3 * @author Lunaixsky (zelong56@gmail.com)
4 * @brief Lunaix virtual file system - an abstraction layer for all file system.
8 * @copyright Copyright (c) 2022
12 // Welcome to The Mountain O'Shit! :)
15 TODO vfs & device todos checklist
17 It is overseen by Twilight Sparkle ;)
19 1. Get inodes hooked into lru (CHECKED)
20 2. Get dnodes hooked into lru (CHECKED)
21 3. Get inodes properly hashed so they can be reused by underling fs (CHECKED)
22 4. (lru) Add a callback function (or destructor) for eviction. (CHECKED)
23 [good idea] or a constructor/destructor pattern in cake allocator ?
24 5. (mount) Figure out a way to identify a busy mount point before unmount
25 maybe a unified mount_point structure that maintain a referencing
26 counter on any dnodes within the subtree? Such a counter will only
27 increament if a file is opened or a dnode is being used as working
28 directory and decreamenting conversely. (CHECKED)
29 6. (mount) Ability to track all mount points (including sub-mounts)
30 so we can be confident to clean up everything when we
32 7. (mount) Figure out a way to acquire the device represented by a dnode.
33 so it can be used to mount. (e.g. we wish to get `struct device*`
34 out of the dnode at /dev/sda)
35 [tip] we should pay attention at twifs and add a private_data field
36 under struct v_dnode? (CHECKED)
37 8. (mount) Then, we should refactor on mount/unmount mechanism. (CHECKED)
38 9. (mount) (future) Ability to mount any thing? e.g. Linux can mount a disk
39 image file using a so called "loopback" pseudo device. Maybe
40 we can do similar thing in Lunaix? A block device emulation
41 above the regular file when we mount it on.
42 10. (device) device number (dev_t) allocation
43 [good idea] <class>:<subclass>:<uniq_id> composition (CHECKED)
46 #include <klibc/string.h>
47 #include <lunaix/foptions.h>
48 #include <lunaix/fs.h>
49 #include <lunaix/mm/cake.h>
50 #include <lunaix/mm/page.h>
51 #include <lunaix/mm/valloc.h>
52 #include <lunaix/process.h>
53 #include <lunaix/spike.h>
54 #include <lunaix/syscall.h>
55 #include <lunaix/syscall_utils.h>
57 #include <lunaix/fs/twifs.h>
59 #include <usr/lunaix/dirent_defs.h>
61 static struct cake_pile* dnode_pile;
62 static struct cake_pile* inode_pile;
63 static struct cake_pile* file_pile;
64 static struct cake_pile* superblock_pile;
65 static struct cake_pile* fd_pile;
67 struct v_dnode* vfs_sysroot;
68 static struct hbucket* dnode_cache;
70 struct lru_zone *dnode_lru, *inode_lru;
72 struct hstr vfs_ddot = HSTR("..", 2);
73 struct hstr vfs_dot = HSTR(".", 1);
74 struct hstr vfs_empty = HSTR("", 0);
80 vfs_sb_free(struct v_superblock* sb);
83 __vfs_try_evict_dnode(struct lru_node* obj);
86 __vfs_try_evict_inode(struct lru_node* obj);
91 // 为他们专门创建一个蛋糕堆,而不使用valloc,这样我们可以最小化内碎片的产生
92 dnode_pile = cake_new_pile("dnode_cache", sizeof(struct v_dnode), 1, 0);
93 inode_pile = cake_new_pile("inode_cache", sizeof(struct v_inode), 1, 0);
94 file_pile = cake_new_pile("file_cache", sizeof(struct v_file), 1, 0);
95 fd_pile = cake_new_pile("fd_cache", sizeof(struct v_fd), 1, 0);
97 cake_new_pile("sb_cache", sizeof(struct v_superblock), 1, 0);
99 dnode_cache = vzalloc(VFS_HASHTABLE_SIZE * sizeof(struct hbucket));
101 dnode_lru = lru_new_zone(__vfs_try_evict_dnode);
102 inode_lru = lru_new_zone(__vfs_try_evict_inode);
104 hstr_rehash(&vfs_ddot, HSTR_FULL_HASH);
105 hstr_rehash(&vfs_dot, HSTR_FULL_HASH);
108 vfs_sysroot = vfs_d_alloc(NULL, &vfs_empty);
109 vfs_sysroot->parent = vfs_sysroot;
110 atomic_fetch_add(&vfs_sysroot->ref_count, 1);
113 static inline struct hbucket*
114 __dcache_hash(struct v_dnode* parent, u32_t* hash)
118 _hash = _hash ^ (_hash >> VFS_HASHBITS);
119 // 与parent的指针值做加法,来减小碰撞的可能性。
120 _hash += (u32_t)parent;
122 return &dnode_cache[_hash & VFS_HASH_MASK];
126 vfs_dcache_lookup(struct v_dnode* parent, struct hstr* str)
128 if (!str->len || HSTR_EQ(str, &vfs_dot))
131 if (HSTR_EQ(str, &vfs_ddot)) {
132 return parent->parent;
135 u32_t hash = str->hash;
136 struct hbucket* slot = __dcache_hash(parent, &hash);
138 struct v_dnode *pos, *n;
139 hashtable_bucket_foreach(slot, pos, n, hash_list)
141 if (pos->name.hash == hash) {
149 vfs_dcache_add(struct v_dnode* parent, struct v_dnode* dnode)
153 atomic_fetch_add(&dnode->ref_count, 1);
154 dnode->parent = parent;
155 llist_append(&parent->children, &dnode->siblings);
157 struct hbucket* bucket = __dcache_hash(parent, &dnode->name.hash);
158 hlist_add(&bucket->head, &dnode->hash_list);
162 vfs_dcache_remove(struct v_dnode* dnode)
165 assert(dnode->ref_count == 1);
167 llist_delete(&dnode->siblings);
168 llist_delete(&dnode->aka_list);
169 hlist_delete(&dnode->hash_list);
171 dnode->parent = NULL;
172 atomic_fetch_sub(&dnode->ref_count, 1);
176 vfs_dcache_rehash(struct v_dnode* new_parent, struct v_dnode* dnode)
180 hstr_rehash(&dnode->name, HSTR_FULL_HASH);
181 vfs_dcache_remove(dnode);
182 vfs_dcache_add(new_parent, dnode);
186 vfs_open(struct v_dnode* dnode, struct v_file** file)
188 if (!dnode->inode || !dnode->inode->ops->open) {
192 struct v_inode* inode = dnode->inode;
196 struct v_file* vfile = cake_grab(file_pile);
197 memset(vfile, 0, sizeof(*vfile));
199 vfile->dnode = dnode;
200 vfile->inode = inode;
201 vfile->ref_count = ATOMIC_VAR_INIT(1);
202 vfile->ops = inode->default_fops;
204 if ((inode->itype & F_MFILE) && !inode->pg_cache) {
205 struct pcache* pcache = vzalloc(sizeof(struct pcache));
207 pcache->master = inode;
208 inode->pg_cache = pcache;
211 int errno = inode->ops->open(inode, vfile);
213 cake_release(file_pile, vfile);
215 atomic_fetch_add(&dnode->ref_count, 1);
217 mnt_mkbusy(dnode->mnt);
228 vfs_assign_inode(struct v_dnode* assign_to, struct v_inode* inode)
230 if (assign_to->inode) {
231 llist_delete(&assign_to->aka_list);
232 assign_to->inode->link_count--;
234 llist_append(&inode->aka_dnodes, &assign_to->aka_list);
235 assign_to->inode = inode;
240 vfs_link(struct v_dnode* to_link, struct v_dnode* name)
244 if ((errno = vfs_check_writable(to_link))) {
248 lock_inode(to_link->inode);
249 if (to_link->super_block->root != name->super_block->root) {
251 } else if (!to_link->inode->ops->link) {
253 } else if (!(errno = to_link->inode->ops->link(to_link->inode, name))) {
254 vfs_assign_inode(name, to_link->inode);
256 unlock_inode(to_link->inode);
262 vfs_pclose(struct v_file* file, pid_t pid)
265 if (file->ref_count > 1) {
266 atomic_fetch_sub(&file->ref_count, 1);
267 } else if (!(errno = file->ops->close(file))) {
268 atomic_fetch_sub(&file->dnode->ref_count, 1);
269 file->inode->open_count--;
273 * This happened when process is terminated while blocking on read.
274 * In that case, the process is still holding the inode lock and it
275 will never get released.
276 * The unlocking should also include ownership check.
278 * To see why, consider two process both open the same file both with
280 * Process A: busy on reading x
281 * Process B: do nothing with x
282 * Assuming that, after a very short time, process B get terminated
283 * while process A is still busy in it's reading business. By this
284 * design, the inode lock of this file x is get released by B rather
285 * than A. And this will cause a probable race condition on A if other
286 * process is writing to this file later after B exit.
288 if (mutex_on_hold(&file->inode->lock)) {
289 mutex_unlock_for(&file->inode->lock, pid);
291 mnt_chillax(file->dnode->mnt);
293 pcache_commit_all(file->inode);
294 cake_release(file_pile, file);
300 vfs_close(struct v_file* file)
302 return vfs_pclose(file, __current->pid);
306 vfs_free_fd(struct v_fd* fd)
308 cake_release(fd_pile, fd);
312 vfs_fsync(struct v_file* file)
315 if ((errno = vfs_check_writable(file->dnode))) {
319 lock_inode(file->inode);
321 pcache_commit_all(file->inode);
324 if (file->ops->sync) {
325 errno = file->ops->sync(file);
328 unlock_inode(file->inode);
334 vfs_alloc_fdslot(int* fd)
336 for (size_t i = 0; i < VFS_MAX_FD; i++) {
337 if (!__current->fdtable->fds[i]) {
348 struct v_superblock* sb = cake_grab(superblock_pile);
349 memset(sb, 0, sizeof(*sb));
350 llist_init_head(&sb->sb_list);
351 sb->i_cache = vzalloc(VFS_HASHTABLE_SIZE * sizeof(struct hbucket));
356 vfs_sb_free(struct v_superblock* sb)
359 cake_release(superblock_pile, sb);
363 __vfs_try_evict_dnode(struct lru_node* obj)
365 struct v_dnode* dnode = container_of(obj, struct v_dnode, lru);
367 if (!dnode->ref_count) {
375 __vfs_try_evict_inode(struct lru_node* obj)
377 struct v_inode* inode = container_of(obj, struct v_inode, lru);
379 if (!inode->link_count && !inode->open_count) {
387 vfs_d_alloc(struct v_dnode* parent, struct hstr* name)
389 struct v_dnode* dnode = cake_grab(dnode_pile);
391 lru_evict_half(dnode_lru);
393 if (!(dnode = cake_grab(dnode_pile))) {
398 memset(dnode, 0, sizeof(*dnode));
399 llist_init_head(&dnode->children);
400 llist_init_head(&dnode->siblings);
401 llist_init_head(&dnode->aka_list);
402 mutex_init(&dnode->lock);
404 dnode->ref_count = ATOMIC_VAR_INIT(0);
405 dnode->name = HHSTR(vzalloc(VFS_NAME_MAXLEN), 0, 0);
407 hstrcpy(&dnode->name, name);
410 dnode->super_block = parent->super_block;
411 dnode->mnt = parent->mnt;
414 lru_use_one(dnode_lru, &dnode->lru);
420 vfs_d_free(struct v_dnode* dnode)
422 assert(dnode->ref_count == 1);
425 assert(dnode->inode->link_count > 0);
426 dnode->inode->link_count--;
429 vfs_dcache_remove(dnode);
430 // Make sure the children de-referencing their parent.
431 // With lru presented, the eviction will be propagated over the entire
432 // detached subtree eventually
433 struct v_dnode *pos, *n;
434 llist_for_each(pos, n, &dnode->children, siblings)
436 vfs_dcache_remove(pos);
439 vfree((void*)dnode->name.value);
440 cake_release(dnode_pile, dnode);
444 vfs_i_find(struct v_superblock* sb, u32_t i_id)
446 struct hbucket* slot = &sb->i_cache[i_id & VFS_HASH_MASK];
447 struct v_inode *pos, *n;
448 hashtable_bucket_foreach(slot, pos, n, hash_list)
450 if (pos->id == i_id) {
451 lru_use_one(inode_lru, &pos->lru);
460 vfs_i_addhash(struct v_inode* inode)
462 struct hbucket* slot = &inode->sb->i_cache[inode->id & VFS_HASH_MASK];
464 hlist_delete(&inode->hash_list);
465 hlist_add(&slot->head, &inode->hash_list);
469 vfs_i_alloc(struct v_superblock* sb)
471 assert(sb->ops.init_inode);
473 struct v_inode* inode;
474 if (!(inode = cake_grab(inode_pile))) {
475 lru_evict_half(inode_lru);
476 if (!(inode = cake_grab(inode_pile))) {
481 memset(inode, 0, sizeof(*inode));
482 mutex_init(&inode->lock);
483 llist_init_head(&inode->xattrs);
484 llist_init_head(&inode->aka_dnodes);
486 sb->ops.init_inode(sb, inode);
489 inode->ctime = clock_unixtime();
490 inode->atime = inode->ctime;
491 inode->mtime = inode->ctime;
493 lru_use_one(inode_lru, &inode->lru);
498 vfs_i_free(struct v_inode* inode)
500 if (inode->pg_cache) {
501 pcache_release(inode->pg_cache);
502 vfree(inode->pg_cache);
504 // we don't need to sync inode.
505 // If an inode can be free, then it must be properly closed.
506 // Hence it must be synced already!
507 if (inode->destruct) {
508 inode->destruct(inode);
510 hlist_delete(&inode->hash_list);
511 cake_release(inode_pile, inode);
514 /* ---- System call definition and support ---- */
516 #define FLOCATE_CREATE_EMPTY 1
517 #define FLOCATE_CREATE_ONLY 2
518 #define FLOCATE_NOFOLLOW 4
521 vfs_getfd(int fd, struct v_fd** fd_s)
523 if (TEST_FD(fd) && (*fd_s = __current->fdtable->fds[fd])) {
530 __vfs_try_locate_file(const char* path,
531 struct v_dnode** fdir,
532 struct v_dnode** file,
535 char name_str[VFS_NAME_MAXLEN];
536 struct hstr name = HSTR(name_str, 0);
537 int errno, woption = 0;
539 if ((options & FLOCATE_NOFOLLOW)) {
540 woption |= VFS_WALK_NOFOLLOW;
544 if ((errno = vfs_walk_proc(path, fdir, &name, woption | VFS_WALK_PARENT))) {
548 errno = vfs_walk(*fdir, name.value, file, NULL, woption);
550 if (errno != ENOENT && (options & FLOCATE_CREATE_ONLY)) {
554 if (errno != ENOENT ||
555 !(options & (FLOCATE_CREATE_EMPTY | FLOCATE_CREATE_ONLY))) {
559 struct v_dnode* parent = *fdir;
560 struct v_dnode* file_new = vfs_d_alloc(parent, &name);
568 if (!(errno = parent->inode->ops->create(parent->inode, file_new))) {
569 vfs_dcache_add(parent, file_new);
572 vfs_d_free(file_new);
575 unlock_dnode(parent);
581 vfs_do_open(const char* path, int options)
583 int errno, fd, loptions = 0;
584 struct v_dnode *dentry, *file;
585 struct v_file* ofile = NULL;
587 if ((options & FO_CREATE)) {
588 loptions |= FLOCATE_CREATE_EMPTY;
589 } else if ((options & FO_NOFOLLOW)) {
590 loptions |= FLOCATE_NOFOLLOW;
593 errno = __vfs_try_locate_file(path, &dentry, &file, loptions);
595 if (!errno && !(errno = vfs_alloc_fdslot(&fd))) {
597 if (errno || (errno = vfs_open(file, &ofile))) {
601 struct v_fd* fd_s = cake_grab(fd_pile);
602 memset(fd_s, 0, sizeof(*fd_s));
604 ofile->f_pos = ofile->inode->fsize & -((options & FO_APPEND) != 0);
606 fd_s->flags = options;
607 __current->fdtable->fds[fd] = fd_s;
614 __DEFINE_LXSYSCALL2(int, open, const char*, path, int, options)
616 int errno = vfs_do_open(path, options);
617 return DO_STATUS_OR_RETURN(errno);
620 __DEFINE_LXSYSCALL1(int, close, int, fd)
624 if ((errno = vfs_getfd(fd, &fd_s))) {
628 if ((errno = vfs_close(fd_s->file))) {
632 cake_release(fd_pile, fd_s);
633 __current->fdtable->fds[fd] = 0;
636 return DO_STATUS(errno);
640 __vfs_readdir_callback(struct dir_context* dctx,
645 struct lx_dirent* dent = (struct lx_dirent*)dctx->cb_data;
646 strncpy(dent->d_name, name, DIRENT_NAME_MAX_LEN);
648 dent->d_type = dtype;
651 __DEFINE_LXSYSCALL2(int, sys_readdir, int, fd, struct lx_dirent*, dent)
656 if ((errno = vfs_getfd(fd, &fd_s))) {
660 struct v_inode* inode = fd_s->file->inode;
664 if ((inode->itype & F_FILE)) {
667 struct dir_context dctx = (struct dir_context){
669 .index = dent->d_offset,
670 .read_complete_callback = __vfs_readdir_callback};
672 if (dent->d_offset == 0) {
673 __vfs_readdir_callback(&dctx, vfs_dot.value, vfs_dot.len, DT_DIR);
674 } else if (dent->d_offset == 1) {
675 __vfs_readdir_callback(&dctx, vfs_ddot.value, vfs_ddot.len, DT_DIR);
678 if ((errno = fd_s->file->ops->readdir(fd_s->file, &dctx)) != 1) {
689 return DO_STATUS_OR_RETURN(errno);
692 __DEFINE_LXSYSCALL3(int, read, int, fd, void*, buf, size_t, count)
696 if ((errno = vfs_getfd(fd, &fd_s))) {
700 struct v_file* file = fd_s->file;
701 if (!(file->inode->itype & F_FILE)) {
706 lock_inode(file->inode);
708 file->inode->atime = clock_unixtime();
710 if ((file->inode->itype & VFS_IFSEQDEV) || (fd_s->flags & FO_DIRECT)) {
711 errno = file->ops->read(file->inode, buf, count, file->f_pos);
713 errno = pcache_read(file->inode, buf, count, file->f_pos);
717 file->f_pos += errno;
718 unlock_inode(file->inode);
722 unlock_inode(file->inode);
725 return DO_STATUS(errno);
728 __DEFINE_LXSYSCALL3(int, write, int, fd, void*, buf, size_t, count)
732 if ((errno = vfs_getfd(fd, &fd_s))) {
736 struct v_file* file = fd_s->file;
738 if ((errno = vfs_check_writable(file->dnode))) {
742 if (!(file->inode->itype & F_FILE)) {
747 lock_inode(file->inode);
749 file->inode->mtime = clock_unixtime();
751 if ((file->inode->itype & VFS_IFSEQDEV) || (fd_s->flags & FO_DIRECT)) {
752 errno = file->ops->write(file->inode, buf, count, file->f_pos);
754 errno = pcache_write(file->inode, buf, count, file->f_pos);
758 file->f_pos += errno;
759 unlock_inode(file->inode);
763 unlock_inode(file->inode);
766 return DO_STATUS(errno);
769 __DEFINE_LXSYSCALL3(int, lseek, int, fd, int, offset, int, options)
773 if ((errno = vfs_getfd(fd, &fd_s))) {
777 struct v_file* file = fd_s->file;
779 if (!file->ops->seek) {
784 lock_inode(file->inode);
787 int fpos = file->f_pos;
790 overflow = sadd_overflow((int)file->f_pos, offset, &fpos);
793 overflow = sadd_overflow((int)file->inode->fsize, offset, &fpos);
801 } else if (!(errno = file->ops->seek(file->inode, fpos))) {
805 unlock_inode(file->inode);
808 return DO_STATUS(errno);
812 vfs_get_path(struct v_dnode* dnode, char* buf, size_t size, int depth)
824 if (dnode->parent != dnode) {
825 len = vfs_get_path(dnode->parent, buf, size, depth + 1);
832 if (!len || buf[len - 1] != VFS_PATH_DELIM) {
833 buf[len++] = VFS_PATH_DELIM;
836 size_t cpy_size = MIN(dnode->name.len, size - len);
837 strncpy(buf + len, dnode->name.value, cpy_size);
844 vfs_readlink(struct v_dnode* dnode, char* buf, size_t size)
847 struct v_inode* inode = dnode->inode;
848 if (inode->ops->read_symlink) {
851 int errno = inode->ops->read_symlink(inode, &link);
852 strncpy(buf, link, size);
861 vfs_get_dtype(int itype)
863 if ((itype & VFS_IFSYMLINK) == VFS_IFSYMLINK) {
865 } else if (!(itype & VFS_IFFILE)) {
872 __DEFINE_LXSYSCALL3(int, realpathat, int, fd, char*, buf, size_t, size)
876 if ((errno = vfs_getfd(fd, &fd_s))) {
880 struct v_dnode* dnode;
881 errno = vfs_get_path(fd_s->file->dnode, buf, size, 0);
888 return DO_STATUS(errno);
891 __DEFINE_LXSYSCALL3(int, readlink, const char*, path, char*, buf, size_t, size)
894 struct v_dnode* dnode;
895 if (!(errno = vfs_walk_proc(path, &dnode, NULL, VFS_WALK_NOFOLLOW))) {
896 errno = vfs_readlink(dnode, buf, size);
903 return DO_STATUS(errno);
907 int, readlinkat, int, dirfd, const char*, pathname, char*, buf, size_t, size)
911 if ((errno = vfs_getfd(dirfd, &fd_s))) {
915 pathname = pathname ? pathname : "";
917 struct v_dnode* dnode;
918 if (!(errno = vfs_walk(
919 fd_s->file->dnode, pathname, &dnode, NULL, VFS_WALK_NOFOLLOW))) {
920 errno = vfs_readlink(fd_s->file->dnode, buf, size);
928 return DO_STATUS(errno);
933 When we perform operation that could affect the layout of
934 directory (i.e., rename, mkdir, rmdir). We must lock the parent dir
935 whenever possible. This will blocking any ongoing path walking to reach
936 it hence avoid any partial state.
939 __DEFINE_LXSYSCALL1(int, rmdir, const char*, pathname)
942 struct v_dnode* dnode;
943 if ((errno = vfs_walk_proc(pathname, &dnode, NULL, 0))) {
944 return DO_STATUS(errno);
949 if ((errno = vfs_check_writable(dnode))) {
953 if ((dnode->super_block->fs->types & FSTYPE_ROFS)) {
958 if (dnode->ref_count > 1 || dnode->inode->open_count) {
963 if (!llist_empty(&dnode->children)) {
968 struct v_dnode* parent = dnode->parent;
976 lock_inode(parent->inode);
978 if (!(dnode->inode->itype & F_MFILE)) {
979 errno = parent->inode->ops->rmdir(parent->inode, dnode);
981 vfs_dcache_remove(dnode);
987 unlock_inode(parent->inode);
988 unlock_dnode(parent);
992 return DO_STATUS(errno);
995 __DEFINE_LXSYSCALL1(int, mkdir, const char*, path)
998 struct v_dnode *parent, *dir;
999 char name_value[VFS_NAME_MAXLEN];
1000 struct hstr name = HHSTR(name_value, 0, 0);
1002 if ((errno = vfs_walk_proc(path, &parent, &name, VFS_WALK_PARENT))) {
1006 if (!(errno = vfs_walk(parent, name_value, &dir, NULL, 0))) {
1011 if ((errno = vfs_check_writable(parent))) {
1015 if (!(dir = vfs_d_alloc(parent, &name))) {
1021 lock_inode(parent->inode);
1023 if ((parent->super_block->fs->types & FSTYPE_ROFS)) {
1025 } else if (!parent->inode->ops->mkdir) {
1027 } else if ((parent->inode->itype & F_FILE)) {
1029 } else if (!(errno = parent->inode->ops->mkdir(parent->inode, dir))) {
1030 vfs_dcache_add(parent, dir);
1037 unlock_inode(parent->inode);
1038 unlock_dnode(parent);
1040 return DO_STATUS(errno);
1044 __vfs_do_unlink(struct v_dnode* dnode)
1047 struct v_inode* inode = dnode->inode;
1049 if (dnode->ref_count > 1) {
1053 if ((errno = vfs_check_writable(dnode))) {
1059 if (inode->open_count) {
1061 } else if ((inode->itype & F_MFILE)) {
1062 errno = inode->ops->unlink(inode);
1070 unlock_inode(inode);
1075 __DEFINE_LXSYSCALL1(int, unlink, const char*, pathname)
1078 struct v_dnode* dnode;
1079 if ((errno = vfs_walk_proc(pathname, &dnode, NULL, 0))) {
1083 errno = __vfs_do_unlink(dnode);
1086 return DO_STATUS(errno);
1089 __DEFINE_LXSYSCALL2(int, unlinkat, int, fd, const char*, pathname)
1093 if ((errno = vfs_getfd(fd, &fd_s))) {
1097 struct v_dnode* dnode;
1098 if (!(errno = vfs_walk(fd_s->file->dnode, pathname, &dnode, NULL, 0))) {
1099 errno = __vfs_do_unlink(dnode);
1103 return DO_STATUS(errno);
1106 __DEFINE_LXSYSCALL2(int, link, const char*, oldpath, const char*, newpath)
1109 struct v_dnode *dentry, *to_link, *name_dentry, *name_file;
1111 errno = __vfs_try_locate_file(oldpath, &dentry, &to_link, 0);
1113 errno = __vfs_try_locate_file(
1114 newpath, &name_dentry, &name_file, FLOCATE_CREATE_ONLY);
1116 errno = vfs_link(to_link, name_file);
1119 return DO_STATUS(errno);
1122 __DEFINE_LXSYSCALL1(int, fsync, int, fildes)
1127 if (!(errno = vfs_getfd(fildes, &fd_s))) {
1128 errno = vfs_fsync(fd_s->file);
1131 return DO_STATUS(errno);
1135 vfs_dup_fd(struct v_fd* old, struct v_fd** new)
1138 struct v_fd* copied = cake_grab(fd_pile);
1140 memcpy(copied, old, sizeof(struct v_fd));
1142 atomic_fetch_add(&old->file->ref_count, 1);
1150 vfs_dup2(int oldfd, int newfd)
1152 if (newfd == oldfd) {
1157 struct v_fd *oldfd_s, *newfd_s;
1158 if ((errno = vfs_getfd(oldfd, &oldfd_s))) {
1162 if (!TEST_FD(newfd)) {
1167 newfd_s = __current->fdtable->fds[newfd];
1168 if (newfd_s && (errno = vfs_close(newfd_s->file))) {
1172 if (!(errno = vfs_dup_fd(oldfd_s, &newfd_s))) {
1173 __current->fdtable->fds[newfd] = newfd_s;
1178 return DO_STATUS(errno);
1181 __DEFINE_LXSYSCALL2(int, dup2, int, oldfd, int, newfd)
1183 return vfs_dup2(oldfd, newfd);
1186 __DEFINE_LXSYSCALL1(int, dup, int, oldfd)
1189 struct v_fd *oldfd_s, *newfd_s;
1190 if ((errno = vfs_getfd(oldfd, &oldfd_s))) {
1194 if (!(errno = vfs_alloc_fdslot(&newfd)) &&
1195 !(errno = vfs_dup_fd(oldfd_s, &newfd_s))) {
1196 __current->fdtable->fds[newfd] = newfd_s;
1201 return DO_STATUS(errno);
1204 __DEFINE_LXSYSCALL2(
1205 int, symlink, const char*, pathname, const char*, link_target)
1208 struct v_dnode *dnode, *file;
1209 if ((errno = __vfs_try_locate_file(
1210 pathname, &dnode, &file, FLOCATE_CREATE_ONLY))) {
1214 if ((errno = vfs_check_writable(file))) {
1218 if (!file->inode->ops->set_symlink) {
1223 lock_inode(file->inode);
1225 errno = file->inode->ops->set_symlink(file->inode, link_target);
1227 unlock_inode(file->inode);
1230 return DO_STATUS(errno);
1234 vfs_ref_file(struct v_file* file)
1236 atomic_fetch_add(&file->ref_count, 1);
1240 vfs_ref_dnode(struct v_dnode* dnode)
1242 atomic_fetch_add(&dnode->ref_count, 1);
1245 mnt_mkbusy(dnode->mnt);
1250 vfs_unref_dnode(struct v_dnode* dnode)
1252 atomic_fetch_sub(&dnode->ref_count, 1);
1254 mnt_chillax(dnode->mnt);
1259 vfs_do_chdir(struct proc_info* proc, struct v_dnode* dnode)
1265 if ((dnode->inode->itype & F_FILE)) {
1271 vfs_unref_dnode(proc->cwd);
1274 vfs_ref_dnode(dnode);
1277 unlock_dnode(dnode);
1283 __DEFINE_LXSYSCALL1(int, chdir, const char*, path)
1285 struct v_dnode* dnode;
1288 if ((errno = vfs_walk_proc(path, &dnode, NULL, 0))) {
1292 errno = vfs_do_chdir((struct proc_info*)__current, dnode);
1295 return DO_STATUS(errno);
1298 __DEFINE_LXSYSCALL1(int, fchdir, int, fd)
1303 if ((errno = vfs_getfd(fd, &fd_s))) {
1307 errno = vfs_do_chdir((struct proc_info*)__current, fd_s->file->dnode);
1310 return DO_STATUS(errno);
1313 __DEFINE_LXSYSCALL2(char*, getcwd, char*, buf, size_t, size)
1324 if (!__current->cwd) {
1325 *buf = VFS_PATH_DELIM;
1328 len = vfs_get_path(__current->cwd, buf, size, 0);
1340 syscall_result(errno);
1345 vfs_do_rename(struct v_dnode* current, struct v_dnode* target)
1348 if (current->inode->id == target->inode->id) {
1353 if ((errno = vfs_check_writable(current))) {
1357 if (current->ref_count > 1 || target->ref_count > 1) {
1361 if (current->super_block != target->super_block) {
1365 struct v_dnode* oldparent = current->parent;
1366 struct v_dnode* newparent = target->parent;
1368 lock_dnode(current);
1371 lock_dnode(oldparent);
1373 lock_dnode(newparent);
1375 if (!llist_empty(&target->children)) {
1377 unlock_dnode(target);
1382 current->inode->ops->rename(current->inode, current, target))) {
1383 unlock_dnode(target);
1387 // re-position current
1388 hstrcpy(¤t->name, &target->name);
1389 vfs_dcache_rehash(newparent, current);
1394 unlock_dnode(target);
1397 unlock_dnode(current);
1399 unlock_dnode(oldparent);
1401 unlock_dnode(newparent);
1406 __DEFINE_LXSYSCALL2(int, rename, const char*, oldpath, const char*, newpath)
1408 struct v_dnode *cur, *target_parent, *target;
1409 struct hstr name = HSTR(valloc(VFS_NAME_MAXLEN), 0);
1412 if ((errno = vfs_walk_proc(oldpath, &cur, NULL, 0))) {
1416 if ((errno = vfs_walk(
1417 __current->cwd, newpath, &target_parent, &name, VFS_WALK_PARENT))) {
1421 errno = vfs_walk(target_parent, name.value, &target, NULL, 0);
1422 if (errno == ENOENT) {
1423 target = vfs_d_alloc(target_parent, &name);
1424 vfs_dcache_add(target_parent, target);
1434 errno = vfs_do_rename(cur, target);
1437 vfree((void*)name.value);
1438 return DO_STATUS(errno);
1441 __DEFINE_LXSYSCALL2(int, fstat, int, fd, struct file_stat*, stat)
1446 if ((errno = vfs_getfd(fd, &fds))) {
1450 struct v_inode* vino = fds->file->inode;
1451 struct device* fdev = vino->sb->dev;
1453 *stat = (struct file_stat){.st_ino = vino->id,
1454 .st_blocks = vino->lb_usage,
1455 .st_size = vino->fsize,
1456 .mode = vino->itype,
1457 .st_ioblksize = PG_SIZE,
1458 .st_blksize = vino->sb->blksize};
1460 if (VFS_DEVFILE(vino->itype)) {
1461 struct device* rdev = resolve_device(vino->data);
1462 if (!rdev || rdev->magic != DEV_STRUCT_MAGIC) {
1467 stat->st_rdev = (dev_t){.meta = rdev->ident.fn_grp,
1468 .unique = rdev->ident.unique,
1469 .index = rdev->dev_uid};
1473 stat->st_dev = (dev_t){.meta = fdev->ident.fn_grp,
1474 .unique = fdev->ident.unique,
1475 .index = fdev->dev_uid};
1479 return DO_STATUS(errno);