1/* SPDX-License-Identifier: GPL-2.0 */
2#ifndef _LINUX_FS_H
3#define _LINUX_FS_H
4
5#include <linux/linkage.h>
6#include <linux/wait_bit.h>
7#include <linux/kdev_t.h>
8#include <linux/dcache.h>
9#include <linux/path.h>
10#include <linux/stat.h>
11#include <linux/cache.h>
12#include <linux/list.h>
13#include <linux/list_lru.h>
14#include <linux/llist.h>
15#include <linux/radix-tree.h>
16#include <linux/xarray.h>
17#include <linux/rbtree.h>
18#include <linux/init.h>
19#include <linux/pid.h>
20#include <linux/bug.h>
21#include <linux/mutex.h>
22#include <linux/rwsem.h>
23#include <linux/mm_types.h>
24#include <linux/capability.h>
25#include <linux/semaphore.h>
26#include <linux/fcntl.h>
27#include <linux/fiemap.h>
28#include <linux/rculist_bl.h>
29#include <linux/atomic.h>
30#include <linux/shrinker.h>
31#include <linux/migrate_mode.h>
32#include <linux/uidgid.h>
33#include <linux/lockdep.h>
34#include <linux/percpu-rwsem.h>
35#include <linux/workqueue.h>
36#include <linux/delayed_call.h>
37#include <linux/uuid.h>
38#include <linux/errseq.h>
39#include <linux/ioprio.h>
40
41#include <asm/byteorder.h>
42#include <uapi/linux/fs.h>
43
44struct backing_dev_info;
45struct bdi_writeback;
46struct bio;
47struct export_operations;
48struct hd_geometry;
49struct iovec;
50struct kiocb;
51struct kobject;
52struct pipe_inode_info;
53struct poll_table_struct;
54struct kstatfs;
55struct vm_area_struct;
56struct vfsmount;
57struct cred;
58struct swap_info_struct;
59struct seq_file;
60struct workqueue_struct;
61struct iov_iter;
62struct fscrypt_info;
63struct fscrypt_operations;
64
65extern void __init inode_init(void);
66extern void __init inode_init_early(void);
67extern void __init files_init(void);
68extern void __init files_maxfiles_init(void);
69
70extern struct files_stat_struct files_stat;
71extern unsigned long get_max_files(void);
72extern unsigned int sysctl_nr_open;
73extern struct inodes_stat_t inodes_stat;
74extern int leases_enable, lease_break_time;
75extern int sysctl_protected_symlinks;
76extern int sysctl_protected_hardlinks;
77
78typedef __kernel_rwf_t rwf_t;
79
80struct buffer_head;
81typedef int (get_block_t)(struct inode *inode, sector_t iblock,
82 struct buffer_head *bh_result, int create);
83typedef int (dio_iodone_t)(struct kiocb *iocb, loff_t offset,
84 ssize_t bytes, void *private);
85
86#define MAY_EXEC 0x00000001
87#define MAY_WRITE 0x00000002
88#define MAY_READ 0x00000004
89#define MAY_APPEND 0x00000008
90#define MAY_ACCESS 0x00000010
91#define MAY_OPEN 0x00000020
92#define MAY_CHDIR 0x00000040
93/* called from RCU mode, don't block */
94#define MAY_NOT_BLOCK 0x00000080
95
96/*
97 * flags in file.f_mode. Note that FMODE_READ and FMODE_WRITE must correspond
98 * to O_WRONLY and O_RDWR via the strange trick in do_dentry_open()
99 */
100
101/* file is open for reading */
102#define FMODE_READ ((__force fmode_t)0x1)
103/* file is open for writing */
104#define FMODE_WRITE ((__force fmode_t)0x2)
105/* file is seekable */
106#define FMODE_LSEEK ((__force fmode_t)0x4)
107/* file can be accessed using pread */
108#define FMODE_PREAD ((__force fmode_t)0x8)
109/* file can be accessed using pwrite */
110#define FMODE_PWRITE ((__force fmode_t)0x10)
111/* File is opened for execution with sys_execve / sys_uselib */
112#define FMODE_EXEC ((__force fmode_t)0x20)
113/* File is opened with O_NDELAY (only set for block devices) */
114#define FMODE_NDELAY ((__force fmode_t)0x40)
115/* File is opened with O_EXCL (only set for block devices) */
116#define FMODE_EXCL ((__force fmode_t)0x80)
117/* File is opened using open(.., 3, ..) and is writeable only for ioctls
118 (specialy hack for floppy.c) */
119#define FMODE_WRITE_IOCTL ((__force fmode_t)0x100)
120/* 32bit hashes as llseek() offset (for directories) */
121#define FMODE_32BITHASH ((__force fmode_t)0x200)
122/* 64bit hashes as llseek() offset (for directories) */
123#define FMODE_64BITHASH ((__force fmode_t)0x400)
124
125/*
126 * Don't update ctime and mtime.
127 *
128 * Currently a special hack for the XFS open_by_handle ioctl, but we'll
129 * hopefully graduate it to a proper O_CMTIME flag supported by open(2) soon.
130 */
131#define FMODE_NOCMTIME ((__force fmode_t)0x800)
132
133/* Expect random access pattern */
134#define FMODE_RANDOM ((__force fmode_t)0x1000)
135
136/* File is huge (eg. /dev/kmem): treat loff_t as unsigned */
137#define FMODE_UNSIGNED_OFFSET ((__force fmode_t)0x2000)
138
139/* File is opened with O_PATH; almost nothing can be done with it */
140#define FMODE_PATH ((__force fmode_t)0x4000)
141
142/* File needs atomic accesses to f_pos */
143#define FMODE_ATOMIC_POS ((__force fmode_t)0x8000)
144/* Write access to underlying fs */
145#define FMODE_WRITER ((__force fmode_t)0x10000)
146/* Has read method(s) */
147#define FMODE_CAN_READ ((__force fmode_t)0x20000)
148/* Has write method(s) */
149#define FMODE_CAN_WRITE ((__force fmode_t)0x40000)
150
151#define FMODE_OPENED ((__force fmode_t)0x80000)
152#define FMODE_CREATED ((__force fmode_t)0x100000)
153
154/* File was opened by fanotify and shouldn't generate fanotify events */
155#define FMODE_NONOTIFY ((__force fmode_t)0x4000000)
156
157/* File is capable of returning -EAGAIN if I/O will block */
158#define FMODE_NOWAIT ((__force fmode_t)0x8000000)
159
160/* File does not contribute to nr_files count */
161#define FMODE_NOACCOUNT ((__force fmode_t)0x20000000)
162
163/*
164 * Flag for rw_copy_check_uvector and compat_rw_copy_check_uvector
165 * that indicates that they should check the contents of the iovec are
166 * valid, but not check the memory that the iovec elements
167 * points too.
168 */
169#define CHECK_IOVEC_ONLY -1
170
171/*
172 * Attribute flags. These should be or-ed together to figure out what
173 * has been changed!
174 */
175#define ATTR_MODE (1 << 0)
176#define ATTR_UID (1 << 1)
177#define ATTR_GID (1 << 2)
178#define ATTR_SIZE (1 << 3)
179#define ATTR_ATIME (1 << 4)
180#define ATTR_MTIME (1 << 5)
181#define ATTR_CTIME (1 << 6)
182#define ATTR_ATIME_SET (1 << 7)
183#define ATTR_MTIME_SET (1 << 8)
184#define ATTR_FORCE (1 << 9) /* Not a change, but a change it */
185#define ATTR_KILL_SUID (1 << 11)
186#define ATTR_KILL_SGID (1 << 12)
187#define ATTR_FILE (1 << 13)
188#define ATTR_KILL_PRIV (1 << 14)
189#define ATTR_OPEN (1 << 15) /* Truncating from open(O_TRUNC) */
190#define ATTR_TIMES_SET (1 << 16)
191#define ATTR_TOUCH (1 << 17)
192
193/*
194 * Whiteout is represented by a char device. The following constants define the
195 * mode and device number to use.
196 */
197#define WHITEOUT_MODE 0
198#define WHITEOUT_DEV 0
199
200/*
201 * This is the Inode Attributes structure, used for notify_change(). It
202 * uses the above definitions as flags, to know which values have changed.
203 * Also, in this manner, a Filesystem can look at only the values it cares
204 * about. Basically, these are the attributes that the VFS layer can
205 * request to change from the FS layer.
206 *
207 * Derek Atkins <warlord@MIT.EDU> 94-10-20
208 */
209struct iattr {
210 unsigned int ia_valid;
211 umode_t ia_mode;
212 kuid_t ia_uid;
213 kgid_t ia_gid;
214 loff_t ia_size;
215 struct timespec64 ia_atime;
216 struct timespec64 ia_mtime;
217 struct timespec64 ia_ctime;
218
219 /*
220 * Not an attribute, but an auxiliary info for filesystems wanting to
221 * implement an ftruncate() like method. NOTE: filesystem should
222 * check for (ia_valid & ATTR_FILE), and not for (ia_file != NULL).
223 */
224 struct file *ia_file;
225};
226
227/*
228 * Includes for diskquotas.
229 */
230#include <linux/quota.h>
231
232/*
233 * Maximum number of layers of fs stack. Needs to be limited to
234 * prevent kernel stack overflow
235 */
236#define FILESYSTEM_MAX_STACK_DEPTH 2
237
238/**
239 * enum positive_aop_returns - aop return codes with specific semantics
240 *
241 * @AOP_WRITEPAGE_ACTIVATE: Informs the caller that page writeback has
242 * completed, that the page is still locked, and
243 * should be considered active. The VM uses this hint
244 * to return the page to the active list -- it won't
245 * be a candidate for writeback again in the near
246 * future. Other callers must be careful to unlock
247 * the page if they get this return. Returned by
248 * writepage();
249 *
250 * @AOP_TRUNCATED_PAGE: The AOP method that was handed a locked page has
251 * unlocked it and the page might have been truncated.
252 * The caller should back up to acquiring a new page and
253 * trying again. The aop will be taking reasonable
254 * precautions not to livelock. If the caller held a page
255 * reference, it should drop it before retrying. Returned
256 * by readpage().
257 *
258 * address_space_operation functions return these large constants to indicate
259 * special semantics to the caller. These are much larger than the bytes in a
260 * page to allow for functions that return the number of bytes operated on in a
261 * given page.
262 */
263
264enum positive_aop_returns {
265 AOP_WRITEPAGE_ACTIVATE = 0x80000,
266 AOP_TRUNCATED_PAGE = 0x80001,
267};
268
269#define AOP_FLAG_CONT_EXPAND 0x0001 /* called from cont_expand */
270#define AOP_FLAG_NOFS 0x0002 /* used by filesystem to direct
271 * helper code (eg buffer layer)
272 * to clear GFP_FS from alloc */
273
274/*
275 * oh the beauties of C type declarations.
276 */
277struct page;
278struct address_space;
279struct writeback_control;
280
281/*
282 * Write life time hint values.
283 * Stored in struct inode as u8.
284 */
285enum rw_hint {
286 WRITE_LIFE_NOT_SET = 0,
287 WRITE_LIFE_NONE = RWH_WRITE_LIFE_NONE,
288 WRITE_LIFE_SHORT = RWH_WRITE_LIFE_SHORT,
289 WRITE_LIFE_MEDIUM = RWH_WRITE_LIFE_MEDIUM,
290 WRITE_LIFE_LONG = RWH_WRITE_LIFE_LONG,
291 WRITE_LIFE_EXTREME = RWH_WRITE_LIFE_EXTREME,
292};
293
294#define IOCB_EVENTFD (1 << 0)
295#define IOCB_APPEND (1 << 1)
296#define IOCB_DIRECT (1 << 2)
297#define IOCB_HIPRI (1 << 3)
298#define IOCB_DSYNC (1 << 4)
299#define IOCB_SYNC (1 << 5)
300#define IOCB_WRITE (1 << 6)
301#define IOCB_NOWAIT (1 << 7)
302
303struct kiocb {
304 struct file *ki_filp;
305 loff_t ki_pos;
306 void (*ki_complete)(struct kiocb *iocb, long ret, long ret2);
307 void *private;
308 int ki_flags;
309 u16 ki_hint;
310 u16 ki_ioprio; /* See linux/ioprio.h */
311} __randomize_layout;
312
313static inline bool is_sync_kiocb(struct kiocb *kiocb)
314{
315 return kiocb->ki_complete == NULL;
316}
317
318/*
319 * "descriptor" for what we're up to with a read.
320 * This allows us to use the same read code yet
321 * have multiple different users of the data that
322 * we read from a file.
323 *
324 * The simplest case just copies the data to user
325 * mode.
326 */
327typedef struct {
328 size_t written;
329 size_t count;
330 union {
331 char __user *buf;
332 void *data;
333 } arg;
334 int error;
335} read_descriptor_t;
336
337typedef int (*read_actor_t)(read_descriptor_t *, struct page *,
338 unsigned long, unsigned long);
339
340struct address_space_operations {
341 int (*writepage)(struct page *page, struct writeback_control *wbc);
342 int (*readpage)(struct file *, struct page *);
343
344 /* Write back some dirty pages from this mapping. */
345 int (*writepages)(struct address_space *, struct writeback_control *);
346
347 /* Set a page dirty. Return true if this dirtied it */
348 int (*set_page_dirty)(struct page *page);
349
350 /*
351 * Reads in the requested pages. Unlike ->readpage(), this is
352 * PURELY used for read-ahead!.
353 */
354 int (*readpages)(struct file *filp, struct address_space *mapping,
355 struct list_head *pages, unsigned nr_pages);
356
357 int (*write_begin)(struct file *, struct address_space *mapping,
358 loff_t pos, unsigned len, unsigned flags,
359 struct page **pagep, void **fsdata);
360 int (*write_end)(struct file *, struct address_space *mapping,
361 loff_t pos, unsigned len, unsigned copied,
362 struct page *page, void *fsdata);
363
364 /* Unfortunately this kludge is needed for FIBMAP. Don't use it */
365 sector_t (*bmap)(struct address_space *, sector_t);
366 void (*invalidatepage) (struct page *, unsigned int, unsigned int);
367 int (*releasepage) (struct page *, gfp_t);
368 void (*freepage)(struct page *);
369 ssize_t (*direct_IO)(struct kiocb *, struct iov_iter *iter);
370 /*
371 * migrate the contents of a page to the specified target. If
372 * migrate_mode is MIGRATE_ASYNC, it must not block.
373 */
374 int (*migratepage) (struct address_space *,
375 struct page *, struct page *, enum migrate_mode);
376 bool (*isolate_page)(struct page *, isolate_mode_t);
377 void (*putback_page)(struct page *);
378 int (*launder_page) (struct page *);
379 int (*is_partially_uptodate) (struct page *, unsigned long,
380 unsigned long);
381 void (*is_dirty_writeback) (struct page *, bool *, bool *);
382 int (*error_remove_page)(struct address_space *, struct page *);
383
384 /* swapfile support */
385 int (*swap_activate)(struct swap_info_struct *sis, struct file *file,
386 sector_t *span);
387 void (*swap_deactivate)(struct file *file);
388};
389
390extern const struct address_space_operations empty_aops;
391
392/*
393 * pagecache_write_begin/pagecache_write_end must be used by general code
394 * to write into the pagecache.
395 */
396int pagecache_write_begin(struct file *, struct address_space *mapping,
397 loff_t pos, unsigned len, unsigned flags,
398 struct page **pagep, void **fsdata);
399
400int pagecache_write_end(struct file *, struct address_space *mapping,
401 loff_t pos, unsigned len, unsigned copied,
402 struct page *page, void *fsdata);
403
404struct address_space {
405 struct inode *host; /* owner: inode, block_device */
406 struct radix_tree_root i_pages; /* cached pages */
407 atomic_t i_mmap_writable;/* count VM_SHARED mappings */
408 struct rb_root_cached i_mmap; /* tree of private and shared mappings */
409 struct rw_semaphore i_mmap_rwsem; /* protect tree, count, list */
410 /* Protected by the i_pages lock */
411 unsigned long nrpages; /* number of total pages */
412 /* number of shadow or DAX exceptional entries */
413 unsigned long nrexceptional;
414 pgoff_t writeback_index;/* writeback starts here */
415 const struct address_space_operations *a_ops; /* methods */
416 unsigned long flags; /* error bits */
417 spinlock_t private_lock; /* for use by the address_space */
418 gfp_t gfp_mask; /* implicit gfp mask for allocations */
419 struct list_head private_list; /* for use by the address_space */
420 void *private_data; /* ditto */
421 errseq_t wb_err;
422} __attribute__((aligned(sizeof(long)))) __randomize_layout;
423 /*
424 * On most architectures that alignment is already the case; but
425 * must be enforced here for CRIS, to let the least significant bit
426 * of struct page's "mapping" pointer be used for PAGE_MAPPING_ANON.
427 */
428struct request_queue;
429
430struct block_device {
431 dev_t bd_dev; /* not a kdev_t - it's a search key */
432 int bd_openers;
433 struct inode * bd_inode; /* will die */
434 struct super_block * bd_super;
435 struct mutex bd_mutex; /* open/close mutex */
436 void * bd_claiming;
437 void * bd_holder;
438 int bd_holders;
439 bool bd_write_holder;
440#ifdef CONFIG_SYSFS
441 struct list_head bd_holder_disks;
442#endif
443 struct block_device * bd_contains;
444 unsigned bd_block_size;
445 u8 bd_partno;
446 struct hd_struct * bd_part;
447 /* number of times partitions within this device have been opened. */
448 unsigned bd_part_count;
449 int bd_invalidated;
450 struct gendisk * bd_disk;
451 struct request_queue * bd_queue;
452 struct backing_dev_info *bd_bdi;
453 struct list_head bd_list;
454 /*
455 * Private data. You must have bd_claim'ed the block_device
456 * to use this. NOTE: bd_claim allows an owner to claim
457 * the same device multiple times, the owner must take special
458 * care to not mess up bd_private for that case.
459 */
460 unsigned long bd_private;
461
462 /* The counter of freeze processes */
463 int bd_fsfreeze_count;
464 /* Mutex for freeze */
465 struct mutex bd_fsfreeze_mutex;
466} __randomize_layout;
467
468/*
469 * Radix-tree tags, for tagging dirty and writeback pages within the pagecache
470 * radix trees
471 */
472#define PAGECACHE_TAG_DIRTY 0
473#define PAGECACHE_TAG_WRITEBACK 1
474#define PAGECACHE_TAG_TOWRITE 2
475
476int mapping_tagged(struct address_space *mapping, int tag);
477
478static inline void i_mmap_lock_write(struct address_space *mapping)
479{
480 down_write(&mapping->i_mmap_rwsem);
481}
482
483static inline void i_mmap_unlock_write(struct address_space *mapping)
484{
485 up_write(&mapping->i_mmap_rwsem);
486}
487
488static inline void i_mmap_lock_read(struct address_space *mapping)
489{
490 down_read(&mapping->i_mmap_rwsem);
491}
492
493static inline void i_mmap_unlock_read(struct address_space *mapping)
494{
495 up_read(&mapping->i_mmap_rwsem);
496}
497
498/*
499 * Might pages of this file be mapped into userspace?
500 */
501static inline int mapping_mapped(struct address_space *mapping)
502{
503 return !RB_EMPTY_ROOT(&mapping->i_mmap.rb_root);
504}
505
506/*
507 * Might pages of this file have been modified in userspace?
508 * Note that i_mmap_writable counts all VM_SHARED vmas: do_mmap_pgoff
509 * marks vma as VM_SHARED if it is shared, and the file was opened for
510 * writing i.e. vma may be mprotected writable even if now readonly.
511 *
512 * If i_mmap_writable is negative, no new writable mappings are allowed. You
513 * can only deny writable mappings, if none exists right now.
514 */
515static inline int mapping_writably_mapped(struct address_space *mapping)
516{
517 return atomic_read(&mapping->i_mmap_writable) > 0;
518}
519
520static inline int mapping_map_writable(struct address_space *mapping)
521{
522 return atomic_inc_unless_negative(&mapping->i_mmap_writable) ?
523 0 : -EPERM;
524}
525
526static inline void mapping_unmap_writable(struct address_space *mapping)
527{
528 atomic_dec(&mapping->i_mmap_writable);
529}
530
531static inline int mapping_deny_writable(struct address_space *mapping)
532{
533 return atomic_dec_unless_positive(&mapping->i_mmap_writable) ?
534 0 : -EBUSY;
535}
536
537static inline void mapping_allow_writable(struct address_space *mapping)
538{
539 atomic_inc(&mapping->i_mmap_writable);
540}
541
542/*
543 * Use sequence counter to get consistent i_size on 32-bit processors.
544 */
545#if BITS_PER_LONG==32 && defined(CONFIG_SMP)
546#include <linux/seqlock.h>
547#define __NEED_I_SIZE_ORDERED
548#define i_size_ordered_init(inode) seqcount_init(&inode->i_size_seqcount)
549#else
550#define i_size_ordered_init(inode) do { } while (0)
551#endif
552
553struct posix_acl;
554#define ACL_NOT_CACHED ((void *)(-1))
555#define ACL_DONT_CACHE ((void *)(-3))
556
557static inline struct posix_acl *
558uncached_acl_sentinel(struct task_struct *task)
559{
560 return (void *)task + 1;
561}
562
563static inline bool
564is_uncached_acl(struct posix_acl *acl)
565{
566 return (long)acl & 1;
567}
568
569#define IOP_FASTPERM 0x0001
570#define IOP_LOOKUP 0x0002
571#define IOP_NOFOLLOW 0x0004
572#define IOP_XATTR 0x0008
573#define IOP_DEFAULT_READLINK 0x0010
574
575struct fsnotify_mark_connector;
576
577/*
578 * Keep mostly read-only and often accessed (especially for
579 * the RCU path lookup and 'stat' data) fields at the beginning
580 * of the 'struct inode'
581 */
582struct inode {
583 umode_t i_mode;
584 unsigned short i_opflags;
585 kuid_t i_uid;
586 kgid_t i_gid;
587 unsigned int i_flags;
588
589#ifdef CONFIG_FS_POSIX_ACL
590 struct posix_acl *i_acl;
591 struct posix_acl *i_default_acl;
592#endif
593
594 const struct inode_operations *i_op;
595 struct super_block *i_sb;
596 struct address_space *i_mapping;
597
598#ifdef CONFIG_SECURITY
599 void *i_security;
600#endif
601
602 /* Stat data, not accessed from path walking */
603 unsigned long i_ino;
604 /*
605 * Filesystems may only read i_nlink directly. They shall use the
606 * following functions for modification:
607 *
608 * (set|clear|inc|drop)_nlink
609 * inode_(inc|dec)_link_count
610 */
611 union {
612 const unsigned int i_nlink;
613 unsigned int __i_nlink;
614 };
615 dev_t i_rdev;
616 loff_t i_size;
617 struct timespec64 i_atime;
618 struct timespec64 i_mtime;
619 struct timespec64 i_ctime;
620 spinlock_t i_lock; /* i_blocks, i_bytes, maybe i_size */
621 unsigned short i_bytes;
622 u8 i_blkbits;
623 u8 i_write_hint;
624 blkcnt_t i_blocks;
625
626#ifdef __NEED_I_SIZE_ORDERED
627 seqcount_t i_size_seqcount;
628#endif
629
630 /* Misc */
631 unsigned long i_state;
632 struct rw_semaphore i_rwsem;
633
634 unsigned long dirtied_when; /* jiffies of first dirtying */
635 unsigned long dirtied_time_when;
636
637 struct hlist_node i_hash;
638 struct list_head i_io_list; /* backing dev IO list */
639#ifdef CONFIG_CGROUP_WRITEBACK
640 struct bdi_writeback *i_wb; /* the associated cgroup wb */
641
642 /* foreign inode detection, see wbc_detach_inode() */
643 int i_wb_frn_winner;
644 u16 i_wb_frn_avg_time;
645 u16 i_wb_frn_history;
646#endif
647 struct list_head i_lru; /* inode LRU list */
648 struct list_head i_sb_list;
649 struct list_head i_wb_list; /* backing dev writeback list */
650 union {
651 struct hlist_head i_dentry;
652 struct rcu_head i_rcu;
653 };
654 atomic64_t i_version;
655 atomic_t i_count;
656 atomic_t i_dio_count;
657 atomic_t i_writecount;
658#ifdef CONFIG_IMA
659 atomic_t i_readcount; /* struct files open RO */
660#endif
661 const struct file_operations *i_fop; /* former ->i_op->default_file_ops */
662 struct file_lock_context *i_flctx;
663 struct address_space i_data;
664 struct list_head i_devices;
665 union {
666 struct pipe_inode_info *i_pipe;
667 struct block_device *i_bdev;
668 struct cdev *i_cdev;
669 char *i_link;
670 unsigned i_dir_seq;
671 };
672
673 __u32 i_generation;
674
675#ifdef CONFIG_FSNOTIFY
676 __u32 i_fsnotify_mask; /* all events this inode cares about */
677 struct fsnotify_mark_connector __rcu *i_fsnotify_marks;
678#endif
679
680#if IS_ENABLED(CONFIG_FS_ENCRYPTION)
681 struct fscrypt_info *i_crypt_info;
682#endif
683
684 void *i_private; /* fs or device private pointer */
685} __randomize_layout;
686
687static inline unsigned int i_blocksize(const struct inode *node)
688{
689 return (1 << node->i_blkbits);
690}
691
692static inline int inode_unhashed(struct inode *inode)
693{
694 return hlist_unhashed(&inode->i_hash);
695}
696
697/*
698 * __mark_inode_dirty expects inodes to be hashed. Since we don't
699 * want special inodes in the fileset inode space, we make them
700 * appear hashed, but do not put on any lists. hlist_del()
701 * will work fine and require no locking.
702 */
703static inline void inode_fake_hash(struct inode *inode)
704{
705 hlist_add_fake(&inode->i_hash);
706}
707
708/*
709 * inode->i_mutex nesting subclasses for the lock validator:
710 *
711 * 0: the object of the current VFS operation
712 * 1: parent
713 * 2: child/target
714 * 3: xattr
715 * 4: second non-directory
716 * 5: second parent (when locking independent directories in rename)
717 *
718 * I_MUTEX_NONDIR2 is for certain operations (such as rename) which lock two
719 * non-directories at once.
720 *
721 * The locking order between these classes is
722 * parent[2] -> child -> grandchild -> normal -> xattr -> second non-directory
723 */
724enum inode_i_mutex_lock_class
725{
726 I_MUTEX_NORMAL,
727 I_MUTEX_PARENT,
728 I_MUTEX_CHILD,
729 I_MUTEX_XATTR,
730 I_MUTEX_NONDIR2,
731 I_MUTEX_PARENT2,
732};
733
734static inline void inode_lock(struct inode *inode)
735{
736 down_write(&inode->i_rwsem);
737}
738
739static inline void inode_unlock(struct inode *inode)
740{
741 up_write(&inode->i_rwsem);
742}
743
744static inline void inode_lock_shared(struct inode *inode)
745{
746 down_read(&inode->i_rwsem);
747}
748
749static inline void inode_unlock_shared(struct inode *inode)
750{
751 up_read(&inode->i_rwsem);
752}
753
754static inline int inode_trylock(struct inode *inode)
755{
756 return down_write_trylock(&inode->i_rwsem);
757}
758
759static inline int inode_trylock_shared(struct inode *inode)
760{
761 return down_read_trylock(&inode->i_rwsem);
762}
763
764static inline int inode_is_locked(struct inode *inode)
765{
766 return rwsem_is_locked(&inode->i_rwsem);
767}
768
769static inline void inode_lock_nested(struct inode *inode, unsigned subclass)
770{
771 down_write_nested(&inode->i_rwsem, subclass);
772}
773
774static inline void inode_lock_shared_nested(struct inode *inode, unsigned subclass)
775{
776 down_read_nested(&inode->i_rwsem, subclass);
777}
778
779void lock_two_nondirectories(struct inode *, struct inode*);
780void unlock_two_nondirectories(struct inode *, struct inode*);
781
782/*
783 * NOTE: in a 32bit arch with a preemptable kernel and
784 * an UP compile the i_size_read/write must be atomic
785 * with respect to the local cpu (unlike with preempt disabled),
786 * but they don't need to be atomic with respect to other cpus like in
787 * true SMP (so they need either to either locally disable irq around
788 * the read or for example on x86 they can be still implemented as a
789 * cmpxchg8b without the need of the lock prefix). For SMP compiles
790 * and 64bit archs it makes no difference if preempt is enabled or not.
791 */
792static inline loff_t i_size_read(const struct inode *inode)
793{
794#if BITS_PER_LONG==32 && defined(CONFIG_SMP)
795 loff_t i_size;
796 unsigned int seq;
797
798 do {
799 seq = read_seqcount_begin(&inode->i_size_seqcount);
800 i_size = inode->i_size;
801 } while (read_seqcount_retry(&inode->i_size_seqcount, seq));
802 return i_size;
803#elif BITS_PER_LONG==32 && defined(CONFIG_PREEMPT)
804 loff_t i_size;
805
806 preempt_disable();
807 i_size = inode->i_size;
808 preempt_enable();
809 return i_size;
810#else
811 return inode->i_size;
812#endif
813}
814
815/*
816 * NOTE: unlike i_size_read(), i_size_write() does need locking around it
817 * (normally i_mutex), otherwise on 32bit/SMP an update of i_size_seqcount
818 * can be lost, resulting in subsequent i_size_read() calls spinning forever.
819 */
820static inline void i_size_write(struct inode *inode, loff_t i_size)
821{
822#if BITS_PER_LONG==32 && defined(CONFIG_SMP)
823 preempt_disable();
824 write_seqcount_begin(&inode->i_size_seqcount);
825 inode->i_size = i_size;
826 write_seqcount_end(&inode->i_size_seqcount);
827 preempt_enable();
828#elif BITS_PER_LONG==32 && defined(CONFIG_PREEMPT)
829 preempt_disable();
830 inode->i_size = i_size;
831 preempt_enable();
832#else
833 inode->i_size = i_size;
834#endif
835}
836
837static inline unsigned iminor(const struct inode *inode)
838{
839 return MINOR(inode->i_rdev);
840}
841
842static inline unsigned imajor(const struct inode *inode)
843{
844 return MAJOR(inode->i_rdev);
845}
846
847extern struct block_device *I_BDEV(struct inode *inode);
848
849struct fown_struct {
850 rwlock_t lock; /* protects pid, uid, euid fields */
851 struct pid *pid; /* pid or -pgrp where SIGIO should be sent */
852 enum pid_type pid_type; /* Kind of process group SIGIO should be sent to */
853 kuid_t uid, euid; /* uid/euid of process setting the owner */
854 int signum; /* posix.1b rt signal to be delivered on IO */
855};
856
857/*
858 * Track a single file's readahead state
859 */
860struct file_ra_state {
861 pgoff_t start; /* where readahead started */
862 unsigned int size; /* # of readahead pages */
863 unsigned int async_size; /* do asynchronous readahead when
864 there are only # of pages ahead */
865
866 unsigned int ra_pages; /* Maximum readahead window */
867 unsigned int mmap_miss; /* Cache miss stat for mmap accesses */
868 loff_t prev_pos; /* Cache last read() position */
869};
870
871/*
872 * Check if @index falls in the readahead windows.
873 */
874static inline int ra_has_index(struct file_ra_state *ra, pgoff_t index)
875{
876 return (index >= ra->start &&
877 index < ra->start + ra->size);
878}
879
880struct file {
881 union {
882 struct llist_node fu_llist;
883 struct rcu_head fu_rcuhead;
884 } f_u;
885 struct path f_path;
886 struct inode *f_inode; /* cached value */
887 const struct file_operations *f_op;
888
889 /*
890 * Protects f_ep_links, f_flags.
891 * Must not be taken from IRQ context.
892 */
893 spinlock_t f_lock;
894 enum rw_hint f_write_hint;
895 atomic_long_t f_count;
896 unsigned int f_flags;
897 fmode_t f_mode;
898 struct mutex f_pos_lock;
899 loff_t f_pos;
900 struct fown_struct f_owner;
901 const struct cred *f_cred;
902 struct file_ra_state f_ra;
903
904 u64 f_version;
905#ifdef CONFIG_SECURITY
906 void *f_security;
907#endif
908 /* needed for tty driver, and maybe others */
909 void *private_data;
910
911#ifdef CONFIG_EPOLL
912 /* Used by fs/eventpoll.c to link all the hooks to this file */
913 struct list_head f_ep_links;
914 struct list_head f_tfile_llink;
915#endif /* #ifdef CONFIG_EPOLL */
916 struct address_space *f_mapping;
917 errseq_t f_wb_err;
918} __randomize_layout
919 __attribute__((aligned(4))); /* lest something weird decides that 2 is OK */
920
921struct file_handle {
922 __u32 handle_bytes;
923 int handle_type;
924 /* file identifier */
925 unsigned char f_handle[0];
926};
927
928static inline struct file *get_file(struct file *f)
929{
930 atomic_long_inc(&f->f_count);
931 return f;
932}
933#define get_file_rcu(x) atomic_long_inc_not_zero(&(x)->f_count)
934#define fput_atomic(x) atomic_long_add_unless(&(x)->f_count, -1, 1)
935#define file_count(x) atomic_long_read(&(x)->f_count)
936
937#define MAX_NON_LFS ((1UL<<31) - 1)
938
939/* Page cache limit. The filesystems should put that into their s_maxbytes
940 limits, otherwise bad things can happen in VM. */
941#if BITS_PER_LONG==32
942#define MAX_LFS_FILESIZE ((loff_t)ULONG_MAX << PAGE_SHIFT)
943#elif BITS_PER_LONG==64
944#define MAX_LFS_FILESIZE ((loff_t)LLONG_MAX)
945#endif
946
947#define FL_POSIX 1
948#define FL_FLOCK 2
949#define FL_DELEG 4 /* NFSv4 delegation */
950#define FL_ACCESS 8 /* not trying to lock, just looking */
951#define FL_EXISTS 16 /* when unlocking, test for existence */
952#define FL_LEASE 32 /* lease held on this file */
953#define FL_CLOSE 64 /* unlock on close */
954#define FL_SLEEP 128 /* A blocking lock */
955#define FL_DOWNGRADE_PENDING 256 /* Lease is being downgraded */
956#define FL_UNLOCK_PENDING 512 /* Lease is being broken */
957#define FL_OFDLCK 1024 /* lock is "owned" by struct file */
958#define FL_LAYOUT 2048 /* outstanding pNFS layout */
959
960#define FL_CLOSE_POSIX (FL_POSIX | FL_CLOSE)
961
962/*
963 * Special return value from posix_lock_file() and vfs_lock_file() for
964 * asynchronous locking.
965 */
966#define FILE_LOCK_DEFERRED 1
967
968/* legacy typedef, should eventually be removed */
969typedef void *fl_owner_t;
970
971struct file_lock;
972
973struct file_lock_operations {
974 void (*fl_copy_lock)(struct file_lock *, struct file_lock *);
975 void (*fl_release_private)(struct file_lock *);
976};
977
978struct lock_manager_operations {
979 int (*lm_compare_owner)(struct file_lock *, struct file_lock *);
980 unsigned long (*lm_owner_key)(struct file_lock *);
981 fl_owner_t (*lm_get_owner)(fl_owner_t);
982 void (*lm_put_owner)(fl_owner_t);
983 void (*lm_notify)(struct file_lock *); /* unblock callback */
984 int (*lm_grant)(struct file_lock *, int);
985 bool (*lm_break)(struct file_lock *);
986 int (*lm_change)(struct file_lock *, int, struct list_head *);
987 void (*lm_setup)(struct file_lock *, void **);
988};
989
990struct lock_manager {
991 struct list_head list;
992 /*
993 * NFSv4 and up also want opens blocked during the grace period;
994 * NLM doesn't care:
995 */
996 bool block_opens;
997};
998
999struct net;
1000void locks_start_grace(struct net *, struct lock_manager *);
1001void locks_end_grace(struct lock_manager *);
1002bool locks_in_grace(struct net *);
1003bool opens_in_grace(struct net *);
1004
1005/* that will die - we need it for nfs_lock_info */
1006#include <linux/nfs_fs_i.h>
1007
1008/*
1009 * struct file_lock represents a generic "file lock". It's used to represent
1010 * POSIX byte range locks, BSD (flock) locks, and leases. It's important to
1011 * note that the same struct is used to represent both a request for a lock and
1012 * the lock itself, but the same object is never used for both.
1013 *
1014 * FIXME: should we create a separate "struct lock_request" to help distinguish
1015 * these two uses?
1016 *
1017 * The varous i_flctx lists are ordered by:
1018 *
1019 * 1) lock owner
1020 * 2) lock range start
1021 * 3) lock range end
1022 *
1023 * Obviously, the last two criteria only matter for POSIX locks.
1024 */
1025struct file_lock {
1026 struct file_lock *fl_next; /* singly linked list for this inode */
1027 struct list_head fl_list; /* link into file_lock_context */
1028 struct hlist_node fl_link; /* node in global lists */
1029 struct list_head fl_block; /* circular list of blocked processes */
1030 fl_owner_t fl_owner;
1031 unsigned int fl_flags;
1032 unsigned char fl_type;
1033 unsigned int fl_pid;
1034 int fl_link_cpu; /* what cpu's list is this on? */
1035 wait_queue_head_t fl_wait;
1036 struct file *fl_file;
1037 loff_t fl_start;
1038 loff_t fl_end;
1039
1040 struct fasync_struct * fl_fasync; /* for lease break notifications */
1041 /* for lease breaks: */
1042 unsigned long fl_break_time;
1043 unsigned long fl_downgrade_time;
1044
1045 const struct file_lock_operations *fl_ops; /* Callbacks for filesystems */
1046 const struct lock_manager_operations *fl_lmops; /* Callbacks for lockmanagers */
1047 union {
1048 struct nfs_lock_info nfs_fl;
1049 struct nfs4_lock_info nfs4_fl;
1050 struct {
1051 struct list_head link; /* link in AFS vnode's pending_locks list */
1052 int state; /* state of grant or error if -ve */
1053 } afs;
1054 } fl_u;
1055} __randomize_layout;
1056
1057struct file_lock_context {
1058 spinlock_t flc_lock;
1059 struct list_head flc_flock;
1060 struct list_head flc_posix;
1061 struct list_head flc_lease;
1062};
1063
1064/* The following constant reflects the upper bound of the file/locking space */
1065#ifndef OFFSET_MAX
1066#define INT_LIMIT(x) (~((x)1 << (sizeof(x)*8 - 1)))
1067#define OFFSET_MAX INT_LIMIT(loff_t)
1068#define OFFT_OFFSET_MAX INT_LIMIT(off_t)
1069#endif
1070
1071extern void send_sigio(struct fown_struct *fown, int fd, int band);
1072
1073#define locks_inode(f) file_inode(f)
1074
1075#ifdef CONFIG_FILE_LOCKING
1076extern int fcntl_getlk(struct file *, unsigned int, struct flock *);
1077extern int fcntl_setlk(unsigned int, struct file *, unsigned int,
1078 struct flock *);
1079
1080#if BITS_PER_LONG == 32
1081extern int fcntl_getlk64(struct file *, unsigned int, struct flock64 *);
1082extern int fcntl_setlk64(unsigned int, struct file *, unsigned int,
1083 struct flock64 *);
1084#endif
1085
1086extern int fcntl_setlease(unsigned int fd, struct file *filp, long arg);
1087extern int fcntl_getlease(struct file *filp);
1088
1089/* fs/locks.c */
1090void locks_free_lock_context(struct inode *inode);
1091void locks_free_lock(struct file_lock *fl);
1092extern void locks_init_lock(struct file_lock *);
1093extern struct file_lock * locks_alloc_lock(void);
1094extern void locks_copy_lock(struct file_lock *, struct file_lock *);
1095extern void locks_copy_conflock(struct file_lock *, struct file_lock *);
1096extern void locks_remove_posix(struct file *, fl_owner_t);
1097extern void locks_remove_file(struct file *);
1098extern void locks_release_private(struct file_lock *);
1099extern void posix_test_lock(struct file *, struct file_lock *);
1100extern int posix_lock_file(struct file *, struct file_lock *, struct file_lock *);
1101extern int posix_unblock_lock(struct file_lock *);
1102extern int vfs_test_lock(struct file *, struct file_lock *);
1103extern int vfs_lock_file(struct file *, unsigned int, struct file_lock *, struct file_lock *);
1104extern int vfs_cancel_lock(struct file *filp, struct file_lock *fl);
1105extern int locks_lock_inode_wait(struct inode *inode, struct file_lock *fl);
1106extern int __break_lease(struct inode *inode, unsigned int flags, unsigned int type);
1107extern void lease_get_mtime(struct inode *, struct timespec64 *time);
1108extern int generic_setlease(struct file *, long, struct file_lock **, void **priv);
1109extern int vfs_setlease(struct file *, long, struct file_lock **, void **);
1110extern int lease_modify(struct file_lock *, int, struct list_head *);
1111struct files_struct;
1112extern void show_fd_locks(struct seq_file *f,
1113 struct file *filp, struct files_struct *files);
1114#else /* !CONFIG_FILE_LOCKING */
1115static inline int fcntl_getlk(struct file *file, unsigned int cmd,
1116 struct flock __user *user)
1117{
1118 return -EINVAL;
1119}
1120
1121static inline int fcntl_setlk(unsigned int fd, struct file *file,
1122 unsigned int cmd, struct flock __user *user)
1123{
1124 return -EACCES;
1125}
1126
1127#if BITS_PER_LONG == 32
1128static inline int fcntl_getlk64(struct file *file, unsigned int cmd,
1129 struct flock64 __user *user)
1130{
1131 return -EINVAL;
1132}
1133
1134static inline int fcntl_setlk64(unsigned int fd, struct file *file,
1135 unsigned int cmd, struct flock64 __user *user)
1136{
1137 return -EACCES;
1138}
1139#endif
1140static inline int fcntl_setlease(unsigned int fd, struct file *filp, long arg)
1141{
1142 return -EINVAL;
1143}
1144
1145static inline int fcntl_getlease(struct file *filp)
1146{
1147 return F_UNLCK;
1148}
1149
1150static inline void
1151locks_free_lock_context(struct inode *inode)
1152{
1153}
1154
1155static inline void locks_init_lock(struct file_lock *fl)
1156{
1157 return;
1158}
1159
1160static inline void locks_copy_conflock(struct file_lock *new, struct file_lock *fl)
1161{
1162 return;
1163}
1164
1165static inline void locks_copy_lock(struct file_lock *new, struct file_lock *fl)
1166{
1167 return;
1168}
1169
1170static inline void locks_remove_posix(struct file *filp, fl_owner_t owner)
1171{
1172 return;
1173}
1174
1175static inline void locks_remove_file(struct file *filp)
1176{
1177 return;
1178}
1179
1180static inline void posix_test_lock(struct file *filp, struct file_lock *fl)
1181{
1182 return;
1183}
1184
1185static inline int posix_lock_file(struct file *filp, struct file_lock *fl,
1186 struct file_lock *conflock)
1187{
1188 return -ENOLCK;
1189}
1190
1191static inline int posix_unblock_lock(struct file_lock *waiter)
1192{
1193 return -ENOENT;
1194}
1195
1196static inline int vfs_test_lock(struct file *filp, struct file_lock *fl)
1197{
1198 return 0;
1199}
1200
1201static inline int vfs_lock_file(struct file *filp, unsigned int cmd,
1202 struct file_lock *fl, struct file_lock *conf)
1203{
1204 return -ENOLCK;
1205}
1206
1207static inline int vfs_cancel_lock(struct file *filp, struct file_lock *fl)
1208{
1209 return 0;
1210}
1211
1212static inline int locks_lock_inode_wait(struct inode *inode, struct file_lock *fl)
1213{
1214 return -ENOLCK;
1215}
1216
1217static inline int __break_lease(struct inode *inode, unsigned int mode, unsigned int type)
1218{
1219 return 0;
1220}
1221
1222static inline void lease_get_mtime(struct inode *inode,
1223 struct timespec64 *time)
1224{
1225 return;
1226}
1227
1228static inline int generic_setlease(struct file *filp, long arg,
1229 struct file_lock **flp, void **priv)
1230{
1231 return -EINVAL;
1232}
1233
1234static inline int vfs_setlease(struct file *filp, long arg,
1235 struct file_lock **lease, void **priv)
1236{
1237 return -EINVAL;
1238}
1239
1240static inline int lease_modify(struct file_lock *fl, int arg,
1241 struct list_head *dispose)
1242{
1243 return -EINVAL;
1244}
1245
1246struct files_struct;
1247static inline void show_fd_locks(struct seq_file *f,
1248 struct file *filp, struct files_struct *files) {}
1249#endif /* !CONFIG_FILE_LOCKING */
1250
1251static inline struct inode *file_inode(const struct file *f)
1252{
1253 return f->f_inode;
1254}
1255
1256static inline struct dentry *file_dentry(const struct file *file)
1257{
1258 return d_real(file->f_path.dentry, file_inode(file));
1259}
1260
1261static inline int locks_lock_file_wait(struct file *filp, struct file_lock *fl)
1262{
1263 return locks_lock_inode_wait(locks_inode(filp), fl);
1264}
1265
1266struct fasync_struct {
1267 rwlock_t fa_lock;
1268 int magic;
1269 int fa_fd;
1270 struct fasync_struct *fa_next; /* singly linked list */
1271 struct file *fa_file;
1272 struct rcu_head fa_rcu;
1273};
1274
1275#define FASYNC_MAGIC 0x4601
1276
1277/* SMP safe fasync helpers: */
1278extern int fasync_helper(int, struct file *, int, struct fasync_struct **);
1279extern struct fasync_struct *fasync_insert_entry(int, struct file *, struct fasync_struct **, struct fasync_struct *);
1280extern int fasync_remove_entry(struct file *, struct fasync_struct **);
1281extern struct fasync_struct *fasync_alloc(void);
1282extern void fasync_free(struct fasync_struct *);
1283
1284/* can be called from interrupts */
1285extern void kill_fasync(struct fasync_struct **, int, int);
1286
1287extern void __f_setown(struct file *filp, struct pid *, enum pid_type, int force);
1288extern int f_setown(struct file *filp, unsigned long arg, int force);
1289extern void f_delown(struct file *filp);
1290extern pid_t f_getown(struct file *filp);
1291extern int send_sigurg(struct fown_struct *fown);
1292
1293/*
1294 * sb->s_flags. Note that these mirror the equivalent MS_* flags where
1295 * represented in both.
1296 */
1297#define SB_RDONLY 1 /* Mount read-only */
1298#define SB_NOSUID 2 /* Ignore suid and sgid bits */
1299#define SB_NODEV 4 /* Disallow access to device special files */
1300#define SB_NOEXEC 8 /* Disallow program execution */
1301#define SB_SYNCHRONOUS 16 /* Writes are synced at once */
1302#define SB_MANDLOCK 64 /* Allow mandatory locks on an FS */
1303#define SB_DIRSYNC 128 /* Directory modifications are synchronous */
1304#define SB_NOATIME 1024 /* Do not update access times. */
1305#define SB_NODIRATIME 2048 /* Do not update directory access times */
1306#define SB_SILENT 32768
1307#define SB_POSIXACL (1<<16) /* VFS does not apply the umask */
1308#define SB_KERNMOUNT (1<<22) /* this is a kern_mount call */
1309#define SB_I_VERSION (1<<23) /* Update inode I_version field */
1310#define SB_LAZYTIME (1<<25) /* Update the on-disk [acm]times lazily */
1311
1312/* These sb flags are internal to the kernel */
1313#define SB_SUBMOUNT (1<<26)
1314#define SB_NOSEC (1<<28)
1315#define SB_BORN (1<<29)
1316#define SB_ACTIVE (1<<30)
1317#define SB_NOUSER (1<<31)
1318
1319/*
1320 * Umount options
1321 */
1322
1323#define MNT_FORCE 0x00000001 /* Attempt to forcibily umount */
1324#define MNT_DETACH 0x00000002 /* Just detach from the tree */
1325#define MNT_EXPIRE 0x00000004 /* Mark for expiry */
1326#define UMOUNT_NOFOLLOW 0x00000008 /* Don't follow symlink on umount */
1327#define UMOUNT_UNUSED 0x80000000 /* Flag guaranteed to be unused */
1328
1329/* sb->s_iflags */
1330#define SB_I_CGROUPWB 0x00000001 /* cgroup-aware writeback enabled */
1331#define SB_I_NOEXEC 0x00000002 /* Ignore executables on this fs */
1332#define SB_I_NODEV 0x00000004 /* Ignore devices on this fs */
1333#define SB_I_MULTIROOT 0x00000008 /* Multiple roots to the dentry tree */
1334
1335/* sb->s_iflags to limit user namespace mounts */
1336#define SB_I_USERNS_VISIBLE 0x00000010 /* fstype already mounted */
1337#define SB_I_IMA_UNVERIFIABLE_SIGNATURE 0x00000020
1338#define SB_I_UNTRUSTED_MOUNTER 0x00000040
1339
1340/* Possible states of 'frozen' field */
1341enum {
1342 SB_UNFROZEN = 0, /* FS is unfrozen */
1343 SB_FREEZE_WRITE = 1, /* Writes, dir ops, ioctls frozen */
1344 SB_FREEZE_PAGEFAULT = 2, /* Page faults stopped as well */
1345 SB_FREEZE_FS = 3, /* For internal FS use (e.g. to stop
1346 * internal threads if needed) */
1347 SB_FREEZE_COMPLETE = 4, /* ->freeze_fs finished successfully */
1348};
1349
1350#define SB_FREEZE_LEVELS (SB_FREEZE_COMPLETE - 1)
1351
1352struct sb_writers {
1353 int frozen; /* Is sb frozen? */
1354 wait_queue_head_t wait_unfrozen; /* for get_super_thawed() */
1355 struct percpu_rw_semaphore rw_sem[SB_FREEZE_LEVELS];
1356};
1357
1358struct super_block {
1359 struct list_head s_list; /* Keep this first */
1360 dev_t s_dev; /* search index; _not_ kdev_t */
1361 unsigned char s_blocksize_bits;
1362 unsigned long s_blocksize;
1363 loff_t s_maxbytes; /* Max file size */
1364 struct file_system_type *s_type;
1365 const struct super_operations *s_op;
1366 const struct dquot_operations *dq_op;
1367 const struct quotactl_ops *s_qcop;
1368 const struct export_operations *s_export_op;
1369 unsigned long s_flags;
1370 unsigned long s_iflags; /* internal SB_I_* flags */
1371 unsigned long s_magic;
1372 struct dentry *s_root;
1373 struct rw_semaphore s_umount;
1374 int s_count;
1375 atomic_t s_active;
1376#ifdef CONFIG_SECURITY
1377 void *s_security;
1378#endif
1379 const struct xattr_handler **s_xattr;
1380#if IS_ENABLED(CONFIG_FS_ENCRYPTION)
1381 const struct fscrypt_operations *s_cop;
1382#endif
1383 struct hlist_bl_head s_roots; /* alternate root dentries for NFS */
1384 struct list_head s_mounts; /* list of mounts; _not_ for fs use */
1385 struct block_device *s_bdev;
1386 struct backing_dev_info *s_bdi;
1387 struct mtd_info *s_mtd;
1388 struct hlist_node s_instances;
1389 unsigned int s_quota_types; /* Bitmask of supported quota types */
1390 struct quota_info s_dquot; /* Diskquota specific options */
1391
1392 struct sb_writers s_writers;
1393
1394 char s_id[32]; /* Informational name */
1395 uuid_t s_uuid; /* UUID */
1396
1397 void *s_fs_info; /* Filesystem private info */
1398 unsigned int s_max_links;
1399 fmode_t s_mode;
1400
1401 /* Granularity of c/m/atime in ns.
1402 Cannot be worse than a second */
1403 u32 s_time_gran;
1404
1405 /*
1406 * The next field is for VFS *only*. No filesystems have any business
1407 * even looking at it. You had been warned.
1408 */
1409 struct mutex s_vfs_rename_mutex; /* Kludge */
1410
1411 /*
1412 * Filesystem subtype. If non-empty the filesystem type field
1413 * in /proc/mounts will be "type.subtype"
1414 */
1415 char *s_subtype;
1416
1417 const struct dentry_operations *s_d_op; /* default d_op for dentries */
1418
1419 /*
1420 * Saved pool identifier for cleancache (-1 means none)
1421 */
1422 int cleancache_poolid;
1423
1424 struct shrinker s_shrink; /* per-sb shrinker handle */
1425
1426 /* Number of inodes with nlink == 0 but still referenced */
1427 atomic_long_t s_remove_count;
1428
1429 /* Being remounted read-only */
1430 int s_readonly_remount;
1431
1432 /* AIO completions deferred from interrupt context */
1433 struct workqueue_struct *s_dio_done_wq;
1434 struct hlist_head s_pins;
1435
1436 /*
1437 * Owning user namespace and default context in which to
1438 * interpret filesystem uids, gids, quotas, device nodes,
1439 * xattrs and security labels.
1440 */
1441 struct user_namespace *s_user_ns;
1442
1443 /*
1444 * Keep the lru lists last in the structure so they always sit on their
1445 * own individual cachelines.
1446 */
1447 struct list_lru s_dentry_lru ____cacheline_aligned_in_smp;
1448 struct list_lru s_inode_lru ____cacheline_aligned_in_smp;
1449 struct rcu_head rcu;
1450 struct work_struct destroy_work;
1451
1452 struct mutex s_sync_lock; /* sync serialisation lock */
1453
1454 /*
1455 * Indicates how deep in a filesystem stack this SB is
1456 */
1457 int s_stack_depth;
1458
1459 /* s_inode_list_lock protects s_inodes */
1460 spinlock_t s_inode_list_lock ____cacheline_aligned_in_smp;
1461 struct list_head s_inodes; /* all inodes */
1462
1463 spinlock_t s_inode_wblist_lock;
1464 struct list_head s_inodes_wb; /* writeback inodes */
1465} __randomize_layout;
1466
1467/* Helper functions so that in most cases filesystems will
1468 * not need to deal directly with kuid_t and kgid_t and can
1469 * instead deal with the raw numeric values that are stored
1470 * in the filesystem.
1471 */
1472static inline uid_t i_uid_read(const struct inode *inode)
1473{
1474 return from_kuid(inode->i_sb->s_user_ns, inode->i_uid);
1475}
1476
1477static inline gid_t i_gid_read(const struct inode *inode)
1478{
1479 return from_kgid(inode->i_sb->s_user_ns, inode->i_gid);
1480}
1481
1482static inline void i_uid_write(struct inode *inode, uid_t uid)
1483{
1484 inode->i_uid = make_kuid(inode->i_sb->s_user_ns, uid);
1485}
1486
1487static inline void i_gid_write(struct inode *inode, gid_t gid)
1488{
1489 inode->i_gid = make_kgid(inode->i_sb->s_user_ns, gid);
1490}
1491
1492extern struct timespec64 timespec64_trunc(struct timespec64 t, unsigned gran);
1493extern struct timespec64 current_time(struct inode *inode);
1494
1495/*
1496 * Snapshotting support.
1497 */
1498
1499void __sb_end_write(struct super_block *sb, int level);
1500int __sb_start_write(struct super_block *sb, int level, bool wait);
1501
1502#define __sb_writers_acquired(sb, lev) \
1503 percpu_rwsem_acquire(&(sb)->s_writers.rw_sem[(lev)-1], 1, _THIS_IP_)
1504#define __sb_writers_release(sb, lev) \
1505 percpu_rwsem_release(&(sb)->s_writers.rw_sem[(lev)-1], 1, _THIS_IP_)
1506
1507/**
1508 * sb_end_write - drop write access to a superblock
1509 * @sb: the super we wrote to
1510 *
1511 * Decrement number of writers to the filesystem. Wake up possible waiters
1512 * wanting to freeze the filesystem.
1513 */
1514static inline void sb_end_write(struct super_block *sb)
1515{
1516 __sb_end_write(sb, SB_FREEZE_WRITE);
1517}
1518
1519/**
1520 * sb_end_pagefault - drop write access to a superblock from a page fault
1521 * @sb: the super we wrote to
1522 *
1523 * Decrement number of processes handling write page fault to the filesystem.
1524 * Wake up possible waiters wanting to freeze the filesystem.
1525 */
1526static inline void sb_end_pagefault(struct super_block *sb)
1527{
1528 __sb_end_write(sb, SB_FREEZE_PAGEFAULT);
1529}
1530
1531/**
1532 * sb_end_intwrite - drop write access to a superblock for internal fs purposes
1533 * @sb: the super we wrote to
1534 *
1535 * Decrement fs-internal number of writers to the filesystem. Wake up possible
1536 * waiters wanting to freeze the filesystem.
1537 */
1538static inline void sb_end_intwrite(struct super_block *sb)
1539{
1540 __sb_end_write(sb, SB_FREEZE_FS);
1541}
1542
1543/**
1544 * sb_start_write - get write access to a superblock
1545 * @sb: the super we write to
1546 *
1547 * When a process wants to write data or metadata to a file system (i.e. dirty
1548 * a page or an inode), it should embed the operation in a sb_start_write() -
1549 * sb_end_write() pair to get exclusion against file system freezing. This
1550 * function increments number of writers preventing freezing. If the file
1551 * system is already frozen, the function waits until the file system is
1552 * thawed.
1553 *
1554 * Since freeze protection behaves as a lock, users have to preserve
1555 * ordering of freeze protection and other filesystem locks. Generally,
1556 * freeze protection should be the outermost lock. In particular, we have:
1557 *
1558 * sb_start_write
1559 * -> i_mutex (write path, truncate, directory ops, ...)
1560 * -> s_umount (freeze_super, thaw_super)
1561 */
1562static inline void sb_start_write(struct super_block *sb)
1563{
1564 __sb_start_write(sb, SB_FREEZE_WRITE, true);
1565}
1566
1567static inline int sb_start_write_trylock(struct super_block *sb)
1568{
1569 return __sb_start_write(sb, SB_FREEZE_WRITE, false);
1570}
1571
1572/**
1573 * sb_start_pagefault - get write access to a superblock from a page fault
1574 * @sb: the super we write to
1575 *
1576 * When a process starts handling write page fault, it should embed the
1577 * operation into sb_start_pagefault() - sb_end_pagefault() pair to get
1578 * exclusion against file system freezing. This is needed since the page fault
1579 * is going to dirty a page. This function increments number of running page
1580 * faults preventing freezing. If the file system is already frozen, the
1581 * function waits until the file system is thawed.
1582 *
1583 * Since page fault freeze protection behaves as a lock, users have to preserve
1584 * ordering of freeze protection and other filesystem locks. It is advised to
1585 * put sb_start_pagefault() close to mmap_sem in lock ordering. Page fault
1586 * handling code implies lock dependency:
1587 *
1588 * mmap_sem
1589 * -> sb_start_pagefault
1590 */
1591static inline void sb_start_pagefault(struct super_block *sb)
1592{
1593 __sb_start_write(sb, SB_FREEZE_PAGEFAULT, true);
1594}
1595
1596/*
1597 * sb_start_intwrite - get write access to a superblock for internal fs purposes
1598 * @sb: the super we write to
1599 *
1600 * This is the third level of protection against filesystem freezing. It is
1601 * free for use by a filesystem. The only requirement is that it must rank
1602 * below sb_start_pagefault.
1603 *
1604 * For example filesystem can call sb_start_intwrite() when starting a
1605 * transaction which somewhat eases handling of freezing for internal sources
1606 * of filesystem changes (internal fs threads, discarding preallocation on file
1607 * close, etc.).
1608 */
1609static inline void sb_start_intwrite(struct super_block *sb)
1610{
1611 __sb_start_write(sb, SB_FREEZE_FS, true);
1612}
1613
1614static inline int sb_start_intwrite_trylock(struct super_block *sb)
1615{
1616 return __sb_start_write(sb, SB_FREEZE_FS, false);
1617}
1618
1619
1620extern bool inode_owner_or_capable(const struct inode *inode);
1621
1622/*
1623 * VFS helper functions..
1624 */
1625extern int vfs_create(struct inode *, struct dentry *, umode_t, bool);
1626extern int vfs_mkdir(struct inode *, struct dentry *, umode_t);
1627extern int vfs_mknod(struct inode *, struct dentry *, umode_t, dev_t);
1628extern int vfs_symlink(struct inode *, struct dentry *, const char *);
1629extern int vfs_link(struct dentry *, struct inode *, struct dentry *, struct inode **);
1630extern int vfs_rmdir(struct inode *, struct dentry *);
1631extern int vfs_unlink(struct inode *, struct dentry *, struct inode **);
1632extern int vfs_rename(struct inode *, struct dentry *, struct inode *, struct dentry *, struct inode **, unsigned int);
1633extern int vfs_whiteout(struct inode *, struct dentry *);
1634
1635extern struct dentry *vfs_tmpfile(struct dentry *dentry, umode_t mode,
1636 int open_flag);
1637
1638int vfs_mkobj(struct dentry *, umode_t,
1639 int (*f)(struct dentry *, umode_t, void *),
1640 void *);
1641
1642extern long vfs_ioctl(struct file *file, unsigned int cmd, unsigned long arg);
1643
1644/*
1645 * VFS file helper functions.
1646 */
1647extern void inode_init_owner(struct inode *inode, const struct inode *dir,
1648 umode_t mode);
1649extern bool may_open_dev(const struct path *path);
1650/*
1651 * VFS FS_IOC_FIEMAP helper definitions.
1652 */
1653struct fiemap_extent_info {
1654 unsigned int fi_flags; /* Flags as passed from user */
1655 unsigned int fi_extents_mapped; /* Number of mapped extents */
1656 unsigned int fi_extents_max; /* Size of fiemap_extent array */
1657 struct fiemap_extent __user *fi_extents_start; /* Start of
1658 fiemap_extent array */
1659};
1660int fiemap_fill_next_extent(struct fiemap_extent_info *info, u64 logical,
1661 u64 phys, u64 len, u32 flags);
1662int fiemap_check_flags(struct fiemap_extent_info *fieinfo, u32 fs_flags);
1663
1664/*
1665 * File types
1666 *
1667 * NOTE! These match bits 12..15 of stat.st_mode
1668 * (ie "(i_mode >> 12) & 15").
1669 */
1670#define DT_UNKNOWN 0
1671#define DT_FIFO 1
1672#define DT_CHR 2
1673#define DT_DIR 4
1674#define DT_BLK 6
1675#define DT_REG 8
1676#define DT_LNK 10
1677#define DT_SOCK 12
1678#define DT_WHT 14
1679
1680/*
1681 * This is the "filldir" function type, used by readdir() to let
1682 * the kernel specify what kind of dirent layout it wants to have.
1683 * This allows the kernel to read directories into kernel space or
1684 * to have different dirent layouts depending on the binary type.
1685 */
1686struct dir_context;
1687typedef int (*filldir_t)(struct dir_context *, const char *, int, loff_t, u64,
1688 unsigned);
1689
1690struct dir_context {
1691 filldir_t actor;
1692 loff_t pos;
1693};
1694
1695struct block_device_operations;
1696
1697/* These macros are for out of kernel modules to test that
1698 * the kernel supports the unlocked_ioctl and compat_ioctl
1699 * fields in struct file_operations. */
1700#define HAVE_COMPAT_IOCTL 1
1701#define HAVE_UNLOCKED_IOCTL 1
1702
1703/*
1704 * These flags let !MMU mmap() govern direct device mapping vs immediate
1705 * copying more easily for MAP_PRIVATE, especially for ROM filesystems.
1706 *
1707 * NOMMU_MAP_COPY: Copy can be mapped (MAP_PRIVATE)
1708 * NOMMU_MAP_DIRECT: Can be mapped directly (MAP_SHARED)
1709 * NOMMU_MAP_READ: Can be mapped for reading
1710 * NOMMU_MAP_WRITE: Can be mapped for writing
1711 * NOMMU_MAP_EXEC: Can be mapped for execution
1712 */
1713#define NOMMU_MAP_COPY 0x00000001
1714#define NOMMU_MAP_DIRECT 0x00000008
1715#define NOMMU_MAP_READ VM_MAYREAD
1716#define NOMMU_MAP_WRITE VM_MAYWRITE
1717#define NOMMU_MAP_EXEC VM_MAYEXEC
1718
1719#define NOMMU_VMFLAGS \
1720 (NOMMU_MAP_READ | NOMMU_MAP_WRITE | NOMMU_MAP_EXEC)
1721
1722
1723struct iov_iter;
1724
1725struct file_operations {
1726 struct module *owner;
1727 loff_t (*llseek) (struct file *, loff_t, int);
1728 ssize_t (*read) (struct file *, char __user *, size_t, loff_t *);
1729 ssize_t (*write) (struct file *, const char __user *, size_t, loff_t *);
1730 ssize_t (*read_iter) (struct kiocb *, struct iov_iter *);
1731 ssize_t (*write_iter) (struct kiocb *, struct iov_iter *);
1732 int (*iterate) (struct file *, struct dir_context *);
1733 int (*iterate_shared) (struct file *, struct dir_context *);
1734 __poll_t (*poll) (struct file *, struct poll_table_struct *);
1735 long (*unlocked_ioctl) (struct file *, unsigned int, unsigned long);
1736 long (*compat_ioctl) (struct file *, unsigned int, unsigned long);
1737 int (*mmap) (struct file *, struct vm_area_struct *);
1738 unsigned long mmap_supported_flags;
1739 int (*open) (struct inode *, struct file *);
1740 int (*flush) (struct file *, fl_owner_t id);
1741 int (*release) (struct inode *, struct file *);
1742 int (*fsync) (struct file *, loff_t, loff_t, int datasync);
1743 int (*fasync) (int, struct file *, int);
1744 int (*lock) (struct file *, int, struct file_lock *);
1745 ssize_t (*sendpage) (struct file *, struct page *, int, size_t, loff_t *, int);
1746 unsigned long (*get_unmapped_area)(struct file *, unsigned long, unsigned long, unsigned long, unsigned long);
1747 int (*check_flags)(int);
1748 int (*flock) (struct file *, int, struct file_lock *);
1749 ssize_t (*splice_write)(struct pipe_inode_info *, struct file *, loff_t *, size_t, unsigned int);
1750 ssize_t (*splice_read)(struct file *, loff_t *, struct pipe_inode_info *, size_t, unsigned int);
1751 int (*setlease)(struct file *, long, struct file_lock **, void **);
1752 long (*fallocate)(struct file *file, int mode, loff_t offset,
1753 loff_t len);
1754 void (*show_fdinfo)(struct seq_file *m, struct file *f);
1755#ifndef CONFIG_MMU
1756 unsigned (*mmap_capabilities)(struct file *);
1757#endif
1758 ssize_t (*copy_file_range)(struct file *, loff_t, struct file *,
1759 loff_t, size_t, unsigned int);
1760 int (*clone_file_range)(struct file *, loff_t, struct file *, loff_t,
1761 u64);
1762 int (*dedupe_file_range)(struct file *, loff_t, struct file *, loff_t,
1763 u64);
1764} __randomize_layout;
1765
1766struct inode_operations {
1767 struct dentry * (*lookup) (struct inode *,struct dentry *, unsigned int);
1768 const char * (*get_link) (struct dentry *, struct inode *, struct delayed_call *);
1769 int (*permission) (struct inode *, int);
1770 struct posix_acl * (*get_acl)(struct inode *, int);
1771
1772 int (*readlink) (struct dentry *, char __user *,int);
1773
1774 int (*create) (struct inode *,struct dentry *, umode_t, bool);
1775 int (*link) (struct dentry *,struct inode *,struct dentry *);
1776 int (*unlink) (struct inode *,struct dentry *);
1777 int (*symlink) (struct inode *,struct dentry *,const char *);
1778 int (*mkdir) (struct inode *,struct dentry *,umode_t);
1779 int (*rmdir) (struct inode *,struct dentry *);
1780 int (*mknod) (struct inode *,struct dentry *,umode_t,dev_t);
1781 int (*rename) (struct inode *, struct dentry *,
1782 struct inode *, struct dentry *, unsigned int);
1783 int (*setattr) (struct dentry *, struct iattr *);
1784 int (*getattr) (const struct path *, struct kstat *, u32, unsigned int);
1785 ssize_t (*listxattr) (struct dentry *, char *, size_t);
1786 int (*fiemap)(struct inode *, struct fiemap_extent_info *, u64 start,
1787 u64 len);
1788 int (*update_time)(struct inode *, struct timespec64 *, int);
1789 int (*atomic_open)(struct inode *, struct dentry *,
1790 struct file *, unsigned open_flag,
1791 umode_t create_mode);
1792 int (*tmpfile) (struct inode *, struct dentry *, umode_t);
1793 int (*set_acl)(struct inode *, struct posix_acl *, int);
1794} ____cacheline_aligned;
1795
1796static inline ssize_t call_read_iter(struct file *file, struct kiocb *kio,
1797 struct iov_iter *iter)
1798{
1799 return file->f_op->read_iter(kio, iter);
1800}
1801
1802static inline ssize_t call_write_iter(struct file *file, struct kiocb *kio,
1803 struct iov_iter *iter)
1804{
1805 return file->f_op->write_iter(kio, iter);
1806}
1807
1808static inline int call_mmap(struct file *file, struct vm_area_struct *vma)
1809{
1810 return file->f_op->mmap(file, vma);
1811}
1812
1813ssize_t rw_copy_check_uvector(int type, const struct iovec __user * uvector,
1814 unsigned long nr_segs, unsigned long fast_segs,
1815 struct iovec *fast_pointer,
1816 struct iovec **ret_pointer);
1817
1818extern ssize_t __vfs_read(struct file *, char __user *, size_t, loff_t *);
1819extern ssize_t vfs_read(struct file *, char __user *, size_t, loff_t *);
1820extern ssize_t vfs_write(struct file *, const char __user *, size_t, loff_t *);
1821extern ssize_t vfs_readv(struct file *, const struct iovec __user *,
1822 unsigned long, loff_t *, rwf_t);
1823extern ssize_t vfs_copy_file_range(struct file *, loff_t , struct file *,
1824 loff_t, size_t, unsigned int);
1825extern int vfs_clone_file_prep_inodes(struct inode *inode_in, loff_t pos_in,
1826 struct inode *inode_out, loff_t pos_out,
1827 u64 *len, bool is_dedupe);
1828extern int vfs_clone_file_range(struct file *file_in, loff_t pos_in,
1829 struct file *file_out, loff_t pos_out, u64 len);
1830extern int vfs_dedupe_file_range_compare(struct inode *src, loff_t srcoff,
1831 struct inode *dest, loff_t destoff,
1832 loff_t len, bool *is_same);
1833extern int vfs_dedupe_file_range(struct file *file,
1834 struct file_dedupe_range *same);
1835extern int vfs_dedupe_file_range_one(struct file *src_file, loff_t src_pos,
1836 struct file *dst_file, loff_t dst_pos,
1837 u64 len);
1838
1839
1840struct super_operations {
1841 struct inode *(*alloc_inode)(struct super_block *sb);
1842 void (*destroy_inode)(struct inode *);
1843
1844 void (*dirty_inode) (struct inode *, int flags);
1845 int (*write_inode) (struct inode *, struct writeback_control *wbc);
1846 int (*drop_inode) (struct inode *);
1847 void (*evict_inode) (struct inode *);
1848 void (*put_super) (struct super_block *);
1849 int (*sync_fs)(struct super_block *sb, int wait);
1850 int (*freeze_super) (struct super_block *);
1851 int (*freeze_fs) (struct super_block *);
1852 int (*thaw_super) (struct super_block *);
1853 int (*unfreeze_fs) (struct super_block *);
1854 int (*statfs) (struct dentry *, struct kstatfs *);
1855 int (*remount_fs) (struct super_block *, int *, char *);
1856 void (*umount_begin) (struct super_block *);
1857
1858 int (*show_options)(struct seq_file *, struct dentry *);
1859 int (*show_devname)(struct seq_file *, struct dentry *);
1860 int (*show_path)(struct seq_file *, struct dentry *);
1861 int (*show_stats)(struct seq_file *, struct dentry *);
1862#ifdef CONFIG_QUOTA
1863 ssize_t (*quota_read)(struct super_block *, int, char *, size_t, loff_t);
1864 ssize_t (*quota_write)(struct super_block *, int, const char *, size_t, loff_t);
1865 struct dquot **(*get_dquots)(struct inode *);
1866#endif
1867 int (*bdev_try_to_free_page)(struct super_block*, struct page*, gfp_t);
1868 long (*nr_cached_objects)(struct super_block *,
1869 struct shrink_control *);
1870 long (*free_cached_objects)(struct super_block *,
1871 struct shrink_control *);
1872};
1873
1874/*
1875 * Inode flags - they have no relation to superblock flags now
1876 */
1877#define S_SYNC 1 /* Writes are synced at once */
1878#define S_NOATIME 2 /* Do not update access times */
1879#define S_APPEND 4 /* Append-only file */
1880#define S_IMMUTABLE 8 /* Immutable file */
1881#define S_DEAD 16 /* removed, but still open directory */
1882#define S_NOQUOTA 32 /* Inode is not counted to quota */
1883#define S_DIRSYNC 64 /* Directory modifications are synchronous */
1884#define S_NOCMTIME 128 /* Do not update file c/mtime */
1885#define S_SWAPFILE 256 /* Do not truncate: swapon got its bmaps */
1886#define S_PRIVATE 512 /* Inode is fs-internal */
1887#define S_IMA 1024 /* Inode has an associated IMA struct */
1888#define S_AUTOMOUNT 2048 /* Automount/referral quasi-directory */
1889#define S_NOSEC 4096 /* no suid or xattr security attributes */
1890#ifdef CONFIG_FS_DAX
1891#define S_DAX 8192 /* Direct Access, avoiding the page cache */
1892#else
1893#define S_DAX 0 /* Make all the DAX code disappear */
1894#endif
1895#define S_ENCRYPTED 16384 /* Encrypted file (using fs/crypto/) */
1896
1897/*
1898 * Note that nosuid etc flags are inode-specific: setting some file-system
1899 * flags just means all the inodes inherit those flags by default. It might be
1900 * possible to override it selectively if you really wanted to with some
1901 * ioctl() that is not currently implemented.
1902 *
1903 * Exception: SB_RDONLY is always applied to the entire file system.
1904 *
1905 * Unfortunately, it is possible to change a filesystems flags with it mounted
1906 * with files in use. This means that all of the inodes will not have their
1907 * i_flags updated. Hence, i_flags no longer inherit the superblock mount
1908 * flags, so these have to be checked separately. -- rmk@arm.uk.linux.org
1909 */
1910#define __IS_FLG(inode, flg) ((inode)->i_sb->s_flags & (flg))
1911
1912static inline bool sb_rdonly(const struct super_block *sb) { return sb->s_flags & SB_RDONLY; }
1913#define IS_RDONLY(inode) sb_rdonly((inode)->i_sb)
1914#define IS_SYNC(inode) (__IS_FLG(inode, SB_SYNCHRONOUS) || \
1915 ((inode)->i_flags & S_SYNC))
1916#define IS_DIRSYNC(inode) (__IS_FLG(inode, SB_SYNCHRONOUS|SB_DIRSYNC) || \
1917 ((inode)->i_flags & (S_SYNC|S_DIRSYNC)))
1918#define IS_MANDLOCK(inode) __IS_FLG(inode, SB_MANDLOCK)
1919#define IS_NOATIME(inode) __IS_FLG(inode, SB_RDONLY|SB_NOATIME)
1920#define IS_I_VERSION(inode) __IS_FLG(inode, SB_I_VERSION)
1921
1922#define IS_NOQUOTA(inode) ((inode)->i_flags & S_NOQUOTA)
1923#define IS_APPEND(inode) ((inode)->i_flags & S_APPEND)
1924#define IS_IMMUTABLE(inode) ((inode)->i_flags & S_IMMUTABLE)
1925#define IS_POSIXACL(inode) __IS_FLG(inode, SB_POSIXACL)
1926
1927#define IS_DEADDIR(inode) ((inode)->i_flags & S_DEAD)
1928#define IS_NOCMTIME(inode) ((inode)->i_flags & S_NOCMTIME)
1929#define IS_SWAPFILE(inode) ((inode)->i_flags & S_SWAPFILE)
1930#define IS_PRIVATE(inode) ((inode)->i_flags & S_PRIVATE)
1931#define IS_IMA(inode) ((inode)->i_flags & S_IMA)
1932#define IS_AUTOMOUNT(inode) ((inode)->i_flags & S_AUTOMOUNT)
1933#define IS_NOSEC(inode) ((inode)->i_flags & S_NOSEC)
1934#define IS_DAX(inode) ((inode)->i_flags & S_DAX)
1935#define IS_ENCRYPTED(inode) ((inode)->i_flags & S_ENCRYPTED)
1936
1937#define IS_WHITEOUT(inode) (S_ISCHR(inode->i_mode) && \
1938 (inode)->i_rdev == WHITEOUT_DEV)
1939
1940static inline bool HAS_UNMAPPED_ID(struct inode *inode)
1941{
1942 return !uid_valid(inode->i_uid) || !gid_valid(inode->i_gid);
1943}
1944
1945static inline enum rw_hint file_write_hint(struct file *file)
1946{
1947 if (file->f_write_hint != WRITE_LIFE_NOT_SET)
1948 return file->f_write_hint;
1949
1950 return file_inode(file)->i_write_hint;
1951}
1952
1953static inline int iocb_flags(struct file *file);
1954
1955static inline u16 ki_hint_validate(enum rw_hint hint)
1956{
1957 typeof(((struct kiocb *)0)->ki_hint) max_hint = -1;
1958
1959 if (hint <= max_hint)
1960 return hint;
1961 return 0;
1962}
1963
1964static inline void init_sync_kiocb(struct kiocb *kiocb, struct file *filp)
1965{
1966 *kiocb = (struct kiocb) {
1967 .ki_filp = filp,
1968 .ki_flags = iocb_flags(filp),
1969 .ki_hint = ki_hint_validate(file_write_hint(filp)),
1970 .ki_ioprio = IOPRIO_PRIO_VALUE(IOPRIO_CLASS_NONE, 0),
1971 };
1972}
1973
1974/*
1975 * Inode state bits. Protected by inode->i_lock
1976 *
1977 * Three bits determine the dirty state of the inode, I_DIRTY_SYNC,
1978 * I_DIRTY_DATASYNC and I_DIRTY_PAGES.
1979 *
1980 * Four bits define the lifetime of an inode. Initially, inodes are I_NEW,
1981 * until that flag is cleared. I_WILL_FREE, I_FREEING and I_CLEAR are set at
1982 * various stages of removing an inode.
1983 *
1984 * Two bits are used for locking and completion notification, I_NEW and I_SYNC.
1985 *
1986 * I_DIRTY_SYNC Inode is dirty, but doesn't have to be written on
1987 * fdatasync(). i_atime is the usual cause.
1988 * I_DIRTY_DATASYNC Data-related inode changes pending. We keep track of
1989 * these changes separately from I_DIRTY_SYNC so that we
1990 * don't have to write inode on fdatasync() when only
1991 * mtime has changed in it.
1992 * I_DIRTY_PAGES Inode has dirty pages. Inode itself may be clean.
1993 * I_NEW Serves as both a mutex and completion notification.
1994 * New inodes set I_NEW. If two processes both create
1995 * the same inode, one of them will release its inode and
1996 * wait for I_NEW to be released before returning.
1997 * Inodes in I_WILL_FREE, I_FREEING or I_CLEAR state can
1998 * also cause waiting on I_NEW, without I_NEW actually
1999 * being set. find_inode() uses this to prevent returning
2000 * nearly-dead inodes.
2001 * I_WILL_FREE Must be set when calling write_inode_now() if i_count
2002 * is zero. I_FREEING must be set when I_WILL_FREE is
2003 * cleared.
2004 * I_FREEING Set when inode is about to be freed but still has dirty
2005 * pages or buffers attached or the inode itself is still
2006 * dirty.
2007 * I_CLEAR Added by clear_inode(). In this state the inode is
2008 * clean and can be destroyed. Inode keeps I_FREEING.
2009 *
2010 * Inodes that are I_WILL_FREE, I_FREEING or I_CLEAR are
2011 * prohibited for many purposes. iget() must wait for
2012 * the inode to be completely released, then create it
2013 * anew. Other functions will just ignore such inodes,
2014 * if appropriate. I_NEW is used for waiting.
2015 *
2016 * I_SYNC Writeback of inode is running. The bit is set during
2017 * data writeback, and cleared with a wakeup on the bit
2018 * address once it is done. The bit is also used to pin
2019 * the inode in memory for flusher thread.
2020 *
2021 * I_REFERENCED Marks the inode as recently references on the LRU list.
2022 *
2023 * I_DIO_WAKEUP Never set. Only used as a key for wait_on_bit().
2024 *
2025 * I_WB_SWITCH Cgroup bdi_writeback switching in progress. Used to
2026 * synchronize competing switching instances and to tell
2027 * wb stat updates to grab the i_pages lock. See
2028 * inode_switch_wb_work_fn() for details.
2029 *
2030 * I_OVL_INUSE Used by overlayfs to get exclusive ownership on upper
2031 * and work dirs among overlayfs mounts.
2032 *
2033 * I_CREATING New object's inode in the middle of setting up.
2034 *
2035 * Q: What is the difference between I_WILL_FREE and I_FREEING?
2036 */
2037#define I_DIRTY_SYNC (1 << 0)
2038#define I_DIRTY_DATASYNC (1 << 1)
2039#define I_DIRTY_PAGES (1 << 2)
2040#define __I_NEW 3
2041#define I_NEW (1 << __I_NEW)
2042#define I_WILL_FREE (1 << 4)
2043#define I_FREEING (1 << 5)
2044#define I_CLEAR (1 << 6)
2045#define __I_SYNC 7
2046#define I_SYNC (1 << __I_SYNC)
2047#define I_REFERENCED (1 << 8)
2048#define __I_DIO_WAKEUP 9
2049#define I_DIO_WAKEUP (1 << __I_DIO_WAKEUP)
2050#define I_LINKABLE (1 << 10)
2051#define I_DIRTY_TIME (1 << 11)
2052#define __I_DIRTY_TIME_EXPIRED 12
2053#define I_DIRTY_TIME_EXPIRED (1 << __I_DIRTY_TIME_EXPIRED)
2054#define I_WB_SWITCH (1 << 13)
2055#define I_OVL_INUSE (1 << 14)
2056#define I_CREATING (1 << 15)
2057
2058#define I_DIRTY_INODE (I_DIRTY_SYNC | I_DIRTY_DATASYNC)
2059#define I_DIRTY (I_DIRTY_INODE | I_DIRTY_PAGES)
2060#define I_DIRTY_ALL (I_DIRTY | I_DIRTY_TIME)
2061
2062extern void __mark_inode_dirty(struct inode *, int);
2063static inline void mark_inode_dirty(struct inode *inode)
2064{
2065 __mark_inode_dirty(inode, I_DIRTY);
2066}
2067
2068static inline void mark_inode_dirty_sync(struct inode *inode)
2069{
2070 __mark_inode_dirty(inode, I_DIRTY_SYNC);
2071}
2072
2073extern void inc_nlink(struct inode *inode);
2074extern void drop_nlink(struct inode *inode);
2075extern void clear_nlink(struct inode *inode);
2076extern void set_nlink(struct inode *inode, unsigned int nlink);
2077
2078static inline void inode_inc_link_count(struct inode *inode)
2079{
2080 inc_nlink(inode);
2081 mark_inode_dirty(inode);
2082}
2083
2084static inline void inode_dec_link_count(struct inode *inode)
2085{
2086 drop_nlink(inode);
2087 mark_inode_dirty(inode);
2088}
2089
2090enum file_time_flags {
2091 S_ATIME = 1,
2092 S_MTIME = 2,
2093 S_CTIME = 4,
2094 S_VERSION = 8,
2095};
2096
2097extern bool atime_needs_update(const struct path *, struct inode *);
2098extern void touch_atime(const struct path *);
2099static inline void file_accessed(struct file *file)
2100{
2101 if (!(file->f_flags & O_NOATIME))
2102 touch_atime(&file->f_path);
2103}
2104
2105int sync_inode(struct inode *inode, struct writeback_control *wbc);
2106int sync_inode_metadata(struct inode *inode, int wait);
2107
2108struct file_system_type {
2109 const char *name;
2110 int fs_flags;
2111#define FS_REQUIRES_DEV 1
2112#define FS_BINARY_MOUNTDATA 2
2113#define FS_HAS_SUBTYPE 4
2114#define FS_USERNS_MOUNT 8 /* Can be mounted by userns root */
2115#define FS_RENAME_DOES_D_MOVE 32768 /* FS will handle d_move() during rename() internally. */
2116 struct dentry *(*mount) (struct file_system_type *, int,
2117 const char *, void *);
2118 void (*kill_sb) (struct super_block *);
2119 struct module *owner;
2120 struct file_system_type * next;
2121 struct hlist_head fs_supers;
2122
2123 struct lock_class_key s_lock_key;
2124 struct lock_class_key s_umount_key;
2125 struct lock_class_key s_vfs_rename_key;
2126 struct lock_class_key s_writers_key[SB_FREEZE_LEVELS];
2127
2128 struct lock_class_key i_lock_key;
2129 struct lock_class_key i_mutex_key;
2130 struct lock_class_key i_mutex_dir_key;
2131};
2132
2133#define MODULE_ALIAS_FS(NAME) MODULE_ALIAS("fs-" NAME)
2134
2135extern struct dentry *mount_ns(struct file_system_type *fs_type,
2136 int flags, void *data, void *ns, struct user_namespace *user_ns,
2137 int (*fill_super)(struct super_block *, void *, int));
2138#ifdef CONFIG_BLOCK
2139extern struct dentry *mount_bdev(struct file_system_type *fs_type,
2140 int flags, const char *dev_name, void *data,
2141 int (*fill_super)(struct super_block *, void *, int));
2142#else
2143static inline struct dentry *mount_bdev(struct file_system_type *fs_type,
2144 int flags, const char *dev_name, void *data,
2145 int (*fill_super)(struct super_block *, void *, int))
2146{
2147 return ERR_PTR(-ENODEV);
2148}
2149#endif
2150extern struct dentry *mount_single(struct file_system_type *fs_type,
2151 int flags, void *data,
2152 int (*fill_super)(struct super_block *, void *, int));
2153extern struct dentry *mount_nodev(struct file_system_type *fs_type,
2154 int flags, void *data,
2155 int (*fill_super)(struct super_block *, void *, int));
2156extern struct dentry *mount_subtree(struct vfsmount *mnt, const char *path);
2157void generic_shutdown_super(struct super_block *sb);
2158#ifdef CONFIG_BLOCK
2159void kill_block_super(struct super_block *sb);
2160#else
2161static inline void kill_block_super(struct super_block *sb)
2162{
2163 BUG();
2164}
2165#endif
2166void kill_anon_super(struct super_block *sb);
2167void kill_litter_super(struct super_block *sb);
2168void deactivate_super(struct super_block *sb);
2169void deactivate_locked_super(struct super_block *sb);
2170int set_anon_super(struct super_block *s, void *data);
2171int get_anon_bdev(dev_t *);
2172void free_anon_bdev(dev_t);
2173struct super_block *sget_userns(struct file_system_type *type,
2174 int (*test)(struct super_block *,void *),
2175 int (*set)(struct super_block *,void *),
2176 int flags, struct user_namespace *user_ns,
2177 void *data);
2178struct super_block *sget(struct file_system_type *type,
2179 int (*test)(struct super_block *,void *),
2180 int (*set)(struct super_block *,void *),
2181 int flags, void *data);
2182extern struct dentry *mount_pseudo_xattr(struct file_system_type *, char *,
2183 const struct super_operations *ops,
2184 const struct xattr_handler **xattr,
2185 const struct dentry_operations *dops,
2186 unsigned long);
2187
2188static inline struct dentry *
2189mount_pseudo(struct file_system_type *fs_type, char *name,
2190 const struct super_operations *ops,
2191 const struct dentry_operations *dops, unsigned long magic)
2192{
2193 return mount_pseudo_xattr(fs_type, name, ops, NULL, dops, magic);
2194}
2195
2196/* Alas, no aliases. Too much hassle with bringing module.h everywhere */
2197#define fops_get(fops) \
2198 (((fops) && try_module_get((fops)->owner) ? (fops) : NULL))
2199#define fops_put(fops) \
2200 do { if (fops) module_put((fops)->owner); } while(0)
2201/*
2202 * This one is to be used *ONLY* from ->open() instances.
2203 * fops must be non-NULL, pinned down *and* module dependencies
2204 * should be sufficient to pin the caller down as well.
2205 */
2206#define replace_fops(f, fops) \
2207 do { \
2208 struct file *__file = (f); \
2209 fops_put(__file->f_op); \
2210 BUG_ON(!(__file->f_op = (fops))); \
2211 } while(0)
2212
2213extern int register_filesystem(struct file_system_type *);
2214extern int unregister_filesystem(struct file_system_type *);
2215extern struct vfsmount *kern_mount_data(struct file_system_type *, void *data);
2216#define kern_mount(type) kern_mount_data(type, NULL)
2217extern void kern_unmount(struct vfsmount *mnt);
2218extern int may_umount_tree(struct vfsmount *);
2219extern int may_umount(struct vfsmount *);
2220extern long do_mount(const char *, const char __user *,
2221 const char *, unsigned long, void *);
2222extern struct vfsmount *collect_mounts(const struct path *);
2223extern void drop_collected_mounts(struct vfsmount *);
2224extern int iterate_mounts(int (*)(struct vfsmount *, void *), void *,
2225 struct vfsmount *);
2226extern int vfs_statfs(const struct path *, struct kstatfs *);
2227extern int user_statfs(const char __user *, struct kstatfs *);
2228extern int fd_statfs(int, struct kstatfs *);
2229extern int freeze_super(struct super_block *super);
2230extern int thaw_super(struct super_block *super);
2231extern bool our_mnt(struct vfsmount *mnt);
2232extern __printf(2, 3)
2233int super_setup_bdi_name(struct super_block *sb, char *fmt, ...);
2234extern int super_setup_bdi(struct super_block *sb);
2235
2236extern int current_umask(void);
2237
2238extern void ihold(struct inode * inode);
2239extern void iput(struct inode *);
2240extern int generic_update_time(struct inode *, struct timespec64 *, int);
2241
2242/* /sys/fs */
2243extern struct kobject *fs_kobj;
2244
2245#define MAX_RW_COUNT (INT_MAX & PAGE_MASK)
2246
2247#ifdef CONFIG_MANDATORY_FILE_LOCKING
2248extern int locks_mandatory_locked(struct file *);
2249extern int locks_mandatory_area(struct inode *, struct file *, loff_t, loff_t, unsigned char);
2250
2251/*
2252 * Candidates for mandatory locking have the setgid bit set
2253 * but no group execute bit - an otherwise meaningless combination.
2254 */
2255
2256static inline int __mandatory_lock(struct inode *ino)
2257{
2258 return (ino->i_mode & (S_ISGID | S_IXGRP)) == S_ISGID;
2259}
2260
2261/*
2262 * ... and these candidates should be on SB_MANDLOCK mounted fs,
2263 * otherwise these will be advisory locks
2264 */
2265
2266static inline int mandatory_lock(struct inode *ino)
2267{
2268 return IS_MANDLOCK(ino) && __mandatory_lock(ino);
2269}
2270
2271static inline int locks_verify_locked(struct file *file)
2272{
2273 if (mandatory_lock(locks_inode(file)))
2274 return locks_mandatory_locked(file);
2275 return 0;
2276}
2277
2278static inline int locks_verify_truncate(struct inode *inode,
2279 struct file *f,
2280 loff_t size)
2281{
2282 if (!inode->i_flctx || !mandatory_lock(inode))
2283 return 0;
2284
2285 if (size < inode->i_size) {
2286 return locks_mandatory_area(inode, f, size, inode->i_size - 1,
2287 F_WRLCK);
2288 } else {
2289 return locks_mandatory_area(inode, f, inode->i_size, size - 1,
2290 F_WRLCK);
2291 }
2292}
2293
2294#else /* !CONFIG_MANDATORY_FILE_LOCKING */
2295
2296static inline int locks_mandatory_locked(struct file *file)
2297{
2298 return 0;
2299}
2300
2301static inline int locks_mandatory_area(struct inode *inode, struct file *filp,
2302 loff_t start, loff_t end, unsigned char type)
2303{
2304 return 0;
2305}
2306
2307static inline int __mandatory_lock(struct inode *inode)
2308{
2309 return 0;
2310}
2311
2312static inline int mandatory_lock(struct inode *inode)
2313{
2314 return 0;
2315}
2316
2317static inline int locks_verify_locked(struct file *file)
2318{
2319 return 0;
2320}
2321
2322static inline int locks_verify_truncate(struct inode *inode, struct file *filp,
2323 size_t size)
2324{
2325 return 0;
2326}
2327
2328#endif /* CONFIG_MANDATORY_FILE_LOCKING */
2329
2330
2331#ifdef CONFIG_FILE_LOCKING
2332static inline int break_lease(struct inode *inode, unsigned int mode)
2333{
2334 /*
2335 * Since this check is lockless, we must ensure that any refcounts
2336 * taken are done before checking i_flctx->flc_lease. Otherwise, we
2337 * could end up racing with tasks trying to set a new lease on this
2338 * file.
2339 */
2340 smp_mb();
2341 if (inode->i_flctx && !list_empty_careful(&inode->i_flctx->flc_lease))
2342 return __break_lease(inode, mode, FL_LEASE);
2343 return 0;
2344}
2345
2346static inline int break_deleg(struct inode *inode, unsigned int mode)
2347{
2348 /*
2349 * Since this check is lockless, we must ensure that any refcounts
2350 * taken are done before checking i_flctx->flc_lease. Otherwise, we
2351 * could end up racing with tasks trying to set a new lease on this
2352 * file.
2353 */
2354 smp_mb();
2355 if (inode->i_flctx && !list_empty_careful(&inode->i_flctx->flc_lease))
2356 return __break_lease(inode, mode, FL_DELEG);
2357 return 0;
2358}
2359
2360static inline int try_break_deleg(struct inode *inode, struct inode **delegated_inode)
2361{
2362 int ret;
2363
2364 ret = break_deleg(inode, O_WRONLY|O_NONBLOCK);
2365 if (ret == -EWOULDBLOCK && delegated_inode) {
2366 *delegated_inode = inode;
2367 ihold(inode);
2368 }
2369 return ret;
2370}
2371
2372static inline int break_deleg_wait(struct inode **delegated_inode)
2373{
2374 int ret;
2375
2376 ret = break_deleg(*delegated_inode, O_WRONLY);
2377 iput(*delegated_inode);
2378 *delegated_inode = NULL;
2379 return ret;
2380}
2381
2382static inline int break_layout(struct inode *inode, bool wait)
2383{
2384 smp_mb();
2385 if (inode->i_flctx && !list_empty_careful(&inode->i_flctx->flc_lease))
2386 return __break_lease(inode,
2387 wait ? O_WRONLY : O_WRONLY | O_NONBLOCK,
2388 FL_LAYOUT);
2389 return 0;
2390}
2391
2392#else /* !CONFIG_FILE_LOCKING */
2393static inline int break_lease(struct inode *inode, unsigned int mode)
2394{
2395 return 0;
2396}
2397
2398static inline int break_deleg(struct inode *inode, unsigned int mode)
2399{
2400 return 0;
2401}
2402
2403static inline int try_break_deleg(struct inode *inode, struct inode **delegated_inode)
2404{
2405 return 0;
2406}
2407
2408static inline int break_deleg_wait(struct inode **delegated_inode)
2409{
2410 BUG();
2411 return 0;
2412}
2413
2414static inline int break_layout(struct inode *inode, bool wait)
2415{
2416 return 0;
2417}
2418
2419#endif /* CONFIG_FILE_LOCKING */
2420
2421/* fs/open.c */
2422struct audit_names;
2423struct filename {
2424 const char *name; /* pointer to actual string */
2425 const __user char *uptr; /* original userland pointer */
2426 int refcnt;
2427 struct audit_names *aname;
2428 const char iname[];
2429};
2430
2431extern long vfs_truncate(const struct path *, loff_t);
2432extern int do_truncate(struct dentry *, loff_t start, unsigned int time_attrs,
2433 struct file *filp);
2434extern int vfs_fallocate(struct file *file, int mode, loff_t offset,
2435 loff_t len);
2436extern long do_sys_open(int dfd, const char __user *filename, int flags,
2437 umode_t mode);
2438extern struct file *file_open_name(struct filename *, int, umode_t);
2439extern struct file *filp_open(const char *, int, umode_t);
2440extern struct file *file_open_root(struct dentry *, struct vfsmount *,
2441 const char *, int, umode_t);
2442extern struct file * dentry_open(const struct path *, int, const struct cred *);
2443extern struct file * open_with_fake_path(const struct path *, int,
2444 struct inode*, const struct cred *);
2445static inline struct file *file_clone_open(struct file *file)
2446{
2447 return dentry_open(&file->f_path, file->f_flags, file->f_cred);
2448}
2449extern int filp_close(struct file *, fl_owner_t id);
2450
2451extern struct filename *getname_flags(const char __user *, int, int *);
2452extern struct filename *getname(const char __user *);
2453extern struct filename *getname_kernel(const char *);
2454extern void putname(struct filename *name);
2455
2456extern int finish_open(struct file *file, struct dentry *dentry,
2457 int (*open)(struct inode *, struct file *));
2458extern int finish_no_open(struct file *file, struct dentry *dentry);
2459
2460/* fs/ioctl.c */
2461
2462extern int ioctl_preallocate(struct file *filp, void __user *argp);
2463
2464/* fs/dcache.c */
2465extern void __init vfs_caches_init_early(void);
2466extern void __init vfs_caches_init(void);
2467
2468extern struct kmem_cache *names_cachep;
2469
2470#define __getname() kmem_cache_alloc(names_cachep, GFP_KERNEL)
2471#define __putname(name) kmem_cache_free(names_cachep, (void *)(name))
2472
2473#ifdef CONFIG_BLOCK
2474extern int register_blkdev(unsigned int, const char *);
2475extern void unregister_blkdev(unsigned int, const char *);
2476extern void bdev_unhash_inode(dev_t dev);
2477extern struct block_device *bdget(dev_t);
2478extern struct block_device *bdgrab(struct block_device *bdev);
2479extern void bd_set_size(struct block_device *, loff_t size);
2480extern void bd_forget(struct inode *inode);
2481extern void bdput(struct block_device *);
2482extern void invalidate_bdev(struct block_device *);
2483extern void iterate_bdevs(void (*)(struct block_device *, void *), void *);
2484extern int sync_blockdev(struct block_device *bdev);
2485extern void kill_bdev(struct block_device *);
2486extern struct super_block *freeze_bdev(struct block_device *);
2487extern void emergency_thaw_all(void);
2488extern void emergency_thaw_bdev(struct super_block *sb);
2489extern int thaw_bdev(struct block_device *bdev, struct super_block *sb);
2490extern int fsync_bdev(struct block_device *);
2491
2492extern struct super_block *blockdev_superblock;
2493
2494static inline bool sb_is_blkdev_sb(struct super_block *sb)
2495{
2496 return sb == blockdev_superblock;
2497}
2498#else
2499static inline void bd_forget(struct inode *inode) {}
2500static inline int sync_blockdev(struct block_device *bdev) { return 0; }
2501static inline void kill_bdev(struct block_device *bdev) {}
2502static inline void invalidate_bdev(struct block_device *bdev) {}
2503
2504static inline struct super_block *freeze_bdev(struct block_device *sb)
2505{
2506 return NULL;
2507}
2508
2509static inline int thaw_bdev(struct block_device *bdev, struct super_block *sb)
2510{
2511 return 0;
2512}
2513
2514static inline int emergency_thaw_bdev(struct super_block *sb)
2515{
2516 return 0;
2517}
2518
2519static inline void iterate_bdevs(void (*f)(struct block_device *, void *), void *arg)
2520{
2521}
2522
2523static inline bool sb_is_blkdev_sb(struct super_block *sb)
2524{
2525 return false;
2526}
2527#endif
2528extern int sync_filesystem(struct super_block *);
2529extern const struct file_operations def_blk_fops;
2530extern const struct file_operations def_chr_fops;
2531#ifdef CONFIG_BLOCK
2532extern int ioctl_by_bdev(struct block_device *, unsigned, unsigned long);
2533extern int blkdev_ioctl(struct block_device *, fmode_t, unsigned, unsigned long);
2534extern long compat_blkdev_ioctl(struct file *, unsigned, unsigned long);
2535extern int blkdev_get(struct block_device *bdev, fmode_t mode, void *holder);
2536extern struct block_device *blkdev_get_by_path(const char *path, fmode_t mode,
2537 void *holder);
2538extern struct block_device *blkdev_get_by_dev(dev_t dev, fmode_t mode,
2539 void *holder);
2540extern void blkdev_put(struct block_device *bdev, fmode_t mode);
2541extern int __blkdev_reread_part(struct block_device *bdev);
2542extern int blkdev_reread_part(struct block_device *bdev);
2543
2544#ifdef CONFIG_SYSFS
2545extern int bd_link_disk_holder(struct block_device *bdev, struct gendisk *disk);
2546extern void bd_unlink_disk_holder(struct block_device *bdev,
2547 struct gendisk *disk);
2548#else
2549static inline int bd_link_disk_holder(struct block_device *bdev,
2550 struct gendisk *disk)
2551{
2552 return 0;
2553}
2554static inline void bd_unlink_disk_holder(struct block_device *bdev,
2555 struct gendisk *disk)
2556{
2557}
2558#endif
2559#endif
2560
2561/* fs/char_dev.c */
2562#define CHRDEV_MAJOR_MAX 512
2563/* Marks the bottom of the first segment of free char majors */
2564#define CHRDEV_MAJOR_DYN_END 234
2565/* Marks the top and bottom of the second segment of free char majors */
2566#define CHRDEV_MAJOR_DYN_EXT_START 511
2567#define CHRDEV_MAJOR_DYN_EXT_END 384
2568
2569extern int alloc_chrdev_region(dev_t *, unsigned, unsigned, const char *);
2570extern int register_chrdev_region(dev_t, unsigned, const char *);
2571extern int __register_chrdev(unsigned int major, unsigned int baseminor,
2572 unsigned int count, const char *name,
2573 const struct file_operations *fops);
2574extern void __unregister_chrdev(unsigned int major, unsigned int baseminor,
2575 unsigned int count, const char *name);
2576extern void unregister_chrdev_region(dev_t, unsigned);
2577extern void chrdev_show(struct seq_file *,off_t);
2578
2579static inline int register_chrdev(unsigned int major, const char *name,
2580 const struct file_operations *fops)
2581{
2582 return __register_chrdev(major, 0, 256, name, fops);
2583}
2584
2585static inline void unregister_chrdev(unsigned int major, const char *name)
2586{
2587 __unregister_chrdev(major, 0, 256, name);
2588}
2589
2590/* fs/block_dev.c */
2591#define BDEVNAME_SIZE 32 /* Largest string for a blockdev identifier */
2592#define BDEVT_SIZE 10 /* Largest string for MAJ:MIN for blkdev */
2593
2594#ifdef CONFIG_BLOCK
2595#define BLKDEV_MAJOR_MAX 512
2596extern const char *__bdevname(dev_t, char *buffer);
2597extern const char *bdevname(struct block_device *bdev, char *buffer);
2598extern struct block_device *lookup_bdev(const char *);
2599extern void blkdev_show(struct seq_file *,off_t);
2600
2601#else
2602#define BLKDEV_MAJOR_MAX 0
2603#endif
2604
2605extern void init_special_inode(struct inode *, umode_t, dev_t);
2606
2607/* Invalid inode operations -- fs/bad_inode.c */
2608extern void make_bad_inode(struct inode *);
2609extern bool is_bad_inode(struct inode *);
2610
2611#ifdef CONFIG_BLOCK
2612extern void check_disk_size_change(struct gendisk *disk,
2613 struct block_device *bdev, bool verbose);
2614extern int revalidate_disk(struct gendisk *);
2615extern int check_disk_change(struct block_device *);
2616extern int __invalidate_device(struct block_device *, bool);
2617extern int invalidate_partition(struct gendisk *, int);
2618#endif
2619unsigned long invalidate_mapping_pages(struct address_space *mapping,
2620 pgoff_t start, pgoff_t end);
2621
2622static inline void invalidate_remote_inode(struct inode *inode)
2623{
2624 if (S_ISREG(inode->i_mode) || S_ISDIR(inode->i_mode) ||
2625 S_ISLNK(inode->i_mode))
2626 invalidate_mapping_pages(inode->i_mapping, 0, -1);
2627}
2628extern int invalidate_inode_pages2(struct address_space *mapping);
2629extern int invalidate_inode_pages2_range(struct address_space *mapping,
2630 pgoff_t start, pgoff_t end);
2631extern int write_inode_now(struct inode *, int);
2632extern int filemap_fdatawrite(struct address_space *);
2633extern int filemap_flush(struct address_space *);
2634extern int filemap_fdatawait_keep_errors(struct address_space *mapping);
2635extern int filemap_fdatawait_range(struct address_space *, loff_t lstart,
2636 loff_t lend);
2637
2638static inline int filemap_fdatawait(struct address_space *mapping)
2639{
2640 return filemap_fdatawait_range(mapping, 0, LLONG_MAX);
2641}
2642
2643extern bool filemap_range_has_page(struct address_space *, loff_t lstart,
2644 loff_t lend);
2645extern int filemap_write_and_wait(struct address_space *mapping);
2646extern int filemap_write_and_wait_range(struct address_space *mapping,
2647 loff_t lstart, loff_t lend);
2648extern int __filemap_fdatawrite_range(struct address_space *mapping,
2649 loff_t start, loff_t end, int sync_mode);
2650extern int filemap_fdatawrite_range(struct address_space *mapping,
2651 loff_t start, loff_t end);
2652extern int filemap_check_errors(struct address_space *mapping);
2653extern void __filemap_set_wb_err(struct address_space *mapping, int err);
2654
2655extern int __must_check file_fdatawait_range(struct file *file, loff_t lstart,
2656 loff_t lend);
2657extern int __must_check file_check_and_advance_wb_err(struct file *file);
2658extern int __must_check file_write_and_wait_range(struct file *file,
2659 loff_t start, loff_t end);
2660
2661static inline int file_write_and_wait(struct file *file)
2662{
2663 return file_write_and_wait_range(file, 0, LLONG_MAX);
2664}
2665
2666/**
2667 * filemap_set_wb_err - set a writeback error on an address_space
2668 * @mapping: mapping in which to set writeback error
2669 * @err: error to be set in mapping
2670 *
2671 * When writeback fails in some way, we must record that error so that
2672 * userspace can be informed when fsync and the like are called. We endeavor
2673 * to report errors on any file that was open at the time of the error. Some
2674 * internal callers also need to know when writeback errors have occurred.
2675 *
2676 * When a writeback error occurs, most filesystems will want to call
2677 * filemap_set_wb_err to record the error in the mapping so that it will be
2678 * automatically reported whenever fsync is called on the file.
2679 */
2680static inline void filemap_set_wb_err(struct address_space *mapping, int err)
2681{
2682 /* Fastpath for common case of no error */
2683 if (unlikely(err))
2684 __filemap_set_wb_err(mapping, err);
2685}
2686
2687/**
2688 * filemap_check_wb_error - has an error occurred since the mark was sampled?
2689 * @mapping: mapping to check for writeback errors
2690 * @since: previously-sampled errseq_t
2691 *
2692 * Grab the errseq_t value from the mapping, and see if it has changed "since"
2693 * the given value was sampled.
2694 *
2695 * If it has then report the latest error set, otherwise return 0.
2696 */
2697static inline int filemap_check_wb_err(struct address_space *mapping,
2698 errseq_t since)
2699{
2700 return errseq_check(&mapping->wb_err, since);
2701}
2702
2703/**
2704 * filemap_sample_wb_err - sample the current errseq_t to test for later errors
2705 * @mapping: mapping to be sampled
2706 *
2707 * Writeback errors are always reported relative to a particular sample point
2708 * in the past. This function provides those sample points.
2709 */
2710static inline errseq_t filemap_sample_wb_err(struct address_space *mapping)
2711{
2712 return errseq_sample(&mapping->wb_err);
2713}
2714
2715extern int vfs_fsync_range(struct file *file, loff_t start, loff_t end,
2716 int datasync);
2717extern int vfs_fsync(struct file *file, int datasync);
2718
2719/*
2720 * Sync the bytes written if this was a synchronous write. Expect ki_pos
2721 * to already be updated for the write, and will return either the amount
2722 * of bytes passed in, or an error if syncing the file failed.
2723 */
2724static inline ssize_t generic_write_sync(struct kiocb *iocb, ssize_t count)
2725{
2726 if (iocb->ki_flags & IOCB_DSYNC) {
2727 int ret = vfs_fsync_range(iocb->ki_filp,
2728 iocb->ki_pos - count, iocb->ki_pos - 1,
2729 (iocb->ki_flags & IOCB_SYNC) ? 0 : 1);
2730 if (ret)
2731 return ret;
2732 }
2733
2734 return count;
2735}
2736
2737extern void emergency_sync(void);
2738extern void emergency_remount(void);
2739#ifdef CONFIG_BLOCK
2740extern sector_t bmap(struct inode *, sector_t);
2741#endif
2742extern int notify_change(struct dentry *, struct iattr *, struct inode **);
2743extern int inode_permission(struct inode *, int);
2744extern int generic_permission(struct inode *, int);
2745extern int __check_sticky(struct inode *dir, struct inode *inode);
2746
2747static inline bool execute_ok(struct inode *inode)
2748{
2749 return (inode->i_mode & S_IXUGO) || S_ISDIR(inode->i_mode);
2750}
2751
2752static inline void file_start_write(struct file *file)
2753{
2754 if (!S_ISREG(file_inode(file)->i_mode))
2755 return;
2756 __sb_start_write(file_inode(file)->i_sb, SB_FREEZE_WRITE, true);
2757}
2758
2759static inline bool file_start_write_trylock(struct file *file)
2760{
2761 if (!S_ISREG(file_inode(file)->i_mode))
2762 return true;
2763 return __sb_start_write(file_inode(file)->i_sb, SB_FREEZE_WRITE, false);
2764}
2765
2766static inline void file_end_write(struct file *file)
2767{
2768 if (!S_ISREG(file_inode(file)->i_mode))
2769 return;
2770 __sb_end_write(file_inode(file)->i_sb, SB_FREEZE_WRITE);
2771}
2772
2773static inline int do_clone_file_range(struct file *file_in, loff_t pos_in,
2774 struct file *file_out, loff_t pos_out,
2775 u64 len)
2776{
2777 int ret;
2778
2779 file_start_write(file_out);
2780 ret = vfs_clone_file_range(file_in, pos_in, file_out, pos_out, len);
2781 file_end_write(file_out);
2782
2783 return ret;
2784}
2785
2786/*
2787 * get_write_access() gets write permission for a file.
2788 * put_write_access() releases this write permission.
2789 * This is used for regular files.
2790 * We cannot support write (and maybe mmap read-write shared) accesses and
2791 * MAP_DENYWRITE mmappings simultaneously. The i_writecount field of an inode
2792 * can have the following values:
2793 * 0: no writers, no VM_DENYWRITE mappings
2794 * < 0: (-i_writecount) vm_area_structs with VM_DENYWRITE set exist
2795 * > 0: (i_writecount) users are writing to the file.
2796 *
2797 * Normally we operate on that counter with atomic_{inc,dec} and it's safe
2798 * except for the cases where we don't hold i_writecount yet. Then we need to
2799 * use {get,deny}_write_access() - these functions check the sign and refuse
2800 * to do the change if sign is wrong.
2801 */
2802static inline int get_write_access(struct inode *inode)
2803{
2804 return atomic_inc_unless_negative(&inode->i_writecount) ? 0 : -ETXTBSY;
2805}
2806static inline int deny_write_access(struct file *file)
2807{
2808 struct inode *inode = file_inode(file);
2809 return atomic_dec_unless_positive(&inode->i_writecount) ? 0 : -ETXTBSY;
2810}
2811static inline void put_write_access(struct inode * inode)
2812{
2813 atomic_dec(&inode->i_writecount);
2814}
2815static inline void allow_write_access(struct file *file)
2816{
2817 if (file)
2818 atomic_inc(&file_inode(file)->i_writecount);
2819}
2820static inline bool inode_is_open_for_write(const struct inode *inode)
2821{
2822 return atomic_read(&inode->i_writecount) > 0;
2823}
2824
2825#ifdef CONFIG_IMA
2826static inline void i_readcount_dec(struct inode *inode)
2827{
2828 BUG_ON(!atomic_read(&inode->i_readcount));
2829 atomic_dec(&inode->i_readcount);
2830}
2831static inline void i_readcount_inc(struct inode *inode)
2832{
2833 atomic_inc(&inode->i_readcount);
2834}
2835#else
2836static inline void i_readcount_dec(struct inode *inode)
2837{
2838 return;
2839}
2840static inline void i_readcount_inc(struct inode *inode)
2841{
2842 return;
2843}
2844#endif
2845extern int do_pipe_flags(int *, int);
2846
2847#define __kernel_read_file_id(id) \
2848 id(UNKNOWN, unknown) \
2849 id(FIRMWARE, firmware) \
2850 id(FIRMWARE_PREALLOC_BUFFER, firmware) \
2851 id(MODULE, kernel-module) \
2852 id(KEXEC_IMAGE, kexec-image) \
2853 id(KEXEC_INITRAMFS, kexec-initramfs) \
2854 id(POLICY, security-policy) \
2855 id(X509_CERTIFICATE, x509-certificate) \
2856 id(MAX_ID, )
2857
2858#define __fid_enumify(ENUM, dummy) READING_ ## ENUM,
2859#define __fid_stringify(dummy, str) #str,
2860
2861enum kernel_read_file_id {
2862 __kernel_read_file_id(__fid_enumify)
2863};
2864
2865static const char * const kernel_read_file_str[] = {
2866 __kernel_read_file_id(__fid_stringify)
2867};
2868
2869static inline const char *kernel_read_file_id_str(enum kernel_read_file_id id)
2870{
2871 if ((unsigned)id >= READING_MAX_ID)
2872 return kernel_read_file_str[READING_UNKNOWN];
2873
2874 return kernel_read_file_str[id];
2875}
2876
2877extern int kernel_read_file(struct file *, void **, loff_t *, loff_t,
2878 enum kernel_read_file_id);
2879extern int kernel_read_file_from_path(const char *, void **, loff_t *, loff_t,
2880 enum kernel_read_file_id);
2881extern int kernel_read_file_from_fd(int, void **, loff_t *, loff_t,
2882 enum kernel_read_file_id);
2883extern ssize_t kernel_read(struct file *, void *, size_t, loff_t *);
2884extern ssize_t kernel_write(struct file *, const void *, size_t, loff_t *);
2885extern ssize_t __kernel_write(struct file *, const void *, size_t, loff_t *);
2886extern struct file * open_exec(const char *);
2887
2888/* fs/dcache.c -- generic fs support functions */
2889extern bool is_subdir(struct dentry *, struct dentry *);
2890extern bool path_is_under(const struct path *, const struct path *);
2891
2892extern char *file_path(struct file *, char *, int);
2893
2894#include <linux/err.h>
2895
2896/* needed for stackable file system support */
2897extern loff_t default_llseek(struct file *file, loff_t offset, int whence);
2898
2899extern loff_t vfs_llseek(struct file *file, loff_t offset, int whence);
2900
2901extern int inode_init_always(struct super_block *, struct inode *);
2902extern void inode_init_once(struct inode *);
2903extern void address_space_init_once(struct address_space *mapping);
2904extern struct inode * igrab(struct inode *);
2905extern ino_t iunique(struct super_block *, ino_t);
2906extern int inode_needs_sync(struct inode *inode);
2907extern int generic_delete_inode(struct inode *inode);
2908static inline int generic_drop_inode(struct inode *inode)
2909{
2910 return !inode->i_nlink || inode_unhashed(inode);
2911}
2912
2913extern struct inode *ilookup5_nowait(struct super_block *sb,
2914 unsigned long hashval, int (*test)(struct inode *, void *),
2915 void *data);
2916extern struct inode *ilookup5(struct super_block *sb, unsigned long hashval,
2917 int (*test)(struct inode *, void *), void *data);
2918extern struct inode *ilookup(struct super_block *sb, unsigned long ino);
2919
2920extern struct inode *inode_insert5(struct inode *inode, unsigned long hashval,
2921 int (*test)(struct inode *, void *),
2922 int (*set)(struct inode *, void *),
2923 void *data);
2924extern struct inode * iget5_locked(struct super_block *, unsigned long, int (*test)(struct inode *, void *), int (*set)(struct inode *, void *), void *);
2925extern struct inode * iget_locked(struct super_block *, unsigned long);
2926extern struct inode *find_inode_nowait(struct super_block *,
2927 unsigned long,
2928 int (*match)(struct inode *,
2929 unsigned long, void *),
2930 void *data);
2931extern int insert_inode_locked4(struct inode *, unsigned long, int (*test)(struct inode *, void *), void *);
2932extern int insert_inode_locked(struct inode *);
2933#ifdef CONFIG_DEBUG_LOCK_ALLOC
2934extern void lockdep_annotate_inode_mutex_key(struct inode *inode);
2935#else
2936static inline void lockdep_annotate_inode_mutex_key(struct inode *inode) { };
2937#endif
2938extern void unlock_new_inode(struct inode *);
2939extern void discard_new_inode(struct inode *);
2940extern unsigned int get_next_ino(void);
2941extern void evict_inodes(struct super_block *sb);
2942
2943extern void __iget(struct inode * inode);
2944extern void iget_failed(struct inode *);
2945extern void clear_inode(struct inode *);
2946extern void __destroy_inode(struct inode *);
2947extern struct inode *new_inode_pseudo(struct super_block *sb);
2948extern struct inode *new_inode(struct super_block *sb);
2949extern void free_inode_nonrcu(struct inode *inode);
2950extern int should_remove_suid(struct dentry *);
2951extern int file_remove_privs(struct file *);
2952
2953extern void __insert_inode_hash(struct inode *, unsigned long hashval);
2954static inline void insert_inode_hash(struct inode *inode)
2955{
2956 __insert_inode_hash(inode, inode->i_ino);
2957}
2958
2959extern void __remove_inode_hash(struct inode *);
2960static inline void remove_inode_hash(struct inode *inode)
2961{
2962 if (!inode_unhashed(inode) && !hlist_fake(&inode->i_hash))
2963 __remove_inode_hash(inode);
2964}
2965
2966extern void inode_sb_list_add(struct inode *inode);
2967
2968#ifdef CONFIG_BLOCK
2969extern int bdev_read_only(struct block_device *);
2970#endif
2971extern int set_blocksize(struct block_device *, int);
2972extern int sb_set_blocksize(struct super_block *, int);
2973extern int sb_min_blocksize(struct super_block *, int);
2974
2975extern int generic_file_mmap(struct file *, struct vm_area_struct *);
2976extern int generic_file_readonly_mmap(struct file *, struct vm_area_struct *);
2977extern ssize_t generic_write_checks(struct kiocb *, struct iov_iter *);
2978extern ssize_t generic_file_read_iter(struct kiocb *, struct iov_iter *);
2979extern ssize_t __generic_file_write_iter(struct kiocb *, struct iov_iter *);
2980extern ssize_t generic_file_write_iter(struct kiocb *, struct iov_iter *);
2981extern ssize_t generic_file_direct_write(struct kiocb *, struct iov_iter *);
2982extern ssize_t generic_perform_write(struct file *, struct iov_iter *, loff_t);
2983
2984ssize_t vfs_iter_read(struct file *file, struct iov_iter *iter, loff_t *ppos,
2985 rwf_t flags);
2986ssize_t vfs_iter_write(struct file *file, struct iov_iter *iter, loff_t *ppos,
2987 rwf_t flags);
2988
2989/* fs/block_dev.c */
2990extern ssize_t blkdev_read_iter(struct kiocb *iocb, struct iov_iter *to);
2991extern ssize_t blkdev_write_iter(struct kiocb *iocb, struct iov_iter *from);
2992extern int blkdev_fsync(struct file *filp, loff_t start, loff_t end,
2993 int datasync);
2994extern void block_sync_page(struct page *page);
2995
2996/* fs/splice.c */
2997extern ssize_t generic_file_splice_read(struct file *, loff_t *,
2998 struct pipe_inode_info *, size_t, unsigned int);
2999extern ssize_t iter_file_splice_write(struct pipe_inode_info *,
3000 struct file *, loff_t *, size_t, unsigned int);
3001extern ssize_t generic_splice_sendpage(struct pipe_inode_info *pipe,
3002 struct file *out, loff_t *, size_t len, unsigned int flags);
3003extern long do_splice_direct(struct file *in, loff_t *ppos, struct file *out,
3004 loff_t *opos, size_t len, unsigned int flags);
3005
3006
3007extern void
3008file_ra_state_init(struct file_ra_state *ra, struct address_space *mapping);
3009extern loff_t noop_llseek(struct file *file, loff_t offset, int whence);
3010extern loff_t no_llseek(struct file *file, loff_t offset, int whence);
3011extern loff_t vfs_setpos(struct file *file, loff_t offset, loff_t maxsize);
3012extern loff_t generic_file_llseek(struct file *file, loff_t offset, int whence);
3013extern loff_t generic_file_llseek_size(struct file *file, loff_t offset,
3014 int whence, loff_t maxsize, loff_t eof);
3015extern loff_t fixed_size_llseek(struct file *file, loff_t offset,
3016 int whence, loff_t size);
3017extern loff_t no_seek_end_llseek_size(struct file *, loff_t, int, loff_t);
3018extern loff_t no_seek_end_llseek(struct file *, loff_t, int);
3019extern int generic_file_open(struct inode * inode, struct file * filp);
3020extern int nonseekable_open(struct inode * inode, struct file * filp);
3021
3022#ifdef CONFIG_BLOCK
3023typedef void (dio_submit_t)(struct bio *bio, struct inode *inode,
3024 loff_t file_offset);
3025
3026enum {
3027 /* need locking between buffered and direct access */
3028 DIO_LOCKING = 0x01,
3029
3030 /* filesystem does not support filling holes */
3031 DIO_SKIP_HOLES = 0x02,
3032};
3033
3034void dio_end_io(struct bio *bio);
3035void dio_warn_stale_pagecache(struct file *filp);
3036
3037ssize_t __blockdev_direct_IO(struct kiocb *iocb, struct inode *inode,
3038 struct block_device *bdev, struct iov_iter *iter,
3039 get_block_t get_block,
3040 dio_iodone_t end_io, dio_submit_t submit_io,
3041 int flags);
3042
3043static inline ssize_t blockdev_direct_IO(struct kiocb *iocb,
3044 struct inode *inode,
3045 struct iov_iter *iter,
3046 get_block_t get_block)
3047{
3048 return __blockdev_direct_IO(iocb, inode, inode->i_sb->s_bdev, iter,
3049 get_block, NULL, NULL, DIO_LOCKING | DIO_SKIP_HOLES);
3050}
3051#endif
3052
3053void inode_dio_wait(struct inode *inode);
3054
3055/*
3056 * inode_dio_begin - signal start of a direct I/O requests
3057 * @inode: inode the direct I/O happens on
3058 *
3059 * This is called once we've finished processing a direct I/O request,
3060 * and is used to wake up callers waiting for direct I/O to be quiesced.
3061 */
3062static inline void inode_dio_begin(struct inode *inode)
3063{
3064 atomic_inc(&inode->i_dio_count);
3065}
3066
3067/*
3068 * inode_dio_end - signal finish of a direct I/O requests
3069 * @inode: inode the direct I/O happens on
3070 *
3071 * This is called once we've finished processing a direct I/O request,
3072 * and is used to wake up callers waiting for direct I/O to be quiesced.
3073 */
3074static inline void inode_dio_end(struct inode *inode)
3075{
3076 if (atomic_dec_and_test(&inode->i_dio_count))
3077 wake_up_bit(&inode->i_state, __I_DIO_WAKEUP);
3078}
3079
3080extern void inode_set_flags(struct inode *inode, unsigned int flags,
3081 unsigned int mask);
3082
3083extern const struct file_operations generic_ro_fops;
3084
3085#define special_file(m) (S_ISCHR(m)||S_ISBLK(m)||S_ISFIFO(m)||S_ISSOCK(m))
3086
3087extern int readlink_copy(char __user *, int, const char *);
3088extern int page_readlink(struct dentry *, char __user *, int);
3089extern const char *page_get_link(struct dentry *, struct inode *,
3090 struct delayed_call *);
3091extern void page_put_link(void *);
3092extern int __page_symlink(struct inode *inode, const char *symname, int len,
3093 int nofs);
3094extern int page_symlink(struct inode *inode, const char *symname, int len);
3095extern const struct inode_operations page_symlink_inode_operations;
3096extern void kfree_link(void *);
3097extern void generic_fillattr(struct inode *, struct kstat *);
3098extern int vfs_getattr_nosec(const struct path *, struct kstat *, u32, unsigned int);
3099extern int vfs_getattr(const struct path *, struct kstat *, u32, unsigned int);
3100void __inode_add_bytes(struct inode *inode, loff_t bytes);
3101void inode_add_bytes(struct inode *inode, loff_t bytes);
3102void __inode_sub_bytes(struct inode *inode, loff_t bytes);
3103void inode_sub_bytes(struct inode *inode, loff_t bytes);
3104static inline loff_t __inode_get_bytes(struct inode *inode)
3105{
3106 return (((loff_t)inode->i_blocks) << 9) + inode->i_bytes;
3107}
3108loff_t inode_get_bytes(struct inode *inode);
3109void inode_set_bytes(struct inode *inode, loff_t bytes);
3110const char *simple_get_link(struct dentry *, struct inode *,
3111 struct delayed_call *);
3112extern const struct inode_operations simple_symlink_inode_operations;
3113
3114extern int iterate_dir(struct file *, struct dir_context *);
3115
3116extern int vfs_statx(int, const char __user *, int, struct kstat *, u32);
3117extern int vfs_statx_fd(unsigned int, struct kstat *, u32, unsigned int);
3118
3119static inline int vfs_stat(const char __user *filename, struct kstat *stat)
3120{
3121 return vfs_statx(AT_FDCWD, filename, AT_NO_AUTOMOUNT,
3122 stat, STATX_BASIC_STATS);
3123}
3124static inline int vfs_lstat(const char __user *name, struct kstat *stat)
3125{
3126 return vfs_statx(AT_FDCWD, name, AT_SYMLINK_NOFOLLOW | AT_NO_AUTOMOUNT,
3127 stat, STATX_BASIC_STATS);
3128}
3129static inline int vfs_fstatat(int dfd, const char __user *filename,
3130 struct kstat *stat, int flags)
3131{
3132 return vfs_statx(dfd, filename, flags | AT_NO_AUTOMOUNT,
3133 stat, STATX_BASIC_STATS);
3134}
3135static inline int vfs_fstat(int fd, struct kstat *stat)
3136{
3137 return vfs_statx_fd(fd, stat, STATX_BASIC_STATS, 0);
3138}
3139
3140
3141extern const char *vfs_get_link(struct dentry *, struct delayed_call *);
3142extern int vfs_readlink(struct dentry *, char __user *, int);
3143
3144extern int __generic_block_fiemap(struct inode *inode,
3145 struct fiemap_extent_info *fieinfo,
3146 loff_t start, loff_t len,
3147 get_block_t *get_block);
3148extern int generic_block_fiemap(struct inode *inode,
3149 struct fiemap_extent_info *fieinfo, u64 start,
3150 u64 len, get_block_t *get_block);
3151
3152extern struct file_system_type *get_filesystem(struct file_system_type *fs);
3153extern void put_filesystem(struct file_system_type *fs);
3154extern struct file_system_type *get_fs_type(const char *name);
3155extern struct super_block *get_super(struct block_device *);
3156extern struct super_block *get_super_thawed(struct block_device *);
3157extern struct super_block *get_super_exclusive_thawed(struct block_device *bdev);
3158extern struct super_block *get_active_super(struct block_device *bdev);
3159extern void drop_super(struct super_block *sb);
3160extern void drop_super_exclusive(struct super_block *sb);
3161extern void iterate_supers(void (*)(struct super_block *, void *), void *);
3162extern void iterate_supers_type(struct file_system_type *,
3163 void (*)(struct super_block *, void *), void *);
3164
3165extern int dcache_dir_open(struct inode *, struct file *);
3166extern int dcache_dir_close(struct inode *, struct file *);
3167extern loff_t dcache_dir_lseek(struct file *, loff_t, int);
3168extern int dcache_readdir(struct file *, struct dir_context *);
3169extern int simple_setattr(struct dentry *, struct iattr *);
3170extern int simple_getattr(const struct path *, struct kstat *, u32, unsigned int);
3171extern int simple_statfs(struct dentry *, struct kstatfs *);
3172extern int simple_open(struct inode *inode, struct file *file);
3173extern int simple_link(struct dentry *, struct inode *, struct dentry *);
3174extern int simple_unlink(struct inode *, struct dentry *);
3175extern int simple_rmdir(struct inode *, struct dentry *);
3176extern int simple_rename(struct inode *, struct dentry *,
3177 struct inode *, struct dentry *, unsigned int);
3178extern int noop_fsync(struct file *, loff_t, loff_t, int);
3179extern int noop_set_page_dirty(struct page *page);
3180extern void noop_invalidatepage(struct page *page, unsigned int offset,
3181 unsigned int length);
3182extern ssize_t noop_direct_IO(struct kiocb *iocb, struct iov_iter *iter);
3183extern int simple_empty(struct dentry *);
3184extern int simple_readpage(struct file *file, struct page *page);
3185extern int simple_write_begin(struct file *file, struct address_space *mapping,
3186 loff_t pos, unsigned len, unsigned flags,
3187 struct page **pagep, void **fsdata);
3188extern int simple_write_end(struct file *file, struct address_space *mapping,
3189 loff_t pos, unsigned len, unsigned copied,
3190 struct page *page, void *fsdata);
3191extern int always_delete_dentry(const struct dentry *);
3192extern struct inode *alloc_anon_inode(struct super_block *);
3193extern int simple_nosetlease(struct file *, long, struct file_lock **, void **);
3194extern const struct dentry_operations simple_dentry_operations;
3195
3196extern struct dentry *simple_lookup(struct inode *, struct dentry *, unsigned int flags);
3197extern ssize_t generic_read_dir(struct file *, char __user *, size_t, loff_t *);
3198extern const struct file_operations simple_dir_operations;
3199extern const struct inode_operations simple_dir_inode_operations;
3200extern void make_empty_dir_inode(struct inode *inode);
3201extern bool is_empty_dir_inode(struct inode *inode);
3202struct tree_descr { const char *name; const struct file_operations *ops; int mode; };
3203struct dentry *d_alloc_name(struct dentry *, const char *);
3204extern int simple_fill_super(struct super_block *, unsigned long,
3205 const struct tree_descr *);
3206extern int simple_pin_fs(struct file_system_type *, struct vfsmount **mount, int *count);
3207extern void simple_release_fs(struct vfsmount **mount, int *count);
3208
3209extern ssize_t simple_read_from_buffer(void __user *to, size_t count,
3210 loff_t *ppos, const void *from, size_t available);
3211extern ssize_t simple_write_to_buffer(void *to, size_t available, loff_t *ppos,
3212 const void __user *from, size_t count);
3213
3214extern int __generic_file_fsync(struct file *, loff_t, loff_t, int);
3215extern int generic_file_fsync(struct file *, loff_t, loff_t, int);
3216
3217extern int generic_check_addressable(unsigned, u64);
3218
3219#ifdef CONFIG_MIGRATION
3220extern int buffer_migrate_page(struct address_space *,
3221 struct page *, struct page *,
3222 enum migrate_mode);
3223#else
3224#define buffer_migrate_page NULL
3225#endif
3226
3227extern int setattr_prepare(struct dentry *, struct iattr *);
3228extern int inode_newsize_ok(const struct inode *, loff_t offset);
3229extern void setattr_copy(struct inode *inode, const struct iattr *attr);
3230
3231extern int file_update_time(struct file *file);
3232
3233static inline bool io_is_direct(struct file *filp)
3234{
3235 return (filp->f_flags & O_DIRECT) || IS_DAX(filp->f_mapping->host);
3236}
3237
3238static inline bool vma_is_dax(struct vm_area_struct *vma)
3239{
3240 return vma->vm_file && IS_DAX(vma->vm_file->f_mapping->host);
3241}
3242
3243static inline bool vma_is_fsdax(struct vm_area_struct *vma)
3244{
3245 struct inode *inode;
3246
3247 if (!vma->vm_file)
3248 return false;
3249 if (!vma_is_dax(vma))
3250 return false;
3251 inode = file_inode(vma->vm_file);
3252 if (S_ISCHR(inode->i_mode))
3253 return false; /* device-dax */
3254 return true;
3255}
3256
3257static inline int iocb_flags(struct file *file)
3258{
3259 int res = 0;
3260 if (file->f_flags & O_APPEND)
3261 res |= IOCB_APPEND;
3262 if (io_is_direct(file))
3263 res |= IOCB_DIRECT;
3264 if ((file->f_flags & O_DSYNC) || IS_SYNC(file->f_mapping->host))
3265 res |= IOCB_DSYNC;
3266 if (file->f_flags & __O_SYNC)
3267 res |= IOCB_SYNC;
3268 return res;
3269}
3270
3271static inline int kiocb_set_rw_flags(struct kiocb *ki, rwf_t flags)
3272{
3273 if (unlikely(flags & ~RWF_SUPPORTED))
3274 return -EOPNOTSUPP;
3275
3276 if (flags & RWF_NOWAIT) {
3277 if (!(ki->ki_filp->f_mode & FMODE_NOWAIT))
3278 return -EOPNOTSUPP;
3279 ki->ki_flags |= IOCB_NOWAIT;
3280 }
3281 if (flags & RWF_HIPRI)
3282 ki->ki_flags |= IOCB_HIPRI;
3283 if (flags & RWF_DSYNC)
3284 ki->ki_flags |= IOCB_DSYNC;
3285 if (flags & RWF_SYNC)
3286 ki->ki_flags |= (IOCB_DSYNC | IOCB_SYNC);
3287 if (flags & RWF_APPEND)
3288 ki->ki_flags |= IOCB_APPEND;
3289 return 0;
3290}
3291
3292static inline ino_t parent_ino(struct dentry *dentry)
3293{
3294 ino_t res;
3295
3296 /*
3297 * Don't strictly need d_lock here? If the parent ino could change
3298 * then surely we'd have a deeper race in the caller?
3299 */
3300 spin_lock(&dentry->d_lock);
3301 res = dentry->d_parent->d_inode->i_ino;
3302 spin_unlock(&dentry->d_lock);
3303 return res;
3304}
3305
3306/* Transaction based IO helpers */
3307
3308/*
3309 * An argresp is stored in an allocated page and holds the
3310 * size of the argument or response, along with its content
3311 */
3312struct simple_transaction_argresp {
3313 ssize_t size;
3314 char data[0];
3315};
3316
3317#define SIMPLE_TRANSACTION_LIMIT (PAGE_SIZE - sizeof(struct simple_transaction_argresp))
3318
3319char *simple_transaction_get(struct file *file, const char __user *buf,
3320 size_t size);
3321ssize_t simple_transaction_read(struct file *file, char __user *buf,
3322 size_t size, loff_t *pos);
3323int simple_transaction_release(struct inode *inode, struct file *file);
3324
3325void simple_transaction_set(struct file *file, size_t n);
3326
3327/*
3328 * simple attribute files
3329 *
3330 * These attributes behave similar to those in sysfs:
3331 *
3332 * Writing to an attribute immediately sets a value, an open file can be
3333 * written to multiple times.
3334 *
3335 * Reading from an attribute creates a buffer from the value that might get
3336 * read with multiple read calls. When the attribute has been read
3337 * completely, no further read calls are possible until the file is opened
3338 * again.
3339 *
3340 * All attributes contain a text representation of a numeric value
3341 * that are accessed with the get() and set() functions.
3342 */
3343#define DEFINE_SIMPLE_ATTRIBUTE(__fops, __get, __set, __fmt) \
3344static int __fops ## _open(struct inode *inode, struct file *file) \
3345{ \
3346 __simple_attr_check_format(__fmt, 0ull); \
3347 return simple_attr_open(inode, file, __get, __set, __fmt); \
3348} \
3349static const struct file_operations __fops = { \
3350 .owner = THIS_MODULE, \
3351 .open = __fops ## _open, \
3352 .release = simple_attr_release, \
3353 .read = simple_attr_read, \
3354 .write = simple_attr_write, \
3355 .llseek = generic_file_llseek, \
3356}
3357
3358static inline __printf(1, 2)
3359void __simple_attr_check_format(const char *fmt, ...)
3360{
3361 /* don't do anything, just let the compiler check the arguments; */
3362}
3363
3364int simple_attr_open(struct inode *inode, struct file *file,
3365 int (*get)(void *, u64 *), int (*set)(void *, u64),
3366 const char *fmt);
3367int simple_attr_release(struct inode *inode, struct file *file);
3368ssize_t simple_attr_read(struct file *file, char __user *buf,
3369 size_t len, loff_t *ppos);
3370ssize_t simple_attr_write(struct file *file, const char __user *buf,
3371 size_t len, loff_t *ppos);
3372
3373struct ctl_table;
3374int proc_nr_files(struct ctl_table *table, int write,
3375 void __user *buffer, size_t *lenp, loff_t *ppos);
3376int proc_nr_dentry(struct ctl_table *table, int write,
3377 void __user *buffer, size_t *lenp, loff_t *ppos);
3378int proc_nr_inodes(struct ctl_table *table, int write,
3379 void __user *buffer, size_t *lenp, loff_t *ppos);
3380int __init get_filesystem_list(char *buf);
3381
3382#define __FMODE_EXEC ((__force int) FMODE_EXEC)
3383#define __FMODE_NONOTIFY ((__force int) FMODE_NONOTIFY)
3384
3385#define ACC_MODE(x) ("\004\002\006\006"[(x)&O_ACCMODE])
3386#define OPEN_FMODE(flag) ((__force fmode_t)(((flag + 1) & O_ACCMODE) | \
3387 (flag & __FMODE_NONOTIFY)))
3388
3389static inline bool is_sxid(umode_t mode)
3390{
3391 return (mode & S_ISUID) || ((mode & S_ISGID) && (mode & S_IXGRP));
3392}
3393
3394static inline int check_sticky(struct inode *dir, struct inode *inode)
3395{
3396 if (!(dir->i_mode & S_ISVTX))
3397 return 0;
3398
3399 return __check_sticky(dir, inode);
3400}
3401
3402static inline void inode_has_no_xattr(struct inode *inode)
3403{
3404 if (!is_sxid(inode->i_mode) && (inode->i_sb->s_flags & SB_NOSEC))
3405 inode->i_flags |= S_NOSEC;
3406}
3407
3408static inline bool is_root_inode(struct inode *inode)
3409{
3410 return inode == inode->i_sb->s_root->d_inode;
3411}
3412
3413static inline bool dir_emit(struct dir_context *ctx,
3414 const char *name, int namelen,
3415 u64 ino, unsigned type)
3416{
3417 return ctx->actor(ctx, name, namelen, ctx->pos, ino, type) == 0;
3418}
3419static inline bool dir_emit_dot(struct file *file, struct dir_context *ctx)
3420{
3421 return ctx->actor(ctx, ".", 1, ctx->pos,
3422 file->f_path.dentry->d_inode->i_ino, DT_DIR) == 0;
3423}
3424static inline bool dir_emit_dotdot(struct file *file, struct dir_context *ctx)
3425{
3426 return ctx->actor(ctx, "..", 2, ctx->pos,
3427 parent_ino(file->f_path.dentry), DT_DIR) == 0;
3428}
3429static inline bool dir_emit_dots(struct file *file, struct dir_context *ctx)
3430{
3431 if (ctx->pos == 0) {
3432 if (!dir_emit_dot(file, ctx))
3433 return false;
3434 ctx->pos = 1;
3435 }
3436 if (ctx->pos == 1) {
3437 if (!dir_emit_dotdot(file, ctx))
3438 return false;
3439 ctx->pos = 2;
3440 }
3441 return true;
3442}
3443static inline bool dir_relax(struct inode *inode)
3444{
3445 inode_unlock(inode);
3446 inode_lock(inode);
3447 return !IS_DEADDIR(inode);
3448}
3449
3450static inline bool dir_relax_shared(struct inode *inode)
3451{
3452 inode_unlock_shared(inode);
3453 inode_lock_shared(inode);
3454 return !IS_DEADDIR(inode);
3455}
3456
3457extern bool path_noexec(const struct path *path);
3458extern void inode_nohighmem(struct inode *inode);
3459
3460#endif /* _LINUX_FS_H */
3461