Merge branch 'evm-fixes' of git://git.kernel.org/pub/scm/linux/kernel/git/kasatkin...
[linux-2.6.git] / kernel / sysctl.c
1 /*
2  * sysctl.c: General linux system control interface
3  *
4  * Begun 24 March 1995, Stephen Tweedie
5  * Added /proc support, Dec 1995
6  * Added bdflush entry and intvec min/max checking, 2/23/96, Tom Dyas.
7  * Added hooks for /proc/sys/net (minor, minor patch), 96/4/1, Mike Shaver.
8  * Added kernel/java-{interpreter,appletviewer}, 96/5/10, Mike Shaver.
9  * Dynamic registration fixes, Stephen Tweedie.
10  * Added kswapd-interval, ctrl-alt-del, printk stuff, 1/8/97, Chris Horn.
11  * Made sysctl support optional via CONFIG_SYSCTL, 1/10/97, Chris
12  *  Horn.
13  * Added proc_doulongvec_ms_jiffies_minmax, 09/08/99, Carlos H. Bauer.
14  * Added proc_doulongvec_minmax, 09/08/99, Carlos H. Bauer.
15  * Changed linked lists to use list.h instead of lists.h, 02/24/00, Bill
16  *  Wendling.
17  * The list_for_each() macro wasn't appropriate for the sysctl loop.
18  *  Removed it and replaced it with older style, 03/23/00, Bill Wendling
19  */
20
21 #include <linux/module.h>
22 #include <linux/mm.h>
23 #include <linux/swap.h>
24 #include <linux/slab.h>
25 #include <linux/sysctl.h>
26 #include <linux/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         {
807                 .procname       = "bootloader_type",
808                 .data           = &bootloader_type,
809                 .maxlen         = sizeof (int),
810                 .mode           = 0444,
811                 .proc_handler   = proc_dointvec,
812         },
813         {
814                 .procname       = "bootloader_version",
815                 .data           = &bootloader_version,
816                 .maxlen         = sizeof (int),
817                 .mode           = 0444,
818                 .proc_handler   = proc_dointvec,
819         },
820         {
821                 .procname       = "kstack_depth_to_print",
822                 .data           = &kstack_depth_to_print,
823                 .maxlen         = sizeof(int),
824                 .mode           = 0644,
825                 .proc_handler   = proc_dointvec,
826         },
827         {
828                 .procname       = "io_delay_type",
829                 .data           = &io_delay_type,
830                 .maxlen         = sizeof(int),
831                 .mode           = 0644,
832                 .proc_handler   = proc_dointvec,
833         },
834 #endif
835 #if defined(CONFIG_MMU)
836         {
837                 .procname       = "randomize_va_space",
838                 .data           = &randomize_va_space,
839                 .maxlen         = sizeof(int),
840                 .mode           = 0644,
841                 .proc_handler   = proc_dointvec,
842         },
843 #endif
844 #if defined(CONFIG_S390) && defined(CONFIG_SMP)
845         {
846                 .procname       = "spin_retry",
847                 .data           = &spin_retry,
848                 .maxlen         = sizeof (int),
849                 .mode           = 0644,
850                 .proc_handler   = proc_dointvec,
851         },
852 #endif
853 #if     defined(CONFIG_ACPI_SLEEP) && defined(CONFIG_X86)
854         {
855                 .procname       = "acpi_video_flags",
856                 .data           = &acpi_realmode_flags,
857                 .maxlen         = sizeof (unsigned long),
858                 .mode           = 0644,
859                 .proc_handler   = proc_doulongvec_minmax,
860         },
861 #endif
862 #ifdef CONFIG_IA64
863         {
864                 .procname       = "ignore-unaligned-usertrap",
865                 .data           = &no_unaligned_warning,
866                 .maxlen         = sizeof (int),
867                 .mode           = 0644,
868                 .proc_handler   = proc_dointvec,
869         },
870         {
871                 .procname       = "unaligned-dump-stack",
872                 .data           = &unaligned_dump_stack,
873                 .maxlen         = sizeof (int),
874                 .mode           = 0644,
875                 .proc_handler   = proc_dointvec,
876         },
877 #endif
878 #ifdef CONFIG_DETECT_HUNG_TASK
879         {
880                 .procname       = "hung_task_panic",
881                 .data           = &sysctl_hung_task_panic,
882                 .maxlen         = sizeof(int),
883                 .mode           = 0644,
884                 .proc_handler   = proc_dointvec_minmax,
885                 .extra1         = &zero,
886                 .extra2         = &one,
887         },
888         {
889                 .procname       = "hung_task_check_count",
890                 .data           = &sysctl_hung_task_check_count,
891                 .maxlen         = sizeof(unsigned long),
892                 .mode           = 0644,
893                 .proc_handler   = proc_doulongvec_minmax,
894         },
895         {
896                 .procname       = "hung_task_timeout_secs",
897                 .data           = &sysctl_hung_task_timeout_secs,
898                 .maxlen         = sizeof(unsigned long),
899                 .mode           = 0644,
900                 .proc_handler   = proc_dohung_task_timeout_secs,
901         },
902         {
903                 .procname       = "hung_task_warnings",
904                 .data           = &sysctl_hung_task_warnings,
905                 .maxlen         = sizeof(unsigned long),
906                 .mode           = 0644,
907                 .proc_handler   = proc_doulongvec_minmax,
908         },
909 #endif
910 #ifdef CONFIG_COMPAT
911         {
912                 .procname       = "compat-log",
913                 .data           = &compat_log,
914                 .maxlen         = sizeof (int),
915                 .mode           = 0644,
916                 .proc_handler   = proc_dointvec,
917         },
918 #endif
919 #ifdef CONFIG_RT_MUTEXES
920         {
921                 .procname       = "max_lock_depth",
922                 .data           = &max_lock_depth,
923                 .maxlen         = sizeof(int),
924                 .mode           = 0644,
925                 .proc_handler   = proc_dointvec,
926         },
927 #endif
928         {
929                 .procname       = "poweroff_cmd",
930                 .data           = &poweroff_cmd,
931                 .maxlen         = POWEROFF_CMD_PATH_LEN,
932                 .mode           = 0644,
933                 .proc_handler   = proc_dostring,
934         },
935 #ifdef CONFIG_KEYS
936         {
937                 .procname       = "keys",
938                 .mode           = 0555,
939                 .child          = key_sysctls,
940         },
941 #endif
942 #ifdef CONFIG_RCU_TORTURE_TEST
943         {
944                 .procname       = "rcutorture_runnable",
945                 .data           = &rcutorture_runnable,
946                 .maxlen         = sizeof(int),
947                 .mode           = 0644,
948                 .proc_handler   = proc_dointvec,
949         },
950 #endif
951 #ifdef CONFIG_PERF_EVENTS
952         /*
953          * User-space scripts rely on the existence of this file
954          * as a feature check for perf_events being enabled.
955          *
956          * So it's an ABI, do not remove!
957          */
958         {
959                 .procname       = "perf_event_paranoid",
960                 .data           = &sysctl_perf_event_paranoid,
961                 .maxlen         = sizeof(sysctl_perf_event_paranoid),
962                 .mode           = 0644,
963                 .proc_handler   = proc_dointvec,
964         },
965         {
966                 .procname       = "perf_event_mlock_kb",
967                 .data           = &sysctl_perf_event_mlock,
968                 .maxlen         = sizeof(sysctl_perf_event_mlock),
969                 .mode           = 0644,
970                 .proc_handler   = proc_dointvec,
971         },
972         {
973                 .procname       = "perf_event_max_sample_rate",
974                 .data           = &sysctl_perf_event_sample_rate,
975                 .maxlen         = sizeof(sysctl_perf_event_sample_rate),
976                 .mode           = 0644,
977                 .proc_handler   = perf_proc_update_handler,
978         },
979 #endif
980 #ifdef CONFIG_KMEMCHECK
981         {
982                 .procname       = "kmemcheck",
983                 .data           = &kmemcheck_enabled,
984                 .maxlen         = sizeof(int),
985                 .mode           = 0644,
986                 .proc_handler   = proc_dointvec,
987         },
988 #endif
989 #ifdef CONFIG_BLOCK
990         {
991                 .procname       = "blk_iopoll",
992                 .data           = &blk_iopoll_enabled,
993                 .maxlen         = sizeof(int),
994                 .mode           = 0644,
995                 .proc_handler   = proc_dointvec,
996         },
997 #endif
998         { }
999 };
1000
1001 static struct ctl_table vm_table[] = {
1002         {
1003                 .procname       = "overcommit_memory",
1004                 .data           = &sysctl_overcommit_memory,
1005                 .maxlen         = sizeof(sysctl_overcommit_memory),
1006                 .mode           = 0644,
1007                 .proc_handler   = proc_dointvec_minmax,
1008                 .extra1         = &zero,
1009                 .extra2         = &two,
1010         },
1011         {
1012                 .procname       = "panic_on_oom",
1013                 .data           = &sysctl_panic_on_oom,
1014                 .maxlen         = sizeof(sysctl_panic_on_oom),
1015                 .mode           = 0644,
1016                 .proc_handler   = proc_dointvec_minmax,
1017                 .extra1         = &zero,
1018                 .extra2         = &two,
1019         },
1020         {
1021                 .procname       = "oom_kill_allocating_task",
1022                 .data           = &sysctl_oom_kill_allocating_task,
1023                 .maxlen         = sizeof(sysctl_oom_kill_allocating_task),
1024                 .mode           = 0644,
1025                 .proc_handler   = proc_dointvec,
1026         },
1027         {
1028                 .procname       = "oom_dump_tasks",
1029                 .data           = &sysctl_oom_dump_tasks,
1030                 .maxlen         = sizeof(sysctl_oom_dump_tasks),
1031                 .mode           = 0644,
1032                 .proc_handler   = proc_dointvec,
1033         },
1034         {
1035                 .procname       = "overcommit_ratio",
1036                 .data           = &sysctl_overcommit_ratio,
1037                 .maxlen         = sizeof(sysctl_overcommit_ratio),
1038                 .mode           = 0644,
1039                 .proc_handler   = proc_dointvec,
1040         },
1041         {
1042                 .procname       = "page-cluster", 
1043                 .data           = &page_cluster,
1044                 .maxlen         = sizeof(int),
1045                 .mode           = 0644,
1046                 .proc_handler   = proc_dointvec_minmax,
1047                 .extra1         = &zero,
1048         },
1049         {
1050                 .procname       = "dirty_background_ratio",
1051                 .data           = &dirty_background_ratio,
1052                 .maxlen         = sizeof(dirty_background_ratio),
1053                 .mode           = 0644,
1054                 .proc_handler   = dirty_background_ratio_handler,
1055                 .extra1         = &zero,
1056                 .extra2         = &one_hundred,
1057         },
1058         {
1059                 .procname       = "dirty_background_bytes",
1060                 .data           = &dirty_background_bytes,
1061                 .maxlen         = sizeof(dirty_background_bytes),
1062                 .mode           = 0644,
1063                 .proc_handler   = dirty_background_bytes_handler,
1064                 .extra1         = &one_ul,
1065         },
1066         {
1067                 .procname       = "dirty_ratio",
1068                 .data           = &vm_dirty_ratio,
1069                 .maxlen         = sizeof(vm_dirty_ratio),
1070                 .mode           = 0644,
1071                 .proc_handler   = dirty_ratio_handler,
1072                 .extra1         = &zero,
1073                 .extra2         = &one_hundred,
1074         },
1075         {
1076                 .procname       = "dirty_bytes",
1077                 .data           = &vm_dirty_bytes,
1078                 .maxlen         = sizeof(vm_dirty_bytes),
1079                 .mode           = 0644,
1080                 .proc_handler   = dirty_bytes_handler,
1081                 .extra1         = &dirty_bytes_min,
1082         },
1083         {
1084                 .procname       = "dirty_writeback_centisecs",
1085                 .data           = &dirty_writeback_interval,
1086                 .maxlen         = sizeof(dirty_writeback_interval),
1087                 .mode           = 0644,
1088                 .proc_handler   = dirty_writeback_centisecs_handler,
1089         },
1090         {
1091                 .procname       = "dirty_expire_centisecs",
1092                 .data           = &dirty_expire_interval,
1093                 .maxlen         = sizeof(dirty_expire_interval),
1094                 .mode           = 0644,
1095                 .proc_handler   = proc_dointvec_minmax,
1096                 .extra1         = &zero,
1097         },
1098         {
1099                 .procname       = "nr_pdflush_threads",
1100                 .data           = &nr_pdflush_threads,
1101                 .maxlen         = sizeof nr_pdflush_threads,
1102                 .mode           = 0444 /* read-only*/,
1103                 .proc_handler   = proc_dointvec,
1104         },
1105         {
1106                 .procname       = "swappiness",
1107                 .data           = &vm_swappiness,
1108                 .maxlen         = sizeof(vm_swappiness),
1109                 .mode           = 0644,
1110                 .proc_handler   = proc_dointvec_minmax,
1111                 .extra1         = &zero,
1112                 .extra2         = &one_hundred,
1113         },
1114 #ifdef CONFIG_HUGETLB_PAGE
1115         {
1116                 .procname       = "nr_hugepages",
1117                 .data           = NULL,
1118                 .maxlen         = sizeof(unsigned long),
1119                 .mode           = 0644,
1120                 .proc_handler   = hugetlb_sysctl_handler,
1121                 .extra1         = (void *)&hugetlb_zero,
1122                 .extra2         = (void *)&hugetlb_infinity,
1123         },
1124 #ifdef CONFIG_NUMA
1125         {
1126                 .procname       = "nr_hugepages_mempolicy",
1127                 .data           = NULL,
1128                 .maxlen         = sizeof(unsigned long),
1129                 .mode           = 0644,
1130                 .proc_handler   = &hugetlb_mempolicy_sysctl_handler,
1131                 .extra1         = (void *)&hugetlb_zero,
1132                 .extra2         = (void *)&hugetlb_infinity,
1133         },
1134 #endif
1135          {
1136                 .procname       = "hugetlb_shm_group",
1137                 .data           = &sysctl_hugetlb_shm_group,
1138                 .maxlen         = sizeof(gid_t),
1139                 .mode           = 0644,
1140                 .proc_handler   = proc_dointvec,
1141          },
1142          {
1143                 .procname       = "hugepages_treat_as_movable",
1144                 .data           = &hugepages_treat_as_movable,
1145                 .maxlen         = sizeof(int),
1146                 .mode           = 0644,
1147                 .proc_handler   = hugetlb_treat_movable_handler,
1148         },
1149         {
1150                 .procname       = "nr_overcommit_hugepages",
1151                 .data           = NULL,
1152                 .maxlen         = sizeof(unsigned long),
1153                 .mode           = 0644,
1154                 .proc_handler   = hugetlb_overcommit_handler,
1155                 .extra1         = (void *)&hugetlb_zero,
1156                 .extra2         = (void *)&hugetlb_infinity,
1157         },
1158 #endif
1159         {
1160                 .procname       = "lowmem_reserve_ratio",
1161                 .data           = &sysctl_lowmem_reserve_ratio,
1162                 .maxlen         = sizeof(sysctl_lowmem_reserve_ratio),
1163                 .mode           = 0644,
1164                 .proc_handler   = lowmem_reserve_ratio_sysctl_handler,
1165         },
1166         {
1167                 .procname       = "drop_caches",
1168                 .data           = &sysctl_drop_caches,
1169                 .maxlen         = sizeof(int),
1170                 .mode           = 0644,
1171                 .proc_handler   = drop_caches_sysctl_handler,
1172                 .extra1         = &one,
1173                 .extra2         = &three,
1174         },
1175 #ifdef CONFIG_COMPACTION
1176         {
1177                 .procname       = "compact_memory",
1178                 .data           = &sysctl_compact_memory,
1179                 .maxlen         = sizeof(int),
1180                 .mode           = 0200,
1181                 .proc_handler   = sysctl_compaction_handler,
1182         },
1183         {
1184                 .procname       = "extfrag_threshold",
1185                 .data           = &sysctl_extfrag_threshold,
1186                 .maxlen         = sizeof(int),
1187                 .mode           = 0644,
1188                 .proc_handler   = sysctl_extfrag_handler,
1189                 .extra1         = &min_extfrag_threshold,
1190                 .extra2         = &max_extfrag_threshold,
1191         },
1192
1193 #endif /* CONFIG_COMPACTION */
1194         {
1195                 .procname       = "min_free_kbytes",
1196                 .data           = &min_free_kbytes,
1197                 .maxlen         = sizeof(min_free_kbytes),
1198                 .mode           = 0644,
1199                 .proc_handler   = min_free_kbytes_sysctl_handler,
1200                 .extra1         = &zero,
1201         },
1202         {
1203                 .procname       = "percpu_pagelist_fraction",
1204                 .data           = &percpu_pagelist_fraction,
1205                 .maxlen         = sizeof(percpu_pagelist_fraction),
1206                 .mode           = 0644,
1207                 .proc_handler   = percpu_pagelist_fraction_sysctl_handler,
1208                 .extra1         = &min_percpu_pagelist_fract,
1209         },
1210 #ifdef CONFIG_MMU
1211         {
1212                 .procname       = "max_map_count",
1213                 .data           = &sysctl_max_map_count,
1214                 .maxlen         = sizeof(sysctl_max_map_count),
1215                 .mode           = 0644,
1216                 .proc_handler   = proc_dointvec_minmax,
1217                 .extra1         = &zero,
1218         },
1219 #else
1220         {
1221                 .procname       = "nr_trim_pages",
1222                 .data           = &sysctl_nr_trim_pages,
1223                 .maxlen         = sizeof(sysctl_nr_trim_pages),
1224                 .mode           = 0644,
1225                 .proc_handler   = proc_dointvec_minmax,
1226                 .extra1         = &zero,
1227         },
1228 #endif
1229         {
1230                 .procname       = "laptop_mode",
1231                 .data           = &laptop_mode,
1232                 .maxlen         = sizeof(laptop_mode),
1233                 .mode           = 0644,
1234                 .proc_handler   = proc_dointvec_jiffies,
1235         },
1236         {
1237                 .procname       = "block_dump",
1238                 .data           = &block_dump,
1239                 .maxlen         = sizeof(block_dump),
1240                 .mode           = 0644,
1241                 .proc_handler   = proc_dointvec,
1242                 .extra1         = &zero,
1243         },
1244         {
1245                 .procname       = "vfs_cache_pressure",
1246                 .data           = &sysctl_vfs_cache_pressure,
1247                 .maxlen         = sizeof(sysctl_vfs_cache_pressure),
1248                 .mode           = 0644,
1249                 .proc_handler   = proc_dointvec,
1250                 .extra1         = &zero,
1251         },
1252 #ifdef HAVE_ARCH_PICK_MMAP_LAYOUT
1253         {
1254                 .procname       = "legacy_va_layout",
1255                 .data           = &sysctl_legacy_va_layout,
1256                 .maxlen         = sizeof(sysctl_legacy_va_layout),
1257                 .mode           = 0644,
1258                 .proc_handler   = proc_dointvec,
1259                 .extra1         = &zero,
1260         },
1261 #endif
1262 #ifdef CONFIG_NUMA
1263         {
1264                 .procname       = "zone_reclaim_mode",
1265                 .data           = &zone_reclaim_mode,
1266                 .maxlen         = sizeof(zone_reclaim_mode),
1267                 .mode           = 0644,
1268                 .proc_handler   = proc_dointvec,
1269                 .extra1         = &zero,
1270         },
1271         {
1272                 .procname       = "min_unmapped_ratio",
1273                 .data           = &sysctl_min_unmapped_ratio,
1274                 .maxlen         = sizeof(sysctl_min_unmapped_ratio),
1275                 .mode           = 0644,
1276                 .proc_handler   = sysctl_min_unmapped_ratio_sysctl_handler,
1277                 .extra1         = &zero,
1278                 .extra2         = &one_hundred,
1279         },
1280         {
1281                 .procname       = "min_slab_ratio",
1282                 .data           = &sysctl_min_slab_ratio,
1283                 .maxlen         = sizeof(sysctl_min_slab_ratio),
1284                 .mode           = 0644,
1285                 .proc_handler   = sysctl_min_slab_ratio_sysctl_handler,
1286                 .extra1         = &zero,
1287                 .extra2         = &one_hundred,
1288         },
1289 #endif
1290 #ifdef CONFIG_SMP
1291         {
1292                 .procname       = "stat_interval",
1293                 .data           = &sysctl_stat_interval,
1294                 .maxlen         = sizeof(sysctl_stat_interval),
1295                 .mode           = 0644,
1296                 .proc_handler   = proc_dointvec_jiffies,
1297         },
1298 #endif
1299 #ifdef CONFIG_MMU
1300         {
1301                 .procname       = "mmap_min_addr",
1302                 .data           = &dac_mmap_min_addr,
1303                 .maxlen         = sizeof(unsigned long),
1304                 .mode           = 0644,
1305                 .proc_handler   = mmap_min_addr_handler,
1306         },
1307 #endif
1308 #ifdef CONFIG_NUMA
1309         {
1310                 .procname       = "numa_zonelist_order",
1311                 .data           = &numa_zonelist_order,
1312                 .maxlen         = NUMA_ZONELIST_ORDER_LEN,
1313                 .mode           = 0644,
1314                 .proc_handler   = numa_zonelist_order_handler,
1315         },
1316 #endif
1317 #if (defined(CONFIG_X86_32) && !defined(CONFIG_UML))|| \
1318    (defined(CONFIG_SUPERH) && defined(CONFIG_VSYSCALL))
1319         {
1320                 .procname       = "vdso_enabled",
1321                 .data           = &vdso_enabled,
1322                 .maxlen         = sizeof(vdso_enabled),
1323                 .mode           = 0644,
1324                 .proc_handler   = proc_dointvec,
1325                 .extra1         = &zero,
1326         },
1327 #endif
1328 #ifdef CONFIG_HIGHMEM
1329         {
1330                 .procname       = "highmem_is_dirtyable",
1331                 .data           = &vm_highmem_is_dirtyable,
1332                 .maxlen         = sizeof(vm_highmem_is_dirtyable),
1333                 .mode           = 0644,
1334                 .proc_handler   = proc_dointvec_minmax,
1335                 .extra1         = &zero,
1336                 .extra2         = &one,
1337         },
1338 #endif
1339         {
1340                 .procname       = "scan_unevictable_pages",
1341                 .data           = &scan_unevictable_pages,
1342                 .maxlen         = sizeof(scan_unevictable_pages),
1343                 .mode           = 0644,
1344                 .proc_handler   = scan_unevictable_handler,
1345         },
1346 #ifdef CONFIG_MEMORY_FAILURE
1347         {
1348                 .procname       = "memory_failure_early_kill",
1349                 .data           = &sysctl_memory_failure_early_kill,
1350                 .maxlen         = sizeof(sysctl_memory_failure_early_kill),
1351                 .mode           = 0644,
1352                 .proc_handler   = proc_dointvec_minmax,
1353                 .extra1         = &zero,
1354                 .extra2         = &one,
1355         },
1356         {
1357                 .procname       = "memory_failure_recovery",
1358                 .data           = &sysctl_memory_failure_recovery,
1359                 .maxlen         = sizeof(sysctl_memory_failure_recovery),
1360                 .mode           = 0644,
1361                 .proc_handler   = proc_dointvec_minmax,
1362                 .extra1         = &zero,
1363                 .extra2         = &one,
1364         },
1365 #endif
1366         { }
1367 };
1368
1369 #if defined(CONFIG_BINFMT_MISC) || defined(CONFIG_BINFMT_MISC_MODULE)
1370 static struct ctl_table binfmt_misc_table[] = {
1371         { }
1372 };
1373 #endif
1374
1375 static struct ctl_table fs_table[] = {
1376         {
1377                 .procname       = "inode-nr",
1378                 .data           = &inodes_stat,
1379                 .maxlen         = 2*sizeof(int),
1380                 .mode           = 0444,
1381                 .proc_handler   = proc_nr_inodes,
1382         },
1383         {
1384                 .procname       = "inode-state",
1385                 .data           = &inodes_stat,
1386                 .maxlen         = 7*sizeof(int),
1387                 .mode           = 0444,
1388                 .proc_handler   = proc_nr_inodes,
1389         },
1390         {
1391                 .procname       = "file-nr",
1392                 .data           = &files_stat,
1393                 .maxlen         = sizeof(files_stat),
1394                 .mode           = 0444,
1395                 .proc_handler   = proc_nr_files,
1396         },
1397         {
1398                 .procname       = "file-max",
1399                 .data           = &files_stat.max_files,
1400                 .maxlen         = sizeof(files_stat.max_files),
1401                 .mode           = 0644,
1402                 .proc_handler   = proc_doulongvec_minmax,
1403         },
1404         {
1405                 .procname       = "nr_open",
1406                 .data           = &sysctl_nr_open,
1407                 .maxlen         = sizeof(int),
1408                 .mode           = 0644,
1409                 .proc_handler   = proc_dointvec_minmax,
1410                 .extra1         = &sysctl_nr_open_min,
1411                 .extra2         = &sysctl_nr_open_max,
1412         },
1413         {
1414                 .procname       = "dentry-state",
1415                 .data           = &dentry_stat,
1416                 .maxlen         = 6*sizeof(int),
1417                 .mode           = 0444,
1418                 .proc_handler   = proc_nr_dentry,
1419         },
1420         {
1421                 .procname       = "overflowuid",
1422                 .data           = &fs_overflowuid,
1423                 .maxlen         = sizeof(int),
1424                 .mode           = 0644,
1425                 .proc_handler   = proc_dointvec_minmax,
1426                 .extra1         = &minolduid,
1427                 .extra2         = &maxolduid,
1428         },
1429         {
1430                 .procname       = "overflowgid",
1431                 .data           = &fs_overflowgid,
1432                 .maxlen         = sizeof(int),
1433                 .mode           = 0644,
1434                 .proc_handler   = proc_dointvec_minmax,
1435                 .extra1         = &minolduid,
1436                 .extra2         = &maxolduid,
1437         },
1438 #ifdef CONFIG_FILE_LOCKING
1439         {
1440                 .procname       = "leases-enable",
1441                 .data           = &leases_enable,
1442                 .maxlen         = sizeof(int),
1443                 .mode           = 0644,
1444                 .proc_handler   = proc_dointvec,
1445         },
1446 #endif
1447 #ifdef CONFIG_DNOTIFY
1448         {
1449                 .procname       = "dir-notify-enable",
1450                 .data           = &dir_notify_enable,
1451                 .maxlen         = sizeof(int),
1452                 .mode           = 0644,
1453                 .proc_handler   = proc_dointvec,
1454         },
1455 #endif
1456 #ifdef CONFIG_MMU
1457 #ifdef CONFIG_FILE_LOCKING
1458         {
1459                 .procname       = "lease-break-time",
1460                 .data           = &lease_break_time,
1461                 .maxlen         = sizeof(int),
1462                 .mode           = 0644,
1463                 .proc_handler   = proc_dointvec,
1464         },
1465 #endif
1466 #ifdef CONFIG_AIO
1467         {
1468                 .procname       = "aio-nr",
1469                 .data           = &aio_nr,
1470                 .maxlen         = sizeof(aio_nr),
1471                 .mode           = 0444,
1472                 .proc_handler   = proc_doulongvec_minmax,
1473         },
1474         {
1475                 .procname       = "aio-max-nr",
1476                 .data           = &aio_max_nr,
1477                 .maxlen         = sizeof(aio_max_nr),
1478                 .mode           = 0644,
1479                 .proc_handler   = proc_doulongvec_minmax,
1480         },
1481 #endif /* CONFIG_AIO */
1482 #ifdef CONFIG_INOTIFY_USER
1483         {
1484                 .procname       = "inotify",
1485                 .mode           = 0555,
1486                 .child          = inotify_table,
1487         },
1488 #endif  
1489 #ifdef CONFIG_EPOLL
1490         {
1491                 .procname       = "epoll",
1492                 .mode           = 0555,
1493                 .child          = epoll_table,
1494         },
1495 #endif
1496 #endif
1497         {
1498                 .procname       = "suid_dumpable",
1499                 .data           = &suid_dumpable,
1500                 .maxlen         = sizeof(int),
1501                 .mode           = 0644,
1502                 .proc_handler   = proc_dointvec_minmax,
1503                 .extra1         = &zero,
1504                 .extra2         = &two,
1505         },
1506 #if defined(CONFIG_BINFMT_MISC) || defined(CONFIG_BINFMT_MISC_MODULE)
1507         {
1508                 .procname       = "binfmt_misc",
1509                 .mode           = 0555,
1510                 .child          = binfmt_misc_table,
1511         },
1512 #endif
1513         {
1514                 .procname       = "pipe-max-size",
1515                 .data           = &pipe_max_size,
1516                 .maxlen         = sizeof(int),
1517                 .mode           = 0644,
1518                 .proc_handler   = &pipe_proc_fn,
1519                 .extra1         = &pipe_min_size,
1520         },
1521         { }
1522 };
1523
1524 static struct ctl_table debug_table[] = {
1525 #if defined(CONFIG_X86) || defined(CONFIG_PPC) || defined(CONFIG_SPARC) || \
1526     defined(CONFIG_S390) || defined(CONFIG_TILE)
1527         {
1528                 .procname       = "exception-trace",
1529                 .data           = &show_unhandled_signals,
1530                 .maxlen         = sizeof(int),
1531                 .mode           = 0644,
1532                 .proc_handler   = proc_dointvec
1533         },
1534 #endif
1535 #if defined(CONFIG_OPTPROBES)
1536         {
1537                 .procname       = "kprobes-optimization",
1538                 .data           = &sysctl_kprobes_optimization,
1539                 .maxlen         = sizeof(int),
1540                 .mode           = 0644,
1541                 .proc_handler   = proc_kprobes_optimization_handler,
1542                 .extra1         = &zero,
1543                 .extra2         = &one,
1544         },
1545 #endif
1546         { }
1547 };
1548
1549 static struct ctl_table dev_table[] = {
1550         { }
1551 };
1552
1553 static DEFINE_SPINLOCK(sysctl_lock);
1554
1555 /* called under sysctl_lock */
1556 static int use_table(struct ctl_table_header *p)
1557 {
1558         if (unlikely(p->unregistering))
1559                 return 0;
1560         p->used++;
1561         return 1;
1562 }
1563
1564 /* called under sysctl_lock */
1565 static void unuse_table(struct ctl_table_header *p)
1566 {
1567         if (!--p->used)
1568                 if (unlikely(p->unregistering))
1569                         complete(p->unregistering);
1570 }
1571
1572 /* called under sysctl_lock, will reacquire if has to wait */
1573 static void start_unregistering(struct ctl_table_header *p)
1574 {
1575         /*
1576          * if p->used is 0, nobody will ever touch that entry again;
1577          * we'll eliminate all paths to it before dropping sysctl_lock
1578          */
1579         if (unlikely(p->used)) {
1580                 struct completion wait;
1581                 init_completion(&wait);
1582                 p->unregistering = &wait;
1583                 spin_unlock(&sysctl_lock);
1584                 wait_for_completion(&wait);
1585                 spin_lock(&sysctl_lock);
1586         } else {
1587                 /* anything non-NULL; we'll never dereference it */
1588                 p->unregistering = ERR_PTR(-EINVAL);
1589         }
1590         /*
1591          * do not remove from the list until nobody holds it; walking the
1592          * list in do_sysctl() relies on that.
1593          */
1594         list_del_init(&p->ctl_entry);
1595 }
1596
1597 void sysctl_head_get(struct ctl_table_header *head)
1598 {
1599         spin_lock(&sysctl_lock);
1600         head->count++;
1601         spin_unlock(&sysctl_lock);
1602 }
1603
1604 void sysctl_head_put(struct ctl_table_header *head)
1605 {
1606         spin_lock(&sysctl_lock);
1607         if (!--head->count)
1608                 kfree_rcu(head, rcu);
1609         spin_unlock(&sysctl_lock);
1610 }
1611
1612 struct ctl_table_header *sysctl_head_grab(struct ctl_table_header *head)
1613 {
1614         if (!head)
1615                 BUG();
1616         spin_lock(&sysctl_lock);
1617         if (!use_table(head))
1618                 head = ERR_PTR(-ENOENT);
1619         spin_unlock(&sysctl_lock);
1620         return head;
1621 }
1622
1623 void sysctl_head_finish(struct ctl_table_header *head)
1624 {
1625         if (!head)
1626                 return;
1627         spin_lock(&sysctl_lock);
1628         unuse_table(head);
1629         spin_unlock(&sysctl_lock);
1630 }
1631
1632 static struct ctl_table_set *
1633 lookup_header_set(struct ctl_table_root *root, struct nsproxy *namespaces)
1634 {
1635         struct ctl_table_set *set = &root->default_set;
1636         if (root->lookup)
1637                 set = root->lookup(root, namespaces);
1638         return set;
1639 }
1640
1641 static struct list_head *
1642 lookup_header_list(struct ctl_table_root *root, struct nsproxy *namespaces)
1643 {
1644         struct ctl_table_set *set = lookup_header_set(root, namespaces);
1645         return &set->list;
1646 }
1647
1648 struct ctl_table_header *__sysctl_head_next(struct nsproxy *namespaces,
1649                                             struct ctl_table_header *prev)
1650 {
1651         struct ctl_table_root *root;
1652         struct list_head *header_list;
1653         struct ctl_table_header *head;
1654         struct list_head *tmp;
1655
1656         spin_lock(&sysctl_lock);
1657         if (prev) {
1658                 head = prev;
1659                 tmp = &prev->ctl_entry;
1660                 unuse_table(prev);
1661                 goto next;
1662         }
1663         tmp = &root_table_header.ctl_entry;
1664         for (;;) {
1665                 head = list_entry(tmp, struct ctl_table_header, ctl_entry);
1666
1667                 if (!use_table(head))
1668                         goto next;
1669                 spin_unlock(&sysctl_lock);
1670                 return head;
1671         next:
1672                 root = head->root;
1673                 tmp = tmp->next;
1674                 header_list = lookup_header_list(root, namespaces);
1675                 if (tmp != header_list)
1676                         continue;
1677
1678                 do {
1679                         root = list_entry(root->root_list.next,
1680                                         struct ctl_table_root, root_list);
1681                         if (root == &sysctl_table_root)
1682                                 goto out;
1683                         header_list = lookup_header_list(root, namespaces);
1684                 } while (list_empty(header_list));
1685                 tmp = header_list->next;
1686         }
1687 out:
1688         spin_unlock(&sysctl_lock);
1689         return NULL;
1690 }
1691
1692 struct ctl_table_header *sysctl_head_next(struct ctl_table_header *prev)
1693 {
1694         return __sysctl_head_next(current->nsproxy, prev);
1695 }
1696
1697 void register_sysctl_root(struct ctl_table_root *root)
1698 {
1699         spin_lock(&sysctl_lock);
1700         list_add_tail(&root->root_list, &sysctl_table_root.root_list);
1701         spin_unlock(&sysctl_lock);
1702 }
1703
1704 /*
1705  * sysctl_perm does NOT grant the superuser all rights automatically, because
1706  * some sysctl variables are readonly even to root.
1707  */
1708
1709 static int test_perm(int mode, int op)
1710 {
1711         if (!current_euid())
1712                 mode >>= 6;
1713         else if (in_egroup_p(0))
1714                 mode >>= 3;
1715         if ((op & ~mode & (MAY_READ|MAY_WRITE|MAY_EXEC)) == 0)
1716                 return 0;
1717         return -EACCES;
1718 }
1719
1720 int sysctl_perm(struct ctl_table_root *root, struct ctl_table *table, int op)
1721 {
1722         int mode;
1723
1724         if (root->permissions)
1725                 mode = root->permissions(root, current->nsproxy, table);
1726         else
1727                 mode = table->mode;
1728
1729         return test_perm(mode, op);
1730 }
1731
1732 static void sysctl_set_parent(struct ctl_table *parent, struct ctl_table *table)
1733 {
1734         for (; table->procname; table++) {
1735                 table->parent = parent;
1736                 if (table->child)
1737                         sysctl_set_parent(table, table->child);
1738         }
1739 }
1740
1741 static __init int sysctl_init(void)
1742 {
1743         sysctl_set_parent(NULL, root_table);
1744 #ifdef CONFIG_SYSCTL_SYSCALL_CHECK
1745         sysctl_check_table(current->nsproxy, root_table);
1746 #endif
1747         return 0;
1748 }
1749
1750 core_initcall(sysctl_init);
1751
1752 static struct ctl_table *is_branch_in(struct ctl_table *branch,
1753                                       struct ctl_table *table)
1754 {
1755         struct ctl_table *p;
1756         const char *s = branch->procname;
1757
1758         /* branch should have named subdirectory as its first element */
1759         if (!s || !branch->child)
1760                 return NULL;
1761
1762         /* ... and nothing else */
1763         if (branch[1].procname)
1764                 return NULL;
1765
1766         /* table should contain subdirectory with the same name */
1767         for (p = table; p->procname; p++) {
1768                 if (!p->child)
1769                         continue;
1770                 if (p->procname && strcmp(p->procname, s) == 0)
1771                         return p;
1772         }
1773         return NULL;
1774 }
1775
1776 /* see if attaching q to p would be an improvement */
1777 static void try_attach(struct ctl_table_header *p, struct ctl_table_header *q)
1778 {
1779         struct ctl_table *to = p->ctl_table, *by = q->ctl_table;
1780         struct ctl_table *next;
1781         int is_better = 0;
1782         int not_in_parent = !p->attached_by;
1783
1784         while ((next = is_branch_in(by, to)) != NULL) {
1785                 if (by == q->attached_by)
1786                         is_better = 1;
1787                 if (to == p->attached_by)
1788                         not_in_parent = 1;
1789                 by = by->child;
1790                 to = next->child;
1791         }
1792
1793         if (is_better && not_in_parent) {
1794                 q->attached_by = by;
1795                 q->attached_to = to;
1796                 q->parent = p;
1797         }
1798 }
1799
1800 /**
1801  * __register_sysctl_paths - register a sysctl hierarchy
1802  * @root: List of sysctl headers to register on
1803  * @namespaces: Data to compute which lists of sysctl entries are visible
1804  * @path: The path to the directory the sysctl table is in.
1805  * @table: the top-level table structure
1806  *
1807  * Register a sysctl table hierarchy. @table should be a filled in ctl_table
1808  * array. A completely 0 filled entry terminates the table.
1809  *
1810  * The members of the &struct ctl_table structure are used as follows:
1811  *
1812  * procname - the name of the sysctl file under /proc/sys. Set to %NULL to not
1813  *            enter a sysctl file
1814  *
1815  * data - a pointer to data for use by proc_handler
1816  *
1817  * maxlen - the maximum size in bytes of the data
1818  *
1819  * mode - the file permissions for the /proc/sys file, and for sysctl(2)
1820  *
1821  * child - a pointer to the child sysctl table if this entry is a directory, or
1822  *         %NULL.
1823  *
1824  * proc_handler - the text handler routine (described below)
1825  *
1826  * de - for internal use by the sysctl routines
1827  *
1828  * extra1, extra2 - extra pointers usable by the proc handler routines
1829  *
1830  * Leaf nodes in the sysctl tree will be represented by a single file
1831  * under /proc; non-leaf nodes will be represented by directories.
1832  *
1833  * sysctl(2) can automatically manage read and write requests through
1834  * the sysctl table.  The data and maxlen fields of the ctl_table
1835  * struct enable minimal validation of the values being written to be
1836  * performed, and the mode field allows minimal authentication.
1837  *
1838  * There must be a proc_handler routine for any terminal nodes
1839  * mirrored under /proc/sys (non-terminals are handled by a built-in
1840  * directory handler).  Several default handlers are available to
1841  * cover common cases -
1842  *
1843  * proc_dostring(), proc_dointvec(), proc_dointvec_jiffies(),
1844  * proc_dointvec_userhz_jiffies(), proc_dointvec_minmax(), 
1845  * proc_doulongvec_ms_jiffies_minmax(), proc_doulongvec_minmax()
1846  *
1847  * It is the handler's job to read the input buffer from user memory
1848  * and process it. The handler should return 0 on success.
1849  *
1850  * This routine returns %NULL on a failure to register, and a pointer
1851  * to the table header on success.
1852  */
1853 struct ctl_table_header *__register_sysctl_paths(
1854         struct ctl_table_root *root,
1855         struct nsproxy *namespaces,
1856         const struct ctl_path *path, struct ctl_table *table)
1857 {
1858         struct ctl_table_header *header;
1859         struct ctl_table *new, **prevp;
1860         unsigned int n, npath;
1861         struct ctl_table_set *set;
1862
1863         /* Count the path components */
1864         for (npath = 0; path[npath].procname; ++npath)
1865                 ;
1866
1867         /*
1868          * For each path component, allocate a 2-element ctl_table array.
1869          * The first array element will be filled with the sysctl entry
1870          * for this, the second will be the sentinel (procname == 0).
1871          *
1872          * We allocate everything in one go so that we don't have to
1873          * worry about freeing additional memory in unregister_sysctl_table.
1874          */
1875         header = kzalloc(sizeof(struct ctl_table_header) +
1876                          (2 * npath * sizeof(struct ctl_table)), GFP_KERNEL);
1877         if (!header)
1878                 return NULL;
1879
1880         new = (struct ctl_table *) (header + 1);
1881
1882         /* Now connect the dots */
1883         prevp = &header->ctl_table;
1884         for (n = 0; n < npath; ++n, ++path) {
1885                 /* Copy the procname */
1886                 new->procname = path->procname;
1887                 new->mode     = 0555;
1888
1889                 *prevp = new;
1890                 prevp = &new->child;
1891
1892                 new += 2;
1893         }
1894         *prevp = table;
1895         header->ctl_table_arg = table;
1896
1897         INIT_LIST_HEAD(&header->ctl_entry);
1898         header->used = 0;
1899         header->unregistering = NULL;
1900         header->root = root;
1901         sysctl_set_parent(NULL, header->ctl_table);
1902         header->count = 1;
1903 #ifdef CONFIG_SYSCTL_SYSCALL_CHECK
1904         if (sysctl_check_table(namespaces, header->ctl_table)) {
1905                 kfree(header);
1906                 return NULL;
1907         }
1908 #endif
1909         spin_lock(&sysctl_lock);
1910         header->set = lookup_header_set(root, namespaces);
1911         header->attached_by = header->ctl_table;
1912         header->attached_to = root_table;
1913         header->parent = &root_table_header;
1914         for (set = header->set; set; set = set->parent) {
1915                 struct ctl_table_header *p;
1916                 list_for_each_entry(p, &set->list, ctl_entry) {
1917                         if (p->unregistering)
1918                                 continue;
1919                         try_attach(p, header);
1920                 }
1921         }
1922         header->parent->count++;
1923         list_add_tail(&header->ctl_entry, &header->set->list);
1924         spin_unlock(&sysctl_lock);
1925
1926         return header;
1927 }
1928
1929 /**
1930  * register_sysctl_table_path - register a sysctl table hierarchy
1931  * @path: The path to the directory the sysctl table is in.
1932  * @table: the top-level table structure
1933  *
1934  * Register a sysctl table hierarchy. @table should be a filled in ctl_table
1935  * array. A completely 0 filled entry terminates the table.
1936  *
1937  * See __register_sysctl_paths for more details.
1938  */
1939 struct ctl_table_header *register_sysctl_paths(const struct ctl_path *path,
1940                                                 struct ctl_table *table)
1941 {
1942         return __register_sysctl_paths(&sysctl_table_root, current->nsproxy,
1943                                         path, table);
1944 }
1945
1946 /**
1947  * register_sysctl_table - register a sysctl table hierarchy
1948  * @table: the top-level table structure
1949  *
1950  * Register a sysctl table hierarchy. @table should be a filled in ctl_table
1951  * array. A completely 0 filled entry terminates the table.
1952  *
1953  * See register_sysctl_paths for more details.
1954  */
1955 struct ctl_table_header *register_sysctl_table(struct ctl_table *table)
1956 {
1957         static const struct ctl_path null_path[] = { {} };
1958
1959         return register_sysctl_paths(null_path, table);
1960 }
1961
1962 /**
1963  * unregister_sysctl_table - unregister a sysctl table hierarchy
1964  * @header: the header returned from register_sysctl_table
1965  *
1966  * Unregisters the sysctl table and all children. proc entries may not
1967  * actually be removed until they are no longer used by anyone.
1968  */
1969 void unregister_sysctl_table(struct ctl_table_header * header)
1970 {
1971         might_sleep();
1972
1973         if (header == NULL)
1974                 return;
1975
1976         spin_lock(&sysctl_lock);
1977         start_unregistering(header);
1978         if (!--header->parent->count) {
1979                 WARN_ON(1);
1980                 kfree_rcu(header->parent, rcu);
1981         }
1982         if (!--header->count)
1983                 kfree_rcu(header, rcu);
1984         spin_unlock(&sysctl_lock);
1985 }
1986
1987 int sysctl_is_seen(struct ctl_table_header *p)
1988 {
1989         struct ctl_table_set *set = p->set;
1990         int res;
1991         spin_lock(&sysctl_lock);
1992         if (p->unregistering)
1993                 res = 0;
1994         else if (!set->is_seen)
1995                 res = 1;
1996         else
1997                 res = set->is_seen(set);
1998         spin_unlock(&sysctl_lock);
1999         return res;
2000 }
2001
2002 void setup_sysctl_set(struct ctl_table_set *p,
2003         struct ctl_table_set *parent,
2004         int (*is_seen)(struct ctl_table_set *))
2005 {
2006         INIT_LIST_HEAD(&p->list);
2007         p->parent = parent ? parent : &sysctl_table_root.default_set;
2008         p->is_seen = is_seen;
2009 }
2010
2011 #else /* !CONFIG_SYSCTL */
2012 struct ctl_table_header *register_sysctl_table(struct ctl_table * table)
2013 {
2014         return NULL;
2015 }
2016
2017 struct ctl_table_header *register_sysctl_paths(const struct ctl_path *path,
2018                                                     struct ctl_table *table)
2019 {
2020         return NULL;
2021 }
2022
2023 void unregister_sysctl_table(struct ctl_table_header * table)
2024 {
2025 }
2026
2027 void setup_sysctl_set(struct ctl_table_set *p,
2028         struct ctl_table_set *parent,
2029         int (*is_seen)(struct ctl_table_set *))
2030 {
2031 }
2032
2033 void sysctl_head_put(struct ctl_table_header *head)
2034 {
2035 }
2036
2037 #endif /* CONFIG_SYSCTL */
2038
2039 /*
2040  * /proc/sys support
2041  */
2042
2043 #ifdef CONFIG_PROC_SYSCTL
2044
2045 static int _proc_do_string(void* data, int maxlen, int write,
2046                            void __user *buffer,
2047                            size_t *lenp, loff_t *ppos)
2048 {
2049         size_t len;
2050         char __user *p;
2051         char c;
2052
2053         if (!data || !maxlen || !*lenp) {
2054                 *lenp = 0;
2055                 return 0;
2056         }
2057
2058         if (write) {
2059                 len = 0;
2060                 p = buffer;
2061                 while (len < *lenp) {
2062                         if (get_user(c, p++))
2063                                 return -EFAULT;
2064                         if (c == 0 || c == '\n')
2065                                 break;
2066                         len++;
2067                 }
2068                 if (len >= maxlen)
2069                         len = maxlen-1;
2070                 if(copy_from_user(data, buffer, len))
2071                         return -EFAULT;
2072                 ((char *) data)[len] = 0;
2073                 *ppos += *lenp;
2074         } else {
2075                 len = strlen(data);
2076                 if (len > maxlen)
2077                         len = maxlen;
2078
2079                 if (*ppos > len) {
2080                         *lenp = 0;
2081                         return 0;
2082                 }
2083
2084                 data += *ppos;
2085                 len  -= *ppos;
2086
2087                 if (len > *lenp)
2088                         len = *lenp;
2089                 if (len)
2090                         if(copy_to_user(buffer, data, len))
2091                                 return -EFAULT;
2092                 if (len < *lenp) {
2093                         if(put_user('\n', ((char __user *) buffer) + len))
2094                                 return -EFAULT;
2095                         len++;
2096                 }
2097                 *lenp = len;
2098                 *ppos += len;
2099         }
2100         return 0;
2101 }
2102
2103 /**
2104  * proc_dostring - read a string sysctl
2105  * @table: the sysctl table
2106  * @write: %TRUE if this is a write to the sysctl file
2107  * @buffer: the user buffer
2108  * @lenp: the size of the user buffer
2109  * @ppos: file position
2110  *
2111  * Reads/writes a string from/to the user buffer. If the kernel
2112  * buffer provided is not large enough to hold the string, the
2113  * string is truncated. The copied string is %NULL-terminated.
2114  * If the string is being read by the user process, it is copied
2115  * and a newline '\n' is added. It is truncated if the buffer is
2116  * not large enough.
2117  *
2118  * Returns 0 on success.
2119  */
2120 int proc_dostring(struct ctl_table *table, int write,
2121                   void __user *buffer, size_t *lenp, loff_t *ppos)
2122 {
2123         return _proc_do_string(table->data, table->maxlen, write,
2124                                buffer, lenp, ppos);
2125 }
2126
2127 static size_t proc_skip_spaces(char **buf)
2128 {
2129         size_t ret;
2130         char *tmp = skip_spaces(*buf);
2131         ret = tmp - *buf;
2132         *buf = tmp;
2133         return ret;
2134 }
2135
2136 static void proc_skip_char(char **buf, size_t *size, const char v)
2137 {
2138         while (*size) {
2139                 if (**buf != v)
2140                         break;
2141                 (*size)--;
2142                 (*buf)++;
2143         }
2144 }
2145
2146 #define TMPBUFLEN 22
2147 /**
2148  * proc_get_long - reads an ASCII formatted integer from a user buffer
2149  *
2150  * @buf: a kernel buffer
2151  * @size: size of the kernel buffer
2152  * @val: this is where the number will be stored
2153  * @neg: set to %TRUE if number is negative
2154  * @perm_tr: a vector which contains the allowed trailers
2155  * @perm_tr_len: size of the perm_tr vector
2156  * @tr: pointer to store the trailer character
2157  *
2158  * In case of success %0 is returned and @buf and @size are updated with
2159  * the amount of bytes read. If @tr is non-NULL and a trailing
2160  * character exists (size is non-zero after returning from this
2161  * function), @tr is updated with the trailing character.
2162  */
2163 static int proc_get_long(char **buf, size_t *size,
2164                           unsigned long *val, bool *neg,
2165                           const char *perm_tr, unsigned perm_tr_len, char *tr)
2166 {
2167         int len;
2168         char *p, tmp[TMPBUFLEN];
2169
2170         if (!*size)
2171                 return -EINVAL;
2172
2173         len = *size;
2174         if (len > TMPBUFLEN - 1)
2175                 len = TMPBUFLEN - 1;
2176
2177         memcpy(tmp, *buf, len);
2178
2179         tmp[len] = 0;
2180         p = tmp;
2181         if (*p == '-' && *size > 1) {
2182                 *neg = true;
2183                 p++;
2184         } else
2185                 *neg = false;
2186         if (!isdigit(*p))
2187                 return -EINVAL;
2188
2189         *val = simple_strtoul(p, &p, 0);
2190
2191         len = p - tmp;
2192
2193         /* We don't know if the next char is whitespace thus we may accept
2194          * invalid integers (e.g. 1234...a) or two integers instead of one
2195          * (e.g. 123...1). So lets not allow such large numbers. */
2196         if (len == TMPBUFLEN - 1)
2197                 return -EINVAL;
2198
2199         if (len < *size && perm_tr_len && !memchr(perm_tr, *p, perm_tr_len))
2200                 return -EINVAL;
2201
2202         if (tr && (len < *size))
2203                 *tr = *p;
2204
2205         *buf += len;
2206         *size -= len;
2207
2208         return 0;
2209 }
2210
2211 /**
2212  * proc_put_long - converts an integer to a decimal ASCII formatted string
2213  *
2214  * @buf: the user buffer
2215  * @size: the size of the user buffer
2216  * @val: the integer to be converted
2217  * @neg: sign of the number, %TRUE for negative
2218  *
2219  * In case of success %0 is returned and @buf and @size are updated with
2220  * the amount of bytes written.
2221  */
2222 static int proc_put_long(void __user **buf, size_t *size, unsigned long val,
2223                           bool neg)
2224 {
2225         int len;
2226         char tmp[TMPBUFLEN], *p = tmp;
2227
2228         sprintf(p, "%s%lu", neg ? "-" : "", val);
2229         len = strlen(tmp);
2230         if (len > *size)
2231                 len = *size;
2232         if (copy_to_user(*buf, tmp, len))
2233                 return -EFAULT;
2234         *size -= len;
2235         *buf += len;
2236         return 0;
2237 }
2238 #undef TMPBUFLEN
2239
2240 static int proc_put_char(void __user **buf, size_t *size, char c)
2241 {
2242         if (*size) {
2243                 char __user **buffer = (char __user **)buf;
2244                 if (put_user(c, *buffer))
2245                         return -EFAULT;
2246                 (*size)--, (*buffer)++;
2247                 *buf = *buffer;
2248         }
2249         return 0;
2250 }
2251
2252 static int do_proc_dointvec_conv(bool *negp, unsigned long *lvalp,
2253                                  int *valp,
2254                                  int write, void *data)
2255 {
2256         if (write) {
2257                 *valp = *negp ? -*lvalp : *lvalp;
2258         } else {
2259                 int val = *valp;
2260                 if (val < 0) {
2261                         *negp = true;
2262                         *lvalp = (unsigned long)-val;
2263                 } else {
2264                         *negp = false;
2265                         *lvalp = (unsigned long)val;
2266                 }
2267         }
2268         return 0;
2269 }
2270
2271 static const char proc_wspace_sep[] = { ' ', '\t', '\n' };
2272
2273 static int __do_proc_dointvec(void *tbl_data, struct ctl_table *table,
2274                   int write, void __user *buffer,
2275                   size_t *lenp, loff_t *ppos,
2276                   int (*conv)(bool *negp, unsigned long *lvalp, int *valp,
2277                               int write, void *data),
2278                   void *data)
2279 {
2280         int *i, vleft, first = 1, err = 0;
2281         unsigned long page = 0;
2282         size_t left;
2283         char *kbuf;
2284         
2285         if (!tbl_data || !table->maxlen || !*lenp || (*ppos && !write)) {
2286                 *lenp = 0;
2287                 return 0;
2288         }
2289         
2290         i = (int *) tbl_data;
2291         vleft = table->maxlen / sizeof(*i);
2292         left = *lenp;
2293
2294         if (!conv)
2295                 conv = do_proc_dointvec_conv;
2296
2297         if (write) {
2298                 if (left > PAGE_SIZE - 1)
2299                         left = PAGE_SIZE - 1;
2300                 page = __get_free_page(GFP_TEMPORARY);
2301                 kbuf = (char *) page;
2302                 if (!kbuf)
2303                         return -ENOMEM;
2304                 if (copy_from_user(kbuf, buffer, left)) {
2305                         err = -EFAULT;
2306                         goto free;
2307                 }
2308                 kbuf[left] = 0;
2309         }
2310
2311         for (; left && vleft--; i++, first=0) {
2312                 unsigned long lval;
2313                 bool neg;
2314
2315                 if (write) {
2316                         left -= proc_skip_spaces(&kbuf);
2317
2318                         if (!left)
2319                                 break;
2320                         err = proc_get_long(&kbuf, &left, &lval, &neg,
2321                                              proc_wspace_sep,
2322                                              sizeof(proc_wspace_sep), NULL);
2323                         if (err)
2324                                 break;
2325                         if (conv(&neg, &lval, i, 1, data)) {
2326                                 err = -EINVAL;
2327                                 break;
2328                         }
2329                 } else {
2330                         if (conv(&neg, &lval, i, 0, data)) {
2331                                 err = -EINVAL;
2332                                 break;
2333                         }
2334                         if (!first)
2335                                 err = proc_put_char(&buffer, &left, '\t');
2336                         if (err)
2337                                 break;
2338                         err = proc_put_long(&buffer, &left, lval, neg);
2339                         if (err)
2340                                 break;
2341                 }
2342         }
2343
2344         if (!write && !first && left && !err)
2345                 err = proc_put_char(&buffer, &left, '\n');
2346         if (write && !err && left)
2347                 left -= proc_skip_spaces(&kbuf);
2348 free:
2349         if (write) {
2350                 free_page(page);
2351                 if (first)
2352                         return err ? : -EINVAL;
2353         }
2354         *lenp -= left;
2355         *ppos += *lenp;
2356         return err;
2357 }
2358
2359 static int do_proc_dointvec(struct ctl_table *table, int write,
2360                   void __user *buffer, size_t *lenp, loff_t *ppos,
2361                   int (*conv)(bool *negp, unsigned long *lvalp, int *valp,
2362                               int write, void *data),
2363                   void *data)
2364 {
2365         return __do_proc_dointvec(table->data, table, write,
2366                         buffer, lenp, ppos, conv, data);
2367 }
2368
2369 /**
2370  * proc_dointvec - read a vector of integers
2371  * @table: the sysctl table
2372  * @write: %TRUE if this is a write to the sysctl file
2373  * @buffer: the user buffer
2374  * @lenp: the size of the user buffer
2375  * @ppos: file position
2376  *
2377  * Reads/writes up to table->maxlen/sizeof(unsigned int) integer
2378  * values from/to the user buffer, treated as an ASCII string. 
2379  *
2380  * Returns 0 on success.
2381  */
2382 int proc_dointvec(struct ctl_table *table, int write,
2383                      void __user *buffer, size_t *lenp, loff_t *ppos)
2384 {
2385     return do_proc_dointvec(table,write,buffer,lenp,ppos,
2386                             NULL,NULL);
2387 }
2388
2389 /*
2390  * Taint values can only be increased
2391  * This means we can safely use a temporary.
2392  */
2393 static int proc_taint(struct ctl_table *table, int write,
2394                                void __user *buffer, size_t *lenp, loff_t *ppos)
2395 {
2396         struct ctl_table t;
2397         unsigned long tmptaint = get_taint();
2398         int err;
2399
2400         if (write && !capable(CAP_SYS_ADMIN))
2401                 return -EPERM;
2402
2403         t = *table;
2404         t.data = &tmptaint;
2405         err = proc_doulongvec_minmax(&t, write, buffer, lenp, ppos);
2406         if (err < 0)
2407                 return err;
2408
2409         if (write) {
2410                 /*
2411                  * Poor man's atomic or. Not worth adding a primitive
2412                  * to everyone's atomic.h for this
2413                  */
2414                 int i;
2415                 for (i = 0; i < BITS_PER_LONG && tmptaint >> i; i++) {
2416                         if ((tmptaint >> i) & 1)
2417                                 add_taint(i);
2418                 }
2419         }
2420
2421         return err;
2422 }
2423
2424 #ifdef CONFIG_PRINTK
2425 static int proc_dmesg_restrict(struct ctl_table *table, int write,
2426                                 void __user *buffer, size_t *lenp, loff_t *ppos)
2427 {
2428         if (write && !capable(CAP_SYS_ADMIN))
2429                 return -EPERM;
2430
2431         return proc_dointvec_minmax(table, write, buffer, lenp, ppos);
2432 }
2433 #endif
2434
2435 struct do_proc_dointvec_minmax_conv_param {
2436         int *min;
2437         int *max;
2438 };
2439
2440 static int do_proc_dointvec_minmax_conv(bool *negp, unsigned long *lvalp,
2441                                         int *valp,
2442                                         int write, void *data)
2443 {
2444         struct do_proc_dointvec_minmax_conv_param *param = data;
2445         if (write) {
2446                 int val = *negp ? -*lvalp : *lvalp;
2447                 if ((param->min && *param->min > val) ||
2448                     (param->max && *param->max < val))
2449                         return -EINVAL;
2450                 *valp = val;
2451         } else {
2452                 int val = *valp;
2453                 if (val < 0) {
2454                         *negp = true;
2455                         *lvalp = (unsigned long)-val;
2456                 } else {
2457                         *negp = false;
2458                         *lvalp = (unsigned long)val;
2459                 }
2460         }
2461         return 0;
2462 }
2463
2464 /**
2465  * proc_dointvec_minmax - read a vector of integers with min/max values
2466  * @table: the sysctl table
2467  * @write: %TRUE if this is a write to the sysctl file
2468  * @buffer: the user buffer
2469  * @lenp: the size of the user buffer
2470  * @ppos: file position
2471  *
2472  * Reads/writes up to table->maxlen/sizeof(unsigned int) integer
2473  * values from/to the user buffer, treated as an ASCII string.
2474  *
2475  * This routine will ensure the values are within the range specified by
2476  * table->extra1 (min) and table->extra2 (max).
2477  *
2478  * Returns 0 on success.
2479  */
2480 int proc_dointvec_minmax(struct ctl_table *table, int write,
2481                   void __user *buffer, size_t *lenp, loff_t *ppos)
2482 {
2483         struct do_proc_dointvec_minmax_conv_param param = {
2484                 .min = (int *) table->extra1,
2485                 .max = (int *) table->extra2,
2486         };
2487         return do_proc_dointvec(table, write, buffer, lenp, ppos,
2488                                 do_proc_dointvec_minmax_conv, &param);
2489 }
2490
2491 static int __do_proc_doulongvec_minmax(void *data, struct ctl_table *table, int write,
2492                                      void __user *buffer,
2493                                      size_t *lenp, loff_t *ppos,
2494                                      unsigned long convmul,
2495                                      unsigned long convdiv)
2496 {
2497         unsigned long *i, *min, *max;
2498         int vleft, first = 1, err = 0;
2499         unsigned long page = 0;
2500         size_t left;
2501         char *kbuf;
2502
2503         if (!data || !table->maxlen || !*lenp || (*ppos && !write)) {
2504                 *lenp = 0;
2505                 return 0;
2506         }
2507
2508         i = (unsigned long *) data;
2509         min = (unsigned long *) table->extra1;
2510         max = (unsigned long *) table->extra2;
2511         vleft = table->maxlen / sizeof(unsigned long);
2512         left = *lenp;
2513
2514         if (write) {
2515                 if (left > PAGE_SIZE - 1)
2516                         left = PAGE_SIZE - 1;
2517                 page = __get_free_page(GFP_TEMPORARY);
2518                 kbuf = (char *) page;
2519                 if (!kbuf)
2520                         return -ENOMEM;
2521                 if (copy_from_user(kbuf, buffer, left)) {
2522                         err = -EFAULT;
2523                         goto free;
2524                 }
2525                 kbuf[left] = 0;
2526         }
2527
2528         for (; left && vleft--; i++, first = 0) {
2529                 unsigned long val;
2530
2531                 if (write) {
2532                         bool neg;
2533
2534                         left -= proc_skip_spaces(&kbuf);
2535
2536                         err = proc_get_long(&kbuf, &left, &val, &neg,
2537                                              proc_wspace_sep,
2538                                              sizeof(proc_wspace_sep), NULL);
2539                         if (err)
2540                                 break;
2541                         if (neg)
2542                                 continue;
2543                         if ((min && val < *min) || (max && val > *max))
2544                                 continue;
2545                         *i = val;
2546                 } else {
2547                         val = convdiv * (*i) / convmul;
2548                         if (!first)
2549                                 err = proc_put_char(&buffer, &left, '\t');
2550                         err = proc_put_long(&buffer, &left, val, false);
2551                         if (err)
2552                                 break;
2553                 }
2554         }
2555
2556         if (!write && !first && left && !err)
2557                 err = proc_put_char(&buffer, &left, '\n');
2558         if (write && !err)
2559                 left -= proc_skip_spaces(&kbuf);
2560 free:
2561         if (write) {
2562                 free_page(page);
2563                 if (first)
2564                         return err ? : -EINVAL;
2565         }
2566         *lenp -= left;
2567         *ppos += *lenp;
2568         return err;
2569 }
2570
2571 static int do_proc_doulongvec_minmax(struct ctl_table *table, int write,
2572                                      void __user *buffer,
2573                                      size_t *lenp, loff_t *ppos,
2574                                      unsigned long convmul,
2575                                      unsigned long convdiv)
2576 {
2577         return __do_proc_doulongvec_minmax(table->data, table, write,
2578                         buffer, lenp, ppos, convmul, convdiv);
2579 }
2580
2581 /**
2582  * proc_doulongvec_minmax - read a vector of long integers with min/max values
2583  * @table: the sysctl table
2584  * @write: %TRUE if this is a write to the sysctl file
2585  * @buffer: the user buffer
2586  * @lenp: the size of the user buffer
2587  * @ppos: file position
2588  *
2589  * Reads/writes up to table->maxlen/sizeof(unsigned long) unsigned long
2590  * values from/to the user buffer, treated as an ASCII string.
2591  *
2592  * This routine will ensure the values are within the range specified by
2593  * table->extra1 (min) and table->extra2 (max).
2594  *
2595  * Returns 0 on success.
2596  */
2597 int proc_doulongvec_minmax(struct ctl_table *table, int write,
2598                            void __user *buffer, size_t *lenp, loff_t *ppos)
2599 {
2600     return do_proc_doulongvec_minmax(table, write, buffer, lenp, ppos, 1l, 1l);
2601 }
2602
2603 /**
2604  * proc_doulongvec_ms_jiffies_minmax - read a vector of millisecond values with min/max values
2605  * @table: the sysctl table
2606  * @write: %TRUE if this is a write to the sysctl file
2607  * @buffer: the user buffer
2608  * @lenp: the size of the user buffer
2609  * @ppos: file position
2610  *
2611  * Reads/writes up to table->maxlen/sizeof(unsigned long) unsigned long
2612  * values from/to the user buffer, treated as an ASCII string. The values
2613  * are treated as milliseconds, and converted to jiffies when they are stored.
2614  *
2615  * This routine will ensure the values are within the range specified by
2616  * table->extra1 (min) and table->extra2 (max).
2617  *
2618  * Returns 0 on success.
2619  */
2620 int proc_doulongvec_ms_jiffies_minmax(struct ctl_table *table, int write,
2621                                       void __user *buffer,
2622                                       size_t *lenp, loff_t *ppos)
2623 {
2624     return do_proc_doulongvec_minmax(table, write, buffer,
2625                                      lenp, ppos, HZ, 1000l);
2626 }
2627
2628
2629 static int do_proc_dointvec_jiffies_conv(bool *negp, unsigned long *lvalp,
2630                                          int *valp,
2631                                          int write, void *data)
2632 {
2633         if (write) {
2634                 if (*lvalp > LONG_MAX / HZ)
2635                         return 1;
2636                 *valp = *negp ? -(*lvalp*HZ) : (*lvalp*HZ);
2637         } else {
2638                 int val = *valp;
2639                 unsigned long lval;
2640                 if (val < 0) {
2641                         *negp = true;
2642                         lval = (unsigned long)-val;
2643                 } else {
2644                         *negp = false;
2645                         lval = (unsigned long)val;
2646                 }
2647                 *lvalp = lval / HZ;
2648         }
2649         return 0;
2650 }
2651
2652 static int do_proc_dointvec_userhz_jiffies_conv(bool *negp, unsigned long *lvalp,
2653                                                 int *valp,
2654                                                 int write, void *data)
2655 {
2656         if (write) {
2657                 if (USER_HZ < HZ && *lvalp > (LONG_MAX / HZ) * USER_HZ)
2658                         return 1;
2659                 *valp = clock_t_to_jiffies(*negp ? -*lvalp : *lvalp);
2660         } else {
2661                 int val = *valp;
2662                 unsigned long lval;
2663                 if (val < 0) {
2664                         *negp = true;
2665                         lval = (unsigned long)-val;
2666                 } else {
2667                         *negp = false;
2668                         lval = (unsigned long)val;
2669                 }
2670                 *lvalp = jiffies_to_clock_t(lval);
2671         }
2672         return 0;
2673 }
2674
2675 static int do_proc_dointvec_ms_jiffies_conv(bool *negp, unsigned long *lvalp,
2676                                             int *valp,
2677                                             int write, void *data)
2678 {
2679         if (write) {
2680                 *valp = msecs_to_jiffies(*negp ? -*lvalp : *lvalp);
2681         } else {
2682                 int val = *valp;
2683                 unsigned long lval;
2684                 if (val < 0) {
2685                         *negp = true;
2686                         lval = (unsigned long)-val;
2687                 } else {
2688                         *negp = false;
2689                         lval = (unsigned long)val;
2690                 }
2691                 *lvalp = jiffies_to_msecs(lval);
2692         }
2693         return 0;
2694 }
2695
2696 /**
2697  * proc_dointvec_jiffies - read a vector of integers as seconds
2698  * @table: the sysctl table
2699  * @write: %TRUE if this is a write to the sysctl file
2700  * @buffer: the user buffer
2701  * @lenp: the size of the user buffer
2702  * @ppos: file position
2703  *
2704  * Reads/writes up to table->maxlen/sizeof(unsigned int) integer
2705  * values from/to the user buffer, treated as an ASCII string. 
2706  * The values read are assumed to be in seconds, and are converted into
2707  * jiffies.
2708  *
2709  * Returns 0 on success.
2710  */
2711 int proc_dointvec_jiffies(struct ctl_table *table, int write,
2712                           void __user *buffer, size_t *lenp, loff_t *ppos)
2713 {
2714     return do_proc_dointvec(table,write,buffer,lenp,ppos,
2715                             do_proc_dointvec_jiffies_conv,NULL);
2716 }
2717
2718 /**
2719  * proc_dointvec_userhz_jiffies - read a vector of integers as 1/USER_HZ seconds
2720  * @table: the sysctl table
2721  * @write: %TRUE if this is a write to the sysctl file
2722  * @buffer: the user buffer
2723  * @lenp: the size of the user buffer
2724  * @ppos: pointer to the file position
2725  *
2726  * Reads/writes up to table->maxlen/sizeof(unsigned int) integer
2727  * values from/to the user buffer, treated as an ASCII string. 
2728  * The values read are assumed to be in 1/USER_HZ seconds, and 
2729  * are converted into jiffies.
2730  *
2731  * Returns 0 on success.
2732  */
2733 int proc_dointvec_userhz_jiffies(struct ctl_table *table, int write,
2734                                  void __user *buffer, size_t *lenp, loff_t *ppos)
2735 {
2736     return do_proc_dointvec(table,write,buffer,lenp,ppos,
2737                             do_proc_dointvec_userhz_jiffies_conv,NULL);
2738 }
2739
2740 /**
2741  * proc_dointvec_ms_jiffies - read a vector of integers as 1 milliseconds
2742  * @table: the sysctl table
2743  * @write: %TRUE if this is a write to the sysctl file
2744  * @buffer: the user buffer
2745  * @lenp: the size of the user buffer
2746  * @ppos: file position
2747  * @ppos: the current position in the file
2748  *
2749  * Reads/writes up to table->maxlen/sizeof(unsigned int) integer
2750  * values from/to the user buffer, treated as an ASCII string. 
2751  * The values read are assumed to be in 1/1000 seconds, and 
2752  * are converted into jiffies.
2753  *
2754  * Returns 0 on success.
2755  */
2756 int proc_dointvec_ms_jiffies(struct ctl_table *table, int write,
2757                              void __user *buffer, size_t *lenp, loff_t *ppos)
2758 {
2759         return do_proc_dointvec(table, write, buffer, lenp, ppos,
2760                                 do_proc_dointvec_ms_jiffies_conv, NULL);
2761 }
2762
2763 static int proc_do_cad_pid(struct ctl_table *table, int write,
2764                            void __user *buffer, size_t *lenp, loff_t *ppos)
2765 {
2766         struct pid *new_pid;
2767         pid_t tmp;
2768         int r;
2769
2770         tmp = pid_vnr(cad_pid);
2771
2772         r = __do_proc_dointvec(&tmp, table, write, buffer,
2773                                lenp, ppos, NULL, NULL);
2774         if (r || !write)
2775                 return r;
2776
2777         new_pid = find_get_pid(tmp);
2778         if (!new_pid)
2779                 return -ESRCH;
2780
2781         put_pid(xchg(&cad_pid, new_pid));
2782         return 0;
2783 }
2784
2785 /**
2786  * proc_do_large_bitmap - read/write from/to a large bitmap
2787  * @table: the sysctl table
2788  * @write: %TRUE if this is a write to the sysctl file
2789  * @buffer: the user buffer
2790  * @lenp: the size of the user buffer
2791  * @ppos: file position
2792  *
2793  * The bitmap is stored at table->data and the bitmap length (in bits)
2794  * in table->maxlen.
2795  *
2796  * We use a range comma separated format (e.g. 1,3-4,10-10) so that
2797  * large bitmaps may be represented in a compact manner. Writing into
2798  * the file will clear the bitmap then update it with the given input.
2799  *
2800  * Returns 0 on success.
2801  */
2802 int proc_do_large_bitmap(struct ctl_table *table, int write,
2803                          void __user *buffer, size_t *lenp, loff_t *ppos)
2804 {
2805         int err = 0;
2806         bool first = 1;
2807         size_t left = *lenp;
2808         unsigned long bitmap_len = table->maxlen;
2809         unsigned long *bitmap = (unsigned long *) table->data;
2810         unsigned long *tmp_bitmap = NULL;
2811         char tr_a[] = { '-', ',', '\n' }, tr_b[] = { ',', '\n', 0 }, c;
2812
2813         if (!bitmap_len || !left || (*ppos && !write)) {
2814                 *lenp = 0;
2815                 return 0;
2816         }
2817
2818         if (write) {
2819                 unsigned long page = 0;
2820                 char *kbuf;
2821
2822                 if (left > PAGE_SIZE - 1)
2823                         left = PAGE_SIZE - 1;
2824
2825                 page = __get_free_page(GFP_TEMPORARY);
2826                 kbuf = (char *) page;
2827                 if (!kbuf)
2828                         return -ENOMEM;
2829                 if (copy_from_user(kbuf, buffer, left)) {
2830                         free_page(page);
2831                         return -EFAULT;
2832                 }
2833                 kbuf[left] = 0;
2834
2835                 tmp_bitmap = kzalloc(BITS_TO_LONGS(bitmap_len) * sizeof(unsigned long),
2836                                      GFP_KERNEL);
2837                 if (!tmp_bitmap) {
2838                         free_page(page);
2839                         return -ENOMEM;
2840                 }
2841                 proc_skip_char(&kbuf, &left, '\n');
2842                 while (!err && left) {
2843                         unsigned long val_a, val_b;
2844                         bool neg;
2845
2846                         err = proc_get_long(&kbuf, &left, &val_a, &neg, tr_a,
2847                                              sizeof(tr_a), &c);
2848                         if (err)
2849                                 break;
2850                         if (val_a >= bitmap_len || neg) {
2851                                 err = -EINVAL;
2852                                 break;
2853                         }
2854
2855                         val_b = val_a;
2856                         if (left) {
2857                                 kbuf++;
2858                                 left--;
2859                         }
2860
2861                         if (c == '-') {
2862                                 err = proc_get_long(&kbuf, &left, &val_b,
2863                                                      &neg, tr_b, sizeof(tr_b),
2864                                                      &c);
2865                                 if (err)
2866                                         break;
2867                                 if (val_b >= bitmap_len || neg ||
2868                                     val_a > val_b) {
2869                                         err = -EINVAL;
2870                                         break;
2871                                 }
2872                                 if (left) {
2873                                         kbuf++;
2874                                         left--;
2875                                 }
2876                         }
2877
2878                         while (val_a <= val_b)
2879                                 set_bit(val_a++, tmp_bitmap);
2880
2881                         first = 0;
2882                         proc_skip_char(&kbuf, &left, '\n');
2883                 }
2884                 free_page(page);
2885         } else {
2886                 unsigned long bit_a, bit_b = 0;
2887
2888                 while (left) {
2889                         bit_a = find_next_bit(bitmap, bitmap_len, bit_b);
2890                         if (bit_a >= bitmap_len)
2891                                 break;
2892                         bit_b = find_next_zero_bit(bitmap, bitmap_len,
2893                                                    bit_a + 1) - 1;
2894
2895                         if (!first) {
2896                                 err = proc_put_char(&buffer, &left, ',');
2897                                 if (err)
2898                                         break;
2899                         }
2900                         err = proc_put_long(&buffer, &left, bit_a, false);
2901                         if (err)
2902                                 break;
2903                         if (bit_a != bit_b) {
2904                                 err = proc_put_char(&buffer, &left, '-');
2905                                 if (err)
2906                                         break;
2907                                 err = proc_put_long(&buffer, &left, bit_b, false);
2908                                 if (err)
2909                                         break;
2910                         }
2911
2912                         first = 0; bit_b++;
2913                 }
2914                 if (!err)
2915                         err = proc_put_char(&buffer, &left, '\n');
2916         }
2917
2918         if (!err) {
2919                 if (write) {
2920                         if (*ppos)
2921                                 bitmap_or(bitmap, bitmap, tmp_bitmap, bitmap_len);
2922                         else
2923                                 memcpy(bitmap, tmp_bitmap,
2924                                         BITS_TO_LONGS(bitmap_len) * sizeof(unsigned long));
2925                 }
2926                 kfree(tmp_bitmap);
2927                 *lenp -= left;
2928                 *ppos += *lenp;
2929                 return 0;
2930         } else {
2931                 kfree(tmp_bitmap);
2932                 return err;
2933         }
2934 }
2935
2936 #else /* CONFIG_PROC_SYSCTL */
2937
2938 int proc_dostring(struct ctl_table *table, int write,
2939                   void __user *buffer, size_t *lenp, loff_t *ppos)
2940 {
2941         return -ENOSYS;
2942 }
2943
2944 int proc_dointvec(struct ctl_table *table, int write,
2945                   void __user *buffer, size_t *lenp, loff_t *ppos)
2946 {
2947         return -ENOSYS;
2948 }
2949
2950 int proc_dointvec_minmax(struct ctl_table *table, int write,
2951                     void __user *buffer, size_t *lenp, loff_t *ppos)
2952 {
2953         return -ENOSYS;
2954 }
2955
2956 int proc_dointvec_jiffies(struct ctl_table *table, int write,
2957                     void __user *buffer, size_t *lenp, loff_t *ppos)
2958 {
2959         return -ENOSYS;
2960 }
2961
2962 int proc_dointvec_userhz_jiffies(struct ctl_table *table, int write,
2963                     void __user *buffer, size_t *lenp, loff_t *ppos)
2964 {
2965         return -ENOSYS;
2966 }
2967
2968 int proc_dointvec_ms_jiffies(struct ctl_table *table, int write,
2969                              void __user *buffer, size_t *lenp, loff_t *ppos)
2970 {
2971         return -ENOSYS;
2972 }
2973
2974 int proc_doulongvec_minmax(struct ctl_table *table, int write,
2975                     void __user *buffer, size_t *lenp, loff_t *ppos)
2976 {
2977         return -ENOSYS;
2978 }
2979
2980 int proc_doulongvec_ms_jiffies_minmax(struct ctl_table *table, int write,
2981                                       void __user *buffer,
2982                                       size_t *lenp, loff_t *ppos)
2983 {
2984     return -ENOSYS;
2985 }
2986
2987
2988 #endif /* CONFIG_PROC_SYSCTL */
2989
2990 /*
2991  * No sense putting this after each symbol definition, twice,
2992  * exception granted :-)
2993  */
2994 EXPORT_SYMBOL(proc_dointvec);
2995 EXPORT_SYMBOL(proc_dointvec_jiffies);
2996 EXPORT_SYMBOL(proc_dointvec_minmax);
2997 EXPORT_SYMBOL(proc_dointvec_userhz_jiffies);
2998 EXPORT_SYMBOL(proc_dointvec_ms_jiffies);
2999 EXPORT_SYMBOL(proc_dostring);
3000 EXPORT_SYMBOL(proc_doulongvec_minmax);
3001 EXPORT_SYMBOL(proc_doulongvec_ms_jiffies_minmax);
3002 EXPORT_SYMBOL(register_sysctl_table);
3003 EXPORT_SYMBOL(register_sysctl_paths);
3004 EXPORT_SYMBOL(unregister_sysctl_table);