xref: /linux/fs/proc/internal.h (revision 260f6f4fda93c8485c8037865c941b42b9cba5d2)
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 
77 static inline bool pde_is_permanent(const struct proc_dir_entry *pde)
78 {
79 	return pde->flags & PROC_ENTRY_PERMANENT;
80 }
81 
82 static inline void pde_make_permanent(struct proc_dir_entry *pde)
83 {
84 	pde->flags |= PROC_ENTRY_PERMANENT;
85 }
86 
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 
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 
101 extern struct kmem_cache *proc_dir_entry_cache;
102 void pde_free(struct proc_dir_entry *pde);
103 
104 union proc_op {
105 	int (*proc_get_link)(struct dentry *, struct path *);
106 	int (*proc_show)(struct seq_file *m,
107 		struct pid_namespace *ns, struct pid *pid,
108 		struct task_struct *task);
109 	int lsmid;
110 };
111 
112 struct proc_inode {
113 	struct pid *pid;
114 	unsigned int fd;
115 	union proc_op op;
116 	struct proc_dir_entry *pde;
117 	struct ctl_table_header *sysctl;
118 	const struct ctl_table *sysctl_entry;
119 	struct hlist_node sibling_inodes;
120 	const struct proc_ns_operations *ns_ops;
121 	struct inode vfs_inode;
122 } __randomize_layout;
123 
124 /*
125  * General functions
126  */
127 static inline struct proc_inode *PROC_I(const struct inode *inode)
128 {
129 	return container_of(inode, struct proc_inode, vfs_inode);
130 }
131 
132 static inline struct proc_dir_entry *PDE(const struct inode *inode)
133 {
134 	return PROC_I(inode)->pde;
135 }
136 
137 static inline struct pid *proc_pid(const struct inode *inode)
138 {
139 	return PROC_I(inode)->pid;
140 }
141 
142 static inline struct task_struct *get_proc_task(const struct inode *inode)
143 {
144 	return get_pid_task(proc_pid(inode), PIDTYPE_PID);
145 }
146 
147 void task_dump_owner(struct task_struct *task, umode_t mode,
148 		     kuid_t *ruid, kgid_t *rgid);
149 
150 unsigned name_to_int(const struct qstr *qstr);
151 /*
152  * Offset of the first process in the /proc root directory..
153  */
154 #define FIRST_PROCESS_ENTRY 256
155 
156 /* Worst case buffer size needed for holding an integer. */
157 #define PROC_NUMBUF 13
158 
159 #ifdef CONFIG_PAGE_MAPCOUNT
160 /**
161  * folio_precise_page_mapcount() - Number of mappings of this folio page.
162  * @folio: The folio.
163  * @page: The page.
164  *
165  * The number of present user page table entries that reference this page
166  * as tracked via the RMAP: either referenced directly (PTE) or as part of
167  * a larger area that covers this page (e.g., PMD).
168  *
169  * Use this function only for the calculation of existing statistics
170  * (USS, PSS, mapcount_max) and for debugging purposes (/proc/kpagecount).
171  *
172  * Do not add new users.
173  *
174  * Returns: The number of mappings of this folio page. 0 for
175  * folios that are not mapped to user space or are not tracked via the RMAP
176  * (e.g., shared zeropage).
177  */
178 static inline int folio_precise_page_mapcount(struct folio *folio,
179 		struct page *page)
180 {
181 	int mapcount = atomic_read(&page->_mapcount) + 1;
182 
183 	if (page_mapcount_is_type(mapcount))
184 		mapcount = 0;
185 	if (folio_test_large(folio))
186 		mapcount += folio_entire_mapcount(folio);
187 
188 	return mapcount;
189 }
190 #else /* !CONFIG_PAGE_MAPCOUNT */
191 static inline int folio_precise_page_mapcount(struct folio *folio,
192 		struct page *page)
193 {
194 	BUILD_BUG();
195 }
196 #endif /* CONFIG_PAGE_MAPCOUNT */
197 
198 /**
199  * folio_average_page_mapcount() - Average number of mappings per page in this
200  *				   folio
201  * @folio: The folio.
202  *
203  * The average number of user page table entries that reference each page in
204  * this folio as tracked via the RMAP: either referenced directly (PTE) or
205  * as part of a larger area that covers this page (e.g., PMD).
206  *
207  * The average is calculated by rounding to the nearest integer; however,
208  * to avoid duplicated code in current callers, the average is at least
209  * 1 if any page of the folio is mapped.
210  *
211  * Returns: The average number of mappings per page in this folio.
212  */
213 static inline int folio_average_page_mapcount(struct folio *folio)
214 {
215 	int mapcount, entire_mapcount, avg;
216 
217 	if (!folio_test_large(folio))
218 		return atomic_read(&folio->_mapcount) + 1;
219 
220 	mapcount = folio_large_mapcount(folio);
221 	if (unlikely(mapcount <= 0))
222 		return 0;
223 	entire_mapcount = folio_entire_mapcount(folio);
224 	if (mapcount <= entire_mapcount)
225 		return entire_mapcount;
226 	mapcount -= entire_mapcount;
227 
228 	/* Round to closest integer ... */
229 	avg = ((unsigned int)mapcount + folio_large_nr_pages(folio) / 2) >> folio_large_order(folio);
230 	/* ... but return at least 1. */
231 	return max_t(int, avg + entire_mapcount, 1);
232 }
233 /*
234  * array.c
235  */
236 extern const struct file_operations proc_tid_children_operations;
237 
238 extern void proc_task_name(struct seq_file *m, struct task_struct *p,
239 			   bool escape);
240 extern int proc_tid_stat(struct seq_file *, struct pid_namespace *,
241 			 struct pid *, struct task_struct *);
242 extern int proc_tgid_stat(struct seq_file *, struct pid_namespace *,
243 			  struct pid *, struct task_struct *);
244 extern int proc_pid_status(struct seq_file *, struct pid_namespace *,
245 			   struct pid *, struct task_struct *);
246 extern int proc_pid_statm(struct seq_file *, struct pid_namespace *,
247 			  struct pid *, struct task_struct *);
248 
249 /*
250  * base.c
251  */
252 extern const struct dentry_operations pid_dentry_operations;
253 extern int pid_getattr(struct mnt_idmap *, const struct path *,
254 		       struct kstat *, u32, unsigned int);
255 extern int proc_setattr(struct mnt_idmap *, struct dentry *,
256 			struct iattr *);
257 extern void proc_pid_evict_inode(struct proc_inode *);
258 extern struct inode *proc_pid_make_inode(struct super_block *, struct task_struct *, umode_t);
259 extern void pid_update_inode(struct task_struct *, struct inode *);
260 extern int pid_delete_dentry(const struct dentry *);
261 extern int proc_pid_readdir(struct file *, struct dir_context *);
262 struct dentry *proc_pid_lookup(struct dentry *, unsigned int);
263 extern loff_t mem_lseek(struct file *, loff_t, int);
264 
265 /* Lookups */
266 typedef struct dentry *instantiate_t(struct dentry *,
267 				     struct task_struct *, const void *);
268 bool proc_fill_cache(struct file *, struct dir_context *, const char *, unsigned int,
269 			   instantiate_t, struct task_struct *, const void *);
270 
271 /*
272  * generic.c
273  */
274 struct proc_dir_entry *proc_create_reg(const char *name, umode_t mode,
275 		struct proc_dir_entry **parent, void *data);
276 struct proc_dir_entry *proc_register(struct proc_dir_entry *dir,
277 		struct proc_dir_entry *dp);
278 extern struct dentry *proc_lookup(struct inode *, struct dentry *, unsigned int);
279 struct dentry *proc_lookup_de(struct inode *, struct dentry *, struct proc_dir_entry *);
280 extern int proc_readdir(struct file *, struct dir_context *);
281 int proc_readdir_de(struct file *, struct dir_context *, struct proc_dir_entry *);
282 
283 static inline void pde_get(struct proc_dir_entry *pde)
284 {
285 	refcount_inc(&pde->refcnt);
286 }
287 extern void pde_put(struct proc_dir_entry *);
288 
289 static inline bool is_empty_pde(const struct proc_dir_entry *pde)
290 {
291 	return S_ISDIR(pde->mode) && !pde->proc_iops;
292 }
293 extern ssize_t proc_simple_write(struct file *, const char __user *, size_t, loff_t *);
294 
295 /*
296  * inode.c
297  */
298 struct pde_opener {
299 	struct list_head lh;
300 	struct file *file;
301 	bool closing;
302 	struct completion *c;
303 } __randomize_layout;
304 extern const struct inode_operations proc_link_inode_operations;
305 extern const struct inode_operations proc_pid_link_inode_operations;
306 extern const struct super_operations proc_sops;
307 
308 void proc_init_kmemcache(void);
309 void proc_invalidate_siblings_dcache(struct hlist_head *inodes, spinlock_t *lock);
310 void set_proc_pid_nlink(void);
311 extern struct inode *proc_get_inode(struct super_block *, struct proc_dir_entry *);
312 extern void proc_entry_rundown(struct proc_dir_entry *);
313 
314 /*
315  * proc_namespaces.c
316  */
317 extern const struct inode_operations proc_ns_dir_inode_operations;
318 extern const struct file_operations proc_ns_dir_operations;
319 
320 /*
321  * proc_net.c
322  */
323 extern const struct file_operations proc_net_operations;
324 extern const struct inode_operations proc_net_inode_operations;
325 
326 #ifdef CONFIG_NET
327 extern int proc_net_init(void);
328 #else
329 static inline int proc_net_init(void) { return 0; }
330 #endif
331 
332 /*
333  * proc_self.c
334  */
335 extern int proc_setup_self(struct super_block *);
336 
337 /*
338  * proc_thread_self.c
339  */
340 extern int proc_setup_thread_self(struct super_block *);
341 extern void proc_thread_self_init(void);
342 
343 /*
344  * proc_sysctl.c
345  */
346 #ifdef CONFIG_PROC_SYSCTL
347 extern int proc_sys_init(void);
348 extern void proc_sys_evict_inode(struct inode *inode,
349 				 struct ctl_table_header *head);
350 #else
351 static inline void proc_sys_init(void) { }
352 static inline void proc_sys_evict_inode(struct  inode *inode,
353 					struct ctl_table_header *head) { }
354 #endif
355 
356 /*
357  * proc_tty.c
358  */
359 #ifdef CONFIG_TTY
360 extern void proc_tty_init(void);
361 #else
362 static inline void proc_tty_init(void) {}
363 #endif
364 
365 /*
366  * root.c
367  */
368 extern struct proc_dir_entry proc_root;
369 
370 extern void proc_self_init(void);
371 
372 /*
373  * task_[no]mmu.c
374  */
375 struct mem_size_stats;
376 struct proc_maps_private {
377 	struct inode *inode;
378 	struct task_struct *task;
379 	struct mm_struct *mm;
380 	struct vma_iterator iter;
381 #ifdef CONFIG_NUMA
382 	struct mempolicy *task_mempolicy;
383 #endif
384 } __randomize_layout;
385 
386 struct mm_struct *proc_mem_open(struct inode *inode, unsigned int mode);
387 
388 extern const struct file_operations proc_pid_maps_operations;
389 extern const struct file_operations proc_pid_numa_maps_operations;
390 extern const struct file_operations proc_pid_smaps_operations;
391 extern const struct file_operations proc_pid_smaps_rollup_operations;
392 extern const struct file_operations proc_clear_refs_operations;
393 extern const struct file_operations proc_pagemap_operations;
394 
395 extern unsigned long task_vsize(struct mm_struct *);
396 extern unsigned long task_statm(struct mm_struct *,
397 				unsigned long *, unsigned long *,
398 				unsigned long *, unsigned long *);
399 extern void task_mem(struct seq_file *, struct mm_struct *);
400 
401 extern const struct dentry_operations proc_net_dentry_ops;
402 static inline void pde_force_lookup(struct proc_dir_entry *pde)
403 {
404 	/* /proc/net/ entries can be changed under us by setns(CLONE_NEWNET) */
405 	pde->flags |= PROC_ENTRY_FORCE_LOOKUP;
406 }
407 
408 /*
409  * Add a new procfs dentry that can't serve as a mountpoint. That should
410  * encompass anything that is ephemeral and can just disappear while the
411  * process is still around.
412  */
413 static inline struct dentry *proc_splice_unmountable(struct inode *inode,
414 		struct dentry *dentry, const struct dentry_operations *d_ops)
415 {
416 	dont_mount(dentry);
417 	return d_splice_alias_ops(inode, dentry, d_ops);
418 }
419