padata: Fix cpu hotplug
[linux-flexiantxendom0-3.2.10.git] / kernel / sysctl.c
1 /*
2  * sysctl.c: General linux system control interface
3  *
4  * Begun 24 March 1995, Stephen Tweedie
5  * Added /proc support, Dec 1995
6  * Added bdflush entry and intvec min/max checking, 2/23/96, Tom Dyas.
7  * Added hooks for /proc/sys/net (minor, minor patch), 96/4/1, Mike Shaver.
8  * Added kernel/java-{interpreter,appletviewer}, 96/5/10, Mike Shaver.
9  * Dynamic registration fixes, Stephen Tweedie.
10  * Added kswapd-interval, ctrl-alt-del, printk stuff, 1/8/97, Chris Horn.
11  * Made sysctl support optional via CONFIG_SYSCTL, 1/10/97, Chris
12  *  Horn.
13  * Added proc_doulongvec_ms_jiffies_minmax, 09/08/99, Carlos H. Bauer.
14  * Added proc_doulongvec_minmax, 09/08/99, Carlos H. Bauer.
15  * Changed linked lists to use list.h instead of lists.h, 02/24/00, Bill
16  *  Wendling.
17  * The list_for_each() macro wasn't appropriate for the sysctl loop.
18  *  Removed it and replaced it with older style, 03/23/00, Bill Wendling
19  */
20
21 #include <linux/module.h>
22 #include <linux/mm.h>
23 #include <linux/swap.h>
24 #include <linux/slab.h>
25 #include <linux/sysctl.h>
26 #include <linux/signal.h>
27 #include <linux/printk.h>
28 #include <linux/proc_fs.h>
29 #include <linux/security.h>
30 #include <linux/ctype.h>
31 #include <linux/kmemcheck.h>
32 #include <linux/fs.h>
33 #include <linux/init.h>
34 #include <linux/kernel.h>
35 #include <linux/kobject.h>
36 #include <linux/net.h>
37 #include <linux/sysrq.h>
38 #include <linux/highuid.h>
39 #include <linux/writeback.h>
40 #include <linux/ratelimit.h>
41 #include <linux/compaction.h>
42 #include <linux/hugetlb.h>
43 #include <linux/initrd.h>
44 #include <linux/key.h>
45 #include <linux/times.h>
46 #include <linux/limits.h>
47 #include <linux/dcache.h>
48 #include <linux/dnotify.h>
49 #include <linux/syscalls.h>
50 #include <linux/vmstat.h>
51 #include <linux/nfs_fs.h>
52 #include <linux/acpi.h>
53 #include <linux/reboot.h>
54 #include <linux/ftrace.h>
55 #include <linux/perf_event.h>
56 #include <linux/kprobes.h>
57 #include <linux/pipe_fs_i.h>
58 #include <linux/oom.h>
59 #include <linux/kmod.h>
60 #include <linux/capability.h>
61 #include <linux/binfmts.h>
62
63 #include <asm/uaccess.h>
64 #include <asm/processor.h>
65
66 #ifdef CONFIG_X86
67 #include <asm/nmi.h>
68 #include <asm/stacktrace.h>
69 #include <asm/io.h>
70 #endif
71 #ifdef CONFIG_BSD_PROCESS_ACCT
72 #include <linux/acct.h>
73 #endif
74 #ifdef CONFIG_RT_MUTEXES
75 #include <linux/rtmutex.h>
76 #endif
77 #if defined(CONFIG_PROVE_LOCKING) || defined(CONFIG_LOCK_STAT)
78 #include <linux/lockdep.h>
79 #endif
80 #ifdef CONFIG_CHR_DEV_SG
81 #include <scsi/sg.h>
82 #endif
83
84 #ifdef CONFIG_LOCKUP_DETECTOR
85 #include <linux/nmi.h>
86 #endif
87
88
89 #if defined(CONFIG_SYSCTL)
90
91 /* External variables not in a header file. */
92 extern int sysctl_overcommit_memory;
93 extern int sysctl_overcommit_ratio;
94 extern int max_threads;
95 extern int core_uses_pid;
96 extern int suid_dumpable;
97 extern char core_pattern[];
98 extern unsigned int core_pipe_limit;
99 extern int pid_max;
100 extern int min_free_kbytes;
101 extern int pid_max_min, pid_max_max;
102 extern int sysctl_drop_caches;
103 extern int percpu_pagelist_fraction;
104 extern int compat_log;
105 extern int latencytop_enabled;
106 extern int sysctl_nr_open_min, sysctl_nr_open_max;
107 #ifndef CONFIG_MMU
108 extern int sysctl_nr_trim_pages;
109 #endif
110 #ifdef CONFIG_BLOCK
111 extern int blk_iopoll_enabled;
112 #endif
113
114 /* Constants used for minimum and  maximum */
115 #ifdef CONFIG_LOCKUP_DETECTOR
116 static int sixty = 60;
117 static int neg_one = -1;
118 #endif
119
120 static int zero;
121 static int __maybe_unused one = 1;
122 static int __maybe_unused two = 2;
123 static int __maybe_unused three = 3;
124 static unsigned long one_ul = 1;
125 static int one_hundred = 100;
126 #ifdef CONFIG_PRINTK
127 static int ten_thousand = 10000;
128 #endif
129
130 /* this is needed for the proc_doulongvec_minmax of vm_dirty_bytes */
131 static unsigned long dirty_bytes_min = 2 * PAGE_SIZE;
132
133 /* this is needed for the proc_dointvec_minmax for [fs_]overflow UID and GID */
134 static int maxolduid = 65535;
135 static int minolduid;
136 static int min_percpu_pagelist_fract = 8;
137
138 static int ngroups_max = NGROUPS_MAX;
139 static const int cap_last_cap = CAP_LAST_CAP;
140
141 #ifdef CONFIG_INOTIFY_USER
142 #include <linux/inotify.h>
143 #endif
144 #ifdef CONFIG_SPARC
145 #include <asm/system.h>
146 #endif
147
148 #ifdef CONFIG_SPARC64
149 extern int sysctl_tsb_ratio;
150 #endif
151
152 #ifdef __hppa__
153 extern int pwrsw_enabled;
154 extern int unaligned_enabled;
155 #endif
156
157 #ifdef CONFIG_IA64
158 extern int no_unaligned_warning;
159 extern int unaligned_dump_stack;
160 #endif
161
162 #ifdef CONFIG_PROC_SYSCTL
163 static int proc_do_cad_pid(struct ctl_table *table, int write,
164                   void __user *buffer, size_t *lenp, loff_t *ppos);
165 static int proc_taint(struct ctl_table *table, int write,
166                                void __user *buffer, size_t *lenp, loff_t *ppos);
167 #endif
168
169 #ifdef CONFIG_PRINTK
170 static int proc_dmesg_restrict(struct ctl_table *table, int write,
171                                 void __user *buffer, size_t *lenp, loff_t *ppos);
172 #endif
173
174 #ifdef CONFIG_MAGIC_SYSRQ
175 /* Note: sysrq code uses it's own private copy */
176 static int __sysrq_enabled = SYSRQ_DEFAULT_ENABLE;
177
178 static int sysrq_sysctl_handler(ctl_table *table, int write,
179                                 void __user *buffer, size_t *lenp,
180                                 loff_t *ppos)
181 {
182         int error;
183
184         error = proc_dointvec(table, write, buffer, lenp, ppos);
185         if (error)
186                 return error;
187
188         if (write)
189                 sysrq_toggle_support(__sysrq_enabled);
190
191         return 0;
192 }
193
194 #endif
195
196 static struct ctl_table root_table[];
197 static struct ctl_table_root sysctl_table_root;
198 static struct ctl_table_header root_table_header = {
199         {{.count = 1,
200         .ctl_table = root_table,
201         .ctl_entry = LIST_HEAD_INIT(sysctl_table_root.default_set.list),}},
202         .root = &sysctl_table_root,
203         .set = &sysctl_table_root.default_set,
204 };
205 static struct ctl_table_root sysctl_table_root = {
206         .root_list = LIST_HEAD_INIT(sysctl_table_root.root_list),
207         .default_set.list = LIST_HEAD_INIT(root_table_header.ctl_entry),
208 };
209
210 static struct ctl_table kern_table[];
211 static struct ctl_table vm_table[];
212 static struct ctl_table fs_table[];
213 static struct ctl_table debug_table[];
214 static struct ctl_table dev_table[];
215 extern struct ctl_table random_table[];
216 #ifdef CONFIG_EPOLL
217 extern struct ctl_table epoll_table[];
218 #endif
219
220 #ifdef HAVE_ARCH_PICK_MMAP_LAYOUT
221 int sysctl_legacy_va_layout;
222 #endif
223
224 /* The default sysctl tables: */
225
226 static struct ctl_table root_table[] = {
227         {
228                 .procname       = "kernel",
229                 .mode           = 0555,
230                 .child          = kern_table,
231         },
232         {
233                 .procname       = "vm",
234                 .mode           = 0555,
235                 .child          = vm_table,
236         },
237         {
238                 .procname       = "fs",
239                 .mode           = 0555,
240                 .child          = fs_table,
241         },
242         {
243                 .procname       = "debug",
244                 .mode           = 0555,
245                 .child          = debug_table,
246         },
247         {
248                 .procname       = "dev",
249                 .mode           = 0555,
250                 .child          = dev_table,
251         },
252         { }
253 };
254
255 #ifdef CONFIG_SCHED_DEBUG
256 static int min_sched_granularity_ns = 100000;           /* 100 usecs */
257 static int max_sched_granularity_ns = NSEC_PER_SEC;     /* 1 second */
258 static int min_wakeup_granularity_ns;                   /* 0 usecs */
259 static int max_wakeup_granularity_ns = NSEC_PER_SEC;    /* 1 second */
260 static int min_sched_tunable_scaling = SCHED_TUNABLESCALING_NONE;
261 static int max_sched_tunable_scaling = SCHED_TUNABLESCALING_END-1;
262 #endif
263
264 #ifdef CONFIG_COMPACTION
265 static int min_extfrag_threshold;
266 static int max_extfrag_threshold = 1000;
267 #endif
268
269 static struct ctl_table kern_table[] = {
270         {
271                 .procname       = "sched_child_runs_first",
272                 .data           = &sysctl_sched_child_runs_first,
273                 .maxlen         = sizeof(unsigned int),
274                 .mode           = 0644,
275                 .proc_handler   = proc_dointvec,
276         },
277 #ifdef CONFIG_SCHED_DEBUG
278         {
279                 .procname       = "sched_min_granularity_ns",
280                 .data           = &sysctl_sched_min_granularity,
281                 .maxlen         = sizeof(unsigned int),
282                 .mode           = 0644,
283                 .proc_handler   = sched_proc_update_handler,
284                 .extra1         = &min_sched_granularity_ns,
285                 .extra2         = &max_sched_granularity_ns,
286         },
287         {
288                 .procname       = "sched_latency_ns",
289                 .data           = &sysctl_sched_latency,
290                 .maxlen         = sizeof(unsigned int),
291                 .mode           = 0644,
292                 .proc_handler   = sched_proc_update_handler,
293                 .extra1         = &min_sched_granularity_ns,
294                 .extra2         = &max_sched_granularity_ns,
295         },
296         {
297                 .procname       = "sched_wakeup_granularity_ns",
298                 .data           = &sysctl_sched_wakeup_granularity,
299                 .maxlen         = sizeof(unsigned int),
300                 .mode           = 0644,
301                 .proc_handler   = sched_proc_update_handler,
302                 .extra1         = &min_wakeup_granularity_ns,
303                 .extra2         = &max_wakeup_granularity_ns,
304         },
305         {
306                 .procname       = "sched_tunable_scaling",
307                 .data           = &sysctl_sched_tunable_scaling,
308                 .maxlen         = sizeof(enum sched_tunable_scaling),
309                 .mode           = 0644,
310                 .proc_handler   = sched_proc_update_handler,
311                 .extra1         = &min_sched_tunable_scaling,
312                 .extra2         = &max_sched_tunable_scaling,
313         },
314         {
315                 .procname       = "sched_migration_cost",
316                 .data           = &sysctl_sched_migration_cost,
317                 .maxlen         = sizeof(unsigned int),
318                 .mode           = 0644,
319                 .proc_handler   = proc_dointvec,
320         },
321         {
322                 .procname       = "sched_nr_migrate",
323                 .data           = &sysctl_sched_nr_migrate,
324                 .maxlen         = sizeof(unsigned int),
325                 .mode           = 0644,
326                 .proc_handler   = proc_dointvec,
327         },
328         {
329                 .procname       = "sched_time_avg",
330                 .data           = &sysctl_sched_time_avg,
331                 .maxlen         = sizeof(unsigned int),
332                 .mode           = 0644,
333                 .proc_handler   = proc_dointvec,
334         },
335         {
336                 .procname       = "sched_shares_window",
337                 .data           = &sysctl_sched_shares_window,
338                 .maxlen         = sizeof(unsigned int),
339                 .mode           = 0644,
340                 .proc_handler   = proc_dointvec,
341         },
342         {
343                 .procname       = "timer_migration",
344                 .data           = &sysctl_timer_migration,
345                 .maxlen         = sizeof(unsigned int),
346                 .mode           = 0644,
347                 .proc_handler   = proc_dointvec_minmax,
348                 .extra1         = &zero,
349                 .extra2         = &one,
350         },
351 #endif
352         {
353                 .procname       = "sched_rt_period_us",
354                 .data           = &sysctl_sched_rt_period,
355                 .maxlen         = sizeof(unsigned int),
356                 .mode           = 0644,
357                 .proc_handler   = sched_rt_handler,
358         },
359         {
360                 .procname       = "sched_rt_runtime_us",
361                 .data           = &sysctl_sched_rt_runtime,
362                 .maxlen         = sizeof(int),
363                 .mode           = 0644,
364                 .proc_handler   = sched_rt_handler,
365         },
366 #ifdef CONFIG_SCHED_AUTOGROUP
367         {
368                 .procname       = "sched_autogroup_enabled",
369                 .data           = &sysctl_sched_autogroup_enabled,
370                 .maxlen         = sizeof(unsigned int),
371                 .mode           = 0644,
372                 .proc_handler   = proc_dointvec_minmax,
373                 .extra1         = &zero,
374                 .extra2         = &one,
375         },
376 #endif
377 #ifdef CONFIG_CFS_BANDWIDTH
378         {
379                 .procname       = "sched_cfs_bandwidth_slice_us",
380                 .data           = &sysctl_sched_cfs_bandwidth_slice,
381                 .maxlen         = sizeof(unsigned int),
382                 .mode           = 0644,
383                 .proc_handler   = proc_dointvec_minmax,
384                 .extra1         = &one,
385         },
386 #endif
387 #ifdef CONFIG_PROVE_LOCKING
388         {
389                 .procname       = "prove_locking",
390                 .data           = &prove_locking,
391                 .maxlen         = sizeof(int),
392                 .mode           = 0644,
393                 .proc_handler   = proc_dointvec,
394         },
395 #endif
396 #ifdef CONFIG_LOCK_STAT
397         {
398                 .procname       = "lock_stat",
399                 .data           = &lock_stat,
400                 .maxlen         = sizeof(int),
401                 .mode           = 0644,
402                 .proc_handler   = proc_dointvec,
403         },
404 #endif
405         {
406                 .procname       = "panic",
407                 .data           = &panic_timeout,
408                 .maxlen         = sizeof(int),
409                 .mode           = 0644,
410                 .proc_handler   = proc_dointvec,
411         },
412         {
413                 .procname       = "core_uses_pid",
414                 .data           = &core_uses_pid,
415                 .maxlen         = sizeof(int),
416                 .mode           = 0644,
417                 .proc_handler   = proc_dointvec,
418         },
419         {
420                 .procname       = "core_pattern",
421                 .data           = core_pattern,
422                 .maxlen         = CORENAME_MAX_SIZE,
423                 .mode           = 0644,
424                 .proc_handler   = proc_dostring,
425         },
426         {
427                 .procname       = "core_pipe_limit",
428                 .data           = &core_pipe_limit,
429                 .maxlen         = sizeof(unsigned int),
430                 .mode           = 0644,
431                 .proc_handler   = proc_dointvec,
432         },
433 #ifdef CONFIG_PROC_SYSCTL
434         {
435                 .procname       = "tainted",
436                 .maxlen         = sizeof(long),
437                 .mode           = 0644,
438                 .proc_handler   = proc_taint,
439         },
440 #endif
441 #ifdef CONFIG_LATENCYTOP
442         {
443                 .procname       = "latencytop",
444                 .data           = &latencytop_enabled,
445                 .maxlen         = sizeof(int),
446                 .mode           = 0644,
447                 .proc_handler   = proc_dointvec,
448         },
449 #endif
450 #ifdef CONFIG_BLK_DEV_INITRD
451         {
452                 .procname       = "real-root-dev",
453                 .data           = &real_root_dev,
454                 .maxlen         = sizeof(int),
455                 .mode           = 0644,
456                 .proc_handler   = proc_dointvec,
457         },
458 #endif
459         {
460                 .procname       = "print-fatal-signals",
461                 .data           = &print_fatal_signals,
462                 .maxlen         = sizeof(int),
463                 .mode           = 0644,
464                 .proc_handler   = proc_dointvec,
465         },
466 #ifdef CONFIG_SPARC
467         {
468                 .procname       = "reboot-cmd",
469                 .data           = reboot_command,
470                 .maxlen         = 256,
471                 .mode           = 0644,
472                 .proc_handler   = proc_dostring,
473         },
474         {
475                 .procname       = "stop-a",
476                 .data           = &stop_a_enabled,
477                 .maxlen         = sizeof (int),
478                 .mode           = 0644,
479                 .proc_handler   = proc_dointvec,
480         },
481         {
482                 .procname       = "scons-poweroff",
483                 .data           = &scons_pwroff,
484                 .maxlen         = sizeof (int),
485                 .mode           = 0644,
486                 .proc_handler   = proc_dointvec,
487         },
488 #endif
489 #ifdef CONFIG_SPARC64
490         {
491                 .procname       = "tsb-ratio",
492                 .data           = &sysctl_tsb_ratio,
493                 .maxlen         = sizeof (int),
494                 .mode           = 0644,
495                 .proc_handler   = proc_dointvec,
496         },
497 #endif
498 #ifdef __hppa__
499         {
500                 .procname       = "soft-power",
501                 .data           = &pwrsw_enabled,
502                 .maxlen         = sizeof (int),
503                 .mode           = 0644,
504                 .proc_handler   = proc_dointvec,
505         },
506         {
507                 .procname       = "unaligned-trap",
508                 .data           = &unaligned_enabled,
509                 .maxlen         = sizeof (int),
510                 .mode           = 0644,
511                 .proc_handler   = proc_dointvec,
512         },
513 #endif
514         {
515                 .procname       = "ctrl-alt-del",
516                 .data           = &C_A_D,
517                 .maxlen         = sizeof(int),
518                 .mode           = 0644,
519                 .proc_handler   = proc_dointvec,
520         },
521 #ifdef CONFIG_FUNCTION_TRACER
522         {
523                 .procname       = "ftrace_enabled",
524                 .data           = &ftrace_enabled,
525                 .maxlen         = sizeof(int),
526                 .mode           = 0644,
527                 .proc_handler   = ftrace_enable_sysctl,
528         },
529 #endif
530 #ifdef CONFIG_STACK_TRACER
531         {
532                 .procname       = "stack_tracer_enabled",
533                 .data           = &stack_tracer_enabled,
534                 .maxlen         = sizeof(int),
535                 .mode           = 0644,
536                 .proc_handler   = stack_trace_sysctl,
537         },
538 #endif
539 #ifdef CONFIG_TRACING
540         {
541                 .procname       = "ftrace_dump_on_oops",
542                 .data           = &ftrace_dump_on_oops,
543                 .maxlen         = sizeof(int),
544                 .mode           = 0644,
545                 .proc_handler   = proc_dointvec,
546         },
547 #endif
548 #ifdef CONFIG_MODULES
549         {
550                 .procname       = "modprobe",
551                 .data           = &modprobe_path,
552                 .maxlen         = KMOD_PATH_LEN,
553                 .mode           = 0644,
554                 .proc_handler   = proc_dostring,
555         },
556         {
557                 .procname       = "modules_disabled",
558                 .data           = &modules_disabled,
559                 .maxlen         = sizeof(int),
560                 .mode           = 0644,
561                 /* only handle a transition from default "0" to "1" */
562                 .proc_handler   = proc_dointvec_minmax,
563                 .extra1         = &one,
564                 .extra2         = &one,
565         },
566 #endif
567 #ifdef CONFIG_HOTPLUG
568         {
569                 .procname       = "hotplug",
570                 .data           = &uevent_helper,
571                 .maxlen         = UEVENT_HELPER_PATH_LEN,
572                 .mode           = 0644,
573                 .proc_handler   = proc_dostring,
574         },
575 #endif
576 #ifdef CONFIG_CHR_DEV_SG
577         {
578                 .procname       = "sg-big-buff",
579                 .data           = &sg_big_buff,
580                 .maxlen         = sizeof (int),
581                 .mode           = 0444,
582                 .proc_handler   = proc_dointvec,
583         },
584 #endif
585 #ifdef CONFIG_BSD_PROCESS_ACCT
586         {
587                 .procname       = "acct",
588                 .data           = &acct_parm,
589                 .maxlen         = 3*sizeof(int),
590                 .mode           = 0644,
591                 .proc_handler   = proc_dointvec,
592         },
593 #endif
594 #ifdef CONFIG_MAGIC_SYSRQ
595         {
596                 .procname       = "sysrq",
597                 .data           = &__sysrq_enabled,
598                 .maxlen         = sizeof (int),
599                 .mode           = 0644,
600                 .proc_handler   = sysrq_sysctl_handler,
601         },
602 #endif
603 #ifdef CONFIG_PROC_SYSCTL
604         {
605                 .procname       = "cad_pid",
606                 .data           = NULL,
607                 .maxlen         = sizeof (int),
608                 .mode           = 0600,
609                 .proc_handler   = proc_do_cad_pid,
610         },
611 #endif
612         {
613                 .procname       = "threads-max",
614                 .data           = &max_threads,
615                 .maxlen         = sizeof(int),
616                 .mode           = 0644,
617                 .proc_handler   = proc_dointvec,
618         },
619         {
620                 .procname       = "random",
621                 .mode           = 0555,
622                 .child          = random_table,
623         },
624         {
625                 .procname       = "usermodehelper",
626                 .mode           = 0555,
627                 .child          = usermodehelper_table,
628         },
629         {
630                 .procname       = "overflowuid",
631                 .data           = &overflowuid,
632                 .maxlen         = sizeof(int),
633                 .mode           = 0644,
634                 .proc_handler   = proc_dointvec_minmax,
635                 .extra1         = &minolduid,
636                 .extra2         = &maxolduid,
637         },
638         {
639                 .procname       = "overflowgid",
640                 .data           = &overflowgid,
641                 .maxlen         = sizeof(int),
642                 .mode           = 0644,
643                 .proc_handler   = proc_dointvec_minmax,
644                 .extra1         = &minolduid,
645                 .extra2         = &maxolduid,
646         },
647 #ifdef CONFIG_S390
648 #ifdef CONFIG_MATHEMU
649         {
650                 .procname       = "ieee_emulation_warnings",
651                 .data           = &sysctl_ieee_emulation_warnings,
652                 .maxlen         = sizeof(int),
653                 .mode           = 0644,
654                 .proc_handler   = proc_dointvec,
655         },
656 #endif
657         {
658                 .procname       = "userprocess_debug",
659                 .data           = &show_unhandled_signals,
660                 .maxlen         = sizeof(int),
661                 .mode           = 0644,
662                 .proc_handler   = proc_dointvec,
663         },
664 #endif
665         {
666                 .procname       = "pid_max",
667                 .data           = &pid_max,
668                 .maxlen         = sizeof (int),
669                 .mode           = 0644,
670                 .proc_handler   = proc_dointvec_minmax,
671                 .extra1         = &pid_max_min,
672                 .extra2         = &pid_max_max,
673         },
674         {
675                 .procname       = "panic_on_oops",
676                 .data           = &panic_on_oops,
677                 .maxlen         = sizeof(int),
678                 .mode           = 0644,
679                 .proc_handler   = proc_dointvec,
680         },
681 #if defined CONFIG_PRINTK
682         {
683                 .procname       = "printk",
684                 .data           = &console_loglevel,
685                 .maxlen         = 4*sizeof(int),
686                 .mode           = 0644,
687                 .proc_handler   = proc_dointvec,
688         },
689         {
690                 .procname       = "printk_ratelimit",
691                 .data           = &printk_ratelimit_state.interval,
692                 .maxlen         = sizeof(int),
693                 .mode           = 0644,
694                 .proc_handler   = proc_dointvec_jiffies,
695         },
696         {
697                 .procname       = "printk_ratelimit_burst",
698                 .data           = &printk_ratelimit_state.burst,
699                 .maxlen         = sizeof(int),
700                 .mode           = 0644,
701                 .proc_handler   = proc_dointvec,
702         },
703         {
704                 .procname       = "printk_delay",
705                 .data           = &printk_delay_msec,
706                 .maxlen         = sizeof(int),
707                 .mode           = 0644,
708                 .proc_handler   = proc_dointvec_minmax,
709                 .extra1         = &zero,
710                 .extra2         = &ten_thousand,
711         },
712         {
713                 .procname       = "dmesg_restrict",
714                 .data           = &dmesg_restrict,
715                 .maxlen         = sizeof(int),
716                 .mode           = 0644,
717                 .proc_handler   = proc_dointvec_minmax,
718                 .extra1         = &zero,
719                 .extra2         = &one,
720         },
721         {
722                 .procname       = "kptr_restrict",
723                 .data           = &kptr_restrict,
724                 .maxlen         = sizeof(int),
725                 .mode           = 0644,
726                 .proc_handler   = proc_dmesg_restrict,
727                 .extra1         = &zero,
728                 .extra2         = &two,
729         },
730 #endif
731         {
732                 .procname       = "ngroups_max",
733                 .data           = &ngroups_max,
734                 .maxlen         = sizeof (int),
735                 .mode           = 0444,
736                 .proc_handler   = proc_dointvec,
737         },
738         {
739                 .procname       = "cap_last_cap",
740                 .data           = (void *)&cap_last_cap,
741                 .maxlen         = sizeof(int),
742                 .mode           = 0444,
743                 .proc_handler   = proc_dointvec,
744         },
745 #if defined(CONFIG_LOCKUP_DETECTOR)
746         {
747                 .procname       = "watchdog",
748                 .data           = &watchdog_enabled,
749                 .maxlen         = sizeof (int),
750                 .mode           = 0644,
751                 .proc_handler   = proc_dowatchdog,
752                 .extra1         = &zero,
753                 .extra2         = &one,
754         },
755         {
756                 .procname       = "watchdog_thresh",
757                 .data           = &watchdog_thresh,
758                 .maxlen         = sizeof(int),
759                 .mode           = 0644,
760                 .proc_handler   = proc_dowatchdog,
761                 .extra1         = &neg_one,
762                 .extra2         = &sixty,
763         },
764         {
765                 .procname       = "softlockup_panic",
766                 .data           = &softlockup_panic,
767                 .maxlen         = sizeof(int),
768                 .mode           = 0644,
769                 .proc_handler   = proc_dointvec_minmax,
770                 .extra1         = &zero,
771                 .extra2         = &one,
772         },
773         {
774                 .procname       = "nmi_watchdog",
775                 .data           = &watchdog_enabled,
776                 .maxlen         = sizeof (int),
777                 .mode           = 0644,
778                 .proc_handler   = proc_dowatchdog,
779                 .extra1         = &zero,
780                 .extra2         = &one,
781         },
782 #endif
783 #if defined(CONFIG_X86_LOCAL_APIC) && defined(CONFIG_X86)
784         {
785                 .procname       = "unknown_nmi_panic",
786                 .data           = &unknown_nmi_panic,
787                 .maxlen         = sizeof (int),
788                 .mode           = 0644,
789                 .proc_handler   = proc_dointvec,
790         },
791 #endif
792 #if defined(CONFIG_X86)
793         {
794                 .procname       = "panic_on_unrecovered_nmi",
795                 .data           = &panic_on_unrecovered_nmi,
796                 .maxlen         = sizeof(int),
797                 .mode           = 0644,
798                 .proc_handler   = proc_dointvec,
799         },
800         {
801                 .procname       = "panic_on_io_nmi",
802                 .data           = &panic_on_io_nmi,
803                 .maxlen         = sizeof(int),
804                 .mode           = 0644,
805                 .proc_handler   = proc_dointvec,
806         },
807 #ifdef CONFIG_DEBUG_STACKOVERFLOW
808         {
809                 .procname       = "panic_on_stackoverflow",
810                 .data           = &sysctl_panic_on_stackoverflow,
811                 .maxlen         = sizeof(int),
812                 .mode           = 0644,
813                 .proc_handler   = proc_dointvec,
814         },
815 #endif
816         {
817                 .procname       = "bootloader_type",
818                 .data           = &bootloader_type,
819                 .maxlen         = sizeof (int),
820                 .mode           = 0444,
821                 .proc_handler   = proc_dointvec,
822         },
823         {
824                 .procname       = "bootloader_version",
825                 .data           = &bootloader_version,
826                 .maxlen         = sizeof (int),
827                 .mode           = 0444,
828                 .proc_handler   = proc_dointvec,
829         },
830         {
831                 .procname       = "kstack_depth_to_print",
832                 .data           = &kstack_depth_to_print,
833                 .maxlen         = sizeof(int),
834                 .mode           = 0644,
835                 .proc_handler   = proc_dointvec,
836         },
837         {
838                 .procname       = "io_delay_type",
839                 .data           = &io_delay_type,
840                 .maxlen         = sizeof(int),
841                 .mode           = 0644,
842                 .proc_handler   = proc_dointvec,
843         },
844 #endif
845 #if defined(CONFIG_MMU)
846         {
847                 .procname       = "randomize_va_space",
848                 .data           = &randomize_va_space,
849                 .maxlen         = sizeof(int),
850                 .mode           = 0644,
851                 .proc_handler   = proc_dointvec,
852         },
853 #endif
854 #if defined(CONFIG_S390) && defined(CONFIG_SMP)
855         {
856                 .procname       = "spin_retry",
857                 .data           = &spin_retry,
858                 .maxlen         = sizeof (int),
859                 .mode           = 0644,
860                 .proc_handler   = proc_dointvec,
861         },
862 #endif
863 #if     defined(CONFIG_ACPI_SLEEP) && defined(CONFIG_X86)
864         {
865                 .procname       = "acpi_video_flags",
866                 .data           = &acpi_realmode_flags,
867                 .maxlen         = sizeof (unsigned long),
868                 .mode           = 0644,
869                 .proc_handler   = proc_doulongvec_minmax,
870         },
871 #endif
872 #ifdef CONFIG_IA64
873         {
874                 .procname       = "ignore-unaligned-usertrap",
875                 .data           = &no_unaligned_warning,
876                 .maxlen         = sizeof (int),
877                 .mode           = 0644,
878                 .proc_handler   = proc_dointvec,
879         },
880         {
881                 .procname       = "unaligned-dump-stack",
882                 .data           = &unaligned_dump_stack,
883                 .maxlen         = sizeof (int),
884                 .mode           = 0644,
885                 .proc_handler   = proc_dointvec,
886         },
887 #endif
888 #ifdef CONFIG_DETECT_HUNG_TASK
889         {
890                 .procname       = "hung_task_panic",
891                 .data           = &sysctl_hung_task_panic,
892                 .maxlen         = sizeof(int),
893                 .mode           = 0644,
894                 .proc_handler   = proc_dointvec_minmax,
895                 .extra1         = &zero,
896                 .extra2         = &one,
897         },
898         {
899                 .procname       = "hung_task_check_count",
900                 .data           = &sysctl_hung_task_check_count,
901                 .maxlen         = sizeof(unsigned long),
902                 .mode           = 0644,
903                 .proc_handler   = proc_doulongvec_minmax,
904         },
905         {
906                 .procname       = "hung_task_timeout_secs",
907                 .data           = &sysctl_hung_task_timeout_secs,
908                 .maxlen         = sizeof(unsigned long),
909                 .mode           = 0644,
910                 .proc_handler   = proc_dohung_task_timeout_secs,
911         },
912         {
913                 .procname       = "hung_task_warnings",
914                 .data           = &sysctl_hung_task_warnings,
915                 .maxlen         = sizeof(unsigned long),
916                 .mode           = 0644,
917                 .proc_handler   = proc_doulongvec_minmax,
918         },
919 #endif
920 #ifdef CONFIG_COMPAT
921         {
922                 .procname       = "compat-log",
923                 .data           = &compat_log,
924                 .maxlen         = sizeof (int),
925                 .mode           = 0644,
926                 .proc_handler   = proc_dointvec,
927         },
928 #endif
929 #ifdef CONFIG_RT_MUTEXES
930         {
931                 .procname       = "max_lock_depth",
932                 .data           = &max_lock_depth,
933                 .maxlen         = sizeof(int),
934                 .mode           = 0644,
935                 .proc_handler   = proc_dointvec,
936         },
937 #endif
938         {
939                 .procname       = "poweroff_cmd",
940                 .data           = &poweroff_cmd,
941                 .maxlen         = POWEROFF_CMD_PATH_LEN,
942                 .mode           = 0644,
943                 .proc_handler   = proc_dostring,
944         },
945 #ifdef CONFIG_KEYS
946         {
947                 .procname       = "keys",
948                 .mode           = 0555,
949                 .child          = key_sysctls,
950         },
951 #endif
952 #ifdef CONFIG_RCU_TORTURE_TEST
953         {
954                 .procname       = "rcutorture_runnable",
955                 .data           = &rcutorture_runnable,
956                 .maxlen         = sizeof(int),
957                 .mode           = 0644,
958                 .proc_handler   = proc_dointvec,
959         },
960 #endif
961 #ifdef CONFIG_PERF_EVENTS
962         /*
963          * User-space scripts rely on the existence of this file
964          * as a feature check for perf_events being enabled.
965          *
966          * So it's an ABI, do not remove!
967          */
968         {
969                 .procname       = "perf_event_paranoid",
970                 .data           = &sysctl_perf_event_paranoid,
971                 .maxlen         = sizeof(sysctl_perf_event_paranoid),
972                 .mode           = 0644,
973                 .proc_handler   = proc_dointvec,
974         },
975         {
976                 .procname       = "perf_event_mlock_kb",
977                 .data           = &sysctl_perf_event_mlock,
978                 .maxlen         = sizeof(sysctl_perf_event_mlock),
979                 .mode           = 0644,
980                 .proc_handler   = proc_dointvec,
981         },
982         {
983                 .procname       = "perf_event_max_sample_rate",
984                 .data           = &sysctl_perf_event_sample_rate,
985                 .maxlen         = sizeof(sysctl_perf_event_sample_rate),
986                 .mode           = 0644,
987                 .proc_handler   = perf_proc_update_handler,
988         },
989 #endif
990 #ifdef CONFIG_KMEMCHECK
991         {
992                 .procname       = "kmemcheck",
993                 .data           = &kmemcheck_enabled,
994                 .maxlen         = sizeof(int),
995                 .mode           = 0644,
996                 .proc_handler   = proc_dointvec,
997         },
998 #endif
999 #ifdef CONFIG_BLOCK
1000         {
1001                 .procname       = "blk_iopoll",
1002                 .data           = &blk_iopoll_enabled,
1003                 .maxlen         = sizeof(int),
1004                 .mode           = 0644,
1005                 .proc_handler   = proc_dointvec,
1006         },
1007 #endif
1008         { }
1009 };
1010
1011 static struct ctl_table vm_table[] = {
1012         {
1013                 .procname       = "overcommit_memory",
1014                 .data           = &sysctl_overcommit_memory,
1015                 .maxlen         = sizeof(sysctl_overcommit_memory),
1016                 .mode           = 0644,
1017                 .proc_handler   = proc_dointvec_minmax,
1018                 .extra1         = &zero,
1019                 .extra2         = &two,
1020         },
1021         {
1022                 .procname       = "panic_on_oom",
1023                 .data           = &sysctl_panic_on_oom,
1024                 .maxlen         = sizeof(sysctl_panic_on_oom),
1025                 .mode           = 0644,
1026                 .proc_handler   = proc_dointvec_minmax,
1027                 .extra1         = &zero,
1028                 .extra2         = &two,
1029         },
1030         {
1031                 .procname       = "oom_kill_allocating_task",
1032                 .data           = &sysctl_oom_kill_allocating_task,
1033                 .maxlen         = sizeof(sysctl_oom_kill_allocating_task),
1034                 .mode           = 0644,
1035                 .proc_handler   = proc_dointvec,
1036         },
1037         {
1038                 .procname       = "oom_dump_tasks",
1039                 .data           = &sysctl_oom_dump_tasks,
1040                 .maxlen         = sizeof(sysctl_oom_dump_tasks),
1041                 .mode           = 0644,
1042                 .proc_handler   = proc_dointvec,
1043         },
1044         {
1045                 .procname       = "overcommit_ratio",
1046                 .data           = &sysctl_overcommit_ratio,
1047                 .maxlen         = sizeof(sysctl_overcommit_ratio),
1048                 .mode           = 0644,
1049                 .proc_handler   = proc_dointvec,
1050         },
1051         {
1052                 .procname       = "page-cluster", 
1053                 .data           = &page_cluster,
1054                 .maxlen         = sizeof(int),
1055                 .mode           = 0644,
1056                 .proc_handler   = proc_dointvec_minmax,
1057                 .extra1         = &zero,
1058         },
1059         {
1060                 .procname       = "dirty_background_ratio",
1061                 .data           = &dirty_background_ratio,
1062                 .maxlen         = sizeof(dirty_background_ratio),
1063                 .mode           = 0644,
1064                 .proc_handler   = dirty_background_ratio_handler,
1065                 .extra1         = &zero,
1066                 .extra2         = &one_hundred,
1067         },
1068         {
1069                 .procname       = "dirty_background_bytes",
1070                 .data           = &dirty_background_bytes,
1071                 .maxlen         = sizeof(dirty_background_bytes),
1072                 .mode           = 0644,
1073                 .proc_handler   = dirty_background_bytes_handler,
1074                 .extra1         = &one_ul,
1075         },
1076         {
1077                 .procname       = "dirty_ratio",
1078                 .data           = &vm_dirty_ratio,
1079                 .maxlen         = sizeof(vm_dirty_ratio),
1080                 .mode           = 0644,
1081                 .proc_handler   = dirty_ratio_handler,
1082                 .extra1         = &zero,
1083                 .extra2         = &one_hundred,
1084         },
1085         {
1086                 .procname       = "dirty_bytes",
1087                 .data           = &vm_dirty_bytes,
1088                 .maxlen         = sizeof(vm_dirty_bytes),
1089                 .mode           = 0644,
1090                 .proc_handler   = dirty_bytes_handler,
1091                 .extra1         = &dirty_bytes_min,
1092         },
1093         {
1094                 .procname       = "dirty_writeback_centisecs",
1095                 .data           = &dirty_writeback_interval,
1096                 .maxlen         = sizeof(dirty_writeback_interval),
1097                 .mode           = 0644,
1098                 .proc_handler   = dirty_writeback_centisecs_handler,
1099         },
1100         {
1101                 .procname       = "dirty_expire_centisecs",
1102                 .data           = &dirty_expire_interval,
1103                 .maxlen         = sizeof(dirty_expire_interval),
1104                 .mode           = 0644,
1105                 .proc_handler   = proc_dointvec_minmax,
1106                 .extra1         = &zero,
1107         },
1108         {
1109                 .procname       = "nr_pdflush_threads",
1110                 .data           = &nr_pdflush_threads,
1111                 .maxlen         = sizeof nr_pdflush_threads,
1112                 .mode           = 0444 /* read-only*/,
1113                 .proc_handler   = proc_dointvec,
1114         },
1115         {
1116                 .procname       = "swappiness",
1117                 .data           = &vm_swappiness,
1118                 .maxlen         = sizeof(vm_swappiness),
1119                 .mode           = 0644,
1120                 .proc_handler   = proc_dointvec_minmax,
1121                 .extra1         = &zero,
1122                 .extra2         = &one_hundred,
1123         },
1124 #ifdef CONFIG_HUGETLB_PAGE
1125         {
1126                 .procname       = "nr_hugepages",
1127                 .data           = NULL,
1128                 .maxlen         = sizeof(unsigned long),
1129                 .mode           = 0644,
1130                 .proc_handler   = hugetlb_sysctl_handler,
1131                 .extra1         = (void *)&hugetlb_zero,
1132                 .extra2         = (void *)&hugetlb_infinity,
1133         },
1134 #ifdef CONFIG_NUMA
1135         {
1136                 .procname       = "nr_hugepages_mempolicy",
1137                 .data           = NULL,
1138                 .maxlen         = sizeof(unsigned long),
1139                 .mode           = 0644,
1140                 .proc_handler   = &hugetlb_mempolicy_sysctl_handler,
1141                 .extra1         = (void *)&hugetlb_zero,
1142                 .extra2         = (void *)&hugetlb_infinity,
1143         },
1144 #endif
1145          {
1146                 .procname       = "hugetlb_shm_group",
1147                 .data           = &sysctl_hugetlb_shm_group,
1148                 .maxlen         = sizeof(gid_t),
1149                 .mode           = 0644,
1150                 .proc_handler   = proc_dointvec,
1151          },
1152          {
1153                 .procname       = "hugepages_treat_as_movable",
1154                 .data           = &hugepages_treat_as_movable,
1155                 .maxlen         = sizeof(int),
1156                 .mode           = 0644,
1157                 .proc_handler   = hugetlb_treat_movable_handler,
1158         },
1159         {
1160                 .procname       = "nr_overcommit_hugepages",
1161                 .data           = NULL,
1162                 .maxlen         = sizeof(unsigned long),
1163                 .mode           = 0644,
1164                 .proc_handler   = hugetlb_overcommit_handler,
1165                 .extra1         = (void *)&hugetlb_zero,
1166                 .extra2         = (void *)&hugetlb_infinity,
1167         },
1168 #endif
1169         {
1170                 .procname       = "lowmem_reserve_ratio",
1171                 .data           = &sysctl_lowmem_reserve_ratio,
1172                 .maxlen         = sizeof(sysctl_lowmem_reserve_ratio),
1173                 .mode           = 0644,
1174                 .proc_handler   = lowmem_reserve_ratio_sysctl_handler,
1175         },
1176         {
1177                 .procname       = "drop_caches",
1178                 .data           = &sysctl_drop_caches,
1179                 .maxlen         = sizeof(int),
1180                 .mode           = 0644,
1181                 .proc_handler   = drop_caches_sysctl_handler,
1182                 .extra1         = &one,
1183                 .extra2         = &three,
1184         },
1185 #ifdef CONFIG_COMPACTION
1186         {
1187                 .procname       = "compact_memory",
1188                 .data           = &sysctl_compact_memory,
1189                 .maxlen         = sizeof(int),
1190                 .mode           = 0200,
1191                 .proc_handler   = sysctl_compaction_handler,
1192         },
1193         {
1194                 .procname       = "extfrag_threshold",
1195                 .data           = &sysctl_extfrag_threshold,
1196                 .maxlen         = sizeof(int),
1197                 .mode           = 0644,
1198                 .proc_handler   = sysctl_extfrag_handler,
1199                 .extra1         = &min_extfrag_threshold,
1200                 .extra2         = &max_extfrag_threshold,
1201         },
1202
1203 #endif /* CONFIG_COMPACTION */
1204         {
1205                 .procname       = "min_free_kbytes",
1206                 .data           = &min_free_kbytes,
1207                 .maxlen         = sizeof(min_free_kbytes),
1208                 .mode           = 0644,
1209                 .proc_handler   = min_free_kbytes_sysctl_handler,
1210                 .extra1         = &zero,
1211         },
1212         {
1213                 .procname       = "percpu_pagelist_fraction",
1214                 .data           = &percpu_pagelist_fraction,
1215                 .maxlen         = sizeof(percpu_pagelist_fraction),
1216                 .mode           = 0644,
1217                 .proc_handler   = percpu_pagelist_fraction_sysctl_handler,
1218                 .extra1         = &min_percpu_pagelist_fract,
1219         },
1220 #ifdef CONFIG_MMU
1221         {
1222                 .procname       = "max_map_count",
1223                 .data           = &sysctl_max_map_count,
1224                 .maxlen         = sizeof(sysctl_max_map_count),
1225                 .mode           = 0644,
1226                 .proc_handler   = proc_dointvec_minmax,
1227                 .extra1         = &zero,
1228         },
1229 #else
1230         {
1231                 .procname       = "nr_trim_pages",
1232                 .data           = &sysctl_nr_trim_pages,
1233                 .maxlen         = sizeof(sysctl_nr_trim_pages),
1234                 .mode           = 0644,
1235                 .proc_handler   = proc_dointvec_minmax,
1236                 .extra1         = &zero,
1237         },
1238 #endif
1239         {
1240                 .procname       = "laptop_mode",
1241                 .data           = &laptop_mode,
1242                 .maxlen         = sizeof(laptop_mode),
1243                 .mode           = 0644,
1244                 .proc_handler   = proc_dointvec_jiffies,
1245         },
1246         {
1247                 .procname       = "block_dump",
1248                 .data           = &block_dump,
1249                 .maxlen         = sizeof(block_dump),
1250                 .mode           = 0644,
1251                 .proc_handler   = proc_dointvec,
1252                 .extra1         = &zero,
1253         },
1254         {
1255                 .procname       = "vfs_cache_pressure",
1256                 .data           = &sysctl_vfs_cache_pressure,
1257                 .maxlen         = sizeof(sysctl_vfs_cache_pressure),
1258                 .mode           = 0644,
1259                 .proc_handler   = proc_dointvec,
1260                 .extra1         = &zero,
1261         },
1262 #ifdef HAVE_ARCH_PICK_MMAP_LAYOUT
1263         {
1264                 .procname       = "legacy_va_layout",
1265                 .data           = &sysctl_legacy_va_layout,
1266                 .maxlen         = sizeof(sysctl_legacy_va_layout),
1267                 .mode           = 0644,
1268                 .proc_handler   = proc_dointvec,
1269                 .extra1         = &zero,
1270         },
1271 #endif
1272 #ifdef CONFIG_NUMA
1273         {
1274                 .procname       = "zone_reclaim_mode",
1275                 .data           = &zone_reclaim_mode,
1276                 .maxlen         = sizeof(zone_reclaim_mode),
1277                 .mode           = 0644,
1278                 .proc_handler   = proc_dointvec,
1279                 .extra1         = &zero,
1280         },
1281         {
1282                 .procname       = "min_unmapped_ratio",
1283                 .data           = &sysctl_min_unmapped_ratio,
1284                 .maxlen         = sizeof(sysctl_min_unmapped_ratio),
1285                 .mode           = 0644,
1286                 .proc_handler   = sysctl_min_unmapped_ratio_sysctl_handler,
1287                 .extra1         = &zero,
1288                 .extra2         = &one_hundred,
1289         },
1290         {
1291                 .procname       = "min_slab_ratio",
1292                 .data           = &sysctl_min_slab_ratio,
1293                 .maxlen         = sizeof(sysctl_min_slab_ratio),
1294                 .mode           = 0644,
1295                 .proc_handler   = sysctl_min_slab_ratio_sysctl_handler,
1296                 .extra1         = &zero,
1297                 .extra2         = &one_hundred,
1298         },
1299 #endif
1300 #ifdef CONFIG_SMP
1301         {
1302                 .procname       = "stat_interval",
1303                 .data           = &sysctl_stat_interval,
1304                 .maxlen         = sizeof(sysctl_stat_interval),
1305                 .mode           = 0644,
1306                 .proc_handler   = proc_dointvec_jiffies,
1307         },
1308 #endif
1309 #ifdef CONFIG_MMU
1310         {
1311                 .procname       = "mmap_min_addr",
1312                 .data           = &dac_mmap_min_addr,
1313                 .maxlen         = sizeof(unsigned long),
1314                 .mode           = 0644,
1315                 .proc_handler   = mmap_min_addr_handler,
1316         },
1317 #endif
1318 #ifdef CONFIG_NUMA
1319         {
1320                 .procname       = "numa_zonelist_order",
1321                 .data           = &numa_zonelist_order,
1322                 .maxlen         = NUMA_ZONELIST_ORDER_LEN,
1323                 .mode           = 0644,
1324                 .proc_handler   = numa_zonelist_order_handler,
1325         },
1326 #endif
1327 #if (defined(CONFIG_X86_32) && !defined(CONFIG_UML))|| \
1328    (defined(CONFIG_SUPERH) && defined(CONFIG_VSYSCALL))
1329         {
1330                 .procname       = "vdso_enabled",
1331                 .data           = &vdso_enabled,
1332                 .maxlen         = sizeof(vdso_enabled),
1333                 .mode           = 0644,
1334                 .proc_handler   = proc_dointvec,
1335                 .extra1         = &zero,
1336         },
1337 #endif
1338 #ifdef CONFIG_HIGHMEM
1339         {
1340                 .procname       = "highmem_is_dirtyable",
1341                 .data           = &vm_highmem_is_dirtyable,
1342                 .maxlen         = sizeof(vm_highmem_is_dirtyable),
1343                 .mode           = 0644,
1344                 .proc_handler   = proc_dointvec_minmax,
1345                 .extra1         = &zero,
1346                 .extra2         = &one,
1347         },
1348 #endif
1349         {
1350                 .procname       = "scan_unevictable_pages",
1351                 .data           = &scan_unevictable_pages,
1352                 .maxlen         = sizeof(scan_unevictable_pages),
1353                 .mode           = 0644,
1354                 .proc_handler   = scan_unevictable_handler,
1355         },
1356 #ifdef CONFIG_MEMORY_FAILURE
1357         {
1358                 .procname       = "memory_failure_early_kill",
1359                 .data           = &sysctl_memory_failure_early_kill,
1360                 .maxlen         = sizeof(sysctl_memory_failure_early_kill),
1361                 .mode           = 0644,
1362                 .proc_handler   = proc_dointvec_minmax,
1363                 .extra1         = &zero,
1364                 .extra2         = &one,
1365         },
1366         {
1367                 .procname       = "memory_failure_recovery",
1368                 .data           = &sysctl_memory_failure_recovery,
1369                 .maxlen         = sizeof(sysctl_memory_failure_recovery),
1370                 .mode           = 0644,
1371                 .proc_handler   = proc_dointvec_minmax,
1372                 .extra1         = &zero,
1373                 .extra2         = &one,
1374         },
1375 #endif
1376         { }
1377 };
1378
1379 #if defined(CONFIG_BINFMT_MISC) || defined(CONFIG_BINFMT_MISC_MODULE)
1380 static struct ctl_table binfmt_misc_table[] = {
1381         { }
1382 };
1383 #endif
1384
1385 static struct ctl_table fs_table[] = {
1386         {
1387                 .procname       = "inode-nr",
1388                 .data           = &inodes_stat,
1389                 .maxlen         = 2*sizeof(int),
1390                 .mode           = 0444,
1391                 .proc_handler   = proc_nr_inodes,
1392         },
1393         {
1394                 .procname       = "inode-state",
1395                 .data           = &inodes_stat,
1396                 .maxlen         = 7*sizeof(int),
1397                 .mode           = 0444,
1398                 .proc_handler   = proc_nr_inodes,
1399         },
1400         {
1401                 .procname       = "file-nr",
1402                 .data           = &files_stat,
1403                 .maxlen         = sizeof(files_stat),
1404                 .mode           = 0444,
1405                 .proc_handler   = proc_nr_files,
1406         },
1407         {
1408                 .procname       = "file-max",
1409                 .data           = &files_stat.max_files,
1410                 .maxlen         = sizeof(files_stat.max_files),
1411                 .mode           = 0644,
1412                 .proc_handler   = proc_doulongvec_minmax,
1413         },
1414         {
1415                 .procname       = "nr_open",
1416                 .data           = &sysctl_nr_open,
1417                 .maxlen         = sizeof(int),
1418                 .mode           = 0644,
1419                 .proc_handler   = proc_dointvec_minmax,
1420                 .extra1         = &sysctl_nr_open_min,
1421                 .extra2         = &sysctl_nr_open_max,
1422         },
1423         {
1424                 .procname       = "dentry-state",
1425                 .data           = &dentry_stat,
1426                 .maxlen         = 6*sizeof(int),
1427                 .mode           = 0444,
1428                 .proc_handler   = proc_nr_dentry,
1429         },
1430         {
1431                 .procname       = "overflowuid",
1432                 .data           = &fs_overflowuid,
1433                 .maxlen         = sizeof(int),
1434                 .mode           = 0644,
1435                 .proc_handler   = proc_dointvec_minmax,
1436                 .extra1         = &minolduid,
1437                 .extra2         = &maxolduid,
1438         },
1439         {
1440                 .procname       = "overflowgid",
1441                 .data           = &fs_overflowgid,
1442                 .maxlen         = sizeof(int),
1443                 .mode           = 0644,
1444                 .proc_handler   = proc_dointvec_minmax,
1445                 .extra1         = &minolduid,
1446                 .extra2         = &maxolduid,
1447         },
1448 #ifdef CONFIG_FILE_LOCKING
1449         {
1450                 .procname       = "leases-enable",
1451                 .data           = &leases_enable,
1452                 .maxlen         = sizeof(int),
1453                 .mode           = 0644,
1454                 .proc_handler   = proc_dointvec,
1455         },
1456 #endif
1457 #ifdef CONFIG_DNOTIFY
1458         {
1459                 .procname       = "dir-notify-enable",
1460                 .data           = &dir_notify_enable,
1461                 .maxlen         = sizeof(int),
1462                 .mode           = 0644,
1463                 .proc_handler   = proc_dointvec,
1464         },
1465 #endif
1466 #ifdef CONFIG_MMU
1467 #ifdef CONFIG_FILE_LOCKING
1468         {
1469                 .procname       = "lease-break-time",
1470                 .data           = &lease_break_time,
1471                 .maxlen         = sizeof(int),
1472                 .mode           = 0644,
1473                 .proc_handler   = proc_dointvec,
1474         },
1475 #endif
1476 #ifdef CONFIG_AIO
1477         {
1478                 .procname       = "aio-nr",
1479                 .data           = &aio_nr,
1480                 .maxlen         = sizeof(aio_nr),
1481                 .mode           = 0444,
1482                 .proc_handler   = proc_doulongvec_minmax,
1483         },
1484         {
1485                 .procname       = "aio-max-nr",
1486                 .data           = &aio_max_nr,
1487                 .maxlen         = sizeof(aio_max_nr),
1488                 .mode           = 0644,
1489                 .proc_handler   = proc_doulongvec_minmax,
1490         },
1491 #endif /* CONFIG_AIO */
1492 #ifdef CONFIG_INOTIFY_USER
1493         {
1494                 .procname       = "inotify",
1495                 .mode           = 0555,
1496                 .child          = inotify_table,
1497         },
1498 #endif  
1499 #ifdef CONFIG_EPOLL
1500         {
1501                 .procname       = "epoll",
1502                 .mode           = 0555,
1503                 .child          = epoll_table,
1504         },
1505 #endif
1506 #endif
1507         {
1508                 .procname       = "suid_dumpable",
1509                 .data           = &suid_dumpable,
1510                 .maxlen         = sizeof(int),
1511                 .mode           = 0644,
1512                 .proc_handler   = proc_dointvec_minmax,
1513                 .extra1         = &zero,
1514                 .extra2         = &two,
1515         },
1516 #if defined(CONFIG_BINFMT_MISC) || defined(CONFIG_BINFMT_MISC_MODULE)
1517         {
1518                 .procname       = "binfmt_misc",
1519                 .mode           = 0555,
1520                 .child          = binfmt_misc_table,
1521         },
1522 #endif
1523         {
1524                 .procname       = "pipe-max-size",
1525                 .data           = &pipe_max_size,
1526                 .maxlen         = sizeof(int),
1527                 .mode           = 0644,
1528                 .proc_handler   = &pipe_proc_fn,
1529                 .extra1         = &pipe_min_size,
1530         },
1531         { }
1532 };
1533
1534 static struct ctl_table debug_table[] = {
1535 #if defined(CONFIG_X86) || defined(CONFIG_PPC) || defined(CONFIG_SPARC) || \
1536     defined(CONFIG_S390) || defined(CONFIG_TILE)
1537         {
1538                 .procname       = "exception-trace",
1539                 .data           = &show_unhandled_signals,
1540                 .maxlen         = sizeof(int),
1541                 .mode           = 0644,
1542                 .proc_handler   = proc_dointvec
1543         },
1544 #endif
1545 #if defined(CONFIG_OPTPROBES)
1546         {
1547                 .procname       = "kprobes-optimization",
1548                 .data           = &sysctl_kprobes_optimization,
1549                 .maxlen         = sizeof(int),
1550                 .mode           = 0644,
1551                 .proc_handler   = proc_kprobes_optimization_handler,
1552                 .extra1         = &zero,
1553                 .extra2         = &one,
1554         },
1555 #endif
1556         { }
1557 };
1558
1559 static struct ctl_table dev_table[] = {
1560         { }
1561 };
1562
1563 static DEFINE_SPINLOCK(sysctl_lock);
1564
1565 /* called under sysctl_lock */
1566 static int use_table(struct ctl_table_header *p)
1567 {
1568         if (unlikely(p->unregistering))
1569                 return 0;
1570         p->used++;
1571         return 1;
1572 }
1573
1574 /* called under sysctl_lock */
1575 static void unuse_table(struct ctl_table_header *p)
1576 {
1577         if (!--p->used)
1578                 if (unlikely(p->unregistering))
1579                         complete(p->unregistering);
1580 }
1581
1582 /* called under sysctl_lock, will reacquire if has to wait */
1583 static void start_unregistering(struct ctl_table_header *p)
1584 {
1585         /*
1586          * if p->used is 0, nobody will ever touch that entry again;
1587          * we'll eliminate all paths to it before dropping sysctl_lock
1588          */
1589         if (unlikely(p->used)) {
1590                 struct completion wait;
1591                 init_completion(&wait);
1592                 p->unregistering = &wait;
1593                 spin_unlock(&sysctl_lock);
1594                 wait_for_completion(&wait);
1595                 spin_lock(&sysctl_lock);
1596         } else {
1597                 /* anything non-NULL; we'll never dereference it */
1598                 p->unregistering = ERR_PTR(-EINVAL);
1599         }
1600         /*
1601          * do not remove from the list until nobody holds it; walking the
1602          * list in do_sysctl() relies on that.
1603          */
1604         list_del_init(&p->ctl_entry);
1605 }
1606
1607 void sysctl_head_get(struct ctl_table_header *head)
1608 {
1609         spin_lock(&sysctl_lock);
1610         head->count++;
1611         spin_unlock(&sysctl_lock);
1612 }
1613
1614 void sysctl_head_put(struct ctl_table_header *head)
1615 {
1616         spin_lock(&sysctl_lock);
1617         if (!--head->count)
1618                 kfree_rcu(head, rcu);
1619         spin_unlock(&sysctl_lock);
1620 }
1621
1622 struct ctl_table_header *sysctl_head_grab(struct ctl_table_header *head)
1623 {
1624         if (!head)
1625                 BUG();
1626         spin_lock(&sysctl_lock);
1627         if (!use_table(head))
1628                 head = ERR_PTR(-ENOENT);
1629         spin_unlock(&sysctl_lock);
1630         return head;
1631 }
1632
1633 void sysctl_head_finish(struct ctl_table_header *head)
1634 {
1635         if (!head)
1636                 return;
1637         spin_lock(&sysctl_lock);
1638         unuse_table(head);
1639         spin_unlock(&sysctl_lock);
1640 }
1641
1642 static struct ctl_table_set *
1643 lookup_header_set(struct ctl_table_root *root, struct nsproxy *namespaces)
1644 {
1645         struct ctl_table_set *set = &root->default_set;
1646         if (root->lookup)
1647                 set = root->lookup(root, namespaces);
1648         return set;
1649 }
1650
1651 static struct list_head *
1652 lookup_header_list(struct ctl_table_root *root, struct nsproxy *namespaces)
1653 {
1654         struct ctl_table_set *set = lookup_header_set(root, namespaces);
1655         return &set->list;
1656 }
1657
1658 struct ctl_table_header *__sysctl_head_next(struct nsproxy *namespaces,
1659                                             struct ctl_table_header *prev)
1660 {
1661         struct ctl_table_root *root;
1662         struct list_head *header_list;
1663         struct ctl_table_header *head;
1664         struct list_head *tmp;
1665
1666         spin_lock(&sysctl_lock);
1667         if (prev) {
1668                 head = prev;
1669                 tmp = &prev->ctl_entry;
1670                 unuse_table(prev);
1671                 goto next;
1672         }
1673         tmp = &root_table_header.ctl_entry;
1674         for (;;) {
1675                 head = list_entry(tmp, struct ctl_table_header, ctl_entry);
1676
1677                 if (!use_table(head))
1678                         goto next;
1679                 spin_unlock(&sysctl_lock);
1680                 return head;
1681         next:
1682                 root = head->root;
1683                 tmp = tmp->next;
1684                 header_list = lookup_header_list(root, namespaces);
1685                 if (tmp != header_list)
1686                         continue;
1687
1688                 do {
1689                         root = list_entry(root->root_list.next,
1690                                         struct ctl_table_root, root_list);
1691                         if (root == &sysctl_table_root)
1692                                 goto out;
1693                         header_list = lookup_header_list(root, namespaces);
1694                 } while (list_empty(header_list));
1695                 tmp = header_list->next;
1696         }
1697 out:
1698         spin_unlock(&sysctl_lock);
1699         return NULL;
1700 }
1701
1702 struct ctl_table_header *sysctl_head_next(struct ctl_table_header *prev)
1703 {
1704         return __sysctl_head_next(current->nsproxy, prev);
1705 }
1706
1707 void register_sysctl_root(struct ctl_table_root *root)
1708 {
1709         spin_lock(&sysctl_lock);
1710         list_add_tail(&root->root_list, &sysctl_table_root.root_list);
1711         spin_unlock(&sysctl_lock);
1712 }
1713
1714 /*
1715  * sysctl_perm does NOT grant the superuser all rights automatically, because
1716  * some sysctl variables are readonly even to root.
1717  */
1718
1719 static int test_perm(int mode, int op)
1720 {
1721         if (!current_euid())
1722                 mode >>= 6;
1723         else if (in_egroup_p(0))
1724                 mode >>= 3;
1725         if ((op & ~mode & (MAY_READ|MAY_WRITE|MAY_EXEC)) == 0)
1726                 return 0;
1727         return -EACCES;
1728 }
1729
1730 int sysctl_perm(struct ctl_table_root *root, struct ctl_table *table, int op)
1731 {
1732         int mode;
1733
1734         if (root->permissions)
1735                 mode = root->permissions(root, current->nsproxy, table);
1736         else
1737                 mode = table->mode;
1738
1739         return test_perm(mode, op);
1740 }
1741
1742 static void sysctl_set_parent(struct ctl_table *parent, struct ctl_table *table)
1743 {
1744         for (; table->procname; table++) {
1745                 table->parent = parent;
1746                 if (table->child)
1747                         sysctl_set_parent(table, table->child);
1748         }
1749 }
1750
1751 static __init int sysctl_init(void)
1752 {
1753         sysctl_set_parent(NULL, root_table);
1754 #ifdef CONFIG_SYSCTL_SYSCALL_CHECK
1755         sysctl_check_table(current->nsproxy, root_table);
1756 #endif
1757         return 0;
1758 }
1759
1760 core_initcall(sysctl_init);
1761
1762 static struct ctl_table *is_branch_in(struct ctl_table *branch,
1763                                       struct ctl_table *table)
1764 {
1765         struct ctl_table *p;
1766         const char *s = branch->procname;
1767
1768         /* branch should have named subdirectory as its first element */
1769         if (!s || !branch->child)
1770                 return NULL;
1771
1772         /* ... and nothing else */
1773         if (branch[1].procname)
1774                 return NULL;
1775
1776         /* table should contain subdirectory with the same name */
1777         for (p = table; p->procname; p++) {
1778                 if (!p->child)
1779                         continue;
1780                 if (p->procname && strcmp(p->procname, s) == 0)
1781                         return p;
1782         }
1783         return NULL;
1784 }
1785
1786 /* see if attaching q to p would be an improvement */
1787 static void try_attach(struct ctl_table_header *p, struct ctl_table_header *q)
1788 {
1789         struct ctl_table *to = p->ctl_table, *by = q->ctl_table;
1790         struct ctl_table *next;
1791         int is_better = 0;
1792         int not_in_parent = !p->attached_by;
1793
1794         while ((next = is_branch_in(by, to)) != NULL) {
1795                 if (by == q->attached_by)
1796                         is_better = 1;
1797                 if (to == p->attached_by)
1798                         not_in_parent = 1;
1799                 by = by->child;
1800                 to = next->child;
1801         }
1802
1803         if (is_better && not_in_parent) {
1804                 q->attached_by = by;
1805                 q->attached_to = to;
1806                 q->parent = p;
1807         }
1808 }
1809
1810 /**
1811  * __register_sysctl_paths - register a sysctl hierarchy
1812  * @root: List of sysctl headers to register on
1813  * @namespaces: Data to compute which lists of sysctl entries are visible
1814  * @path: The path to the directory the sysctl table is in.
1815  * @table: the top-level table structure
1816  *
1817  * Register a sysctl table hierarchy. @table should be a filled in ctl_table
1818  * array. A completely 0 filled entry terminates the table.
1819  *
1820  * The members of the &struct ctl_table structure are used as follows:
1821  *
1822  * procname - the name of the sysctl file under /proc/sys. Set to %NULL to not
1823  *            enter a sysctl file
1824  *
1825  * data - a pointer to data for use by proc_handler
1826  *
1827  * maxlen - the maximum size in bytes of the data
1828  *
1829  * mode - the file permissions for the /proc/sys file, and for sysctl(2)
1830  *
1831  * child - a pointer to the child sysctl table if this entry is a directory, or
1832  *         %NULL.
1833  *
1834  * proc_handler - the text handler routine (described below)
1835  *
1836  * de - for internal use by the sysctl routines
1837  *
1838  * extra1, extra2 - extra pointers usable by the proc handler routines
1839  *
1840  * Leaf nodes in the sysctl tree will be represented by a single file
1841  * under /proc; non-leaf nodes will be represented by directories.
1842  *
1843  * sysctl(2) can automatically manage read and write requests through
1844  * the sysctl table.  The data and maxlen fields of the ctl_table
1845  * struct enable minimal validation of the values being written to be
1846  * performed, and the mode field allows minimal authentication.
1847  *
1848  * There must be a proc_handler routine for any terminal nodes
1849  * mirrored under /proc/sys (non-terminals are handled by a built-in
1850  * directory handler).  Several default handlers are available to
1851  * cover common cases -
1852  *
1853  * proc_dostring(), proc_dointvec(), proc_dointvec_jiffies(),
1854  * proc_dointvec_userhz_jiffies(), proc_dointvec_minmax(), 
1855  * proc_doulongvec_ms_jiffies_minmax(), proc_doulongvec_minmax()
1856  *
1857  * It is the handler's job to read the input buffer from user memory
1858  * and process it. The handler should return 0 on success.
1859  *
1860  * This routine returns %NULL on a failure to register, and a pointer
1861  * to the table header on success.
1862  */
1863 struct ctl_table_header *__register_sysctl_paths(
1864         struct ctl_table_root *root,
1865         struct nsproxy *namespaces,
1866         const struct ctl_path *path, struct ctl_table *table)
1867 {
1868         struct ctl_table_header *header;
1869         struct ctl_table *new, **prevp;
1870         unsigned int n, npath;
1871         struct ctl_table_set *set;
1872
1873         /* Count the path components */
1874         for (npath = 0; path[npath].procname; ++npath)
1875                 ;
1876
1877         /*
1878          * For each path component, allocate a 2-element ctl_table array.
1879          * The first array element will be filled with the sysctl entry
1880          * for this, the second will be the sentinel (procname == 0).
1881          *
1882          * We allocate everything in one go so that we don't have to
1883          * worry about freeing additional memory in unregister_sysctl_table.
1884          */
1885         header = kzalloc(sizeof(struct ctl_table_header) +
1886                          (2 * npath * sizeof(struct ctl_table)), GFP_KERNEL);
1887         if (!header)
1888                 return NULL;
1889
1890         new = (struct ctl_table *) (header + 1);
1891
1892         /* Now connect the dots */
1893         prevp = &header->ctl_table;
1894         for (n = 0; n < npath; ++n, ++path) {
1895                 /* Copy the procname */
1896                 new->procname = path->procname;
1897                 new->mode     = 0555;
1898
1899                 *prevp = new;
1900                 prevp = &new->child;
1901
1902                 new += 2;
1903         }
1904         *prevp = table;
1905         header->ctl_table_arg = table;
1906
1907         INIT_LIST_HEAD(&header->ctl_entry);
1908         header->used = 0;
1909         header->unregistering = NULL;
1910         header->root = root;
1911         sysctl_set_parent(NULL, header->ctl_table);
1912         header->count = 1;
1913 #ifdef CONFIG_SYSCTL_SYSCALL_CHECK
1914         if (sysctl_check_table(namespaces, header->ctl_table)) {
1915                 kfree(header);
1916                 return NULL;
1917         }
1918 #endif
1919         spin_lock(&sysctl_lock);
1920         header->set = lookup_header_set(root, namespaces);
1921         header->attached_by = header->ctl_table;
1922         header->attached_to = root_table;
1923         header->parent = &root_table_header;
1924         for (set = header->set; set; set = set->parent) {
1925                 struct ctl_table_header *p;
1926                 list_for_each_entry(p, &set->list, ctl_entry) {
1927                         if (p->unregistering)
1928                                 continue;
1929                         try_attach(p, header);
1930                 }
1931         }
1932         header->parent->count++;
1933         list_add_tail(&header->ctl_entry, &header->set->list);
1934         spin_unlock(&sysctl_lock);
1935
1936         return header;
1937 }
1938
1939 /**
1940  * register_sysctl_table_path - register a sysctl table hierarchy
1941  * @path: The path to the directory the sysctl table is in.
1942  * @table: the top-level table structure
1943  *
1944  * Register a sysctl table hierarchy. @table should be a filled in ctl_table
1945  * array. A completely 0 filled entry terminates the table.
1946  *
1947  * See __register_sysctl_paths for more details.
1948  */
1949 struct ctl_table_header *register_sysctl_paths(const struct ctl_path *path,
1950                                                 struct ctl_table *table)
1951 {
1952         return __register_sysctl_paths(&sysctl_table_root, current->nsproxy,
1953                                         path, table);
1954 }
1955
1956 /**
1957  * register_sysctl_table - register a sysctl table hierarchy
1958  * @table: the top-level table structure
1959  *
1960  * Register a sysctl table hierarchy. @table should be a filled in ctl_table
1961  * array. A completely 0 filled entry terminates the table.
1962  *
1963  * See register_sysctl_paths for more details.
1964  */
1965 struct ctl_table_header *register_sysctl_table(struct ctl_table *table)
1966 {
1967         static const struct ctl_path null_path[] = { {} };
1968
1969         return register_sysctl_paths(null_path, table);
1970 }
1971
1972 /**
1973  * unregister_sysctl_table - unregister a sysctl table hierarchy
1974  * @header: the header returned from register_sysctl_table
1975  *
1976  * Unregisters the sysctl table and all children. proc entries may not
1977  * actually be removed until they are no longer used by anyone.
1978  */
1979 void unregister_sysctl_table(struct ctl_table_header * header)
1980 {
1981         might_sleep();
1982
1983         if (header == NULL)
1984                 return;
1985
1986         spin_lock(&sysctl_lock);
1987         start_unregistering(header);
1988         if (!--header->parent->count) {
1989                 WARN_ON(1);
1990                 kfree_rcu(header->parent, rcu);
1991         }
1992         if (!--header->count)
1993                 kfree_rcu(header, rcu);
1994         spin_unlock(&sysctl_lock);
1995 }
1996
1997 int sysctl_is_seen(struct ctl_table_header *p)
1998 {
1999         struct ctl_table_set *set = p->set;
2000         int res;
2001         spin_lock(&sysctl_lock);
2002         if (p->unregistering)
2003                 res = 0;
2004         else if (!set->is_seen)
2005                 res = 1;
2006         else
2007                 res = set->is_seen(set);
2008         spin_unlock(&sysctl_lock);
2009         return res;
2010 }
2011
2012 void setup_sysctl_set(struct ctl_table_set *p,
2013         struct ctl_table_set *parent,
2014         int (*is_seen)(struct ctl_table_set *))
2015 {
2016         INIT_LIST_HEAD(&p->list);
2017         p->parent = parent ? parent : &sysctl_table_root.default_set;
2018         p->is_seen = is_seen;
2019 }
2020
2021 #else /* !CONFIG_SYSCTL */
2022 struct ctl_table_header *register_sysctl_table(struct ctl_table * table)
2023 {
2024         return NULL;
2025 }
2026
2027 struct ctl_table_header *register_sysctl_paths(const struct ctl_path *path,
2028                                                     struct ctl_table *table)
2029 {
2030         return NULL;
2031 }
2032
2033 void unregister_sysctl_table(struct ctl_table_header * table)
2034 {
2035 }
2036
2037 void setup_sysctl_set(struct ctl_table_set *p,
2038         struct ctl_table_set *parent,
2039         int (*is_seen)(struct ctl_table_set *))
2040 {
2041 }
2042
2043 void sysctl_head_put(struct ctl_table_header *head)
2044 {
2045 }
2046
2047 #endif /* CONFIG_SYSCTL */
2048
2049 /*
2050  * /proc/sys support
2051  */
2052
2053 #ifdef CONFIG_PROC_SYSCTL
2054
2055 static int _proc_do_string(void* data, int maxlen, int write,
2056                            void __user *buffer,
2057                            size_t *lenp, loff_t *ppos)
2058 {
2059         size_t len;
2060         char __user *p;
2061         char c;
2062
2063         if (!data || !maxlen || !*lenp) {
2064                 *lenp = 0;
2065                 return 0;
2066         }
2067
2068         if (write) {
2069                 len = 0;
2070                 p = buffer;
2071                 while (len < *lenp) {
2072                         if (get_user(c, p++))
2073                                 return -EFAULT;
2074                         if (c == 0 || c == '\n')
2075                                 break;
2076                         len++;
2077                 }
2078                 if (len >= maxlen)
2079                         len = maxlen-1;
2080                 if(copy_from_user(data, buffer, len))
2081                         return -EFAULT;
2082                 ((char *) data)[len] = 0;
2083                 *ppos += *lenp;
2084         } else {
2085                 len = strlen(data);
2086                 if (len > maxlen)
2087                         len = maxlen;
2088
2089                 if (*ppos > len) {
2090                         *lenp = 0;
2091                         return 0;
2092                 }
2093
2094                 data += *ppos;
2095                 len  -= *ppos;
2096
2097                 if (len > *lenp)
2098                         len = *lenp;
2099                 if (len)
2100                         if(copy_to_user(buffer, data, len))
2101                                 return -EFAULT;
2102                 if (len < *lenp) {
2103                         if(put_user('\n', ((char __user *) buffer) + len))
2104                                 return -EFAULT;
2105                         len++;
2106                 }
2107                 *lenp = len;
2108                 *ppos += len;
2109         }
2110         return 0;
2111 }
2112
2113 /**
2114  * proc_dostring - read a string sysctl
2115  * @table: the sysctl table
2116  * @write: %TRUE if this is a write to the sysctl file
2117  * @buffer: the user buffer
2118  * @lenp: the size of the user buffer
2119  * @ppos: file position
2120  *
2121  * Reads/writes a string from/to the user buffer. If the kernel
2122  * buffer provided is not large enough to hold the string, the
2123  * string is truncated. The copied string is %NULL-terminated.
2124  * If the string is being read by the user process, it is copied
2125  * and a newline '\n' is added. It is truncated if the buffer is
2126  * not large enough.
2127  *
2128  * Returns 0 on success.
2129  */
2130 int proc_dostring(struct ctl_table *table, int write,
2131                   void __user *buffer, size_t *lenp, loff_t *ppos)
2132 {
2133         return _proc_do_string(table->data, table->maxlen, write,
2134                                buffer, lenp, ppos);
2135 }
2136
2137 static size_t proc_skip_spaces(char **buf)
2138 {
2139         size_t ret;
2140         char *tmp = skip_spaces(*buf);
2141         ret = tmp - *buf;
2142         *buf = tmp;
2143         return ret;
2144 }
2145
2146 static void proc_skip_char(char **buf, size_t *size, const char v)
2147 {
2148         while (*size) {
2149                 if (**buf != v)
2150                         break;
2151                 (*size)--;
2152                 (*buf)++;
2153         }
2154 }
2155
2156 #define TMPBUFLEN 22
2157 /**
2158  * proc_get_long - reads an ASCII formatted integer from a user buffer
2159  *
2160  * @buf: a kernel buffer
2161  * @size: size of the kernel buffer
2162  * @val: this is where the number will be stored
2163  * @neg: set to %TRUE if number is negative
2164  * @perm_tr: a vector which contains the allowed trailers
2165  * @perm_tr_len: size of the perm_tr vector
2166  * @tr: pointer to store the trailer character
2167  *
2168  * In case of success %0 is returned and @buf and @size are updated with
2169  * the amount of bytes read. If @tr is non-NULL and a trailing
2170  * character exists (size is non-zero after returning from this
2171  * function), @tr is updated with the trailing character.
2172  */
2173 static int proc_get_long(char **buf, size_t *size,
2174                           unsigned long *val, bool *neg,
2175                           const char *perm_tr, unsigned perm_tr_len, char *tr)
2176 {
2177         int len;
2178         char *p, tmp[TMPBUFLEN];
2179
2180         if (!*size)
2181                 return -EINVAL;
2182
2183         len = *size;
2184         if (len > TMPBUFLEN - 1)
2185                 len = TMPBUFLEN - 1;
2186
2187         memcpy(tmp, *buf, len);
2188
2189         tmp[len] = 0;
2190         p = tmp;
2191         if (*p == '-' && *size > 1) {
2192                 *neg = true;
2193                 p++;
2194         } else
2195                 *neg = false;
2196         if (!isdigit(*p))
2197                 return -EINVAL;
2198
2199         *val = simple_strtoul(p, &p, 0);
2200
2201         len = p - tmp;
2202
2203         /* We don't know if the next char is whitespace thus we may accept
2204          * invalid integers (e.g. 1234...a) or two integers instead of one
2205          * (e.g. 123...1). So lets not allow such large numbers. */
2206         if (len == TMPBUFLEN - 1)
2207                 return -EINVAL;
2208
2209         if (len < *size && perm_tr_len && !memchr(perm_tr, *p, perm_tr_len))
2210                 return -EINVAL;
2211
2212         if (tr && (len < *size))
2213                 *tr = *p;
2214
2215         *buf += len;
2216         *size -= len;
2217
2218         return 0;
2219 }
2220
2221 /**
2222  * proc_put_long - converts an integer to a decimal ASCII formatted string
2223  *
2224  * @buf: the user buffer
2225  * @size: the size of the user buffer
2226  * @val: the integer to be converted
2227  * @neg: sign of the number, %TRUE for negative
2228  *
2229  * In case of success %0 is returned and @buf and @size are updated with
2230  * the amount of bytes written.
2231  */
2232 static int proc_put_long(void __user **buf, size_t *size, unsigned long val,
2233                           bool neg)
2234 {
2235         int len;
2236         char tmp[TMPBUFLEN], *p = tmp;
2237
2238         sprintf(p, "%s%lu", neg ? "-" : "", val);
2239         len = strlen(tmp);
2240         if (len > *size)
2241                 len = *size;
2242         if (copy_to_user(*buf, tmp, len))
2243                 return -EFAULT;
2244         *size -= len;
2245         *buf += len;
2246         return 0;
2247 }
2248 #undef TMPBUFLEN
2249
2250 static int proc_put_char(void __user **buf, size_t *size, char c)
2251 {
2252         if (*size) {
2253                 char __user **buffer = (char __user **)buf;
2254                 if (put_user(c, *buffer))
2255                         return -EFAULT;
2256                 (*size)--, (*buffer)++;
2257                 *buf = *buffer;
2258         }
2259         return 0;
2260 }
2261
2262 static int do_proc_dointvec_conv(bool *negp, unsigned long *lvalp,
2263                                  int *valp,
2264                                  int write, void *data)
2265 {
2266         if (write) {
2267                 *valp = *negp ? -*lvalp : *lvalp;
2268         } else {
2269                 int val = *valp;
2270                 if (val < 0) {
2271                         *negp = true;
2272                         *lvalp = (unsigned long)-val;
2273                 } else {
2274                         *negp = false;
2275                         *lvalp = (unsigned long)val;
2276                 }
2277         }
2278         return 0;
2279 }
2280
2281 static const char proc_wspace_sep[] = { ' ', '\t', '\n' };
2282
2283 static int __do_proc_dointvec(void *tbl_data, struct ctl_table *table,
2284                   int write, void __user *buffer,
2285                   size_t *lenp, loff_t *ppos,
2286                   int (*conv)(bool *negp, unsigned long *lvalp, int *valp,
2287                               int write, void *data),
2288                   void *data)
2289 {
2290         int *i, vleft, first = 1, err = 0;
2291         unsigned long page = 0;
2292         size_t left;
2293         char *kbuf;
2294         
2295         if (!tbl_data || !table->maxlen || !*lenp || (*ppos && !write)) {
2296                 *lenp = 0;
2297                 return 0;
2298         }
2299         
2300         i = (int *) tbl_data;
2301         vleft = table->maxlen / sizeof(*i);
2302         left = *lenp;
2303
2304         if (!conv)
2305                 conv = do_proc_dointvec_conv;
2306
2307         if (write) {
2308                 if (left > PAGE_SIZE - 1)
2309                         left = PAGE_SIZE - 1;
2310                 page = __get_free_page(GFP_TEMPORARY);
2311                 kbuf = (char *) page;
2312                 if (!kbuf)
2313                         return -ENOMEM;
2314                 if (copy_from_user(kbuf, buffer, left)) {
2315                         err = -EFAULT;
2316                         goto free;
2317                 }
2318                 kbuf[left] = 0;
2319         }
2320
2321         for (; left && vleft--; i++, first=0) {
2322                 unsigned long lval;
2323                 bool neg;
2324
2325                 if (write) {
2326                         left -= proc_skip_spaces(&kbuf);
2327
2328                         if (!left)
2329                                 break;
2330                         err = proc_get_long(&kbuf, &left, &lval, &neg,
2331                                              proc_wspace_sep,
2332                                              sizeof(proc_wspace_sep), NULL);
2333                         if (err)
2334                                 break;
2335                         if (conv(&neg, &lval, i, 1, data)) {
2336                                 err = -EINVAL;
2337                                 break;
2338                         }
2339                 } else {
2340                         if (conv(&neg, &lval, i, 0, data)) {
2341                                 err = -EINVAL;
2342                                 break;
2343                         }
2344                         if (!first)
2345                                 err = proc_put_char(&buffer, &left, '\t');
2346                         if (err)
2347                                 break;
2348                         err = proc_put_long(&buffer, &left, lval, neg);
2349                         if (err)
2350                                 break;
2351                 }
2352         }
2353
2354         if (!write && !first && left && !err)
2355                 err = proc_put_char(&buffer, &left, '\n');
2356         if (write && !err && left)
2357                 left -= proc_skip_spaces(&kbuf);
2358 free:
2359         if (write) {
2360                 free_page(page);
2361                 if (first)
2362                         return err ? : -EINVAL;
2363         }
2364         *lenp -= left;
2365         *ppos += *lenp;
2366         return err;
2367 }
2368
2369 static int do_proc_dointvec(struct ctl_table *table, int write,
2370                   void __user *buffer, size_t *lenp, loff_t *ppos,
2371                   int (*conv)(bool *negp, unsigned long *lvalp, int *valp,
2372                               int write, void *data),
2373                   void *data)
2374 {
2375         return __do_proc_dointvec(table->data, table, write,
2376                         buffer, lenp, ppos, conv, data);
2377 }
2378
2379 /**
2380  * proc_dointvec - read a vector of integers
2381  * @table: the sysctl table
2382  * @write: %TRUE if this is a write to the sysctl file
2383  * @buffer: the user buffer
2384  * @lenp: the size of the user buffer
2385  * @ppos: file position
2386  *
2387  * Reads/writes up to table->maxlen/sizeof(unsigned int) integer
2388  * values from/to the user buffer, treated as an ASCII string. 
2389  *
2390  * Returns 0 on success.
2391  */
2392 int proc_dointvec(struct ctl_table *table, int write,
2393                      void __user *buffer, size_t *lenp, loff_t *ppos)
2394 {
2395     return do_proc_dointvec(table,write,buffer,lenp,ppos,
2396                             NULL,NULL);
2397 }
2398
2399 /*
2400  * Taint values can only be increased
2401  * This means we can safely use a temporary.
2402  */
2403 static int proc_taint(struct ctl_table *table, int write,
2404                                void __user *buffer, size_t *lenp, loff_t *ppos)
2405 {
2406         struct ctl_table t;
2407         unsigned long tmptaint = get_taint();
2408         int err;
2409
2410         if (write && !capable(CAP_SYS_ADMIN))
2411                 return -EPERM;
2412
2413         t = *table;
2414         t.data = &tmptaint;
2415         err = proc_doulongvec_minmax(&t, write, buffer, lenp, ppos);
2416         if (err < 0)
2417                 return err;
2418
2419         if (write) {
2420                 /*
2421                  * Poor man's atomic or. Not worth adding a primitive
2422                  * to everyone's atomic.h for this
2423                  */
2424                 int i;
2425                 for (i = 0; i < BITS_PER_LONG && tmptaint >> i; i++) {
2426                         if ((tmptaint >> i) & 1)
2427                                 add_taint(i);
2428                 }
2429         }
2430
2431         return err;
2432 }
2433
2434 #ifdef CONFIG_PRINTK
2435 static int proc_dmesg_restrict(struct ctl_table *table, int write,
2436                                 void __user *buffer, size_t *lenp, loff_t *ppos)
2437 {
2438         if (write && !capable(CAP_SYS_ADMIN))
2439                 return -EPERM;
2440
2441         return proc_dointvec_minmax(table, write, buffer, lenp, ppos);
2442 }
2443 #endif
2444
2445 struct do_proc_dointvec_minmax_conv_param {
2446         int *min;
2447         int *max;
2448 };
2449
2450 static int do_proc_dointvec_minmax_conv(bool *negp, unsigned long *lvalp,
2451                                         int *valp,
2452                                         int write, void *data)
2453 {
2454         struct do_proc_dointvec_minmax_conv_param *param = data;
2455         if (write) {
2456                 int val = *negp ? -*lvalp : *lvalp;
2457                 if ((param->min && *param->min > val) ||
2458                     (param->max && *param->max < val))
2459                         return -EINVAL;
2460                 *valp = val;
2461         } else {
2462                 int val = *valp;
2463                 if (val < 0) {
2464                         *negp = true;
2465                         *lvalp = (unsigned long)-val;
2466                 } else {
2467                         *negp = false;
2468                         *lvalp = (unsigned long)val;
2469                 }
2470         }
2471         return 0;
2472 }
2473
2474 /**
2475  * proc_dointvec_minmax - read a vector of integers with min/max values
2476  * @table: the sysctl table
2477  * @write: %TRUE if this is a write to the sysctl file
2478  * @buffer: the user buffer
2479  * @lenp: the size of the user buffer
2480  * @ppos: file position
2481  *
2482  * Reads/writes up to table->maxlen/sizeof(unsigned int) integer
2483  * values from/to the user buffer, treated as an ASCII string.
2484  *
2485  * This routine will ensure the values are within the range specified by
2486  * table->extra1 (min) and table->extra2 (max).
2487  *
2488  * Returns 0 on success.
2489  */
2490 int proc_dointvec_minmax(struct ctl_table *table, int write,
2491                   void __user *buffer, size_t *lenp, loff_t *ppos)
2492 {
2493         struct do_proc_dointvec_minmax_conv_param param = {
2494                 .min = (int *) table->extra1,
2495                 .max = (int *) table->extra2,
2496         };
2497         return do_proc_dointvec(table, write, buffer, lenp, ppos,
2498                                 do_proc_dointvec_minmax_conv, &param);
2499 }
2500
2501 static int __do_proc_doulongvec_minmax(void *data, struct ctl_table *table, int write,
2502                                      void __user *buffer,
2503                                      size_t *lenp, loff_t *ppos,
2504                                      unsigned long convmul,
2505                                      unsigned long convdiv)
2506 {
2507         unsigned long *i, *min, *max;
2508         int vleft, first = 1, err = 0;
2509         unsigned long page = 0;
2510         size_t left;
2511         char *kbuf;
2512
2513         if (!data || !table->maxlen || !*lenp || (*ppos && !write)) {
2514                 *lenp = 0;
2515                 return 0;
2516         }
2517
2518         i = (unsigned long *) data;
2519         min = (unsigned long *) table->extra1;
2520         max = (unsigned long *) table->extra2;
2521         vleft = table->maxlen / sizeof(unsigned long);
2522         left = *lenp;
2523
2524         if (write) {
2525                 if (left > PAGE_SIZE - 1)
2526                         left = PAGE_SIZE - 1;
2527                 page = __get_free_page(GFP_TEMPORARY);
2528                 kbuf = (char *) page;
2529                 if (!kbuf)
2530                         return -ENOMEM;
2531                 if (copy_from_user(kbuf, buffer, left)) {
2532                         err = -EFAULT;
2533                         goto free;
2534                 }
2535                 kbuf[left] = 0;
2536         }
2537
2538         for (; left && vleft--; i++, first = 0) {
2539                 unsigned long val;
2540
2541                 if (write) {
2542                         bool neg;
2543
2544                         left -= proc_skip_spaces(&kbuf);
2545
2546                         err = proc_get_long(&kbuf, &left, &val, &neg,
2547                                              proc_wspace_sep,
2548                                              sizeof(proc_wspace_sep), NULL);
2549                         if (err)
2550                                 break;
2551                         if (neg)
2552                                 continue;
2553                         if ((min && val < *min) || (max && val > *max))
2554                                 continue;
2555                         *i = val;
2556                 } else {
2557                         val = convdiv * (*i) / convmul;
2558                         if (!first)
2559                                 err = proc_put_char(&buffer, &left, '\t');
2560                         err = proc_put_long(&buffer, &left, val, false);
2561                         if (err)
2562                                 break;
2563                 }
2564         }
2565
2566         if (!write && !first && left && !err)
2567                 err = proc_put_char(&buffer, &left, '\n');
2568         if (write && !err)
2569                 left -= proc_skip_spaces(&kbuf);
2570 free:
2571         if (write) {
2572                 free_page(page);
2573                 if (first)
2574                         return err ? : -EINVAL;
2575         }
2576         *lenp -= left;
2577         *ppos += *lenp;
2578         return err;
2579 }
2580
2581 static int do_proc_doulongvec_minmax(struct ctl_table *table, int write,
2582                                      void __user *buffer,
2583                                      size_t *lenp, loff_t *ppos,
2584                                      unsigned long convmul,
2585                                      unsigned long convdiv)
2586 {
2587         return __do_proc_doulongvec_minmax(table->data, table, write,
2588                         buffer, lenp, ppos, convmul, convdiv);
2589 }
2590
2591 /**
2592  * proc_doulongvec_minmax - read a vector of long integers with min/max values
2593  * @table: the sysctl table
2594  * @write: %TRUE if this is a write to the sysctl file
2595  * @buffer: the user buffer
2596  * @lenp: the size of the user buffer
2597  * @ppos: file position
2598  *
2599  * Reads/writes up to table->maxlen/sizeof(unsigned long) unsigned long
2600  * values from/to the user buffer, treated as an ASCII string.
2601  *
2602  * This routine will ensure the values are within the range specified by
2603  * table->extra1 (min) and table->extra2 (max).
2604  *
2605  * Returns 0 on success.
2606  */
2607 int proc_doulongvec_minmax(struct ctl_table *table, int write,
2608                            void __user *buffer, size_t *lenp, loff_t *ppos)
2609 {
2610     return do_proc_doulongvec_minmax(table, write, buffer, lenp, ppos, 1l, 1l);
2611 }
2612
2613 /**
2614  * proc_doulongvec_ms_jiffies_minmax - read a vector of millisecond values with min/max values
2615  * @table: the sysctl table
2616  * @write: %TRUE if this is a write to the sysctl file
2617  * @buffer: the user buffer
2618  * @lenp: the size of the user buffer
2619  * @ppos: file position
2620  *
2621  * Reads/writes up to table->maxlen/sizeof(unsigned long) unsigned long
2622  * values from/to the user buffer, treated as an ASCII string. The values
2623  * are treated as milliseconds, and converted to jiffies when they are stored.
2624  *
2625  * This routine will ensure the values are within the range specified by
2626  * table->extra1 (min) and table->extra2 (max).
2627  *
2628  * Returns 0 on success.
2629  */
2630 int proc_doulongvec_ms_jiffies_minmax(struct ctl_table *table, int write,
2631                                       void __user *buffer,
2632                                       size_t *lenp, loff_t *ppos)
2633 {
2634     return do_proc_doulongvec_minmax(table, write, buffer,
2635                                      lenp, ppos, HZ, 1000l);
2636 }
2637
2638
2639 static int do_proc_dointvec_jiffies_conv(bool *negp, unsigned long *lvalp,
2640                                          int *valp,
2641                                          int write, void *data)
2642 {
2643         if (write) {
2644                 if (*lvalp > LONG_MAX / HZ)
2645                         return 1;
2646                 *valp = *negp ? -(*lvalp*HZ) : (*lvalp*HZ);
2647         } else {
2648                 int val = *valp;
2649                 unsigned long lval;
2650                 if (val < 0) {
2651                         *negp = true;
2652                         lval = (unsigned long)-val;
2653                 } else {
2654                         *negp = false;
2655                         lval = (unsigned long)val;
2656                 }
2657                 *lvalp = lval / HZ;
2658         }
2659         return 0;
2660 }
2661
2662 static int do_proc_dointvec_userhz_jiffies_conv(bool *negp, unsigned long *lvalp,
2663                                                 int *valp,
2664                                                 int write, void *data)
2665 {
2666         if (write) {
2667                 if (USER_HZ < HZ && *lvalp > (LONG_MAX / HZ) * USER_HZ)
2668                         return 1;
2669                 *valp = clock_t_to_jiffies(*negp ? -*lvalp : *lvalp);
2670         } else {
2671                 int val = *valp;
2672                 unsigned long lval;
2673                 if (val < 0) {
2674                         *negp = true;
2675                         lval = (unsigned long)-val;
2676                 } else {
2677                         *negp = false;
2678                         lval = (unsigned long)val;
2679                 }
2680                 *lvalp = jiffies_to_clock_t(lval);
2681         }
2682         return 0;
2683 }
2684
2685 static int do_proc_dointvec_ms_jiffies_conv(bool *negp, unsigned long *lvalp,
2686                                             int *valp,
2687                                             int write, void *data)
2688 {
2689         if (write) {
2690                 *valp = msecs_to_jiffies(*negp ? -*lvalp : *lvalp);
2691         } else {
2692                 int val = *valp;
2693                 unsigned long lval;
2694                 if (val < 0) {
2695                         *negp = true;
2696                         lval = (unsigned long)-val;
2697                 } else {
2698                         *negp = false;
2699                         lval = (unsigned long)val;
2700                 }
2701                 *lvalp = jiffies_to_msecs(lval);
2702         }
2703         return 0;
2704 }
2705
2706 /**
2707  * proc_dointvec_jiffies - read a vector of integers as seconds
2708  * @table: the sysctl table
2709  * @write: %TRUE if this is a write to the sysctl file
2710  * @buffer: the user buffer
2711  * @lenp: the size of the user buffer
2712  * @ppos: file position
2713  *
2714  * Reads/writes up to table->maxlen/sizeof(unsigned int) integer
2715  * values from/to the user buffer, treated as an ASCII string. 
2716  * The values read are assumed to be in seconds, and are converted into
2717  * jiffies.
2718  *
2719  * Returns 0 on success.
2720  */
2721 int proc_dointvec_jiffies(struct ctl_table *table, int write,
2722                           void __user *buffer, size_t *lenp, loff_t *ppos)
2723 {
2724     return do_proc_dointvec(table,write,buffer,lenp,ppos,
2725                             do_proc_dointvec_jiffies_conv,NULL);
2726 }
2727
2728 /**
2729  * proc_dointvec_userhz_jiffies - read a vector of integers as 1/USER_HZ seconds
2730  * @table: the sysctl table
2731  * @write: %TRUE if this is a write to the sysctl file
2732  * @buffer: the user buffer
2733  * @lenp: the size of the user buffer
2734  * @ppos: pointer to the file position
2735  *
2736  * Reads/writes up to table->maxlen/sizeof(unsigned int) integer
2737  * values from/to the user buffer, treated as an ASCII string. 
2738  * The values read are assumed to be in 1/USER_HZ seconds, and 
2739  * are converted into jiffies.
2740  *
2741  * Returns 0 on success.
2742  */
2743 int proc_dointvec_userhz_jiffies(struct ctl_table *table, int write,
2744                                  void __user *buffer, size_t *lenp, loff_t *ppos)
2745 {
2746     return do_proc_dointvec(table,write,buffer,lenp,ppos,
2747                             do_proc_dointvec_userhz_jiffies_conv,NULL);
2748 }
2749
2750 /**
2751  * proc_dointvec_ms_jiffies - read a vector of integers as 1 milliseconds
2752  * @table: the sysctl table
2753  * @write: %TRUE if this is a write to the sysctl file
2754  * @buffer: the user buffer
2755  * @lenp: the size of the user buffer
2756  * @ppos: file position
2757  * @ppos: the current position in the file
2758  *
2759  * Reads/writes up to table->maxlen/sizeof(unsigned int) integer
2760  * values from/to the user buffer, treated as an ASCII string. 
2761  * The values read are assumed to be in 1/1000 seconds, and 
2762  * are converted into jiffies.
2763  *
2764  * Returns 0 on success.
2765  */
2766 int proc_dointvec_ms_jiffies(struct ctl_table *table, int write,
2767                              void __user *buffer, size_t *lenp, loff_t *ppos)
2768 {
2769         return do_proc_dointvec(table, write, buffer, lenp, ppos,
2770                                 do_proc_dointvec_ms_jiffies_conv, NULL);
2771 }
2772
2773 static int proc_do_cad_pid(struct ctl_table *table, int write,
2774                            void __user *buffer, size_t *lenp, loff_t *ppos)
2775 {
2776         struct pid *new_pid;
2777         pid_t tmp;
2778         int r;
2779
2780         tmp = pid_vnr(cad_pid);
2781
2782         r = __do_proc_dointvec(&tmp, table, write, buffer,
2783                                lenp, ppos, NULL, NULL);
2784         if (r || !write)
2785                 return r;
2786
2787         new_pid = find_get_pid(tmp);
2788         if (!new_pid)
2789                 return -ESRCH;
2790
2791         put_pid(xchg(&cad_pid, new_pid));
2792         return 0;
2793 }
2794
2795 /**
2796  * proc_do_large_bitmap - read/write from/to a large bitmap
2797  * @table: the sysctl table
2798  * @write: %TRUE if this is a write to the sysctl file
2799  * @buffer: the user buffer
2800  * @lenp: the size of the user buffer
2801  * @ppos: file position
2802  *
2803  * The bitmap is stored at table->data and the bitmap length (in bits)
2804  * in table->maxlen.
2805  *
2806  * We use a range comma separated format (e.g. 1,3-4,10-10) so that
2807  * large bitmaps may be represented in a compact manner. Writing into
2808  * the file will clear the bitmap then update it with the given input.
2809  *
2810  * Returns 0 on success.
2811  */
2812 int proc_do_large_bitmap(struct ctl_table *table, int write,
2813                          void __user *buffer, size_t *lenp, loff_t *ppos)
2814 {
2815         int err = 0;
2816         bool first = 1;
2817         size_t left = *lenp;
2818         unsigned long bitmap_len = table->maxlen;
2819         unsigned long *bitmap = (unsigned long *) table->data;
2820         unsigned long *tmp_bitmap = NULL;
2821         char tr_a[] = { '-', ',', '\n' }, tr_b[] = { ',', '\n', 0 }, c;
2822
2823         if (!bitmap_len || !left || (*ppos && !write)) {
2824                 *lenp = 0;
2825                 return 0;
2826         }
2827
2828         if (write) {
2829                 unsigned long page = 0;
2830                 char *kbuf;
2831
2832                 if (left > PAGE_SIZE - 1)
2833                         left = PAGE_SIZE - 1;
2834
2835                 page = __get_free_page(GFP_TEMPORARY);
2836                 kbuf = (char *) page;
2837                 if (!kbuf)
2838                         return -ENOMEM;
2839                 if (copy_from_user(kbuf, buffer, left)) {
2840                         free_page(page);
2841                         return -EFAULT;
2842                 }
2843                 kbuf[left] = 0;
2844
2845                 tmp_bitmap = kzalloc(BITS_TO_LONGS(bitmap_len) * sizeof(unsigned long),
2846                                      GFP_KERNEL);
2847                 if (!tmp_bitmap) {
2848                         free_page(page);
2849                         return -ENOMEM;
2850                 }
2851                 proc_skip_char(&kbuf, &left, '\n');
2852                 while (!err && left) {
2853                         unsigned long val_a, val_b;
2854                         bool neg;
2855
2856                         err = proc_get_long(&kbuf, &left, &val_a, &neg, tr_a,
2857                                              sizeof(tr_a), &c);
2858                         if (err)
2859                                 break;
2860                         if (val_a >= bitmap_len || neg) {
2861                                 err = -EINVAL;
2862                                 break;
2863                         }
2864
2865                         val_b = val_a;
2866                         if (left) {
2867                                 kbuf++;
2868                                 left--;
2869                         }
2870
2871                         if (c == '-') {
2872                                 err = proc_get_long(&kbuf, &left, &val_b,
2873                                                      &neg, tr_b, sizeof(tr_b),
2874                                                      &c);
2875                                 if (err)
2876                                         break;
2877                                 if (val_b >= bitmap_len || neg ||
2878                                     val_a > val_b) {
2879                                         err = -EINVAL;
2880                                         break;
2881                                 }
2882                                 if (left) {
2883                                         kbuf++;
2884                                         left--;
2885                                 }
2886                         }
2887
2888                         while (val_a <= val_b)
2889                                 set_bit(val_a++, tmp_bitmap);
2890
2891                         first = 0;
2892                         proc_skip_char(&kbuf, &left, '\n');
2893                 }
2894                 free_page(page);
2895         } else {
2896                 unsigned long bit_a, bit_b = 0;
2897
2898                 while (left) {
2899                         bit_a = find_next_bit(bitmap, bitmap_len, bit_b);
2900                         if (bit_a >= bitmap_len)
2901                                 break;
2902                         bit_b = find_next_zero_bit(bitmap, bitmap_len,
2903                                                    bit_a + 1) - 1;
2904
2905                         if (!first) {
2906                                 err = proc_put_char(&buffer, &left, ',');
2907                                 if (err)
2908                                         break;
2909                         }
2910                         err = proc_put_long(&buffer, &left, bit_a, false);
2911                         if (err)
2912                                 break;
2913                         if (bit_a != bit_b) {
2914                                 err = proc_put_char(&buffer, &left, '-');
2915                                 if (err)
2916                                         break;
2917                                 err = proc_put_long(&buffer, &left, bit_b, false);
2918                                 if (err)
2919                                         break;
2920                         }
2921
2922                         first = 0; bit_b++;
2923                 }
2924                 if (!err)
2925                         err = proc_put_char(&buffer, &left, '\n');
2926         }
2927
2928         if (!err) {
2929                 if (write) {
2930                         if (*ppos)
2931                                 bitmap_or(bitmap, bitmap, tmp_bitmap, bitmap_len);
2932                         else
2933                                 memcpy(bitmap, tmp_bitmap,
2934                                         BITS_TO_LONGS(bitmap_len) * sizeof(unsigned long));
2935                 }
2936                 kfree(tmp_bitmap);
2937                 *lenp -= left;
2938                 *ppos += *lenp;
2939                 return 0;
2940         } else {
2941                 kfree(tmp_bitmap);
2942                 return err;
2943         }
2944 }
2945
2946 #else /* CONFIG_PROC_SYSCTL */
2947
2948 int proc_dostring(struct ctl_table *table, int write,
2949                   void __user *buffer, size_t *lenp, loff_t *ppos)
2950 {
2951         return -ENOSYS;
2952 }
2953
2954 int proc_dointvec(struct ctl_table *table, int write,
2955                   void __user *buffer, size_t *lenp, loff_t *ppos)
2956 {
2957         return -ENOSYS;
2958 }
2959
2960 int proc_dointvec_minmax(struct ctl_table *table, int write,
2961                     void __user *buffer, size_t *lenp, loff_t *ppos)
2962 {
2963         return -ENOSYS;
2964 }
2965
2966 int proc_dointvec_jiffies(struct ctl_table *table, int write,
2967                     void __user *buffer, size_t *lenp, loff_t *ppos)
2968 {
2969         return -ENOSYS;
2970 }
2971
2972 int proc_dointvec_userhz_jiffies(struct ctl_table *table, int write,
2973                     void __user *buffer, size_t *lenp, loff_t *ppos)
2974 {
2975         return -ENOSYS;
2976 }
2977
2978 int proc_dointvec_ms_jiffies(struct ctl_table *table, int write,
2979                              void __user *buffer, size_t *lenp, loff_t *ppos)
2980 {
2981         return -ENOSYS;
2982 }
2983
2984 int proc_doulongvec_minmax(struct ctl_table *table, int write,
2985                     void __user *buffer, size_t *lenp, loff_t *ppos)
2986 {
2987         return -ENOSYS;
2988 }
2989
2990 int proc_doulongvec_ms_jiffies_minmax(struct ctl_table *table, int write,
2991                                       void __user *buffer,
2992                                       size_t *lenp, loff_t *ppos)
2993 {
2994     return -ENOSYS;
2995 }
2996
2997
2998 #endif /* CONFIG_PROC_SYSCTL */
2999
3000 /*
3001  * No sense putting this after each symbol definition, twice,
3002  * exception granted :-)
3003  */
3004 EXPORT_SYMBOL(proc_dointvec);
3005 EXPORT_SYMBOL(proc_dointvec_jiffies);
3006 EXPORT_SYMBOL(proc_dointvec_minmax);
3007 EXPORT_SYMBOL(proc_dointvec_userhz_jiffies);
3008 EXPORT_SYMBOL(proc_dointvec_ms_jiffies);
3009 EXPORT_SYMBOL(proc_dostring);
3010 EXPORT_SYMBOL(proc_doulongvec_minmax);
3011 EXPORT_SYMBOL(proc_doulongvec_ms_jiffies_minmax);
3012 EXPORT_SYMBOL(register_sysctl_table);
3013 EXPORT_SYMBOL(register_sysctl_paths);
3014 EXPORT_SYMBOL(unregister_sysctl_table);