【ARM】CODE RO RW ZI 程序分布

Relationship between the default armclang-generated sections and scatter-loading input sections

How the default sections that armclang generates relate to the sections in an image depends on the attributes of the sections. Without a scatter file, the scatter-loading mechanism maps the sections using default input section selectors. However, you can modify the mapping with a scatter file.

The following table shows the relationship between armclang-generated sections and scatter-loading input sections:

Table 1. Relationship between the default armclang-generated sections and input section selectors

Default section names that armclang generates	Input section selectors for scatter-loading	Corresponding execution regions for scatter-loading without a scatter file	Example
.bss	.bss, ZI, or BSS	ER_ZI	Zero-Initialized data int b=0; and uninitialized data int a;
.data	.data, RW, RW-DATA, or DATA	ER_RW	Read/write data int z=5;
.constdata	.constdata, RO, RO-DATA, or CONST	ER_RO	Read-only constants.
.rodata	.rodata, RO, RO-DATA, or CONST	ER_RO	Read-only numerical constants int const x=5
.rodata.str1.1	.rodata.str1.1, RO, RO-DATA, or CONST	ER_RO	String contained in printf("const string");
.text	.text, RO, RO-CODE, or CODE	ER_RO	Source that translates to instructions.
.text with SHF_ARM_NOREAD flag	XO	ER_XO	Source code built with -mexecute_only
.llvmbc (LLVM bitcode)	RO-CODE, or CODE	ER_RO	Section containing bitcode generated by the link time optimizer.

Note

This table shows the default section names that armclang generates. You can create sections with different names using the __attribute__((section("name"))) function and variable attribute, for example.

armlink prioritizes the most specific selector first, with no ambiguity allowed. The input section selector .rodata* also selects .rodata.str1.1. Specifying both *(.rodata*) and *(.rodata.str1.*) matches *(.rodata.str1.*) sections then any remaining RO data sections with *(.rodata*). For more information, see How the linker resolves multiple matches when processing scatter files.

Example

The following example shows the placement of code and data, with default section names and user-specified section names:

int x1 = 5;                                                // in .data.x1 (default) - RW
int y1[100];                                               // in .bss.y1 (default) - ZI
int const z1[3] = {1,2,3};                                 // in .rodata.z1 (default) - RO
 
int x2 __attribute__((section("foo"))) = 5;                // in foo (data part of region) - RW
int y2[100];                                               // in .bss.y2 (default) - ZI
int const z2[3] __attribute__((section("bar"))) = {1,2,3}; // in bar - RW
char *s2 __attribute__((section("foo"))) = "abc";          // s2 in foo, "abc" in .rodata.str1.1 - RO
 
int x3 __attribute__((section("foo"))) = 5;                // in foo (data part of region) - RW
int y3[100];                                               // in .bss.y3 (default) - RW
int const z3[3] = {1,2,3};                                 // in .rodata.z3 (default) - RO
char *s3 __attribute__((section("foo"))) = "def";          // s3 in foo, "def" in .rodata.str1.1 - RW
 
int add1(int x) __attribute__((section("foo")));
int add1(int x)                                            // in foo (code part of region) - CODE
{
    return x+1;
}

 

【来源】