[PATCH] Support piping into commands in /proc/sys/kernel/core_pattern
[linux-2.6.git] / kernel / sysctl.c
1 /*
2  * sysctl.c: General linux system control interface
3  *
4  * Begun 24 March 1995, Stephen Tweedie
5  * Added /proc support, Dec 1995
6  * Added bdflush entry and intvec min/max checking, 2/23/96, Tom Dyas.
7  * Added hooks for /proc/sys/net (minor, minor patch), 96/4/1, Mike Shaver.
8  * Added kernel/java-{interpreter,appletviewer}, 96/5/10, Mike Shaver.
9  * Dynamic registration fixes, Stephen Tweedie.
10  * Added kswapd-interval, ctrl-alt-del, printk stuff, 1/8/97, Chris Horn.
11  * Made sysctl support optional via CONFIG_SYSCTL, 1/10/97, Chris
12  *  Horn.
13  * Added proc_doulongvec_ms_jiffies_minmax, 09/08/99, Carlos H. Bauer.
14  * Added proc_doulongvec_minmax, 09/08/99, Carlos H. Bauer.
15  * Changed linked lists to use list.h instead of lists.h, 02/24/00, Bill
16  *  Wendling.
17  * The list_for_each() macro wasn't appropriate for the sysctl loop.
18  *  Removed it and replaced it with older style, 03/23/00, Bill Wendling
19  */
20
21 #include <linux/module.h>
22 #include <linux/mm.h>
23 #include <linux/swap.h>
24 #include <linux/slab.h>
25 #include <linux/sysctl.h>
26 #include <linux/proc_fs.h>
27 #include <linux/capability.h>
28 #include <linux/ctype.h>
29 #include <linux/utsname.h>
30 #include <linux/capability.h>
31 #include <linux/smp_lock.h>
32 #include <linux/init.h>
33 #include <linux/kernel.h>
34 #include <linux/kobject.h>
35 #include <linux/net.h>
36 #include <linux/sysrq.h>
37 #include <linux/highuid.h>
38 #include <linux/writeback.h>
39 #include <linux/hugetlb.h>
40 #include <linux/security.h>
41 #include <linux/initrd.h>
42 #include <linux/times.h>
43 #include <linux/limits.h>
44 #include <linux/dcache.h>
45 #include <linux/syscalls.h>
46 #include <linux/nfs_fs.h>
47 #include <linux/acpi.h>
48
49 #include <asm/uaccess.h>
50 #include <asm/processor.h>
51
52 extern int proc_nr_files(ctl_table *table, int write, struct file *filp,
53                      void __user *buffer, size_t *lenp, loff_t *ppos);
54
55 #ifdef CONFIG_X86
56 #include <asm/nmi.h>
57 #endif
58
59 #if defined(CONFIG_SYSCTL)
60
61 /* External variables not in a header file. */
62 extern int C_A_D;
63 extern int sysctl_overcommit_memory;
64 extern int sysctl_overcommit_ratio;
65 extern int sysctl_panic_on_oom;
66 extern int max_threads;
67 extern int sysrq_enabled;
68 extern int core_uses_pid;
69 extern int suid_dumpable;
70 extern char core_pattern[];
71 extern int cad_pid;
72 extern int pid_max;
73 extern int min_free_kbytes;
74 extern int printk_ratelimit_jiffies;
75 extern int printk_ratelimit_burst;
76 extern int pid_max_min, pid_max_max;
77 extern int sysctl_drop_caches;
78 extern int percpu_pagelist_fraction;
79 extern int compat_log;
80
81 /* this is needed for the proc_dointvec_minmax for [fs_]overflow UID and GID */
82 static int maxolduid = 65535;
83 static int minolduid;
84 static int min_percpu_pagelist_fract = 8;
85
86 static int ngroups_max = NGROUPS_MAX;
87
88 #ifdef CONFIG_KMOD
89 extern char modprobe_path[];
90 #endif
91 #ifdef CONFIG_CHR_DEV_SG
92 extern int sg_big_buff;
93 #endif
94 #ifdef CONFIG_SYSVIPC
95 extern size_t shm_ctlmax;
96 extern size_t shm_ctlall;
97 extern int shm_ctlmni;
98 extern int msg_ctlmax;
99 extern int msg_ctlmnb;
100 extern int msg_ctlmni;
101 extern int sem_ctls[];
102 #endif
103
104 #ifdef __sparc__
105 extern char reboot_command [];
106 extern int stop_a_enabled;
107 extern int scons_pwroff;
108 #endif
109
110 #ifdef __hppa__
111 extern int pwrsw_enabled;
112 extern int unaligned_enabled;
113 #endif
114
115 #ifdef CONFIG_S390
116 #ifdef CONFIG_MATHEMU
117 extern int sysctl_ieee_emulation_warnings;
118 #endif
119 extern int sysctl_userprocess_debug;
120 extern int spin_retry;
121 #endif
122
123 extern int sysctl_hz_timer;
124
125 #ifdef CONFIG_BSD_PROCESS_ACCT
126 extern int acct_parm[];
127 #endif
128
129 #ifdef CONFIG_IA64
130 extern int no_unaligned_warning;
131 #endif
132
133 #ifdef CONFIG_RT_MUTEXES
134 extern int max_lock_depth;
135 #endif
136
137 #ifdef CONFIG_SYSCTL_SYSCALL
138 static int parse_table(int __user *, int, void __user *, size_t __user *,
139                 void __user *, size_t, ctl_table *, void **);
140 #endif
141
142 static int proc_doutsstring(ctl_table *table, int write, struct file *filp,
143                   void __user *buffer, size_t *lenp, loff_t *ppos);
144
145 static ctl_table root_table[];
146 static struct ctl_table_header root_table_header =
147         { root_table, LIST_HEAD_INIT(root_table_header.ctl_entry) };
148
149 static ctl_table kern_table[];
150 static ctl_table vm_table[];
151 static ctl_table fs_table[];
152 static ctl_table debug_table[];
153 static ctl_table dev_table[];
154 extern ctl_table random_table[];
155 #ifdef CONFIG_UNIX98_PTYS
156 extern ctl_table pty_table[];
157 #endif
158 #ifdef CONFIG_INOTIFY_USER
159 extern ctl_table inotify_table[];
160 #endif
161
162 #ifdef HAVE_ARCH_PICK_MMAP_LAYOUT
163 int sysctl_legacy_va_layout;
164 #endif
165
166 /* /proc declarations: */
167
168 #ifdef CONFIG_PROC_SYSCTL
169
170 static ssize_t proc_readsys(struct file *, char __user *, size_t, loff_t *);
171 static ssize_t proc_writesys(struct file *, const char __user *, size_t, loff_t *);
172 static int proc_opensys(struct inode *, struct file *);
173
174 struct file_operations proc_sys_file_operations = {
175         .open           = proc_opensys,
176         .read           = proc_readsys,
177         .write          = proc_writesys,
178 };
179
180 extern struct proc_dir_entry *proc_sys_root;
181
182 static void register_proc_table(ctl_table *, struct proc_dir_entry *, void *);
183 static void unregister_proc_table(ctl_table *, struct proc_dir_entry *);
184 #endif
185
186 /* The default sysctl tables: */
187
188 static ctl_table root_table[] = {
189         {
190                 .ctl_name       = CTL_KERN,
191                 .procname       = "kernel",
192                 .mode           = 0555,
193                 .child          = kern_table,
194         },
195         {
196                 .ctl_name       = CTL_VM,
197                 .procname       = "vm",
198                 .mode           = 0555,
199                 .child          = vm_table,
200         },
201 #ifdef CONFIG_NET
202         {
203                 .ctl_name       = CTL_NET,
204                 .procname       = "net",
205                 .mode           = 0555,
206                 .child          = net_table,
207         },
208 #endif
209         {
210                 .ctl_name       = CTL_FS,
211                 .procname       = "fs",
212                 .mode           = 0555,
213                 .child          = fs_table,
214         },
215         {
216                 .ctl_name       = CTL_DEBUG,
217                 .procname       = "debug",
218                 .mode           = 0555,
219                 .child          = debug_table,
220         },
221         {
222                 .ctl_name       = CTL_DEV,
223                 .procname       = "dev",
224                 .mode           = 0555,
225                 .child          = dev_table,
226         },
227
228         { .ctl_name = 0 }
229 };
230
231 static ctl_table kern_table[] = {
232         {
233                 .ctl_name       = KERN_OSTYPE,
234                 .procname       = "ostype",
235                 .data           = system_utsname.sysname,
236                 .maxlen         = sizeof(system_utsname.sysname),
237                 .mode           = 0444,
238                 .proc_handler   = &proc_doutsstring,
239                 .strategy       = &sysctl_string,
240         },
241         {
242                 .ctl_name       = KERN_OSRELEASE,
243                 .procname       = "osrelease",
244                 .data           = system_utsname.release,
245                 .maxlen         = sizeof(system_utsname.release),
246                 .mode           = 0444,
247                 .proc_handler   = &proc_doutsstring,
248                 .strategy       = &sysctl_string,
249         },
250         {
251                 .ctl_name       = KERN_VERSION,
252                 .procname       = "version",
253                 .data           = system_utsname.version,
254                 .maxlen         = sizeof(system_utsname.version),
255                 .mode           = 0444,
256                 .proc_handler   = &proc_doutsstring,
257                 .strategy       = &sysctl_string,
258         },
259         {
260                 .ctl_name       = KERN_NODENAME,
261                 .procname       = "hostname",
262                 .data           = system_utsname.nodename,
263                 .maxlen         = sizeof(system_utsname.nodename),
264                 .mode           = 0644,
265                 .proc_handler   = &proc_doutsstring,
266                 .strategy       = &sysctl_string,
267         },
268         {
269                 .ctl_name       = KERN_DOMAINNAME,
270                 .procname       = "domainname",
271                 .data           = system_utsname.domainname,
272                 .maxlen         = sizeof(system_utsname.domainname),
273                 .mode           = 0644,
274                 .proc_handler   = &proc_doutsstring,
275                 .strategy       = &sysctl_string,
276         },
277         {
278                 .ctl_name       = KERN_PANIC,
279                 .procname       = "panic",
280                 .data           = &panic_timeout,
281                 .maxlen         = sizeof(int),
282                 .mode           = 0644,
283                 .proc_handler   = &proc_dointvec,
284         },
285         {
286                 .ctl_name       = KERN_CORE_USES_PID,
287                 .procname       = "core_uses_pid",
288                 .data           = &core_uses_pid,
289                 .maxlen         = sizeof(int),
290                 .mode           = 0644,
291                 .proc_handler   = &proc_dointvec,
292         },
293         {
294                 .ctl_name       = KERN_CORE_PATTERN,
295                 .procname       = "core_pattern",
296                 .data           = core_pattern,
297                 .maxlen         = 128,
298                 .mode           = 0644,
299                 .proc_handler   = &proc_dostring,
300                 .strategy       = &sysctl_string,
301         },
302         {
303                 .ctl_name       = KERN_TAINTED,
304                 .procname       = "tainted",
305                 .data           = &tainted,
306                 .maxlen         = sizeof(int),
307                 .mode           = 0444,
308                 .proc_handler   = &proc_dointvec,
309         },
310         {
311                 .ctl_name       = KERN_CAP_BSET,
312                 .procname       = "cap-bound",
313                 .data           = &cap_bset,
314                 .maxlen         = sizeof(kernel_cap_t),
315                 .mode           = 0600,
316                 .proc_handler   = &proc_dointvec_bset,
317         },
318 #ifdef CONFIG_BLK_DEV_INITRD
319         {
320                 .ctl_name       = KERN_REALROOTDEV,
321                 .procname       = "real-root-dev",
322                 .data           = &real_root_dev,
323                 .maxlen         = sizeof(int),
324                 .mode           = 0644,
325                 .proc_handler   = &proc_dointvec,
326         },
327 #endif
328 #ifdef __sparc__
329         {
330                 .ctl_name       = KERN_SPARC_REBOOT,
331                 .procname       = "reboot-cmd",
332                 .data           = reboot_command,
333                 .maxlen         = 256,
334                 .mode           = 0644,
335                 .proc_handler   = &proc_dostring,
336                 .strategy       = &sysctl_string,
337         },
338         {
339                 .ctl_name       = KERN_SPARC_STOP_A,
340                 .procname       = "stop-a",
341                 .data           = &stop_a_enabled,
342                 .maxlen         = sizeof (int),
343                 .mode           = 0644,
344                 .proc_handler   = &proc_dointvec,
345         },
346         {
347                 .ctl_name       = KERN_SPARC_SCONS_PWROFF,
348                 .procname       = "scons-poweroff",
349                 .data           = &scons_pwroff,
350                 .maxlen         = sizeof (int),
351                 .mode           = 0644,
352                 .proc_handler   = &proc_dointvec,
353         },
354 #endif
355 #ifdef __hppa__
356         {
357                 .ctl_name       = KERN_HPPA_PWRSW,
358                 .procname       = "soft-power",
359                 .data           = &pwrsw_enabled,
360                 .maxlen         = sizeof (int),
361                 .mode           = 0644,
362                 .proc_handler   = &proc_dointvec,
363         },
364         {
365                 .ctl_name       = KERN_HPPA_UNALIGNED,
366                 .procname       = "unaligned-trap",
367                 .data           = &unaligned_enabled,
368                 .maxlen         = sizeof (int),
369                 .mode           = 0644,
370                 .proc_handler   = &proc_dointvec,
371         },
372 #endif
373         {
374                 .ctl_name       = KERN_CTLALTDEL,
375                 .procname       = "ctrl-alt-del",
376                 .data           = &C_A_D,
377                 .maxlen         = sizeof(int),
378                 .mode           = 0644,
379                 .proc_handler   = &proc_dointvec,
380         },
381         {
382                 .ctl_name       = KERN_PRINTK,
383                 .procname       = "printk",
384                 .data           = &console_loglevel,
385                 .maxlen         = 4*sizeof(int),
386                 .mode           = 0644,
387                 .proc_handler   = &proc_dointvec,
388         },
389 #ifdef CONFIG_KMOD
390         {
391                 .ctl_name       = KERN_MODPROBE,
392                 .procname       = "modprobe",
393                 .data           = &modprobe_path,
394                 .maxlen         = KMOD_PATH_LEN,
395                 .mode           = 0644,
396                 .proc_handler   = &proc_dostring,
397                 .strategy       = &sysctl_string,
398         },
399 #endif
400 #if defined(CONFIG_HOTPLUG) && defined(CONFIG_NET)
401         {
402                 .ctl_name       = KERN_HOTPLUG,
403                 .procname       = "hotplug",
404                 .data           = &uevent_helper,
405                 .maxlen         = UEVENT_HELPER_PATH_LEN,
406                 .mode           = 0644,
407                 .proc_handler   = &proc_dostring,
408                 .strategy       = &sysctl_string,
409         },
410 #endif
411 #ifdef CONFIG_CHR_DEV_SG
412         {
413                 .ctl_name       = KERN_SG_BIG_BUFF,
414                 .procname       = "sg-big-buff",
415                 .data           = &sg_big_buff,
416                 .maxlen         = sizeof (int),
417                 .mode           = 0444,
418                 .proc_handler   = &proc_dointvec,
419         },
420 #endif
421 #ifdef CONFIG_BSD_PROCESS_ACCT
422         {
423                 .ctl_name       = KERN_ACCT,
424                 .procname       = "acct",
425                 .data           = &acct_parm,
426                 .maxlen         = 3*sizeof(int),
427                 .mode           = 0644,
428                 .proc_handler   = &proc_dointvec,
429         },
430 #endif
431 #ifdef CONFIG_SYSVIPC
432         {
433                 .ctl_name       = KERN_SHMMAX,
434                 .procname       = "shmmax",
435                 .data           = &shm_ctlmax,
436                 .maxlen         = sizeof (size_t),
437                 .mode           = 0644,
438                 .proc_handler   = &proc_doulongvec_minmax,
439         },
440         {
441                 .ctl_name       = KERN_SHMALL,
442                 .procname       = "shmall",
443                 .data           = &shm_ctlall,
444                 .maxlen         = sizeof (size_t),
445                 .mode           = 0644,
446                 .proc_handler   = &proc_doulongvec_minmax,
447         },
448         {
449                 .ctl_name       = KERN_SHMMNI,
450                 .procname       = "shmmni",
451                 .data           = &shm_ctlmni,
452                 .maxlen         = sizeof (int),
453                 .mode           = 0644,
454                 .proc_handler   = &proc_dointvec,
455         },
456         {
457                 .ctl_name       = KERN_MSGMAX,
458                 .procname       = "msgmax",
459                 .data           = &msg_ctlmax,
460                 .maxlen         = sizeof (int),
461                 .mode           = 0644,
462                 .proc_handler   = &proc_dointvec,
463         },
464         {
465                 .ctl_name       = KERN_MSGMNI,
466                 .procname       = "msgmni",
467                 .data           = &msg_ctlmni,
468                 .maxlen         = sizeof (int),
469                 .mode           = 0644,
470                 .proc_handler   = &proc_dointvec,
471         },
472         {
473                 .ctl_name       = KERN_MSGMNB,
474                 .procname       =  "msgmnb",
475                 .data           = &msg_ctlmnb,
476                 .maxlen         = sizeof (int),
477                 .mode           = 0644,
478                 .proc_handler   = &proc_dointvec,
479         },
480         {
481                 .ctl_name       = KERN_SEM,
482                 .procname       = "sem",
483                 .data           = &sem_ctls,
484                 .maxlen         = 4*sizeof (int),
485                 .mode           = 0644,
486                 .proc_handler   = &proc_dointvec,
487         },
488 #endif
489 #ifdef CONFIG_MAGIC_SYSRQ
490         {
491                 .ctl_name       = KERN_SYSRQ,
492                 .procname       = "sysrq",
493                 .data           = &sysrq_enabled,
494                 .maxlen         = sizeof (int),
495                 .mode           = 0644,
496                 .proc_handler   = &proc_dointvec,
497         },
498 #endif
499         {
500                 .ctl_name       = KERN_CADPID,
501                 .procname       = "cad_pid",
502                 .data           = &cad_pid,
503                 .maxlen         = sizeof (int),
504                 .mode           = 0600,
505                 .proc_handler   = &proc_dointvec,
506         },
507         {
508                 .ctl_name       = KERN_MAX_THREADS,
509                 .procname       = "threads-max",
510                 .data           = &max_threads,
511                 .maxlen         = sizeof(int),
512                 .mode           = 0644,
513                 .proc_handler   = &proc_dointvec,
514         },
515         {
516                 .ctl_name       = KERN_RANDOM,
517                 .procname       = "random",
518                 .mode           = 0555,
519                 .child          = random_table,
520         },
521 #ifdef CONFIG_UNIX98_PTYS
522         {
523                 .ctl_name       = KERN_PTY,
524                 .procname       = "pty",
525                 .mode           = 0555,
526                 .child          = pty_table,
527         },
528 #endif
529         {
530                 .ctl_name       = KERN_OVERFLOWUID,
531                 .procname       = "overflowuid",
532                 .data           = &overflowuid,
533                 .maxlen         = sizeof(int),
534                 .mode           = 0644,
535                 .proc_handler   = &proc_dointvec_minmax,
536                 .strategy       = &sysctl_intvec,
537                 .extra1         = &minolduid,
538                 .extra2         = &maxolduid,
539         },
540         {
541                 .ctl_name       = KERN_OVERFLOWGID,
542                 .procname       = "overflowgid",
543                 .data           = &overflowgid,
544                 .maxlen         = sizeof(int),
545                 .mode           = 0644,
546                 .proc_handler   = &proc_dointvec_minmax,
547                 .strategy       = &sysctl_intvec,
548                 .extra1         = &minolduid,
549                 .extra2         = &maxolduid,
550         },
551 #ifdef CONFIG_S390
552 #ifdef CONFIG_MATHEMU
553         {
554                 .ctl_name       = KERN_IEEE_EMULATION_WARNINGS,
555                 .procname       = "ieee_emulation_warnings",
556                 .data           = &sysctl_ieee_emulation_warnings,
557                 .maxlen         = sizeof(int),
558                 .mode           = 0644,
559                 .proc_handler   = &proc_dointvec,
560         },
561 #endif
562 #ifdef CONFIG_NO_IDLE_HZ
563         {
564                 .ctl_name       = KERN_HZ_TIMER,
565                 .procname       = "hz_timer",
566                 .data           = &sysctl_hz_timer,
567                 .maxlen         = sizeof(int),
568                 .mode           = 0644,
569                 .proc_handler   = &proc_dointvec,
570         },
571 #endif
572         {
573                 .ctl_name       = KERN_S390_USER_DEBUG_LOGGING,
574                 .procname       = "userprocess_debug",
575                 .data           = &sysctl_userprocess_debug,
576                 .maxlen         = sizeof(int),
577                 .mode           = 0644,
578                 .proc_handler   = &proc_dointvec,
579         },
580 #endif
581         {
582                 .ctl_name       = KERN_PIDMAX,
583                 .procname       = "pid_max",
584                 .data           = &pid_max,
585                 .maxlen         = sizeof (int),
586                 .mode           = 0644,
587                 .proc_handler   = &proc_dointvec_minmax,
588                 .strategy       = sysctl_intvec,
589                 .extra1         = &pid_max_min,
590                 .extra2         = &pid_max_max,
591         },
592         {
593                 .ctl_name       = KERN_PANIC_ON_OOPS,
594                 .procname       = "panic_on_oops",
595                 .data           = &panic_on_oops,
596                 .maxlen         = sizeof(int),
597                 .mode           = 0644,
598                 .proc_handler   = &proc_dointvec,
599         },
600         {
601                 .ctl_name       = KERN_PRINTK_RATELIMIT,
602                 .procname       = "printk_ratelimit",
603                 .data           = &printk_ratelimit_jiffies,
604                 .maxlen         = sizeof(int),
605                 .mode           = 0644,
606                 .proc_handler   = &proc_dointvec_jiffies,
607                 .strategy       = &sysctl_jiffies,
608         },
609         {
610                 .ctl_name       = KERN_PRINTK_RATELIMIT_BURST,
611                 .procname       = "printk_ratelimit_burst",
612                 .data           = &printk_ratelimit_burst,
613                 .maxlen         = sizeof(int),
614                 .mode           = 0644,
615                 .proc_handler   = &proc_dointvec,
616         },
617         {
618                 .ctl_name       = KERN_NGROUPS_MAX,
619                 .procname       = "ngroups_max",
620                 .data           = &ngroups_max,
621                 .maxlen         = sizeof (int),
622                 .mode           = 0444,
623                 .proc_handler   = &proc_dointvec,
624         },
625 #if defined(CONFIG_X86_LOCAL_APIC) && defined(CONFIG_X86)
626         {
627                 .ctl_name       = KERN_UNKNOWN_NMI_PANIC,
628                 .procname       = "unknown_nmi_panic",
629                 .data           = &unknown_nmi_panic,
630                 .maxlen         = sizeof (int),
631                 .mode           = 0644,
632                 .proc_handler   = &proc_dointvec,
633         },
634         {
635                 .ctl_name       = KERN_NMI_WATCHDOG,
636                 .procname       = "nmi_watchdog",
637                 .data           = &nmi_watchdog_enabled,
638                 .maxlen         = sizeof (int),
639                 .mode           = 0644,
640                 .proc_handler   = &proc_nmi_enabled,
641         },
642 #endif
643 #if defined(CONFIG_X86)
644         {
645                 .ctl_name       = KERN_PANIC_ON_NMI,
646                 .procname       = "panic_on_unrecovered_nmi",
647                 .data           = &panic_on_unrecovered_nmi,
648                 .maxlen         = sizeof(int),
649                 .mode           = 0644,
650                 .proc_handler   = &proc_dointvec,
651         },
652         {
653                 .ctl_name       = KERN_BOOTLOADER_TYPE,
654                 .procname       = "bootloader_type",
655                 .data           = &bootloader_type,
656                 .maxlen         = sizeof (int),
657                 .mode           = 0444,
658                 .proc_handler   = &proc_dointvec,
659         },
660 #endif
661 #if defined(CONFIG_MMU)
662         {
663                 .ctl_name       = KERN_RANDOMIZE,
664                 .procname       = "randomize_va_space",
665                 .data           = &randomize_va_space,
666                 .maxlen         = sizeof(int),
667                 .mode           = 0644,
668                 .proc_handler   = &proc_dointvec,
669         },
670 #endif
671 #if defined(CONFIG_S390) && defined(CONFIG_SMP)
672         {
673                 .ctl_name       = KERN_SPIN_RETRY,
674                 .procname       = "spin_retry",
675                 .data           = &spin_retry,
676                 .maxlen         = sizeof (int),
677                 .mode           = 0644,
678                 .proc_handler   = &proc_dointvec,
679         },
680 #endif
681 #ifdef CONFIG_ACPI_SLEEP
682         {
683                 .ctl_name       = KERN_ACPI_VIDEO_FLAGS,
684                 .procname       = "acpi_video_flags",
685                 .data           = &acpi_video_flags,
686                 .maxlen         = sizeof (unsigned long),
687                 .mode           = 0644,
688                 .proc_handler   = &proc_doulongvec_minmax,
689         },
690 #endif
691 #ifdef CONFIG_IA64
692         {
693                 .ctl_name       = KERN_IA64_UNALIGNED,
694                 .procname       = "ignore-unaligned-usertrap",
695                 .data           = &no_unaligned_warning,
696                 .maxlen         = sizeof (int),
697                 .mode           = 0644,
698                 .proc_handler   = &proc_dointvec,
699         },
700 #endif
701 #ifdef CONFIG_COMPAT
702         {
703                 .ctl_name       = KERN_COMPAT_LOG,
704                 .procname       = "compat-log",
705                 .data           = &compat_log,
706                 .maxlen         = sizeof (int),
707                 .mode           = 0644,
708                 .proc_handler   = &proc_dointvec,
709         },
710 #endif
711 #ifdef CONFIG_RT_MUTEXES
712         {
713                 .ctl_name       = KERN_MAX_LOCK_DEPTH,
714                 .procname       = "max_lock_depth",
715                 .data           = &max_lock_depth,
716                 .maxlen         = sizeof(int),
717                 .mode           = 0644,
718                 .proc_handler   = &proc_dointvec,
719         },
720 #endif
721
722         { .ctl_name = 0 }
723 };
724
725 /* Constants for minimum and maximum testing in vm_table.
726    We use these as one-element integer vectors. */
727 static int zero;
728 static int one_hundred = 100;
729
730
731 static ctl_table vm_table[] = {
732         {
733                 .ctl_name       = VM_OVERCOMMIT_MEMORY,
734                 .procname       = "overcommit_memory",
735                 .data           = &sysctl_overcommit_memory,
736                 .maxlen         = sizeof(sysctl_overcommit_memory),
737                 .mode           = 0644,
738                 .proc_handler   = &proc_dointvec,
739         },
740         {
741                 .ctl_name       = VM_PANIC_ON_OOM,
742                 .procname       = "panic_on_oom",
743                 .data           = &sysctl_panic_on_oom,
744                 .maxlen         = sizeof(sysctl_panic_on_oom),
745                 .mode           = 0644,
746                 .proc_handler   = &proc_dointvec,
747         },
748         {
749                 .ctl_name       = VM_OVERCOMMIT_RATIO,
750                 .procname       = "overcommit_ratio",
751                 .data           = &sysctl_overcommit_ratio,
752                 .maxlen         = sizeof(sysctl_overcommit_ratio),
753                 .mode           = 0644,
754                 .proc_handler   = &proc_dointvec,
755         },
756         {
757                 .ctl_name       = VM_PAGE_CLUSTER,
758                 .procname       = "page-cluster", 
759                 .data           = &page_cluster,
760                 .maxlen         = sizeof(int),
761                 .mode           = 0644,
762                 .proc_handler   = &proc_dointvec,
763         },
764         {
765                 .ctl_name       = VM_DIRTY_BACKGROUND,
766                 .procname       = "dirty_background_ratio",
767                 .data           = &dirty_background_ratio,
768                 .maxlen         = sizeof(dirty_background_ratio),
769                 .mode           = 0644,
770                 .proc_handler   = &proc_dointvec_minmax,
771                 .strategy       = &sysctl_intvec,
772                 .extra1         = &zero,
773                 .extra2         = &one_hundred,
774         },
775         {
776                 .ctl_name       = VM_DIRTY_RATIO,
777                 .procname       = "dirty_ratio",
778                 .data           = &vm_dirty_ratio,
779                 .maxlen         = sizeof(vm_dirty_ratio),
780                 .mode           = 0644,
781                 .proc_handler   = &proc_dointvec_minmax,
782                 .strategy       = &sysctl_intvec,
783                 .extra1         = &zero,
784                 .extra2         = &one_hundred,
785         },
786         {
787                 .ctl_name       = VM_DIRTY_WB_CS,
788                 .procname       = "dirty_writeback_centisecs",
789                 .data           = &dirty_writeback_interval,
790                 .maxlen         = sizeof(dirty_writeback_interval),
791                 .mode           = 0644,
792                 .proc_handler   = &dirty_writeback_centisecs_handler,
793         },
794         {
795                 .ctl_name       = VM_DIRTY_EXPIRE_CS,
796                 .procname       = "dirty_expire_centisecs",
797                 .data           = &dirty_expire_interval,
798                 .maxlen         = sizeof(dirty_expire_interval),
799                 .mode           = 0644,
800                 .proc_handler   = &proc_dointvec_userhz_jiffies,
801         },
802         {
803                 .ctl_name       = VM_NR_PDFLUSH_THREADS,
804                 .procname       = "nr_pdflush_threads",
805                 .data           = &nr_pdflush_threads,
806                 .maxlen         = sizeof nr_pdflush_threads,
807                 .mode           = 0444 /* read-only*/,
808                 .proc_handler   = &proc_dointvec,
809         },
810         {
811                 .ctl_name       = VM_SWAPPINESS,
812                 .procname       = "swappiness",
813                 .data           = &vm_swappiness,
814                 .maxlen         = sizeof(vm_swappiness),
815                 .mode           = 0644,
816                 .proc_handler   = &proc_dointvec_minmax,
817                 .strategy       = &sysctl_intvec,
818                 .extra1         = &zero,
819                 .extra2         = &one_hundred,
820         },
821 #ifdef CONFIG_HUGETLB_PAGE
822          {
823                 .ctl_name       = VM_HUGETLB_PAGES,
824                 .procname       = "nr_hugepages",
825                 .data           = &max_huge_pages,
826                 .maxlen         = sizeof(unsigned long),
827                 .mode           = 0644,
828                 .proc_handler   = &hugetlb_sysctl_handler,
829                 .extra1         = (void *)&hugetlb_zero,
830                 .extra2         = (void *)&hugetlb_infinity,
831          },
832          {
833                 .ctl_name       = VM_HUGETLB_GROUP,
834                 .procname       = "hugetlb_shm_group",
835                 .data           = &sysctl_hugetlb_shm_group,
836                 .maxlen         = sizeof(gid_t),
837                 .mode           = 0644,
838                 .proc_handler   = &proc_dointvec,
839          },
840 #endif
841         {
842                 .ctl_name       = VM_LOWMEM_RESERVE_RATIO,
843                 .procname       = "lowmem_reserve_ratio",
844                 .data           = &sysctl_lowmem_reserve_ratio,
845                 .maxlen         = sizeof(sysctl_lowmem_reserve_ratio),
846                 .mode           = 0644,
847                 .proc_handler   = &lowmem_reserve_ratio_sysctl_handler,
848                 .strategy       = &sysctl_intvec,
849         },
850         {
851                 .ctl_name       = VM_DROP_PAGECACHE,
852                 .procname       = "drop_caches",
853                 .data           = &sysctl_drop_caches,
854                 .maxlen         = sizeof(int),
855                 .mode           = 0644,
856                 .proc_handler   = drop_caches_sysctl_handler,
857                 .strategy       = &sysctl_intvec,
858         },
859         {
860                 .ctl_name       = VM_MIN_FREE_KBYTES,
861                 .procname       = "min_free_kbytes",
862                 .data           = &min_free_kbytes,
863                 .maxlen         = sizeof(min_free_kbytes),
864                 .mode           = 0644,
865                 .proc_handler   = &min_free_kbytes_sysctl_handler,
866                 .strategy       = &sysctl_intvec,
867                 .extra1         = &zero,
868         },
869         {
870                 .ctl_name       = VM_PERCPU_PAGELIST_FRACTION,
871                 .procname       = "percpu_pagelist_fraction",
872                 .data           = &percpu_pagelist_fraction,
873                 .maxlen         = sizeof(percpu_pagelist_fraction),
874                 .mode           = 0644,
875                 .proc_handler   = &percpu_pagelist_fraction_sysctl_handler,
876                 .strategy       = &sysctl_intvec,
877                 .extra1         = &min_percpu_pagelist_fract,
878         },
879 #ifdef CONFIG_MMU
880         {
881                 .ctl_name       = VM_MAX_MAP_COUNT,
882                 .procname       = "max_map_count",
883                 .data           = &sysctl_max_map_count,
884                 .maxlen         = sizeof(sysctl_max_map_count),
885                 .mode           = 0644,
886                 .proc_handler   = &proc_dointvec
887         },
888 #endif
889         {
890                 .ctl_name       = VM_LAPTOP_MODE,
891                 .procname       = "laptop_mode",
892                 .data           = &laptop_mode,
893                 .maxlen         = sizeof(laptop_mode),
894                 .mode           = 0644,
895                 .proc_handler   = &proc_dointvec_jiffies,
896                 .strategy       = &sysctl_jiffies,
897         },
898         {
899                 .ctl_name       = VM_BLOCK_DUMP,
900                 .procname       = "block_dump",
901                 .data           = &block_dump,
902                 .maxlen         = sizeof(block_dump),
903                 .mode           = 0644,
904                 .proc_handler   = &proc_dointvec,
905                 .strategy       = &sysctl_intvec,
906                 .extra1         = &zero,
907         },
908         {
909                 .ctl_name       = VM_VFS_CACHE_PRESSURE,
910                 .procname       = "vfs_cache_pressure",
911                 .data           = &sysctl_vfs_cache_pressure,
912                 .maxlen         = sizeof(sysctl_vfs_cache_pressure),
913                 .mode           = 0644,
914                 .proc_handler   = &proc_dointvec,
915                 .strategy       = &sysctl_intvec,
916                 .extra1         = &zero,
917         },
918 #ifdef HAVE_ARCH_PICK_MMAP_LAYOUT
919         {
920                 .ctl_name       = VM_LEGACY_VA_LAYOUT,
921                 .procname       = "legacy_va_layout",
922                 .data           = &sysctl_legacy_va_layout,
923                 .maxlen         = sizeof(sysctl_legacy_va_layout),
924                 .mode           = 0644,
925                 .proc_handler   = &proc_dointvec,
926                 .strategy       = &sysctl_intvec,
927                 .extra1         = &zero,
928         },
929 #endif
930 #ifdef CONFIG_SWAP
931         {
932                 .ctl_name       = VM_SWAP_TOKEN_TIMEOUT,
933                 .procname       = "swap_token_timeout",
934                 .data           = &swap_token_default_timeout,
935                 .maxlen         = sizeof(swap_token_default_timeout),
936                 .mode           = 0644,
937                 .proc_handler   = &proc_dointvec_jiffies,
938                 .strategy       = &sysctl_jiffies,
939         },
940 #endif
941 #ifdef CONFIG_NUMA
942         {
943                 .ctl_name       = VM_ZONE_RECLAIM_MODE,
944                 .procname       = "zone_reclaim_mode",
945                 .data           = &zone_reclaim_mode,
946                 .maxlen         = sizeof(zone_reclaim_mode),
947                 .mode           = 0644,
948                 .proc_handler   = &proc_dointvec,
949                 .strategy       = &sysctl_intvec,
950                 .extra1         = &zero,
951         },
952         {
953                 .ctl_name       = VM_MIN_UNMAPPED,
954                 .procname       = "min_unmapped_ratio",
955                 .data           = &sysctl_min_unmapped_ratio,
956                 .maxlen         = sizeof(sysctl_min_unmapped_ratio),
957                 .mode           = 0644,
958                 .proc_handler   = &sysctl_min_unmapped_ratio_sysctl_handler,
959                 .strategy       = &sysctl_intvec,
960                 .extra1         = &zero,
961                 .extra2         = &one_hundred,
962         },
963         {
964                 .ctl_name       = VM_MIN_SLAB,
965                 .procname       = "min_slab_ratio",
966                 .data           = &sysctl_min_slab_ratio,
967                 .maxlen         = sizeof(sysctl_min_slab_ratio),
968                 .mode           = 0644,
969                 .proc_handler   = &sysctl_min_slab_ratio_sysctl_handler,
970                 .strategy       = &sysctl_intvec,
971                 .extra1         = &zero,
972                 .extra2         = &one_hundred,
973         },
974 #endif
975 #ifdef CONFIG_X86_32
976         {
977                 .ctl_name       = VM_VDSO_ENABLED,
978                 .procname       = "vdso_enabled",
979                 .data           = &vdso_enabled,
980                 .maxlen         = sizeof(vdso_enabled),
981                 .mode           = 0644,
982                 .proc_handler   = &proc_dointvec,
983                 .strategy       = &sysctl_intvec,
984                 .extra1         = &zero,
985         },
986 #endif
987         { .ctl_name = 0 }
988 };
989
990 static ctl_table fs_table[] = {
991         {
992                 .ctl_name       = FS_NRINODE,
993                 .procname       = "inode-nr",
994                 .data           = &inodes_stat,
995                 .maxlen         = 2*sizeof(int),
996                 .mode           = 0444,
997                 .proc_handler   = &proc_dointvec,
998         },
999         {
1000                 .ctl_name       = FS_STATINODE,
1001                 .procname       = "inode-state",
1002                 .data           = &inodes_stat,
1003                 .maxlen         = 7*sizeof(int),
1004                 .mode           = 0444,
1005                 .proc_handler   = &proc_dointvec,
1006         },
1007         {
1008                 .ctl_name       = FS_NRFILE,
1009                 .procname       = "file-nr",
1010                 .data           = &files_stat,
1011                 .maxlen         = 3*sizeof(int),
1012                 .mode           = 0444,
1013                 .proc_handler   = &proc_nr_files,
1014         },
1015         {
1016                 .ctl_name       = FS_MAXFILE,
1017                 .procname       = "file-max",
1018                 .data           = &files_stat.max_files,
1019                 .maxlen         = sizeof(int),
1020                 .mode           = 0644,
1021                 .proc_handler   = &proc_dointvec,
1022         },
1023         {
1024                 .ctl_name       = FS_DENTRY,
1025                 .procname       = "dentry-state",
1026                 .data           = &dentry_stat,
1027                 .maxlen         = 6*sizeof(int),
1028                 .mode           = 0444,
1029                 .proc_handler   = &proc_dointvec,
1030         },
1031         {
1032                 .ctl_name       = FS_OVERFLOWUID,
1033                 .procname       = "overflowuid",
1034                 .data           = &fs_overflowuid,
1035                 .maxlen         = sizeof(int),
1036                 .mode           = 0644,
1037                 .proc_handler   = &proc_dointvec_minmax,
1038                 .strategy       = &sysctl_intvec,
1039                 .extra1         = &minolduid,
1040                 .extra2         = &maxolduid,
1041         },
1042         {
1043                 .ctl_name       = FS_OVERFLOWGID,
1044                 .procname       = "overflowgid",
1045                 .data           = &fs_overflowgid,
1046                 .maxlen         = sizeof(int),
1047                 .mode           = 0644,
1048                 .proc_handler   = &proc_dointvec_minmax,
1049                 .strategy       = &sysctl_intvec,
1050                 .extra1         = &minolduid,
1051                 .extra2         = &maxolduid,
1052         },
1053         {
1054                 .ctl_name       = FS_LEASES,
1055                 .procname       = "leases-enable",
1056                 .data           = &leases_enable,
1057                 .maxlen         = sizeof(int),
1058                 .mode           = 0644,
1059                 .proc_handler   = &proc_dointvec,
1060         },
1061 #ifdef CONFIG_DNOTIFY
1062         {
1063                 .ctl_name       = FS_DIR_NOTIFY,
1064                 .procname       = "dir-notify-enable",
1065                 .data           = &dir_notify_enable,
1066                 .maxlen         = sizeof(int),
1067                 .mode           = 0644,
1068                 .proc_handler   = &proc_dointvec,
1069         },
1070 #endif
1071 #ifdef CONFIG_MMU
1072         {
1073                 .ctl_name       = FS_LEASE_TIME,
1074                 .procname       = "lease-break-time",
1075                 .data           = &lease_break_time,
1076                 .maxlen         = sizeof(int),
1077                 .mode           = 0644,
1078                 .proc_handler   = &proc_dointvec,
1079         },
1080         {
1081                 .ctl_name       = FS_AIO_NR,
1082                 .procname       = "aio-nr",
1083                 .data           = &aio_nr,
1084                 .maxlen         = sizeof(aio_nr),
1085                 .mode           = 0444,
1086                 .proc_handler   = &proc_doulongvec_minmax,
1087         },
1088         {
1089                 .ctl_name       = FS_AIO_MAX_NR,
1090                 .procname       = "aio-max-nr",
1091                 .data           = &aio_max_nr,
1092                 .maxlen         = sizeof(aio_max_nr),
1093                 .mode           = 0644,
1094                 .proc_handler   = &proc_doulongvec_minmax,
1095         },
1096 #ifdef CONFIG_INOTIFY_USER
1097         {
1098                 .ctl_name       = FS_INOTIFY,
1099                 .procname       = "inotify",
1100                 .mode           = 0555,
1101                 .child          = inotify_table,
1102         },
1103 #endif  
1104 #endif
1105         {
1106                 .ctl_name       = KERN_SETUID_DUMPABLE,
1107                 .procname       = "suid_dumpable",
1108                 .data           = &suid_dumpable,
1109                 .maxlen         = sizeof(int),
1110                 .mode           = 0644,
1111                 .proc_handler   = &proc_dointvec,
1112         },
1113         { .ctl_name = 0 }
1114 };
1115
1116 static ctl_table debug_table[] = {
1117         { .ctl_name = 0 }
1118 };
1119
1120 static ctl_table dev_table[] = {
1121         { .ctl_name = 0 }
1122 };
1123
1124 extern void init_irq_proc (void);
1125
1126 static DEFINE_SPINLOCK(sysctl_lock);
1127
1128 /* called under sysctl_lock */
1129 static int use_table(struct ctl_table_header *p)
1130 {
1131         if (unlikely(p->unregistering))
1132                 return 0;
1133         p->used++;
1134         return 1;
1135 }
1136
1137 /* called under sysctl_lock */
1138 static void unuse_table(struct ctl_table_header *p)
1139 {
1140         if (!--p->used)
1141                 if (unlikely(p->unregistering))
1142                         complete(p->unregistering);
1143 }
1144
1145 /* called under sysctl_lock, will reacquire if has to wait */
1146 static void start_unregistering(struct ctl_table_header *p)
1147 {
1148         /*
1149          * if p->used is 0, nobody will ever touch that entry again;
1150          * we'll eliminate all paths to it before dropping sysctl_lock
1151          */
1152         if (unlikely(p->used)) {
1153                 struct completion wait;
1154                 init_completion(&wait);
1155                 p->unregistering = &wait;
1156                 spin_unlock(&sysctl_lock);
1157                 wait_for_completion(&wait);
1158                 spin_lock(&sysctl_lock);
1159         }
1160         /*
1161          * do not remove from the list until nobody holds it; walking the
1162          * list in do_sysctl() relies on that.
1163          */
1164         list_del_init(&p->ctl_entry);
1165 }
1166
1167 void __init sysctl_init(void)
1168 {
1169 #ifdef CONFIG_PROC_SYSCTL
1170         register_proc_table(root_table, proc_sys_root, &root_table_header);
1171         init_irq_proc();
1172 #endif
1173 }
1174
1175 #ifdef CONFIG_SYSCTL_SYSCALL
1176 int do_sysctl(int __user *name, int nlen, void __user *oldval, size_t __user *oldlenp,
1177                void __user *newval, size_t newlen)
1178 {
1179         struct list_head *tmp;
1180         int error = -ENOTDIR;
1181
1182         if (nlen <= 0 || nlen >= CTL_MAXNAME)
1183                 return -ENOTDIR;
1184         if (oldval) {
1185                 int old_len;
1186                 if (!oldlenp || get_user(old_len, oldlenp))
1187                         return -EFAULT;
1188         }
1189         spin_lock(&sysctl_lock);
1190         tmp = &root_table_header.ctl_entry;
1191         do {
1192                 struct ctl_table_header *head =
1193                         list_entry(tmp, struct ctl_table_header, ctl_entry);
1194                 void *context = NULL;
1195
1196                 if (!use_table(head))
1197                         continue;
1198
1199                 spin_unlock(&sysctl_lock);
1200
1201                 error = parse_table(name, nlen, oldval, oldlenp, 
1202                                         newval, newlen, head->ctl_table,
1203                                         &context);
1204                 kfree(context);
1205
1206                 spin_lock(&sysctl_lock);
1207                 unuse_table(head);
1208                 if (error != -ENOTDIR)
1209                         break;
1210         } while ((tmp = tmp->next) != &root_table_header.ctl_entry);
1211         spin_unlock(&sysctl_lock);
1212         return error;
1213 }
1214
1215 asmlinkage long sys_sysctl(struct __sysctl_args __user *args)
1216 {
1217         struct __sysctl_args tmp;
1218         int error;
1219
1220         if (copy_from_user(&tmp, args, sizeof(tmp)))
1221                 return -EFAULT;
1222
1223         lock_kernel();
1224         error = do_sysctl(tmp.name, tmp.nlen, tmp.oldval, tmp.oldlenp,
1225                           tmp.newval, tmp.newlen);
1226         unlock_kernel();
1227         return error;
1228 }
1229 #endif /* CONFIG_SYSCTL_SYSCALL */
1230
1231 /*
1232  * ctl_perm does NOT grant the superuser all rights automatically, because
1233  * some sysctl variables are readonly even to root.
1234  */
1235
1236 static int test_perm(int mode, int op)
1237 {
1238         if (!current->euid)
1239                 mode >>= 6;
1240         else if (in_egroup_p(0))
1241                 mode >>= 3;
1242         if ((mode & op & 0007) == op)
1243                 return 0;
1244         return -EACCES;
1245 }
1246
1247 static inline int ctl_perm(ctl_table *table, int op)
1248 {
1249         int error;
1250         error = security_sysctl(table, op);
1251         if (error)
1252                 return error;
1253         return test_perm(table->mode, op);
1254 }
1255
1256 #ifdef CONFIG_SYSCTL_SYSCALL
1257 static int parse_table(int __user *name, int nlen,
1258                        void __user *oldval, size_t __user *oldlenp,
1259                        void __user *newval, size_t newlen,
1260                        ctl_table *table, void **context)
1261 {
1262         int n;
1263 repeat:
1264         if (!nlen)
1265                 return -ENOTDIR;
1266         if (get_user(n, name))
1267                 return -EFAULT;
1268         for ( ; table->ctl_name; table++) {
1269                 if (n == table->ctl_name || table->ctl_name == CTL_ANY) {
1270                         int error;
1271                         if (table->child) {
1272                                 if (ctl_perm(table, 001))
1273                                         return -EPERM;
1274                                 if (table->strategy) {
1275                                         error = table->strategy(
1276                                                 table, name, nlen,
1277                                                 oldval, oldlenp,
1278                                                 newval, newlen, context);
1279                                         if (error)
1280                                                 return error;
1281                                 }
1282                                 name++;
1283                                 nlen--;
1284                                 table = table->child;
1285                                 goto repeat;
1286                         }
1287                         error = do_sysctl_strategy(table, name, nlen,
1288                                                    oldval, oldlenp,
1289                                                    newval, newlen, context);
1290                         return error;
1291                 }
1292         }
1293         return -ENOTDIR;
1294 }
1295
1296 /* Perform the actual read/write of a sysctl table entry. */
1297 int do_sysctl_strategy (ctl_table *table, 
1298                         int __user *name, int nlen,
1299                         void __user *oldval, size_t __user *oldlenp,
1300                         void __user *newval, size_t newlen, void **context)
1301 {
1302         int op = 0, rc;
1303         size_t len;
1304
1305         if (oldval)
1306                 op |= 004;
1307         if (newval) 
1308                 op |= 002;
1309         if (ctl_perm(table, op))
1310                 return -EPERM;
1311
1312         if (table->strategy) {
1313                 rc = table->strategy(table, name, nlen, oldval, oldlenp,
1314                                      newval, newlen, context);
1315                 if (rc < 0)
1316                         return rc;
1317                 if (rc > 0)
1318                         return 0;
1319         }
1320
1321         /* If there is no strategy routine, or if the strategy returns
1322          * zero, proceed with automatic r/w */
1323         if (table->data && table->maxlen) {
1324                 if (oldval && oldlenp) {
1325                         if (get_user(len, oldlenp))
1326                                 return -EFAULT;
1327                         if (len) {
1328                                 if (len > table->maxlen)
1329                                         len = table->maxlen;
1330                                 if(copy_to_user(oldval, table->data, len))
1331                                         return -EFAULT;
1332                                 if(put_user(len, oldlenp))
1333                                         return -EFAULT;
1334                         }
1335                 }
1336                 if (newval && newlen) {
1337                         len = newlen;
1338                         if (len > table->maxlen)
1339                                 len = table->maxlen;
1340                         if(copy_from_user(table->data, newval, len))
1341                                 return -EFAULT;
1342                 }
1343         }
1344         return 0;
1345 }
1346 #endif /* CONFIG_SYSCTL_SYSCALL */
1347
1348 /**
1349  * register_sysctl_table - register a sysctl hierarchy
1350  * @table: the top-level table structure
1351  * @insert_at_head: whether the entry should be inserted in front or at the end
1352  *
1353  * Register a sysctl table hierarchy. @table should be a filled in ctl_table
1354  * array. An entry with a ctl_name of 0 terminates the table. 
1355  *
1356  * The members of the &ctl_table structure are used as follows:
1357  *
1358  * ctl_name - This is the numeric sysctl value used by sysctl(2). The number
1359  *            must be unique within that level of sysctl
1360  *
1361  * procname - the name of the sysctl file under /proc/sys. Set to %NULL to not
1362  *            enter a sysctl file
1363  *
1364  * data - a pointer to data for use by proc_handler
1365  *
1366  * maxlen - the maximum size in bytes of the data
1367  *
1368  * mode - the file permissions for the /proc/sys file, and for sysctl(2)
1369  *
1370  * child - a pointer to the child sysctl table if this entry is a directory, or
1371  *         %NULL.
1372  *
1373  * proc_handler - the text handler routine (described below)
1374  *
1375  * strategy - the strategy routine (described below)
1376  *
1377  * de - for internal use by the sysctl routines
1378  *
1379  * extra1, extra2 - extra pointers usable by the proc handler routines
1380  *
1381  * Leaf nodes in the sysctl tree will be represented by a single file
1382  * under /proc; non-leaf nodes will be represented by directories.
1383  *
1384  * sysctl(2) can automatically manage read and write requests through
1385  * the sysctl table.  The data and maxlen fields of the ctl_table
1386  * struct enable minimal validation of the values being written to be
1387  * performed, and the mode field allows minimal authentication.
1388  *
1389  * More sophisticated management can be enabled by the provision of a
1390  * strategy routine with the table entry.  This will be called before
1391  * any automatic read or write of the data is performed.
1392  *
1393  * The strategy routine may return
1394  *
1395  * < 0 - Error occurred (error is passed to user process)
1396  *
1397  * 0   - OK - proceed with automatic read or write.
1398  *
1399  * > 0 - OK - read or write has been done by the strategy routine, so
1400  *       return immediately.
1401  *
1402  * There must be a proc_handler routine for any terminal nodes
1403  * mirrored under /proc/sys (non-terminals are handled by a built-in
1404  * directory handler).  Several default handlers are available to
1405  * cover common cases -
1406  *
1407  * proc_dostring(), proc_dointvec(), proc_dointvec_jiffies(),
1408  * proc_dointvec_userhz_jiffies(), proc_dointvec_minmax(), 
1409  * proc_doulongvec_ms_jiffies_minmax(), proc_doulongvec_minmax()
1410  *
1411  * It is the handler's job to read the input buffer from user memory
1412  * and process it. The handler should return 0 on success.
1413  *
1414  * This routine returns %NULL on a failure to register, and a pointer
1415  * to the table header on success.
1416  */
1417 struct ctl_table_header *register_sysctl_table(ctl_table * table, 
1418                                                int insert_at_head)
1419 {
1420         struct ctl_table_header *tmp;
1421         tmp = kmalloc(sizeof(struct ctl_table_header), GFP_KERNEL);
1422         if (!tmp)
1423                 return NULL;
1424         tmp->ctl_table = table;
1425         INIT_LIST_HEAD(&tmp->ctl_entry);
1426         tmp->used = 0;
1427         tmp->unregistering = NULL;
1428         spin_lock(&sysctl_lock);
1429         if (insert_at_head)
1430                 list_add(&tmp->ctl_entry, &root_table_header.ctl_entry);
1431         else
1432                 list_add_tail(&tmp->ctl_entry, &root_table_header.ctl_entry);
1433         spin_unlock(&sysctl_lock);
1434 #ifdef CONFIG_PROC_SYSCTL
1435         register_proc_table(table, proc_sys_root, tmp);
1436 #endif
1437         return tmp;
1438 }
1439
1440 /**
1441  * unregister_sysctl_table - unregister a sysctl table hierarchy
1442  * @header: the header returned from register_sysctl_table
1443  *
1444  * Unregisters the sysctl table and all children. proc entries may not
1445  * actually be removed until they are no longer used by anyone.
1446  */
1447 void unregister_sysctl_table(struct ctl_table_header * header)
1448 {
1449         might_sleep();
1450         spin_lock(&sysctl_lock);
1451         start_unregistering(header);
1452 #ifdef CONFIG_PROC_SYSCTL
1453         unregister_proc_table(header->ctl_table, proc_sys_root);
1454 #endif
1455         spin_unlock(&sysctl_lock);
1456         kfree(header);
1457 }
1458
1459 #else /* !CONFIG_SYSCTL */
1460 struct ctl_table_header * register_sysctl_table(ctl_table * table,
1461                                                 int insert_at_head)
1462 {
1463         return NULL;
1464 }
1465
1466 void unregister_sysctl_table(struct ctl_table_header * table)
1467 {
1468 }
1469
1470 #endif /* CONFIG_SYSCTL */
1471
1472 /*
1473  * /proc/sys support
1474  */
1475
1476 #ifdef CONFIG_PROC_SYSCTL
1477
1478 /* Scan the sysctl entries in table and add them all into /proc */
1479 static void register_proc_table(ctl_table * table, struct proc_dir_entry *root, void *set)
1480 {
1481         struct proc_dir_entry *de;
1482         int len;
1483         mode_t mode;
1484         
1485         for (; table->ctl_name; table++) {
1486                 /* Can't do anything without a proc name. */
1487                 if (!table->procname)
1488                         continue;
1489                 /* Maybe we can't do anything with it... */
1490                 if (!table->proc_handler && !table->child) {
1491                         printk(KERN_WARNING "SYSCTL: Can't register %s\n",
1492                                 table->procname);
1493                         continue;
1494                 }
1495
1496                 len = strlen(table->procname);
1497                 mode = table->mode;
1498
1499                 de = NULL;
1500                 if (table->proc_handler)
1501                         mode |= S_IFREG;
1502                 else {
1503                         mode |= S_IFDIR;
1504                         for (de = root->subdir; de; de = de->next) {
1505                                 if (proc_match(len, table->procname, de))
1506                                         break;
1507                         }
1508                         /* If the subdir exists already, de is non-NULL */
1509                 }
1510
1511                 if (!de) {
1512                         de = create_proc_entry(table->procname, mode, root);
1513                         if (!de)
1514                                 continue;
1515                         de->set = set;
1516                         de->data = (void *) table;
1517                         if (table->proc_handler)
1518                                 de->proc_fops = &proc_sys_file_operations;
1519                 }
1520                 table->de = de;
1521                 if (de->mode & S_IFDIR)
1522                         register_proc_table(table->child, de, set);
1523         }
1524 }
1525
1526 /*
1527  * Unregister a /proc sysctl table and any subdirectories.
1528  */
1529 static void unregister_proc_table(ctl_table * table, struct proc_dir_entry *root)
1530 {
1531         struct proc_dir_entry *de;
1532         for (; table->ctl_name; table++) {
1533                 if (!(de = table->de))
1534                         continue;
1535                 if (de->mode & S_IFDIR) {
1536                         if (!table->child) {
1537                                 printk (KERN_ALERT "Help - malformed sysctl tree on free\n");
1538                                 continue;
1539                         }
1540                         unregister_proc_table(table->child, de);
1541
1542                         /* Don't unregister directories which still have entries.. */
1543                         if (de->subdir)
1544                                 continue;
1545                 }
1546
1547                 /*
1548                  * In any case, mark the entry as goner; we'll keep it
1549                  * around if it's busy, but we'll know to do nothing with
1550                  * its fields.  We are under sysctl_lock here.
1551                  */
1552                 de->data = NULL;
1553
1554                 /* Don't unregister proc entries that are still being used.. */
1555                 if (atomic_read(&de->count))
1556                         continue;
1557
1558                 table->de = NULL;
1559                 remove_proc_entry(table->procname, root);
1560         }
1561 }
1562
1563 static ssize_t do_rw_proc(int write, struct file * file, char __user * buf,
1564                           size_t count, loff_t *ppos)
1565 {
1566         int op;
1567         struct proc_dir_entry *de = PDE(file->f_dentry->d_inode);
1568         struct ctl_table *table;
1569         size_t res;
1570         ssize_t error = -ENOTDIR;
1571         
1572         spin_lock(&sysctl_lock);
1573         if (de && de->data && use_table(de->set)) {
1574                 /*
1575                  * at that point we know that sysctl was not unregistered
1576                  * and won't be until we finish
1577                  */
1578                 spin_unlock(&sysctl_lock);
1579                 table = (struct ctl_table *) de->data;
1580                 if (!table || !table->proc_handler)
1581                         goto out;
1582                 error = -EPERM;
1583                 op = (write ? 002 : 004);
1584                 if (ctl_perm(table, op))
1585                         goto out;
1586                 
1587                 /* careful: calling conventions are nasty here */
1588                 res = count;
1589                 error = (*table->proc_handler)(table, write, file,
1590                                                 buf, &res, ppos);
1591                 if (!error)
1592                         error = res;
1593         out:
1594                 spin_lock(&sysctl_lock);
1595                 unuse_table(de->set);
1596         }
1597         spin_unlock(&sysctl_lock);
1598         return error;
1599 }
1600
1601 static int proc_opensys(struct inode *inode, struct file *file)
1602 {
1603         if (file->f_mode & FMODE_WRITE) {
1604                 /*
1605                  * sysctl entries that are not writable,
1606                  * are _NOT_ writable, capabilities or not.
1607                  */
1608                 if (!(inode->i_mode & S_IWUSR))
1609                         return -EPERM;
1610         }
1611
1612         return 0;
1613 }
1614
1615 static ssize_t proc_readsys(struct file * file, char __user * buf,
1616                             size_t count, loff_t *ppos)
1617 {
1618         return do_rw_proc(0, file, buf, count, ppos);
1619 }
1620
1621 static ssize_t proc_writesys(struct file * file, const char __user * buf,
1622                              size_t count, loff_t *ppos)
1623 {
1624         return do_rw_proc(1, file, (char __user *) buf, count, ppos);
1625 }
1626
1627 /**
1628  * proc_dostring - read a string sysctl
1629  * @table: the sysctl table
1630  * @write: %TRUE if this is a write to the sysctl file
1631  * @filp: the file structure
1632  * @buffer: the user buffer
1633  * @lenp: the size of the user buffer
1634  * @ppos: file position
1635  *
1636  * Reads/writes a string from/to the user buffer. If the kernel
1637  * buffer provided is not large enough to hold the string, the
1638  * string is truncated. The copied string is %NULL-terminated.
1639  * If the string is being read by the user process, it is copied
1640  * and a newline '\n' is added. It is truncated if the buffer is
1641  * not large enough.
1642  *
1643  * Returns 0 on success.
1644  */
1645 int proc_dostring(ctl_table *table, int write, struct file *filp,
1646                   void __user *buffer, size_t *lenp, loff_t *ppos)
1647 {
1648         size_t len;
1649         char __user *p;
1650         char c;
1651         
1652         if (!table->data || !table->maxlen || !*lenp ||
1653             (*ppos && !write)) {
1654                 *lenp = 0;
1655                 return 0;
1656         }
1657         
1658         if (write) {
1659                 len = 0;
1660                 p = buffer;
1661                 while (len < *lenp) {
1662                         if (get_user(c, p++))
1663                                 return -EFAULT;
1664                         if (c == 0 || c == '\n')
1665                                 break;
1666                         len++;
1667                 }
1668                 if (len >= table->maxlen)
1669                         len = table->maxlen-1;
1670                 if(copy_from_user(table->data, buffer, len))
1671                         return -EFAULT;
1672                 ((char *) table->data)[len] = 0;
1673                 *ppos += *lenp;
1674         } else {
1675                 len = strlen(table->data);
1676                 if (len > table->maxlen)
1677                         len = table->maxlen;
1678                 if (len > *lenp)
1679                         len = *lenp;
1680                 if (len)
1681                         if(copy_to_user(buffer, table->data, len))
1682                                 return -EFAULT;
1683                 if (len < *lenp) {
1684                         if(put_user('\n', ((char __user *) buffer) + len))
1685                                 return -EFAULT;
1686                         len++;
1687                 }
1688                 *lenp = len;
1689                 *ppos += len;
1690         }
1691         return 0;
1692 }
1693
1694 /*
1695  *      Special case of dostring for the UTS structure. This has locks
1696  *      to observe. Should this be in kernel/sys.c ????
1697  */
1698  
1699 static int proc_doutsstring(ctl_table *table, int write, struct file *filp,
1700                   void __user *buffer, size_t *lenp, loff_t *ppos)
1701 {
1702         int r;
1703
1704         if (!write) {
1705                 down_read(&uts_sem);
1706                 r=proc_dostring(table,0,filp,buffer,lenp, ppos);
1707                 up_read(&uts_sem);
1708         } else {
1709                 down_write(&uts_sem);
1710                 r=proc_dostring(table,1,filp,buffer,lenp, ppos);
1711                 up_write(&uts_sem);
1712         }
1713         return r;
1714 }
1715
1716 static int do_proc_dointvec_conv(int *negp, unsigned long *lvalp,
1717                                  int *valp,
1718                                  int write, void *data)
1719 {
1720         if (write) {
1721                 *valp = *negp ? -*lvalp : *lvalp;
1722         } else {
1723                 int val = *valp;
1724                 if (val < 0) {
1725                         *negp = -1;
1726                         *lvalp = (unsigned long)-val;
1727                 } else {
1728                         *negp = 0;
1729                         *lvalp = (unsigned long)val;
1730                 }
1731         }
1732         return 0;
1733 }
1734
1735 static int do_proc_dointvec(ctl_table *table, int write, struct file *filp,
1736                   void __user *buffer, size_t *lenp, loff_t *ppos,
1737                   int (*conv)(int *negp, unsigned long *lvalp, int *valp,
1738                               int write, void *data),
1739                   void *data)
1740 {
1741 #define TMPBUFLEN 21
1742         int *i, vleft, first=1, neg, val;
1743         unsigned long lval;
1744         size_t left, len;
1745         
1746         char buf[TMPBUFLEN], *p;
1747         char __user *s = buffer;
1748         
1749         if (!table->data || !table->maxlen || !*lenp ||
1750             (*ppos && !write)) {
1751                 *lenp = 0;
1752                 return 0;
1753         }
1754         
1755         i = (int *) table->data;
1756         vleft = table->maxlen / sizeof(*i);
1757         left = *lenp;
1758
1759         if (!conv)
1760                 conv = do_proc_dointvec_conv;
1761
1762         for (; left && vleft--; i++, first=0) {
1763                 if (write) {
1764                         while (left) {
1765                                 char c;
1766                                 if (get_user(c, s))
1767                                         return -EFAULT;
1768                                 if (!isspace(c))
1769                                         break;
1770                                 left--;
1771                                 s++;
1772                         }
1773                         if (!left)
1774                                 break;
1775                         neg = 0;
1776                         len = left;
1777                         if (len > sizeof(buf) - 1)
1778                                 len = sizeof(buf) - 1;
1779                         if (copy_from_user(buf, s, len))
1780                                 return -EFAULT;
1781                         buf[len] = 0;
1782                         p = buf;
1783                         if (*p == '-' && left > 1) {
1784                                 neg = 1;
1785                                 left--, p++;
1786                         }
1787                         if (*p < '0' || *p > '9')
1788                                 break;
1789
1790                         lval = simple_strtoul(p, &p, 0);
1791
1792                         len = p-buf;
1793                         if ((len < left) && *p && !isspace(*p))
1794                                 break;
1795                         if (neg)
1796                                 val = -val;
1797                         s += len;
1798                         left -= len;
1799
1800                         if (conv(&neg, &lval, i, 1, data))
1801                                 break;
1802                 } else {
1803                         p = buf;
1804                         if (!first)
1805                                 *p++ = '\t';
1806         
1807                         if (conv(&neg, &lval, i, 0, data))
1808                                 break;
1809
1810                         sprintf(p, "%s%lu", neg ? "-" : "", lval);
1811                         len = strlen(buf);
1812                         if (len > left)
1813                                 len = left;
1814                         if(copy_to_user(s, buf, len))
1815                                 return -EFAULT;
1816                         left -= len;
1817                         s += len;
1818                 }
1819         }
1820
1821         if (!write && !first && left) {
1822                 if(put_user('\n', s))
1823                         return -EFAULT;
1824                 left--, s++;
1825         }
1826         if (write) {
1827                 while (left) {
1828                         char c;
1829                         if (get_user(c, s++))
1830                                 return -EFAULT;
1831                         if (!isspace(c))
1832                                 break;
1833                         left--;
1834                 }
1835         }
1836         if (write && first)
1837                 return -EINVAL;
1838         *lenp -= left;
1839         *ppos += *lenp;
1840         return 0;
1841 #undef TMPBUFLEN
1842 }
1843
1844 /**
1845  * proc_dointvec - read a vector of integers
1846  * @table: the sysctl table
1847  * @write: %TRUE if this is a write to the sysctl file
1848  * @filp: the file structure
1849  * @buffer: the user buffer
1850  * @lenp: the size of the user buffer
1851  * @ppos: file position
1852  *
1853  * Reads/writes up to table->maxlen/sizeof(unsigned int) integer
1854  * values from/to the user buffer, treated as an ASCII string. 
1855  *
1856  * Returns 0 on success.
1857  */
1858 int proc_dointvec(ctl_table *table, int write, struct file *filp,
1859                      void __user *buffer, size_t *lenp, loff_t *ppos)
1860 {
1861     return do_proc_dointvec(table,write,filp,buffer,lenp,ppos,
1862                             NULL,NULL);
1863 }
1864
1865 #define OP_SET  0
1866 #define OP_AND  1
1867 #define OP_OR   2
1868 #define OP_MAX  3
1869 #define OP_MIN  4
1870
1871 static int do_proc_dointvec_bset_conv(int *negp, unsigned long *lvalp,
1872                                       int *valp,
1873                                       int write, void *data)
1874 {
1875         int op = *(int *)data;
1876         if (write) {
1877                 int val = *negp ? -*lvalp : *lvalp;
1878                 switch(op) {
1879                 case OP_SET:    *valp = val; break;
1880                 case OP_AND:    *valp &= val; break;
1881                 case OP_OR:     *valp |= val; break;
1882                 case OP_MAX:    if(*valp < val)
1883                                         *valp = val;
1884                                 break;
1885                 case OP_MIN:    if(*valp > val)
1886                                 *valp = val;
1887                                 break;
1888                 }
1889         } else {
1890                 int val = *valp;
1891                 if (val < 0) {
1892                         *negp = -1;
1893                         *lvalp = (unsigned long)-val;
1894                 } else {
1895                         *negp = 0;
1896                         *lvalp = (unsigned long)val;
1897                 }
1898         }
1899         return 0;
1900 }
1901
1902 /*
1903  *      init may raise the set.
1904  */
1905  
1906 int proc_dointvec_bset(ctl_table *table, int write, struct file *filp,
1907                         void __user *buffer, size_t *lenp, loff_t *ppos)
1908 {
1909         int op;
1910
1911         if (!capable(CAP_SYS_MODULE)) {
1912                 return -EPERM;
1913         }
1914
1915         op = is_init(current) ? OP_SET : OP_AND;
1916         return do_proc_dointvec(table,write,filp,buffer,lenp,ppos,
1917                                 do_proc_dointvec_bset_conv,&op);
1918 }
1919
1920 struct do_proc_dointvec_minmax_conv_param {
1921         int *min;
1922         int *max;
1923 };
1924
1925 static int do_proc_dointvec_minmax_conv(int *negp, unsigned long *lvalp, 
1926                                         int *valp, 
1927                                         int write, void *data)
1928 {
1929         struct do_proc_dointvec_minmax_conv_param *param = data;
1930         if (write) {
1931                 int val = *negp ? -*lvalp : *lvalp;
1932                 if ((param->min && *param->min > val) ||
1933                     (param->max && *param->max < val))
1934                         return -EINVAL;
1935                 *valp = val;
1936         } else {
1937                 int val = *valp;
1938                 if (val < 0) {
1939                         *negp = -1;
1940                         *lvalp = (unsigned long)-val;
1941                 } else {
1942                         *negp = 0;
1943                         *lvalp = (unsigned long)val;
1944                 }
1945         }
1946         return 0;
1947 }
1948
1949 /**
1950  * proc_dointvec_minmax - read a vector of integers with min/max values
1951  * @table: the sysctl table
1952  * @write: %TRUE if this is a write to the sysctl file
1953  * @filp: the file structure
1954  * @buffer: the user buffer
1955  * @lenp: the size of the user buffer
1956  * @ppos: file position
1957  *
1958  * Reads/writes up to table->maxlen/sizeof(unsigned int) integer
1959  * values from/to the user buffer, treated as an ASCII string.
1960  *
1961  * This routine will ensure the values are within the range specified by
1962  * table->extra1 (min) and table->extra2 (max).
1963  *
1964  * Returns 0 on success.
1965  */
1966 int proc_dointvec_minmax(ctl_table *table, int write, struct file *filp,
1967                   void __user *buffer, size_t *lenp, loff_t *ppos)
1968 {
1969         struct do_proc_dointvec_minmax_conv_param param = {
1970                 .min = (int *) table->extra1,
1971                 .max = (int *) table->extra2,
1972         };
1973         return do_proc_dointvec(table, write, filp, buffer, lenp, ppos,
1974                                 do_proc_dointvec_minmax_conv, &param);
1975 }
1976
1977 static int do_proc_doulongvec_minmax(ctl_table *table, int write,
1978                                      struct file *filp,
1979                                      void __user *buffer,
1980                                      size_t *lenp, loff_t *ppos,
1981                                      unsigned long convmul,
1982                                      unsigned long convdiv)
1983 {
1984 #define TMPBUFLEN 21
1985         unsigned long *i, *min, *max, val;
1986         int vleft, first=1, neg;
1987         size_t len, left;
1988         char buf[TMPBUFLEN], *p;
1989         char __user *s = buffer;
1990         
1991         if (!table->data || !table->maxlen || !*lenp ||
1992             (*ppos && !write)) {
1993                 *lenp = 0;
1994                 return 0;
1995         }
1996         
1997         i = (unsigned long *) table->data;
1998         min = (unsigned long *) table->extra1;
1999         max = (unsigned long *) table->extra2;
2000         vleft = table->maxlen / sizeof(unsigned long);
2001         left = *lenp;
2002         
2003         for (; left && vleft--; i++, min++, max++, first=0) {
2004                 if (write) {
2005                         while (left) {
2006                                 char c;
2007                                 if (get_user(c, s))
2008                                         return -EFAULT;
2009                                 if (!isspace(c))
2010                                         break;
2011                                 left--;
2012                                 s++;
2013                         }
2014                         if (!left)
2015                                 break;
2016                         neg = 0;
2017                         len = left;
2018                         if (len > TMPBUFLEN-1)
2019                                 len = TMPBUFLEN-1;
2020                         if (copy_from_user(buf, s, len))
2021                                 return -EFAULT;
2022                         buf[len] = 0;
2023                         p = buf;
2024                         if (*p == '-' && left > 1) {
2025                                 neg = 1;
2026                                 left--, p++;
2027                         }
2028                         if (*p < '0' || *p > '9')
2029                                 break;
2030                         val = simple_strtoul(p, &p, 0) * convmul / convdiv ;
2031                         len = p-buf;
2032                         if ((len < left) && *p && !isspace(*p))
2033                                 break;
2034                         if (neg)
2035                                 val = -val;
2036                         s += len;
2037                         left -= len;
2038
2039                         if(neg)
2040                                 continue;
2041                         if ((min && val < *min) || (max && val > *max))
2042                                 continue;
2043                         *i = val;
2044                 } else {
2045                         p = buf;
2046                         if (!first)
2047                                 *p++ = '\t';
2048                         sprintf(p, "%lu", convdiv * (*i) / convmul);
2049                         len = strlen(buf);
2050                         if (len > left)
2051                                 len = left;
2052                         if(copy_to_user(s, buf, len))
2053                                 return -EFAULT;
2054                         left -= len;
2055                         s += len;
2056                 }
2057         }
2058
2059         if (!write && !first && left) {
2060                 if(put_user('\n', s))
2061                         return -EFAULT;
2062                 left--, s++;
2063         }
2064         if (write) {
2065                 while (left) {
2066                         char c;
2067                         if (get_user(c, s++))
2068                                 return -EFAULT;
2069                         if (!isspace(c))
2070                                 break;
2071                         left--;
2072                 }
2073         }
2074         if (write && first)
2075                 return -EINVAL;
2076         *lenp -= left;
2077         *ppos += *lenp;
2078         return 0;
2079 #undef TMPBUFLEN
2080 }
2081
2082 /**
2083  * proc_doulongvec_minmax - read a vector of long integers with min/max values
2084  * @table: the sysctl table
2085  * @write: %TRUE if this is a write to the sysctl file
2086  * @filp: the file structure
2087  * @buffer: the user buffer
2088  * @lenp: the size of the user buffer
2089  * @ppos: file position
2090  *
2091  * Reads/writes up to table->maxlen/sizeof(unsigned long) unsigned long
2092  * values from/to the user buffer, treated as an ASCII string.
2093  *
2094  * This routine will ensure the values are within the range specified by
2095  * table->extra1 (min) and table->extra2 (max).
2096  *
2097  * Returns 0 on success.
2098  */
2099 int proc_doulongvec_minmax(ctl_table *table, int write, struct file *filp,
2100                            void __user *buffer, size_t *lenp, loff_t *ppos)
2101 {
2102     return do_proc_doulongvec_minmax(table, write, filp, buffer, lenp, ppos, 1l, 1l);
2103 }
2104
2105 /**
2106  * proc_doulongvec_ms_jiffies_minmax - read a vector of millisecond values with min/max values
2107  * @table: the sysctl table
2108  * @write: %TRUE if this is a write to the sysctl file
2109  * @filp: the file structure
2110  * @buffer: the user buffer
2111  * @lenp: the size of the user buffer
2112  * @ppos: file position
2113  *
2114  * Reads/writes up to table->maxlen/sizeof(unsigned long) unsigned long
2115  * values from/to the user buffer, treated as an ASCII string. The values
2116  * are treated as milliseconds, and converted to jiffies when they are stored.
2117  *
2118  * This routine will ensure the values are within the range specified by
2119  * table->extra1 (min) and table->extra2 (max).
2120  *
2121  * Returns 0 on success.
2122  */
2123 int proc_doulongvec_ms_jiffies_minmax(ctl_table *table, int write,
2124                                       struct file *filp,
2125                                       void __user *buffer,
2126                                       size_t *lenp, loff_t *ppos)
2127 {
2128     return do_proc_doulongvec_minmax(table, write, filp, buffer,
2129                                      lenp, ppos, HZ, 1000l);
2130 }
2131
2132
2133 static int do_proc_dointvec_jiffies_conv(int *negp, unsigned long *lvalp,
2134                                          int *valp,
2135                                          int write, void *data)
2136 {
2137         if (write) {
2138                 if (*lvalp > LONG_MAX / HZ)
2139                         return 1;
2140                 *valp = *negp ? -(*lvalp*HZ) : (*lvalp*HZ);
2141         } else {
2142                 int val = *valp;
2143                 unsigned long lval;
2144                 if (val < 0) {
2145                         *negp = -1;
2146                         lval = (unsigned long)-val;
2147                 } else {
2148                         *negp = 0;
2149                         lval = (unsigned long)val;
2150                 }
2151                 *lvalp = lval / HZ;
2152         }
2153         return 0;
2154 }
2155
2156 static int do_proc_dointvec_userhz_jiffies_conv(int *negp, unsigned long *lvalp,
2157                                                 int *valp,
2158                                                 int write, void *data)
2159 {
2160         if (write) {
2161                 if (USER_HZ < HZ && *lvalp > (LONG_MAX / HZ) * USER_HZ)
2162                         return 1;
2163                 *valp = clock_t_to_jiffies(*negp ? -*lvalp : *lvalp);
2164         } else {
2165                 int val = *valp;
2166                 unsigned long lval;
2167                 if (val < 0) {
2168                         *negp = -1;
2169                         lval = (unsigned long)-val;
2170                 } else {
2171                         *negp = 0;
2172                         lval = (unsigned long)val;
2173                 }
2174                 *lvalp = jiffies_to_clock_t(lval);
2175         }
2176         return 0;
2177 }
2178
2179 static int do_proc_dointvec_ms_jiffies_conv(int *negp, unsigned long *lvalp,
2180                                             int *valp,
2181                                             int write, void *data)
2182 {
2183         if (write) {
2184                 *valp = msecs_to_jiffies(*negp ? -*lvalp : *lvalp);
2185         } else {
2186                 int val = *valp;
2187                 unsigned long lval;
2188                 if (val < 0) {
2189                         *negp = -1;
2190                         lval = (unsigned long)-val;
2191                 } else {
2192                         *negp = 0;
2193                         lval = (unsigned long)val;
2194                 }
2195                 *lvalp = jiffies_to_msecs(lval);
2196         }
2197         return 0;
2198 }
2199
2200 /**
2201  * proc_dointvec_jiffies - read a vector of integers as seconds
2202  * @table: the sysctl table
2203  * @write: %TRUE if this is a write to the sysctl file
2204  * @filp: the file structure
2205  * @buffer: the user buffer
2206  * @lenp: the size of the user buffer
2207  * @ppos: file position
2208  *
2209  * Reads/writes up to table->maxlen/sizeof(unsigned int) integer
2210  * values from/to the user buffer, treated as an ASCII string. 
2211  * The values read are assumed to be in seconds, and are converted into
2212  * jiffies.
2213  *
2214  * Returns 0 on success.
2215  */
2216 int proc_dointvec_jiffies(ctl_table *table, int write, struct file *filp,
2217                           void __user *buffer, size_t *lenp, loff_t *ppos)
2218 {
2219     return do_proc_dointvec(table,write,filp,buffer,lenp,ppos,
2220                             do_proc_dointvec_jiffies_conv,NULL);
2221 }
2222
2223 /**
2224  * proc_dointvec_userhz_jiffies - read a vector of integers as 1/USER_HZ seconds
2225  * @table: the sysctl table
2226  * @write: %TRUE if this is a write to the sysctl file
2227  * @filp: the file structure
2228  * @buffer: the user buffer
2229  * @lenp: the size of the user buffer
2230  * @ppos: pointer to the file position
2231  *
2232  * Reads/writes up to table->maxlen/sizeof(unsigned int) integer
2233  * values from/to the user buffer, treated as an ASCII string. 
2234  * The values read are assumed to be in 1/USER_HZ seconds, and 
2235  * are converted into jiffies.
2236  *
2237  * Returns 0 on success.
2238  */
2239 int proc_dointvec_userhz_jiffies(ctl_table *table, int write, struct file *filp,
2240                                  void __user *buffer, size_t *lenp, loff_t *ppos)
2241 {
2242     return do_proc_dointvec(table,write,filp,buffer,lenp,ppos,
2243                             do_proc_dointvec_userhz_jiffies_conv,NULL);
2244 }
2245
2246 /**
2247  * proc_dointvec_ms_jiffies - read a vector of integers as 1 milliseconds
2248  * @table: the sysctl table
2249  * @write: %TRUE if this is a write to the sysctl file
2250  * @filp: the file structure
2251  * @buffer: the user buffer
2252  * @lenp: the size of the user buffer
2253  * @ppos: file position
2254  * @ppos: the current position in the file
2255  *
2256  * Reads/writes up to table->maxlen/sizeof(unsigned int) integer
2257  * values from/to the user buffer, treated as an ASCII string. 
2258  * The values read are assumed to be in 1/1000 seconds, and 
2259  * are converted into jiffies.
2260  *
2261  * Returns 0 on success.
2262  */
2263 int proc_dointvec_ms_jiffies(ctl_table *table, int write, struct file *filp,
2264                              void __user *buffer, size_t *lenp, loff_t *ppos)
2265 {
2266         return do_proc_dointvec(table, write, filp, buffer, lenp, ppos,
2267                                 do_proc_dointvec_ms_jiffies_conv, NULL);
2268 }
2269
2270 #else /* CONFIG_PROC_FS */
2271
2272 int proc_dostring(ctl_table *table, int write, struct file *filp,
2273                   void __user *buffer, size_t *lenp, loff_t *ppos)
2274 {
2275         return -ENOSYS;
2276 }
2277
2278 static int proc_doutsstring(ctl_table *table, int write, struct file *filp,
2279                             void __user *buffer, size_t *lenp, loff_t *ppos)
2280 {
2281         return -ENOSYS;
2282 }
2283
2284 int proc_dointvec(ctl_table *table, int write, struct file *filp,
2285                   void __user *buffer, size_t *lenp, loff_t *ppos)
2286 {
2287         return -ENOSYS;
2288 }
2289
2290 int proc_dointvec_bset(ctl_table *table, int write, struct file *filp,
2291                         void __user *buffer, size_t *lenp, loff_t *ppos)
2292 {
2293         return -ENOSYS;
2294 }
2295
2296 int proc_dointvec_minmax(ctl_table *table, int write, struct file *filp,
2297                     void __user *buffer, size_t *lenp, loff_t *ppos)
2298 {
2299         return -ENOSYS;
2300 }
2301
2302 int proc_dointvec_jiffies(ctl_table *table, int write, struct file *filp,
2303                     void __user *buffer, size_t *lenp, loff_t *ppos)
2304 {
2305         return -ENOSYS;
2306 }
2307
2308 int proc_dointvec_userhz_jiffies(ctl_table *table, int write, struct file *filp,
2309                     void __user *buffer, size_t *lenp, loff_t *ppos)
2310 {
2311         return -ENOSYS;
2312 }
2313
2314 int proc_dointvec_ms_jiffies(ctl_table *table, int write, struct file *filp,
2315                              void __user *buffer, size_t *lenp, loff_t *ppos)
2316 {
2317         return -ENOSYS;
2318 }
2319
2320 int proc_doulongvec_minmax(ctl_table *table, int write, struct file *filp,
2321                     void __user *buffer, size_t *lenp, loff_t *ppos)
2322 {
2323         return -ENOSYS;
2324 }
2325
2326 int proc_doulongvec_ms_jiffies_minmax(ctl_table *table, int write,
2327                                       struct file *filp,
2328                                       void __user *buffer,
2329                                       size_t *lenp, loff_t *ppos)
2330 {
2331     return -ENOSYS;
2332 }
2333
2334
2335 #endif /* CONFIG_PROC_FS */
2336
2337
2338 #ifdef CONFIG_SYSCTL_SYSCALL
2339 /*
2340  * General sysctl support routines 
2341  */
2342
2343 /* The generic string strategy routine: */
2344 int sysctl_string(ctl_table *table, int __user *name, int nlen,
2345                   void __user *oldval, size_t __user *oldlenp,
2346                   void __user *newval, size_t newlen, void **context)
2347 {
2348         if (!table->data || !table->maxlen) 
2349                 return -ENOTDIR;
2350         
2351         if (oldval && oldlenp) {
2352                 size_t bufsize;
2353                 if (get_user(bufsize, oldlenp))
2354                         return -EFAULT;
2355                 if (bufsize) {
2356                         size_t len = strlen(table->data), copied;
2357
2358                         /* This shouldn't trigger for a well-formed sysctl */
2359                         if (len > table->maxlen)
2360                                 len = table->maxlen;
2361
2362                         /* Copy up to a max of bufsize-1 bytes of the string */
2363                         copied = (len >= bufsize) ? bufsize - 1 : len;
2364
2365                         if (copy_to_user(oldval, table->data, copied) ||
2366                             put_user(0, (char __user *)(oldval + copied)))
2367                                 return -EFAULT;
2368                         if (put_user(len, oldlenp))
2369                                 return -EFAULT;
2370                 }
2371         }
2372         if (newval && newlen) {
2373                 size_t len = newlen;
2374                 if (len > table->maxlen)
2375                         len = table->maxlen;
2376                 if(copy_from_user(table->data, newval, len))
2377                         return -EFAULT;
2378                 if (len == table->maxlen)
2379                         len--;
2380                 ((char *) table->data)[len] = 0;
2381         }
2382         return 1;
2383 }
2384
2385 /*
2386  * This function makes sure that all of the integers in the vector
2387  * are between the minimum and maximum values given in the arrays
2388  * table->extra1 and table->extra2, respectively.
2389  */
2390 int sysctl_intvec(ctl_table *table, int __user *name, int nlen,
2391                 void __user *oldval, size_t __user *oldlenp,
2392                 void __user *newval, size_t newlen, void **context)
2393 {
2394
2395         if (newval && newlen) {
2396                 int __user *vec = (int __user *) newval;
2397                 int *min = (int *) table->extra1;
2398                 int *max = (int *) table->extra2;
2399                 size_t length;
2400                 int i;
2401
2402                 if (newlen % sizeof(int) != 0)
2403                         return -EINVAL;
2404
2405                 if (!table->extra1 && !table->extra2)
2406                         return 0;
2407
2408                 if (newlen > table->maxlen)
2409                         newlen = table->maxlen;
2410                 length = newlen / sizeof(int);
2411
2412                 for (i = 0; i < length; i++) {
2413                         int value;
2414                         if (get_user(value, vec + i))
2415                                 return -EFAULT;
2416                         if (min && value < min[i])
2417                                 return -EINVAL;
2418                         if (max && value > max[i])
2419                                 return -EINVAL;
2420                 }
2421         }
2422         return 0;
2423 }
2424
2425 /* Strategy function to convert jiffies to seconds */ 
2426 int sysctl_jiffies(ctl_table *table, int __user *name, int nlen,
2427                 void __user *oldval, size_t __user *oldlenp,
2428                 void __user *newval, size_t newlen, void **context)
2429 {
2430         if (oldval) {
2431                 size_t olen;
2432                 if (oldlenp) { 
2433                         if (get_user(olen, oldlenp))
2434                                 return -EFAULT;
2435                         if (olen!=sizeof(int))
2436                                 return -EINVAL; 
2437                 }
2438                 if (put_user(*(int *)(table->data)/HZ, (int __user *)oldval) ||
2439                     (oldlenp && put_user(sizeof(int),oldlenp)))
2440                         return -EFAULT;
2441         }
2442         if (newval && newlen) { 
2443                 int new;
2444                 if (newlen != sizeof(int))
2445                         return -EINVAL; 
2446                 if (get_user(new, (int __user *)newval))
2447                         return -EFAULT;
2448                 *(int *)(table->data) = new*HZ; 
2449         }
2450         return 1;
2451 }
2452
2453 /* Strategy function to convert jiffies to seconds */ 
2454 int sysctl_ms_jiffies(ctl_table *table, int __user *name, int nlen,
2455                 void __user *oldval, size_t __user *oldlenp,
2456                 void __user *newval, size_t newlen, void **context)
2457 {
2458         if (oldval) {
2459                 size_t olen;
2460                 if (oldlenp) { 
2461                         if (get_user(olen, oldlenp))
2462                                 return -EFAULT;
2463                         if (olen!=sizeof(int))
2464                                 return -EINVAL; 
2465                 }
2466                 if (put_user(jiffies_to_msecs(*(int *)(table->data)), (int __user *)oldval) ||
2467                     (oldlenp && put_user(sizeof(int),oldlenp)))
2468                         return -EFAULT;
2469         }
2470         if (newval && newlen) { 
2471                 int new;
2472                 if (newlen != sizeof(int))
2473                         return -EINVAL; 
2474                 if (get_user(new, (int __user *)newval))
2475                         return -EFAULT;
2476                 *(int *)(table->data) = msecs_to_jiffies(new);
2477         }
2478         return 1;
2479 }
2480
2481 #else /* CONFIG_SYSCTL_SYSCALL */
2482
2483
2484 asmlinkage long sys_sysctl(struct __sysctl_args __user *args)
2485 {
2486         static int msg_count;
2487
2488         if (msg_count < 5) {
2489                 msg_count++;
2490                 printk(KERN_INFO
2491                         "warning: process `%s' used the removed sysctl "
2492                         "system call\n", current->comm);
2493         }
2494         return -ENOSYS;
2495 }
2496
2497 int sysctl_string(ctl_table *table, int __user *name, int nlen,
2498                   void __user *oldval, size_t __user *oldlenp,
2499                   void __user *newval, size_t newlen, void **context)
2500 {
2501         return -ENOSYS;
2502 }
2503
2504 int sysctl_intvec(ctl_table *table, int __user *name, int nlen,
2505                 void __user *oldval, size_t __user *oldlenp,
2506                 void __user *newval, size_t newlen, void **context)
2507 {
2508         return -ENOSYS;
2509 }
2510
2511 int sysctl_jiffies(ctl_table *table, int __user *name, int nlen,
2512                 void __user *oldval, size_t __user *oldlenp,
2513                 void __user *newval, size_t newlen, void **context)
2514 {
2515         return -ENOSYS;
2516 }
2517
2518 int sysctl_ms_jiffies(ctl_table *table, int __user *name, int nlen,
2519                 void __user *oldval, size_t __user *oldlenp,
2520                 void __user *newval, size_t newlen, void **context)
2521 {
2522         return -ENOSYS;
2523 }
2524
2525 #endif /* CONFIG_SYSCTL_SYSCALL */
2526
2527 /*
2528  * No sense putting this after each symbol definition, twice,
2529  * exception granted :-)
2530  */
2531 EXPORT_SYMBOL(proc_dointvec);
2532 EXPORT_SYMBOL(proc_dointvec_jiffies);
2533 EXPORT_SYMBOL(proc_dointvec_minmax);
2534 EXPORT_SYMBOL(proc_dointvec_userhz_jiffies);
2535 EXPORT_SYMBOL(proc_dointvec_ms_jiffies);
2536 EXPORT_SYMBOL(proc_dostring);
2537 EXPORT_SYMBOL(proc_doulongvec_minmax);
2538 EXPORT_SYMBOL(proc_doulongvec_ms_jiffies_minmax);
2539 EXPORT_SYMBOL(register_sysctl_table);
2540 EXPORT_SYMBOL(sysctl_intvec);
2541 EXPORT_SYMBOL(sysctl_jiffies);
2542 EXPORT_SYMBOL(sysctl_ms_jiffies);
2543 EXPORT_SYMBOL(sysctl_string);
2544 EXPORT_SYMBOL(unregister_sysctl_table);