Dynamic Loading of Cortex-M3 1 (address-independent code implementation)

This article explains how to implement address independence. Then, you can check the Assembly and CPU instruction manual and analyze and solve your doubts.

The -- acps/ropi option needs to be set when compiling C code, as shown in the following example:

SystemInit (fun_for_sub (j =; j >=; j -- main ((}C-example

 

Compile:

armcc  -c --cpu Cortex-M3 -O0 --apcs=interwork --apcs /ropi/rwpi -o main.o main.c

 

Use fromelf to view assembly code

fromelf.exe -s -c main.o

 

The assembly code generated by text segments is as follows:

** Section Size: 14 bytes (alignment 2 Address: 0x00000000 0x00000000: 4770 0x00000002: 4770 0x00000004 0x00000006: f7fffffe .... BL fun_for_sub; 0x2 Section 0x0000000a: 205a z movs r0, 0x0000000c: bd00 .. POP {pc}Assembly instructions

 

View the compilation code for calling the function fun_for_sub:

0x00000006:    f7fffffe       ....    BL       fun_for_sub ; 0x2 Section 

 

Find the arm ddi0403d_arm_ubunture_v7m_reference_manual_errata_markup_00000.pdf. The explanation of the BL command is as follows:

Branch with Link (immediate) calls a subroutine at a PC-relative address.

We know that BL is a PC-related instruction.

For details, refer to the composition of the BL command:

 

 

According to the command f7fffffe,

Corresponding:

f7ff  :  15  14  13  12  11  10  9  8  7  6  5  4  3  2  1        1   1   1   1   0   1  1  1  1  1  1  1  1  1  1

fffe  :  15  14  13  12  11  10  9  8  7  6  5  4  3  2  1        1   1   1   1   1   1  1  1  1  1  1  1  1  1  0

 

Symbol bit S = 1, J1 = 1, J2 = 1, imm10 = 11 1111 1111, imm11 = 111 1111 1110

 

So I1 =! (J1 ~ S) = 1, I2 =! (J2 ~ S) = 1,

Imm32 = SignExtend (S: I1: I2: imm10: imm11: '0', 32) = SignExtend (1: 1: 1: 11 1111 11: 111 1111 1110: '0 ', 32) = 1111 1111 1111 1111 1111 1111 1111 1100 = 0 xfffffffc.

0xfffffffc is the complement code of-4, and the current PC is 0x00000006,

Then, based on the last step of the above Operation, BranchWritePC (PC + imm32)

The final jump jumps to the address 0x6 + (-4) = 0x2, that is, the address of the function fun_for_sub. Therefore, the address-independent code is implemented based on the current PC.

Similar principles are also applied in the X86 platform, which is based on PC-related jump commands. Programmer self-cultivation-links, loads and libraries.