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 static inline bool pde_has_proc_lseek(const struct proc_dir_entry *pde) 103 { 104 return pde->flags & PROC_ENTRY_proc_lseek; 105 } 106 107 extern struct kmem_cache *proc_dir_entry_cache; 108 void pde_free(struct proc_dir_entry *pde); 109 110 union proc_op { 111 int (*proc_get_link)(struct dentry *, struct path *); 112 int (*proc_show)(struct seq_file *m, 113 struct pid_namespace *ns, struct pid *pid, 114 struct task_struct *task); 115 int lsmid; 116 }; 117 118 struct proc_inode { 119 struct pid *pid; 120 unsigned int fd; 121 union proc_op op; 122 struct proc_dir_entry *pde; 123 struct ctl_table_header *sysctl; 124 const struct ctl_table *sysctl_entry; 125 struct hlist_node sibling_inodes; 126 const struct proc_ns_operations *ns_ops; 127 struct inode vfs_inode; 128 } __randomize_layout; 129 130 /* 131 * General functions 132 */ 133 static inline struct proc_inode *PROC_I(const struct inode *inode) 134 { 135 return container_of(inode, struct proc_inode, vfs_inode); 136 } 137 138 static inline struct proc_dir_entry *PDE(const struct inode *inode) 139 { 140 return PROC_I(inode)->pde; 141 } 142 143 static inline struct pid *proc_pid(const struct inode *inode) 144 { 145 return PROC_I(inode)->pid; 146 } 147 148 static inline struct task_struct *get_proc_task(const struct inode *inode) 149 { 150 return get_pid_task(proc_pid(inode), PIDTYPE_PID); 151 } 152 153 void task_dump_owner(struct task_struct *task, umode_t mode, 154 kuid_t *ruid, kgid_t *rgid); 155 156 unsigned name_to_int(const struct qstr *qstr); 157 /* 158 * Offset of the first process in the /proc root directory.. 159 */ 160 #define FIRST_PROCESS_ENTRY 256 161 162 /* Worst case buffer size needed for holding an integer. */ 163 #define PROC_NUMBUF 13 164 165 #ifdef CONFIG_PAGE_MAPCOUNT 166 /** 167 * folio_precise_page_mapcount() - Number of mappings of this folio page. 168 * @folio: The folio. 169 * @page: The page. 170 * 171 * The number of present user page table entries that reference this page 172 * as tracked via the RMAP: either referenced directly (PTE) or as part of 173 * a larger area that covers this page (e.g., PMD). 174 * 175 * Use this function only for the calculation of existing statistics 176 * (USS, PSS, mapcount_max) and for debugging purposes (/proc/kpagecount). 177 * 178 * Do not add new users. 179 * 180 * Returns: The number of mappings of this folio page. 0 for 181 * folios that are not mapped to user space or are not tracked via the RMAP 182 * (e.g., shared zeropage). 183 */ 184 static inline int folio_precise_page_mapcount(struct folio *folio, 185 struct page *page) 186 { 187 int mapcount = atomic_read(&page->_mapcount) + 1; 188 189 if (page_mapcount_is_type(mapcount)) 190 mapcount = 0; 191 if (folio_test_large(folio)) 192 mapcount += folio_entire_mapcount(folio); 193 194 return mapcount; 195 } 196 #else /* !CONFIG_PAGE_MAPCOUNT */ 197 static inline int folio_precise_page_mapcount(struct folio *folio, 198 struct page *page) 199 { 200 BUILD_BUG(); 201 } 202 #endif /* CONFIG_PAGE_MAPCOUNT */ 203 204 /** 205 * folio_average_page_mapcount() - Average number of mappings per page in this 206 * folio 207 * @folio: The folio. 208 * 209 * The average number of user page table entries that reference each page in 210 * this folio as tracked via the RMAP: either referenced directly (PTE) or 211 * as part of a larger area that covers this page (e.g., PMD). 212 * 213 * The average is calculated by rounding to the nearest integer; however, 214 * to avoid duplicated code in current callers, the average is at least 215 * 1 if any page of the folio is mapped. 216 * 217 * Returns: The average number of mappings per page in this folio. 218 */ 219 static inline int folio_average_page_mapcount(struct folio *folio) 220 { 221 int mapcount, entire_mapcount, avg; 222 223 if (!folio_test_large(folio)) 224 return atomic_read(&folio->_mapcount) + 1; 225 226 mapcount = folio_large_mapcount(folio); 227 if (unlikely(mapcount <= 0)) 228 return 0; 229 entire_mapcount = folio_entire_mapcount(folio); 230 if (mapcount <= entire_mapcount) 231 return entire_mapcount; 232 mapcount -= entire_mapcount; 233 234 /* Round to closest integer ... */ 235 avg = ((unsigned int)mapcount + folio_large_nr_pages(folio) / 2) >> folio_large_order(folio); 236 /* ... but return at least 1. */ 237 return max_t(int, avg + entire_mapcount, 1); 238 } 239 /* 240 * array.c 241 */ 242 extern const struct file_operations proc_tid_children_operations; 243 244 extern void proc_task_name(struct seq_file *m, struct task_struct *p, 245 bool escape); 246 extern int proc_tid_stat(struct seq_file *, struct pid_namespace *, 247 struct pid *, struct task_struct *); 248 extern int proc_tgid_stat(struct seq_file *, struct pid_namespace *, 249 struct pid *, struct task_struct *); 250 extern int proc_pid_status(struct seq_file *, struct pid_namespace *, 251 struct pid *, struct task_struct *); 252 extern int proc_pid_statm(struct seq_file *, struct pid_namespace *, 253 struct pid *, struct task_struct *); 254 255 /* 256 * base.c 257 */ 258 extern const struct dentry_operations pid_dentry_operations; 259 extern int pid_getattr(struct mnt_idmap *, const struct path *, 260 struct kstat *, u32, unsigned int); 261 extern int proc_setattr(struct mnt_idmap *, struct dentry *, 262 struct iattr *); 263 extern void proc_pid_evict_inode(struct proc_inode *); 264 extern struct inode *proc_pid_make_inode(struct super_block *, struct task_struct *, umode_t); 265 extern void pid_update_inode(struct task_struct *, struct inode *); 266 extern int pid_delete_dentry(const struct dentry *); 267 extern int proc_pid_readdir(struct file *, struct dir_context *); 268 struct dentry *proc_pid_lookup(struct dentry *, unsigned int); 269 extern loff_t mem_lseek(struct file *, loff_t, int); 270 271 /* Lookups */ 272 typedef struct dentry *instantiate_t(struct dentry *, 273 struct task_struct *, const void *); 274 bool proc_fill_cache(struct file *, struct dir_context *, const char *, unsigned int, 275 instantiate_t, struct task_struct *, const void *); 276 277 /* 278 * generic.c 279 */ 280 struct proc_dir_entry *proc_create_reg(const char *name, umode_t mode, 281 struct proc_dir_entry **parent, void *data); 282 struct proc_dir_entry *proc_register(struct proc_dir_entry *dir, 283 struct proc_dir_entry *dp); 284 extern struct dentry *proc_lookup(struct inode *, struct dentry *, unsigned int); 285 struct dentry *proc_lookup_de(struct inode *, struct dentry *, struct proc_dir_entry *); 286 extern int proc_readdir(struct file *, struct dir_context *); 287 int proc_readdir_de(struct file *, struct dir_context *, struct proc_dir_entry *); 288 289 static inline void pde_get(struct proc_dir_entry *pde) 290 { 291 refcount_inc(&pde->refcnt); 292 } 293 extern void pde_put(struct proc_dir_entry *); 294 295 static inline bool is_empty_pde(const struct proc_dir_entry *pde) 296 { 297 return S_ISDIR(pde->mode) && !pde->proc_iops; 298 } 299 extern ssize_t proc_simple_write(struct file *, const char __user *, size_t, loff_t *); 300 301 /* 302 * inode.c 303 */ 304 struct pde_opener { 305 struct list_head lh; 306 struct file *file; 307 bool closing; 308 struct completion *c; 309 } __randomize_layout; 310 extern const struct inode_operations proc_link_inode_operations; 311 extern const struct inode_operations proc_pid_link_inode_operations; 312 extern const struct super_operations proc_sops; 313 314 void proc_init_kmemcache(void); 315 void proc_invalidate_siblings_dcache(struct hlist_head *inodes, spinlock_t *lock); 316 void set_proc_pid_nlink(void); 317 extern struct inode *proc_get_inode(struct super_block *, struct proc_dir_entry *); 318 extern void proc_entry_rundown(struct proc_dir_entry *); 319 320 /* 321 * proc_namespaces.c 322 */ 323 extern const struct inode_operations proc_ns_dir_inode_operations; 324 extern const struct file_operations proc_ns_dir_operations; 325 326 /* 327 * proc_net.c 328 */ 329 extern const struct file_operations proc_net_operations; 330 extern const struct inode_operations proc_net_inode_operations; 331 332 #ifdef CONFIG_NET 333 extern int proc_net_init(void); 334 #else 335 static inline int proc_net_init(void) { return 0; } 336 #endif 337 338 /* 339 * proc_self.c 340 */ 341 extern int proc_setup_self(struct super_block *); 342 343 /* 344 * proc_thread_self.c 345 */ 346 extern int proc_setup_thread_self(struct super_block *); 347 extern void proc_thread_self_init(void); 348 349 /* 350 * proc_sysctl.c 351 */ 352 #ifdef CONFIG_PROC_SYSCTL 353 extern int proc_sys_init(void); 354 extern void proc_sys_evict_inode(struct inode *inode, 355 struct ctl_table_header *head); 356 #else 357 static inline void proc_sys_init(void) { } 358 static inline void proc_sys_evict_inode(struct inode *inode, 359 struct ctl_table_header *head) { } 360 #endif 361 362 /* 363 * proc_tty.c 364 */ 365 #ifdef CONFIG_TTY 366 extern void proc_tty_init(void); 367 #else 368 static inline void proc_tty_init(void) {} 369 #endif 370 371 /* 372 * root.c 373 */ 374 extern struct proc_dir_entry proc_root; 375 376 extern void proc_self_init(void); 377 378 /* 379 * task_[no]mmu.c 380 */ 381 struct mem_size_stats; 382 struct proc_maps_private { 383 struct inode *inode; 384 struct task_struct *task; 385 struct mm_struct *mm; 386 struct vma_iterator iter; 387 #ifdef CONFIG_NUMA 388 struct mempolicy *task_mempolicy; 389 #endif 390 } __randomize_layout; 391 392 struct mm_struct *proc_mem_open(struct inode *inode, unsigned int mode); 393 394 extern const struct file_operations proc_pid_maps_operations; 395 extern const struct file_operations proc_pid_numa_maps_operations; 396 extern const struct file_operations proc_pid_smaps_operations; 397 extern const struct file_operations proc_pid_smaps_rollup_operations; 398 extern const struct file_operations proc_clear_refs_operations; 399 extern const struct file_operations proc_pagemap_operations; 400 401 extern unsigned long task_vsize(struct mm_struct *); 402 extern unsigned long task_statm(struct mm_struct *, 403 unsigned long *, unsigned long *, 404 unsigned long *, unsigned long *); 405 extern void task_mem(struct seq_file *, struct mm_struct *); 406 407 extern const struct dentry_operations proc_net_dentry_ops; 408 static inline void pde_force_lookup(struct proc_dir_entry *pde) 409 { 410 /* /proc/net/ entries can be changed under us by setns(CLONE_NEWNET) */ 411 pde->proc_dops = &proc_net_dentry_ops; 412 } 413 414 /* 415 * Add a new procfs dentry that can't serve as a mountpoint. That should 416 * encompass anything that is ephemeral and can just disappear while the 417 * process is still around. 418 */ 419 static inline struct dentry *proc_splice_unmountable(struct inode *inode, 420 struct dentry *dentry, const struct dentry_operations *d_ops) 421 { 422 d_set_d_op(dentry, d_ops); 423 dont_mount(dentry); 424 return d_splice_alias(inode, dentry); 425 } 426