4 #include <lunaix/clock.h>
5 #include <lunaix/device.h>
6 #include <lunaix/ds/btrie.h>
7 #include <lunaix/ds/hashtable.h>
8 #include <lunaix/ds/hstr.h>
9 #include <lunaix/ds/ldga.h>
10 #include <lunaix/ds/llist.h>
11 #include <lunaix/ds/lru.h>
12 #include <lunaix/ds/mutex.h>
13 #include <lunaix/status.h>
14 #include <lunaix/spike.h>
15 #include <lunaix/bcache.h>
16 #include <lunaix/fs_acl.h>
18 #include <usr/lunaix/fstypes.h>
20 #define VFS_NAME_MAXLEN 128
23 #define VFS_IFFILE F_FILE
24 #define VFS_IFDIR (F_FILE | F_DIR )
25 #define VFS_IFDEV (F_FILE | F_DEV )
26 #define VFS_IFSYMLINK (F_FILE | F_SYMLINK)
27 #define VFS_IFVOLDEV (F_FILE | F_SVDEV )
28 #define VFS_IFSEQDEV VFS_IFDEV
30 // Walk, mkdir if component encountered is non-exists.
31 #define VFS_WALK_MKPARENT 0x1
33 // Walk, relative to current FS.
34 #define VFS_WALK_FSRELATIVE 0x2
37 Terminate the walk on the immediate parent,
38 name of child (last component) is returned through `component`
40 #define VFS_WALK_PARENT 0x4
42 // Do not follow the symbolic link
43 #define VFS_WALK_NOFOLLOW 0x8
45 #define VFS_HASHTABLE_BITS 10
46 #define VFS_HASHTABLE_SIZE (1 << VFS_HASHTABLE_BITS)
47 #define VFS_HASH_MASK (VFS_HASHTABLE_SIZE - 1)
48 #define VFS_HASHBITS (32 - VFS_HASHTABLE_BITS)
50 #define VFS_PATH_DELIM '/'
52 #define FSTYPE_ROFS 0b00000001
53 #define FSTYPE_PSEUDO 0x00000010
55 #define TEST_FD(fd) (fd >= 0 && fd < VFS_MAX_FD)
57 #define EXPORT_FILE_SYSTEM(fs_id, init_fn) \
58 export_ldga_el(fs, fs_id, ptr_t, init_fn)
60 #define VFS_VALID_CHAR(chr) \
61 (('A' <= (chr) && (chr) <= 'Z') || ('a' <= (chr) && (chr) <= 'z') || \
62 ('0' <= (chr) && (chr) <= '9') || (chr) == '.' || (chr) == '_' || \
63 (chr) == '-' || (chr) == ':')
65 #define unlock_inode(inode) mutex_unlock(&inode->lock)
66 #define lock_inode(inode) \
68 mutex_lock(&inode->lock); \
69 lru_use_one(inode_lru, &inode->lru); \
72 #define unlock_dnode(dnode) mutex_unlock(&dnode->lock)
73 #define lock_dnode(dnode) \
75 mutex_lock(&dnode->lock); \
76 lru_use_one(dnode_lru, &dnode->lru); \
79 #define assert_fs(cond) assert_p(cond, "FS")
80 #define fail_fs(msg) fail_p(msg, "FS")
82 typedef u32_t inode_t;
94 extern struct v_file_ops default_file_ops;
95 extern struct v_inode_ops default_inode_ops;
97 extern struct hstr vfs_ddot;
98 extern struct hstr vfs_dot;
99 extern struct v_dnode* vfs_sysroot;
101 typedef int (*mntops_mnt)(struct v_superblock* vsb, struct v_dnode* mount_point);
102 typedef int (*mntops_umnt)(struct v_superblock* vsb);
106 struct llist_header fs_flat;
107 struct hlist_node fs_list;
117 struct filesystem* fs;
122 struct llist_header sb_list;
124 struct v_dnode* root;
125 struct filesystem* fs;
126 struct blkbuf_cache* blks;
127 struct hbucket* i_cache;
128 struct hbucket* d_cache;
131 unsigned int ref_count;
135 size_t (*read_capacity)(struct v_superblock* vsb);
136 size_t (*read_usage)(struct v_superblock* vsb);
137 void (*init_inode)(struct v_superblock* vsb, struct v_inode* inode);
138 void (*release)(struct v_superblock* vsb);
145 void (*read_complete_callback)(struct dir_context* dctx,
153 int (*write)(struct v_inode* inode, void* buffer, size_t len, size_t fpos);
154 int (*read)(struct v_inode* inode, void* buffer, size_t len, size_t fpos);
156 // for operatiosn {write|read}_page, following are true:
157 // + `len` always equals to PG_SIZE
158 // + `fpos` always PG_SIZE aligned.
159 // These additional operations allow underlying fs to use more specialized
160 // and optimized code.
162 int (*write_page)(struct v_inode* inode, void* pg, size_t fpos);
163 int (*read_page)(struct v_inode* inode, void* pg, size_t fpos);
165 int (*readdir)(struct v_file* file, struct dir_context* dctx);
166 int (*seek)(struct v_file* file, size_t offset);
167 int (*close)(struct v_file* file);
168 int (*sync)(struct v_file* file);
173 int (*create)(struct v_inode* this, struct v_dnode* dnode,
176 int (*open)(struct v_inode* this, struct v_file* file);
177 int (*sync)(struct v_inode* this);
179 int (*mkdir)(struct v_inode* this, struct v_dnode* dnode);
180 int (*rmdir)(struct v_inode* this, struct v_dnode* dnode);
181 int (*unlink)(struct v_inode* this, struct v_dnode* name);
182 int (*link)(struct v_inode* this, struct v_dnode* new_name);
184 int (*read_symlink)(struct v_inode* this, const char** path_out);
185 int (*set_symlink)(struct v_inode* this, const char* target);
187 int (*dir_lookup)(struct v_inode* this, struct v_dnode* dnode);
189 int (*rename)(struct v_inode* from_inode,
190 struct v_dnode* from_dnode,
191 struct v_dnode* to_dnode);
193 int (*getxattr)(struct v_inode* this,
194 struct v_xattr_entry* entry); // optional
195 int (*setxattr)(struct v_inode* this,
196 struct v_xattr_entry* entry); // optional
197 int (*delxattr)(struct v_inode* this,
198 struct v_xattr_entry* entry); // optional
203 struct llist_header entries;
211 struct v_inode* inode;
212 struct v_dnode* dnode;
213 struct llist_header* f_list;
215 unsigned long ref_count;
217 struct v_file_ops* ops; // for caching
226 // [v_inode::aka_nodes]
227 // how do we invalidate corresponding v_dnodes given the v_inode?
229 Consider taskfs, which is Lunaix's speak of Linux's procfs, that allow
230 info of every process being accessible via file system. Each process's
231 creation will result a creation of a directory under the root of task fs
232 with it's pid as name. But that dir must delete when process is killed, and
233 such deletion does not mediated by vfs itself, so there is a need of cache
235 And this is also the case of all ramfs where argumentation to file tree is
236 performed by third party.
257 void* data; // 允许底层FS绑定他的一些专有数据
258 struct llist_header aka_dnodes;
259 struct llist_header xattrs;
260 struct v_superblock* sb;
261 struct hlist_node hash_list;
263 struct pcache* pg_cache;
264 struct v_inode_ops* ops;
265 struct v_file_ops* default_fops;
267 void (*destruct)(struct v_inode* inode);
273 struct llist_header list;
274 struct llist_header submnts;
275 struct llist_header sibmnts;
276 struct v_mount* parent;
277 struct v_dnode* mnt_point;
278 struct v_superblock* super_block;
285 mutex_t lock; // sync the path walking
288 struct v_inode* inode;
289 struct v_dnode* parent;
290 struct hlist_node hash_list;
291 struct llist_header aka_list;
292 struct llist_header children;
293 struct llist_header siblings;
295 struct v_superblock* super_block;
298 unsigned long ref_count;
302 void (*destruct)(struct v_dnode* dnode);
307 struct v_fd* fds[VFS_MAX_FD];
312 struct v_inode* master;
314 struct llist_header dirty;
321 struct llist_header dirty_list;
335 check_itype_any(struct v_inode* inode, unsigned int type_mask)
337 return !!(inode->itype & type_mask) || !type_mask;
347 fsm_new_fs(char* name, size_t name_len);
350 fsm_register(struct filesystem* fs);
353 fsm_get(const char* fs_name);
356 fsm_itbegin(struct fs_iter* iterator);
359 fsm_itnext(struct fs_iter* iterator);
362 fsm_itend(struct fs_iter* iterator)
371 vfs_export_attributes();
374 vfs_dcache_lookup(struct v_dnode* parent, struct hstr* str);
377 vfs_dcache_add(struct v_dnode* parent, struct v_dnode* dnode);
380 vfs_dcache_rehash(struct v_dnode* new_parent, struct v_dnode* dnode);
383 vfs_dcache_remove(struct v_dnode* dnode);
386 vfs_walk(struct v_dnode* start,
388 struct v_dnode** dentry,
389 struct hstr* component,
393 vfs_walk_proc(const char* path,
394 struct v_dnode** dentry,
395 struct hstr* component,
399 vfs_mount(const char* target,
401 struct device* device,
405 vfs_unmount(const char* target);
408 vfs_mount_at(const char* fs_name,
409 struct device* device,
410 struct v_dnode* mnt_point,
414 vfs_unmount_at(struct v_dnode* mnt_point);
417 vfs_mkdir(const char* path, struct v_dnode** dentry);
420 vfs_open(struct v_dnode* dnode, struct v_file** file);
423 vfs_pclose(struct v_file* file, pid_t pid);
426 vfs_close(struct v_file* file);
429 vfs_free_fd(struct v_fd* fd);
432 vfs_fsync(struct v_file* file);
435 vfs_assign_inode(struct v_dnode* assign_to, struct v_inode* inode);
441 vfs_sb_unref(struct v_superblock* sb);
444 vfs_sb_ref(struct v_superblock* sb);
446 #define vfs_assign_sb(sb_accessor, sb) \
448 if (likely(sb_accessor)) { \
449 vfs_sb_unref(sb_accessor); \
451 vfs_sb_ref(((sb_accessor) = (sb))); \
455 vfs_i_assign_sb(struct v_inode* inode, struct v_superblock* sb)
457 vfs_assign_sb(inode->sb, sb);
461 vfs_d_assign_sb(struct v_dnode* dnode, struct v_superblock* sb)
463 vfs_assign_sb(dnode->super_block, sb);
467 vfs_d_assign_vmnt(struct v_dnode* dnode, struct v_mount* vmnt)
470 assert_msg(dnode->mnt->mnt_point != dnode,
471 "vmnt must be detached first");
477 vfs_d_assign_sb(dnode, vmnt->super_block);
481 vfs_vmnt_assign_sb(struct v_mount* vmnt, struct v_superblock* sb)
483 vfs_assign_sb(vmnt->super_block, sb);
487 vfs_d_alloc(struct v_dnode* parent, struct hstr* name);
490 vfs_d_free(struct v_dnode* dnode);
493 vfs_i_find(struct v_superblock* sb, u32_t i_id);
496 vfs_i_addhash(struct v_inode* inode);
499 vfs_i_alloc(struct v_superblock* sb);
502 vfs_i_free(struct v_inode* inode);
505 vfs_dup_fd(struct v_fd* old, struct v_fd** new);
508 vfs_getfd(int fd, struct v_fd** fd_s);
511 vfs_get_dtype(int itype);
514 vfs_get_path(struct v_dnode* dnode, char* buf, size_t size, int depth);
517 pcache_init(struct pcache* pcache);
520 pcache_write(struct v_inode* inode, void* data, u32_t len, u32_t fpos);
523 pcache_read(struct v_inode* inode, void* data, u32_t len, u32_t fpos);
526 pcache_release(struct pcache* pcache);
529 pcache_commit(struct v_inode* inode, struct pcache_pg* page);
532 pcache_commit_all(struct v_inode* inode);
535 pcache_invalidate(struct pcache* pcache, struct pcache_pg* page);
543 mnt_mkbusy(struct v_mount* mnt);
551 mnt_chillax(struct v_mount* mnt);
554 vfs_mount_root(const char* fs_name, struct device* device);
557 mnt_check_busy(struct v_mount* mnt)
559 return mnt->busy_counter > 1;
563 vfs_ref_dnode(struct v_dnode* dnode)
567 if (likely(dnode->mnt)) {
568 mnt_mkbusy(dnode->mnt);
573 vfs_unref_dnode(struct v_dnode* dnode)
577 if (likely(dnode->mnt)) {
578 mnt_chillax(dnode->mnt);
583 vfs_ref_file(struct v_file* file)
589 vfs_unref_file(struct v_file* file)
595 vfs_check_duped_file(struct v_file* file)
597 return file->ref_count > 1;
601 vfs_check_writable(struct v_dnode* dnode);
604 default_file_read(struct v_inode* inode, void* buffer, size_t len, size_t fpos);
607 default_file_write(struct v_inode* inode,
613 default_file_read_page(struct v_inode* inode, void* buffer, size_t fpos);
616 default_file_write_page(struct v_inode* inode, void* buffer, size_t fpos);
619 default_file_readdir(struct v_file* file, struct dir_context* dctx);
622 default_inode_dirlookup(struct v_inode* this, struct v_dnode* dnode);
625 default_inode_rename(struct v_inode* from_inode,
626 struct v_dnode* from_dnode,
627 struct v_dnode* to_dnode);
630 default_file_close(struct v_file* file);
633 default_file_seek(struct v_file* file, size_t offset);
636 default_inode_open(struct v_inode* this, struct v_file* file);
639 default_inode_rmdir(struct v_inode* this, struct v_dnode* dir);
642 default_inode_mkdir(struct v_inode* this, struct v_dnode* dir);
644 struct v_xattr_entry*
645 xattr_new(struct hstr* name);
647 struct v_xattr_entry*
648 xattr_getcache(struct v_inode* inode, struct hstr* name);
651 xattr_addcache(struct v_inode* inode, struct v_xattr_entry* xattr);
654 /* --- misc stuff --- */
656 #define check_itype(to_check, itype) \
657 (((to_check) & (itype)) == (itype))
660 * @brief Check if node represent a regular file (nothing but a file)
667 check_regfile_node(struct v_inode* inode)
669 return inode->itype == VFS_IFFILE;
673 * @brief Check if node represent a file.
674 * This is basically everything within file system (dir, dev, etc.)
681 check_file_node(struct v_inode* inode)
683 return check_itype(inode->itype, VFS_IFFILE);
687 check_directory_node(struct v_inode* inode)
689 return check_itype(inode->itype, VFS_IFDIR);
693 check_device_node(struct v_inode* inode)
695 return check_itype(inode->itype, VFS_IFDEV);
699 check_seqdev_node(struct v_inode* inode)
701 return check_device_node(inode);
705 check_voldev_node(struct v_inode* inode)
707 return check_itype(inode->itype, VFS_IFVOLDEV);
711 check_symlink_node(struct v_inode* inode)
713 return check_itype(inode->itype, VFS_IFSYMLINK);
717 check_allow_ops(struct v_inode* inode, unsigned int perm)
719 return fsacl_allow_ops(perm, inode->acl, inode->uid, inode->gid);
723 check_allow_read(struct v_inode* inode)
725 return check_allow_ops(inode, FSACL_aR);
729 check_allow_write(struct v_inode* inode)
731 return check_allow_ops(inode, FSACL_aW);
735 check_allow_execute(struct v_inode* inode)
737 return check_allow_ops(inode, FSACL_aX);
740 #endif /* __LUNAIX_VFS_H */