X-Git-Url: https://scm.lunaixsky.com/lunaix-os.git/blobdiff_plain/003db17f7a36a8098827f40df8fedc087b5a26f6..2687264120b44704961375849a51d69b17ce585a:/lunaix-os/kernel/fs/vfs.c

diff --git a/lunaix-os/kernel/fs/vfs.c b/lunaix-os/kernel/fs/vfs.c
index 902f21e..82f409a 100644
--- a/lunaix-os/kernel/fs/vfs.c
+++ b/lunaix-os/kernel/fs/vfs.c
@@ -26,6 +26,12 @@
 #define DNODE_HASH_MASK (DNODE_HASHTABLE_SIZE - 1)
 #define DNODE_HASHBITS (32 - DNODE_HASHTABLE_BITS)
 
+#define lock_inode(inode) mutex_lock(&inode->lock)
+#define unlock_inode(inode) mutex_unlock(&inode->lock)
+
+#define lock_dnode(dnode) mutex_lock(&dnode->lock)
+#define unlock_dnode(dnode) mutex_unlock(&dnode->lock)
+
 static struct cake_pile* dnode_pile;
 static struct cake_pile* inode_pile;
 static struct cake_pile* file_pile;
@@ -37,6 +43,10 @@ static struct hbucket* dnode_cache;
 
 static int fs_id = 0;
 
+struct hstr vfs_ddot = HSTR("..", 2);
+struct hstr vfs_dot = HSTR(".", 1);
+struct hstr vfs_empty = HSTR("", 0);
+
 struct v_dnode*
 vfs_d_alloc();
 
@@ -62,9 +72,13 @@ vfs_init()
 
     dnode_cache = vzalloc(DNODE_HASHTABLE_SIZE * sizeof(struct hbucket));
 
+    hstr_rehash(&vfs_ddot, HSTR_FULL_HASH);
+    hstr_rehash(&vfs_dot, HSTR_FULL_HASH);
+
     // 创建一个根superblock，用来蕴含我们的根目录。
     root_sb = vfs_sb_alloc();
     root_sb->root = vfs_d_alloc();
+    root_sb->root->inode = vfs_i_alloc();
 }
 
 inline struct hbucket*
@@ -80,6 +94,13 @@ __dcache_get_bucket(struct v_dnode* parent, unsigned int hash)
 struct v_dnode*
 vfs_dcache_lookup(struct v_dnode* parent, struct hstr* str)
 {
+    if (!str->len || HSTR_EQ(str, &vfs_dot))
+        return parent;
+
+    if (HSTR_EQ(str, &vfs_ddot)) {
+        return parent->parent ? parent->parent : parent;
+    }
+
     struct hbucket* slot = __dcache_get_bucket(parent, str->hash);
 
     struct v_dnode *pos, *n;
@@ -100,16 +121,16 @@ vfs_dcache_add(struct v_dnode* parent, struct v_dnode* dnode)
 }
 
 int
-vfs_walk(struct v_dnode* start,
-         const char* path,
-         struct v_dnode** dentry,
-         struct hstr* component,
-         int walk_options)
+__vfs_walk(struct v_dnode* start,
+           const char* path,
+           struct v_dnode** dentry,
+           struct hstr* component,
+           int walk_options)
 {
     int errno = 0;
     int i = 0, j = 0;
 
-    if (path[0] == PATH_DELIM) {
+    if (path[0] == PATH_DELIM || !start) {
         if ((walk_options & VFS_WALK_FSRELATIVE) && start) {
             start = start->super_block->root;
         } else {
@@ -132,7 +153,7 @@ vfs_walk(struct v_dnode* start,
                 return ENAMETOOLONG;
             }
             if (!VFS_VALID_CHAR(current)) {
-                return VFS_EINVLD;
+                return EINVAL;
             }
             name_content[j++] = current;
             if (lookahead) {
@@ -145,7 +166,10 @@ vfs_walk(struct v_dnode* start,
             goto cont;
         }
 
+        lock_dnode(current_level);
+
         name_content[j] = 0;
+        name.len = j;
         hstr_rehash(&name, HSTR_FULL_HASH);
 
         if (!lookahead && (walk_options & VFS_WALK_PARENT)) {
@@ -154,6 +178,7 @@ vfs_walk(struct v_dnode* start,
                 component->len = j;
                 strcpy(component->value, name_content);
             }
+            unlock_dnode(current_level);
             break;
         }
 
@@ -161,11 +186,12 @@ vfs_walk(struct v_dnode* start,
 
         if (!dnode) {
             dnode = vfs_d_alloc();
-            dnode->name = HSTR(valloc(VFS_NAME_MAXLEN), j);
-            dnode->name.hash = name.hash;
+            dnode->name = HHSTR(valloc(VFS_NAME_MAXLEN), j, name.hash);
 
             strcpy(dnode->name.value, name_content);
 
+            lock_inode(current_level->inode);
+
             errno =
               current_level->inode->ops.dir_lookup(current_level->inode, dnode);
 
@@ -178,7 +204,11 @@ vfs_walk(struct v_dnode* start,
                 }
             }
 
+            unlock_inode(current_level->inode);
+
             if (errno) {
+                unlock_dnode(current_level);
+                vfree(dnode->name.value);
                 goto error;
             }
 
@@ -188,6 +218,8 @@ vfs_walk(struct v_dnode* start,
             llist_append(&current_level->children, &dnode->siblings);
         }
 
+        unlock_dnode(current_level);
+
         j = 0;
         current_level = dnode;
     cont:
@@ -200,6 +232,45 @@ vfs_walk(struct v_dnode* start,
 error:
     vfree(dnode->name.value);
     vfs_d_free(dnode);
+    *dentry = NULL;
+    return errno;
+}
+
+#define VFS_MAX_SYMLINK 16
+
+int
+vfs_walk(struct v_dnode* start,
+         const char* path,
+         struct v_dnode** dentry,
+         struct hstr* component,
+         int options)
+{
+    struct v_dnode* interim;
+    const char* pathname = path;
+    int errno = __vfs_walk(start, path, &interim, component, options);
+    int counter = 0;
+
+    while (!errno) {
+        if (counter >= VFS_MAX_SYMLINK) {
+            errno = ELOOP;
+            continue;
+        }
+        if ((interim->inode->itype & VFS_IFSYMLINK) &&
+            !(options & VFS_WALK_NOFOLLOW) &&
+            interim->inode->ops.read_symlink) {
+            errno = interim->inode->ops.read_symlink(interim->inode, &pathname);
+            if (errno) {
+                break;
+            }
+        } else {
+            break;
+        }
+        errno = __vfs_walk(start, pathname, &interim, component, options);
+        counter++;
+    }
+
+    *dentry = errno ? 0 : interim;
+
     return errno;
 }
 
@@ -209,7 +280,7 @@ vfs_mount(const char* target, const char* fs_name, bdev_t device)
     int errno;
     struct v_dnode* mnt;
 
-    if (!(errno = vfs_walk(NULL, target, &mnt, NULL, 0))) {
+    if (!(errno = vfs_walk(__current->cwd, target, &mnt, NULL, 0))) {
         errno = vfs_mount_at(fs_name, device, mnt);
     }
 
@@ -222,7 +293,7 @@ vfs_unmount(const char* target)
     int errno;
     struct v_dnode* mnt;
 
-    if (!(errno = vfs_walk(NULL, target, &mnt, NULL, 0))) {
+    if (!(errno = vfs_walk(__current->cwd, target, &mnt, NULL, 0))) {
         errno = vfs_unmount_at(mnt);
     }
 
@@ -234,7 +305,7 @@ vfs_mount_at(const char* fs_name, bdev_t device, struct v_dnode* mnt_point)
 {
     struct filesystem* fs = fsm_get(fs_name);
     if (!fs)
-        return VFS_ENOFS;
+        return ENODEV;
     struct v_superblock* sb = vfs_sb_alloc();
     sb->dev = device;
     sb->fs_id = fs_id++;
@@ -256,7 +327,7 @@ vfs_unmount_at(struct v_dnode* mnt_point)
     int errno = 0;
     struct v_superblock* sb = mnt_point->super_block;
     if (!sb) {
-        return VFS_EBADMNT;
+        return EINVAL;
     }
     if (!(errno = sb->fs->unmount(sb))) {
         struct v_dnode* fs_root = sb->root;
@@ -274,27 +345,63 @@ vfs_open(struct v_dnode* dnode, struct v_file** file)
         return ENOTSUP;
     }
 
+    struct v_inode* inode = dnode->inode;
     struct v_file* vfile = cake_grab(file_pile);
     memset(vfile, 0, sizeof(*vfile));
 
-    int errno = dnode->inode->ops.open(dnode->inode, vfile);
+    vfile->dnode = dnode;
+    vfile->inode = inode;
+    vfile->ref_count = ATOMIC_VAR_INIT(1);
+    vfile->ops = inode->default_fops;
+
+    if ((inode->itype & VFS_IFFILE) && !inode->pg_cache) {
+        struct pcache* pcache = vzalloc(sizeof(struct pcache));
+        pcache_init(pcache);
+        pcache->master = inode;
+        inode->pg_cache = pcache;
+    }
+
+    int errno = inode->ops.open(inode, vfile);
     if (errno) {
         cake_release(file_pile, vfile);
     } else {
+        atomic_fetch_add(&dnode->ref_count, 1);
+        inode->open_count++;
+
         *file = vfile;
     }
+
     return errno;
 }
 
 int
-vfs_close(struct v_file* file)
+vfs_link(struct v_dnode* to_link, struct v_dnode* name)
 {
-    if (!file->ops.close) {
-        return ENOTSUP;
+    int errno;
+
+    lock_inode(to_link->inode);
+    if (to_link->super_block->root != name->super_block->root) {
+        errno = EXDEV;
+    } else if (!to_link->inode->ops.link) {
+        errno = ENOTSUP;
+    } else if (!(errno = to_link->inode->ops.link(to_link->inode, name))) {
+        name->inode = to_link->inode;
+        atomic_fetch_add(&to_link->inode->link_count, 1);
     }
+    unlock_inode(to_link->inode);
 
-    int errno = file->ops.close(file);
-    if (!errno) {
+    return errno;
+}
+
+int
+vfs_close(struct v_file* file)
+{
+    int errno = 0;
+    if (!file->ops.close || !(errno = file->ops.close(file))) {
+        atomic_fetch_sub(&file->dnode->ref_count, 1);
+        file->inode->open_count--;
+
+        pcache_commit_all(file->inode);
         cake_release(file_pile, file);
     }
     return errno;
@@ -303,13 +410,16 @@ vfs_close(struct v_file* file)
 int
 vfs_fsync(struct v_file* file)
 {
+    lock_inode(file->inode);
+
     int errno = ENOTSUP;
+    pcache_commit_all(file->inode);
     if (file->ops.sync) {
-        errno = file->ops.sync(file);
-    }
-    if (!errno && file->inode->ops.sync) {
-        return file->inode->ops.sync(file->inode);
+        errno = file->ops.sync(file->inode);
     }
+
+    unlock_inode(file->inode);
+
     return errno;
 }
 
@@ -344,7 +454,14 @@ struct v_dnode*
 vfs_d_alloc()
 {
     struct v_dnode* dnode = cake_grab(dnode_pile);
+    memset(dnode, 0, sizeof(*dnode));
     llist_init_head(&dnode->children);
+    mutex_init(&dnode->lock);
+
+    dnode->ref_count = ATOMIC_VAR_INIT(0);
+
+    dnode->name = vfs_empty;
+    return dnode;
 }
 
 void
@@ -361,6 +478,7 @@ vfs_i_alloc()
 {
     struct v_inode* inode = cake_grab(inode_pile);
     memset(inode, 0, sizeof(*inode));
+    mutex_init(&inode->lock);
 
     return inode;
 }
@@ -371,50 +489,119 @@ vfs_i_free(struct v_inode* inode)
     cake_release(inode_pile, inode);
 }
 
-__DEFINE_LXSYSCALL2(int, open, const char*, path, int, options)
+/* ---- System call definition and support ---- */
+
+#define FLOCATE_CREATE_EMPTY 1
+
+#define DO_STATUS(errno) SYSCALL_ESTATUS(__current->k_status = errno)
+#define DO_STATUS_OR_RETURN(errno) ({ errno < 0 ? DO_STATUS(errno) : errno; })
+
+#define TEST_FD(fd) (fd >= 0 && fd < VFS_MAX_FD)
+
+int
+__vfs_getfd(int fd, struct v_fd** fd_s)
+{
+    if (TEST_FD(fd) && (*fd_s = __current->fdtable->fds[fd])) {
+        return 0;
+    }
+    return EBADF;
+}
+
+int
+__vfs_try_locate_file(const char* path,
+                      struct v_dnode** fdir,
+                      struct v_dnode** file,
+                      int options)
 {
     char name_str[VFS_NAME_MAXLEN];
     struct hstr name = HSTR(name_str, 0);
-    struct v_dnode *dentry, *file;
-    int errno, fd;
-    if ((errno = vfs_walk(NULL, path, &dentry, &name, VFS_WALK_PARENT))) {
-        return -1;
+    int errno;
+    if ((errno =
+           vfs_walk(__current->cwd, path, fdir, &name, VFS_WALK_PARENT))) {
+        return errno;
     }
 
-    struct v_file* opened_file = 0;
-    if (!(file = vfs_dcache_lookup(dentry, &name)) && (options & FO_CREATE)) {
-        errno = dentry->inode->ops.create(dentry->inode, opened_file);
-    } else if (!file) {
-        errno = EEXIST;
+    errno = vfs_walk(*fdir, name.value, file, NULL, 0);
+    if (errno != ENOENT || !(options & FLOCATE_CREATE_EMPTY)) {
+        return errno;
+    }
+
+    struct v_dnode* parent = *fdir;
+    struct v_dnode* file_new = vfs_d_alloc();
+    file_new->name = HHSTR(valloc(VFS_NAME_MAXLEN), name.len, name.hash);
+    strcpy(file_new->name.value, name_str);
+
+    if (!(errno = parent->inode->ops.create(parent->inode, file_new))) {
+        *file = file_new;
+
+        vfs_dcache_add(parent, file_new);
+        llist_append(&parent->children, &file_new->siblings);
     } else {
-        errno = vfs_open(file, &opened_file);
+        vfree(file_new->name.value);
+        vfs_d_free(file_new);
     }
 
-    __current->k_status = errno;
+    return errno;
+}
+
+int
+vfs_do_open(const char* path, int options)
+{
+    int errno, fd;
+    struct v_dnode *dentry, *file;
+    struct v_file* ofile = 0;
+
+    errno = __vfs_try_locate_file(
+      path, &dentry, &file, (options & FO_CREATE) ? FLOCATE_CREATE_EMPTY : 0);
+
+    if (errno || (errno = vfs_open(file, &ofile))) {
+        return errno;
+    }
+
+    struct v_inode* o_inode = ofile->inode;
+    if (!(o_inode->itype & VFS_IFSEQDEV) && !(options & FO_DIRECT)) {
+        // XXX Change here accordingly when signature of pcache_r/w changed.
+        ofile->ops.read = pcache_read;
+        ofile->ops.write = pcache_write;
+    }
 
     if (!errno && !(errno = vfs_alloc_fdslot(&fd))) {
         struct v_fd* fd_s = vzalloc(sizeof(*fd_s));
-        fd_s->file = opened_file;
-        fd_s->pos = file->inode->fsize & -((options & FO_APPEND) == 0);
+        ofile->f_pos = ofile->inode->fsize & -((options & FO_APPEND) != 0);
+        fd_s->file = ofile;
+        fd_s->flags = options;
         __current->fdtable->fds[fd] = fd_s;
+        return fd;
     }
 
-    return SYSCALL_ESTATUS(errno);
+    return errno;
+}
+
+__DEFINE_LXSYSCALL2(int, open, const char*, path, int, options)
+{
+    int errno = vfs_do_open(path, options);
+    return DO_STATUS_OR_RETURN(errno);
 }
 
 __DEFINE_LXSYSCALL1(int, close, int, fd)
 {
     struct v_fd* fd_s;
-    int errno;
-    if (fd < 0 || fd >= VFS_MAX_FD || !(fd_s = __current->fdtable->fds[fd])) {
-        errno = EBADF;
-    } else if (!(errno = vfs_close(fd_s->file))) {
-        vfree(fd_s);
+    int errno = 0;
+    if ((errno = __vfs_getfd(fd, &fd_s))) {
+        goto done_err;
     }
 
-    __current->k_status = errno;
+    if (fd_s->file->ref_count > 1) {
+        fd_s->file->ref_count--;
+    } else if ((errno = vfs_close(fd_s->file))) {
+        goto done_err;
+    }
 
-    return SYSCALL_ESTATUS(errno);
+    vfree(fd_s);
+    __current->fdtable->fds[fd] = 0;
+
+done_err:
+    return DO_STATUS(errno);
 }
 
 void
@@ -424,7 +611,7 @@ __vfs_readdir_callback(struct dir_context* dctx,
                        const int dtype)
 {
     struct dirent* dent = (struct dirent*)dctx->cb_data;
-    strcpy(dent->d_name, name);
+    strncpy(dent->d_name, name, DIRENT_NAME_MAX_LEN);
     dent->d_nlen = len;
     dent->d_type = dtype;
 }
@@ -433,9 +620,16 @@ __DEFINE_LXSYSCALL2(int, readdir, int, fd, struct dirent*, dent)
 {
     struct v_fd* fd_s;
     int errno;
-    if (fd < 0 || fd >= VFS_MAX_FD || !(fd_s = __current->fdtable->fds[fd])) {
-        errno = EBADF;
-    } else if (!(fd_s->file->inode->itype & VFS_INODE_TYPE_DIR)) {
+
+    if ((errno = __vfs_getfd(fd, &fd_s))) {
+        goto done;
+    }
+
+    struct v_inode* inode = fd_s->file->inode;
+
+    lock_inode(inode);
+
+    if (!(fd_s->file->inode->itype & VFS_IFDIR)) {
         errno = ENOTDIR;
     } else {
         struct dir_context dctx =
@@ -443,13 +637,288 @@ __DEFINE_LXSYSCALL2(int, readdir, int, fd, struct dirent*, dent)
                                 .index = dent->d_offset,
                                 .read_complete_callback =
                                   __vfs_readdir_callback };
-        if (!(errno = fd_s->file->ops.readdir(fd_s->file, &dctx))) {
-            dent->d_offset++;
+        if (dent->d_offset == 0) {
+            __vfs_readdir_callback(&dctx, vfs_dot.value, vfs_dot.len, 0);
+        } else if (dent->d_offset == 1) {
+            __vfs_readdir_callback(&dctx, vfs_ddot.value, vfs_ddot.len, 0);
+        } else {
+            dctx.index -= 2;
+            if ((errno = fd_s->file->ops.readdir(inode, &dctx))) {
+                unlock_inode(inode);
+                goto done;
+            }
         }
+        errno = 0;
+        dent->d_offset++;
     }
 
-    __current->k_status = errno;
-    return SYSCALL_ESTATUS(errno);
+    unlock_inode(inode);
+
+done:
+    return DO_STATUS(errno);
+}
+
+__DEFINE_LXSYSCALL3(int, read, int, fd, void*, buf, size_t, count)
+{
+    int errno = 0;
+    struct v_fd* fd_s;
+    if ((errno = __vfs_getfd(fd, &fd_s))) {
+        goto done;
+    }
+
+    struct v_file* file = fd_s->file;
+    if ((file->inode->itype & VFS_IFDIR)) {
+        errno = EISDIR;
+        goto done;
+    }
+
+    lock_inode(file->inode);
+
+    file->inode->atime = clock_unixtime();
+
+    __SYSCALL_INTERRUPTIBLE(
+      { errno = file->ops.read(file->inode, buf, count, file->f_pos); })
+
+    unlock_inode(file->inode);
+
+    if (errno > 0) {
+        file->f_pos += errno;
+        return errno;
+    }
+
+done:
+    return DO_STATUS(errno);
+}
+
+__DEFINE_LXSYSCALL3(int, write, int, fd, void*, buf, size_t, count)
+{
+    int errno = 0;
+    struct v_fd* fd_s;
+    if ((errno = __vfs_getfd(fd, &fd_s))) {
+        goto done;
+    }
+
+    struct v_file* file = fd_s->file;
+    if ((file->inode->itype & VFS_IFDIR)) {
+        errno = EISDIR;
+        goto done;
+    }
+
+    lock_inode(file->inode);
+
+    file->inode->mtime = clock_unixtime();
+
+    __SYSCALL_INTERRUPTIBLE(
+      { errno = file->ops.write(file->inode, buf, count, file->f_pos); })
+
+    unlock_inode(file->inode);
+
+    if (errno > 0) {
+        file->f_pos += errno;
+        return errno;
+    }
+
+done:
+    return DO_STATUS(errno);
+}
+
+__DEFINE_LXSYSCALL3(int, lseek, int, fd, int, offset, int, options)
+{
+    int errno = 0;
+    struct v_fd* fd_s;
+    if ((errno = __vfs_getfd(fd, &fd_s))) {
+        goto done;
+    }
+
+    struct v_file* file = fd_s->file;
+
+    lock_inode(file->inode);
+
+    size_t fpos = file->f_pos;
+    switch (options) {
+        case FSEEK_CUR:
+            fpos = (size_t)((int)file->f_pos + offset);
+            break;
+        case FSEEK_END:
+            fpos = (size_t)((int)file->inode->fsize + offset);
+            break;
+        case FSEEK_SET:
+            fpos = offset;
+            break;
+    }
+    if (!file->ops.seek || !(errno = file->ops.seek(file->inode, fpos))) {
+        file->f_pos = fpos;
+    }
+
+    unlock_inode(file->inode);
+
+done:
+    return DO_STATUS(errno);
+}
+
+int
+vfs_get_path(struct v_dnode* dnode, char* buf, size_t size, int depth)
+{
+    if (!dnode) {
+        return 0;
+    }
+
+    if (depth > 64) {
+        return ELOOP;
+    }
+
+    size_t len = vfs_get_path(dnode->parent, buf, size, depth + 1);
+
+    if (len >= size) {
+        return len;
+    }
+
+    size_t cpy_size = MIN(dnode->name.len, size - len);
+    strncpy(buf + len, dnode->name.value, cpy_size);
+    len += cpy_size;
+
+    if (len < size) {
+        buf[len++] = PATH_DELIM;
+    }
+
+    return len;
+}
+
+int
+vfs_readlink(struct v_dnode* dnode, char* buf, size_t size)
+{
+    const char* link;
+    struct v_inode* inode = dnode->inode;
+    if (inode->ops.read_symlink) {
+        lock_inode(inode);
+
+        int errno = inode->ops.read_symlink(inode, &link);
+        strncpy(buf, link, size);
+
+        unlock_inode(inode);
+        return errno;
+    }
+    return 0;
+}
+
+__DEFINE_LXSYSCALL3(int, realpathat, int, fd, char*, buf, size_t, size)
+{
+    int errno;
+    struct v_fd* fd_s;
+    if ((errno = __vfs_getfd(fd, &fd_s))) {
+        goto done;
+    }
+
+    struct v_dnode* dnode;
+    errno = vfs_get_path(fd_s->file->dnode, buf, size, 0);
+
+    if (errno >= 0) {
+        return errno;
+    }
+
+done:
+    return DO_STATUS(errno);
+}
+
+__DEFINE_LXSYSCALL3(int, readlink, const char*, path, char*, buf, size_t, size)
+{
+    int errno;
+    struct v_dnode* dnode;
+    if (!(errno =
+            vfs_walk(__current->cwd, path, &dnode, NULL, VFS_WALK_NOFOLLOW))) {
+        errno = vfs_readlink(dnode, buf, size);
+    }
+
+    if (errno >= 0) {
+        return errno;
+    }
+
+    return DO_STATUS(errno);
+}
+
+__DEFINE_LXSYSCALL4(int,
+                    readlinkat,
+                    int,
+                    dirfd,
+                    const char*,
+                    pathname,
+                    char*,
+                    buf,
+                    size_t,
+                    size)
+{
+    int errno;
+    struct v_fd* fd_s;
+    if ((errno = __vfs_getfd(dirfd, &fd_s))) {
+        goto done;
+    }
+
+    struct v_dnode* dnode;
+    if (!(errno = vfs_walk(
+            fd_s->file->dnode, pathname, &dnode, NULL, VFS_WALK_NOFOLLOW))) {
+        errno = vfs_readlink(fd_s->file->dnode, buf, size);
+    }
+
+    if (errno >= 0) {
+        return errno;
+    }
+
+done:
+    return DO_STATUS(errno);
+}
+
+/*
+    NOTE
+    When we perform operation that could affect the layout of
+    directory (i.e., rename, mkdir, rmdir). We must lock the parent dir
+    whenever possible. This will blocking any ongoing path walking to reach
+    it hence avoid any partial state.
+*/
+
+__DEFINE_LXSYSCALL1(int, rmdir, const char*, pathname)
+{
+    int errno;
+    struct v_dnode* dnode;
+    if ((errno = vfs_walk(__current->cwd, pathname, &dnode, NULL, 0))) {
+        return DO_STATUS(errno);
+    }
+
+    lock_dnode(dnode);
+
+    if (dnode->parent)
+        lock_dnode(dnode->parent);
+
+    if ((dnode->super_block->fs->types & FSTYPE_ROFS)) {
+        errno = EROFS;
+        goto done;
+    }
+
+    if (dnode->ref_count || dnode->inode->open_count) {
+        errno = EBUSY;
+        goto done;
+    }
+
+    if (!llist_empty(&dnode->children)) {
+        errno = ENOTEMPTY;
+        goto done;
+    }
+
+    lock_inode(dnode->inode);
+
+    if ((dnode->inode->itype & VFS_IFDIR)) {
+        errno = dnode->inode->ops.rmdir(dnode->inode);
+        // FIXME remove the dnode from cache & parent.
+    } else {
+        errno = ENOTDIR;
+    }
+
+    unlock_inode(dnode->inode);
+
+done:
+    unlock_dnode(dnode);
+    if (dnode->parent)
+        unlock_dnode(dnode->parent);
+    return DO_STATUS(errno);
 }
 
 __DEFINE_LXSYSCALL1(int, mkdir, const char*, path)
@@ -457,14 +926,19 @@ __DEFINE_LXSYSCALL1(int, mkdir, const char*, path)
     struct v_dnode *parent, *dir;
     struct hstr component = HSTR(valloc(VFS_NAME_MAXLEN), 0);
     int errno =
-      vfs_walk(root_sb->root, path, &parent, &component, VFS_WALK_PARENT);
+      vfs_walk(__current->cwd, path, &parent, &component, VFS_WALK_PARENT);
     if (errno) {
         goto done;
     }
 
-    if (!parent->inode->ops.mkdir) {
+    lock_dnode(parent);
+    lock_inode(parent->inode);
+
+    if ((parent->super_block->fs->types & FSTYPE_ROFS)) {
         errno = ENOTSUP;
-    } else if (!(parent->inode->itype & VFS_INODE_TYPE_DIR)) {
+    } else if (!parent->inode->ops.mkdir) {
+        errno = ENOTSUP;
+    } else if (!(parent->inode->itype & VFS_IFDIR)) {
         errno = ENOTDIR;
     } else {
         dir = vfs_d_alloc();
@@ -477,7 +951,293 @@ __DEFINE_LXSYSCALL1(int, mkdir, const char*, path)
         }
     }
 
+    unlock_inode(parent->inode);
+    unlock_dnode(parent);
+
+done:
+    return DO_STATUS(errno);
+}
+
+int
+__vfs_do_unlink(struct v_dnode* dnode)
+{
+    struct v_inode* inode = dnode->inode;
+
+    if (dnode->ref_count) {
+        return EBUSY;
+    }
+
+    lock_inode(inode);
+
+    int errno;
+    if (inode->open_count) {
+        errno = EBUSY;
+    } else if (!(inode->itype & VFS_IFDIR)) {
+        // The underlying unlink implementation should handle
+        //  symlink case
+        errno = inode->ops.unlink(inode);
+        if (!errno) {
+            inode->link_count--;
+            // FIXME remove the dnode from cache & parent
+        }
+    } else {
+        errno = EISDIR;
+    }
+
+    unlock_inode(inode);
+
+    return errno;
+}
+
+__DEFINE_LXSYSCALL1(int, unlink, const char*, pathname)
+{
+    int errno;
+    struct v_dnode* dnode;
+    if ((errno = vfs_walk(__current->cwd, pathname, &dnode, NULL, 0))) {
+        goto done;
+    }
+    if ((dnode->super_block->fs->types & FSTYPE_ROFS)) {
+        errno = EROFS;
+        goto done;
+    }
+
+    errno = __vfs_do_unlink(dnode);
+
+done:
+    return DO_STATUS(errno);
+}
+
+__DEFINE_LXSYSCALL2(int, unlinkat, int, fd, const char*, pathname)
+{
+    int errno;
+    struct v_fd* fd_s;
+    if ((errno = __vfs_getfd(fd, &fd_s))) {
+        goto done;
+    }
+
+    struct v_dnode* dnode;
+    if (!(errno = vfs_walk(fd_s->file->dnode, pathname, &dnode, NULL, 0))) {
+        errno = __vfs_do_unlink(dnode);
+    }
+
+done:
+    return DO_STATUS(errno);
+}
+
+__DEFINE_LXSYSCALL2(int, link, const char*, oldpath, const char*, newpath)
+{
+    int errno;
+    struct v_dnode *dentry, *to_link, *name_dentry, *name_file;
+
+    errno = __vfs_try_locate_file(oldpath, &dentry, &to_link, 0);
+    if (!errno) {
+        errno = __vfs_try_locate_file(
+          newpath, &name_dentry, &name_file, FLOCATE_CREATE_EMPTY);
+        if (!errno) {
+            errno = EEXIST;
+        } else if (name_file) {
+            errno = vfs_link(to_link, name_file);
+        }
+    }
+    return DO_STATUS(errno);
+}
+
+__DEFINE_LXSYSCALL1(int, fsync, int, fildes)
+{
+    int errno;
+    struct v_fd* fd_s;
+    if (!(errno = __vfs_getfd(fildes, &fd_s))) {
+        errno = vfs_fsync(fd_s->file);
+    }
+
+    return DO_STATUS(errno);
+}
+
+int
+vfs_dup_fd(struct v_fd* old, struct v_fd** new)
+{
+    int errno = 0;
+    struct v_fd* copied = cake_grab(fd_pile);
+
+    memcpy(copied, old, sizeof(struct v_fd));
+
+    atomic_fetch_add(&old->file->ref_count, 1);
+
+    *new = copied;
+
+    return errno;
+}
+
+int
+vfs_dup2(int oldfd, int newfd)
+{
+    if (newfd == oldfd) {
+        return newfd;
+    }
+
+    int errno;
+    struct v_fd *oldfd_s, *newfd_s;
+    if ((errno = __vfs_getfd(oldfd, &oldfd_s))) {
+        goto done;
+    }
+
+    if (!TEST_FD(newfd)) {
+        errno = EBADF;
+        goto done;
+    }
+
+    newfd_s = __current->fdtable->fds[newfd];
+    if (newfd_s && (errno = vfs_close(newfd_s->file))) {
+        goto done;
+    }
+
+    if (!(errno = vfs_dup_fd(oldfd_s, &newfd_s))) {
+        __current->fdtable->fds[newfd] = newfd_s;
+        return newfd;
+    }
+
+done:
+    return DO_STATUS(errno);
+}
+
+__DEFINE_LXSYSCALL2(int, dup2, int, oldfd, int, newfd)
+{
+    return vfs_dup2(oldfd, newfd);
+}
+
+__DEFINE_LXSYSCALL1(int, dup, int, oldfd)
+{
+    int errno, newfd;
+    struct v_fd *oldfd_s, *newfd_s;
+    if ((errno = __vfs_getfd(oldfd, &oldfd_s))) {
+        goto done;
+    }
+
+    if (!(errno = vfs_alloc_fdslot(&newfd)) &&
+        !(errno = vfs_dup_fd(oldfd_s, &newfd_s))) {
+        __current->fdtable->fds[newfd] = newfd_s;
+        return newfd;
+    }
+
+done:
+    return DO_STATUS(errno);
+}
+
+__DEFINE_LXSYSCALL2(int,
+                    symlink,
+                    const char*,
+                    pathname,
+                    const char*,
+                    link_target)
+{
+    int errno;
+    struct v_dnode* dnode;
+    if ((errno = vfs_walk(__current->cwd, pathname, &dnode, NULL, 0))) {
+        goto done;
+    }
+    if ((dnode->super_block->fs->types & FSTYPE_ROFS)) {
+        errno = EROFS;
+        goto done;
+    }
+    if (!dnode->inode->ops.symlink) {
+        errno = ENOTSUP;
+        goto done;
+    }
+
+    lock_inode(dnode->inode);
+
+    errno = dnode->inode->ops.symlink(dnode->inode, link_target);
+
+    unlock_inode(dnode->inode);
+
+done:
+    return DO_STATUS(errno);
+}
+
+int
+__vfs_do_chdir(struct v_dnode* dnode)
+{
+    int errno = 0;
+
+    lock_dnode(dnode);
+
+    if (!(dnode->inode->itype & VFS_IFDIR)) {
+        errno = ENOTDIR;
+        goto done;
+    }
+
+    if (__current->cwd) {
+        __current->cwd->ref_count--;
+    }
+
+    dnode->ref_count--;
+    __current->cwd = dnode;
+
+    unlock_dnode(dnode);
+
+done:
+    return errno;
+}
+
+__DEFINE_LXSYSCALL1(int, chdir, const char*, path)
+{
+    struct v_dnode* dnode;
+    int errno = 0;
+
+    if ((errno = vfs_walk(__current->cwd, path, &dnode, NULL, 0))) {
+        goto done;
+    }
+
+    errno = __vfs_do_chdir(dnode);
+
+done:
+    return DO_STATUS(errno);
+}
+
+__DEFINE_LXSYSCALL1(int, fchdir, int, fd)
+{
+    struct v_fd* fd_s;
+    int errno = 0;
+
+    if ((errno = __vfs_getfd(fd, &fd_s))) {
+        goto done;
+    }
+
+    errno = __vfs_do_chdir(fd_s->file->dnode);
+
+done:
+    return DO_STATUS(errno);
+}
+
+__DEFINE_LXSYSCALL2(char*, getcwd, char*, buf, size_t, size)
+{
+    int errno = 0;
+    char* ret_ptr = 0;
+    if (size < 2) {
+        errno = ERANGE;
+        goto done;
+    }
+
+    size_t len = 0;
+
+    if (!__current->cwd) {
+        *buf = PATH_DELIM;
+        len = 1;
+    } else {
+        len = vfs_get_path(__current->cwd, buf, size, 0);
+        if (len == size) {
+            errno = ERANGE;
+            goto done;
+        }
+    }
+
+    buf[len + 1] = '\0';
+
+    ret_ptr = buf;
+
 done:
     __current->k_status = errno;
-    return SYSCALL_ESTATUS(errno);
-}
\ No newline at end of file
+    return ret_ptr;
+}
+
+// TODO rename syscall
\ No newline at end of file