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  
pde_is_permanent(const struct proc_dir_entry * pde)78  static inline bool pde_is_permanent(const struct proc_dir_entry *pde)
79  {
80  	return pde->flags & PROC_ENTRY_PERMANENT;
81  }
82  
pde_make_permanent(struct proc_dir_entry * pde)83  static inline void pde_make_permanent(struct proc_dir_entry *pde)
84  {
85  	pde->flags |= PROC_ENTRY_PERMANENT;
86  }
87  
pde_has_proc_read_iter(const struct proc_dir_entry * pde)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  
pde_has_proc_compat_ioctl(const struct proc_dir_entry * pde)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   */
PROC_I(const struct inode * inode)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  
PDE(const struct inode * inode)133  static inline struct proc_dir_entry *PDE(const struct inode *inode)
134  {
135  	return PROC_I(inode)->pde;
136  }
137  
proc_pid(const struct inode * inode)138  static inline struct pid *proc_pid(const struct inode *inode)
139  {
140  	return PROC_I(inode)->pid;
141  }
142  
get_proc_task(const struct inode * inode)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  /**
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   */
folio_precise_page_mapcount(struct folio * folio,struct page * page)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  
191  /*
192   * array.c
193   */
194  extern const struct file_operations proc_tid_children_operations;
195  
196  extern void proc_task_name(struct seq_file *m, struct task_struct *p,
197  			   bool escape);
198  extern int proc_tid_stat(struct seq_file *, struct pid_namespace *,
199  			 struct pid *, struct task_struct *);
200  extern int proc_tgid_stat(struct seq_file *, struct pid_namespace *,
201  			  struct pid *, struct task_struct *);
202  extern int proc_pid_status(struct seq_file *, struct pid_namespace *,
203  			   struct pid *, struct task_struct *);
204  extern int proc_pid_statm(struct seq_file *, struct pid_namespace *,
205  			  struct pid *, struct task_struct *);
206  
207  /*
208   * base.c
209   */
210  extern const struct dentry_operations pid_dentry_operations;
211  extern int pid_getattr(struct mnt_idmap *, const struct path *,
212  		       struct kstat *, u32, unsigned int);
213  extern int proc_setattr(struct mnt_idmap *, struct dentry *,
214  			struct iattr *);
215  extern void proc_pid_evict_inode(struct proc_inode *);
216  extern struct inode *proc_pid_make_inode(struct super_block *, struct task_struct *, umode_t);
217  extern void pid_update_inode(struct task_struct *, struct inode *);
218  extern int pid_delete_dentry(const struct dentry *);
219  extern int proc_pid_readdir(struct file *, struct dir_context *);
220  struct dentry *proc_pid_lookup(struct dentry *, unsigned int);
221  extern loff_t mem_lseek(struct file *, loff_t, int);
222  
223  /* Lookups */
224  typedef struct dentry *instantiate_t(struct dentry *,
225  				     struct task_struct *, const void *);
226  bool proc_fill_cache(struct file *, struct dir_context *, const char *, unsigned int,
227  			   instantiate_t, struct task_struct *, const void *);
228  
229  /*
230   * generic.c
231   */
232  struct proc_dir_entry *proc_create_reg(const char *name, umode_t mode,
233  		struct proc_dir_entry **parent, void *data);
234  struct proc_dir_entry *proc_register(struct proc_dir_entry *dir,
235  		struct proc_dir_entry *dp);
236  extern struct dentry *proc_lookup(struct inode *, struct dentry *, unsigned int);
237  struct dentry *proc_lookup_de(struct inode *, struct dentry *, struct proc_dir_entry *);
238  extern int proc_readdir(struct file *, struct dir_context *);
239  int proc_readdir_de(struct file *, struct dir_context *, struct proc_dir_entry *);
240  
pde_get(struct proc_dir_entry * pde)241  static inline void pde_get(struct proc_dir_entry *pde)
242  {
243  	refcount_inc(&pde->refcnt);
244  }
245  extern void pde_put(struct proc_dir_entry *);
246  
is_empty_pde(const struct proc_dir_entry * pde)247  static inline bool is_empty_pde(const struct proc_dir_entry *pde)
248  {
249  	return S_ISDIR(pde->mode) && !pde->proc_iops;
250  }
251  extern ssize_t proc_simple_write(struct file *, const char __user *, size_t, loff_t *);
252  
253  /*
254   * inode.c
255   */
256  struct pde_opener {
257  	struct list_head lh;
258  	struct file *file;
259  	bool closing;
260  	struct completion *c;
261  } __randomize_layout;
262  extern const struct inode_operations proc_link_inode_operations;
263  extern const struct inode_operations proc_pid_link_inode_operations;
264  extern const struct super_operations proc_sops;
265  
266  void proc_init_kmemcache(void);
267  void proc_invalidate_siblings_dcache(struct hlist_head *inodes, spinlock_t *lock);
268  void set_proc_pid_nlink(void);
269  extern struct inode *proc_get_inode(struct super_block *, struct proc_dir_entry *);
270  extern void proc_entry_rundown(struct proc_dir_entry *);
271  
272  /*
273   * proc_namespaces.c
274   */
275  extern const struct inode_operations proc_ns_dir_inode_operations;
276  extern const struct file_operations proc_ns_dir_operations;
277  
278  /*
279   * proc_net.c
280   */
281  extern const struct file_operations proc_net_operations;
282  extern const struct inode_operations proc_net_inode_operations;
283  
284  #ifdef CONFIG_NET
285  extern int proc_net_init(void);
286  #else
proc_net_init(void)287  static inline int proc_net_init(void) { return 0; }
288  #endif
289  
290  /*
291   * proc_self.c
292   */
293  extern int proc_setup_self(struct super_block *);
294  
295  /*
296   * proc_thread_self.c
297   */
298  extern int proc_setup_thread_self(struct super_block *);
299  extern void proc_thread_self_init(void);
300  
301  /*
302   * proc_sysctl.c
303   */
304  #ifdef CONFIG_PROC_SYSCTL
305  extern int proc_sys_init(void);
306  extern void proc_sys_evict_inode(struct inode *inode,
307  				 struct ctl_table_header *head);
308  #else
proc_sys_init(void)309  static inline void proc_sys_init(void) { }
proc_sys_evict_inode(struct inode * inode,struct ctl_table_header * head)310  static inline void proc_sys_evict_inode(struct  inode *inode,
311  					struct ctl_table_header *head) { }
312  #endif
313  
314  /*
315   * proc_tty.c
316   */
317  #ifdef CONFIG_TTY
318  extern void proc_tty_init(void);
319  #else
proc_tty_init(void)320  static inline void proc_tty_init(void) {}
321  #endif
322  
323  /*
324   * root.c
325   */
326  extern struct proc_dir_entry proc_root;
327  
328  extern void proc_self_init(void);
329  
330  /*
331   * task_[no]mmu.c
332   */
333  struct mem_size_stats;
334  struct proc_maps_private {
335  	struct inode *inode;
336  	struct task_struct *task;
337  	struct mm_struct *mm;
338  	struct vma_iterator iter;
339  #ifdef CONFIG_NUMA
340  	struct mempolicy *task_mempolicy;
341  #endif
342  } __randomize_layout;
343  
344  struct mm_struct *proc_mem_open(struct inode *inode, unsigned int mode);
345  
346  extern const struct file_operations proc_pid_maps_operations;
347  extern const struct file_operations proc_pid_numa_maps_operations;
348  extern const struct file_operations proc_pid_smaps_operations;
349  extern const struct file_operations proc_pid_smaps_rollup_operations;
350  extern const struct file_operations proc_clear_refs_operations;
351  extern const struct file_operations proc_pagemap_operations;
352  
353  extern unsigned long task_vsize(struct mm_struct *);
354  extern unsigned long task_statm(struct mm_struct *,
355  				unsigned long *, unsigned long *,
356  				unsigned long *, unsigned long *);
357  extern void task_mem(struct seq_file *, struct mm_struct *);
358  
359  extern const struct dentry_operations proc_net_dentry_ops;
pde_force_lookup(struct proc_dir_entry * pde)360  static inline void pde_force_lookup(struct proc_dir_entry *pde)
361  {
362  	/* /proc/net/ entries can be changed under us by setns(CLONE_NEWNET) */
363  	pde->proc_dops = &proc_net_dentry_ops;
364  }
365  
366  /*
367   * Add a new procfs dentry that can't serve as a mountpoint. That should
368   * encompass anything that is ephemeral and can just disappear while the
369   * process is still around.
370   */
proc_splice_unmountable(struct inode * inode,struct dentry * dentry,const struct dentry_operations * d_ops)371  static inline struct dentry *proc_splice_unmountable(struct inode *inode,
372  		struct dentry *dentry, const struct dentry_operations *d_ops)
373  {
374  	d_set_d_op(dentry, d_ops);
375  	dont_mount(dentry);
376  	return d_splice_alias(inode, dentry);
377  }
378