# The Virtual Filesystem

kernel開給user-space用的file interface.


往下cover掉各種不同的file system or underlying physical medium. 往上給使用者提供open(),read(),write()等等system call，使用者不會察覺就算底下硬體有什麼不一樣，簡單的使用上層的wrapper就可以對file作讀寫。

* 就算底層各種file system如何翻新或變動，上層寫好的程式也不用改


# Unix Filesystems

* files (file_system_type)
* is an ordered string of bytes.
* first byte marks the beginning of the file
* last byte marks the end of the file.
* operations
* read
* write
* create
* delete
* mount
* inodes(index node) (file metadata)
* access permissions
* size
* owner
* creation time
* ......
* directory
* 裡面放一堆files
* subdirectories
* directory自己也是file

* mount points (vfsmount struct)
* Unix把filesystems mount在namepsace
* Windows把namespace拆分成
* C:
* D: (放A片的地方) :)
* superblock
* data structure containing information about the filesytem as a whole.

# VFS Objects and Their Data Structures
(用C語言寫出來的OOP架構，所以以下很多稱objects)

the four primary object types of the VFS are
* superblock
* mounted filesystem
* inode
* specific file
* dentry
* directory entry
* which is a single component of a path
* file
* an open file as associated with a process
* operations object
* super_operations
* invoke on a specific filesytem
* write_inode()
* sync_fs()
* inode_operations
* invoke on a specific file
* create()
* link()
* dentry_operations\
* invoke on a specific directory entry
* d_compare()
* d_delete()
* file_operations
* invoke on an open file
* read()
* write()

# The Superblock Object
(filesystem control block)
存放在disk上某個sector內

Filesystems 不是描述disk-based的
是放給VM(virtual memory based)看得
* ex sysfs
* super_block struct

裡面最重要的是 s_op這個 objects，代表了所有對superblock的操作

# Superblock Operations

裡面全都是function pointer

ex:
a filesystem want to write to its superblock,
sb: a pointer to the filesystem's superblock
這是C語言不support oop所以很醜
(沒有繼承觀念，也沒有override觀念)
in c++就可以直接這樣做了
 (因為有繼承)

In C, there is no way for the method to easily obtain its parent, so you have to pass it

super_operations function實際用途如下

# The Inode Object

inode提供了kernel操作file跟directory的一切資訊
(如果某個filesystem沒有inodes架構，則需要額外跟kernel充分溝通你的file介面長什麼樣，要定protocal來轉換成inode)

inode就是file的描述啦，但inode只會在files are accessed時才會被construct in memory，包括special files, such as device files or pipes.


下面這三種type用union存因為一個file(inode)同一時間只會有一種形態

* i_pipe:
* points to a named pepe data structure
* i_bdev:
* points to a block device structure
* i_cdev:
* points to a character device structure.

# Inode Operations

just like superblock operations
inode operations are like this


operations詳解


# The Dentry Object

directroies 也是一種file
ex: in the path /bin/vi.
bin跟vi都是一個file
* 但bin是special directory file
* vi是 regular file.

linux用dentry來描述path上每一個component
ex: /bin/vi can break into 3 component
* /
* bin
* vi



Dentry State
* used
* 代表一個valid inode
* d_inode point to it
* d_count 指出有多少users use it.
* unused
* 也代表一個valid inode
* d_count = 0
* currently not used.
* negative
* is not associated with a valid inode
* d_inode = NULL
* ex:
* 想像有個daemon持續的open read/write一個現在不在memory的config file，open() system call持續回傳ENOENT, 但kernel不會去constructs the path, walks the on-disk directory structure, verifies the file's inexistence. 因為這個成本還是很高的，我們caching這個negative是值得的

# The File Object

The file object is used to represent a file opened by a process.

* Process是直接跟files object溝通的，而不是透過inodes, dentries, superblocks..
* in-memory representation of an open file.
* 呼叫open()的時候被create
* 呼叫close()的時候被destroy
* 同時間可能會有多個對象存取同一個file
* 所以同時間一個file可能會有多個file objects存在
* 這個objects其實就代表process看一個open file的樣子



Here are the individual operations


ioctl:
* unlocked_ioctl()
* 跟ioctl()一樣
* 但是沒有Big kernel Lock版本
* 所以author要自己做好lock機制
* 因為ioctl用了BKL很沒效率所以現在都規定要用unlocked_ioctl()了
* compat_ioctl()
* 一樣沒用BKL
* provide 32-bits compatible ioctl method for 64-bits systems.
* 要注意portable問題

# Data Structures Associated with Filesystems

因為support很多不同的filesystems. kernel必須有一個特別的結構來描述每種filesystems的能力跟行為

這個用來描述不同filesystem間的不同

這個用來描述mounted instance of a filesystem

他會追蹤filesystem跟所有的mount points之間的關係

# Data Structures Associated with a Process

每個process都會有自己的一份
* list of open files
* root filesystem
* current working directory
* mount points
* ...

VFS跟process之間的3種structure
* files_struct
* task_struct裡的files point to here
* 
* 

* fs_struct
* 

* namespace
* 
*