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