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