blob: f92d5dd578a4a5289aab9e29dcd37b09ee775c46 [file] [log] [blame]
Eric W. Biederman77b14db2007-02-14 00:34:12 -08001/*
2 * /proc/sys support
3 */
Alexey Dobriyan1e0edd32008-10-17 05:07:44 +04004#include <linux/init.h>
Eric W. Biederman77b14db2007-02-14 00:34:12 -08005#include <linux/sysctl.h>
Lucas De Marchif1ecf062011-11-02 13:39:22 -07006#include <linux/poll.h>
Eric W. Biederman77b14db2007-02-14 00:34:12 -08007#include <linux/proc_fs.h>
Andrew Morton87ebdc02013-02-27 17:03:16 -08008#include <linux/printk.h>
Eric W. Biederman77b14db2007-02-14 00:34:12 -08009#include <linux/security.h>
Al Viro40401532012-02-13 03:58:52 +000010#include <linux/sched.h>
Nick Piggin34286d62011-01-07 17:49:57 +110011#include <linux/namei.h>
Al Viro40401532012-02-13 03:58:52 +000012#include <linux/mm.h>
Eric W. Biederman1f87f0b2012-01-06 04:07:15 -080013#include <linux/module.h>
Eric W. Biederman77b14db2007-02-14 00:34:12 -080014#include "internal.h"
15
Al Virod72f71e2009-02-20 05:58:47 +000016static const struct dentry_operations proc_sys_dentry_operations;
Eric W. Biederman77b14db2007-02-14 00:34:12 -080017static const struct file_operations proc_sys_file_operations;
Jan Engelhardt03a44822008-02-08 04:21:19 -080018static const struct inode_operations proc_sys_inode_operations;
Al Viro9043476f2008-07-15 08:54:06 -040019static const struct file_operations proc_sys_dir_file_operations;
20static const struct inode_operations proc_sys_dir_operations;
Eric W. Biederman77b14db2007-02-14 00:34:12 -080021
Lucas De Marchif1ecf062011-11-02 13:39:22 -070022void proc_sys_poll_notify(struct ctl_table_poll *poll)
23{
24 if (!poll)
25 return;
26
27 atomic_inc(&poll->event);
28 wake_up_interruptible(&poll->wait);
29}
30
Eric W. Biedermana1945582012-01-21 17:51:48 -080031static struct ctl_table root_table[] = {
32 {
33 .procname = "",
Eric W. Biederman7ec66d02011-12-29 08:24:29 -080034 .mode = S_IFDIR|S_IRUGO|S_IXUGO,
Eric W. Biedermana1945582012-01-21 17:51:48 -080035 },
36 { }
37};
Eric W. Biederman0e47c992012-01-07 23:24:30 -080038static struct ctl_table_root sysctl_table_root = {
Eric W. Biederman0e47c992012-01-07 23:24:30 -080039 .default_set.dir.header = {
Eric W. Biederman7ec66d02011-12-29 08:24:29 -080040 {{.count = 1,
41 .nreg = 1,
Eric W. Biedermanac13ac62012-01-09 17:24:30 -080042 .ctl_table = root_table }},
Eric W. Biederman0e47c992012-01-07 23:24:30 -080043 .ctl_table_arg = root_table,
Eric W. Biederman7ec66d02011-12-29 08:24:29 -080044 .root = &sysctl_table_root,
45 .set = &sysctl_table_root.default_set,
46 },
Eric W. Biederman1f87f0b2012-01-06 04:07:15 -080047};
Eric W. Biederman1f87f0b2012-01-06 04:07:15 -080048
49static DEFINE_SPINLOCK(sysctl_lock);
50
Eric W. Biederman7ec66d02011-12-29 08:24:29 -080051static void drop_sysctl_table(struct ctl_table_header *header);
Eric W. Biederman0e47c992012-01-07 23:24:30 -080052static int sysctl_follow_link(struct ctl_table_header **phead,
53 struct ctl_table **pentry, struct nsproxy *namespaces);
54static int insert_links(struct ctl_table_header *head);
55static void put_links(struct ctl_table_header *header);
Eric W. Biederman7ec66d02011-12-29 08:24:29 -080056
Eric W. Biederman69801282012-01-21 20:09:45 -080057static void sysctl_print_dir(struct ctl_dir *dir)
58{
59 if (dir->header.parent)
60 sysctl_print_dir(dir->header.parent);
Andrew Morton87ebdc02013-02-27 17:03:16 -080061 pr_cont("%s/", dir->header.ctl_table[0].procname);
Eric W. Biederman69801282012-01-21 20:09:45 -080062}
63
Eric W. Biederman076c3ee2012-01-09 21:42:02 -080064static int namecmp(const char *name1, int len1, const char *name2, int len2)
65{
66 int minlen;
67 int cmp;
68
69 minlen = len1;
70 if (minlen > len2)
71 minlen = len2;
72
73 cmp = memcmp(name1, name2, minlen);
74 if (cmp == 0)
75 cmp = len1 - len2;
76 return cmp;
77}
78
Eric W. Biederman60f126d2012-01-30 21:23:52 -080079/* Called under sysctl_lock */
Eric W. Biederman076c3ee2012-01-09 21:42:02 -080080static struct ctl_table *find_entry(struct ctl_table_header **phead,
Eric W. Biederman0e47c992012-01-07 23:24:30 -080081 struct ctl_dir *dir, const char *name, int namelen)
Eric W. Biederman076c3ee2012-01-09 21:42:02 -080082{
83 struct ctl_table_header *head;
84 struct ctl_table *entry;
Eric W. Biedermanac13ac62012-01-09 17:24:30 -080085 struct rb_node *node = dir->root.rb_node;
Eric W. Biederman076c3ee2012-01-09 21:42:02 -080086
Eric W. Biedermanac13ac62012-01-09 17:24:30 -080087 while (node)
88 {
89 struct ctl_node *ctl_node;
90 const char *procname;
91 int cmp;
92
93 ctl_node = rb_entry(node, struct ctl_node, node);
94 head = ctl_node->header;
95 entry = &head->ctl_table[ctl_node - head->node];
96 procname = entry->procname;
97
98 cmp = namecmp(name, namelen, procname, strlen(procname));
99 if (cmp < 0)
100 node = node->rb_left;
101 else if (cmp > 0)
102 node = node->rb_right;
103 else {
104 *phead = head;
105 return entry;
Eric W. Biederman076c3ee2012-01-09 21:42:02 -0800106 }
107 }
108 return NULL;
109}
110
Eric W. Biedermanac13ac62012-01-09 17:24:30 -0800111static int insert_entry(struct ctl_table_header *head, struct ctl_table *entry)
112{
113 struct rb_node *node = &head->node[entry - head->ctl_table].node;
114 struct rb_node **p = &head->parent->root.rb_node;
115 struct rb_node *parent = NULL;
116 const char *name = entry->procname;
117 int namelen = strlen(name);
118
119 while (*p) {
120 struct ctl_table_header *parent_head;
121 struct ctl_table *parent_entry;
122 struct ctl_node *parent_node;
123 const char *parent_name;
124 int cmp;
125
126 parent = *p;
127 parent_node = rb_entry(parent, struct ctl_node, node);
128 parent_head = parent_node->header;
129 parent_entry = &parent_head->ctl_table[parent_node - parent_head->node];
130 parent_name = parent_entry->procname;
131
132 cmp = namecmp(name, namelen, parent_name, strlen(parent_name));
133 if (cmp < 0)
134 p = &(*p)->rb_left;
135 else if (cmp > 0)
136 p = &(*p)->rb_right;
137 else {
Andrew Morton87ebdc02013-02-27 17:03:16 -0800138 pr_err("sysctl duplicate entry: ");
Eric W. Biedermanac13ac62012-01-09 17:24:30 -0800139 sysctl_print_dir(head->parent);
Andrew Morton87ebdc02013-02-27 17:03:16 -0800140 pr_cont("/%s\n", entry->procname);
Eric W. Biedermanac13ac62012-01-09 17:24:30 -0800141 return -EEXIST;
142 }
143 }
144
145 rb_link_node(node, parent, p);
Michel Lespinasseea5272f2012-10-08 16:30:35 -0700146 rb_insert_color(node, &head->parent->root);
Eric W. Biedermanac13ac62012-01-09 17:24:30 -0800147 return 0;
148}
149
150static void erase_entry(struct ctl_table_header *head, struct ctl_table *entry)
151{
152 struct rb_node *node = &head->node[entry - head->ctl_table].node;
153
154 rb_erase(node, &head->parent->root);
155}
156
Eric W. Biedermane0d04522012-01-09 22:36:41 -0800157static void init_header(struct ctl_table_header *head,
158 struct ctl_table_root *root, struct ctl_table_set *set,
Eric W. Biedermanac13ac62012-01-09 17:24:30 -0800159 struct ctl_node *node, struct ctl_table *table)
Eric W. Biedermane0d04522012-01-09 22:36:41 -0800160{
Eric W. Biederman7ec66d02011-12-29 08:24:29 -0800161 head->ctl_table = table;
Eric W. Biedermane0d04522012-01-09 22:36:41 -0800162 head->ctl_table_arg = table;
Eric W. Biedermane0d04522012-01-09 22:36:41 -0800163 head->used = 0;
164 head->count = 1;
165 head->nreg = 1;
166 head->unregistering = NULL;
167 head->root = root;
168 head->set = set;
169 head->parent = NULL;
Eric W. Biedermanac13ac62012-01-09 17:24:30 -0800170 head->node = node;
171 if (node) {
172 struct ctl_table *entry;
Michel Lespinasse4c199a92012-10-08 16:30:32 -0700173 for (entry = table; entry->procname; entry++, node++)
Eric W. Biedermanac13ac62012-01-09 17:24:30 -0800174 node->header = head;
Eric W. Biedermanac13ac62012-01-09 17:24:30 -0800175 }
Eric W. Biedermane0d04522012-01-09 22:36:41 -0800176}
177
Eric W. Biederman8425d6a2012-01-09 17:35:01 -0800178static void erase_header(struct ctl_table_header *head)
179{
Eric W. Biedermanac13ac62012-01-09 17:24:30 -0800180 struct ctl_table *entry;
181 for (entry = head->ctl_table; entry->procname; entry++)
182 erase_entry(head, entry);
Eric W. Biederman8425d6a2012-01-09 17:35:01 -0800183}
184
Eric W. Biederman0e47c992012-01-07 23:24:30 -0800185static int insert_header(struct ctl_dir *dir, struct ctl_table_header *header)
Eric W. Biederman8425d6a2012-01-09 17:35:01 -0800186{
Eric W. Biedermanac13ac62012-01-09 17:24:30 -0800187 struct ctl_table *entry;
Eric W. Biederman0e47c992012-01-07 23:24:30 -0800188 int err;
189
190 dir->header.nreg++;
Eric W. Biederman7ec66d02011-12-29 08:24:29 -0800191 header->parent = dir;
Eric W. Biederman0e47c992012-01-07 23:24:30 -0800192 err = insert_links(header);
193 if (err)
194 goto fail_links;
Eric W. Biedermanac13ac62012-01-09 17:24:30 -0800195 for (entry = header->ctl_table; entry->procname; entry++) {
196 err = insert_entry(header, entry);
197 if (err)
198 goto fail;
199 }
Eric W. Biederman0e47c992012-01-07 23:24:30 -0800200 return 0;
Eric W. Biedermanac13ac62012-01-09 17:24:30 -0800201fail:
202 erase_header(header);
203 put_links(header);
Eric W. Biederman0e47c992012-01-07 23:24:30 -0800204fail_links:
205 header->parent = NULL;
206 drop_sysctl_table(&dir->header);
207 return err;
Eric W. Biederman8425d6a2012-01-09 17:35:01 -0800208}
209
Eric W. Biederman1f87f0b2012-01-06 04:07:15 -0800210/* called under sysctl_lock */
211static int use_table(struct ctl_table_header *p)
212{
213 if (unlikely(p->unregistering))
214 return 0;
215 p->used++;
216 return 1;
217}
218
219/* called under sysctl_lock */
220static void unuse_table(struct ctl_table_header *p)
221{
222 if (!--p->used)
223 if (unlikely(p->unregistering))
224 complete(p->unregistering);
225}
226
227/* called under sysctl_lock, will reacquire if has to wait */
228static void start_unregistering(struct ctl_table_header *p)
229{
230 /*
231 * if p->used is 0, nobody will ever touch that entry again;
232 * we'll eliminate all paths to it before dropping sysctl_lock
233 */
234 if (unlikely(p->used)) {
235 struct completion wait;
236 init_completion(&wait);
237 p->unregistering = &wait;
238 spin_unlock(&sysctl_lock);
239 wait_for_completion(&wait);
240 spin_lock(&sysctl_lock);
241 } else {
242 /* anything non-NULL; we'll never dereference it */
243 p->unregistering = ERR_PTR(-EINVAL);
244 }
245 /*
246 * do not remove from the list until nobody holds it; walking the
247 * list in do_sysctl() relies on that.
248 */
Eric W. Biederman8425d6a2012-01-09 17:35:01 -0800249 erase_header(p);
Eric W. Biederman1f87f0b2012-01-06 04:07:15 -0800250}
251
252static void sysctl_head_get(struct ctl_table_header *head)
253{
254 spin_lock(&sysctl_lock);
255 head->count++;
256 spin_unlock(&sysctl_lock);
257}
258
259void sysctl_head_put(struct ctl_table_header *head)
260{
261 spin_lock(&sysctl_lock);
262 if (!--head->count)
263 kfree_rcu(head, rcu);
264 spin_unlock(&sysctl_lock);
265}
266
267static struct ctl_table_header *sysctl_head_grab(struct ctl_table_header *head)
268{
Prasad Joshiab4a1f22012-10-04 17:15:45 -0700269 BUG_ON(!head);
Eric W. Biederman1f87f0b2012-01-06 04:07:15 -0800270 spin_lock(&sysctl_lock);
271 if (!use_table(head))
272 head = ERR_PTR(-ENOENT);
273 spin_unlock(&sysctl_lock);
274 return head;
275}
276
277static void sysctl_head_finish(struct ctl_table_header *head)
278{
279 if (!head)
280 return;
281 spin_lock(&sysctl_lock);
282 unuse_table(head);
283 spin_unlock(&sysctl_lock);
284}
285
286static struct ctl_table_set *
287lookup_header_set(struct ctl_table_root *root, struct nsproxy *namespaces)
288{
289 struct ctl_table_set *set = &root->default_set;
290 if (root->lookup)
291 set = root->lookup(root, namespaces);
292 return set;
293}
294
Eric W. Biederman076c3ee2012-01-09 21:42:02 -0800295static struct ctl_table *lookup_entry(struct ctl_table_header **phead,
Eric W. Biederman7ec66d02011-12-29 08:24:29 -0800296 struct ctl_dir *dir,
Eric W. Biederman076c3ee2012-01-09 21:42:02 -0800297 const char *name, int namelen)
298{
299 struct ctl_table_header *head;
300 struct ctl_table *entry;
Eric W. Biederman076c3ee2012-01-09 21:42:02 -0800301
302 spin_lock(&sysctl_lock);
Eric W. Biederman0e47c992012-01-07 23:24:30 -0800303 entry = find_entry(&head, dir, name, namelen);
304 if (entry && use_table(head))
305 *phead = head;
306 else
307 entry = NULL;
Eric W. Biederman076c3ee2012-01-09 21:42:02 -0800308 spin_unlock(&sysctl_lock);
309 return entry;
310}
311
Eric W. Biedermanac13ac62012-01-09 17:24:30 -0800312static struct ctl_node *first_usable_entry(struct rb_node *node)
Eric W. Biederman1f87f0b2012-01-06 04:07:15 -0800313{
Eric W. Biedermanac13ac62012-01-09 17:24:30 -0800314 struct ctl_node *ctl_node;
Eric W. Biederman1f87f0b2012-01-06 04:07:15 -0800315
Eric W. Biedermanac13ac62012-01-09 17:24:30 -0800316 for (;node; node = rb_next(node)) {
317 ctl_node = rb_entry(node, struct ctl_node, node);
318 if (use_table(ctl_node->header))
319 return ctl_node;
Eric W. Biederman1f87f0b2012-01-06 04:07:15 -0800320 }
Eric W. Biederman1f87f0b2012-01-06 04:07:15 -0800321 return NULL;
322}
323
Eric W. Biederman7ec66d02011-12-29 08:24:29 -0800324static void first_entry(struct ctl_dir *dir,
Eric W. Biederman6a75ce12012-01-18 03:15:51 -0800325 struct ctl_table_header **phead, struct ctl_table **pentry)
Eric W. Biederman1f87f0b2012-01-06 04:07:15 -0800326{
Eric W. Biedermanac13ac62012-01-09 17:24:30 -0800327 struct ctl_table_header *head = NULL;
Eric W. Biederman7ec66d02011-12-29 08:24:29 -0800328 struct ctl_table *entry = NULL;
Eric W. Biedermanac13ac62012-01-09 17:24:30 -0800329 struct ctl_node *ctl_node;
Eric W. Biederman6a75ce12012-01-18 03:15:51 -0800330
331 spin_lock(&sysctl_lock);
Eric W. Biedermanac13ac62012-01-09 17:24:30 -0800332 ctl_node = first_usable_entry(rb_first(&dir->root));
Eric W. Biederman6a75ce12012-01-18 03:15:51 -0800333 spin_unlock(&sysctl_lock);
Eric W. Biedermanac13ac62012-01-09 17:24:30 -0800334 if (ctl_node) {
335 head = ctl_node->header;
336 entry = &head->ctl_table[ctl_node - head->node];
337 }
Eric W. Biederman6a75ce12012-01-18 03:15:51 -0800338 *phead = head;
339 *pentry = entry;
340}
341
Eric W. Biederman7ec66d02011-12-29 08:24:29 -0800342static void next_entry(struct ctl_table_header **phead, struct ctl_table **pentry)
Eric W. Biederman6a75ce12012-01-18 03:15:51 -0800343{
344 struct ctl_table_header *head = *phead;
345 struct ctl_table *entry = *pentry;
Eric W. Biedermanac13ac62012-01-09 17:24:30 -0800346 struct ctl_node *ctl_node = &head->node[entry - head->ctl_table];
Eric W. Biederman6a75ce12012-01-18 03:15:51 -0800347
Eric W. Biedermanac13ac62012-01-09 17:24:30 -0800348 spin_lock(&sysctl_lock);
349 unuse_table(head);
350
351 ctl_node = first_usable_entry(rb_next(&ctl_node->node));
352 spin_unlock(&sysctl_lock);
353 head = NULL;
354 if (ctl_node) {
355 head = ctl_node->header;
356 entry = &head->ctl_table[ctl_node - head->node];
Eric W. Biederman6a75ce12012-01-18 03:15:51 -0800357 }
358 *phead = head;
359 *pentry = entry;
Eric W. Biederman1f87f0b2012-01-06 04:07:15 -0800360}
361
362void register_sysctl_root(struct ctl_table_root *root)
363{
Eric W. Biederman1f87f0b2012-01-06 04:07:15 -0800364}
365
366/*
367 * sysctl_perm does NOT grant the superuser all rights automatically, because
368 * some sysctl variables are readonly even to root.
369 */
370
371static int test_perm(int mode, int op)
372{
Eric W. Biederman091bd3e2012-02-13 18:02:50 -0800373 if (uid_eq(current_euid(), GLOBAL_ROOT_UID))
Eric W. Biederman1f87f0b2012-01-06 04:07:15 -0800374 mode >>= 6;
Eric W. Biederman091bd3e2012-02-13 18:02:50 -0800375 else if (in_egroup_p(GLOBAL_ROOT_GID))
Eric W. Biederman1f87f0b2012-01-06 04:07:15 -0800376 mode >>= 3;
377 if ((op & ~mode & (MAY_READ|MAY_WRITE|MAY_EXEC)) == 0)
378 return 0;
379 return -EACCES;
380}
381
Eric W. Biederman73f7ef42012-11-16 03:02:58 +0000382static int sysctl_perm(struct ctl_table_header *head, struct ctl_table *table, int op)
Eric W. Biederman1f87f0b2012-01-06 04:07:15 -0800383{
Eric W. Biederman73f7ef42012-11-16 03:02:58 +0000384 struct ctl_table_root *root = head->root;
Eric W. Biederman1f87f0b2012-01-06 04:07:15 -0800385 int mode;
386
387 if (root->permissions)
Eric W. Biederman73f7ef42012-11-16 03:02:58 +0000388 mode = root->permissions(head, table);
Eric W. Biederman1f87f0b2012-01-06 04:07:15 -0800389 else
390 mode = table->mode;
391
392 return test_perm(mode, op);
393}
394
Al Viro9043476f2008-07-15 08:54:06 -0400395static struct inode *proc_sys_make_inode(struct super_block *sb,
396 struct ctl_table_header *head, struct ctl_table *table)
Eric W. Biederman77b14db2007-02-14 00:34:12 -0800397{
398 struct inode *inode;
Al Viro9043476f2008-07-15 08:54:06 -0400399 struct proc_inode *ei;
Eric W. Biederman77b14db2007-02-14 00:34:12 -0800400
Al Viro9043476f2008-07-15 08:54:06 -0400401 inode = new_inode(sb);
Eric W. Biederman77b14db2007-02-14 00:34:12 -0800402 if (!inode)
403 goto out;
404
Christoph Hellwig85fe4022010-10-23 11:19:54 -0400405 inode->i_ino = get_next_ino();
406
Al Viro9043476f2008-07-15 08:54:06 -0400407 sysctl_head_get(head);
Eric W. Biederman77b14db2007-02-14 00:34:12 -0800408 ei = PROC_I(inode);
Al Viro9043476f2008-07-15 08:54:06 -0400409 ei->sysctl = head;
410 ei->sysctl_entry = table;
411
Eric W. Biederman77b14db2007-02-14 00:34:12 -0800412 inode->i_mtime = inode->i_atime = inode->i_ctime = CURRENT_TIME;
Al Viro9043476f2008-07-15 08:54:06 -0400413 inode->i_mode = table->mode;
Eric W. Biederman7ec66d02011-12-29 08:24:29 -0800414 if (!S_ISDIR(table->mode)) {
Al Viro9043476f2008-07-15 08:54:06 -0400415 inode->i_mode |= S_IFREG;
416 inode->i_op = &proc_sys_inode_operations;
417 inode->i_fop = &proc_sys_file_operations;
418 } else {
419 inode->i_mode |= S_IFDIR;
Al Viro9043476f2008-07-15 08:54:06 -0400420 inode->i_op = &proc_sys_dir_operations;
421 inode->i_fop = &proc_sys_dir_file_operations;
422 }
Eric W. Biederman77b14db2007-02-14 00:34:12 -0800423out:
424 return inode;
425}
426
Adrian Bunk81324362008-10-03 00:33:54 +0400427static struct ctl_table_header *grab_header(struct inode *inode)
Eric W. Biederman77b14db2007-02-14 00:34:12 -0800428{
Eric W. Biederman3cc3e042012-01-07 06:57:47 -0800429 struct ctl_table_header *head = PROC_I(inode)->sysctl;
430 if (!head)
Eric W. Biederman0e47c992012-01-07 23:24:30 -0800431 head = &sysctl_table_root.default_set.dir.header;
Eric W. Biederman3cc3e042012-01-07 06:57:47 -0800432 return sysctl_head_grab(head);
Eric W. Biederman77b14db2007-02-14 00:34:12 -0800433}
434
435static struct dentry *proc_sys_lookup(struct inode *dir, struct dentry *dentry,
Al Viro00cd8dd2012-06-10 17:13:09 -0400436 unsigned int flags)
Eric W. Biederman77b14db2007-02-14 00:34:12 -0800437{
Al Viro9043476f2008-07-15 08:54:06 -0400438 struct ctl_table_header *head = grab_header(dir);
Al Viro9043476f2008-07-15 08:54:06 -0400439 struct ctl_table_header *h = NULL;
440 struct qstr *name = &dentry->d_name;
441 struct ctl_table *p;
Eric W. Biederman77b14db2007-02-14 00:34:12 -0800442 struct inode *inode;
Al Viro9043476f2008-07-15 08:54:06 -0400443 struct dentry *err = ERR_PTR(-ENOENT);
Eric W. Biederman7ec66d02011-12-29 08:24:29 -0800444 struct ctl_dir *ctl_dir;
Eric W. Biederman0e47c992012-01-07 23:24:30 -0800445 int ret;
Eric W. Biederman77b14db2007-02-14 00:34:12 -0800446
Al Viro9043476f2008-07-15 08:54:06 -0400447 if (IS_ERR(head))
448 return ERR_CAST(head);
449
Eric W. Biederman7ec66d02011-12-29 08:24:29 -0800450 ctl_dir = container_of(head, struct ctl_dir, header);
Al Viro9043476f2008-07-15 08:54:06 -0400451
Eric W. Biederman7ec66d02011-12-29 08:24:29 -0800452 p = lookup_entry(&h, ctl_dir, name->name, name->len);
Al Viro9043476f2008-07-15 08:54:06 -0400453 if (!p)
Eric W. Biederman77b14db2007-02-14 00:34:12 -0800454 goto out;
455
Eric W. Biederman4e757322012-01-30 21:24:59 -0800456 if (S_ISLNK(p->mode)) {
457 ret = sysctl_follow_link(&h, &p, current->nsproxy);
458 err = ERR_PTR(ret);
459 if (ret)
460 goto out;
461 }
Eric W. Biederman0e47c992012-01-07 23:24:30 -0800462
Eric W. Biederman77b14db2007-02-14 00:34:12 -0800463 err = ERR_PTR(-ENOMEM);
Al Viro9043476f2008-07-15 08:54:06 -0400464 inode = proc_sys_make_inode(dir->i_sb, h ? h : head, p);
Eric W. Biederman77b14db2007-02-14 00:34:12 -0800465 if (!inode)
466 goto out;
467
468 err = NULL;
Nick Pigginfb045ad2011-01-07 17:49:55 +1100469 d_set_d_op(dentry, &proc_sys_dentry_operations);
Eric W. Biederman77b14db2007-02-14 00:34:12 -0800470 d_add(dentry, inode);
471
472out:
Francesco Ruggeri6bf61042012-09-13 15:03:37 -0700473 if (h)
474 sysctl_head_finish(h);
Eric W. Biederman77b14db2007-02-14 00:34:12 -0800475 sysctl_head_finish(head);
476 return err;
477}
478
Pavel Emelyanov7708bfb2008-04-29 01:02:40 -0700479static ssize_t proc_sys_call_handler(struct file *filp, void __user *buf,
480 size_t count, loff_t *ppos, int write)
Eric W. Biederman77b14db2007-02-14 00:34:12 -0800481{
Al Viro496ad9a2013-01-23 17:07:38 -0500482 struct inode *inode = file_inode(filp);
Al Viro9043476f2008-07-15 08:54:06 -0400483 struct ctl_table_header *head = grab_header(inode);
484 struct ctl_table *table = PROC_I(inode)->sysctl_entry;
David Howells2a2da532007-10-25 15:27:40 +0100485 ssize_t error;
486 size_t res;
Eric W. Biederman77b14db2007-02-14 00:34:12 -0800487
Al Viro9043476f2008-07-15 08:54:06 -0400488 if (IS_ERR(head))
489 return PTR_ERR(head);
Eric W. Biederman77b14db2007-02-14 00:34:12 -0800490
491 /*
492 * At this point we know that the sysctl was not unregistered
493 * and won't be until we finish.
494 */
495 error = -EPERM;
Eric W. Biederman73f7ef42012-11-16 03:02:58 +0000496 if (sysctl_perm(head, table, write ? MAY_WRITE : MAY_READ))
Eric W. Biederman77b14db2007-02-14 00:34:12 -0800497 goto out;
498
Al Viro9043476f2008-07-15 08:54:06 -0400499 /* if that can happen at all, it should be -EINVAL, not -EISDIR */
500 error = -EINVAL;
501 if (!table->proc_handler)
502 goto out;
503
Eric W. Biederman77b14db2007-02-14 00:34:12 -0800504 /* careful: calling conventions are nasty here */
505 res = count;
Alexey Dobriyan8d65af72009-09-23 15:57:19 -0700506 error = table->proc_handler(table, write, buf, &res, ppos);
Eric W. Biederman77b14db2007-02-14 00:34:12 -0800507 if (!error)
508 error = res;
509out:
510 sysctl_head_finish(head);
511
512 return error;
513}
514
Pavel Emelyanov7708bfb2008-04-29 01:02:40 -0700515static ssize_t proc_sys_read(struct file *filp, char __user *buf,
516 size_t count, loff_t *ppos)
517{
518 return proc_sys_call_handler(filp, (void __user *)buf, count, ppos, 0);
519}
520
Eric W. Biederman77b14db2007-02-14 00:34:12 -0800521static ssize_t proc_sys_write(struct file *filp, const char __user *buf,
522 size_t count, loff_t *ppos)
523{
Pavel Emelyanov7708bfb2008-04-29 01:02:40 -0700524 return proc_sys_call_handler(filp, (void __user *)buf, count, ppos, 1);
Eric W. Biederman77b14db2007-02-14 00:34:12 -0800525}
526
Lucas De Marchif1ecf062011-11-02 13:39:22 -0700527static int proc_sys_open(struct inode *inode, struct file *filp)
528{
Lucas De Marchi4e474a02012-03-22 14:42:22 -0700529 struct ctl_table_header *head = grab_header(inode);
Lucas De Marchif1ecf062011-11-02 13:39:22 -0700530 struct ctl_table *table = PROC_I(inode)->sysctl_entry;
531
Lucas De Marchi4e474a02012-03-22 14:42:22 -0700532 /* sysctl was unregistered */
533 if (IS_ERR(head))
534 return PTR_ERR(head);
535
Lucas De Marchif1ecf062011-11-02 13:39:22 -0700536 if (table->poll)
537 filp->private_data = proc_sys_poll_event(table->poll);
538
Lucas De Marchi4e474a02012-03-22 14:42:22 -0700539 sysctl_head_finish(head);
540
Lucas De Marchif1ecf062011-11-02 13:39:22 -0700541 return 0;
542}
543
544static unsigned int proc_sys_poll(struct file *filp, poll_table *wait)
545{
Al Viro496ad9a2013-01-23 17:07:38 -0500546 struct inode *inode = file_inode(filp);
Lucas De Marchi4e474a02012-03-22 14:42:22 -0700547 struct ctl_table_header *head = grab_header(inode);
Lucas De Marchif1ecf062011-11-02 13:39:22 -0700548 struct ctl_table *table = PROC_I(inode)->sysctl_entry;
Lucas De Marchif1ecf062011-11-02 13:39:22 -0700549 unsigned int ret = DEFAULT_POLLMASK;
Lucas De Marchi4e474a02012-03-22 14:42:22 -0700550 unsigned long event;
551
552 /* sysctl was unregistered */
553 if (IS_ERR(head))
554 return POLLERR | POLLHUP;
Lucas De Marchif1ecf062011-11-02 13:39:22 -0700555
556 if (!table->proc_handler)
557 goto out;
558
559 if (!table->poll)
560 goto out;
561
Lucas De Marchi4e474a02012-03-22 14:42:22 -0700562 event = (unsigned long)filp->private_data;
Lucas De Marchif1ecf062011-11-02 13:39:22 -0700563 poll_wait(filp, &table->poll->wait, wait);
564
565 if (event != atomic_read(&table->poll->event)) {
566 filp->private_data = proc_sys_poll_event(table->poll);
567 ret = POLLIN | POLLRDNORM | POLLERR | POLLPRI;
568 }
569
570out:
Lucas De Marchi4e474a02012-03-22 14:42:22 -0700571 sysctl_head_finish(head);
572
Lucas De Marchif1ecf062011-11-02 13:39:22 -0700573 return ret;
574}
Eric W. Biederman77b14db2007-02-14 00:34:12 -0800575
Al Virof0c3b502013-05-16 12:07:31 -0400576static bool proc_sys_fill_cache(struct file *file,
577 struct dir_context *ctx,
Al Viro9043476f2008-07-15 08:54:06 -0400578 struct ctl_table_header *head,
579 struct ctl_table *table)
Eric W. Biederman77b14db2007-02-14 00:34:12 -0800580{
Al Virof0c3b502013-05-16 12:07:31 -0400581 struct dentry *child, *dir = file->f_path.dentry;
Eric W. Biederman77b14db2007-02-14 00:34:12 -0800582 struct inode *inode;
583 struct qstr qname;
584 ino_t ino = 0;
585 unsigned type = DT_UNKNOWN;
Eric W. Biederman77b14db2007-02-14 00:34:12 -0800586
587 qname.name = table->procname;
588 qname.len = strlen(table->procname);
589 qname.hash = full_name_hash(qname.name, qname.len);
590
Eric W. Biederman77b14db2007-02-14 00:34:12 -0800591 child = d_lookup(dir, &qname);
592 if (!child) {
Al Viro9043476f2008-07-15 08:54:06 -0400593 child = d_alloc(dir, &qname);
594 if (child) {
595 inode = proc_sys_make_inode(dir->d_sb, head, table);
596 if (!inode) {
597 dput(child);
Al Virof0c3b502013-05-16 12:07:31 -0400598 return false;
Al Viro9043476f2008-07-15 08:54:06 -0400599 } else {
Nick Pigginfb045ad2011-01-07 17:49:55 +1100600 d_set_d_op(child, &proc_sys_dentry_operations);
Al Viro9043476f2008-07-15 08:54:06 -0400601 d_add(child, inode);
Eric W. Biederman77b14db2007-02-14 00:34:12 -0800602 }
Al Viro9043476f2008-07-15 08:54:06 -0400603 } else {
Al Virof0c3b502013-05-16 12:07:31 -0400604 return false;
Eric W. Biederman77b14db2007-02-14 00:34:12 -0800605 }
606 }
Eric W. Biederman77b14db2007-02-14 00:34:12 -0800607 inode = child->d_inode;
Al Viro9043476f2008-07-15 08:54:06 -0400608 ino = inode->i_ino;
609 type = inode->i_mode >> 12;
Eric W. Biederman77b14db2007-02-14 00:34:12 -0800610 dput(child);
Al Virof0c3b502013-05-16 12:07:31 -0400611 return dir_emit(ctx, qname.name, qname.len, ino, type);
Al Viro9043476f2008-07-15 08:54:06 -0400612}
613
Al Virof0c3b502013-05-16 12:07:31 -0400614static bool proc_sys_link_fill_cache(struct file *file,
615 struct dir_context *ctx,
Eric W. Biederman0e47c992012-01-07 23:24:30 -0800616 struct ctl_table_header *head,
617 struct ctl_table *table)
618{
Al Virof0c3b502013-05-16 12:07:31 -0400619 bool ret = true;
Eric W. Biederman0e47c992012-01-07 23:24:30 -0800620 head = sysctl_head_grab(head);
621
Eric W. Biederman4e757322012-01-30 21:24:59 -0800622 if (S_ISLNK(table->mode)) {
623 /* It is not an error if we can not follow the link ignore it */
Al Virof0c3b502013-05-16 12:07:31 -0400624 int err = sysctl_follow_link(&head, &table, current->nsproxy);
Eric W. Biederman4e757322012-01-30 21:24:59 -0800625 if (err)
626 goto out;
627 }
Eric W. Biederman0e47c992012-01-07 23:24:30 -0800628
Al Virof0c3b502013-05-16 12:07:31 -0400629 ret = proc_sys_fill_cache(file, ctx, head, table);
Eric W. Biederman0e47c992012-01-07 23:24:30 -0800630out:
631 sysctl_head_finish(head);
632 return ret;
633}
634
Joe Perchese5eea092014-08-08 14:22:16 -0700635static int scan(struct ctl_table_header *head, struct ctl_table *table,
Al Viro9043476f2008-07-15 08:54:06 -0400636 unsigned long *pos, struct file *file,
Al Virof0c3b502013-05-16 12:07:31 -0400637 struct dir_context *ctx)
Al Viro9043476f2008-07-15 08:54:06 -0400638{
Al Virof0c3b502013-05-16 12:07:31 -0400639 bool res;
Al Viro9043476f2008-07-15 08:54:06 -0400640
Al Virof0c3b502013-05-16 12:07:31 -0400641 if ((*pos)++ < ctx->pos)
642 return true;
Al Viro9043476f2008-07-15 08:54:06 -0400643
Eric W. Biederman0e47c992012-01-07 23:24:30 -0800644 if (unlikely(S_ISLNK(table->mode)))
Al Virof0c3b502013-05-16 12:07:31 -0400645 res = proc_sys_link_fill_cache(file, ctx, head, table);
Eric W. Biederman0e47c992012-01-07 23:24:30 -0800646 else
Al Virof0c3b502013-05-16 12:07:31 -0400647 res = proc_sys_fill_cache(file, ctx, head, table);
Al Viro9043476f2008-07-15 08:54:06 -0400648
Al Virof0c3b502013-05-16 12:07:31 -0400649 if (res)
650 ctx->pos = *pos;
Al Viro9043476f2008-07-15 08:54:06 -0400651
Eric W. Biederman6a75ce12012-01-18 03:15:51 -0800652 return res;
Eric W. Biederman77b14db2007-02-14 00:34:12 -0800653}
654
Al Virof0c3b502013-05-16 12:07:31 -0400655static int proc_sys_readdir(struct file *file, struct dir_context *ctx)
Eric W. Biederman77b14db2007-02-14 00:34:12 -0800656{
Al Virof0c3b502013-05-16 12:07:31 -0400657 struct ctl_table_header *head = grab_header(file_inode(file));
Al Viro9043476f2008-07-15 08:54:06 -0400658 struct ctl_table_header *h = NULL;
Eric W. Biederman6a75ce12012-01-18 03:15:51 -0800659 struct ctl_table *entry;
Eric W. Biederman7ec66d02011-12-29 08:24:29 -0800660 struct ctl_dir *ctl_dir;
Eric W. Biederman77b14db2007-02-14 00:34:12 -0800661 unsigned long pos;
Eric W. Biederman77b14db2007-02-14 00:34:12 -0800662
Al Viro9043476f2008-07-15 08:54:06 -0400663 if (IS_ERR(head))
664 return PTR_ERR(head);
665
Eric W. Biederman7ec66d02011-12-29 08:24:29 -0800666 ctl_dir = container_of(head, struct ctl_dir, header);
Eric W. Biederman77b14db2007-02-14 00:34:12 -0800667
Al Virof0c3b502013-05-16 12:07:31 -0400668 if (!dir_emit_dots(file, ctx))
669 return 0;
670
Eric W. Biederman77b14db2007-02-14 00:34:12 -0800671 pos = 2;
672
Eric W. Biederman7ec66d02011-12-29 08:24:29 -0800673 for (first_entry(ctl_dir, &h, &entry); h; next_entry(&h, &entry)) {
Al Virof0c3b502013-05-16 12:07:31 -0400674 if (!scan(h, entry, &pos, file, ctx)) {
Al Viro9043476f2008-07-15 08:54:06 -0400675 sysctl_head_finish(h);
676 break;
Eric W. Biederman77b14db2007-02-14 00:34:12 -0800677 }
678 }
Eric W. Biederman77b14db2007-02-14 00:34:12 -0800679 sysctl_head_finish(head);
Al Virof0c3b502013-05-16 12:07:31 -0400680 return 0;
Eric W. Biederman77b14db2007-02-14 00:34:12 -0800681}
682
Al Viro10556cb2011-06-20 19:28:19 -0400683static int proc_sys_permission(struct inode *inode, int mask)
Eric W. Biederman77b14db2007-02-14 00:34:12 -0800684{
685 /*
686 * sysctl entries that are not writeable,
687 * are _NOT_ writeable, capabilities or not.
688 */
Miklos Szeredif696a362008-07-31 13:41:58 +0200689 struct ctl_table_header *head;
690 struct ctl_table *table;
Eric W. Biederman77b14db2007-02-14 00:34:12 -0800691 int error;
692
Miklos Szeredif696a362008-07-31 13:41:58 +0200693 /* Executable files are not allowed under /proc/sys/ */
694 if ((mask & MAY_EXEC) && S_ISREG(inode->i_mode))
695 return -EACCES;
696
697 head = grab_header(inode);
Al Viro9043476f2008-07-15 08:54:06 -0400698 if (IS_ERR(head))
699 return PTR_ERR(head);
Eric W. Biederman77b14db2007-02-14 00:34:12 -0800700
Miklos Szeredif696a362008-07-31 13:41:58 +0200701 table = PROC_I(inode)->sysctl_entry;
Al Viro9043476f2008-07-15 08:54:06 -0400702 if (!table) /* global root - r-xr-xr-x */
703 error = mask & MAY_WRITE ? -EACCES : 0;
704 else /* Use the permissions on the sysctl table entry */
Eric W. Biederman73f7ef42012-11-16 03:02:58 +0000705 error = sysctl_perm(head, table, mask & ~MAY_NOT_BLOCK);
Eric W. Biederman77b14db2007-02-14 00:34:12 -0800706
Eric W. Biederman77b14db2007-02-14 00:34:12 -0800707 sysctl_head_finish(head);
708 return error;
709}
710
711static int proc_sys_setattr(struct dentry *dentry, struct iattr *attr)
712{
713 struct inode *inode = dentry->d_inode;
714 int error;
715
716 if (attr->ia_valid & (ATTR_MODE | ATTR_UID | ATTR_GID))
717 return -EPERM;
718
719 error = inode_change_ok(inode, attr);
Christoph Hellwig10257742010-06-04 11:30:02 +0200720 if (error)
721 return error;
Eric W. Biederman77b14db2007-02-14 00:34:12 -0800722
Christoph Hellwig10257742010-06-04 11:30:02 +0200723 setattr_copy(inode, attr);
724 mark_inode_dirty(inode);
725 return 0;
Eric W. Biederman77b14db2007-02-14 00:34:12 -0800726}
727
Al Viro9043476f2008-07-15 08:54:06 -0400728static int proc_sys_getattr(struct vfsmount *mnt, struct dentry *dentry, struct kstat *stat)
729{
730 struct inode *inode = dentry->d_inode;
731 struct ctl_table_header *head = grab_header(inode);
732 struct ctl_table *table = PROC_I(inode)->sysctl_entry;
733
734 if (IS_ERR(head))
735 return PTR_ERR(head);
736
737 generic_fillattr(inode, stat);
738 if (table)
739 stat->mode = (stat->mode & S_IFMT) | table->mode;
740
741 sysctl_head_finish(head);
742 return 0;
743}
744
Eric W. Biederman77b14db2007-02-14 00:34:12 -0800745static const struct file_operations proc_sys_file_operations = {
Lucas De Marchif1ecf062011-11-02 13:39:22 -0700746 .open = proc_sys_open,
747 .poll = proc_sys_poll,
Eric W. Biederman77b14db2007-02-14 00:34:12 -0800748 .read = proc_sys_read,
749 .write = proc_sys_write,
Arnd Bergmann6038f372010-08-15 18:52:59 +0200750 .llseek = default_llseek,
Al Viro9043476f2008-07-15 08:54:06 -0400751};
752
753static const struct file_operations proc_sys_dir_file_operations = {
Pavel Emelyanov887df072011-11-02 13:38:42 -0700754 .read = generic_read_dir,
Al Virof0c3b502013-05-16 12:07:31 -0400755 .iterate = proc_sys_readdir,
Christoph Hellwig3222a3e2008-09-03 21:53:01 +0200756 .llseek = generic_file_llseek,
Eric W. Biederman77b14db2007-02-14 00:34:12 -0800757};
758
Jan Engelhardt03a44822008-02-08 04:21:19 -0800759static const struct inode_operations proc_sys_inode_operations = {
Al Viro9043476f2008-07-15 08:54:06 -0400760 .permission = proc_sys_permission,
761 .setattr = proc_sys_setattr,
762 .getattr = proc_sys_getattr,
763};
764
765static const struct inode_operations proc_sys_dir_operations = {
Eric W. Biederman77b14db2007-02-14 00:34:12 -0800766 .lookup = proc_sys_lookup,
767 .permission = proc_sys_permission,
768 .setattr = proc_sys_setattr,
Al Viro9043476f2008-07-15 08:54:06 -0400769 .getattr = proc_sys_getattr,
Eric W. Biederman77b14db2007-02-14 00:34:12 -0800770};
771
Al Viro0b728e12012-06-10 16:03:43 -0400772static int proc_sys_revalidate(struct dentry *dentry, unsigned int flags)
Eric W. Biederman77b14db2007-02-14 00:34:12 -0800773{
Al Viro0b728e12012-06-10 16:03:43 -0400774 if (flags & LOOKUP_RCU)
Nick Piggin34286d62011-01-07 17:49:57 +1100775 return -ECHILD;
Al Viro9043476f2008-07-15 08:54:06 -0400776 return !PROC_I(dentry->d_inode)->sysctl->unregistering;
777}
778
Nick Pigginfe15ce42011-01-07 17:49:23 +1100779static int proc_sys_delete(const struct dentry *dentry)
Al Viro9043476f2008-07-15 08:54:06 -0400780{
781 return !!PROC_I(dentry->d_inode)->sysctl->unregistering;
782}
783
Eric W. Biederman1f87f0b2012-01-06 04:07:15 -0800784static int sysctl_is_seen(struct ctl_table_header *p)
785{
786 struct ctl_table_set *set = p->set;
787 int res;
788 spin_lock(&sysctl_lock);
789 if (p->unregistering)
790 res = 0;
791 else if (!set->is_seen)
792 res = 1;
793 else
794 res = set->is_seen(set);
795 spin_unlock(&sysctl_lock);
796 return res;
797}
798
Linus Torvaldsda53be12013-05-21 15:22:44 -0700799static int proc_sys_compare(const struct dentry *parent, const struct dentry *dentry,
Nick Piggin621e1552011-01-07 17:49:27 +1100800 unsigned int len, const char *str, const struct qstr *name)
Al Viro9043476f2008-07-15 08:54:06 -0400801{
Al Virodfef6dcd32011-03-08 01:25:28 -0500802 struct ctl_table_header *head;
Linus Torvaldsda53be12013-05-21 15:22:44 -0700803 struct inode *inode;
804
Nick Piggin31e6b012011-01-07 17:49:52 +1100805 /* Although proc doesn't have negative dentries, rcu-walk means
806 * that inode here can be NULL */
Al Virodfef6dcd32011-03-08 01:25:28 -0500807 /* AV: can it, indeed? */
Linus Torvaldsda53be12013-05-21 15:22:44 -0700808 inode = ACCESS_ONCE(dentry->d_inode);
Nick Piggin31e6b012011-01-07 17:49:52 +1100809 if (!inode)
Al Virodfef6dcd32011-03-08 01:25:28 -0500810 return 1;
Nick Piggin621e1552011-01-07 17:49:27 +1100811 if (name->len != len)
Al Viro9043476f2008-07-15 08:54:06 -0400812 return 1;
Nick Piggin621e1552011-01-07 17:49:27 +1100813 if (memcmp(name->name, str, len))
Al Viro9043476f2008-07-15 08:54:06 -0400814 return 1;
Al Virodfef6dcd32011-03-08 01:25:28 -0500815 head = rcu_dereference(PROC_I(inode)->sysctl);
816 return !head || !sysctl_is_seen(head);
Eric W. Biederman77b14db2007-02-14 00:34:12 -0800817}
818
Al Virod72f71e2009-02-20 05:58:47 +0000819static const struct dentry_operations proc_sys_dentry_operations = {
Eric W. Biederman77b14db2007-02-14 00:34:12 -0800820 .d_revalidate = proc_sys_revalidate,
Al Viro9043476f2008-07-15 08:54:06 -0400821 .d_delete = proc_sys_delete,
822 .d_compare = proc_sys_compare,
Eric W. Biederman77b14db2007-02-14 00:34:12 -0800823};
824
Eric W. Biederman0e47c992012-01-07 23:24:30 -0800825static struct ctl_dir *find_subdir(struct ctl_dir *dir,
826 const char *name, int namelen)
Eric W. Biederman1f87f0b2012-01-06 04:07:15 -0800827{
Eric W. Biederman7ec66d02011-12-29 08:24:29 -0800828 struct ctl_table_header *head;
829 struct ctl_table *entry;
Eric W. Biederman1f87f0b2012-01-06 04:07:15 -0800830
Eric W. Biederman0e47c992012-01-07 23:24:30 -0800831 entry = find_entry(&head, dir, name, namelen);
Eric W. Biederman7ec66d02011-12-29 08:24:29 -0800832 if (!entry)
833 return ERR_PTR(-ENOENT);
Eric W. Biederman51f72f42012-01-30 20:09:33 -0800834 if (!S_ISDIR(entry->mode))
835 return ERR_PTR(-ENOTDIR);
836 return container_of(head, struct ctl_dir, header);
Eric W. Biederman1f87f0b2012-01-06 04:07:15 -0800837}
838
Eric W. Biederman7ec66d02011-12-29 08:24:29 -0800839static struct ctl_dir *new_dir(struct ctl_table_set *set,
Eric W. Biederman0e47c992012-01-07 23:24:30 -0800840 const char *name, int namelen)
Eric W. Biederman1f87f0b2012-01-06 04:07:15 -0800841{
Eric W. Biederman7ec66d02011-12-29 08:24:29 -0800842 struct ctl_table *table;
843 struct ctl_dir *new;
Eric W. Biedermanac13ac62012-01-09 17:24:30 -0800844 struct ctl_node *node;
Eric W. Biederman7ec66d02011-12-29 08:24:29 -0800845 char *new_name;
Eric W. Biederman1f87f0b2012-01-06 04:07:15 -0800846
Eric W. Biedermanac13ac62012-01-09 17:24:30 -0800847 new = kzalloc(sizeof(*new) + sizeof(struct ctl_node) +
848 sizeof(struct ctl_table)*2 + namelen + 1,
849 GFP_KERNEL);
Eric W. Biederman7ec66d02011-12-29 08:24:29 -0800850 if (!new)
851 return NULL;
Eric W. Biederman1f87f0b2012-01-06 04:07:15 -0800852
Eric W. Biedermanac13ac62012-01-09 17:24:30 -0800853 node = (struct ctl_node *)(new + 1);
854 table = (struct ctl_table *)(node + 1);
Eric W. Biederman7ec66d02011-12-29 08:24:29 -0800855 new_name = (char *)(table + 2);
856 memcpy(new_name, name, namelen);
857 new_name[namelen] = '\0';
858 table[0].procname = new_name;
859 table[0].mode = S_IFDIR|S_IRUGO|S_IXUGO;
Eric W. Biedermanac13ac62012-01-09 17:24:30 -0800860 init_header(&new->header, set->dir.header.root, set, node, table);
Eric W. Biederman7ec66d02011-12-29 08:24:29 -0800861
862 return new;
863}
864
Eric W. Biederman60f126d2012-01-30 21:23:52 -0800865/**
866 * get_subdir - find or create a subdir with the specified name.
867 * @dir: Directory to create the subdirectory in
868 * @name: The name of the subdirectory to find or create
869 * @namelen: The length of name
870 *
871 * Takes a directory with an elevated reference count so we know that
872 * if we drop the lock the directory will not go away. Upon success
873 * the reference is moved from @dir to the returned subdirectory.
874 * Upon error an error code is returned and the reference on @dir is
875 * simply dropped.
876 */
Eric W. Biederman0e47c992012-01-07 23:24:30 -0800877static struct ctl_dir *get_subdir(struct ctl_dir *dir,
878 const char *name, int namelen)
Eric W. Biederman7ec66d02011-12-29 08:24:29 -0800879{
Eric W. Biederman0e47c992012-01-07 23:24:30 -0800880 struct ctl_table_set *set = dir->header.set;
Eric W. Biederman7ec66d02011-12-29 08:24:29 -0800881 struct ctl_dir *subdir, *new = NULL;
Eric W. Biederman0eb97f32012-01-30 20:37:51 -0800882 int err;
Eric W. Biederman7ec66d02011-12-29 08:24:29 -0800883
884 spin_lock(&sysctl_lock);
Eric W. Biederman0e47c992012-01-07 23:24:30 -0800885 subdir = find_subdir(dir, name, namelen);
Eric W. Biederman7ec66d02011-12-29 08:24:29 -0800886 if (!IS_ERR(subdir))
887 goto found;
888 if (PTR_ERR(subdir) != -ENOENT)
889 goto failed;
890
891 spin_unlock(&sysctl_lock);
892 new = new_dir(set, name, namelen);
893 spin_lock(&sysctl_lock);
894 subdir = ERR_PTR(-ENOMEM);
895 if (!new)
896 goto failed;
897
Eric W. Biederman60f126d2012-01-30 21:23:52 -0800898 /* Was the subdir added while we dropped the lock? */
Eric W. Biederman0e47c992012-01-07 23:24:30 -0800899 subdir = find_subdir(dir, name, namelen);
Eric W. Biederman7ec66d02011-12-29 08:24:29 -0800900 if (!IS_ERR(subdir))
901 goto found;
902 if (PTR_ERR(subdir) != -ENOENT)
903 goto failed;
904
Eric W. Biederman60f126d2012-01-30 21:23:52 -0800905 /* Nope. Use the our freshly made directory entry. */
Eric W. Biederman0eb97f32012-01-30 20:37:51 -0800906 err = insert_header(dir, &new->header);
907 subdir = ERR_PTR(err);
908 if (err)
Eric W. Biederman0e47c992012-01-07 23:24:30 -0800909 goto failed;
Eric W. Biederman7ec66d02011-12-29 08:24:29 -0800910 subdir = new;
911found:
912 subdir->header.nreg++;
913failed:
914 if (unlikely(IS_ERR(subdir))) {
Andrew Morton87ebdc02013-02-27 17:03:16 -0800915 pr_err("sysctl could not get directory: ");
Eric W. Biederman69801282012-01-21 20:09:45 -0800916 sysctl_print_dir(dir);
Andrew Morton87ebdc02013-02-27 17:03:16 -0800917 pr_cont("/%*.*s %ld\n",
Eric W. Biederman7ec66d02011-12-29 08:24:29 -0800918 namelen, namelen, name, PTR_ERR(subdir));
Eric W. Biederman1f87f0b2012-01-06 04:07:15 -0800919 }
Eric W. Biederman7ec66d02011-12-29 08:24:29 -0800920 drop_sysctl_table(&dir->header);
921 if (new)
922 drop_sysctl_table(&new->header);
923 spin_unlock(&sysctl_lock);
924 return subdir;
Eric W. Biederman1f87f0b2012-01-06 04:07:15 -0800925}
926
Eric W. Biederman0e47c992012-01-07 23:24:30 -0800927static struct ctl_dir *xlate_dir(struct ctl_table_set *set, struct ctl_dir *dir)
928{
929 struct ctl_dir *parent;
930 const char *procname;
931 if (!dir->header.parent)
932 return &set->dir;
933 parent = xlate_dir(set, dir->header.parent);
934 if (IS_ERR(parent))
935 return parent;
936 procname = dir->header.ctl_table[0].procname;
937 return find_subdir(parent, procname, strlen(procname));
938}
939
940static int sysctl_follow_link(struct ctl_table_header **phead,
941 struct ctl_table **pentry, struct nsproxy *namespaces)
942{
943 struct ctl_table_header *head;
944 struct ctl_table_root *root;
945 struct ctl_table_set *set;
946 struct ctl_table *entry;
947 struct ctl_dir *dir;
948 int ret;
949
Eric W. Biederman0e47c992012-01-07 23:24:30 -0800950 ret = 0;
951 spin_lock(&sysctl_lock);
952 root = (*pentry)->data;
953 set = lookup_header_set(root, namespaces);
954 dir = xlate_dir(set, (*phead)->parent);
955 if (IS_ERR(dir))
956 ret = PTR_ERR(dir);
957 else {
958 const char *procname = (*pentry)->procname;
959 head = NULL;
960 entry = find_entry(&head, dir, procname, strlen(procname));
961 ret = -ENOENT;
962 if (entry && use_table(head)) {
963 unuse_table(*phead);
964 *phead = head;
965 *pentry = entry;
966 ret = 0;
967 }
968 }
969
970 spin_unlock(&sysctl_lock);
971 return ret;
972}
973
Eric W. Biederman7c60c482012-01-21 13:34:05 -0800974static int sysctl_err(const char *path, struct ctl_table *table, char *fmt, ...)
975{
976 struct va_format vaf;
977 va_list args;
978
979 va_start(args, fmt);
980 vaf.fmt = fmt;
981 vaf.va = &args;
982
Andrew Morton87ebdc02013-02-27 17:03:16 -0800983 pr_err("sysctl table check failed: %s/%s %pV\n",
984 path, table->procname, &vaf);
Eric W. Biederman7c60c482012-01-21 13:34:05 -0800985
986 va_end(args);
987 return -EINVAL;
988}
989
990static int sysctl_check_table(const char *path, struct ctl_table *table)
991{
992 int err = 0;
993 for (; table->procname; table++) {
994 if (table->child)
995 err = sysctl_err(path, table, "Not a file");
996
997 if ((table->proc_handler == proc_dostring) ||
998 (table->proc_handler == proc_dointvec) ||
999 (table->proc_handler == proc_dointvec_minmax) ||
1000 (table->proc_handler == proc_dointvec_jiffies) ||
1001 (table->proc_handler == proc_dointvec_userhz_jiffies) ||
1002 (table->proc_handler == proc_dointvec_ms_jiffies) ||
1003 (table->proc_handler == proc_doulongvec_minmax) ||
1004 (table->proc_handler == proc_doulongvec_ms_jiffies_minmax)) {
1005 if (!table->data)
1006 err = sysctl_err(path, table, "No data");
1007 if (!table->maxlen)
1008 err = sysctl_err(path, table, "No maxlen");
1009 }
1010 if (!table->proc_handler)
1011 err = sysctl_err(path, table, "No proc_handler");
1012
1013 if ((table->mode & (S_IRUGO|S_IWUGO)) != table->mode)
1014 err = sysctl_err(path, table, "bogus .mode 0%o",
1015 table->mode);
1016 }
1017 return err;
1018}
Eric W. Biederman1f87f0b2012-01-06 04:07:15 -08001019
Eric W. Biederman0e47c992012-01-07 23:24:30 -08001020static struct ctl_table_header *new_links(struct ctl_dir *dir, struct ctl_table *table,
1021 struct ctl_table_root *link_root)
1022{
1023 struct ctl_table *link_table, *entry, *link;
1024 struct ctl_table_header *links;
Eric W. Biedermanac13ac62012-01-09 17:24:30 -08001025 struct ctl_node *node;
Eric W. Biederman0e47c992012-01-07 23:24:30 -08001026 char *link_name;
1027 int nr_entries, name_bytes;
1028
1029 name_bytes = 0;
1030 nr_entries = 0;
1031 for (entry = table; entry->procname; entry++) {
1032 nr_entries++;
1033 name_bytes += strlen(entry->procname) + 1;
1034 }
1035
1036 links = kzalloc(sizeof(struct ctl_table_header) +
Eric W. Biedermanac13ac62012-01-09 17:24:30 -08001037 sizeof(struct ctl_node)*nr_entries +
Eric W. Biederman0e47c992012-01-07 23:24:30 -08001038 sizeof(struct ctl_table)*(nr_entries + 1) +
1039 name_bytes,
1040 GFP_KERNEL);
1041
1042 if (!links)
1043 return NULL;
1044
Eric W. Biedermanac13ac62012-01-09 17:24:30 -08001045 node = (struct ctl_node *)(links + 1);
1046 link_table = (struct ctl_table *)(node + nr_entries);
Eric W. Biederman0e47c992012-01-07 23:24:30 -08001047 link_name = (char *)&link_table[nr_entries + 1];
1048
1049 for (link = link_table, entry = table; entry->procname; link++, entry++) {
1050 int len = strlen(entry->procname) + 1;
1051 memcpy(link_name, entry->procname, len);
1052 link->procname = link_name;
1053 link->mode = S_IFLNK|S_IRWXUGO;
1054 link->data = link_root;
1055 link_name += len;
1056 }
Eric W. Biedermanac13ac62012-01-09 17:24:30 -08001057 init_header(links, dir->header.root, dir->header.set, node, link_table);
Eric W. Biederman0e47c992012-01-07 23:24:30 -08001058 links->nreg = nr_entries;
1059
1060 return links;
1061}
1062
1063static bool get_links(struct ctl_dir *dir,
1064 struct ctl_table *table, struct ctl_table_root *link_root)
1065{
1066 struct ctl_table_header *head;
1067 struct ctl_table *entry, *link;
1068
1069 /* Are there links available for every entry in table? */
1070 for (entry = table; entry->procname; entry++) {
1071 const char *procname = entry->procname;
1072 link = find_entry(&head, dir, procname, strlen(procname));
1073 if (!link)
1074 return false;
1075 if (S_ISDIR(link->mode) && S_ISDIR(entry->mode))
1076 continue;
1077 if (S_ISLNK(link->mode) && (link->data == link_root))
1078 continue;
1079 return false;
1080 }
1081
1082 /* The checks passed. Increase the registration count on the links */
1083 for (entry = table; entry->procname; entry++) {
1084 const char *procname = entry->procname;
1085 link = find_entry(&head, dir, procname, strlen(procname));
1086 head->nreg++;
1087 }
1088 return true;
1089}
1090
1091static int insert_links(struct ctl_table_header *head)
1092{
1093 struct ctl_table_set *root_set = &sysctl_table_root.default_set;
1094 struct ctl_dir *core_parent = NULL;
1095 struct ctl_table_header *links;
1096 int err;
1097
1098 if (head->set == root_set)
1099 return 0;
1100
1101 core_parent = xlate_dir(root_set, head->parent);
1102 if (IS_ERR(core_parent))
1103 return 0;
1104
1105 if (get_links(core_parent, head->ctl_table, head->root))
1106 return 0;
1107
1108 core_parent->header.nreg++;
1109 spin_unlock(&sysctl_lock);
1110
1111 links = new_links(core_parent, head->ctl_table, head->root);
1112
1113 spin_lock(&sysctl_lock);
1114 err = -ENOMEM;
1115 if (!links)
1116 goto out;
1117
1118 err = 0;
1119 if (get_links(core_parent, head->ctl_table, head->root)) {
1120 kfree(links);
1121 goto out;
1122 }
1123
1124 err = insert_header(core_parent, links);
1125 if (err)
1126 kfree(links);
1127out:
1128 drop_sysctl_table(&core_parent->header);
1129 return err;
1130}
1131
Eric W. Biederman1f87f0b2012-01-06 04:07:15 -08001132/**
Eric W. Biedermanf7280192012-01-22 18:22:05 -08001133 * __register_sysctl_table - register a leaf sysctl table
Eric W. Biederman60a47a22012-01-08 00:02:37 -08001134 * @set: Sysctl tree to register on
Eric W. Biederman1f87f0b2012-01-06 04:07:15 -08001135 * @path: The path to the directory the sysctl table is in.
1136 * @table: the top-level table structure
1137 *
1138 * Register a sysctl table hierarchy. @table should be a filled in ctl_table
1139 * array. A completely 0 filled entry terminates the table.
1140 *
1141 * The members of the &struct ctl_table structure are used as follows:
1142 *
1143 * procname - the name of the sysctl file under /proc/sys. Set to %NULL to not
1144 * enter a sysctl file
1145 *
1146 * data - a pointer to data for use by proc_handler
1147 *
1148 * maxlen - the maximum size in bytes of the data
1149 *
Eric W. Biedermanf7280192012-01-22 18:22:05 -08001150 * mode - the file permissions for the /proc/sys file
Eric W. Biederman1f87f0b2012-01-06 04:07:15 -08001151 *
Eric W. Biedermanf7280192012-01-22 18:22:05 -08001152 * child - must be %NULL.
Eric W. Biederman1f87f0b2012-01-06 04:07:15 -08001153 *
1154 * proc_handler - the text handler routine (described below)
1155 *
Eric W. Biederman1f87f0b2012-01-06 04:07:15 -08001156 * extra1, extra2 - extra pointers usable by the proc handler routines
1157 *
1158 * Leaf nodes in the sysctl tree will be represented by a single file
1159 * under /proc; non-leaf nodes will be represented by directories.
1160 *
Eric W. Biedermanf7280192012-01-22 18:22:05 -08001161 * There must be a proc_handler routine for any terminal nodes.
1162 * Several default handlers are available to cover common cases -
Eric W. Biederman1f87f0b2012-01-06 04:07:15 -08001163 *
1164 * proc_dostring(), proc_dointvec(), proc_dointvec_jiffies(),
1165 * proc_dointvec_userhz_jiffies(), proc_dointvec_minmax(),
1166 * proc_doulongvec_ms_jiffies_minmax(), proc_doulongvec_minmax()
1167 *
1168 * It is the handler's job to read the input buffer from user memory
1169 * and process it. The handler should return 0 on success.
1170 *
1171 * This routine returns %NULL on a failure to register, and a pointer
1172 * to the table header on success.
1173 */
Eric W. Biederman6e9d5162012-01-21 10:26:26 -08001174struct ctl_table_header *__register_sysctl_table(
Eric W. Biederman60a47a22012-01-08 00:02:37 -08001175 struct ctl_table_set *set,
Eric W. Biederman6e9d5162012-01-21 10:26:26 -08001176 const char *path, struct ctl_table *table)
Eric W. Biederman1f87f0b2012-01-06 04:07:15 -08001177{
Eric W. Biederman60a47a22012-01-08 00:02:37 -08001178 struct ctl_table_root *root = set->dir.header.root;
Eric W. Biederman1f87f0b2012-01-06 04:07:15 -08001179 struct ctl_table_header *header;
Eric W. Biederman6e9d5162012-01-21 10:26:26 -08001180 const char *name, *nextname;
Eric W. Biederman7ec66d02011-12-29 08:24:29 -08001181 struct ctl_dir *dir;
Eric W. Biedermanac13ac62012-01-09 17:24:30 -08001182 struct ctl_table *entry;
1183 struct ctl_node *node;
1184 int nr_entries = 0;
Eric W. Biederman1f87f0b2012-01-06 04:07:15 -08001185
Eric W. Biedermanac13ac62012-01-09 17:24:30 -08001186 for (entry = table; entry->procname; entry++)
1187 nr_entries++;
1188
1189 header = kzalloc(sizeof(struct ctl_table_header) +
1190 sizeof(struct ctl_node)*nr_entries, GFP_KERNEL);
Eric W. Biederman1f87f0b2012-01-06 04:07:15 -08001191 if (!header)
1192 return NULL;
1193
Eric W. Biedermanac13ac62012-01-09 17:24:30 -08001194 node = (struct ctl_node *)(header + 1);
1195 init_header(header, root, set, node, table);
Eric W. Biederman7c60c482012-01-21 13:34:05 -08001196 if (sysctl_check_table(path, table))
1197 goto fail;
Eric W. Biederman8d6ecfc2012-01-06 11:55:30 -08001198
Eric W. Biederman1f87f0b2012-01-06 04:07:15 -08001199 spin_lock(&sysctl_lock);
Eric W. Biederman0e47c992012-01-07 23:24:30 -08001200 dir = &set->dir;
Eric W. Biederman60f126d2012-01-30 21:23:52 -08001201 /* Reference moved down the diretory tree get_subdir */
Eric W. Biederman7ec66d02011-12-29 08:24:29 -08001202 dir->header.nreg++;
1203 spin_unlock(&sysctl_lock);
1204
1205 /* Find the directory for the ctl_table */
1206 for (name = path; name; name = nextname) {
1207 int namelen;
1208 nextname = strchr(name, '/');
1209 if (nextname) {
1210 namelen = nextname - name;
1211 nextname++;
1212 } else {
1213 namelen = strlen(name);
Eric W. Biederman1f87f0b2012-01-06 04:07:15 -08001214 }
Eric W. Biederman7ec66d02011-12-29 08:24:29 -08001215 if (namelen == 0)
1216 continue;
1217
Eric W. Biederman0e47c992012-01-07 23:24:30 -08001218 dir = get_subdir(dir, name, namelen);
Eric W. Biederman7ec66d02011-12-29 08:24:29 -08001219 if (IS_ERR(dir))
1220 goto fail;
Eric W. Biederman1f87f0b2012-01-06 04:07:15 -08001221 }
Eric W. Biederman0e47c992012-01-07 23:24:30 -08001222
Eric W. Biederman7ec66d02011-12-29 08:24:29 -08001223 spin_lock(&sysctl_lock);
Eric W. Biederman0e47c992012-01-07 23:24:30 -08001224 if (insert_header(dir, header))
1225 goto fail_put_dir_locked;
1226
Eric W. Biederman7ec66d02011-12-29 08:24:29 -08001227 drop_sysctl_table(&dir->header);
Eric W. Biederman1f87f0b2012-01-06 04:07:15 -08001228 spin_unlock(&sysctl_lock);
1229
1230 return header;
Eric W. Biederman0e47c992012-01-07 23:24:30 -08001231
Eric W. Biederman7ec66d02011-12-29 08:24:29 -08001232fail_put_dir_locked:
1233 drop_sysctl_table(&dir->header);
Eric W. Biederman7c60c482012-01-21 13:34:05 -08001234 spin_unlock(&sysctl_lock);
1235fail:
1236 kfree(header);
1237 dump_stack();
1238 return NULL;
Eric W. Biederman1f87f0b2012-01-06 04:07:15 -08001239}
1240
Eric W. Biedermanfea478d2012-01-20 21:47:03 -08001241/**
1242 * register_sysctl - register a sysctl table
1243 * @path: The path to the directory the sysctl table is in.
1244 * @table: the table structure
1245 *
1246 * Register a sysctl table. @table should be a filled in ctl_table
1247 * array. A completely 0 filled entry terminates the table.
1248 *
1249 * See __register_sysctl_table for more details.
1250 */
1251struct ctl_table_header *register_sysctl(const char *path, struct ctl_table *table)
1252{
1253 return __register_sysctl_table(&sysctl_table_root.default_set,
1254 path, table);
1255}
1256EXPORT_SYMBOL(register_sysctl);
1257
Eric W. Biederman6e9d5162012-01-21 10:26:26 -08001258static char *append_path(const char *path, char *pos, const char *name)
1259{
1260 int namelen;
1261 namelen = strlen(name);
1262 if (((pos - path) + namelen + 2) >= PATH_MAX)
1263 return NULL;
1264 memcpy(pos, name, namelen);
1265 pos[namelen] = '/';
1266 pos[namelen + 1] = '\0';
1267 pos += namelen + 1;
1268 return pos;
1269}
1270
Eric W. Biedermanf7280192012-01-22 18:22:05 -08001271static int count_subheaders(struct ctl_table *table)
1272{
1273 int has_files = 0;
1274 int nr_subheaders = 0;
1275 struct ctl_table *entry;
1276
1277 /* special case: no directory and empty directory */
1278 if (!table || !table->procname)
1279 return 1;
1280
1281 for (entry = table; entry->procname; entry++) {
1282 if (entry->child)
1283 nr_subheaders += count_subheaders(entry->child);
1284 else
1285 has_files = 1;
1286 }
1287 return nr_subheaders + has_files;
1288}
1289
1290static int register_leaf_sysctl_tables(const char *path, char *pos,
Eric W. Biederman60a47a22012-01-08 00:02:37 -08001291 struct ctl_table_header ***subheader, struct ctl_table_set *set,
Eric W. Biedermanf7280192012-01-22 18:22:05 -08001292 struct ctl_table *table)
1293{
1294 struct ctl_table *ctl_table_arg = NULL;
1295 struct ctl_table *entry, *files;
1296 int nr_files = 0;
1297 int nr_dirs = 0;
1298 int err = -ENOMEM;
1299
1300 for (entry = table; entry->procname; entry++) {
1301 if (entry->child)
1302 nr_dirs++;
1303 else
1304 nr_files++;
1305 }
1306
1307 files = table;
1308 /* If there are mixed files and directories we need a new table */
1309 if (nr_dirs && nr_files) {
1310 struct ctl_table *new;
1311 files = kzalloc(sizeof(struct ctl_table) * (nr_files + 1),
1312 GFP_KERNEL);
1313 if (!files)
1314 goto out;
1315
1316 ctl_table_arg = files;
1317 for (new = files, entry = table; entry->procname; entry++) {
1318 if (entry->child)
1319 continue;
1320 *new = *entry;
1321 new++;
1322 }
1323 }
1324
1325 /* Register everything except a directory full of subdirectories */
1326 if (nr_files || !nr_dirs) {
1327 struct ctl_table_header *header;
Eric W. Biederman60a47a22012-01-08 00:02:37 -08001328 header = __register_sysctl_table(set, path, files);
Eric W. Biedermanf7280192012-01-22 18:22:05 -08001329 if (!header) {
1330 kfree(ctl_table_arg);
1331 goto out;
1332 }
1333
1334 /* Remember if we need to free the file table */
1335 header->ctl_table_arg = ctl_table_arg;
1336 **subheader = header;
1337 (*subheader)++;
1338 }
1339
1340 /* Recurse into the subdirectories. */
1341 for (entry = table; entry->procname; entry++) {
1342 char *child_pos;
1343
1344 if (!entry->child)
1345 continue;
1346
1347 err = -ENAMETOOLONG;
1348 child_pos = append_path(path, pos, entry->procname);
1349 if (!child_pos)
1350 goto out;
1351
1352 err = register_leaf_sysctl_tables(path, child_pos, subheader,
Eric W. Biederman60a47a22012-01-08 00:02:37 -08001353 set, entry->child);
Eric W. Biedermanf7280192012-01-22 18:22:05 -08001354 pos[0] = '\0';
1355 if (err)
1356 goto out;
1357 }
1358 err = 0;
1359out:
1360 /* On failure our caller will unregister all registered subheaders */
1361 return err;
1362}
1363
Eric W. Biederman6e9d5162012-01-21 10:26:26 -08001364/**
1365 * __register_sysctl_paths - register a sysctl table hierarchy
Eric W. Biederman60a47a22012-01-08 00:02:37 -08001366 * @set: Sysctl tree to register on
Eric W. Biederman6e9d5162012-01-21 10:26:26 -08001367 * @path: The path to the directory the sysctl table is in.
1368 * @table: the top-level table structure
1369 *
1370 * Register a sysctl table hierarchy. @table should be a filled in ctl_table
1371 * array. A completely 0 filled entry terminates the table.
1372 *
1373 * See __register_sysctl_table for more details.
1374 */
1375struct ctl_table_header *__register_sysctl_paths(
Eric W. Biederman60a47a22012-01-08 00:02:37 -08001376 struct ctl_table_set *set,
Eric W. Biederman6e9d5162012-01-21 10:26:26 -08001377 const struct ctl_path *path, struct ctl_table *table)
1378{
Eric W. Biedermanec6a5262012-01-21 12:35:23 -08001379 struct ctl_table *ctl_table_arg = table;
Eric W. Biedermanf7280192012-01-22 18:22:05 -08001380 int nr_subheaders = count_subheaders(table);
1381 struct ctl_table_header *header = NULL, **subheaders, **subheader;
Eric W. Biederman6e9d5162012-01-21 10:26:26 -08001382 const struct ctl_path *component;
1383 char *new_path, *pos;
1384
1385 pos = new_path = kmalloc(PATH_MAX, GFP_KERNEL);
1386 if (!new_path)
1387 return NULL;
1388
1389 pos[0] = '\0';
1390 for (component = path; component->procname; component++) {
1391 pos = append_path(new_path, pos, component->procname);
1392 if (!pos)
1393 goto out;
1394 }
Eric W. Biedermanec6a5262012-01-21 12:35:23 -08001395 while (table->procname && table->child && !table[1].procname) {
1396 pos = append_path(new_path, pos, table->procname);
1397 if (!pos)
1398 goto out;
1399 table = table->child;
1400 }
Eric W. Biedermanf7280192012-01-22 18:22:05 -08001401 if (nr_subheaders == 1) {
Eric W. Biederman60a47a22012-01-08 00:02:37 -08001402 header = __register_sysctl_table(set, new_path, table);
Eric W. Biedermanf7280192012-01-22 18:22:05 -08001403 if (header)
1404 header->ctl_table_arg = ctl_table_arg;
1405 } else {
1406 header = kzalloc(sizeof(*header) +
1407 sizeof(*subheaders)*nr_subheaders, GFP_KERNEL);
1408 if (!header)
1409 goto out;
1410
1411 subheaders = (struct ctl_table_header **) (header + 1);
1412 subheader = subheaders;
Eric W. Biedermanec6a5262012-01-21 12:35:23 -08001413 header->ctl_table_arg = ctl_table_arg;
Eric W. Biedermanf7280192012-01-22 18:22:05 -08001414
1415 if (register_leaf_sysctl_tables(new_path, pos, &subheader,
Eric W. Biederman60a47a22012-01-08 00:02:37 -08001416 set, table))
Eric W. Biedermanf7280192012-01-22 18:22:05 -08001417 goto err_register_leaves;
1418 }
1419
Eric W. Biederman6e9d5162012-01-21 10:26:26 -08001420out:
1421 kfree(new_path);
1422 return header;
Eric W. Biedermanf7280192012-01-22 18:22:05 -08001423
1424err_register_leaves:
1425 while (subheader > subheaders) {
1426 struct ctl_table_header *subh = *(--subheader);
1427 struct ctl_table *table = subh->ctl_table_arg;
1428 unregister_sysctl_table(subh);
1429 kfree(table);
1430 }
1431 kfree(header);
1432 header = NULL;
1433 goto out;
Eric W. Biederman6e9d5162012-01-21 10:26:26 -08001434}
1435
Eric W. Biederman1f87f0b2012-01-06 04:07:15 -08001436/**
1437 * register_sysctl_table_path - register a sysctl table hierarchy
1438 * @path: The path to the directory the sysctl table is in.
1439 * @table: the top-level table structure
1440 *
1441 * Register a sysctl table hierarchy. @table should be a filled in ctl_table
1442 * array. A completely 0 filled entry terminates the table.
1443 *
1444 * See __register_sysctl_paths for more details.
1445 */
1446struct ctl_table_header *register_sysctl_paths(const struct ctl_path *path,
1447 struct ctl_table *table)
1448{
Eric W. Biederman60a47a22012-01-08 00:02:37 -08001449 return __register_sysctl_paths(&sysctl_table_root.default_set,
Eric W. Biederman1f87f0b2012-01-06 04:07:15 -08001450 path, table);
1451}
1452EXPORT_SYMBOL(register_sysctl_paths);
1453
1454/**
1455 * register_sysctl_table - register a sysctl table hierarchy
1456 * @table: the top-level table structure
1457 *
1458 * Register a sysctl table hierarchy. @table should be a filled in ctl_table
1459 * array. A completely 0 filled entry terminates the table.
1460 *
1461 * See register_sysctl_paths for more details.
1462 */
1463struct ctl_table_header *register_sysctl_table(struct ctl_table *table)
1464{
1465 static const struct ctl_path null_path[] = { {} };
1466
1467 return register_sysctl_paths(null_path, table);
1468}
1469EXPORT_SYMBOL(register_sysctl_table);
1470
Eric W. Biederman0e47c992012-01-07 23:24:30 -08001471static void put_links(struct ctl_table_header *header)
1472{
1473 struct ctl_table_set *root_set = &sysctl_table_root.default_set;
1474 struct ctl_table_root *root = header->root;
1475 struct ctl_dir *parent = header->parent;
1476 struct ctl_dir *core_parent;
1477 struct ctl_table *entry;
1478
1479 if (header->set == root_set)
1480 return;
1481
1482 core_parent = xlate_dir(root_set, parent);
1483 if (IS_ERR(core_parent))
1484 return;
1485
1486 for (entry = header->ctl_table; entry->procname; entry++) {
1487 struct ctl_table_header *link_head;
1488 struct ctl_table *link;
1489 const char *name = entry->procname;
1490
1491 link = find_entry(&link_head, core_parent, name, strlen(name));
1492 if (link &&
1493 ((S_ISDIR(link->mode) && S_ISDIR(entry->mode)) ||
1494 (S_ISLNK(link->mode) && (link->data == root)))) {
1495 drop_sysctl_table(link_head);
1496 }
1497 else {
Andrew Morton87ebdc02013-02-27 17:03:16 -08001498 pr_err("sysctl link missing during unregister: ");
Eric W. Biederman0e47c992012-01-07 23:24:30 -08001499 sysctl_print_dir(parent);
Andrew Morton87ebdc02013-02-27 17:03:16 -08001500 pr_cont("/%s\n", name);
Eric W. Biederman0e47c992012-01-07 23:24:30 -08001501 }
1502 }
1503}
1504
Eric W. Biederman938aaa42012-01-09 17:24:30 -08001505static void drop_sysctl_table(struct ctl_table_header *header)
1506{
Eric W. Biederman7ec66d02011-12-29 08:24:29 -08001507 struct ctl_dir *parent = header->parent;
1508
Eric W. Biederman938aaa42012-01-09 17:24:30 -08001509 if (--header->nreg)
1510 return;
1511
Eric W. Biederman0e47c992012-01-07 23:24:30 -08001512 put_links(header);
Eric W. Biederman938aaa42012-01-09 17:24:30 -08001513 start_unregistering(header);
Eric W. Biederman938aaa42012-01-09 17:24:30 -08001514 if (!--header->count)
1515 kfree_rcu(header, rcu);
Eric W. Biederman7ec66d02011-12-29 08:24:29 -08001516
1517 if (parent)
1518 drop_sysctl_table(&parent->header);
Eric W. Biederman938aaa42012-01-09 17:24:30 -08001519}
1520
Eric W. Biederman1f87f0b2012-01-06 04:07:15 -08001521/**
1522 * unregister_sysctl_table - unregister a sysctl table hierarchy
1523 * @header: the header returned from register_sysctl_table
1524 *
1525 * Unregisters the sysctl table and all children. proc entries may not
1526 * actually be removed until they are no longer used by anyone.
1527 */
1528void unregister_sysctl_table(struct ctl_table_header * header)
1529{
Eric W. Biedermanf7280192012-01-22 18:22:05 -08001530 int nr_subheaders;
Eric W. Biederman1f87f0b2012-01-06 04:07:15 -08001531 might_sleep();
1532
1533 if (header == NULL)
1534 return;
1535
Eric W. Biedermanf7280192012-01-22 18:22:05 -08001536 nr_subheaders = count_subheaders(header->ctl_table_arg);
1537 if (unlikely(nr_subheaders > 1)) {
1538 struct ctl_table_header **subheaders;
1539 int i;
1540
1541 subheaders = (struct ctl_table_header **)(header + 1);
1542 for (i = nr_subheaders -1; i >= 0; i--) {
1543 struct ctl_table_header *subh = subheaders[i];
1544 struct ctl_table *table = subh->ctl_table_arg;
1545 unregister_sysctl_table(subh);
1546 kfree(table);
1547 }
1548 kfree(header);
1549 return;
1550 }
1551
Eric W. Biederman1f87f0b2012-01-06 04:07:15 -08001552 spin_lock(&sysctl_lock);
Eric W. Biederman938aaa42012-01-09 17:24:30 -08001553 drop_sysctl_table(header);
Eric W. Biederman1f87f0b2012-01-06 04:07:15 -08001554 spin_unlock(&sysctl_lock);
1555}
1556EXPORT_SYMBOL(unregister_sysctl_table);
1557
Eric W. Biederman0e47c992012-01-07 23:24:30 -08001558void setup_sysctl_set(struct ctl_table_set *set,
Eric W. Biederman9eb47c22012-01-22 21:26:00 -08001559 struct ctl_table_root *root,
Eric W. Biederman1f87f0b2012-01-06 04:07:15 -08001560 int (*is_seen)(struct ctl_table_set *))
1561{
Dan Carpenter13474402012-01-30 16:40:29 +03001562 memset(set, 0, sizeof(*set));
Eric W. Biederman0e47c992012-01-07 23:24:30 -08001563 set->is_seen = is_seen;
Eric W. Biedermanac13ac62012-01-09 17:24:30 -08001564 init_header(&set->dir.header, root, set, NULL, root_table);
Eric W. Biederman1f87f0b2012-01-06 04:07:15 -08001565}
1566
Eric W. Biederman97324cd82012-01-09 22:19:13 -08001567void retire_sysctl_set(struct ctl_table_set *set)
1568{
Eric W. Biedermanac13ac62012-01-09 17:24:30 -08001569 WARN_ON(!RB_EMPTY_ROOT(&set->dir.root));
Eric W. Biederman97324cd82012-01-09 22:19:13 -08001570}
Eric W. Biederman1f87f0b2012-01-06 04:07:15 -08001571
Alexey Dobriyan1e0edd32008-10-17 05:07:44 +04001572int __init proc_sys_init(void)
Eric W. Biederman77b14db2007-02-14 00:34:12 -08001573{
Alexey Dobriyane1675232008-10-03 00:23:32 +04001574 struct proc_dir_entry *proc_sys_root;
1575
Eric W. Biederman77b14db2007-02-14 00:34:12 -08001576 proc_sys_root = proc_mkdir("sys", NULL);
Al Viro9043476f2008-07-15 08:54:06 -04001577 proc_sys_root->proc_iops = &proc_sys_dir_operations;
1578 proc_sys_root->proc_fops = &proc_sys_dir_file_operations;
Eric W. Biederman77b14db2007-02-14 00:34:12 -08001579 proc_sys_root->nlink = 0;
Eric W. Biedermande4e83bd2012-01-06 03:34:20 -08001580
1581 return sysctl_init();
Eric W. Biederman77b14db2007-02-14 00:34:12 -08001582}