1 //
2 // linux/fs.h
3 // linux文件系统最外围结构之一，超级块
4 //
5 struct super_block {
6 //
7 // 双向链表，所有文件系统的超级快通过这个链接在一起
8 //
9 struct list_head s_list; /* Keep this first */
10 //
11 // 设备???目前不明
12 //
13 dev_t s_dev; /* search index; _not_ kdev_t */
14 unsigned char s_blocksize_bits;
15 unsigned long s_blocksize;
16 //
17 //该文件系统最大文件大小
18 //
19 loff_t s_maxbytes; /* Max file size */
20 //
21 // 文件系统指针，在一个文件系统驱动注册的时候填写好这个
22 // 结构体，通过 register_filesystem(); 注册进来!!!!!
23 //
24 struct file_system_type *s_type;
25 //
26 // 最重要的结构之一，超级块的操作
27 //
28 const struct super_operations *s_op;
29 const struct dquot_operations *dq_op;
30 const struct quotactl_ops *s_qcop;
31 const struct export_operations *s_export_op;
32 unsigned long s_flags;
33 unsigned long s_iflags; /* internal SB_I_* flags */
34 //
35 // 超级块的魔数
36 //
37 unsigned long s_magic;
38 //
39 // 文件系统挂载的目录项
40 //
41 struct dentry *s_root;
42 struct rw_semaphore s_umount;
43 int s_count;
44 atomic_t s_active;
45 #ifdef CONFIG_SECURITY
46 void *s_security;
47 #endif
48 const struct xattr_handler **s_xattr;
49 #ifdef CONFIG_FS_ENCRYPTION
50 const struct fscrypt_operations *s_cop;
51 #endif
52 struct hlist_bl_head s_roots; /* alternate root dentries for NFS */
53 struct list_head s_mounts; /* list of mounts; _not_ for fs use */
54 struct block_device *s_bdev;
55 struct backing_dev_info *s_bdi;
56 struct mtd_info *s_mtd;
57 struct hlist_node s_instances;
58 unsigned int s_quota_types; /* Bitmask of supported quota types */
59 struct quota_info s_dquot; /* Diskquota specific options */
60
61 struct sb_writers s_writers;
62
63 /*
64 * Keep s_fs_info, s_time_gran, s_fsnotify_mask, and
65 * s_fsnotify_marks together for cache efficiency. They are frequently
66 * accessed and rarely modified.
67 */
68 void *s_fs_info; /* Filesystem private info */
69
70 /* Granularity of c/m/atime in ns (cannot be worse than a second) */
71 u32 s_time_gran;
72 #ifdef CONFIG_FSNOTIFY
73 __u32 s_fsnotify_mask;
74 struct fsnotify_mark_connector __rcu *s_fsnotify_marks;
75 #endif
76
77 char s_id[32]; /* Informational name */
78 uuid_t s_uuid; /* UUID */
79
80 unsigned int s_max_links;
81 fmode_t s_mode;
82
83 /*
84 * The next field is for VFS *only*. No filesystems have any business
85 * even looking at it. You had been warned.
86 */
87 struct mutex s_vfs_rename_mutex; /* Kludge */
88
89 /*
90 * Filesystem subtype. If non-empty the filesystem type field
91 * in /proc/mounts will be "type.subtype"
92 */
93 const char *s_subtype;
94 //
95 // 目录项的操作方法
96 //
97 const struct dentry_operations *s_d_op; /* default d_op for dentries */
98
99 /*
100 * Saved pool identifier for cleancache (-1 means none)
101 */
102 int cleancache_poolid;
103
104 struct shrinker s_shrink; /* per-sb shrinker handle */
105
106 /* Number of inodes with nlink == 0 but still referenced */
107 atomic_long_t s_remove_count;
108
109 /* Pending fsnotify inode refs */
110 atomic_long_t s_fsnotify_inode_refs;
111
112 /* Being remounted read-only */
113 int s_readonly_remount;
114
115 /* AIO completions deferred from interrupt context */
116 struct workqueue_struct *s_dio_done_wq;
117 struct hlist_head s_pins;
118
119 /*
120 * Owning user namespace and default context in which to
121 * interpret filesystem uids, gids, quotas, device nodes,
122 * xattrs and security labels.
123 */
124 struct user_namespace *s_user_ns;
125
126 /*
127 * The list_lru structure is essentially just a pointer to a table
128 * of per-node lru lists, each of which has its own spinlock.
129 * There is no need to put them into separate cachelines.
130 */
131 //
132 // 应该是未使用的dentry和inode链表
133 //
134 struct list_lru s_dentry_lru;
135 struct list_lru s_inode_lru;
136 struct rcu_head rcu;
137 struct work_struct destroy_work;
138
139 struct mutex s_sync_lock; /* sync serialisation lock */
140
141 /*
142 * Indicates how deep in a filesystem stack this SB is
143 */
144 int s_stack_depth;
145
146 /* s_inode_list_lock protects s_inodes */
147 spinlock_t s_inode_list_lock ____cacheline_aligned_in_smp;
148 struct list_head s_inodes; /* all inodes */
149
150 spinlock_t s_inode_wblist_lock;
151 struct list_head s_inodes_wb; /* writeback inodes */
152 } __randomize_layout;
153
154
155 //
156 // 对应 super_block->s_op
157 //
158 struct super_operations {
159 //
160 // 初始化新的索引节点, inode
161 //
162 struct inode *(*alloc_inode)(struct super_block *sb);
163 void (*destroy_inode)(struct inode *);
164 void (*free_inode)(struct inode *);
165
166 void (*dirty_inode) (struct inode *, int flags);
167 int (*write_inode) (struct inode *, struct writeback_control *wbc);
168 int (*drop_inode) (struct inode *);
169 void (*evict_inode) (struct inode *);
170 void (*put_super) (struct super_block *);
171 int (*sync_fs)(struct super_block *sb, int wait);
172 int (*freeze_super) (struct super_block *);
173 int (*freeze_fs) (struct super_block *);
174 int (*thaw_super) (struct super_block *);
175 int (*unfreeze_fs) (struct super_block *);
176 int (*statfs) (struct dentry *, struct kstatfs *);
177 int (*remount_fs) (struct super_block *, int *, char *);
178 void (*umount_begin) (struct super_block *);
179
180 int (*show_options)(struct seq_file *, struct dentry *);
181 int (*show_devname)(struct seq_file *, struct dentry *);
182 int (*show_path)(struct seq_file *, struct dentry *);
183 int (*show_stats)(struct seq_file *, struct dentry *);
184 #ifdef CONFIG_QUOTA
185 ssize_t (*quota_read)(struct super_block *, int, char *, size_t, loff_t);
186 ssize_t (*quota_write)(struct super_block *, int, const char *, size_t, loff_t);
187 struct dquot **(*get_dquots)(struct inode *);
188 #endif
189 int (*bdev_try_to_free_page)(struct super_block*, struct page*, gfp_t);
190 long (*nr_cached_objects)(struct super_block *,
191 struct shrink_control *);
192 long (*free_cached_objects)(struct super_block *,
193 struct shrink_control *);
194 };
195
196
197 /*
198 * Keep mostly read-only and often accessed (especially for
199 * the RCU path lookup and 'stat' data) fields at the beginning
200 * of the 'struct inode'
201 */
202
203 //
204 // 文件系统inode节点
205 //
206
207 struct inode {
208 umode_t i_mode;
209 unsigned short i_opflags;
210 kuid_t i_uid;
211 kgid_t i_gid;
212 unsigned int i_flags;
213
214 //
215 // 当内核配置了POSIX ACL后开启，一般开启
216 //
217 #ifdef CONFIG_FS_POSIX_ACL
218 struct posix_acl *i_acl;
219 struct posix_acl *i_default_acl;
220 #endif
221 //
222 // inode操作
223 //
224 const struct inode_operations *i_op;
225 //
226 // 指向超级块
227 //
228 struct super_block *i_sb;
229 //
230 // 指向address_space
231 //
232 struct address_space *i_mapping;
233
234 #ifdef CONFIG_SECURITY
235 void *i_security;
236 #endif
237 //
238 // inode number
239 //
240 /* Stat data, not accessed from path walking */
241 unsigned long i_ino;
242 /*
243 * Filesystems may only read i_nlink directly. They shall use the
244 * following functions for modification:
245 *
246 * (set|clear|inc|drop)_nlink
247 * inode_(inc|dec)_link_count
248 */
249 union {
250 const unsigned int i_nlink;
251 unsigned int __i_nlink;
252 };
253 dev_t i_rdev;
254 //
255 // 文件大小，多少个字节
256 //
257 loff_t i_size;
258 // i_atime access time
259 // i_mtime modify time
260 // i_ctime create time
261 struct timespec64 i_atime;
262 struct timespec64 i_mtime;
263 struct timespec64 i_ctime;
264 spinlock_t i_lock; /* i_blocks, i_bytes, maybe i_size */
265 unsigned short i_bytes;
266 u8 i_blkbits;
267 u8 i_write_hint;
268 blkcnt_t i_blocks;
269
270 #ifdef __NEED_I_SIZE_ORDERED
271 seqcount_t i_size_seqcount;
272 #endif
273
274 /* Misc */
275 unsigned long i_state;
276 struct rw_semaphore i_rwsem;
277
278 unsigned long dirtied_when; /* jiffies of first dirtying */
279 unsigned long dirtied_time_when;
280
281 struct hlist_node i_hash;
282 struct list_head i_io_list; /* backing dev IO list */
283 #ifdef CONFIG_CGROUP_WRITEBACK
284 struct bdi_writeback *i_wb; /* the associated cgroup wb */
285
286 /* foreign inode detection, see wbc_detach_inode() */
287 int i_wb_frn_winner;
288 u16 i_wb_frn_avg_time;
289 u16 i_wb_frn_history;
290 #endif
291 struct list_head i_lru; /* inode LRU list */
292 struct list_head i_sb_list;
293 struct list_head i_wb_list; /* backing dev writeback list */
294 union {
295 struct hlist_head i_dentry;
296 struct rcu_head i_rcu;
297 };
298 atomic64_t i_version;
299 atomic_t i_count;
300 atomic_t i_dio_count;
301 atomic_t i_writecount;
302 #if defined(CONFIG_IMA) || defined(CONFIG_FILE_LOCKING)
303 atomic_t i_readcount; /* struct files open RO */
304 #endif
305 union {
306 const struct file_operations *i_fop; /* former ->i_op->default_file_ops */
307 void (*free_inode)(struct inode *);
308 };
309 struct file_lock_context *i_flctx;
310 struct address_space i_data;
311 struct list_head i_devices;
312 union {
313 struct pipe_inode_info *i_pipe;
314 struct block_device *i_bdev;
315 struct cdev *i_cdev;
316 char *i_link;
317 unsigned i_dir_seq;
318 };
319
320 __u32 i_generation;
321
322 #ifdef CONFIG_FSNOTIFY
323 __u32 i_fsnotify_mask; /* all events this inode cares about */
324 struct fsnotify_mark_connector __rcu *i_fsnotify_marks;
325 #endif
326
327 #ifdef CONFIG_FS_ENCRYPTION
328 struct fscrypt_info *i_crypt_info;
329 #endif
330 //
331 // 私有结构
332 //
333 void *i_private; /* fs or device private pointer */
334 } __randomize_layout;
335
336 //
337 // inode 节点操作方法
338 //
339 struct inode_operations {
340 //查找
341 struct dentry * (*lookup) (struct inode *,struct dentry *, unsigned int);
342 const char * (*get_link) (struct dentry *, struct inode *, struct delayed_call *);
343 int (*permission) (struct inode *, int);
344 struct posix_acl * (*get_acl)(struct inode *, int);
345
346 int (*readlink) (struct dentry *, char __user *,int);
347 //
348 // 创建一个新的索引节点
349 //
350 int (*create) (struct inode *,struct dentry *, umode_t, bool);
351 //
352 // 创建硬链接
353 //
354 int (*link) (struct dentry *,struct inode *,struct dentry *);
355 //
356 // 取消硬链接
357 //
358 int (*unlink) (struct inode *,struct dentry *);
359 //
360 // 创建符号链接
361 //
362 int (*symlink) (struct inode *,struct dentry *,const char *);
363 //
364 // 创建目录
365 //
366 int (*mkdir) (struct inode *,struct dentry *,umode_t);
367 //
368 // 删除目录
369 //
370 int (*rmdir) (struct inode *,struct dentry *);
371 //
372 // 创建块设备节点
373 //
374 int (*mknod) (struct inode *,struct dentry *,umode_t,dev_t);
375 //
376 // 重命名
377 //
378 int (*rename) (struct inode *, struct dentry *,
379 struct inode *, struct dentry *, unsigned int);
380 //
381 // 设置属性
382 //
383 int (*setattr) (struct dentry *, struct iattr *);
384 //
385 // 得到属性
386 //
387 int (*getattr) (const struct path *, struct kstat *, u32, unsigned int);
388 ssize_t (*listxattr) (struct dentry *, char *, size_t);
389 //
390 // 从函数名来看是文件映射
391 //
392 int (*fiemap)(struct inode *, struct fiemap_extent_info *, u64 start,
393 u64 len);
394 int (*update_time)(struct inode *, struct timespec64 *, int);
395 int (*atomic_open)(struct inode *, struct dentry *,
396 struct file *, unsigned open_flag,
397 umode_t create_mode);
398 int (*tmpfile) (struct inode *, struct dentry *, umode_t);
399 int (*set_acl)(struct inode *, struct posix_acl *, int);
400 } ____cacheline_aligned;
401
402 //
403 // dentry --> directory entry 目录项
404 // 一般结构是: file ==> dentry ==> inode
405 // 对于进程来说: task_struct 有一个结构体叫
406 // task_struct -> files_struct -> fdtable->file
407 //
408 //
409 //
410 //
411
412
413 //
414 // dentry 定义在dcache.h
415 //
416 struct dentry {
417 /* RCU lookup touched fields */
418 unsigned int d_flags; /* protected by d_lock */
419 seqcount_t d_seq; /* per dentry seqlock */
420 //
421 // 查找hash表
422 //
423 struct hlist_bl_node d_hash; /* lookup hash list */
424 //
425 // 父目录项
426 //
427 struct dentry *d_parent; /* parent directory */
428 struct qstr d_name;
429 //
430 // dentry 对应inode节点
431 //
432 struct inode *d_inode; /* Where the name belongs to - NULL is
433 * negative */
434 unsigned char d_iname[DNAME_INLINE_LEN]; /* small names */
435
436 /* Ref lookup also touches following */
437 struct lockref d_lockref; /* per-dentry lock and refcount */
438 const struct dentry_operations *d_op;
439 struct super_block *d_sb; /* The root of the dentry tree */
440 unsigned long d_time; /* used by d_revalidate */
441 void *d_fsdata; /* fs-specific data */
442
443 union {
444 struct list_head d_lru; /* LRU list */
445 wait_queue_head_t *d_wait; /* in-lookup ones only */
446 };
447 //
448 // 父目录目录项链表
449 //
450 struct list_head d_child; /* child of parent list */
451 //
452 // 自己子目录链表
453 //
454 struct list_head d_subdirs; /* our children */
455 /*
456 * d_alias and d_rcu can share memory
457 */
458 union {
459 struct hlist_node d_alias; /* inode alias list */
460 struct hlist_bl_node d_in_lookup_hash; /* only for in-lookup ones */
461 struct rcu_head d_rcu;
462 } d_u;
463 } __randomize_layout;
464
465
466 struct dentry_operations {
467 //
468 // 判断目录项是否实效
469 //
470 int (*d_revalidate)(struct dentry *, unsigned int);
471 //
472 //
473 //
474 int (*d_weak_revalidate)(struct dentry *, unsigned int);
475 //
476 // 生成hash,VFS插入目录项时会计算hash
477 //
478 int (*d_hash)(const struct dentry *, struct qstr *);
479 //
480 // 比较字符串
481 //
482 int (*d_compare)(const struct dentry *,
483 unsigned int, const char *, const struct qstr *);
484 //
485 // 当d_count 为零是调用
486 //
487 int (*d_delete)(const struct dentry *);
488 int (*d_init)(struct dentry *);
489 //
490 // 文件被释放时被调用
491 //
492 void (*d_release)(struct dentry *);
493 void (*d_prune)(struct dentry *);
494 //
495 // 当目录项丢失inode节点时候调用
496 //
497 void (*d_iput)(struct dentry *, struct inode *);
498 char *(*d_dname)(struct dentry *, char *, int);
499 struct vfsmount *(*d_automount)(struct path *);
500 int (*d_manage)(const struct path *, bool);
501 struct dentry *(*d_real)(struct dentry *, const struct inode *);
502 } ____cacheline_aligned;
503
504 //
505 // path
506 //
507 struct path {
508 struct vfsmount *mnt;
509 struct dentry *dentry;
510 } __randomize_layout;
511
512 struct vfsmount {
513 struct dentry *mnt_root; /* root of the mounted tree */
514 struct super_block *mnt_sb; /* pointer to superblock */
515 int mnt_flags;
516 } __randomize_layout;
517
518 //
519 // struct file, 文件对象
520 //
521 struct file {
522 union {
523 //
524 // 文件对象链表
525 //
526 struct llist_node fu_llist;
527 struct rcu_head fu_rcuhead;
528 } f_u;
529 //
530 // path struct path {
531 // struct vfsmount *mnt; // vfs挂载点
532 // struct dentry *dentry; // 对应目录项
533 //} __randomize_layout;
534
535 //
536 struct path f_path;
537 struct inode *f_inode; /* cached value */
538 //
539 // 文件操作函数
540 //
541 const struct file_operations *f_op;
542
543 /*
544 * Protects f_ep_links, f_flags.
545 * Must not be taken from IRQ context.
546 */
547 spinlock_t f_lock;
548 enum rw_hint f_write_hint;
549 //
550 // 文件被多少进程使用
551 //
552 atomic_long_t f_count;
553 unsigned int f_flags;
554 fmode_t f_mode;
555 struct mutex f_pos_lock;
556 //
557 // 文件指针位置
558 //
559 loff_t f_pos;
560 struct fown_struct f_owner;
561 const struct cred *f_cred;
562 struct file_ra_state f_ra;
563
564 u64 f_version;
565 #ifdef CONFIG_SECURITY
566 void *f_security;
567 #endif
568 /* needed for tty driver, and maybe others */
569 void *private_data;
570
571 #ifdef CONFIG_EPOLL
572 /* Used by fs/eventpoll.c to link all the hooks to this file */
573 struct list_head f_ep_links;
574 struct list_head f_tfile_llink;
575 #endif /* #ifdef CONFIG_EPOLL */
576 //
577 // address_space
578 //
579 struct address_space *f_mapping;
580 errseq_t f_wb_err;
581 } __randomize_layout
582 __attribute__((aligned(4))); /* lest something weird decides that 2 is OK */
583
584 struct file_handle {
585 __u32 handle_bytes;
586 int handle_type;
587 /* file identifier */
588 unsigned char f_handle[0];
589 };
590
591 //
592 //
593 //
594 struct file_operations {
595 struct module *owner;
596 loff_t (*llseek) (struct file *, loff_t, int);
597 ssize_t (*read) (struct file *, char __user *, size_t, loff_t *);
598 ssize_t (*write) (struct file *, const char __user *, size_t, loff_t *);
599 ssize_t (*read_iter) (struct kiocb *, struct iov_iter *);
600 ssize_t (*write_iter) (struct kiocb *, struct iov_iter *);
601 int (*iopoll)(struct kiocb *kiocb, bool spin);
602 int (*iterate) (struct file *, struct dir_context *);
603 int (*iterate_shared) (struct file *, struct dir_context *);
604 __poll_t (*poll) (struct file *, struct poll_table_struct *);
605 long (*unlocked_ioctl) (struct file *, unsigned int, unsigned long);
606 long (*compat_ioctl) (struct file *, unsigned int, unsigned long);
607 int (*mmap) (struct file *, struct vm_area_struct *);
608 unsigned long mmap_supported_flags;
609 int (*open) (struct inode *, struct file *);
610 int (*flush) (struct file *, fl_owner_t id);
611 int (*release) (struct inode *, struct file *);
612 int (*fsync) (struct file *, loff_t, loff_t, int datasync);
613 int (*fasync) (int, struct file *, int);
614 int (*lock) (struct file *, int, struct file_lock *);
615 ssize_t (*sendpage) (struct file *, struct page *, int, size_t, loff_t *, int);
616 unsigned long (*get_unmapped_area)(struct file *, unsigned long, unsigned long, unsigned long, unsigned long);
617 int (*check_flags)(int);
618 int (*flock) (struct file *, int, struct file_lock *);
619 ssize_t (*splice_write)(struct pipe_inode_info *, struct file *, loff_t *, size_t, unsigned int);
620 ssize_t (*splice_read)(struct file *, loff_t *, struct pipe_inode_info *, size_t, unsigned int);
621 int (*setlease)(struct file *, long, struct file_lock **, void **);
622 long (*fallocate)(struct file *file, int mode, loff_t offset,
623 loff_t len);
624 void (*show_fdinfo)(struct seq_file *m, struct file *f);
625 #ifndef CONFIG_MMU
626 unsigned (*mmap_capabilities)(struct file *);
627 #endif
628 ssize_t (*copy_file_range)(struct file *, loff_t, struct file *,
629 loff_t, size_t, unsigned int);
630 loff_t (*remap_file_range)(struct file *file_in, loff_t pos_in,
631 struct file *file_out, loff_t pos_out,
632 loff_t len, unsigned int remap_flags);
633 int (*fadvise)(struct file *, loff_t, loff_t, int);
634 } __randomize_layout;
635
636 //
637 // 文件系统类型
638 //
639 struct file_system_type {
640 const char *name;
641 int fs_flags;
642 #define FS_REQUIRES_DEV 1
643 #define FS_BINARY_MOUNTDATA 2
644 #define FS_HAS_SUBTYPE 4
645 #define FS_USERNS_MOUNT 8 /* Can be mounted by userns root */
646 #define FS_DISALLOW_NOTIFY_PERM 16 /* Disable fanotify permission events */
647 #define FS_RENAME_DOES_D_MOVE 32768 /* FS will handle d_move() during rename() internally. */
648 int (*init_fs_context)(struct fs_context *);
649 const struct fs_parameter_description *parameters;
650 struct dentry *(*mount) (struct file_system_type *, int,
651 const char *, void *);
652 void (*kill_sb) (struct super_block *);
653 struct module *owner;
654 struct file_system_type * next;
655 struct hlist_head fs_supers;
656
657 struct lock_class_key s_lock_key;
658 struct lock_class_key s_umount_key;
659 struct lock_class_key s_vfs_rename_key;
660 struct lock_class_key s_writers_key[SB_FREEZE_LEVELS];
661
662 struct lock_class_key i_lock_key;
663 struct lock_class_key i_mutex_key;
664 struct lock_class_key i_mutex_dir_key;
665 };
666 //
667 // 挂载点
668 //
669 struct vfsmount {
670 struct dentry *mnt_root; /* root of the mounted tree */
671 struct super_block *mnt_sb; /* pointer to superblock */
672 int mnt_flags;
673 } __randomize_layout;
