struct ctl_table *table, int write, struct file *filp,
2721                              void __user *buffer, size_t *lenp, loff_t *ppos)
2722 {
2723         return do_proc_dointvec(table, write, filp, buffer, lenp, ppos,
2724                                 do_proc_dointvec_ms_jiffies_conv, NULL);
2725 }
2726
2727 static int proc_do_cad_pid(struct ctl_table *table, int write, struct file *filp,
2728                            void __user *buffer, size_t *lenp, loff_t *ppos)
2729 {
2730         struct pid *new_pid;
2731         pid_t tmp;
2732         int r;
2733
2734         tmp = pid_vnr(cad_pid);
2735
2736         r = __do_proc_dointvec(&tmp, table, write, filp, buffer,
2737                                lenp, ppos, NULL, NULL);
2738         if (r || !write)
2739                 return r;
2740
2741         new_pid = find_get_pid(tmp);
2742         if (!new_pid)
2743                 return -ESRCH;
2744
2745         put_pid(xchg(&cad_pid, new_pid));
2746         return 0;
2747 }
2748
2749 #else /* CONFIG_PROC_FS */
2750
2751 int proc_dostring(struct ctl_table *table, int write, struct file *filp,
2752                   void __user *buffer, size_t *lenp, loff_t *ppos)
2753 {
2754         return -ENOSYS;
2755 }
2756
2757 int proc_dointvec(struct ctl_table *table, int write, struct file *filp,
2758                   void __user *buffer, size_t *lenp, loff_t *ppos)
2759 {
2760         return -ENOSYS;
2761 }
2762
2763 int proc_dointvec_minmax(struct ctl_table *table, int write, struct file *filp,
2764                     void __user *buffer, size_t *lenp, loff_t *ppos)
2765 {
2766         return -ENOSYS;
2767 }
2768
2769 int proc_dointvec_jiffies(struct ctl_table *table, int write, struct file *filp,
2770                     void __user *buffer, size_t *lenp, loff_t *ppos)
2771 {
2772         return -ENOSYS;
2773 }
2774
2775 int proc_dointvec_userhz_jiffies(struct ctl_table *table, int write, struct file *filp,
2776                     void __user *buffer, size_t *lenp, loff_t *ppos)
2777 {
2778         return -ENOSYS;
2779 }
2780
2781 int proc_dointvec_ms_jiffies(struct ctl_table *table, int write, struct file *filp,
2782                              void __user *buffer, size_t *lenp, loff_t *ppos)
2783 {
2784         return -ENOSYS;
2785 }
2786
2787 int proc_doulongvec_minmax(struct ctl_table *table, int write, struct file *filp,
2788                     void __user *buffer, size_t *lenp, loff_t *ppos)
2789 {
2790         return -ENOSYS;
2791 }
2792
2793 int proc_doulongvec_ms_jiffies_minmax(struct ctl_table *table, int write,
2794                                       struct file *filp,
2795                                       void __user *buffer,
2796                                       size_t *lenp, loff_t *ppos)
2797 {
2798     return -ENOSYS;
2799 }
2800
2801
2802 #endif /* CONFIG_PROC_FS */
2803
2804
2805 #ifdef CONFIG_SYSCTL_SYSCALL
2806 /*
2807  * General sysctl support routines 
2808  */
2809
2810 /* The generic sysctl data routine (used if no strategy routine supplied) */
2811 int sysctl_data(struct ctl_table *table,
2812                 void __user *oldval, size_t __user *oldlenp,
2813                 void __user *newval, size_t newlen)
2814 {
2815         size_t len;
2816
2817         /* Get out of I don't have a variable */
2818         if (!table->data || !table->maxlen)
2819                 return -ENOTDIR;
2820
2821         if (oldval && oldlenp) {
2822                 if (get_user(len, oldlenp))
2823                         return -EFAULT;
2824                 if (len) {
2825                         if (len > table->maxlen)
2826                                 len = table->maxlen;
2827                         if (copy_to_user(oldval, table->data, len))
2828                                 return -EFAULT;
2829                         if (put_user(len, oldlenp))
2830                                 return -EFAULT;
2831                 }
2832         }
2833
2834         if (newval && newlen) {
2835                 if (newlen > table->maxlen)
2836                         newlen = table->maxlen;
2837
2838                 if (copy_from_user(table->data, newval, newlen))
2839                         return -EFAULT;
2840         }
2841         return 1;
2842 }
2843
2844 /* The generic string strategy routine: */
2845 int sysctl_string(struct ctl_table *table,
2846                   void __user *oldval, size_t __user *oldlenp,
2847                   void __user *newval, size_t newlen)
2848 {
2849         if (!table->data || !table->maxlen) 
2850                 return -ENOTDIR;
2851         
2852         if (oldval && oldlenp) {
2853                 size_t bufsize;
2854                 if (get_user(bufsize, oldlenp))
2855                         return -EFAULT;
2856                 if (bufsize) {
2857                         size_t len = strlen(table->data), copied;
2858
2859                         /* This shouldn't trigger for a well-formed sysctl */
2860                         if (len > table->maxlen)
2861                                 len = table->maxlen;
2862
2863                         /* Copy up to a max of bufsize-1 bytes of the string */
2864                         copied = (len >= bufsize) ? bufsize - 1 : len;
2865
2866                         if (copy_to_user(oldval, table->data, copied) ||
2867                             put_user(0, (char __user *)(oldval + copied)))
2868                                 return -EFAULT;
2869                         if (put_user(len, oldlenp))
2870                                 return -EFAULT;
2871                 }
2872         }
2873         if (newval && newlen) {
2874                 size_t len = newlen;
2875                 if (len > table->maxlen)
2876                         len = table->maxlen;
2877                 if(copy_from_user(table->data, newval, len))
2878                         return -EFAULT;
2879                 if (len == table->maxlen)
2880                         len--;
2881                 ((char *) table->data)[len] = 0;
2882         }
2883         return 1;
2884 }
2885
2886 /*
2887  * This function makes sure that all of the integers in the vector
2888  * are between the minimum and maximum values given in the arrays
2889  * table->extra1 and table->extra2, respectively.
2890  */
2891 int sysctl_intvec(struct ctl_table *table,
2892                 void __user *oldval, size_t __user *oldlenp,
2893                 void __user *newval, size_t newlen)
2894 {
2895
2896         if (newval && newlen) {
2897                 int __user *vec = (int __user *) newval;
2898                 int *min = (int *) table->extra1;
2899                 int *max = (int *) table->extra2;
2900                 size_t length;
2901                 int i;
2902
2903                 if (newlen % sizeof(int) != 0)
2904                         return -EINVAL;
2905
2906                 if (!table->extra1 && !table->extra2)
2907                         return 0;
2908
2909                 if (newlen > table->maxlen)
2910                         newlen = table->maxlen;
2911                 length = newlen / sizeof(int);
2912
2913                 for (i = 0; i < length; i++) {
2914                         int value;
2915                         if (get_user(value, vec + i))
2916                                 return -EFAULT;
2917                         if (min && value < min[i])
2918                                 return -EINVAL;
2919                         if (max && value > max[i])
2920                                 return -EINVAL;
2921                 }
2922         }
2923         return 0;
2924 }
2925
2926 /* Strategy function to convert jiffies to seconds */ 
2927 int sysctl_jiffies(struct ctl_table *table,
2928                 void __user *oldval, size_t __user *oldlenp,
2929                 void __user *newval, size_t newlen)
2930 {
2931         if (oldval && oldlenp) {
2932                 size_t olen;
2933
2934                 if (get_user(olen, oldlenp))
2935                         return -EFAULT;
2936                 if (olen) {
2937                         int val;
2938
2939                         if (olen < sizeof(int))
2940                                 return -EINVAL;
2941
2942                         val = *(int *)(table->data) / HZ;
2943                         if (put_user(val, (int __user *)oldval))
2944                                 return -EFAULT;
2945                         if (put_user(sizeof(int), oldlenp))
2946                                 return -EFAULT;
2947                 }
2948         }
2949         if (newval && newlen) { 
2950                 int new;
2951                 if (newlen != sizeof(int))
2952                         return -EINVAL; 
2953                 if (get_user(new, (int __user *)newval))
2954                         return -EFAULT;
2955                 *(int *)(table->data) = new*HZ; 
2956         }
2957         return 1;
2958 }
2959
2960 /* Strategy function to convert jiffies to seconds */ 
2961 int sysctl_ms_jiffies(struct ctl_table *table,
2962                 void __user *oldval, size_t __user *oldlenp,
2963                 void __user *newval, size_t newlen)
2964 {
2965         if (oldval && oldlenp) {
2966                 size_t olen;
2967
2968                 if (get_user(olen, oldlenp))
2969                         return -EFAULT;
2970                 if (olen) {
2971                         int val;
2972
2973                         if (olen < sizeof(int))
2974                                 return -EINVAL;
2975
2976                         val = jiffies_to_msecs(*(int *)(table->data));
2977                         if (put_user(val, (int __user *)oldval))
2978                                 return -EFAULT;
2979                         if (put_user(sizeof(int), oldlenp))
2980                                 return -EFAULT;
2981                 }
2982         }
2983         if (newval && newlen) { 
2984                 int new;
2985                 if (newlen != sizeof(int))
2986                         return -EINVAL; 
2987                 if (get_user(new, (int __user *)newval))
2988                         return -EFAULT;
2989                 *(int *)(table->data) = msecs_to_jiffies(new);
2990         }
2991         return 1;
2992 }
2993
2994
2995
2996 #else /* CONFIG_SYSCTL_SYSCALL */
2997
2998
2999 SYSCALL_DEFINE1(sysctl, struct __sysctl_args __user *, args)
3000 {
3001         struct __sysctl_args tmp;
3002         int error;
3003
3004         if (copy_from_user(&tmp, args, sizeof(tmp)))
3005                 return -EFAULT;
3006
3007         error = deprecated_sysctl_warning(&tmp);
3008
3009         /* If no error reading the parameters then just -ENOSYS ... */
3010         if (!error)
3011                 error = -ENOSYS;
3012
3013         return error;
3014 }
3015
3016 int sysctl_data(struct ctl_table *table,
3017                   void __user *oldval, size_t __user *oldlenp,
3018                   void __user *newval, size_t newlen)
3019 {
3020         return -ENOSYS;
3021 }
3022
3023 int sysctl_string(struct ctl_table *table,
3024                   void __user *oldval, size_t __user *oldlenp,
3025                   void __user *newval, size_t newlen)
3026 {
3027         return -ENOSYS;
3028 }
3029
3030 int sysctl_intvec(struct ctl_table *table,
3031                 void __user *oldval, size_t __user *oldlenp,
3032                 void __user *newval, size_t newlen)
3033 {
3034         return -ENOSYS;
3035 }
3036
3037 int sysctl_jiffies(struct ctl_table *table,
3038                 void __user *oldval, size_t __user *oldlenp,
3039                 void __user *newval, size_t newlen)
3040 {
3041         return -ENOSYS;
3042 }
3043
3044 int sysctl_ms_jiffies(struct ctl_table *table,
3045                 void __user *oldval, size_t __user *oldlenp,
3046                 void __user *newval, size_t newlen)
3047 {
3048         return -ENOSYS;
3049 }
3050
3051 #endif /* CONFIG_SYSCTL_SYSCALL */
3052
3053 static int deprecated_sysctl_warning(struct __sysctl_args *args)
3054 {
3055         static int msg_count;
3056         int name[CTL_MAXNAME];
3057         int i;
3058
3059         /* Check args->nlen. */
3060         if (args->nlen < 0 || args->nlen > CTL_MAXNAME)
3061                 return -ENOTDIR;
3062
3063         /* Read in the sysctl name for better debug message logging */
3064         for (i = 0; i < args->nlen; i++)
3065                 if (get_user(name[i], args->name + i))
3066                         return -EFAULT;
3067
3068         /* Ignore accesses to kernel.version */
3069         if ((args->nlen == 2) && (name[0] == CTL_KERN) && (name[1] == KERN_VERSION))
3070                 return 0;
3071
3072         if (msg_count < 5) {
3073                 msg_count++;
3074                 printk(KERN_INFO
3075                         "warning: process `%s' used the deprecated sysctl "
3076                         "system call with ", current->comm);
3077                 for (i = 0; i < args->nlen; i++)
3078                         printk("%d.", name[i]);
3079                 printk("\n");
3080         }
3081         return 0;
3082 }
3083
3084 /*
3085  * No sense putting this after each symbol definition, twice,
3086  * exception granted :-)
3087  */
3088 EXPORT_SYMBOL(proc_dointvec);
3089 EXPORT_SYMBOL(proc_dointvec_jiffies);
3090 EXPORT_SYMBOL(proc_dointvec_minmax);
3091 EXPORT_SYMBOL(proc_dointvec_userhz_jiffies);
3092 EXPORT_SYMBOL(proc_dointvec_ms_jiffies);
3093 EXPORT_SYMBOL(proc_dostring);
3094 EXPORT_SYMBOL(proc_doulongvec_minmax);
3095 EXPORT_SYMBOL(proc_doulongvec_ms_jiffies_minmax);
3096 EXPORT_SYMBOL(register_sysctl_table);
3097 EXPORT_SYMBOL(register_sysctl_paths);
3098 EXPORT_SYMBOL(sysctl_intvec);
3099 EXPORT_SYMBOL(sysctl_jiffies);
3100 EXPORT_SYMBOL(sysctl_ms_jiffies);
3101 EXPORT_SYMBOL(sysctl_string);
3102 EXPORT_SYMBOL(sysctl_data);
3103 EXPORT_SYMBOL(unregister_sysctl_table);