projects/linux·blob v2.6.19-rc6 · refs

2762 lines · 64.5 KiB

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
52extern 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. */
62extern int C_A_D;
63extern int sysctl_overcommit_memory;
64extern int sysctl_overcommit_ratio;
65extern int sysctl_panic_on_oom;
66extern int max_threads;
67extern int sysrq_enabled;
68extern int core_uses_pid;
69extern int suid_dumpable;
70extern char core_pattern[];
71extern int pid_max;
72extern int min_free_kbytes;
73extern int printk_ratelimit_jiffies;
74extern int printk_ratelimit_burst;
75extern int pid_max_min, pid_max_max;
76extern int sysctl_drop_caches;
77extern int percpu_pagelist_fraction;
78extern int compat_log;
79
80/* this is needed for the proc_dointvec_minmax for [fs_]overflow UID and GID */
81static int maxolduid = 65535;
82static int minolduid;
83static int min_percpu_pagelist_fract = 8;
84
85static int ngroups_max = NGROUPS_MAX;
86
87#ifdef CONFIG_KMOD
88extern char modprobe_path[];
89#endif
90#ifdef CONFIG_CHR_DEV_SG
91extern int sg_big_buff;
92#endif
93#ifdef CONFIG_SYSVIPC
94static int proc_do_ipc_string(ctl_table *table, int write, struct file *filp,
95 void __user *buffer, size_t *lenp, loff_t *ppos);
96#endif
97
98#ifdef __sparc__
99extern char reboot_command [];
100extern int stop_a_enabled;
101extern int scons_pwroff;
102#endif
103
104#ifdef __hppa__
105extern int pwrsw_enabled;
106extern int unaligned_enabled;
107#endif
108
109#ifdef CONFIG_S390
110#ifdef CONFIG_MATHEMU
111extern int sysctl_ieee_emulation_warnings;
112#endif
113extern int sysctl_userprocess_debug;
114extern int spin_retry;
115#endif
116
117extern int sysctl_hz_timer;
118
119#ifdef CONFIG_BSD_PROCESS_ACCT
120extern int acct_parm[];
121#endif
122
123#ifdef CONFIG_IA64
124extern int no_unaligned_warning;
125#endif
126
127#ifdef CONFIG_RT_MUTEXES
128extern int max_lock_depth;
129#endif
130
131#ifdef CONFIG_SYSCTL_SYSCALL
132static int parse_table(int __user *, int, void __user *, size_t __user *,
133 void __user *, size_t, ctl_table *, void **);
134#endif
135
136static int proc_do_uts_string(ctl_table *table, int write, struct file *filp,
137 void __user *buffer, size_t *lenp, loff_t *ppos);
138
139#ifdef CONFIG_PROC_SYSCTL
140static int proc_do_cad_pid(ctl_table *table, int write, struct file *filp,
141 void __user *buffer, size_t *lenp, loff_t *ppos);
142#endif
143
144static ctl_table root_table[];
145static struct ctl_table_header root_table_header =
146 { root_table, LIST_HEAD_INIT(root_table_header.ctl_entry) };
147
148static ctl_table kern_table[];
149static ctl_table vm_table[];
150static ctl_table fs_table[];
151static ctl_table debug_table[];
152static ctl_table dev_table[];
153extern ctl_table random_table[];
154#ifdef CONFIG_UNIX98_PTYS
155extern ctl_table pty_table[];
156#endif
157#ifdef CONFIG_INOTIFY_USER
158extern ctl_table inotify_table[];
159#endif
160
161#ifdef HAVE_ARCH_PICK_MMAP_LAYOUT
162int sysctl_legacy_va_layout;
163#endif
164
165/* /proc declarations: */
166
167#ifdef CONFIG_PROC_SYSCTL
168
169static ssize_t proc_readsys(struct file *, char __user *, size_t, loff_t *);
170static ssize_t proc_writesys(struct file *, const char __user *, size_t, loff_t *);
171static int proc_opensys(struct inode *, struct file *);
172
173struct file_operations proc_sys_file_operations = {
174 .open = proc_opensys,
175 .read = proc_readsys,
176 .write = proc_writesys,
177};
178
179extern struct proc_dir_entry *proc_sys_root;
180
181static void register_proc_table(ctl_table *, struct proc_dir_entry *, void *);
182static void unregister_proc_table(ctl_table *, struct proc_dir_entry *);
183#endif
184
185/* The default sysctl tables: */
186
187static ctl_table root_table[] = {
188 {
189 .ctl_name = CTL_KERN,
190 .procname = "kernel",
191 .mode = 0555,
192 .child = kern_table,
193 },
194 {
195 .ctl_name = CTL_VM,
196 .procname = "vm",
197 .mode = 0555,
198 .child = vm_table,
199 },
200#ifdef CONFIG_NET
201 {
202 .ctl_name = CTL_NET,
203 .procname = "net",
204 .mode = 0555,
205 .child = net_table,
206 },
207#endif
208 {
209 .ctl_name = CTL_FS,
210 .procname = "fs",
211 .mode = 0555,
212 .child = fs_table,
213 },
214 {
215 .ctl_name = CTL_DEBUG,
216 .procname = "debug",
217 .mode = 0555,
218 .child = debug_table,
219 },
220 {
221 .ctl_name = CTL_DEV,
222 .procname = "dev",
223 .mode = 0555,
224 .child = dev_table,
225 },
226
227 { .ctl_name = 0 }
228};
229
230static ctl_table kern_table[] = {
231#ifndef CONFIG_UTS_NS
232 {
233 .ctl_name = KERN_OSTYPE,
234 .procname = "ostype",
235 .data = init_uts_ns.name.sysname,
236 .maxlen = sizeof(init_uts_ns.name.sysname),
237 .mode = 0444,
238 .proc_handler = &proc_do_uts_string,
239 .strategy = &sysctl_string,
240 },
241 {
242 .ctl_name = KERN_OSRELEASE,
243 .procname = "osrelease",
244 .data = init_uts_ns.name.release,
245 .maxlen = sizeof(init_uts_ns.name.release),
246 .mode = 0444,
247 .proc_handler = &proc_do_uts_string,
248 .strategy = &sysctl_string,
249 },
250 {
251 .ctl_name = KERN_VERSION,
252 .procname = "version",
253 .data = init_uts_ns.name.version,
254 .maxlen = sizeof(init_uts_ns.name.version),
255 .mode = 0444,
256 .proc_handler = &proc_do_uts_string,
257 .strategy = &sysctl_string,
258 },
259 {
260 .ctl_name = KERN_NODENAME,
261 .procname = "hostname",
262 .data = init_uts_ns.name.nodename,
263 .maxlen = sizeof(init_uts_ns.name.nodename),
264 .mode = 0644,
265 .proc_handler = &proc_do_uts_string,
266 .strategy = &sysctl_string,
267 },
268 {
269 .ctl_name = KERN_DOMAINNAME,
270 .procname = "domainname",
271 .data = init_uts_ns.name.domainname,
272 .maxlen = sizeof(init_uts_ns.name.domainname),
273 .mode = 0644,
274 .proc_handler = &proc_do_uts_string,
275 .strategy = &sysctl_string,
276 },
277#else /* !CONFIG_UTS_NS */
278 {
279 .ctl_name = KERN_OSTYPE,
280 .procname = "ostype",
281 .data = NULL,
282 /* could maybe use __NEW_UTS_LEN here? */
283 .maxlen = FIELD_SIZEOF(struct new_utsname, sysname),
284 .mode = 0444,
285 .proc_handler = &proc_do_uts_string,
286 .strategy = &sysctl_string,
287 },
288 {
289 .ctl_name = KERN_OSRELEASE,
290 .procname = "osrelease",
291 .data = NULL,
292 .maxlen = FIELD_SIZEOF(struct new_utsname, release),
293 .mode = 0444,
294 .proc_handler = &proc_do_uts_string,
295 .strategy = &sysctl_string,
296 },
297 {
298 .ctl_name = KERN_VERSION,
299 .procname = "version",
300 .data = NULL,
301 .maxlen = FIELD_SIZEOF(struct new_utsname, version),
302 .mode = 0444,
303 .proc_handler = &proc_do_uts_string,
304 .strategy = &sysctl_string,
305 },
306 {
307 .ctl_name = KERN_NODENAME,
308 .procname = "hostname",
309 .data = NULL,
310 .maxlen = FIELD_SIZEOF(struct new_utsname, nodename),
311 .mode = 0644,
312 .proc_handler = &proc_do_uts_string,
313 .strategy = &sysctl_string,
314 },
315 {
316 .ctl_name = KERN_DOMAINNAME,
317 .procname = "domainname",
318 .data = NULL,
319 .maxlen = FIELD_SIZEOF(struct new_utsname, domainname),
320 .mode = 0644,
321 .proc_handler = &proc_do_uts_string,
322 .strategy = &sysctl_string,
323 },
324#endif /* !CONFIG_UTS_NS */
325 {
326 .ctl_name = KERN_PANIC,
327 .procname = "panic",
328 .data = &panic_timeout,
329 .maxlen = sizeof(int),
330 .mode = 0644,
331 .proc_handler = &proc_dointvec,
332 },
333 {
334 .ctl_name = KERN_CORE_USES_PID,
335 .procname = "core_uses_pid",
336 .data = &core_uses_pid,
337 .maxlen = sizeof(int),
338 .mode = 0644,
339 .proc_handler = &proc_dointvec,
340 },
341 {
342 .ctl_name = KERN_CORE_PATTERN,
343 .procname = "core_pattern",
344 .data = core_pattern,
345 .maxlen = 128,
346 .mode = 0644,
347 .proc_handler = &proc_dostring,
348 .strategy = &sysctl_string,
349 },
350 {
351 .ctl_name = KERN_TAINTED,
352 .procname = "tainted",
353 .data = &tainted,
354 .maxlen = sizeof(int),
355 .mode = 0444,
356 .proc_handler = &proc_dointvec,
357 },
358 {
359 .ctl_name = KERN_CAP_BSET,
360 .procname = "cap-bound",
361 .data = &cap_bset,
362 .maxlen = sizeof(kernel_cap_t),
363 .mode = 0600,
364 .proc_handler = &proc_dointvec_bset,
365 },
366#ifdef CONFIG_BLK_DEV_INITRD
367 {
368 .ctl_name = KERN_REALROOTDEV,
369 .procname = "real-root-dev",
370 .data = &real_root_dev,
371 .maxlen = sizeof(int),
372 .mode = 0644,
373 .proc_handler = &proc_dointvec,
374 },
375#endif
376#ifdef __sparc__
377 {
378 .ctl_name = KERN_SPARC_REBOOT,
379 .procname = "reboot-cmd",
380 .data = reboot_command,
381 .maxlen = 256,
382 .mode = 0644,
383 .proc_handler = &proc_dostring,
384 .strategy = &sysctl_string,
385 },
386 {
387 .ctl_name = KERN_SPARC_STOP_A,
388 .procname = "stop-a",
389 .data = &stop_a_enabled,
390 .maxlen = sizeof (int),
391 .mode = 0644,
392 .proc_handler = &proc_dointvec,
393 },
394 {
395 .ctl_name = KERN_SPARC_SCONS_PWROFF,
396 .procname = "scons-poweroff",
397 .data = &scons_pwroff,
398 .maxlen = sizeof (int),
399 .mode = 0644,
400 .proc_handler = &proc_dointvec,
401 },
402#endif
403#ifdef __hppa__
404 {
405 .ctl_name = KERN_HPPA_PWRSW,
406 .procname = "soft-power",
407 .data = &pwrsw_enabled,
408 .maxlen = sizeof (int),
409 .mode = 0644,
410 .proc_handler = &proc_dointvec,
411 },
412 {
413 .ctl_name = KERN_HPPA_UNALIGNED,
414 .procname = "unaligned-trap",
415 .data = &unaligned_enabled,
416 .maxlen = sizeof (int),
417 .mode = 0644,
418 .proc_handler = &proc_dointvec,
419 },
420#endif
421 {
422 .ctl_name = KERN_CTLALTDEL,
423 .procname = "ctrl-alt-del",
424 .data = &C_A_D,
425 .maxlen = sizeof(int),
426 .mode = 0644,
427 .proc_handler = &proc_dointvec,
428 },
429 {
430 .ctl_name = KERN_PRINTK,
431 .procname = "printk",
432 .data = &console_loglevel,
433 .maxlen = 4*sizeof(int),
434 .mode = 0644,
435 .proc_handler = &proc_dointvec,
436 },
437#ifdef CONFIG_KMOD
438 {
439 .ctl_name = KERN_MODPROBE,
440 .procname = "modprobe",
441 .data = &modprobe_path,
442 .maxlen = KMOD_PATH_LEN,
443 .mode = 0644,
444 .proc_handler = &proc_dostring,
445 .strategy = &sysctl_string,
446 },
447#endif
448#if defined(CONFIG_HOTPLUG) && defined(CONFIG_NET)
449 {
450 .ctl_name = KERN_HOTPLUG,
451 .procname = "hotplug",
452 .data = &uevent_helper,
453 .maxlen = UEVENT_HELPER_PATH_LEN,
454 .mode = 0644,
455 .proc_handler = &proc_dostring,
456 .strategy = &sysctl_string,
457 },
458#endif
459#ifdef CONFIG_CHR_DEV_SG
460 {
461 .ctl_name = KERN_SG_BIG_BUFF,
462 .procname = "sg-big-buff",
463 .data = &sg_big_buff,
464 .maxlen = sizeof (int),
465 .mode = 0444,
466 .proc_handler = &proc_dointvec,
467 },
468#endif
469#ifdef CONFIG_BSD_PROCESS_ACCT
470 {
471 .ctl_name = KERN_ACCT,
472 .procname = "acct",
473 .data = &acct_parm,
474 .maxlen = 3*sizeof(int),
475 .mode = 0644,
476 .proc_handler = &proc_dointvec,
477 },
478#endif
479#ifdef CONFIG_SYSVIPC
480 {
481 .ctl_name = KERN_SHMMAX,
482 .procname = "shmmax",
483 .data = NULL,
484 .maxlen = sizeof (size_t),
485 .mode = 0644,
486 .proc_handler = &proc_do_ipc_string,
487 },
488 {
489 .ctl_name = KERN_SHMALL,
490 .procname = "shmall",
491 .data = NULL,
492 .maxlen = sizeof (size_t),
493 .mode = 0644,
494 .proc_handler = &proc_do_ipc_string,
495 },
496 {
497 .ctl_name = KERN_SHMMNI,
498 .procname = "shmmni",
499 .data = NULL,
500 .maxlen = sizeof (int),
501 .mode = 0644,
502 .proc_handler = &proc_do_ipc_string,
503 },
504 {
505 .ctl_name = KERN_MSGMAX,
506 .procname = "msgmax",
507 .data = NULL,
508 .maxlen = sizeof (int),
509 .mode = 0644,
510 .proc_handler = &proc_do_ipc_string,
511 },
512 {
513 .ctl_name = KERN_MSGMNI,
514 .procname = "msgmni",
515 .data = NULL,
516 .maxlen = sizeof (int),
517 .mode = 0644,
518 .proc_handler = &proc_do_ipc_string,
519 },
520 {
521 .ctl_name = KERN_MSGMNB,
522 .procname = "msgmnb",
523 .data = NULL,
524 .maxlen = sizeof (int),
525 .mode = 0644,
526 .proc_handler = &proc_do_ipc_string,
527 },
528 {
529 .ctl_name = KERN_SEM,
530 .procname = "sem",
531 .data = NULL,
532 .maxlen = 4*sizeof (int),
533 .mode = 0644,
534 .proc_handler = &proc_do_ipc_string,
535 },
536#endif
537#ifdef CONFIG_MAGIC_SYSRQ
538 {
539 .ctl_name = KERN_SYSRQ,
540 .procname = "sysrq",
541 .data = &sysrq_enabled,
542 .maxlen = sizeof (int),
543 .mode = 0644,
544 .proc_handler = &proc_dointvec,
545 },
546#endif
547#ifdef CONFIG_PROC_SYSCTL
548 {
549 .ctl_name = KERN_CADPID,
550 .procname = "cad_pid",
551 .data = NULL,
552 .maxlen = sizeof (int),
553 .mode = 0600,
554 .proc_handler = &proc_do_cad_pid,
555 },
556#endif
557 {
558 .ctl_name = KERN_MAX_THREADS,
559 .procname = "threads-max",
560 .data = &max_threads,
561 .maxlen = sizeof(int),
562 .mode = 0644,
563 .proc_handler = &proc_dointvec,
564 },
565 {
566 .ctl_name = KERN_RANDOM,
567 .procname = "random",
568 .mode = 0555,
569 .child = random_table,
570 },
571#ifdef CONFIG_UNIX98_PTYS
572 {
573 .ctl_name = KERN_PTY,
574 .procname = "pty",
575 .mode = 0555,
576 .child = pty_table,
577 },
578#endif
579 {
580 .ctl_name = KERN_OVERFLOWUID,
581 .procname = "overflowuid",
582 .data = &overflowuid,
583 .maxlen = sizeof(int),
584 .mode = 0644,
585 .proc_handler = &proc_dointvec_minmax,
586 .strategy = &sysctl_intvec,
587 .extra1 = &minolduid,
588 .extra2 = &maxolduid,
589 },
590 {
591 .ctl_name = KERN_OVERFLOWGID,
592 .procname = "overflowgid",
593 .data = &overflowgid,
594 .maxlen = sizeof(int),
595 .mode = 0644,
596 .proc_handler = &proc_dointvec_minmax,
597 .strategy = &sysctl_intvec,
598 .extra1 = &minolduid,
599 .extra2 = &maxolduid,
600 },
601#ifdef CONFIG_S390
602#ifdef CONFIG_MATHEMU
603 {
604 .ctl_name = KERN_IEEE_EMULATION_WARNINGS,
605 .procname = "ieee_emulation_warnings",
606 .data = &sysctl_ieee_emulation_warnings,
607 .maxlen = sizeof(int),
608 .mode = 0644,
609 .proc_handler = &proc_dointvec,
610 },
611#endif
612#ifdef CONFIG_NO_IDLE_HZ
613 {
614 .ctl_name = KERN_HZ_TIMER,
615 .procname = "hz_timer",
616 .data = &sysctl_hz_timer,
617 .maxlen = sizeof(int),
618 .mode = 0644,
619 .proc_handler = &proc_dointvec,
620 },
621#endif
622 {
623 .ctl_name = KERN_S390_USER_DEBUG_LOGGING,
624 .procname = "userprocess_debug",
625 .data = &sysctl_userprocess_debug,
626 .maxlen = sizeof(int),
627 .mode = 0644,
628 .proc_handler = &proc_dointvec,
629 },
630#endif
631 {
632 .ctl_name = KERN_PIDMAX,
633 .procname = "pid_max",
634 .data = &pid_max,
635 .maxlen = sizeof (int),
636 .mode = 0644,
637 .proc_handler = &proc_dointvec_minmax,
638 .strategy = sysctl_intvec,
639 .extra1 = &pid_max_min,
640 .extra2 = &pid_max_max,
641 },
642 {
643 .ctl_name = KERN_PANIC_ON_OOPS,
644 .procname = "panic_on_oops",
645 .data = &panic_on_oops,
646 .maxlen = sizeof(int),
647 .mode = 0644,
648 .proc_handler = &proc_dointvec,
649 },
650 {
651 .ctl_name = KERN_PRINTK_RATELIMIT,
652 .procname = "printk_ratelimit",
653 .data = &printk_ratelimit_jiffies,
654 .maxlen = sizeof(int),
655 .mode = 0644,
656 .proc_handler = &proc_dointvec_jiffies,
657 .strategy = &sysctl_jiffies,
658 },
659 {
660 .ctl_name = KERN_PRINTK_RATELIMIT_BURST,
661 .procname = "printk_ratelimit_burst",
662 .data = &printk_ratelimit_burst,
663 .maxlen = sizeof(int),
664 .mode = 0644,
665 .proc_handler = &proc_dointvec,
666 },
667 {
668 .ctl_name = KERN_NGROUPS_MAX,
669 .procname = "ngroups_max",
670 .data = &ngroups_max,
671 .maxlen = sizeof (int),
672 .mode = 0444,
673 .proc_handler = &proc_dointvec,
674 },
675#if defined(CONFIG_X86_LOCAL_APIC) && defined(CONFIG_X86)
676 {
677 .ctl_name = KERN_UNKNOWN_NMI_PANIC,
678 .procname = "unknown_nmi_panic",
679 .data = &unknown_nmi_panic,
680 .maxlen = sizeof (int),
681 .mode = 0644,
682 .proc_handler = &proc_dointvec,
683 },
684 {
685 .ctl_name = KERN_NMI_WATCHDOG,
686 .procname = "nmi_watchdog",
687 .data = &nmi_watchdog_enabled,
688 .maxlen = sizeof (int),
689 .mode = 0644,
690 .proc_handler = &proc_nmi_enabled,
691 },
692#endif
693#if defined(CONFIG_X86)
694 {
695 .ctl_name = KERN_PANIC_ON_NMI,
696 .procname = "panic_on_unrecovered_nmi",
697 .data = &panic_on_unrecovered_nmi,
698 .maxlen = sizeof(int),
699 .mode = 0644,
700 .proc_handler = &proc_dointvec,
701 },
702 {
703 .ctl_name = KERN_BOOTLOADER_TYPE,
704 .procname = "bootloader_type",
705 .data = &bootloader_type,
706 .maxlen = sizeof (int),
707 .mode = 0444,
708 .proc_handler = &proc_dointvec,
709 },
710#endif
711#if defined(CONFIG_MMU)
712 {
713 .ctl_name = KERN_RANDOMIZE,
714 .procname = "randomize_va_space",
715 .data = &randomize_va_space,
716 .maxlen = sizeof(int),
717 .mode = 0644,
718 .proc_handler = &proc_dointvec,
719 },
720#endif
721#if defined(CONFIG_S390) && defined(CONFIG_SMP)
722 {
723 .ctl_name = KERN_SPIN_RETRY,
724 .procname = "spin_retry",
725 .data = &spin_retry,
726 .maxlen = sizeof (int),
727 .mode = 0644,
728 .proc_handler = &proc_dointvec,
729 },
730#endif
731#ifdef CONFIG_ACPI_SLEEP
732 {
733 .ctl_name = KERN_ACPI_VIDEO_FLAGS,
734 .procname = "acpi_video_flags",
735 .data = &acpi_video_flags,
736 .maxlen = sizeof (unsigned long),
737 .mode = 0644,
738 .proc_handler = &proc_doulongvec_minmax,
739 },
740#endif
741#ifdef CONFIG_IA64
742 {
743 .ctl_name = KERN_IA64_UNALIGNED,
744 .procname = "ignore-unaligned-usertrap",
745 .data = &no_unaligned_warning,
746 .maxlen = sizeof (int),
747 .mode = 0644,
748 .proc_handler = &proc_dointvec,
749 },
750#endif
751#ifdef CONFIG_COMPAT
752 {
753 .ctl_name = KERN_COMPAT_LOG,
754 .procname = "compat-log",
755 .data = &compat_log,
756 .maxlen = sizeof (int),
757 .mode = 0644,
758 .proc_handler = &proc_dointvec,
759 },
760#endif
761#ifdef CONFIG_RT_MUTEXES
762 {
763 .ctl_name = KERN_MAX_LOCK_DEPTH,
764 .procname = "max_lock_depth",
765 .data = &max_lock_depth,
766 .maxlen = sizeof(int),
767 .mode = 0644,
768 .proc_handler = &proc_dointvec,
769 },
770#endif
771
772 { .ctl_name = 0 }
773};
774
775/* Constants for minimum and maximum testing in vm_table.
776 We use these as one-element integer vectors. */
777static int zero;
778static int one_hundred = 100;
779
780
781static ctl_table vm_table[] = {
782 {
783 .ctl_name = VM_OVERCOMMIT_MEMORY,
784 .procname = "overcommit_memory",
785 .data = &sysctl_overcommit_memory,
786 .maxlen = sizeof(sysctl_overcommit_memory),
787 .mode = 0644,
788 .proc_handler = &proc_dointvec,
789 },
790 {
791 .ctl_name = VM_PANIC_ON_OOM,
792 .procname = "panic_on_oom",
793 .data = &sysctl_panic_on_oom,
794 .maxlen = sizeof(sysctl_panic_on_oom),
795 .mode = 0644,
796 .proc_handler = &proc_dointvec,
797 },
798 {
799 .ctl_name = VM_OVERCOMMIT_RATIO,
800 .procname = "overcommit_ratio",
801 .data = &sysctl_overcommit_ratio,
802 .maxlen = sizeof(sysctl_overcommit_ratio),
803 .mode = 0644,
804 .proc_handler = &proc_dointvec,
805 },
806 {
807 .ctl_name = VM_PAGE_CLUSTER,
808 .procname = "page-cluster",
809 .data = &page_cluster,
810 .maxlen = sizeof(int),
811 .mode = 0644,
812 .proc_handler = &proc_dointvec,
813 },
814 {
815 .ctl_name = VM_DIRTY_BACKGROUND,
816 .procname = "dirty_background_ratio",
817 .data = &dirty_background_ratio,
818 .maxlen = sizeof(dirty_background_ratio),
819 .mode = 0644,
820 .proc_handler = &proc_dointvec_minmax,
821 .strategy = &sysctl_intvec,
822 .extra1 = &zero,
823 .extra2 = &one_hundred,
824 },
825 {
826 .ctl_name = VM_DIRTY_RATIO,
827 .procname = "dirty_ratio",
828 .data = &vm_dirty_ratio,
829 .maxlen = sizeof(vm_dirty_ratio),
830 .mode = 0644,
831 .proc_handler = &proc_dointvec_minmax,
832 .strategy = &sysctl_intvec,
833 .extra1 = &zero,
834 .extra2 = &one_hundred,
835 },
836 {
837 .ctl_name = VM_DIRTY_WB_CS,
838 .procname = "dirty_writeback_centisecs",
839 .data = &dirty_writeback_interval,
840 .maxlen = sizeof(dirty_writeback_interval),
841 .mode = 0644,
842 .proc_handler = &dirty_writeback_centisecs_handler,
843 },
844 {
845 .ctl_name = VM_DIRTY_EXPIRE_CS,
846 .procname = "dirty_expire_centisecs",
847 .data = &dirty_expire_interval,
848 .maxlen = sizeof(dirty_expire_interval),
849 .mode = 0644,
850 .proc_handler = &proc_dointvec_userhz_jiffies,
851 },
852 {
853 .ctl_name = VM_NR_PDFLUSH_THREADS,
854 .procname = "nr_pdflush_threads",
855 .data = &nr_pdflush_threads,
856 .maxlen = sizeof nr_pdflush_threads,
857 .mode = 0444 /* read-only*/,
858 .proc_handler = &proc_dointvec,
859 },
860 {
861 .ctl_name = VM_SWAPPINESS,
862 .procname = "swappiness",
863 .data = &vm_swappiness,
864 .maxlen = sizeof(vm_swappiness),
865 .mode = 0644,
866 .proc_handler = &proc_dointvec_minmax,
867 .strategy = &sysctl_intvec,
868 .extra1 = &zero,
869 .extra2 = &one_hundred,
870 },
871#ifdef CONFIG_HUGETLB_PAGE
872 {
873 .ctl_name = VM_HUGETLB_PAGES,
874 .procname = "nr_hugepages",
875 .data = &max_huge_pages,
876 .maxlen = sizeof(unsigned long),
877 .mode = 0644,
878 .proc_handler = &hugetlb_sysctl_handler,
879 .extra1 = (void *)&hugetlb_zero,
880 .extra2 = (void *)&hugetlb_infinity,
881 },
882 {
883 .ctl_name = VM_HUGETLB_GROUP,
884 .procname = "hugetlb_shm_group",
885 .data = &sysctl_hugetlb_shm_group,
886 .maxlen = sizeof(gid_t),
887 .mode = 0644,
888 .proc_handler = &proc_dointvec,
889 },
890#endif
891 {
892 .ctl_name = VM_LOWMEM_RESERVE_RATIO,
893 .procname = "lowmem_reserve_ratio",
894 .data = &sysctl_lowmem_reserve_ratio,
895 .maxlen = sizeof(sysctl_lowmem_reserve_ratio),
896 .mode = 0644,
897 .proc_handler = &lowmem_reserve_ratio_sysctl_handler,
898 .strategy = &sysctl_intvec,
899 },
900 {
901 .ctl_name = VM_DROP_PAGECACHE,
902 .procname = "drop_caches",
903 .data = &sysctl_drop_caches,
904 .maxlen = sizeof(int),
905 .mode = 0644,
906 .proc_handler = drop_caches_sysctl_handler,
907 .strategy = &sysctl_intvec,
908 },
909 {
910 .ctl_name = VM_MIN_FREE_KBYTES,
911 .procname = "min_free_kbytes",
912 .data = &min_free_kbytes,
913 .maxlen = sizeof(min_free_kbytes),
914 .mode = 0644,
915 .proc_handler = &min_free_kbytes_sysctl_handler,
916 .strategy = &sysctl_intvec,
917 .extra1 = &zero,
918 },
919 {
920 .ctl_name = VM_PERCPU_PAGELIST_FRACTION,
921 .procname = "percpu_pagelist_fraction",
922 .data = &percpu_pagelist_fraction,
923 .maxlen = sizeof(percpu_pagelist_fraction),
924 .mode = 0644,
925 .proc_handler = &percpu_pagelist_fraction_sysctl_handler,
926 .strategy = &sysctl_intvec,
927 .extra1 = &min_percpu_pagelist_fract,
928 },
929#ifdef CONFIG_MMU
930 {
931 .ctl_name = VM_MAX_MAP_COUNT,
932 .procname = "max_map_count",
933 .data = &sysctl_max_map_count,
934 .maxlen = sizeof(sysctl_max_map_count),
935 .mode = 0644,
936 .proc_handler = &proc_dointvec
937 },
938#endif
939 {
940 .ctl_name = VM_LAPTOP_MODE,
941 .procname = "laptop_mode",
942 .data = &laptop_mode,
943 .maxlen = sizeof(laptop_mode),
944 .mode = 0644,
945 .proc_handler = &proc_dointvec_jiffies,
946 .strategy = &sysctl_jiffies,
947 },
948 {
949 .ctl_name = VM_BLOCK_DUMP,
950 .procname = "block_dump",
951 .data = &block_dump,
952 .maxlen = sizeof(block_dump),
953 .mode = 0644,
954 .proc_handler = &proc_dointvec,
955 .strategy = &sysctl_intvec,
956 .extra1 = &zero,
957 },
958 {
959 .ctl_name = VM_VFS_CACHE_PRESSURE,
960 .procname = "vfs_cache_pressure",
961 .data = &sysctl_vfs_cache_pressure,
962 .maxlen = sizeof(sysctl_vfs_cache_pressure),
963 .mode = 0644,
964 .proc_handler = &proc_dointvec,
965 .strategy = &sysctl_intvec,
966 .extra1 = &zero,
967 },
968#ifdef HAVE_ARCH_PICK_MMAP_LAYOUT
969 {
970 .ctl_name = VM_LEGACY_VA_LAYOUT,
971 .procname = "legacy_va_layout",
972 .data = &sysctl_legacy_va_layout,
973 .maxlen = sizeof(sysctl_legacy_va_layout),
974 .mode = 0644,
975 .proc_handler = &proc_dointvec,
976 .strategy = &sysctl_intvec,
977 .extra1 = &zero,
978 },
979#endif
980#ifdef CONFIG_SWAP
981 {
982 .ctl_name = VM_SWAP_TOKEN_TIMEOUT,
983 .procname = "swap_token_timeout",
984 .data = &swap_token_default_timeout,
985 .maxlen = sizeof(swap_token_default_timeout),
986 .mode = 0644,
987 .proc_handler = &proc_dointvec_jiffies,
988 .strategy = &sysctl_jiffies,
989 },
990#endif
991#ifdef CONFIG_NUMA
992 {
993 .ctl_name = VM_ZONE_RECLAIM_MODE,
994 .procname = "zone_reclaim_mode",
995 .data = &zone_reclaim_mode,
996 .maxlen = sizeof(zone_reclaim_mode),
997 .mode = 0644,
998 .proc_handler = &proc_dointvec,
999 .strategy = &sysctl_intvec,
1000 .extra1 = &zero,
1001 },
1002 {
1003 .ctl_name = VM_MIN_UNMAPPED,
1004 .procname = "min_unmapped_ratio",
1005 .data = &sysctl_min_unmapped_ratio,
1006 .maxlen = sizeof(sysctl_min_unmapped_ratio),
1007 .mode = 0644,
1008 .proc_handler = &sysctl_min_unmapped_ratio_sysctl_handler,
1009 .strategy = &sysctl_intvec,
1010 .extra1 = &zero,
1011 .extra2 = &one_hundred,
1012 },
1013 {
1014 .ctl_name = VM_MIN_SLAB,
1015 .procname = "min_slab_ratio",
1016 .data = &sysctl_min_slab_ratio,
1017 .maxlen = sizeof(sysctl_min_slab_ratio),
1018 .mode = 0644,
1019 .proc_handler = &sysctl_min_slab_ratio_sysctl_handler,
1020 .strategy = &sysctl_intvec,
1021 .extra1 = &zero,
1022 .extra2 = &one_hundred,
1023 },
1024#endif
1025#ifdef CONFIG_X86_32
1026 {
1027 .ctl_name = VM_VDSO_ENABLED,
1028 .procname = "vdso_enabled",
1029 .data = &vdso_enabled,
1030 .maxlen = sizeof(vdso_enabled),
1031 .mode = 0644,
1032 .proc_handler = &proc_dointvec,
1033 .strategy = &sysctl_intvec,
1034 .extra1 = &zero,
1035 },
1036#endif
1037 { .ctl_name = 0 }
1038};
1039
1040static ctl_table fs_table[] = {
1041 {
1042 .ctl_name = FS_NRINODE,
1043 .procname = "inode-nr",
1044 .data = &inodes_stat,
1045 .maxlen = 2*sizeof(int),
1046 .mode = 0444,
1047 .proc_handler = &proc_dointvec,
1048 },
1049 {
1050 .ctl_name = FS_STATINODE,
1051 .procname = "inode-state",
1052 .data = &inodes_stat,
1053 .maxlen = 7*sizeof(int),
1054 .mode = 0444,
1055 .proc_handler = &proc_dointvec,
1056 },
1057 {
1058 .ctl_name = FS_NRFILE,
1059 .procname = "file-nr",
1060 .data = &files_stat,
1061 .maxlen = 3*sizeof(int),
1062 .mode = 0444,
1063 .proc_handler = &proc_nr_files,
1064 },
1065 {
1066 .ctl_name = FS_MAXFILE,
1067 .procname = "file-max",
1068 .data = &files_stat.max_files,
1069 .maxlen = sizeof(int),
1070 .mode = 0644,
1071 .proc_handler = &proc_dointvec,
1072 },
1073 {
1074 .ctl_name = FS_DENTRY,
1075 .procname = "dentry-state",
1076 .data = &dentry_stat,
1077 .maxlen = 6*sizeof(int),
1078 .mode = 0444,
1079 .proc_handler = &proc_dointvec,
1080 },
1081 {
1082 .ctl_name = FS_OVERFLOWUID,
1083 .procname = "overflowuid",
1084 .data = &fs_overflowuid,
1085 .maxlen = sizeof(int),
1086 .mode = 0644,
1087 .proc_handler = &proc_dointvec_minmax,
1088 .strategy = &sysctl_intvec,
1089 .extra1 = &minolduid,
1090 .extra2 = &maxolduid,
1091 },
1092 {
1093 .ctl_name = FS_OVERFLOWGID,
1094 .procname = "overflowgid",
1095 .data = &fs_overflowgid,
1096 .maxlen = sizeof(int),
1097 .mode = 0644,
1098 .proc_handler = &proc_dointvec_minmax,
1099 .strategy = &sysctl_intvec,
1100 .extra1 = &minolduid,
1101 .extra2 = &maxolduid,
1102 },
1103 {
1104 .ctl_name = FS_LEASES,
1105 .procname = "leases-enable",
1106 .data = &leases_enable,
1107 .maxlen = sizeof(int),
1108 .mode = 0644,
1109 .proc_handler = &proc_dointvec,
1110 },
1111#ifdef CONFIG_DNOTIFY
1112 {
1113 .ctl_name = FS_DIR_NOTIFY,
1114 .procname = "dir-notify-enable",
1115 .data = &dir_notify_enable,
1116 .maxlen = sizeof(int),
1117 .mode = 0644,
1118 .proc_handler = &proc_dointvec,
1119 },
1120#endif
1121#ifdef CONFIG_MMU
1122 {
1123 .ctl_name = FS_LEASE_TIME,
1124 .procname = "lease-break-time",
1125 .data = &lease_break_time,
1126 .maxlen = sizeof(int),
1127 .mode = 0644,
1128 .proc_handler = &proc_dointvec,
1129 },
1130 {
1131 .ctl_name = FS_AIO_NR,
1132 .procname = "aio-nr",
1133 .data = &aio_nr,
1134 .maxlen = sizeof(aio_nr),
1135 .mode = 0444,
1136 .proc_handler = &proc_doulongvec_minmax,
1137 },
1138 {
1139 .ctl_name = FS_AIO_MAX_NR,
1140 .procname = "aio-max-nr",
1141 .data = &aio_max_nr,
1142 .maxlen = sizeof(aio_max_nr),
1143 .mode = 0644,
1144 .proc_handler = &proc_doulongvec_minmax,
1145 },
1146#ifdef CONFIG_INOTIFY_USER
1147 {
1148 .ctl_name = FS_INOTIFY,
1149 .procname = "inotify",
1150 .mode = 0555,
1151 .child = inotify_table,
1152 },
1153#endif
1154#endif
1155 {
1156 .ctl_name = KERN_SETUID_DUMPABLE,
1157 .procname = "suid_dumpable",
1158 .data = &suid_dumpable,
1159 .maxlen = sizeof(int),
1160 .mode = 0644,
1161 .proc_handler = &proc_dointvec,
1162 },
1163 { .ctl_name = 0 }
1164};
1165
1166static ctl_table debug_table[] = {
1167 { .ctl_name = 0 }
1168};
1169
1170static ctl_table dev_table[] = {
1171 { .ctl_name = 0 }
1172};
1173
1174extern void init_irq_proc (void);
1175
1176static DEFINE_SPINLOCK(sysctl_lock);
1177
1178/* called under sysctl_lock */
1179static int use_table(struct ctl_table_header *p)
1180{
1181 if (unlikely(p->unregistering))
1182 return 0;
1183 p->used++;
1184 return 1;
1185}
1186
1187/* called under sysctl_lock */
1188static void unuse_table(struct ctl_table_header *p)
1189{
1190 if (!--p->used)
1191 if (unlikely(p->unregistering))
1192 complete(p->unregistering);
1193}
1194
1195/* called under sysctl_lock, will reacquire if has to wait */
1196static void start_unregistering(struct ctl_table_header *p)
1197{
1198 /*
1199 * if p->used is 0, nobody will ever touch that entry again;
1200 * we'll eliminate all paths to it before dropping sysctl_lock
1201 */
1202 if (unlikely(p->used)) {
1203 struct completion wait;
1204 init_completion(&wait);
1205 p->unregistering = &wait;
1206 spin_unlock(&sysctl_lock);
1207 wait_for_completion(&wait);
1208 spin_lock(&sysctl_lock);
1209 }
1210 /*
1211 * do not remove from the list until nobody holds it; walking the
1212 * list in do_sysctl() relies on that.
1213 */
1214 list_del_init(&p->ctl_entry);
1215}
1216
1217void __init sysctl_init(void)
1218{
1219#ifdef CONFIG_PROC_SYSCTL
1220 register_proc_table(root_table, proc_sys_root, &root_table_header);
1221 init_irq_proc();
1222#endif
1223}
1224
1225#ifdef CONFIG_SYSCTL_SYSCALL
1226int do_sysctl(int __user *name, int nlen, void __user *oldval, size_t __user *oldlenp,
1227 void __user *newval, size_t newlen)
1228{
1229 struct list_head *tmp;
1230 int error = -ENOTDIR;
1231
1232 if (nlen <= 0 || nlen >= CTL_MAXNAME)
1233 return -ENOTDIR;
1234 if (oldval) {
1235 int old_len;
1236 if (!oldlenp || get_user(old_len, oldlenp))
1237 return -EFAULT;
1238 }
1239 spin_lock(&sysctl_lock);
1240 tmp = &root_table_header.ctl_entry;
1241 do {
1242 struct ctl_table_header *head =
1243 list_entry(tmp, struct ctl_table_header, ctl_entry);
1244 void *context = NULL;
1245
1246 if (!use_table(head))
1247 continue;
1248
1249 spin_unlock(&sysctl_lock);
1250
1251 error = parse_table(name, nlen, oldval, oldlenp,
1252 newval, newlen, head->ctl_table,
1253 &context);
1254 kfree(context);
1255
1256 spin_lock(&sysctl_lock);
1257 unuse_table(head);
1258 if (error != -ENOTDIR)
1259 break;
1260 } while ((tmp = tmp->next) != &root_table_header.ctl_entry);
1261 spin_unlock(&sysctl_lock);
1262 return error;
1263}
1264
1265asmlinkage long sys_sysctl(struct __sysctl_args __user *args)
1266{
1267 struct __sysctl_args tmp;
1268 int error;
1269
1270 if (copy_from_user(&tmp, args, sizeof(tmp)))
1271 return -EFAULT;
1272
1273 lock_kernel();
1274 error = do_sysctl(tmp.name, tmp.nlen, tmp.oldval, tmp.oldlenp,
1275 tmp.newval, tmp.newlen);
1276 unlock_kernel();
1277 return error;
1278}
1279#endif /* CONFIG_SYSCTL_SYSCALL */
1280
1281/*
1282 * ctl_perm does NOT grant the superuser all rights automatically, because
1283 * some sysctl variables are readonly even to root.
1284 */
1285
1286static int test_perm(int mode, int op)
1287{
1288 if (!current->euid)
1289 mode >>= 6;
1290 else if (in_egroup_p(0))
1291 mode >>= 3;
1292 if ((mode & op & 0007) == op)
1293 return 0;
1294 return -EACCES;
1295}
1296
1297static inline int ctl_perm(ctl_table *table, int op)
1298{
1299 int error;
1300 error = security_sysctl(table, op);
1301 if (error)
1302 return error;
1303 return test_perm(table->mode, op);
1304}
1305
1306#ifdef CONFIG_SYSCTL_SYSCALL
1307static int parse_table(int __user *name, int nlen,
1308 void __user *oldval, size_t __user *oldlenp,
1309 void __user *newval, size_t newlen,
1310 ctl_table *table, void **context)
1311{
1312 int n;
1313repeat:
1314 if (!nlen)
1315 return -ENOTDIR;
1316 if (get_user(n, name))
1317 return -EFAULT;
1318 for ( ; table->ctl_name || table->procname; table++) {
1319 if (!table->ctl_name)
1320 continue;
1321 if (n == table->ctl_name || table->ctl_name == CTL_ANY) {
1322 int error;
1323 if (table->child) {
1324 if (ctl_perm(table, 001))
1325 return -EPERM;
1326 if (table->strategy) {
1327 error = table->strategy(
1328 table, name, nlen,
1329 oldval, oldlenp,
1330 newval, newlen, context);
1331 if (error)
1332 return error;
1333 }
1334 name++;
1335 nlen--;
1336 table = table->child;
1337 goto repeat;
1338 }
1339 error = do_sysctl_strategy(table, name, nlen,
1340 oldval, oldlenp,
1341 newval, newlen, context);
1342 return error;
1343 }
1344 }
1345 return -ENOTDIR;
1346}
1347
1348/* Perform the actual read/write of a sysctl table entry. */
1349int do_sysctl_strategy (ctl_table *table,
1350 int __user *name, int nlen,
1351 void __user *oldval, size_t __user *oldlenp,
1352 void __user *newval, size_t newlen, void **context)
1353{
1354 int op = 0, rc;
1355 size_t len;
1356
1357 if (oldval)
1358 op |= 004;
1359 if (newval)
1360 op |= 002;
1361 if (ctl_perm(table, op))
1362 return -EPERM;
1363
1364 if (table->strategy) {
1365 rc = table->strategy(table, name, nlen, oldval, oldlenp,
1366 newval, newlen, context);
1367 if (rc < 0)
1368 return rc;
1369 if (rc > 0)
1370 return 0;
1371 }
1372
1373 /* If there is no strategy routine, or if the strategy returns
1374 * zero, proceed with automatic r/w */
1375 if (table->data && table->maxlen) {
1376 if (oldval && oldlenp) {
1377 if (get_user(len, oldlenp))
1378 return -EFAULT;
1379 if (len) {
1380 if (len > table->maxlen)
1381 len = table->maxlen;
1382 if(copy_to_user(oldval, table->data, len))
1383 return -EFAULT;
1384 if(put_user(len, oldlenp))
1385 return -EFAULT;
1386 }
1387 }
1388 if (newval && newlen) {
1389 len = newlen;
1390 if (len > table->maxlen)
1391 len = table->maxlen;
1392 if(copy_from_user(table->data, newval, len))
1393 return -EFAULT;
1394 }
1395 }
1396 return 0;
1397}
1398#endif /* CONFIG_SYSCTL_SYSCALL */
1399
1400/**
1401 * register_sysctl_table - register a sysctl hierarchy
1402 * @table: the top-level table structure
1403 * @insert_at_head: whether the entry should be inserted in front or at the end
1404 *
1405 * Register a sysctl table hierarchy. @table should be a filled in ctl_table
1406 * array. An entry with a ctl_name of 0 terminates the table.
1407 *
1408 * The members of the &ctl_table structure are used as follows:
1409 *
1410 * ctl_name - This is the numeric sysctl value used by sysctl(2). The number
1411 * must be unique within that level of sysctl
1412 *
1413 * procname - the name of the sysctl file under /proc/sys. Set to %NULL to not
1414 * enter a sysctl file
1415 *
1416 * data - a pointer to data for use by proc_handler
1417 *
1418 * maxlen - the maximum size in bytes of the data
1419 *
1420 * mode - the file permissions for the /proc/sys file, and for sysctl(2)
1421 *
1422 * child - a pointer to the child sysctl table if this entry is a directory, or
1423 * %NULL.
1424 *
1425 * proc_handler - the text handler routine (described below)
1426 *
1427 * strategy - the strategy routine (described below)
1428 *
1429 * de - for internal use by the sysctl routines
1430 *
1431 * extra1, extra2 - extra pointers usable by the proc handler routines
1432 *
1433 * Leaf nodes in the sysctl tree will be represented by a single file
1434 * under /proc; non-leaf nodes will be represented by directories.
1435 *
1436 * sysctl(2) can automatically manage read and write requests through
1437 * the sysctl table. The data and maxlen fields of the ctl_table
1438 * struct enable minimal validation of the values being written to be
1439 * performed, and the mode field allows minimal authentication.
1440 *
1441 * More sophisticated management can be enabled by the provision of a
1442 * strategy routine with the table entry. This will be called before
1443 * any automatic read or write of the data is performed.
1444 *
1445 * The strategy routine may return
1446 *
1447 * < 0 - Error occurred (error is passed to user process)
1448 *
1449 * 0 - OK - proceed with automatic read or write.
1450 *
1451 * > 0 - OK - read or write has been done by the strategy routine, so
1452 * return immediately.
1453 *
1454 * There must be a proc_handler routine for any terminal nodes
1455 * mirrored under /proc/sys (non-terminals are handled by a built-in
1456 * directory handler). Several default handlers are available to
1457 * cover common cases -
1458 *
1459 * proc_dostring(), proc_dointvec(), proc_dointvec_jiffies(),
1460 * proc_dointvec_userhz_jiffies(), proc_dointvec_minmax(),
1461 * proc_doulongvec_ms_jiffies_minmax(), proc_doulongvec_minmax()
1462 *
1463 * It is the handler's job to read the input buffer from user memory
1464 * and process it. The handler should return 0 on success.
1465 *
1466 * This routine returns %NULL on a failure to register, and a pointer
1467 * to the table header on success.
1468 */
1469struct ctl_table_header *register_sysctl_table(ctl_table * table,
1470 int insert_at_head)
1471{
1472 struct ctl_table_header *tmp;
1473 tmp = kmalloc(sizeof(struct ctl_table_header), GFP_KERNEL);
1474 if (!tmp)
1475 return NULL;
1476 tmp->ctl_table = table;
1477 INIT_LIST_HEAD(&tmp->ctl_entry);
1478 tmp->used = 0;
1479 tmp->unregistering = NULL;
1480 spin_lock(&sysctl_lock);
1481 if (insert_at_head)
1482 list_add(&tmp->ctl_entry, &root_table_header.ctl_entry);
1483 else
1484 list_add_tail(&tmp->ctl_entry, &root_table_header.ctl_entry);
1485 spin_unlock(&sysctl_lock);
1486#ifdef CONFIG_PROC_SYSCTL
1487 register_proc_table(table, proc_sys_root, tmp);
1488#endif
1489 return tmp;
1490}
1491
1492/**
1493 * unregister_sysctl_table - unregister a sysctl table hierarchy
1494 * @header: the header returned from register_sysctl_table
1495 *
1496 * Unregisters the sysctl table and all children. proc entries may not
1497 * actually be removed until they are no longer used by anyone.
1498 */
1499void unregister_sysctl_table(struct ctl_table_header * header)
1500{
1501 might_sleep();
1502 spin_lock(&sysctl_lock);
1503 start_unregistering(header);
1504#ifdef CONFIG_PROC_SYSCTL
1505 unregister_proc_table(header->ctl_table, proc_sys_root);
1506#endif
1507 spin_unlock(&sysctl_lock);
1508 kfree(header);
1509}
1510
1511#else /* !CONFIG_SYSCTL */
1512struct ctl_table_header * register_sysctl_table(ctl_table * table,
1513 int insert_at_head)
1514{
1515 return NULL;
1516}
1517
1518void unregister_sysctl_table(struct ctl_table_header * table)
1519{
1520}
1521
1522#endif /* CONFIG_SYSCTL */
1523
1524/*
1525 * /proc/sys support
1526 */
1527
1528#ifdef CONFIG_PROC_SYSCTL
1529
1530/* Scan the sysctl entries in table and add them all into /proc */
1531static void register_proc_table(ctl_table * table, struct proc_dir_entry *root, void *set)
1532{
1533 struct proc_dir_entry *de;
1534 int len;
1535 mode_t mode;
1536
1537 for (; table->ctl_name || table->procname; table++) {
1538 /* Can't do anything without a proc name. */
1539 if (!table->procname)
1540 continue;
1541 /* Maybe we can't do anything with it... */
1542 if (!table->proc_handler && !table->child) {
1543 printk(KERN_WARNING "SYSCTL: Can't register %s\n",
1544 table->procname);
1545 continue;
1546 }
1547
1548 len = strlen(table->procname);
1549 mode = table->mode;
1550
1551 de = NULL;
1552 if (table->proc_handler)
1553 mode |= S_IFREG;
1554 else {
1555 mode |= S_IFDIR;
1556 for (de = root->subdir; de; de = de->next) {
1557 if (proc_match(len, table->procname, de))
1558 break;
1559 }
1560 /* If the subdir exists already, de is non-NULL */
1561 }
1562
1563 if (!de) {
1564 de = create_proc_entry(table->procname, mode, root);
1565 if (!de)
1566 continue;
1567 de->set = set;
1568 de->data = (void *) table;
1569 if (table->proc_handler)
1570 de->proc_fops = &proc_sys_file_operations;
1571 }
1572 table->de = de;
1573 if (de->mode & S_IFDIR)
1574 register_proc_table(table->child, de, set);
1575 }
1576}
1577
1578/*
1579 * Unregister a /proc sysctl table and any subdirectories.
1580 */
1581static void unregister_proc_table(ctl_table * table, struct proc_dir_entry *root)
1582{
1583 struct proc_dir_entry *de;
1584 for (; table->ctl_name || table->procname; table++) {
1585 if (!(de = table->de))
1586 continue;
1587 if (de->mode & S_IFDIR) {
1588 if (!table->child) {
1589 printk (KERN_ALERT "Help - malformed sysctl tree on free\n");
1590 continue;
1591 }
1592 unregister_proc_table(table->child, de);
1593
1594 /* Don't unregister directories which still have entries.. */
1595 if (de->subdir)
1596 continue;
1597 }
1598
1599 /*
1600 * In any case, mark the entry as goner; we'll keep it
1601 * around if it's busy, but we'll know to do nothing with
1602 * its fields. We are under sysctl_lock here.
1603 */
1604 de->data = NULL;
1605
1606 /* Don't unregister proc entries that are still being used.. */
1607 if (atomic_read(&de->count))
1608 continue;
1609
1610 table->de = NULL;
1611 remove_proc_entry(table->procname, root);
1612 }
1613}
1614
1615static ssize_t do_rw_proc(int write, struct file * file, char __user * buf,
1616 size_t count, loff_t *ppos)
1617{
1618 int op;
1619 struct proc_dir_entry *de = PDE(file->f_dentry->d_inode);
1620 struct ctl_table *table;
1621 size_t res;
1622 ssize_t error = -ENOTDIR;
1623
1624 spin_lock(&sysctl_lock);
1625 if (de && de->data && use_table(de->set)) {
1626 /*
1627 * at that point we know that sysctl was not unregistered
1628 * and won't be until we finish
1629 */
1630 spin_unlock(&sysctl_lock);
1631 table = (struct ctl_table *) de->data;
1632 if (!table || !table->proc_handler)
1633 goto out;
1634 error = -EPERM;
1635 op = (write ? 002 : 004);
1636 if (ctl_perm(table, op))
1637 goto out;
1638
1639 /* careful: calling conventions are nasty here */
1640 res = count;
1641 error = (*table->proc_handler)(table, write, file,
1642 buf, &res, ppos);
1643 if (!error)
1644 error = res;
1645 out:
1646 spin_lock(&sysctl_lock);
1647 unuse_table(de->set);
1648 }
1649 spin_unlock(&sysctl_lock);
1650 return error;
1651}
1652
1653static int proc_opensys(struct inode *inode, struct file *file)
1654{
1655 if (file->f_mode & FMODE_WRITE) {
1656 /*
1657 * sysctl entries that are not writable,
1658 * are _NOT_ writable, capabilities or not.
1659 */
1660 if (!(inode->i_mode & S_IWUSR))
1661 return -EPERM;
1662 }
1663
1664 return 0;
1665}
1666
1667static ssize_t proc_readsys(struct file * file, char __user * buf,
1668 size_t count, loff_t *ppos)
1669{
1670 return do_rw_proc(0, file, buf, count, ppos);
1671}
1672
1673static ssize_t proc_writesys(struct file * file, const char __user * buf,
1674 size_t count, loff_t *ppos)
1675{
1676 return do_rw_proc(1, file, (char __user *) buf, count, ppos);
1677}
1678
1679static int _proc_do_string(void* data, int maxlen, int write,
1680 struct file *filp, void __user *buffer,
1681 size_t *lenp, loff_t *ppos)
1682{
1683 size_t len;
1684 char __user *p;
1685 char c;
1686
1687 if (!data || !maxlen || !*lenp ||
1688 (*ppos && !write)) {
1689 *lenp = 0;
1690 return 0;
1691 }
1692
1693 if (write) {
1694 len = 0;
1695 p = buffer;
1696 while (len < *lenp) {
1697 if (get_user(c, p++))
1698 return -EFAULT;
1699 if (c == 0 || c == '\n')
1700 break;
1701 len++;
1702 }
1703 if (len >= maxlen)
1704 len = maxlen-1;
1705 if(copy_from_user(data, buffer, len))
1706 return -EFAULT;
1707 ((char *) data)[len] = 0;
1708 *ppos += *lenp;
1709 } else {
1710 len = strlen(data);
1711 if (len > maxlen)
1712 len = maxlen;
1713 if (len > *lenp)
1714 len = *lenp;
1715 if (len)
1716 if(copy_to_user(buffer, data, len))
1717 return -EFAULT;
1718 if (len < *lenp) {
1719 if(put_user('\n', ((char __user *) buffer) + len))
1720 return -EFAULT;
1721 len++;
1722 }
1723 *lenp = len;
1724 *ppos += len;
1725 }
1726 return 0;
1727}
1728
1729/**
1730 * proc_dostring - read a string sysctl
1731 * @table: the sysctl table
1732 * @write: %TRUE if this is a write to the sysctl file
1733 * @filp: the file structure
1734 * @buffer: the user buffer
1735 * @lenp: the size of the user buffer
1736 * @ppos: file position
1737 *
1738 * Reads/writes a string from/to the user buffer. If the kernel
1739 * buffer provided is not large enough to hold the string, the
1740 * string is truncated. The copied string is %NULL-terminated.
1741 * If the string is being read by the user process, it is copied
1742 * and a newline '\n' is added. It is truncated if the buffer is
1743 * not large enough.
1744 *
1745 * Returns 0 on success.
1746 */
1747int proc_dostring(ctl_table *table, int write, struct file *filp,
1748 void __user *buffer, size_t *lenp, loff_t *ppos)
1749{
1750 return _proc_do_string(table->data, table->maxlen, write, filp,
1751 buffer, lenp, ppos);
1752}
1753
1754/*
1755 * Special case of dostring for the UTS structure. This has locks
1756 * to observe. Should this be in kernel/sys.c ????
1757 */
1758
1759#ifndef CONFIG_UTS_NS
1760static int proc_do_uts_string(ctl_table *table, int write, struct file *filp,
1761 void __user *buffer, size_t *lenp, loff_t *ppos)
1762{
1763 int r;
1764
1765 if (!write) {
1766 down_read(&uts_sem);
1767 r=proc_dostring(table,0,filp,buffer,lenp, ppos);
1768 up_read(&uts_sem);
1769 } else {
1770 down_write(&uts_sem);
1771 r=proc_dostring(table,1,filp,buffer,lenp, ppos);
1772 up_write(&uts_sem);
1773 }
1774 return r;
1775}
1776#else /* !CONFIG_UTS_NS */
1777static int proc_do_uts_string(ctl_table *table, int write, struct file *filp,
1778 void __user *buffer, size_t *lenp, loff_t *ppos)
1779{
1780 int r;
1781 struct uts_namespace* uts_ns = current->nsproxy->uts_ns;
1782 char* which;
1783
1784 switch (table->ctl_name) {
1785 case KERN_OSTYPE:
1786 which = uts_ns->name.sysname;
1787 break;
1788 case KERN_NODENAME:
1789 which = uts_ns->name.nodename;
1790 break;
1791 case KERN_OSRELEASE:
1792 which = uts_ns->name.release;
1793 break;
1794 case KERN_VERSION:
1795 which = uts_ns->name.version;
1796 break;
1797 case KERN_DOMAINNAME:
1798 which = uts_ns->name.domainname;
1799 break;
1800 default:
1801 r = -EINVAL;
1802 goto out;
1803 }
1804
1805 if (!write) {
1806 down_read(&uts_sem);
1807 r=_proc_do_string(which,table->maxlen,0,filp,buffer,lenp, ppos);
1808 up_read(&uts_sem);
1809 } else {
1810 down_write(&uts_sem);
1811 r=_proc_do_string(which,table->maxlen,1,filp,buffer,lenp, ppos);
1812 up_write(&uts_sem);
1813 }
1814 out:
1815 return r;
1816}
1817#endif /* !CONFIG_UTS_NS */
1818
1819static int do_proc_dointvec_conv(int *negp, unsigned long *lvalp,
1820 int *valp,
1821 int write, void *data)
1822{
1823 if (write) {
1824 *valp = *negp ? -*lvalp : *lvalp;
1825 } else {
1826 int val = *valp;
1827 if (val < 0) {
1828 *negp = -1;
1829 *lvalp = (unsigned long)-val;
1830 } else {
1831 *negp = 0;
1832 *lvalp = (unsigned long)val;
1833 }
1834 }
1835 return 0;
1836}
1837
1838static int __do_proc_dointvec(void *tbl_data, ctl_table *table,
1839 int write, struct file *filp, void __user *buffer,
1840 size_t *lenp, loff_t *ppos,
1841 int (*conv)(int *negp, unsigned long *lvalp, int *valp,
1842 int write, void *data),
1843 void *data)
1844{
1845#define TMPBUFLEN 21
1846 int *i, vleft, first=1, neg, val;
1847 unsigned long lval;
1848 size_t left, len;
1849
1850 char buf[TMPBUFLEN], *p;
1851 char __user *s = buffer;
1852
1853 if (!tbl_data || !table->maxlen || !*lenp ||
1854 (*ppos && !write)) {
1855 *lenp = 0;
1856 return 0;
1857 }
1858
1859 i = (int *) tbl_data;
1860 vleft = table->maxlen / sizeof(*i);
1861 left = *lenp;
1862
1863 if (!conv)
1864 conv = do_proc_dointvec_conv;
1865
1866 for (; left && vleft--; i++, first=0) {
1867 if (write) {
1868 while (left) {
1869 char c;
1870 if (get_user(c, s))
1871 return -EFAULT;
1872 if (!isspace(c))
1873 break;
1874 left--;
1875 s++;
1876 }
1877 if (!left)
1878 break;
1879 neg = 0;
1880 len = left;
1881 if (len > sizeof(buf) - 1)
1882 len = sizeof(buf) - 1;
1883 if (copy_from_user(buf, s, len))
1884 return -EFAULT;
1885 buf[len] = 0;
1886 p = buf;
1887 if (*p == '-' && left > 1) {
1888 neg = 1;
1889 left--, p++;
1890 }
1891 if (*p < '0' || *p > '9')
1892 break;
1893
1894 lval = simple_strtoul(p, &p, 0);
1895
1896 len = p-buf;
1897 if ((len < left) && *p && !isspace(*p))
1898 break;
1899 if (neg)
1900 val = -val;
1901 s += len;
1902 left -= len;
1903
1904 if (conv(&neg, &lval, i, 1, data))
1905 break;
1906 } else {
1907 p = buf;
1908 if (!first)
1909 *p++ = '\t';
1910
1911 if (conv(&neg, &lval, i, 0, data))
1912 break;
1913
1914 sprintf(p, "%s%lu", neg ? "-" : "", lval);
1915 len = strlen(buf);
1916 if (len > left)
1917 len = left;
1918 if(copy_to_user(s, buf, len))
1919 return -EFAULT;
1920 left -= len;
1921 s += len;
1922 }
1923 }
1924
1925 if (!write && !first && left) {
1926 if(put_user('\n', s))
1927 return -EFAULT;
1928 left--, s++;
1929 }
1930 if (write) {
1931 while (left) {
1932 char c;
1933 if (get_user(c, s++))
1934 return -EFAULT;
1935 if (!isspace(c))
1936 break;
1937 left--;
1938 }
1939 }
1940 if (write && first)
1941 return -EINVAL;
1942 *lenp -= left;
1943 *ppos += *lenp;
1944 return 0;
1945#undef TMPBUFLEN
1946}
1947
1948static int do_proc_dointvec(ctl_table *table, int write, struct file *filp,
1949 void __user *buffer, size_t *lenp, loff_t *ppos,
1950 int (*conv)(int *negp, unsigned long *lvalp, int *valp,
1951 int write, void *data),
1952 void *data)
1953{
1954 return __do_proc_dointvec(table->data, table, write, filp,
1955 buffer, lenp, ppos, conv, data);
1956}
1957
1958/**
1959 * proc_dointvec - read a vector of integers
1960 * @table: the sysctl table
1961 * @write: %TRUE if this is a write to the sysctl file
1962 * @filp: the file structure
1963 * @buffer: the user buffer
1964 * @lenp: the size of the user buffer
1965 * @ppos: file position
1966 *
1967 * Reads/writes up to table->maxlen/sizeof(unsigned int) integer
1968 * values from/to the user buffer, treated as an ASCII string.
1969 *
1970 * Returns 0 on success.
1971 */
1972int proc_dointvec(ctl_table *table, int write, struct file *filp,
1973 void __user *buffer, size_t *lenp, loff_t *ppos)
1974{
1975 return do_proc_dointvec(table,write,filp,buffer,lenp,ppos,
1976 NULL,NULL);
1977}
1978
1979#define OP_SET 0
1980#define OP_AND 1
1981#define OP_OR 2
1982#define OP_MAX 3
1983#define OP_MIN 4
1984
1985static int do_proc_dointvec_bset_conv(int *negp, unsigned long *lvalp,
1986 int *valp,
1987 int write, void *data)
1988{
1989 int op = *(int *)data;
1990 if (write) {
1991 int val = *negp ? -*lvalp : *lvalp;
1992 switch(op) {
1993 case OP_SET: *valp = val; break;
1994 case OP_AND: *valp &= val; break;
1995 case OP_OR: *valp |= val; break;
1996 case OP_MAX: if(*valp < val)
1997 *valp = val;
1998 break;
1999 case OP_MIN: if(*valp > val)
2000 *valp = val;
2001 break;
2002 }
2003 } else {
2004 int val = *valp;
2005 if (val < 0) {
2006 *negp = -1;
2007 *lvalp = (unsigned long)-val;
2008 } else {
2009 *negp = 0;
2010 *lvalp = (unsigned long)val;
2011 }
2012 }
2013 return 0;
2014}
2015
2016/*
2017 * init may raise the set.
2018 */
2019
2020int proc_dointvec_bset(ctl_table *table, int write, struct file *filp,
2021 void __user *buffer, size_t *lenp, loff_t *ppos)
2022{
2023 int op;
2024
2025 if (!capable(CAP_SYS_MODULE)) {
2026 return -EPERM;
2027 }
2028
2029 op = is_init(current) ? OP_SET : OP_AND;
2030 return do_proc_dointvec(table,write,filp,buffer,lenp,ppos,
2031 do_proc_dointvec_bset_conv,&op);
2032}
2033
2034struct do_proc_dointvec_minmax_conv_param {
2035 int *min;
2036 int *max;
2037};
2038
2039static int do_proc_dointvec_minmax_conv(int *negp, unsigned long *lvalp,
2040 int *valp,
2041 int write, void *data)
2042{
2043 struct do_proc_dointvec_minmax_conv_param *param = data;
2044 if (write) {
2045 int val = *negp ? -*lvalp : *lvalp;
2046 if ((param->min && *param->min > val) ||
2047 (param->max && *param->max < val))
2048 return -EINVAL;
2049 *valp = val;
2050 } else {
2051 int val = *valp;
2052 if (val < 0) {
2053 *negp = -1;
2054 *lvalp = (unsigned long)-val;
2055 } else {
2056 *negp = 0;
2057 *lvalp = (unsigned long)val;
2058 }
2059 }
2060 return 0;
2061}
2062
2063/**
2064 * proc_dointvec_minmax - read a vector of integers with min/max values
2065 * @table: the sysctl table
2066 * @write: %TRUE if this is a write to the sysctl file
2067 * @filp: the file structure
2068 * @buffer: the user buffer
2069 * @lenp: the size of the user buffer
2070 * @ppos: file position
2071 *
2072 * Reads/writes up to table->maxlen/sizeof(unsigned int) integer
2073 * values from/to the user buffer, treated as an ASCII string.
2074 *
2075 * This routine will ensure the values are within the range specified by
2076 * table->extra1 (min) and table->extra2 (max).
2077 *
2078 * Returns 0 on success.
2079 */
2080int proc_dointvec_minmax(ctl_table *table, int write, struct file *filp,
2081 void __user *buffer, size_t *lenp, loff_t *ppos)
2082{
2083 struct do_proc_dointvec_minmax_conv_param param = {
2084 .min = (int *) table->extra1,
2085 .max = (int *) table->extra2,
2086 };
2087 return do_proc_dointvec(table, write, filp, buffer, lenp, ppos,
2088 do_proc_dointvec_minmax_conv, &param);
2089}
2090
2091static int __do_proc_doulongvec_minmax(void *data, ctl_table *table, int write,
2092 struct file *filp,
2093 void __user *buffer,
2094 size_t *lenp, loff_t *ppos,
2095 unsigned long convmul,
2096 unsigned long convdiv)
2097{
2098#define TMPBUFLEN 21
2099 unsigned long *i, *min, *max, val;
2100 int vleft, first=1, neg;
2101 size_t len, left;
2102 char buf[TMPBUFLEN], *p;
2103 char __user *s = buffer;
2104
2105 if (!data || !table->maxlen || !*lenp ||
2106 (*ppos && !write)) {
2107 *lenp = 0;
2108 return 0;
2109 }
2110
2111 i = (unsigned long *) data;
2112 min = (unsigned long *) table->extra1;
2113 max = (unsigned long *) table->extra2;
2114 vleft = table->maxlen / sizeof(unsigned long);
2115 left = *lenp;
2116
2117 for (; left && vleft--; i++, min++, max++, first=0) {
2118 if (write) {
2119 while (left) {
2120 char c;
2121 if (get_user(c, s))
2122 return -EFAULT;
2123 if (!isspace(c))
2124 break;
2125 left--;
2126 s++;
2127 }
2128 if (!left)
2129 break;
2130 neg = 0;
2131 len = left;
2132 if (len > TMPBUFLEN-1)
2133 len = TMPBUFLEN-1;
2134 if (copy_from_user(buf, s, len))
2135 return -EFAULT;
2136 buf[len] = 0;
2137 p = buf;
2138 if (*p == '-' && left > 1) {
2139 neg = 1;
2140 left--, p++;
2141 }
2142 if (*p < '0' || *p > '9')
2143 break;
2144 val = simple_strtoul(p, &p, 0) * convmul / convdiv ;
2145 len = p-buf;
2146 if ((len < left) && *p && !isspace(*p))
2147 break;
2148 if (neg)
2149 val = -val;
2150 s += len;
2151 left -= len;
2152
2153 if(neg)
2154 continue;
2155 if ((min && val < *min) || (max && val > *max))
2156 continue;
2157 *i = val;
2158 } else {
2159 p = buf;
2160 if (!first)
2161 *p++ = '\t';
2162 sprintf(p, "%lu", convdiv * (*i) / convmul);
2163 len = strlen(buf);
2164 if (len > left)
2165 len = left;
2166 if(copy_to_user(s, buf, len))
2167 return -EFAULT;
2168 left -= len;
2169 s += len;
2170 }
2171 }
2172
2173 if (!write && !first && left) {
2174 if(put_user('\n', s))
2175 return -EFAULT;
2176 left--, s++;
2177 }
2178 if (write) {
2179 while (left) {
2180 char c;
2181 if (get_user(c, s++))
2182 return -EFAULT;
2183 if (!isspace(c))
2184 break;
2185 left--;
2186 }
2187 }
2188 if (write && first)
2189 return -EINVAL;
2190 *lenp -= left;
2191 *ppos += *lenp;
2192 return 0;
2193#undef TMPBUFLEN
2194}
2195
2196static int do_proc_doulongvec_minmax(ctl_table *table, int write,
2197 struct file *filp,
2198 void __user *buffer,
2199 size_t *lenp, loff_t *ppos,
2200 unsigned long convmul,
2201 unsigned long convdiv)
2202{
2203 return __do_proc_doulongvec_minmax(table->data, table, write,
2204 filp, buffer, lenp, ppos, convmul, convdiv);
2205}
2206
2207/**
2208 * proc_doulongvec_minmax - read a vector of long integers with min/max values
2209 * @table: the sysctl table
2210 * @write: %TRUE if this is a write to the sysctl file
2211 * @filp: the file structure
2212 * @buffer: the user buffer
2213 * @lenp: the size of the user buffer
2214 * @ppos: file position
2215 *
2216 * Reads/writes up to table->maxlen/sizeof(unsigned long) unsigned long
2217 * values from/to the user buffer, treated as an ASCII string.
2218 *
2219 * This routine will ensure the values are within the range specified by
2220 * table->extra1 (min) and table->extra2 (max).
2221 *
2222 * Returns 0 on success.
2223 */
2224int proc_doulongvec_minmax(ctl_table *table, int write, struct file *filp,
2225 void __user *buffer, size_t *lenp, loff_t *ppos)
2226{
2227 return do_proc_doulongvec_minmax(table, write, filp, buffer, lenp, ppos, 1l, 1l);
2228}
2229
2230/**
2231 * proc_doulongvec_ms_jiffies_minmax - read a vector of millisecond values with min/max values
2232 * @table: the sysctl table
2233 * @write: %TRUE if this is a write to the sysctl file
2234 * @filp: the file structure
2235 * @buffer: the user buffer
2236 * @lenp: the size of the user buffer
2237 * @ppos: file position
2238 *
2239 * Reads/writes up to table->maxlen/sizeof(unsigned long) unsigned long
2240 * values from/to the user buffer, treated as an ASCII string. The values
2241 * are treated as milliseconds, and converted to jiffies when they are stored.
2242 *
2243 * This routine will ensure the values are within the range specified by
2244 * table->extra1 (min) and table->extra2 (max).
2245 *
2246 * Returns 0 on success.
2247 */
2248int proc_doulongvec_ms_jiffies_minmax(ctl_table *table, int write,
2249 struct file *filp,
2250 void __user *buffer,
2251 size_t *lenp, loff_t *ppos)
2252{
2253 return do_proc_doulongvec_minmax(table, write, filp, buffer,
2254 lenp, ppos, HZ, 1000l);
2255}
2256
2257
2258static int do_proc_dointvec_jiffies_conv(int *negp, unsigned long *lvalp,
2259 int *valp,
2260 int write, void *data)
2261{
2262 if (write) {
2263 if (*lvalp > LONG_MAX / HZ)
2264 return 1;
2265 *valp = *negp ? -(*lvalp*HZ) : (*lvalp*HZ);
2266 } else {
2267 int val = *valp;
2268 unsigned long lval;
2269 if (val < 0) {
2270 *negp = -1;
2271 lval = (unsigned long)-val;
2272 } else {
2273 *negp = 0;
2274 lval = (unsigned long)val;
2275 }
2276 *lvalp = lval / HZ;
2277 }
2278 return 0;
2279}
2280
2281static int do_proc_dointvec_userhz_jiffies_conv(int *negp, unsigned long *lvalp,
2282 int *valp,
2283 int write, void *data)
2284{
2285 if (write) {
2286 if (USER_HZ < HZ && *lvalp > (LONG_MAX / HZ) * USER_HZ)
2287 return 1;
2288 *valp = clock_t_to_jiffies(*negp ? -*lvalp : *lvalp);
2289 } else {
2290 int val = *valp;
2291 unsigned long lval;
2292 if (val < 0) {
2293 *negp = -1;
2294 lval = (unsigned long)-val;
2295 } else {
2296 *negp = 0;
2297 lval = (unsigned long)val;
2298 }
2299 *lvalp = jiffies_to_clock_t(lval);
2300 }
2301 return 0;
2302}
2303
2304static int do_proc_dointvec_ms_jiffies_conv(int *negp, unsigned long *lvalp,
2305 int *valp,
2306 int write, void *data)
2307{
2308 if (write) {
2309 *valp = msecs_to_jiffies(*negp ? -*lvalp : *lvalp);
2310 } else {
2311 int val = *valp;
2312 unsigned long lval;
2313 if (val < 0) {
2314 *negp = -1;
2315 lval = (unsigned long)-val;
2316 } else {
2317 *negp = 0;
2318 lval = (unsigned long)val;
2319 }
2320 *lvalp = jiffies_to_msecs(lval);
2321 }
2322 return 0;
2323}
2324
2325/**
2326 * proc_dointvec_jiffies - read a vector of integers as seconds
2327 * @table: the sysctl table
2328 * @write: %TRUE if this is a write to the sysctl file
2329 * @filp: the file structure
2330 * @buffer: the user buffer
2331 * @lenp: the size of the user buffer
2332 * @ppos: file position
2333 *
2334 * Reads/writes up to table->maxlen/sizeof(unsigned int) integer
2335 * values from/to the user buffer, treated as an ASCII string.
2336 * The values read are assumed to be in seconds, and are converted into
2337 * jiffies.
2338 *
2339 * Returns 0 on success.
2340 */
2341int proc_dointvec_jiffies(ctl_table *table, int write, struct file *filp,
2342 void __user *buffer, size_t *lenp, loff_t *ppos)
2343{
2344 return do_proc_dointvec(table,write,filp,buffer,lenp,ppos,
2345 do_proc_dointvec_jiffies_conv,NULL);
2346}
2347
2348/**
2349 * proc_dointvec_userhz_jiffies - read a vector of integers as 1/USER_HZ seconds
2350 * @table: the sysctl table
2351 * @write: %TRUE if this is a write to the sysctl file
2352 * @filp: the file structure
2353 * @buffer: the user buffer
2354 * @lenp: the size of the user buffer
2355 * @ppos: pointer to the file position
2356 *
2357 * Reads/writes up to table->maxlen/sizeof(unsigned int) integer
2358 * values from/to the user buffer, treated as an ASCII string.
2359 * The values read are assumed to be in 1/USER_HZ seconds, and
2360 * are converted into jiffies.
2361 *
2362 * Returns 0 on success.
2363 */
2364int proc_dointvec_userhz_jiffies(ctl_table *table, int write, struct file *filp,
2365 void __user *buffer, size_t *lenp, loff_t *ppos)
2366{
2367 return do_proc_dointvec(table,write,filp,buffer,lenp,ppos,
2368 do_proc_dointvec_userhz_jiffies_conv,NULL);
2369}
2370
2371/**
2372 * proc_dointvec_ms_jiffies - read a vector of integers as 1 milliseconds
2373 * @table: the sysctl table
2374 * @write: %TRUE if this is a write to the sysctl file
2375 * @filp: the file structure
2376 * @buffer: the user buffer
2377 * @lenp: the size of the user buffer
2378 * @ppos: file position
2379 * @ppos: the current position in the file
2380 *
2381 * Reads/writes up to table->maxlen/sizeof(unsigned int) integer
2382 * values from/to the user buffer, treated as an ASCII string.
2383 * The values read are assumed to be in 1/1000 seconds, and
2384 * are converted into jiffies.
2385 *
2386 * Returns 0 on success.
2387 */
2388int proc_dointvec_ms_jiffies(ctl_table *table, int write, struct file *filp,
2389 void __user *buffer, size_t *lenp, loff_t *ppos)
2390{
2391 return do_proc_dointvec(table, write, filp, buffer, lenp, ppos,
2392 do_proc_dointvec_ms_jiffies_conv, NULL);
2393}
2394
2395#ifdef CONFIG_SYSVIPC
2396static int proc_do_ipc_string(ctl_table *table, int write, struct file *filp,
2397 void __user *buffer, size_t *lenp, loff_t *ppos)
2398{
2399 void *data;
2400 struct ipc_namespace *ns;
2401
2402 ns = current->nsproxy->ipc_ns;
2403
2404 switch (table->ctl_name) {
2405 case KERN_SHMMAX:
2406 data = &ns->shm_ctlmax;
2407 goto proc_minmax;
2408 case KERN_SHMALL:
2409 data = &ns->shm_ctlall;
2410 goto proc_minmax;
2411 case KERN_SHMMNI:
2412 data = &ns->shm_ctlmni;
2413 break;
2414 case KERN_MSGMAX:
2415 data = &ns->msg_ctlmax;
2416 break;
2417 case KERN_MSGMNI:
2418 data = &ns->msg_ctlmni;
2419 break;
2420 case KERN_MSGMNB:
2421 data = &ns->msg_ctlmnb;
2422 break;
2423 case KERN_SEM:
2424 data = &ns->sem_ctls;
2425 break;
2426 default:
2427 return -EINVAL;
2428 }
2429
2430 return __do_proc_dointvec(data, table, write, filp, buffer,
2431 lenp, ppos, NULL, NULL);
2432proc_minmax:
2433 return __do_proc_doulongvec_minmax(data, table, write, filp, buffer,
2434 lenp, ppos, 1l, 1l);
2435}
2436#endif
2437
2438static int proc_do_cad_pid(ctl_table *table, int write, struct file *filp,
2439 void __user *buffer, size_t *lenp, loff_t *ppos)
2440{
2441 struct pid *new_pid;
2442 pid_t tmp;
2443 int r;
2444
2445 tmp = pid_nr(cad_pid);
2446
2447 r = __do_proc_dointvec(&tmp, table, write, filp, buffer,
2448 lenp, ppos, NULL, NULL);
2449 if (r || !write)
2450 return r;
2451
2452 new_pid = find_get_pid(tmp);
2453 if (!new_pid)
2454 return -ESRCH;
2455
2456 put_pid(xchg(&cad_pid, new_pid));
2457 return 0;
2458}
2459
2460#else /* CONFIG_PROC_FS */
2461
2462int proc_dostring(ctl_table *table, int write, struct file *filp,
2463 void __user *buffer, size_t *lenp, loff_t *ppos)
2464{
2465 return -ENOSYS;
2466}
2467
2468static int proc_do_uts_string(ctl_table *table, int write, struct file *filp,
2469 void __user *buffer, size_t *lenp, loff_t *ppos)
2470{
2471 return -ENOSYS;
2472}
2473
2474#ifdef CONFIG_SYSVIPC
2475static int proc_do_ipc_string(ctl_table *table, int write, struct file *filp,
2476 void __user *buffer, size_t *lenp, loff_t *ppos)
2477{
2478 return -ENOSYS;
2479}
2480#endif
2481
2482int proc_dointvec(ctl_table *table, int write, struct file *filp,
2483 void __user *buffer, size_t *lenp, loff_t *ppos)
2484{
2485 return -ENOSYS;
2486}
2487
2488int proc_dointvec_bset(ctl_table *table, int write, struct file *filp,
2489 void __user *buffer, size_t *lenp, loff_t *ppos)
2490{
2491 return -ENOSYS;
2492}
2493
2494int proc_dointvec_minmax(ctl_table *table, int write, struct file *filp,
2495 void __user *buffer, size_t *lenp, loff_t *ppos)
2496{
2497 return -ENOSYS;
2498}
2499
2500int proc_dointvec_jiffies(ctl_table *table, int write, struct file *filp,
2501 void __user *buffer, size_t *lenp, loff_t *ppos)
2502{
2503 return -ENOSYS;
2504}
2505
2506int proc_dointvec_userhz_jiffies(ctl_table *table, int write, struct file *filp,
2507 void __user *buffer, size_t *lenp, loff_t *ppos)
2508{
2509 return -ENOSYS;
2510}
2511
2512int proc_dointvec_ms_jiffies(ctl_table *table, int write, struct file *filp,
2513 void __user *buffer, size_t *lenp, loff_t *ppos)
2514{
2515 return -ENOSYS;
2516}
2517
2518int proc_doulongvec_minmax(ctl_table *table, int write, struct file *filp,
2519 void __user *buffer, size_t *lenp, loff_t *ppos)
2520{
2521 return -ENOSYS;
2522}
2523
2524int proc_doulongvec_ms_jiffies_minmax(ctl_table *table, int write,
2525 struct file *filp,
2526 void __user *buffer,
2527 size_t *lenp, loff_t *ppos)
2528{
2529 return -ENOSYS;
2530}
2531
2532
2533#endif /* CONFIG_PROC_FS */
2534
2535
2536#ifdef CONFIG_SYSCTL_SYSCALL
2537/*
2538 * General sysctl support routines
2539 */
2540
2541/* The generic string strategy routine: */
2542int sysctl_string(ctl_table *table, int __user *name, int nlen,
2543 void __user *oldval, size_t __user *oldlenp,
2544 void __user *newval, size_t newlen, void **context)
2545{
2546 if (!table->data || !table->maxlen)
2547 return -ENOTDIR;
2548
2549 if (oldval && oldlenp) {
2550 size_t bufsize;
2551 if (get_user(bufsize, oldlenp))
2552 return -EFAULT;
2553 if (bufsize) {
2554 size_t len = strlen(table->data), copied;
2555
2556 /* This shouldn't trigger for a well-formed sysctl */
2557 if (len > table->maxlen)
2558 len = table->maxlen;
2559
2560 /* Copy up to a max of bufsize-1 bytes of the string */
2561 copied = (len >= bufsize) ? bufsize - 1 : len;
2562
2563 if (copy_to_user(oldval, table->data, copied) ||
2564 put_user(0, (char __user *)(oldval + copied)))
2565 return -EFAULT;
2566 if (put_user(len, oldlenp))
2567 return -EFAULT;
2568 }
2569 }
2570 if (newval && newlen) {
2571 size_t len = newlen;
2572 if (len > table->maxlen)
2573 len = table->maxlen;
2574 if(copy_from_user(table->data, newval, len))
2575 return -EFAULT;
2576 if (len == table->maxlen)
2577 len--;
2578 ((char *) table->data)[len] = 0;
2579 }
2580 return 1;
2581}
2582
2583/*
2584 * This function makes sure that all of the integers in the vector
2585 * are between the minimum and maximum values given in the arrays
2586 * table->extra1 and table->extra2, respectively.
2587 */
2588int sysctl_intvec(ctl_table *table, int __user *name, int nlen,
2589 void __user *oldval, size_t __user *oldlenp,
2590 void __user *newval, size_t newlen, void **context)
2591{
2592
2593 if (newval && newlen) {
2594 int __user *vec = (int __user *) newval;
2595 int *min = (int *) table->extra1;
2596 int *max = (int *) table->extra2;
2597 size_t length;
2598 int i;
2599
2600 if (newlen % sizeof(int) != 0)
2601 return -EINVAL;
2602
2603 if (!table->extra1 && !table->extra2)
2604 return 0;
2605
2606 if (newlen > table->maxlen)
2607 newlen = table->maxlen;
2608 length = newlen / sizeof(int);
2609
2610 for (i = 0; i < length; i++) {
2611 int value;
2612 if (get_user(value, vec + i))
2613 return -EFAULT;
2614 if (min && value < min[i])
2615 return -EINVAL;
2616 if (max && value > max[i])
2617 return -EINVAL;
2618 }
2619 }
2620 return 0;
2621}
2622
2623/* Strategy function to convert jiffies to seconds */
2624int sysctl_jiffies(ctl_table *table, int __user *name, int nlen,
2625 void __user *oldval, size_t __user *oldlenp,
2626 void __user *newval, size_t newlen, void **context)
2627{
2628 if (oldval) {
2629 size_t olen;
2630 if (oldlenp) {
2631 if (get_user(olen, oldlenp))
2632 return -EFAULT;
2633 if (olen!=sizeof(int))
2634 return -EINVAL;
2635 }
2636 if (put_user(*(int *)(table->data)/HZ, (int __user *)oldval) ||
2637 (oldlenp && put_user(sizeof(int),oldlenp)))
2638 return -EFAULT;
2639 }
2640 if (newval && newlen) {
2641 int new;
2642 if (newlen != sizeof(int))
2643 return -EINVAL;
2644 if (get_user(new, (int __user *)newval))
2645 return -EFAULT;
2646 *(int *)(table->data) = new*HZ;
2647 }
2648 return 1;
2649}
2650
2651/* Strategy function to convert jiffies to seconds */
2652int sysctl_ms_jiffies(ctl_table *table, int __user *name, int nlen,
2653 void __user *oldval, size_t __user *oldlenp,
2654 void __user *newval, size_t newlen, void **context)
2655{
2656 if (oldval) {
2657 size_t olen;
2658 if (oldlenp) {
2659 if (get_user(olen, oldlenp))
2660 return -EFAULT;
2661 if (olen!=sizeof(int))
2662 return -EINVAL;
2663 }
2664 if (put_user(jiffies_to_msecs(*(int *)(table->data)), (int __user *)oldval) ||
2665 (oldlenp && put_user(sizeof(int),oldlenp)))
2666 return -EFAULT;
2667 }
2668 if (newval && newlen) {
2669 int new;
2670 if (newlen != sizeof(int))
2671 return -EINVAL;
2672 if (get_user(new, (int __user *)newval))
2673 return -EFAULT;
2674 *(int *)(table->data) = msecs_to_jiffies(new);
2675 }
2676 return 1;
2677}
2678
2679#else /* CONFIG_SYSCTL_SYSCALL */
2680
2681
2682asmlinkage long sys_sysctl(struct __sysctl_args __user *args)
2683{
2684 static int msg_count;
2685 struct __sysctl_args tmp;
2686 int name[CTL_MAXNAME];
2687 int i;
2688
2689 /* Read in the sysctl name for better debug message logging */
2690 if (copy_from_user(&tmp, args, sizeof(tmp)))
2691 return -EFAULT;
2692 if (tmp.nlen <= 0 || tmp.nlen >= CTL_MAXNAME)
2693 return -ENOTDIR;
2694 for (i = 0; i < tmp.nlen; i++)
2695 if (get_user(name[i], tmp.name + i))
2696 return -EFAULT;
2697
2698 /* Ignore accesses to kernel.version */
2699 if ((tmp.nlen == 2) && (name[0] == CTL_KERN) && (name[1] == KERN_VERSION))
2700 goto out;
2701
2702 if (msg_count < 5) {
2703 msg_count++;
2704 printk(KERN_INFO
2705 "warning: process `%s' used the removed sysctl "
2706 "system call with ", current->comm);
2707 for (i = 0; i < tmp.nlen; i++)
2708 printk("%d.", name[i]);
2709 printk("\n");
2710 }
2711out:
2712 return -ENOSYS;
2713}
2714
2715int sysctl_string(ctl_table *table, int __user *name, int nlen,
2716 void __user *oldval, size_t __user *oldlenp,
2717 void __user *newval, size_t newlen, void **context)
2718{
2719 return -ENOSYS;
2720}
2721
2722int sysctl_intvec(ctl_table *table, int __user *name, int nlen,
2723 void __user *oldval, size_t __user *oldlenp,
2724 void __user *newval, size_t newlen, void **context)
2725{
2726 return -ENOSYS;
2727}
2728
2729int sysctl_jiffies(ctl_table *table, int __user *name, int nlen,
2730 void __user *oldval, size_t __user *oldlenp,
2731 void __user *newval, size_t newlen, void **context)
2732{
2733 return -ENOSYS;
2734}
2735
2736int sysctl_ms_jiffies(ctl_table *table, int __user *name, int nlen,
2737 void __user *oldval, size_t __user *oldlenp,
2738 void __user *newval, size_t newlen, void **context)
2739{
2740 return -ENOSYS;
2741}
2742
2743#endif /* CONFIG_SYSCTL_SYSCALL */
2744
2745/*
2746 * No sense putting this after each symbol definition, twice,
2747 * exception granted :-)
2748 */
2749EXPORT_SYMBOL(proc_dointvec);
2750EXPORT_SYMBOL(proc_dointvec_jiffies);
2751EXPORT_SYMBOL(proc_dointvec_minmax);
2752EXPORT_SYMBOL(proc_dointvec_userhz_jiffies);
2753EXPORT_SYMBOL(proc_dointvec_ms_jiffies);
2754EXPORT_SYMBOL(proc_dostring);
2755EXPORT_SYMBOL(proc_doulongvec_minmax);
2756EXPORT_SYMBOL(proc_doulongvec_ms_jiffies_minmax);
2757EXPORT_SYMBOL(register_sysctl_table);
2758EXPORT_SYMBOL(sysctl_intvec);
2759EXPORT_SYMBOL(sysctl_jiffies);
2760EXPORT_SYMBOL(sysctl_ms_jiffies);
2761EXPORT_SYMBOL(sysctl_string);
2762EXPORT_SYMBOL(unregister_sysctl_table);