1 /* SPDX-License-Identifier: GPL-2.0-or-later */
2 /* Internal procfs definitions
3 *
4 * Copyright (C) 2004 Red Hat, Inc. All Rights Reserved.
5 * Written by David Howells (dhowells@redhat.com)
6 */
7
8 #include <linux/proc_fs.h>
9 #include <linux/proc_ns.h>
10 #include <linux/refcount.h>
11 #include <linux/spinlock.h>
12 #include <linux/atomic.h>
13 #include <linux/binfmts.h>
14 #include <linux/sched/coredump.h>
15 #include <linux/sched/task.h>
16 #include <linux/mm.h>
17
18 struct ctl_table_header;
19 struct mempolicy;
20
21 /*
22 * This is not completely implemented yet. The idea is to
23 * create an in-memory tree (like the actual /proc filesystem
24 * tree) of these proc_dir_entries, so that we can dynamically
25 * add new files to /proc.
26 *
27 * parent/subdir are used for the directory structure (every /proc file has a
28 * parent, but "subdir" is empty for all non-directory entries).
29 * subdir_node is used to build the rb tree "subdir" of the parent.
30 */
31 struct proc_dir_entry {
32 /*
33 * number of callers into module in progress;
34 * negative -> it's going away RSN
35 */
36 atomic_t in_use;
37 refcount_t refcnt;
38 struct list_head pde_openers; /* who did ->open, but not ->release */
39 /* protects ->pde_openers and all struct pde_opener instances */
40 spinlock_t pde_unload_lock;
41 struct completion *pde_unload_completion;
42 const struct inode_operations *proc_iops;
43 union {
44 const struct proc_ops *proc_ops;
45 const struct file_operations *proc_dir_ops;
46 };
47 union {
48 const struct seq_operations *seq_ops;
49 int (*single_show)(struct seq_file *, void *);
50 };
51 proc_write_t write;
52 void *data;
53 unsigned int state_size;
54 unsigned int low_ino;
55 nlink_t nlink;
56 kuid_t uid;
57 kgid_t gid;
58 loff_t size;
59 struct proc_dir_entry *parent;
60 struct rb_root subdir;
61 struct rb_node subdir_node;
62 char *name;
63 umode_t mode;
64 u8 flags;
65 u8 namelen;
66 char inline_name[];
67 } __randomize_layout;
68
69 #define SIZEOF_PDE ( \
70 sizeof(struct proc_dir_entry) < 128 ? 128 : \
71 sizeof(struct proc_dir_entry) < 192 ? 192 : \
72 sizeof(struct proc_dir_entry) < 256 ? 256 : \
73 sizeof(struct proc_dir_entry) < 512 ? 512 : \
74 0)
75 #define SIZEOF_PDE_INLINE_NAME (SIZEOF_PDE - sizeof(struct proc_dir_entry))
76
pde_is_permanent(const struct proc_dir_entry * pde)77 static inline bool pde_is_permanent(const struct proc_dir_entry *pde)
78 {
79 return pde->flags & PROC_ENTRY_PERMANENT;
80 }
81
pde_make_permanent(struct proc_dir_entry * pde)82 static inline void pde_make_permanent(struct proc_dir_entry *pde)
83 {
84 pde->flags |= PROC_ENTRY_PERMANENT;
85 }
86
pde_has_proc_read_iter(const struct proc_dir_entry * pde)87 static inline bool pde_has_proc_read_iter(const struct proc_dir_entry *pde)
88 {
89 return pde->flags & PROC_ENTRY_proc_read_iter;
90 }
91
pde_has_proc_compat_ioctl(const struct proc_dir_entry * pde)92 static inline bool pde_has_proc_compat_ioctl(const struct proc_dir_entry *pde)
93 {
94 #ifdef CONFIG_COMPAT
95 return pde->flags & PROC_ENTRY_proc_compat_ioctl;
96 #else
97 return false;
98 #endif
99 }
100
pde_has_proc_lseek(const struct proc_dir_entry * pde)101 static inline bool pde_has_proc_lseek(const struct proc_dir_entry *pde)
102 {
103 return pde->flags & PROC_ENTRY_proc_lseek;
104 }
105
106 extern struct kmem_cache *proc_dir_entry_cache;
107 void pde_free(struct proc_dir_entry *pde);
108
109 union proc_op {
110 int (*proc_get_link)(struct dentry *, struct path *);
111 int (*proc_show)(struct seq_file *m,
112 struct pid_namespace *ns, struct pid *pid,
113 struct task_struct *task);
114 int lsmid;
115 };
116
117 struct proc_inode {
118 struct pid *pid;
119 unsigned int fd;
120 union proc_op op;
121 struct proc_dir_entry *pde;
122 struct ctl_table_header *sysctl;
123 const struct ctl_table *sysctl_entry;
124 struct hlist_node sibling_inodes;
125 const struct proc_ns_operations *ns_ops;
126 struct inode vfs_inode;
127 } __randomize_layout;
128
129 /*
130 * General functions
131 */
PROC_I(const struct inode * inode)132 static inline struct proc_inode *PROC_I(const struct inode *inode)
133 {
134 return container_of(inode, struct proc_inode, vfs_inode);
135 }
136
PDE(const struct inode * inode)137 static inline struct proc_dir_entry *PDE(const struct inode *inode)
138 {
139 return PROC_I(inode)->pde;
140 }
141
proc_pid(const struct inode * inode)142 static inline struct pid *proc_pid(const struct inode *inode)
143 {
144 return PROC_I(inode)->pid;
145 }
146
get_proc_task(const struct inode * inode)147 static inline struct task_struct *get_proc_task(const struct inode *inode)
148 {
149 return get_pid_task(proc_pid(inode), PIDTYPE_PID);
150 }
151
152 void task_dump_owner(struct task_struct *task, umode_t mode,
153 kuid_t *ruid, kgid_t *rgid);
154
155 unsigned name_to_int(const struct qstr *qstr);
156 /*
157 * Offset of the first process in the /proc root directory..
158 */
159 #define FIRST_PROCESS_ENTRY 256
160
161 /* Worst case buffer size needed for holding an integer. */
162 #define PROC_NUMBUF 13
163
164 #ifdef CONFIG_PAGE_MAPCOUNT
165 /**
166 * folio_precise_page_mapcount() - Number of mappings of this folio page.
167 * @folio: The folio.
168 * @page: The page.
169 *
170 * The number of present user page table entries that reference this page
171 * as tracked via the RMAP: either referenced directly (PTE) or as part of
172 * a larger area that covers this page (e.g., PMD).
173 *
174 * Use this function only for the calculation of existing statistics
175 * (USS, PSS, mapcount_max) and for debugging purposes (/proc/kpagecount).
176 *
177 * Do not add new users.
178 *
179 * Returns: The number of mappings of this folio page. 0 for
180 * folios that are not mapped to user space or are not tracked via the RMAP
181 * (e.g., shared zeropage).
182 */
folio_precise_page_mapcount(struct folio * folio,struct page * page)183 static inline int folio_precise_page_mapcount(struct folio *folio,
184 struct page *page)
185 {
186 int mapcount = atomic_read(&page->_mapcount) + 1;
187
188 if (page_mapcount_is_type(mapcount))
189 mapcount = 0;
190 if (folio_test_large(folio))
191 mapcount += folio_entire_mapcount(folio);
192
193 return mapcount;
194 }
195 #else /* !CONFIG_PAGE_MAPCOUNT */
folio_precise_page_mapcount(struct folio * folio,struct page * page)196 static inline int folio_precise_page_mapcount(struct folio *folio,
197 struct page *page)
198 {
199 BUILD_BUG();
200 }
201 #endif /* CONFIG_PAGE_MAPCOUNT */
202
203 /**
204 * folio_average_page_mapcount() - Average number of mappings per page in this
205 * folio
206 * @folio: The folio.
207 *
208 * The average number of user page table entries that reference each page in
209 * this folio as tracked via the RMAP: either referenced directly (PTE) or
210 * as part of a larger area that covers this page (e.g., PMD).
211 *
212 * The average is calculated by rounding to the nearest integer; however,
213 * to avoid duplicated code in current callers, the average is at least
214 * 1 if any page of the folio is mapped.
215 *
216 * Returns: The average number of mappings per page in this folio.
217 */
folio_average_page_mapcount(struct folio * folio)218 static inline int folio_average_page_mapcount(struct folio *folio)
219 {
220 int mapcount, entire_mapcount, avg;
221
222 if (!folio_test_large(folio))
223 return atomic_read(&folio->_mapcount) + 1;
224
225 mapcount = folio_large_mapcount(folio);
226 if (unlikely(mapcount <= 0))
227 return 0;
228 entire_mapcount = folio_entire_mapcount(folio);
229 if (mapcount <= entire_mapcount)
230 return entire_mapcount;
231 mapcount -= entire_mapcount;
232
233 /* Round to closest integer ... */
234 avg = ((unsigned int)mapcount + folio_large_nr_pages(folio) / 2) >> folio_large_order(folio);
235 /* ... but return at least 1. */
236 return max_t(int, avg + entire_mapcount, 1);
237 }
238 /*
239 * array.c
240 */
241 extern const struct file_operations proc_tid_children_operations;
242
243 extern void proc_task_name(struct seq_file *m, struct task_struct *p,
244 bool escape);
245 extern int proc_tid_stat(struct seq_file *, struct pid_namespace *,
246 struct pid *, struct task_struct *);
247 extern int proc_tgid_stat(struct seq_file *, struct pid_namespace *,
248 struct pid *, struct task_struct *);
249 extern int proc_pid_status(struct seq_file *, struct pid_namespace *,
250 struct pid *, struct task_struct *);
251 extern int proc_pid_statm(struct seq_file *, struct pid_namespace *,
252 struct pid *, struct task_struct *);
253
254 /*
255 * base.c
256 */
257 extern const struct dentry_operations pid_dentry_operations;
258 extern int pid_getattr(struct mnt_idmap *, const struct path *,
259 struct kstat *, u32, unsigned int);
260 extern int proc_setattr(struct mnt_idmap *, struct dentry *,
261 struct iattr *);
262 extern void proc_pid_evict_inode(struct proc_inode *);
263 extern struct inode *proc_pid_make_inode(struct super_block *, struct task_struct *, umode_t);
264 extern void pid_update_inode(struct task_struct *, struct inode *);
265 extern int pid_delete_dentry(const struct dentry *);
266 extern int proc_pid_readdir(struct file *, struct dir_context *);
267 struct dentry *proc_pid_lookup(struct dentry *, unsigned int);
268 extern loff_t mem_lseek(struct file *, loff_t, int);
269
270 /* Lookups */
271 typedef struct dentry *instantiate_t(struct dentry *,
272 struct task_struct *, const void *);
273 bool proc_fill_cache(struct file *, struct dir_context *, const char *, unsigned int,
274 instantiate_t, struct task_struct *, const void *);
275
276 /*
277 * generic.c
278 */
279 struct proc_dir_entry *proc_create_reg(const char *name, umode_t mode,
280 struct proc_dir_entry **parent, void *data);
281 struct proc_dir_entry *proc_register(struct proc_dir_entry *dir,
282 struct proc_dir_entry *dp);
283 extern struct dentry *proc_lookup(struct inode *, struct dentry *, unsigned int);
284 struct dentry *proc_lookup_de(struct inode *, struct dentry *, struct proc_dir_entry *);
285 extern int proc_readdir(struct file *, struct dir_context *);
286 int proc_readdir_de(struct file *, struct dir_context *, struct proc_dir_entry *);
287
pde_get(struct proc_dir_entry * pde)288 static inline void pde_get(struct proc_dir_entry *pde)
289 {
290 refcount_inc(&pde->refcnt);
291 }
292 extern void pde_put(struct proc_dir_entry *);
293
is_empty_pde(const struct proc_dir_entry * pde)294 static inline bool is_empty_pde(const struct proc_dir_entry *pde)
295 {
296 return S_ISDIR(pde->mode) && !pde->proc_iops;
297 }
298 extern ssize_t proc_simple_write(struct file *, const char __user *, size_t, loff_t *);
299
300 /*
301 * inode.c
302 */
303 struct pde_opener {
304 struct list_head lh;
305 struct file *file;
306 bool closing;
307 struct completion *c;
308 } __randomize_layout;
309 extern const struct inode_operations proc_link_inode_operations;
310 extern const struct inode_operations proc_pid_link_inode_operations;
311 extern const struct super_operations proc_sops;
312
313 void proc_init_kmemcache(void);
314 void proc_invalidate_siblings_dcache(struct hlist_head *inodes, spinlock_t *lock);
315 void set_proc_pid_nlink(void);
316 extern struct inode *proc_get_inode(struct super_block *, struct proc_dir_entry *);
317 extern void proc_entry_rundown(struct proc_dir_entry *);
318
319 /*
320 * proc_namespaces.c
321 */
322 extern const struct inode_operations proc_ns_dir_inode_operations;
323 extern const struct file_operations proc_ns_dir_operations;
324
325 /*
326 * proc_net.c
327 */
328 extern const struct file_operations proc_net_operations;
329 extern const struct inode_operations proc_net_inode_operations;
330
331 #ifdef CONFIG_NET
332 extern int proc_net_init(void);
333 #else
proc_net_init(void)334 static inline int proc_net_init(void) { return 0; }
335 #endif
336
337 /*
338 * proc_self.c
339 */
340 extern int proc_setup_self(struct super_block *);
341
342 /*
343 * proc_thread_self.c
344 */
345 extern int proc_setup_thread_self(struct super_block *);
346 extern void proc_thread_self_init(void);
347
348 /*
349 * proc_sysctl.c
350 */
351 #ifdef CONFIG_PROC_SYSCTL
352 extern int proc_sys_init(void);
353 extern void proc_sys_evict_inode(struct inode *inode,
354 struct ctl_table_header *head);
355 #else
proc_sys_init(void)356 static inline void proc_sys_init(void) { }
proc_sys_evict_inode(struct inode * inode,struct ctl_table_header * head)357 static inline void proc_sys_evict_inode(struct inode *inode,
358 struct ctl_table_header *head) { }
359 #endif
360
361 /*
362 * proc_tty.c
363 */
364 #ifdef CONFIG_TTY
365 extern void proc_tty_init(void);
366 #else
proc_tty_init(void)367 static inline void proc_tty_init(void) {}
368 #endif
369
370 /*
371 * root.c
372 */
373 extern struct proc_dir_entry proc_root;
374
375 extern void proc_self_init(void);
376
377 /*
378 * task_[no]mmu.c
379 */
380 struct mem_size_stats;
381 struct proc_maps_private {
382 struct inode *inode;
383 struct task_struct *task;
384 struct mm_struct *mm;
385 struct vma_iterator iter;
386 loff_t last_pos;
387 #ifdef CONFIG_PER_VMA_LOCK
388 bool mmap_locked;
389 struct vm_area_struct *locked_vma;
390 #endif
391 #ifdef CONFIG_NUMA
392 struct mempolicy *task_mempolicy;
393 #endif
394 } __randomize_layout;
395
396 struct mm_struct *proc_mem_open(struct inode *inode, unsigned int mode);
397
398 extern const struct file_operations proc_pid_maps_operations;
399 extern const struct file_operations proc_pid_numa_maps_operations;
400 extern const struct file_operations proc_pid_smaps_operations;
401 extern const struct file_operations proc_pid_smaps_rollup_operations;
402 extern const struct file_operations proc_clear_refs_operations;
403 extern const struct file_operations proc_pagemap_operations;
404
405 extern unsigned long task_vsize(struct mm_struct *);
406 extern unsigned long task_statm(struct mm_struct *,
407 unsigned long *, unsigned long *,
408 unsigned long *, unsigned long *);
409 extern void task_mem(struct seq_file *, struct mm_struct *);
410
411 extern const struct dentry_operations proc_net_dentry_ops;
pde_force_lookup(struct proc_dir_entry * pde)412 static inline void pde_force_lookup(struct proc_dir_entry *pde)
413 {
414 /* /proc/net/ entries can be changed under us by setns(CLONE_NEWNET) */
415 pde->flags |= PROC_ENTRY_FORCE_LOOKUP;
416 }
417
418 /*
419 * Add a new procfs dentry that can't serve as a mountpoint. That should
420 * encompass anything that is ephemeral and can just disappear while the
421 * process is still around.
422 */
proc_splice_unmountable(struct inode * inode,struct dentry * dentry,const struct dentry_operations * d_ops)423 static inline struct dentry *proc_splice_unmountable(struct inode *inode,
424 struct dentry *dentry, const struct dentry_operations *d_ops)
425 {
426 dont_mount(dentry);
427 return d_splice_alias_ops(inode, dentry, d_ops);
428 }
429