xref: /linux/kernel/sysctl.c (revision 7f71507851fc7764b36a3221839607d3a45c2025)
1 // SPDX-License-Identifier: GPL-2.0-only
2 /*
3  * sysctl.c: General linux system control interface
4  *
5  * Begun 24 March 1995, Stephen Tweedie
6  * Added /proc support, Dec 1995
7  * Added bdflush entry and intvec min/max checking, 2/23/96, Tom Dyas.
8  * Added hooks for /proc/sys/net (minor, minor patch), 96/4/1, Mike Shaver.
9  * Added kernel/java-{interpreter,appletviewer}, 96/5/10, Mike Shaver.
10  * Dynamic registration fixes, Stephen Tweedie.
11  * Added kswapd-interval, ctrl-alt-del, printk stuff, 1/8/97, Chris Horn.
12  * Made sysctl support optional via CONFIG_SYSCTL, 1/10/97, Chris
13  *  Horn.
14  * Added proc_doulongvec_ms_jiffies_minmax, 09/08/99, Carlos H. Bauer.
15  * Added proc_doulongvec_minmax, 09/08/99, Carlos H. Bauer.
16  * Changed linked lists to use list.h instead of lists.h, 02/24/00, Bill
17  *  Wendling.
18  * The list_for_each() macro wasn't appropriate for the sysctl loop.
19  *  Removed it and replaced it with older style, 03/23/00, Bill Wendling
20  */
21 
22 #include <linux/module.h>
23 #include <linux/mm.h>
24 #include <linux/swap.h>
25 #include <linux/slab.h>
26 #include <linux/sysctl.h>
27 #include <linux/bitmap.h>
28 #include <linux/signal.h>
29 #include <linux/panic.h>
30 #include <linux/printk.h>
31 #include <linux/proc_fs.h>
32 #include <linux/security.h>
33 #include <linux/ctype.h>
34 #include <linux/kmemleak.h>
35 #include <linux/filter.h>
36 #include <linux/fs.h>
37 #include <linux/init.h>
38 #include <linux/kernel.h>
39 #include <linux/kobject.h>
40 #include <linux/net.h>
41 #include <linux/sysrq.h>
42 #include <linux/highuid.h>
43 #include <linux/writeback.h>
44 #include <linux/ratelimit.h>
45 #include <linux/hugetlb.h>
46 #include <linux/initrd.h>
47 #include <linux/key.h>
48 #include <linux/times.h>
49 #include <linux/limits.h>
50 #include <linux/dcache.h>
51 #include <linux/syscalls.h>
52 #include <linux/vmstat.h>
53 #include <linux/nfs_fs.h>
54 #include <linux/acpi.h>
55 #include <linux/reboot.h>
56 #include <linux/ftrace.h>
57 #include <linux/perf_event.h>
58 #include <linux/oom.h>
59 #include <linux/kmod.h>
60 #include <linux/capability.h>
61 #include <linux/binfmts.h>
62 #include <linux/sched/sysctl.h>
63 #include <linux/mount.h>
64 #include <linux/userfaultfd_k.h>
65 #include <linux/pid.h>
66 
67 #include "../lib/kstrtox.h"
68 
69 #include <linux/uaccess.h>
70 #include <asm/processor.h>
71 
72 #ifdef CONFIG_X86
73 #include <asm/nmi.h>
74 #include <asm/stacktrace.h>
75 #include <asm/io.h>
76 #endif
77 #ifdef CONFIG_SPARC
78 #include <asm/setup.h>
79 #endif
80 #ifdef CONFIG_RT_MUTEXES
81 #include <linux/rtmutex.h>
82 #endif
83 
84 /* shared constants to be used in various sysctls */
85 const int sysctl_vals[] = { 0, 1, 2, 3, 4, 100, 200, 1000, 3000, INT_MAX, 65535, -1 };
86 EXPORT_SYMBOL(sysctl_vals);
87 
88 const unsigned long sysctl_long_vals[] = { 0, 1, LONG_MAX };
89 EXPORT_SYMBOL_GPL(sysctl_long_vals);
90 
91 #if defined(CONFIG_SYSCTL)
92 
93 /* Constants used for minimum and maximum */
94 
95 #ifdef CONFIG_PERF_EVENTS
96 static const int six_hundred_forty_kb = 640 * 1024;
97 #endif
98 
99 
100 static const int ngroups_max = NGROUPS_MAX;
101 static const int cap_last_cap = CAP_LAST_CAP;
102 
103 #ifdef CONFIG_PROC_SYSCTL
104 
105 /**
106  * enum sysctl_writes_mode - supported sysctl write modes
107  *
108  * @SYSCTL_WRITES_LEGACY: each write syscall must fully contain the sysctl value
109  *	to be written, and multiple writes on the same sysctl file descriptor
110  *	will rewrite the sysctl value, regardless of file position. No warning
111  *	is issued when the initial position is not 0.
112  * @SYSCTL_WRITES_WARN: same as above but warn when the initial file position is
113  *	not 0.
114  * @SYSCTL_WRITES_STRICT: writes to numeric sysctl entries must always be at
115  *	file position 0 and the value must be fully contained in the buffer
116  *	sent to the write syscall. If dealing with strings respect the file
117  *	position, but restrict this to the max length of the buffer, anything
118  *	passed the max length will be ignored. Multiple writes will append
119  *	to the buffer.
120  *
121  * These write modes control how current file position affects the behavior of
122  * updating sysctl values through the proc interface on each write.
123  */
124 enum sysctl_writes_mode {
125 	SYSCTL_WRITES_LEGACY		= -1,
126 	SYSCTL_WRITES_WARN		= 0,
127 	SYSCTL_WRITES_STRICT		= 1,
128 };
129 
130 static enum sysctl_writes_mode sysctl_writes_strict = SYSCTL_WRITES_STRICT;
131 #endif /* CONFIG_PROC_SYSCTL */
132 
133 #if defined(HAVE_ARCH_PICK_MMAP_LAYOUT) || \
134     defined(CONFIG_ARCH_WANT_DEFAULT_TOPDOWN_MMAP_LAYOUT)
135 int sysctl_legacy_va_layout;
136 #endif
137 
138 #endif /* CONFIG_SYSCTL */
139 
140 /*
141  * /proc/sys support
142  */
143 
144 #ifdef CONFIG_PROC_SYSCTL
145 
146 static int _proc_do_string(char *data, int maxlen, int write,
147 		char *buffer, size_t *lenp, loff_t *ppos)
148 {
149 	size_t len;
150 	char c, *p;
151 
152 	if (!data || !maxlen || !*lenp) {
153 		*lenp = 0;
154 		return 0;
155 	}
156 
157 	if (write) {
158 		if (sysctl_writes_strict == SYSCTL_WRITES_STRICT) {
159 			/* Only continue writes not past the end of buffer. */
160 			len = strlen(data);
161 			if (len > maxlen - 1)
162 				len = maxlen - 1;
163 
164 			if (*ppos > len)
165 				return 0;
166 			len = *ppos;
167 		} else {
168 			/* Start writing from beginning of buffer. */
169 			len = 0;
170 		}
171 
172 		*ppos += *lenp;
173 		p = buffer;
174 		while ((p - buffer) < *lenp && len < maxlen - 1) {
175 			c = *(p++);
176 			if (c == 0 || c == '\n')
177 				break;
178 			data[len++] = c;
179 		}
180 		data[len] = 0;
181 	} else {
182 		len = strlen(data);
183 		if (len > maxlen)
184 			len = maxlen;
185 
186 		if (*ppos > len) {
187 			*lenp = 0;
188 			return 0;
189 		}
190 
191 		data += *ppos;
192 		len  -= *ppos;
193 
194 		if (len > *lenp)
195 			len = *lenp;
196 		if (len)
197 			memcpy(buffer, data, len);
198 		if (len < *lenp) {
199 			buffer[len] = '\n';
200 			len++;
201 		}
202 		*lenp = len;
203 		*ppos += len;
204 	}
205 	return 0;
206 }
207 
208 static void warn_sysctl_write(const struct ctl_table *table)
209 {
210 	pr_warn_once("%s wrote to %s when file position was not 0!\n"
211 		"This will not be supported in the future. To silence this\n"
212 		"warning, set kernel.sysctl_writes_strict = -1\n",
213 		current->comm, table->procname);
214 }
215 
216 /**
217  * proc_first_pos_non_zero_ignore - check if first position is allowed
218  * @ppos: file position
219  * @table: the sysctl table
220  *
221  * Returns true if the first position is non-zero and the sysctl_writes_strict
222  * mode indicates this is not allowed for numeric input types. String proc
223  * handlers can ignore the return value.
224  */
225 static bool proc_first_pos_non_zero_ignore(loff_t *ppos,
226 					   const struct ctl_table *table)
227 {
228 	if (!*ppos)
229 		return false;
230 
231 	switch (sysctl_writes_strict) {
232 	case SYSCTL_WRITES_STRICT:
233 		return true;
234 	case SYSCTL_WRITES_WARN:
235 		warn_sysctl_write(table);
236 		return false;
237 	default:
238 		return false;
239 	}
240 }
241 
242 /**
243  * proc_dostring - read a string sysctl
244  * @table: the sysctl table
245  * @write: %TRUE if this is a write to the sysctl file
246  * @buffer: the user buffer
247  * @lenp: the size of the user buffer
248  * @ppos: file position
249  *
250  * Reads/writes a string from/to the user buffer. If the kernel
251  * buffer provided is not large enough to hold the string, the
252  * string is truncated. The copied string is %NULL-terminated.
253  * If the string is being read by the user process, it is copied
254  * and a newline '\n' is added. It is truncated if the buffer is
255  * not large enough.
256  *
257  * Returns 0 on success.
258  */
259 int proc_dostring(const struct ctl_table *table, int write,
260 		  void *buffer, size_t *lenp, loff_t *ppos)
261 {
262 	if (write)
263 		proc_first_pos_non_zero_ignore(ppos, table);
264 
265 	return _proc_do_string(table->data, table->maxlen, write, buffer, lenp,
266 			ppos);
267 }
268 
269 static void proc_skip_spaces(char **buf, size_t *size)
270 {
271 	while (*size) {
272 		if (!isspace(**buf))
273 			break;
274 		(*size)--;
275 		(*buf)++;
276 	}
277 }
278 
279 static void proc_skip_char(char **buf, size_t *size, const char v)
280 {
281 	while (*size) {
282 		if (**buf != v)
283 			break;
284 		(*size)--;
285 		(*buf)++;
286 	}
287 }
288 
289 /**
290  * strtoul_lenient - parse an ASCII formatted integer from a buffer and only
291  *                   fail on overflow
292  *
293  * @cp: kernel buffer containing the string to parse
294  * @endp: pointer to store the trailing characters
295  * @base: the base to use
296  * @res: where the parsed integer will be stored
297  *
298  * In case of success 0 is returned and @res will contain the parsed integer,
299  * @endp will hold any trailing characters.
300  * This function will fail the parse on overflow. If there wasn't an overflow
301  * the function will defer the decision what characters count as invalid to the
302  * caller.
303  */
304 static int strtoul_lenient(const char *cp, char **endp, unsigned int base,
305 			   unsigned long *res)
306 {
307 	unsigned long long result;
308 	unsigned int rv;
309 
310 	cp = _parse_integer_fixup_radix(cp, &base);
311 	rv = _parse_integer(cp, base, &result);
312 	if ((rv & KSTRTOX_OVERFLOW) || (result != (unsigned long)result))
313 		return -ERANGE;
314 
315 	cp += rv;
316 
317 	if (endp)
318 		*endp = (char *)cp;
319 
320 	*res = (unsigned long)result;
321 	return 0;
322 }
323 
324 #define TMPBUFLEN 22
325 /**
326  * proc_get_long - reads an ASCII formatted integer from a user buffer
327  *
328  * @buf: a kernel buffer
329  * @size: size of the kernel buffer
330  * @val: this is where the number will be stored
331  * @neg: set to %TRUE if number is negative
332  * @perm_tr: a vector which contains the allowed trailers
333  * @perm_tr_len: size of the perm_tr vector
334  * @tr: pointer to store the trailer character
335  *
336  * In case of success %0 is returned and @buf and @size are updated with
337  * the amount of bytes read. If @tr is non-NULL and a trailing
338  * character exists (size is non-zero after returning from this
339  * function), @tr is updated with the trailing character.
340  */
341 static int proc_get_long(char **buf, size_t *size,
342 			  unsigned long *val, bool *neg,
343 			  const char *perm_tr, unsigned perm_tr_len, char *tr)
344 {
345 	char *p, tmp[TMPBUFLEN];
346 	ssize_t len = *size;
347 
348 	if (len <= 0)
349 		return -EINVAL;
350 
351 	if (len > TMPBUFLEN - 1)
352 		len = TMPBUFLEN - 1;
353 
354 	memcpy(tmp, *buf, len);
355 
356 	tmp[len] = 0;
357 	p = tmp;
358 	if (*p == '-' && *size > 1) {
359 		*neg = true;
360 		p++;
361 	} else
362 		*neg = false;
363 	if (!isdigit(*p))
364 		return -EINVAL;
365 
366 	if (strtoul_lenient(p, &p, 0, val))
367 		return -EINVAL;
368 
369 	len = p - tmp;
370 
371 	/* We don't know if the next char is whitespace thus we may accept
372 	 * invalid integers (e.g. 1234...a) or two integers instead of one
373 	 * (e.g. 123...1). So lets not allow such large numbers. */
374 	if (len == TMPBUFLEN - 1)
375 		return -EINVAL;
376 
377 	if (len < *size && perm_tr_len && !memchr(perm_tr, *p, perm_tr_len))
378 		return -EINVAL;
379 
380 	if (tr && (len < *size))
381 		*tr = *p;
382 
383 	*buf += len;
384 	*size -= len;
385 
386 	return 0;
387 }
388 
389 /**
390  * proc_put_long - converts an integer to a decimal ASCII formatted string
391  *
392  * @buf: the user buffer
393  * @size: the size of the user buffer
394  * @val: the integer to be converted
395  * @neg: sign of the number, %TRUE for negative
396  *
397  * In case of success @buf and @size are updated with the amount of bytes
398  * written.
399  */
400 static void proc_put_long(void **buf, size_t *size, unsigned long val, bool neg)
401 {
402 	int len;
403 	char tmp[TMPBUFLEN], *p = tmp;
404 
405 	sprintf(p, "%s%lu", neg ? "-" : "", val);
406 	len = strlen(tmp);
407 	if (len > *size)
408 		len = *size;
409 	memcpy(*buf, tmp, len);
410 	*size -= len;
411 	*buf += len;
412 }
413 #undef TMPBUFLEN
414 
415 static void proc_put_char(void **buf, size_t *size, char c)
416 {
417 	if (*size) {
418 		char **buffer = (char **)buf;
419 		**buffer = c;
420 
421 		(*size)--;
422 		(*buffer)++;
423 		*buf = *buffer;
424 	}
425 }
426 
427 static int do_proc_dointvec_conv(bool *negp, unsigned long *lvalp,
428 				 int *valp,
429 				 int write, void *data)
430 {
431 	if (write) {
432 		if (*negp) {
433 			if (*lvalp > (unsigned long) INT_MAX + 1)
434 				return -EINVAL;
435 			WRITE_ONCE(*valp, -*lvalp);
436 		} else {
437 			if (*lvalp > (unsigned long) INT_MAX)
438 				return -EINVAL;
439 			WRITE_ONCE(*valp, *lvalp);
440 		}
441 	} else {
442 		int val = READ_ONCE(*valp);
443 		if (val < 0) {
444 			*negp = true;
445 			*lvalp = -(unsigned long)val;
446 		} else {
447 			*negp = false;
448 			*lvalp = (unsigned long)val;
449 		}
450 	}
451 	return 0;
452 }
453 
454 static int do_proc_douintvec_conv(unsigned long *lvalp,
455 				  unsigned int *valp,
456 				  int write, void *data)
457 {
458 	if (write) {
459 		if (*lvalp > UINT_MAX)
460 			return -EINVAL;
461 		WRITE_ONCE(*valp, *lvalp);
462 	} else {
463 		unsigned int val = READ_ONCE(*valp);
464 		*lvalp = (unsigned long)val;
465 	}
466 	return 0;
467 }
468 
469 static const char proc_wspace_sep[] = { ' ', '\t', '\n' };
470 
471 static int __do_proc_dointvec(void *tbl_data, const struct ctl_table *table,
472 		  int write, void *buffer,
473 		  size_t *lenp, loff_t *ppos,
474 		  int (*conv)(bool *negp, unsigned long *lvalp, int *valp,
475 			      int write, void *data),
476 		  void *data)
477 {
478 	int *i, vleft, first = 1, err = 0;
479 	size_t left;
480 	char *p;
481 
482 	if (!tbl_data || !table->maxlen || !*lenp || (*ppos && !write)) {
483 		*lenp = 0;
484 		return 0;
485 	}
486 
487 	i = (int *) tbl_data;
488 	vleft = table->maxlen / sizeof(*i);
489 	left = *lenp;
490 
491 	if (!conv)
492 		conv = do_proc_dointvec_conv;
493 
494 	if (write) {
495 		if (proc_first_pos_non_zero_ignore(ppos, table))
496 			goto out;
497 
498 		if (left > PAGE_SIZE - 1)
499 			left = PAGE_SIZE - 1;
500 		p = buffer;
501 	}
502 
503 	for (; left && vleft--; i++, first=0) {
504 		unsigned long lval;
505 		bool neg;
506 
507 		if (write) {
508 			proc_skip_spaces(&p, &left);
509 
510 			if (!left)
511 				break;
512 			err = proc_get_long(&p, &left, &lval, &neg,
513 					     proc_wspace_sep,
514 					     sizeof(proc_wspace_sep), NULL);
515 			if (err)
516 				break;
517 			if (conv(&neg, &lval, i, 1, data)) {
518 				err = -EINVAL;
519 				break;
520 			}
521 		} else {
522 			if (conv(&neg, &lval, i, 0, data)) {
523 				err = -EINVAL;
524 				break;
525 			}
526 			if (!first)
527 				proc_put_char(&buffer, &left, '\t');
528 			proc_put_long(&buffer, &left, lval, neg);
529 		}
530 	}
531 
532 	if (!write && !first && left && !err)
533 		proc_put_char(&buffer, &left, '\n');
534 	if (write && !err && left)
535 		proc_skip_spaces(&p, &left);
536 	if (write && first)
537 		return err ? : -EINVAL;
538 	*lenp -= left;
539 out:
540 	*ppos += *lenp;
541 	return err;
542 }
543 
544 static int do_proc_dointvec(const struct ctl_table *table, int write,
545 		  void *buffer, size_t *lenp, loff_t *ppos,
546 		  int (*conv)(bool *negp, unsigned long *lvalp, int *valp,
547 			      int write, void *data),
548 		  void *data)
549 {
550 	return __do_proc_dointvec(table->data, table, write,
551 			buffer, lenp, ppos, conv, data);
552 }
553 
554 static int do_proc_douintvec_w(unsigned int *tbl_data,
555 			       const struct ctl_table *table,
556 			       void *buffer,
557 			       size_t *lenp, loff_t *ppos,
558 			       int (*conv)(unsigned long *lvalp,
559 					   unsigned int *valp,
560 					   int write, void *data),
561 			       void *data)
562 {
563 	unsigned long lval;
564 	int err = 0;
565 	size_t left;
566 	bool neg;
567 	char *p = buffer;
568 
569 	left = *lenp;
570 
571 	if (proc_first_pos_non_zero_ignore(ppos, table))
572 		goto bail_early;
573 
574 	if (left > PAGE_SIZE - 1)
575 		left = PAGE_SIZE - 1;
576 
577 	proc_skip_spaces(&p, &left);
578 	if (!left) {
579 		err = -EINVAL;
580 		goto out_free;
581 	}
582 
583 	err = proc_get_long(&p, &left, &lval, &neg,
584 			     proc_wspace_sep,
585 			     sizeof(proc_wspace_sep), NULL);
586 	if (err || neg) {
587 		err = -EINVAL;
588 		goto out_free;
589 	}
590 
591 	if (conv(&lval, tbl_data, 1, data)) {
592 		err = -EINVAL;
593 		goto out_free;
594 	}
595 
596 	if (!err && left)
597 		proc_skip_spaces(&p, &left);
598 
599 out_free:
600 	if (err)
601 		return -EINVAL;
602 
603 	return 0;
604 
605 	/* This is in keeping with old __do_proc_dointvec() */
606 bail_early:
607 	*ppos += *lenp;
608 	return err;
609 }
610 
611 static int do_proc_douintvec_r(unsigned int *tbl_data, void *buffer,
612 			       size_t *lenp, loff_t *ppos,
613 			       int (*conv)(unsigned long *lvalp,
614 					   unsigned int *valp,
615 					   int write, void *data),
616 			       void *data)
617 {
618 	unsigned long lval;
619 	int err = 0;
620 	size_t left;
621 
622 	left = *lenp;
623 
624 	if (conv(&lval, tbl_data, 0, data)) {
625 		err = -EINVAL;
626 		goto out;
627 	}
628 
629 	proc_put_long(&buffer, &left, lval, false);
630 	if (!left)
631 		goto out;
632 
633 	proc_put_char(&buffer, &left, '\n');
634 
635 out:
636 	*lenp -= left;
637 	*ppos += *lenp;
638 
639 	return err;
640 }
641 
642 static int __do_proc_douintvec(void *tbl_data, const struct ctl_table *table,
643 			       int write, void *buffer,
644 			       size_t *lenp, loff_t *ppos,
645 			       int (*conv)(unsigned long *lvalp,
646 					   unsigned int *valp,
647 					   int write, void *data),
648 			       void *data)
649 {
650 	unsigned int *i, vleft;
651 
652 	if (!tbl_data || !table->maxlen || !*lenp || (*ppos && !write)) {
653 		*lenp = 0;
654 		return 0;
655 	}
656 
657 	i = (unsigned int *) tbl_data;
658 	vleft = table->maxlen / sizeof(*i);
659 
660 	/*
661 	 * Arrays are not supported, keep this simple. *Do not* add
662 	 * support for them.
663 	 */
664 	if (vleft != 1) {
665 		*lenp = 0;
666 		return -EINVAL;
667 	}
668 
669 	if (!conv)
670 		conv = do_proc_douintvec_conv;
671 
672 	if (write)
673 		return do_proc_douintvec_w(i, table, buffer, lenp, ppos,
674 					   conv, data);
675 	return do_proc_douintvec_r(i, buffer, lenp, ppos, conv, data);
676 }
677 
678 int do_proc_douintvec(const struct ctl_table *table, int write,
679 		      void *buffer, size_t *lenp, loff_t *ppos,
680 		      int (*conv)(unsigned long *lvalp,
681 				  unsigned int *valp,
682 				  int write, void *data),
683 		      void *data)
684 {
685 	return __do_proc_douintvec(table->data, table, write,
686 				   buffer, lenp, ppos, conv, data);
687 }
688 
689 /**
690  * proc_dobool - read/write a bool
691  * @table: the sysctl table
692  * @write: %TRUE if this is a write to the sysctl file
693  * @buffer: the user buffer
694  * @lenp: the size of the user buffer
695  * @ppos: file position
696  *
697  * Reads/writes one integer value from/to the user buffer,
698  * treated as an ASCII string.
699  *
700  * table->data must point to a bool variable and table->maxlen must
701  * be sizeof(bool).
702  *
703  * Returns 0 on success.
704  */
705 int proc_dobool(const struct ctl_table *table, int write, void *buffer,
706 		size_t *lenp, loff_t *ppos)
707 {
708 	struct ctl_table tmp;
709 	bool *data = table->data;
710 	int res, val;
711 
712 	/* Do not support arrays yet. */
713 	if (table->maxlen != sizeof(bool))
714 		return -EINVAL;
715 
716 	tmp = *table;
717 	tmp.maxlen = sizeof(val);
718 	tmp.data = &val;
719 
720 	val = READ_ONCE(*data);
721 	res = proc_dointvec(&tmp, write, buffer, lenp, ppos);
722 	if (res)
723 		return res;
724 	if (write)
725 		WRITE_ONCE(*data, val);
726 	return 0;
727 }
728 
729 /**
730  * proc_dointvec - read a vector of integers
731  * @table: the sysctl table
732  * @write: %TRUE if this is a write to the sysctl file
733  * @buffer: the user buffer
734  * @lenp: the size of the user buffer
735  * @ppos: file position
736  *
737  * Reads/writes up to table->maxlen/sizeof(unsigned int) integer
738  * values from/to the user buffer, treated as an ASCII string.
739  *
740  * Returns 0 on success.
741  */
742 int proc_dointvec(const struct ctl_table *table, int write, void *buffer,
743 		  size_t *lenp, loff_t *ppos)
744 {
745 	return do_proc_dointvec(table, write, buffer, lenp, ppos, NULL, NULL);
746 }
747 
748 /**
749  * proc_douintvec - read a vector of unsigned integers
750  * @table: the sysctl table
751  * @write: %TRUE if this is a write to the sysctl file
752  * @buffer: the user buffer
753  * @lenp: the size of the user buffer
754  * @ppos: file position
755  *
756  * Reads/writes up to table->maxlen/sizeof(unsigned int) unsigned integer
757  * values from/to the user buffer, treated as an ASCII string.
758  *
759  * Returns 0 on success.
760  */
761 int proc_douintvec(const struct ctl_table *table, int write, void *buffer,
762 		size_t *lenp, loff_t *ppos)
763 {
764 	return do_proc_douintvec(table, write, buffer, lenp, ppos,
765 				 do_proc_douintvec_conv, NULL);
766 }
767 
768 /*
769  * Taint values can only be increased
770  * This means we can safely use a temporary.
771  */
772 static int proc_taint(const struct ctl_table *table, int write,
773 			       void *buffer, size_t *lenp, loff_t *ppos)
774 {
775 	struct ctl_table t;
776 	unsigned long tmptaint = get_taint();
777 	int err;
778 
779 	if (write && !capable(CAP_SYS_ADMIN))
780 		return -EPERM;
781 
782 	t = *table;
783 	t.data = &tmptaint;
784 	err = proc_doulongvec_minmax(&t, write, buffer, lenp, ppos);
785 	if (err < 0)
786 		return err;
787 
788 	if (write) {
789 		int i;
790 
791 		/*
792 		 * If we are relying on panic_on_taint not producing
793 		 * false positives due to userspace input, bail out
794 		 * before setting the requested taint flags.
795 		 */
796 		if (panic_on_taint_nousertaint && (tmptaint & panic_on_taint))
797 			return -EINVAL;
798 
799 		/*
800 		 * Poor man's atomic or. Not worth adding a primitive
801 		 * to everyone's atomic.h for this
802 		 */
803 		for (i = 0; i < TAINT_FLAGS_COUNT; i++)
804 			if ((1UL << i) & tmptaint)
805 				add_taint(i, LOCKDEP_STILL_OK);
806 	}
807 
808 	return err;
809 }
810 
811 /**
812  * struct do_proc_dointvec_minmax_conv_param - proc_dointvec_minmax() range checking structure
813  * @min: pointer to minimum allowable value
814  * @max: pointer to maximum allowable value
815  *
816  * The do_proc_dointvec_minmax_conv_param structure provides the
817  * minimum and maximum values for doing range checking for those sysctl
818  * parameters that use the proc_dointvec_minmax() handler.
819  */
820 struct do_proc_dointvec_minmax_conv_param {
821 	int *min;
822 	int *max;
823 };
824 
825 static int do_proc_dointvec_minmax_conv(bool *negp, unsigned long *lvalp,
826 					int *valp,
827 					int write, void *data)
828 {
829 	int tmp, ret;
830 	struct do_proc_dointvec_minmax_conv_param *param = data;
831 	/*
832 	 * If writing, first do so via a temporary local int so we can
833 	 * bounds-check it before touching *valp.
834 	 */
835 	int *ip = write ? &tmp : valp;
836 
837 	ret = do_proc_dointvec_conv(negp, lvalp, ip, write, data);
838 	if (ret)
839 		return ret;
840 
841 	if (write) {
842 		if ((param->min && *param->min > tmp) ||
843 		    (param->max && *param->max < tmp))
844 			return -EINVAL;
845 		WRITE_ONCE(*valp, tmp);
846 	}
847 
848 	return 0;
849 }
850 
851 /**
852  * proc_dointvec_minmax - read a vector of integers with min/max values
853  * @table: the sysctl table
854  * @write: %TRUE if this is a write to the sysctl file
855  * @buffer: the user buffer
856  * @lenp: the size of the user buffer
857  * @ppos: file position
858  *
859  * Reads/writes up to table->maxlen/sizeof(unsigned int) integer
860  * values from/to the user buffer, treated as an ASCII string.
861  *
862  * This routine will ensure the values are within the range specified by
863  * table->extra1 (min) and table->extra2 (max).
864  *
865  * Returns 0 on success or -EINVAL on write when the range check fails.
866  */
867 int proc_dointvec_minmax(const struct ctl_table *table, int write,
868 		  void *buffer, size_t *lenp, loff_t *ppos)
869 {
870 	struct do_proc_dointvec_minmax_conv_param param = {
871 		.min = (int *) table->extra1,
872 		.max = (int *) table->extra2,
873 	};
874 	return do_proc_dointvec(table, write, buffer, lenp, ppos,
875 				do_proc_dointvec_minmax_conv, &param);
876 }
877 
878 /**
879  * struct do_proc_douintvec_minmax_conv_param - proc_douintvec_minmax() range checking structure
880  * @min: pointer to minimum allowable value
881  * @max: pointer to maximum allowable value
882  *
883  * The do_proc_douintvec_minmax_conv_param structure provides the
884  * minimum and maximum values for doing range checking for those sysctl
885  * parameters that use the proc_douintvec_minmax() handler.
886  */
887 struct do_proc_douintvec_minmax_conv_param {
888 	unsigned int *min;
889 	unsigned int *max;
890 };
891 
892 static int do_proc_douintvec_minmax_conv(unsigned long *lvalp,
893 					 unsigned int *valp,
894 					 int write, void *data)
895 {
896 	int ret;
897 	unsigned int tmp;
898 	struct do_proc_douintvec_minmax_conv_param *param = data;
899 	/* write via temporary local uint for bounds-checking */
900 	unsigned int *up = write ? &tmp : valp;
901 
902 	ret = do_proc_douintvec_conv(lvalp, up, write, data);
903 	if (ret)
904 		return ret;
905 
906 	if (write) {
907 		if ((param->min && *param->min > tmp) ||
908 		    (param->max && *param->max < tmp))
909 			return -ERANGE;
910 
911 		WRITE_ONCE(*valp, tmp);
912 	}
913 
914 	return 0;
915 }
916 
917 /**
918  * proc_douintvec_minmax - read a vector of unsigned ints with min/max values
919  * @table: the sysctl table
920  * @write: %TRUE if this is a write to the sysctl file
921  * @buffer: the user buffer
922  * @lenp: the size of the user buffer
923  * @ppos: file position
924  *
925  * Reads/writes up to table->maxlen/sizeof(unsigned int) unsigned integer
926  * values from/to the user buffer, treated as an ASCII string. Negative
927  * strings are not allowed.
928  *
929  * This routine will ensure the values are within the range specified by
930  * table->extra1 (min) and table->extra2 (max). There is a final sanity
931  * check for UINT_MAX to avoid having to support wrap around uses from
932  * userspace.
933  *
934  * Returns 0 on success or -ERANGE on write when the range check fails.
935  */
936 int proc_douintvec_minmax(const struct ctl_table *table, int write,
937 			  void *buffer, size_t *lenp, loff_t *ppos)
938 {
939 	struct do_proc_douintvec_minmax_conv_param param = {
940 		.min = (unsigned int *) table->extra1,
941 		.max = (unsigned int *) table->extra2,
942 	};
943 	return do_proc_douintvec(table, write, buffer, lenp, ppos,
944 				 do_proc_douintvec_minmax_conv, &param);
945 }
946 
947 /**
948  * proc_dou8vec_minmax - read a vector of unsigned chars with min/max values
949  * @table: the sysctl table
950  * @write: %TRUE if this is a write to the sysctl file
951  * @buffer: the user buffer
952  * @lenp: the size of the user buffer
953  * @ppos: file position
954  *
955  * Reads/writes up to table->maxlen/sizeof(u8) unsigned chars
956  * values from/to the user buffer, treated as an ASCII string. Negative
957  * strings are not allowed.
958  *
959  * This routine will ensure the values are within the range specified by
960  * table->extra1 (min) and table->extra2 (max).
961  *
962  * Returns 0 on success or an error on write when the range check fails.
963  */
964 int proc_dou8vec_minmax(const struct ctl_table *table, int write,
965 			void *buffer, size_t *lenp, loff_t *ppos)
966 {
967 	struct ctl_table tmp;
968 	unsigned int min = 0, max = 255U, val;
969 	u8 *data = table->data;
970 	struct do_proc_douintvec_minmax_conv_param param = {
971 		.min = &min,
972 		.max = &max,
973 	};
974 	int res;
975 
976 	/* Do not support arrays yet. */
977 	if (table->maxlen != sizeof(u8))
978 		return -EINVAL;
979 
980 	if (table->extra1)
981 		min = *(unsigned int *) table->extra1;
982 	if (table->extra2)
983 		max = *(unsigned int *) table->extra2;
984 
985 	tmp = *table;
986 
987 	tmp.maxlen = sizeof(val);
988 	tmp.data = &val;
989 	val = READ_ONCE(*data);
990 	res = do_proc_douintvec(&tmp, write, buffer, lenp, ppos,
991 				do_proc_douintvec_minmax_conv, &param);
992 	if (res)
993 		return res;
994 	if (write)
995 		WRITE_ONCE(*data, val);
996 	return 0;
997 }
998 EXPORT_SYMBOL_GPL(proc_dou8vec_minmax);
999 
1000 #ifdef CONFIG_MAGIC_SYSRQ
1001 static int sysrq_sysctl_handler(const struct ctl_table *table, int write,
1002 				void *buffer, size_t *lenp, loff_t *ppos)
1003 {
1004 	int tmp, ret;
1005 
1006 	tmp = sysrq_mask();
1007 
1008 	ret = __do_proc_dointvec(&tmp, table, write, buffer,
1009 			       lenp, ppos, NULL, NULL);
1010 	if (ret || !write)
1011 		return ret;
1012 
1013 	if (write)
1014 		sysrq_toggle_support(tmp);
1015 
1016 	return 0;
1017 }
1018 #endif
1019 
1020 static int __do_proc_doulongvec_minmax(void *data,
1021 		const struct ctl_table *table, int write,
1022 		void *buffer, size_t *lenp, loff_t *ppos,
1023 		unsigned long convmul, unsigned long convdiv)
1024 {
1025 	unsigned long *i, *min, *max;
1026 	int vleft, first = 1, err = 0;
1027 	size_t left;
1028 	char *p;
1029 
1030 	if (!data || !table->maxlen || !*lenp || (*ppos && !write)) {
1031 		*lenp = 0;
1032 		return 0;
1033 	}
1034 
1035 	i = data;
1036 	min = table->extra1;
1037 	max = table->extra2;
1038 	vleft = table->maxlen / sizeof(unsigned long);
1039 	left = *lenp;
1040 
1041 	if (write) {
1042 		if (proc_first_pos_non_zero_ignore(ppos, table))
1043 			goto out;
1044 
1045 		if (left > PAGE_SIZE - 1)
1046 			left = PAGE_SIZE - 1;
1047 		p = buffer;
1048 	}
1049 
1050 	for (; left && vleft--; i++, first = 0) {
1051 		unsigned long val;
1052 
1053 		if (write) {
1054 			bool neg;
1055 
1056 			proc_skip_spaces(&p, &left);
1057 			if (!left)
1058 				break;
1059 
1060 			err = proc_get_long(&p, &left, &val, &neg,
1061 					     proc_wspace_sep,
1062 					     sizeof(proc_wspace_sep), NULL);
1063 			if (err || neg) {
1064 				err = -EINVAL;
1065 				break;
1066 			}
1067 
1068 			val = convmul * val / convdiv;
1069 			if ((min && val < *min) || (max && val > *max)) {
1070 				err = -EINVAL;
1071 				break;
1072 			}
1073 			WRITE_ONCE(*i, val);
1074 		} else {
1075 			val = convdiv * READ_ONCE(*i) / convmul;
1076 			if (!first)
1077 				proc_put_char(&buffer, &left, '\t');
1078 			proc_put_long(&buffer, &left, val, false);
1079 		}
1080 	}
1081 
1082 	if (!write && !first && left && !err)
1083 		proc_put_char(&buffer, &left, '\n');
1084 	if (write && !err)
1085 		proc_skip_spaces(&p, &left);
1086 	if (write && first)
1087 		return err ? : -EINVAL;
1088 	*lenp -= left;
1089 out:
1090 	*ppos += *lenp;
1091 	return err;
1092 }
1093 
1094 static int do_proc_doulongvec_minmax(const struct ctl_table *table, int write,
1095 		void *buffer, size_t *lenp, loff_t *ppos, unsigned long convmul,
1096 		unsigned long convdiv)
1097 {
1098 	return __do_proc_doulongvec_minmax(table->data, table, write,
1099 			buffer, lenp, ppos, convmul, convdiv);
1100 }
1101 
1102 /**
1103  * proc_doulongvec_minmax - read a vector of long integers with min/max values
1104  * @table: the sysctl table
1105  * @write: %TRUE if this is a write to the sysctl file
1106  * @buffer: the user buffer
1107  * @lenp: the size of the user buffer
1108  * @ppos: file position
1109  *
1110  * Reads/writes up to table->maxlen/sizeof(unsigned long) unsigned long
1111  * values from/to the user buffer, treated as an ASCII string.
1112  *
1113  * This routine will ensure the values are within the range specified by
1114  * table->extra1 (min) and table->extra2 (max).
1115  *
1116  * Returns 0 on success.
1117  */
1118 int proc_doulongvec_minmax(const struct ctl_table *table, int write,
1119 			   void *buffer, size_t *lenp, loff_t *ppos)
1120 {
1121     return do_proc_doulongvec_minmax(table, write, buffer, lenp, ppos, 1l, 1l);
1122 }
1123 
1124 /**
1125  * proc_doulongvec_ms_jiffies_minmax - read a vector of millisecond values with min/max values
1126  * @table: the sysctl table
1127  * @write: %TRUE if this is a write to the sysctl file
1128  * @buffer: the user buffer
1129  * @lenp: the size of the user buffer
1130  * @ppos: file position
1131  *
1132  * Reads/writes up to table->maxlen/sizeof(unsigned long) unsigned long
1133  * values from/to the user buffer, treated as an ASCII string. The values
1134  * are treated as milliseconds, and converted to jiffies when they are stored.
1135  *
1136  * This routine will ensure the values are within the range specified by
1137  * table->extra1 (min) and table->extra2 (max).
1138  *
1139  * Returns 0 on success.
1140  */
1141 int proc_doulongvec_ms_jiffies_minmax(const struct ctl_table *table, int write,
1142 				      void *buffer, size_t *lenp, loff_t *ppos)
1143 {
1144     return do_proc_doulongvec_minmax(table, write, buffer,
1145 				     lenp, ppos, HZ, 1000l);
1146 }
1147 
1148 
1149 static int do_proc_dointvec_jiffies_conv(bool *negp, unsigned long *lvalp,
1150 					 int *valp,
1151 					 int write, void *data)
1152 {
1153 	if (write) {
1154 		if (*lvalp > INT_MAX / HZ)
1155 			return 1;
1156 		if (*negp)
1157 			WRITE_ONCE(*valp, -*lvalp * HZ);
1158 		else
1159 			WRITE_ONCE(*valp, *lvalp * HZ);
1160 	} else {
1161 		int val = READ_ONCE(*valp);
1162 		unsigned long lval;
1163 		if (val < 0) {
1164 			*negp = true;
1165 			lval = -(unsigned long)val;
1166 		} else {
1167 			*negp = false;
1168 			lval = (unsigned long)val;
1169 		}
1170 		*lvalp = lval / HZ;
1171 	}
1172 	return 0;
1173 }
1174 
1175 static int do_proc_dointvec_userhz_jiffies_conv(bool *negp, unsigned long *lvalp,
1176 						int *valp,
1177 						int write, void *data)
1178 {
1179 	if (write) {
1180 		if (USER_HZ < HZ && *lvalp > (LONG_MAX / HZ) * USER_HZ)
1181 			return 1;
1182 		*valp = clock_t_to_jiffies(*negp ? -*lvalp : *lvalp);
1183 	} else {
1184 		int val = *valp;
1185 		unsigned long lval;
1186 		if (val < 0) {
1187 			*negp = true;
1188 			lval = -(unsigned long)val;
1189 		} else {
1190 			*negp = false;
1191 			lval = (unsigned long)val;
1192 		}
1193 		*lvalp = jiffies_to_clock_t(lval);
1194 	}
1195 	return 0;
1196 }
1197 
1198 static int do_proc_dointvec_ms_jiffies_conv(bool *negp, unsigned long *lvalp,
1199 					    int *valp,
1200 					    int write, void *data)
1201 {
1202 	if (write) {
1203 		unsigned long jif = msecs_to_jiffies(*negp ? -*lvalp : *lvalp);
1204 
1205 		if (jif > INT_MAX)
1206 			return 1;
1207 		WRITE_ONCE(*valp, (int)jif);
1208 	} else {
1209 		int val = READ_ONCE(*valp);
1210 		unsigned long lval;
1211 		if (val < 0) {
1212 			*negp = true;
1213 			lval = -(unsigned long)val;
1214 		} else {
1215 			*negp = false;
1216 			lval = (unsigned long)val;
1217 		}
1218 		*lvalp = jiffies_to_msecs(lval);
1219 	}
1220 	return 0;
1221 }
1222 
1223 static int do_proc_dointvec_ms_jiffies_minmax_conv(bool *negp, unsigned long *lvalp,
1224 						int *valp, int write, void *data)
1225 {
1226 	int tmp, ret;
1227 	struct do_proc_dointvec_minmax_conv_param *param = data;
1228 	/*
1229 	 * If writing, first do so via a temporary local int so we can
1230 	 * bounds-check it before touching *valp.
1231 	 */
1232 	int *ip = write ? &tmp : valp;
1233 
1234 	ret = do_proc_dointvec_ms_jiffies_conv(negp, lvalp, ip, write, data);
1235 	if (ret)
1236 		return ret;
1237 
1238 	if (write) {
1239 		if ((param->min && *param->min > tmp) ||
1240 				(param->max && *param->max < tmp))
1241 			return -EINVAL;
1242 		*valp = tmp;
1243 	}
1244 	return 0;
1245 }
1246 
1247 /**
1248  * proc_dointvec_jiffies - read a vector of integers as seconds
1249  * @table: the sysctl table
1250  * @write: %TRUE if this is a write to the sysctl file
1251  * @buffer: the user buffer
1252  * @lenp: the size of the user buffer
1253  * @ppos: file position
1254  *
1255  * Reads/writes up to table->maxlen/sizeof(unsigned int) integer
1256  * values from/to the user buffer, treated as an ASCII string.
1257  * The values read are assumed to be in seconds, and are converted into
1258  * jiffies.
1259  *
1260  * Returns 0 on success.
1261  */
1262 int proc_dointvec_jiffies(const struct ctl_table *table, int write,
1263 			  void *buffer, size_t *lenp, loff_t *ppos)
1264 {
1265     return do_proc_dointvec(table,write,buffer,lenp,ppos,
1266 		    	    do_proc_dointvec_jiffies_conv,NULL);
1267 }
1268 
1269 int proc_dointvec_ms_jiffies_minmax(const struct ctl_table *table, int write,
1270 			  void *buffer, size_t *lenp, loff_t *ppos)
1271 {
1272 	struct do_proc_dointvec_minmax_conv_param param = {
1273 		.min = (int *) table->extra1,
1274 		.max = (int *) table->extra2,
1275 	};
1276 	return do_proc_dointvec(table, write, buffer, lenp, ppos,
1277 			do_proc_dointvec_ms_jiffies_minmax_conv, &param);
1278 }
1279 
1280 /**
1281  * proc_dointvec_userhz_jiffies - read a vector of integers as 1/USER_HZ seconds
1282  * @table: the sysctl table
1283  * @write: %TRUE if this is a write to the sysctl file
1284  * @buffer: the user buffer
1285  * @lenp: the size of the user buffer
1286  * @ppos: pointer to the file position
1287  *
1288  * Reads/writes up to table->maxlen/sizeof(unsigned int) integer
1289  * values from/to the user buffer, treated as an ASCII string.
1290  * The values read are assumed to be in 1/USER_HZ seconds, and
1291  * are converted into jiffies.
1292  *
1293  * Returns 0 on success.
1294  */
1295 int proc_dointvec_userhz_jiffies(const struct ctl_table *table, int write,
1296 				 void *buffer, size_t *lenp, loff_t *ppos)
1297 {
1298 	return do_proc_dointvec(table, write, buffer, lenp, ppos,
1299 				do_proc_dointvec_userhz_jiffies_conv, NULL);
1300 }
1301 
1302 /**
1303  * proc_dointvec_ms_jiffies - read a vector of integers as 1 milliseconds
1304  * @table: the sysctl table
1305  * @write: %TRUE if this is a write to the sysctl file
1306  * @buffer: the user buffer
1307  * @lenp: the size of the user buffer
1308  * @ppos: the current position in the file
1309  *
1310  * Reads/writes up to table->maxlen/sizeof(unsigned int) integer
1311  * values from/to the user buffer, treated as an ASCII string.
1312  * The values read are assumed to be in 1/1000 seconds, and
1313  * are converted into jiffies.
1314  *
1315  * Returns 0 on success.
1316  */
1317 int proc_dointvec_ms_jiffies(const struct ctl_table *table, int write, void *buffer,
1318 		size_t *lenp, loff_t *ppos)
1319 {
1320 	return do_proc_dointvec(table, write, buffer, lenp, ppos,
1321 				do_proc_dointvec_ms_jiffies_conv, NULL);
1322 }
1323 
1324 static int proc_do_cad_pid(const struct ctl_table *table, int write, void *buffer,
1325 		size_t *lenp, loff_t *ppos)
1326 {
1327 	struct pid *new_pid;
1328 	pid_t tmp;
1329 	int r;
1330 
1331 	tmp = pid_vnr(cad_pid);
1332 
1333 	r = __do_proc_dointvec(&tmp, table, write, buffer,
1334 			       lenp, ppos, NULL, NULL);
1335 	if (r || !write)
1336 		return r;
1337 
1338 	new_pid = find_get_pid(tmp);
1339 	if (!new_pid)
1340 		return -ESRCH;
1341 
1342 	put_pid(xchg(&cad_pid, new_pid));
1343 	return 0;
1344 }
1345 
1346 /**
1347  * proc_do_large_bitmap - read/write from/to a large bitmap
1348  * @table: the sysctl table
1349  * @write: %TRUE if this is a write to the sysctl file
1350  * @buffer: the user buffer
1351  * @lenp: the size of the user buffer
1352  * @ppos: file position
1353  *
1354  * The bitmap is stored at table->data and the bitmap length (in bits)
1355  * in table->maxlen.
1356  *
1357  * We use a range comma separated format (e.g. 1,3-4,10-10) so that
1358  * large bitmaps may be represented in a compact manner. Writing into
1359  * the file will clear the bitmap then update it with the given input.
1360  *
1361  * Returns 0 on success.
1362  */
1363 int proc_do_large_bitmap(const struct ctl_table *table, int write,
1364 			 void *buffer, size_t *lenp, loff_t *ppos)
1365 {
1366 	int err = 0;
1367 	size_t left = *lenp;
1368 	unsigned long bitmap_len = table->maxlen;
1369 	unsigned long *bitmap = *(unsigned long **) table->data;
1370 	unsigned long *tmp_bitmap = NULL;
1371 	char tr_a[] = { '-', ',', '\n' }, tr_b[] = { ',', '\n', 0 }, c;
1372 
1373 	if (!bitmap || !bitmap_len || !left || (*ppos && !write)) {
1374 		*lenp = 0;
1375 		return 0;
1376 	}
1377 
1378 	if (write) {
1379 		char *p = buffer;
1380 		size_t skipped = 0;
1381 
1382 		if (left > PAGE_SIZE - 1) {
1383 			left = PAGE_SIZE - 1;
1384 			/* How much of the buffer we'll skip this pass */
1385 			skipped = *lenp - left;
1386 		}
1387 
1388 		tmp_bitmap = bitmap_zalloc(bitmap_len, GFP_KERNEL);
1389 		if (!tmp_bitmap)
1390 			return -ENOMEM;
1391 		proc_skip_char(&p, &left, '\n');
1392 		while (!err && left) {
1393 			unsigned long val_a, val_b;
1394 			bool neg;
1395 			size_t saved_left;
1396 
1397 			/* In case we stop parsing mid-number, we can reset */
1398 			saved_left = left;
1399 			err = proc_get_long(&p, &left, &val_a, &neg, tr_a,
1400 					     sizeof(tr_a), &c);
1401 			/*
1402 			 * If we consumed the entirety of a truncated buffer or
1403 			 * only one char is left (may be a "-"), then stop here,
1404 			 * reset, & come back for more.
1405 			 */
1406 			if ((left <= 1) && skipped) {
1407 				left = saved_left;
1408 				break;
1409 			}
1410 
1411 			if (err)
1412 				break;
1413 			if (val_a >= bitmap_len || neg) {
1414 				err = -EINVAL;
1415 				break;
1416 			}
1417 
1418 			val_b = val_a;
1419 			if (left) {
1420 				p++;
1421 				left--;
1422 			}
1423 
1424 			if (c == '-') {
1425 				err = proc_get_long(&p, &left, &val_b,
1426 						     &neg, tr_b, sizeof(tr_b),
1427 						     &c);
1428 				/*
1429 				 * If we consumed all of a truncated buffer or
1430 				 * then stop here, reset, & come back for more.
1431 				 */
1432 				if (!left && skipped) {
1433 					left = saved_left;
1434 					break;
1435 				}
1436 
1437 				if (err)
1438 					break;
1439 				if (val_b >= bitmap_len || neg ||
1440 				    val_a > val_b) {
1441 					err = -EINVAL;
1442 					break;
1443 				}
1444 				if (left) {
1445 					p++;
1446 					left--;
1447 				}
1448 			}
1449 
1450 			bitmap_set(tmp_bitmap, val_a, val_b - val_a + 1);
1451 			proc_skip_char(&p, &left, '\n');
1452 		}
1453 		left += skipped;
1454 	} else {
1455 		unsigned long bit_a, bit_b = 0;
1456 		bool first = 1;
1457 
1458 		while (left) {
1459 			bit_a = find_next_bit(bitmap, bitmap_len, bit_b);
1460 			if (bit_a >= bitmap_len)
1461 				break;
1462 			bit_b = find_next_zero_bit(bitmap, bitmap_len,
1463 						   bit_a + 1) - 1;
1464 
1465 			if (!first)
1466 				proc_put_char(&buffer, &left, ',');
1467 			proc_put_long(&buffer, &left, bit_a, false);
1468 			if (bit_a != bit_b) {
1469 				proc_put_char(&buffer, &left, '-');
1470 				proc_put_long(&buffer, &left, bit_b, false);
1471 			}
1472 
1473 			first = 0; bit_b++;
1474 		}
1475 		proc_put_char(&buffer, &left, '\n');
1476 	}
1477 
1478 	if (!err) {
1479 		if (write) {
1480 			if (*ppos)
1481 				bitmap_or(bitmap, bitmap, tmp_bitmap, bitmap_len);
1482 			else
1483 				bitmap_copy(bitmap, tmp_bitmap, bitmap_len);
1484 		}
1485 		*lenp -= left;
1486 		*ppos += *lenp;
1487 	}
1488 
1489 	bitmap_free(tmp_bitmap);
1490 	return err;
1491 }
1492 
1493 #else /* CONFIG_PROC_SYSCTL */
1494 
1495 int proc_dostring(const struct ctl_table *table, int write,
1496 		  void *buffer, size_t *lenp, loff_t *ppos)
1497 {
1498 	return -ENOSYS;
1499 }
1500 
1501 int proc_dobool(const struct ctl_table *table, int write,
1502 		void *buffer, size_t *lenp, loff_t *ppos)
1503 {
1504 	return -ENOSYS;
1505 }
1506 
1507 int proc_dointvec(const struct ctl_table *table, int write,
1508 		  void *buffer, size_t *lenp, loff_t *ppos)
1509 {
1510 	return -ENOSYS;
1511 }
1512 
1513 int proc_douintvec(const struct ctl_table *table, int write,
1514 		  void *buffer, size_t *lenp, loff_t *ppos)
1515 {
1516 	return -ENOSYS;
1517 }
1518 
1519 int proc_dointvec_minmax(const struct ctl_table *table, int write,
1520 		    void *buffer, size_t *lenp, loff_t *ppos)
1521 {
1522 	return -ENOSYS;
1523 }
1524 
1525 int proc_douintvec_minmax(const struct ctl_table *table, int write,
1526 			  void *buffer, size_t *lenp, loff_t *ppos)
1527 {
1528 	return -ENOSYS;
1529 }
1530 
1531 int proc_dou8vec_minmax(const struct ctl_table *table, int write,
1532 			void *buffer, size_t *lenp, loff_t *ppos)
1533 {
1534 	return -ENOSYS;
1535 }
1536 
1537 int proc_dointvec_jiffies(const struct ctl_table *table, int write,
1538 		    void *buffer, size_t *lenp, loff_t *ppos)
1539 {
1540 	return -ENOSYS;
1541 }
1542 
1543 int proc_dointvec_ms_jiffies_minmax(const struct ctl_table *table, int write,
1544 				    void *buffer, size_t *lenp, loff_t *ppos)
1545 {
1546 	return -ENOSYS;
1547 }
1548 
1549 int proc_dointvec_userhz_jiffies(const struct ctl_table *table, int write,
1550 		    void *buffer, size_t *lenp, loff_t *ppos)
1551 {
1552 	return -ENOSYS;
1553 }
1554 
1555 int proc_dointvec_ms_jiffies(const struct ctl_table *table, int write,
1556 			     void *buffer, size_t *lenp, loff_t *ppos)
1557 {
1558 	return -ENOSYS;
1559 }
1560 
1561 int proc_doulongvec_minmax(const struct ctl_table *table, int write,
1562 		    void *buffer, size_t *lenp, loff_t *ppos)
1563 {
1564 	return -ENOSYS;
1565 }
1566 
1567 int proc_doulongvec_ms_jiffies_minmax(const struct ctl_table *table, int write,
1568 				      void *buffer, size_t *lenp, loff_t *ppos)
1569 {
1570 	return -ENOSYS;
1571 }
1572 
1573 int proc_do_large_bitmap(const struct ctl_table *table, int write,
1574 			 void *buffer, size_t *lenp, loff_t *ppos)
1575 {
1576 	return -ENOSYS;
1577 }
1578 
1579 #endif /* CONFIG_PROC_SYSCTL */
1580 
1581 #if defined(CONFIG_SYSCTL)
1582 int proc_do_static_key(const struct ctl_table *table, int write,
1583 		       void *buffer, size_t *lenp, loff_t *ppos)
1584 {
1585 	struct static_key *key = (struct static_key *)table->data;
1586 	static DEFINE_MUTEX(static_key_mutex);
1587 	int val, ret;
1588 	struct ctl_table tmp = {
1589 		.data   = &val,
1590 		.maxlen = sizeof(val),
1591 		.mode   = table->mode,
1592 		.extra1 = SYSCTL_ZERO,
1593 		.extra2 = SYSCTL_ONE,
1594 	};
1595 
1596 	if (write && !capable(CAP_SYS_ADMIN))
1597 		return -EPERM;
1598 
1599 	mutex_lock(&static_key_mutex);
1600 	val = static_key_enabled(key);
1601 	ret = proc_dointvec_minmax(&tmp, write, buffer, lenp, ppos);
1602 	if (write && !ret) {
1603 		if (val)
1604 			static_key_enable(key);
1605 		else
1606 			static_key_disable(key);
1607 	}
1608 	mutex_unlock(&static_key_mutex);
1609 	return ret;
1610 }
1611 
1612 static struct ctl_table kern_table[] = {
1613 	{
1614 		.procname	= "panic",
1615 		.data		= &panic_timeout,
1616 		.maxlen		= sizeof(int),
1617 		.mode		= 0644,
1618 		.proc_handler	= proc_dointvec,
1619 	},
1620 #ifdef CONFIG_PROC_SYSCTL
1621 	{
1622 		.procname	= "tainted",
1623 		.maxlen 	= sizeof(long),
1624 		.mode		= 0644,
1625 		.proc_handler	= proc_taint,
1626 	},
1627 	{
1628 		.procname	= "sysctl_writes_strict",
1629 		.data		= &sysctl_writes_strict,
1630 		.maxlen		= sizeof(int),
1631 		.mode		= 0644,
1632 		.proc_handler	= proc_dointvec_minmax,
1633 		.extra1		= SYSCTL_NEG_ONE,
1634 		.extra2		= SYSCTL_ONE,
1635 	},
1636 #endif
1637 	{
1638 		.procname	= "print-fatal-signals",
1639 		.data		= &print_fatal_signals,
1640 		.maxlen		= sizeof(int),
1641 		.mode		= 0644,
1642 		.proc_handler	= proc_dointvec,
1643 	},
1644 #ifdef CONFIG_SPARC
1645 	{
1646 		.procname	= "reboot-cmd",
1647 		.data		= reboot_command,
1648 		.maxlen		= 256,
1649 		.mode		= 0644,
1650 		.proc_handler	= proc_dostring,
1651 	},
1652 	{
1653 		.procname	= "stop-a",
1654 		.data		= &stop_a_enabled,
1655 		.maxlen		= sizeof (int),
1656 		.mode		= 0644,
1657 		.proc_handler	= proc_dointvec,
1658 	},
1659 	{
1660 		.procname	= "scons-poweroff",
1661 		.data		= &scons_pwroff,
1662 		.maxlen		= sizeof (int),
1663 		.mode		= 0644,
1664 		.proc_handler	= proc_dointvec,
1665 	},
1666 #endif
1667 #ifdef CONFIG_SPARC64
1668 	{
1669 		.procname	= "tsb-ratio",
1670 		.data		= &sysctl_tsb_ratio,
1671 		.maxlen		= sizeof (int),
1672 		.mode		= 0644,
1673 		.proc_handler	= proc_dointvec,
1674 	},
1675 #endif
1676 #ifdef CONFIG_PARISC
1677 	{
1678 		.procname	= "soft-power",
1679 		.data		= &pwrsw_enabled,
1680 		.maxlen		= sizeof (int),
1681 		.mode		= 0644,
1682 		.proc_handler	= proc_dointvec,
1683 	},
1684 #endif
1685 #ifdef CONFIG_SYSCTL_ARCH_UNALIGN_ALLOW
1686 	{
1687 		.procname	= "unaligned-trap",
1688 		.data		= &unaligned_enabled,
1689 		.maxlen		= sizeof (int),
1690 		.mode		= 0644,
1691 		.proc_handler	= proc_dointvec,
1692 	},
1693 #endif
1694 #ifdef CONFIG_STACK_TRACER
1695 	{
1696 		.procname	= "stack_tracer_enabled",
1697 		.data		= &stack_tracer_enabled,
1698 		.maxlen		= sizeof(int),
1699 		.mode		= 0644,
1700 		.proc_handler	= stack_trace_sysctl,
1701 	},
1702 #endif
1703 #ifdef CONFIG_TRACING
1704 	{
1705 		.procname	= "ftrace_dump_on_oops",
1706 		.data		= &ftrace_dump_on_oops,
1707 		.maxlen		= MAX_TRACER_SIZE,
1708 		.mode		= 0644,
1709 		.proc_handler	= proc_dostring,
1710 	},
1711 	{
1712 		.procname	= "traceoff_on_warning",
1713 		.data		= &__disable_trace_on_warning,
1714 		.maxlen		= sizeof(__disable_trace_on_warning),
1715 		.mode		= 0644,
1716 		.proc_handler	= proc_dointvec,
1717 	},
1718 	{
1719 		.procname	= "tracepoint_printk",
1720 		.data		= &tracepoint_printk,
1721 		.maxlen		= sizeof(tracepoint_printk),
1722 		.mode		= 0644,
1723 		.proc_handler	= tracepoint_printk_sysctl,
1724 	},
1725 #endif
1726 #ifdef CONFIG_MODULES
1727 	{
1728 		.procname	= "modprobe",
1729 		.data		= &modprobe_path,
1730 		.maxlen		= KMOD_PATH_LEN,
1731 		.mode		= 0644,
1732 		.proc_handler	= proc_dostring,
1733 	},
1734 	{
1735 		.procname	= "modules_disabled",
1736 		.data		= &modules_disabled,
1737 		.maxlen		= sizeof(int),
1738 		.mode		= 0644,
1739 		/* only handle a transition from default "0" to "1" */
1740 		.proc_handler	= proc_dointvec_minmax,
1741 		.extra1		= SYSCTL_ONE,
1742 		.extra2		= SYSCTL_ONE,
1743 	},
1744 #endif
1745 #ifdef CONFIG_UEVENT_HELPER
1746 	{
1747 		.procname	= "hotplug",
1748 		.data		= &uevent_helper,
1749 		.maxlen		= UEVENT_HELPER_PATH_LEN,
1750 		.mode		= 0644,
1751 		.proc_handler	= proc_dostring,
1752 	},
1753 #endif
1754 #ifdef CONFIG_MAGIC_SYSRQ
1755 	{
1756 		.procname	= "sysrq",
1757 		.data		= NULL,
1758 		.maxlen		= sizeof (int),
1759 		.mode		= 0644,
1760 		.proc_handler	= sysrq_sysctl_handler,
1761 	},
1762 #endif
1763 #ifdef CONFIG_PROC_SYSCTL
1764 	{
1765 		.procname	= "cad_pid",
1766 		.data		= NULL,
1767 		.maxlen		= sizeof (int),
1768 		.mode		= 0600,
1769 		.proc_handler	= proc_do_cad_pid,
1770 	},
1771 #endif
1772 	{
1773 		.procname	= "threads-max",
1774 		.data		= NULL,
1775 		.maxlen		= sizeof(int),
1776 		.mode		= 0644,
1777 		.proc_handler	= sysctl_max_threads,
1778 	},
1779 	{
1780 		.procname	= "overflowuid",
1781 		.data		= &overflowuid,
1782 		.maxlen		= sizeof(int),
1783 		.mode		= 0644,
1784 		.proc_handler	= proc_dointvec_minmax,
1785 		.extra1		= SYSCTL_ZERO,
1786 		.extra2		= SYSCTL_MAXOLDUID,
1787 	},
1788 	{
1789 		.procname	= "overflowgid",
1790 		.data		= &overflowgid,
1791 		.maxlen		= sizeof(int),
1792 		.mode		= 0644,
1793 		.proc_handler	= proc_dointvec_minmax,
1794 		.extra1		= SYSCTL_ZERO,
1795 		.extra2		= SYSCTL_MAXOLDUID,
1796 	},
1797 #ifdef CONFIG_S390
1798 	{
1799 		.procname	= "userprocess_debug",
1800 		.data		= &show_unhandled_signals,
1801 		.maxlen		= sizeof(int),
1802 		.mode		= 0644,
1803 		.proc_handler	= proc_dointvec,
1804 	},
1805 #endif
1806 	{
1807 		.procname	= "pid_max",
1808 		.data		= &pid_max,
1809 		.maxlen		= sizeof (int),
1810 		.mode		= 0644,
1811 		.proc_handler	= proc_dointvec_minmax,
1812 		.extra1		= &pid_max_min,
1813 		.extra2		= &pid_max_max,
1814 	},
1815 	{
1816 		.procname	= "panic_on_oops",
1817 		.data		= &panic_on_oops,
1818 		.maxlen		= sizeof(int),
1819 		.mode		= 0644,
1820 		.proc_handler	= proc_dointvec,
1821 	},
1822 	{
1823 		.procname	= "panic_print",
1824 		.data		= &panic_print,
1825 		.maxlen		= sizeof(unsigned long),
1826 		.mode		= 0644,
1827 		.proc_handler	= proc_doulongvec_minmax,
1828 	},
1829 	{
1830 		.procname	= "ngroups_max",
1831 		.data		= (void *)&ngroups_max,
1832 		.maxlen		= sizeof (int),
1833 		.mode		= 0444,
1834 		.proc_handler	= proc_dointvec,
1835 	},
1836 	{
1837 		.procname	= "cap_last_cap",
1838 		.data		= (void *)&cap_last_cap,
1839 		.maxlen		= sizeof(int),
1840 		.mode		= 0444,
1841 		.proc_handler	= proc_dointvec,
1842 	},
1843 #if defined(CONFIG_X86_LOCAL_APIC) && defined(CONFIG_X86)
1844 	{
1845 		.procname       = "unknown_nmi_panic",
1846 		.data           = &unknown_nmi_panic,
1847 		.maxlen         = sizeof (int),
1848 		.mode           = 0644,
1849 		.proc_handler   = proc_dointvec,
1850 	},
1851 #endif
1852 
1853 #if (defined(CONFIG_X86_32) || defined(CONFIG_PARISC)) && \
1854 	defined(CONFIG_DEBUG_STACKOVERFLOW)
1855 	{
1856 		.procname	= "panic_on_stackoverflow",
1857 		.data		= &sysctl_panic_on_stackoverflow,
1858 		.maxlen		= sizeof(int),
1859 		.mode		= 0644,
1860 		.proc_handler	= proc_dointvec,
1861 	},
1862 #endif
1863 #if defined(CONFIG_X86)
1864 	{
1865 		.procname	= "panic_on_unrecovered_nmi",
1866 		.data		= &panic_on_unrecovered_nmi,
1867 		.maxlen		= sizeof(int),
1868 		.mode		= 0644,
1869 		.proc_handler	= proc_dointvec,
1870 	},
1871 	{
1872 		.procname	= "panic_on_io_nmi",
1873 		.data		= &panic_on_io_nmi,
1874 		.maxlen		= sizeof(int),
1875 		.mode		= 0644,
1876 		.proc_handler	= proc_dointvec,
1877 	},
1878 	{
1879 		.procname	= "bootloader_type",
1880 		.data		= &bootloader_type,
1881 		.maxlen		= sizeof (int),
1882 		.mode		= 0444,
1883 		.proc_handler	= proc_dointvec,
1884 	},
1885 	{
1886 		.procname	= "bootloader_version",
1887 		.data		= &bootloader_version,
1888 		.maxlen		= sizeof (int),
1889 		.mode		= 0444,
1890 		.proc_handler	= proc_dointvec,
1891 	},
1892 	{
1893 		.procname	= "io_delay_type",
1894 		.data		= &io_delay_type,
1895 		.maxlen		= sizeof(int),
1896 		.mode		= 0644,
1897 		.proc_handler	= proc_dointvec,
1898 	},
1899 #endif
1900 #if defined(CONFIG_MMU)
1901 	{
1902 		.procname	= "randomize_va_space",
1903 		.data		= &randomize_va_space,
1904 		.maxlen		= sizeof(int),
1905 		.mode		= 0644,
1906 		.proc_handler	= proc_dointvec,
1907 	},
1908 #endif
1909 #if defined(CONFIG_S390) && defined(CONFIG_SMP)
1910 	{
1911 		.procname	= "spin_retry",
1912 		.data		= &spin_retry,
1913 		.maxlen		= sizeof (int),
1914 		.mode		= 0644,
1915 		.proc_handler	= proc_dointvec,
1916 	},
1917 #endif
1918 #if	defined(CONFIG_ACPI_SLEEP) && defined(CONFIG_X86)
1919 	{
1920 		.procname	= "acpi_video_flags",
1921 		.data		= &acpi_realmode_flags,
1922 		.maxlen		= sizeof (unsigned long),
1923 		.mode		= 0644,
1924 		.proc_handler	= proc_doulongvec_minmax,
1925 	},
1926 #endif
1927 #ifdef CONFIG_SYSCTL_ARCH_UNALIGN_NO_WARN
1928 	{
1929 		.procname	= "ignore-unaligned-usertrap",
1930 		.data		= &no_unaligned_warning,
1931 		.maxlen		= sizeof (int),
1932 		.mode		= 0644,
1933 		.proc_handler	= proc_dointvec,
1934 	},
1935 #endif
1936 #ifdef CONFIG_RT_MUTEXES
1937 	{
1938 		.procname	= "max_lock_depth",
1939 		.data		= &max_lock_depth,
1940 		.maxlen		= sizeof(int),
1941 		.mode		= 0644,
1942 		.proc_handler	= proc_dointvec,
1943 	},
1944 #endif
1945 #ifdef CONFIG_PERF_EVENTS
1946 	/*
1947 	 * User-space scripts rely on the existence of this file
1948 	 * as a feature check for perf_events being enabled.
1949 	 *
1950 	 * So it's an ABI, do not remove!
1951 	 */
1952 	{
1953 		.procname	= "perf_event_paranoid",
1954 		.data		= &sysctl_perf_event_paranoid,
1955 		.maxlen		= sizeof(sysctl_perf_event_paranoid),
1956 		.mode		= 0644,
1957 		.proc_handler	= proc_dointvec,
1958 	},
1959 	{
1960 		.procname	= "perf_event_mlock_kb",
1961 		.data		= &sysctl_perf_event_mlock,
1962 		.maxlen		= sizeof(sysctl_perf_event_mlock),
1963 		.mode		= 0644,
1964 		.proc_handler	= proc_dointvec,
1965 	},
1966 	{
1967 		.procname	= "perf_event_max_sample_rate",
1968 		.data		= &sysctl_perf_event_sample_rate,
1969 		.maxlen		= sizeof(sysctl_perf_event_sample_rate),
1970 		.mode		= 0644,
1971 		.proc_handler	= perf_event_max_sample_rate_handler,
1972 		.extra1		= SYSCTL_ONE,
1973 	},
1974 	{
1975 		.procname	= "perf_cpu_time_max_percent",
1976 		.data		= &sysctl_perf_cpu_time_max_percent,
1977 		.maxlen		= sizeof(sysctl_perf_cpu_time_max_percent),
1978 		.mode		= 0644,
1979 		.proc_handler	= perf_cpu_time_max_percent_handler,
1980 		.extra1		= SYSCTL_ZERO,
1981 		.extra2		= SYSCTL_ONE_HUNDRED,
1982 	},
1983 	{
1984 		.procname	= "perf_event_max_stack",
1985 		.data		= &sysctl_perf_event_max_stack,
1986 		.maxlen		= sizeof(sysctl_perf_event_max_stack),
1987 		.mode		= 0644,
1988 		.proc_handler	= perf_event_max_stack_handler,
1989 		.extra1		= SYSCTL_ZERO,
1990 		.extra2		= (void *)&six_hundred_forty_kb,
1991 	},
1992 	{
1993 		.procname	= "perf_event_max_contexts_per_stack",
1994 		.data		= &sysctl_perf_event_max_contexts_per_stack,
1995 		.maxlen		= sizeof(sysctl_perf_event_max_contexts_per_stack),
1996 		.mode		= 0644,
1997 		.proc_handler	= perf_event_max_stack_handler,
1998 		.extra1		= SYSCTL_ZERO,
1999 		.extra2		= SYSCTL_ONE_THOUSAND,
2000 	},
2001 #endif
2002 	{
2003 		.procname	= "panic_on_warn",
2004 		.data		= &panic_on_warn,
2005 		.maxlen		= sizeof(int),
2006 		.mode		= 0644,
2007 		.proc_handler	= proc_dointvec_minmax,
2008 		.extra1		= SYSCTL_ZERO,
2009 		.extra2		= SYSCTL_ONE,
2010 	},
2011 #ifdef CONFIG_TREE_RCU
2012 	{
2013 		.procname	= "panic_on_rcu_stall",
2014 		.data		= &sysctl_panic_on_rcu_stall,
2015 		.maxlen		= sizeof(sysctl_panic_on_rcu_stall),
2016 		.mode		= 0644,
2017 		.proc_handler	= proc_dointvec_minmax,
2018 		.extra1		= SYSCTL_ZERO,
2019 		.extra2		= SYSCTL_ONE,
2020 	},
2021 	{
2022 		.procname	= "max_rcu_stall_to_panic",
2023 		.data		= &sysctl_max_rcu_stall_to_panic,
2024 		.maxlen		= sizeof(sysctl_max_rcu_stall_to_panic),
2025 		.mode		= 0644,
2026 		.proc_handler	= proc_dointvec_minmax,
2027 		.extra1		= SYSCTL_ONE,
2028 		.extra2		= SYSCTL_INT_MAX,
2029 	},
2030 #endif
2031 };
2032 
2033 static struct ctl_table vm_table[] = {
2034 	{
2035 		.procname	= "overcommit_memory",
2036 		.data		= &sysctl_overcommit_memory,
2037 		.maxlen		= sizeof(sysctl_overcommit_memory),
2038 		.mode		= 0644,
2039 		.proc_handler	= overcommit_policy_handler,
2040 		.extra1		= SYSCTL_ZERO,
2041 		.extra2		= SYSCTL_TWO,
2042 	},
2043 	{
2044 		.procname	= "overcommit_ratio",
2045 		.data		= &sysctl_overcommit_ratio,
2046 		.maxlen		= sizeof(sysctl_overcommit_ratio),
2047 		.mode		= 0644,
2048 		.proc_handler	= overcommit_ratio_handler,
2049 	},
2050 	{
2051 		.procname	= "overcommit_kbytes",
2052 		.data		= &sysctl_overcommit_kbytes,
2053 		.maxlen		= sizeof(sysctl_overcommit_kbytes),
2054 		.mode		= 0644,
2055 		.proc_handler	= overcommit_kbytes_handler,
2056 	},
2057 	{
2058 		.procname	= "page-cluster",
2059 		.data		= &page_cluster,
2060 		.maxlen		= sizeof(int),
2061 		.mode		= 0644,
2062 		.proc_handler	= proc_dointvec_minmax,
2063 		.extra1		= SYSCTL_ZERO,
2064 		.extra2		= (void *)&page_cluster_max,
2065 	},
2066 	{
2067 		.procname	= "dirtytime_expire_seconds",
2068 		.data		= &dirtytime_expire_interval,
2069 		.maxlen		= sizeof(dirtytime_expire_interval),
2070 		.mode		= 0644,
2071 		.proc_handler	= dirtytime_interval_handler,
2072 		.extra1		= SYSCTL_ZERO,
2073 	},
2074 	{
2075 		.procname	= "swappiness",
2076 		.data		= &vm_swappiness,
2077 		.maxlen		= sizeof(vm_swappiness),
2078 		.mode		= 0644,
2079 		.proc_handler	= proc_dointvec_minmax,
2080 		.extra1		= SYSCTL_ZERO,
2081 		.extra2		= SYSCTL_TWO_HUNDRED,
2082 	},
2083 #ifdef CONFIG_NUMA
2084 	{
2085 		.procname	= "numa_stat",
2086 		.data		= &sysctl_vm_numa_stat,
2087 		.maxlen		= sizeof(int),
2088 		.mode		= 0644,
2089 		.proc_handler	= sysctl_vm_numa_stat_handler,
2090 		.extra1		= SYSCTL_ZERO,
2091 		.extra2		= SYSCTL_ONE,
2092 	},
2093 #endif
2094 	{
2095 		.procname	= "drop_caches",
2096 		.data		= &sysctl_drop_caches,
2097 		.maxlen		= sizeof(int),
2098 		.mode		= 0200,
2099 		.proc_handler	= drop_caches_sysctl_handler,
2100 		.extra1		= SYSCTL_ONE,
2101 		.extra2		= SYSCTL_FOUR,
2102 	},
2103 	{
2104 		.procname	= "page_lock_unfairness",
2105 		.data		= &sysctl_page_lock_unfairness,
2106 		.maxlen		= sizeof(sysctl_page_lock_unfairness),
2107 		.mode		= 0644,
2108 		.proc_handler	= proc_dointvec_minmax,
2109 		.extra1		= SYSCTL_ZERO,
2110 	},
2111 #ifdef CONFIG_MMU
2112 	{
2113 		.procname	= "max_map_count",
2114 		.data		= &sysctl_max_map_count,
2115 		.maxlen		= sizeof(sysctl_max_map_count),
2116 		.mode		= 0644,
2117 		.proc_handler	= proc_dointvec_minmax,
2118 		.extra1		= SYSCTL_ZERO,
2119 	},
2120 #else
2121 	{
2122 		.procname	= "nr_trim_pages",
2123 		.data		= &sysctl_nr_trim_pages,
2124 		.maxlen		= sizeof(sysctl_nr_trim_pages),
2125 		.mode		= 0644,
2126 		.proc_handler	= proc_dointvec_minmax,
2127 		.extra1		= SYSCTL_ZERO,
2128 	},
2129 #endif
2130 	{
2131 		.procname	= "vfs_cache_pressure",
2132 		.data		= &sysctl_vfs_cache_pressure,
2133 		.maxlen		= sizeof(sysctl_vfs_cache_pressure),
2134 		.mode		= 0644,
2135 		.proc_handler	= proc_dointvec_minmax,
2136 		.extra1		= SYSCTL_ZERO,
2137 	},
2138 #if defined(HAVE_ARCH_PICK_MMAP_LAYOUT) || \
2139     defined(CONFIG_ARCH_WANT_DEFAULT_TOPDOWN_MMAP_LAYOUT)
2140 	{
2141 		.procname	= "legacy_va_layout",
2142 		.data		= &sysctl_legacy_va_layout,
2143 		.maxlen		= sizeof(sysctl_legacy_va_layout),
2144 		.mode		= 0644,
2145 		.proc_handler	= proc_dointvec_minmax,
2146 		.extra1		= SYSCTL_ZERO,
2147 	},
2148 #endif
2149 #ifdef CONFIG_NUMA
2150 	{
2151 		.procname	= "zone_reclaim_mode",
2152 		.data		= &node_reclaim_mode,
2153 		.maxlen		= sizeof(node_reclaim_mode),
2154 		.mode		= 0644,
2155 		.proc_handler	= proc_dointvec_minmax,
2156 		.extra1		= SYSCTL_ZERO,
2157 	},
2158 #endif
2159 #ifdef CONFIG_SMP
2160 	{
2161 		.procname	= "stat_interval",
2162 		.data		= &sysctl_stat_interval,
2163 		.maxlen		= sizeof(sysctl_stat_interval),
2164 		.mode		= 0644,
2165 		.proc_handler	= proc_dointvec_jiffies,
2166 	},
2167 	{
2168 		.procname	= "stat_refresh",
2169 		.data		= NULL,
2170 		.maxlen		= 0,
2171 		.mode		= 0600,
2172 		.proc_handler	= vmstat_refresh,
2173 	},
2174 #endif
2175 #ifdef CONFIG_MMU
2176 	{
2177 		.procname	= "mmap_min_addr",
2178 		.data		= &dac_mmap_min_addr,
2179 		.maxlen		= sizeof(unsigned long),
2180 		.mode		= 0644,
2181 		.proc_handler	= mmap_min_addr_handler,
2182 	},
2183 #endif
2184 #if (defined(CONFIG_X86_32) && !defined(CONFIG_UML))|| \
2185    (defined(CONFIG_SUPERH) && defined(CONFIG_VSYSCALL))
2186 	{
2187 		.procname	= "vdso_enabled",
2188 #ifdef CONFIG_X86_32
2189 		.data		= &vdso32_enabled,
2190 		.maxlen		= sizeof(vdso32_enabled),
2191 #else
2192 		.data		= &vdso_enabled,
2193 		.maxlen		= sizeof(vdso_enabled),
2194 #endif
2195 		.mode		= 0644,
2196 		.proc_handler	= proc_dointvec,
2197 		.extra1		= SYSCTL_ZERO,
2198 	},
2199 #endif
2200 	{
2201 		.procname	= "user_reserve_kbytes",
2202 		.data		= &sysctl_user_reserve_kbytes,
2203 		.maxlen		= sizeof(sysctl_user_reserve_kbytes),
2204 		.mode		= 0644,
2205 		.proc_handler	= proc_doulongvec_minmax,
2206 	},
2207 	{
2208 		.procname	= "admin_reserve_kbytes",
2209 		.data		= &sysctl_admin_reserve_kbytes,
2210 		.maxlen		= sizeof(sysctl_admin_reserve_kbytes),
2211 		.mode		= 0644,
2212 		.proc_handler	= proc_doulongvec_minmax,
2213 	},
2214 #ifdef CONFIG_HAVE_ARCH_MMAP_RND_BITS
2215 	{
2216 		.procname	= "mmap_rnd_bits",
2217 		.data		= &mmap_rnd_bits,
2218 		.maxlen		= sizeof(mmap_rnd_bits),
2219 		.mode		= 0600,
2220 		.proc_handler	= proc_dointvec_minmax,
2221 		.extra1		= (void *)&mmap_rnd_bits_min,
2222 		.extra2		= (void *)&mmap_rnd_bits_max,
2223 	},
2224 #endif
2225 #ifdef CONFIG_HAVE_ARCH_MMAP_RND_COMPAT_BITS
2226 	{
2227 		.procname	= "mmap_rnd_compat_bits",
2228 		.data		= &mmap_rnd_compat_bits,
2229 		.maxlen		= sizeof(mmap_rnd_compat_bits),
2230 		.mode		= 0600,
2231 		.proc_handler	= proc_dointvec_minmax,
2232 		.extra1		= (void *)&mmap_rnd_compat_bits_min,
2233 		.extra2		= (void *)&mmap_rnd_compat_bits_max,
2234 	},
2235 #endif
2236 };
2237 
2238 int __init sysctl_init_bases(void)
2239 {
2240 	register_sysctl_init("kernel", kern_table);
2241 	register_sysctl_init("vm", vm_table);
2242 
2243 	return 0;
2244 }
2245 #endif /* CONFIG_SYSCTL */
2246 /*
2247  * No sense putting this after each symbol definition, twice,
2248  * exception granted :-)
2249  */
2250 EXPORT_SYMBOL(proc_dobool);
2251 EXPORT_SYMBOL(proc_dointvec);
2252 EXPORT_SYMBOL(proc_douintvec);
2253 EXPORT_SYMBOL(proc_dointvec_jiffies);
2254 EXPORT_SYMBOL(proc_dointvec_minmax);
2255 EXPORT_SYMBOL_GPL(proc_douintvec_minmax);
2256 EXPORT_SYMBOL(proc_dointvec_userhz_jiffies);
2257 EXPORT_SYMBOL(proc_dointvec_ms_jiffies);
2258 EXPORT_SYMBOL(proc_dostring);
2259 EXPORT_SYMBOL(proc_doulongvec_minmax);
2260 EXPORT_SYMBOL(proc_doulongvec_ms_jiffies_minmax);
2261 EXPORT_SYMBOL(proc_do_large_bitmap);
2262