Kernel link script parsing

The original text can be found in Kernel link script parsing Check out , Subsequent updates will be updated to Personal blog

---

The specific code can be seen in Warehouse , Here, choose a few to talk about .

Output architecture

OUTPUT_ARCH This command is used to specify the system architecture of the output file , What we use here is riscv.

entry point

ENTRY Defines the entry point of the program ,xv6 The default entry point for is _entry. The entry point code is entry.S Inside .

Position counter

Get into SECTIONS after , The default initial address of the current program is 0x0. Here with Linux The file system of , use “.” To represent the current address . The name of this symbol is Position counter (location counter), Indicates the following paragraph / The memory address corresponding to the variable . This means that if you give it directly “.” assignment , It is equivalent to switching the working directory , The relative position of subsequent segments will change .

Here is a simple example

SECTIONS
{
  . = 0x10000;
  .text : { *(.text) }
  . = 0x8000000;
  .data : { *(.data) }
  .bss : { *(.bss) }
}

Get into SECTIONS after , The position counter will be set to 0x10000, So the linker will .text The address of is set to 0x10000. complete .text After setting , The position counter will be changed to 0x8000000,.data The address of will be set to 0x8000000. turn .bss When , It will follow .data paragraph , At this time, the value of the position counter is 0x8000000 add .data The size of the segment .

xv6 Initial address setting

As soon as you enter SECTIONS The script will set the current initial path to 0x80000000. The teacher said in class that this address is qemu Approved address , The first instruction that must be placed is it . Book 2.6 section The saying is 0x0 To 0x80000000 This address is used to place IO equipment , So the address of the first instruction must be 0x80000000.

Segment content parsing

  .text : {
    *(.text .text.*)
    . = ALIGN(0x1000);
    _trampoline = .;
    *(trampsec)
    . = ALIGN(0x1000);
    ASSERT(. - _trampoline == 0x1000, "error: trampoline larger than one page");
    PROVIDE(etext = .);
  }

The segment settings of the kernel are similar , So look at one .text That's all right. .

.text The first expression of the segment *(.text .text.*), Among them * The wildcard . It represents .text and .text.* Segments are placed in the output file .text In the paragraph .

Then there are orders ALIGN(exp), This command will return the position counter to align to the next exp Address of boundary , It does not modify the value of the position counter .

PROVIDE This order is similar to GCC Medium attribute((weak)). Now we only need to understand it simply as defining a symbol .

remainder _trampoline and trampsec The chapter of the page table will cover , Then you can go and understand .

Reference resources

Link script

GNU Link script 0 - Basic introduction to link script