zero . Preface

In the previous section, we learned the operation when the file has been loaded into memory , So how does the operating system find files on disk and load them into memory ？ This section will cover this .

One 、 disk

Understand the storage of files when they are not opened , We need to know the hardware that stores it ： disk .
In computers, disks belong to the category of peripherals , It is a mechanical device inside the computer .

Disks are divided into mechanical disks and solid-state disks , At present, most laptops are loaded with solid-state disks .
The basic unit of disk is sector , Each sector probably stores 512k The amount of data .
But when we understand disk, we should think of it as a strip structure . That is to regard the circular disk as a rectangle .

meanwhile , We also divide the long disk into areas of different sizes , It's called C disc ,D disc ,E disc .

Most computers have only one disk , Just divide it into different areas .
meanwhile , Because each disk has a large space , Therefore, we need to continue to divide management .

We will D The disk is divided into many groups for management , Its group head has a BootBlock This is related to startup .
And we have to divide and manage each group , We ：

among SuperBlock Is related to the file structure and GroupDescriptorTable Is group related . But these two are not the focus of this article .
The focus of this article is to understand BlockBitmap,inodeBitmap,inodeTable and DataBlocks.

Two 、 File attributes and data storage

1.inodeTable And Datablocks

When we store a file , We should not only store its attribute information, but also store its data information . among InodeTable And Datablocks The attribute information and data information of the file are stored separately .
stay InodeTable There are many small pieces , Every small piece is one inode. It stores the attribute information of a file .

stay Datablocks neutralization inodeTable It's similar , It contains a lot of block unit , every last block The unit stores information of a certain size ：

The operating system uses inode Number to find , Not by file name , The file name is only used by the user layer .Linux The system really identifies a file , It's through inode Number . Every file has a inode Number . At the same time inode The file corresponding to block Number in block .
If it's a catalog file , It's stored in DataBlock The contents in are the file name in the directory and inode The mapping relation of .

2.InodeBitmap And BlockBitmap

InodeBitmap and BlockBitmap Deposit is ,InodeTable and DataBlocks Bitmap . The position of the bit represents the number , The contents of bits indicate whether they are occupied .
such as ： stay InodeBitmap The number in is 0000 1010, That means 0 Number one inode Not occupied , and 1 Number one inode It's already occupied .
Blockmap Empathy .

3. give an example

We can understand the above content by giving an example ：

mkdir tmp
cd tmp
touch hello.c
echo "hello bit" >hello.c
cat hello.c
rm hello.c

touch hello.c： stay InodeBitmap Found unoccupied inode unit , Store the attribute information .
echo “hello bit”>hello.c： see BlockBitp Not used in block unit , Put the rest inode Building mapping , And store the contents of the document .
cat hello.c： adopt tmp Of inode Find and view tmp Of Datablock, Find and hello.c File names form a mapping relationship inode value , Through this inode Find the corresponding hello.c Of documents Datablock, So as to print the contents .
rm hello.c： stay tmp Of DataBlock Find the file name hello.c Corresponding inode, Only need to inodeBitmap It's time to inode The number of positions is determined by 1 Set as 0 that will do . There is no need for inodeTable And DataBlock To operate .

This is why the deletion of files is very fast , Because attributes and data are not deleted , Just delete whether it is effective . And compared with windows Recycle bin of the system , In fact, the files in the recycle bin have not been deleted , Just changed a directory . When we accidentally delete the file , The best way is to do nothing , Avoid content being overwritten .
Empathy ,mv Instructions are actually just moving file names and inode The mapping relation of .

3、 ... and 、 Hard and soft links

stay Linux Software and hardware links under the system , We can understand it as a shortcut .
Use ln Option to create hard and soft links ：

ln -s// Create soft link
ln // Create a hard link

Here we create a about log.txt The soft links log_s.

Create a hard link log_h.

If it's a soft link , Soft link inode And documents inode Is different . Note the soft link creates a new file , It has its own attribute information and data blocks ( What is saved is the path of the file + file name ). Hard link inode And the original file inode It's the same , It indicates that the hard link does not create a file , Instead, a directory is added DataBlock File name and inode Correspondence of .
The previous red box indicates the number of hard links , When we create hard links , Both the hard link and the file become 2. When the value is 0 It means that the file is deleted .
When we create files , The value is 1, But when we create a directory , The value is 2.

This is because test In the catalog . It is also a hard link to the directory . When you create another directory under this directory , The value is 3, This is because … Also said test Catalog .

Four 、 Time attribute of the file

We can go through stat To view the basic properties of the file , Let's build a new one file.txt. And write something into it .

stat file.txt

among access,modify,change It represents the three time attributes of the file .

1.Access

Access Indicates the last access time of the file , But we go through cat Instruction access file discovery , Of documents access Time has not changed .

This is because opening a file is a high-frequency operation , To avoid a large number of repeated refreshes , In the newer Linux The kernel ,Access Will not be refreshed immediately , But there are certain time intervals OS Will automatically update the time .

2.Modify

Modify Indicates the latest modification time , Time is updated in real time .

echo “hello bit”>>file.txt

3.change

The last time the file properties were modified ：

chomd u+x file.txt

Be careful , When modifying file attributes ,Modify Nothing has changed . But when modifying the contents of the file ,change It's possible to change . This is because modifying the contents of a file generally modifies the file properties , For example, adding content will modify the size attribute of the file .

4.Makefile Judge the way of file update

mytest:test.c    
  gcc $^ -o [email protected]    
.PHONY:clean    
clean:    
  rm mytest   

among test.c by mytest Dependent source file . When Makefile Conduct make When , Will check the mytest and test.c Of Modify Time , Normally speaking, it must be test.c Of Modify Time ratio mytest Earlier ( because mytest By test.c To generate ). If we change test.c The content in , that test.c Of Modify Time will be better than mytest Late , here Makefile Will execute the corresponding dependent methods .
We can also verify ： No change test.c In the case of touch(touch An existing file indicates the modification time ), At this point, we can do make 了 .

5、 ... and 、 summary

We use the creation of the entire file , Use and delete to summarize file operations .

1.touch file.txt

create file , First, the operating system is InodeBitmap and BlockBitmap Find not for 1 The number of , Take its subscript as the Inode and block The number of . And set it to 1, among Inode only one ,block There can be multiple .

2.echo “I am the king of Asgard”>>file.txt

Will call echo process , Turn off the standard output first (close(1)) In this process file.txt Files use system call functions open open ,
First, through the Blocktable, find file.txt And inode Mapping . Then find file.txt This file and its contents , And load it into memory , formation file The structure becomes a file of the process .
file.txt Get a file descriptor , Stored in the process files In the structure , This file descriptor is the file descriptor of the original standard output ：1.
The process passes the print function to the file descriptor as 1 The contents printed in the document , At this time, it is to file.txt Print content in .

3.rm file.txt

take InodeBitmap In the said file.txt The number of the file 1 Set as 0.