[IA64] sysctl option to silence unaligned trap warnings
[linux-2.6.git] / kernel / sysctl.c
1 /*
2  * sysctl.c: General linux system control interface
3  *
4  * Begun 24 March 1995, Stephen Tweedie
5  * Added /proc support, Dec 1995
6  * Added bdflush entry and intvec min/max checking, 2/23/96, Tom Dyas.
7  * Added hooks for /proc/sys/net (minor, minor patch), 96/4/1, Mike Shaver.
8  * Added kernel/java-{interpreter,appletviewer}, 96/5/10, Mike Shaver.
9  * Dynamic registration fixes, Stephen Tweedie.
10  * Added kswapd-interval, ctrl-alt-del, printk stuff, 1/8/97, Chris Horn.
11  * Made sysctl support optional via CONFIG_SYSCTL, 1/10/97, Chris
12  *  Horn.
13  * Added proc_doulongvec_ms_jiffies_minmax, 09/08/99, Carlos H. Bauer.
14  * Added proc_doulongvec_minmax, 09/08/99, Carlos H. Bauer.
15  * Changed linked lists to use list.h instead of lists.h, 02/24/00, Bill
16  *  Wendling.
17  * The list_for_each() macro wasn't appropriate for the sysctl loop.
18  *  Removed it and replaced it with older style, 03/23/00, Bill Wendling
19  */
20
21 #include <linux/config.h>
22 #include <linux/module.h>
23 #include <linux/mm.h>
24 #include <linux/swap.h>
25 #include <linux/slab.h>
26 #include <linux/sysctl.h>
27 #include <linux/proc_fs.h>
28 #include <linux/capability.h>
29 #include <linux/ctype.h>
30 #include <linux/utsname.h>
31 #include <linux/capability.h>
32 #include <linux/smp_lock.h>
33 #include <linux/init.h>
34 #include <linux/kernel.h>
35 #include <linux/kobject.h>
36 #include <linux/net.h>
37 #include <linux/sysrq.h>
38 #include <linux/highuid.h>
39 #include <linux/writeback.h>
40 #include <linux/hugetlb.h>
41 #include <linux/security.h>
42 #include <linux/initrd.h>
43 #include <linux/times.h>
44 #include <linux/limits.h>
45 #include <linux/dcache.h>
46 #include <linux/syscalls.h>
47 #include <linux/nfs_fs.h>
48 #include <linux/acpi.h>
49
50 #include <asm/uaccess.h>
51 #include <asm/processor.h>
52
53 #if defined(CONFIG_SYSCTL)
54
55 /* External variables not in a header file. */
56 extern int C_A_D;
57 extern int sysctl_overcommit_memory;
58 extern int sysctl_overcommit_ratio;
59 extern int max_threads;
60 extern int sysrq_enabled;
61 extern int core_uses_pid;
62 extern int suid_dumpable;
63 extern char core_pattern[];
64 extern int cad_pid;
65 extern int pid_max;
66 extern int min_free_kbytes;
67 extern int printk_ratelimit_jiffies;
68 extern int printk_ratelimit_burst;
69 extern int pid_max_min, pid_max_max;
70 extern int sysctl_drop_caches;
71 extern int percpu_pagelist_fraction;
72
73 #if defined(CONFIG_X86_LOCAL_APIC) && defined(CONFIG_X86)
74 int unknown_nmi_panic;
75 extern int proc_unknown_nmi_panic(ctl_table *, int, struct file *,
76                                   void __user *, size_t *, loff_t *);
77 #endif
78
79 /* this is needed for the proc_dointvec_minmax for [fs_]overflow UID and GID */
80 static int maxolduid = 65535;
81 static int minolduid;
82 static int min_percpu_pagelist_fract = 8;
83
84 static int ngroups_max = NGROUPS_MAX;
85
86 #ifdef CONFIG_KMOD
87 extern char modprobe_path[];
88 #endif
89 #ifdef CONFIG_CHR_DEV_SG
90 extern int sg_big_buff;
91 #endif
92 #ifdef CONFIG_SYSVIPC
93 extern size_t shm_ctlmax;
94 extern size_t shm_ctlall;
95 extern int shm_ctlmni;
96 extern int msg_ctlmax;
97 extern int msg_ctlmnb;
98 extern int msg_ctlmni;
99 extern int sem_ctls[];
100 #endif
101
102 #ifdef __sparc__
103 extern char reboot_command [];
104 extern int stop_a_enabled;
105 extern int scons_pwroff;
106 #endif
107
108 #ifdef __hppa__
109 extern int pwrsw_enabled;
110 extern int unaligned_enabled;
111 #endif
112
113 #ifdef CONFIG_S390
114 #ifdef CONFIG_MATHEMU
115 extern int sysctl_ieee_emulation_warnings;
116 #endif
117 extern int sysctl_userprocess_debug;
118 extern int spin_retry;
119 #endif
120
121 extern int sysctl_hz_timer;
122
123 #ifdef CONFIG_BSD_PROCESS_ACCT
124 extern int acct_parm[];
125 #endif
126
127 #ifdef CONFIG_IA64
128 extern int no_unaligned_warning;
129 #endif
130
131 static int parse_table(int __user *, int, void __user *, size_t __user *, void __user *, size_t,
132                        ctl_table *, void **);
133 static int proc_doutsstring(ctl_table *table, int write, struct file *filp,
134                   void __user *buffer, size_t *lenp, loff_t *ppos);
135
136 static ctl_table root_table[];
137 static struct ctl_table_header root_table_header =
138         { root_table, LIST_HEAD_INIT(root_table_header.ctl_entry) };
139
140 static ctl_table kern_table[];
141 static ctl_table vm_table[];
142 static ctl_table proc_table[];
143 static ctl_table fs_table[];
144 static ctl_table debug_table[];
145 static ctl_table dev_table[];
146 extern ctl_table random_table[];
147 #ifdef CONFIG_UNIX98_PTYS
148 extern ctl_table pty_table[];
149 #endif
150 #ifdef CONFIG_INOTIFY
151 extern ctl_table inotify_table[];
152 #endif
153
154 #ifdef HAVE_ARCH_PICK_MMAP_LAYOUT
155 int sysctl_legacy_va_layout;
156 #endif
157
158 /* /proc declarations: */
159
160 #ifdef CONFIG_PROC_FS
161
162 static ssize_t proc_readsys(struct file *, char __user *, size_t, loff_t *);
163 static ssize_t proc_writesys(struct file *, const char __user *, size_t, loff_t *);
164 static int proc_opensys(struct inode *, struct file *);
165
166 struct file_operations proc_sys_file_operations = {
167         .open           = proc_opensys,
168         .read           = proc_readsys,
169         .write          = proc_writesys,
170 };
171
172 extern struct proc_dir_entry *proc_sys_root;
173
174 static void register_proc_table(ctl_table *, struct proc_dir_entry *, void *);
175 static void unregister_proc_table(ctl_table *, struct proc_dir_entry *);
176 #endif
177
178 /* The default sysctl tables: */
179
180 static ctl_table root_table[] = {
181         {
182                 .ctl_name       = CTL_KERN,
183                 .procname       = "kernel",
184                 .mode           = 0555,
185                 .child          = kern_table,
186         },
187         {
188                 .ctl_name       = CTL_VM,
189                 .procname       = "vm",
190                 .mode           = 0555,
191                 .child          = vm_table,
192         },
193 #ifdef CONFIG_NET
194         {
195                 .ctl_name       = CTL_NET,
196                 .procname       = "net",
197                 .mode           = 0555,
198                 .child          = net_table,
199         },
200 #endif
201         {
202                 .ctl_name       = CTL_PROC,
203                 .procname       = "proc",
204                 .mode           = 0555,
205                 .child          = proc_table,
206         },
207         {
208                 .ctl_name       = CTL_FS,
209                 .procname       = "fs",
210                 .mode           = 0555,
211                 .child          = fs_table,
212         },
213         {
214                 .ctl_name       = CTL_DEBUG,
215                 .procname       = "debug",
216                 .mode           = 0555,
217                 .child          = debug_table,
218         },
219         {
220                 .ctl_name       = CTL_DEV,
221                 .procname       = "dev",
222                 .mode           = 0555,
223                 .child          = dev_table,
224         },
225
226         { .ctl_name = 0 }
227 };
228
229 static ctl_table kern_table[] = {
230         {
231                 .ctl_name       = KERN_OSTYPE,
232                 .procname       = "ostype",
233                 .data           = system_utsname.sysname,
234                 .maxlen         = sizeof(system_utsname.sysname),
235                 .mode           = 0444,
236                 .proc_handler   = &proc_doutsstring,
237                 .strategy       = &sysctl_string,
238         },
239         {
240                 .ctl_name       = KERN_OSRELEASE,
241                 .procname       = "osrelease",
242                 .data           = system_utsname.release,
243                 .maxlen         = sizeof(system_utsname.release),
244                 .mode           = 0444,
245                 .proc_handler   = &proc_doutsstring,
246                 .strategy       = &sysctl_string,
247         },
248         {
249                 .ctl_name       = KERN_VERSION,
250                 .procname       = "version",
251                 .data           = system_utsname.version,
252                 .maxlen         = sizeof(system_utsname.version),
253                 .mode           = 0444,
254                 .proc_handler   = &proc_doutsstring,
255                 .strategy       = &sysctl_string,
256         },
257         {
258                 .ctl_name       = KERN_NODENAME,
259                 .procname       = "hostname",
260                 .data           = system_utsname.nodename,
261                 .maxlen         = sizeof(system_utsname.nodename),
262                 .mode           = 0644,
263                 .proc_handler   = &proc_doutsstring,
264                 .strategy       = &sysctl_string,
265         },
266         {
267                 .ctl_name       = KERN_DOMAINNAME,
268                 .procname       = "domainname",
269                 .data           = system_utsname.domainname,
270                 .maxlen         = sizeof(system_utsname.domainname),
271                 .mode           = 0644,
272                 .proc_handler   = &proc_doutsstring,
273                 .strategy       = &sysctl_string,
274         },
275         {
276                 .ctl_name       = KERN_PANIC,
277                 .procname       = "panic",
278                 .data           = &panic_timeout,
279                 .maxlen         = sizeof(int),
280                 .mode           = 0644,
281                 .proc_handler   = &proc_dointvec,
282         },
283         {
284                 .ctl_name       = KERN_CORE_USES_PID,
285                 .procname       = "core_uses_pid",
286                 .data           = &core_uses_pid,
287                 .maxlen         = sizeof(int),
288                 .mode           = 0644,
289                 .proc_handler   = &proc_dointvec,
290         },
291         {
292                 .ctl_name       = KERN_CORE_PATTERN,
293                 .procname       = "core_pattern",
294                 .data           = core_pattern,
295                 .maxlen         = 64,
296                 .mode           = 0644,
297                 .proc_handler   = &proc_dostring,
298                 .strategy       = &sysctl_string,
299         },
300         {
301                 .ctl_name       = KERN_TAINTED,
302                 .procname       = "tainted",
303                 .data           = &tainted,
304                 .maxlen         = sizeof(int),
305                 .mode           = 0444,
306                 .proc_handler   = &proc_dointvec,
307         },
308         {
309                 .ctl_name       = KERN_CAP_BSET,
310                 .procname       = "cap-bound",
311                 .data           = &cap_bset,
312                 .maxlen         = sizeof(kernel_cap_t),
313                 .mode           = 0600,
314                 .proc_handler   = &proc_dointvec_bset,
315         },
316 #ifdef CONFIG_BLK_DEV_INITRD
317         {
318                 .ctl_name       = KERN_REALROOTDEV,
319                 .procname       = "real-root-dev",
320                 .data           = &real_root_dev,
321                 .maxlen         = sizeof(int),
322                 .mode           = 0644,
323                 .proc_handler   = &proc_dointvec,
324         },
325 #endif
326 #ifdef __sparc__
327         {
328                 .ctl_name       = KERN_SPARC_REBOOT,
329                 .procname       = "reboot-cmd",
330                 .data           = reboot_command,
331                 .maxlen         = 256,
332                 .mode           = 0644,
333                 .proc_handler   = &proc_dostring,
334                 .strategy       = &sysctl_string,
335         },
336         {
337                 .ctl_name       = KERN_SPARC_STOP_A,
338                 .procname       = "stop-a",
339                 .data           = &stop_a_enabled,
340                 .maxlen         = sizeof (int),
341                 .mode           = 0644,
342                 .proc_handler   = &proc_dointvec,
343         },
344         {
345                 .ctl_name       = KERN_SPARC_SCONS_PWROFF,
346                 .procname       = "scons-poweroff",
347                 .data           = &scons_pwroff,
348                 .maxlen         = sizeof (int),
349                 .mode           = 0644,
350                 .proc_handler   = &proc_dointvec,
351         },
352 #endif
353 #ifdef __hppa__
354         {
355                 .ctl_name       = KERN_HPPA_PWRSW,
356                 .procname       = "soft-power",
357                 .data           = &pwrsw_enabled,
358                 .maxlen         = sizeof (int),
359                 .mode           = 0644,
360                 .proc_handler   = &proc_dointvec,
361         },
362         {
363                 .ctl_name       = KERN_HPPA_UNALIGNED,
364                 .procname       = "unaligned-trap",
365                 .data           = &unaligned_enabled,
366                 .maxlen         = sizeof (int),
367                 .mode           = 0644,
368                 .proc_handler   = &proc_dointvec,
369         },
370 #endif
371         {
372                 .ctl_name       = KERN_CTLALTDEL,
373                 .procname       = "ctrl-alt-del",
374                 .data           = &C_A_D,
375                 .maxlen         = sizeof(int),
376                 .mode           = 0644,
377                 .proc_handler   = &proc_dointvec,
378         },
379         {
380                 .ctl_name       = KERN_PRINTK,
381                 .procname       = "printk",
382                 .data           = &console_loglevel,
383                 .maxlen         = 4*sizeof(int),
384                 .mode           = 0644,
385                 .proc_handler   = &proc_dointvec,
386         },
387 #ifdef CONFIG_KMOD
388         {
389                 .ctl_name       = KERN_MODPROBE,
390                 .procname       = "modprobe",
391                 .data           = &modprobe_path,
392                 .maxlen         = KMOD_PATH_LEN,
393                 .mode           = 0644,
394                 .proc_handler   = &proc_dostring,
395                 .strategy       = &sysctl_string,
396         },
397 #endif
398 #ifdef CONFIG_HOTPLUG
399         {
400                 .ctl_name       = KERN_HOTPLUG,
401                 .procname       = "hotplug",
402                 .data           = &uevent_helper,
403                 .maxlen         = UEVENT_HELPER_PATH_LEN,
404                 .mode           = 0644,
405                 .proc_handler   = &proc_dostring,
406                 .strategy       = &sysctl_string,
407         },
408 #endif
409 #ifdef CONFIG_CHR_DEV_SG
410         {
411                 .ctl_name       = KERN_SG_BIG_BUFF,
412                 .procname       = "sg-big-buff",
413                 .data           = &sg_big_buff,
414                 .maxlen         = sizeof (int),
415                 .mode           = 0444,
416                 .proc_handler   = &proc_dointvec,
417         },
418 #endif
419 #ifdef CONFIG_BSD_PROCESS_ACCT
420         {
421                 .ctl_name       = KERN_ACCT,
422                 .procname       = "acct",
423                 .data           = &acct_parm,
424                 .maxlen         = 3*sizeof(int),
425                 .mode           = 0644,
426                 .proc_handler   = &proc_dointvec,
427         },
428 #endif
429 #ifdef CONFIG_SYSVIPC
430         {
431                 .ctl_name       = KERN_SHMMAX,
432                 .procname       = "shmmax",
433                 .data           = &shm_ctlmax,
434                 .maxlen         = sizeof (size_t),
435                 .mode           = 0644,
436                 .proc_handler   = &proc_doulongvec_minmax,
437         },
438         {
439                 .ctl_name       = KERN_SHMALL,
440                 .procname       = "shmall",
441                 .data           = &shm_ctlall,
442                 .maxlen         = sizeof (size_t),
443                 .mode           = 0644,
444                 .proc_handler   = &proc_doulongvec_minmax,
445         },
446         {
447                 .ctl_name       = KERN_SHMMNI,
448                 .procname       = "shmmni",
449                 .data           = &shm_ctlmni,
450                 .maxlen         = sizeof (int),
451                 .mode           = 0644,
452                 .proc_handler   = &proc_dointvec,
453         },
454         {
455                 .ctl_name       = KERN_MSGMAX,
456                 .procname       = "msgmax",
457                 .data           = &msg_ctlmax,
458                 .maxlen         = sizeof (int),
459                 .mode           = 0644,
460                 .proc_handler   = &proc_dointvec,
461         },
462         {
463                 .ctl_name       = KERN_MSGMNI,
464                 .procname       = "msgmni",
465                 .data           = &msg_ctlmni,
466                 .maxlen         = sizeof (int),
467                 .mode           = 0644,
468                 .proc_handler   = &proc_dointvec,
469         },
470         {
471                 .ctl_name       = KERN_MSGMNB,
472                 .procname       =  "msgmnb",
473                 .data           = &msg_ctlmnb,
474                 .maxlen         = sizeof (int),
475                 .mode           = 0644,
476                 .proc_handler   = &proc_dointvec,
477         },
478         {
479                 .ctl_name       = KERN_SEM,
480                 .procname       = "sem",
481                 .data           = &sem_ctls,
482                 .maxlen         = 4*sizeof (int),
483                 .mode           = 0644,
484                 .proc_handler   = &proc_dointvec,
485         },
486 #endif
487 #ifdef CONFIG_MAGIC_SYSRQ
488         {
489                 .ctl_name       = KERN_SYSRQ,
490                 .procname       = "sysrq",
491                 .data           = &sysrq_enabled,
492                 .maxlen         = sizeof (int),
493                 .mode           = 0644,
494                 .proc_handler   = &proc_dointvec,
495         },
496 #endif
497         {
498                 .ctl_name       = KERN_CADPID,
499                 .procname       = "cad_pid",
500                 .data           = &cad_pid,
501                 .maxlen         = sizeof (int),
502                 .mode           = 0600,
503                 .proc_handler   = &proc_dointvec,
504         },
505         {
506                 .ctl_name       = KERN_MAX_THREADS,
507                 .procname       = "threads-max",
508                 .data           = &max_threads,
509                 .maxlen         = sizeof(int),
510                 .mode           = 0644,
511                 .proc_handler   = &proc_dointvec,
512         },
513         {
514                 .ctl_name       = KERN_RANDOM,
515                 .procname       = "random",
516                 .mode           = 0555,
517                 .child          = random_table,
518         },
519 #ifdef CONFIG_UNIX98_PTYS
520         {
521                 .ctl_name       = KERN_PTY,
522                 .procname       = "pty",
523                 .mode           = 0555,
524                 .child          = pty_table,
525         },
526 #endif
527         {
528                 .ctl_name       = KERN_OVERFLOWUID,
529                 .procname       = "overflowuid",
530                 .data           = &overflowuid,
531                 .maxlen         = sizeof(int),
532                 .mode           = 0644,
533                 .proc_handler   = &proc_dointvec_minmax,
534                 .strategy       = &sysctl_intvec,
535                 .extra1         = &minolduid,
536                 .extra2         = &maxolduid,
537         },
538         {
539                 .ctl_name       = KERN_OVERFLOWGID,
540                 .procname       = "overflowgid",
541                 .data           = &overflowgid,
542                 .maxlen         = sizeof(int),
543                 .mode           = 0644,
544                 .proc_handler   = &proc_dointvec_minmax,
545                 .strategy       = &sysctl_intvec,
546                 .extra1         = &minolduid,
547                 .extra2         = &maxolduid,
548         },
549 #ifdef CONFIG_S390
550 #ifdef CONFIG_MATHEMU
551         {
552                 .ctl_name       = KERN_IEEE_EMULATION_WARNINGS,
553                 .procname       = "ieee_emulation_warnings",
554                 .data           = &sysctl_ieee_emulation_warnings,
555                 .maxlen         = sizeof(int),
556                 .mode           = 0644,
557                 .proc_handler   = &proc_dointvec,
558         },
559 #endif
560 #ifdef CONFIG_NO_IDLE_HZ
561         {
562                 .ctl_name       = KERN_HZ_TIMER,
563                 .procname       = "hz_timer",
564                 .data           = &sysctl_hz_timer,
565                 .maxlen         = sizeof(int),
566                 .mode           = 0644,
567                 .proc_handler   = &proc_dointvec,
568         },
569 #endif
570         {
571                 .ctl_name       = KERN_S390_USER_DEBUG_LOGGING,
572                 .procname       = "userprocess_debug",
573                 .data           = &sysctl_userprocess_debug,
574                 .maxlen         = sizeof(int),
575                 .mode           = 0644,
576                 .proc_handler   = &proc_dointvec,
577         },
578 #endif
579         {
580                 .ctl_name       = KERN_PIDMAX,
581                 .procname       = "pid_max",
582                 .data           = &pid_max,
583                 .maxlen         = sizeof (int),
584                 .mode           = 0644,
585                 .proc_handler   = &proc_dointvec_minmax,
586                 .strategy       = sysctl_intvec,
587                 .extra1         = &pid_max_min,
588                 .extra2         = &pid_max_max,
589         },
590         {
591                 .ctl_name       = KERN_PANIC_ON_OOPS,
592                 .procname       = "panic_on_oops",
593                 .data           = &panic_on_oops,
594                 .maxlen         = sizeof(int),
595                 .mode           = 0644,
596                 .proc_handler   = &proc_dointvec,
597         },
598         {
599                 .ctl_name       = KERN_PRINTK_RATELIMIT,
600                 .procname       = "printk_ratelimit",
601                 .data           = &printk_ratelimit_jiffies,
602                 .maxlen         = sizeof(int),
603                 .mode           = 0644,
604                 .proc_handler   = &proc_dointvec_jiffies,
605                 .strategy       = &sysctl_jiffies,
606         },
607         {
608                 .ctl_name       = KERN_PRINTK_RATELIMIT_BURST,
609                 .procname       = "printk_ratelimit_burst",
610                 .data           = &printk_ratelimit_burst,
611                 .maxlen         = sizeof(int),
612                 .mode           = 0644,
613                 .proc_handler   = &proc_dointvec,
614         },
615         {
616                 .ctl_name       = KERN_NGROUPS_MAX,
617                 .procname       = "ngroups_max",
618                 .data           = &ngroups_max,
619                 .maxlen         = sizeof (int),
620                 .mode           = 0444,
621                 .proc_handler   = &proc_dointvec,
622         },
623 #if defined(CONFIG_X86_LOCAL_APIC) && defined(CONFIG_X86)
624         {
625                 .ctl_name       = KERN_UNKNOWN_NMI_PANIC,
626                 .procname       = "unknown_nmi_panic",
627                 .data           = &unknown_nmi_panic,
628                 .maxlen         = sizeof (int),
629                 .mode           = 0644,
630                 .proc_handler   = &proc_unknown_nmi_panic,
631         },
632 #endif
633 #if defined(CONFIG_X86)
634         {
635                 .ctl_name       = KERN_BOOTLOADER_TYPE,
636                 .procname       = "bootloader_type",
637                 .data           = &bootloader_type,
638                 .maxlen         = sizeof (int),
639                 .mode           = 0444,
640                 .proc_handler   = &proc_dointvec,
641         },
642 #endif
643 #if defined(CONFIG_MMU)
644         {
645                 .ctl_name       = KERN_RANDOMIZE,
646                 .procname       = "randomize_va_space",
647                 .data           = &randomize_va_space,
648                 .maxlen         = sizeof(int),
649                 .mode           = 0644,
650                 .proc_handler   = &proc_dointvec,
651         },
652 #endif
653 #if defined(CONFIG_S390) && defined(CONFIG_SMP)
654         {
655                 .ctl_name       = KERN_SPIN_RETRY,
656                 .procname       = "spin_retry",
657                 .data           = &spin_retry,
658                 .maxlen         = sizeof (int),
659                 .mode           = 0644,
660                 .proc_handler   = &proc_dointvec,
661         },
662 #endif
663 #ifdef CONFIG_ACPI_SLEEP
664         {
665                 .ctl_name       = KERN_ACPI_VIDEO_FLAGS,
666                 .procname       = "acpi_video_flags",
667                 .data           = &acpi_video_flags,
668                 .maxlen         = sizeof (unsigned long),
669                 .mode           = 0644,
670                 .proc_handler   = &proc_dointvec,
671         },
672 #endif
673 #ifdef CONFIG_IA64
674         {
675                 .ctl_name       = KERN_IA64_UNALIGNED,
676                 .procname       = "ignore-unaligned-usertrap",
677                 .data           = &no_unaligned_warning,
678                 .maxlen         = sizeof (int),
679                 .mode           = 0644,
680                 .proc_handler   = &proc_dointvec,
681         },
682 #endif
683         { .ctl_name = 0 }
684 };
685
686 /* Constants for minimum and maximum testing in vm_table.
687    We use these as one-element integer vectors. */
688 static int zero;
689 static int one_hundred = 100;
690
691
692 static ctl_table vm_table[] = {
693         {
694                 .ctl_name       = VM_OVERCOMMIT_MEMORY,
695                 .procname       = "overcommit_memory",
696                 .data           = &sysctl_overcommit_memory,
697                 .maxlen         = sizeof(sysctl_overcommit_memory),
698                 .mode           = 0644,
699                 .proc_handler   = &proc_dointvec,
700         },
701         {
702                 .ctl_name       = VM_OVERCOMMIT_RATIO,
703                 .procname       = "overcommit_ratio",
704                 .data           = &sysctl_overcommit_ratio,
705                 .maxlen         = sizeof(sysctl_overcommit_ratio),
706                 .mode           = 0644,
707                 .proc_handler   = &proc_dointvec,
708         },
709         {
710                 .ctl_name       = VM_PAGE_CLUSTER,
711                 .procname       = "page-cluster", 
712                 .data           = &page_cluster,
713                 .maxlen         = sizeof(int),
714                 .mode           = 0644,
715                 .proc_handler   = &proc_dointvec,
716         },
717         {
718                 .ctl_name       = VM_DIRTY_BACKGROUND,
719                 .procname       = "dirty_background_ratio",
720                 .data           = &dirty_background_ratio,
721                 .maxlen         = sizeof(dirty_background_ratio),
722                 .mode           = 0644,
723                 .proc_handler   = &proc_dointvec_minmax,
724                 .strategy       = &sysctl_intvec,
725                 .extra1         = &zero,
726                 .extra2         = &one_hundred,
727         },
728         {
729                 .ctl_name       = VM_DIRTY_RATIO,
730                 .procname       = "dirty_ratio",
731                 .data           = &vm_dirty_ratio,
732                 .maxlen         = sizeof(vm_dirty_ratio),
733                 .mode           = 0644,
734                 .proc_handler   = &proc_dointvec_minmax,
735                 .strategy       = &sysctl_intvec,
736                 .extra1         = &zero,
737                 .extra2         = &one_hundred,
738         },
739         {
740                 .ctl_name       = VM_DIRTY_WB_CS,
741                 .procname       = "dirty_writeback_centisecs",
742                 .data           = &dirty_writeback_centisecs,
743                 .maxlen         = sizeof(dirty_writeback_centisecs),
744                 .mode           = 0644,
745                 .proc_handler   = &dirty_writeback_centisecs_handler,
746         },
747         {
748                 .ctl_name       = VM_DIRTY_EXPIRE_CS,
749                 .procname       = "dirty_expire_centisecs",
750                 .data           = &dirty_expire_centisecs,
751                 .maxlen         = sizeof(dirty_expire_centisecs),
752                 .mode           = 0644,
753                 .proc_handler   = &proc_dointvec,
754         },
755         {
756                 .ctl_name       = VM_NR_PDFLUSH_THREADS,
757                 .procname       = "nr_pdflush_threads",
758                 .data           = &nr_pdflush_threads,
759                 .maxlen         = sizeof nr_pdflush_threads,
760                 .mode           = 0444 /* read-only*/,
761                 .proc_handler   = &proc_dointvec,
762         },
763         {
764                 .ctl_name       = VM_SWAPPINESS,
765                 .procname       = "swappiness",
766                 .data           = &vm_swappiness,
767                 .maxlen         = sizeof(vm_swappiness),
768                 .mode           = 0644,
769                 .proc_handler   = &proc_dointvec_minmax,
770                 .strategy       = &sysctl_intvec,
771                 .extra1         = &zero,
772                 .extra2         = &one_hundred,
773         },
774 #ifdef CONFIG_HUGETLB_PAGE
775          {
776                 .ctl_name       = VM_HUGETLB_PAGES,
777                 .procname       = "nr_hugepages",
778                 .data           = &max_huge_pages,
779                 .maxlen         = sizeof(unsigned long),
780                 .mode           = 0644,
781                 .proc_handler   = &hugetlb_sysctl_handler,
782                 .extra1         = (void *)&hugetlb_zero,
783                 .extra2         = (void *)&hugetlb_infinity,
784          },
785          {
786                 .ctl_name       = VM_HUGETLB_GROUP,
787                 .procname       = "hugetlb_shm_group",
788                 .data           = &sysctl_hugetlb_shm_group,
789                 .maxlen         = sizeof(gid_t),
790                 .mode           = 0644,
791                 .proc_handler   = &proc_dointvec,
792          },
793 #endif
794         {
795                 .ctl_name       = VM_LOWMEM_RESERVE_RATIO,
796                 .procname       = "lowmem_reserve_ratio",
797                 .data           = &sysctl_lowmem_reserve_ratio,
798                 .maxlen         = sizeof(sysctl_lowmem_reserve_ratio),
799                 .mode           = 0644,
800                 .proc_handler   = &lowmem_reserve_ratio_sysctl_handler,
801                 .strategy       = &sysctl_intvec,
802         },
803         {
804                 .ctl_name       = VM_DROP_PAGECACHE,
805                 .procname       = "drop_caches",
806                 .data           = &sysctl_drop_caches,
807                 .maxlen         = sizeof(int),
808                 .mode           = 0644,
809                 .proc_handler   = drop_caches_sysctl_handler,
810                 .strategy       = &sysctl_intvec,
811         },
812         {
813                 .ctl_name       = VM_MIN_FREE_KBYTES,
814                 .procname       = "min_free_kbytes",
815                 .data           = &min_free_kbytes,
816                 .maxlen         = sizeof(min_free_kbytes),
817                 .mode           = 0644,
818                 .proc_handler   = &min_free_kbytes_sysctl_handler,
819                 .strategy       = &sysctl_intvec,
820                 .extra1         = &zero,
821         },
822         {
823                 .ctl_name       = VM_PERCPU_PAGELIST_FRACTION,
824                 .procname       = "percpu_pagelist_fraction",
825                 .data           = &percpu_pagelist_fraction,
826                 .maxlen         = sizeof(percpu_pagelist_fraction),
827                 .mode           = 0644,
828                 .proc_handler   = &percpu_pagelist_fraction_sysctl_handler,
829                 .strategy       = &sysctl_intvec,
830                 .extra1         = &min_percpu_pagelist_fract,
831         },
832 #ifdef CONFIG_MMU
833         {
834                 .ctl_name       = VM_MAX_MAP_COUNT,
835                 .procname       = "max_map_count",
836                 .data           = &sysctl_max_map_count,
837                 .maxlen         = sizeof(sysctl_max_map_count),
838                 .mode           = 0644,
839                 .proc_handler   = &proc_dointvec
840         },
841 #endif
842         {
843                 .ctl_name       = VM_LAPTOP_MODE,
844                 .procname       = "laptop_mode",
845                 .data           = &laptop_mode,
846                 .maxlen         = sizeof(laptop_mode),
847                 .mode           = 0644,
848                 .proc_handler   = &proc_dointvec,
849                 .strategy       = &sysctl_intvec,
850                 .extra1         = &zero,
851         },
852         {
853                 .ctl_name       = VM_BLOCK_DUMP,
854                 .procname       = "block_dump",
855                 .data           = &block_dump,
856                 .maxlen         = sizeof(block_dump),
857                 .mode           = 0644,
858                 .proc_handler   = &proc_dointvec,
859                 .strategy       = &sysctl_intvec,
860                 .extra1         = &zero,
861         },
862         {
863                 .ctl_name       = VM_VFS_CACHE_PRESSURE,
864                 .procname       = "vfs_cache_pressure",
865                 .data           = &sysctl_vfs_cache_pressure,
866                 .maxlen         = sizeof(sysctl_vfs_cache_pressure),
867                 .mode           = 0644,
868                 .proc_handler   = &proc_dointvec,
869                 .strategy       = &sysctl_intvec,
870                 .extra1         = &zero,
871         },
872 #ifdef HAVE_ARCH_PICK_MMAP_LAYOUT
873         {
874                 .ctl_name       = VM_LEGACY_VA_LAYOUT,
875                 .procname       = "legacy_va_layout",
876                 .data           = &sysctl_legacy_va_layout,
877                 .maxlen         = sizeof(sysctl_legacy_va_layout),
878                 .mode           = 0644,
879                 .proc_handler   = &proc_dointvec,
880                 .strategy       = &sysctl_intvec,
881                 .extra1         = &zero,
882         },
883 #endif
884 #ifdef CONFIG_SWAP
885         {
886                 .ctl_name       = VM_SWAP_TOKEN_TIMEOUT,
887                 .procname       = "swap_token_timeout",
888                 .data           = &swap_token_default_timeout,
889                 .maxlen         = sizeof(swap_token_default_timeout),
890                 .mode           = 0644,
891                 .proc_handler   = &proc_dointvec_jiffies,
892                 .strategy       = &sysctl_jiffies,
893         },
894 #endif
895 #ifdef CONFIG_NUMA
896         {
897                 .ctl_name       = VM_ZONE_RECLAIM_MODE,
898                 .procname       = "zone_reclaim_mode",
899                 .data           = &zone_reclaim_mode,
900                 .maxlen         = sizeof(zone_reclaim_mode),
901                 .mode           = 0644,
902                 .proc_handler   = &proc_dointvec,
903                 .strategy       = &sysctl_intvec,
904                 .extra1         = &zero,
905         },
906         {
907                 .ctl_name       = VM_ZONE_RECLAIM_INTERVAL,
908                 .procname       = "zone_reclaim_interval",
909                 .data           = &zone_reclaim_interval,
910                 .maxlen         = sizeof(zone_reclaim_interval),
911                 .mode           = 0644,
912                 .proc_handler   = &proc_dointvec_jiffies,
913                 .strategy       = &sysctl_jiffies,
914         },
915 #endif
916         { .ctl_name = 0 }
917 };
918
919 static ctl_table proc_table[] = {
920         { .ctl_name = 0 }
921 };
922
923 static ctl_table fs_table[] = {
924         {
925                 .ctl_name       = FS_NRINODE,
926                 .procname       = "inode-nr",
927                 .data           = &inodes_stat,
928                 .maxlen         = 2*sizeof(int),
929                 .mode           = 0444,
930                 .proc_handler   = &proc_dointvec,
931         },
932         {
933                 .ctl_name       = FS_STATINODE,
934                 .procname       = "inode-state",
935                 .data           = &inodes_stat,
936                 .maxlen         = 7*sizeof(int),
937                 .mode           = 0444,
938                 .proc_handler   = &proc_dointvec,
939         },
940         {
941                 .ctl_name       = FS_NRFILE,
942                 .procname       = "file-nr",
943                 .data           = &files_stat,
944                 .maxlen         = 3*sizeof(int),
945                 .mode           = 0444,
946                 .proc_handler   = &proc_dointvec,
947         },
948         {
949                 .ctl_name       = FS_MAXFILE,
950                 .procname       = "file-max",
951                 .data           = &files_stat.max_files,
952                 .maxlen         = sizeof(int),
953                 .mode           = 0644,
954                 .proc_handler   = &proc_dointvec,
955         },
956         {
957                 .ctl_name       = FS_DENTRY,
958                 .procname       = "dentry-state",
959                 .data           = &dentry_stat,
960                 .maxlen         = 6*sizeof(int),
961                 .mode           = 0444,
962                 .proc_handler   = &proc_dointvec,
963         },
964         {
965                 .ctl_name       = FS_OVERFLOWUID,
966                 .procname       = "overflowuid",
967                 .data           = &fs_overflowuid,
968                 .maxlen         = sizeof(int),
969                 .mode           = 0644,
970                 .proc_handler   = &proc_dointvec_minmax,
971                 .strategy       = &sysctl_intvec,
972                 .extra1         = &minolduid,
973                 .extra2         = &maxolduid,
974         },
975         {
976                 .ctl_name       = FS_OVERFLOWGID,
977                 .procname       = "overflowgid",
978                 .data           = &fs_overflowgid,
979                 .maxlen         = sizeof(int),
980                 .mode           = 0644,
981                 .proc_handler   = &proc_dointvec_minmax,
982                 .strategy       = &sysctl_intvec,
983                 .extra1         = &minolduid,
984                 .extra2         = &maxolduid,
985         },
986         {
987                 .ctl_name       = FS_LEASES,
988                 .procname       = "leases-enable",
989                 .data           = &leases_enable,
990                 .maxlen         = sizeof(int),
991                 .mode           = 0644,
992                 .proc_handler   = &proc_dointvec,
993         },
994 #ifdef CONFIG_DNOTIFY
995         {
996                 .ctl_name       = FS_DIR_NOTIFY,
997                 .procname       = "dir-notify-enable",
998                 .data           = &dir_notify_enable,
999                 .maxlen         = sizeof(int),
1000                 .mode           = 0644,
1001                 .proc_handler   = &proc_dointvec,
1002         },
1003 #endif
1004 #ifdef CONFIG_MMU
1005         {
1006                 .ctl_name       = FS_LEASE_TIME,
1007                 .procname       = "lease-break-time",
1008                 .data           = &lease_break_time,
1009                 .maxlen         = sizeof(int),
1010                 .mode           = 0644,
1011                 .proc_handler   = &proc_dointvec,
1012         },
1013         {
1014                 .ctl_name       = FS_AIO_NR,
1015                 .procname       = "aio-nr",
1016                 .data           = &aio_nr,
1017                 .maxlen         = sizeof(aio_nr),
1018                 .mode           = 0444,
1019                 .proc_handler   = &proc_doulongvec_minmax,
1020         },
1021         {
1022                 .ctl_name       = FS_AIO_MAX_NR,
1023                 .procname       = "aio-max-nr",
1024                 .data           = &aio_max_nr,
1025                 .maxlen         = sizeof(aio_max_nr),
1026                 .mode           = 0644,
1027                 .proc_handler   = &proc_doulongvec_minmax,
1028         },
1029 #ifdef CONFIG_INOTIFY
1030         {
1031                 .ctl_name       = FS_INOTIFY,
1032                 .procname       = "inotify",
1033                 .mode           = 0555,
1034                 .child          = inotify_table,
1035         },
1036 #endif  
1037 #endif
1038         {
1039                 .ctl_name       = KERN_SETUID_DUMPABLE,
1040                 .procname       = "suid_dumpable",
1041                 .data           = &suid_dumpable,
1042                 .maxlen         = sizeof(int),
1043                 .mode           = 0644,
1044                 .proc_handler   = &proc_dointvec,
1045         },
1046         { .ctl_name = 0 }
1047 };
1048
1049 static ctl_table debug_table[] = {
1050         { .ctl_name = 0 }
1051 };
1052
1053 static ctl_table dev_table[] = {
1054         { .ctl_name = 0 }
1055 };
1056
1057 extern void init_irq_proc (void);
1058
1059 static DEFINE_SPINLOCK(sysctl_lock);
1060
1061 /* called under sysctl_lock */
1062 static int use_table(struct ctl_table_header *p)
1063 {
1064         if (unlikely(p->unregistering))
1065                 return 0;
1066         p->used++;
1067         return 1;
1068 }
1069
1070 /* called under sysctl_lock */
1071 static void unuse_table(struct ctl_table_header *p)
1072 {
1073         if (!--p->used)
1074                 if (unlikely(p->unregistering))
1075                         complete(p->unregistering);
1076 }
1077
1078 /* called under sysctl_lock, will reacquire if has to wait */
1079 static void start_unregistering(struct ctl_table_header *p)
1080 {
1081         /*
1082          * if p->used is 0, nobody will ever touch that entry again;
1083          * we'll eliminate all paths to it before dropping sysctl_lock
1084          */
1085         if (unlikely(p->used)) {
1086                 struct completion wait;
1087                 init_completion(&wait);
1088                 p->unregistering = &wait;
1089                 spin_unlock(&sysctl_lock);
1090                 wait_for_completion(&wait);
1091                 spin_lock(&sysctl_lock);
1092         }
1093         /*
1094          * do not remove from the list until nobody holds it; walking the
1095          * list in do_sysctl() relies on that.
1096          */
1097         list_del_init(&p->ctl_entry);
1098 }
1099
1100 void __init sysctl_init(void)
1101 {
1102 #ifdef CONFIG_PROC_FS
1103         register_proc_table(root_table, proc_sys_root, &root_table_header);
1104         init_irq_proc();
1105 #endif
1106 }
1107
1108 int do_sysctl(int __user *name, int nlen, void __user *oldval, size_t __user *oldlenp,
1109                void __user *newval, size_t newlen)
1110 {
1111         struct list_head *tmp;
1112         int error = -ENOTDIR;
1113
1114         if (nlen <= 0 || nlen >= CTL_MAXNAME)
1115                 return -ENOTDIR;
1116         if (oldval) {
1117                 int old_len;
1118                 if (!oldlenp || get_user(old_len, oldlenp))
1119                         return -EFAULT;
1120         }
1121         spin_lock(&sysctl_lock);
1122         tmp = &root_table_header.ctl_entry;
1123         do {
1124                 struct ctl_table_header *head =
1125                         list_entry(tmp, struct ctl_table_header, ctl_entry);
1126                 void *context = NULL;
1127
1128                 if (!use_table(head))
1129                         continue;
1130
1131                 spin_unlock(&sysctl_lock);
1132
1133                 error = parse_table(name, nlen, oldval, oldlenp, 
1134                                         newval, newlen, head->ctl_table,
1135                                         &context);
1136                 kfree(context);
1137
1138                 spin_lock(&sysctl_lock);
1139                 unuse_table(head);
1140                 if (error != -ENOTDIR)
1141                         break;
1142         } while ((tmp = tmp->next) != &root_table_header.ctl_entry);
1143         spin_unlock(&sysctl_lock);
1144         return error;
1145 }
1146
1147 asmlinkage long sys_sysctl(struct __sysctl_args __user *args)
1148 {
1149         struct __sysctl_args tmp;
1150         int error;
1151
1152         if (copy_from_user(&tmp, args, sizeof(tmp)))
1153                 return -EFAULT;
1154
1155         lock_kernel();
1156         error = do_sysctl(tmp.name, tmp.nlen, tmp.oldval, tmp.oldlenp,
1157                           tmp.newval, tmp.newlen);
1158         unlock_kernel();
1159         return error;
1160 }
1161
1162 /*
1163  * ctl_perm does NOT grant the superuser all rights automatically, because
1164  * some sysctl variables are readonly even to root.
1165  */
1166
1167 static int test_perm(int mode, int op)
1168 {
1169         if (!current->euid)
1170                 mode >>= 6;
1171         else if (in_egroup_p(0))
1172                 mode >>= 3;
1173         if ((mode & op & 0007) == op)
1174                 return 0;
1175         return -EACCES;
1176 }
1177
1178 static inline int ctl_perm(ctl_table *table, int op)
1179 {
1180         int error;
1181         error = security_sysctl(table, op);
1182         if (error)
1183                 return error;
1184         return test_perm(table->mode, op);
1185 }
1186
1187 static int parse_table(int __user *name, int nlen,
1188                        void __user *oldval, size_t __user *oldlenp,
1189                        void __user *newval, size_t newlen,
1190                        ctl_table *table, void **context)
1191 {
1192         int n;
1193 repeat:
1194         if (!nlen)
1195                 return -ENOTDIR;
1196         if (get_user(n, name))
1197                 return -EFAULT;
1198         for ( ; table->ctl_name; table++) {
1199                 if (n == table->ctl_name || table->ctl_name == CTL_ANY) {
1200                         int error;
1201                         if (table->child) {
1202                                 if (ctl_perm(table, 001))
1203                                         return -EPERM;
1204                                 if (table->strategy) {
1205                                         error = table->strategy(
1206                                                 table, name, nlen,
1207                                                 oldval, oldlenp,
1208                                                 newval, newlen, context);
1209                                         if (error)
1210                                                 return error;
1211                                 }
1212                                 name++;
1213                                 nlen--;
1214                                 table = table->child;
1215                                 goto repeat;
1216                         }
1217                         error = do_sysctl_strategy(table, name, nlen,
1218                                                    oldval, oldlenp,
1219                                                    newval, newlen, context);
1220                         return error;
1221                 }
1222         }
1223         return -ENOTDIR;
1224 }
1225
1226 /* Perform the actual read/write of a sysctl table entry. */
1227 int do_sysctl_strategy (ctl_table *table, 
1228                         int __user *name, int nlen,
1229                         void __user *oldval, size_t __user *oldlenp,
1230                         void __user *newval, size_t newlen, void **context)
1231 {
1232         int op = 0, rc;
1233         size_t len;
1234
1235         if (oldval)
1236                 op |= 004;
1237         if (newval) 
1238                 op |= 002;
1239         if (ctl_perm(table, op))
1240                 return -EPERM;
1241
1242         if (table->strategy) {
1243                 rc = table->strategy(table, name, nlen, oldval, oldlenp,
1244                                      newval, newlen, context);
1245                 if (rc < 0)
1246                         return rc;
1247                 if (rc > 0)
1248                         return 0;
1249         }
1250
1251         /* If there is no strategy routine, or if the strategy returns
1252          * zero, proceed with automatic r/w */
1253         if (table->data && table->maxlen) {
1254                 if (oldval && oldlenp) {
1255                         if (get_user(len, oldlenp))
1256                                 return -EFAULT;
1257                         if (len) {
1258                                 if (len > table->maxlen)
1259                                         len = table->maxlen;
1260                                 if(copy_to_user(oldval, table->data, len))
1261                                         return -EFAULT;
1262                                 if(put_user(len, oldlenp))
1263                                         return -EFAULT;
1264                         }
1265                 }
1266                 if (newval && newlen) {
1267                         len = newlen;
1268                         if (len > table->maxlen)
1269                                 len = table->maxlen;
1270                         if(copy_from_user(table->data, newval, len))
1271                                 return -EFAULT;
1272                 }
1273         }
1274         return 0;
1275 }
1276
1277 /**
1278  * register_sysctl_table - register a sysctl hierarchy
1279  * @table: the top-level table structure
1280  * @insert_at_head: whether the entry should be inserted in front or at the end
1281  *
1282  * Register a sysctl table hierarchy. @table should be a filled in ctl_table
1283  * array. An entry with a ctl_name of 0 terminates the table. 
1284  *
1285  * The members of the &ctl_table structure are used as follows:
1286  *
1287  * ctl_name - This is the numeric sysctl value used by sysctl(2). The number
1288  *            must be unique within that level of sysctl
1289  *
1290  * procname - the name of the sysctl file under /proc/sys. Set to %NULL to not
1291  *            enter a sysctl file
1292  *
1293  * data - a pointer to data for use by proc_handler
1294  *
1295  * maxlen - the maximum size in bytes of the data
1296  *
1297  * mode - the file permissions for the /proc/sys file, and for sysctl(2)
1298  *
1299  * child - a pointer to the child sysctl table if this entry is a directory, or
1300  *         %NULL.
1301  *
1302  * proc_handler - the text handler routine (described below)
1303  *
1304  * strategy - the strategy routine (described below)
1305  *
1306  * de - for internal use by the sysctl routines
1307  *
1308  * extra1, extra2 - extra pointers usable by the proc handler routines
1309  *
1310  * Leaf nodes in the sysctl tree will be represented by a single file
1311  * under /proc; non-leaf nodes will be represented by directories.
1312  *
1313  * sysctl(2) can automatically manage read and write requests through
1314  * the sysctl table.  The data and maxlen fields of the ctl_table
1315  * struct enable minimal validation of the values being written to be
1316  * performed, and the mode field allows minimal authentication.
1317  *
1318  * More sophisticated management can be enabled by the provision of a
1319  * strategy routine with the table entry.  This will be called before
1320  * any automatic read or write of the data is performed.
1321  *
1322  * The strategy routine may return
1323  *
1324  * < 0 - Error occurred (error is passed to user process)
1325  *
1326  * 0   - OK - proceed with automatic read or write.
1327  *
1328  * > 0 - OK - read or write has been done by the strategy routine, so
1329  *       return immediately.
1330  *
1331  * There must be a proc_handler routine for any terminal nodes
1332  * mirrored under /proc/sys (non-terminals are handled by a built-in
1333  * directory handler).  Several default handlers are available to
1334  * cover common cases -
1335  *
1336  * proc_dostring(), proc_dointvec(), proc_dointvec_jiffies(),
1337  * proc_dointvec_userhz_jiffies(), proc_dointvec_minmax(), 
1338  * proc_doulongvec_ms_jiffies_minmax(), proc_doulongvec_minmax()
1339  *
1340  * It is the handler's job to read the input buffer from user memory
1341  * and process it. The handler should return 0 on success.
1342  *
1343  * This routine returns %NULL on a failure to register, and a pointer
1344  * to the table header on success.
1345  */
1346 struct ctl_table_header *register_sysctl_table(ctl_table * table, 
1347                                                int insert_at_head)
1348 {
1349         struct ctl_table_header *tmp;
1350         tmp = kmalloc(sizeof(struct ctl_table_header), GFP_KERNEL);
1351         if (!tmp)
1352                 return NULL;
1353         tmp->ctl_table = table;
1354         INIT_LIST_HEAD(&tmp->ctl_entry);
1355         tmp->used = 0;
1356         tmp->unregistering = NULL;
1357         spin_lock(&sysctl_lock);
1358         if (insert_at_head)
1359                 list_add(&tmp->ctl_entry, &root_table_header.ctl_entry);
1360         else
1361                 list_add_tail(&tmp->ctl_entry, &root_table_header.ctl_entry);
1362         spin_unlock(&sysctl_lock);
1363 #ifdef CONFIG_PROC_FS
1364         register_proc_table(table, proc_sys_root, tmp);
1365 #endif
1366         return tmp;
1367 }
1368
1369 /**
1370  * unregister_sysctl_table - unregister a sysctl table hierarchy
1371  * @header: the header returned from register_sysctl_table
1372  *
1373  * Unregisters the sysctl table and all children. proc entries may not
1374  * actually be removed until they are no longer used by anyone.
1375  */
1376 void unregister_sysctl_table(struct ctl_table_header * header)
1377 {
1378         might_sleep();
1379         spin_lock(&sysctl_lock);
1380         start_unregistering(header);
1381 #ifdef CONFIG_PROC_FS
1382         unregister_proc_table(header->ctl_table, proc_sys_root);
1383 #endif
1384         spin_unlock(&sysctl_lock);
1385         kfree(header);
1386 }
1387
1388 /*
1389  * /proc/sys support
1390  */
1391
1392 #ifdef CONFIG_PROC_FS
1393
1394 /* Scan the sysctl entries in table and add them all into /proc */
1395 static void register_proc_table(ctl_table * table, struct proc_dir_entry *root, void *set)
1396 {
1397         struct proc_dir_entry *de;
1398         int len;
1399         mode_t mode;
1400         
1401         for (; table->ctl_name; table++) {
1402                 /* Can't do anything without a proc name. */
1403                 if (!table->procname)
1404                         continue;
1405                 /* Maybe we can't do anything with it... */
1406                 if (!table->proc_handler && !table->child) {
1407                         printk(KERN_WARNING "SYSCTL: Can't register %s\n",
1408                                 table->procname);
1409                         continue;
1410                 }
1411
1412                 len = strlen(table->procname);
1413                 mode = table->mode;
1414
1415                 de = NULL;
1416                 if (table->proc_handler)
1417                         mode |= S_IFREG;
1418                 else {
1419                         mode |= S_IFDIR;
1420                         for (de = root->subdir; de; de = de->next) {
1421                                 if (proc_match(len, table->procname, de))
1422                                         break;
1423                         }
1424                         /* If the subdir exists already, de is non-NULL */
1425                 }
1426
1427                 if (!de) {
1428                         de = create_proc_entry(table->procname, mode, root);
1429                         if (!de)
1430                                 continue;
1431                         de->set = set;
1432                         de->data = (void *) table;
1433                         if (table->proc_handler)
1434                                 de->proc_fops = &proc_sys_file_operations;
1435                 }
1436                 table->de = de;
1437                 if (de->mode & S_IFDIR)
1438                         register_proc_table(table->child, de, set);
1439         }
1440 }
1441
1442 /*
1443  * Unregister a /proc sysctl table and any subdirectories.
1444  */
1445 static void unregister_proc_table(ctl_table * table, struct proc_dir_entry *root)
1446 {
1447         struct proc_dir_entry *de;
1448         for (; table->ctl_name; table++) {
1449                 if (!(de = table->de))
1450                         continue;
1451                 if (de->mode & S_IFDIR) {
1452                         if (!table->child) {
1453                                 printk (KERN_ALERT "Help - malformed sysctl tree on free\n");
1454                                 continue;
1455                         }
1456                         unregister_proc_table(table->child, de);
1457
1458                         /* Don't unregister directories which still have entries.. */
1459                         if (de->subdir)
1460                                 continue;
1461                 }
1462
1463                 /*
1464                  * In any case, mark the entry as goner; we'll keep it
1465                  * around if it's busy, but we'll know to do nothing with
1466                  * its fields.  We are under sysctl_lock here.
1467                  */
1468                 de->data = NULL;
1469
1470                 /* Don't unregister proc entries that are still being used.. */
1471                 if (atomic_read(&de->count))
1472                         continue;
1473
1474                 table->de = NULL;
1475                 remove_proc_entry(table->procname, root);
1476         }
1477 }
1478
1479 static ssize_t do_rw_proc(int write, struct file * file, char __user * buf,
1480                           size_t count, loff_t *ppos)
1481 {
1482         int op;
1483         struct proc_dir_entry *de = PDE(file->f_dentry->d_inode);
1484         struct ctl_table *table;
1485         size_t res;
1486         ssize_t error = -ENOTDIR;
1487         
1488         spin_lock(&sysctl_lock);
1489         if (de && de->data && use_table(de->set)) {
1490                 /*
1491                  * at that point we know that sysctl was not unregistered
1492                  * and won't be until we finish
1493                  */
1494                 spin_unlock(&sysctl_lock);
1495                 table = (struct ctl_table *) de->data;
1496                 if (!table || !table->proc_handler)
1497                         goto out;
1498                 error = -EPERM;
1499                 op = (write ? 002 : 004);
1500                 if (ctl_perm(table, op))
1501                         goto out;
1502                 
1503                 /* careful: calling conventions are nasty here */
1504                 res = count;
1505                 error = (*table->proc_handler)(table, write, file,
1506                                                 buf, &res, ppos);
1507                 if (!error)
1508                         error = res;
1509         out:
1510                 spin_lock(&sysctl_lock);
1511                 unuse_table(de->set);
1512         }
1513         spin_unlock(&sysctl_lock);
1514         return error;
1515 }
1516
1517 static int proc_opensys(struct inode *inode, struct file *file)
1518 {
1519         if (file->f_mode & FMODE_WRITE) {
1520                 /*
1521                  * sysctl entries that are not writable,
1522                  * are _NOT_ writable, capabilities or not.
1523                  */
1524                 if (!(inode->i_mode & S_IWUSR))
1525                         return -EPERM;
1526         }
1527
1528         return 0;
1529 }
1530
1531 static ssize_t proc_readsys(struct file * file, char __user * buf,
1532                             size_t count, loff_t *ppos)
1533 {
1534         return do_rw_proc(0, file, buf, count, ppos);
1535 }
1536
1537 static ssize_t proc_writesys(struct file * file, const char __user * buf,
1538                              size_t count, loff_t *ppos)
1539 {
1540         return do_rw_proc(1, file, (char __user *) buf, count, ppos);
1541 }
1542
1543 /**
1544  * proc_dostring - read a string sysctl
1545  * @table: the sysctl table
1546  * @write: %TRUE if this is a write to the sysctl file
1547  * @filp: the file structure
1548  * @buffer: the user buffer
1549  * @lenp: the size of the user buffer
1550  * @ppos: file position
1551  *
1552  * Reads/writes a string from/to the user buffer. If the kernel
1553  * buffer provided is not large enough to hold the string, the
1554  * string is truncated. The copied string is %NULL-terminated.
1555  * If the string is being read by the user process, it is copied
1556  * and a newline '\n' is added. It is truncated if the buffer is
1557  * not large enough.
1558  *
1559  * Returns 0 on success.
1560  */
1561 int proc_dostring(ctl_table *table, int write, struct file *filp,
1562                   void __user *buffer, size_t *lenp, loff_t *ppos)
1563 {
1564         size_t len;
1565         char __user *p;
1566         char c;
1567         
1568         if (!table->data || !table->maxlen || !*lenp ||
1569             (*ppos && !write)) {
1570                 *lenp = 0;
1571                 return 0;
1572         }
1573         
1574         if (write) {
1575                 len = 0;
1576                 p = buffer;
1577                 while (len < *lenp) {
1578                         if (get_user(c, p++))
1579                                 return -EFAULT;
1580                         if (c == 0 || c == '\n')
1581                                 break;
1582                         len++;
1583                 }
1584                 if (len >= table->maxlen)
1585                         len = table->maxlen-1;
1586                 if(copy_from_user(table->data, buffer, len))
1587                         return -EFAULT;
1588                 ((char *) table->data)[len] = 0;
1589                 *ppos += *lenp;
1590         } else {
1591                 len = strlen(table->data);
1592                 if (len > table->maxlen)
1593                         len = table->maxlen;
1594                 if (len > *lenp)
1595                         len = *lenp;
1596                 if (len)
1597                         if(copy_to_user(buffer, table->data, len))
1598                                 return -EFAULT;
1599                 if (len < *lenp) {
1600                         if(put_user('\n', ((char __user *) buffer) + len))
1601                                 return -EFAULT;
1602                         len++;
1603                 }
1604                 *lenp = len;
1605                 *ppos += len;
1606         }
1607         return 0;
1608 }
1609
1610 /*
1611  *      Special case of dostring for the UTS structure. This has locks
1612  *      to observe. Should this be in kernel/sys.c ????
1613  */
1614  
1615 static int proc_doutsstring(ctl_table *table, int write, struct file *filp,
1616                   void __user *buffer, size_t *lenp, loff_t *ppos)
1617 {
1618         int r;
1619
1620         if (!write) {
1621                 down_read(&uts_sem);
1622                 r=proc_dostring(table,0,filp,buffer,lenp, ppos);
1623                 up_read(&uts_sem);
1624         } else {
1625                 down_write(&uts_sem);
1626                 r=proc_dostring(table,1,filp,buffer,lenp, ppos);
1627                 up_write(&uts_sem);
1628         }
1629         return r;
1630 }
1631
1632 static int do_proc_dointvec_conv(int *negp, unsigned long *lvalp,
1633                                  int *valp,
1634                                  int write, void *data)
1635 {
1636         if (write) {
1637                 *valp = *negp ? -*lvalp : *lvalp;
1638         } else {
1639                 int val = *valp;
1640                 if (val < 0) {
1641                         *negp = -1;
1642                         *lvalp = (unsigned long)-val;
1643                 } else {
1644                         *negp = 0;
1645                         *lvalp = (unsigned long)val;
1646                 }
1647         }
1648         return 0;
1649 }
1650
1651 static int do_proc_dointvec(ctl_table *table, int write, struct file *filp,
1652                   void __user *buffer, size_t *lenp, loff_t *ppos,
1653                   int (*conv)(int *negp, unsigned long *lvalp, int *valp,
1654                               int write, void *data),
1655                   void *data)
1656 {
1657 #define TMPBUFLEN 21
1658         int *i, vleft, first=1, neg, val;
1659         unsigned long lval;
1660         size_t left, len;
1661         
1662         char buf[TMPBUFLEN], *p;
1663         char __user *s = buffer;
1664         
1665         if (!table->data || !table->maxlen || !*lenp ||
1666             (*ppos && !write)) {
1667                 *lenp = 0;
1668                 return 0;
1669         }
1670         
1671         i = (int *) table->data;
1672         vleft = table->maxlen / sizeof(*i);
1673         left = *lenp;
1674
1675         if (!conv)
1676                 conv = do_proc_dointvec_conv;
1677
1678         for (; left && vleft--; i++, first=0) {
1679                 if (write) {
1680                         while (left) {
1681                                 char c;
1682                                 if (get_user(c, s))
1683                                         return -EFAULT;
1684                                 if (!isspace(c))
1685                                         break;
1686                                 left--;
1687                                 s++;
1688                         }
1689                         if (!left)
1690                                 break;
1691                         neg = 0;
1692                         len = left;
1693                         if (len > sizeof(buf) - 1)
1694                                 len = sizeof(buf) - 1;
1695                         if (copy_from_user(buf, s, len))
1696                                 return -EFAULT;
1697                         buf[len] = 0;
1698                         p = buf;
1699                         if (*p == '-' && left > 1) {
1700                                 neg = 1;
1701                                 left--, p++;
1702                         }
1703                         if (*p < '0' || *p > '9')
1704                                 break;
1705
1706                         lval = simple_strtoul(p, &p, 0);
1707
1708                         len = p-buf;
1709                         if ((len < left) && *p && !isspace(*p))
1710                                 break;
1711                         if (neg)
1712                                 val = -val;
1713                         s += len;
1714                         left -= len;
1715
1716                         if (conv(&neg, &lval, i, 1, data))
1717                                 break;
1718                 } else {
1719                         p = buf;
1720                         if (!first)
1721                                 *p++ = '\t';
1722         
1723                         if (conv(&neg, &lval, i, 0, data))
1724                                 break;
1725
1726                         sprintf(p, "%s%lu", neg ? "-" : "", lval);
1727                         len = strlen(buf);
1728                         if (len > left)
1729                                 len = left;
1730                         if(copy_to_user(s, buf, len))
1731                                 return -EFAULT;
1732                         left -= len;
1733                         s += len;
1734                 }
1735         }
1736
1737         if (!write && !first && left) {
1738                 if(put_user('\n', s))
1739                         return -EFAULT;
1740                 left--, s++;
1741         }
1742         if (write) {
1743                 while (left) {
1744                         char c;
1745                         if (get_user(c, s++))
1746                                 return -EFAULT;
1747                         if (!isspace(c))
1748                                 break;
1749                         left--;
1750                 }
1751         }
1752         if (write && first)
1753                 return -EINVAL;
1754         *lenp -= left;
1755         *ppos += *lenp;
1756         return 0;
1757 #undef TMPBUFLEN
1758 }
1759
1760 /**
1761  * proc_dointvec - read a vector of integers
1762  * @table: the sysctl table
1763  * @write: %TRUE if this is a write to the sysctl file
1764  * @filp: the file structure
1765  * @buffer: the user buffer
1766  * @lenp: the size of the user buffer
1767  * @ppos: file position
1768  *
1769  * Reads/writes up to table->maxlen/sizeof(unsigned int) integer
1770  * values from/to the user buffer, treated as an ASCII string. 
1771  *
1772  * Returns 0 on success.
1773  */
1774 int proc_dointvec(ctl_table *table, int write, struct file *filp,
1775                      void __user *buffer, size_t *lenp, loff_t *ppos)
1776 {
1777     return do_proc_dointvec(table,write,filp,buffer,lenp,ppos,
1778                             NULL,NULL);
1779 }
1780
1781 #define OP_SET  0
1782 #define OP_AND  1
1783 #define OP_OR   2
1784 #define OP_MAX  3
1785 #define OP_MIN  4
1786
1787 static int do_proc_dointvec_bset_conv(int *negp, unsigned long *lvalp,
1788                                       int *valp,
1789                                       int write, void *data)
1790 {
1791         int op = *(int *)data;
1792         if (write) {
1793                 int val = *negp ? -*lvalp : *lvalp;
1794                 switch(op) {
1795                 case OP_SET:    *valp = val; break;
1796                 case OP_AND:    *valp &= val; break;
1797                 case OP_OR:     *valp |= val; break;
1798                 case OP_MAX:    if(*valp < val)
1799                                         *valp = val;
1800                                 break;
1801                 case OP_MIN:    if(*valp > val)
1802                                 *valp = val;
1803                                 break;
1804                 }
1805         } else {
1806                 int val = *valp;
1807                 if (val < 0) {
1808                         *negp = -1;
1809                         *lvalp = (unsigned long)-val;
1810                 } else {
1811                         *negp = 0;
1812                         *lvalp = (unsigned long)val;
1813                 }
1814         }
1815         return 0;
1816 }
1817
1818 /*
1819  *      init may raise the set.
1820  */
1821  
1822 int proc_dointvec_bset(ctl_table *table, int write, struct file *filp,
1823                         void __user *buffer, size_t *lenp, loff_t *ppos)
1824 {
1825         int op;
1826
1827         if (!capable(CAP_SYS_MODULE)) {
1828                 return -EPERM;
1829         }
1830
1831         op = (current->pid == 1) ? OP_SET : OP_AND;
1832         return do_proc_dointvec(table,write,filp,buffer,lenp,ppos,
1833                                 do_proc_dointvec_bset_conv,&op);
1834 }
1835
1836 struct do_proc_dointvec_minmax_conv_param {
1837         int *min;
1838         int *max;
1839 };
1840
1841 static int do_proc_dointvec_minmax_conv(int *negp, unsigned long *lvalp, 
1842                                         int *valp, 
1843                                         int write, void *data)
1844 {
1845         struct do_proc_dointvec_minmax_conv_param *param = data;
1846         if (write) {
1847                 int val = *negp ? -*lvalp : *lvalp;
1848                 if ((param->min && *param->min > val) ||
1849                     (param->max && *param->max < val))
1850                         return -EINVAL;
1851                 *valp = val;
1852         } else {
1853                 int val = *valp;
1854                 if (val < 0) {
1855                         *negp = -1;
1856                         *lvalp = (unsigned long)-val;
1857                 } else {
1858                         *negp = 0;
1859                         *lvalp = (unsigned long)val;
1860                 }
1861         }
1862         return 0;
1863 }
1864
1865 /**
1866  * proc_dointvec_minmax - read a vector of integers with min/max values
1867  * @table: the sysctl table
1868  * @write: %TRUE if this is a write to the sysctl file
1869  * @filp: the file structure
1870  * @buffer: the user buffer
1871  * @lenp: the size of the user buffer
1872  * @ppos: file position
1873  *
1874  * Reads/writes up to table->maxlen/sizeof(unsigned int) integer
1875  * values from/to the user buffer, treated as an ASCII string.
1876  *
1877  * This routine will ensure the values are within the range specified by
1878  * table->extra1 (min) and table->extra2 (max).
1879  *
1880  * Returns 0 on success.
1881  */
1882 int proc_dointvec_minmax(ctl_table *table, int write, struct file *filp,
1883                   void __user *buffer, size_t *lenp, loff_t *ppos)
1884 {
1885         struct do_proc_dointvec_minmax_conv_param param = {
1886                 .min = (int *) table->extra1,
1887                 .max = (int *) table->extra2,
1888         };
1889         return do_proc_dointvec(table, write, filp, buffer, lenp, ppos,
1890                                 do_proc_dointvec_minmax_conv, &param);
1891 }
1892
1893 static int do_proc_doulongvec_minmax(ctl_table *table, int write,
1894                                      struct file *filp,
1895                                      void __user *buffer,
1896                                      size_t *lenp, loff_t *ppos,
1897                                      unsigned long convmul,
1898                                      unsigned long convdiv)
1899 {
1900 #define TMPBUFLEN 21
1901         unsigned long *i, *min, *max, val;
1902         int vleft, first=1, neg;
1903         size_t len, left;
1904         char buf[TMPBUFLEN], *p;
1905         char __user *s = buffer;
1906         
1907         if (!table->data || !table->maxlen || !*lenp ||
1908             (*ppos && !write)) {
1909                 *lenp = 0;
1910                 return 0;
1911         }
1912         
1913         i = (unsigned long *) table->data;
1914         min = (unsigned long *) table->extra1;
1915         max = (unsigned long *) table->extra2;
1916         vleft = table->maxlen / sizeof(unsigned long);
1917         left = *lenp;
1918         
1919         for (; left && vleft--; i++, min++, max++, first=0) {
1920                 if (write) {
1921                         while (left) {
1922                                 char c;
1923                                 if (get_user(c, s))
1924                                         return -EFAULT;
1925                                 if (!isspace(c))
1926                                         break;
1927                                 left--;
1928                                 s++;
1929                         }
1930                         if (!left)
1931                                 break;
1932                         neg = 0;
1933                         len = left;
1934                         if (len > TMPBUFLEN-1)
1935                                 len = TMPBUFLEN-1;
1936                         if (copy_from_user(buf, s, len))
1937                                 return -EFAULT;
1938                         buf[len] = 0;
1939                         p = buf;
1940                         if (*p == '-' && left > 1) {
1941                                 neg = 1;
1942                                 left--, p++;
1943                         }
1944                         if (*p < '0' || *p > '9')
1945                                 break;
1946                         val = simple_strtoul(p, &p, 0) * convmul / convdiv ;
1947                         len = p-buf;
1948                         if ((len < left) && *p && !isspace(*p))
1949                                 break;
1950                         if (neg)
1951                                 val = -val;
1952                         s += len;
1953                         left -= len;
1954
1955                         if(neg)
1956                                 continue;
1957                         if ((min && val < *min) || (max && val > *max))
1958                                 continue;
1959                         *i = val;
1960                 } else {
1961                         p = buf;
1962                         if (!first)
1963                                 *p++ = '\t';
1964                         sprintf(p, "%lu", convdiv * (*i) / convmul);
1965                         len = strlen(buf);
1966                         if (len > left)
1967                                 len = left;
1968                         if(copy_to_user(s, buf, len))
1969                                 return -EFAULT;
1970                         left -= len;
1971                         s += len;
1972                 }
1973         }
1974
1975         if (!write && !first && left) {
1976                 if(put_user('\n', s))
1977                         return -EFAULT;
1978                 left--, s++;
1979         }
1980         if (write) {
1981                 while (left) {
1982                         char c;
1983                         if (get_user(c, s++))
1984                                 return -EFAULT;
1985                         if (!isspace(c))
1986                                 break;
1987                         left--;
1988                 }
1989         }
1990         if (write && first)
1991                 return -EINVAL;
1992         *lenp -= left;
1993         *ppos += *lenp;
1994         return 0;
1995 #undef TMPBUFLEN
1996 }
1997
1998 /**
1999  * proc_doulongvec_minmax - read a vector of long integers with min/max values
2000  * @table: the sysctl table
2001  * @write: %TRUE if this is a write to the sysctl file
2002  * @filp: the file structure
2003  * @buffer: the user buffer
2004  * @lenp: the size of the user buffer
2005  * @ppos: file position
2006  *
2007  * Reads/writes up to table->maxlen/sizeof(unsigned long) unsigned long
2008  * values from/to the user buffer, treated as an ASCII string.
2009  *
2010  * This routine will ensure the values are within the range specified by
2011  * table->extra1 (min) and table->extra2 (max).
2012  *
2013  * Returns 0 on success.
2014  */
2015 int proc_doulongvec_minmax(ctl_table *table, int write, struct file *filp,
2016                            void __user *buffer, size_t *lenp, loff_t *ppos)
2017 {
2018     return do_proc_doulongvec_minmax(table, write, filp, buffer, lenp, ppos, 1l, 1l);
2019 }
2020
2021 /**
2022  * proc_doulongvec_ms_jiffies_minmax - read a vector of millisecond values with min/max values
2023  * @table: the sysctl table
2024  * @write: %TRUE if this is a write to the sysctl file
2025  * @filp: the file structure
2026  * @buffer: the user buffer
2027  * @lenp: the size of the user buffer
2028  * @ppos: file position
2029  *
2030  * Reads/writes up to table->maxlen/sizeof(unsigned long) unsigned long
2031  * values from/to the user buffer, treated as an ASCII string. The values
2032  * are treated as milliseconds, and converted to jiffies when they are stored.
2033  *
2034  * This routine will ensure the values are within the range specified by
2035  * table->extra1 (min) and table->extra2 (max).
2036  *
2037  * Returns 0 on success.
2038  */
2039 int proc_doulongvec_ms_jiffies_minmax(ctl_table *table, int write,
2040                                       struct file *filp,
2041                                       void __user *buffer,
2042                                       size_t *lenp, loff_t *ppos)
2043 {
2044     return do_proc_doulongvec_minmax(table, write, filp, buffer,
2045                                      lenp, ppos, HZ, 1000l);
2046 }
2047
2048
2049 static int do_proc_dointvec_jiffies_conv(int *negp, unsigned long *lvalp,
2050                                          int *valp,
2051                                          int write, void *data)
2052 {
2053         if (write) {
2054                 *valp = *negp ? -(*lvalp*HZ) : (*lvalp*HZ);
2055         } else {
2056                 int val = *valp;
2057                 unsigned long lval;
2058                 if (val < 0) {
2059                         *negp = -1;
2060                         lval = (unsigned long)-val;
2061                 } else {
2062                         *negp = 0;
2063                         lval = (unsigned long)val;
2064                 }
2065                 *lvalp = lval / HZ;
2066         }
2067         return 0;
2068 }
2069
2070 static int do_proc_dointvec_userhz_jiffies_conv(int *negp, unsigned long *lvalp,
2071                                                 int *valp,
2072                                                 int write, void *data)
2073 {
2074         if (write) {
2075                 *valp = clock_t_to_jiffies(*negp ? -*lvalp : *lvalp);
2076         } else {
2077                 int val = *valp;
2078                 unsigned long lval;
2079                 if (val < 0) {
2080                         *negp = -1;
2081                         lval = (unsigned long)-val;
2082                 } else {
2083                         *negp = 0;
2084                         lval = (unsigned long)val;
2085                 }
2086                 *lvalp = jiffies_to_clock_t(lval);
2087         }
2088         return 0;
2089 }
2090
2091 static int do_proc_dointvec_ms_jiffies_conv(int *negp, unsigned long *lvalp,
2092                                             int *valp,
2093                                             int write, void *data)
2094 {
2095         if (write) {
2096                 *valp = msecs_to_jiffies(*negp ? -*lvalp : *lvalp);
2097         } else {
2098                 int val = *valp;
2099                 unsigned long lval;
2100                 if (val < 0) {
2101                         *negp = -1;
2102                         lval = (unsigned long)-val;
2103                 } else {
2104                         *negp = 0;
2105                         lval = (unsigned long)val;
2106                 }
2107                 *lvalp = jiffies_to_msecs(lval);
2108         }
2109         return 0;
2110 }
2111
2112 /**
2113  * proc_dointvec_jiffies - read a vector of integers as seconds
2114  * @table: the sysctl table
2115  * @write: %TRUE if this is a write to the sysctl file
2116  * @filp: the file structure
2117  * @buffer: the user buffer
2118  * @lenp: the size of the user buffer
2119  * @ppos: file position
2120  *
2121  * Reads/writes up to table->maxlen/sizeof(unsigned int) integer
2122  * values from/to the user buffer, treated as an ASCII string. 
2123  * The values read are assumed to be in seconds, and are converted into
2124  * jiffies.
2125  *
2126  * Returns 0 on success.
2127  */
2128 int proc_dointvec_jiffies(ctl_table *table, int write, struct file *filp,
2129                           void __user *buffer, size_t *lenp, loff_t *ppos)
2130 {
2131     return do_proc_dointvec(table,write,filp,buffer,lenp,ppos,
2132                             do_proc_dointvec_jiffies_conv,NULL);
2133 }
2134
2135 /**
2136  * proc_dointvec_userhz_jiffies - read a vector of integers as 1/USER_HZ seconds
2137  * @table: the sysctl table
2138  * @write: %TRUE if this is a write to the sysctl file
2139  * @filp: the file structure
2140  * @buffer: the user buffer
2141  * @lenp: the size of the user buffer
2142  * @ppos: pointer to the file position
2143  *
2144  * Reads/writes up to table->maxlen/sizeof(unsigned int) integer
2145  * values from/to the user buffer, treated as an ASCII string. 
2146  * The values read are assumed to be in 1/USER_HZ seconds, and 
2147  * are converted into jiffies.
2148  *
2149  * Returns 0 on success.
2150  */
2151 int proc_dointvec_userhz_jiffies(ctl_table *table, int write, struct file *filp,
2152                                  void __user *buffer, size_t *lenp, loff_t *ppos)
2153 {
2154     return do_proc_dointvec(table,write,filp,buffer,lenp,ppos,
2155                             do_proc_dointvec_userhz_jiffies_conv,NULL);
2156 }
2157
2158 /**
2159  * proc_dointvec_ms_jiffies - read a vector of integers as 1 milliseconds
2160  * @table: the sysctl table
2161  * @write: %TRUE if this is a write to the sysctl file
2162  * @filp: the file structure
2163  * @buffer: the user buffer
2164  * @lenp: the size of the user buffer
2165  * @ppos: file position
2166  * @ppos: the current position in the file
2167  *
2168  * Reads/writes up to table->maxlen/sizeof(unsigned int) integer
2169  * values from/to the user buffer, treated as an ASCII string. 
2170  * The values read are assumed to be in 1/1000 seconds, and 
2171  * are converted into jiffies.
2172  *
2173  * Returns 0 on success.
2174  */
2175 int proc_dointvec_ms_jiffies(ctl_table *table, int write, struct file *filp,
2176                              void __user *buffer, size_t *lenp, loff_t *ppos)
2177 {
2178         return do_proc_dointvec(table, write, filp, buffer, lenp, ppos,
2179                                 do_proc_dointvec_ms_jiffies_conv, NULL);
2180 }
2181
2182 #else /* CONFIG_PROC_FS */
2183
2184 int proc_dostring(ctl_table *table, int write, struct file *filp,
2185                   void __user *buffer, size_t *lenp, loff_t *ppos)
2186 {
2187         return -ENOSYS;
2188 }
2189
2190 static int proc_doutsstring(ctl_table *table, int write, struct file *filp,
2191                             void __user *buffer, size_t *lenp, loff_t *ppos)
2192 {
2193         return -ENOSYS;
2194 }
2195
2196 int proc_dointvec(ctl_table *table, int write, struct file *filp,
2197                   void __user *buffer, size_t *lenp, loff_t *ppos)
2198 {
2199         return -ENOSYS;
2200 }
2201
2202 int proc_dointvec_bset(ctl_table *table, int write, struct file *filp,
2203                         void __user *buffer, size_t *lenp, loff_t *ppos)
2204 {
2205         return -ENOSYS;
2206 }
2207
2208 int proc_dointvec_minmax(ctl_table *table, int write, struct file *filp,
2209                     void __user *buffer, size_t *lenp, loff_t *ppos)
2210 {
2211         return -ENOSYS;
2212 }
2213
2214 int proc_dointvec_jiffies(ctl_table *table, int write, struct file *filp,
2215                     void __user *buffer, size_t *lenp, loff_t *ppos)
2216 {
2217         return -ENOSYS;
2218 }
2219
2220 int proc_dointvec_userhz_jiffies(ctl_table *table, int write, struct file *filp,
2221                     void __user *buffer, size_t *lenp, loff_t *ppos)
2222 {
2223         return -ENOSYS;
2224 }
2225
2226 int proc_dointvec_ms_jiffies(ctl_table *table, int write, struct file *filp,
2227                              void __user *buffer, size_t *lenp, loff_t *ppos)
2228 {
2229         return -ENOSYS;
2230 }
2231
2232 int proc_doulongvec_minmax(ctl_table *table, int write, struct file *filp,
2233                     void __user *buffer, size_t *lenp, loff_t *ppos)
2234 {
2235         return -ENOSYS;
2236 }
2237
2238 int proc_doulongvec_ms_jiffies_minmax(ctl_table *table, int write,
2239                                       struct file *filp,
2240                                       void __user *buffer,
2241                                       size_t *lenp, loff_t *ppos)
2242 {
2243     return -ENOSYS;
2244 }
2245
2246
2247 #endif /* CONFIG_PROC_FS */
2248
2249
2250 /*
2251  * General sysctl support routines 
2252  */
2253
2254 /* The generic string strategy routine: */
2255 int sysctl_string(ctl_table *table, int __user *name, int nlen,
2256                   void __user *oldval, size_t __user *oldlenp,
2257                   void __user *newval, size_t newlen, void **context)
2258 {
2259         if (!table->data || !table->maxlen) 
2260                 return -ENOTDIR;
2261         
2262         if (oldval && oldlenp) {
2263                 size_t bufsize;
2264                 if (get_user(bufsize, oldlenp))
2265                         return -EFAULT;
2266                 if (bufsize) {
2267                         size_t len = strlen(table->data), copied;
2268
2269                         /* This shouldn't trigger for a well-formed sysctl */
2270                         if (len > table->maxlen)
2271                                 len = table->maxlen;
2272
2273                         /* Copy up to a max of bufsize-1 bytes of the string */
2274                         copied = (len >= bufsize) ? bufsize - 1 : len;
2275
2276                         if (copy_to_user(oldval, table->data, copied) ||
2277                             put_user(0, (char __user *)(oldval + copied)))
2278                                 return -EFAULT;
2279                         if (put_user(len, oldlenp))
2280                                 return -EFAULT;
2281                 }
2282         }
2283         if (newval && newlen) {
2284                 size_t len = newlen;
2285                 if (len > table->maxlen)
2286                         len = table->maxlen;
2287                 if(copy_from_user(table->data, newval, len))
2288                         return -EFAULT;
2289                 if (len == table->maxlen)
2290                         len--;
2291                 ((char *) table->data)[len] = 0;
2292         }
2293         return 1;
2294 }
2295
2296 /*
2297  * This function makes sure that all of the integers in the vector
2298  * are between the minimum and maximum values given in the arrays
2299  * table->extra1 and table->extra2, respectively.
2300  */
2301 int sysctl_intvec(ctl_table *table, int __user *name, int nlen,
2302                 void __user *oldval, size_t __user *oldlenp,
2303                 void __user *newval, size_t newlen, void **context)
2304 {
2305
2306         if (newval && newlen) {
2307                 int __user *vec = (int __user *) newval;
2308                 int *min = (int *) table->extra1;
2309                 int *max = (int *) table->extra2;
2310                 size_t length;
2311                 int i;
2312
2313                 if (newlen % sizeof(int) != 0)
2314                         return -EINVAL;
2315
2316                 if (!table->extra1 && !table->extra2)
2317                         return 0;
2318
2319                 if (newlen > table->maxlen)
2320                         newlen = table->maxlen;
2321                 length = newlen / sizeof(int);
2322
2323                 for (i = 0; i < length; i++) {
2324                         int value;
2325                         if (get_user(value, vec + i))
2326                                 return -EFAULT;
2327                         if (min && value < min[i])
2328                                 return -EINVAL;
2329                         if (max && value > max[i])
2330                                 return -EINVAL;
2331                 }
2332         }
2333         return 0;
2334 }
2335
2336 /* Strategy function to convert jiffies to seconds */ 
2337 int sysctl_jiffies(ctl_table *table, int __user *name, int nlen,
2338                 void __user *oldval, size_t __user *oldlenp,
2339                 void __user *newval, size_t newlen, void **context)
2340 {
2341         if (oldval) {
2342                 size_t olen;
2343                 if (oldlenp) { 
2344                         if (get_user(olen, oldlenp))
2345                                 return -EFAULT;
2346                         if (olen!=sizeof(int))
2347                                 return -EINVAL; 
2348                 }
2349                 if (put_user(*(int *)(table->data)/HZ, (int __user *)oldval) ||
2350                     (oldlenp && put_user(sizeof(int),oldlenp)))
2351                         return -EFAULT;
2352         }
2353         if (newval && newlen) { 
2354                 int new;
2355                 if (newlen != sizeof(int))
2356                         return -EINVAL; 
2357                 if (get_user(new, (int __user *)newval))
2358                         return -EFAULT;
2359                 *(int *)(table->data) = new*HZ; 
2360         }
2361         return 1;
2362 }
2363
2364 /* Strategy function to convert jiffies to seconds */ 
2365 int sysctl_ms_jiffies(ctl_table *table, int __user *name, int nlen,
2366                 void __user *oldval, size_t __user *oldlenp,
2367                 void __user *newval, size_t newlen, void **context)
2368 {
2369         if (oldval) {
2370                 size_t olen;
2371                 if (oldlenp) { 
2372                         if (get_user(olen, oldlenp))
2373                                 return -EFAULT;
2374                         if (olen!=sizeof(int))
2375                                 return -EINVAL; 
2376                 }
2377                 if (put_user(jiffies_to_msecs(*(int *)(table->data)), (int __user *)oldval) ||
2378                     (oldlenp && put_user(sizeof(int),oldlenp)))
2379                         return -EFAULT;
2380         }
2381         if (newval && newlen) { 
2382                 int new;
2383                 if (newlen != sizeof(int))
2384                         return -EINVAL; 
2385                 if (get_user(new, (int __user *)newval))
2386                         return -EFAULT;
2387                 *(int *)(table->data) = msecs_to_jiffies(new);
2388         }
2389         return 1;
2390 }
2391
2392 #else /* CONFIG_SYSCTL */
2393
2394
2395 asmlinkage long sys_sysctl(struct __sysctl_args __user *args)
2396 {
2397         return -ENOSYS;
2398 }
2399
2400 int sysctl_string(ctl_table *table, int __user *name, int nlen,
2401                   void __user *oldval, size_t __user *oldlenp,
2402                   void __user *newval, size_t newlen, void **context)
2403 {
2404         return -ENOSYS;
2405 }
2406
2407 int sysctl_intvec(ctl_table *table, int __user *name, int nlen,
2408                 void __user *oldval, size_t __user *oldlenp,
2409                 void __user *newval, size_t newlen, void **context)
2410 {
2411         return -ENOSYS;
2412 }
2413
2414 int sysctl_jiffies(ctl_table *table, int __user *name, int nlen,
2415                 void __user *oldval, size_t __user *oldlenp,
2416                 void __user *newval, size_t newlen, void **context)
2417 {
2418         return -ENOSYS;
2419 }
2420
2421 int sysctl_ms_jiffies(ctl_table *table, int __user *name, int nlen,
2422                 void __user *oldval, size_t __user *oldlenp,
2423                 void __user *newval, size_t newlen, void **context)
2424 {
2425         return -ENOSYS;
2426 }
2427
2428 int proc_dostring(ctl_table *table, int write, struct file *filp,
2429                   void __user *buffer, size_t *lenp, loff_t *ppos)
2430 {
2431         return -ENOSYS;
2432 }
2433
2434 int proc_dointvec(ctl_table *table, int write, struct file *filp,
2435                   void __user *buffer, size_t *lenp, loff_t *ppos)
2436 {
2437         return -ENOSYS;
2438 }
2439
2440 int proc_dointvec_bset(ctl_table *table, int write, struct file *filp,
2441                         void __user *buffer, size_t *lenp, loff_t *ppos)
2442 {
2443         return -ENOSYS;
2444 }
2445
2446 int proc_dointvec_minmax(ctl_table *table, int write, struct file *filp,
2447                     void __user *buffer, size_t *lenp, loff_t *ppos)
2448 {
2449         return -ENOSYS;
2450 }
2451
2452 int proc_dointvec_jiffies(ctl_table *table, int write, struct file *filp,
2453                           void __user *buffer, size_t *lenp, loff_t *ppos)
2454 {
2455         return -ENOSYS;
2456 }
2457
2458 int proc_dointvec_userhz_jiffies(ctl_table *table, int write, struct file *filp,
2459                           void __user *buffer, size_t *lenp, loff_t *ppos)
2460 {
2461         return -ENOSYS;
2462 }
2463
2464 int proc_dointvec_ms_jiffies(ctl_table *table, int write, struct file *filp,
2465                              void __user *buffer, size_t *lenp, loff_t *ppos)
2466 {
2467         return -ENOSYS;
2468 }
2469
2470 int proc_doulongvec_minmax(ctl_table *table, int write, struct file *filp,
2471                     void __user *buffer, size_t *lenp, loff_t *ppos)
2472 {
2473         return -ENOSYS;
2474 }
2475
2476 int proc_doulongvec_ms_jiffies_minmax(ctl_table *table, int write,
2477                                       struct file *filp,
2478                                       void __user *buffer,
2479                                       size_t *lenp, loff_t *ppos)
2480 {
2481     return -ENOSYS;
2482 }
2483
2484 struct ctl_table_header * register_sysctl_table(ctl_table * table, 
2485                                                 int insert_at_head)
2486 {
2487         return NULL;
2488 }
2489
2490 void unregister_sysctl_table(struct ctl_table_header * table)
2491 {
2492 }
2493
2494 #endif /* CONFIG_SYSCTL */
2495
2496 /*
2497  * No sense putting this after each symbol definition, twice,
2498  * exception granted :-)
2499  */
2500 EXPORT_SYMBOL(proc_dointvec);
2501 EXPORT_SYMBOL(proc_dointvec_jiffies);
2502 EXPORT_SYMBOL(proc_dointvec_minmax);
2503 EXPORT_SYMBOL(proc_dointvec_userhz_jiffies);
2504 EXPORT_SYMBOL(proc_dointvec_ms_jiffies);
2505 EXPORT_SYMBOL(proc_dostring);
2506 EXPORT_SYMBOL(proc_doulongvec_minmax);
2507 EXPORT_SYMBOL(proc_doulongvec_ms_jiffies_minmax);
2508 EXPORT_SYMBOL(register_sysctl_table);
2509 EXPORT_SYMBOL(sysctl_intvec);
2510 EXPORT_SYMBOL(sysctl_jiffies);
2511 EXPORT_SYMBOL(sysctl_ms_jiffies);
2512 EXPORT_SYMBOL(sysctl_string);
2513 EXPORT_SYMBOL(unregister_sysctl_table);