?  search:
summary | log | graphiclog1 | graphiclog2 | commit | commitdiff | tree | refs | edit | fork
blame | history | raw | HEAD
[PATCH] ext3 balloc: say rb_entry not list_entry
[linux-2.6.22.y-op.git] / kernel / sysctl.c
blob758dbbf972a544d0889d7230464983bb52ce6c41
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 pid_max;
72 extern int min_free_kbytes;
73 extern int printk_ratelimit_jiffies;
74 extern int printk_ratelimit_burst;
75 extern int pid_max_min, pid_max_max;
76 extern int sysctl_drop_caches;
77 extern int percpu_pagelist_fraction;
78 extern int compat_log;
79
80 /* this is needed for the proc_dointvec_minmax for [fs_]overflow UID and GID */
81 static int maxolduid = 65535;
82 static int minolduid;
83 static int min_percpu_pagelist_fract = 8;
84
85 static int ngroups_max = NGROUPS_MAX;
86
87 #ifdef CONFIG_KMOD
88 extern char modprobe_path[];
89 #endif
90 #ifdef CONFIG_CHR_DEV_SG
91 extern int sg_big_buff;
92 #endif
93 #ifdef CONFIG_SYSVIPC
94 static 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__
99 extern char reboot_command [];
100 extern int stop_a_enabled;
101 extern int scons_pwroff;
102 #endif
103
104 #ifdef __hppa__
105 extern int pwrsw_enabled;
106 extern int unaligned_enabled;
107 #endif
108
109 #ifdef CONFIG_S390
110 #ifdef CONFIG_MATHEMU
111 extern int sysctl_ieee_emulation_warnings;
112 #endif
113 extern int sysctl_userprocess_debug;
114 extern int spin_retry;
115 #endif
116
117 extern int sysctl_hz_timer;
118
119 #ifdef CONFIG_BSD_PROCESS_ACCT
120 extern int acct_parm[];
121 #endif
122
123 #ifdef CONFIG_IA64
124 extern int no_unaligned_warning;
125 #endif
126
127 #ifdef CONFIG_RT_MUTEXES
128 extern int max_lock_depth;
129 #endif
130
131 #ifdef CONFIG_SYSCTL_SYSCALL
132 static int parse_table(int __user *, int, void __user *, size_t __user *,
133 void __user *, size_t, ctl_table *, void **);
134 #endif
135
136 static 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
140 static 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
144 static ctl_table root_table[];
145 static struct ctl_table_header root_table_header =
146 { root_table, LIST_HEAD_INIT(root_table_header.ctl_entry) };
147
148 static ctl_table kern_table[];
149 static ctl_table vm_table[];
150 static ctl_table fs_table[];
151 static ctl_table debug_table[];
152 static ctl_table dev_table[];
153 extern ctl_table random_table[];
154 #ifdef CONFIG_UNIX98_PTYS
155 extern ctl_table pty_table[];
156 #endif
157 #ifdef CONFIG_INOTIFY_USER
158 extern ctl_table inotify_table[];
159 #endif
160
161 #ifdef HAVE_ARCH_PICK_MMAP_LAYOUT
162 int sysctl_legacy_va_layout;
163 #endif
164
165 /* /proc declarations: */
166
167 #ifdef CONFIG_PROC_SYSCTL
168
169 static ssize_t proc_readsys(struct file *, char __user *, size_t, loff_t *);
170 static ssize_t proc_writesys(struct file *, const char __user *, size_t, loff_t *);
171 static int proc_opensys(struct inode *, struct file *);
172
173 const struct file_operations proc_sys_file_operations = {
174 .open = proc_opensys,
175 .read = proc_readsys,
176 .write = proc_writesys,
177 };
178
179 extern struct proc_dir_entry *proc_sys_root;
180
181 static void register_proc_table(ctl_table *, struct proc_dir_entry *, void *);
182 static void unregister_proc_table(ctl_table *, struct proc_dir_entry *);
183 #endif
184
185 /* The default sysctl tables: */
186
187 static 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
230 static 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. */
777 static int zero;
778 static int one_hundred = 100;
779
780
781 static 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_NUMA
981 {
982 .ctl_name = VM_ZONE_RECLAIM_MODE,
983 .procname = "zone_reclaim_mode",
984 .data = &zone_reclaim_mode,
985 .maxlen = sizeof(zone_reclaim_mode),
986 .mode = 0644,
987 .proc_handler = &proc_dointvec,
988 .strategy = &sysctl_intvec,
989 .extra1 = &zero,
990 },
991 {
992 .ctl_name = VM_MIN_UNMAPPED,
993 .procname = "min_unmapped_ratio",
994 .data = &sysctl_min_unmapped_ratio,
995 .maxlen = sizeof(sysctl_min_unmapped_ratio),
996 .mode = 0644,
997 .proc_handler = &sysctl_min_unmapped_ratio_sysctl_handler,
998 .strategy = &sysctl_intvec,
999 .extra1 = &zero,
1000 .extra2 = &one_hundred,
1001 },
1002 {
1003 .ctl_name = VM_MIN_SLAB,
1004 .procname = "min_slab_ratio",
1005 .data = &sysctl_min_slab_ratio,
1006 .maxlen = sizeof(sysctl_min_slab_ratio),
1007 .mode = 0644,
1008 .proc_handler = &sysctl_min_slab_ratio_sysctl_handler,
1009 .strategy = &sysctl_intvec,
1010 .extra1 = &zero,
1011 .extra2 = &one_hundred,
1012 },
1013 #endif
1014 #ifdef CONFIG_X86_32
1015 {
1016 .ctl_name = VM_VDSO_ENABLED,
1017 .procname = "vdso_enabled",
1018 .data = &vdso_enabled,
1019 .maxlen = sizeof(vdso_enabled),
1020 .mode = 0644,
1021 .proc_handler = &proc_dointvec,
1022 .strategy = &sysctl_intvec,
1023 .extra1 = &zero,
1024 },
1025 #endif
1026 { .ctl_name = 0 }
1027 };
1028
1029 static ctl_table fs_table[] = {
1030 {
1031 .ctl_name = FS_NRINODE,
1032 .procname = "inode-nr",
1033 .data = &inodes_stat,
1034 .maxlen = 2*sizeof(int),
1035 .mode = 0444,
1036 .proc_handler = &proc_dointvec,
1037 },
1038 {
1039 .ctl_name = FS_STATINODE,
1040 .procname = "inode-state",
1041 .data = &inodes_stat,
1042 .maxlen = 7*sizeof(int),
1043 .mode = 0444,
1044 .proc_handler = &proc_dointvec,
1045 },
1046 {
1047 .ctl_name = FS_NRFILE,
1048 .procname = "file-nr",
1049 .data = &files_stat,
1050 .maxlen = 3*sizeof(int),
1051 .mode = 0444,
1052 .proc_handler = &proc_nr_files,
1053 },
1054 {
1055 .ctl_name = FS_MAXFILE,
1056 .procname = "file-max",
1057 .data = &files_stat.max_files,
1058 .maxlen = sizeof(int),
1059 .mode = 0644,
1060 .proc_handler = &proc_dointvec,
1061 },
1062 {
1063 .ctl_name = FS_DENTRY,
1064 .procname = "dentry-state",
1065 .data = &dentry_stat,
1066 .maxlen = 6*sizeof(int),
1067 .mode = 0444,
1068 .proc_handler = &proc_dointvec,
1069 },
1070 {
1071 .ctl_name = FS_OVERFLOWUID,
1072 .procname = "overflowuid",
1073 .data = &fs_overflowuid,
1074 .maxlen = sizeof(int),
1075 .mode = 0644,
1076 .proc_handler = &proc_dointvec_minmax,
1077 .strategy = &sysctl_intvec,
1078 .extra1 = &minolduid,
1079 .extra2 = &maxolduid,
1080 },
1081 {
1082 .ctl_name = FS_OVERFLOWGID,
1083 .procname = "overflowgid",
1084 .data = &fs_overflowgid,
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_LEASES,
1094 .procname = "leases-enable",
1095 .data = &leases_enable,
1096 .maxlen = sizeof(int),
1097 .mode = 0644,
1098 .proc_handler = &proc_dointvec,
1099 },
1100 #ifdef CONFIG_DNOTIFY
1101 {
1102 .ctl_name = FS_DIR_NOTIFY,
1103 .procname = "dir-notify-enable",
1104 .data = &dir_notify_enable,
1105 .maxlen = sizeof(int),
1106 .mode = 0644,
1107 .proc_handler = &proc_dointvec,
1108 },
1109 #endif
1110 #ifdef CONFIG_MMU
1111 {
1112 .ctl_name = FS_LEASE_TIME,
1113 .procname = "lease-break-time",
1114 .data = &lease_break_time,
1115 .maxlen = sizeof(int),
1116 .mode = 0644,
1117 .proc_handler = &proc_dointvec,
1118 },
1119 {
1120 .ctl_name = FS_AIO_NR,
1121 .procname = "aio-nr",
1122 .data = &aio_nr,
1123 .maxlen = sizeof(aio_nr),
1124 .mode = 0444,
1125 .proc_handler = &proc_doulongvec_minmax,
1126 },
1127 {
1128 .ctl_name = FS_AIO_MAX_NR,
1129 .procname = "aio-max-nr",
1130 .data = &aio_max_nr,
1131 .maxlen = sizeof(aio_max_nr),
1132 .mode = 0644,
1133 .proc_handler = &proc_doulongvec_minmax,
1134 },
1135 #ifdef CONFIG_INOTIFY_USER
1136 {
1137 .ctl_name = FS_INOTIFY,
1138 .procname = "inotify",
1139 .mode = 0555,
1140 .child = inotify_table,
1141 },
1142 #endif
1143 #endif
1144 {
1145 .ctl_name = KERN_SETUID_DUMPABLE,
1146 .procname = "suid_dumpable",
1147 .data = &suid_dumpable,
1148 .maxlen = sizeof(int),
1149 .mode = 0644,
1150 .proc_handler = &proc_dointvec,
1151 },
1152 { .ctl_name = 0 }
1153 };
1154
1155 static ctl_table debug_table[] = {
1156 { .ctl_name = 0 }
1157 };
1158
1159 static ctl_table dev_table[] = {
1160 { .ctl_name = 0 }
1161 };
1162
1163 extern void init_irq_proc (void);
1164
1165 static DEFINE_SPINLOCK(sysctl_lock);
1166
1167 /* called under sysctl_lock */
1168 static int use_table(struct ctl_table_header *p)
1169 {
1170 if (unlikely(p->unregistering))
1171 return 0;
1172 p->used++;
1173 return 1;
1174 }
1175
1176 /* called under sysctl_lock */
1177 static void unuse_table(struct ctl_table_header *p)
1178 {
1179 if (!--p->used)
1180 if (unlikely(p->unregistering))
1181 complete(p->unregistering);
1182 }
1183
1184 /* called under sysctl_lock, will reacquire if has to wait */
1185 static void start_unregistering(struct ctl_table_header *p)
1186 {
1187 /*
1188 * if p->used is 0, nobody will ever touch that entry again;
1189 * we'll eliminate all paths to it before dropping sysctl_lock
1190 */
1191 if (unlikely(p->used)) {
1192 struct completion wait;
1193 init_completion(&wait);
1194 p->unregistering = &wait;
1195 spin_unlock(&sysctl_lock);
1196 wait_for_completion(&wait);
1197 spin_lock(&sysctl_lock);
1198 }
1199 /*
1200 * do not remove from the list until nobody holds it; walking the
1201 * list in do_sysctl() relies on that.
1202 */
1203 list_del_init(&p->ctl_entry);
1204 }
1205
1206 void __init sysctl_init(void)
1207 {
1208 #ifdef CONFIG_PROC_SYSCTL
1209 register_proc_table(root_table, proc_sys_root, &root_table_header);
1210 init_irq_proc();
1211 #endif
1212 }
1213
1214 #ifdef CONFIG_SYSCTL_SYSCALL
1215 int do_sysctl(int __user *name, int nlen, void __user *oldval, size_t __user *oldlenp,
1216 void __user *newval, size_t newlen)
1217 {
1218 struct list_head *tmp;
1219 int error = -ENOTDIR;
1220
1221 if (nlen <= 0 || nlen >= CTL_MAXNAME)
1222 return -ENOTDIR;
1223 if (oldval) {
1224 int old_len;
1225 if (!oldlenp || get_user(old_len, oldlenp))
1226 return -EFAULT;
1227 }
1228 spin_lock(&sysctl_lock);
1229 tmp = &root_table_header.ctl_entry;
1230 do {
1231 struct ctl_table_header *head =
1232 list_entry(tmp, struct ctl_table_header, ctl_entry);
1233 void *context = NULL;
1234
1235 if (!use_table(head))
1236 continue;
1237
1238 spin_unlock(&sysctl_lock);
1239
1240 error = parse_table(name, nlen, oldval, oldlenp,
1241 newval, newlen, head->ctl_table,
1242 &context);
1243 kfree(context);
1244
1245 spin_lock(&sysctl_lock);
1246 unuse_table(head);
1247 if (error != -ENOTDIR)
1248 break;
1249 } while ((tmp = tmp->next) != &root_table_header.ctl_entry);
1250 spin_unlock(&sysctl_lock);
1251 return error;
1252 }
1253
1254 asmlinkage long sys_sysctl(struct __sysctl_args __user *args)
1255 {
1256 struct __sysctl_args tmp;
1257 int error;
1258
1259 if (copy_from_user(&tmp, args, sizeof(tmp)))
1260 return -EFAULT;
1261
1262 lock_kernel();
1263 error = do_sysctl(tmp.name, tmp.nlen, tmp.oldval, tmp.oldlenp,
1264 tmp.newval, tmp.newlen);
1265 unlock_kernel();
1266 return error;
1267 }
1268 #endif /* CONFIG_SYSCTL_SYSCALL */
1269
1270 /*
1271 * ctl_perm does NOT grant the superuser all rights automatically, because
1272 * some sysctl variables are readonly even to root.
1273 */
1274
1275 static int test_perm(int mode, int op)
1276 {
1277 if (!current->euid)
1278 mode >>= 6;
1279 else if (in_egroup_p(0))
1280 mode >>= 3;
1281 if ((mode & op & 0007) == op)
1282 return 0;
1283 return -EACCES;
1284 }
1285
1286 static inline int ctl_perm(ctl_table *table, int op)
1287 {
1288 int error;
1289 error = security_sysctl(table, op);
1290 if (error)
1291 return error;
1292 return test_perm(table->mode, op);
1293 }
1294
1295 #ifdef CONFIG_SYSCTL_SYSCALL
1296 static int parse_table(int __user *name, int nlen,
1297 void __user *oldval, size_t __user *oldlenp,
1298 void __user *newval, size_t newlen,
1299 ctl_table *table, void **context)
1300 {
1301 int n;
1302 repeat:
1303 if (!nlen)
1304 return -ENOTDIR;
1305 if (get_user(n, name))
1306 return -EFAULT;
1307 for ( ; table->ctl_name || table->procname; table++) {
1308 if (!table->ctl_name)
1309 continue;
1310 if (n == table->ctl_name || table->ctl_name == CTL_ANY) {
1311 int error;
1312 if (table->child) {
1313 if (ctl_perm(table, 001))
1314 return -EPERM;
1315 if (table->strategy) {
1316 error = table->strategy(
1317 table, name, nlen,
1318 oldval, oldlenp,
1319 newval, newlen, context);
1320 if (error)
1321 return error;
1322 }
1323 name++;
1324 nlen--;
1325 table = table->child;
1326 goto repeat;
1327 }
1328 error = do_sysctl_strategy(table, name, nlen,
1329 oldval, oldlenp,
1330 newval, newlen, context);
1331 return error;
1332 }
1333 }
1334 return -ENOTDIR;
1335 }
1336
1337 /* Perform the actual read/write of a sysctl table entry. */
1338 int do_sysctl_strategy (ctl_table *table,
1339 int __user *name, int nlen,
1340 void __user *oldval, size_t __user *oldlenp,
1341 void __user *newval, size_t newlen, void **context)
1342 {
1343 int op = 0, rc;
1344 size_t len;
1345
1346 if (oldval)
1347 op |= 004;
1348 if (newval)
1349 op |= 002;
1350 if (ctl_perm(table, op))
1351 return -EPERM;
1352
1353 if (table->strategy) {
1354 rc = table->strategy(table, name, nlen, oldval, oldlenp,
1355 newval, newlen, context);
1356 if (rc < 0)
1357 return rc;
1358 if (rc > 0)
1359 return 0;
1360 }
1361
1362 /* If there is no strategy routine, or if the strategy returns
1363 * zero, proceed with automatic r/w */
1364 if (table->data && table->maxlen) {
1365 if (oldval && oldlenp) {
1366 if (get_user(len, oldlenp))
1367 return -EFAULT;
1368 if (len) {
1369 if (len > table->maxlen)
1370 len = table->maxlen;
1371 if(copy_to_user(oldval, table->data, len))
1372 return -EFAULT;
1373 if(put_user(len, oldlenp))
1374 return -EFAULT;
1375 }
1376 }
1377 if (newval && newlen) {
1378 len = newlen;
1379 if (len > table->maxlen)
1380 len = table->maxlen;
1381 if(copy_from_user(table->data, newval, len))
1382 return -EFAULT;
1383 }
1384 }
1385 return 0;
1386 }
1387 #endif /* CONFIG_SYSCTL_SYSCALL */
1388
1389 /**
1390 * register_sysctl_table - register a sysctl hierarchy
1391 * @table: the top-level table structure
1392 * @insert_at_head: whether the entry should be inserted in front or at the end
1393 *
1394 * Register a sysctl table hierarchy. @table should be a filled in ctl_table
1395 * array. An entry with a ctl_name of 0 terminates the table.
1396 *
1397 * The members of the &ctl_table structure are used as follows:
1398 *
1399 * ctl_name - This is the numeric sysctl value used by sysctl(2). The number
1400 * must be unique within that level of sysctl
1401 *
1402 * procname - the name of the sysctl file under /proc/sys. Set to %NULL to not
1403 * enter a sysctl file
1404 *
1405 * data - a pointer to data for use by proc_handler
1406 *
1407 * maxlen - the maximum size in bytes of the data
1408 *
1409 * mode - the file permissions for the /proc/sys file, and for sysctl(2)
1410 *
1411 * child - a pointer to the child sysctl table if this entry is a directory, or
1412 * %NULL.
1413 *
1414 * proc_handler - the text handler routine (described below)
1415 *
1416 * strategy - the strategy routine (described below)
1417 *
1418 * de - for internal use by the sysctl routines
1419 *
1420 * extra1, extra2 - extra pointers usable by the proc handler routines
1421 *
1422 * Leaf nodes in the sysctl tree will be represented by a single file
1423 * under /proc; non-leaf nodes will be represented by directories.
1424 *
1425 * sysctl(2) can automatically manage read and write requests through
1426 * the sysctl table. The data and maxlen fields of the ctl_table
1427 * struct enable minimal validation of the values being written to be
1428 * performed, and the mode field allows minimal authentication.
1429 *
1430 * More sophisticated management can be enabled by the provision of a
1431 * strategy routine with the table entry. This will be called before
1432 * any automatic read or write of the data is performed.
1433 *
1434 * The strategy routine may return
1435 *
1436 * < 0 - Error occurred (error is passed to user process)
1437 *
1438 * 0 - OK - proceed with automatic read or write.
1439 *
1440 * > 0 - OK - read or write has been done by the strategy routine, so
1441 * return immediately.
1442 *
1443 * There must be a proc_handler routine for any terminal nodes
1444 * mirrored under /proc/sys (non-terminals are handled by a built-in
1445 * directory handler). Several default handlers are available to
1446 * cover common cases -
1447 *
1448 * proc_dostring(), proc_dointvec(), proc_dointvec_jiffies(),
1449 * proc_dointvec_userhz_jiffies(), proc_dointvec_minmax(),
1450 * proc_doulongvec_ms_jiffies_minmax(), proc_doulongvec_minmax()
1451 *
1452 * It is the handler's job to read the input buffer from user memory
1453 * and process it. The handler should return 0 on success.
1454 *
1455 * This routine returns %NULL on a failure to register, and a pointer
1456 * to the table header on success.
1457 */
1458 struct ctl_table_header *register_sysctl_table(ctl_table * table,
1459 int insert_at_head)
1460 {
1461 struct ctl_table_header *tmp;
1462 tmp = kmalloc(sizeof(struct ctl_table_header), GFP_KERNEL);
1463 if (!tmp)
1464 return NULL;
1465 tmp->ctl_table = table;
1466 INIT_LIST_HEAD(&tmp->ctl_entry);
1467 tmp->used = 0;
1468 tmp->unregistering = NULL;
1469 spin_lock(&sysctl_lock);
1470 if (insert_at_head)
1471 list_add(&tmp->ctl_entry, &root_table_header.ctl_entry);
1472 else
1473 list_add_tail(&tmp->ctl_entry, &root_table_header.ctl_entry);
1474 spin_unlock(&sysctl_lock);
1475 #ifdef CONFIG_PROC_SYSCTL
1476 register_proc_table(table, proc_sys_root, tmp);
1477 #endif
1478 return tmp;
1479 }
1480
1481 /**
1482 * unregister_sysctl_table - unregister a sysctl table hierarchy
1483 * @header: the header returned from register_sysctl_table
1484 *
1485 * Unregisters the sysctl table and all children. proc entries may not
1486 * actually be removed until they are no longer used by anyone.
1487 */
1488 void unregister_sysctl_table(struct ctl_table_header * header)
1489 {
1490 might_sleep();
1491 spin_lock(&sysctl_lock);
1492 start_unregistering(header);
1493 #ifdef CONFIG_PROC_SYSCTL
1494 unregister_proc_table(header->ctl_table, proc_sys_root);
1495 #endif
1496 spin_unlock(&sysctl_lock);
1497 kfree(header);
1498 }
1499
1500 #else /* !CONFIG_SYSCTL */
1501 struct ctl_table_header * register_sysctl_table(ctl_table * table,
1502 int insert_at_head)
1503 {
1504 return NULL;
1505 }
1506
1507 void unregister_sysctl_table(struct ctl_table_header * table)
1508 {
1509 }
1510
1511 #endif /* CONFIG_SYSCTL */
1512
1513 /*
1514 * /proc/sys support
1515 */
1516
1517 #ifdef CONFIG_PROC_SYSCTL
1518
1519 /* Scan the sysctl entries in table and add them all into /proc */
1520 static void register_proc_table(ctl_table * table, struct proc_dir_entry *root, void *set)
1521 {
1522 struct proc_dir_entry *de;
1523 int len;
1524 mode_t mode;
1525
1526 for (; table->ctl_name || table->procname; table++) {
1527 /* Can't do anything without a proc name. */
1528 if (!table->procname)
1529 continue;
1530 /* Maybe we can't do anything with it... */
1531 if (!table->proc_handler && !table->child) {
1532 printk(KERN_WARNING "SYSCTL: Can't register %s\n",
1533 table->procname);
1534 continue;
1535 }
1536
1537 len = strlen(table->procname);
1538 mode = table->mode;
1539
1540 de = NULL;
1541 if (table->proc_handler)
1542 mode |= S_IFREG;
1543 else {
1544 mode |= S_IFDIR;
1545 for (de = root->subdir; de; de = de->next) {
1546 if (proc_match(len, table->procname, de))
1547 break;
1548 }
1549 /* If the subdir exists already, de is non-NULL */
1550 }
1551
1552 if (!de) {
1553 de = create_proc_entry(table->procname, mode, root);
1554 if (!de)
1555 continue;
1556 de->set = set;
1557 de->data = (void *) table;
1558 if (table->proc_handler)
1559 de->proc_fops = &proc_sys_file_operations;
1560 }
1561 table->de = de;
1562 if (de->mode & S_IFDIR)
1563 register_proc_table(table->child, de, set);
1564 }
1565 }
1566
1567 /*
1568 * Unregister a /proc sysctl table and any subdirectories.
1569 */
1570 static void unregister_proc_table(ctl_table * table, struct proc_dir_entry *root)
1571 {
1572 struct proc_dir_entry *de;
1573 for (; table->ctl_name || table->procname; table++) {
1574 if (!(de = table->de))
1575 continue;
1576 if (de->mode & S_IFDIR) {
1577 if (!table->child) {
1578 printk (KERN_ALERT "Help - malformed sysctl tree on free\n");
1579 continue;
1580 }
1581 unregister_proc_table(table->child, de);
1582
1583 /* Don't unregister directories which still have entries.. */
1584 if (de->subdir)
1585 continue;
1586 }
1587
1588 /*
1589 * In any case, mark the entry as goner; we'll keep it
1590 * around if it's busy, but we'll know to do nothing with
1591 * its fields. We are under sysctl_lock here.
1592 */
1593 de->data = NULL;
1594
1595 /* Don't unregister proc entries that are still being used.. */
1596 if (atomic_read(&de->count))
1597 continue;
1598
1599 table->de = NULL;
1600 remove_proc_entry(table->procname, root);
1601 }
1602 }
1603
1604 static ssize_t do_rw_proc(int write, struct file * file, char __user * buf,
1605 size_t count, loff_t *ppos)
1606 {
1607 int op;
1608 struct proc_dir_entry *de = PDE(file->f_dentry->d_inode);
1609 struct ctl_table *table;
1610 size_t res;
1611 ssize_t error = -ENOTDIR;
1612
1613 spin_lock(&sysctl_lock);
1614 if (de && de->data && use_table(de->set)) {
1615 /*
1616 * at that point we know that sysctl was not unregistered
1617 * and won't be until we finish
1618 */
1619 spin_unlock(&sysctl_lock);
1620 table = (struct ctl_table *) de->data;
1621 if (!table || !table->proc_handler)
1622 goto out;
1623 error = -EPERM;
1624 op = (write ? 002 : 004);
1625 if (ctl_perm(table, op))
1626 goto out;
1627
1628 /* careful: calling conventions are nasty here */
1629 res = count;
1630 error = (*table->proc_handler)(table, write, file,
1631 buf, &res, ppos);
1632 if (!error)
1633 error = res;
1634 out:
1635 spin_lock(&sysctl_lock);
1636 unuse_table(de->set);
1637 }
1638 spin_unlock(&sysctl_lock);
1639 return error;
1640 }
1641
1642 static int proc_opensys(struct inode *inode, struct file *file)
1643 {
1644 if (file->f_mode & FMODE_WRITE) {
1645 /*
1646 * sysctl entries that are not writable,
1647 * are _NOT_ writable, capabilities or not.
1648 */
1649 if (!(inode->i_mode & S_IWUSR))
1650 return -EPERM;
1651 }
1652
1653 return 0;
1654 }
1655
1656 static ssize_t proc_readsys(struct file * file, char __user * buf,
1657 size_t count, loff_t *ppos)
1658 {
1659 return do_rw_proc(0, file, buf, count, ppos);
1660 }
1661
1662 static ssize_t proc_writesys(struct file * file, const char __user * buf,
1663 size_t count, loff_t *ppos)
1664 {
1665 return do_rw_proc(1, file, (char __user *) buf, count, ppos);
1666 }
1667
1668 static int _proc_do_string(void* data, int maxlen, int write,
1669 struct file *filp, void __user *buffer,
1670 size_t *lenp, loff_t *ppos)
1671 {
1672 size_t len;
1673 char __user *p;
1674 char c;
1675
1676 if (!data || !maxlen || !*lenp ||
1677 (*ppos && !write)) {
1678 *lenp = 0;
1679 return 0;
1680 }
1681
1682 if (write) {
1683 len = 0;
1684 p = buffer;
1685 while (len < *lenp) {
1686 if (get_user(c, p++))
1687 return -EFAULT;
1688 if (c == 0 || c == '\n')
1689 break;
1690 len++;
1691 }
1692 if (len >= maxlen)
1693 len = maxlen-1;
1694 if(copy_from_user(data, buffer, len))
1695 return -EFAULT;
1696 ((char *) data)[len] = 0;
1697 *ppos += *lenp;
1698 } else {
1699 len = strlen(data);
1700 if (len > maxlen)
1701 len = maxlen;
1702 if (len > *lenp)
1703 len = *lenp;
1704 if (len)
1705 if(copy_to_user(buffer, data, len))
1706 return -EFAULT;
1707 if (len < *lenp) {
1708 if(put_user('\n', ((char __user *) buffer) + len))
1709 return -EFAULT;
1710 len++;
1711 }
1712 *lenp = len;
1713 *ppos += len;
1714 }
1715 return 0;
1716 }
1717
1718 /**
1719 * proc_dostring - read a string sysctl
1720 * @table: the sysctl table
1721 * @write: %TRUE if this is a write to the sysctl file
1722 * @filp: the file structure
1723 * @buffer: the user buffer
1724 * @lenp: the size of the user buffer
1725 * @ppos: file position
1726 *
1727 * Reads/writes a string from/to the user buffer. If the kernel
1728 * buffer provided is not large enough to hold the string, the
1729 * string is truncated. The copied string is %NULL-terminated.
1730 * If the string is being read by the user process, it is copied
1731 * and a newline '\n' is added. It is truncated if the buffer is
1732 * not large enough.
1733 *
1734 * Returns 0 on success.
1735 */
1736 int proc_dostring(ctl_table *table, int write, struct file *filp,
1737 void __user *buffer, size_t *lenp, loff_t *ppos)
1738 {
1739 return _proc_do_string(table->data, table->maxlen, write, filp,
1740 buffer, lenp, ppos);
1741 }
1742
1743 /*
1744 * Special case of dostring for the UTS structure. This has locks
1745 * to observe. Should this be in kernel/sys.c ????
1746 */
1747
1748 #ifndef CONFIG_UTS_NS
1749 static int proc_do_uts_string(ctl_table *table, int write, struct file *filp,
1750 void __user *buffer, size_t *lenp, loff_t *ppos)
1751 {
1752 int r;
1753
1754 if (!write) {
1755 down_read(&uts_sem);
1756 r=proc_dostring(table,0,filp,buffer,lenp, ppos);
1757 up_read(&uts_sem);
1758 } else {
1759 down_write(&uts_sem);
1760 r=proc_dostring(table,1,filp,buffer,lenp, ppos);
1761 up_write(&uts_sem);
1762 }
1763 return r;
1764 }
1765 #else /* !CONFIG_UTS_NS */
1766 static int proc_do_uts_string(ctl_table *table, int write, struct file *filp,
1767 void __user *buffer, size_t *lenp, loff_t *ppos)
1768 {
1769 int r;
1770 struct uts_namespace* uts_ns = current->nsproxy->uts_ns;
1771 char* which;
1772
1773 switch (table->ctl_name) {
1774 case KERN_OSTYPE:
1775 which = uts_ns->name.sysname;
1776 break;
1777 case KERN_NODENAME:
1778 which = uts_ns->name.nodename;
1779 break;
1780 case KERN_OSRELEASE:
1781 which = uts_ns->name.release;
1782 break;
1783 case KERN_VERSION:
1784 which = uts_ns->name.version;
1785 break;
1786 case KERN_DOMAINNAME:
1787 which = uts_ns->name.domainname;
1788 break;
1789 default:
1790 r = -EINVAL;
1791 goto out;
1792 }
1793
1794 if (!write) {
1795 down_read(&uts_sem);
1796 r=_proc_do_string(which,table->maxlen,0,filp,buffer,lenp, ppos);
1797 up_read(&uts_sem);
1798 } else {
1799 down_write(&uts_sem);
1800 r=_proc_do_string(which,table->maxlen,1,filp,buffer,lenp, ppos);
1801 up_write(&uts_sem);
1802 }
1803 out:
1804 return r;
1805 }
1806 #endif /* !CONFIG_UTS_NS */
1807
1808 static int do_proc_dointvec_conv(int *negp, unsigned long *lvalp,
1809 int *valp,
1810 int write, void *data)
1811 {
1812 if (write) {
1813 *valp = *negp ? -*lvalp : *lvalp;
1814 } else {
1815 int val = *valp;
1816 if (val < 0) {
1817 *negp = -1;
1818 *lvalp = (unsigned long)-val;
1819 } else {
1820 *negp = 0;
1821 *lvalp = (unsigned long)val;
1822 }
1823 }
1824 return 0;
1825 }
1826
1827 static int __do_proc_dointvec(void *tbl_data, ctl_table *table,
1828 int write, struct file *filp, void __user *buffer,
1829 size_t *lenp, loff_t *ppos,
1830 int (*conv)(int *negp, unsigned long *lvalp, int *valp,
1831 int write, void *data),
1832 void *data)
1833 {
1834 #define TMPBUFLEN 21
1835 int *i, vleft, first=1, neg, val;
1836 unsigned long lval;
1837 size_t left, len;
1838
1839 char buf[TMPBUFLEN], *p;
1840 char __user *s = buffer;
1841
1842 if (!tbl_data || !table->maxlen || !*lenp ||
1843 (*ppos && !write)) {
1844 *lenp = 0;
1845 return 0;
1846 }
1847
1848 i = (int *) tbl_data;
1849 vleft = table->maxlen / sizeof(*i);
1850 left = *lenp;
1851
1852 if (!conv)
1853 conv = do_proc_dointvec_conv;
1854
1855 for (; left && vleft--; i++, first=0) {
1856 if (write) {
1857 while (left) {
1858 char c;
1859 if (get_user(c, s))
1860 return -EFAULT;
1861 if (!isspace(c))
1862 break;
1863 left--;
1864 s++;
1865 }
1866 if (!left)
1867 break;
1868 neg = 0;
1869 len = left;
1870 if (len > sizeof(buf) - 1)
1871 len = sizeof(buf) - 1;
1872 if (copy_from_user(buf, s, len))
1873 return -EFAULT;
1874 buf[len] = 0;
1875 p = buf;
1876 if (*p == '-' && left > 1) {
1877 neg = 1;
1878 p++;
1879 }
1880 if (*p < '0' || *p > '9')
1881 break;
1882
1883 lval = simple_strtoul(p, &p, 0);
1884
1885 len = p-buf;
1886 if ((len < left) && *p && !isspace(*p))
1887 break;
1888 if (neg)
1889 val = -val;
1890 s += len;
1891 left -= len;
1892
1893 if (conv(&neg, &lval, i, 1, data))
1894 break;
1895 } else {
1896 p = buf;
1897 if (!first)
1898 *p++ = '\t';
1899
1900 if (conv(&neg, &lval, i, 0, data))
1901 break;
1902
1903 sprintf(p, "%s%lu", neg ? "-" : "", lval);
1904 len = strlen(buf);
1905 if (len > left)
1906 len = left;
1907 if(copy_to_user(s, buf, len))
1908 return -EFAULT;
1909 left -= len;
1910 s += len;
1911 }
1912 }
1913
1914 if (!write && !first && left) {
1915 if(put_user('\n', s))
1916 return -EFAULT;
1917 left--, s++;
1918 }
1919 if (write) {
1920 while (left) {
1921 char c;
1922 if (get_user(c, s++))
1923 return -EFAULT;
1924 if (!isspace(c))
1925 break;
1926 left--;
1927 }
1928 }
1929 if (write && first)
1930 return -EINVAL;
1931 *lenp -= left;
1932 *ppos += *lenp;
1933 return 0;
1934 #undef TMPBUFLEN
1935 }
1936
1937 static int do_proc_dointvec(ctl_table *table, int write, struct file *filp,
1938 void __user *buffer, size_t *lenp, loff_t *ppos,
1939 int (*conv)(int *negp, unsigned long *lvalp, int *valp,
1940 int write, void *data),
1941 void *data)
1942 {
1943 return __do_proc_dointvec(table->data, table, write, filp,
1944 buffer, lenp, ppos, conv, data);
1945 }
1946
1947 /**
1948 * proc_dointvec - read a vector of integers
1949 * @table: the sysctl table
1950 * @write: %TRUE if this is a write to the sysctl file
1951 * @filp: the file structure
1952 * @buffer: the user buffer
1953 * @lenp: the size of the user buffer
1954 * @ppos: file position
1955 *
1956 * Reads/writes up to table->maxlen/sizeof(unsigned int) integer
1957 * values from/to the user buffer, treated as an ASCII string.
1958 *
1959 * Returns 0 on success.
1960 */
1961 int proc_dointvec(ctl_table *table, int write, struct file *filp,
1962 void __user *buffer, size_t *lenp, loff_t *ppos)
1963 {
1964 return do_proc_dointvec(table,write,filp,buffer,lenp,ppos,
1965 NULL,NULL);
1966 }
1967
1968 #define OP_SET 0
1969 #define OP_AND 1
1970 #define OP_OR 2
1971 #define OP_MAX 3
1972 #define OP_MIN 4
1973
1974 static int do_proc_dointvec_bset_conv(int *negp, unsigned long *lvalp,
1975 int *valp,
1976 int write, void *data)
1977 {
1978 int op = *(int *)data;
1979 if (write) {
1980 int val = *negp ? -*lvalp : *lvalp;
1981 switch(op) {
1982 case OP_SET: *valp = val; break;
1983 case OP_AND: *valp &= val; break;
1984 case OP_OR: *valp |= val; break;
1985 case OP_MAX: if(*valp < val)
1986 *valp = val;
1987 break;
1988 case OP_MIN: if(*valp > val)
1989 *valp = val;
1990 break;
1991 }
1992 } else {
1993 int val = *valp;
1994 if (val < 0) {
1995 *negp = -1;
1996 *lvalp = (unsigned long)-val;
1997 } else {
1998 *negp = 0;
1999 *lvalp = (unsigned long)val;
2000 }
2001 }
2002 return 0;
2003 }
2004
2005 /*
2006 * init may raise the set.
2007 */
2008
2009 int proc_dointvec_bset(ctl_table *table, int write, struct file *filp,
2010 void __user *buffer, size_t *lenp, loff_t *ppos)
2011 {
2012 int op;
2013
2014 if (!capable(CAP_SYS_MODULE)) {
2015 return -EPERM;
2016 }
2017
2018 op = is_init(current) ? OP_SET : OP_AND;
2019 return do_proc_dointvec(table,write,filp,buffer,lenp,ppos,
2020 do_proc_dointvec_bset_conv,&op);
2021 }
2022
2023 struct do_proc_dointvec_minmax_conv_param {
2024 int *min;
2025 int *max;
2026 };
2027
2028 static int do_proc_dointvec_minmax_conv(int *negp, unsigned long *lvalp,
2029 int *valp,
2030 int write, void *data)
2031 {
2032 struct do_proc_dointvec_minmax_conv_param *param = data;
2033 if (write) {
2034 int val = *negp ? -*lvalp : *lvalp;
2035 if ((param->min && *param->min > val) ||
2036 (param->max && *param->max < val))
2037 return -EINVAL;
2038 *valp = val;
2039 } else {
2040 int val = *valp;
2041 if (val < 0) {
2042 *negp = -1;
2043 *lvalp = (unsigned long)-val;
2044 } else {
2045 *negp = 0;
2046 *lvalp = (unsigned long)val;
2047 }
2048 }
2049 return 0;
2050 }
2051
2052 /**
2053 * proc_dointvec_minmax - read a vector of integers with min/max values
2054 * @table: the sysctl table
2055 * @write: %TRUE if this is a write to the sysctl file
2056 * @filp: the file structure
2057 * @buffer: the user buffer
2058 * @lenp: the size of the user buffer
2059 * @ppos: file position
2060 *
2061 * Reads/writes up to table->maxlen/sizeof(unsigned int) integer
2062 * values from/to the user buffer, treated as an ASCII string.
2063 *
2064 * This routine will ensure the values are within the range specified by
2065 * table->extra1 (min) and table->extra2 (max).
2066 *
2067 * Returns 0 on success.
2068 */
2069 int proc_dointvec_minmax(ctl_table *table, int write, struct file *filp,
2070 void __user *buffer, size_t *lenp, loff_t *ppos)
2071 {
2072 struct do_proc_dointvec_minmax_conv_param param = {
2073 .min = (int *) table->extra1,
2074 .max = (int *) table->extra2,
2075 };
2076 return do_proc_dointvec(table, write, filp, buffer, lenp, ppos,
2077 do_proc_dointvec_minmax_conv, &param);
2078 }
2079
2080 static int __do_proc_doulongvec_minmax(void *data, ctl_table *table, int write,
2081 struct file *filp,
2082 void __user *buffer,
2083 size_t *lenp, loff_t *ppos,
2084 unsigned long convmul,
2085 unsigned long convdiv)
2086 {
2087 #define TMPBUFLEN 21
2088 unsigned long *i, *min, *max, val;
2089 int vleft, first=1, neg;
2090 size_t len, left;
2091 char buf[TMPBUFLEN], *p;
2092 char __user *s = buffer;
2093
2094 if (!data || !table->maxlen || !*lenp ||
2095 (*ppos && !write)) {
2096 *lenp = 0;
2097 return 0;
2098 }
2099
2100 i = (unsigned long *) data;
2101 min = (unsigned long *) table->extra1;
2102 max = (unsigned long *) table->extra2;
2103 vleft = table->maxlen / sizeof(unsigned long);
2104 left = *lenp;
2105
2106 for (; left && vleft--; i++, min++, max++, first=0) {
2107 if (write) {
2108 while (left) {
2109 char c;
2110 if (get_user(c, s))
2111 return -EFAULT;
2112 if (!isspace(c))
2113 break;
2114 left--;
2115 s++;
2116 }
2117 if (!left)
2118 break;
2119 neg = 0;
2120 len = left;
2121 if (len > TMPBUFLEN-1)
2122 len = TMPBUFLEN-1;
2123 if (copy_from_user(buf, s, len))
2124 return -EFAULT;
2125 buf[len] = 0;
2126 p = buf;
2127 if (*p == '-' && left > 1) {
2128 neg = 1;
2129 p++;
2130 }
2131 if (*p < '0' || *p > '9')
2132 break;
2133 val = simple_strtoul(p, &p, 0) * convmul / convdiv ;
2134 len = p-buf;
2135 if ((len < left) && *p && !isspace(*p))
2136 break;
2137 if (neg)
2138 val = -val;
2139 s += len;
2140 left -= len;
2141
2142 if(neg)
2143 continue;
2144 if ((min && val < *min) || (max && val > *max))
2145 continue;
2146 *i = val;
2147 } else {
2148 p = buf;
2149 if (!first)
2150 *p++ = '\t';
2151 sprintf(p, "%lu", convdiv * (*i) / convmul);
2152 len = strlen(buf);
2153 if (len > left)
2154 len = left;
2155 if(copy_to_user(s, buf, len))
2156 return -EFAULT;
2157 left -= len;
2158 s += len;
2159 }
2160 }
2161
2162 if (!write && !first && left) {
2163 if(put_user('\n', s))
2164 return -EFAULT;
2165 left--, s++;
2166 }
2167 if (write) {
2168 while (left) {
2169 char c;
2170 if (get_user(c, s++))
2171 return -EFAULT;
2172 if (!isspace(c))
2173 break;
2174 left--;
2175 }
2176 }
2177 if (write && first)
2178 return -EINVAL;
2179 *lenp -= left;
2180 *ppos += *lenp;
2181 return 0;
2182 #undef TMPBUFLEN
2183 }
2184
2185 static int do_proc_doulongvec_minmax(ctl_table *table, int write,
2186 struct file *filp,
2187 void __user *buffer,
2188 size_t *lenp, loff_t *ppos,
2189 unsigned long convmul,
2190 unsigned long convdiv)
2191 {
2192 return __do_proc_doulongvec_minmax(table->data, table, write,
2193 filp, buffer, lenp, ppos, convmul, convdiv);
2194 }
2195
2196 /**
2197 * proc_doulongvec_minmax - read a vector of long integers with min/max values
2198 * @table: the sysctl table
2199 * @write: %TRUE if this is a write to the sysctl file
2200 * @filp: the file structure
2201 * @buffer: the user buffer
2202 * @lenp: the size of the user buffer
2203 * @ppos: file position
2204 *
2205 * Reads/writes up to table->maxlen/sizeof(unsigned long) unsigned long
2206 * values from/to the user buffer, treated as an ASCII string.
2207 *
2208 * This routine will ensure the values are within the range specified by
2209 * table->extra1 (min) and table->extra2 (max).
2210 *
2211 * Returns 0 on success.
2212 */
2213 int proc_doulongvec_minmax(ctl_table *table, int write, struct file *filp,
2214 void __user *buffer, size_t *lenp, loff_t *ppos)
2215 {
2216 return do_proc_doulongvec_minmax(table, write, filp, buffer, lenp, ppos, 1l, 1l);
2217 }
2218
2219 /**
2220 * proc_doulongvec_ms_jiffies_minmax - read a vector of millisecond values with min/max values
2221 * @table: the sysctl table
2222 * @write: %TRUE if this is a write to the sysctl file
2223 * @filp: the file structure
2224 * @buffer: the user buffer
2225 * @lenp: the size of the user buffer
2226 * @ppos: file position
2227 *
2228 * Reads/writes up to table->maxlen/sizeof(unsigned long) unsigned long
2229 * values from/to the user buffer, treated as an ASCII string. The values
2230 * are treated as milliseconds, and converted to jiffies when they are stored.
2231 *
2232 * This routine will ensure the values are within the range specified by
2233 * table->extra1 (min) and table->extra2 (max).
2234 *
2235 * Returns 0 on success.
2236 */
2237 int proc_doulongvec_ms_jiffies_minmax(ctl_table *table, int write,
2238 struct file *filp,
2239 void __user *buffer,
2240 size_t *lenp, loff_t *ppos)
2241 {
2242 return do_proc_doulongvec_minmax(table, write, filp, buffer,
2243 lenp, ppos, HZ, 1000l);
2244 }
2245
2246
2247 static int do_proc_dointvec_jiffies_conv(int *negp, unsigned long *lvalp,
2248 int *valp,
2249 int write, void *data)
2250 {
2251 if (write) {
2252 if (*lvalp > LONG_MAX / HZ)
2253 return 1;
2254 *valp = *negp ? -(*lvalp*HZ) : (*lvalp*HZ);
2255 } else {
2256 int val = *valp;
2257 unsigned long lval;
2258 if (val < 0) {
2259 *negp = -1;
2260 lval = (unsigned long)-val;
2261 } else {
2262 *negp = 0;
2263 lval = (unsigned long)val;
2264 }
2265 *lvalp = lval / HZ;
2266 }
2267 return 0;
2268 }
2269
2270 static int do_proc_dointvec_userhz_jiffies_conv(int *negp, unsigned long *lvalp,
2271 int *valp,
2272 int write, void *data)
2273 {
2274 if (write) {
2275 if (USER_HZ < HZ && *lvalp > (LONG_MAX / HZ) * USER_HZ)
2276 return 1;
2277 *valp = clock_t_to_jiffies(*negp ? -*lvalp : *lvalp);
2278 } else {
2279 int val = *valp;
2280 unsigned long lval;
2281 if (val < 0) {
2282 *negp = -1;
2283 lval = (unsigned long)-val;
2284 } else {
2285 *negp = 0;
2286 lval = (unsigned long)val;
2287 }
2288 *lvalp = jiffies_to_clock_t(lval);
2289 }
2290 return 0;
2291 }
2292
2293 static int do_proc_dointvec_ms_jiffies_conv(int *negp, unsigned long *lvalp,
2294 int *valp,
2295 int write, void *data)
2296 {
2297 if (write) {
2298 *valp = msecs_to_jiffies(*negp ? -*lvalp : *lvalp);
2299 } else {
2300 int val = *valp;
2301 unsigned long lval;
2302 if (val < 0) {
2303 *negp = -1;
2304 lval = (unsigned long)-val;
2305 } else {
2306 *negp = 0;
2307 lval = (unsigned long)val;
2308 }
2309 *lvalp = jiffies_to_msecs(lval);
2310 }
2311 return 0;
2312 }
2313
2314 /**
2315 * proc_dointvec_jiffies - read a vector of integers as seconds
2316 * @table: the sysctl table
2317 * @write: %TRUE if this is a write to the sysctl file
2318 * @filp: the file structure
2319 * @buffer: the user buffer
2320 * @lenp: the size of the user buffer
2321 * @ppos: file position
2322 *
2323 * Reads/writes up to table->maxlen/sizeof(unsigned int) integer
2324 * values from/to the user buffer, treated as an ASCII string.
2325 * The values read are assumed to be in seconds, and are converted into
2326 * jiffies.
2327 *
2328 * Returns 0 on success.
2329 */
2330 int proc_dointvec_jiffies(ctl_table *table, int write, struct file *filp,
2331 void __user *buffer, size_t *lenp, loff_t *ppos)
2332 {
2333 return do_proc_dointvec(table,write,filp,buffer,lenp,ppos,
2334 do_proc_dointvec_jiffies_conv,NULL);
2335 }
2336
2337 /**
2338 * proc_dointvec_userhz_jiffies - read a vector of integers as 1/USER_HZ seconds
2339 * @table: the sysctl table
2340 * @write: %TRUE if this is a write to the sysctl file
2341 * @filp: the file structure
2342 * @buffer: the user buffer
2343 * @lenp: the size of the user buffer
2344 * @ppos: pointer to the file position
2345 *
2346 * Reads/writes up to table->maxlen/sizeof(unsigned int) integer
2347 * values from/to the user buffer, treated as an ASCII string.
2348 * The values read are assumed to be in 1/USER_HZ seconds, and
2349 * are converted into jiffies.
2350 *
2351 * Returns 0 on success.
2352 */
2353 int proc_dointvec_userhz_jiffies(ctl_table *table, int write, struct file *filp,
2354 void __user *buffer, size_t *lenp, loff_t *ppos)
2355 {
2356 return do_proc_dointvec(table,write,filp,buffer,lenp,ppos,
2357 do_proc_dointvec_userhz_jiffies_conv,NULL);
2358 }
2359
2360 /**
2361 * proc_dointvec_ms_jiffies - read a vector of integers as 1 milliseconds
2362 * @table: the sysctl table
2363 * @write: %TRUE if this is a write to the sysctl file
2364 * @filp: the file structure
2365 * @buffer: the user buffer
2366 * @lenp: the size of the user buffer
2367 * @ppos: file position
2368 * @ppos: the current position in the file
2369 *
2370 * Reads/writes up to table->maxlen/sizeof(unsigned int) integer
2371 * values from/to the user buffer, treated as an ASCII string.
2372 * The values read are assumed to be in 1/1000 seconds, and
2373 * are converted into jiffies.
2374 *
2375 * Returns 0 on success.
2376 */
2377 int proc_dointvec_ms_jiffies(ctl_table *table, int write, struct file *filp,
2378 void __user *buffer, size_t *lenp, loff_t *ppos)
2379 {
2380 return do_proc_dointvec(table, write, filp, buffer, lenp, ppos,
2381 do_proc_dointvec_ms_jiffies_conv, NULL);
2382 }
2383
2384 #ifdef CONFIG_SYSVIPC
2385 static int proc_do_ipc_string(ctl_table *table, int write, struct file *filp,
2386 void __user *buffer, size_t *lenp, loff_t *ppos)
2387 {
2388 void *data;
2389 struct ipc_namespace *ns;
2390
2391 ns = current->nsproxy->ipc_ns;
2392
2393 switch (table->ctl_name) {
2394 case KERN_SHMMAX:
2395 data = &ns->shm_ctlmax;
2396 goto proc_minmax;
2397 case KERN_SHMALL:
2398 data = &ns->shm_ctlall;
2399 goto proc_minmax;
2400 case KERN_SHMMNI:
2401 data = &ns->shm_ctlmni;
2402 break;
2403 case KERN_MSGMAX:
2404 data = &ns->msg_ctlmax;
2405 break;
2406 case KERN_MSGMNI:
2407 data = &ns->msg_ctlmni;
2408 break;
2409 case KERN_MSGMNB:
2410 data = &ns->msg_ctlmnb;
2411 break;
2412 case KERN_SEM:
2413 data = &ns->sem_ctls;
2414 break;
2415 default:
2416 return -EINVAL;
2417 }
2418
2419 return __do_proc_dointvec(data, table, write, filp, buffer,
2420 lenp, ppos, NULL, NULL);
2421 proc_minmax:
2422 return __do_proc_doulongvec_minmax(data, table, write, filp, buffer,
2423 lenp, ppos, 1l, 1l);
2424 }
2425 #endif
2426
2427 static int proc_do_cad_pid(ctl_table *table, int write, struct file *filp,
2428 void __user *buffer, size_t *lenp, loff_t *ppos)
2429 {
2430 struct pid *new_pid;
2431 pid_t tmp;
2432 int r;
2433
2434 tmp = pid_nr(cad_pid);
2435
2436 r = __do_proc_dointvec(&tmp, table, write, filp, buffer,
2437 lenp, ppos, NULL, NULL);
2438 if (r || !write)
2439 return r;
2440
2441 new_pid = find_get_pid(tmp);
2442 if (!new_pid)
2443 return -ESRCH;
2444
2445 put_pid(xchg(&cad_pid, new_pid));
2446 return 0;
2447 }
2448
2449 #else /* CONFIG_PROC_FS */
2450
2451 int proc_dostring(ctl_table *table, int write, struct file *filp,
2452 void __user *buffer, size_t *lenp, loff_t *ppos)
2453 {
2454 return -ENOSYS;
2455 }
2456
2457 static int proc_do_uts_string(ctl_table *table, int write, struct file *filp,
2458 void __user *buffer, size_t *lenp, loff_t *ppos)
2459 {
2460 return -ENOSYS;
2461 }
2462
2463 #ifdef CONFIG_SYSVIPC
2464 static int proc_do_ipc_string(ctl_table *table, int write, struct file *filp,
2465 void __user *buffer, size_t *lenp, loff_t *ppos)
2466 {
2467 return -ENOSYS;
2468 }
2469 #endif
2470
2471 int proc_dointvec(ctl_table *table, int write, struct file *filp,
2472 void __user *buffer, size_t *lenp, loff_t *ppos)
2473 {
2474 return -ENOSYS;
2475 }
2476
2477 int proc_dointvec_bset(ctl_table *table, int write, struct file *filp,
2478 void __user *buffer, size_t *lenp, loff_t *ppos)
2479 {
2480 return -ENOSYS;
2481 }
2482
2483 int proc_dointvec_minmax(ctl_table *table, int write, struct file *filp,
2484 void __user *buffer, size_t *lenp, loff_t *ppos)
2485 {
2486 return -ENOSYS;
2487 }
2488
2489 int proc_dointvec_jiffies(ctl_table *table, int write, struct file *filp,
2490 void __user *buffer, size_t *lenp, loff_t *ppos)
2491 {
2492 return -ENOSYS;
2493 }
2494
2495 int proc_dointvec_userhz_jiffies(ctl_table *table, int write, struct file *filp,
2496 void __user *buffer, size_t *lenp, loff_t *ppos)
2497 {
2498 return -ENOSYS;
2499 }
2500
2501 int proc_dointvec_ms_jiffies(ctl_table *table, int write, struct file *filp,
2502 void __user *buffer, size_t *lenp, loff_t *ppos)
2503 {
2504 return -ENOSYS;
2505 }
2506
2507 int proc_doulongvec_minmax(ctl_table *table, int write, struct file *filp,
2508 void __user *buffer, size_t *lenp, loff_t *ppos)
2509 {
2510 return -ENOSYS;
2511 }
2512
2513 int proc_doulongvec_ms_jiffies_minmax(ctl_table *table, int write,
2514 struct file *filp,
2515 void __user *buffer,
2516 size_t *lenp, loff_t *ppos)
2517 {
2518 return -ENOSYS;
2519 }
2520
2521
2522 #endif /* CONFIG_PROC_FS */
2523
2524
2525 #ifdef CONFIG_SYSCTL_SYSCALL
2526 /*
2527 * General sysctl support routines
2528 */
2529
2530 /* The generic string strategy routine: */
2531 int sysctl_string(ctl_table *table, int __user *name, int nlen,
2532 void __user *oldval, size_t __user *oldlenp,
2533 void __user *newval, size_t newlen, void **context)
2534 {
2535 if (!table->data || !table->maxlen)
2536 return -ENOTDIR;
2537
2538 if (oldval && oldlenp) {
2539 size_t bufsize;
2540 if (get_user(bufsize, oldlenp))
2541 return -EFAULT;
2542 if (bufsize) {
2543 size_t len = strlen(table->data), copied;
2544
2545 /* This shouldn't trigger for a well-formed sysctl */
2546 if (len > table->maxlen)
2547 len = table->maxlen;
2548
2549 /* Copy up to a max of bufsize-1 bytes of the string */
2550 copied = (len >= bufsize) ? bufsize - 1 : len;
2551
2552 if (copy_to_user(oldval, table->data, copied) ||
2553 put_user(0, (char __user *)(oldval + copied)))
2554 return -EFAULT;
2555 if (put_user(len, oldlenp))
2556 return -EFAULT;
2557 }
2558 }
2559 if (newval && newlen) {
2560 size_t len = newlen;
2561 if (len > table->maxlen)
2562 len = table->maxlen;
2563 if(copy_from_user(table->data, newval, len))
2564 return -EFAULT;
2565 if (len == table->maxlen)
2566 len--;
2567 ((char *) table->data)[len] = 0;
2568 }
2569 return 1;
2570 }
2571
2572 /*
2573 * This function makes sure that all of the integers in the vector
2574 * are between the minimum and maximum values given in the arrays
2575 * table->extra1 and table->extra2, respectively.
2576 */
2577 int sysctl_intvec(ctl_table *table, int __user *name, int nlen,
2578 void __user *oldval, size_t __user *oldlenp,
2579 void __user *newval, size_t newlen, void **context)
2580 {
2581
2582 if (newval && newlen) {
2583 int __user *vec = (int __user *) newval;
2584 int *min = (int *) table->extra1;
2585 int *max = (int *) table->extra2;
2586 size_t length;
2587 int i;
2588
2589 if (newlen % sizeof(int) != 0)
2590 return -EINVAL;
2591
2592 if (!table->extra1 && !table->extra2)
2593 return 0;
2594
2595 if (newlen > table->maxlen)
2596 newlen = table->maxlen;
2597 length = newlen / sizeof(int);
2598
2599 for (i = 0; i < length; i++) {
2600 int value;
2601 if (get_user(value, vec + i))
2602 return -EFAULT;
2603 if (min && value < min[i])
2604 return -EINVAL;
2605 if (max && value > max[i])
2606 return -EINVAL;
2607 }
2608 }
2609 return 0;
2610 }
2611
2612 /* Strategy function to convert jiffies to seconds */
2613 int sysctl_jiffies(ctl_table *table, int __user *name, int nlen,
2614 void __user *oldval, size_t __user *oldlenp,
2615 void __user *newval, size_t newlen, void **context)
2616 {
2617 if (oldval) {
2618 size_t olen;
2619 if (oldlenp) {
2620 if (get_user(olen, oldlenp))
2621 return -EFAULT;
2622 if (olen!=sizeof(int))
2623 return -EINVAL;
2624 }
2625 if (put_user(*(int *)(table->data)/HZ, (int __user *)oldval) ||
2626 (oldlenp && put_user(sizeof(int),oldlenp)))
2627 return -EFAULT;
2628 }
2629 if (newval && newlen) {
2630 int new;
2631 if (newlen != sizeof(int))
2632 return -EINVAL;
2633 if (get_user(new, (int __user *)newval))
2634 return -EFAULT;
2635 *(int *)(table->data) = new*HZ;
2636 }
2637 return 1;
2638 }
2639
2640 /* Strategy function to convert jiffies to seconds */
2641 int sysctl_ms_jiffies(ctl_table *table, int __user *name, int nlen,
2642 void __user *oldval, size_t __user *oldlenp,
2643 void __user *newval, size_t newlen, void **context)
2644 {
2645 if (oldval) {
2646 size_t olen;
2647 if (oldlenp) {
2648 if (get_user(olen, oldlenp))
2649 return -EFAULT;
2650 if (olen!=sizeof(int))
2651 return -EINVAL;
2652 }
2653 if (put_user(jiffies_to_msecs(*(int *)(table->data)), (int __user *)oldval) ||
2654 (oldlenp && put_user(sizeof(int),oldlenp)))
2655 return -EFAULT;
2656 }
2657 if (newval && newlen) {
2658 int new;
2659 if (newlen != sizeof(int))
2660 return -EINVAL;
2661 if (get_user(new, (int __user *)newval))
2662 return -EFAULT;
2663 *(int *)(table->data) = msecs_to_jiffies(new);
2664 }
2665 return 1;
2666 }
2667
2668 #else /* CONFIG_SYSCTL_SYSCALL */
2669
2670
2671 asmlinkage long sys_sysctl(struct __sysctl_args __user *args)
2672 {
2673 static int msg_count;
2674 struct __sysctl_args tmp;
2675 int name[CTL_MAXNAME];
2676 int i;
2677
2678 /* Read in the sysctl name for better debug message logging */
2679 if (copy_from_user(&tmp, args, sizeof(tmp)))
2680 return -EFAULT;
2681 if (tmp.nlen <= 0 || tmp.nlen >= CTL_MAXNAME)
2682 return -ENOTDIR;
2683 for (i = 0; i < tmp.nlen; i++)
2684 if (get_user(name[i], tmp.name + i))
2685 return -EFAULT;
2686
2687 /* Ignore accesses to kernel.version */
2688 if ((tmp.nlen == 2) && (name[0] == CTL_KERN) && (name[1] == KERN_VERSION))
2689 goto out;
2690
2691 if (msg_count < 5) {
2692 msg_count++;
2693 printk(KERN_INFO
2694 "warning: process `%s' used the removed sysctl "
2695 "system call with ", current->comm);
2696 for (i = 0; i < tmp.nlen; i++)
2697 printk("%d.", name[i]);
2698 printk("\n");
2699 }
2700 out:
2701 return -ENOSYS;
2702 }
2703
2704 int sysctl_string(ctl_table *table, int __user *name, int nlen,
2705 void __user *oldval, size_t __user *oldlenp,
2706 void __user *newval, size_t newlen, void **context)
2707 {
2708 return -ENOSYS;
2709 }
2710
2711 int sysctl_intvec(ctl_table *table, int __user *name, int nlen,
2712 void __user *oldval, size_t __user *oldlenp,
2713 void __user *newval, size_t newlen, void **context)
2714 {
2715 return -ENOSYS;
2716 }
2717
2718 int sysctl_jiffies(ctl_table *table, int __user *name, int nlen,
2719 void __user *oldval, size_t __user *oldlenp,
2720 void __user *newval, size_t newlen, void **context)
2721 {
2722 return -ENOSYS;
2723 }
2724
2725 int sysctl_ms_jiffies(ctl_table *table, int __user *name, int nlen,
2726 void __user *oldval, size_t __user *oldlenp,
2727 void __user *newval, size_t newlen, void **context)
2728 {
2729 return -ENOSYS;
2730 }
2731
2732 #endif /* CONFIG_SYSCTL_SYSCALL */
2733
2734 /*
2735 * No sense putting this after each symbol definition, twice,
2736 * exception granted :-)
2737 */
2738 EXPORT_SYMBOL(proc_dointvec);
2739 EXPORT_SYMBOL(proc_dointvec_jiffies);
2740 EXPORT_SYMBOL(proc_dointvec_minmax);
2741 EXPORT_SYMBOL(proc_dointvec_userhz_jiffies);
2742 EXPORT_SYMBOL(proc_dointvec_ms_jiffies);
2743 EXPORT_SYMBOL(proc_dostring);
2744 EXPORT_SYMBOL(proc_doulongvec_minmax);
2745 EXPORT_SYMBOL(proc_doulongvec_ms_jiffies_minmax);
2746 EXPORT_SYMBOL(register_sysctl_table);
2747 EXPORT_SYMBOL(sysctl_intvec);
2748 EXPORT_SYMBOL(sysctl_jiffies);
2749 EXPORT_SYMBOL(sysctl_ms_jiffies);
2750 EXPORT_SYMBOL(sysctl_string);
2751 EXPORT_SYMBOL(unregister_sysctl_table);
linux v2.6.22.y-op