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