Pull sony-2.6.24 into release branch
[linux-2.6] / 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/proc_fs.h>
27 #include <linux/capability.h>
28 #include <linux/ctype.h>
29 #include <linux/utsname.h>
30 #include <linux/smp_lock.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/hugetlb.h>
40 #include <linux/security.h>
41 #include <linux/initrd.h>
42 #include <linux/times.h>
43 #include <linux/limits.h>
44 #include <linux/dcache.h>
45 #include <linux/syscalls.h>
46 #include <linux/nfs_fs.h>
47 #include <linux/acpi.h>
48 #include <linux/reboot.h>
49
50 #include <asm/uaccess.h>
51 #include <asm/processor.h>
52
53 #ifdef CONFIG_X86
54 #include <asm/nmi.h>
55 #include <asm/stacktrace.h>
56 #endif
57
58 #if defined(CONFIG_SYSCTL)
59
60 /* External variables not in a header file. */
61 extern int C_A_D;
62 extern int print_fatal_signals;
63 extern int sysctl_overcommit_memory;
64 extern int sysctl_overcommit_ratio;
65 extern int sysctl_panic_on_oom;
66 extern int max_threads;
67 extern int core_uses_pid;
68 extern int suid_dumpable;
69 extern char core_pattern[];
70 extern int pid_max;
71 extern int min_free_kbytes;
72 extern int printk_ratelimit_jiffies;
73 extern int printk_ratelimit_burst;
74 extern int pid_max_min, pid_max_max;
75 extern int sysctl_drop_caches;
76 extern int percpu_pagelist_fraction;
77 extern int compat_log;
78 extern int maps_protect;
79 extern int sysctl_stat_interval;
80 extern int audit_argv_kb;
81
82 /* this is needed for the proc_dointvec_minmax for [fs_]overflow UID and GID */
83 static int maxolduid = 65535;
84 static int minolduid;
85 static int min_percpu_pagelist_fract = 8;
86
87 static int ngroups_max = NGROUPS_MAX;
88
89 #ifdef CONFIG_KMOD
90 extern char modprobe_path[];
91 #endif
92 #ifdef CONFIG_CHR_DEV_SG
93 extern int sg_big_buff;
94 #endif
95
96 #ifdef __sparc__
97 extern char reboot_command [];
98 extern int stop_a_enabled;
99 extern int scons_pwroff;
100 #endif
101
102 #ifdef __hppa__
103 extern int pwrsw_enabled;
104 extern int unaligned_enabled;
105 #endif
106
107 #ifdef CONFIG_S390
108 #ifdef CONFIG_MATHEMU
109 extern int sysctl_ieee_emulation_warnings;
110 #endif
111 extern int sysctl_userprocess_debug;
112 extern int spin_retry;
113 #endif
114
115 extern int sysctl_hz_timer;
116
117 #ifdef CONFIG_BSD_PROCESS_ACCT
118 extern int acct_parm[];
119 #endif
120
121 #ifdef CONFIG_IA64
122 extern int no_unaligned_warning;
123 #endif
124
125 #ifdef CONFIG_RT_MUTEXES
126 extern int max_lock_depth;
127 #endif
128
129 #ifdef CONFIG_SYSCTL_SYSCALL
130 static int parse_table(int __user *, int, void __user *, size_t __user *,
131                 void __user *, size_t, ctl_table *);
132 #endif
133
134
135 #ifdef CONFIG_PROC_SYSCTL
136 static int proc_do_cad_pid(ctl_table *table, int write, struct file *filp,
137                   void __user *buffer, size_t *lenp, loff_t *ppos);
138 static int proc_dointvec_taint(ctl_table *table, int write, struct file *filp,
139                                void __user *buffer, size_t *lenp, loff_t *ppos);
140 #endif
141
142 static ctl_table root_table[];
143 static struct ctl_table_header root_table_header =
144         { root_table, LIST_HEAD_INIT(root_table_header.ctl_entry) };
145
146 static ctl_table kern_table[];
147 static ctl_table vm_table[];
148 static ctl_table fs_table[];
149 static ctl_table debug_table[];
150 static ctl_table dev_table[];
151 extern ctl_table random_table[];
152 #ifdef CONFIG_UNIX98_PTYS
153 extern ctl_table pty_table[];
154 #endif
155 #ifdef CONFIG_INOTIFY_USER
156 extern ctl_table inotify_table[];
157 #endif
158
159 #ifdef HAVE_ARCH_PICK_MMAP_LAYOUT
160 int sysctl_legacy_va_layout;
161 #endif
162
163 extern int prove_locking;
164 extern int lock_stat;
165
166 /* The default sysctl tables: */
167
168 static ctl_table root_table[] = {
169         {
170                 .ctl_name       = CTL_KERN,
171                 .procname       = "kernel",
172                 .mode           = 0555,
173                 .child          = kern_table,
174         },
175         {
176                 .ctl_name       = CTL_VM,
177                 .procname       = "vm",
178                 .mode           = 0555,
179                 .child          = vm_table,
180         },
181 #ifdef CONFIG_NET
182         {
183                 .ctl_name       = CTL_NET,
184                 .procname       = "net",
185                 .mode           = 0555,
186                 .child          = net_table,
187         },
188 #endif
189         {
190                 .ctl_name       = CTL_FS,
191                 .procname       = "fs",
192                 .mode           = 0555,
193                 .child          = fs_table,
194         },
195         {
196                 .ctl_name       = CTL_DEBUG,
197                 .procname       = "debug",
198                 .mode           = 0555,
199                 .child          = debug_table,
200         },
201         {
202                 .ctl_name       = CTL_DEV,
203                 .procname       = "dev",
204                 .mode           = 0555,
205                 .child          = dev_table,
206         },
207 /*
208  * NOTE: do not add new entries to this table unless you have read
209  * Documentation/sysctl/ctl_unnumbered.txt
210  */
211         { .ctl_name = 0 }
212 };
213
214 #ifdef CONFIG_SCHED_DEBUG
215 static unsigned long min_sched_granularity_ns = 100000;         /* 100 usecs */
216 static unsigned long max_sched_granularity_ns = 1000000000;     /* 1 second */
217 static unsigned long min_wakeup_granularity_ns;                 /* 0 usecs */
218 static unsigned long max_wakeup_granularity_ns = 1000000000;    /* 1 second */
219 #endif
220
221 static ctl_table kern_table[] = {
222 #ifdef CONFIG_SCHED_DEBUG
223         {
224                 .ctl_name       = CTL_UNNUMBERED,
225                 .procname       = "sched_min_granularity_ns",
226                 .data           = &sysctl_sched_min_granularity,
227                 .maxlen         = sizeof(unsigned int),
228                 .mode           = 0644,
229                 .proc_handler   = &proc_dointvec_minmax,
230                 .strategy       = &sysctl_intvec,
231                 .extra1         = &min_sched_granularity_ns,
232                 .extra2         = &max_sched_granularity_ns,
233         },
234         {
235                 .ctl_name       = CTL_UNNUMBERED,
236                 .procname       = "sched_latency_ns",
237                 .data           = &sysctl_sched_latency,
238                 .maxlen         = sizeof(unsigned int),
239                 .mode           = 0644,
240                 .proc_handler   = &proc_dointvec_minmax,
241                 .strategy       = &sysctl_intvec,
242                 .extra1         = &min_sched_granularity_ns,
243                 .extra2         = &max_sched_granularity_ns,
244         },
245         {
246                 .ctl_name       = CTL_UNNUMBERED,
247                 .procname       = "sched_wakeup_granularity_ns",
248                 .data           = &sysctl_sched_wakeup_granularity,
249                 .maxlen         = sizeof(unsigned int),
250                 .mode           = 0644,
251                 .proc_handler   = &proc_dointvec_minmax,
252                 .strategy       = &sysctl_intvec,
253                 .extra1         = &min_wakeup_granularity_ns,
254                 .extra2         = &max_wakeup_granularity_ns,
255         },
256         {
257                 .ctl_name       = CTL_UNNUMBERED,
258                 .procname       = "sched_batch_wakeup_granularity_ns",
259                 .data           = &sysctl_sched_batch_wakeup_granularity,
260                 .maxlen         = sizeof(unsigned int),
261                 .mode           = 0644,
262                 .proc_handler   = &proc_dointvec_minmax,
263                 .strategy       = &sysctl_intvec,
264                 .extra1         = &min_wakeup_granularity_ns,
265                 .extra2         = &max_wakeup_granularity_ns,
266         },
267         {
268                 .ctl_name       = CTL_UNNUMBERED,
269                 .procname       = "sched_stat_granularity_ns",
270                 .data           = &sysctl_sched_stat_granularity,
271                 .maxlen         = sizeof(unsigned int),
272                 .mode           = 0644,
273                 .proc_handler   = &proc_dointvec_minmax,
274                 .strategy       = &sysctl_intvec,
275                 .extra1         = &min_wakeup_granularity_ns,
276                 .extra2         = &max_wakeup_granularity_ns,
277         },
278         {
279                 .ctl_name       = CTL_UNNUMBERED,
280                 .procname       = "sched_runtime_limit_ns",
281                 .data           = &sysctl_sched_runtime_limit,
282                 .maxlen         = sizeof(unsigned int),
283                 .mode           = 0644,
284                 .proc_handler   = &proc_dointvec_minmax,
285                 .strategy       = &sysctl_intvec,
286                 .extra1         = &min_sched_granularity_ns,
287                 .extra2         = &max_sched_granularity_ns,
288         },
289         {
290                 .ctl_name       = CTL_UNNUMBERED,
291                 .procname       = "sched_child_runs_first",
292                 .data           = &sysctl_sched_child_runs_first,
293                 .maxlen         = sizeof(unsigned int),
294                 .mode           = 0644,
295                 .proc_handler   = &proc_dointvec,
296         },
297         {
298                 .ctl_name       = CTL_UNNUMBERED,
299                 .procname       = "sched_features",
300                 .data           = &sysctl_sched_features,
301                 .maxlen         = sizeof(unsigned int),
302                 .mode           = 0644,
303                 .proc_handler   = &proc_dointvec,
304         },
305 #endif
306         {
307                 .ctl_name       = CTL_UNNUMBERED,
308                 .procname       = "sched_compat_yield",
309                 .data           = &sysctl_sched_compat_yield,
310                 .maxlen         = sizeof(unsigned int),
311                 .mode           = 0644,
312                 .proc_handler   = &proc_dointvec,
313         },
314 #ifdef CONFIG_PROVE_LOCKING
315         {
316                 .ctl_name       = CTL_UNNUMBERED,
317                 .procname       = "prove_locking",
318                 .data           = &prove_locking,
319                 .maxlen         = sizeof(int),
320                 .mode           = 0644,
321                 .proc_handler   = &proc_dointvec,
322         },
323 #endif
324 #ifdef CONFIG_LOCK_STAT
325         {
326                 .ctl_name       = CTL_UNNUMBERED,
327                 .procname       = "lock_stat",
328                 .data           = &lock_stat,
329                 .maxlen         = sizeof(int),
330                 .mode           = 0644,
331                 .proc_handler   = &proc_dointvec,
332         },
333 #endif
334         {
335                 .ctl_name       = KERN_PANIC,
336                 .procname       = "panic",
337                 .data           = &panic_timeout,
338                 .maxlen         = sizeof(int),
339                 .mode           = 0644,
340                 .proc_handler   = &proc_dointvec,
341         },
342         {
343                 .ctl_name       = KERN_CORE_USES_PID,
344                 .procname       = "core_uses_pid",
345                 .data           = &core_uses_pid,
346                 .maxlen         = sizeof(int),
347                 .mode           = 0644,
348                 .proc_handler   = &proc_dointvec,
349         },
350 #ifdef CONFIG_AUDITSYSCALL
351         {
352                 .ctl_name       = CTL_UNNUMBERED,
353                 .procname       = "audit_argv_kb",
354                 .data           = &audit_argv_kb,
355                 .maxlen         = sizeof(int),
356                 .mode           = 0644,
357                 .proc_handler   = &proc_dointvec,
358         },
359 #endif
360         {
361                 .ctl_name       = KERN_CORE_PATTERN,
362                 .procname       = "core_pattern",
363                 .data           = core_pattern,
364                 .maxlen         = CORENAME_MAX_SIZE,
365                 .mode           = 0644,
366                 .proc_handler   = &proc_dostring,
367                 .strategy       = &sysctl_string,
368         },
369 #ifdef CONFIG_PROC_SYSCTL
370         {
371                 .ctl_name       = KERN_TAINTED,
372                 .procname       = "tainted",
373                 .data           = &tainted,
374                 .maxlen         = sizeof(int),
375                 .mode           = 0644,
376                 .proc_handler   = &proc_dointvec_taint,
377         },
378 #endif
379         {
380                 .ctl_name       = KERN_CAP_BSET,
381                 .procname       = "cap-bound",
382                 .data           = &cap_bset,
383                 .maxlen         = sizeof(kernel_cap_t),
384                 .mode           = 0600,
385                 .proc_handler   = &proc_dointvec_bset,
386         },
387 #ifdef CONFIG_BLK_DEV_INITRD
388         {
389                 .ctl_name       = KERN_REALROOTDEV,
390                 .procname       = "real-root-dev",
391                 .data           = &real_root_dev,
392                 .maxlen         = sizeof(int),
393                 .mode           = 0644,
394                 .proc_handler   = &proc_dointvec,
395         },
396 #endif
397         {
398                 .ctl_name       = CTL_UNNUMBERED,
399                 .procname       = "print-fatal-signals",
400                 .data           = &print_fatal_signals,
401                 .maxlen         = sizeof(int),
402                 .mode           = 0644,
403                 .proc_handler   = &proc_dointvec,
404         },
405 #ifdef __sparc__
406         {
407                 .ctl_name       = KERN_SPARC_REBOOT,
408                 .procname       = "reboot-cmd",
409                 .data           = reboot_command,
410                 .maxlen         = 256,
411                 .mode           = 0644,
412                 .proc_handler   = &proc_dostring,
413                 .strategy       = &sysctl_string,
414         },
415         {
416                 .ctl_name       = KERN_SPARC_STOP_A,
417                 .procname       = "stop-a",
418                 .data           = &stop_a_enabled,
419                 .maxlen         = sizeof (int),
420                 .mode           = 0644,
421                 .proc_handler   = &proc_dointvec,
422         },
423         {
424                 .ctl_name       = KERN_SPARC_SCONS_PWROFF,
425                 .procname       = "scons-poweroff",
426                 .data           = &scons_pwroff,
427                 .maxlen         = sizeof (int),
428                 .mode           = 0644,
429                 .proc_handler   = &proc_dointvec,
430         },
431 #endif
432 #ifdef __hppa__
433         {
434                 .ctl_name       = KERN_HPPA_PWRSW,
435                 .procname       = "soft-power",
436                 .data           = &pwrsw_enabled,
437                 .maxlen         = sizeof (int),
438                 .mode           = 0644,
439                 .proc_handler   = &proc_dointvec,
440         },
441         {
442                 .ctl_name       = KERN_HPPA_UNALIGNED,
443                 .procname       = "unaligned-trap",
444                 .data           = &unaligned_enabled,
445                 .maxlen         = sizeof (int),
446                 .mode           = 0644,
447                 .proc_handler   = &proc_dointvec,
448         },
449 #endif
450         {
451                 .ctl_name       = KERN_CTLALTDEL,
452                 .procname       = "ctrl-alt-del",
453                 .data           = &C_A_D,
454                 .maxlen         = sizeof(int),
455                 .mode           = 0644,
456                 .proc_handler   = &proc_dointvec,
457         },
458         {
459                 .ctl_name       = KERN_PRINTK,
460                 .procname       = "printk",
461                 .data           = &console_loglevel,
462                 .maxlen         = 4*sizeof(int),
463                 .mode           = 0644,
464                 .proc_handler   = &proc_dointvec,
465         },
466 #ifdef CONFIG_KMOD
467         {
468                 .ctl_name       = KERN_MODPROBE,
469                 .procname       = "modprobe",
470                 .data           = &modprobe_path,
471                 .maxlen         = KMOD_PATH_LEN,
472                 .mode           = 0644,
473                 .proc_handler   = &proc_dostring,
474                 .strategy       = &sysctl_string,
475         },
476 #endif
477 #if defined(CONFIG_HOTPLUG) && defined(CONFIG_NET)
478         {
479                 .ctl_name       = KERN_HOTPLUG,
480                 .procname       = "hotplug",
481                 .data           = &uevent_helper,
482                 .maxlen         = UEVENT_HELPER_PATH_LEN,
483                 .mode           = 0644,
484                 .proc_handler   = &proc_dostring,
485                 .strategy       = &sysctl_string,
486         },
487 #endif
488 #ifdef CONFIG_CHR_DEV_SG
489         {
490                 .ctl_name       = KERN_SG_BIG_BUFF,
491                 .procname       = "sg-big-buff",
492                 .data           = &sg_big_buff,
493                 .maxlen         = sizeof (int),
494                 .mode           = 0444,
495                 .proc_handler   = &proc_dointvec,
496         },
497 #endif
498 #ifdef CONFIG_BSD_PROCESS_ACCT
499         {
500                 .ctl_name       = KERN_ACCT,
501                 .procname       = "acct",
502                 .data           = &acct_parm,
503                 .maxlen         = 3*sizeof(int),
504                 .mode           = 0644,
505                 .proc_handler   = &proc_dointvec,
506         },
507 #endif
508 #ifdef CONFIG_MAGIC_SYSRQ
509         {
510                 .ctl_name       = KERN_SYSRQ,
511                 .procname       = "sysrq",
512                 .data           = &__sysrq_enabled,
513                 .maxlen         = sizeof (int),
514                 .mode           = 0644,
515                 .proc_handler   = &proc_dointvec,
516         },
517 #endif
518 #ifdef CONFIG_PROC_SYSCTL
519         {
520                 .ctl_name       = KERN_CADPID,
521                 .procname       = "cad_pid",
522                 .data           = NULL,
523                 .maxlen         = sizeof (int),
524                 .mode           = 0600,
525                 .proc_handler   = &proc_do_cad_pid,
526         },
527 #endif
528         {
529                 .ctl_name       = KERN_MAX_THREADS,
530                 .procname       = "threads-max",
531                 .data           = &max_threads,
532                 .maxlen         = sizeof(int),
533                 .mode           = 0644,
534                 .proc_handler   = &proc_dointvec,
535         },
536         {
537                 .ctl_name       = KERN_RANDOM,
538                 .procname       = "random",
539                 .mode           = 0555,
540                 .child          = random_table,
541         },
542 #ifdef CONFIG_UNIX98_PTYS
543         {
544                 .ctl_name       = KERN_PTY,
545                 .procname       = "pty",
546                 .mode           = 0555,
547                 .child          = pty_table,
548         },
549 #endif
550         {
551                 .ctl_name       = KERN_OVERFLOWUID,
552                 .procname       = "overflowuid",
553                 .data           = &overflowuid,
554                 .maxlen         = sizeof(int),
555                 .mode           = 0644,
556                 .proc_handler   = &proc_dointvec_minmax,
557                 .strategy       = &sysctl_intvec,
558                 .extra1         = &minolduid,
559                 .extra2         = &maxolduid,
560         },
561         {
562                 .ctl_name       = KERN_OVERFLOWGID,
563                 .procname       = "overflowgid",
564                 .data           = &overflowgid,
565                 .maxlen         = sizeof(int),
566                 .mode           = 0644,
567                 .proc_handler   = &proc_dointvec_minmax,
568                 .strategy       = &sysctl_intvec,
569                 .extra1         = &minolduid,
570                 .extra2         = &maxolduid,
571         },
572 #ifdef CONFIG_S390
573 #ifdef CONFIG_MATHEMU
574         {
575                 .ctl_name       = KERN_IEEE_EMULATION_WARNINGS,
576                 .procname       = "ieee_emulation_warnings",
577                 .data           = &sysctl_ieee_emulation_warnings,
578                 .maxlen         = sizeof(int),
579                 .mode           = 0644,
580                 .proc_handler   = &proc_dointvec,
581         },
582 #endif
583 #ifdef CONFIG_NO_IDLE_HZ
584         {
585                 .ctl_name       = KERN_HZ_TIMER,
586                 .procname       = "hz_timer",
587                 .data           = &sysctl_hz_timer,
588                 .maxlen         = sizeof(int),
589                 .mode           = 0644,
590                 .proc_handler   = &proc_dointvec,
591         },
592 #endif
593         {
594                 .ctl_name       = KERN_S390_USER_DEBUG_LOGGING,
595                 .procname       = "userprocess_debug",
596                 .data           = &sysctl_userprocess_debug,
597                 .maxlen         = sizeof(int),
598                 .mode           = 0644,
599                 .proc_handler   = &proc_dointvec,
600         },
601 #endif
602         {
603                 .ctl_name       = KERN_PIDMAX,
604                 .procname       = "pid_max",
605                 .data           = &pid_max,
606                 .maxlen         = sizeof (int),
607                 .mode           = 0644,
608                 .proc_handler   = &proc_dointvec_minmax,
609                 .strategy       = sysctl_intvec,
610                 .extra1         = &pid_max_min,
611                 .extra2         = &pid_max_max,
612         },
613         {
614                 .ctl_name       = KERN_PANIC_ON_OOPS,
615                 .procname       = "panic_on_oops",
616                 .data           = &panic_on_oops,
617                 .maxlen         = sizeof(int),
618                 .mode           = 0644,
619                 .proc_handler   = &proc_dointvec,
620         },
621         {
622                 .ctl_name       = KERN_PRINTK_RATELIMIT,
623                 .procname       = "printk_ratelimit",
624                 .data           = &printk_ratelimit_jiffies,
625                 .maxlen         = sizeof(int),
626                 .mode           = 0644,
627                 .proc_handler   = &proc_dointvec_jiffies,
628                 .strategy       = &sysctl_jiffies,
629         },
630         {
631                 .ctl_name       = KERN_PRINTK_RATELIMIT_BURST,
632                 .procname       = "printk_ratelimit_burst",
633                 .data           = &printk_ratelimit_burst,
634                 .maxlen         = sizeof(int),
635                 .mode           = 0644,
636                 .proc_handler   = &proc_dointvec,
637         },
638         {
639                 .ctl_name       = KERN_NGROUPS_MAX,
640                 .procname       = "ngroups_max",
641                 .data           = &ngroups_max,
642                 .maxlen         = sizeof (int),
643                 .mode           = 0444,
644                 .proc_handler   = &proc_dointvec,
645         },
646 #if defined(CONFIG_X86_LOCAL_APIC) && defined(CONFIG_X86)
647         {
648                 .ctl_name       = KERN_UNKNOWN_NMI_PANIC,
649                 .procname       = "unknown_nmi_panic",
650                 .data           = &unknown_nmi_panic,
651                 .maxlen         = sizeof (int),
652                 .mode           = 0644,
653                 .proc_handler   = &proc_dointvec,
654         },
655         {
656                 .ctl_name       = KERN_NMI_WATCHDOG,
657                 .procname       = "nmi_watchdog",
658                 .data           = &nmi_watchdog_enabled,
659                 .maxlen         = sizeof (int),
660                 .mode           = 0644,
661                 .proc_handler   = &proc_nmi_enabled,
662         },
663 #endif
664 #if defined(CONFIG_X86)
665         {
666                 .ctl_name       = KERN_PANIC_ON_NMI,
667                 .procname       = "panic_on_unrecovered_nmi",
668                 .data           = &panic_on_unrecovered_nmi,
669                 .maxlen         = sizeof(int),
670                 .mode           = 0644,
671                 .proc_handler   = &proc_dointvec,
672         },
673         {
674                 .ctl_name       = KERN_BOOTLOADER_TYPE,
675                 .procname       = "bootloader_type",
676                 .data           = &bootloader_type,
677                 .maxlen         = sizeof (int),
678                 .mode           = 0444,
679                 .proc_handler   = &proc_dointvec,
680         },
681         {
682                 .ctl_name       = CTL_UNNUMBERED,
683                 .procname       = "kstack_depth_to_print",
684                 .data           = &kstack_depth_to_print,
685                 .maxlen         = sizeof(int),
686                 .mode           = 0644,
687                 .proc_handler   = &proc_dointvec,
688         },
689 #endif
690 #if defined(CONFIG_MMU)
691         {
692                 .ctl_name       = KERN_RANDOMIZE,
693                 .procname       = "randomize_va_space",
694                 .data           = &randomize_va_space,
695                 .maxlen         = sizeof(int),
696                 .mode           = 0644,
697                 .proc_handler   = &proc_dointvec,
698         },
699 #endif
700 #if defined(CONFIG_S390) && defined(CONFIG_SMP)
701         {
702                 .ctl_name       = KERN_SPIN_RETRY,
703                 .procname       = "spin_retry",
704                 .data           = &spin_retry,
705                 .maxlen         = sizeof (int),
706                 .mode           = 0644,
707                 .proc_handler   = &proc_dointvec,
708         },
709 #endif
710 #if     defined(CONFIG_ACPI_SLEEP) && defined(CONFIG_X86)
711         {
712                 .ctl_name       = KERN_ACPI_VIDEO_FLAGS,
713                 .procname       = "acpi_video_flags",
714                 .data           = &acpi_realmode_flags,
715                 .maxlen         = sizeof (unsigned long),
716                 .mode           = 0644,
717                 .proc_handler   = &proc_doulongvec_minmax,
718         },
719 #endif
720 #ifdef CONFIG_IA64
721         {
722                 .ctl_name       = KERN_IA64_UNALIGNED,
723                 .procname       = "ignore-unaligned-usertrap",
724                 .data           = &no_unaligned_warning,
725                 .maxlen         = sizeof (int),
726                 .mode           = 0644,
727                 .proc_handler   = &proc_dointvec,
728         },
729 #endif
730 #ifdef CONFIG_COMPAT
731         {
732                 .ctl_name       = KERN_COMPAT_LOG,
733                 .procname       = "compat-log",
734                 .data           = &compat_log,
735                 .maxlen         = sizeof (int),
736                 .mode           = 0644,
737                 .proc_handler   = &proc_dointvec,
738         },
739 #endif
740 #ifdef CONFIG_RT_MUTEXES
741         {
742                 .ctl_name       = KERN_MAX_LOCK_DEPTH,
743                 .procname       = "max_lock_depth",
744                 .data           = &max_lock_depth,
745                 .maxlen         = sizeof(int),
746                 .mode           = 0644,
747                 .proc_handler   = &proc_dointvec,
748         },
749 #endif
750 #ifdef CONFIG_PROC_FS
751         {
752                 .ctl_name       = CTL_UNNUMBERED,
753                 .procname       = "maps_protect",
754                 .data           = &maps_protect,
755                 .maxlen         = sizeof(int),
756                 .mode           = 0644,
757                 .proc_handler   = &proc_dointvec,
758         },
759 #endif
760         {
761                 .ctl_name       = CTL_UNNUMBERED,
762                 .procname       = "poweroff_cmd",
763                 .data           = &poweroff_cmd,
764                 .maxlen         = POWEROFF_CMD_PATH_LEN,
765                 .mode           = 0644,
766                 .proc_handler   = &proc_dostring,
767                 .strategy       = &sysctl_string,
768         },
769 /*
770  * NOTE: do not add new entries to this table unless you have read
771  * Documentation/sysctl/ctl_unnumbered.txt
772  */
773         { .ctl_name = 0 }
774 };
775
776 /* Constants for minimum and maximum testing in vm_table.
777    We use these as one-element integer vectors. */
778 static int zero;
779 static int two = 2;
780 static int one_hundred = 100;
781
782
783 static ctl_table vm_table[] = {
784         {
785                 .ctl_name       = VM_OVERCOMMIT_MEMORY,
786                 .procname       = "overcommit_memory",
787                 .data           = &sysctl_overcommit_memory,
788                 .maxlen         = sizeof(sysctl_overcommit_memory),
789                 .mode           = 0644,
790                 .proc_handler   = &proc_dointvec,
791         },
792         {
793                 .ctl_name       = VM_PANIC_ON_OOM,
794                 .procname       = "panic_on_oom",
795                 .data           = &sysctl_panic_on_oom,
796                 .maxlen         = sizeof(sysctl_panic_on_oom),
797                 .mode           = 0644,
798                 .proc_handler   = &proc_dointvec,
799         },
800         {
801                 .ctl_name       = VM_OVERCOMMIT_RATIO,
802                 .procname       = "overcommit_ratio",
803                 .data           = &sysctl_overcommit_ratio,
804                 .maxlen         = sizeof(sysctl_overcommit_ratio),
805                 .mode           = 0644,
806                 .proc_handler   = &proc_dointvec,
807         },
808         {
809                 .ctl_name       = VM_PAGE_CLUSTER,
810                 .procname       = "page-cluster", 
811                 .data           = &page_cluster,
812                 .maxlen         = sizeof(int),
813                 .mode           = 0644,
814                 .proc_handler   = &proc_dointvec,
815         },
816         {
817                 .ctl_name       = VM_DIRTY_BACKGROUND,
818                 .procname       = "dirty_background_ratio",
819                 .data           = &dirty_background_ratio,
820                 .maxlen         = sizeof(dirty_background_ratio),
821                 .mode           = 0644,
822                 .proc_handler   = &proc_dointvec_minmax,
823                 .strategy       = &sysctl_intvec,
824                 .extra1         = &zero,
825                 .extra2         = &one_hundred,
826         },
827         {
828                 .ctl_name       = VM_DIRTY_RATIO,
829                 .procname       = "dirty_ratio",
830                 .data           = &vm_dirty_ratio,
831                 .maxlen         = sizeof(vm_dirty_ratio),
832                 .mode           = 0644,
833                 .proc_handler   = &proc_dointvec_minmax,
834                 .strategy       = &sysctl_intvec,
835                 .extra1         = &zero,
836                 .extra2         = &one_hundred,
837         },
838         {
839                 .ctl_name       = VM_DIRTY_WB_CS,
840                 .procname       = "dirty_writeback_centisecs",
841                 .data           = &dirty_writeback_interval,
842                 .maxlen         = sizeof(dirty_writeback_interval),
843                 .mode           = 0644,
844                 .proc_handler   = &dirty_writeback_centisecs_handler,
845         },
846         {
847                 .ctl_name       = VM_DIRTY_EXPIRE_CS,
848                 .procname       = "dirty_expire_centisecs",
849                 .data           = &dirty_expire_interval,
850                 .maxlen         = sizeof(dirty_expire_interval),
851                 .mode           = 0644,
852                 .proc_handler   = &proc_dointvec_userhz_jiffies,
853         },
854         {
855                 .ctl_name       = VM_NR_PDFLUSH_THREADS,
856                 .procname       = "nr_pdflush_threads",
857                 .data           = &nr_pdflush_threads,
858                 .maxlen         = sizeof nr_pdflush_threads,
859                 .mode           = 0444 /* read-only*/,
860                 .proc_handler   = &proc_dointvec,
861         },
862         {
863                 .ctl_name       = VM_SWAPPINESS,
864                 .procname       = "swappiness",
865                 .data           = &vm_swappiness,
866                 .maxlen         = sizeof(vm_swappiness),
867                 .mode           = 0644,
868                 .proc_handler   = &proc_dointvec_minmax,
869                 .strategy       = &sysctl_intvec,
870                 .extra1         = &zero,
871                 .extra2         = &one_hundred,
872         },
873 #ifdef CONFIG_HUGETLB_PAGE
874          {
875                 .ctl_name       = VM_HUGETLB_PAGES,
876                 .procname       = "nr_hugepages",
877                 .data           = &max_huge_pages,
878                 .maxlen         = sizeof(unsigned long),
879                 .mode           = 0644,
880                 .proc_handler   = &hugetlb_sysctl_handler,
881                 .extra1         = (void *)&hugetlb_zero,
882                 .extra2         = (void *)&hugetlb_infinity,
883          },
884          {
885                 .ctl_name       = VM_HUGETLB_GROUP,
886                 .procname       = "hugetlb_shm_group",
887                 .data           = &sysctl_hugetlb_shm_group,
888                 .maxlen         = sizeof(gid_t),
889                 .mode           = 0644,
890                 .proc_handler   = &proc_dointvec,
891          },
892          {
893                 .ctl_name       = CTL_UNNUMBERED,
894                 .procname       = "hugepages_treat_as_movable",
895                 .data           = &hugepages_treat_as_movable,
896                 .maxlen         = sizeof(int),
897                 .mode           = 0644,
898                 .proc_handler   = &hugetlb_treat_movable_handler,
899         },
900 #endif
901         {
902                 .ctl_name       = VM_LOWMEM_RESERVE_RATIO,
903                 .procname       = "lowmem_reserve_ratio",
904                 .data           = &sysctl_lowmem_reserve_ratio,
905                 .maxlen         = sizeof(sysctl_lowmem_reserve_ratio),
906                 .mode           = 0644,
907                 .proc_handler   = &lowmem_reserve_ratio_sysctl_handler,
908                 .strategy       = &sysctl_intvec,
909         },
910         {
911                 .ctl_name       = VM_DROP_PAGECACHE,
912                 .procname       = "drop_caches",
913                 .data           = &sysctl_drop_caches,
914                 .maxlen         = sizeof(int),
915                 .mode           = 0644,
916                 .proc_handler   = drop_caches_sysctl_handler,
917                 .strategy       = &sysctl_intvec,
918         },
919         {
920                 .ctl_name       = VM_MIN_FREE_KBYTES,
921                 .procname       = "min_free_kbytes",
922                 .data           = &min_free_kbytes,
923                 .maxlen         = sizeof(min_free_kbytes),
924                 .mode           = 0644,
925                 .proc_handler   = &min_free_kbytes_sysctl_handler,
926                 .strategy       = &sysctl_intvec,
927                 .extra1         = &zero,
928         },
929         {
930                 .ctl_name       = VM_PERCPU_PAGELIST_FRACTION,
931                 .procname       = "percpu_pagelist_fraction",
932                 .data           = &percpu_pagelist_fraction,
933                 .maxlen         = sizeof(percpu_pagelist_fraction),
934                 .mode           = 0644,
935                 .proc_handler   = &percpu_pagelist_fraction_sysctl_handler,
936                 .strategy       = &sysctl_intvec,
937                 .extra1         = &min_percpu_pagelist_fract,
938         },
939 #ifdef CONFIG_MMU
940         {
941                 .ctl_name       = VM_MAX_MAP_COUNT,
942                 .procname       = "max_map_count",
943                 .data           = &sysctl_max_map_count,
944                 .maxlen         = sizeof(sysctl_max_map_count),
945                 .mode           = 0644,
946                 .proc_handler   = &proc_dointvec
947         },
948 #endif
949         {
950                 .ctl_name       = VM_LAPTOP_MODE,
951                 .procname       = "laptop_mode",
952                 .data           = &laptop_mode,
953                 .maxlen         = sizeof(laptop_mode),
954                 .mode           = 0644,
955                 .proc_handler   = &proc_dointvec_jiffies,
956                 .strategy       = &sysctl_jiffies,
957         },
958         {
959                 .ctl_name       = VM_BLOCK_DUMP,
960                 .procname       = "block_dump",
961                 .data           = &block_dump,
962                 .maxlen         = sizeof(block_dump),
963                 .mode           = 0644,
964                 .proc_handler   = &proc_dointvec,
965                 .strategy       = &sysctl_intvec,
966                 .extra1         = &zero,
967         },
968         {
969                 .ctl_name       = VM_VFS_CACHE_PRESSURE,
970                 .procname       = "vfs_cache_pressure",
971                 .data           = &sysctl_vfs_cache_pressure,
972                 .maxlen         = sizeof(sysctl_vfs_cache_pressure),
973                 .mode           = 0644,
974                 .proc_handler   = &proc_dointvec,
975                 .strategy       = &sysctl_intvec,
976                 .extra1         = &zero,
977         },
978 #ifdef HAVE_ARCH_PICK_MMAP_LAYOUT
979         {
980                 .ctl_name       = VM_LEGACY_VA_LAYOUT,
981                 .procname       = "legacy_va_layout",
982                 .data           = &sysctl_legacy_va_layout,
983                 .maxlen         = sizeof(sysctl_legacy_va_layout),
984                 .mode           = 0644,
985                 .proc_handler   = &proc_dointvec,
986                 .strategy       = &sysctl_intvec,
987                 .extra1         = &zero,
988         },
989 #endif
990 #ifdef CONFIG_NUMA
991         {
992                 .ctl_name       = VM_ZONE_RECLAIM_MODE,
993                 .procname       = "zone_reclaim_mode",
994                 .data           = &zone_reclaim_mode,
995                 .maxlen         = sizeof(zone_reclaim_mode),
996                 .mode           = 0644,
997                 .proc_handler   = &proc_dointvec,
998                 .strategy       = &sysctl_intvec,
999                 .extra1         = &zero,
1000         },
1001         {
1002                 .ctl_name       = VM_MIN_UNMAPPED,
1003                 .procname       = "min_unmapped_ratio",
1004                 .data           = &sysctl_min_unmapped_ratio,
1005                 .maxlen         = sizeof(sysctl_min_unmapped_ratio),
1006                 .mode           = 0644,
1007                 .proc_handler   = &sysctl_min_unmapped_ratio_sysctl_handler,
1008                 .strategy       = &sysctl_intvec,
1009                 .extra1         = &zero,
1010                 .extra2         = &one_hundred,
1011         },
1012         {
1013                 .ctl_name       = VM_MIN_SLAB,
1014                 .procname       = "min_slab_ratio",
1015                 .data           = &sysctl_min_slab_ratio,
1016                 .maxlen         = sizeof(sysctl_min_slab_ratio),
1017                 .mode           = 0644,
1018                 .proc_handler   = &sysctl_min_slab_ratio_sysctl_handler,
1019                 .strategy       = &sysctl_intvec,
1020                 .extra1         = &zero,
1021                 .extra2         = &one_hundred,
1022         },
1023 #endif
1024 #ifdef CONFIG_SMP
1025         {
1026                 .ctl_name       = CTL_UNNUMBERED,
1027                 .procname       = "stat_interval",
1028                 .data           = &sysctl_stat_interval,
1029                 .maxlen         = sizeof(sysctl_stat_interval),
1030                 .mode           = 0644,
1031                 .proc_handler   = &proc_dointvec_jiffies,
1032                 .strategy       = &sysctl_jiffies,
1033         },
1034 #endif
1035 #ifdef CONFIG_SECURITY
1036         {
1037                 .ctl_name       = CTL_UNNUMBERED,
1038                 .procname       = "mmap_min_addr",
1039                 .data           = &mmap_min_addr,
1040                 .maxlen         = sizeof(unsigned long),
1041                 .mode           = 0644,
1042                 .proc_handler   = &proc_doulongvec_minmax,
1043         },
1044 #endif
1045 #ifdef CONFIG_NUMA
1046         {
1047                 .ctl_name       = CTL_UNNUMBERED,
1048                 .procname       = "numa_zonelist_order",
1049                 .data           = &numa_zonelist_order,
1050                 .maxlen         = NUMA_ZONELIST_ORDER_LEN,
1051                 .mode           = 0644,
1052                 .proc_handler   = &numa_zonelist_order_handler,
1053                 .strategy       = &sysctl_string,
1054         },
1055 #endif
1056 #if defined(CONFIG_X86_32) || \
1057    (defined(CONFIG_SUPERH) && defined(CONFIG_VSYSCALL))
1058         {
1059                 .ctl_name       = VM_VDSO_ENABLED,
1060                 .procname       = "vdso_enabled",
1061                 .data           = &vdso_enabled,
1062                 .maxlen         = sizeof(vdso_enabled),
1063                 .mode           = 0644,
1064                 .proc_handler   = &proc_dointvec,
1065                 .strategy       = &sysctl_intvec,
1066                 .extra1         = &zero,
1067         },
1068 #endif
1069 /*
1070  * NOTE: do not add new entries to this table unless you have read
1071  * Documentation/sysctl/ctl_unnumbered.txt
1072  */
1073         { .ctl_name = 0 }
1074 };
1075
1076 #if defined(CONFIG_BINFMT_MISC) || defined(CONFIG_BINFMT_MISC_MODULE)
1077 static ctl_table binfmt_misc_table[] = {
1078         { .ctl_name = 0 }
1079 };
1080 #endif
1081
1082 static ctl_table fs_table[] = {
1083         {
1084                 .ctl_name       = FS_NRINODE,
1085                 .procname       = "inode-nr",
1086                 .data           = &inodes_stat,
1087                 .maxlen         = 2*sizeof(int),
1088                 .mode           = 0444,
1089                 .proc_handler   = &proc_dointvec,
1090         },
1091         {
1092                 .ctl_name       = FS_STATINODE,
1093                 .procname       = "inode-state",
1094                 .data           = &inodes_stat,
1095                 .maxlen         = 7*sizeof(int),
1096                 .mode           = 0444,
1097                 .proc_handler   = &proc_dointvec,
1098         },
1099         {
1100                 .ctl_name       = FS_NRFILE,
1101                 .procname       = "file-nr",
1102                 .data           = &files_stat,
1103                 .maxlen         = 3*sizeof(int),
1104                 .mode           = 0444,
1105                 .proc_handler   = &proc_nr_files,
1106         },
1107         {
1108                 .ctl_name       = FS_MAXFILE,
1109                 .procname       = "file-max",
1110                 .data           = &files_stat.max_files,
1111                 .maxlen         = sizeof(int),
1112                 .mode           = 0644,
1113                 .proc_handler   = &proc_dointvec,
1114         },
1115         {
1116                 .ctl_name       = FS_DENTRY,
1117                 .procname       = "dentry-state",
1118                 .data           = &dentry_stat,
1119                 .maxlen         = 6*sizeof(int),
1120                 .mode           = 0444,
1121                 .proc_handler   = &proc_dointvec,
1122         },
1123         {
1124                 .ctl_name       = FS_OVERFLOWUID,
1125                 .procname       = "overflowuid",
1126                 .data           = &fs_overflowuid,
1127                 .maxlen         = sizeof(int),
1128                 .mode           = 0644,
1129                 .proc_handler   = &proc_dointvec_minmax,
1130                 .strategy       = &sysctl_intvec,
1131                 .extra1         = &minolduid,
1132                 .extra2         = &maxolduid,
1133         },
1134         {
1135                 .ctl_name       = FS_OVERFLOWGID,
1136                 .procname       = "overflowgid",
1137                 .data           = &fs_overflowgid,
1138                 .maxlen         = sizeof(int),
1139                 .mode           = 0644,
1140                 .proc_handler   = &proc_dointvec_minmax,
1141                 .strategy       = &sysctl_intvec,
1142                 .extra1         = &minolduid,
1143                 .extra2         = &maxolduid,
1144         },
1145         {
1146                 .ctl_name       = FS_LEASES,
1147                 .procname       = "leases-enable",
1148                 .data           = &leases_enable,
1149                 .maxlen         = sizeof(int),
1150                 .mode           = 0644,
1151                 .proc_handler   = &proc_dointvec,
1152         },
1153 #ifdef CONFIG_DNOTIFY
1154         {
1155                 .ctl_name       = FS_DIR_NOTIFY,
1156                 .procname       = "dir-notify-enable",
1157                 .data           = &dir_notify_enable,
1158                 .maxlen         = sizeof(int),
1159                 .mode           = 0644,
1160                 .proc_handler   = &proc_dointvec,
1161         },
1162 #endif
1163 #ifdef CONFIG_MMU
1164         {
1165                 .ctl_name       = FS_LEASE_TIME,
1166                 .procname       = "lease-break-time",
1167                 .data           = &lease_break_time,
1168                 .maxlen         = sizeof(int),
1169                 .mode           = 0644,
1170                 .proc_handler   = &proc_dointvec_minmax,
1171                 .strategy       = &sysctl_intvec,
1172                 .extra1         = &zero,
1173                 .extra2         = &two,
1174         },
1175         {
1176                 .ctl_name       = FS_AIO_NR,
1177                 .procname       = "aio-nr",
1178                 .data           = &aio_nr,
1179                 .maxlen         = sizeof(aio_nr),
1180                 .mode           = 0444,
1181                 .proc_handler   = &proc_doulongvec_minmax,
1182         },
1183         {
1184                 .ctl_name       = FS_AIO_MAX_NR,
1185                 .procname       = "aio-max-nr",
1186                 .data           = &aio_max_nr,
1187                 .maxlen         = sizeof(aio_max_nr),
1188                 .mode           = 0644,
1189                 .proc_handler   = &proc_doulongvec_minmax,
1190         },
1191 #ifdef CONFIG_INOTIFY_USER
1192         {
1193                 .ctl_name       = FS_INOTIFY,
1194                 .procname       = "inotify",
1195                 .mode           = 0555,
1196                 .child          = inotify_table,
1197         },
1198 #endif  
1199 #endif
1200         {
1201                 .ctl_name       = KERN_SETUID_DUMPABLE,
1202                 .procname       = "suid_dumpable",
1203                 .data           = &suid_dumpable,
1204                 .maxlen         = sizeof(int),
1205                 .mode           = 0644,
1206                 .proc_handler   = &proc_dointvec,
1207         },
1208 #if defined(CONFIG_BINFMT_MISC) || defined(CONFIG_BINFMT_MISC_MODULE)
1209         {
1210                 .ctl_name       = CTL_UNNUMBERED,
1211                 .procname       = "binfmt_misc",
1212                 .mode           = 0555,
1213                 .child          = binfmt_misc_table,
1214         },
1215 #endif
1216 /*
1217  * NOTE: do not add new entries to this table unless you have read
1218  * Documentation/sysctl/ctl_unnumbered.txt
1219  */
1220         { .ctl_name = 0 }
1221 };
1222
1223 static ctl_table debug_table[] = {
1224 #ifdef CONFIG_X86
1225         {
1226                 .ctl_name       = CTL_UNNUMBERED,
1227                 .procname       = "exception-trace",
1228                 .data           = &show_unhandled_signals,
1229                 .maxlen         = sizeof(int),
1230                 .mode           = 0644,
1231                 .proc_handler   = proc_dointvec
1232         },
1233 #endif
1234         { .ctl_name = 0 }
1235 };
1236
1237 static ctl_table dev_table[] = {
1238         { .ctl_name = 0 }
1239 };
1240
1241 static DEFINE_SPINLOCK(sysctl_lock);
1242
1243 /* called under sysctl_lock */
1244 static int use_table(struct ctl_table_header *p)
1245 {
1246         if (unlikely(p->unregistering))
1247                 return 0;
1248         p->used++;
1249         return 1;
1250 }
1251
1252 /* called under sysctl_lock */
1253 static void unuse_table(struct ctl_table_header *p)
1254 {
1255         if (!--p->used)
1256                 if (unlikely(p->unregistering))
1257                         complete(p->unregistering);
1258 }
1259
1260 /* called under sysctl_lock, will reacquire if has to wait */
1261 static void start_unregistering(struct ctl_table_header *p)
1262 {
1263         /*
1264          * if p->used is 0, nobody will ever touch that entry again;
1265          * we'll eliminate all paths to it before dropping sysctl_lock
1266          */
1267         if (unlikely(p->used)) {
1268                 struct completion wait;
1269                 init_completion(&wait);
1270                 p->unregistering = &wait;
1271                 spin_unlock(&sysctl_lock);
1272                 wait_for_completion(&wait);
1273                 spin_lock(&sysctl_lock);
1274         }
1275         /*
1276          * do not remove from the list until nobody holds it; walking the
1277          * list in do_sysctl() relies on that.
1278          */
1279         list_del_init(&p->ctl_entry);
1280 }
1281
1282 void sysctl_head_finish(struct ctl_table_header *head)
1283 {
1284         if (!head)
1285                 return;
1286         spin_lock(&sysctl_lock);
1287         unuse_table(head);
1288         spin_unlock(&sysctl_lock);
1289 }
1290
1291 struct ctl_table_header *sysctl_head_next(struct ctl_table_header *prev)
1292 {
1293         struct ctl_table_header *head;
1294         struct list_head *tmp;
1295         spin_lock(&sysctl_lock);
1296         if (prev) {
1297                 tmp = &prev->ctl_entry;
1298                 unuse_table(prev);
1299                 goto next;
1300         }
1301         tmp = &root_table_header.ctl_entry;
1302         for (;;) {
1303                 head = list_entry(tmp, struct ctl_table_header, ctl_entry);
1304
1305                 if (!use_table(head))
1306                         goto next;
1307                 spin_unlock(&sysctl_lock);
1308                 return head;
1309         next:
1310                 tmp = tmp->next;
1311                 if (tmp == &root_table_header.ctl_entry)
1312                         break;
1313         }
1314         spin_unlock(&sysctl_lock);
1315         return NULL;
1316 }
1317
1318 #ifdef CONFIG_SYSCTL_SYSCALL
1319 int do_sysctl(int __user *name, int nlen, void __user *oldval, size_t __user *oldlenp,
1320                void __user *newval, size_t newlen)
1321 {
1322         struct ctl_table_header *head;
1323         int error = -ENOTDIR;
1324
1325         if (nlen <= 0 || nlen >= CTL_MAXNAME)
1326                 return -ENOTDIR;
1327         if (oldval) {
1328                 int old_len;
1329                 if (!oldlenp || get_user(old_len, oldlenp))
1330                         return -EFAULT;
1331         }
1332
1333         for (head = sysctl_head_next(NULL); head;
1334                         head = sysctl_head_next(head)) {
1335                 error = parse_table(name, nlen, oldval, oldlenp, 
1336                                         newval, newlen, head->ctl_table);
1337                 if (error != -ENOTDIR) {
1338                         sysctl_head_finish(head);
1339                         break;
1340                 }
1341         }
1342         return error;
1343 }
1344
1345 asmlinkage long sys_sysctl(struct __sysctl_args __user *args)
1346 {
1347         struct __sysctl_args tmp;
1348         int error;
1349
1350         if (copy_from_user(&tmp, args, sizeof(tmp)))
1351                 return -EFAULT;
1352
1353         lock_kernel();
1354         error = do_sysctl(tmp.name, tmp.nlen, tmp.oldval, tmp.oldlenp,
1355                           tmp.newval, tmp.newlen);
1356         unlock_kernel();
1357         return error;
1358 }
1359 #endif /* CONFIG_SYSCTL_SYSCALL */
1360
1361 /*
1362  * sysctl_perm does NOT grant the superuser all rights automatically, because
1363  * some sysctl variables are readonly even to root.
1364  */
1365
1366 static int test_perm(int mode, int op)
1367 {
1368         if (!current->euid)
1369                 mode >>= 6;
1370         else if (in_egroup_p(0))
1371                 mode >>= 3;
1372         if ((mode & op & 0007) == op)
1373                 return 0;
1374         return -EACCES;
1375 }
1376
1377 int sysctl_perm(ctl_table *table, int op)
1378 {
1379         int error;
1380         error = security_sysctl(table, op);
1381         if (error)
1382                 return error;
1383         return test_perm(table->mode, op);
1384 }
1385
1386 #ifdef CONFIG_SYSCTL_SYSCALL
1387 static int parse_table(int __user *name, int nlen,
1388                        void __user *oldval, size_t __user *oldlenp,
1389                        void __user *newval, size_t newlen,
1390                        ctl_table *table)
1391 {
1392         int n;
1393 repeat:
1394         if (!nlen)
1395                 return -ENOTDIR;
1396         if (get_user(n, name))
1397                 return -EFAULT;
1398         for ( ; table->ctl_name || table->procname; table++) {
1399                 if (!table->ctl_name)
1400                         continue;
1401                 if (n == table->ctl_name) {
1402                         int error;
1403                         if (table->child) {
1404                                 if (sysctl_perm(table, 001))
1405                                         return -EPERM;
1406                                 name++;
1407                                 nlen--;
1408                                 table = table->child;
1409                                 goto repeat;
1410                         }
1411                         error = do_sysctl_strategy(table, name, nlen,
1412                                                    oldval, oldlenp,
1413                                                    newval, newlen);
1414                         return error;
1415                 }
1416         }
1417         return -ENOTDIR;
1418 }
1419
1420 /* Perform the actual read/write of a sysctl table entry. */
1421 int do_sysctl_strategy (ctl_table *table, 
1422                         int __user *name, int nlen,
1423                         void __user *oldval, size_t __user *oldlenp,
1424                         void __user *newval, size_t newlen)
1425 {
1426         int op = 0, rc;
1427         size_t len;
1428
1429         if (oldval)
1430                 op |= 004;
1431         if (newval) 
1432                 op |= 002;
1433         if (sysctl_perm(table, op))
1434                 return -EPERM;
1435
1436         if (table->strategy) {
1437                 rc = table->strategy(table, name, nlen, oldval, oldlenp,
1438                                      newval, newlen);
1439                 if (rc < 0)
1440                         return rc;
1441                 if (rc > 0)
1442                         return 0;
1443         }
1444
1445         /* If there is no strategy routine, or if the strategy returns
1446          * zero, proceed with automatic r/w */
1447         if (table->data && table->maxlen) {
1448                 if (oldval && oldlenp) {
1449                         if (get_user(len, oldlenp))
1450                                 return -EFAULT;
1451                         if (len) {
1452                                 if (len > table->maxlen)
1453                                         len = table->maxlen;
1454                                 if(copy_to_user(oldval, table->data, len))
1455                                         return -EFAULT;
1456                                 if(put_user(len, oldlenp))
1457                                         return -EFAULT;
1458                         }
1459                 }
1460                 if (newval && newlen) {
1461                         len = newlen;
1462                         if (len > table->maxlen)
1463                                 len = table->maxlen;
1464                         if(copy_from_user(table->data, newval, len))
1465                                 return -EFAULT;
1466                 }
1467         }
1468         return 0;
1469 }
1470 #endif /* CONFIG_SYSCTL_SYSCALL */
1471
1472 static void sysctl_set_parent(struct ctl_table *parent, struct ctl_table *table)
1473 {
1474         for (; table->ctl_name || table->procname; table++) {
1475                 table->parent = parent;
1476                 if (table->child)
1477                         sysctl_set_parent(table, table->child);
1478         }
1479 }
1480
1481 static __init int sysctl_init(void)
1482 {
1483         sysctl_set_parent(NULL, root_table);
1484         return 0;
1485 }
1486
1487 core_initcall(sysctl_init);
1488
1489 /**
1490  * register_sysctl_table - register a sysctl hierarchy
1491  * @table: the top-level table structure
1492  *
1493  * Register a sysctl table hierarchy. @table should be a filled in ctl_table
1494  * array. An entry with a ctl_name of 0 terminates the table. 
1495  *
1496  * The members of the &ctl_table structure are used as follows:
1497  *
1498  * ctl_name - This is the numeric sysctl value used by sysctl(2). The number
1499  *            must be unique within that level of sysctl
1500  *
1501  * procname - the name of the sysctl file under /proc/sys. Set to %NULL to not
1502  *            enter a sysctl file
1503  *
1504  * data - a pointer to data for use by proc_handler
1505  *
1506  * maxlen - the maximum size in bytes of the data
1507  *
1508  * mode - the file permissions for the /proc/sys file, and for sysctl(2)
1509  *
1510  * child - a pointer to the child sysctl table if this entry is a directory, or
1511  *         %NULL.
1512  *
1513  * proc_handler - the text handler routine (described below)
1514  *
1515  * strategy - the strategy routine (described below)
1516  *
1517  * de - for internal use by the sysctl routines
1518  *
1519  * extra1, extra2 - extra pointers usable by the proc handler routines
1520  *
1521  * Leaf nodes in the sysctl tree will be represented by a single file
1522  * under /proc; non-leaf nodes will be represented by directories.
1523  *
1524  * sysctl(2) can automatically manage read and write requests through
1525  * the sysctl table.  The data and maxlen fields of the ctl_table
1526  * struct enable minimal validation of the values being written to be
1527  * performed, and the mode field allows minimal authentication.
1528  *
1529  * More sophisticated management can be enabled by the provision of a
1530  * strategy routine with the table entry.  This will be called before
1531  * any automatic read or write of the data is performed.
1532  *
1533  * The strategy routine may return
1534  *
1535  * < 0 - Error occurred (error is passed to user process)
1536  *
1537  * 0   - OK - proceed with automatic read or write.
1538  *
1539  * > 0 - OK - read or write has been done by the strategy routine, so
1540  *       return immediately.
1541  *
1542  * There must be a proc_handler routine for any terminal nodes
1543  * mirrored under /proc/sys (non-terminals are handled by a built-in
1544  * directory handler).  Several default handlers are available to
1545  * cover common cases -
1546  *
1547  * proc_dostring(), proc_dointvec(), proc_dointvec_jiffies(),
1548  * proc_dointvec_userhz_jiffies(), proc_dointvec_minmax(), 
1549  * proc_doulongvec_ms_jiffies_minmax(), proc_doulongvec_minmax()
1550  *
1551  * It is the handler's job to read the input buffer from user memory
1552  * and process it. The handler should return 0 on success.
1553  *
1554  * This routine returns %NULL on a failure to register, and a pointer
1555  * to the table header on success.
1556  */
1557 struct ctl_table_header *register_sysctl_table(ctl_table * table)
1558 {
1559         struct ctl_table_header *tmp;
1560         tmp = kmalloc(sizeof(struct ctl_table_header), GFP_KERNEL);
1561         if (!tmp)
1562                 return NULL;
1563         tmp->ctl_table = table;
1564         INIT_LIST_HEAD(&tmp->ctl_entry);
1565         tmp->used = 0;
1566         tmp->unregistering = NULL;
1567         sysctl_set_parent(NULL, table);
1568         spin_lock(&sysctl_lock);
1569         list_add_tail(&tmp->ctl_entry, &root_table_header.ctl_entry);
1570         spin_unlock(&sysctl_lock);
1571         return tmp;
1572 }
1573
1574 /**
1575  * unregister_sysctl_table - unregister a sysctl table hierarchy
1576  * @header: the header returned from register_sysctl_table
1577  *
1578  * Unregisters the sysctl table and all children. proc entries may not
1579  * actually be removed until they are no longer used by anyone.
1580  */
1581 void unregister_sysctl_table(struct ctl_table_header * header)
1582 {
1583         might_sleep();
1584         spin_lock(&sysctl_lock);
1585         start_unregistering(header);
1586         spin_unlock(&sysctl_lock);
1587         kfree(header);
1588 }
1589
1590 #else /* !CONFIG_SYSCTL */
1591 struct ctl_table_header *register_sysctl_table(ctl_table * table)
1592 {
1593         return NULL;
1594 }
1595
1596 void unregister_sysctl_table(struct ctl_table_header * table)
1597 {
1598 }
1599
1600 #endif /* CONFIG_SYSCTL */
1601
1602 /*
1603  * /proc/sys support
1604  */
1605
1606 #ifdef CONFIG_PROC_SYSCTL
1607
1608 static int _proc_do_string(void* data, int maxlen, int write,
1609                            struct file *filp, void __user *buffer,
1610                            size_t *lenp, loff_t *ppos)
1611 {
1612         size_t len;
1613         char __user *p;
1614         char c;
1615
1616         if (!data || !maxlen || !*lenp) {
1617                 *lenp = 0;
1618                 return 0;
1619         }
1620
1621         if (write) {
1622                 len = 0;
1623                 p = buffer;
1624                 while (len < *lenp) {
1625                         if (get_user(c, p++))
1626                                 return -EFAULT;
1627                         if (c == 0 || c == '\n')
1628                                 break;
1629                         len++;
1630                 }
1631                 if (len >= maxlen)
1632                         len = maxlen-1;
1633                 if(copy_from_user(data, buffer, len))
1634                         return -EFAULT;
1635                 ((char *) data)[len] = 0;
1636                 *ppos += *lenp;
1637         } else {
1638                 len = strlen(data);
1639                 if (len > maxlen)
1640                         len = maxlen;
1641
1642                 if (*ppos > len) {
1643                         *lenp = 0;
1644                         return 0;
1645                 }
1646
1647                 data += *ppos;
1648                 len  -= *ppos;
1649
1650                 if (len > *lenp)
1651                         len = *lenp;
1652                 if (len)
1653                         if(copy_to_user(buffer, data, len))
1654                                 return -EFAULT;
1655                 if (len < *lenp) {
1656                         if(put_user('\n', ((char __user *) buffer) + len))
1657                                 return -EFAULT;
1658                         len++;
1659                 }
1660                 *lenp = len;
1661                 *ppos += len;
1662         }
1663         return 0;
1664 }
1665
1666 /**
1667  * proc_dostring - read a string sysctl
1668  * @table: the sysctl table
1669  * @write: %TRUE if this is a write to the sysctl file
1670  * @filp: the file structure
1671  * @buffer: the user buffer
1672  * @lenp: the size of the user buffer
1673  * @ppos: file position
1674  *
1675  * Reads/writes a string from/to the user buffer. If the kernel
1676  * buffer provided is not large enough to hold the string, the
1677  * string is truncated. The copied string is %NULL-terminated.
1678  * If the string is being read by the user process, it is copied
1679  * and a newline '\n' is added. It is truncated if the buffer is
1680  * not large enough.
1681  *
1682  * Returns 0 on success.
1683  */
1684 int proc_dostring(ctl_table *table, int write, struct file *filp,
1685                   void __user *buffer, size_t *lenp, loff_t *ppos)
1686 {
1687         return _proc_do_string(table->data, table->maxlen, write, filp,
1688                                buffer, lenp, ppos);
1689 }
1690
1691
1692 static int do_proc_dointvec_conv(int *negp, unsigned long *lvalp,
1693                                  int *valp,
1694                                  int write, void *data)
1695 {
1696         if (write) {
1697                 *valp = *negp ? -*lvalp : *lvalp;
1698         } else {
1699                 int val = *valp;
1700                 if (val < 0) {
1701                         *negp = -1;
1702                         *lvalp = (unsigned long)-val;
1703                 } else {
1704                         *negp = 0;
1705                         *lvalp = (unsigned long)val;
1706                 }
1707         }
1708         return 0;
1709 }
1710
1711 static int __do_proc_dointvec(void *tbl_data, ctl_table *table,
1712                   int write, struct file *filp, void __user *buffer,
1713                   size_t *lenp, loff_t *ppos,
1714                   int (*conv)(int *negp, unsigned long *lvalp, int *valp,
1715                               int write, void *data),
1716                   void *data)
1717 {
1718 #define TMPBUFLEN 21
1719         int *i, vleft, first=1, neg, val;
1720         unsigned long lval;
1721         size_t left, len;
1722         
1723         char buf[TMPBUFLEN], *p;
1724         char __user *s = buffer;
1725         
1726         if (!tbl_data || !table->maxlen || !*lenp ||
1727             (*ppos && !write)) {
1728                 *lenp = 0;
1729                 return 0;
1730         }
1731         
1732         i = (int *) tbl_data;
1733         vleft = table->maxlen / sizeof(*i);
1734         left = *lenp;
1735
1736         if (!conv)
1737                 conv = do_proc_dointvec_conv;
1738
1739         for (; left && vleft--; i++, first=0) {
1740                 if (write) {
1741                         while (left) {
1742                                 char c;
1743                                 if (get_user(c, s))
1744                                         return -EFAULT;
1745                                 if (!isspace(c))
1746                                         break;
1747                                 left--;
1748                                 s++;
1749                         }
1750                         if (!left)
1751                                 break;
1752                         neg = 0;
1753                         len = left;
1754                         if (len > sizeof(buf) - 1)
1755                                 len = sizeof(buf) - 1;
1756                         if (copy_from_user(buf, s, len))
1757                                 return -EFAULT;
1758                         buf[len] = 0;
1759                         p = buf;
1760                         if (*p == '-' && left > 1) {
1761                                 neg = 1;
1762                                 p++;
1763                         }
1764                         if (*p < '0' || *p > '9')
1765                                 break;
1766
1767                         lval = simple_strtoul(p, &p, 0);
1768
1769                         len = p-buf;
1770                         if ((len < left) && *p && !isspace(*p))
1771                                 break;
1772                         if (neg)
1773                                 val = -val;
1774                         s += len;
1775                         left -= len;
1776
1777                         if (conv(&neg, &lval, i, 1, data))
1778                                 break;
1779                 } else {
1780                         p = buf;
1781                         if (!first)
1782                                 *p++ = '\t';
1783         
1784                         if (conv(&neg, &lval, i, 0, data))
1785                                 break;
1786
1787                         sprintf(p, "%s%lu", neg ? "-" : "", lval);
1788                         len = strlen(buf);
1789                         if (len > left)
1790                                 len = left;
1791                         if(copy_to_user(s, buf, len))
1792                                 return -EFAULT;
1793                         left -= len;
1794                         s += len;
1795                 }
1796         }
1797
1798         if (!write && !first && left) {
1799                 if(put_user('\n', s))
1800                         return -EFAULT;
1801                 left--, s++;
1802         }
1803         if (write) {
1804                 while (left) {
1805                         char c;
1806                         if (get_user(c, s++))
1807                                 return -EFAULT;
1808                         if (!isspace(c))
1809                                 break;
1810                         left--;
1811                 }
1812         }
1813         if (write && first)
1814                 return -EINVAL;
1815         *lenp -= left;
1816         *ppos += *lenp;
1817         return 0;
1818 #undef TMPBUFLEN
1819 }
1820
1821 static int do_proc_dointvec(ctl_table *table, int write, struct file *filp,
1822                   void __user *buffer, size_t *lenp, loff_t *ppos,
1823                   int (*conv)(int *negp, unsigned long *lvalp, int *valp,
1824                               int write, void *data),
1825                   void *data)
1826 {
1827         return __do_proc_dointvec(table->data, table, write, filp,
1828                         buffer, lenp, ppos, conv, data);
1829 }
1830
1831 /**
1832  * proc_dointvec - read a vector of integers
1833  * @table: the sysctl table
1834  * @write: %TRUE if this is a write to the sysctl file
1835  * @filp: the file structure
1836  * @buffer: the user buffer
1837  * @lenp: the size of the user buffer
1838  * @ppos: file position
1839  *
1840  * Reads/writes up to table->maxlen/sizeof(unsigned int) integer
1841  * values from/to the user buffer, treated as an ASCII string. 
1842  *
1843  * Returns 0 on success.
1844  */
1845 int proc_dointvec(ctl_table *table, int write, struct file *filp,
1846                      void __user *buffer, size_t *lenp, loff_t *ppos)
1847 {
1848     return do_proc_dointvec(table,write,filp,buffer,lenp,ppos,
1849                             NULL,NULL);
1850 }
1851
1852 #define OP_SET  0
1853 #define OP_AND  1
1854 #define OP_OR   2
1855
1856 static int do_proc_dointvec_bset_conv(int *negp, unsigned long *lvalp,
1857                                       int *valp,
1858                                       int write, void *data)
1859 {
1860         int op = *(int *)data;
1861         if (write) {
1862                 int val = *negp ? -*lvalp : *lvalp;
1863                 switch(op) {
1864                 case OP_SET:    *valp = val; break;
1865                 case OP_AND:    *valp &= val; break;
1866                 case OP_OR:     *valp |= val; break;
1867                 }
1868         } else {
1869                 int val = *valp;
1870                 if (val < 0) {
1871                         *negp = -1;
1872                         *lvalp = (unsigned long)-val;
1873                 } else {
1874                         *negp = 0;
1875                         *lvalp = (unsigned long)val;
1876                 }
1877         }
1878         return 0;
1879 }
1880
1881 /*
1882  *      init may raise the set.
1883  */
1884  
1885 int proc_dointvec_bset(ctl_table *table, int write, struct file *filp,
1886                         void __user *buffer, size_t *lenp, loff_t *ppos)
1887 {
1888         int op;
1889
1890         if (write && !capable(CAP_SYS_MODULE)) {
1891                 return -EPERM;
1892         }
1893
1894         op = is_init(current) ? OP_SET : OP_AND;
1895         return do_proc_dointvec(table,write,filp,buffer,lenp,ppos,
1896                                 do_proc_dointvec_bset_conv,&op);
1897 }
1898
1899 /*
1900  *      Taint values can only be increased
1901  */
1902 static int proc_dointvec_taint(ctl_table *table, int write, struct file *filp,
1903                                void __user *buffer, size_t *lenp, loff_t *ppos)
1904 {
1905         int op;
1906
1907         if (write && !capable(CAP_SYS_ADMIN))
1908                 return -EPERM;
1909
1910         op = OP_OR;
1911         return do_proc_dointvec(table,write,filp,buffer,lenp,ppos,
1912                                 do_proc_dointvec_bset_conv,&op);
1913 }
1914
1915 struct do_proc_dointvec_minmax_conv_param {
1916         int *min;
1917         int *max;
1918 };
1919
1920 static int do_proc_dointvec_minmax_conv(int *negp, unsigned long *lvalp, 
1921                                         int *valp, 
1922                                         int write, void *data)
1923 {
1924         struct do_proc_dointvec_minmax_conv_param *param = data;
1925         if (write) {
1926                 int val = *negp ? -*lvalp : *lvalp;
1927                 if ((param->min && *param->min > val) ||
1928                     (param->max && *param->max < val))
1929                         return -EINVAL;
1930                 *valp = val;
1931         } else {
1932                 int val = *valp;
1933                 if (val < 0) {
1934                         *negp = -1;
1935                         *lvalp = (unsigned long)-val;
1936                 } else {
1937                         *negp = 0;
1938                         *lvalp = (unsigned long)val;
1939                 }
1940         }
1941         return 0;
1942 }
1943
1944 /**
1945  * proc_dointvec_minmax - read a vector of integers with min/max values
1946  * @table: the sysctl table
1947  * @write: %TRUE if this is a write to the sysctl file
1948  * @filp: the file structure
1949  * @buffer: the user buffer
1950  * @lenp: the size of the user buffer
1951  * @ppos: file position
1952  *
1953  * Reads/writes up to table->maxlen/sizeof(unsigned int) integer
1954  * values from/to the user buffer, treated as an ASCII string.
1955  *
1956  * This routine will ensure the values are within the range specified by
1957  * table->extra1 (min) and table->extra2 (max).
1958  *
1959  * Returns 0 on success.
1960  */
1961 int proc_dointvec_minmax(ctl_table *table, int write, struct file *filp,
1962                   void __user *buffer, size_t *lenp, loff_t *ppos)
1963 {
1964         struct do_proc_dointvec_minmax_conv_param param = {
1965                 .min = (int *) table->extra1,
1966                 .max = (int *) table->extra2,
1967         };
1968         return do_proc_dointvec(table, write, filp, buffer, lenp, ppos,
1969                                 do_proc_dointvec_minmax_conv, &param);
1970 }
1971
1972 static int __do_proc_doulongvec_minmax(void *data, ctl_table *table, int write,
1973                                      struct file *filp,
1974                                      void __user *buffer,
1975                                      size_t *lenp, loff_t *ppos,
1976                                      unsigned long convmul,
1977                                      unsigned long convdiv)
1978 {
1979 #define TMPBUFLEN 21
1980         unsigned long *i, *min, *max, val;
1981         int vleft, first=1, neg;
1982         size_t len, left;
1983         char buf[TMPBUFLEN], *p;
1984         char __user *s = buffer;
1985         
1986         if (!data || !table->maxlen || !*lenp ||
1987             (*ppos && !write)) {
1988                 *lenp = 0;
1989                 return 0;
1990         }
1991         
1992         i = (unsigned long *) data;
1993         min = (unsigned long *) table->extra1;
1994         max = (unsigned long *) table->extra2;
1995         vleft = table->maxlen / sizeof(unsigned long);
1996         left = *lenp;
1997         
1998         for (; left && vleft--; i++, min++, max++, first=0) {
1999                 if (write) {
2000                         while (left) {
2001                                 char c;
2002                                 if (get_user(c, s))
2003                                         return -EFAULT;
2004                                 if (!isspace(c))
2005                                         break;
2006                                 left--;
2007                                 s++;
2008                         }
2009                         if (!left)
2010                                 break;
2011                         neg = 0;
2012                         len = left;
2013                         if (len > TMPBUFLEN-1)
2014                                 len = TMPBUFLEN-1;
2015                         if (copy_from_user(buf, s, len))
2016                                 return -EFAULT;
2017                         buf[len] = 0;
2018                         p = buf;
2019                         if (*p == '-' && left > 1) {
2020                                 neg = 1;
2021                                 p++;
2022                         }
2023                         if (*p < '0' || *p > '9')
2024                                 break;
2025                         val = simple_strtoul(p, &p, 0) * convmul / convdiv ;
2026                         len = p-buf;
2027                         if ((len < left) && *p && !isspace(*p))
2028                                 break;
2029                         if (neg)
2030                                 val = -val;
2031                         s += len;
2032                         left -= len;
2033
2034                         if(neg)
2035                                 continue;
2036                         if ((min && val < *min) || (max && val > *max))
2037                                 continue;
2038                         *i = val;
2039                 } else {
2040                         p = buf;
2041                         if (!first)
2042                                 *p++ = '\t';
2043                         sprintf(p, "%lu", convdiv * (*i) / convmul);
2044                         len = strlen(buf);
2045                         if (len > left)
2046                                 len = left;
2047                         if(copy_to_user(s, buf, len))
2048                                 return -EFAULT;
2049                         left -= len;
2050                         s += len;
2051                 }
2052         }
2053
2054         if (!write && !first && left) {
2055                 if(put_user('\n', s))
2056                         return -EFAULT;
2057                 left--, s++;
2058         }
2059         if (write) {
2060                 while (left) {
2061                         char c;
2062                         if (get_user(c, s++))
2063                                 return -EFAULT;
2064                         if (!isspace(c))
2065                                 break;
2066                         left--;
2067                 }
2068         }
2069         if (write && first)
2070                 return -EINVAL;
2071         *lenp -= left;
2072         *ppos += *lenp;
2073         return 0;
2074 #undef TMPBUFLEN
2075 }
2076
2077 static int do_proc_doulongvec_minmax(ctl_table *table, int write,
2078                                      struct file *filp,
2079                                      void __user *buffer,
2080                                      size_t *lenp, loff_t *ppos,
2081                                      unsigned long convmul,
2082                                      unsigned long convdiv)
2083 {
2084         return __do_proc_doulongvec_minmax(table->data, table, write,
2085                         filp, buffer, lenp, ppos, convmul, convdiv);
2086 }
2087
2088 /**
2089  * proc_doulongvec_minmax - read a vector of long integers with min/max values
2090  * @table: the sysctl table
2091  * @write: %TRUE if this is a write to the sysctl file
2092  * @filp: the file structure
2093  * @buffer: the user buffer
2094  * @lenp: the size of the user buffer
2095  * @ppos: file position
2096  *
2097  * Reads/writes up to table->maxlen/sizeof(unsigned long) unsigned long
2098  * values from/to the user buffer, treated as an ASCII string.
2099  *
2100  * This routine will ensure the values are within the range specified by
2101  * table->extra1 (min) and table->extra2 (max).
2102  *
2103  * Returns 0 on success.
2104  */
2105 int proc_doulongvec_minmax(ctl_table *table, int write, struct file *filp,
2106                            void __user *buffer, size_t *lenp, loff_t *ppos)
2107 {
2108     return do_proc_doulongvec_minmax(table, write, filp, buffer, lenp, ppos, 1l, 1l);
2109 }
2110
2111 /**
2112  * proc_doulongvec_ms_jiffies_minmax - read a vector of millisecond values with min/max values
2113  * @table: the sysctl table
2114  * @write: %TRUE if this is a write to the sysctl file
2115  * @filp: the file structure
2116  * @buffer: the user buffer
2117  * @lenp: the size of the user buffer
2118  * @ppos: file position
2119  *
2120  * Reads/writes up to table->maxlen/sizeof(unsigned long) unsigned long
2121  * values from/to the user buffer, treated as an ASCII string. The values
2122  * are treated as milliseconds, and converted to jiffies when they are stored.
2123  *
2124  * This routine will ensure the values are within the range specified by
2125  * table->extra1 (min) and table->extra2 (max).
2126  *
2127  * Returns 0 on success.
2128  */
2129 int proc_doulongvec_ms_jiffies_minmax(ctl_table *table, int write,
2130                                       struct file *filp,
2131                                       void __user *buffer,
2132                                       size_t *lenp, loff_t *ppos)
2133 {
2134     return do_proc_doulongvec_minmax(table, write, filp, buffer,
2135                                      lenp, ppos, HZ, 1000l);
2136 }
2137
2138
2139 static int do_proc_dointvec_jiffies_conv(int *negp, unsigned long *lvalp,
2140                                          int *valp,
2141                                          int write, void *data)
2142 {
2143         if (write) {
2144                 if (*lvalp > LONG_MAX / HZ)
2145                         return 1;
2146                 *valp = *negp ? -(*lvalp*HZ) : (*lvalp*HZ);
2147         } else {
2148                 int val = *valp;
2149                 unsigned long lval;
2150                 if (val < 0) {
2151                         *negp = -1;
2152                         lval = (unsigned long)-val;
2153                 } else {
2154                         *negp = 0;
2155                         lval = (unsigned long)val;
2156                 }
2157                 *lvalp = lval / HZ;
2158         }
2159         return 0;
2160 }
2161
2162 static int do_proc_dointvec_userhz_jiffies_conv(int *negp, unsigned long *lvalp,
2163                                                 int *valp,
2164                                                 int write, void *data)
2165 {
2166         if (write) {
2167                 if (USER_HZ < HZ && *lvalp > (LONG_MAX / HZ) * USER_HZ)
2168                         return 1;
2169                 *valp = clock_t_to_jiffies(*negp ? -*lvalp : *lvalp);
2170         } else {
2171                 int val = *valp;
2172                 unsigned long lval;
2173                 if (val < 0) {
2174                         *negp = -1;
2175                         lval = (unsigned long)-val;
2176                 } else {
2177                         *negp = 0;
2178                         lval = (unsigned long)val;
2179                 }
2180                 *lvalp = jiffies_to_clock_t(lval);
2181         }
2182         return 0;
2183 }
2184
2185 static int do_proc_dointvec_ms_jiffies_conv(int *negp, unsigned long *lvalp,
2186                                             int *valp,
2187                                             int write, void *data)
2188 {
2189         if (write) {
2190                 *valp = msecs_to_jiffies(*negp ? -*lvalp : *lvalp);
2191         } else {
2192                 int val = *valp;
2193                 unsigned long lval;
2194                 if (val < 0) {
2195                         *negp = -1;
2196                         lval = (unsigned long)-val;
2197                 } else {
2198                         *negp = 0;
2199                         lval = (unsigned long)val;
2200                 }
2201                 *lvalp = jiffies_to_msecs(lval);
2202         }
2203         return 0;
2204 }
2205
2206 /**
2207  * proc_dointvec_jiffies - read a vector of integers as seconds
2208  * @table: the sysctl table
2209  * @write: %TRUE if this is a write to the sysctl file
2210  * @filp: the file structure
2211  * @buffer: the user buffer
2212  * @lenp: the size of the user buffer
2213  * @ppos: file position
2214  *
2215  * Reads/writes up to table->maxlen/sizeof(unsigned int) integer
2216  * values from/to the user buffer, treated as an ASCII string. 
2217  * The values read are assumed to be in seconds, and are converted into
2218  * jiffies.
2219  *
2220  * Returns 0 on success.
2221  */
2222 int proc_dointvec_jiffies(ctl_table *table, int write, struct file *filp,
2223                           void __user *buffer, size_t *lenp, loff_t *ppos)
2224 {
2225     return do_proc_dointvec(table,write,filp,buffer,lenp,ppos,
2226                             do_proc_dointvec_jiffies_conv,NULL);
2227 }
2228
2229 /**
2230  * proc_dointvec_userhz_jiffies - read a vector of integers as 1/USER_HZ seconds
2231  * @table: the sysctl table
2232  * @write: %TRUE if this is a write to the sysctl file
2233  * @filp: the file structure
2234  * @buffer: the user buffer
2235  * @lenp: the size of the user buffer
2236  * @ppos: pointer to the file position
2237  *
2238  * Reads/writes up to table->maxlen/sizeof(unsigned int) integer
2239  * values from/to the user buffer, treated as an ASCII string. 
2240  * The values read are assumed to be in 1/USER_HZ seconds, and 
2241  * are converted into jiffies.
2242  *
2243  * Returns 0 on success.
2244  */
2245 int proc_dointvec_userhz_jiffies(ctl_table *table, int write, struct file *filp,
2246                                  void __user *buffer, size_t *lenp, loff_t *ppos)
2247 {
2248     return do_proc_dointvec(table,write,filp,buffer,lenp,ppos,
2249                             do_proc_dointvec_userhz_jiffies_conv,NULL);
2250 }
2251
2252 /**
2253  * proc_dointvec_ms_jiffies - read a vector of integers as 1 milliseconds
2254  * @table: the sysctl table
2255  * @write: %TRUE if this is a write to the sysctl file
2256  * @filp: the file structure
2257  * @buffer: the user buffer
2258  * @lenp: the size of the user buffer
2259  * @ppos: file position
2260  * @ppos: the current position in the file
2261  *
2262  * Reads/writes up to table->maxlen/sizeof(unsigned int) integer
2263  * values from/to the user buffer, treated as an ASCII string. 
2264  * The values read are assumed to be in 1/1000 seconds, and 
2265  * are converted into jiffies.
2266  *
2267  * Returns 0 on success.
2268  */
2269 int proc_dointvec_ms_jiffies(ctl_table *table, int write, struct file *filp,
2270                              void __user *buffer, size_t *lenp, loff_t *ppos)
2271 {
2272         return do_proc_dointvec(table, write, filp, buffer, lenp, ppos,
2273                                 do_proc_dointvec_ms_jiffies_conv, NULL);
2274 }
2275
2276 static int proc_do_cad_pid(ctl_table *table, int write, struct file *filp,
2277                            void __user *buffer, size_t *lenp, loff_t *ppos)
2278 {
2279         struct pid *new_pid;
2280         pid_t tmp;
2281         int r;
2282
2283         tmp = pid_nr(cad_pid);
2284
2285         r = __do_proc_dointvec(&tmp, table, write, filp, buffer,
2286                                lenp, ppos, NULL, NULL);
2287         if (r || !write)
2288                 return r;
2289
2290         new_pid = find_get_pid(tmp);
2291         if (!new_pid)
2292                 return -ESRCH;
2293
2294         put_pid(xchg(&cad_pid, new_pid));
2295         return 0;
2296 }
2297
2298 #else /* CONFIG_PROC_FS */
2299
2300 int proc_dostring(ctl_table *table, int write, struct file *filp,
2301                   void __user *buffer, size_t *lenp, loff_t *ppos)
2302 {
2303         return -ENOSYS;
2304 }
2305
2306 int proc_dointvec(ctl_table *table, int write, struct file *filp,
2307                   void __user *buffer, size_t *lenp, loff_t *ppos)
2308 {
2309         return -ENOSYS;
2310 }
2311
2312 int proc_dointvec_bset(ctl_table *table, int write, struct file *filp,
2313                         void __user *buffer, size_t *lenp, loff_t *ppos)
2314 {
2315         return -ENOSYS;
2316 }
2317
2318 int proc_dointvec_minmax(ctl_table *table, int write, struct file *filp,
2319                     void __user *buffer, size_t *lenp, loff_t *ppos)
2320 {
2321         return -ENOSYS;
2322 }
2323
2324 int proc_dointvec_jiffies(ctl_table *table, int write, struct file *filp,
2325                     void __user *buffer, size_t *lenp, loff_t *ppos)
2326 {
2327         return -ENOSYS;
2328 }
2329
2330 int proc_dointvec_userhz_jiffies(ctl_table *table, int write, struct file *filp,
2331                     void __user *buffer, size_t *lenp, loff_t *ppos)
2332 {
2333         return -ENOSYS;
2334 }
2335
2336 int proc_dointvec_ms_jiffies(ctl_table *table, int write, struct file *filp,
2337                              void __user *buffer, size_t *lenp, loff_t *ppos)
2338 {
2339         return -ENOSYS;
2340 }
2341
2342 int proc_doulongvec_minmax(ctl_table *table, int write, struct file *filp,
2343                     void __user *buffer, size_t *lenp, loff_t *ppos)
2344 {
2345         return -ENOSYS;
2346 }
2347
2348 int proc_doulongvec_ms_jiffies_minmax(ctl_table *table, int write,
2349                                       struct file *filp,
2350                                       void __user *buffer,
2351                                       size_t *lenp, loff_t *ppos)
2352 {
2353     return -ENOSYS;
2354 }
2355
2356
2357 #endif /* CONFIG_PROC_FS */
2358
2359
2360 #ifdef CONFIG_SYSCTL_SYSCALL
2361 /*
2362  * General sysctl support routines 
2363  */
2364
2365 /* The generic string strategy routine: */
2366 int sysctl_string(ctl_table *table, int __user *name, int nlen,
2367                   void __user *oldval, size_t __user *oldlenp,
2368                   void __user *newval, size_t newlen)
2369 {
2370         if (!table->data || !table->maxlen) 
2371                 return -ENOTDIR;
2372         
2373         if (oldval && oldlenp) {
2374                 size_t bufsize;
2375                 if (get_user(bufsize, oldlenp))
2376                         return -EFAULT;
2377                 if (bufsize) {
2378                         size_t len = strlen(table->data), copied;
2379
2380                         /* This shouldn't trigger for a well-formed sysctl */
2381                         if (len > table->maxlen)
2382                                 len = table->maxlen;
2383
2384                         /* Copy up to a max of bufsize-1 bytes of the string */
2385                         copied = (len >= bufsize) ? bufsize - 1 : len;
2386
2387                         if (copy_to_user(oldval, table->data, copied) ||
2388                             put_user(0, (char __user *)(oldval + copied)))
2389                                 return -EFAULT;
2390                         if (put_user(len, oldlenp))
2391                                 return -EFAULT;
2392                 }
2393         }
2394         if (newval && newlen) {
2395                 size_t len = newlen;
2396                 if (len > table->maxlen)
2397                         len = table->maxlen;
2398                 if(copy_from_user(table->data, newval, len))
2399                         return -EFAULT;
2400                 if (len == table->maxlen)
2401                         len--;
2402                 ((char *) table->data)[len] = 0;
2403         }
2404         return 1;
2405 }
2406
2407 /*
2408  * This function makes sure that all of the integers in the vector
2409  * are between the minimum and maximum values given in the arrays
2410  * table->extra1 and table->extra2, respectively.
2411  */
2412 int sysctl_intvec(ctl_table *table, int __user *name, int nlen,
2413                 void __user *oldval, size_t __user *oldlenp,
2414                 void __user *newval, size_t newlen)
2415 {
2416
2417         if (newval && newlen) {
2418                 int __user *vec = (int __user *) newval;
2419                 int *min = (int *) table->extra1;
2420                 int *max = (int *) table->extra2;
2421                 size_t length;
2422                 int i;
2423
2424                 if (newlen % sizeof(int) != 0)
2425                         return -EINVAL;
2426
2427                 if (!table->extra1 && !table->extra2)
2428                         return 0;
2429
2430                 if (newlen > table->maxlen)
2431                         newlen = table->maxlen;
2432                 length = newlen / sizeof(int);
2433
2434                 for (i = 0; i < length; i++) {
2435                         int value;
2436                         if (get_user(value, vec + i))
2437                                 return -EFAULT;
2438                         if (min && value < min[i])
2439                                 return -EINVAL;
2440                         if (max && value > max[i])
2441                                 return -EINVAL;
2442                 }
2443         }
2444         return 0;
2445 }
2446
2447 /* Strategy function to convert jiffies to seconds */ 
2448 int sysctl_jiffies(ctl_table *table, int __user *name, int nlen,
2449                 void __user *oldval, size_t __user *oldlenp,
2450                 void __user *newval, size_t newlen)
2451 {
2452         if (oldval && oldlenp) {
2453                 size_t olen;
2454
2455                 if (get_user(olen, oldlenp))
2456                         return -EFAULT;
2457                 if (olen) {
2458                         int val;
2459
2460                         if (olen < sizeof(int))
2461                                 return -EINVAL;
2462
2463                         val = *(int *)(table->data) / HZ;
2464                         if (put_user(val, (int __user *)oldval))
2465                                 return -EFAULT;
2466                         if (put_user(sizeof(int), oldlenp))
2467                                 return -EFAULT;
2468                 }
2469         }
2470         if (newval && newlen) { 
2471                 int new;
2472                 if (newlen != sizeof(int))
2473                         return -EINVAL; 
2474                 if (get_user(new, (int __user *)newval))
2475                         return -EFAULT;
2476                 *(int *)(table->data) = new*HZ; 
2477         }
2478         return 1;
2479 }
2480
2481 /* Strategy function to convert jiffies to seconds */ 
2482 int sysctl_ms_jiffies(ctl_table *table, int __user *name, int nlen,
2483                 void __user *oldval, size_t __user *oldlenp,
2484                 void __user *newval, size_t newlen)
2485 {
2486         if (oldval && oldlenp) {
2487                 size_t olen;
2488
2489                 if (get_user(olen, oldlenp))
2490                         return -EFAULT;
2491                 if (olen) {
2492                         int val;
2493
2494                         if (olen < sizeof(int))
2495                                 return -EINVAL;
2496
2497                         val = jiffies_to_msecs(*(int *)(table->data));
2498                         if (put_user(val, (int __user *)oldval))
2499                                 return -EFAULT;
2500                         if (put_user(sizeof(int), oldlenp))
2501                                 return -EFAULT;
2502                 }
2503         }
2504         if (newval && newlen) { 
2505                 int new;
2506                 if (newlen != sizeof(int))
2507                         return -EINVAL; 
2508                 if (get_user(new, (int __user *)newval))
2509                         return -EFAULT;
2510                 *(int *)(table->data) = msecs_to_jiffies(new);
2511         }
2512         return 1;
2513 }
2514
2515
2516
2517 #else /* CONFIG_SYSCTL_SYSCALL */
2518
2519
2520 asmlinkage long sys_sysctl(struct __sysctl_args __user *args)
2521 {
2522         static int msg_count;
2523         struct __sysctl_args tmp;
2524         int name[CTL_MAXNAME];
2525         int i;
2526
2527         /* Read in the sysctl name for better debug message logging */
2528         if (copy_from_user(&tmp, args, sizeof(tmp)))
2529                 return -EFAULT;
2530         if (tmp.nlen <= 0 || tmp.nlen >= CTL_MAXNAME)
2531                 return -ENOTDIR;
2532         for (i = 0; i < tmp.nlen; i++)
2533                 if (get_user(name[i], tmp.name + i))
2534                         return -EFAULT;
2535
2536         /* Ignore accesses to kernel.version */
2537         if ((tmp.nlen == 2) && (name[0] == CTL_KERN) && (name[1] == KERN_VERSION))
2538                 goto out;
2539
2540         if (msg_count < 5) {
2541                 msg_count++;
2542                 printk(KERN_INFO
2543                         "warning: process `%s' used the removed sysctl "
2544                         "system call with ", current->comm);
2545                 for (i = 0; i < tmp.nlen; i++)
2546                         printk("%d.", name[i]);
2547                 printk("\n");
2548         }
2549 out:
2550         return -ENOSYS;
2551 }
2552
2553 int sysctl_string(ctl_table *table, int __user *name, int nlen,
2554                   void __user *oldval, size_t __user *oldlenp,
2555                   void __user *newval, size_t newlen)
2556 {
2557         return -ENOSYS;
2558 }
2559
2560 int sysctl_intvec(ctl_table *table, int __user *name, int nlen,
2561                 void __user *oldval, size_t __user *oldlenp,
2562                 void __user *newval, size_t newlen)
2563 {
2564         return -ENOSYS;
2565 }
2566
2567 int sysctl_jiffies(ctl_table *table, int __user *name, int nlen,
2568                 void __user *oldval, size_t __user *oldlenp,
2569                 void __user *newval, size_t newlen)
2570 {
2571         return -ENOSYS;
2572 }
2573
2574 int sysctl_ms_jiffies(ctl_table *table, int __user *name, int nlen,
2575                 void __user *oldval, size_t __user *oldlenp,
2576                 void __user *newval, size_t newlen)
2577 {
2578         return -ENOSYS;
2579 }
2580
2581 #endif /* CONFIG_SYSCTL_SYSCALL */
2582
2583 /*
2584  * No sense putting this after each symbol definition, twice,
2585  * exception granted :-)
2586  */
2587 EXPORT_SYMBOL(proc_dointvec);
2588 EXPORT_SYMBOL(proc_dointvec_jiffies);
2589 EXPORT_SYMBOL(proc_dointvec_minmax);
2590 EXPORT_SYMBOL(proc_dointvec_userhz_jiffies);
2591 EXPORT_SYMBOL(proc_dointvec_ms_jiffies);
2592 EXPORT_SYMBOL(proc_dostring);
2593 EXPORT_SYMBOL(proc_doulongvec_minmax);
2594 EXPORT_SYMBOL(proc_doulongvec_ms_jiffies_minmax);
2595 EXPORT_SYMBOL(register_sysctl_table);
2596 EXPORT_SYMBOL(sysctl_intvec);
2597 EXPORT_SYMBOL(sysctl_jiffies);
2598 EXPORT_SYMBOL(sysctl_ms_jiffies);
2599 EXPORT_SYMBOL(sysctl_string);
2600 EXPORT_SYMBOL(unregister_sysctl_table);