hugetlb: introduce nr_overcommit_hugepages sysctl
[linux-2.6-block.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/security.h>
41 #include <linux/initrd.h>
42 #include <linux/times.h>
43 #include <linux/limits.h>
44 #include <linux/dcache.h>
45 #include <linux/syscalls.h>
46 #include <linux/nfs_fs.h>
47 #include <linux/acpi.h>
48 #include <linux/reboot.h>
49
50 #include <asm/uaccess.h>
51 #include <asm/processor.h>
52
53 #ifdef CONFIG_X86
54 #include <asm/nmi.h>
55 #include <asm/stacktrace.h>
56 #endif
57
58 static int deprecated_sysctl_warning(struct __sysctl_args *args);
59
60 #if defined(CONFIG_SYSCTL)
61
62 /* External variables not in a header file. */
63 extern int C_A_D;
64 extern int print_fatal_signals;
65 extern int sysctl_overcommit_memory;
66 extern int sysctl_overcommit_ratio;
67 extern int sysctl_panic_on_oom;
68 extern int sysctl_oom_kill_allocating_task;
69 extern int max_threads;
70 extern int core_uses_pid;
71 extern int suid_dumpable;
72 extern char core_pattern[];
73 extern int pid_max;
74 extern int min_free_kbytes;
75 extern int printk_ratelimit_jiffies;
76 extern int printk_ratelimit_burst;
77 extern int pid_max_min, pid_max_max;
78 extern int sysctl_drop_caches;
79 extern int percpu_pagelist_fraction;
80 extern int compat_log;
81 extern int maps_protect;
82 extern int sysctl_stat_interval;
83 extern int audit_argv_kb;
84
85 /* Constants used for minimum and  maximum */
86 #ifdef CONFIG_DETECT_SOFTLOCKUP
87 static int one = 1;
88 static int sixty = 60;
89 #endif
90
91 #ifdef CONFIG_MMU
92 static int two = 2;
93 #endif
94
95 static int zero;
96 static int one_hundred = 100;
97
98 /* this is needed for the proc_dointvec_minmax for [fs_]overflow UID and GID */
99 static int maxolduid = 65535;
100 static int minolduid;
101 static int min_percpu_pagelist_fract = 8;
102
103 static int ngroups_max = NGROUPS_MAX;
104
105 #ifdef CONFIG_KMOD
106 extern char modprobe_path[];
107 #endif
108 #ifdef CONFIG_CHR_DEV_SG
109 extern int sg_big_buff;
110 #endif
111
112 #ifdef __sparc__
113 extern char reboot_command [];
114 extern int stop_a_enabled;
115 extern int scons_pwroff;
116 #endif
117
118 #ifdef __hppa__
119 extern int pwrsw_enabled;
120 extern int unaligned_enabled;
121 #endif
122
123 #ifdef CONFIG_S390
124 #ifdef CONFIG_MATHEMU
125 extern int sysctl_ieee_emulation_warnings;
126 #endif
127 extern int sysctl_userprocess_debug;
128 extern int spin_retry;
129 #endif
130
131 extern int sysctl_hz_timer;
132
133 #ifdef CONFIG_BSD_PROCESS_ACCT
134 extern int acct_parm[];
135 #endif
136
137 #ifdef CONFIG_IA64
138 extern int no_unaligned_warning;
139 #endif
140
141 #ifdef CONFIG_RT_MUTEXES
142 extern int max_lock_depth;
143 #endif
144
145 #ifdef CONFIG_SYSCTL_SYSCALL
146 static int parse_table(int __user *, int, void __user *, size_t __user *,
147                 void __user *, size_t, struct ctl_table *);
148 #endif
149
150
151 #ifdef CONFIG_PROC_SYSCTL
152 static int proc_do_cad_pid(struct ctl_table *table, int write, struct file *filp,
153                   void __user *buffer, size_t *lenp, loff_t *ppos);
154 static int proc_dointvec_taint(struct ctl_table *table, int write, struct file *filp,
155                                void __user *buffer, size_t *lenp, loff_t *ppos);
156 #endif
157
158 static struct ctl_table root_table[];
159 static struct ctl_table_header root_table_header =
160         { root_table, LIST_HEAD_INIT(root_table_header.ctl_entry) };
161
162 static struct ctl_table kern_table[];
163 static struct ctl_table vm_table[];
164 static struct ctl_table fs_table[];
165 static struct ctl_table debug_table[];
166 static struct ctl_table dev_table[];
167 extern struct ctl_table random_table[];
168 #ifdef CONFIG_INOTIFY_USER
169 extern struct ctl_table inotify_table[];
170 #endif
171
172 #ifdef HAVE_ARCH_PICK_MMAP_LAYOUT
173 int sysctl_legacy_va_layout;
174 #endif
175
176 extern int prove_locking;
177 extern int lock_stat;
178
179 /* The default sysctl tables: */
180
181 static struct ctl_table root_table[] = {
182         {
183                 .ctl_name       = CTL_KERN,
184                 .procname       = "kernel",
185                 .mode           = 0555,
186                 .child          = kern_table,
187         },
188         {
189                 .ctl_name       = CTL_VM,
190                 .procname       = "vm",
191                 .mode           = 0555,
192                 .child          = vm_table,
193         },
194 #ifdef CONFIG_NET
195         {
196                 .ctl_name       = CTL_NET,
197                 .procname       = "net",
198                 .mode           = 0555,
199                 .child          = net_table,
200         },
201 #endif
202         {
203                 .ctl_name       = CTL_FS,
204                 .procname       = "fs",
205                 .mode           = 0555,
206                 .child          = fs_table,
207         },
208         {
209                 .ctl_name       = CTL_DEBUG,
210                 .procname       = "debug",
211                 .mode           = 0555,
212                 .child          = debug_table,
213         },
214         {
215                 .ctl_name       = CTL_DEV,
216                 .procname       = "dev",
217                 .mode           = 0555,
218                 .child          = dev_table,
219         },
220 /*
221  * NOTE: do not add new entries to this table unless you have read
222  * Documentation/sysctl/ctl_unnumbered.txt
223  */
224         { .ctl_name = 0 }
225 };
226
227 #ifdef CONFIG_SCHED_DEBUG
228 static unsigned long min_sched_granularity_ns = 100000;         /* 100 usecs */
229 static unsigned long max_sched_granularity_ns = NSEC_PER_SEC;   /* 1 second */
230 static unsigned long min_wakeup_granularity_ns;                 /* 0 usecs */
231 static unsigned long max_wakeup_granularity_ns = NSEC_PER_SEC;  /* 1 second */
232 #endif
233
234 static struct ctl_table kern_table[] = {
235 #ifdef CONFIG_SCHED_DEBUG
236         {
237                 .ctl_name       = CTL_UNNUMBERED,
238                 .procname       = "sched_min_granularity_ns",
239                 .data           = &sysctl_sched_min_granularity,
240                 .maxlen         = sizeof(unsigned int),
241                 .mode           = 0644,
242                 .proc_handler   = &sched_nr_latency_handler,
243                 .strategy       = &sysctl_intvec,
244                 .extra1         = &min_sched_granularity_ns,
245                 .extra2         = &max_sched_granularity_ns,
246         },
247         {
248                 .ctl_name       = CTL_UNNUMBERED,
249                 .procname       = "sched_latency_ns",
250                 .data           = &sysctl_sched_latency,
251                 .maxlen         = sizeof(unsigned int),
252                 .mode           = 0644,
253                 .proc_handler   = &sched_nr_latency_handler,
254                 .strategy       = &sysctl_intvec,
255                 .extra1         = &min_sched_granularity_ns,
256                 .extra2         = &max_sched_granularity_ns,
257         },
258         {
259                 .ctl_name       = CTL_UNNUMBERED,
260                 .procname       = "sched_wakeup_granularity_ns",
261                 .data           = &sysctl_sched_wakeup_granularity,
262                 .maxlen         = sizeof(unsigned int),
263                 .mode           = 0644,
264                 .proc_handler   = &proc_dointvec_minmax,
265                 .strategy       = &sysctl_intvec,
266                 .extra1         = &min_wakeup_granularity_ns,
267                 .extra2         = &max_wakeup_granularity_ns,
268         },
269         {
270                 .ctl_name       = CTL_UNNUMBERED,
271                 .procname       = "sched_batch_wakeup_granularity_ns",
272                 .data           = &sysctl_sched_batch_wakeup_granularity,
273                 .maxlen         = sizeof(unsigned int),
274                 .mode           = 0644,
275                 .proc_handler   = &proc_dointvec_minmax,
276                 .strategy       = &sysctl_intvec,
277                 .extra1         = &min_wakeup_granularity_ns,
278                 .extra2         = &max_wakeup_granularity_ns,
279         },
280         {
281                 .ctl_name       = CTL_UNNUMBERED,
282                 .procname       = "sched_child_runs_first",
283                 .data           = &sysctl_sched_child_runs_first,
284                 .maxlen         = sizeof(unsigned int),
285                 .mode           = 0644,
286                 .proc_handler   = &proc_dointvec,
287         },
288         {
289                 .ctl_name       = CTL_UNNUMBERED,
290                 .procname       = "sched_features",
291                 .data           = &sysctl_sched_features,
292                 .maxlen         = sizeof(unsigned int),
293                 .mode           = 0644,
294                 .proc_handler   = &proc_dointvec,
295         },
296         {
297                 .ctl_name       = CTL_UNNUMBERED,
298                 .procname       = "sched_migration_cost",
299                 .data           = &sysctl_sched_migration_cost,
300                 .maxlen         = sizeof(unsigned int),
301                 .mode           = 0644,
302                 .proc_handler   = &proc_dointvec,
303         },
304         {
305                 .ctl_name       = CTL_UNNUMBERED,
306                 .procname       = "sched_nr_migrate",
307                 .data           = &sysctl_sched_nr_migrate,
308                 .maxlen         = sizeof(unsigned int),
309                 .mode           = 644,
310                 .proc_handler   = &proc_dointvec,
311         },
312 #endif
313         {
314                 .ctl_name       = CTL_UNNUMBERED,
315                 .procname       = "sched_compat_yield",
316                 .data           = &sysctl_sched_compat_yield,
317                 .maxlen         = sizeof(unsigned int),
318                 .mode           = 0644,
319                 .proc_handler   = &proc_dointvec,
320         },
321 #ifdef CONFIG_PROVE_LOCKING
322         {
323                 .ctl_name       = CTL_UNNUMBERED,
324                 .procname       = "prove_locking",
325                 .data           = &prove_locking,
326                 .maxlen         = sizeof(int),
327                 .mode           = 0644,
328                 .proc_handler   = &proc_dointvec,
329         },
330 #endif
331 #ifdef CONFIG_LOCK_STAT
332         {
333                 .ctl_name       = CTL_UNNUMBERED,
334                 .procname       = "lock_stat",
335                 .data           = &lock_stat,
336                 .maxlen         = sizeof(int),
337                 .mode           = 0644,
338                 .proc_handler   = &proc_dointvec,
339         },
340 #endif
341         {
342                 .ctl_name       = KERN_PANIC,
343                 .procname       = "panic",
344                 .data           = &panic_timeout,
345                 .maxlen         = sizeof(int),
346                 .mode           = 0644,
347                 .proc_handler   = &proc_dointvec,
348         },
349         {
350                 .ctl_name       = KERN_CORE_USES_PID,
351                 .procname       = "core_uses_pid",
352                 .data           = &core_uses_pid,
353                 .maxlen         = sizeof(int),
354                 .mode           = 0644,
355                 .proc_handler   = &proc_dointvec,
356         },
357 #ifdef CONFIG_AUDITSYSCALL
358         {
359                 .ctl_name       = CTL_UNNUMBERED,
360                 .procname       = "audit_argv_kb",
361                 .data           = &audit_argv_kb,
362                 .maxlen         = sizeof(int),
363                 .mode           = 0644,
364                 .proc_handler   = &proc_dointvec,
365         },
366 #endif
367         {
368                 .ctl_name       = KERN_CORE_PATTERN,
369                 .procname       = "core_pattern",
370                 .data           = core_pattern,
371                 .maxlen         = CORENAME_MAX_SIZE,
372                 .mode           = 0644,
373                 .proc_handler   = &proc_dostring,
374                 .strategy       = &sysctl_string,
375         },
376 #ifdef CONFIG_PROC_SYSCTL
377         {
378                 .procname       = "tainted",
379                 .data           = &tainted,
380                 .maxlen         = sizeof(int),
381                 .mode           = 0644,
382                 .proc_handler   = &proc_dointvec_taint,
383         },
384 #endif
385 #ifdef CONFIG_SECURITY_CAPABILITIES
386         {
387                 .procname       = "cap-bound",
388                 .data           = &cap_bset,
389                 .maxlen         = sizeof(kernel_cap_t),
390                 .mode           = 0600,
391                 .proc_handler   = &proc_dointvec_bset,
392         },
393 #endif /* def CONFIG_SECURITY_CAPABILITIES */
394 #ifdef CONFIG_BLK_DEV_INITRD
395         {
396                 .ctl_name       = KERN_REALROOTDEV,
397                 .procname       = "real-root-dev",
398                 .data           = &real_root_dev,
399                 .maxlen         = sizeof(int),
400                 .mode           = 0644,
401                 .proc_handler   = &proc_dointvec,
402         },
403 #endif
404         {
405                 .ctl_name       = CTL_UNNUMBERED,
406                 .procname       = "print-fatal-signals",
407                 .data           = &print_fatal_signals,
408                 .maxlen         = sizeof(int),
409                 .mode           = 0644,
410                 .proc_handler   = &proc_dointvec,
411         },
412 #ifdef __sparc__
413         {
414                 .ctl_name       = KERN_SPARC_REBOOT,
415                 .procname       = "reboot-cmd",
416                 .data           = reboot_command,
417                 .maxlen         = 256,
418                 .mode           = 0644,
419                 .proc_handler   = &proc_dostring,
420                 .strategy       = &sysctl_string,
421         },
422         {
423                 .ctl_name       = KERN_SPARC_STOP_A,
424                 .procname       = "stop-a",
425                 .data           = &stop_a_enabled,
426                 .maxlen         = sizeof (int),
427                 .mode           = 0644,
428                 .proc_handler   = &proc_dointvec,
429         },
430         {
431                 .ctl_name       = KERN_SPARC_SCONS_PWROFF,
432                 .procname       = "scons-poweroff",
433                 .data           = &scons_pwroff,
434                 .maxlen         = sizeof (int),
435                 .mode           = 0644,
436                 .proc_handler   = &proc_dointvec,
437         },
438 #endif
439 #ifdef __hppa__
440         {
441                 .ctl_name       = KERN_HPPA_PWRSW,
442                 .procname       = "soft-power",
443                 .data           = &pwrsw_enabled,
444                 .maxlen         = sizeof (int),
445                 .mode           = 0644,
446                 .proc_handler   = &proc_dointvec,
447         },
448         {
449                 .ctl_name       = KERN_HPPA_UNALIGNED,
450                 .procname       = "unaligned-trap",
451                 .data           = &unaligned_enabled,
452                 .maxlen         = sizeof (int),
453                 .mode           = 0644,
454                 .proc_handler   = &proc_dointvec,
455         },
456 #endif
457         {
458                 .ctl_name       = KERN_CTLALTDEL,
459                 .procname       = "ctrl-alt-del",
460                 .data           = &C_A_D,
461                 .maxlen         = sizeof(int),
462                 .mode           = 0644,
463                 .proc_handler   = &proc_dointvec,
464         },
465         {
466                 .ctl_name       = KERN_PRINTK,
467                 .procname       = "printk",
468                 .data           = &console_loglevel,
469                 .maxlen         = 4*sizeof(int),
470                 .mode           = 0644,
471                 .proc_handler   = &proc_dointvec,
472         },
473 #ifdef CONFIG_KMOD
474         {
475                 .ctl_name       = KERN_MODPROBE,
476                 .procname       = "modprobe",
477                 .data           = &modprobe_path,
478                 .maxlen         = KMOD_PATH_LEN,
479                 .mode           = 0644,
480                 .proc_handler   = &proc_dostring,
481                 .strategy       = &sysctl_string,
482         },
483 #endif
484 #if defined(CONFIG_HOTPLUG) && defined(CONFIG_NET)
485         {
486                 .ctl_name       = KERN_HOTPLUG,
487                 .procname       = "hotplug",
488                 .data           = &uevent_helper,
489                 .maxlen         = UEVENT_HELPER_PATH_LEN,
490                 .mode           = 0644,
491                 .proc_handler   = &proc_dostring,
492                 .strategy       = &sysctl_string,
493         },
494 #endif
495 #ifdef CONFIG_CHR_DEV_SG
496         {
497                 .ctl_name       = KERN_SG_BIG_BUFF,
498                 .procname       = "sg-big-buff",
499                 .data           = &sg_big_buff,
500                 .maxlen         = sizeof (int),
501                 .mode           = 0444,
502                 .proc_handler   = &proc_dointvec,
503         },
504 #endif
505 #ifdef CONFIG_BSD_PROCESS_ACCT
506         {
507                 .ctl_name       = KERN_ACCT,
508                 .procname       = "acct",
509                 .data           = &acct_parm,
510                 .maxlen         = 3*sizeof(int),
511                 .mode           = 0644,
512                 .proc_handler   = &proc_dointvec,
513         },
514 #endif
515 #ifdef CONFIG_MAGIC_SYSRQ
516         {
517                 .ctl_name       = KERN_SYSRQ,
518                 .procname       = "sysrq",
519                 .data           = &__sysrq_enabled,
520                 .maxlen         = sizeof (int),
521                 .mode           = 0644,
522                 .proc_handler   = &proc_dointvec,
523         },
524 #endif
525 #ifdef CONFIG_PROC_SYSCTL
526         {
527                 .procname       = "cad_pid",
528                 .data           = NULL,
529                 .maxlen         = sizeof (int),
530                 .mode           = 0600,
531                 .proc_handler   = &proc_do_cad_pid,
532         },
533 #endif
534         {
535                 .ctl_name       = KERN_MAX_THREADS,
536                 .procname       = "threads-max",
537                 .data           = &max_threads,
538                 .maxlen         = sizeof(int),
539                 .mode           = 0644,
540                 .proc_handler   = &proc_dointvec,
541         },
542         {
543                 .ctl_name       = KERN_RANDOM,
544                 .procname       = "random",
545                 .mode           = 0555,
546                 .child          = random_table,
547         },
548         {
549                 .ctl_name       = KERN_OVERFLOWUID,
550                 .procname       = "overflowuid",
551                 .data           = &overflowuid,
552                 .maxlen         = sizeof(int),
553                 .mode           = 0644,
554                 .proc_handler   = &proc_dointvec_minmax,
555                 .strategy       = &sysctl_intvec,
556                 .extra1         = &minolduid,
557                 .extra2         = &maxolduid,
558         },
559         {
560                 .ctl_name       = KERN_OVERFLOWGID,
561                 .procname       = "overflowgid",
562                 .data           = &overflowgid,
563                 .maxlen         = sizeof(int),
564                 .mode           = 0644,
565                 .proc_handler   = &proc_dointvec_minmax,
566                 .strategy       = &sysctl_intvec,
567                 .extra1         = &minolduid,
568                 .extra2         = &maxolduid,
569         },
570 #ifdef CONFIG_S390
571 #ifdef CONFIG_MATHEMU
572         {
573                 .ctl_name       = KERN_IEEE_EMULATION_WARNINGS,
574                 .procname       = "ieee_emulation_warnings",
575                 .data           = &sysctl_ieee_emulation_warnings,
576                 .maxlen         = sizeof(int),
577                 .mode           = 0644,
578                 .proc_handler   = &proc_dointvec,
579         },
580 #endif
581 #ifdef CONFIG_NO_IDLE_HZ
582         {
583                 .ctl_name       = KERN_HZ_TIMER,
584                 .procname       = "hz_timer",
585                 .data           = &sysctl_hz_timer,
586                 .maxlen         = sizeof(int),
587                 .mode           = 0644,
588                 .proc_handler   = &proc_dointvec,
589         },
590 #endif
591         {
592                 .ctl_name       = KERN_S390_USER_DEBUG_LOGGING,
593                 .procname       = "userprocess_debug",
594                 .data           = &sysctl_userprocess_debug,
595                 .maxlen         = sizeof(int),
596                 .mode           = 0644,
597                 .proc_handler   = &proc_dointvec,
598         },
599 #endif
600         {
601                 .ctl_name       = KERN_PIDMAX,
602                 .procname       = "pid_max",
603                 .data           = &pid_max,
604                 .maxlen         = sizeof (int),
605                 .mode           = 0644,
606                 .proc_handler   = &proc_dointvec_minmax,
607                 .strategy       = sysctl_intvec,
608                 .extra1         = &pid_max_min,
609                 .extra2         = &pid_max_max,
610         },
611         {
612                 .ctl_name       = KERN_PANIC_ON_OOPS,
613                 .procname       = "panic_on_oops",
614                 .data           = &panic_on_oops,
615                 .maxlen         = sizeof(int),
616                 .mode           = 0644,
617                 .proc_handler   = &proc_dointvec,
618         },
619         {
620                 .ctl_name       = KERN_PRINTK_RATELIMIT,
621                 .procname       = "printk_ratelimit",
622                 .data           = &printk_ratelimit_jiffies,
623                 .maxlen         = sizeof(int),
624                 .mode           = 0644,
625                 .proc_handler   = &proc_dointvec_jiffies,
626                 .strategy       = &sysctl_jiffies,
627         },
628         {
629                 .ctl_name       = KERN_PRINTK_RATELIMIT_BURST,
630                 .procname       = "printk_ratelimit_burst",
631                 .data           = &printk_ratelimit_burst,
632                 .maxlen         = sizeof(int),
633                 .mode           = 0644,
634                 .proc_handler   = &proc_dointvec,
635         },
636         {
637                 .ctl_name       = KERN_NGROUPS_MAX,
638                 .procname       = "ngroups_max",
639                 .data           = &ngroups_max,
640                 .maxlen         = sizeof (int),
641                 .mode           = 0444,
642                 .proc_handler   = &proc_dointvec,
643         },
644 #if defined(CONFIG_X86_LOCAL_APIC) && defined(CONFIG_X86)
645         {
646                 .ctl_name       = KERN_UNKNOWN_NMI_PANIC,
647                 .procname       = "unknown_nmi_panic",
648                 .data           = &unknown_nmi_panic,
649                 .maxlen         = sizeof (int),
650                 .mode           = 0644,
651                 .proc_handler   = &proc_dointvec,
652         },
653         {
654                 .procname       = "nmi_watchdog",
655                 .data           = &nmi_watchdog_enabled,
656                 .maxlen         = sizeof (int),
657                 .mode           = 0644,
658                 .proc_handler   = &proc_nmi_enabled,
659         },
660 #endif
661 #if defined(CONFIG_X86)
662         {
663                 .ctl_name       = KERN_PANIC_ON_NMI,
664                 .procname       = "panic_on_unrecovered_nmi",
665                 .data           = &panic_on_unrecovered_nmi,
666                 .maxlen         = sizeof(int),
667                 .mode           = 0644,
668                 .proc_handler   = &proc_dointvec,
669         },
670         {
671                 .ctl_name       = KERN_BOOTLOADER_TYPE,
672                 .procname       = "bootloader_type",
673                 .data           = &bootloader_type,
674                 .maxlen         = sizeof (int),
675                 .mode           = 0444,
676                 .proc_handler   = &proc_dointvec,
677         },
678         {
679                 .ctl_name       = CTL_UNNUMBERED,
680                 .procname       = "kstack_depth_to_print",
681                 .data           = &kstack_depth_to_print,
682                 .maxlen         = sizeof(int),
683                 .mode           = 0644,
684                 .proc_handler   = &proc_dointvec,
685         },
686 #endif
687 #if defined(CONFIG_MMU)
688         {
689                 .ctl_name       = KERN_RANDOMIZE,
690                 .procname       = "randomize_va_space",
691                 .data           = &randomize_va_space,
692                 .maxlen         = sizeof(int),
693                 .mode           = 0644,
694                 .proc_handler   = &proc_dointvec,
695         },
696 #endif
697 #if defined(CONFIG_S390) && defined(CONFIG_SMP)
698         {
699                 .ctl_name       = KERN_SPIN_RETRY,
700                 .procname       = "spin_retry",
701                 .data           = &spin_retry,
702                 .maxlen         = sizeof (int),
703                 .mode           = 0644,
704                 .proc_handler   = &proc_dointvec,
705         },
706 #endif
707 #if     defined(CONFIG_ACPI_SLEEP) && defined(CONFIG_X86)
708         {
709                 .procname       = "acpi_video_flags",
710                 .data           = &acpi_realmode_flags,
711                 .maxlen         = sizeof (unsigned long),
712                 .mode           = 0644,
713                 .proc_handler   = &proc_doulongvec_minmax,
714         },
715 #endif
716 #ifdef CONFIG_IA64
717         {
718                 .ctl_name       = KERN_IA64_UNALIGNED,
719                 .procname       = "ignore-unaligned-usertrap",
720                 .data           = &no_unaligned_warning,
721                 .maxlen         = sizeof (int),
722                 .mode           = 0644,
723                 .proc_handler   = &proc_dointvec,
724         },
725 #endif
726 #ifdef CONFIG_DETECT_SOFTLOCKUP
727         {
728                 .ctl_name       = CTL_UNNUMBERED,
729                 .procname       = "softlockup_thresh",
730                 .data           = &softlockup_thresh,
731                 .maxlen         = sizeof(int),
732                 .mode           = 0644,
733                 .proc_handler   = &proc_dointvec_minmax,
734                 .strategy       = &sysctl_intvec,
735                 .extra1         = &one,
736                 .extra2         = &sixty,
737         },
738 #endif
739 #ifdef CONFIG_COMPAT
740         {
741                 .ctl_name       = KERN_COMPAT_LOG,
742                 .procname       = "compat-log",
743                 .data           = &compat_log,
744                 .maxlen         = sizeof (int),
745                 .mode           = 0644,
746                 .proc_handler   = &proc_dointvec,
747         },
748 #endif
749 #ifdef CONFIG_RT_MUTEXES
750         {
751                 .ctl_name       = KERN_MAX_LOCK_DEPTH,
752                 .procname       = "max_lock_depth",
753                 .data           = &max_lock_depth,
754                 .maxlen         = sizeof(int),
755                 .mode           = 0644,
756                 .proc_handler   = &proc_dointvec,
757         },
758 #endif
759 #ifdef CONFIG_PROC_FS
760         {
761                 .ctl_name       = CTL_UNNUMBERED,
762                 .procname       = "maps_protect",
763                 .data           = &maps_protect,
764                 .maxlen         = sizeof(int),
765                 .mode           = 0644,
766                 .proc_handler   = &proc_dointvec,
767         },
768 #endif
769         {
770                 .ctl_name       = CTL_UNNUMBERED,
771                 .procname       = "poweroff_cmd",
772                 .data           = &poweroff_cmd,
773                 .maxlen         = POWEROFF_CMD_PATH_LEN,
774                 .mode           = 0644,
775                 .proc_handler   = &proc_dostring,
776                 .strategy       = &sysctl_string,
777         },
778 /*
779  * NOTE: do not add new entries to this table unless you have read
780  * Documentation/sysctl/ctl_unnumbered.txt
781  */
782         { .ctl_name = 0 }
783 };
784
785 static struct ctl_table vm_table[] = {
786         {
787                 .ctl_name       = VM_OVERCOMMIT_MEMORY,
788                 .procname       = "overcommit_memory",
789                 .data           = &sysctl_overcommit_memory,
790                 .maxlen         = sizeof(sysctl_overcommit_memory),
791                 .mode           = 0644,
792                 .proc_handler   = &proc_dointvec,
793         },
794         {
795                 .ctl_name       = VM_PANIC_ON_OOM,
796                 .procname       = "panic_on_oom",
797                 .data           = &sysctl_panic_on_oom,
798                 .maxlen         = sizeof(sysctl_panic_on_oom),
799                 .mode           = 0644,
800                 .proc_handler   = &proc_dointvec,
801         },
802         {
803                 .ctl_name       = CTL_UNNUMBERED,
804                 .procname       = "oom_kill_allocating_task",
805                 .data           = &sysctl_oom_kill_allocating_task,
806                 .maxlen         = sizeof(sysctl_oom_kill_allocating_task),
807                 .mode           = 0644,
808                 .proc_handler   = &proc_dointvec,
809         },
810         {
811                 .ctl_name       = VM_OVERCOMMIT_RATIO,
812                 .procname       = "overcommit_ratio",
813                 .data           = &sysctl_overcommit_ratio,
814                 .maxlen         = sizeof(sysctl_overcommit_ratio),
815                 .mode           = 0644,
816                 .proc_handler   = &proc_dointvec,
817         },
818         {
819                 .ctl_name       = VM_PAGE_CLUSTER,
820                 .procname       = "page-cluster", 
821                 .data           = &page_cluster,
822                 .maxlen         = sizeof(int),
823                 .mode           = 0644,
824                 .proc_handler   = &proc_dointvec,
825         },
826         {
827                 .ctl_name       = VM_DIRTY_BACKGROUND,
828                 .procname       = "dirty_background_ratio",
829                 .data           = &dirty_background_ratio,
830                 .maxlen         = sizeof(dirty_background_ratio),
831                 .mode           = 0644,
832                 .proc_handler   = &proc_dointvec_minmax,
833                 .strategy       = &sysctl_intvec,
834                 .extra1         = &zero,
835                 .extra2         = &one_hundred,
836         },
837         {
838                 .ctl_name       = VM_DIRTY_RATIO,
839                 .procname       = "dirty_ratio",
840                 .data           = &vm_dirty_ratio,
841                 .maxlen         = sizeof(vm_dirty_ratio),
842                 .mode           = 0644,
843                 .proc_handler   = &dirty_ratio_handler,
844                 .strategy       = &sysctl_intvec,
845                 .extra1         = &zero,
846                 .extra2         = &one_hundred,
847         },
848         {
849                 .procname       = "dirty_writeback_centisecs",
850                 .data           = &dirty_writeback_interval,
851                 .maxlen         = sizeof(dirty_writeback_interval),
852                 .mode           = 0644,
853                 .proc_handler   = &dirty_writeback_centisecs_handler,
854         },
855         {
856                 .procname       = "dirty_expire_centisecs",
857                 .data           = &dirty_expire_interval,
858                 .maxlen         = sizeof(dirty_expire_interval),
859                 .mode           = 0644,
860                 .proc_handler   = &proc_dointvec_userhz_jiffies,
861         },
862         {
863                 .ctl_name       = VM_NR_PDFLUSH_THREADS,
864                 .procname       = "nr_pdflush_threads",
865                 .data           = &nr_pdflush_threads,
866                 .maxlen         = sizeof nr_pdflush_threads,
867                 .mode           = 0444 /* read-only*/,
868                 .proc_handler   = &proc_dointvec,
869         },
870         {
871                 .ctl_name       = VM_SWAPPINESS,
872                 .procname       = "swappiness",
873                 .data           = &vm_swappiness,
874                 .maxlen         = sizeof(vm_swappiness),
875                 .mode           = 0644,
876                 .proc_handler   = &proc_dointvec_minmax,
877                 .strategy       = &sysctl_intvec,
878                 .extra1         = &zero,
879                 .extra2         = &one_hundred,
880         },
881 #ifdef CONFIG_HUGETLB_PAGE
882          {
883                 .procname       = "nr_hugepages",
884                 .data           = &max_huge_pages,
885                 .maxlen         = sizeof(unsigned long),
886                 .mode           = 0644,
887                 .proc_handler   = &hugetlb_sysctl_handler,
888                 .extra1         = (void *)&hugetlb_zero,
889                 .extra2         = (void *)&hugetlb_infinity,
890          },
891          {
892                 .ctl_name       = VM_HUGETLB_GROUP,
893                 .procname       = "hugetlb_shm_group",
894                 .data           = &sysctl_hugetlb_shm_group,
895                 .maxlen         = sizeof(gid_t),
896                 .mode           = 0644,
897                 .proc_handler   = &proc_dointvec,
898          },
899          {
900                 .ctl_name       = CTL_UNNUMBERED,
901                 .procname       = "hugepages_treat_as_movable",
902                 .data           = &hugepages_treat_as_movable,
903                 .maxlen         = sizeof(int),
904                 .mode           = 0644,
905                 .proc_handler   = &hugetlb_treat_movable_handler,
906         },
907         {
908                 .ctl_name       = CTL_UNNUMBERED,
909                 .procname       = "hugetlb_dynamic_pool",
910                 .data           = &hugetlb_dynamic_pool,
911                 .maxlen         = sizeof(hugetlb_dynamic_pool),
912                 .mode           = 0644,
913                 .proc_handler   = &proc_dointvec,
914         },
915         {
916                 .ctl_name       = CTL_UNNUMBERED,
917                 .procname       = "nr_overcommit_hugepages",
918                 .data           = &nr_overcommit_huge_pages,
919                 .maxlen         = sizeof(nr_overcommit_huge_pages),
920                 .mode           = 0644,
921                 .proc_handler   = &proc_doulongvec_minmax,
922         },
923 #endif
924         {
925                 .ctl_name       = VM_LOWMEM_RESERVE_RATIO,
926                 .procname       = "lowmem_reserve_ratio",
927                 .data           = &sysctl_lowmem_reserve_ratio,
928                 .maxlen         = sizeof(sysctl_lowmem_reserve_ratio),
929                 .mode           = 0644,
930                 .proc_handler   = &lowmem_reserve_ratio_sysctl_handler,
931                 .strategy       = &sysctl_intvec,
932         },
933         {
934                 .ctl_name       = VM_DROP_PAGECACHE,
935                 .procname       = "drop_caches",
936                 .data           = &sysctl_drop_caches,
937                 .maxlen         = sizeof(int),
938                 .mode           = 0644,
939                 .proc_handler   = drop_caches_sysctl_handler,
940                 .strategy       = &sysctl_intvec,
941         },
942         {
943                 .ctl_name       = VM_MIN_FREE_KBYTES,
944                 .procname       = "min_free_kbytes",
945                 .data           = &min_free_kbytes,
946                 .maxlen         = sizeof(min_free_kbytes),
947                 .mode           = 0644,
948                 .proc_handler   = &min_free_kbytes_sysctl_handler,
949                 .strategy       = &sysctl_intvec,
950                 .extra1         = &zero,
951         },
952         {
953                 .ctl_name       = VM_PERCPU_PAGELIST_FRACTION,
954                 .procname       = "percpu_pagelist_fraction",
955                 .data           = &percpu_pagelist_fraction,
956                 .maxlen         = sizeof(percpu_pagelist_fraction),
957                 .mode           = 0644,
958                 .proc_handler   = &percpu_pagelist_fraction_sysctl_handler,
959                 .strategy       = &sysctl_intvec,
960                 .extra1         = &min_percpu_pagelist_fract,
961         },
962 #ifdef CONFIG_MMU
963         {
964                 .ctl_name       = VM_MAX_MAP_COUNT,
965                 .procname       = "max_map_count",
966                 .data           = &sysctl_max_map_count,
967                 .maxlen         = sizeof(sysctl_max_map_count),
968                 .mode           = 0644,
969                 .proc_handler   = &proc_dointvec
970         },
971 #endif
972         {
973                 .ctl_name       = VM_LAPTOP_MODE,
974                 .procname       = "laptop_mode",
975                 .data           = &laptop_mode,
976                 .maxlen         = sizeof(laptop_mode),
977                 .mode           = 0644,
978                 .proc_handler   = &proc_dointvec_jiffies,
979                 .strategy       = &sysctl_jiffies,
980         },
981         {
982                 .ctl_name       = VM_BLOCK_DUMP,
983                 .procname       = "block_dump",
984                 .data           = &block_dump,
985                 .maxlen         = sizeof(block_dump),
986                 .mode           = 0644,
987                 .proc_handler   = &proc_dointvec,
988                 .strategy       = &sysctl_intvec,
989                 .extra1         = &zero,
990         },
991         {
992                 .ctl_name       = VM_VFS_CACHE_PRESSURE,
993                 .procname       = "vfs_cache_pressure",
994                 .data           = &sysctl_vfs_cache_pressure,
995                 .maxlen         = sizeof(sysctl_vfs_cache_pressure),
996                 .mode           = 0644,
997                 .proc_handler   = &proc_dointvec,
998                 .strategy       = &sysctl_intvec,
999                 .extra1         = &zero,
1000         },
1001 #ifdef HAVE_ARCH_PICK_MMAP_LAYOUT
1002         {
1003                 .ctl_name       = VM_LEGACY_VA_LAYOUT,
1004                 .procname       = "legacy_va_layout",
1005                 .data           = &sysctl_legacy_va_layout,
1006                 .maxlen         = sizeof(sysctl_legacy_va_layout),
1007                 .mode           = 0644,
1008                 .proc_handler   = &proc_dointvec,
1009                 .strategy       = &sysctl_intvec,
1010                 .extra1         = &zero,
1011         },
1012 #endif
1013 #ifdef CONFIG_NUMA
1014         {
1015                 .ctl_name       = VM_ZONE_RECLAIM_MODE,
1016                 .procname       = "zone_reclaim_mode",
1017                 .data           = &zone_reclaim_mode,
1018                 .maxlen         = sizeof(zone_reclaim_mode),
1019                 .mode           = 0644,
1020                 .proc_handler   = &proc_dointvec,
1021                 .strategy       = &sysctl_intvec,
1022                 .extra1         = &zero,
1023         },
1024         {
1025                 .ctl_name       = VM_MIN_UNMAPPED,
1026                 .procname       = "min_unmapped_ratio",
1027                 .data           = &sysctl_min_unmapped_ratio,
1028                 .maxlen         = sizeof(sysctl_min_unmapped_ratio),
1029                 .mode           = 0644,
1030                 .proc_handler   = &sysctl_min_unmapped_ratio_sysctl_handler,
1031                 .strategy       = &sysctl_intvec,
1032                 .extra1         = &zero,
1033                 .extra2         = &one_hundred,
1034         },
1035         {
1036                 .ctl_name       = VM_MIN_SLAB,
1037                 .procname       = "min_slab_ratio",
1038                 .data           = &sysctl_min_slab_ratio,
1039                 .maxlen         = sizeof(sysctl_min_slab_ratio),
1040                 .mode           = 0644,
1041                 .proc_handler   = &sysctl_min_slab_ratio_sysctl_handler,
1042                 .strategy       = &sysctl_intvec,
1043                 .extra1         = &zero,
1044                 .extra2         = &one_hundred,
1045         },
1046 #endif
1047 #ifdef CONFIG_SMP
1048         {
1049                 .ctl_name       = CTL_UNNUMBERED,
1050                 .procname       = "stat_interval",
1051                 .data           = &sysctl_stat_interval,
1052                 .maxlen         = sizeof(sysctl_stat_interval),
1053                 .mode           = 0644,
1054                 .proc_handler   = &proc_dointvec_jiffies,
1055                 .strategy       = &sysctl_jiffies,
1056         },
1057 #endif
1058 #ifdef CONFIG_SECURITY
1059         {
1060                 .ctl_name       = CTL_UNNUMBERED,
1061                 .procname       = "mmap_min_addr",
1062                 .data           = &mmap_min_addr,
1063                 .maxlen         = sizeof(unsigned long),
1064                 .mode           = 0644,
1065                 .proc_handler   = &proc_doulongvec_minmax,
1066         },
1067 #endif
1068 #ifdef CONFIG_NUMA
1069         {
1070                 .ctl_name       = CTL_UNNUMBERED,
1071                 .procname       = "numa_zonelist_order",
1072                 .data           = &numa_zonelist_order,
1073                 .maxlen         = NUMA_ZONELIST_ORDER_LEN,
1074                 .mode           = 0644,
1075                 .proc_handler   = &numa_zonelist_order_handler,
1076                 .strategy       = &sysctl_string,
1077         },
1078 #endif
1079 #if (defined(CONFIG_X86_32) && !defined(CONFIG_UML))|| \
1080    (defined(CONFIG_SUPERH) && defined(CONFIG_VSYSCALL))
1081         {
1082                 .ctl_name       = VM_VDSO_ENABLED,
1083                 .procname       = "vdso_enabled",
1084                 .data           = &vdso_enabled,
1085                 .maxlen         = sizeof(vdso_enabled),
1086                 .mode           = 0644,
1087                 .proc_handler   = &proc_dointvec,
1088                 .strategy       = &sysctl_intvec,
1089                 .extra1         = &zero,
1090         },
1091 #endif
1092 /*
1093  * NOTE: do not add new entries to this table unless you have read
1094  * Documentation/sysctl/ctl_unnumbered.txt
1095  */
1096         { .ctl_name = 0 }
1097 };
1098
1099 #if defined(CONFIG_BINFMT_MISC) || defined(CONFIG_BINFMT_MISC_MODULE)
1100 static struct ctl_table binfmt_misc_table[] = {
1101         { .ctl_name = 0 }
1102 };
1103 #endif
1104
1105 static struct ctl_table fs_table[] = {
1106         {
1107                 .ctl_name       = FS_NRINODE,
1108                 .procname       = "inode-nr",
1109                 .data           = &inodes_stat,
1110                 .maxlen         = 2*sizeof(int),
1111                 .mode           = 0444,
1112                 .proc_handler   = &proc_dointvec,
1113         },
1114         {
1115                 .ctl_name       = FS_STATINODE,
1116                 .procname       = "inode-state",
1117                 .data           = &inodes_stat,
1118                 .maxlen         = 7*sizeof(int),
1119                 .mode           = 0444,
1120                 .proc_handler   = &proc_dointvec,
1121         },
1122         {
1123                 .procname       = "file-nr",
1124                 .data           = &files_stat,
1125                 .maxlen         = 3*sizeof(int),
1126                 .mode           = 0444,
1127                 .proc_handler   = &proc_nr_files,
1128         },
1129         {
1130                 .ctl_name       = FS_MAXFILE,
1131                 .procname       = "file-max",
1132                 .data           = &files_stat.max_files,
1133                 .maxlen         = sizeof(int),
1134                 .mode           = 0644,
1135                 .proc_handler   = &proc_dointvec,
1136         },
1137         {
1138                 .ctl_name       = FS_DENTRY,
1139                 .procname       = "dentry-state",
1140                 .data           = &dentry_stat,
1141                 .maxlen         = 6*sizeof(int),
1142                 .mode           = 0444,
1143                 .proc_handler   = &proc_dointvec,
1144         },
1145         {
1146                 .ctl_name       = FS_OVERFLOWUID,
1147                 .procname       = "overflowuid",
1148                 .data           = &fs_overflowuid,
1149                 .maxlen         = sizeof(int),
1150                 .mode           = 0644,
1151                 .proc_handler   = &proc_dointvec_minmax,
1152                 .strategy       = &sysctl_intvec,
1153                 .extra1         = &minolduid,
1154                 .extra2         = &maxolduid,
1155         },
1156         {
1157                 .ctl_name       = FS_OVERFLOWGID,
1158                 .procname       = "overflowgid",
1159                 .data           = &fs_overflowgid,
1160                 .maxlen         = sizeof(int),
1161                 .mode           = 0644,
1162                 .proc_handler   = &proc_dointvec_minmax,
1163                 .strategy       = &sysctl_intvec,
1164                 .extra1         = &minolduid,
1165                 .extra2         = &maxolduid,
1166         },
1167         {
1168                 .ctl_name       = FS_LEASES,
1169                 .procname       = "leases-enable",
1170                 .data           = &leases_enable,
1171                 .maxlen         = sizeof(int),
1172                 .mode           = 0644,
1173                 .proc_handler   = &proc_dointvec,
1174         },
1175 #ifdef CONFIG_DNOTIFY
1176         {
1177                 .ctl_name       = FS_DIR_NOTIFY,
1178                 .procname       = "dir-notify-enable",
1179                 .data           = &dir_notify_enable,
1180                 .maxlen         = sizeof(int),
1181                 .mode           = 0644,
1182                 .proc_handler   = &proc_dointvec,
1183         },
1184 #endif
1185 #ifdef CONFIG_MMU
1186         {
1187                 .ctl_name       = FS_LEASE_TIME,
1188                 .procname       = "lease-break-time",
1189                 .data           = &lease_break_time,
1190                 .maxlen         = sizeof(int),
1191                 .mode           = 0644,
1192                 .proc_handler   = &proc_dointvec_minmax,
1193                 .strategy       = &sysctl_intvec,
1194                 .extra1         = &zero,
1195                 .extra2         = &two,
1196         },
1197         {
1198                 .procname       = "aio-nr",
1199                 .data           = &aio_nr,
1200                 .maxlen         = sizeof(aio_nr),
1201                 .mode           = 0444,
1202                 .proc_handler   = &proc_doulongvec_minmax,
1203         },
1204         {
1205                 .procname       = "aio-max-nr",
1206                 .data           = &aio_max_nr,
1207                 .maxlen         = sizeof(aio_max_nr),
1208                 .mode           = 0644,
1209                 .proc_handler   = &proc_doulongvec_minmax,
1210         },
1211 #ifdef CONFIG_INOTIFY_USER
1212         {
1213                 .ctl_name       = FS_INOTIFY,
1214                 .procname       = "inotify",
1215                 .mode           = 0555,
1216                 .child          = inotify_table,
1217         },
1218 #endif  
1219 #endif
1220         {
1221                 .ctl_name       = KERN_SETUID_DUMPABLE,
1222                 .procname       = "suid_dumpable",
1223                 .data           = &suid_dumpable,
1224                 .maxlen         = sizeof(int),
1225                 .mode           = 0644,
1226                 .proc_handler   = &proc_dointvec,
1227         },
1228 #if defined(CONFIG_BINFMT_MISC) || defined(CONFIG_BINFMT_MISC_MODULE)
1229         {
1230                 .ctl_name       = CTL_UNNUMBERED,
1231                 .procname       = "binfmt_misc",
1232                 .mode           = 0555,
1233                 .child          = binfmt_misc_table,
1234         },
1235 #endif
1236 /*
1237  * NOTE: do not add new entries to this table unless you have read
1238  * Documentation/sysctl/ctl_unnumbered.txt
1239  */
1240         { .ctl_name = 0 }
1241 };
1242
1243 static struct ctl_table debug_table[] = {
1244 #if defined(CONFIG_X86) || defined(CONFIG_PPC)
1245         {
1246                 .ctl_name       = CTL_UNNUMBERED,
1247                 .procname       = "exception-trace",
1248                 .data           = &show_unhandled_signals,
1249                 .maxlen         = sizeof(int),
1250                 .mode           = 0644,
1251                 .proc_handler   = proc_dointvec
1252         },
1253 #endif
1254         { .ctl_name = 0 }
1255 };
1256
1257 static struct ctl_table dev_table[] = {
1258         { .ctl_name = 0 }
1259 };
1260
1261 static DEFINE_SPINLOCK(sysctl_lock);
1262
1263 /* called under sysctl_lock */
1264 static int use_table(struct ctl_table_header *p)
1265 {
1266         if (unlikely(p->unregistering))
1267                 return 0;
1268         p->used++;
1269         return 1;
1270 }
1271
1272 /* called under sysctl_lock */
1273 static void unuse_table(struct ctl_table_header *p)
1274 {
1275         if (!--p->used)
1276                 if (unlikely(p->unregistering))
1277                         complete(p->unregistering);
1278 }
1279
1280 /* called under sysctl_lock, will reacquire if has to wait */
1281 static void start_unregistering(struct ctl_table_header *p)
1282 {
1283         /*
1284          * if p->used is 0, nobody will ever touch that entry again;
1285          * we'll eliminate all paths to it before dropping sysctl_lock
1286          */
1287         if (unlikely(p->used)) {
1288                 struct completion wait;
1289                 init_completion(&wait);
1290                 p->unregistering = &wait;
1291                 spin_unlock(&sysctl_lock);
1292                 wait_for_completion(&wait);
1293                 spin_lock(&sysctl_lock);
1294         }
1295         /*
1296          * do not remove from the list until nobody holds it; walking the
1297          * list in do_sysctl() relies on that.
1298          */
1299         list_del_init(&p->ctl_entry);
1300 }
1301
1302 void sysctl_head_finish(struct ctl_table_header *head)
1303 {
1304         if (!head)
1305                 return;
1306         spin_lock(&sysctl_lock);
1307         unuse_table(head);
1308         spin_unlock(&sysctl_lock);
1309 }
1310
1311 struct ctl_table_header *sysctl_head_next(struct ctl_table_header *prev)
1312 {
1313         struct ctl_table_header *head;
1314         struct list_head *tmp;
1315         spin_lock(&sysctl_lock);
1316         if (prev) {
1317                 tmp = &prev->ctl_entry;
1318                 unuse_table(prev);
1319                 goto next;
1320         }
1321         tmp = &root_table_header.ctl_entry;
1322         for (;;) {
1323                 head = list_entry(tmp, struct ctl_table_header, ctl_entry);
1324
1325                 if (!use_table(head))
1326                         goto next;
1327                 spin_unlock(&sysctl_lock);
1328                 return head;
1329         next:
1330                 tmp = tmp->next;
1331                 if (tmp == &root_table_header.ctl_entry)
1332                         break;
1333         }
1334         spin_unlock(&sysctl_lock);
1335         return NULL;
1336 }
1337
1338 #ifdef CONFIG_SYSCTL_SYSCALL
1339 int do_sysctl(int __user *name, int nlen, void __user *oldval, size_t __user *oldlenp,
1340                void __user *newval, size_t newlen)
1341 {
1342         struct ctl_table_header *head;
1343         int error = -ENOTDIR;
1344
1345         if (nlen <= 0 || nlen >= CTL_MAXNAME)
1346                 return -ENOTDIR;
1347         if (oldval) {
1348                 int old_len;
1349                 if (!oldlenp || get_user(old_len, oldlenp))
1350                         return -EFAULT;
1351         }
1352
1353         for (head = sysctl_head_next(NULL); head;
1354                         head = sysctl_head_next(head)) {
1355                 error = parse_table(name, nlen, oldval, oldlenp, 
1356                                         newval, newlen, head->ctl_table);
1357                 if (error != -ENOTDIR) {
1358                         sysctl_head_finish(head);
1359                         break;
1360                 }
1361         }
1362         return error;
1363 }
1364
1365 asmlinkage long sys_sysctl(struct __sysctl_args __user *args)
1366 {
1367         struct __sysctl_args tmp;
1368         int error;
1369
1370         if (copy_from_user(&tmp, args, sizeof(tmp)))
1371                 return -EFAULT;
1372
1373         error = deprecated_sysctl_warning(&tmp);
1374         if (error)
1375                 goto out;
1376
1377         lock_kernel();
1378         error = do_sysctl(tmp.name, tmp.nlen, tmp.oldval, tmp.oldlenp,
1379                           tmp.newval, tmp.newlen);
1380         unlock_kernel();
1381 out:
1382         return error;
1383 }
1384 #endif /* CONFIG_SYSCTL_SYSCALL */
1385
1386 /*
1387  * sysctl_perm does NOT grant the superuser all rights automatically, because
1388  * some sysctl variables are readonly even to root.
1389  */
1390
1391 static int test_perm(int mode, int op)
1392 {
1393         if (!current->euid)
1394                 mode >>= 6;
1395         else if (in_egroup_p(0))
1396                 mode >>= 3;
1397         if ((mode & op & 0007) == op)
1398                 return 0;
1399         return -EACCES;
1400 }
1401
1402 int sysctl_perm(struct ctl_table *table, int op)
1403 {
1404         int error;
1405         error = security_sysctl(table, op);
1406         if (error)
1407                 return error;
1408         return test_perm(table->mode, op);
1409 }
1410
1411 #ifdef CONFIG_SYSCTL_SYSCALL
1412 static int parse_table(int __user *name, int nlen,
1413                        void __user *oldval, size_t __user *oldlenp,
1414                        void __user *newval, size_t newlen,
1415                        struct ctl_table *table)
1416 {
1417         int n;
1418 repeat:
1419         if (!nlen)
1420                 return -ENOTDIR;
1421         if (get_user(n, name))
1422                 return -EFAULT;
1423         for ( ; table->ctl_name || table->procname; table++) {
1424                 if (!table->ctl_name)
1425                         continue;
1426                 if (n == table->ctl_name) {
1427                         int error;
1428                         if (table->child) {
1429                                 if (sysctl_perm(table, 001))
1430                                         return -EPERM;
1431                                 name++;
1432                                 nlen--;
1433                                 table = table->child;
1434                                 goto repeat;
1435                         }
1436                         error = do_sysctl_strategy(table, name, nlen,
1437                                                    oldval, oldlenp,
1438                                                    newval, newlen);
1439                         return error;
1440                 }
1441         }
1442         return -ENOTDIR;
1443 }
1444
1445 /* Perform the actual read/write of a sysctl table entry. */
1446 int do_sysctl_strategy (struct ctl_table *table,
1447                         int __user *name, int nlen,
1448                         void __user *oldval, size_t __user *oldlenp,
1449                         void __user *newval, size_t newlen)
1450 {
1451         int op = 0, rc;
1452
1453         if (oldval)
1454                 op |= 004;
1455         if (newval) 
1456                 op |= 002;
1457         if (sysctl_perm(table, op))
1458                 return -EPERM;
1459
1460         if (table->strategy) {
1461                 rc = table->strategy(table, name, nlen, oldval, oldlenp,
1462                                      newval, newlen);
1463                 if (rc < 0)
1464                         return rc;
1465                 if (rc > 0)
1466                         return 0;
1467         }
1468
1469         /* If there is no strategy routine, or if the strategy returns
1470          * zero, proceed with automatic r/w */
1471         if (table->data && table->maxlen) {
1472                 rc = sysctl_data(table, name, nlen, oldval, oldlenp,
1473                                  newval, newlen);
1474                 if (rc < 0)
1475                         return rc;
1476         }
1477         return 0;
1478 }
1479 #endif /* CONFIG_SYSCTL_SYSCALL */
1480
1481 static void sysctl_set_parent(struct ctl_table *parent, struct ctl_table *table)
1482 {
1483         for (; table->ctl_name || table->procname; table++) {
1484                 table->parent = parent;
1485                 if (table->child)
1486                         sysctl_set_parent(table, table->child);
1487         }
1488 }
1489
1490 static __init int sysctl_init(void)
1491 {
1492         int err;
1493         sysctl_set_parent(NULL, root_table);
1494         err = sysctl_check_table(root_table);
1495         return 0;
1496 }
1497
1498 core_initcall(sysctl_init);
1499
1500 /**
1501  * register_sysctl_table - register a sysctl hierarchy
1502  * @table: the top-level table structure
1503  *
1504  * Register a sysctl table hierarchy. @table should be a filled in ctl_table
1505  * array. An entry with a ctl_name of 0 terminates the table. 
1506  *
1507  * The members of the &struct ctl_table structure are used as follows:
1508  *
1509  * ctl_name - This is the numeric sysctl value used by sysctl(2). The number
1510  *            must be unique within that level of sysctl
1511  *
1512  * procname - the name of the sysctl file under /proc/sys. Set to %NULL to not
1513  *            enter a sysctl file
1514  *
1515  * data - a pointer to data for use by proc_handler
1516  *
1517  * maxlen - the maximum size in bytes of the data
1518  *
1519  * mode - the file permissions for the /proc/sys file, and for sysctl(2)
1520  *
1521  * child - a pointer to the child sysctl table if this entry is a directory, or
1522  *         %NULL.
1523  *
1524  * proc_handler - the text handler routine (described below)
1525  *
1526  * strategy - the strategy routine (described below)
1527  *
1528  * de - for internal use by the sysctl routines
1529  *
1530  * extra1, extra2 - extra pointers usable by the proc handler routines
1531  *
1532  * Leaf nodes in the sysctl tree will be represented by a single file
1533  * under /proc; non-leaf nodes will be represented by directories.
1534  *
1535  * sysctl(2) can automatically manage read and write requests through
1536  * the sysctl table.  The data and maxlen fields of the ctl_table
1537  * struct enable minimal validation of the values being written to be
1538  * performed, and the mode field allows minimal authentication.
1539  *
1540  * More sophisticated management can be enabled by the provision of a
1541  * strategy routine with the table entry.  This will be called before
1542  * any automatic read or write of the data is performed.
1543  *
1544  * The strategy routine may return
1545  *
1546  * < 0 - Error occurred (error is passed to user process)
1547  *
1548  * 0   - OK - proceed with automatic read or write.
1549  *
1550  * > 0 - OK - read or write has been done by the strategy routine, so
1551  *       return immediately.
1552  *
1553  * There must be a proc_handler routine for any terminal nodes
1554  * mirrored under /proc/sys (non-terminals are handled by a built-in
1555  * directory handler).  Several default handlers are available to
1556  * cover common cases -
1557  *
1558  * proc_dostring(), proc_dointvec(), proc_dointvec_jiffies(),
1559  * proc_dointvec_userhz_jiffies(), proc_dointvec_minmax(), 
1560  * proc_doulongvec_ms_jiffies_minmax(), proc_doulongvec_minmax()
1561  *
1562  * It is the handler's job to read the input buffer from user memory
1563  * and process it. The handler should return 0 on success.
1564  *
1565  * This routine returns %NULL on a failure to register, and a pointer
1566  * to the table header on success.
1567  */
1568 struct ctl_table_header *register_sysctl_table(struct ctl_table * table)
1569 {
1570         struct ctl_table_header *tmp;
1571         tmp = kmalloc(sizeof(struct ctl_table_header), GFP_KERNEL);
1572         if (!tmp)
1573                 return NULL;
1574         tmp->ctl_table = table;
1575         INIT_LIST_HEAD(&tmp->ctl_entry);
1576         tmp->used = 0;
1577         tmp->unregistering = NULL;
1578         sysctl_set_parent(NULL, table);
1579         if (sysctl_check_table(tmp->ctl_table)) {
1580                 kfree(tmp);
1581                 return NULL;
1582         }
1583         spin_lock(&sysctl_lock);
1584         list_add_tail(&tmp->ctl_entry, &root_table_header.ctl_entry);
1585         spin_unlock(&sysctl_lock);
1586         return tmp;
1587 }
1588
1589 /**
1590  * unregister_sysctl_table - unregister a sysctl table hierarchy
1591  * @header: the header returned from register_sysctl_table
1592  *
1593  * Unregisters the sysctl table and all children. proc entries may not
1594  * actually be removed until they are no longer used by anyone.
1595  */
1596 void unregister_sysctl_table(struct ctl_table_header * header)
1597 {
1598         might_sleep();
1599
1600         if (header == NULL)
1601                 return;
1602
1603         spin_lock(&sysctl_lock);
1604         start_unregistering(header);
1605         spin_unlock(&sysctl_lock);
1606         kfree(header);
1607 }
1608
1609 #else /* !CONFIG_SYSCTL */
1610 struct ctl_table_header *register_sysctl_table(struct ctl_table * table)
1611 {
1612         return NULL;
1613 }
1614
1615 void unregister_sysctl_table(struct ctl_table_header * table)
1616 {
1617 }
1618
1619 #endif /* CONFIG_SYSCTL */
1620
1621 /*
1622  * /proc/sys support
1623  */
1624
1625 #ifdef CONFIG_PROC_SYSCTL
1626
1627 static int _proc_do_string(void* data, int maxlen, int write,
1628                            struct file *filp, void __user *buffer,
1629                            size_t *lenp, loff_t *ppos)
1630 {
1631         size_t len;
1632         char __user *p;
1633         char c;
1634
1635         if (!data || !maxlen || !*lenp) {
1636                 *lenp = 0;
1637                 return 0;
1638         }
1639
1640         if (write) {
1641                 len = 0;
1642                 p = buffer;
1643                 while (len < *lenp) {
1644                         if (get_user(c, p++))
1645                                 return -EFAULT;
1646                         if (c == 0 || c == '\n')
1647                                 break;
1648                         len++;
1649                 }
1650                 if (len >= maxlen)
1651                         len = maxlen-1;
1652                 if(copy_from_user(data, buffer, len))
1653                         return -EFAULT;
1654                 ((char *) data)[len] = 0;
1655                 *ppos += *lenp;
1656         } else {
1657                 len = strlen(data);
1658                 if (len > maxlen)
1659                         len = maxlen;
1660
1661                 if (*ppos > len) {
1662                         *lenp = 0;
1663                         return 0;
1664                 }
1665
1666                 data += *ppos;
1667                 len  -= *ppos;
1668
1669                 if (len > *lenp)
1670                         len = *lenp;
1671                 if (len)
1672                         if(copy_to_user(buffer, data, len))
1673                                 return -EFAULT;
1674                 if (len < *lenp) {
1675                         if(put_user('\n', ((char __user *) buffer) + len))
1676                                 return -EFAULT;
1677                         len++;
1678                 }
1679                 *lenp = len;
1680                 *ppos += len;
1681         }
1682         return 0;
1683 }
1684
1685 /**
1686  * proc_dostring - read a string sysctl
1687  * @table: the sysctl table
1688  * @write: %TRUE if this is a write to the sysctl file
1689  * @filp: the file structure
1690  * @buffer: the user buffer
1691  * @lenp: the size of the user buffer
1692  * @ppos: file position
1693  *
1694  * Reads/writes a string from/to the user buffer. If the kernel
1695  * buffer provided is not large enough to hold the string, the
1696  * string is truncated. The copied string is %NULL-terminated.
1697  * If the string is being read by the user process, it is copied
1698  * and a newline '\n' is added. It is truncated if the buffer is
1699  * not large enough.
1700  *
1701  * Returns 0 on success.
1702  */
1703 int proc_dostring(struct ctl_table *table, int write, struct file *filp,
1704                   void __user *buffer, size_t *lenp, loff_t *ppos)
1705 {
1706         return _proc_do_string(table->data, table->maxlen, write, filp,
1707                                buffer, lenp, ppos);
1708 }
1709
1710
1711 static int do_proc_dointvec_conv(int *negp, unsigned long *lvalp,
1712                                  int *valp,
1713                                  int write, void *data)
1714 {
1715         if (write) {
1716                 *valp = *negp ? -*lvalp : *lvalp;
1717         } else {
1718                 int val = *valp;
1719                 if (val < 0) {
1720                         *negp = -1;
1721                         *lvalp = (unsigned long)-val;
1722                 } else {
1723                         *negp = 0;
1724                         *lvalp = (unsigned long)val;
1725                 }
1726         }
1727         return 0;
1728 }
1729
1730 static int __do_proc_dointvec(void *tbl_data, struct ctl_table *table,
1731                   int write, struct file *filp, void __user *buffer,
1732                   size_t *lenp, loff_t *ppos,
1733                   int (*conv)(int *negp, unsigned long *lvalp, int *valp,
1734                               int write, void *data),
1735                   void *data)
1736 {
1737 #define TMPBUFLEN 21
1738         int *i, vleft, first=1, neg, val;
1739         unsigned long lval;
1740         size_t left, len;
1741         
1742         char buf[TMPBUFLEN], *p;
1743         char __user *s = buffer;
1744         
1745         if (!tbl_data || !table->maxlen || !*lenp ||
1746             (*ppos && !write)) {
1747                 *lenp = 0;
1748                 return 0;
1749         }
1750         
1751         i = (int *) tbl_data;
1752         vleft = table->maxlen / sizeof(*i);
1753         left = *lenp;
1754
1755         if (!conv)
1756                 conv = do_proc_dointvec_conv;
1757
1758         for (; left && vleft--; i++, first=0) {
1759                 if (write) {
1760                         while (left) {
1761                                 char c;
1762                                 if (get_user(c, s))
1763                                         return -EFAULT;
1764                                 if (!isspace(c))
1765                                         break;
1766                                 left--;
1767                                 s++;
1768                         }
1769                         if (!left)
1770                                 break;
1771                         neg = 0;
1772                         len = left;
1773                         if (len > sizeof(buf) - 1)
1774                                 len = sizeof(buf) - 1;
1775                         if (copy_from_user(buf, s, len))
1776                                 return -EFAULT;
1777                         buf[len] = 0;
1778                         p = buf;
1779                         if (*p == '-' && left > 1) {
1780                                 neg = 1;
1781                                 p++;
1782                         }
1783                         if (*p < '0' || *p > '9')
1784                                 break;
1785
1786                         lval = simple_strtoul(p, &p, 0);
1787
1788                         len = p-buf;
1789                         if ((len < left) && *p && !isspace(*p))
1790                                 break;
1791                         if (neg)
1792                                 val = -val;
1793                         s += len;
1794                         left -= len;
1795
1796                         if (conv(&neg, &lval, i, 1, data))
1797                                 break;
1798                 } else {
1799                         p = buf;
1800                         if (!first)
1801                                 *p++ = '\t';
1802         
1803                         if (conv(&neg, &lval, i, 0, data))
1804                                 break;
1805
1806                         sprintf(p, "%s%lu", neg ? "-" : "", lval);
1807                         len = strlen(buf);
1808                         if (len > left)
1809                                 len = left;
1810                         if(copy_to_user(s, buf, len))
1811                                 return -EFAULT;
1812                         left -= len;
1813                         s += len;
1814                 }
1815         }
1816
1817         if (!write && !first && left) {
1818                 if(put_user('\n', s))
1819                         return -EFAULT;
1820                 left--, s++;
1821         }
1822         if (write) {
1823                 while (left) {
1824                         char c;
1825                         if (get_user(c, s++))
1826                                 return -EFAULT;
1827                         if (!isspace(c))
1828                                 break;
1829                         left--;
1830                 }
1831         }
1832         if (write && first)
1833                 return -EINVAL;
1834         *lenp -= left;
1835         *ppos += *lenp;
1836         return 0;
1837 #undef TMPBUFLEN
1838 }
1839
1840 static int do_proc_dointvec(struct ctl_table *table, int write, struct file *filp,
1841                   void __user *buffer, size_t *lenp, loff_t *ppos,
1842                   int (*conv)(int *negp, unsigned long *lvalp, int *valp,
1843                               int write, void *data),
1844                   void *data)
1845 {
1846         return __do_proc_dointvec(table->data, table, write, filp,
1847                         buffer, lenp, ppos, conv, data);
1848 }
1849
1850 /**
1851  * proc_dointvec - read a vector of integers
1852  * @table: the sysctl table
1853  * @write: %TRUE if this is a write to the sysctl file
1854  * @filp: the file structure
1855  * @buffer: the user buffer
1856  * @lenp: the size of the user buffer
1857  * @ppos: file position
1858  *
1859  * Reads/writes up to table->maxlen/sizeof(unsigned int) integer
1860  * values from/to the user buffer, treated as an ASCII string. 
1861  *
1862  * Returns 0 on success.
1863  */
1864 int proc_dointvec(struct ctl_table *table, int write, struct file *filp,
1865                      void __user *buffer, size_t *lenp, loff_t *ppos)
1866 {
1867     return do_proc_dointvec(table,write,filp,buffer,lenp,ppos,
1868                             NULL,NULL);
1869 }
1870
1871 #define OP_SET  0
1872 #define OP_AND  1
1873 #define OP_OR   2
1874
1875 static int do_proc_dointvec_bset_conv(int *negp, unsigned long *lvalp,
1876                                       int *valp,
1877                                       int write, void *data)
1878 {
1879         int op = *(int *)data;
1880         if (write) {
1881                 int val = *negp ? -*lvalp : *lvalp;
1882                 switch(op) {
1883                 case OP_SET:    *valp = val; break;
1884                 case OP_AND:    *valp &= val; break;
1885                 case OP_OR:     *valp |= val; break;
1886                 }
1887         } else {
1888                 int val = *valp;
1889                 if (val < 0) {
1890                         *negp = -1;
1891                         *lvalp = (unsigned long)-val;
1892                 } else {
1893                         *negp = 0;
1894                         *lvalp = (unsigned long)val;
1895                 }
1896         }
1897         return 0;
1898 }
1899
1900 #ifdef CONFIG_SECURITY_CAPABILITIES
1901 /*
1902  *      init may raise the set.
1903  */
1904
1905 int proc_dointvec_bset(struct ctl_table *table, int write, struct file *filp,
1906                         void __user *buffer, size_t *lenp, loff_t *ppos)
1907 {
1908         int op;
1909
1910         if (write && !capable(CAP_SYS_MODULE)) {
1911                 return -EPERM;
1912         }
1913
1914         op = is_global_init(current) ? OP_SET : OP_AND;
1915         return do_proc_dointvec(table,write,filp,buffer,lenp,ppos,
1916                                 do_proc_dointvec_bset_conv,&op);
1917 }
1918 #endif /* def CONFIG_SECURITY_CAPABILITIES */
1919
1920 /*
1921  *      Taint values can only be increased
1922  */
1923 static int proc_dointvec_taint(struct ctl_table *table, int write, struct file *filp,
1924                                void __user *buffer, size_t *lenp, loff_t *ppos)
1925 {
1926         int op;
1927
1928         if (write && !capable(CAP_SYS_ADMIN))
1929                 return -EPERM;
1930
1931         op = OP_OR;
1932         return do_proc_dointvec(table,write,filp,buffer,lenp,ppos,
1933                                 do_proc_dointvec_bset_conv,&op);
1934 }
1935
1936 struct do_proc_dointvec_minmax_conv_param {
1937         int *min;
1938         int *max;
1939 };
1940
1941 static int do_proc_dointvec_minmax_conv(int *negp, unsigned long *lvalp, 
1942                                         int *valp, 
1943                                         int write, void *data)
1944 {
1945         struct do_proc_dointvec_minmax_conv_param *param = data;
1946         if (write) {
1947                 int val = *negp ? -*lvalp : *lvalp;
1948                 if ((param->min && *param->min > val) ||
1949                     (param->max && *param->max < val))
1950                         return -EINVAL;
1951                 *valp = val;
1952         } else {
1953                 int val = *valp;
1954                 if (val < 0) {
1955                         *negp = -1;
1956                         *lvalp = (unsigned long)-val;
1957                 } else {
1958                         *negp = 0;
1959                         *lvalp = (unsigned long)val;
1960                 }
1961         }
1962         return 0;
1963 }
1964
1965 /**
1966  * proc_dointvec_minmax - read a vector of integers with min/max values
1967  * @table: the sysctl table
1968  * @write: %TRUE if this is a write to the sysctl file
1969  * @filp: the file structure
1970  * @buffer: the user buffer
1971  * @lenp: the size of the user buffer
1972  * @ppos: file position
1973  *
1974  * Reads/writes up to table->maxlen/sizeof(unsigned int) integer
1975  * values from/to the user buffer, treated as an ASCII string.
1976  *
1977  * This routine will ensure the values are within the range specified by
1978  * table->extra1 (min) and table->extra2 (max).
1979  *
1980  * Returns 0 on success.
1981  */
1982 int proc_dointvec_minmax(struct ctl_table *table, int write, struct file *filp,
1983                   void __user *buffer, size_t *lenp, loff_t *ppos)
1984 {
1985         struct do_proc_dointvec_minmax_conv_param param = {
1986                 .min = (int *) table->extra1,
1987                 .max = (int *) table->extra2,
1988         };
1989         return do_proc_dointvec(table, write, filp, buffer, lenp, ppos,
1990                                 do_proc_dointvec_minmax_conv, &param);
1991 }
1992
1993 static int __do_proc_doulongvec_minmax(void *data, struct ctl_table *table, int write,
1994                                      struct file *filp,
1995                                      void __user *buffer,
1996                                      size_t *lenp, loff_t *ppos,
1997                                      unsigned long convmul,
1998                                      unsigned long convdiv)
1999 {
2000 #define TMPBUFLEN 21
2001         unsigned long *i, *min, *max, val;
2002         int vleft, first=1, neg;
2003         size_t len, left;
2004         char buf[TMPBUFLEN], *p;
2005         char __user *s = buffer;
2006         
2007         if (!data || !table->maxlen || !*lenp ||
2008             (*ppos && !write)) {
2009                 *lenp = 0;
2010                 return 0;
2011         }
2012         
2013         i = (unsigned long *) data;
2014         min = (unsigned long *) table->extra1;
2015         max = (unsigned long *) table->extra2;
2016         vleft = table->maxlen / sizeof(unsigned long);
2017         left = *lenp;
2018         
2019         for (; left && vleft--; i++, min++, max++, first=0) {
2020                 if (write) {
2021                         while (left) {
2022                                 char c;
2023                                 if (get_user(c, s))
2024                                         return -EFAULT;
2025                                 if (!isspace(c))
2026                                         break;
2027                                 left--;
2028                                 s++;
2029                         }
2030                         if (!left)
2031                                 break;
2032                         neg = 0;
2033                         len = left;
2034                         if (len > TMPBUFLEN-1)
2035                                 len = TMPBUFLEN-1;
2036                         if (copy_from_user(buf, s, len))
2037                                 return -EFAULT;
2038                         buf[len] = 0;
2039                         p = buf;
2040                         if (*p == '-' && left > 1) {
2041                                 neg = 1;
2042                                 p++;
2043                         }
2044                         if (*p < '0' || *p > '9')
2045                                 break;
2046                         val = simple_strtoul(p, &p, 0) * convmul / convdiv ;
2047                         len = p-buf;
2048                         if ((len < left) && *p && !isspace(*p))
2049                                 break;
2050                         if (neg)
2051                                 val = -val;
2052                         s += len;
2053                         left -= len;
2054
2055                         if(neg)
2056                                 continue;
2057                         if ((min && val < *min) || (max && val > *max))
2058                                 continue;
2059                         *i = val;
2060                 } else {
2061                         p = buf;
2062                         if (!first)
2063                                 *p++ = '\t';
2064                         sprintf(p, "%lu", convdiv * (*i) / convmul);
2065                         len = strlen(buf);
2066                         if (len > left)
2067                                 len = left;
2068                         if(copy_to_user(s, buf, len))
2069                                 return -EFAULT;
2070                         left -= len;
2071                         s += len;
2072                 }
2073         }
2074
2075         if (!write && !first && left) {
2076                 if(put_user('\n', s))
2077                         return -EFAULT;
2078                 left--, s++;
2079         }
2080         if (write) {
2081                 while (left) {
2082                         char c;
2083                         if (get_user(c, s++))
2084                                 return -EFAULT;
2085                         if (!isspace(c))
2086                                 break;
2087                         left--;
2088                 }
2089         }
2090         if (write && first)
2091                 return -EINVAL;
2092         *lenp -= left;
2093         *ppos += *lenp;
2094         return 0;
2095 #undef TMPBUFLEN
2096 }
2097
2098 static int do_proc_doulongvec_minmax(struct ctl_table *table, int write,
2099                                      struct file *filp,
2100                                      void __user *buffer,
2101                                      size_t *lenp, loff_t *ppos,
2102                                      unsigned long convmul,
2103                                      unsigned long convdiv)
2104 {
2105         return __do_proc_doulongvec_minmax(table->data, table, write,
2106                         filp, buffer, lenp, ppos, convmul, convdiv);
2107 }
2108
2109 /**
2110  * proc_doulongvec_minmax - read a vector of long integers with min/max values
2111  * @table: the sysctl table
2112  * @write: %TRUE if this is a write to the sysctl file
2113  * @filp: the file structure
2114  * @buffer: the user buffer
2115  * @lenp: the size of the user buffer
2116  * @ppos: file position
2117  *
2118  * Reads/writes up to table->maxlen/sizeof(unsigned long) unsigned long
2119  * values from/to the user buffer, treated as an ASCII string.
2120  *
2121  * This routine will ensure the values are within the range specified by
2122  * table->extra1 (min) and table->extra2 (max).
2123  *
2124  * Returns 0 on success.
2125  */
2126 int proc_doulongvec_minmax(struct ctl_table *table, int write, struct file *filp,
2127                            void __user *buffer, size_t *lenp, loff_t *ppos)
2128 {
2129     return do_proc_doulongvec_minmax(table, write, filp, buffer, lenp, ppos, 1l, 1l);
2130 }
2131
2132 /**
2133  * proc_doulongvec_ms_jiffies_minmax - read a vector of millisecond values with min/max values
2134  * @table: the sysctl table
2135  * @write: %TRUE if this is a write to the sysctl file
2136  * @filp: the file structure
2137  * @buffer: the user buffer
2138  * @lenp: the size of the user buffer
2139  * @ppos: file position
2140  *
2141  * Reads/writes up to table->maxlen/sizeof(unsigned long) unsigned long
2142  * values from/to the user buffer, treated as an ASCII string. The values
2143  * are treated as milliseconds, and converted to jiffies when they are stored.
2144  *
2145  * This routine will ensure the values are within the range specified by
2146  * table->extra1 (min) and table->extra2 (max).
2147  *
2148  * Returns 0 on success.
2149  */
2150 int proc_doulongvec_ms_jiffies_minmax(struct ctl_table *table, int write,
2151                                       struct file *filp,
2152                                       void __user *buffer,
2153                                       size_t *lenp, loff_t *ppos)
2154 {
2155     return do_proc_doulongvec_minmax(table, write, filp, buffer,
2156                                      lenp, ppos, HZ, 1000l);
2157 }
2158
2159
2160 static int do_proc_dointvec_jiffies_conv(int *negp, unsigned long *lvalp,
2161                                          int *valp,
2162                                          int write, void *data)
2163 {
2164         if (write) {
2165                 if (*lvalp > LONG_MAX / HZ)
2166                         return 1;
2167                 *valp = *negp ? -(*lvalp*HZ) : (*lvalp*HZ);
2168         } else {
2169                 int val = *valp;
2170                 unsigned long lval;
2171                 if (val < 0) {
2172                         *negp = -1;
2173                         lval = (unsigned long)-val;
2174                 } else {
2175                         *negp = 0;
2176                         lval = (unsigned long)val;
2177                 }
2178                 *lvalp = lval / HZ;
2179         }
2180         return 0;
2181 }
2182
2183 static int do_proc_dointvec_userhz_jiffies_conv(int *negp, unsigned long *lvalp,
2184                                                 int *valp,
2185                                                 int write, void *data)
2186 {
2187         if (write) {
2188                 if (USER_HZ < HZ && *lvalp > (LONG_MAX / HZ) * USER_HZ)
2189                         return 1;
2190                 *valp = clock_t_to_jiffies(*negp ? -*lvalp : *lvalp);
2191         } else {
2192                 int val = *valp;
2193                 unsigned long lval;
2194                 if (val < 0) {
2195                         *negp = -1;
2196                         lval = (unsigned long)-val;
2197                 } else {
2198                         *negp = 0;
2199                         lval = (unsigned long)val;
2200                 }
2201                 *lvalp = jiffies_to_clock_t(lval);
2202         }
2203         return 0;
2204 }
2205
2206 static int do_proc_dointvec_ms_jiffies_conv(int *negp, unsigned long *lvalp,
2207                                             int *valp,
2208                                             int write, void *data)
2209 {
2210         if (write) {
2211                 *valp = msecs_to_jiffies(*negp ? -*lvalp : *lvalp);
2212         } else {
2213                 int val = *valp;
2214                 unsigned long lval;
2215                 if (val < 0) {
2216                         *negp = -1;
2217                         lval = (unsigned long)-val;
2218                 } else {
2219                         *negp = 0;
2220                         lval = (unsigned long)val;
2221                 }
2222                 *lvalp = jiffies_to_msecs(lval);
2223         }
2224         return 0;
2225 }
2226
2227 /**
2228  * proc_dointvec_jiffies - read a vector of integers as seconds
2229  * @table: the sysctl table
2230  * @write: %TRUE if this is a write to the sysctl file
2231  * @filp: the file structure
2232  * @buffer: the user buffer
2233  * @lenp: the size of the user buffer
2234  * @ppos: file position
2235  *
2236  * Reads/writes up to table->maxlen/sizeof(unsigned int) integer
2237  * values from/to the user buffer, treated as an ASCII string. 
2238  * The values read are assumed to be in seconds, and are converted into
2239  * jiffies.
2240  *
2241  * Returns 0 on success.
2242  */
2243 int proc_dointvec_jiffies(struct ctl_table *table, int write, struct file *filp,
2244                           void __user *buffer, size_t *lenp, loff_t *ppos)
2245 {
2246     return do_proc_dointvec(table,write,filp,buffer,lenp,ppos,
2247                             do_proc_dointvec_jiffies_conv,NULL);
2248 }
2249
2250 /**
2251  * proc_dointvec_userhz_jiffies - read a vector of integers as 1/USER_HZ seconds
2252  * @table: the sysctl table
2253  * @write: %TRUE if this is a write to the sysctl file
2254  * @filp: the file structure
2255  * @buffer: the user buffer
2256  * @lenp: the size of the user buffer
2257  * @ppos: pointer to the file position
2258  *
2259  * Reads/writes up to table->maxlen/sizeof(unsigned int) integer
2260  * values from/to the user buffer, treated as an ASCII string. 
2261  * The values read are assumed to be in 1/USER_HZ seconds, and 
2262  * are converted into jiffies.
2263  *
2264  * Returns 0 on success.
2265  */
2266 int proc_dointvec_userhz_jiffies(struct ctl_table *table, int write, struct file *filp,
2267                                  void __user *buffer, size_t *lenp, loff_t *ppos)
2268 {
2269     return do_proc_dointvec(table,write,filp,buffer,lenp,ppos,
2270                             do_proc_dointvec_userhz_jiffies_conv,NULL);
2271 }
2272
2273 /**
2274  * proc_dointvec_ms_jiffies - read a vector of integers as 1 milliseconds
2275  * @table: the sysctl table
2276  * @write: %TRUE if this is a write to the sysctl file
2277  * @filp: the file structure
2278  * @buffer: the user buffer
2279  * @lenp: the size of the user buffer
2280  * @ppos: file position
2281  * @ppos: the current position in the file
2282  *
2283  * Reads/writes up to table->maxlen/sizeof(unsigned int) integer
2284  * values from/to the user buffer, treated as an ASCII string. 
2285  * The values read are assumed to be in 1/1000 seconds, and 
2286  * are converted into jiffies.
2287  *
2288  * Returns 0 on success.
2289  */
2290 int proc_dointvec_ms_jiffies(struct ctl_table *table, int write, struct file *filp,
2291                              void __user *buffer, size_t *lenp, loff_t *ppos)
2292 {
2293         return do_proc_dointvec(table, write, filp, buffer, lenp, ppos,
2294                                 do_proc_dointvec_ms_jiffies_conv, NULL);
2295 }
2296
2297 static int proc_do_cad_pid(struct ctl_table *table, int write, struct file *filp,
2298                            void __user *buffer, size_t *lenp, loff_t *ppos)
2299 {
2300         struct pid *new_pid;
2301         pid_t tmp;
2302         int r;
2303
2304         tmp = pid_nr_ns(cad_pid, current->nsproxy->pid_ns);
2305
2306         r = __do_proc_dointvec(&tmp, table, write, filp, buffer,
2307                                lenp, ppos, NULL, NULL);
2308         if (r || !write)
2309                 return r;
2310
2311         new_pid = find_get_pid(tmp);
2312         if (!new_pid)
2313                 return -ESRCH;
2314
2315         put_pid(xchg(&cad_pid, new_pid));
2316         return 0;
2317 }
2318
2319 #else /* CONFIG_PROC_FS */
2320
2321 int proc_dostring(struct ctl_table *table, int write, struct file *filp,
2322                   void __user *buffer, size_t *lenp, loff_t *ppos)
2323 {
2324         return -ENOSYS;
2325 }
2326
2327 int proc_dointvec(struct ctl_table *table, int write, struct file *filp,
2328                   void __user *buffer, size_t *lenp, loff_t *ppos)
2329 {
2330         return -ENOSYS;
2331 }
2332
2333 int proc_dointvec_bset(struct ctl_table *table, int write, struct file *filp,
2334                         void __user *buffer, size_t *lenp, loff_t *ppos)
2335 {
2336         return -ENOSYS;
2337 }
2338
2339 int proc_dointvec_minmax(struct ctl_table *table, int write, struct file *filp,
2340                     void __user *buffer, size_t *lenp, loff_t *ppos)
2341 {
2342         return -ENOSYS;
2343 }
2344
2345 int proc_dointvec_jiffies(struct ctl_table *table, int write, struct file *filp,
2346                     void __user *buffer, size_t *lenp, loff_t *ppos)
2347 {
2348         return -ENOSYS;
2349 }
2350
2351 int proc_dointvec_userhz_jiffies(struct ctl_table *table, int write, struct file *filp,
2352                     void __user *buffer, size_t *lenp, loff_t *ppos)
2353 {
2354         return -ENOSYS;
2355 }
2356
2357 int proc_dointvec_ms_jiffies(struct ctl_table *table, int write, struct file *filp,
2358                              void __user *buffer, size_t *lenp, loff_t *ppos)
2359 {
2360         return -ENOSYS;
2361 }
2362
2363 int proc_doulongvec_minmax(struct ctl_table *table, int write, struct file *filp,
2364                     void __user *buffer, size_t *lenp, loff_t *ppos)
2365 {
2366         return -ENOSYS;
2367 }
2368
2369 int proc_doulongvec_ms_jiffies_minmax(struct ctl_table *table, int write,
2370                                       struct file *filp,
2371                                       void __user *buffer,
2372                                       size_t *lenp, loff_t *ppos)
2373 {
2374     return -ENOSYS;
2375 }
2376
2377
2378 #endif /* CONFIG_PROC_FS */
2379
2380
2381 #ifdef CONFIG_SYSCTL_SYSCALL
2382 /*
2383  * General sysctl support routines 
2384  */
2385
2386 /* The generic sysctl data routine (used if no strategy routine supplied) */
2387 int sysctl_data(struct ctl_table *table, int __user *name, int nlen,
2388                 void __user *oldval, size_t __user *oldlenp,
2389                 void __user *newval, size_t newlen)
2390 {
2391         size_t len;
2392
2393         /* Get out of I don't have a variable */
2394         if (!table->data || !table->maxlen)
2395                 return -ENOTDIR;
2396
2397         if (oldval && oldlenp) {
2398                 if (get_user(len, oldlenp))
2399                         return -EFAULT;
2400                 if (len) {
2401                         if (len > table->maxlen)
2402                                 len = table->maxlen;
2403                         if (copy_to_user(oldval, table->data, len))
2404                                 return -EFAULT;
2405                         if (put_user(len, oldlenp))
2406                                 return -EFAULT;
2407                 }
2408         }
2409
2410         if (newval && newlen) {
2411                 if (newlen > table->maxlen)
2412                         newlen = table->maxlen;
2413
2414                 if (copy_from_user(table->data, newval, newlen))
2415                         return -EFAULT;
2416         }
2417         return 1;
2418 }
2419
2420 /* The generic string strategy routine: */
2421 int sysctl_string(struct ctl_table *table, int __user *name, int nlen,
2422                   void __user *oldval, size_t __user *oldlenp,
2423                   void __user *newval, size_t newlen)
2424 {
2425         if (!table->data || !table->maxlen) 
2426                 return -ENOTDIR;
2427         
2428         if (oldval && oldlenp) {
2429                 size_t bufsize;
2430                 if (get_user(bufsize, oldlenp))
2431                         return -EFAULT;
2432                 if (bufsize) {
2433                         size_t len = strlen(table->data), copied;
2434
2435                         /* This shouldn't trigger for a well-formed sysctl */
2436                         if (len > table->maxlen)
2437                                 len = table->maxlen;
2438
2439                         /* Copy up to a max of bufsize-1 bytes of the string */
2440                         copied = (len >= bufsize) ? bufsize - 1 : len;
2441
2442                         if (copy_to_user(oldval, table->data, copied) ||
2443                             put_user(0, (char __user *)(oldval + copied)))
2444                                 return -EFAULT;
2445                         if (put_user(len, oldlenp))
2446                                 return -EFAULT;
2447                 }
2448         }
2449         if (newval && newlen) {
2450                 size_t len = newlen;
2451                 if (len > table->maxlen)
2452                         len = table->maxlen;
2453                 if(copy_from_user(table->data, newval, len))
2454                         return -EFAULT;
2455                 if (len == table->maxlen)
2456                         len--;
2457                 ((char *) table->data)[len] = 0;
2458         }
2459         return 1;
2460 }
2461
2462 /*
2463  * This function makes sure that all of the integers in the vector
2464  * are between the minimum and maximum values given in the arrays
2465  * table->extra1 and table->extra2, respectively.
2466  */
2467 int sysctl_intvec(struct ctl_table *table, int __user *name, int nlen,
2468                 void __user *oldval, size_t __user *oldlenp,
2469                 void __user *newval, size_t newlen)
2470 {
2471
2472         if (newval && newlen) {
2473                 int __user *vec = (int __user *) newval;
2474                 int *min = (int *) table->extra1;
2475                 int *max = (int *) table->extra2;
2476                 size_t length;
2477                 int i;
2478
2479                 if (newlen % sizeof(int) != 0)
2480                         return -EINVAL;
2481
2482                 if (!table->extra1 && !table->extra2)
2483                         return 0;
2484
2485                 if (newlen > table->maxlen)
2486                         newlen = table->maxlen;
2487                 length = newlen / sizeof(int);
2488
2489                 for (i = 0; i < length; i++) {
2490                         int value;
2491                         if (get_user(value, vec + i))
2492                                 return -EFAULT;
2493                         if (min && value < min[i])
2494                                 return -EINVAL;
2495                         if (max && value > max[i])
2496                                 return -EINVAL;
2497                 }
2498         }
2499         return 0;
2500 }
2501
2502 /* Strategy function to convert jiffies to seconds */ 
2503 int sysctl_jiffies(struct ctl_table *table, int __user *name, int nlen,
2504                 void __user *oldval, size_t __user *oldlenp,
2505                 void __user *newval, size_t newlen)
2506 {
2507         if (oldval && oldlenp) {
2508                 size_t olen;
2509
2510                 if (get_user(olen, oldlenp))
2511                         return -EFAULT;
2512                 if (olen) {
2513                         int val;
2514
2515                         if (olen < sizeof(int))
2516                                 return -EINVAL;
2517
2518                         val = *(int *)(table->data) / HZ;
2519                         if (put_user(val, (int __user *)oldval))
2520                                 return -EFAULT;
2521                         if (put_user(sizeof(int), oldlenp))
2522                                 return -EFAULT;
2523                 }
2524         }
2525         if (newval && newlen) { 
2526                 int new;
2527                 if (newlen != sizeof(int))
2528                         return -EINVAL; 
2529                 if (get_user(new, (int __user *)newval))
2530                         return -EFAULT;
2531                 *(int *)(table->data) = new*HZ; 
2532         }
2533         return 1;
2534 }
2535
2536 /* Strategy function to convert jiffies to seconds */ 
2537 int sysctl_ms_jiffies(struct ctl_table *table, int __user *name, int nlen,
2538                 void __user *oldval, size_t __user *oldlenp,
2539                 void __user *newval, size_t newlen)
2540 {
2541         if (oldval && oldlenp) {
2542                 size_t olen;
2543
2544                 if (get_user(olen, oldlenp))
2545                         return -EFAULT;
2546                 if (olen) {
2547                         int val;
2548
2549                         if (olen < sizeof(int))
2550                                 return -EINVAL;
2551
2552                         val = jiffies_to_msecs(*(int *)(table->data));
2553                         if (put_user(val, (int __user *)oldval))
2554                                 return -EFAULT;
2555                         if (put_user(sizeof(int), oldlenp))
2556                                 return -EFAULT;
2557                 }
2558         }
2559         if (newval && newlen) { 
2560                 int new;
2561                 if (newlen != sizeof(int))
2562                         return -EINVAL; 
2563                 if (get_user(new, (int __user *)newval))
2564                         return -EFAULT;
2565                 *(int *)(table->data) = msecs_to_jiffies(new);
2566         }
2567         return 1;
2568 }
2569
2570
2571
2572 #else /* CONFIG_SYSCTL_SYSCALL */
2573
2574
2575 asmlinkage long sys_sysctl(struct __sysctl_args __user *args)
2576 {
2577         struct __sysctl_args tmp;
2578         int error;
2579
2580         if (copy_from_user(&tmp, args, sizeof(tmp)))
2581                 return -EFAULT;
2582
2583         error = deprecated_sysctl_warning(&tmp);
2584
2585         /* If no error reading the parameters then just -ENOSYS ... */
2586         if (!error)
2587                 error = -ENOSYS;
2588
2589         return error;
2590 }
2591
2592 int sysctl_data(struct ctl_table *table, int __user *name, int nlen,
2593                   void __user *oldval, size_t __user *oldlenp,
2594                   void __user *newval, size_t newlen)
2595 {
2596         return -ENOSYS;
2597 }
2598
2599 int sysctl_string(struct ctl_table *table, int __user *name, int nlen,
2600                   void __user *oldval, size_t __user *oldlenp,
2601                   void __user *newval, size_t newlen)
2602 {
2603         return -ENOSYS;
2604 }
2605
2606 int sysctl_intvec(struct ctl_table *table, int __user *name, int nlen,
2607                 void __user *oldval, size_t __user *oldlenp,
2608                 void __user *newval, size_t newlen)
2609 {
2610         return -ENOSYS;
2611 }
2612
2613 int sysctl_jiffies(struct ctl_table *table, int __user *name, int nlen,
2614                 void __user *oldval, size_t __user *oldlenp,
2615                 void __user *newval, size_t newlen)
2616 {
2617         return -ENOSYS;
2618 }
2619
2620 int sysctl_ms_jiffies(struct ctl_table *table, int __user *name, int nlen,
2621                 void __user *oldval, size_t __user *oldlenp,
2622                 void __user *newval, size_t newlen)
2623 {
2624         return -ENOSYS;
2625 }
2626
2627 #endif /* CONFIG_SYSCTL_SYSCALL */
2628
2629 static int deprecated_sysctl_warning(struct __sysctl_args *args)
2630 {
2631         static int msg_count;
2632         int name[CTL_MAXNAME];
2633         int i;
2634
2635         /* Check args->nlen. */
2636         if (args->nlen < 0 || args->nlen > CTL_MAXNAME)
2637                 return -ENOTDIR;
2638
2639         /* Read in the sysctl name for better debug message logging */
2640         for (i = 0; i < args->nlen; i++)
2641                 if (get_user(name[i], args->name + i))
2642                         return -EFAULT;
2643
2644         /* Ignore accesses to kernel.version */
2645         if ((args->nlen == 2) && (name[0] == CTL_KERN) && (name[1] == KERN_VERSION))
2646                 return 0;
2647
2648         if (msg_count < 5) {
2649                 msg_count++;
2650                 printk(KERN_INFO
2651                         "warning: process `%s' used the deprecated sysctl "
2652                         "system call with ", current->comm);
2653                 for (i = 0; i < args->nlen; i++)
2654                         printk("%d.", name[i]);
2655                 printk("\n");
2656         }
2657         return 0;
2658 }
2659
2660 /*
2661  * No sense putting this after each symbol definition, twice,
2662  * exception granted :-)
2663  */
2664 EXPORT_SYMBOL(proc_dointvec);
2665 EXPORT_SYMBOL(proc_dointvec_jiffies);
2666 EXPORT_SYMBOL(proc_dointvec_minmax);
2667 EXPORT_SYMBOL(proc_dointvec_userhz_jiffies);
2668 EXPORT_SYMBOL(proc_dointvec_ms_jiffies);
2669 EXPORT_SYMBOL(proc_dostring);
2670 EXPORT_SYMBOL(proc_doulongvec_minmax);
2671 EXPORT_SYMBOL(proc_doulongvec_ms_jiffies_minmax);
2672 EXPORT_SYMBOL(register_sysctl_table);
2673 EXPORT_SYMBOL(sysctl_intvec);
2674 EXPORT_SYMBOL(sysctl_jiffies);
2675 EXPORT_SYMBOL(sysctl_ms_jiffies);
2676 EXPORT_SYMBOL(sysctl_string);
2677 EXPORT_SYMBOL(sysctl_data);
2678 EXPORT_SYMBOL(unregister_sysctl_table);