]> git.openfabrics.org - ~shefty/rdma-dev.git/blob - kernel/sysctl.c
48d9d689498fbaa394e4d6ec8f4403c90767d7f1
[~shefty/rdma-dev.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/proc_fs.h>
28 #include <linux/security.h>
29 #include <linux/ctype.h>
30 #include <linux/kmemcheck.h>
31 #include <linux/fs.h>
32 #include <linux/init.h>
33 #include <linux/kernel.h>
34 #include <linux/kobject.h>
35 #include <linux/net.h>
36 #include <linux/sysrq.h>
37 #include <linux/highuid.h>
38 #include <linux/writeback.h>
39 #include <linux/ratelimit.h>
40 #include <linux/compaction.h>
41 #include <linux/hugetlb.h>
42 #include <linux/initrd.h>
43 #include <linux/key.h>
44 #include <linux/times.h>
45 #include <linux/limits.h>
46 #include <linux/dcache.h>
47 #include <linux/dnotify.h>
48 #include <linux/syscalls.h>
49 #include <linux/vmstat.h>
50 #include <linux/nfs_fs.h>
51 #include <linux/acpi.h>
52 #include <linux/reboot.h>
53 #include <linux/ftrace.h>
54 #include <linux/perf_event.h>
55 #include <linux/kprobes.h>
56 #include <linux/pipe_fs_i.h>
57 #include <linux/oom.h>
58
59 #include <asm/uaccess.h>
60 #include <asm/processor.h>
61
62 #ifdef CONFIG_X86
63 #include <asm/nmi.h>
64 #include <asm/stacktrace.h>
65 #include <asm/io.h>
66 #endif
67 #ifdef CONFIG_BSD_PROCESS_ACCT
68 #include <linux/acct.h>
69 #endif
70 #ifdef CONFIG_RT_MUTEXES
71 #include <linux/rtmutex.h>
72 #endif
73 #if defined(CONFIG_PROVE_LOCKING) || defined(CONFIG_LOCK_STAT)
74 #include <linux/lockdep.h>
75 #endif
76 #ifdef CONFIG_CHR_DEV_SG
77 #include <scsi/sg.h>
78 #endif
79
80 #ifdef CONFIG_LOCKUP_DETECTOR
81 #include <linux/nmi.h>
82 #endif
83
84
85 #if defined(CONFIG_SYSCTL)
86
87 /* External variables not in a header file. */
88 extern int sysctl_overcommit_memory;
89 extern int sysctl_overcommit_ratio;
90 extern int max_threads;
91 extern int core_uses_pid;
92 extern int suid_dumpable;
93 extern char core_pattern[];
94 extern unsigned int core_pipe_limit;
95 extern int pid_max;
96 extern int min_free_kbytes;
97 extern int pid_max_min, pid_max_max;
98 extern int sysctl_drop_caches;
99 extern int percpu_pagelist_fraction;
100 extern int compat_log;
101 extern int latencytop_enabled;
102 extern int sysctl_nr_open_min, sysctl_nr_open_max;
103 #ifndef CONFIG_MMU
104 extern int sysctl_nr_trim_pages;
105 #endif
106 #ifdef CONFIG_BLOCK
107 extern int blk_iopoll_enabled;
108 #endif
109
110 /* Constants used for minimum and  maximum */
111 #ifdef CONFIG_LOCKUP_DETECTOR
112 static int sixty = 60;
113 static int neg_one = -1;
114 #endif
115
116 static int zero;
117 static int __maybe_unused one = 1;
118 static int __maybe_unused two = 2;
119 static unsigned long one_ul = 1;
120 static int one_hundred = 100;
121 #ifdef CONFIG_PRINTK
122 static int ten_thousand = 10000;
123 #endif
124
125 /* this is needed for the proc_doulongvec_minmax of vm_dirty_bytes */
126 static unsigned long dirty_bytes_min = 2 * PAGE_SIZE;
127
128 /* this is needed for the proc_dointvec_minmax for [fs_]overflow UID and GID */
129 static int maxolduid = 65535;
130 static int minolduid;
131 static int min_percpu_pagelist_fract = 8;
132
133 static int ngroups_max = NGROUPS_MAX;
134
135 #ifdef CONFIG_INOTIFY_USER
136 #include <linux/inotify.h>
137 #endif
138 #ifdef CONFIG_SPARC
139 #include <asm/system.h>
140 #endif
141
142 #ifdef CONFIG_SPARC64
143 extern int sysctl_tsb_ratio;
144 #endif
145
146 #ifdef __hppa__
147 extern int pwrsw_enabled;
148 extern int unaligned_enabled;
149 #endif
150
151 #ifdef CONFIG_S390
152 #ifdef CONFIG_MATHEMU
153 extern int sysctl_ieee_emulation_warnings;
154 #endif
155 extern int sysctl_userprocess_debug;
156 extern int spin_retry;
157 #endif
158
159 #ifdef CONFIG_IA64
160 extern int no_unaligned_warning;
161 extern int unaligned_dump_stack;
162 #endif
163
164 #ifdef CONFIG_PROC_SYSCTL
165 static int proc_do_cad_pid(struct ctl_table *table, int write,
166                   void __user *buffer, size_t *lenp, loff_t *ppos);
167 static int proc_taint(struct ctl_table *table, int write,
168                                void __user *buffer, size_t *lenp, loff_t *ppos);
169 #endif
170
171 #ifdef CONFIG_MAGIC_SYSRQ
172 static int __sysrq_enabled; /* Note: sysrq code ises it's own private copy */
173
174 static int sysrq_sysctl_handler(ctl_table *table, int write,
175                                 void __user *buffer, size_t *lenp,
176                                 loff_t *ppos)
177 {
178         int error;
179
180         error = proc_dointvec(table, write, buffer, lenp, ppos);
181         if (error)
182                 return error;
183
184         if (write)
185                 sysrq_toggle_support(__sysrq_enabled);
186
187         return 0;
188 }
189
190 #endif
191
192 static struct ctl_table root_table[];
193 static struct ctl_table_root sysctl_table_root;
194 static struct ctl_table_header root_table_header = {
195         .count = 1,
196         .ctl_table = root_table,
197         .ctl_entry = LIST_HEAD_INIT(sysctl_table_root.default_set.list),
198         .root = &sysctl_table_root,
199         .set = &sysctl_table_root.default_set,
200 };
201 static struct ctl_table_root sysctl_table_root = {
202         .root_list = LIST_HEAD_INIT(sysctl_table_root.root_list),
203         .default_set.list = LIST_HEAD_INIT(root_table_header.ctl_entry),
204 };
205
206 static struct ctl_table kern_table[];
207 static struct ctl_table vm_table[];
208 static struct ctl_table fs_table[];
209 static struct ctl_table debug_table[];
210 static struct ctl_table dev_table[];
211 extern struct ctl_table random_table[];
212 #ifdef CONFIG_EPOLL
213 extern struct ctl_table epoll_table[];
214 #endif
215
216 #ifdef HAVE_ARCH_PICK_MMAP_LAYOUT
217 int sysctl_legacy_va_layout;
218 #endif
219
220 /* The default sysctl tables: */
221
222 static struct ctl_table root_table[] = {
223         {
224                 .procname       = "kernel",
225                 .mode           = 0555,
226                 .child          = kern_table,
227         },
228         {
229                 .procname       = "vm",
230                 .mode           = 0555,
231                 .child          = vm_table,
232         },
233         {
234                 .procname       = "fs",
235                 .mode           = 0555,
236                 .child          = fs_table,
237         },
238         {
239                 .procname       = "debug",
240                 .mode           = 0555,
241                 .child          = debug_table,
242         },
243         {
244                 .procname       = "dev",
245                 .mode           = 0555,
246                 .child          = dev_table,
247         },
248 /*
249  * NOTE: do not add new entries to this table unless you have read
250  * Documentation/sysctl/ctl_unnumbered.txt
251  */
252         { }
253 };
254
255 #ifdef CONFIG_SCHED_DEBUG
256 static int min_sched_granularity_ns = 100000;           /* 100 usecs */
257 static int max_sched_granularity_ns = NSEC_PER_SEC;     /* 1 second */
258 static int min_wakeup_granularity_ns;                   /* 0 usecs */
259 static int max_wakeup_granularity_ns = NSEC_PER_SEC;    /* 1 second */
260 static int min_sched_tunable_scaling = SCHED_TUNABLESCALING_NONE;
261 static int max_sched_tunable_scaling = SCHED_TUNABLESCALING_END-1;
262 static int min_sched_shares_ratelimit = 100000; /* 100 usec */
263 static int max_sched_shares_ratelimit = NSEC_PER_SEC; /* 1 second */
264 #endif
265
266 #ifdef CONFIG_COMPACTION
267 static int min_extfrag_threshold;
268 static int max_extfrag_threshold = 1000;
269 #endif
270
271 static struct ctl_table kern_table[] = {
272         {
273                 .procname       = "sched_child_runs_first",
274                 .data           = &sysctl_sched_child_runs_first,
275                 .maxlen         = sizeof(unsigned int),
276                 .mode           = 0644,
277                 .proc_handler   = proc_dointvec,
278         },
279 #ifdef CONFIG_SCHED_DEBUG
280         {
281                 .procname       = "sched_min_granularity_ns",
282                 .data           = &sysctl_sched_min_granularity,
283                 .maxlen         = sizeof(unsigned int),
284                 .mode           = 0644,
285                 .proc_handler   = sched_proc_update_handler,
286                 .extra1         = &min_sched_granularity_ns,
287                 .extra2         = &max_sched_granularity_ns,
288         },
289         {
290                 .procname       = "sched_latency_ns",
291                 .data           = &sysctl_sched_latency,
292                 .maxlen         = sizeof(unsigned int),
293                 .mode           = 0644,
294                 .proc_handler   = sched_proc_update_handler,
295                 .extra1         = &min_sched_granularity_ns,
296                 .extra2         = &max_sched_granularity_ns,
297         },
298         {
299                 .procname       = "sched_wakeup_granularity_ns",
300                 .data           = &sysctl_sched_wakeup_granularity,
301                 .maxlen         = sizeof(unsigned int),
302                 .mode           = 0644,
303                 .proc_handler   = sched_proc_update_handler,
304                 .extra1         = &min_wakeup_granularity_ns,
305                 .extra2         = &max_wakeup_granularity_ns,
306         },
307         {
308                 .procname       = "sched_shares_ratelimit",
309                 .data           = &sysctl_sched_shares_ratelimit,
310                 .maxlen         = sizeof(unsigned int),
311                 .mode           = 0644,
312                 .proc_handler   = sched_proc_update_handler,
313                 .extra1         = &min_sched_shares_ratelimit,
314                 .extra2         = &max_sched_shares_ratelimit,
315         },
316         {
317                 .procname       = "sched_tunable_scaling",
318                 .data           = &sysctl_sched_tunable_scaling,
319                 .maxlen         = sizeof(enum sched_tunable_scaling),
320                 .mode           = 0644,
321                 .proc_handler   = sched_proc_update_handler,
322                 .extra1         = &min_sched_tunable_scaling,
323                 .extra2         = &max_sched_tunable_scaling,
324         },
325         {
326                 .procname       = "sched_shares_thresh",
327                 .data           = &sysctl_sched_shares_thresh,
328                 .maxlen         = sizeof(unsigned int),
329                 .mode           = 0644,
330                 .proc_handler   = proc_dointvec_minmax,
331                 .extra1         = &zero,
332         },
333         {
334                 .procname       = "sched_migration_cost",
335                 .data           = &sysctl_sched_migration_cost,
336                 .maxlen         = sizeof(unsigned int),
337                 .mode           = 0644,
338                 .proc_handler   = proc_dointvec,
339         },
340         {
341                 .procname       = "sched_nr_migrate",
342                 .data           = &sysctl_sched_nr_migrate,
343                 .maxlen         = sizeof(unsigned int),
344                 .mode           = 0644,
345                 .proc_handler   = proc_dointvec,
346         },
347         {
348                 .procname       = "sched_time_avg",
349                 .data           = &sysctl_sched_time_avg,
350                 .maxlen         = sizeof(unsigned int),
351                 .mode           = 0644,
352                 .proc_handler   = proc_dointvec,
353         },
354         {
355                 .procname       = "timer_migration",
356                 .data           = &sysctl_timer_migration,
357                 .maxlen         = sizeof(unsigned int),
358                 .mode           = 0644,
359                 .proc_handler   = proc_dointvec_minmax,
360                 .extra1         = &zero,
361                 .extra2         = &one,
362         },
363 #endif
364         {
365                 .procname       = "sched_rt_period_us",
366                 .data           = &sysctl_sched_rt_period,
367                 .maxlen         = sizeof(unsigned int),
368                 .mode           = 0644,
369                 .proc_handler   = sched_rt_handler,
370         },
371         {
372                 .procname       = "sched_rt_runtime_us",
373                 .data           = &sysctl_sched_rt_runtime,
374                 .maxlen         = sizeof(int),
375                 .mode           = 0644,
376                 .proc_handler   = sched_rt_handler,
377         },
378         {
379                 .procname       = "sched_compat_yield",
380                 .data           = &sysctl_sched_compat_yield,
381                 .maxlen         = sizeof(unsigned int),
382                 .mode           = 0644,
383                 .proc_handler   = proc_dointvec,
384         },
385 #ifdef CONFIG_PROVE_LOCKING
386         {
387                 .procname       = "prove_locking",
388                 .data           = &prove_locking,
389                 .maxlen         = sizeof(int),
390                 .mode           = 0644,
391                 .proc_handler   = proc_dointvec,
392         },
393 #endif
394 #ifdef CONFIG_LOCK_STAT
395         {
396                 .procname       = "lock_stat",
397                 .data           = &lock_stat,
398                 .maxlen         = sizeof(int),
399                 .mode           = 0644,
400                 .proc_handler   = proc_dointvec,
401         },
402 #endif
403         {
404                 .procname       = "panic",
405                 .data           = &panic_timeout,
406                 .maxlen         = sizeof(int),
407                 .mode           = 0644,
408                 .proc_handler   = proc_dointvec,
409         },
410         {
411                 .procname       = "core_uses_pid",
412                 .data           = &core_uses_pid,
413                 .maxlen         = sizeof(int),
414                 .mode           = 0644,
415                 .proc_handler   = proc_dointvec,
416         },
417         {
418                 .procname       = "core_pattern",
419                 .data           = core_pattern,
420                 .maxlen         = CORENAME_MAX_SIZE,
421                 .mode           = 0644,
422                 .proc_handler   = proc_dostring,
423         },
424         {
425                 .procname       = "core_pipe_limit",
426                 .data           = &core_pipe_limit,
427                 .maxlen         = sizeof(unsigned int),
428                 .mode           = 0644,
429                 .proc_handler   = proc_dointvec,
430         },
431 #ifdef CONFIG_PROC_SYSCTL
432         {
433                 .procname       = "tainted",
434                 .maxlen         = sizeof(long),
435                 .mode           = 0644,
436                 .proc_handler   = proc_taint,
437         },
438 #endif
439 #ifdef CONFIG_LATENCYTOP
440         {
441                 .procname       = "latencytop",
442                 .data           = &latencytop_enabled,
443                 .maxlen         = sizeof(int),
444                 .mode           = 0644,
445                 .proc_handler   = proc_dointvec,
446         },
447 #endif
448 #ifdef CONFIG_BLK_DEV_INITRD
449         {
450                 .procname       = "real-root-dev",
451                 .data           = &real_root_dev,
452                 .maxlen         = sizeof(int),
453                 .mode           = 0644,
454                 .proc_handler   = proc_dointvec,
455         },
456 #endif
457         {
458                 .procname       = "print-fatal-signals",
459                 .data           = &print_fatal_signals,
460                 .maxlen         = sizeof(int),
461                 .mode           = 0644,
462                 .proc_handler   = proc_dointvec,
463         },
464 #ifdef CONFIG_SPARC
465         {
466                 .procname       = "reboot-cmd",
467                 .data           = reboot_command,
468                 .maxlen         = 256,
469                 .mode           = 0644,
470                 .proc_handler   = proc_dostring,
471         },
472         {
473                 .procname       = "stop-a",
474                 .data           = &stop_a_enabled,
475                 .maxlen         = sizeof (int),
476                 .mode           = 0644,
477                 .proc_handler   = proc_dointvec,
478         },
479         {
480                 .procname       = "scons-poweroff",
481                 .data           = &scons_pwroff,
482                 .maxlen         = sizeof (int),
483                 .mode           = 0644,
484                 .proc_handler   = proc_dointvec,
485         },
486 #endif
487 #ifdef CONFIG_SPARC64
488         {
489                 .procname       = "tsb-ratio",
490                 .data           = &sysctl_tsb_ratio,
491                 .maxlen         = sizeof (int),
492                 .mode           = 0644,
493                 .proc_handler   = proc_dointvec,
494         },
495 #endif
496 #ifdef __hppa__
497         {
498                 .procname       = "soft-power",
499                 .data           = &pwrsw_enabled,
500                 .maxlen         = sizeof (int),
501                 .mode           = 0644,
502                 .proc_handler   = proc_dointvec,
503         },
504         {
505                 .procname       = "unaligned-trap",
506                 .data           = &unaligned_enabled,
507                 .maxlen         = sizeof (int),
508                 .mode           = 0644,
509                 .proc_handler   = proc_dointvec,
510         },
511 #endif
512         {
513                 .procname       = "ctrl-alt-del",
514                 .data           = &C_A_D,
515                 .maxlen         = sizeof(int),
516                 .mode           = 0644,
517                 .proc_handler   = proc_dointvec,
518         },
519 #ifdef CONFIG_FUNCTION_TRACER
520         {
521                 .procname       = "ftrace_enabled",
522                 .data           = &ftrace_enabled,
523                 .maxlen         = sizeof(int),
524                 .mode           = 0644,
525                 .proc_handler   = ftrace_enable_sysctl,
526         },
527 #endif
528 #ifdef CONFIG_STACK_TRACER
529         {
530                 .procname       = "stack_tracer_enabled",
531                 .data           = &stack_tracer_enabled,
532                 .maxlen         = sizeof(int),
533                 .mode           = 0644,
534                 .proc_handler   = stack_trace_sysctl,
535         },
536 #endif
537 #ifdef CONFIG_TRACING
538         {
539                 .procname       = "ftrace_dump_on_oops",
540                 .data           = &ftrace_dump_on_oops,
541                 .maxlen         = sizeof(int),
542                 .mode           = 0644,
543                 .proc_handler   = proc_dointvec,
544         },
545 #endif
546 #ifdef CONFIG_MODULES
547         {
548                 .procname       = "modprobe",
549                 .data           = &modprobe_path,
550                 .maxlen         = KMOD_PATH_LEN,
551                 .mode           = 0644,
552                 .proc_handler   = proc_dostring,
553         },
554         {
555                 .procname       = "modules_disabled",
556                 .data           = &modules_disabled,
557                 .maxlen         = sizeof(int),
558                 .mode           = 0644,
559                 /* only handle a transition from default "0" to "1" */
560                 .proc_handler   = proc_dointvec_minmax,
561                 .extra1         = &one,
562                 .extra2         = &one,
563         },
564 #endif
565 #ifdef CONFIG_HOTPLUG
566         {
567                 .procname       = "hotplug",
568                 .data           = &uevent_helper,
569                 .maxlen         = UEVENT_HELPER_PATH_LEN,
570                 .mode           = 0644,
571                 .proc_handler   = proc_dostring,
572         },
573 #endif
574 #ifdef CONFIG_CHR_DEV_SG
575         {
576                 .procname       = "sg-big-buff",
577                 .data           = &sg_big_buff,
578                 .maxlen         = sizeof (int),
579                 .mode           = 0444,
580                 .proc_handler   = proc_dointvec,
581         },
582 #endif
583 #ifdef CONFIG_BSD_PROCESS_ACCT
584         {
585                 .procname       = "acct",
586                 .data           = &acct_parm,
587                 .maxlen         = 3*sizeof(int),
588                 .mode           = 0644,
589                 .proc_handler   = proc_dointvec,
590         },
591 #endif
592 #ifdef CONFIG_MAGIC_SYSRQ
593         {
594                 .procname       = "sysrq",
595                 .data           = &__sysrq_enabled,
596                 .maxlen         = sizeof (int),
597                 .mode           = 0644,
598                 .proc_handler   = sysrq_sysctl_handler,
599         },
600 #endif
601 #ifdef CONFIG_PROC_SYSCTL
602         {
603                 .procname       = "cad_pid",
604                 .data           = NULL,
605                 .maxlen         = sizeof (int),
606                 .mode           = 0600,
607                 .proc_handler   = proc_do_cad_pid,
608         },
609 #endif
610         {
611                 .procname       = "threads-max",
612                 .data           = &max_threads,
613                 .maxlen         = sizeof(int),
614                 .mode           = 0644,
615                 .proc_handler   = proc_dointvec,
616         },
617         {
618                 .procname       = "random",
619                 .mode           = 0555,
620                 .child          = random_table,
621         },
622         {
623                 .procname       = "overflowuid",
624                 .data           = &overflowuid,
625                 .maxlen         = sizeof(int),
626                 .mode           = 0644,
627                 .proc_handler   = proc_dointvec_minmax,
628                 .extra1         = &minolduid,
629                 .extra2         = &maxolduid,
630         },
631         {
632                 .procname       = "overflowgid",
633                 .data           = &overflowgid,
634                 .maxlen         = sizeof(int),
635                 .mode           = 0644,
636                 .proc_handler   = proc_dointvec_minmax,
637                 .extra1         = &minolduid,
638                 .extra2         = &maxolduid,
639         },
640 #ifdef CONFIG_S390
641 #ifdef CONFIG_MATHEMU
642         {
643                 .procname       = "ieee_emulation_warnings",
644                 .data           = &sysctl_ieee_emulation_warnings,
645                 .maxlen         = sizeof(int),
646                 .mode           = 0644,
647                 .proc_handler   = proc_dointvec,
648         },
649 #endif
650         {
651                 .procname       = "userprocess_debug",
652                 .data           = &show_unhandled_signals,
653                 .maxlen         = sizeof(int),
654                 .mode           = 0644,
655                 .proc_handler   = proc_dointvec,
656         },
657 #endif
658         {
659                 .procname       = "pid_max",
660                 .data           = &pid_max,
661                 .maxlen         = sizeof (int),
662                 .mode           = 0644,
663                 .proc_handler   = proc_dointvec_minmax,
664                 .extra1         = &pid_max_min,
665                 .extra2         = &pid_max_max,
666         },
667         {
668                 .procname       = "panic_on_oops",
669                 .data           = &panic_on_oops,
670                 .maxlen         = sizeof(int),
671                 .mode           = 0644,
672                 .proc_handler   = proc_dointvec,
673         },
674 #if defined CONFIG_PRINTK
675         {
676                 .procname       = "printk",
677                 .data           = &console_loglevel,
678                 .maxlen         = 4*sizeof(int),
679                 .mode           = 0644,
680                 .proc_handler   = proc_dointvec,
681         },
682         {
683                 .procname       = "printk_ratelimit",
684                 .data           = &printk_ratelimit_state.interval,
685                 .maxlen         = sizeof(int),
686                 .mode           = 0644,
687                 .proc_handler   = proc_dointvec_jiffies,
688         },
689         {
690                 .procname       = "printk_ratelimit_burst",
691                 .data           = &printk_ratelimit_state.burst,
692                 .maxlen         = sizeof(int),
693                 .mode           = 0644,
694                 .proc_handler   = proc_dointvec,
695         },
696         {
697                 .procname       = "printk_delay",
698                 .data           = &printk_delay_msec,
699                 .maxlen         = sizeof(int),
700                 .mode           = 0644,
701                 .proc_handler   = proc_dointvec_minmax,
702                 .extra1         = &zero,
703                 .extra2         = &ten_thousand,
704         },
705 #endif
706         {
707                 .procname       = "ngroups_max",
708                 .data           = &ngroups_max,
709                 .maxlen         = sizeof (int),
710                 .mode           = 0444,
711                 .proc_handler   = proc_dointvec,
712         },
713 #if defined(CONFIG_LOCKUP_DETECTOR)
714         {
715                 .procname       = "watchdog",
716                 .data           = &watchdog_enabled,
717                 .maxlen         = sizeof (int),
718                 .mode           = 0644,
719                 .proc_handler   = proc_dowatchdog_enabled,
720         },
721         {
722                 .procname       = "watchdog_thresh",
723                 .data           = &softlockup_thresh,
724                 .maxlen         = sizeof(int),
725                 .mode           = 0644,
726                 .proc_handler   = proc_dowatchdog_thresh,
727                 .extra1         = &neg_one,
728                 .extra2         = &sixty,
729         },
730         {
731                 .procname       = "softlockup_panic",
732                 .data           = &softlockup_panic,
733                 .maxlen         = sizeof(int),
734                 .mode           = 0644,
735                 .proc_handler   = proc_dointvec_minmax,
736                 .extra1         = &zero,
737                 .extra2         = &one,
738         },
739 #endif
740 #if defined(CONFIG_X86_LOCAL_APIC) && defined(CONFIG_X86) && !defined(CONFIG_LOCKUP_DETECTOR)
741         {
742                 .procname       = "unknown_nmi_panic",
743                 .data           = &unknown_nmi_panic,
744                 .maxlen         = sizeof (int),
745                 .mode           = 0644,
746                 .proc_handler   = proc_dointvec,
747         },
748         {
749                 .procname       = "nmi_watchdog",
750                 .data           = &nmi_watchdog_enabled,
751                 .maxlen         = sizeof (int),
752                 .mode           = 0644,
753                 .proc_handler   = proc_nmi_enabled,
754         },
755 #endif
756 #if defined(CONFIG_X86)
757         {
758                 .procname       = "panic_on_unrecovered_nmi",
759                 .data           = &panic_on_unrecovered_nmi,
760                 .maxlen         = sizeof(int),
761                 .mode           = 0644,
762                 .proc_handler   = proc_dointvec,
763         },
764         {
765                 .procname       = "panic_on_io_nmi",
766                 .data           = &panic_on_io_nmi,
767                 .maxlen         = sizeof(int),
768                 .mode           = 0644,
769                 .proc_handler   = proc_dointvec,
770         },
771         {
772                 .procname       = "bootloader_type",
773                 .data           = &bootloader_type,
774                 .maxlen         = sizeof (int),
775                 .mode           = 0444,
776                 .proc_handler   = proc_dointvec,
777         },
778         {
779                 .procname       = "bootloader_version",
780                 .data           = &bootloader_version,
781                 .maxlen         = sizeof (int),
782                 .mode           = 0444,
783                 .proc_handler   = proc_dointvec,
784         },
785         {
786                 .procname       = "kstack_depth_to_print",
787                 .data           = &kstack_depth_to_print,
788                 .maxlen         = sizeof(int),
789                 .mode           = 0644,
790                 .proc_handler   = proc_dointvec,
791         },
792         {
793                 .procname       = "io_delay_type",
794                 .data           = &io_delay_type,
795                 .maxlen         = sizeof(int),
796                 .mode           = 0644,
797                 .proc_handler   = proc_dointvec,
798         },
799 #endif
800 #if defined(CONFIG_MMU)
801         {
802                 .procname       = "randomize_va_space",
803                 .data           = &randomize_va_space,
804                 .maxlen         = sizeof(int),
805                 .mode           = 0644,
806                 .proc_handler   = proc_dointvec,
807         },
808 #endif
809 #if defined(CONFIG_S390) && defined(CONFIG_SMP)
810         {
811                 .procname       = "spin_retry",
812                 .data           = &spin_retry,
813                 .maxlen         = sizeof (int),
814                 .mode           = 0644,
815                 .proc_handler   = proc_dointvec,
816         },
817 #endif
818 #if     defined(CONFIG_ACPI_SLEEP) && defined(CONFIG_X86)
819         {
820                 .procname       = "acpi_video_flags",
821                 .data           = &acpi_realmode_flags,
822                 .maxlen         = sizeof (unsigned long),
823                 .mode           = 0644,
824                 .proc_handler   = proc_doulongvec_minmax,
825         },
826 #endif
827 #ifdef CONFIG_IA64
828         {
829                 .procname       = "ignore-unaligned-usertrap",
830                 .data           = &no_unaligned_warning,
831                 .maxlen         = sizeof (int),
832                 .mode           = 0644,
833                 .proc_handler   = proc_dointvec,
834         },
835         {
836                 .procname       = "unaligned-dump-stack",
837                 .data           = &unaligned_dump_stack,
838                 .maxlen         = sizeof (int),
839                 .mode           = 0644,
840                 .proc_handler   = proc_dointvec,
841         },
842 #endif
843 #ifdef CONFIG_DETECT_HUNG_TASK
844         {
845                 .procname       = "hung_task_panic",
846                 .data           = &sysctl_hung_task_panic,
847                 .maxlen         = sizeof(int),
848                 .mode           = 0644,
849                 .proc_handler   = proc_dointvec_minmax,
850                 .extra1         = &zero,
851                 .extra2         = &one,
852         },
853         {
854                 .procname       = "hung_task_check_count",
855                 .data           = &sysctl_hung_task_check_count,
856                 .maxlen         = sizeof(unsigned long),
857                 .mode           = 0644,
858                 .proc_handler   = proc_doulongvec_minmax,
859         },
860         {
861                 .procname       = "hung_task_timeout_secs",
862                 .data           = &sysctl_hung_task_timeout_secs,
863                 .maxlen         = sizeof(unsigned long),
864                 .mode           = 0644,
865                 .proc_handler   = proc_dohung_task_timeout_secs,
866         },
867         {
868                 .procname       = "hung_task_warnings",
869                 .data           = &sysctl_hung_task_warnings,
870                 .maxlen         = sizeof(unsigned long),
871                 .mode           = 0644,
872                 .proc_handler   = proc_doulongvec_minmax,
873         },
874 #endif
875 #ifdef CONFIG_COMPAT
876         {
877                 .procname       = "compat-log",
878                 .data           = &compat_log,
879                 .maxlen         = sizeof (int),
880                 .mode           = 0644,
881                 .proc_handler   = proc_dointvec,
882         },
883 #endif
884 #ifdef CONFIG_RT_MUTEXES
885         {
886                 .procname       = "max_lock_depth",
887                 .data           = &max_lock_depth,
888                 .maxlen         = sizeof(int),
889                 .mode           = 0644,
890                 .proc_handler   = proc_dointvec,
891         },
892 #endif
893         {
894                 .procname       = "poweroff_cmd",
895                 .data           = &poweroff_cmd,
896                 .maxlen         = POWEROFF_CMD_PATH_LEN,
897                 .mode           = 0644,
898                 .proc_handler   = proc_dostring,
899         },
900 #ifdef CONFIG_KEYS
901         {
902                 .procname       = "keys",
903                 .mode           = 0555,
904                 .child          = key_sysctls,
905         },
906 #endif
907 #ifdef CONFIG_RCU_TORTURE_TEST
908         {
909                 .procname       = "rcutorture_runnable",
910                 .data           = &rcutorture_runnable,
911                 .maxlen         = sizeof(int),
912                 .mode           = 0644,
913                 .proc_handler   = proc_dointvec,
914         },
915 #endif
916 #ifdef CONFIG_PERF_EVENTS
917         {
918                 .procname       = "perf_event_paranoid",
919                 .data           = &sysctl_perf_event_paranoid,
920                 .maxlen         = sizeof(sysctl_perf_event_paranoid),
921                 .mode           = 0644,
922                 .proc_handler   = proc_dointvec,
923         },
924         {
925                 .procname       = "perf_event_mlock_kb",
926                 .data           = &sysctl_perf_event_mlock,
927                 .maxlen         = sizeof(sysctl_perf_event_mlock),
928                 .mode           = 0644,
929                 .proc_handler   = proc_dointvec,
930         },
931         {
932                 .procname       = "perf_event_max_sample_rate",
933                 .data           = &sysctl_perf_event_sample_rate,
934                 .maxlen         = sizeof(sysctl_perf_event_sample_rate),
935                 .mode           = 0644,
936                 .proc_handler   = proc_dointvec,
937         },
938 #endif
939 #ifdef CONFIG_KMEMCHECK
940         {
941                 .procname       = "kmemcheck",
942                 .data           = &kmemcheck_enabled,
943                 .maxlen         = sizeof(int),
944                 .mode           = 0644,
945                 .proc_handler   = proc_dointvec,
946         },
947 #endif
948 #ifdef CONFIG_BLOCK
949         {
950                 .procname       = "blk_iopoll",
951                 .data           = &blk_iopoll_enabled,
952                 .maxlen         = sizeof(int),
953                 .mode           = 0644,
954                 .proc_handler   = proc_dointvec,
955         },
956 #endif
957 /*
958  * NOTE: do not add new entries to this table unless you have read
959  * Documentation/sysctl/ctl_unnumbered.txt
960  */
961         { }
962 };
963
964 static struct ctl_table vm_table[] = {
965         {
966                 .procname       = "overcommit_memory",
967                 .data           = &sysctl_overcommit_memory,
968                 .maxlen         = sizeof(sysctl_overcommit_memory),
969                 .mode           = 0644,
970                 .proc_handler   = proc_dointvec,
971         },
972         {
973                 .procname       = "panic_on_oom",
974                 .data           = &sysctl_panic_on_oom,
975                 .maxlen         = sizeof(sysctl_panic_on_oom),
976                 .mode           = 0644,
977                 .proc_handler   = proc_dointvec,
978         },
979         {
980                 .procname       = "oom_kill_allocating_task",
981                 .data           = &sysctl_oom_kill_allocating_task,
982                 .maxlen         = sizeof(sysctl_oom_kill_allocating_task),
983                 .mode           = 0644,
984                 .proc_handler   = proc_dointvec,
985         },
986         {
987                 .procname       = "oom_dump_tasks",
988                 .data           = &sysctl_oom_dump_tasks,
989                 .maxlen         = sizeof(sysctl_oom_dump_tasks),
990                 .mode           = 0644,
991                 .proc_handler   = proc_dointvec,
992         },
993         {
994                 .procname       = "overcommit_ratio",
995                 .data           = &sysctl_overcommit_ratio,
996                 .maxlen         = sizeof(sysctl_overcommit_ratio),
997                 .mode           = 0644,
998                 .proc_handler   = proc_dointvec,
999         },
1000         {
1001                 .procname       = "page-cluster", 
1002                 .data           = &page_cluster,
1003                 .maxlen         = sizeof(int),
1004                 .mode           = 0644,
1005                 .proc_handler   = proc_dointvec,
1006         },
1007         {
1008                 .procname       = "dirty_background_ratio",
1009                 .data           = &dirty_background_ratio,
1010                 .maxlen         = sizeof(dirty_background_ratio),
1011                 .mode           = 0644,
1012                 .proc_handler   = dirty_background_ratio_handler,
1013                 .extra1         = &zero,
1014                 .extra2         = &one_hundred,
1015         },
1016         {
1017                 .procname       = "dirty_background_bytes",
1018                 .data           = &dirty_background_bytes,
1019                 .maxlen         = sizeof(dirty_background_bytes),
1020                 .mode           = 0644,
1021                 .proc_handler   = dirty_background_bytes_handler,
1022                 .extra1         = &one_ul,
1023         },
1024         {
1025                 .procname       = "dirty_ratio",
1026                 .data           = &vm_dirty_ratio,
1027                 .maxlen         = sizeof(vm_dirty_ratio),
1028                 .mode           = 0644,
1029                 .proc_handler   = dirty_ratio_handler,
1030                 .extra1         = &zero,
1031                 .extra2         = &one_hundred,
1032         },
1033         {
1034                 .procname       = "dirty_bytes",
1035                 .data           = &vm_dirty_bytes,
1036                 .maxlen         = sizeof(vm_dirty_bytes),
1037                 .mode           = 0644,
1038                 .proc_handler   = dirty_bytes_handler,
1039                 .extra1         = &dirty_bytes_min,
1040         },
1041         {
1042                 .procname       = "dirty_writeback_centisecs",
1043                 .data           = &dirty_writeback_interval,
1044                 .maxlen         = sizeof(dirty_writeback_interval),
1045                 .mode           = 0644,
1046                 .proc_handler   = dirty_writeback_centisecs_handler,
1047         },
1048         {
1049                 .procname       = "dirty_expire_centisecs",
1050                 .data           = &dirty_expire_interval,
1051                 .maxlen         = sizeof(dirty_expire_interval),
1052                 .mode           = 0644,
1053                 .proc_handler   = proc_dointvec,
1054         },
1055         {
1056                 .procname       = "nr_pdflush_threads",
1057                 .data           = &nr_pdflush_threads,
1058                 .maxlen         = sizeof nr_pdflush_threads,
1059                 .mode           = 0444 /* read-only*/,
1060                 .proc_handler   = proc_dointvec,
1061         },
1062         {
1063                 .procname       = "swappiness",
1064                 .data           = &vm_swappiness,
1065                 .maxlen         = sizeof(vm_swappiness),
1066                 .mode           = 0644,
1067                 .proc_handler   = proc_dointvec_minmax,
1068                 .extra1         = &zero,
1069                 .extra2         = &one_hundred,
1070         },
1071 #ifdef CONFIG_HUGETLB_PAGE
1072         {
1073                 .procname       = "nr_hugepages",
1074                 .data           = NULL,
1075                 .maxlen         = sizeof(unsigned long),
1076                 .mode           = 0644,
1077                 .proc_handler   = hugetlb_sysctl_handler,
1078                 .extra1         = (void *)&hugetlb_zero,
1079                 .extra2         = (void *)&hugetlb_infinity,
1080         },
1081 #ifdef CONFIG_NUMA
1082         {
1083                 .procname       = "nr_hugepages_mempolicy",
1084                 .data           = NULL,
1085                 .maxlen         = sizeof(unsigned long),
1086                 .mode           = 0644,
1087                 .proc_handler   = &hugetlb_mempolicy_sysctl_handler,
1088                 .extra1         = (void *)&hugetlb_zero,
1089                 .extra2         = (void *)&hugetlb_infinity,
1090         },
1091 #endif
1092          {
1093                 .procname       = "hugetlb_shm_group",
1094                 .data           = &sysctl_hugetlb_shm_group,
1095                 .maxlen         = sizeof(gid_t),
1096                 .mode           = 0644,
1097                 .proc_handler   = proc_dointvec,
1098          },
1099          {
1100                 .procname       = "hugepages_treat_as_movable",
1101                 .data           = &hugepages_treat_as_movable,
1102                 .maxlen         = sizeof(int),
1103                 .mode           = 0644,
1104                 .proc_handler   = hugetlb_treat_movable_handler,
1105         },
1106         {
1107                 .procname       = "nr_overcommit_hugepages",
1108                 .data           = NULL,
1109                 .maxlen         = sizeof(unsigned long),
1110                 .mode           = 0644,
1111                 .proc_handler   = hugetlb_overcommit_handler,
1112                 .extra1         = (void *)&hugetlb_zero,
1113                 .extra2         = (void *)&hugetlb_infinity,
1114         },
1115 #endif
1116         {
1117                 .procname       = "lowmem_reserve_ratio",
1118                 .data           = &sysctl_lowmem_reserve_ratio,
1119                 .maxlen         = sizeof(sysctl_lowmem_reserve_ratio),
1120                 .mode           = 0644,
1121                 .proc_handler   = lowmem_reserve_ratio_sysctl_handler,
1122         },
1123         {
1124                 .procname       = "drop_caches",
1125                 .data           = &sysctl_drop_caches,
1126                 .maxlen         = sizeof(int),
1127                 .mode           = 0644,
1128                 .proc_handler   = drop_caches_sysctl_handler,
1129         },
1130 #ifdef CONFIG_COMPACTION
1131         {
1132                 .procname       = "compact_memory",
1133                 .data           = &sysctl_compact_memory,
1134                 .maxlen         = sizeof(int),
1135                 .mode           = 0200,
1136                 .proc_handler   = sysctl_compaction_handler,
1137         },
1138         {
1139                 .procname       = "extfrag_threshold",
1140                 .data           = &sysctl_extfrag_threshold,
1141                 .maxlen         = sizeof(int),
1142                 .mode           = 0644,
1143                 .proc_handler   = sysctl_extfrag_handler,
1144                 .extra1         = &min_extfrag_threshold,
1145                 .extra2         = &max_extfrag_threshold,
1146         },
1147
1148 #endif /* CONFIG_COMPACTION */
1149         {
1150                 .procname       = "min_free_kbytes",
1151                 .data           = &min_free_kbytes,
1152                 .maxlen         = sizeof(min_free_kbytes),
1153                 .mode           = 0644,
1154                 .proc_handler   = min_free_kbytes_sysctl_handler,
1155                 .extra1         = &zero,
1156         },
1157         {
1158                 .procname       = "percpu_pagelist_fraction",
1159                 .data           = &percpu_pagelist_fraction,
1160                 .maxlen         = sizeof(percpu_pagelist_fraction),
1161                 .mode           = 0644,
1162                 .proc_handler   = percpu_pagelist_fraction_sysctl_handler,
1163                 .extra1         = &min_percpu_pagelist_fract,
1164         },
1165 #ifdef CONFIG_MMU
1166         {
1167                 .procname       = "max_map_count",
1168                 .data           = &sysctl_max_map_count,
1169                 .maxlen         = sizeof(sysctl_max_map_count),
1170                 .mode           = 0644,
1171                 .proc_handler   = proc_dointvec_minmax,
1172                 .extra1         = &zero,
1173         },
1174 #else
1175         {
1176                 .procname       = "nr_trim_pages",
1177                 .data           = &sysctl_nr_trim_pages,
1178                 .maxlen         = sizeof(sysctl_nr_trim_pages),
1179                 .mode           = 0644,
1180                 .proc_handler   = proc_dointvec_minmax,
1181                 .extra1         = &zero,
1182         },
1183 #endif
1184         {
1185                 .procname       = "laptop_mode",
1186                 .data           = &laptop_mode,
1187                 .maxlen         = sizeof(laptop_mode),
1188                 .mode           = 0644,
1189                 .proc_handler   = proc_dointvec_jiffies,
1190         },
1191         {
1192                 .procname       = "block_dump",
1193                 .data           = &block_dump,
1194                 .maxlen         = sizeof(block_dump),
1195                 .mode           = 0644,
1196                 .proc_handler   = proc_dointvec,
1197                 .extra1         = &zero,
1198         },
1199         {
1200                 .procname       = "vfs_cache_pressure",
1201                 .data           = &sysctl_vfs_cache_pressure,
1202                 .maxlen         = sizeof(sysctl_vfs_cache_pressure),
1203                 .mode           = 0644,
1204                 .proc_handler   = proc_dointvec,
1205                 .extra1         = &zero,
1206         },
1207 #ifdef HAVE_ARCH_PICK_MMAP_LAYOUT
1208         {
1209                 .procname       = "legacy_va_layout",
1210                 .data           = &sysctl_legacy_va_layout,
1211                 .maxlen         = sizeof(sysctl_legacy_va_layout),
1212                 .mode           = 0644,
1213                 .proc_handler   = proc_dointvec,
1214                 .extra1         = &zero,
1215         },
1216 #endif
1217 #ifdef CONFIG_NUMA
1218         {
1219                 .procname       = "zone_reclaim_mode",
1220                 .data           = &zone_reclaim_mode,
1221                 .maxlen         = sizeof(zone_reclaim_mode),
1222                 .mode           = 0644,
1223                 .proc_handler   = proc_dointvec,
1224                 .extra1         = &zero,
1225         },
1226         {
1227                 .procname       = "min_unmapped_ratio",
1228                 .data           = &sysctl_min_unmapped_ratio,
1229                 .maxlen         = sizeof(sysctl_min_unmapped_ratio),
1230                 .mode           = 0644,
1231                 .proc_handler   = sysctl_min_unmapped_ratio_sysctl_handler,
1232                 .extra1         = &zero,
1233                 .extra2         = &one_hundred,
1234         },
1235         {
1236                 .procname       = "min_slab_ratio",
1237                 .data           = &sysctl_min_slab_ratio,
1238                 .maxlen         = sizeof(sysctl_min_slab_ratio),
1239                 .mode           = 0644,
1240                 .proc_handler   = sysctl_min_slab_ratio_sysctl_handler,
1241                 .extra1         = &zero,
1242                 .extra2         = &one_hundred,
1243         },
1244 #endif
1245 #ifdef CONFIG_SMP
1246         {
1247                 .procname       = "stat_interval",
1248                 .data           = &sysctl_stat_interval,
1249                 .maxlen         = sizeof(sysctl_stat_interval),
1250                 .mode           = 0644,
1251                 .proc_handler   = proc_dointvec_jiffies,
1252         },
1253 #endif
1254 #ifdef CONFIG_MMU
1255         {
1256                 .procname       = "mmap_min_addr",
1257                 .data           = &dac_mmap_min_addr,
1258                 .maxlen         = sizeof(unsigned long),
1259                 .mode           = 0644,
1260                 .proc_handler   = mmap_min_addr_handler,
1261         },
1262 #endif
1263 #ifdef CONFIG_NUMA
1264         {
1265                 .procname       = "numa_zonelist_order",
1266                 .data           = &numa_zonelist_order,
1267                 .maxlen         = NUMA_ZONELIST_ORDER_LEN,
1268                 .mode           = 0644,
1269                 .proc_handler   = numa_zonelist_order_handler,
1270         },
1271 #endif
1272 #if (defined(CONFIG_X86_32) && !defined(CONFIG_UML))|| \
1273    (defined(CONFIG_SUPERH) && defined(CONFIG_VSYSCALL))
1274         {
1275                 .procname       = "vdso_enabled",
1276                 .data           = &vdso_enabled,
1277                 .maxlen         = sizeof(vdso_enabled),
1278                 .mode           = 0644,
1279                 .proc_handler   = proc_dointvec,
1280                 .extra1         = &zero,
1281         },
1282 #endif
1283 #ifdef CONFIG_HIGHMEM
1284         {
1285                 .procname       = "highmem_is_dirtyable",
1286                 .data           = &vm_highmem_is_dirtyable,
1287                 .maxlen         = sizeof(vm_highmem_is_dirtyable),
1288                 .mode           = 0644,
1289                 .proc_handler   = proc_dointvec_minmax,
1290                 .extra1         = &zero,
1291                 .extra2         = &one,
1292         },
1293 #endif
1294         {
1295                 .procname       = "scan_unevictable_pages",
1296                 .data           = &scan_unevictable_pages,
1297                 .maxlen         = sizeof(scan_unevictable_pages),
1298                 .mode           = 0644,
1299                 .proc_handler   = scan_unevictable_handler,
1300         },
1301 #ifdef CONFIG_MEMORY_FAILURE
1302         {
1303                 .procname       = "memory_failure_early_kill",
1304                 .data           = &sysctl_memory_failure_early_kill,
1305                 .maxlen         = sizeof(sysctl_memory_failure_early_kill),
1306                 .mode           = 0644,
1307                 .proc_handler   = proc_dointvec_minmax,
1308                 .extra1         = &zero,
1309                 .extra2         = &one,
1310         },
1311         {
1312                 .procname       = "memory_failure_recovery",
1313                 .data           = &sysctl_memory_failure_recovery,
1314                 .maxlen         = sizeof(sysctl_memory_failure_recovery),
1315                 .mode           = 0644,
1316                 .proc_handler   = proc_dointvec_minmax,
1317                 .extra1         = &zero,
1318                 .extra2         = &one,
1319         },
1320 #endif
1321
1322 /*
1323  * NOTE: do not add new entries to this table unless you have read
1324  * Documentation/sysctl/ctl_unnumbered.txt
1325  */
1326         { }
1327 };
1328
1329 #if defined(CONFIG_BINFMT_MISC) || defined(CONFIG_BINFMT_MISC_MODULE)
1330 static struct ctl_table binfmt_misc_table[] = {
1331         { }
1332 };
1333 #endif
1334
1335 static struct ctl_table fs_table[] = {
1336         {
1337                 .procname       = "inode-nr",
1338                 .data           = &inodes_stat,
1339                 .maxlen         = 2*sizeof(int),
1340                 .mode           = 0444,
1341                 .proc_handler   = proc_dointvec,
1342         },
1343         {
1344                 .procname       = "inode-state",
1345                 .data           = &inodes_stat,
1346                 .maxlen         = 7*sizeof(int),
1347                 .mode           = 0444,
1348                 .proc_handler   = proc_dointvec,
1349         },
1350         {
1351                 .procname       = "file-nr",
1352                 .data           = &files_stat,
1353                 .maxlen         = sizeof(files_stat),
1354                 .mode           = 0444,
1355                 .proc_handler   = proc_nr_files,
1356         },
1357         {
1358                 .procname       = "file-max",
1359                 .data           = &files_stat.max_files,
1360                 .maxlen         = sizeof(files_stat.max_files),
1361                 .mode           = 0644,
1362                 .proc_handler   = proc_doulongvec_minmax,
1363         },
1364         {
1365                 .procname       = "nr_open",
1366                 .data           = &sysctl_nr_open,
1367                 .maxlen         = sizeof(int),
1368                 .mode           = 0644,
1369                 .proc_handler   = proc_dointvec_minmax,
1370                 .extra1         = &sysctl_nr_open_min,
1371                 .extra2         = &sysctl_nr_open_max,
1372         },
1373         {
1374                 .procname       = "dentry-state",
1375                 .data           = &dentry_stat,
1376                 .maxlen         = 6*sizeof(int),
1377                 .mode           = 0444,
1378                 .proc_handler   = proc_dointvec,
1379         },
1380         {
1381                 .procname       = "overflowuid",
1382                 .data           = &fs_overflowuid,
1383                 .maxlen         = sizeof(int),
1384                 .mode           = 0644,
1385                 .proc_handler   = proc_dointvec_minmax,
1386                 .extra1         = &minolduid,
1387                 .extra2         = &maxolduid,
1388         },
1389         {
1390                 .procname       = "overflowgid",
1391                 .data           = &fs_overflowgid,
1392                 .maxlen         = sizeof(int),
1393                 .mode           = 0644,
1394                 .proc_handler   = proc_dointvec_minmax,
1395                 .extra1         = &minolduid,
1396                 .extra2         = &maxolduid,
1397         },
1398 #ifdef CONFIG_FILE_LOCKING
1399         {
1400                 .procname       = "leases-enable",
1401                 .data           = &leases_enable,
1402                 .maxlen         = sizeof(int),
1403                 .mode           = 0644,
1404                 .proc_handler   = proc_dointvec,
1405         },
1406 #endif
1407 #ifdef CONFIG_DNOTIFY
1408         {
1409                 .procname       = "dir-notify-enable",
1410                 .data           = &dir_notify_enable,
1411                 .maxlen         = sizeof(int),
1412                 .mode           = 0644,
1413                 .proc_handler   = proc_dointvec,
1414         },
1415 #endif
1416 #ifdef CONFIG_MMU
1417 #ifdef CONFIG_FILE_LOCKING
1418         {
1419                 .procname       = "lease-break-time",
1420                 .data           = &lease_break_time,
1421                 .maxlen         = sizeof(int),
1422                 .mode           = 0644,
1423                 .proc_handler   = proc_dointvec,
1424         },
1425 #endif
1426 #ifdef CONFIG_AIO
1427         {
1428                 .procname       = "aio-nr",
1429                 .data           = &aio_nr,
1430                 .maxlen         = sizeof(aio_nr),
1431                 .mode           = 0444,
1432                 .proc_handler   = proc_doulongvec_minmax,
1433         },
1434         {
1435                 .procname       = "aio-max-nr",
1436                 .data           = &aio_max_nr,
1437                 .maxlen         = sizeof(aio_max_nr),
1438                 .mode           = 0644,
1439                 .proc_handler   = proc_doulongvec_minmax,
1440         },
1441 #endif /* CONFIG_AIO */
1442 #ifdef CONFIG_INOTIFY_USER
1443         {
1444                 .procname       = "inotify",
1445                 .mode           = 0555,
1446                 .child          = inotify_table,
1447         },
1448 #endif  
1449 #ifdef CONFIG_EPOLL
1450         {
1451                 .procname       = "epoll",
1452                 .mode           = 0555,
1453                 .child          = epoll_table,
1454         },
1455 #endif
1456 #endif
1457         {
1458                 .procname       = "suid_dumpable",
1459                 .data           = &suid_dumpable,
1460                 .maxlen         = sizeof(int),
1461                 .mode           = 0644,
1462                 .proc_handler   = proc_dointvec_minmax,
1463                 .extra1         = &zero,
1464                 .extra2         = &two,
1465         },
1466 #if defined(CONFIG_BINFMT_MISC) || defined(CONFIG_BINFMT_MISC_MODULE)
1467         {
1468                 .procname       = "binfmt_misc",
1469                 .mode           = 0555,
1470                 .child          = binfmt_misc_table,
1471         },
1472 #endif
1473         {
1474                 .procname       = "pipe-max-size",
1475                 .data           = &pipe_max_size,
1476                 .maxlen         = sizeof(int),
1477                 .mode           = 0644,
1478                 .proc_handler   = &pipe_proc_fn,
1479                 .extra1         = &pipe_min_size,
1480         },
1481 /*
1482  * NOTE: do not add new entries to this table unless you have read
1483  * Documentation/sysctl/ctl_unnumbered.txt
1484  */
1485         { }
1486 };
1487
1488 static struct ctl_table debug_table[] = {
1489 #if defined(CONFIG_X86) || defined(CONFIG_PPC) || defined(CONFIG_SPARC) || \
1490     defined(CONFIG_S390)
1491         {
1492                 .procname       = "exception-trace",
1493                 .data           = &show_unhandled_signals,
1494                 .maxlen         = sizeof(int),
1495                 .mode           = 0644,
1496                 .proc_handler   = proc_dointvec
1497         },
1498 #endif
1499 #if defined(CONFIG_OPTPROBES)
1500         {
1501                 .procname       = "kprobes-optimization",
1502                 .data           = &sysctl_kprobes_optimization,
1503                 .maxlen         = sizeof(int),
1504                 .mode           = 0644,
1505                 .proc_handler   = proc_kprobes_optimization_handler,
1506                 .extra1         = &zero,
1507                 .extra2         = &one,
1508         },
1509 #endif
1510         { }
1511 };
1512
1513 static struct ctl_table dev_table[] = {
1514         { }
1515 };
1516
1517 static DEFINE_SPINLOCK(sysctl_lock);
1518
1519 /* called under sysctl_lock */
1520 static int use_table(struct ctl_table_header *p)
1521 {
1522         if (unlikely(p->unregistering))
1523                 return 0;
1524         p->used++;
1525         return 1;
1526 }
1527
1528 /* called under sysctl_lock */
1529 static void unuse_table(struct ctl_table_header *p)
1530 {
1531         if (!--p->used)
1532                 if (unlikely(p->unregistering))
1533                         complete(p->unregistering);
1534 }
1535
1536 /* called under sysctl_lock, will reacquire if has to wait */
1537 static void start_unregistering(struct ctl_table_header *p)
1538 {
1539         /*
1540          * if p->used is 0, nobody will ever touch that entry again;
1541          * we'll eliminate all paths to it before dropping sysctl_lock
1542          */
1543         if (unlikely(p->used)) {
1544                 struct completion wait;
1545                 init_completion(&wait);
1546                 p->unregistering = &wait;
1547                 spin_unlock(&sysctl_lock);
1548                 wait_for_completion(&wait);
1549                 spin_lock(&sysctl_lock);
1550         } else {
1551                 /* anything non-NULL; we'll never dereference it */
1552                 p->unregistering = ERR_PTR(-EINVAL);
1553         }
1554         /*
1555          * do not remove from the list until nobody holds it; walking the
1556          * list in do_sysctl() relies on that.
1557          */
1558         list_del_init(&p->ctl_entry);
1559 }
1560
1561 void sysctl_head_get(struct ctl_table_header *head)
1562 {
1563         spin_lock(&sysctl_lock);
1564         head->count++;
1565         spin_unlock(&sysctl_lock);
1566 }
1567
1568 void sysctl_head_put(struct ctl_table_header *head)
1569 {
1570         spin_lock(&sysctl_lock);
1571         if (!--head->count)
1572                 kfree(head);
1573         spin_unlock(&sysctl_lock);
1574 }
1575
1576 struct ctl_table_header *sysctl_head_grab(struct ctl_table_header *head)
1577 {
1578         if (!head)
1579                 BUG();
1580         spin_lock(&sysctl_lock);
1581         if (!use_table(head))
1582                 head = ERR_PTR(-ENOENT);
1583         spin_unlock(&sysctl_lock);
1584         return head;
1585 }
1586
1587 void sysctl_head_finish(struct ctl_table_header *head)
1588 {
1589         if (!head)
1590                 return;
1591         spin_lock(&sysctl_lock);
1592         unuse_table(head);
1593         spin_unlock(&sysctl_lock);
1594 }
1595
1596 static struct ctl_table_set *
1597 lookup_header_set(struct ctl_table_root *root, struct nsproxy *namespaces)
1598 {
1599         struct ctl_table_set *set = &root->default_set;
1600         if (root->lookup)
1601                 set = root->lookup(root, namespaces);
1602         return set;
1603 }
1604
1605 static struct list_head *
1606 lookup_header_list(struct ctl_table_root *root, struct nsproxy *namespaces)
1607 {
1608         struct ctl_table_set *set = lookup_header_set(root, namespaces);
1609         return &set->list;
1610 }
1611
1612 struct ctl_table_header *__sysctl_head_next(struct nsproxy *namespaces,
1613                                             struct ctl_table_header *prev)
1614 {
1615         struct ctl_table_root *root;
1616         struct list_head *header_list;
1617         struct ctl_table_header *head;
1618         struct list_head *tmp;
1619
1620         spin_lock(&sysctl_lock);
1621         if (prev) {
1622                 head = prev;
1623                 tmp = &prev->ctl_entry;
1624                 unuse_table(prev);
1625                 goto next;
1626         }
1627         tmp = &root_table_header.ctl_entry;
1628         for (;;) {
1629                 head = list_entry(tmp, struct ctl_table_header, ctl_entry);
1630
1631                 if (!use_table(head))
1632                         goto next;
1633                 spin_unlock(&sysctl_lock);
1634                 return head;
1635         next:
1636                 root = head->root;
1637                 tmp = tmp->next;
1638                 header_list = lookup_header_list(root, namespaces);
1639                 if (tmp != header_list)
1640                         continue;
1641
1642                 do {
1643                         root = list_entry(root->root_list.next,
1644                                         struct ctl_table_root, root_list);
1645                         if (root == &sysctl_table_root)
1646                                 goto out;
1647                         header_list = lookup_header_list(root, namespaces);
1648                 } while (list_empty(header_list));
1649                 tmp = header_list->next;
1650         }
1651 out:
1652         spin_unlock(&sysctl_lock);
1653         return NULL;
1654 }
1655
1656 struct ctl_table_header *sysctl_head_next(struct ctl_table_header *prev)
1657 {
1658         return __sysctl_head_next(current->nsproxy, prev);
1659 }
1660
1661 void register_sysctl_root(struct ctl_table_root *root)
1662 {
1663         spin_lock(&sysctl_lock);
1664         list_add_tail(&root->root_list, &sysctl_table_root.root_list);
1665         spin_unlock(&sysctl_lock);
1666 }
1667
1668 /*
1669  * sysctl_perm does NOT grant the superuser all rights automatically, because
1670  * some sysctl variables are readonly even to root.
1671  */
1672
1673 static int test_perm(int mode, int op)
1674 {
1675         if (!current_euid())
1676                 mode >>= 6;
1677         else if (in_egroup_p(0))
1678                 mode >>= 3;
1679         if ((op & ~mode & (MAY_READ|MAY_WRITE|MAY_EXEC)) == 0)
1680                 return 0;
1681         return -EACCES;
1682 }
1683
1684 int sysctl_perm(struct ctl_table_root *root, struct ctl_table *table, int op)
1685 {
1686         int error;
1687         int mode;
1688
1689         error = security_sysctl(table, op & (MAY_READ | MAY_WRITE | MAY_EXEC));
1690         if (error)
1691                 return error;
1692
1693         if (root->permissions)
1694                 mode = root->permissions(root, current->nsproxy, table);
1695         else
1696                 mode = table->mode;
1697
1698         return test_perm(mode, op);
1699 }
1700
1701 static void sysctl_set_parent(struct ctl_table *parent, struct ctl_table *table)
1702 {
1703         for (; table->procname; table++) {
1704                 table->parent = parent;
1705                 if (table->child)
1706                         sysctl_set_parent(table, table->child);
1707         }
1708 }
1709
1710 static __init int sysctl_init(void)
1711 {
1712         sysctl_set_parent(NULL, root_table);
1713 #ifdef CONFIG_SYSCTL_SYSCALL_CHECK
1714         sysctl_check_table(current->nsproxy, root_table);
1715 #endif
1716         return 0;
1717 }
1718
1719 core_initcall(sysctl_init);
1720
1721 static struct ctl_table *is_branch_in(struct ctl_table *branch,
1722                                       struct ctl_table *table)
1723 {
1724         struct ctl_table *p;
1725         const char *s = branch->procname;
1726
1727         /* branch should have named subdirectory as its first element */
1728         if (!s || !branch->child)
1729                 return NULL;
1730
1731         /* ... and nothing else */
1732         if (branch[1].procname)
1733                 return NULL;
1734
1735         /* table should contain subdirectory with the same name */
1736         for (p = table; p->procname; p++) {
1737                 if (!p->child)
1738                         continue;
1739                 if (p->procname && strcmp(p->procname, s) == 0)
1740                         return p;
1741         }
1742         return NULL;
1743 }
1744
1745 /* see if attaching q to p would be an improvement */
1746 static void try_attach(struct ctl_table_header *p, struct ctl_table_header *q)
1747 {
1748         struct ctl_table *to = p->ctl_table, *by = q->ctl_table;
1749         struct ctl_table *next;
1750         int is_better = 0;
1751         int not_in_parent = !p->attached_by;
1752
1753         while ((next = is_branch_in(by, to)) != NULL) {
1754                 if (by == q->attached_by)
1755                         is_better = 1;
1756                 if (to == p->attached_by)
1757                         not_in_parent = 1;
1758                 by = by->child;
1759                 to = next->child;
1760         }
1761
1762         if (is_better && not_in_parent) {
1763                 q->attached_by = by;
1764                 q->attached_to = to;
1765                 q->parent = p;
1766         }
1767 }
1768
1769 /**
1770  * __register_sysctl_paths - register a sysctl hierarchy
1771  * @root: List of sysctl headers to register on
1772  * @namespaces: Data to compute which lists of sysctl entries are visible
1773  * @path: The path to the directory the sysctl table is in.
1774  * @table: the top-level table structure
1775  *
1776  * Register a sysctl table hierarchy. @table should be a filled in ctl_table
1777  * array. A completely 0 filled entry terminates the table.
1778  *
1779  * The members of the &struct ctl_table structure are used as follows:
1780  *
1781  * procname - the name of the sysctl file under /proc/sys. Set to %NULL to not
1782  *            enter a sysctl file
1783  *
1784  * data - a pointer to data for use by proc_handler
1785  *
1786  * maxlen - the maximum size in bytes of the data
1787  *
1788  * mode - the file permissions for the /proc/sys file, and for sysctl(2)
1789  *
1790  * child - a pointer to the child sysctl table if this entry is a directory, or
1791  *         %NULL.
1792  *
1793  * proc_handler - the text handler routine (described below)
1794  *
1795  * de - for internal use by the sysctl routines
1796  *
1797  * extra1, extra2 - extra pointers usable by the proc handler routines
1798  *
1799  * Leaf nodes in the sysctl tree will be represented by a single file
1800  * under /proc; non-leaf nodes will be represented by directories.
1801  *
1802  * sysctl(2) can automatically manage read and write requests through
1803  * the sysctl table.  The data and maxlen fields of the ctl_table
1804  * struct enable minimal validation of the values being written to be
1805  * performed, and the mode field allows minimal authentication.
1806  *
1807  * There must be a proc_handler routine for any terminal nodes
1808  * mirrored under /proc/sys (non-terminals are handled by a built-in
1809  * directory handler).  Several default handlers are available to
1810  * cover common cases -
1811  *
1812  * proc_dostring(), proc_dointvec(), proc_dointvec_jiffies(),
1813  * proc_dointvec_userhz_jiffies(), proc_dointvec_minmax(), 
1814  * proc_doulongvec_ms_jiffies_minmax(), proc_doulongvec_minmax()
1815  *
1816  * It is the handler's job to read the input buffer from user memory
1817  * and process it. The handler should return 0 on success.
1818  *
1819  * This routine returns %NULL on a failure to register, and a pointer
1820  * to the table header on success.
1821  */
1822 struct ctl_table_header *__register_sysctl_paths(
1823         struct ctl_table_root *root,
1824         struct nsproxy *namespaces,
1825         const struct ctl_path *path, struct ctl_table *table)
1826 {
1827         struct ctl_table_header *header;
1828         struct ctl_table *new, **prevp;
1829         unsigned int n, npath;
1830         struct ctl_table_set *set;
1831
1832         /* Count the path components */
1833         for (npath = 0; path[npath].procname; ++npath)
1834                 ;
1835
1836         /*
1837          * For each path component, allocate a 2-element ctl_table array.
1838          * The first array element will be filled with the sysctl entry
1839          * for this, the second will be the sentinel (procname == 0).
1840          *
1841          * We allocate everything in one go so that we don't have to
1842          * worry about freeing additional memory in unregister_sysctl_table.
1843          */
1844         header = kzalloc(sizeof(struct ctl_table_header) +
1845                          (2 * npath * sizeof(struct ctl_table)), GFP_KERNEL);
1846         if (!header)
1847                 return NULL;
1848
1849         new = (struct ctl_table *) (header + 1);
1850
1851         /* Now connect the dots */
1852         prevp = &header->ctl_table;
1853         for (n = 0; n < npath; ++n, ++path) {
1854                 /* Copy the procname */
1855                 new->procname = path->procname;
1856                 new->mode     = 0555;
1857
1858                 *prevp = new;
1859                 prevp = &new->child;
1860
1861                 new += 2;
1862         }
1863         *prevp = table;
1864         header->ctl_table_arg = table;
1865
1866         INIT_LIST_HEAD(&header->ctl_entry);
1867         header->used = 0;
1868         header->unregistering = NULL;
1869         header->root = root;
1870         sysctl_set_parent(NULL, header->ctl_table);
1871         header->count = 1;
1872 #ifdef CONFIG_SYSCTL_SYSCALL_CHECK
1873         if (sysctl_check_table(namespaces, header->ctl_table)) {
1874                 kfree(header);
1875                 return NULL;
1876         }
1877 #endif
1878         spin_lock(&sysctl_lock);
1879         header->set = lookup_header_set(root, namespaces);
1880         header->attached_by = header->ctl_table;
1881         header->attached_to = root_table;
1882         header->parent = &root_table_header;
1883         for (set = header->set; set; set = set->parent) {
1884                 struct ctl_table_header *p;
1885                 list_for_each_entry(p, &set->list, ctl_entry) {
1886                         if (p->unregistering)
1887                                 continue;
1888                         try_attach(p, header);
1889                 }
1890         }
1891         header->parent->count++;
1892         list_add_tail(&header->ctl_entry, &header->set->list);
1893         spin_unlock(&sysctl_lock);
1894
1895         return header;
1896 }
1897
1898 /**
1899  * register_sysctl_table_path - register a sysctl table hierarchy
1900  * @path: The path to the directory the sysctl table is in.
1901  * @table: the top-level table structure
1902  *
1903  * Register a sysctl table hierarchy. @table should be a filled in ctl_table
1904  * array. A completely 0 filled entry terminates the table.
1905  *
1906  * See __register_sysctl_paths for more details.
1907  */
1908 struct ctl_table_header *register_sysctl_paths(const struct ctl_path *path,
1909                                                 struct ctl_table *table)
1910 {
1911         return __register_sysctl_paths(&sysctl_table_root, current->nsproxy,
1912                                         path, table);
1913 }
1914
1915 /**
1916  * register_sysctl_table - register a sysctl table hierarchy
1917  * @table: the top-level table structure
1918  *
1919  * Register a sysctl table hierarchy. @table should be a filled in ctl_table
1920  * array. A completely 0 filled entry terminates the table.
1921  *
1922  * See register_sysctl_paths for more details.
1923  */
1924 struct ctl_table_header *register_sysctl_table(struct ctl_table *table)
1925 {
1926         static const struct ctl_path null_path[] = { {} };
1927
1928         return register_sysctl_paths(null_path, table);
1929 }
1930
1931 /**
1932  * unregister_sysctl_table - unregister a sysctl table hierarchy
1933  * @header: the header returned from register_sysctl_table
1934  *
1935  * Unregisters the sysctl table and all children. proc entries may not
1936  * actually be removed until they are no longer used by anyone.
1937  */
1938 void unregister_sysctl_table(struct ctl_table_header * header)
1939 {
1940         might_sleep();
1941
1942         if (header == NULL)
1943                 return;
1944
1945         spin_lock(&sysctl_lock);
1946         start_unregistering(header);
1947         if (!--header->parent->count) {
1948                 WARN_ON(1);
1949                 kfree(header->parent);
1950         }
1951         if (!--header->count)
1952                 kfree(header);
1953         spin_unlock(&sysctl_lock);
1954 }
1955
1956 int sysctl_is_seen(struct ctl_table_header *p)
1957 {
1958         struct ctl_table_set *set = p->set;
1959         int res;
1960         spin_lock(&sysctl_lock);
1961         if (p->unregistering)
1962                 res = 0;
1963         else if (!set->is_seen)
1964                 res = 1;
1965         else
1966                 res = set->is_seen(set);
1967         spin_unlock(&sysctl_lock);
1968         return res;
1969 }
1970
1971 void setup_sysctl_set(struct ctl_table_set *p,
1972         struct ctl_table_set *parent,
1973         int (*is_seen)(struct ctl_table_set *))
1974 {
1975         INIT_LIST_HEAD(&p->list);
1976         p->parent = parent ? parent : &sysctl_table_root.default_set;
1977         p->is_seen = is_seen;
1978 }
1979
1980 #else /* !CONFIG_SYSCTL */
1981 struct ctl_table_header *register_sysctl_table(struct ctl_table * table)
1982 {
1983         return NULL;
1984 }
1985
1986 struct ctl_table_header *register_sysctl_paths(const struct ctl_path *path,
1987                                                     struct ctl_table *table)
1988 {
1989         return NULL;
1990 }
1991
1992 void unregister_sysctl_table(struct ctl_table_header * table)
1993 {
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 }
2001
2002 void sysctl_head_put(struct ctl_table_header *head)
2003 {
2004 }
2005
2006 #endif /* CONFIG_SYSCTL */
2007
2008 /*
2009  * /proc/sys support
2010  */
2011
2012 #ifdef CONFIG_PROC_SYSCTL
2013
2014 static int _proc_do_string(void* data, int maxlen, int write,
2015                            void __user *buffer,
2016                            size_t *lenp, loff_t *ppos)
2017 {
2018         size_t len;
2019         char __user *p;
2020         char c;
2021
2022         if (!data || !maxlen || !*lenp) {
2023                 *lenp = 0;
2024                 return 0;
2025         }
2026
2027         if (write) {
2028                 len = 0;
2029                 p = buffer;
2030                 while (len < *lenp) {
2031                         if (get_user(c, p++))
2032                                 return -EFAULT;
2033                         if (c == 0 || c == '\n')
2034                                 break;
2035                         len++;
2036                 }
2037                 if (len >= maxlen)
2038                         len = maxlen-1;
2039                 if(copy_from_user(data, buffer, len))
2040                         return -EFAULT;
2041                 ((char *) data)[len] = 0;
2042                 *ppos += *lenp;
2043         } else {
2044                 len = strlen(data);
2045                 if (len > maxlen)
2046                         len = maxlen;
2047
2048                 if (*ppos > len) {
2049                         *lenp = 0;
2050                         return 0;
2051                 }
2052
2053                 data += *ppos;
2054                 len  -= *ppos;
2055
2056                 if (len > *lenp)
2057                         len = *lenp;
2058                 if (len)
2059                         if(copy_to_user(buffer, data, len))
2060                                 return -EFAULT;
2061                 if (len < *lenp) {
2062                         if(put_user('\n', ((char __user *) buffer) + len))
2063                                 return -EFAULT;
2064                         len++;
2065                 }
2066                 *lenp = len;
2067                 *ppos += len;
2068         }
2069         return 0;
2070 }
2071
2072 /**
2073  * proc_dostring - read a string sysctl
2074  * @table: the sysctl table
2075  * @write: %TRUE if this is a write to the sysctl file
2076  * @buffer: the user buffer
2077  * @lenp: the size of the user buffer
2078  * @ppos: file position
2079  *
2080  * Reads/writes a string from/to the user buffer. If the kernel
2081  * buffer provided is not large enough to hold the string, the
2082  * string is truncated. The copied string is %NULL-terminated.
2083  * If the string is being read by the user process, it is copied
2084  * and a newline '\n' is added. It is truncated if the buffer is
2085  * not large enough.
2086  *
2087  * Returns 0 on success.
2088  */
2089 int proc_dostring(struct ctl_table *table, int write,
2090                   void __user *buffer, size_t *lenp, loff_t *ppos)
2091 {
2092         return _proc_do_string(table->data, table->maxlen, write,
2093                                buffer, lenp, ppos);
2094 }
2095
2096 static size_t proc_skip_spaces(char **buf)
2097 {
2098         size_t ret;
2099         char *tmp = skip_spaces(*buf);
2100         ret = tmp - *buf;
2101         *buf = tmp;
2102         return ret;
2103 }
2104
2105 static void proc_skip_char(char **buf, size_t *size, const char v)
2106 {
2107         while (*size) {
2108                 if (**buf != v)
2109                         break;
2110                 (*size)--;
2111                 (*buf)++;
2112         }
2113 }
2114
2115 #define TMPBUFLEN 22
2116 /**
2117  * proc_get_long - reads an ASCII formatted integer from a user buffer
2118  *
2119  * @buf: a kernel buffer
2120  * @size: size of the kernel buffer
2121  * @val: this is where the number will be stored
2122  * @neg: set to %TRUE if number is negative
2123  * @perm_tr: a vector which contains the allowed trailers
2124  * @perm_tr_len: size of the perm_tr vector
2125  * @tr: pointer to store the trailer character
2126  *
2127  * In case of success %0 is returned and @buf and @size are updated with
2128  * the amount of bytes read. If @tr is non-NULL and a trailing
2129  * character exists (size is non-zero after returning from this
2130  * function), @tr is updated with the trailing character.
2131  */
2132 static int proc_get_long(char **buf, size_t *size,
2133                           unsigned long *val, bool *neg,
2134                           const char *perm_tr, unsigned perm_tr_len, char *tr)
2135 {
2136         int len;
2137         char *p, tmp[TMPBUFLEN];
2138
2139         if (!*size)
2140                 return -EINVAL;
2141
2142         len = *size;
2143         if (len > TMPBUFLEN - 1)
2144                 len = TMPBUFLEN - 1;
2145
2146         memcpy(tmp, *buf, len);
2147
2148         tmp[len] = 0;
2149         p = tmp;
2150         if (*p == '-' && *size > 1) {
2151                 *neg = true;
2152                 p++;
2153         } else
2154                 *neg = false;
2155         if (!isdigit(*p))
2156                 return -EINVAL;
2157
2158         *val = simple_strtoul(p, &p, 0);
2159
2160         len = p - tmp;
2161
2162         /* We don't know if the next char is whitespace thus we may accept
2163          * invalid integers (e.g. 1234...a) or two integers instead of one
2164          * (e.g. 123...1). So lets not allow such large numbers. */
2165         if (len == TMPBUFLEN - 1)
2166                 return -EINVAL;
2167
2168         if (len < *size && perm_tr_len && !memchr(perm_tr, *p, perm_tr_len))
2169                 return -EINVAL;
2170
2171         if (tr && (len < *size))
2172                 *tr = *p;
2173
2174         *buf += len;
2175         *size -= len;
2176
2177         return 0;
2178 }
2179
2180 /**
2181  * proc_put_long - converts an integer to a decimal ASCII formatted string
2182  *
2183  * @buf: the user buffer
2184  * @size: the size of the user buffer
2185  * @val: the integer to be converted
2186  * @neg: sign of the number, %TRUE for negative
2187  *
2188  * In case of success %0 is returned and @buf and @size are updated with
2189  * the amount of bytes written.
2190  */
2191 static int proc_put_long(void __user **buf, size_t *size, unsigned long val,
2192                           bool neg)
2193 {
2194         int len;
2195         char tmp[TMPBUFLEN], *p = tmp;
2196
2197         sprintf(p, "%s%lu", neg ? "-" : "", val);
2198         len = strlen(tmp);
2199         if (len > *size)
2200                 len = *size;
2201         if (copy_to_user(*buf, tmp, len))
2202                 return -EFAULT;
2203         *size -= len;
2204         *buf += len;
2205         return 0;
2206 }
2207 #undef TMPBUFLEN
2208
2209 static int proc_put_char(void __user **buf, size_t *size, char c)
2210 {
2211         if (*size) {
2212                 char __user **buffer = (char __user **)buf;
2213                 if (put_user(c, *buffer))
2214                         return -EFAULT;
2215                 (*size)--, (*buffer)++;
2216                 *buf = *buffer;
2217         }
2218         return 0;
2219 }
2220
2221 static int do_proc_dointvec_conv(bool *negp, unsigned long *lvalp,
2222                                  int *valp,
2223                                  int write, void *data)
2224 {
2225         if (write) {
2226                 *valp = *negp ? -*lvalp : *lvalp;
2227         } else {
2228                 int val = *valp;
2229                 if (val < 0) {
2230                         *negp = true;
2231                         *lvalp = (unsigned long)-val;
2232                 } else {
2233                         *negp = false;
2234                         *lvalp = (unsigned long)val;
2235                 }
2236         }
2237         return 0;
2238 }
2239
2240 static const char proc_wspace_sep[] = { ' ', '\t', '\n' };
2241
2242 static int __do_proc_dointvec(void *tbl_data, struct ctl_table *table,
2243                   int write, void __user *buffer,
2244                   size_t *lenp, loff_t *ppos,
2245                   int (*conv)(bool *negp, unsigned long *lvalp, int *valp,
2246                               int write, void *data),
2247                   void *data)
2248 {
2249         int *i, vleft, first = 1, err = 0;
2250         unsigned long page = 0;
2251         size_t left;
2252         char *kbuf;
2253         
2254         if (!tbl_data || !table->maxlen || !*lenp || (*ppos && !write)) {
2255                 *lenp = 0;
2256                 return 0;
2257         }
2258         
2259         i = (int *) tbl_data;
2260         vleft = table->maxlen / sizeof(*i);
2261         left = *lenp;
2262
2263         if (!conv)
2264                 conv = do_proc_dointvec_conv;
2265
2266         if (write) {
2267                 if (left > PAGE_SIZE - 1)
2268                         left = PAGE_SIZE - 1;
2269                 page = __get_free_page(GFP_TEMPORARY);
2270                 kbuf = (char *) page;
2271                 if (!kbuf)
2272                         return -ENOMEM;
2273                 if (copy_from_user(kbuf, buffer, left)) {
2274                         err = -EFAULT;
2275                         goto free;
2276                 }
2277                 kbuf[left] = 0;
2278         }
2279
2280         for (; left && vleft--; i++, first=0) {
2281                 unsigned long lval;
2282                 bool neg;
2283
2284                 if (write) {
2285                         left -= proc_skip_spaces(&kbuf);
2286
2287                         if (!left)
2288                                 break;
2289                         err = proc_get_long(&kbuf, &left, &lval, &neg,
2290                                              proc_wspace_sep,
2291                                              sizeof(proc_wspace_sep), NULL);
2292                         if (err)
2293                                 break;
2294                         if (conv(&neg, &lval, i, 1, data)) {
2295                                 err = -EINVAL;
2296                                 break;
2297                         }
2298                 } else {
2299                         if (conv(&neg, &lval, i, 0, data)) {
2300                                 err = -EINVAL;
2301                                 break;
2302                         }
2303                         if (!first)
2304                                 err = proc_put_char(&buffer, &left, '\t');
2305                         if (err)
2306                                 break;
2307                         err = proc_put_long(&buffer, &left, lval, neg);
2308                         if (err)
2309                                 break;
2310                 }
2311         }
2312
2313         if (!write && !first && left && !err)
2314                 err = proc_put_char(&buffer, &left, '\n');
2315         if (write && !err && left)
2316                 left -= proc_skip_spaces(&kbuf);
2317 free:
2318         if (write) {
2319                 free_page(page);
2320                 if (first)
2321                         return err ? : -EINVAL;
2322         }
2323         *lenp -= left;
2324         *ppos += *lenp;
2325         return err;
2326 }
2327
2328 static int do_proc_dointvec(struct ctl_table *table, int write,
2329                   void __user *buffer, size_t *lenp, loff_t *ppos,
2330                   int (*conv)(bool *negp, unsigned long *lvalp, int *valp,
2331                               int write, void *data),
2332                   void *data)
2333 {
2334         return __do_proc_dointvec(table->data, table, write,
2335                         buffer, lenp, ppos, conv, data);
2336 }
2337
2338 /**
2339  * proc_dointvec - read a vector of integers
2340  * @table: the sysctl table
2341  * @write: %TRUE if this is a write to the sysctl file
2342  * @buffer: the user buffer
2343  * @lenp: the size of the user buffer
2344  * @ppos: file position
2345  *
2346  * Reads/writes up to table->maxlen/sizeof(unsigned int) integer
2347  * values from/to the user buffer, treated as an ASCII string. 
2348  *
2349  * Returns 0 on success.
2350  */
2351 int proc_dointvec(struct ctl_table *table, int write,
2352                      void __user *buffer, size_t *lenp, loff_t *ppos)
2353 {
2354     return do_proc_dointvec(table,write,buffer,lenp,ppos,
2355                             NULL,NULL);
2356 }
2357
2358 /*
2359  * Taint values can only be increased
2360  * This means we can safely use a temporary.
2361  */
2362 static int proc_taint(struct ctl_table *table, int write,
2363                                void __user *buffer, size_t *lenp, loff_t *ppos)
2364 {
2365         struct ctl_table t;
2366         unsigned long tmptaint = get_taint();
2367         int err;
2368
2369         if (write && !capable(CAP_SYS_ADMIN))
2370                 return -EPERM;
2371
2372         t = *table;
2373         t.data = &tmptaint;
2374         err = proc_doulongvec_minmax(&t, write, buffer, lenp, ppos);
2375         if (err < 0)
2376                 return err;
2377
2378         if (write) {
2379                 /*
2380                  * Poor man's atomic or. Not worth adding a primitive
2381                  * to everyone's atomic.h for this
2382                  */
2383                 int i;
2384                 for (i = 0; i < BITS_PER_LONG && tmptaint >> i; i++) {
2385                         if ((tmptaint >> i) & 1)
2386                                 add_taint(i);
2387                 }
2388         }
2389
2390         return err;
2391 }
2392
2393 struct do_proc_dointvec_minmax_conv_param {
2394         int *min;
2395         int *max;
2396 };
2397
2398 static int do_proc_dointvec_minmax_conv(bool *negp, unsigned long *lvalp,
2399                                         int *valp,
2400                                         int write, void *data)
2401 {
2402         struct do_proc_dointvec_minmax_conv_param *param = data;
2403         if (write) {
2404                 int val = *negp ? -*lvalp : *lvalp;
2405                 if ((param->min && *param->min > val) ||
2406                     (param->max && *param->max < val))
2407                         return -EINVAL;
2408                 *valp = val;
2409         } else {
2410                 int val = *valp;
2411                 if (val < 0) {
2412                         *negp = true;
2413                         *lvalp = (unsigned long)-val;
2414                 } else {
2415                         *negp = false;
2416                         *lvalp = (unsigned long)val;
2417                 }
2418         }
2419         return 0;
2420 }
2421
2422 /**
2423  * proc_dointvec_minmax - read a vector of integers with min/max values
2424  * @table: the sysctl table
2425  * @write: %TRUE if this is a write to the sysctl file
2426  * @buffer: the user buffer
2427  * @lenp: the size of the user buffer
2428  * @ppos: file position
2429  *
2430  * Reads/writes up to table->maxlen/sizeof(unsigned int) integer
2431  * values from/to the user buffer, treated as an ASCII string.
2432  *
2433  * This routine will ensure the values are within the range specified by
2434  * table->extra1 (min) and table->extra2 (max).
2435  *
2436  * Returns 0 on success.
2437  */
2438 int proc_dointvec_minmax(struct ctl_table *table, int write,
2439                   void __user *buffer, size_t *lenp, loff_t *ppos)
2440 {
2441         struct do_proc_dointvec_minmax_conv_param param = {
2442                 .min = (int *) table->extra1,
2443                 .max = (int *) table->extra2,
2444         };
2445         return do_proc_dointvec(table, write, buffer, lenp, ppos,
2446                                 do_proc_dointvec_minmax_conv, &param);
2447 }
2448
2449 static int __do_proc_doulongvec_minmax(void *data, struct ctl_table *table, int write,
2450                                      void __user *buffer,
2451                                      size_t *lenp, loff_t *ppos,
2452                                      unsigned long convmul,
2453                                      unsigned long convdiv)
2454 {
2455         unsigned long *i, *min, *max;
2456         int vleft, first = 1, err = 0;
2457         unsigned long page = 0;
2458         size_t left;
2459         char *kbuf;
2460
2461         if (!data || !table->maxlen || !*lenp || (*ppos && !write)) {
2462                 *lenp = 0;
2463                 return 0;
2464         }
2465
2466         i = (unsigned long *) data;
2467         min = (unsigned long *) table->extra1;
2468         max = (unsigned long *) table->extra2;
2469         vleft = table->maxlen / sizeof(unsigned long);
2470         left = *lenp;
2471
2472         if (write) {
2473                 if (left > PAGE_SIZE - 1)
2474                         left = PAGE_SIZE - 1;
2475                 page = __get_free_page(GFP_TEMPORARY);
2476                 kbuf = (char *) page;
2477                 if (!kbuf)
2478                         return -ENOMEM;
2479                 if (copy_from_user(kbuf, buffer, left)) {
2480                         err = -EFAULT;
2481                         goto free;
2482                 }
2483                 kbuf[left] = 0;
2484         }
2485
2486         for (; left && vleft--; i++, first = 0) {
2487                 unsigned long val;
2488
2489                 if (write) {
2490                         bool neg;
2491
2492                         left -= proc_skip_spaces(&kbuf);
2493
2494                         err = proc_get_long(&kbuf, &left, &val, &neg,
2495                                              proc_wspace_sep,
2496                                              sizeof(proc_wspace_sep), NULL);
2497                         if (err)
2498                                 break;
2499                         if (neg)
2500                                 continue;
2501                         if ((min && val < *min) || (max && val > *max))
2502                                 continue;
2503                         *i = val;
2504                 } else {
2505                         val = convdiv * (*i) / convmul;
2506                         if (!first)
2507                                 err = proc_put_char(&buffer, &left, '\t');
2508                         err = proc_put_long(&buffer, &left, val, false);
2509                         if (err)
2510                                 break;
2511                 }
2512         }
2513
2514         if (!write && !first && left && !err)
2515                 err = proc_put_char(&buffer, &left, '\n');
2516         if (write && !err)
2517                 left -= proc_skip_spaces(&kbuf);
2518 free:
2519         if (write) {
2520                 free_page(page);
2521                 if (first)
2522                         return err ? : -EINVAL;
2523         }
2524         *lenp -= left;
2525         *ppos += *lenp;
2526         return err;
2527 }
2528
2529 static int do_proc_doulongvec_minmax(struct ctl_table *table, int write,
2530                                      void __user *buffer,
2531                                      size_t *lenp, loff_t *ppos,
2532                                      unsigned long convmul,
2533                                      unsigned long convdiv)
2534 {
2535         return __do_proc_doulongvec_minmax(table->data, table, write,
2536                         buffer, lenp, ppos, convmul, convdiv);
2537 }
2538
2539 /**
2540  * proc_doulongvec_minmax - read a vector of long integers with min/max values
2541  * @table: the sysctl table
2542  * @write: %TRUE if this is a write to the sysctl file
2543  * @buffer: the user buffer
2544  * @lenp: the size of the user buffer
2545  * @ppos: file position
2546  *
2547  * Reads/writes up to table->maxlen/sizeof(unsigned long) unsigned long
2548  * values from/to the user buffer, treated as an ASCII string.
2549  *
2550  * This routine will ensure the values are within the range specified by
2551  * table->extra1 (min) and table->extra2 (max).
2552  *
2553  * Returns 0 on success.
2554  */
2555 int proc_doulongvec_minmax(struct ctl_table *table, int write,
2556                            void __user *buffer, size_t *lenp, loff_t *ppos)
2557 {
2558     return do_proc_doulongvec_minmax(table, write, buffer, lenp, ppos, 1l, 1l);
2559 }
2560
2561 /**
2562  * proc_doulongvec_ms_jiffies_minmax - read a vector of millisecond values with min/max values
2563  * @table: the sysctl table
2564  * @write: %TRUE if this is a write to the sysctl file
2565  * @buffer: the user buffer
2566  * @lenp: the size of the user buffer
2567  * @ppos: file position
2568  *
2569  * Reads/writes up to table->maxlen/sizeof(unsigned long) unsigned long
2570  * values from/to the user buffer, treated as an ASCII string. The values
2571  * are treated as milliseconds, and converted to jiffies when they are stored.
2572  *
2573  * This routine will ensure the values are within the range specified by
2574  * table->extra1 (min) and table->extra2 (max).
2575  *
2576  * Returns 0 on success.
2577  */
2578 int proc_doulongvec_ms_jiffies_minmax(struct ctl_table *table, int write,
2579                                       void __user *buffer,
2580                                       size_t *lenp, loff_t *ppos)
2581 {
2582     return do_proc_doulongvec_minmax(table, write, buffer,
2583                                      lenp, ppos, HZ, 1000l);
2584 }
2585
2586
2587 static int do_proc_dointvec_jiffies_conv(bool *negp, unsigned long *lvalp,
2588                                          int *valp,
2589                                          int write, void *data)
2590 {
2591         if (write) {
2592                 if (*lvalp > LONG_MAX / HZ)
2593                         return 1;
2594                 *valp = *negp ? -(*lvalp*HZ) : (*lvalp*HZ);
2595         } else {
2596                 int val = *valp;
2597                 unsigned long lval;
2598                 if (val < 0) {
2599                         *negp = true;
2600                         lval = (unsigned long)-val;
2601                 } else {
2602                         *negp = false;
2603                         lval = (unsigned long)val;
2604                 }
2605                 *lvalp = lval / HZ;
2606         }
2607         return 0;
2608 }
2609
2610 static int do_proc_dointvec_userhz_jiffies_conv(bool *negp, unsigned long *lvalp,
2611                                                 int *valp,
2612                                                 int write, void *data)
2613 {
2614         if (write) {
2615                 if (USER_HZ < HZ && *lvalp > (LONG_MAX / HZ) * USER_HZ)
2616                         return 1;
2617                 *valp = clock_t_to_jiffies(*negp ? -*lvalp : *lvalp);
2618         } else {
2619                 int val = *valp;
2620                 unsigned long lval;
2621                 if (val < 0) {
2622                         *negp = true;
2623                         lval = (unsigned long)-val;
2624                 } else {
2625                         *negp = false;
2626                         lval = (unsigned long)val;
2627                 }
2628                 *lvalp = jiffies_to_clock_t(lval);
2629         }
2630         return 0;
2631 }
2632
2633 static int do_proc_dointvec_ms_jiffies_conv(bool *negp, unsigned long *lvalp,
2634                                             int *valp,
2635                                             int write, void *data)
2636 {
2637         if (write) {
2638                 *valp = msecs_to_jiffies(*negp ? -*lvalp : *lvalp);
2639         } else {
2640                 int val = *valp;
2641                 unsigned long lval;
2642                 if (val < 0) {
2643                         *negp = true;
2644                         lval = (unsigned long)-val;
2645                 } else {
2646                         *negp = false;
2647                         lval = (unsigned long)val;
2648                 }
2649                 *lvalp = jiffies_to_msecs(lval);
2650         }
2651         return 0;
2652 }
2653
2654 /**
2655  * proc_dointvec_jiffies - read a vector of integers as seconds
2656  * @table: the sysctl table
2657  * @write: %TRUE if this is a write to the sysctl file
2658  * @buffer: the user buffer
2659  * @lenp: the size of the user buffer
2660  * @ppos: file position
2661  *
2662  * Reads/writes up to table->maxlen/sizeof(unsigned int) integer
2663  * values from/to the user buffer, treated as an ASCII string. 
2664  * The values read are assumed to be in seconds, and are converted into
2665  * jiffies.
2666  *
2667  * Returns 0 on success.
2668  */
2669 int proc_dointvec_jiffies(struct ctl_table *table, int write,
2670                           void __user *buffer, size_t *lenp, loff_t *ppos)
2671 {
2672     return do_proc_dointvec(table,write,buffer,lenp,ppos,
2673                             do_proc_dointvec_jiffies_conv,NULL);
2674 }
2675
2676 /**
2677  * proc_dointvec_userhz_jiffies - read a vector of integers as 1/USER_HZ seconds
2678  * @table: the sysctl table
2679  * @write: %TRUE if this is a write to the sysctl file
2680  * @buffer: the user buffer
2681  * @lenp: the size of the user buffer
2682  * @ppos: pointer to the file position
2683  *
2684  * Reads/writes up to table->maxlen/sizeof(unsigned int) integer
2685  * values from/to the user buffer, treated as an ASCII string. 
2686  * The values read are assumed to be in 1/USER_HZ seconds, and 
2687  * are converted into jiffies.
2688  *
2689  * Returns 0 on success.
2690  */
2691 int proc_dointvec_userhz_jiffies(struct ctl_table *table, int write,
2692                                  void __user *buffer, size_t *lenp, loff_t *ppos)
2693 {
2694     return do_proc_dointvec(table,write,buffer,lenp,ppos,
2695                             do_proc_dointvec_userhz_jiffies_conv,NULL);
2696 }
2697
2698 /**
2699  * proc_dointvec_ms_jiffies - read a vector of integers as 1 milliseconds
2700  * @table: the sysctl table
2701  * @write: %TRUE if this is a write to the sysctl file
2702  * @buffer: the user buffer
2703  * @lenp: the size of the user buffer
2704  * @ppos: file position
2705  * @ppos: the current position in the file
2706  *
2707  * Reads/writes up to table->maxlen/sizeof(unsigned int) integer
2708  * values from/to the user buffer, treated as an ASCII string. 
2709  * The values read are assumed to be in 1/1000 seconds, and 
2710  * are converted into jiffies.
2711  *
2712  * Returns 0 on success.
2713  */
2714 int proc_dointvec_ms_jiffies(struct ctl_table *table, int write,
2715                              void __user *buffer, size_t *lenp, loff_t *ppos)
2716 {
2717         return do_proc_dointvec(table, write, buffer, lenp, ppos,
2718                                 do_proc_dointvec_ms_jiffies_conv, NULL);
2719 }
2720
2721 static int proc_do_cad_pid(struct ctl_table *table, int write,
2722                            void __user *buffer, size_t *lenp, loff_t *ppos)
2723 {
2724         struct pid *new_pid;
2725         pid_t tmp;
2726         int r;
2727
2728         tmp = pid_vnr(cad_pid);
2729
2730         r = __do_proc_dointvec(&tmp, table, write, buffer,
2731                                lenp, ppos, NULL, NULL);
2732         if (r || !write)
2733                 return r;
2734
2735         new_pid = find_get_pid(tmp);
2736         if (!new_pid)
2737                 return -ESRCH;
2738
2739         put_pid(xchg(&cad_pid, new_pid));
2740         return 0;
2741 }
2742
2743 /**
2744  * proc_do_large_bitmap - read/write from/to a large bitmap
2745  * @table: the sysctl table
2746  * @write: %TRUE if this is a write to the sysctl file
2747  * @buffer: the user buffer
2748  * @lenp: the size of the user buffer
2749  * @ppos: file position
2750  *
2751  * The bitmap is stored at table->data and the bitmap length (in bits)
2752  * in table->maxlen.
2753  *
2754  * We use a range comma separated format (e.g. 1,3-4,10-10) so that
2755  * large bitmaps may be represented in a compact manner. Writing into
2756  * the file will clear the bitmap then update it with the given input.
2757  *
2758  * Returns 0 on success.
2759  */
2760 int proc_do_large_bitmap(struct ctl_table *table, int write,
2761                          void __user *buffer, size_t *lenp, loff_t *ppos)
2762 {
2763         int err = 0;
2764         bool first = 1;
2765         size_t left = *lenp;
2766         unsigned long bitmap_len = table->maxlen;
2767         unsigned long *bitmap = (unsigned long *) table->data;
2768         unsigned long *tmp_bitmap = NULL;
2769         char tr_a[] = { '-', ',', '\n' }, tr_b[] = { ',', '\n', 0 }, c;
2770
2771         if (!bitmap_len || !left || (*ppos && !write)) {
2772                 *lenp = 0;
2773                 return 0;
2774         }
2775
2776         if (write) {
2777                 unsigned long page = 0;
2778                 char *kbuf;
2779
2780                 if (left > PAGE_SIZE - 1)
2781                         left = PAGE_SIZE - 1;
2782
2783                 page = __get_free_page(GFP_TEMPORARY);
2784                 kbuf = (char *) page;
2785                 if (!kbuf)
2786                         return -ENOMEM;
2787                 if (copy_from_user(kbuf, buffer, left)) {
2788                         free_page(page);
2789                         return -EFAULT;
2790                 }
2791                 kbuf[left] = 0;
2792
2793                 tmp_bitmap = kzalloc(BITS_TO_LONGS(bitmap_len) * sizeof(unsigned long),
2794                                      GFP_KERNEL);
2795                 if (!tmp_bitmap) {
2796                         free_page(page);
2797                         return -ENOMEM;
2798                 }
2799                 proc_skip_char(&kbuf, &left, '\n');
2800                 while (!err && left) {
2801                         unsigned long val_a, val_b;
2802                         bool neg;
2803
2804                         err = proc_get_long(&kbuf, &left, &val_a, &neg, tr_a,
2805                                              sizeof(tr_a), &c);
2806                         if (err)
2807                                 break;
2808                         if (val_a >= bitmap_len || neg) {
2809                                 err = -EINVAL;
2810                                 break;
2811                         }
2812
2813                         val_b = val_a;
2814                         if (left) {
2815                                 kbuf++;
2816                                 left--;
2817                         }
2818
2819                         if (c == '-') {
2820                                 err = proc_get_long(&kbuf, &left, &val_b,
2821                                                      &neg, tr_b, sizeof(tr_b),
2822                                                      &c);
2823                                 if (err)
2824                                         break;
2825                                 if (val_b >= bitmap_len || neg ||
2826                                     val_a > val_b) {
2827                                         err = -EINVAL;
2828                                         break;
2829                                 }
2830                                 if (left) {
2831                                         kbuf++;
2832                                         left--;
2833                                 }
2834                         }
2835
2836                         while (val_a <= val_b)
2837                                 set_bit(val_a++, tmp_bitmap);
2838
2839                         first = 0;
2840                         proc_skip_char(&kbuf, &left, '\n');
2841                 }
2842                 free_page(page);
2843         } else {
2844                 unsigned long bit_a, bit_b = 0;
2845
2846                 while (left) {
2847                         bit_a = find_next_bit(bitmap, bitmap_len, bit_b);
2848                         if (bit_a >= bitmap_len)
2849                                 break;
2850                         bit_b = find_next_zero_bit(bitmap, bitmap_len,
2851                                                    bit_a + 1) - 1;
2852
2853                         if (!first) {
2854                                 err = proc_put_char(&buffer, &left, ',');
2855                                 if (err)
2856                                         break;
2857                         }
2858                         err = proc_put_long(&buffer, &left, bit_a, false);
2859                         if (err)
2860                                 break;
2861                         if (bit_a != bit_b) {
2862                                 err = proc_put_char(&buffer, &left, '-');
2863                                 if (err)
2864                                         break;
2865                                 err = proc_put_long(&buffer, &left, bit_b, false);
2866                                 if (err)
2867                                         break;
2868                         }
2869
2870                         first = 0; bit_b++;
2871                 }
2872                 if (!err)
2873                         err = proc_put_char(&buffer, &left, '\n');
2874         }
2875
2876         if (!err) {
2877                 if (write) {
2878                         if (*ppos)
2879                                 bitmap_or(bitmap, bitmap, tmp_bitmap, bitmap_len);
2880                         else
2881                                 memcpy(bitmap, tmp_bitmap,
2882                                         BITS_TO_LONGS(bitmap_len) * sizeof(unsigned long));
2883                 }
2884                 kfree(tmp_bitmap);
2885                 *lenp -= left;
2886                 *ppos += *lenp;
2887                 return 0;
2888         } else {
2889                 kfree(tmp_bitmap);
2890                 return err;
2891         }
2892 }
2893
2894 #else /* CONFIG_PROC_FS */
2895
2896 int proc_dostring(struct ctl_table *table, int write,
2897                   void __user *buffer, size_t *lenp, loff_t *ppos)
2898 {
2899         return -ENOSYS;
2900 }
2901
2902 int proc_dointvec(struct ctl_table *table, int write,
2903                   void __user *buffer, size_t *lenp, loff_t *ppos)
2904 {
2905         return -ENOSYS;
2906 }
2907
2908 int proc_dointvec_minmax(struct ctl_table *table, int write,
2909                     void __user *buffer, size_t *lenp, loff_t *ppos)
2910 {
2911         return -ENOSYS;
2912 }
2913
2914 int proc_dointvec_jiffies(struct ctl_table *table, int write,
2915                     void __user *buffer, size_t *lenp, loff_t *ppos)
2916 {
2917         return -ENOSYS;
2918 }
2919
2920 int proc_dointvec_userhz_jiffies(struct ctl_table *table, int write,
2921                     void __user *buffer, size_t *lenp, loff_t *ppos)
2922 {
2923         return -ENOSYS;
2924 }
2925
2926 int proc_dointvec_ms_jiffies(struct ctl_table *table, int write,
2927                              void __user *buffer, size_t *lenp, loff_t *ppos)
2928 {
2929         return -ENOSYS;
2930 }
2931
2932 int proc_doulongvec_minmax(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_doulongvec_ms_jiffies_minmax(struct ctl_table *table, int write,
2939                                       void __user *buffer,
2940                                       size_t *lenp, loff_t *ppos)
2941 {
2942     return -ENOSYS;
2943 }
2944
2945
2946 #endif /* CONFIG_PROC_FS */
2947
2948 /*
2949  * No sense putting this after each symbol definition, twice,
2950  * exception granted :-)
2951  */
2952 EXPORT_SYMBOL(proc_dointvec);
2953 EXPORT_SYMBOL(proc_dointvec_jiffies);
2954 EXPORT_SYMBOL(proc_dointvec_minmax);
2955 EXPORT_SYMBOL(proc_dointvec_userhz_jiffies);
2956 EXPORT_SYMBOL(proc_dointvec_ms_jiffies);
2957 EXPORT_SYMBOL(proc_dostring);
2958 EXPORT_SYMBOL(proc_doulongvec_minmax);
2959 EXPORT_SYMBOL(proc_doulongvec_ms_jiffies_minmax);
2960 EXPORT_SYMBOL(register_sysctl_table);
2961 EXPORT_SYMBOL(register_sysctl_paths);
2962 EXPORT_SYMBOL(unregister_sysctl_table);