The kernel of the decompression process steps

1、Differences between various kernel images

参考博客：《vmlinuz/vmlinux、Image、zImage与uImage的区别》;

2、Why does the kernel need to decompress

(1)Decompression is required because compression was done when making the kernel image,The compressed image cannot be directly executed,内核镜像 = uncompressed header + 压缩的内核;
(2)The main function of the uncompressed header is to decompress the compressed kernel image and place it at a suitable address in memory,Then call the decompressed kernel;
(3)The reason for compressing the kernel is that the uncompressed kernel is too large,Compression saves storage space.
(4)The cost of saving space is that the kernel boot process is more complicated,Decompression also causes the kernel to take longer to boot.In today's more complex embedded devices,flashThere are dozens of them at every turnG,The few megabytes of space saved by compressing the kernel don't really matter,
So personally I think the compression kernel is rightflashSmall equipment is cost-effective,对flashLarge equipment is not cost-effective,但是为了统一,All are compressed.

3、ubootPreliminary work to start the kernel

(1)Relocate the kernel image into memory,Parse the header of the image,Find the entry address of the kernel image;
(2)Get the machine code ready,tagparameter address ordtb数据的地址;
(3)call kernel entry,Pass the data prepared in the previous step to the kernel;

4、Unzip the files involved

(1)以armTake the chip of the architecture as an example："arch/arm/boot/comperssed/"目录;
(2)Analyze the program entry of the mirror and the composition structure of the mirror,查看vmlinux.lds;
(3)The entire mirror program entry is inhead.S文件中;

5、Confirm the storage address of the decompressed kernel

5.1、汇编代码

#ifdef CONFIG_AUTO_ZRELADDR
		mov	r4, pc	//当前PC的值保存到r4寄存器
		and	r4, r4, #0xf8000000		//只保留高5位
		/* Determine final kernel image address. */
		add	r4, r4, #TEXT_OFFSET	//加上偏移量#TEXT_OFFSET
#else
		ldr	r4, =zreladdr	//Specify the unpacked kernel address in the configuration file
#endif

5.2、CONFIG_AUTO_ZRELADDR宏

(1)The function of the above assembly code is to calculate the storage address of the decompressed kernel,然后保存到r4寄存器中;
(2)定义了CONFIG_AUTO_ZRELADDRThe macro representation is based on the currentPCThe value of , calculates the address where the kernel will be stored after decompression,然后存放到r4寄存器;
(3)如果没有定义CONFIG_AUTO_ZRELADDR宏,则把zreladdr变量的值赋值给r4寄存器;

5.3、TEXT_OFFSET宏

(1)The offset where the kernel code is stored relative to the memory start,通常设为 32k (0x8000);
(2)TEXT_OFFSET宏分析"arch/arm/boot/comperssed/Makefile"和"arch/arm/Makefile";

5.4、zreladdr变量

//arch/arm/boot/comperssed/Makefile
ifneq ($(CONFIG_AUTO_ZRELADDR),y)
LDFLAGS_vmlinux += --defsym zreladdr=$(ZRELADDR)
endif

//arch/arm/boot/Makefile
ZRELADDR    := $(zreladdr-y)

//arch/arm/Mach-xxx/ Makefile.boot
在这里指定zreladdr-y的值

The address stored by the kernel after decompression,It is also the starting address of the kernel after the final decompression,可以在配置文件中指定;

6、Kernel decompression function

6.1、汇编代码部分

 * The C runtime environment should now be setup sufficiently.
 * Set up some pointers, and start decompressing.
 *   r4  = kernel execution address
 *   r7  = architecture ID
 *   r8  = atags pointer
 */
		mov	r0, r4	//内核解压后存放的地址
		mov	r1, sp			@ malloc space above stack
		add	r2, sp, #0x10000	@ 64k max
		mov	r3, r7	//architecture ID：机器码
		bl	decompress_kernel	//执行解压zImage为elf格式的可执行kernel镜像
		bl	cache_clean_flush
		bl	cache_off
		mov	r1, r7			@ restore architecture number
		mov	r2, r8			@ restore atags pointer

//函数原型
void decompress_kernel(unsigned long output_start, unsigned long free_mem_ptr_p, \
		unsigned long free_mem_ptr_end_p,
		int arch_id)

decompress_kernel函数传参	含义
output_start	The starting memory address where the data is stored after decompression by the kernel
free_mem_ptr_p	Available memory starting address
free_mem_ptr_end_p	The address of the end of available memory
arch_id	机器码

(1)r4Before this, the memory address stored by the kernel after decompression has been stored,这里将r4寄存器的值赋值给r0;
(2)sp是栈寄存器,r1The value of the register is the current stack address,r2The value of the register is the current stack address+64KB,The effect is to allocate on the stack64KBmemory space to pass todecompress_kernel()函数;
(3)r3registers are assignedr7寄存器的值,是机器码;
(4)decompress_kernel()函数的4Each parameter corresponds respectivelyr0-r3四个寄存器;

6.2、decompress_kernel()函数源码分析

//arch/arm/boot/compressed/piggy.S
	.section .piggydata,#alloc
	.globl	input_data
input_data:
	.incbin	"arch/arm/boot/compressed/piggy_data"
	.globl	input_data_end
input_data_end:

void decompress_kernel(unsigned long output_start, unsigned long free_mem_ptr_p,
						unsigned long free_mem_ptr_end_p, int arch_id)
{
    
	int ret;

	__stack_chk_guard_setup();

	//Assign values ​​to some global variables
	output_data		= (unsigned char *)output_start;	//The memory address stored by the kernel after decompression
	free_mem_ptr		= free_mem_ptr_p;	//The starting address of available memory
	free_mem_end_ptr	= free_mem_ptr_end_p;	//可用内存的结束地址
	__machine_arch_type	= arch_id;	//机器码

	//Decompression module initialization
	arch_decomp_setup();

	putstr("Uncompressing Linux...");
	
	//(The memory address where the compressed kernel image is located, Compressed kernel image size, The storage address of the decompressed image, Error handling function pointer)
	ret = do_decompress(input_data, input_data_end - input_data,
			    output_data, error);
	if (ret)
		error("decompressor returned an error");
	else
		putstr(" done, booting the kernel.\n");
}

(1)do_decompress()The function is the function that actually does the decompression,Only the address and length of the compressed kernel image are required、Decompress the kernel storage address and other information and pass it in,Inside the function, the function corresponding to the decompression method will be called according to the configuration item,The principle is to use the corresponding decompression method for whatever compression method is used;
(2)The start and end addresses where the compressed kernel image is located are stored in input_data、input_data_end变量中,这在piggy.S文件中指定,Specifically, the link script can be analyzed;
(3)piggy_dataIt is the compressed kernel image,The general process of generation：vmlinux->Image->piggy_data->zImage->zImage-dtb;

7、Invoke the unpacked kernel

b	__enter_kernel

__enter_kernel:
		mov	r0, #0			@ must be 0
 ARM(		mov	pc, r4		)	@ call kernel Call the decompressed kernel image
 M_CLASS(	add	r4, r4, #1	)	@ enter in Thumb mode for M class
 THUMB(		bx	r4		)	@ entry point is always ARM for A/R classes

During the kernel decompression stage,The decompressed kernel has been put inr4The memory address where the register executes,这里将r4寄存器的值赋值给PC寄存器,下面就是执行r4寄存器指向的内存地址,That is, the first sentence of the kernel code.