Linux kernel mirror (for testing)
git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git
kernel
os
linux
1#ifndef _LINUX_FS_H
2#define _LINUX_FS_H
3
4
5#include <linux/linkage.h>
6#include <linux/wait.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/rbtree.h>
17#include <linux/init.h>
18#include <linux/pid.h>
19#include <linux/bug.h>
20#include <linux/mutex.h>
21#include <linux/rwsem.h>
22#include <linux/capability.h>
23#include <linux/semaphore.h>
24#include <linux/fiemap.h>
25#include <linux/rculist_bl.h>
26#include <linux/atomic.h>
27#include <linux/shrinker.h>
28#include <linux/migrate_mode.h>
29#include <linux/uidgid.h>
30#include <linux/lockdep.h>
31#include <linux/percpu-rwsem.h>
32#include <linux/blk_types.h>
33
34#include <asm/byteorder.h>
35#include <uapi/linux/fs.h>
36
37struct export_operations;
38struct hd_geometry;
39struct iovec;
40struct nameidata;
41struct kiocb;
42struct kobject;
43struct pipe_inode_info;
44struct poll_table_struct;
45struct kstatfs;
46struct vm_area_struct;
47struct vfsmount;
48struct cred;
49struct swap_info_struct;
50struct seq_file;
51struct workqueue_struct;
52struct iov_iter;
53
54extern void __init inode_init(void);
55extern void __init inode_init_early(void);
56extern void __init files_init(unsigned long);
57
58extern struct files_stat_struct files_stat;
59extern unsigned long get_max_files(void);
60extern int sysctl_nr_open;
61extern struct inodes_stat_t inodes_stat;
62extern int leases_enable, lease_break_time;
63extern int sysctl_protected_symlinks;
64extern int sysctl_protected_hardlinks;
65
66struct buffer_head;
67typedef int (get_block_t)(struct inode *inode, sector_t iblock,
68 struct buffer_head *bh_result, int create);
69typedef void (dio_iodone_t)(struct kiocb *iocb, loff_t offset,
70 ssize_t bytes, void *private);
71
72#define MAY_EXEC 0x00000001
73#define MAY_WRITE 0x00000002
74#define MAY_READ 0x00000004
75#define MAY_APPEND 0x00000008
76#define MAY_ACCESS 0x00000010
77#define MAY_OPEN 0x00000020
78#define MAY_CHDIR 0x00000040
79/* called from RCU mode, don't block */
80#define MAY_NOT_BLOCK 0x00000080
81
82/*
83 * flags in file.f_mode. Note that FMODE_READ and FMODE_WRITE must correspond
84 * to O_WRONLY and O_RDWR via the strange trick in __dentry_open()
85 */
86
87/* file is open for reading */
88#define FMODE_READ ((__force fmode_t)0x1)
89/* file is open for writing */
90#define FMODE_WRITE ((__force fmode_t)0x2)
91/* file is seekable */
92#define FMODE_LSEEK ((__force fmode_t)0x4)
93/* file can be accessed using pread */
94#define FMODE_PREAD ((__force fmode_t)0x8)
95/* file can be accessed using pwrite */
96#define FMODE_PWRITE ((__force fmode_t)0x10)
97/* File is opened for execution with sys_execve / sys_uselib */
98#define FMODE_EXEC ((__force fmode_t)0x20)
99/* File is opened with O_NDELAY (only set for block devices) */
100#define FMODE_NDELAY ((__force fmode_t)0x40)
101/* File is opened with O_EXCL (only set for block devices) */
102#define FMODE_EXCL ((__force fmode_t)0x80)
103/* File is opened using open(.., 3, ..) and is writeable only for ioctls
104 (specialy hack for floppy.c) */
105#define FMODE_WRITE_IOCTL ((__force fmode_t)0x100)
106/* 32bit hashes as llseek() offset (for directories) */
107#define FMODE_32BITHASH ((__force fmode_t)0x200)
108/* 64bit hashes as llseek() offset (for directories) */
109#define FMODE_64BITHASH ((__force fmode_t)0x400)
110
111/*
112 * Don't update ctime and mtime.
113 *
114 * Currently a special hack for the XFS open_by_handle ioctl, but we'll
115 * hopefully graduate it to a proper O_CMTIME flag supported by open(2) soon.
116 */
117#define FMODE_NOCMTIME ((__force fmode_t)0x800)
118
119/* Expect random access pattern */
120#define FMODE_RANDOM ((__force fmode_t)0x1000)
121
122/* File is huge (eg. /dev/kmem): treat loff_t as unsigned */
123#define FMODE_UNSIGNED_OFFSET ((__force fmode_t)0x2000)
124
125/* File is opened with O_PATH; almost nothing can be done with it */
126#define FMODE_PATH ((__force fmode_t)0x4000)
127
128/* File needs atomic accesses to f_pos */
129#define FMODE_ATOMIC_POS ((__force fmode_t)0x8000)
130/* Write access to underlying fs */
131#define FMODE_WRITER ((__force fmode_t)0x10000)
132/* Has read method(s) */
133#define FMODE_CAN_READ ((__force fmode_t)0x20000)
134/* Has write method(s) */
135#define FMODE_CAN_WRITE ((__force fmode_t)0x40000)
136
137/* File was opened by fanotify and shouldn't generate fanotify events */
138#define FMODE_NONOTIFY ((__force fmode_t)0x4000000)
139
140/*
141 * Flag for rw_copy_check_uvector and compat_rw_copy_check_uvector
142 * that indicates that they should check the contents of the iovec are
143 * valid, but not check the memory that the iovec elements
144 * points too.
145 */
146#define CHECK_IOVEC_ONLY -1
147
148/*
149 * The below are the various read and write types that we support. Some of
150 * them include behavioral modifiers that send information down to the
151 * block layer and IO scheduler. Terminology:
152 *
153 * The block layer uses device plugging to defer IO a little bit, in
154 * the hope that we will see more IO very shortly. This increases
155 * coalescing of adjacent IO and thus reduces the number of IOs we
156 * have to send to the device. It also allows for better queuing,
157 * if the IO isn't mergeable. If the caller is going to be waiting
158 * for the IO, then he must ensure that the device is unplugged so
159 * that the IO is dispatched to the driver.
160 *
161 * All IO is handled async in Linux. This is fine for background
162 * writes, but for reads or writes that someone waits for completion
163 * on, we want to notify the block layer and IO scheduler so that they
164 * know about it. That allows them to make better scheduling
165 * decisions. So when the below references 'sync' and 'async', it
166 * is referencing this priority hint.
167 *
168 * With that in mind, the available types are:
169 *
170 * READ A normal read operation. Device will be plugged.
171 * READ_SYNC A synchronous read. Device is not plugged, caller can
172 * immediately wait on this read without caring about
173 * unplugging.
174 * READA Used for read-ahead operations. Lower priority, and the
175 * block layer could (in theory) choose to ignore this
176 * request if it runs into resource problems.
177 * WRITE A normal async write. Device will be plugged.
178 * WRITE_SYNC Synchronous write. Identical to WRITE, but passes down
179 * the hint that someone will be waiting on this IO
180 * shortly. The write equivalent of READ_SYNC.
181 * WRITE_ODIRECT Special case write for O_DIRECT only.
182 * WRITE_FLUSH Like WRITE_SYNC but with preceding cache flush.
183 * WRITE_FUA Like WRITE_SYNC but data is guaranteed to be on
184 * non-volatile media on completion.
185 * WRITE_FLUSH_FUA Combination of WRITE_FLUSH and FUA. The IO is preceded
186 * by a cache flush and data is guaranteed to be on
187 * non-volatile media on completion.
188 *
189 */
190#define RW_MASK REQ_WRITE
191#define RWA_MASK REQ_RAHEAD
192
193#define READ 0
194#define WRITE RW_MASK
195#define READA RWA_MASK
196
197#define READ_SYNC (READ | REQ_SYNC)
198#define WRITE_SYNC (WRITE | REQ_SYNC | REQ_NOIDLE)
199#define WRITE_ODIRECT (WRITE | REQ_SYNC)
200#define WRITE_FLUSH (WRITE | REQ_SYNC | REQ_NOIDLE | REQ_FLUSH)
201#define WRITE_FUA (WRITE | REQ_SYNC | REQ_NOIDLE | REQ_FUA)
202#define WRITE_FLUSH_FUA (WRITE | REQ_SYNC | REQ_NOIDLE | REQ_FLUSH | REQ_FUA)
203
204/*
205 * Attribute flags. These should be or-ed together to figure out what
206 * has been changed!
207 */
208#define ATTR_MODE (1 << 0)
209#define ATTR_UID (1 << 1)
210#define ATTR_GID (1 << 2)
211#define ATTR_SIZE (1 << 3)
212#define ATTR_ATIME (1 << 4)
213#define ATTR_MTIME (1 << 5)
214#define ATTR_CTIME (1 << 6)
215#define ATTR_ATIME_SET (1 << 7)
216#define ATTR_MTIME_SET (1 << 8)
217#define ATTR_FORCE (1 << 9) /* Not a change, but a change it */
218#define ATTR_ATTR_FLAG (1 << 10)
219#define ATTR_KILL_SUID (1 << 11)
220#define ATTR_KILL_SGID (1 << 12)
221#define ATTR_FILE (1 << 13)
222#define ATTR_KILL_PRIV (1 << 14)
223#define ATTR_OPEN (1 << 15) /* Truncating from open(O_TRUNC) */
224#define ATTR_TIMES_SET (1 << 16)
225
226/*
227 * Whiteout is represented by a char device. The following constants define the
228 * mode and device number to use.
229 */
230#define WHITEOUT_MODE 0
231#define WHITEOUT_DEV 0
232
233/*
234 * This is the Inode Attributes structure, used for notify_change(). It
235 * uses the above definitions as flags, to know which values have changed.
236 * Also, in this manner, a Filesystem can look at only the values it cares
237 * about. Basically, these are the attributes that the VFS layer can
238 * request to change from the FS layer.
239 *
240 * Derek Atkins <warlord@MIT.EDU> 94-10-20
241 */
242struct iattr {
243 unsigned int ia_valid;
244 umode_t ia_mode;
245 kuid_t ia_uid;
246 kgid_t ia_gid;
247 loff_t ia_size;
248 struct timespec ia_atime;
249 struct timespec ia_mtime;
250 struct timespec ia_ctime;
251
252 /*
253 * Not an attribute, but an auxiliary info for filesystems wanting to
254 * implement an ftruncate() like method. NOTE: filesystem should
255 * check for (ia_valid & ATTR_FILE), and not for (ia_file != NULL).
256 */
257 struct file *ia_file;
258};
259
260/*
261 * Includes for diskquotas.
262 */
263#include <linux/quota.h>
264
265/*
266 * Maximum number of layers of fs stack. Needs to be limited to
267 * prevent kernel stack overflow
268 */
269#define FILESYSTEM_MAX_STACK_DEPTH 2
270
271/**
272 * enum positive_aop_returns - aop return codes with specific semantics
273 *
274 * @AOP_WRITEPAGE_ACTIVATE: Informs the caller that page writeback has
275 * completed, that the page is still locked, and
276 * should be considered active. The VM uses this hint
277 * to return the page to the active list -- it won't
278 * be a candidate for writeback again in the near
279 * future. Other callers must be careful to unlock
280 * the page if they get this return. Returned by
281 * writepage();
282 *
283 * @AOP_TRUNCATED_PAGE: The AOP method that was handed a locked page has
284 * unlocked it and the page might have been truncated.
285 * The caller should back up to acquiring a new page and
286 * trying again. The aop will be taking reasonable
287 * precautions not to livelock. If the caller held a page
288 * reference, it should drop it before retrying. Returned
289 * by readpage().
290 *
291 * address_space_operation functions return these large constants to indicate
292 * special semantics to the caller. These are much larger than the bytes in a
293 * page to allow for functions that return the number of bytes operated on in a
294 * given page.
295 */
296
297enum positive_aop_returns {
298 AOP_WRITEPAGE_ACTIVATE = 0x80000,
299 AOP_TRUNCATED_PAGE = 0x80001,
300};
301
302#define AOP_FLAG_UNINTERRUPTIBLE 0x0001 /* will not do a short write */
303#define AOP_FLAG_CONT_EXPAND 0x0002 /* called from cont_expand */
304#define AOP_FLAG_NOFS 0x0004 /* used by filesystem to direct
305 * helper code (eg buffer layer)
306 * to clear GFP_FS from alloc */
307
308/*
309 * oh the beauties of C type declarations.
310 */
311struct page;
312struct address_space;
313struct writeback_control;
314
315/*
316 * "descriptor" for what we're up to with a read.
317 * This allows us to use the same read code yet
318 * have multiple different users of the data that
319 * we read from a file.
320 *
321 * The simplest case just copies the data to user
322 * mode.
323 */
324typedef struct {
325 size_t written;
326 size_t count;
327 union {
328 char __user *buf;
329 void *data;
330 } arg;
331 int error;
332} read_descriptor_t;
333
334typedef int (*read_actor_t)(read_descriptor_t *, struct page *,
335 unsigned long, unsigned long);
336
337struct address_space_operations {
338 int (*writepage)(struct page *page, struct writeback_control *wbc);
339 int (*readpage)(struct file *, struct page *);
340
341 /* Write back some dirty pages from this mapping. */
342 int (*writepages)(struct address_space *, struct writeback_control *);
343
344 /* Set a page dirty. Return true if this dirtied it */
345 int (*set_page_dirty)(struct page *page);
346
347 int (*readpages)(struct file *filp, struct address_space *mapping,
348 struct list_head *pages, unsigned nr_pages);
349
350 int (*write_begin)(struct file *, struct address_space *mapping,
351 loff_t pos, unsigned len, unsigned flags,
352 struct page **pagep, void **fsdata);
353 int (*write_end)(struct file *, struct address_space *mapping,
354 loff_t pos, unsigned len, unsigned copied,
355 struct page *page, void *fsdata);
356
357 /* Unfortunately this kludge is needed for FIBMAP. Don't use it */
358 sector_t (*bmap)(struct address_space *, sector_t);
359 void (*invalidatepage) (struct page *, unsigned int, unsigned int);
360 int (*releasepage) (struct page *, gfp_t);
361 void (*freepage)(struct page *);
362 ssize_t (*direct_IO)(int, struct kiocb *, struct iov_iter *iter, loff_t offset);
363 int (*get_xip_mem)(struct address_space *, pgoff_t, int,
364 void **, unsigned long *);
365 /*
366 * migrate the contents of a page to the specified target. If
367 * migrate_mode is MIGRATE_ASYNC, it must not block.
368 */
369 int (*migratepage) (struct address_space *,
370 struct page *, struct page *, enum migrate_mode);
371 int (*launder_page) (struct page *);
372 int (*is_partially_uptodate) (struct page *, unsigned long,
373 unsigned long);
374 void (*is_dirty_writeback) (struct page *, bool *, bool *);
375 int (*error_remove_page)(struct address_space *, struct page *);
376
377 /* swapfile support */
378 int (*swap_activate)(struct swap_info_struct *sis, struct file *file,
379 sector_t *span);
380 void (*swap_deactivate)(struct file *file);
381};
382
383extern const struct address_space_operations empty_aops;
384
385/*
386 * pagecache_write_begin/pagecache_write_end must be used by general code
387 * to write into the pagecache.
388 */
389int pagecache_write_begin(struct file *, struct address_space *mapping,
390 loff_t pos, unsigned len, unsigned flags,
391 struct page **pagep, void **fsdata);
392
393int pagecache_write_end(struct file *, struct address_space *mapping,
394 loff_t pos, unsigned len, unsigned copied,
395 struct page *page, void *fsdata);
396
397struct backing_dev_info;
398struct address_space {
399 struct inode *host; /* owner: inode, block_device */
400 struct radix_tree_root page_tree; /* radix tree of all pages */
401 spinlock_t tree_lock; /* and lock protecting it */
402 atomic_t i_mmap_writable;/* count VM_SHARED mappings */
403 struct rb_root i_mmap; /* tree of private and shared mappings */
404 struct list_head i_mmap_nonlinear;/*list VM_NONLINEAR mappings */
405 struct rw_semaphore i_mmap_rwsem; /* protect tree, count, list */
406 /* Protected by tree_lock together with the radix tree */
407 unsigned long nrpages; /* number of total pages */
408 unsigned long nrshadows; /* number of shadow entries */
409 pgoff_t writeback_index;/* writeback starts here */
410 const struct address_space_operations *a_ops; /* methods */
411 unsigned long flags; /* error bits/gfp mask */
412 struct backing_dev_info *backing_dev_info; /* device readahead, etc */
413 spinlock_t private_lock; /* for use by the address_space */
414 struct list_head private_list; /* ditto */
415 void *private_data; /* ditto */
416} __attribute__((aligned(sizeof(long))));
417 /*
418 * On most architectures that alignment is already the case; but
419 * must be enforced here for CRIS, to let the least significant bit
420 * of struct page's "mapping" pointer be used for PAGE_MAPPING_ANON.
421 */
422struct request_queue;
423
424struct block_device {
425 dev_t bd_dev; /* not a kdev_t - it's a search key */
426 int bd_openers;
427 struct inode * bd_inode; /* will die */
428 struct super_block * bd_super;
429 struct mutex bd_mutex; /* open/close mutex */
430 struct list_head bd_inodes;
431 void * bd_claiming;
432 void * bd_holder;
433 int bd_holders;
434 bool bd_write_holder;
435#ifdef CONFIG_SYSFS
436 struct list_head bd_holder_disks;
437#endif
438 struct block_device * bd_contains;
439 unsigned bd_block_size;
440 struct hd_struct * bd_part;
441 /* number of times partitions within this device have been opened. */
442 unsigned bd_part_count;
443 int bd_invalidated;
444 struct gendisk * bd_disk;
445 struct request_queue * bd_queue;
446 struct list_head bd_list;
447 /*
448 * Private data. You must have bd_claim'ed the block_device
449 * to use this. NOTE: bd_claim allows an owner to claim
450 * the same device multiple times, the owner must take special
451 * care to not mess up bd_private for that case.
452 */
453 unsigned long bd_private;
454
455 /* The counter of freeze processes */
456 int bd_fsfreeze_count;
457 /* Mutex for freeze */
458 struct mutex bd_fsfreeze_mutex;
459};
460
461/*
462 * Radix-tree tags, for tagging dirty and writeback pages within the pagecache
463 * radix trees
464 */
465#define PAGECACHE_TAG_DIRTY 0
466#define PAGECACHE_TAG_WRITEBACK 1
467#define PAGECACHE_TAG_TOWRITE 2
468
469int mapping_tagged(struct address_space *mapping, int tag);
470
471static inline void i_mmap_lock_write(struct address_space *mapping)
472{
473 down_write(&mapping->i_mmap_rwsem);
474}
475
476static inline void i_mmap_unlock_write(struct address_space *mapping)
477{
478 up_write(&mapping->i_mmap_rwsem);
479}
480
481static inline void i_mmap_lock_read(struct address_space *mapping)
482{
483 down_read(&mapping->i_mmap_rwsem);
484}
485
486static inline void i_mmap_unlock_read(struct address_space *mapping)
487{
488 up_read(&mapping->i_mmap_rwsem);
489}
490
491/*
492 * Might pages of this file be mapped into userspace?
493 */
494static inline int mapping_mapped(struct address_space *mapping)
495{
496 return !RB_EMPTY_ROOT(&mapping->i_mmap) ||
497 !list_empty(&mapping->i_mmap_nonlinear);
498}
499
500/*
501 * Might pages of this file have been modified in userspace?
502 * Note that i_mmap_writable counts all VM_SHARED vmas: do_mmap_pgoff
503 * marks vma as VM_SHARED if it is shared, and the file was opened for
504 * writing i.e. vma may be mprotected writable even if now readonly.
505 *
506 * If i_mmap_writable is negative, no new writable mappings are allowed. You
507 * can only deny writable mappings, if none exists right now.
508 */
509static inline int mapping_writably_mapped(struct address_space *mapping)
510{
511 return atomic_read(&mapping->i_mmap_writable) > 0;
512}
513
514static inline int mapping_map_writable(struct address_space *mapping)
515{
516 return atomic_inc_unless_negative(&mapping->i_mmap_writable) ?
517 0 : -EPERM;
518}
519
520static inline void mapping_unmap_writable(struct address_space *mapping)
521{
522 atomic_dec(&mapping->i_mmap_writable);
523}
524
525static inline int mapping_deny_writable(struct address_space *mapping)
526{
527 return atomic_dec_unless_positive(&mapping->i_mmap_writable) ?
528 0 : -EBUSY;
529}
530
531static inline void mapping_allow_writable(struct address_space *mapping)
532{
533 atomic_inc(&mapping->i_mmap_writable);
534}
535
536/*
537 * Use sequence counter to get consistent i_size on 32-bit processors.
538 */
539#if BITS_PER_LONG==32 && defined(CONFIG_SMP)
540#include <linux/seqlock.h>
541#define __NEED_I_SIZE_ORDERED
542#define i_size_ordered_init(inode) seqcount_init(&inode->i_size_seqcount)
543#else
544#define i_size_ordered_init(inode) do { } while (0)
545#endif
546
547struct posix_acl;
548#define ACL_NOT_CACHED ((void *)(-1))
549
550#define IOP_FASTPERM 0x0001
551#define IOP_LOOKUP 0x0002
552#define IOP_NOFOLLOW 0x0004
553
554/*
555 * Keep mostly read-only and often accessed (especially for
556 * the RCU path lookup and 'stat' data) fields at the beginning
557 * of the 'struct inode'
558 */
559struct inode {
560 umode_t i_mode;
561 unsigned short i_opflags;
562 kuid_t i_uid;
563 kgid_t i_gid;
564 unsigned int i_flags;
565
566#ifdef CONFIG_FS_POSIX_ACL
567 struct posix_acl *i_acl;
568 struct posix_acl *i_default_acl;
569#endif
570
571 const struct inode_operations *i_op;
572 struct super_block *i_sb;
573 struct address_space *i_mapping;
574
575#ifdef CONFIG_SECURITY
576 void *i_security;
577#endif
578
579 /* Stat data, not accessed from path walking */
580 unsigned long i_ino;
581 /*
582 * Filesystems may only read i_nlink directly. They shall use the
583 * following functions for modification:
584 *
585 * (set|clear|inc|drop)_nlink
586 * inode_(inc|dec)_link_count
587 */
588 union {
589 const unsigned int i_nlink;
590 unsigned int __i_nlink;
591 };
592 dev_t i_rdev;
593 loff_t i_size;
594 struct timespec i_atime;
595 struct timespec i_mtime;
596 struct timespec i_ctime;
597 spinlock_t i_lock; /* i_blocks, i_bytes, maybe i_size */
598 unsigned short i_bytes;
599 unsigned int i_blkbits;
600 blkcnt_t i_blocks;
601
602#ifdef __NEED_I_SIZE_ORDERED
603 seqcount_t i_size_seqcount;
604#endif
605
606 /* Misc */
607 unsigned long i_state;
608 struct mutex i_mutex;
609
610 unsigned long dirtied_when; /* jiffies of first dirtying */
611
612 struct hlist_node i_hash;
613 struct list_head i_wb_list; /* backing dev IO list */
614 struct list_head i_lru; /* inode LRU list */
615 struct list_head i_sb_list;
616 union {
617 struct hlist_head i_dentry;
618 struct rcu_head i_rcu;
619 };
620 u64 i_version;
621 atomic_t i_count;
622 atomic_t i_dio_count;
623 atomic_t i_writecount;
624#ifdef CONFIG_IMA
625 atomic_t i_readcount; /* struct files open RO */
626#endif
627 const struct file_operations *i_fop; /* former ->i_op->default_file_ops */
628 struct file_lock *i_flock;
629 struct address_space i_data;
630 struct list_head i_devices;
631 union {
632 struct pipe_inode_info *i_pipe;
633 struct block_device *i_bdev;
634 struct cdev *i_cdev;
635 };
636
637 __u32 i_generation;
638
639#ifdef CONFIG_FSNOTIFY
640 __u32 i_fsnotify_mask; /* all events this inode cares about */
641 struct hlist_head i_fsnotify_marks;
642#endif
643
644 void *i_private; /* fs or device private pointer */
645};
646
647static inline int inode_unhashed(struct inode *inode)
648{
649 return hlist_unhashed(&inode->i_hash);
650}
651
652/*
653 * inode->i_mutex nesting subclasses for the lock validator:
654 *
655 * 0: the object of the current VFS operation
656 * 1: parent
657 * 2: child/target
658 * 3: xattr
659 * 4: second non-directory
660 * 5: second parent (when locking independent directories in rename)
661 *
662 * I_MUTEX_NONDIR2 is for certain operations (such as rename) which lock two
663 * non-directories at once.
664 *
665 * The locking order between these classes is
666 * parent[2] -> child -> grandchild -> normal -> xattr -> second non-directory
667 */
668enum inode_i_mutex_lock_class
669{
670 I_MUTEX_NORMAL,
671 I_MUTEX_PARENT,
672 I_MUTEX_CHILD,
673 I_MUTEX_XATTR,
674 I_MUTEX_NONDIR2,
675 I_MUTEX_PARENT2,
676};
677
678void lock_two_nondirectories(struct inode *, struct inode*);
679void unlock_two_nondirectories(struct inode *, struct inode*);
680
681/*
682 * NOTE: in a 32bit arch with a preemptable kernel and
683 * an UP compile the i_size_read/write must be atomic
684 * with respect to the local cpu (unlike with preempt disabled),
685 * but they don't need to be atomic with respect to other cpus like in
686 * true SMP (so they need either to either locally disable irq around
687 * the read or for example on x86 they can be still implemented as a
688 * cmpxchg8b without the need of the lock prefix). For SMP compiles
689 * and 64bit archs it makes no difference if preempt is enabled or not.
690 */
691static inline loff_t i_size_read(const struct inode *inode)
692{
693#if BITS_PER_LONG==32 && defined(CONFIG_SMP)
694 loff_t i_size;
695 unsigned int seq;
696
697 do {
698 seq = read_seqcount_begin(&inode->i_size_seqcount);
699 i_size = inode->i_size;
700 } while (read_seqcount_retry(&inode->i_size_seqcount, seq));
701 return i_size;
702#elif BITS_PER_LONG==32 && defined(CONFIG_PREEMPT)
703 loff_t i_size;
704
705 preempt_disable();
706 i_size = inode->i_size;
707 preempt_enable();
708 return i_size;
709#else
710 return inode->i_size;
711#endif
712}
713
714/*
715 * NOTE: unlike i_size_read(), i_size_write() does need locking around it
716 * (normally i_mutex), otherwise on 32bit/SMP an update of i_size_seqcount
717 * can be lost, resulting in subsequent i_size_read() calls spinning forever.
718 */
719static inline void i_size_write(struct inode *inode, loff_t i_size)
720{
721#if BITS_PER_LONG==32 && defined(CONFIG_SMP)
722 preempt_disable();
723 write_seqcount_begin(&inode->i_size_seqcount);
724 inode->i_size = i_size;
725 write_seqcount_end(&inode->i_size_seqcount);
726 preempt_enable();
727#elif BITS_PER_LONG==32 && defined(CONFIG_PREEMPT)
728 preempt_disable();
729 inode->i_size = i_size;
730 preempt_enable();
731#else
732 inode->i_size = i_size;
733#endif
734}
735
736/* Helper functions so that in most cases filesystems will
737 * not need to deal directly with kuid_t and kgid_t and can
738 * instead deal with the raw numeric values that are stored
739 * in the filesystem.
740 */
741static inline uid_t i_uid_read(const struct inode *inode)
742{
743 return from_kuid(&init_user_ns, inode->i_uid);
744}
745
746static inline gid_t i_gid_read(const struct inode *inode)
747{
748 return from_kgid(&init_user_ns, inode->i_gid);
749}
750
751static inline void i_uid_write(struct inode *inode, uid_t uid)
752{
753 inode->i_uid = make_kuid(&init_user_ns, uid);
754}
755
756static inline void i_gid_write(struct inode *inode, gid_t gid)
757{
758 inode->i_gid = make_kgid(&init_user_ns, gid);
759}
760
761static inline unsigned iminor(const struct inode *inode)
762{
763 return MINOR(inode->i_rdev);
764}
765
766static inline unsigned imajor(const struct inode *inode)
767{
768 return MAJOR(inode->i_rdev);
769}
770
771extern struct block_device *I_BDEV(struct inode *inode);
772
773struct fown_struct {
774 rwlock_t lock; /* protects pid, uid, euid fields */
775 struct pid *pid; /* pid or -pgrp where SIGIO should be sent */
776 enum pid_type pid_type; /* Kind of process group SIGIO should be sent to */
777 kuid_t uid, euid; /* uid/euid of process setting the owner */
778 int signum; /* posix.1b rt signal to be delivered on IO */
779};
780
781/*
782 * Track a single file's readahead state
783 */
784struct file_ra_state {
785 pgoff_t start; /* where readahead started */
786 unsigned int size; /* # of readahead pages */
787 unsigned int async_size; /* do asynchronous readahead when
788 there are only # of pages ahead */
789
790 unsigned int ra_pages; /* Maximum readahead window */
791 unsigned int mmap_miss; /* Cache miss stat for mmap accesses */
792 loff_t prev_pos; /* Cache last read() position */
793};
794
795/*
796 * Check if @index falls in the readahead windows.
797 */
798static inline int ra_has_index(struct file_ra_state *ra, pgoff_t index)
799{
800 return (index >= ra->start &&
801 index < ra->start + ra->size);
802}
803
804struct file {
805 union {
806 struct llist_node fu_llist;
807 struct rcu_head fu_rcuhead;
808 } f_u;
809 struct path f_path;
810 struct inode *f_inode; /* cached value */
811 const struct file_operations *f_op;
812
813 /*
814 * Protects f_ep_links, f_flags.
815 * Must not be taken from IRQ context.
816 */
817 spinlock_t f_lock;
818 atomic_long_t f_count;
819 unsigned int f_flags;
820 fmode_t f_mode;
821 struct mutex f_pos_lock;
822 loff_t f_pos;
823 struct fown_struct f_owner;
824 const struct cred *f_cred;
825 struct file_ra_state f_ra;
826
827 u64 f_version;
828#ifdef CONFIG_SECURITY
829 void *f_security;
830#endif
831 /* needed for tty driver, and maybe others */
832 void *private_data;
833
834#ifdef CONFIG_EPOLL
835 /* Used by fs/eventpoll.c to link all the hooks to this file */
836 struct list_head f_ep_links;
837 struct list_head f_tfile_llink;
838#endif /* #ifdef CONFIG_EPOLL */
839 struct address_space *f_mapping;
840} __attribute__((aligned(4))); /* lest something weird decides that 2 is OK */
841
842struct file_handle {
843 __u32 handle_bytes;
844 int handle_type;
845 /* file identifier */
846 unsigned char f_handle[0];
847};
848
849static inline struct file *get_file(struct file *f)
850{
851 atomic_long_inc(&f->f_count);
852 return f;
853}
854#define fput_atomic(x) atomic_long_add_unless(&(x)->f_count, -1, 1)
855#define file_count(x) atomic_long_read(&(x)->f_count)
856
857#define MAX_NON_LFS ((1UL<<31) - 1)
858
859/* Page cache limit. The filesystems should put that into their s_maxbytes
860 limits, otherwise bad things can happen in VM. */
861#if BITS_PER_LONG==32
862#define MAX_LFS_FILESIZE (((loff_t)PAGE_CACHE_SIZE << (BITS_PER_LONG-1))-1)
863#elif BITS_PER_LONG==64
864#define MAX_LFS_FILESIZE ((loff_t)0x7fffffffffffffffLL)
865#endif
866
867#define FL_POSIX 1
868#define FL_FLOCK 2
869#define FL_DELEG 4 /* NFSv4 delegation */
870#define FL_ACCESS 8 /* not trying to lock, just looking */
871#define FL_EXISTS 16 /* when unlocking, test for existence */
872#define FL_LEASE 32 /* lease held on this file */
873#define FL_CLOSE 64 /* unlock on close */
874#define FL_SLEEP 128 /* A blocking lock */
875#define FL_DOWNGRADE_PENDING 256 /* Lease is being downgraded */
876#define FL_UNLOCK_PENDING 512 /* Lease is being broken */
877#define FL_OFDLCK 1024 /* lock is "owned" by struct file */
878
879/*
880 * Special return value from posix_lock_file() and vfs_lock_file() for
881 * asynchronous locking.
882 */
883#define FILE_LOCK_DEFERRED 1
884
885/* legacy typedef, should eventually be removed */
886typedef void *fl_owner_t;
887
888struct file_lock_operations {
889 void (*fl_copy_lock)(struct file_lock *, struct file_lock *);
890 void (*fl_release_private)(struct file_lock *);
891};
892
893struct lock_manager_operations {
894 int (*lm_compare_owner)(struct file_lock *, struct file_lock *);
895 unsigned long (*lm_owner_key)(struct file_lock *);
896 void (*lm_get_owner)(struct file_lock *, struct file_lock *);
897 void (*lm_put_owner)(struct file_lock *);
898 void (*lm_notify)(struct file_lock *); /* unblock callback */
899 int (*lm_grant)(struct file_lock *, int);
900 bool (*lm_break)(struct file_lock *);
901 int (*lm_change)(struct file_lock **, int, struct list_head *);
902 void (*lm_setup)(struct file_lock *, void **);
903};
904
905struct lock_manager {
906 struct list_head list;
907};
908
909struct net;
910void locks_start_grace(struct net *, struct lock_manager *);
911void locks_end_grace(struct lock_manager *);
912int locks_in_grace(struct net *);
913
914/* that will die - we need it for nfs_lock_info */
915#include <linux/nfs_fs_i.h>
916
917/*
918 * struct file_lock represents a generic "file lock". It's used to represent
919 * POSIX byte range locks, BSD (flock) locks, and leases. It's important to
920 * note that the same struct is used to represent both a request for a lock and
921 * the lock itself, but the same object is never used for both.
922 *
923 * FIXME: should we create a separate "struct lock_request" to help distinguish
924 * these two uses?
925 *
926 * The i_flock list is ordered by:
927 *
928 * 1) lock type -- FL_LEASEs first, then FL_FLOCK, and finally FL_POSIX
929 * 2) lock owner
930 * 3) lock range start
931 * 4) lock range end
932 *
933 * Obviously, the last two criteria only matter for POSIX locks.
934 */
935struct file_lock {
936 struct file_lock *fl_next; /* singly linked list for this inode */
937 struct hlist_node fl_link; /* node in global lists */
938 struct list_head fl_block; /* circular list of blocked processes */
939 fl_owner_t fl_owner;
940 unsigned int fl_flags;
941 unsigned char fl_type;
942 unsigned int fl_pid;
943 int fl_link_cpu; /* what cpu's list is this on? */
944 struct pid *fl_nspid;
945 wait_queue_head_t fl_wait;
946 struct file *fl_file;
947 loff_t fl_start;
948 loff_t fl_end;
949
950 struct fasync_struct * fl_fasync; /* for lease break notifications */
951 /* for lease breaks: */
952 unsigned long fl_break_time;
953 unsigned long fl_downgrade_time;
954
955 const struct file_lock_operations *fl_ops; /* Callbacks for filesystems */
956 const struct lock_manager_operations *fl_lmops; /* Callbacks for lockmanagers */
957 union {
958 struct nfs_lock_info nfs_fl;
959 struct nfs4_lock_info nfs4_fl;
960 struct {
961 struct list_head link; /* link in AFS vnode's pending_locks list */
962 int state; /* state of grant or error if -ve */
963 } afs;
964 } fl_u;
965};
966
967/* The following constant reflects the upper bound of the file/locking space */
968#ifndef OFFSET_MAX
969#define INT_LIMIT(x) (~((x)1 << (sizeof(x)*8 - 1)))
970#define OFFSET_MAX INT_LIMIT(loff_t)
971#define OFFT_OFFSET_MAX INT_LIMIT(off_t)
972#endif
973
974#include <linux/fcntl.h>
975
976extern void send_sigio(struct fown_struct *fown, int fd, int band);
977
978#ifdef CONFIG_FILE_LOCKING
979extern int fcntl_getlk(struct file *, unsigned int, struct flock __user *);
980extern int fcntl_setlk(unsigned int, struct file *, unsigned int,
981 struct flock __user *);
982
983#if BITS_PER_LONG == 32
984extern int fcntl_getlk64(struct file *, unsigned int, struct flock64 __user *);
985extern int fcntl_setlk64(unsigned int, struct file *, unsigned int,
986 struct flock64 __user *);
987#endif
988
989extern int fcntl_setlease(unsigned int fd, struct file *filp, long arg);
990extern int fcntl_getlease(struct file *filp);
991
992/* fs/locks.c */
993void locks_free_lock(struct file_lock *fl);
994extern void locks_init_lock(struct file_lock *);
995extern struct file_lock * locks_alloc_lock(void);
996extern void locks_copy_lock(struct file_lock *, struct file_lock *);
997extern void locks_copy_conflock(struct file_lock *, struct file_lock *);
998extern void locks_remove_posix(struct file *, fl_owner_t);
999extern void locks_remove_file(struct file *);
1000extern void locks_release_private(struct file_lock *);
1001extern void posix_test_lock(struct file *, struct file_lock *);
1002extern int posix_lock_file(struct file *, struct file_lock *, struct file_lock *);
1003extern int posix_lock_file_wait(struct file *, struct file_lock *);
1004extern int posix_unblock_lock(struct file_lock *);
1005extern int vfs_test_lock(struct file *, struct file_lock *);
1006extern int vfs_lock_file(struct file *, unsigned int, struct file_lock *, struct file_lock *);
1007extern int vfs_cancel_lock(struct file *filp, struct file_lock *fl);
1008extern int flock_lock_file_wait(struct file *filp, struct file_lock *fl);
1009extern int __break_lease(struct inode *inode, unsigned int flags, unsigned int type);
1010extern void lease_get_mtime(struct inode *, struct timespec *time);
1011extern int generic_setlease(struct file *, long, struct file_lock **, void **priv);
1012extern int vfs_setlease(struct file *, long, struct file_lock **, void **);
1013extern int lease_modify(struct file_lock **, int, struct list_head *);
1014#else /* !CONFIG_FILE_LOCKING */
1015static inline int fcntl_getlk(struct file *file, unsigned int cmd,
1016 struct flock __user *user)
1017{
1018 return -EINVAL;
1019}
1020
1021static inline int fcntl_setlk(unsigned int fd, struct file *file,
1022 unsigned int cmd, struct flock __user *user)
1023{
1024 return -EACCES;
1025}
1026
1027#if BITS_PER_LONG == 32
1028static inline int fcntl_getlk64(struct file *file, unsigned int cmd,
1029 struct flock64 __user *user)
1030{
1031 return -EINVAL;
1032}
1033
1034static inline int fcntl_setlk64(unsigned int fd, struct file *file,
1035 unsigned int cmd, struct flock64 __user *user)
1036{
1037 return -EACCES;
1038}
1039#endif
1040static inline int fcntl_setlease(unsigned int fd, struct file *filp, long arg)
1041{
1042 return -EINVAL;
1043}
1044
1045static inline int fcntl_getlease(struct file *filp)
1046{
1047 return F_UNLCK;
1048}
1049
1050static inline void locks_init_lock(struct file_lock *fl)
1051{
1052 return;
1053}
1054
1055static inline void locks_copy_conflock(struct file_lock *new, struct file_lock *fl)
1056{
1057 return;
1058}
1059
1060static inline void locks_copy_lock(struct file_lock *new, struct file_lock *fl)
1061{
1062 return;
1063}
1064
1065static inline void locks_remove_posix(struct file *filp, fl_owner_t owner)
1066{
1067 return;
1068}
1069
1070static inline void locks_remove_file(struct file *filp)
1071{
1072 return;
1073}
1074
1075static inline void posix_test_lock(struct file *filp, struct file_lock *fl)
1076{
1077 return;
1078}
1079
1080static inline int posix_lock_file(struct file *filp, struct file_lock *fl,
1081 struct file_lock *conflock)
1082{
1083 return -ENOLCK;
1084}
1085
1086static inline int posix_lock_file_wait(struct file *filp, struct file_lock *fl)
1087{
1088 return -ENOLCK;
1089}
1090
1091static inline int posix_unblock_lock(struct file_lock *waiter)
1092{
1093 return -ENOENT;
1094}
1095
1096static inline int vfs_test_lock(struct file *filp, struct file_lock *fl)
1097{
1098 return 0;
1099}
1100
1101static inline int vfs_lock_file(struct file *filp, unsigned int cmd,
1102 struct file_lock *fl, struct file_lock *conf)
1103{
1104 return -ENOLCK;
1105}
1106
1107static inline int vfs_cancel_lock(struct file *filp, struct file_lock *fl)
1108{
1109 return 0;
1110}
1111
1112static inline int flock_lock_file_wait(struct file *filp,
1113 struct file_lock *request)
1114{
1115 return -ENOLCK;
1116}
1117
1118static inline int __break_lease(struct inode *inode, unsigned int mode, unsigned int type)
1119{
1120 return 0;
1121}
1122
1123static inline void lease_get_mtime(struct inode *inode, struct timespec *time)
1124{
1125 return;
1126}
1127
1128static inline int generic_setlease(struct file *filp, long arg,
1129 struct file_lock **flp, void **priv)
1130{
1131 return -EINVAL;
1132}
1133
1134static inline int vfs_setlease(struct file *filp, long arg,
1135 struct file_lock **lease, void **priv)
1136{
1137 return -EINVAL;
1138}
1139
1140static inline int lease_modify(struct file_lock **before, int arg,
1141 struct list_head *dispose)
1142{
1143 return -EINVAL;
1144}
1145#endif /* !CONFIG_FILE_LOCKING */
1146
1147
1148struct fasync_struct {
1149 spinlock_t fa_lock;
1150 int magic;
1151 int fa_fd;
1152 struct fasync_struct *fa_next; /* singly linked list */
1153 struct file *fa_file;
1154 struct rcu_head fa_rcu;
1155};
1156
1157#define FASYNC_MAGIC 0x4601
1158
1159/* SMP safe fasync helpers: */
1160extern int fasync_helper(int, struct file *, int, struct fasync_struct **);
1161extern struct fasync_struct *fasync_insert_entry(int, struct file *, struct fasync_struct **, struct fasync_struct *);
1162extern int fasync_remove_entry(struct file *, struct fasync_struct **);
1163extern struct fasync_struct *fasync_alloc(void);
1164extern void fasync_free(struct fasync_struct *);
1165
1166/* can be called from interrupts */
1167extern void kill_fasync(struct fasync_struct **, int, int);
1168
1169extern void __f_setown(struct file *filp, struct pid *, enum pid_type, int force);
1170extern void f_setown(struct file *filp, unsigned long arg, int force);
1171extern void f_delown(struct file *filp);
1172extern pid_t f_getown(struct file *filp);
1173extern int send_sigurg(struct fown_struct *fown);
1174
1175struct mm_struct;
1176
1177/*
1178 * Umount options
1179 */
1180
1181#define MNT_FORCE 0x00000001 /* Attempt to forcibily umount */
1182#define MNT_DETACH 0x00000002 /* Just detach from the tree */
1183#define MNT_EXPIRE 0x00000004 /* Mark for expiry */
1184#define UMOUNT_NOFOLLOW 0x00000008 /* Don't follow symlink on umount */
1185#define UMOUNT_UNUSED 0x80000000 /* Flag guaranteed to be unused */
1186
1187extern struct list_head super_blocks;
1188extern spinlock_t sb_lock;
1189
1190/* Possible states of 'frozen' field */
1191enum {
1192 SB_UNFROZEN = 0, /* FS is unfrozen */
1193 SB_FREEZE_WRITE = 1, /* Writes, dir ops, ioctls frozen */
1194 SB_FREEZE_PAGEFAULT = 2, /* Page faults stopped as well */
1195 SB_FREEZE_FS = 3, /* For internal FS use (e.g. to stop
1196 * internal threads if needed) */
1197 SB_FREEZE_COMPLETE = 4, /* ->freeze_fs finished successfully */
1198};
1199
1200#define SB_FREEZE_LEVELS (SB_FREEZE_COMPLETE - 1)
1201
1202struct sb_writers {
1203 /* Counters for counting writers at each level */
1204 struct percpu_counter counter[SB_FREEZE_LEVELS];
1205 wait_queue_head_t wait; /* queue for waiting for
1206 writers / faults to finish */
1207 int frozen; /* Is sb frozen? */
1208 wait_queue_head_t wait_unfrozen; /* queue for waiting for
1209 sb to be thawed */
1210#ifdef CONFIG_DEBUG_LOCK_ALLOC
1211 struct lockdep_map lock_map[SB_FREEZE_LEVELS];
1212#endif
1213};
1214
1215struct super_block {
1216 struct list_head s_list; /* Keep this first */
1217 dev_t s_dev; /* search index; _not_ kdev_t */
1218 unsigned char s_blocksize_bits;
1219 unsigned long s_blocksize;
1220 loff_t s_maxbytes; /* Max file size */
1221 struct file_system_type *s_type;
1222 const struct super_operations *s_op;
1223 const struct dquot_operations *dq_op;
1224 const struct quotactl_ops *s_qcop;
1225 const struct export_operations *s_export_op;
1226 unsigned long s_flags;
1227 unsigned long s_magic;
1228 struct dentry *s_root;
1229 struct rw_semaphore s_umount;
1230 int s_count;
1231 atomic_t s_active;
1232#ifdef CONFIG_SECURITY
1233 void *s_security;
1234#endif
1235 const struct xattr_handler **s_xattr;
1236
1237 struct list_head s_inodes; /* all inodes */
1238 struct hlist_bl_head s_anon; /* anonymous dentries for (nfs) exporting */
1239 struct list_head s_mounts; /* list of mounts; _not_ for fs use */
1240 struct block_device *s_bdev;
1241 struct backing_dev_info *s_bdi;
1242 struct mtd_info *s_mtd;
1243 struct hlist_node s_instances;
1244 unsigned int s_quota_types; /* Bitmask of supported quota types */
1245 struct quota_info s_dquot; /* Diskquota specific options */
1246
1247 struct sb_writers s_writers;
1248
1249 char s_id[32]; /* Informational name */
1250 u8 s_uuid[16]; /* UUID */
1251
1252 void *s_fs_info; /* Filesystem private info */
1253 unsigned int s_max_links;
1254 fmode_t s_mode;
1255
1256 /* Granularity of c/m/atime in ns.
1257 Cannot be worse than a second */
1258 u32 s_time_gran;
1259
1260 /*
1261 * The next field is for VFS *only*. No filesystems have any business
1262 * even looking at it. You had been warned.
1263 */
1264 struct mutex s_vfs_rename_mutex; /* Kludge */
1265
1266 /*
1267 * Filesystem subtype. If non-empty the filesystem type field
1268 * in /proc/mounts will be "type.subtype"
1269 */
1270 char *s_subtype;
1271
1272 /*
1273 * Saved mount options for lazy filesystems using
1274 * generic_show_options()
1275 */
1276 char __rcu *s_options;
1277 const struct dentry_operations *s_d_op; /* default d_op for dentries */
1278
1279 /*
1280 * Saved pool identifier for cleancache (-1 means none)
1281 */
1282 int cleancache_poolid;
1283
1284 struct shrinker s_shrink; /* per-sb shrinker handle */
1285
1286 /* Number of inodes with nlink == 0 but still referenced */
1287 atomic_long_t s_remove_count;
1288
1289 /* Being remounted read-only */
1290 int s_readonly_remount;
1291
1292 /* AIO completions deferred from interrupt context */
1293 struct workqueue_struct *s_dio_done_wq;
1294 struct hlist_head s_pins;
1295
1296 /*
1297 * Keep the lru lists last in the structure so they always sit on their
1298 * own individual cachelines.
1299 */
1300 struct list_lru s_dentry_lru ____cacheline_aligned_in_smp;
1301 struct list_lru s_inode_lru ____cacheline_aligned_in_smp;
1302 struct rcu_head rcu;
1303
1304 /*
1305 * Indicates how deep in a filesystem stack this SB is
1306 */
1307 int s_stack_depth;
1308};
1309
1310extern struct timespec current_fs_time(struct super_block *sb);
1311
1312/*
1313 * Snapshotting support.
1314 */
1315
1316void __sb_end_write(struct super_block *sb, int level);
1317int __sb_start_write(struct super_block *sb, int level, bool wait);
1318
1319/**
1320 * sb_end_write - drop write access to a superblock
1321 * @sb: the super we wrote to
1322 *
1323 * Decrement number of writers to the filesystem. Wake up possible waiters
1324 * wanting to freeze the filesystem.
1325 */
1326static inline void sb_end_write(struct super_block *sb)
1327{
1328 __sb_end_write(sb, SB_FREEZE_WRITE);
1329}
1330
1331/**
1332 * sb_end_pagefault - drop write access to a superblock from a page fault
1333 * @sb: the super we wrote to
1334 *
1335 * Decrement number of processes handling write page fault to the filesystem.
1336 * Wake up possible waiters wanting to freeze the filesystem.
1337 */
1338static inline void sb_end_pagefault(struct super_block *sb)
1339{
1340 __sb_end_write(sb, SB_FREEZE_PAGEFAULT);
1341}
1342
1343/**
1344 * sb_end_intwrite - drop write access to a superblock for internal fs purposes
1345 * @sb: the super we wrote to
1346 *
1347 * Decrement fs-internal number of writers to the filesystem. Wake up possible
1348 * waiters wanting to freeze the filesystem.
1349 */
1350static inline void sb_end_intwrite(struct super_block *sb)
1351{
1352 __sb_end_write(sb, SB_FREEZE_FS);
1353}
1354
1355/**
1356 * sb_start_write - get write access to a superblock
1357 * @sb: the super we write to
1358 *
1359 * When a process wants to write data or metadata to a file system (i.e. dirty
1360 * a page or an inode), it should embed the operation in a sb_start_write() -
1361 * sb_end_write() pair to get exclusion against file system freezing. This
1362 * function increments number of writers preventing freezing. If the file
1363 * system is already frozen, the function waits until the file system is
1364 * thawed.
1365 *
1366 * Since freeze protection behaves as a lock, users have to preserve
1367 * ordering of freeze protection and other filesystem locks. Generally,
1368 * freeze protection should be the outermost lock. In particular, we have:
1369 *
1370 * sb_start_write
1371 * -> i_mutex (write path, truncate, directory ops, ...)
1372 * -> s_umount (freeze_super, thaw_super)
1373 */
1374static inline void sb_start_write(struct super_block *sb)
1375{
1376 __sb_start_write(sb, SB_FREEZE_WRITE, true);
1377}
1378
1379static inline int sb_start_write_trylock(struct super_block *sb)
1380{
1381 return __sb_start_write(sb, SB_FREEZE_WRITE, false);
1382}
1383
1384/**
1385 * sb_start_pagefault - get write access to a superblock from a page fault
1386 * @sb: the super we write to
1387 *
1388 * When a process starts handling write page fault, it should embed the
1389 * operation into sb_start_pagefault() - sb_end_pagefault() pair to get
1390 * exclusion against file system freezing. This is needed since the page fault
1391 * is going to dirty a page. This function increments number of running page
1392 * faults preventing freezing. If the file system is already frozen, the
1393 * function waits until the file system is thawed.
1394 *
1395 * Since page fault freeze protection behaves as a lock, users have to preserve
1396 * ordering of freeze protection and other filesystem locks. It is advised to
1397 * put sb_start_pagefault() close to mmap_sem in lock ordering. Page fault
1398 * handling code implies lock dependency:
1399 *
1400 * mmap_sem
1401 * -> sb_start_pagefault
1402 */
1403static inline void sb_start_pagefault(struct super_block *sb)
1404{
1405 __sb_start_write(sb, SB_FREEZE_PAGEFAULT, true);
1406}
1407
1408/*
1409 * sb_start_intwrite - get write access to a superblock for internal fs purposes
1410 * @sb: the super we write to
1411 *
1412 * This is the third level of protection against filesystem freezing. It is
1413 * free for use by a filesystem. The only requirement is that it must rank
1414 * below sb_start_pagefault.
1415 *
1416 * For example filesystem can call sb_start_intwrite() when starting a
1417 * transaction which somewhat eases handling of freezing for internal sources
1418 * of filesystem changes (internal fs threads, discarding preallocation on file
1419 * close, etc.).
1420 */
1421static inline void sb_start_intwrite(struct super_block *sb)
1422{
1423 __sb_start_write(sb, SB_FREEZE_FS, true);
1424}
1425
1426
1427extern bool inode_owner_or_capable(const struct inode *inode);
1428
1429/*
1430 * VFS helper functions..
1431 */
1432extern int vfs_create(struct inode *, struct dentry *, umode_t, bool);
1433extern int vfs_mkdir(struct inode *, struct dentry *, umode_t);
1434extern int vfs_mknod(struct inode *, struct dentry *, umode_t, dev_t);
1435extern int vfs_symlink(struct inode *, struct dentry *, const char *);
1436extern int vfs_link(struct dentry *, struct inode *, struct dentry *, struct inode **);
1437extern int vfs_rmdir(struct inode *, struct dentry *);
1438extern int vfs_unlink(struct inode *, struct dentry *, struct inode **);
1439extern int vfs_rename(struct inode *, struct dentry *, struct inode *, struct dentry *, struct inode **, unsigned int);
1440extern int vfs_whiteout(struct inode *, struct dentry *);
1441
1442/*
1443 * VFS dentry helper functions.
1444 */
1445extern void dentry_unhash(struct dentry *dentry);
1446
1447/*
1448 * VFS file helper functions.
1449 */
1450extern void inode_init_owner(struct inode *inode, const struct inode *dir,
1451 umode_t mode);
1452/*
1453 * VFS FS_IOC_FIEMAP helper definitions.
1454 */
1455struct fiemap_extent_info {
1456 unsigned int fi_flags; /* Flags as passed from user */
1457 unsigned int fi_extents_mapped; /* Number of mapped extents */
1458 unsigned int fi_extents_max; /* Size of fiemap_extent array */
1459 struct fiemap_extent __user *fi_extents_start; /* Start of
1460 fiemap_extent array */
1461};
1462int fiemap_fill_next_extent(struct fiemap_extent_info *info, u64 logical,
1463 u64 phys, u64 len, u32 flags);
1464int fiemap_check_flags(struct fiemap_extent_info *fieinfo, u32 fs_flags);
1465
1466/*
1467 * File types
1468 *
1469 * NOTE! These match bits 12..15 of stat.st_mode
1470 * (ie "(i_mode >> 12) & 15").
1471 */
1472#define DT_UNKNOWN 0
1473#define DT_FIFO 1
1474#define DT_CHR 2
1475#define DT_DIR 4
1476#define DT_BLK 6
1477#define DT_REG 8
1478#define DT_LNK 10
1479#define DT_SOCK 12
1480#define DT_WHT 14
1481
1482/*
1483 * This is the "filldir" function type, used by readdir() to let
1484 * the kernel specify what kind of dirent layout it wants to have.
1485 * This allows the kernel to read directories into kernel space or
1486 * to have different dirent layouts depending on the binary type.
1487 */
1488struct dir_context;
1489typedef int (*filldir_t)(struct dir_context *, const char *, int, loff_t, u64,
1490 unsigned);
1491
1492struct dir_context {
1493 const filldir_t actor;
1494 loff_t pos;
1495};
1496
1497struct block_device_operations;
1498
1499/* These macros are for out of kernel modules to test that
1500 * the kernel supports the unlocked_ioctl and compat_ioctl
1501 * fields in struct file_operations. */
1502#define HAVE_COMPAT_IOCTL 1
1503#define HAVE_UNLOCKED_IOCTL 1
1504
1505struct iov_iter;
1506
1507struct file_operations {
1508 struct module *owner;
1509 loff_t (*llseek) (struct file *, loff_t, int);
1510 ssize_t (*read) (struct file *, char __user *, size_t, loff_t *);
1511 ssize_t (*write) (struct file *, const char __user *, size_t, loff_t *);
1512 ssize_t (*aio_read) (struct kiocb *, const struct iovec *, unsigned long, loff_t);
1513 ssize_t (*aio_write) (struct kiocb *, const struct iovec *, unsigned long, loff_t);
1514 ssize_t (*read_iter) (struct kiocb *, struct iov_iter *);
1515 ssize_t (*write_iter) (struct kiocb *, struct iov_iter *);
1516 int (*iterate) (struct file *, struct dir_context *);
1517 unsigned int (*poll) (struct file *, struct poll_table_struct *);
1518 long (*unlocked_ioctl) (struct file *, unsigned int, unsigned long);
1519 long (*compat_ioctl) (struct file *, unsigned int, unsigned long);
1520 int (*mmap) (struct file *, struct vm_area_struct *);
1521 void (*mremap)(struct file *, struct vm_area_struct *);
1522 int (*open) (struct inode *, struct file *);
1523 int (*flush) (struct file *, fl_owner_t id);
1524 int (*release) (struct inode *, struct file *);
1525 int (*fsync) (struct file *, loff_t, loff_t, int datasync);
1526 int (*aio_fsync) (struct kiocb *, int datasync);
1527 int (*fasync) (int, struct file *, int);
1528 int (*lock) (struct file *, int, struct file_lock *);
1529 ssize_t (*sendpage) (struct file *, struct page *, int, size_t, loff_t *, int);
1530 unsigned long (*get_unmapped_area)(struct file *, unsigned long, unsigned long, unsigned long, unsigned long);
1531 int (*check_flags)(int);
1532 int (*flock) (struct file *, int, struct file_lock *);
1533 ssize_t (*splice_write)(struct pipe_inode_info *, struct file *, loff_t *, size_t, unsigned int);
1534 ssize_t (*splice_read)(struct file *, loff_t *, struct pipe_inode_info *, size_t, unsigned int);
1535 int (*setlease)(struct file *, long, struct file_lock **, void **);
1536 long (*fallocate)(struct file *file, int mode, loff_t offset,
1537 loff_t len);
1538 void (*show_fdinfo)(struct seq_file *m, struct file *f);
1539};
1540
1541struct inode_operations {
1542 struct dentry * (*lookup) (struct inode *,struct dentry *, unsigned int);
1543 void * (*follow_link) (struct dentry *, struct nameidata *);
1544 int (*permission) (struct inode *, int);
1545 struct posix_acl * (*get_acl)(struct inode *, int);
1546
1547 int (*readlink) (struct dentry *, char __user *,int);
1548 void (*put_link) (struct dentry *, struct nameidata *, void *);
1549
1550 int (*create) (struct inode *,struct dentry *, umode_t, bool);
1551 int (*link) (struct dentry *,struct inode *,struct dentry *);
1552 int (*unlink) (struct inode *,struct dentry *);
1553 int (*symlink) (struct inode *,struct dentry *,const char *);
1554 int (*mkdir) (struct inode *,struct dentry *,umode_t);
1555 int (*rmdir) (struct inode *,struct dentry *);
1556 int (*mknod) (struct inode *,struct dentry *,umode_t,dev_t);
1557 int (*rename) (struct inode *, struct dentry *,
1558 struct inode *, struct dentry *);
1559 int (*rename2) (struct inode *, struct dentry *,
1560 struct inode *, struct dentry *, unsigned int);
1561 int (*setattr) (struct dentry *, struct iattr *);
1562 int (*getattr) (struct vfsmount *mnt, struct dentry *, struct kstat *);
1563 int (*setxattr) (struct dentry *, const char *,const void *,size_t,int);
1564 ssize_t (*getxattr) (struct dentry *, const char *, void *, size_t);
1565 ssize_t (*listxattr) (struct dentry *, char *, size_t);
1566 int (*removexattr) (struct dentry *, const char *);
1567 int (*fiemap)(struct inode *, struct fiemap_extent_info *, u64 start,
1568 u64 len);
1569 int (*update_time)(struct inode *, struct timespec *, int);
1570 int (*atomic_open)(struct inode *, struct dentry *,
1571 struct file *, unsigned open_flag,
1572 umode_t create_mode, int *opened);
1573 int (*tmpfile) (struct inode *, struct dentry *, umode_t);
1574 int (*set_acl)(struct inode *, struct posix_acl *, int);
1575
1576 /* WARNING: probably going away soon, do not use! */
1577 int (*dentry_open)(struct dentry *, struct file *, const struct cred *);
1578} ____cacheline_aligned;
1579
1580ssize_t rw_copy_check_uvector(int type, const struct iovec __user * uvector,
1581 unsigned long nr_segs, unsigned long fast_segs,
1582 struct iovec *fast_pointer,
1583 struct iovec **ret_pointer);
1584
1585extern ssize_t __vfs_read(struct file *, char __user *, size_t, loff_t *);
1586extern ssize_t vfs_read(struct file *, char __user *, size_t, loff_t *);
1587extern ssize_t vfs_write(struct file *, const char __user *, size_t, loff_t *);
1588extern ssize_t vfs_readv(struct file *, const struct iovec __user *,
1589 unsigned long, loff_t *);
1590extern ssize_t vfs_writev(struct file *, const struct iovec __user *,
1591 unsigned long, loff_t *);
1592
1593struct super_operations {
1594 struct inode *(*alloc_inode)(struct super_block *sb);
1595 void (*destroy_inode)(struct inode *);
1596
1597 void (*dirty_inode) (struct inode *, int flags);
1598 int (*write_inode) (struct inode *, struct writeback_control *wbc);
1599 int (*drop_inode) (struct inode *);
1600 void (*evict_inode) (struct inode *);
1601 void (*put_super) (struct super_block *);
1602 int (*sync_fs)(struct super_block *sb, int wait);
1603 int (*freeze_super) (struct super_block *);
1604 int (*freeze_fs) (struct super_block *);
1605 int (*thaw_super) (struct super_block *);
1606 int (*unfreeze_fs) (struct super_block *);
1607 int (*statfs) (struct dentry *, struct kstatfs *);
1608 int (*remount_fs) (struct super_block *, int *, char *);
1609 void (*umount_begin) (struct super_block *);
1610
1611 int (*show_options)(struct seq_file *, struct dentry *);
1612 int (*show_devname)(struct seq_file *, struct dentry *);
1613 int (*show_path)(struct seq_file *, struct dentry *);
1614 int (*show_stats)(struct seq_file *, struct dentry *);
1615#ifdef CONFIG_QUOTA
1616 ssize_t (*quota_read)(struct super_block *, int, char *, size_t, loff_t);
1617 ssize_t (*quota_write)(struct super_block *, int, const char *, size_t, loff_t);
1618 struct dquot **(*get_dquots)(struct inode *);
1619#endif
1620 int (*bdev_try_to_free_page)(struct super_block*, struct page*, gfp_t);
1621 long (*nr_cached_objects)(struct super_block *, int);
1622 long (*free_cached_objects)(struct super_block *, long, int);
1623};
1624
1625/*
1626 * Inode flags - they have no relation to superblock flags now
1627 */
1628#define S_SYNC 1 /* Writes are synced at once */
1629#define S_NOATIME 2 /* Do not update access times */
1630#define S_APPEND 4 /* Append-only file */
1631#define S_IMMUTABLE 8 /* Immutable file */
1632#define S_DEAD 16 /* removed, but still open directory */
1633#define S_NOQUOTA 32 /* Inode is not counted to quota */
1634#define S_DIRSYNC 64 /* Directory modifications are synchronous */
1635#define S_NOCMTIME 128 /* Do not update file c/mtime */
1636#define S_SWAPFILE 256 /* Do not truncate: swapon got its bmaps */
1637#define S_PRIVATE 512 /* Inode is fs-internal */
1638#define S_IMA 1024 /* Inode has an associated IMA struct */
1639#define S_AUTOMOUNT 2048 /* Automount/referral quasi-directory */
1640#define S_NOSEC 4096 /* no suid or xattr security attributes */
1641
1642/*
1643 * Note that nosuid etc flags are inode-specific: setting some file-system
1644 * flags just means all the inodes inherit those flags by default. It might be
1645 * possible to override it selectively if you really wanted to with some
1646 * ioctl() that is not currently implemented.
1647 *
1648 * Exception: MS_RDONLY is always applied to the entire file system.
1649 *
1650 * Unfortunately, it is possible to change a filesystems flags with it mounted
1651 * with files in use. This means that all of the inodes will not have their
1652 * i_flags updated. Hence, i_flags no longer inherit the superblock mount
1653 * flags, so these have to be checked separately. -- rmk@arm.uk.linux.org
1654 */
1655#define __IS_FLG(inode, flg) ((inode)->i_sb->s_flags & (flg))
1656
1657#define IS_RDONLY(inode) ((inode)->i_sb->s_flags & MS_RDONLY)
1658#define IS_SYNC(inode) (__IS_FLG(inode, MS_SYNCHRONOUS) || \
1659 ((inode)->i_flags & S_SYNC))
1660#define IS_DIRSYNC(inode) (__IS_FLG(inode, MS_SYNCHRONOUS|MS_DIRSYNC) || \
1661 ((inode)->i_flags & (S_SYNC|S_DIRSYNC)))
1662#define IS_MANDLOCK(inode) __IS_FLG(inode, MS_MANDLOCK)
1663#define IS_NOATIME(inode) __IS_FLG(inode, MS_RDONLY|MS_NOATIME)
1664#define IS_I_VERSION(inode) __IS_FLG(inode, MS_I_VERSION)
1665
1666#define IS_NOQUOTA(inode) ((inode)->i_flags & S_NOQUOTA)
1667#define IS_APPEND(inode) ((inode)->i_flags & S_APPEND)
1668#define IS_IMMUTABLE(inode) ((inode)->i_flags & S_IMMUTABLE)
1669#define IS_POSIXACL(inode) __IS_FLG(inode, MS_POSIXACL)
1670
1671#define IS_DEADDIR(inode) ((inode)->i_flags & S_DEAD)
1672#define IS_NOCMTIME(inode) ((inode)->i_flags & S_NOCMTIME)
1673#define IS_SWAPFILE(inode) ((inode)->i_flags & S_SWAPFILE)
1674#define IS_PRIVATE(inode) ((inode)->i_flags & S_PRIVATE)
1675#define IS_IMA(inode) ((inode)->i_flags & S_IMA)
1676#define IS_AUTOMOUNT(inode) ((inode)->i_flags & S_AUTOMOUNT)
1677#define IS_NOSEC(inode) ((inode)->i_flags & S_NOSEC)
1678
1679#define IS_WHITEOUT(inode) (S_ISCHR(inode->i_mode) && \
1680 (inode)->i_rdev == WHITEOUT_DEV)
1681
1682/*
1683 * Inode state bits. Protected by inode->i_lock
1684 *
1685 * Three bits determine the dirty state of the inode, I_DIRTY_SYNC,
1686 * I_DIRTY_DATASYNC and I_DIRTY_PAGES.
1687 *
1688 * Four bits define the lifetime of an inode. Initially, inodes are I_NEW,
1689 * until that flag is cleared. I_WILL_FREE, I_FREEING and I_CLEAR are set at
1690 * various stages of removing an inode.
1691 *
1692 * Two bits are used for locking and completion notification, I_NEW and I_SYNC.
1693 *
1694 * I_DIRTY_SYNC Inode is dirty, but doesn't have to be written on
1695 * fdatasync(). i_atime is the usual cause.
1696 * I_DIRTY_DATASYNC Data-related inode changes pending. We keep track of
1697 * these changes separately from I_DIRTY_SYNC so that we
1698 * don't have to write inode on fdatasync() when only
1699 * mtime has changed in it.
1700 * I_DIRTY_PAGES Inode has dirty pages. Inode itself may be clean.
1701 * I_NEW Serves as both a mutex and completion notification.
1702 * New inodes set I_NEW. If two processes both create
1703 * the same inode, one of them will release its inode and
1704 * wait for I_NEW to be released before returning.
1705 * Inodes in I_WILL_FREE, I_FREEING or I_CLEAR state can
1706 * also cause waiting on I_NEW, without I_NEW actually
1707 * being set. find_inode() uses this to prevent returning
1708 * nearly-dead inodes.
1709 * I_WILL_FREE Must be set when calling write_inode_now() if i_count
1710 * is zero. I_FREEING must be set when I_WILL_FREE is
1711 * cleared.
1712 * I_FREEING Set when inode is about to be freed but still has dirty
1713 * pages or buffers attached or the inode itself is still
1714 * dirty.
1715 * I_CLEAR Added by clear_inode(). In this state the inode is
1716 * clean and can be destroyed. Inode keeps I_FREEING.
1717 *
1718 * Inodes that are I_WILL_FREE, I_FREEING or I_CLEAR are
1719 * prohibited for many purposes. iget() must wait for
1720 * the inode to be completely released, then create it
1721 * anew. Other functions will just ignore such inodes,
1722 * if appropriate. I_NEW is used for waiting.
1723 *
1724 * I_SYNC Writeback of inode is running. The bit is set during
1725 * data writeback, and cleared with a wakeup on the bit
1726 * address once it is done. The bit is also used to pin
1727 * the inode in memory for flusher thread.
1728 *
1729 * I_REFERENCED Marks the inode as recently references on the LRU list.
1730 *
1731 * I_DIO_WAKEUP Never set. Only used as a key for wait_on_bit().
1732 *
1733 * Q: What is the difference between I_WILL_FREE and I_FREEING?
1734 */
1735#define I_DIRTY_SYNC (1 << 0)
1736#define I_DIRTY_DATASYNC (1 << 1)
1737#define I_DIRTY_PAGES (1 << 2)
1738#define __I_NEW 3
1739#define I_NEW (1 << __I_NEW)
1740#define I_WILL_FREE (1 << 4)
1741#define I_FREEING (1 << 5)
1742#define I_CLEAR (1 << 6)
1743#define __I_SYNC 7
1744#define I_SYNC (1 << __I_SYNC)
1745#define I_REFERENCED (1 << 8)
1746#define __I_DIO_WAKEUP 9
1747#define I_DIO_WAKEUP (1 << I_DIO_WAKEUP)
1748#define I_LINKABLE (1 << 10)
1749
1750#define I_DIRTY (I_DIRTY_SYNC | I_DIRTY_DATASYNC | I_DIRTY_PAGES)
1751
1752extern void __mark_inode_dirty(struct inode *, int);
1753static inline void mark_inode_dirty(struct inode *inode)
1754{
1755 __mark_inode_dirty(inode, I_DIRTY);
1756}
1757
1758static inline void mark_inode_dirty_sync(struct inode *inode)
1759{
1760 __mark_inode_dirty(inode, I_DIRTY_SYNC);
1761}
1762
1763extern void inc_nlink(struct inode *inode);
1764extern void drop_nlink(struct inode *inode);
1765extern void clear_nlink(struct inode *inode);
1766extern void set_nlink(struct inode *inode, unsigned int nlink);
1767
1768static inline void inode_inc_link_count(struct inode *inode)
1769{
1770 inc_nlink(inode);
1771 mark_inode_dirty(inode);
1772}
1773
1774static inline void inode_dec_link_count(struct inode *inode)
1775{
1776 drop_nlink(inode);
1777 mark_inode_dirty(inode);
1778}
1779
1780/**
1781 * inode_inc_iversion - increments i_version
1782 * @inode: inode that need to be updated
1783 *
1784 * Every time the inode is modified, the i_version field will be incremented.
1785 * The filesystem has to be mounted with i_version flag
1786 */
1787
1788static inline void inode_inc_iversion(struct inode *inode)
1789{
1790 spin_lock(&inode->i_lock);
1791 inode->i_version++;
1792 spin_unlock(&inode->i_lock);
1793}
1794
1795enum file_time_flags {
1796 S_ATIME = 1,
1797 S_MTIME = 2,
1798 S_CTIME = 4,
1799 S_VERSION = 8,
1800};
1801
1802extern void touch_atime(const struct path *);
1803static inline void file_accessed(struct file *file)
1804{
1805 if (!(file->f_flags & O_NOATIME))
1806 touch_atime(&file->f_path);
1807}
1808
1809int sync_inode(struct inode *inode, struct writeback_control *wbc);
1810int sync_inode_metadata(struct inode *inode, int wait);
1811
1812struct file_system_type {
1813 const char *name;
1814 int fs_flags;
1815#define FS_REQUIRES_DEV 1
1816#define FS_BINARY_MOUNTDATA 2
1817#define FS_HAS_SUBTYPE 4
1818#define FS_USERNS_MOUNT 8 /* Can be mounted by userns root */
1819#define FS_USERNS_DEV_MOUNT 16 /* A userns mount does not imply MNT_NODEV */
1820#define FS_RENAME_DOES_D_MOVE 32768 /* FS will handle d_move() during rename() internally. */
1821 struct dentry *(*mount) (struct file_system_type *, int,
1822 const char *, void *);
1823 void (*kill_sb) (struct super_block *);
1824 struct module *owner;
1825 struct file_system_type * next;
1826 struct hlist_head fs_supers;
1827
1828 struct lock_class_key s_lock_key;
1829 struct lock_class_key s_umount_key;
1830 struct lock_class_key s_vfs_rename_key;
1831 struct lock_class_key s_writers_key[SB_FREEZE_LEVELS];
1832
1833 struct lock_class_key i_lock_key;
1834 struct lock_class_key i_mutex_key;
1835 struct lock_class_key i_mutex_dir_key;
1836};
1837
1838#define MODULE_ALIAS_FS(NAME) MODULE_ALIAS("fs-" NAME)
1839
1840extern struct dentry *mount_ns(struct file_system_type *fs_type, int flags,
1841 void *data, int (*fill_super)(struct super_block *, void *, int));
1842extern struct dentry *mount_bdev(struct file_system_type *fs_type,
1843 int flags, const char *dev_name, void *data,
1844 int (*fill_super)(struct super_block *, void *, int));
1845extern struct dentry *mount_single(struct file_system_type *fs_type,
1846 int flags, void *data,
1847 int (*fill_super)(struct super_block *, void *, int));
1848extern struct dentry *mount_nodev(struct file_system_type *fs_type,
1849 int flags, void *data,
1850 int (*fill_super)(struct super_block *, void *, int));
1851extern struct dentry *mount_subtree(struct vfsmount *mnt, const char *path);
1852void generic_shutdown_super(struct super_block *sb);
1853void kill_block_super(struct super_block *sb);
1854void kill_anon_super(struct super_block *sb);
1855void kill_litter_super(struct super_block *sb);
1856void deactivate_super(struct super_block *sb);
1857void deactivate_locked_super(struct super_block *sb);
1858int set_anon_super(struct super_block *s, void *data);
1859int get_anon_bdev(dev_t *);
1860void free_anon_bdev(dev_t);
1861struct super_block *sget(struct file_system_type *type,
1862 int (*test)(struct super_block *,void *),
1863 int (*set)(struct super_block *,void *),
1864 int flags, void *data);
1865extern struct dentry *mount_pseudo(struct file_system_type *, char *,
1866 const struct super_operations *ops,
1867 const struct dentry_operations *dops,
1868 unsigned long);
1869
1870/* Alas, no aliases. Too much hassle with bringing module.h everywhere */
1871#define fops_get(fops) \
1872 (((fops) && try_module_get((fops)->owner) ? (fops) : NULL))
1873#define fops_put(fops) \
1874 do { if (fops) module_put((fops)->owner); } while(0)
1875/*
1876 * This one is to be used *ONLY* from ->open() instances.
1877 * fops must be non-NULL, pinned down *and* module dependencies
1878 * should be sufficient to pin the caller down as well.
1879 */
1880#define replace_fops(f, fops) \
1881 do { \
1882 struct file *__file = (f); \
1883 fops_put(__file->f_op); \
1884 BUG_ON(!(__file->f_op = (fops))); \
1885 } while(0)
1886
1887extern int register_filesystem(struct file_system_type *);
1888extern int unregister_filesystem(struct file_system_type *);
1889extern struct vfsmount *kern_mount_data(struct file_system_type *, void *data);
1890#define kern_mount(type) kern_mount_data(type, NULL)
1891extern void kern_unmount(struct vfsmount *mnt);
1892extern int may_umount_tree(struct vfsmount *);
1893extern int may_umount(struct vfsmount *);
1894extern long do_mount(const char *, const char __user *,
1895 const char *, unsigned long, void *);
1896extern struct vfsmount *collect_mounts(struct path *);
1897extern void drop_collected_mounts(struct vfsmount *);
1898extern int iterate_mounts(int (*)(struct vfsmount *, void *), void *,
1899 struct vfsmount *);
1900extern int vfs_statfs(struct path *, struct kstatfs *);
1901extern int user_statfs(const char __user *, struct kstatfs *);
1902extern int fd_statfs(int, struct kstatfs *);
1903extern int vfs_ustat(dev_t, struct kstatfs *);
1904extern int freeze_super(struct super_block *super);
1905extern int thaw_super(struct super_block *super);
1906extern bool our_mnt(struct vfsmount *mnt);
1907extern bool fs_fully_visible(struct file_system_type *);
1908
1909extern int current_umask(void);
1910
1911extern void ihold(struct inode * inode);
1912extern void iput(struct inode *);
1913
1914static inline struct inode *file_inode(const struct file *f)
1915{
1916 return f->f_inode;
1917}
1918
1919/* /sys/fs */
1920extern struct kobject *fs_kobj;
1921
1922#define MAX_RW_COUNT (INT_MAX & PAGE_CACHE_MASK)
1923
1924#define FLOCK_VERIFY_READ 1
1925#define FLOCK_VERIFY_WRITE 2
1926
1927#ifdef CONFIG_FILE_LOCKING
1928extern int locks_mandatory_locked(struct file *);
1929extern int locks_mandatory_area(int, struct inode *, struct file *, loff_t, size_t);
1930
1931/*
1932 * Candidates for mandatory locking have the setgid bit set
1933 * but no group execute bit - an otherwise meaningless combination.
1934 */
1935
1936static inline int __mandatory_lock(struct inode *ino)
1937{
1938 return (ino->i_mode & (S_ISGID | S_IXGRP)) == S_ISGID;
1939}
1940
1941/*
1942 * ... and these candidates should be on MS_MANDLOCK mounted fs,
1943 * otherwise these will be advisory locks
1944 */
1945
1946static inline int mandatory_lock(struct inode *ino)
1947{
1948 return IS_MANDLOCK(ino) && __mandatory_lock(ino);
1949}
1950
1951static inline int locks_verify_locked(struct file *file)
1952{
1953 if (mandatory_lock(file_inode(file)))
1954 return locks_mandatory_locked(file);
1955 return 0;
1956}
1957
1958static inline int locks_verify_truncate(struct inode *inode,
1959 struct file *filp,
1960 loff_t size)
1961{
1962 if (inode->i_flock && mandatory_lock(inode))
1963 return locks_mandatory_area(
1964 FLOCK_VERIFY_WRITE, inode, filp,
1965 size < inode->i_size ? size : inode->i_size,
1966 (size < inode->i_size ? inode->i_size - size
1967 : size - inode->i_size)
1968 );
1969 return 0;
1970}
1971
1972static inline int break_lease(struct inode *inode, unsigned int mode)
1973{
1974 /*
1975 * Since this check is lockless, we must ensure that any refcounts
1976 * taken are done before checking inode->i_flock. Otherwise, we could
1977 * end up racing with tasks trying to set a new lease on this file.
1978 */
1979 smp_mb();
1980 if (inode->i_flock)
1981 return __break_lease(inode, mode, FL_LEASE);
1982 return 0;
1983}
1984
1985static inline int break_deleg(struct inode *inode, unsigned int mode)
1986{
1987 /*
1988 * Since this check is lockless, we must ensure that any refcounts
1989 * taken are done before checking inode->i_flock. Otherwise, we could
1990 * end up racing with tasks trying to set a new lease on this file.
1991 */
1992 smp_mb();
1993 if (inode->i_flock)
1994 return __break_lease(inode, mode, FL_DELEG);
1995 return 0;
1996}
1997
1998static inline int try_break_deleg(struct inode *inode, struct inode **delegated_inode)
1999{
2000 int ret;
2001
2002 ret = break_deleg(inode, O_WRONLY|O_NONBLOCK);
2003 if (ret == -EWOULDBLOCK && delegated_inode) {
2004 *delegated_inode = inode;
2005 ihold(inode);
2006 }
2007 return ret;
2008}
2009
2010static inline int break_deleg_wait(struct inode **delegated_inode)
2011{
2012 int ret;
2013
2014 ret = break_deleg(*delegated_inode, O_WRONLY);
2015 iput(*delegated_inode);
2016 *delegated_inode = NULL;
2017 return ret;
2018}
2019
2020#else /* !CONFIG_FILE_LOCKING */
2021static inline int locks_mandatory_locked(struct file *file)
2022{
2023 return 0;
2024}
2025
2026static inline int locks_mandatory_area(int rw, struct inode *inode,
2027 struct file *filp, loff_t offset,
2028 size_t count)
2029{
2030 return 0;
2031}
2032
2033static inline int __mandatory_lock(struct inode *inode)
2034{
2035 return 0;
2036}
2037
2038static inline int mandatory_lock(struct inode *inode)
2039{
2040 return 0;
2041}
2042
2043static inline int locks_verify_locked(struct file *file)
2044{
2045 return 0;
2046}
2047
2048static inline int locks_verify_truncate(struct inode *inode, struct file *filp,
2049 size_t size)
2050{
2051 return 0;
2052}
2053
2054static inline int break_lease(struct inode *inode, unsigned int mode)
2055{
2056 return 0;
2057}
2058
2059static inline int break_deleg(struct inode *inode, unsigned int mode)
2060{
2061 return 0;
2062}
2063
2064static inline int try_break_deleg(struct inode *inode, struct inode **delegated_inode)
2065{
2066 return 0;
2067}
2068
2069static inline int break_deleg_wait(struct inode **delegated_inode)
2070{
2071 BUG();
2072 return 0;
2073}
2074
2075#endif /* CONFIG_FILE_LOCKING */
2076
2077/* fs/open.c */
2078struct audit_names;
2079struct filename {
2080 const char *name; /* pointer to actual string */
2081 const __user char *uptr; /* original userland pointer */
2082 struct audit_names *aname;
2083 bool separate; /* should "name" be freed? */
2084};
2085
2086extern long vfs_truncate(struct path *, loff_t);
2087extern int do_truncate(struct dentry *, loff_t start, unsigned int time_attrs,
2088 struct file *filp);
2089extern int vfs_fallocate(struct file *file, int mode, loff_t offset,
2090 loff_t len);
2091extern long do_sys_open(int dfd, const char __user *filename, int flags,
2092 umode_t mode);
2093extern struct file *file_open_name(struct filename *, int, umode_t);
2094extern struct file *filp_open(const char *, int, umode_t);
2095extern struct file *file_open_root(struct dentry *, struct vfsmount *,
2096 const char *, int);
2097extern int vfs_open(const struct path *, struct file *, const struct cred *);
2098extern struct file * dentry_open(const struct path *, int, const struct cred *);
2099extern int filp_close(struct file *, fl_owner_t id);
2100
2101extern struct filename *getname_flags(const char __user *, int, int *);
2102extern struct filename *getname(const char __user *);
2103extern struct filename *getname_kernel(const char *);
2104
2105enum {
2106 FILE_CREATED = 1,
2107 FILE_OPENED = 2
2108};
2109extern int finish_open(struct file *file, struct dentry *dentry,
2110 int (*open)(struct inode *, struct file *),
2111 int *opened);
2112extern int finish_no_open(struct file *file, struct dentry *dentry);
2113
2114/* fs/ioctl.c */
2115
2116extern int ioctl_preallocate(struct file *filp, void __user *argp);
2117
2118/* fs/dcache.c */
2119extern void __init vfs_caches_init_early(void);
2120extern void __init vfs_caches_init(unsigned long);
2121
2122extern struct kmem_cache *names_cachep;
2123
2124extern void final_putname(struct filename *name);
2125
2126#define __getname() kmem_cache_alloc(names_cachep, GFP_KERNEL)
2127#define __putname(name) kmem_cache_free(names_cachep, (void *)(name))
2128#ifndef CONFIG_AUDITSYSCALL
2129#define putname(name) final_putname(name)
2130#else
2131extern void putname(struct filename *name);
2132#endif
2133
2134#ifdef CONFIG_BLOCK
2135extern int register_blkdev(unsigned int, const char *);
2136extern void unregister_blkdev(unsigned int, const char *);
2137extern struct block_device *bdget(dev_t);
2138extern struct block_device *bdgrab(struct block_device *bdev);
2139extern void bd_set_size(struct block_device *, loff_t size);
2140extern void bd_forget(struct inode *inode);
2141extern void bdput(struct block_device *);
2142extern void invalidate_bdev(struct block_device *);
2143extern void iterate_bdevs(void (*)(struct block_device *, void *), void *);
2144extern int sync_blockdev(struct block_device *bdev);
2145extern void kill_bdev(struct block_device *);
2146extern struct super_block *freeze_bdev(struct block_device *);
2147extern void emergency_thaw_all(void);
2148extern int thaw_bdev(struct block_device *bdev, struct super_block *sb);
2149extern int fsync_bdev(struct block_device *);
2150extern int sb_is_blkdev_sb(struct super_block *sb);
2151#else
2152static inline void bd_forget(struct inode *inode) {}
2153static inline int sync_blockdev(struct block_device *bdev) { return 0; }
2154static inline void kill_bdev(struct block_device *bdev) {}
2155static inline void invalidate_bdev(struct block_device *bdev) {}
2156
2157static inline struct super_block *freeze_bdev(struct block_device *sb)
2158{
2159 return NULL;
2160}
2161
2162static inline int thaw_bdev(struct block_device *bdev, struct super_block *sb)
2163{
2164 return 0;
2165}
2166
2167static inline void iterate_bdevs(void (*f)(struct block_device *, void *), void *arg)
2168{
2169}
2170
2171static inline int sb_is_blkdev_sb(struct super_block *sb)
2172{
2173 return 0;
2174}
2175#endif
2176extern int sync_filesystem(struct super_block *);
2177extern const struct file_operations def_blk_fops;
2178extern const struct file_operations def_chr_fops;
2179#ifdef CONFIG_BLOCK
2180extern int ioctl_by_bdev(struct block_device *, unsigned, unsigned long);
2181extern int blkdev_ioctl(struct block_device *, fmode_t, unsigned, unsigned long);
2182extern long compat_blkdev_ioctl(struct file *, unsigned, unsigned long);
2183extern int blkdev_get(struct block_device *bdev, fmode_t mode, void *holder);
2184extern struct block_device *blkdev_get_by_path(const char *path, fmode_t mode,
2185 void *holder);
2186extern struct block_device *blkdev_get_by_dev(dev_t dev, fmode_t mode,
2187 void *holder);
2188extern void blkdev_put(struct block_device *bdev, fmode_t mode);
2189#ifdef CONFIG_SYSFS
2190extern int bd_link_disk_holder(struct block_device *bdev, struct gendisk *disk);
2191extern void bd_unlink_disk_holder(struct block_device *bdev,
2192 struct gendisk *disk);
2193#else
2194static inline int bd_link_disk_holder(struct block_device *bdev,
2195 struct gendisk *disk)
2196{
2197 return 0;
2198}
2199static inline void bd_unlink_disk_holder(struct block_device *bdev,
2200 struct gendisk *disk)
2201{
2202}
2203#endif
2204#endif
2205
2206/* fs/char_dev.c */
2207#define CHRDEV_MAJOR_HASH_SIZE 255
2208extern int alloc_chrdev_region(dev_t *, unsigned, unsigned, const char *);
2209extern int register_chrdev_region(dev_t, unsigned, const char *);
2210extern int __register_chrdev(unsigned int major, unsigned int baseminor,
2211 unsigned int count, const char *name,
2212 const struct file_operations *fops);
2213extern void __unregister_chrdev(unsigned int major, unsigned int baseminor,
2214 unsigned int count, const char *name);
2215extern void unregister_chrdev_region(dev_t, unsigned);
2216extern void chrdev_show(struct seq_file *,off_t);
2217
2218static inline int register_chrdev(unsigned int major, const char *name,
2219 const struct file_operations *fops)
2220{
2221 return __register_chrdev(major, 0, 256, name, fops);
2222}
2223
2224static inline void unregister_chrdev(unsigned int major, const char *name)
2225{
2226 __unregister_chrdev(major, 0, 256, name);
2227}
2228
2229/* fs/block_dev.c */
2230#define BDEVNAME_SIZE 32 /* Largest string for a blockdev identifier */
2231#define BDEVT_SIZE 10 /* Largest string for MAJ:MIN for blkdev */
2232
2233#ifdef CONFIG_BLOCK
2234#define BLKDEV_MAJOR_HASH_SIZE 255
2235extern const char *__bdevname(dev_t, char *buffer);
2236extern const char *bdevname(struct block_device *bdev, char *buffer);
2237extern struct block_device *lookup_bdev(const char *);
2238extern void blkdev_show(struct seq_file *,off_t);
2239
2240#else
2241#define BLKDEV_MAJOR_HASH_SIZE 0
2242#endif
2243
2244extern void init_special_inode(struct inode *, umode_t, dev_t);
2245
2246/* Invalid inode operations -- fs/bad_inode.c */
2247extern void make_bad_inode(struct inode *);
2248extern int is_bad_inode(struct inode *);
2249
2250#ifdef CONFIG_BLOCK
2251/*
2252 * return READ, READA, or WRITE
2253 */
2254#define bio_rw(bio) ((bio)->bi_rw & (RW_MASK | RWA_MASK))
2255
2256/*
2257 * return data direction, READ or WRITE
2258 */
2259#define bio_data_dir(bio) ((bio)->bi_rw & 1)
2260
2261extern void check_disk_size_change(struct gendisk *disk,
2262 struct block_device *bdev);
2263extern int revalidate_disk(struct gendisk *);
2264extern int check_disk_change(struct block_device *);
2265extern int __invalidate_device(struct block_device *, bool);
2266extern int invalidate_partition(struct gendisk *, int);
2267#endif
2268unsigned long invalidate_mapping_pages(struct address_space *mapping,
2269 pgoff_t start, pgoff_t end);
2270
2271static inline void invalidate_remote_inode(struct inode *inode)
2272{
2273 if (S_ISREG(inode->i_mode) || S_ISDIR(inode->i_mode) ||
2274 S_ISLNK(inode->i_mode))
2275 invalidate_mapping_pages(inode->i_mapping, 0, -1);
2276}
2277extern int invalidate_inode_pages2(struct address_space *mapping);
2278extern int invalidate_inode_pages2_range(struct address_space *mapping,
2279 pgoff_t start, pgoff_t end);
2280extern int write_inode_now(struct inode *, int);
2281extern int filemap_fdatawrite(struct address_space *);
2282extern int filemap_flush(struct address_space *);
2283extern int filemap_fdatawait(struct address_space *);
2284extern int filemap_fdatawait_range(struct address_space *, loff_t lstart,
2285 loff_t lend);
2286extern int filemap_write_and_wait(struct address_space *mapping);
2287extern int filemap_write_and_wait_range(struct address_space *mapping,
2288 loff_t lstart, loff_t lend);
2289extern int __filemap_fdatawrite_range(struct address_space *mapping,
2290 loff_t start, loff_t end, int sync_mode);
2291extern int filemap_fdatawrite_range(struct address_space *mapping,
2292 loff_t start, loff_t end);
2293
2294extern int vfs_fsync_range(struct file *file, loff_t start, loff_t end,
2295 int datasync);
2296extern int vfs_fsync(struct file *file, int datasync);
2297static inline int generic_write_sync(struct file *file, loff_t pos, loff_t count)
2298{
2299 if (!(file->f_flags & O_DSYNC) && !IS_SYNC(file->f_mapping->host))
2300 return 0;
2301 return vfs_fsync_range(file, pos, pos + count - 1,
2302 (file->f_flags & __O_SYNC) ? 0 : 1);
2303}
2304extern void emergency_sync(void);
2305extern void emergency_remount(void);
2306#ifdef CONFIG_BLOCK
2307extern sector_t bmap(struct inode *, sector_t);
2308#endif
2309extern int notify_change(struct dentry *, struct iattr *, struct inode **);
2310extern int inode_permission(struct inode *, int);
2311extern int __inode_permission(struct inode *, int);
2312extern int generic_permission(struct inode *, int);
2313extern int __check_sticky(struct inode *dir, struct inode *inode);
2314
2315static inline bool execute_ok(struct inode *inode)
2316{
2317 return (inode->i_mode & S_IXUGO) || S_ISDIR(inode->i_mode);
2318}
2319
2320static inline void file_start_write(struct file *file)
2321{
2322 if (!S_ISREG(file_inode(file)->i_mode))
2323 return;
2324 __sb_start_write(file_inode(file)->i_sb, SB_FREEZE_WRITE, true);
2325}
2326
2327static inline bool file_start_write_trylock(struct file *file)
2328{
2329 if (!S_ISREG(file_inode(file)->i_mode))
2330 return true;
2331 return __sb_start_write(file_inode(file)->i_sb, SB_FREEZE_WRITE, false);
2332}
2333
2334static inline void file_end_write(struct file *file)
2335{
2336 if (!S_ISREG(file_inode(file)->i_mode))
2337 return;
2338 __sb_end_write(file_inode(file)->i_sb, SB_FREEZE_WRITE);
2339}
2340
2341/*
2342 * get_write_access() gets write permission for a file.
2343 * put_write_access() releases this write permission.
2344 * This is used for regular files.
2345 * We cannot support write (and maybe mmap read-write shared) accesses and
2346 * MAP_DENYWRITE mmappings simultaneously. The i_writecount field of an inode
2347 * can have the following values:
2348 * 0: no writers, no VM_DENYWRITE mappings
2349 * < 0: (-i_writecount) vm_area_structs with VM_DENYWRITE set exist
2350 * > 0: (i_writecount) users are writing to the file.
2351 *
2352 * Normally we operate on that counter with atomic_{inc,dec} and it's safe
2353 * except for the cases where we don't hold i_writecount yet. Then we need to
2354 * use {get,deny}_write_access() - these functions check the sign and refuse
2355 * to do the change if sign is wrong.
2356 */
2357static inline int get_write_access(struct inode *inode)
2358{
2359 return atomic_inc_unless_negative(&inode->i_writecount) ? 0 : -ETXTBSY;
2360}
2361static inline int deny_write_access(struct file *file)
2362{
2363 struct inode *inode = file_inode(file);
2364 return atomic_dec_unless_positive(&inode->i_writecount) ? 0 : -ETXTBSY;
2365}
2366static inline void put_write_access(struct inode * inode)
2367{
2368 atomic_dec(&inode->i_writecount);
2369}
2370static inline void allow_write_access(struct file *file)
2371{
2372 if (file)
2373 atomic_inc(&file_inode(file)->i_writecount);
2374}
2375static inline bool inode_is_open_for_write(const struct inode *inode)
2376{
2377 return atomic_read(&inode->i_writecount) > 0;
2378}
2379
2380#ifdef CONFIG_IMA
2381static inline void i_readcount_dec(struct inode *inode)
2382{
2383 BUG_ON(!atomic_read(&inode->i_readcount));
2384 atomic_dec(&inode->i_readcount);
2385}
2386static inline void i_readcount_inc(struct inode *inode)
2387{
2388 atomic_inc(&inode->i_readcount);
2389}
2390#else
2391static inline void i_readcount_dec(struct inode *inode)
2392{
2393 return;
2394}
2395static inline void i_readcount_inc(struct inode *inode)
2396{
2397 return;
2398}
2399#endif
2400extern int do_pipe_flags(int *, int);
2401
2402extern int kernel_read(struct file *, loff_t, char *, unsigned long);
2403extern ssize_t kernel_write(struct file *, const char *, size_t, loff_t);
2404extern ssize_t __kernel_write(struct file *, const char *, size_t, loff_t *);
2405extern struct file * open_exec(const char *);
2406
2407/* fs/dcache.c -- generic fs support functions */
2408extern int is_subdir(struct dentry *, struct dentry *);
2409extern int path_is_under(struct path *, struct path *);
2410
2411#include <linux/err.h>
2412
2413/* needed for stackable file system support */
2414extern loff_t default_llseek(struct file *file, loff_t offset, int whence);
2415
2416extern loff_t vfs_llseek(struct file *file, loff_t offset, int whence);
2417
2418extern int inode_init_always(struct super_block *, struct inode *);
2419extern void inode_init_once(struct inode *);
2420extern void address_space_init_once(struct address_space *mapping);
2421extern struct inode * igrab(struct inode *);
2422extern ino_t iunique(struct super_block *, ino_t);
2423extern int inode_needs_sync(struct inode *inode);
2424extern int generic_delete_inode(struct inode *inode);
2425static inline int generic_drop_inode(struct inode *inode)
2426{
2427 return !inode->i_nlink || inode_unhashed(inode);
2428}
2429
2430extern struct inode *ilookup5_nowait(struct super_block *sb,
2431 unsigned long hashval, int (*test)(struct inode *, void *),
2432 void *data);
2433extern struct inode *ilookup5(struct super_block *sb, unsigned long hashval,
2434 int (*test)(struct inode *, void *), void *data);
2435extern struct inode *ilookup(struct super_block *sb, unsigned long ino);
2436
2437extern struct inode * iget5_locked(struct super_block *, unsigned long, int (*test)(struct inode *, void *), int (*set)(struct inode *, void *), void *);
2438extern struct inode * iget_locked(struct super_block *, unsigned long);
2439extern int insert_inode_locked4(struct inode *, unsigned long, int (*test)(struct inode *, void *), void *);
2440extern int insert_inode_locked(struct inode *);
2441#ifdef CONFIG_DEBUG_LOCK_ALLOC
2442extern void lockdep_annotate_inode_mutex_key(struct inode *inode);
2443#else
2444static inline void lockdep_annotate_inode_mutex_key(struct inode *inode) { };
2445#endif
2446extern void unlock_new_inode(struct inode *);
2447extern unsigned int get_next_ino(void);
2448
2449extern void __iget(struct inode * inode);
2450extern void iget_failed(struct inode *);
2451extern void clear_inode(struct inode *);
2452extern void __destroy_inode(struct inode *);
2453extern struct inode *new_inode_pseudo(struct super_block *sb);
2454extern struct inode *new_inode(struct super_block *sb);
2455extern void free_inode_nonrcu(struct inode *inode);
2456extern int should_remove_suid(struct dentry *);
2457extern int file_remove_suid(struct file *);
2458
2459extern void __insert_inode_hash(struct inode *, unsigned long hashval);
2460static inline void insert_inode_hash(struct inode *inode)
2461{
2462 __insert_inode_hash(inode, inode->i_ino);
2463}
2464
2465extern void __remove_inode_hash(struct inode *);
2466static inline void remove_inode_hash(struct inode *inode)
2467{
2468 if (!inode_unhashed(inode))
2469 __remove_inode_hash(inode);
2470}
2471
2472extern void inode_sb_list_add(struct inode *inode);
2473
2474#ifdef CONFIG_BLOCK
2475extern void submit_bio(int, struct bio *);
2476extern int bdev_read_only(struct block_device *);
2477#endif
2478extern int set_blocksize(struct block_device *, int);
2479extern int sb_set_blocksize(struct super_block *, int);
2480extern int sb_min_blocksize(struct super_block *, int);
2481
2482extern int generic_file_mmap(struct file *, struct vm_area_struct *);
2483extern int generic_file_readonly_mmap(struct file *, struct vm_area_struct *);
2484extern int generic_file_remap_pages(struct vm_area_struct *, unsigned long addr,
2485 unsigned long size, pgoff_t pgoff);
2486int generic_write_checks(struct file *file, loff_t *pos, size_t *count, int isblk);
2487extern ssize_t generic_file_read_iter(struct kiocb *, struct iov_iter *);
2488extern ssize_t __generic_file_write_iter(struct kiocb *, struct iov_iter *);
2489extern ssize_t generic_file_write_iter(struct kiocb *, struct iov_iter *);
2490extern ssize_t generic_file_direct_write(struct kiocb *, struct iov_iter *, loff_t);
2491extern ssize_t generic_perform_write(struct file *, struct iov_iter *, loff_t);
2492extern ssize_t do_sync_read(struct file *filp, char __user *buf, size_t len, loff_t *ppos);
2493extern ssize_t do_sync_write(struct file *filp, const char __user *buf, size_t len, loff_t *ppos);
2494extern ssize_t new_sync_read(struct file *filp, char __user *buf, size_t len, loff_t *ppos);
2495extern ssize_t new_sync_write(struct file *filp, const char __user *buf, size_t len, loff_t *ppos);
2496
2497/* fs/block_dev.c */
2498extern ssize_t blkdev_read_iter(struct kiocb *iocb, struct iov_iter *to);
2499extern ssize_t blkdev_write_iter(struct kiocb *iocb, struct iov_iter *from);
2500extern int blkdev_fsync(struct file *filp, loff_t start, loff_t end,
2501 int datasync);
2502extern void block_sync_page(struct page *page);
2503
2504/* fs/splice.c */
2505extern ssize_t generic_file_splice_read(struct file *, loff_t *,
2506 struct pipe_inode_info *, size_t, unsigned int);
2507extern ssize_t default_file_splice_read(struct file *, loff_t *,
2508 struct pipe_inode_info *, size_t, unsigned int);
2509extern ssize_t iter_file_splice_write(struct pipe_inode_info *,
2510 struct file *, loff_t *, size_t, unsigned int);
2511extern ssize_t generic_splice_sendpage(struct pipe_inode_info *pipe,
2512 struct file *out, loff_t *, size_t len, unsigned int flags);
2513extern long do_splice_direct(struct file *in, loff_t *ppos, struct file *out,
2514 loff_t *opos, size_t len, unsigned int flags);
2515
2516
2517extern void
2518file_ra_state_init(struct file_ra_state *ra, struct address_space *mapping);
2519extern loff_t noop_llseek(struct file *file, loff_t offset, int whence);
2520extern loff_t no_llseek(struct file *file, loff_t offset, int whence);
2521extern loff_t vfs_setpos(struct file *file, loff_t offset, loff_t maxsize);
2522extern loff_t generic_file_llseek(struct file *file, loff_t offset, int whence);
2523extern loff_t generic_file_llseek_size(struct file *file, loff_t offset,
2524 int whence, loff_t maxsize, loff_t eof);
2525extern loff_t fixed_size_llseek(struct file *file, loff_t offset,
2526 int whence, loff_t size);
2527extern int generic_file_open(struct inode * inode, struct file * filp);
2528extern int nonseekable_open(struct inode * inode, struct file * filp);
2529
2530#ifdef CONFIG_FS_XIP
2531extern ssize_t xip_file_read(struct file *filp, char __user *buf, size_t len,
2532 loff_t *ppos);
2533extern int xip_file_mmap(struct file * file, struct vm_area_struct * vma);
2534extern ssize_t xip_file_write(struct file *filp, const char __user *buf,
2535 size_t len, loff_t *ppos);
2536extern int xip_truncate_page(struct address_space *mapping, loff_t from);
2537#else
2538static inline int xip_truncate_page(struct address_space *mapping, loff_t from)
2539{
2540 return 0;
2541}
2542#endif
2543
2544#ifdef CONFIG_BLOCK
2545typedef void (dio_submit_t)(int rw, struct bio *bio, struct inode *inode,
2546 loff_t file_offset);
2547
2548enum {
2549 /* need locking between buffered and direct access */
2550 DIO_LOCKING = 0x01,
2551
2552 /* filesystem does not support filling holes */
2553 DIO_SKIP_HOLES = 0x02,
2554
2555 /* filesystem can handle aio writes beyond i_size */
2556 DIO_ASYNC_EXTEND = 0x04,
2557};
2558
2559void dio_end_io(struct bio *bio, int error);
2560
2561ssize_t __blockdev_direct_IO(int rw, struct kiocb *iocb, struct inode *inode,
2562 struct block_device *bdev, struct iov_iter *iter, loff_t offset,
2563 get_block_t get_block, dio_iodone_t end_io,
2564 dio_submit_t submit_io, int flags);
2565
2566static inline ssize_t blockdev_direct_IO(int rw, struct kiocb *iocb,
2567 struct inode *inode, struct iov_iter *iter, loff_t offset,
2568 get_block_t get_block)
2569{
2570 return __blockdev_direct_IO(rw, iocb, inode, inode->i_sb->s_bdev, iter,
2571 offset, get_block, NULL, NULL,
2572 DIO_LOCKING | DIO_SKIP_HOLES);
2573}
2574#endif
2575
2576void inode_dio_wait(struct inode *inode);
2577void inode_dio_done(struct inode *inode);
2578
2579extern void inode_set_flags(struct inode *inode, unsigned int flags,
2580 unsigned int mask);
2581
2582extern const struct file_operations generic_ro_fops;
2583
2584#define special_file(m) (S_ISCHR(m)||S_ISBLK(m)||S_ISFIFO(m)||S_ISSOCK(m))
2585
2586extern int readlink_copy(char __user *, int, const char *);
2587extern int page_readlink(struct dentry *, char __user *, int);
2588extern void *page_follow_link_light(struct dentry *, struct nameidata *);
2589extern void page_put_link(struct dentry *, struct nameidata *, void *);
2590extern int __page_symlink(struct inode *inode, const char *symname, int len,
2591 int nofs);
2592extern int page_symlink(struct inode *inode, const char *symname, int len);
2593extern const struct inode_operations page_symlink_inode_operations;
2594extern void kfree_put_link(struct dentry *, struct nameidata *, void *);
2595extern int generic_readlink(struct dentry *, char __user *, int);
2596extern void generic_fillattr(struct inode *, struct kstat *);
2597int vfs_getattr_nosec(struct path *path, struct kstat *stat);
2598extern int vfs_getattr(struct path *, struct kstat *);
2599void __inode_add_bytes(struct inode *inode, loff_t bytes);
2600void inode_add_bytes(struct inode *inode, loff_t bytes);
2601void __inode_sub_bytes(struct inode *inode, loff_t bytes);
2602void inode_sub_bytes(struct inode *inode, loff_t bytes);
2603loff_t inode_get_bytes(struct inode *inode);
2604void inode_set_bytes(struct inode *inode, loff_t bytes);
2605
2606extern int vfs_readdir(struct file *, filldir_t, void *);
2607extern int iterate_dir(struct file *, struct dir_context *);
2608
2609extern int vfs_stat(const char __user *, struct kstat *);
2610extern int vfs_lstat(const char __user *, struct kstat *);
2611extern int vfs_fstat(unsigned int, struct kstat *);
2612extern int vfs_fstatat(int , const char __user *, struct kstat *, int);
2613
2614extern int do_vfs_ioctl(struct file *filp, unsigned int fd, unsigned int cmd,
2615 unsigned long arg);
2616extern int __generic_block_fiemap(struct inode *inode,
2617 struct fiemap_extent_info *fieinfo,
2618 loff_t start, loff_t len,
2619 get_block_t *get_block);
2620extern int generic_block_fiemap(struct inode *inode,
2621 struct fiemap_extent_info *fieinfo, u64 start,
2622 u64 len, get_block_t *get_block);
2623
2624extern void get_filesystem(struct file_system_type *fs);
2625extern void put_filesystem(struct file_system_type *fs);
2626extern struct file_system_type *get_fs_type(const char *name);
2627extern struct super_block *get_super(struct block_device *);
2628extern struct super_block *get_super_thawed(struct block_device *);
2629extern struct super_block *get_active_super(struct block_device *bdev);
2630extern void drop_super(struct super_block *sb);
2631extern void iterate_supers(void (*)(struct super_block *, void *), void *);
2632extern void iterate_supers_type(struct file_system_type *,
2633 void (*)(struct super_block *, void *), void *);
2634
2635extern int dcache_dir_open(struct inode *, struct file *);
2636extern int dcache_dir_close(struct inode *, struct file *);
2637extern loff_t dcache_dir_lseek(struct file *, loff_t, int);
2638extern int dcache_readdir(struct file *, struct dir_context *);
2639extern int simple_setattr(struct dentry *, struct iattr *);
2640extern int simple_getattr(struct vfsmount *, struct dentry *, struct kstat *);
2641extern int simple_statfs(struct dentry *, struct kstatfs *);
2642extern int simple_open(struct inode *inode, struct file *file);
2643extern int simple_link(struct dentry *, struct inode *, struct dentry *);
2644extern int simple_unlink(struct inode *, struct dentry *);
2645extern int simple_rmdir(struct inode *, struct dentry *);
2646extern int simple_rename(struct inode *, struct dentry *, struct inode *, struct dentry *);
2647extern int noop_fsync(struct file *, loff_t, loff_t, int);
2648extern int simple_empty(struct dentry *);
2649extern int simple_readpage(struct file *file, struct page *page);
2650extern int simple_write_begin(struct file *file, struct address_space *mapping,
2651 loff_t pos, unsigned len, unsigned flags,
2652 struct page **pagep, void **fsdata);
2653extern int simple_write_end(struct file *file, struct address_space *mapping,
2654 loff_t pos, unsigned len, unsigned copied,
2655 struct page *page, void *fsdata);
2656extern int always_delete_dentry(const struct dentry *);
2657extern struct inode *alloc_anon_inode(struct super_block *);
2658extern int simple_nosetlease(struct file *, long, struct file_lock **, void **);
2659extern const struct dentry_operations simple_dentry_operations;
2660
2661extern struct dentry *simple_lookup(struct inode *, struct dentry *, unsigned int flags);
2662extern ssize_t generic_read_dir(struct file *, char __user *, size_t, loff_t *);
2663extern const struct file_operations simple_dir_operations;
2664extern const struct inode_operations simple_dir_inode_operations;
2665struct tree_descr { char *name; const struct file_operations *ops; int mode; };
2666struct dentry *d_alloc_name(struct dentry *, const char *);
2667extern int simple_fill_super(struct super_block *, unsigned long, struct tree_descr *);
2668extern int simple_pin_fs(struct file_system_type *, struct vfsmount **mount, int *count);
2669extern void simple_release_fs(struct vfsmount **mount, int *count);
2670
2671extern ssize_t simple_read_from_buffer(void __user *to, size_t count,
2672 loff_t *ppos, const void *from, size_t available);
2673extern ssize_t simple_write_to_buffer(void *to, size_t available, loff_t *ppos,
2674 const void __user *from, size_t count);
2675
2676extern int __generic_file_fsync(struct file *, loff_t, loff_t, int);
2677extern int generic_file_fsync(struct file *, loff_t, loff_t, int);
2678
2679extern int generic_check_addressable(unsigned, u64);
2680
2681#ifdef CONFIG_MIGRATION
2682extern int buffer_migrate_page(struct address_space *,
2683 struct page *, struct page *,
2684 enum migrate_mode);
2685#else
2686#define buffer_migrate_page NULL
2687#endif
2688
2689extern int inode_change_ok(const struct inode *, struct iattr *);
2690extern int inode_newsize_ok(const struct inode *, loff_t offset);
2691extern void setattr_copy(struct inode *inode, const struct iattr *attr);
2692
2693extern int file_update_time(struct file *file);
2694
2695extern int generic_show_options(struct seq_file *m, struct dentry *root);
2696extern void save_mount_options(struct super_block *sb, char *options);
2697extern void replace_mount_options(struct super_block *sb, char *options);
2698
2699static inline ino_t parent_ino(struct dentry *dentry)
2700{
2701 ino_t res;
2702
2703 /*
2704 * Don't strictly need d_lock here? If the parent ino could change
2705 * then surely we'd have a deeper race in the caller?
2706 */
2707 spin_lock(&dentry->d_lock);
2708 res = dentry->d_parent->d_inode->i_ino;
2709 spin_unlock(&dentry->d_lock);
2710 return res;
2711}
2712
2713/* Transaction based IO helpers */
2714
2715/*
2716 * An argresp is stored in an allocated page and holds the
2717 * size of the argument or response, along with its content
2718 */
2719struct simple_transaction_argresp {
2720 ssize_t size;
2721 char data[0];
2722};
2723
2724#define SIMPLE_TRANSACTION_LIMIT (PAGE_SIZE - sizeof(struct simple_transaction_argresp))
2725
2726char *simple_transaction_get(struct file *file, const char __user *buf,
2727 size_t size);
2728ssize_t simple_transaction_read(struct file *file, char __user *buf,
2729 size_t size, loff_t *pos);
2730int simple_transaction_release(struct inode *inode, struct file *file);
2731
2732void simple_transaction_set(struct file *file, size_t n);
2733
2734/*
2735 * simple attribute files
2736 *
2737 * These attributes behave similar to those in sysfs:
2738 *
2739 * Writing to an attribute immediately sets a value, an open file can be
2740 * written to multiple times.
2741 *
2742 * Reading from an attribute creates a buffer from the value that might get
2743 * read with multiple read calls. When the attribute has been read
2744 * completely, no further read calls are possible until the file is opened
2745 * again.
2746 *
2747 * All attributes contain a text representation of a numeric value
2748 * that are accessed with the get() and set() functions.
2749 */
2750#define DEFINE_SIMPLE_ATTRIBUTE(__fops, __get, __set, __fmt) \
2751static int __fops ## _open(struct inode *inode, struct file *file) \
2752{ \
2753 __simple_attr_check_format(__fmt, 0ull); \
2754 return simple_attr_open(inode, file, __get, __set, __fmt); \
2755} \
2756static const struct file_operations __fops = { \
2757 .owner = THIS_MODULE, \
2758 .open = __fops ## _open, \
2759 .release = simple_attr_release, \
2760 .read = simple_attr_read, \
2761 .write = simple_attr_write, \
2762 .llseek = generic_file_llseek, \
2763}
2764
2765static inline __printf(1, 2)
2766void __simple_attr_check_format(const char *fmt, ...)
2767{
2768 /* don't do anything, just let the compiler check the arguments; */
2769}
2770
2771int simple_attr_open(struct inode *inode, struct file *file,
2772 int (*get)(void *, u64 *), int (*set)(void *, u64),
2773 const char *fmt);
2774int simple_attr_release(struct inode *inode, struct file *file);
2775ssize_t simple_attr_read(struct file *file, char __user *buf,
2776 size_t len, loff_t *ppos);
2777ssize_t simple_attr_write(struct file *file, const char __user *buf,
2778 size_t len, loff_t *ppos);
2779
2780struct ctl_table;
2781int proc_nr_files(struct ctl_table *table, int write,
2782 void __user *buffer, size_t *lenp, loff_t *ppos);
2783int proc_nr_dentry(struct ctl_table *table, int write,
2784 void __user *buffer, size_t *lenp, loff_t *ppos);
2785int proc_nr_inodes(struct ctl_table *table, int write,
2786 void __user *buffer, size_t *lenp, loff_t *ppos);
2787int __init get_filesystem_list(char *buf);
2788
2789#define __FMODE_EXEC ((__force int) FMODE_EXEC)
2790#define __FMODE_NONOTIFY ((__force int) FMODE_NONOTIFY)
2791
2792#define ACC_MODE(x) ("\004\002\006\006"[(x)&O_ACCMODE])
2793#define OPEN_FMODE(flag) ((__force fmode_t)(((flag + 1) & O_ACCMODE) | \
2794 (flag & __FMODE_NONOTIFY)))
2795
2796static inline int is_sxid(umode_t mode)
2797{
2798 return (mode & S_ISUID) || ((mode & S_ISGID) && (mode & S_IXGRP));
2799}
2800
2801static inline int check_sticky(struct inode *dir, struct inode *inode)
2802{
2803 if (!(dir->i_mode & S_ISVTX))
2804 return 0;
2805
2806 return __check_sticky(dir, inode);
2807}
2808
2809static inline void inode_has_no_xattr(struct inode *inode)
2810{
2811 if (!is_sxid(inode->i_mode) && (inode->i_sb->s_flags & MS_NOSEC))
2812 inode->i_flags |= S_NOSEC;
2813}
2814
2815static inline bool is_root_inode(struct inode *inode)
2816{
2817 return inode == inode->i_sb->s_root->d_inode;
2818}
2819
2820static inline bool dir_emit(struct dir_context *ctx,
2821 const char *name, int namelen,
2822 u64 ino, unsigned type)
2823{
2824 return ctx->actor(ctx, name, namelen, ctx->pos, ino, type) == 0;
2825}
2826static inline bool dir_emit_dot(struct file *file, struct dir_context *ctx)
2827{
2828 return ctx->actor(ctx, ".", 1, ctx->pos,
2829 file->f_path.dentry->d_inode->i_ino, DT_DIR) == 0;
2830}
2831static inline bool dir_emit_dotdot(struct file *file, struct dir_context *ctx)
2832{
2833 return ctx->actor(ctx, "..", 2, ctx->pos,
2834 parent_ino(file->f_path.dentry), DT_DIR) == 0;
2835}
2836static inline bool dir_emit_dots(struct file *file, struct dir_context *ctx)
2837{
2838 if (ctx->pos == 0) {
2839 if (!dir_emit_dot(file, ctx))
2840 return false;
2841 ctx->pos = 1;
2842 }
2843 if (ctx->pos == 1) {
2844 if (!dir_emit_dotdot(file, ctx))
2845 return false;
2846 ctx->pos = 2;
2847 }
2848 return true;
2849}
2850static inline bool dir_relax(struct inode *inode)
2851{
2852 mutex_unlock(&inode->i_mutex);
2853 mutex_lock(&inode->i_mutex);
2854 return !IS_DEADDIR(inode);
2855}
2856
2857#endif /* _LINUX_FS_H */