1 /* SPDX-License-Identifier: GPL-2.0 */
2 /*
3 * sysctl.h: General linux system control interface
4 *
5 * Begun 24 March 1995, Stephen Tweedie
6 *
7 ****************************************************************
8 ****************************************************************
9 **
10 ** WARNING:
11 ** The values in this file are exported to user space via
12 ** the sysctl() binary interface. Do *NOT* change the
13 ** numbering of any existing values here, and do not change
14 ** any numbers within any one set of values. If you have to
15 ** redefine an existing interface, use a new number for it.
16 ** The kernel will then return -ENOTDIR to any application using
17 ** the old binary interface.
18 **
19 ****************************************************************
20 ****************************************************************
21 */
22 #ifndef _LINUX_SYSCTL_H
23 #define _LINUX_SYSCTL_H
24
25 #include <linux/list.h>
26 #include <linux/rcupdate.h>
27 #include <linux/wait.h>
28 #include <linux/rbtree.h>
29 #include <linux/uidgid.h>
30 #include <uapi/linux/sysctl.h>
31
32 /* For the /proc/sys support */
33 struct completion;
34 struct ctl_table;
35 struct nsproxy;
36 struct ctl_table_root;
37 struct ctl_table_header;
38 struct ctl_dir;
39
40 /* Keep the same order as in fs/proc/proc_sysctl.c */
41 #define SYSCTL_ZERO ((void *)&sysctl_vals[0])
42 #define SYSCTL_ONE ((void *)&sysctl_vals[1])
43 #define SYSCTL_TWO ((void *)&sysctl_vals[2])
44 #define SYSCTL_THREE ((void *)&sysctl_vals[3])
45 #define SYSCTL_FOUR ((void *)&sysctl_vals[4])
46 #define SYSCTL_ONE_HUNDRED ((void *)&sysctl_vals[5])
47 #define SYSCTL_TWO_HUNDRED ((void *)&sysctl_vals[6])
48 #define SYSCTL_ONE_THOUSAND ((void *)&sysctl_vals[7])
49 #define SYSCTL_THREE_THOUSAND ((void *)&sysctl_vals[8])
50 #define SYSCTL_INT_MAX ((void *)&sysctl_vals[9])
51
52 /* this is needed for the proc_dointvec_minmax for [fs_]overflow UID and GID */
53 #define SYSCTL_MAXOLDUID ((void *)&sysctl_vals[10])
54 #define SYSCTL_NEG_ONE ((void *)&sysctl_vals[11])
55
56 extern const int sysctl_vals[];
57
58 #define SYSCTL_LONG_ZERO ((void *)&sysctl_long_vals[0])
59 #define SYSCTL_LONG_ONE ((void *)&sysctl_long_vals[1])
60 #define SYSCTL_LONG_MAX ((void *)&sysctl_long_vals[2])
61
62 #define SYSCTL_CONV_IDENTITY(val) (val)
63 /**
64 *
65 * "dir" originates from read_iter (dir = 0) or write_iter (dir = 1)
66 * in the file_operations struct at proc/proc_sysctl.c. Its value means
67 * one of two things for sysctl:
68 * 1. SYSCTL_USER_TO_KERN(dir) Writing to an internal kernel variable from user
69 * space (dir > 0)
70 * 2. SYSCTL_KERN_TO_USER(dir) Writing to a user space buffer from a kernel
71 * variable (dir == 0).
72 */
73 #define SYSCTL_USER_TO_KERN(dir) (!!(dir))
74 #define SYSCTL_KERN_TO_USER(dir) (!dir)
75
76 #define SYSCTL_USER_TO_KERN_INT_CONV(name, u_ptr_op) \
77 int sysctl_user_to_kern_int_conv##name(const bool *negp, \
78 const unsigned long *u_ptr,\
79 int *k_ptr) \
80 { \
81 unsigned long u = u_ptr_op(*u_ptr); \
82 if (*negp) { \
83 if (u > (unsigned long) INT_MAX + 1) \
84 return -EINVAL; \
85 WRITE_ONCE(*k_ptr, -u); \
86 } else { \
87 if (u > (unsigned long) INT_MAX) \
88 return -EINVAL; \
89 WRITE_ONCE(*k_ptr, u); \
90 } \
91 return 0; \
92 }
93
94 #define SYSCTL_KERN_TO_USER_INT_CONV(name, k_ptr_op) \
95 int sysctl_kern_to_user_int_conv##name(bool *negp, \
96 unsigned long *u_ptr, \
97 const int *k_ptr) \
98 { \
99 int val = READ_ONCE(*k_ptr); \
100 if (val < 0) { \
101 *negp = true; \
102 *u_ptr = -k_ptr_op((unsigned long)val); \
103 } else { \
104 *negp = false; \
105 *u_ptr = k_ptr_op((unsigned long)val); \
106 } \
107 return 0; \
108 }
109
110 /**
111 * To range check on a converted value, use a temp k_ptr
112 * When checking range, value should be within (tbl->extra1, tbl->extra2)
113 */
114 #define SYSCTL_INT_CONV_CUSTOM(name, user_to_kern, kern_to_user, \
115 k_ptr_range_check) \
116 int do_proc_int_conv##name(bool *negp, unsigned long *u_ptr, int *k_ptr,\
117 int dir, const struct ctl_table *tbl) \
118 { \
119 if (SYSCTL_KERN_TO_USER(dir)) \
120 return kern_to_user(negp, u_ptr, k_ptr); \
121 \
122 if (k_ptr_range_check) { \
123 int tmp_k, ret; \
124 if (!tbl) \
125 return -EINVAL; \
126 ret = user_to_kern(negp, u_ptr, &tmp_k); \
127 if (ret) \
128 return ret; \
129 if ((tbl->extra1 && *(int *)tbl->extra1 > tmp_k) || \
130 (tbl->extra2 && *(int *)tbl->extra2 < tmp_k)) \
131 return -EINVAL; \
132 WRITE_ONCE(*k_ptr, tmp_k); \
133 } else \
134 return user_to_kern(negp, u_ptr, k_ptr); \
135 return 0; \
136 }
137
138 #define SYSCTL_USER_TO_KERN_UINT_CONV(name, u_ptr_op) \
139 int sysctl_user_to_kern_uint_conv##name(const unsigned long *u_ptr,\
140 unsigned int *k_ptr) \
141 { \
142 unsigned long u = u_ptr_op(*u_ptr); \
143 if (u > UINT_MAX) \
144 return -EINVAL; \
145 WRITE_ONCE(*k_ptr, u); \
146 return 0; \
147 }
148
149 #define SYSCTL_UINT_CONV_CUSTOM(name, user_to_kern, kern_to_user, \
150 k_ptr_range_check) \
151 int do_proc_uint_conv##name(unsigned long *u_ptr, unsigned int *k_ptr, \
152 int dir, const struct ctl_table *tbl) \
153 { \
154 if (SYSCTL_KERN_TO_USER(dir)) \
155 return kern_to_user(u_ptr, k_ptr); \
156 \
157 if (k_ptr_range_check) { \
158 unsigned int tmp_k; \
159 int ret; \
160 if (!tbl) \
161 return -EINVAL; \
162 ret = user_to_kern(u_ptr, &tmp_k); \
163 if (ret) \
164 return ret; \
165 if ((tbl->extra1 && \
166 *(unsigned int *)tbl->extra1 > tmp_k) || \
167 (tbl->extra2 && \
168 *(unsigned int *)tbl->extra2 < tmp_k)) \
169 return -ERANGE; \
170 WRITE_ONCE(*k_ptr, tmp_k); \
171 } else \
172 return user_to_kern(u_ptr, k_ptr); \
173 return 0; \
174 }
175
176
177 extern const unsigned long sysctl_long_vals[];
178
179 typedef int proc_handler(const struct ctl_table *ctl, int write, void *buffer,
180 size_t *lenp, loff_t *ppos);
181
182 int proc_dostring(const struct ctl_table *, int, void *, size_t *, loff_t *);
183 int proc_dobool(const struct ctl_table *table, int write, void *buffer,
184 size_t *lenp, loff_t *ppos);
185 int proc_dointvec(const struct ctl_table *, int, void *, size_t *, loff_t *);
186 int proc_dointvec_minmax(const struct ctl_table *table, int dir, void *buffer,
187 size_t *lenp, loff_t *ppos);
188 int proc_dointvec_conv(const struct ctl_table *table, int dir, void *buffer,
189 size_t *lenp, loff_t *ppos,
190 int (*conv)(bool *negp, unsigned long *u_ptr, int *k_ptr,
191 int dir, const struct ctl_table *table));
192 int proc_douintvec(const struct ctl_table *, int, void *, size_t *, loff_t *);
193 int proc_douintvec_minmax(const struct ctl_table *table, int write, void *buffer,
194 size_t *lenp, loff_t *ppos);
195 int proc_douintvec_conv(const struct ctl_table *table, int write, void *buffer,
196 size_t *lenp, loff_t *ppos,
197 int (*conv)(unsigned long *lvalp, unsigned int *valp,
198 int write, const struct ctl_table *table));
199
200 int proc_dou8vec_minmax(const struct ctl_table *table, int write, void *buffer,
201 size_t *lenp, loff_t *ppos);
202 int proc_doulongvec_minmax(const struct ctl_table *, int, void *, size_t *, loff_t *);
203 int proc_doulongvec_minmax_conv(const struct ctl_table *table, int dir,
204 void *buffer, size_t *lenp, loff_t *ppos,
205 unsigned long convmul, unsigned long convdiv);
206 int proc_do_large_bitmap(const struct ctl_table *, int, void *, size_t *, loff_t *);
207 int proc_do_static_key(const struct ctl_table *table, int write, void *buffer,
208 size_t *lenp, loff_t *ppos);
209 int sysctl_kern_to_user_uint_conv(unsigned long *u_ptr, const unsigned int *k_ptr);
210
211 /*
212 * Register a set of sysctl names by calling register_sysctl
213 * with an initialised array of struct ctl_table's.
214 *
215 * sysctl names can be mirrored automatically under /proc/sys. The
216 * procname supplied controls /proc naming.
217 *
218 * The table's mode will be honoured for proc-fs access.
219 *
220 * Leaf nodes in the sysctl tree will be represented by a single file
221 * under /proc; non-leaf nodes will be represented by directories. A
222 * null procname disables /proc mirroring at this node.
223 *
224 * The data and maxlen fields of the ctl_table
225 * struct enable minimal validation of the values being written to be
226 * performed, and the mode field allows minimal authentication.
227 *
228 * There must be a proc_handler routine for any terminal nodes
229 * mirrored under /proc/sys (non-terminals are handled by a built-in
230 * directory handler). Several default handlers are available to
231 * cover common cases.
232 */
233
234 /* Support for userspace poll() to watch for changes */
235 struct ctl_table_poll {
236 atomic_t event;
237 wait_queue_head_t wait;
238 };
239
proc_sys_poll_event(struct ctl_table_poll * poll)240 static inline void *proc_sys_poll_event(struct ctl_table_poll *poll)
241 {
242 return (void *)(unsigned long)atomic_read(&poll->event);
243 }
244
245 #define __CTL_TABLE_POLL_INITIALIZER(name) { \
246 .event = ATOMIC_INIT(0), \
247 .wait = __WAIT_QUEUE_HEAD_INITIALIZER(name.wait) }
248
249 #define DEFINE_CTL_TABLE_POLL(name) \
250 struct ctl_table_poll name = __CTL_TABLE_POLL_INITIALIZER(name)
251
252 /* A sysctl table is an array of struct ctl_table: */
253 struct ctl_table {
254 const char *procname; /* Text ID for /proc/sys */
255 void *data;
256 int maxlen;
257 umode_t mode;
258 proc_handler *proc_handler; /* Callback for text formatting */
259 struct ctl_table_poll *poll;
260 void *extra1;
261 void *extra2;
262 } __randomize_layout;
263
264 struct ctl_node {
265 struct rb_node node;
266 struct ctl_table_header *header;
267 };
268
269 /**
270 * struct ctl_table_header - maintains dynamic lists of struct ctl_table trees
271 * @ctl_table: pointer to the first element in ctl_table array
272 * @ctl_table_size: number of elements pointed by @ctl_table
273 * @used: The entry will never be touched when equal to 0.
274 * @count: Upped every time something is added to @inodes and downed every time
275 * something is removed from inodes
276 * @nreg: When nreg drops to 0 the ctl_table_header will be unregistered.
277 * @rcu: Delays the freeing of the inode. Introduced with "unfuck proc_sysctl ->d_compare()"
278 *
279 * @type: Enumeration to differentiate between ctl target types
280 * @type.SYSCTL_TABLE_TYPE_DEFAULT: ctl target with no special considerations
281 * @type.SYSCTL_TABLE_TYPE_PERMANENTLY_EMPTY: Identifies a permanently empty dir
282 * target to serve as a mount point
283 */
284 struct ctl_table_header {
285 union {
286 struct {
287 const struct ctl_table *ctl_table;
288 int ctl_table_size;
289 int used;
290 int count;
291 int nreg;
292 };
293 struct rcu_head rcu;
294 };
295 struct completion *unregistering;
296 const struct ctl_table *ctl_table_arg;
297 struct ctl_table_root *root;
298 struct ctl_table_set *set;
299 struct ctl_dir *parent;
300 struct ctl_node *node;
301 struct hlist_head inodes; /* head for proc_inode->sysctl_inodes */
302 enum {
303 SYSCTL_TABLE_TYPE_DEFAULT,
304 SYSCTL_TABLE_TYPE_PERMANENTLY_EMPTY,
305 } type;
306 };
307
308 struct ctl_dir {
309 /* Header must be at the start of ctl_dir */
310 struct ctl_table_header header;
311 struct rb_root root;
312 };
313
314 struct ctl_table_set {
315 int (*is_seen)(struct ctl_table_set *);
316 struct ctl_dir dir;
317 };
318
319 struct ctl_table_root {
320 struct ctl_table_set default_set;
321 struct ctl_table_set *(*lookup)(struct ctl_table_root *root);
322 void (*set_ownership)(struct ctl_table_header *head,
323 kuid_t *uid, kgid_t *gid);
324 int (*permissions)(struct ctl_table_header *head, const struct ctl_table *table);
325 };
326
327 #define register_sysctl(path, table) \
328 register_sysctl_sz(path, table, ARRAY_SIZE(table))
329
330 #ifdef CONFIG_SYSCTL
331
332 void proc_sys_poll_notify(struct ctl_table_poll *poll);
333
334 extern void setup_sysctl_set(struct ctl_table_set *p,
335 struct ctl_table_root *root,
336 int (*is_seen)(struct ctl_table_set *));
337 extern void retire_sysctl_set(struct ctl_table_set *set);
338
339 struct ctl_table_header *__register_sysctl_table(
340 struct ctl_table_set *set,
341 const char *path, const struct ctl_table *table, size_t table_size);
342 struct ctl_table_header *register_sysctl_sz(const char *path, const struct ctl_table *table,
343 size_t table_size);
344 void unregister_sysctl_table(struct ctl_table_header * table);
345
346 extern int sysctl_init_bases(void);
347 extern void __register_sysctl_init(const char *path, const struct ctl_table *table,
348 const char *table_name, size_t table_size);
349 #define register_sysctl_init(path, table) \
350 __register_sysctl_init(path, table, #table, ARRAY_SIZE(table))
351 extern struct ctl_table_header *register_sysctl_mount_point(const char *path);
352
353 void do_sysctl_args(void);
354 bool sysctl_is_alias(char *param);
355
356 extern int unaligned_enabled;
357 extern int no_unaligned_warning;
358
359 #else /* CONFIG_SYSCTL */
360
register_sysctl_init(const char * path,const struct ctl_table * table)361 static inline void register_sysctl_init(const char *path, const struct ctl_table *table)
362 {
363 }
364
register_sysctl_mount_point(const char * path)365 static inline struct ctl_table_header *register_sysctl_mount_point(const char *path)
366 {
367 return NULL;
368 }
369
register_sysctl_sz(const char * path,const struct ctl_table * table,size_t table_size)370 static inline struct ctl_table_header *register_sysctl_sz(const char *path,
371 const struct ctl_table *table,
372 size_t table_size)
373 {
374 return NULL;
375 }
376
unregister_sysctl_table(struct ctl_table_header * table)377 static inline void unregister_sysctl_table(struct ctl_table_header * table)
378 {
379 }
380
setup_sysctl_set(struct ctl_table_set * p,struct ctl_table_root * root,int (* is_seen)(struct ctl_table_set *))381 static inline void setup_sysctl_set(struct ctl_table_set *p,
382 struct ctl_table_root *root,
383 int (*is_seen)(struct ctl_table_set *))
384 {
385 }
386
do_sysctl_args(void)387 static inline void do_sysctl_args(void)
388 {
389 }
390
sysctl_is_alias(char * param)391 static inline bool sysctl_is_alias(char *param)
392 {
393 return false;
394 }
395 #endif /* CONFIG_SYSCTL */
396
397 #endif /* _LINUX_SYSCTL_H */
398