refactor: synchronize the vfs objects

[lunaix-os.git] / lunaix-os / kernel / fs / vfs.c

diff --git a/lunaix-os/kernel/fs/vfs.c b/lunaix-os/kernel/fs/vfs.c
index 11896ec6730e4377a24d3d4d72b7644622663c51..82f409a6c9996019ccf724fb7ce88ee2f069621e 100644 (file)
--- 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 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;
 static struct cake_pile* dnode_pile;
 static struct cake_pile* inode_pile;
 static struct cake_pile* file_pile;


@@ -160,6 +166,8 @@ __vfs_walk(struct v_dnode* start,
             goto cont;
         }
 
             goto cont;
         }
 

+        lock_dnode(current_level);
+

         name_content[j] = 0;
         name.len = j;
         hstr_rehash(&name, HSTR_FULL_HASH);
         name_content[j] = 0;
         name.len = j;
         hstr_rehash(&name, HSTR_FULL_HASH);


@@ -170,6 +178,7 @@ __vfs_walk(struct v_dnode* start,
                 component->len = j;
                 strcpy(component->value, name_content);
             }
                 component->len = j;
                 strcpy(component->value, name_content);
             }

+            unlock_dnode(current_level);

             break;
         }
 
             break;
         }
 


@@ -181,6 +190,8 @@ __vfs_walk(struct v_dnode* start,
 
             strcpy(dnode->name.value, name_content);
 
 
             strcpy(dnode->name.value, name_content);
 

+            lock_inode(current_level->inode);
+

             errno =
               current_level->inode->ops.dir_lookup(current_level->inode, dnode);
 
             errno =
               current_level->inode->ops.dir_lookup(current_level->inode, dnode);
 


@@ -193,7 +204,11 @@ __vfs_walk(struct v_dnode* start,
                 }
             }
 
                 }
             }
 

+            unlock_inode(current_level->inode);
+

             if (errno) {
             if (errno) {

+                unlock_dnode(current_level);
+                vfree(dnode->name.value);

                 goto error;
             }
 
                 goto error;
             }
 


@@ -203,6 +218,8 @@ __vfs_walk(struct v_dnode* start,
             llist_append(&current_level->children, &dnode->siblings);
         }
 
             llist_append(&current_level->children, &dnode->siblings);
         }
 

+        unlock_dnode(current_level);
+

         j = 0;
         current_level = dnode;
     cont:
         j = 0;
         current_level = dnode;
     cont:


@@ -229,7 +246,7 @@ vfs_walk(struct v_dnode* start,
          int options)
 {
     struct v_dnode* interim;
          int options)
 {
     struct v_dnode* interim;

-    char* pathname = path;
+    const char* pathname = path;

     int errno = __vfs_walk(start, path, &interim, component, options);
     int counter = 0;
 
     int errno = __vfs_walk(start, path, &interim, component, options);
     int counter = 0;
 


@@ -263,7 +280,7 @@ vfs_mount(const char* target, const char* fs_name, bdev_t device)
     int errno;
     struct v_dnode* mnt;
 
     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);
     }
 
         errno = vfs_mount_at(fs_name, device, mnt);
     }
 


@@ -276,7 +293,7 @@ vfs_unmount(const char* target)
     int errno;
     struct v_dnode* mnt;
 
     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);
     }
 
         errno = vfs_unmount_at(mnt);
     }
 


@@ -334,13 +351,13 @@ vfs_open(struct v_dnode* dnode, struct v_file** file)
 
     vfile->dnode = dnode;
     vfile->inode = inode;
 
     vfile->dnode = dnode;
     vfile->inode = inode;

-    vfile->ref_count = 1;
+    vfile->ref_count = ATOMIC_VAR_INIT(1);

     vfile->ops = inode->default_fops;
     vfile->ops = inode->default_fops;

-    inode->open_count++;

 
     if ((inode->itype & VFS_IFFILE) && !inode->pg_cache) {
         struct pcache* pcache = vzalloc(sizeof(struct pcache));
         pcache_init(pcache);
 
     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;
     }
 
         inode->pg_cache = pcache;
     }
 


@@ -348,8 +365,12 @@ vfs_open(struct v_dnode* dnode, struct v_file** file)
     if (errno) {
         cake_release(file_pile, vfile);
     } else {
     if (errno) {
         cake_release(file_pile, vfile);
     } else {

+        atomic_fetch_add(&dnode->ref_count, 1);
+        inode->open_count++;
+

         *file = vfile;
     }
         *file = vfile;
     }

+

     return errno;
 }
 
     return errno;
 }
 


@@ -357,14 +378,17 @@ int
 vfs_link(struct v_dnode* to_link, struct v_dnode* name)
 {
     int errno;
 vfs_link(struct v_dnode* to_link, struct v_dnode* name)
 {
     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;
     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;

-        to_link->inode->link_count++;
+        atomic_fetch_add(&to_link->inode->link_count, 1);

     }
     }

+    unlock_inode(to_link->inode);

 
     return errno;
 }
 
     return errno;
 }


@@ -374,10 +398,10 @@ vfs_close(struct v_file* file)
 {
     int errno = 0;
     if (!file->ops.close || !(errno = file->ops.close(file))) {
 {
     int errno = 0;
     if (!file->ops.close || !(errno = file->ops.close(file))) {

-        if (file->inode->open_count) {
-            file->inode->open_count--;
-        }
-        pcache_commit_all(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;
         cake_release(file_pile, file);
     }
     return errno;


@@ -386,14 +410,16 @@ vfs_close(struct v_file* file)
 int
 vfs_fsync(struct v_file* file)
 {
 int
 vfs_fsync(struct v_file* file)
 {

+    lock_inode(file->inode);
+

     int errno = ENOTSUP;
     int errno = ENOTSUP;

-    pcache_commit_all(file);
+    pcache_commit_all(file->inode);

     if (file->ops.sync) {
     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;
 }
 
     return errno;
 }
 


@@ -430,6 +456,10 @@ vfs_d_alloc()
     struct v_dnode* dnode = cake_grab(dnode_pile);
     memset(dnode, 0, sizeof(*dnode));
     llist_init_head(&dnode->children);
     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;
 }
     dnode->name = vfs_empty;
     return dnode;
 }


@@ -448,6 +478,8 @@ vfs_i_alloc()
 {
     struct v_inode* inode = cake_grab(inode_pile);
     memset(inode, 0, sizeof(*inode));
 {
     struct v_inode* inode = cake_grab(inode_pile);
     memset(inode, 0, sizeof(*inode));

+    mutex_init(&inode->lock);
+

     return inode;
 }
 
     return inode;
 }
 


@@ -464,6 +496,17 @@ vfs_i_free(struct v_inode* inode)
 #define DO_STATUS(errno) SYSCALL_ESTATUS(__current->k_status = errno)
 #define DO_STATUS_OR_RETURN(errno) ({ errno < 0 ? DO_STATUS(errno) : errno; })
 
 #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,
 int
 __vfs_try_locate_file(const char* path,
                       struct v_dnode** fdir,


@@ -473,7 +516,8 @@ __vfs_try_locate_file(const char* path,
     char name_str[VFS_NAME_MAXLEN];
     struct hstr name = HSTR(name_str, 0);
     int errno;
     char name_str[VFS_NAME_MAXLEN];
     struct hstr name = HSTR(name_str, 0);
     int errno;

-    if ((errno = vfs_walk(NULL, path, fdir, &name, VFS_WALK_PARENT))) {
+    if ((errno =
+           vfs_walk(__current->cwd, path, fdir, &name, VFS_WALK_PARENT))) {

         return errno;
     }
 
         return errno;
     }
 


@@ -482,25 +526,24 @@ __vfs_try_locate_file(const char* path,
         return errno;
     }
 
         return errno;
     }
 

-    if (!(errno = (*fdir)->inode->ops.create((*fdir)->inode))) {
-        struct v_dnode* file_new;
-        file_new = vfs_d_alloc();
-        file_new->name = HHSTR(valloc(VFS_NAME_MAXLEN), name.len, name.hash);
-        strcpy(file_new->name.value, name_str);
+    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;
 
         *file = file_new;
 

-        llist_append(&(*fdir)->children, &file_new->siblings);
+        vfs_dcache_add(parent, file_new);
+        llist_append(&parent->children, &file_new->siblings);
+    } else {
+        vfree(file_new->name.value);
+        vfs_d_free(file_new);

     }
 
     return errno;
 }
 
     }
 
     return errno;
 }
 

-int
-__file_cached_read(struct v_file* file, void* buf, size_t len, size_t fpos)
-{
-    return pcache_read(file, buf, len, fpos);
-}
-

 int
 vfs_do_open(const char* path, int options)
 {
 int
 vfs_do_open(const char* path, int options)
 {


@@ -508,13 +551,10 @@ vfs_do_open(const char* path, int options)
     struct v_dnode *dentry, *file;
     struct v_file* ofile = 0;
 
     struct v_dnode *dentry, *file;
     struct v_file* ofile = 0;
 

-    errno = __vfs_try_locate_file(path, &dentry, &file, 0);
-
-    if (errno != ENOENT && (options & FO_CREATE)) {
-        errno = dentry->inode->ops.create(dentry->inode);
-    }
+    errno = __vfs_try_locate_file(
+      path, &dentry, &file, (options & FO_CREATE) ? FLOCATE_CREATE_EMPTY : 0);

 
 

-    if (!errno && (errno = vfs_open(file, &ofile))) {
+    if (errno || (errno = vfs_open(file, &ofile))) {

         return errno;
     }
 
         return errno;
     }
 


@@ -543,15 +583,11 @@ __DEFINE_LXSYSCALL2(int, open, const char*, path, int, options)
     return DO_STATUS_OR_RETURN(errno);
 }
 
     return DO_STATUS_OR_RETURN(errno);
 }
 

-#define TEST_FD(fd) (fd >= 0 && fd < VFS_MAX_FD)
-#define GET_FD(fd, fd_s) (TEST_FD(fd) && (fd_s = __current->fdtable->fds[fd]))
-

 __DEFINE_LXSYSCALL1(int, close, int, fd)
 {
     struct v_fd* fd_s;
     int errno = 0;
 __DEFINE_LXSYSCALL1(int, close, int, fd)
 {
     struct v_fd* fd_s;
     int errno = 0;

-    if (!GET_FD(fd, fd_s)) {
-        errno = EBADF;
+    if ((errno = __vfs_getfd(fd, &fd_s))) {

         goto done_err;
     }
 
         goto done_err;
     }
 


@@ -584,9 +620,16 @@ __DEFINE_LXSYSCALL2(int, readdir, int, fd, struct dirent*, dent)
 {
     struct v_fd* fd_s;
     int errno;
 {
     struct v_fd* fd_s;
     int errno;

-    if (!GET_FD(fd, fd_s)) {
-        errno = EBADF;
-    } else if (!(fd_s->file->inode->itype & VFS_IFDIR)) {
+
+    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 =
         errno = ENOTDIR;
     } else {
         struct dir_context dctx =


@@ -600,7 +643,8 @@ __DEFINE_LXSYSCALL2(int, readdir, int, fd, struct dirent*, dent)
             __vfs_readdir_callback(&dctx, vfs_ddot.value, vfs_ddot.len, 0);
         } else {
             dctx.index -= 2;
             __vfs_readdir_callback(&dctx, vfs_ddot.value, vfs_ddot.len, 0);
         } else {
             dctx.index -= 2;

-            if ((errno = fd_s->file->ops.readdir(fd_s->file, &dctx))) {
+            if ((errno = fd_s->file->ops.readdir(inode, &dctx))) {
+                unlock_inode(inode);

                 goto done;
             }
         }
                 goto done;
             }
         }


@@ -608,35 +652,7 @@ __DEFINE_LXSYSCALL2(int, readdir, int, fd, struct dirent*, dent)
         dent->d_offset++;
     }
 
         dent->d_offset++;
     }
 

-done:
-    return DO_STATUS(errno);
-}
-
-__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(NULL, path, &parent, &component, VFS_WALK_PARENT);
-    if (errno) {
-        goto done;
-    }
-
-    if ((parent->super_block->fs->types & FSTYPE_ROFS)) {
-        errno = ENOTSUP;
-    } else if (!parent->inode->ops.mkdir) {
-        errno = ENOTSUP;
-    } else if (!(parent->inode->itype & VFS_IFDIR)) {
-        errno = ENOTDIR;
-    } else {
-        dir = vfs_d_alloc();
-        dir->name = component;
-        if (!(errno = parent->inode->ops.mkdir(parent->inode, dir))) {
-            llist_append(&parent->children, &dir->siblings);
-        } else {
-            vfs_d_free(dir);
-            vfree(component.value);
-        }
-    }
+    unlock_inode(inode);

 
 done:
     return DO_STATUS(errno);
 
 done:
     return DO_STATUS(errno);


@@ -646,8 +662,7 @@ __DEFINE_LXSYSCALL3(int, read, int, fd, void*, buf, size_t, count)
 {
     int errno = 0;
     struct v_fd* fd_s;
 {
     int errno = 0;
     struct v_fd* fd_s;

-    if (!GET_FD(fd, fd_s)) {
-        errno = EBADF;
+    if ((errno = __vfs_getfd(fd, &fd_s))) {

         goto done;
     }
 
         goto done;
     }
 


@@ -657,9 +672,14 @@ __DEFINE_LXSYSCALL3(int, read, int, fd, void*, buf, size_t, count)
         goto done;
     }
 
         goto done;
     }
 

-    cpu_enable_interrupt();
-    errno = file->ops.read(file, buf, count, file->f_pos);
-    cpu_disable_interrupt();
+    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;
 
     if (errno > 0) {
         file->f_pos += errno;


@@ -674,8 +694,7 @@ __DEFINE_LXSYSCALL3(int, write, int, fd, void*, buf, size_t, count)
 {
     int errno = 0;
     struct v_fd* fd_s;
 {
     int errno = 0;
     struct v_fd* fd_s;

-    if (!GET_FD(fd, fd_s)) {
-        errno = EBADF;
+    if ((errno = __vfs_getfd(fd, &fd_s))) {

         goto done;
     }
 
         goto done;
     }
 


@@ -685,9 +704,14 @@ __DEFINE_LXSYSCALL3(int, write, int, fd, void*, buf, size_t, count)
         goto done;
     }
 
         goto done;
     }
 

-    cpu_enable_interrupt();
-    errno = file->ops.write(file, buf, count, file->f_pos);
-    cpu_disable_interrupt();
+    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;
 
     if (errno > 0) {
         file->f_pos += errno;


@@ -702,27 +726,33 @@ __DEFINE_LXSYSCALL3(int, lseek, int, fd, int, offset, int, options)
 {
     int errno = 0;
     struct v_fd* fd_s;
 {
     int errno = 0;
     struct v_fd* fd_s;

-    if (!GET_FD(fd, fd_s)) {
-        errno = EBADF;
-    } else {
-        struct v_file* file = fd_s->file;
-        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, fpos))) {
-            file->f_pos = fpos;
-        }
+    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);
 }
 
     return DO_STATUS(errno);
 }
 


@@ -757,10 +787,15 @@ vfs_get_path(struct v_dnode* dnode, char* buf, size_t size, int depth)
 int
 vfs_readlink(struct v_dnode* dnode, char* buf, size_t size)
 {
 int
 vfs_readlink(struct v_dnode* dnode, char* buf, size_t size)
 {

-    char* link;
-    if (dnode->inode->ops.read_symlink) {
-        int errno = dnode->inode->ops.read_symlink(dnode->inode, &link);
+    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);
         strncpy(buf, link, size);

+
+        unlock_inode(inode);

         return errno;
     }
     return 0;
         return errno;
     }
     return 0;


@@ -770,17 +805,18 @@ __DEFINE_LXSYSCALL3(int, realpathat, int, fd, char*, buf, size_t, size)
 {
     int errno;
     struct v_fd* fd_s;
 {
     int errno;
     struct v_fd* fd_s;

-    if (!GET_FD(fd, fd_s)) {
-        errno = EBADF;
-    } else {
-        struct v_dnode* dnode;
-        errno = vfs_get_path(fd_s->file->dnode, buf, size, 0);
+    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;
     }
 
     if (errno >= 0) {
         return errno;
     }
 

+done:

     return DO_STATUS(errno);
 }
 
     return DO_STATUS(errno);
 }
 


@@ -788,7 +824,8 @@ __DEFINE_LXSYSCALL3(int, readlink, const char*, path, char*, buf, size_t, size)
 {
     int errno;
     struct v_dnode* dnode;
 {
     int errno;
     struct v_dnode* dnode;

-    if (!(errno = vfs_walk(NULL, path, &dnode, NULL, VFS_WALK_NOFOLLOW))) {
+    if (!(errno =
+            vfs_walk(__current->cwd, path, &dnode, NULL, VFS_WALK_NOFOLLOW))) {

         errno = vfs_readlink(dnode, buf, size);
     }
 
         errno = vfs_readlink(dnode, buf, size);
     }
 


@@ -812,69 +849,143 @@ __DEFINE_LXSYSCALL4(int,
 {
     int errno;
     struct v_fd* fd_s;
 {
     int errno;
     struct v_fd* fd_s;

-    if (!GET_FD(dirfd, fd_s)) {
-        errno = EBADF;
-    } else {
-        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 = __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;
     }
 
     }
 
     if (errno >= 0) {
         return errno;
     }
 

+done:

     return DO_STATUS(errno);
 }
 
     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;
 __DEFINE_LXSYSCALL1(int, rmdir, const char*, pathname)
 {
     int errno;
     struct v_dnode* dnode;

-    if ((errno = vfs_walk(NULL, pathname, &dnode, NULL, 0))) {
-        goto done;
+    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->super_block->fs->types & FSTYPE_ROFS)) {
         errno = EROFS;
         goto done;
     }
 

-    if (dnode->inode->open_count) {
+    if (dnode->ref_count || dnode->inode->open_count) {

         errno = EBUSY;
         goto done;
     }
 
         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);
     if ((dnode->inode->itype & VFS_IFDIR)) {
         errno = dnode->inode->ops.rmdir(dnode->inode);

+        // FIXME remove the dnode from cache & parent.

     } else {
         errno = ENOTDIR;
     }
 
     } 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)
+{
+    struct v_dnode *parent, *dir;
+    struct hstr component = HSTR(valloc(VFS_NAME_MAXLEN), 0);
+    int errno =
+      vfs_walk(__current->cwd, path, &parent, &component, VFS_WALK_PARENT);
+    if (errno) {
+        goto done;
+    }
+
+    lock_dnode(parent);
+    lock_inode(parent->inode);
+
+    if ((parent->super_block->fs->types & FSTYPE_ROFS)) {
+        errno = ENOTSUP;
+    } else if (!parent->inode->ops.mkdir) {
+        errno = ENOTSUP;
+    } else if (!(parent->inode->itype & VFS_IFDIR)) {
+        errno = ENOTDIR;
+    } else {
+        dir = vfs_d_alloc();
+        dir->name = component;
+        if (!(errno = parent->inode->ops.mkdir(parent->inode, dir))) {
+            llist_append(&parent->children, &dir->siblings);
+        } else {
+            vfs_d_free(dir);
+            vfree(component.value);
+        }
+    }
+
+    unlock_inode(parent->inode);
+    unlock_dnode(parent);
+

 done:
     return DO_STATUS(errno);
 }
 
 int
 done:
     return DO_STATUS(errno);
 }
 
 int

-__vfs_do_unlink(struct v_inode* inode)
+__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)) {
     int errno;
     if (inode->open_count) {
         errno = EBUSY;
     } else if (!(inode->itype & VFS_IFDIR)) {

-        // TODO handle symbolic link and type other than regular file
+        // The underlying unlink implementation should handle
+        //  symlink case

         errno = inode->ops.unlink(inode);
         if (!errno) {
             inode->link_count--;
         errno = inode->ops.unlink(inode);
         if (!errno) {
             inode->link_count--;

+            // FIXME remove the dnode from cache & parent

         }
     } else {
         errno = EISDIR;
     }
 
         }
     } else {
         errno = EISDIR;
     }
 

+    unlock_inode(inode);
+

     return errno;
 }
 
     return errno;
 }
 


@@ -882,7 +993,7 @@ __DEFINE_LXSYSCALL1(int, unlink, const char*, pathname)
 {
     int errno;
     struct v_dnode* dnode;
 {
     int errno;
     struct v_dnode* dnode;

-    if ((errno = vfs_walk(NULL, pathname, &dnode, NULL, 0))) {
+    if ((errno = vfs_walk(__current->cwd, pathname, &dnode, NULL, 0))) {

         goto done;
     }
     if ((dnode->super_block->fs->types & FSTYPE_ROFS)) {
         goto done;
     }
     if ((dnode->super_block->fs->types & FSTYPE_ROFS)) {


@@ -890,7 +1001,7 @@ __DEFINE_LXSYSCALL1(int, unlink, const char*, pathname)
         goto done;
     }
 
         goto done;
     }
 

-    errno = __vfs_do_unlink(dnode->inode);
+    errno = __vfs_do_unlink(dnode);

 
 done:
     return DO_STATUS(errno);
 
 done:
     return DO_STATUS(errno);


@@ -900,13 +1011,13 @@ __DEFINE_LXSYSCALL2(int, unlinkat, int, fd, const char*, pathname)
 {
     int errno;
     struct v_fd* fd_s;
 {
     int errno;
     struct v_fd* fd_s;

-    if (!GET_FD(fd, fd_s)) {
-        errno = EBADF;
-    } else {
-        struct v_dnode* dnode;
-        if (!(errno = vfs_walk(fd_s->file->dnode, pathname, &dnode, NULL, 0))) {
-            errno = __vfs_do_unlink(dnode->inode);
-        }
+    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:
     }
 
 done:


@@ -935,9 +1046,7 @@ __DEFINE_LXSYSCALL1(int, fsync, int, fildes)
 {
     int errno;
     struct v_fd* fd_s;
 {
     int errno;
     struct v_fd* fd_s;

-    if (!GET_FD(fildes, fd_s)) {
-        errno = EBADF;
-    } else {
+    if (!(errno = __vfs_getfd(fildes, &fd_s))) {

         errno = vfs_fsync(fd_s->file);
     }
 
         errno = vfs_fsync(fd_s->file);
     }
 


@@ -951,7 +1060,8 @@ vfs_dup_fd(struct v_fd* old, struct v_fd** new)
     struct v_fd* copied = cake_grab(fd_pile);
 
     memcpy(copied, old, sizeof(struct v_fd));
     struct v_fd* copied = cake_grab(fd_pile);
 
     memcpy(copied, old, sizeof(struct v_fd));

-    old->file->ref_count++;
+
+    atomic_fetch_add(&old->file->ref_count, 1);

 
     *new = copied;
 
 
     *new = copied;
 


@@ -967,12 +1077,17 @@ vfs_dup2(int oldfd, int newfd)
 
     int errno;
     struct v_fd *oldfd_s, *newfd_s;
 
     int errno;
     struct v_fd *oldfd_s, *newfd_s;

-    if (!GET_FD(oldfd, oldfd_s) || !TEST_FD(newfd)) {
+    if ((errno = __vfs_getfd(oldfd, &oldfd_s))) {
+        goto done;
+    }
+
+    if (!TEST_FD(newfd)) {

         errno = EBADF;
         goto done;
     }
         errno = EBADF;
         goto done;
     }

+

     newfd_s = __current->fdtable->fds[newfd];
     newfd_s = __current->fdtable->fds[newfd];

-    if (newfd_s && (errno = vfs_close(newfd_s))) {
+    if (newfd_s && (errno = vfs_close(newfd_s->file))) {

         goto done;
     }
 
         goto done;
     }
 


@@ -994,14 +1109,17 @@ __DEFINE_LXSYSCALL1(int, dup, int, oldfd)
 {
     int errno, newfd;
     struct v_fd *oldfd_s, *newfd_s;
 {
     int errno, newfd;
     struct v_fd *oldfd_s, *newfd_s;

-    if (!GET_FD(oldfd, oldfd_s)) {
-        errno = EBADF;
-    } else if (!(errno = vfs_alloc_fdslot(&newfd)) &&
-               !(errno = vfs_dup_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;
     }
 
         __current->fdtable->fds[newfd] = newfd_s;
         return newfd;
     }
 

+done:

     return DO_STATUS(errno);
 }
 
     return DO_STATUS(errno);
 }
 


@@ -1014,7 +1132,7 @@ __DEFINE_LXSYSCALL2(int,
 {
     int errno;
     struct v_dnode* dnode;
 {
     int errno;
     struct v_dnode* dnode;

-    if ((errno = vfs_walk(NULL, pathname, &dnode, NULL, 0))) {
+    if ((errno = vfs_walk(__current->cwd, pathname, &dnode, NULL, 0))) {

         goto done;
     }
     if ((dnode->super_block->fs->types & FSTYPE_ROFS)) {
         goto done;
     }
     if ((dnode->super_block->fs->types & FSTYPE_ROFS)) {


@@ -1026,8 +1144,100 @@ __DEFINE_LXSYSCALL2(int,
         goto done;
     }
 
         goto done;
     }
 

+    lock_inode(dnode->inode);
+

     errno = dnode->inode->ops.symlink(dnode->inode, link_target);
 
     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);
 done:
     return DO_STATUS(errno);

-}
\ No newline at end of file
+}
+
+__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 ret_ptr;
+}
+
+// TODO rename syscall
\ No newline at end of file