我目前正在编写一些小的内核代码。以下是我从某处的内核项目中复制的内容。它包含将内核加载到内存位置 0x1000 并跳转到位置 0x1000 的代码：
;
; The Bootsector Code (First 512 bytes of the floppy)
;
;
; Define Code Segment and Data Segment Rights details for inputting to GDTFill function
;
%define CS_ACCES 10011011b ; CS and DS Access Rights (Details in GDT.INC)
%define DS_ACCES 10010011b
;
; 16 Bit Addressing initially
;
[bits 16]
;
; Code begins at 0x7c00
;
[org 0x7c00]
;
; Bios Jumps to 0xf000:0xffff
; Then it loads the first 512 bytes (BootSector)
; from first boot device to 0x0000:0x7c00
;
jmp boot
;
; Includes
;
%include "GDT.INC"
;
; Define Stack
;
boot:
mov ax,0x07C0
mov ds,ax
mov es,ax
mov ax,0x8FFF
mov ss,ax
;
; Stack begins at 0xf000 and fills from there downwards
;
mov sp,0xFFFF
;
; Note:
; Linear Address = Shift Segment by 1 byte and add Offset to it
;
; Read Kernel From Floppy to Memory location es:bx (0x1000 here)
; Cylinder Head Sector and Buffer are as follows:
;
; es:bx - buffer where to load Kernel to
; ch - track number
; cl - starting sector
; dh - head number
; dl - drive number (0 here)
; Then call interrupt 0x13
;
xor ax,ax
int 0x13
;
; Do the floppy int 13 reading
;
push es
mov ax,0x100
mov es,ax
mov bx,0
mov ah,2
mov al,30
mov ch,0
mov cl,2
mov dh,0
mov dl,0
int 0x13
;
; Now es holds stack addresss
;
pop es
;
; Fill GDT
; Refer to GDT.INC for details
;
GDTFill 0, 0xFFFFF,CS_ACCES,1101b,gdt_cs
GDTFill 0, 0xFFFFF,DS_ACCES,1101b,gdt_ds
;
; Store Limit of GDT beginning at location marked as gdtptr
; This has to be passed on to lgdt instruction
;
mov ax, gdtend
mov bx, gdt
sub ax,bx
mov word [gdtptr], ax
;
; Store Linear address of GDT at gdtptr after allowing space for the previous data
; Linear Address = Shift Segment by 1 byte and add Offset to it
;
xor ax,ax
mov ax,ds
mov bx,gdt
call LinearAdd
mov dword [gdtptr+2], ecx
;
; Load gdt using lgdt. Disable interrupts before that
;
cli
lgdt[gdtptr]
;
; Move to protected mode. Set cr0's first bit to 1 by or'ing it
;
mov eax,cr0
or ax,1
mov cr0,eax
;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;
;
; Once in Protected Mode,
; Except cs all are defined w.r.t DataSegment
; DataSegment Descriptor from beginning of GDT is 8 bytes
; CodeSegment Descriptor from beginning of GDT is 10 bytes
; Stack (Very important! - This is what i messed with initially:
; --------------------------------------------------------------
; Defined w.r.t Data segment
; Beginning - 0x9f000 (Fills downwards)
; Stack size is = 0x9f000 - DataSegment Value (= 0x0) (Have to change this)
; Quite enough for some small operations and LIBC Functions
;
;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;
jmp next
next:
mov ax,0x10
mov ds,ax
mov es,ax
mov fs,ax
mov gs,ax
mov ss,ax
mov esp,0x9F000 ; Stack begins filling at this address
;
; protected mode segmented address = cs:0x1000
; This is nothing but 0x0:0x1000 in protected mode (Have to change these)
; Which is where kernel was loaded earlier from floppy
;
jmp dword 0x8:0x1000
end:
jmp end ; Just in case it slips the earlier step !
;
; Initially fill GDT with 0's
;
gdt:
gdt_null:
dw 0,0,0,0
gdt_cs:
dw 0,0,0,0
gdt_ds:
dw 0,0,0,0
gdtend:
;
; The following is the GDT Pointer
; This is used for passing it on to LGDT Instruction
;
gdtptr:
dw 0x0000 ; 16 bit size of GDT
dd 0 ; 32 bit linear address of GDT
;
; Filling the rest of space with NOP
; Or else boot sector might start executing invalid instructions there
; Because of Junk Data
;
times 510-($-$$) db 144
;
; The signature of boot sector
;
dw 0xaa55
以下在 GDT.INC 中：
; Calculate Linear Address
; Called with following Values:
; ax - has segment number
; bx - has offset number
;
; Output:
; ecx - has Linear Address
;
; We are in 16 bit mode
;
; Linear Address = Shift Segment by 1 byte and add Offset to it
; So 0x07c0:0x0 = 0x7c00
LinearAdd:
xor ecx,ecx
mov cx,ax
shl ecx,4
and ebx,0x0000FFFF
add ecx,ebx
ret
; Filling Global Descriptor Table
; -------------------------------
;
; Note:
; -----
; 1. Order of variables input: Base (32 bits),
; Limit (20 bits), Access Rights(8 bits), Flags(4 bits), Segment Address
; (32 bits)
;
; 2. Variables input to function are moved LS 4 bits First (Right to Left) to table
;
; 3. While reading LS 4 bits or MS 4 bits, read from left to right
;
; Base:
; -----
; Bits in Table Bits in Field 'Base' Location w.r.t Beginning of Segment
;---------------------------------------------------------------------------------------
; 16 - 31 0 - 15 +2 (2 bytes in length)
; 32 - 39 16 - 23 +4 (1 byte in length)
; 55 - 63 24 - 31 +7 (1 byte in length)
;
;
; Howto:
; ------
;
; Below %5 ie., the fifth variable input to function GDTFill is Beginning of GDT's Code Segment
; [%5+2] represents entry 1 in above table
; [%5+4] represents entry 2 in above table
; [$5+7] represents entry 3 in above table
; Entry 1 is 2 bytes (so ax is moved to word [%5+2]
; Entry 2 and 3 are 1 byte (so al is moved to byte[%5+4] and [%5+7] respectively
;
; Limit:
; ------
;
; Bits in Table Bits in Field 'Limit' Location w.r.t Beginning of Segment
;---------------------------------------------------------------------------------------
; 0 - 15 0 - 15 +0 (2 bytes in length)
; 16 - 20 48 - 51 +6 (1 byte in length)
;
; Refer 'Howto' above for detailed description
;
; Access Rights:
; --------------
;
; Access Rights (Bit 40 to Bit 47) = Type (Bit 40 to Bit 43) + System Flag (Bit 44) + DPL (Bit 45 and 46) + Reserved (Bit 47);
; Type - Your Call (Say A is Kernel Code and B is User code (Here it is 11 and 3 respectively)
; System Flag - Both Code and Data Segment have S Flag = 1
; DPL - Privilege level (Ring 0 or 3?) (Ring 0 - 00 and Ring 3 - 11)
;
; Flags:
; ------
; G B O AVL
;
; G - Granularity = 1 here (means Segment Size is 4096 bytes)
; B - Address offsets used for accessing segments are 32 bits long
; O - 0 (Don't know what it is!)
; AVL - 1 here (You Can Ignore it)
;
;
%macro GDTFill 5
push eax
; Base
mov eax,%1
mov word [%5+2],ax
shr eax,16 ; Shift Right to
mov byte [%5+4],al
shr eax,8
mov byte [%5+7],al
; Limit
mov eax,%2
and eax,0x000FFFFF
mov word [%5],ax ; ecrit (0..15)
shr eax,16 ; place (16..19) sur le nibble inferieur
mov byte [%5+6],0 ; initialise flags+lim(16..19) a 0
or [%5+6],al ; ecrit (16..19)
; flags :
mov al,%4
and al,0x0F
shl al,4
or [%5+6],al
; acces :
mov byte [%5+5],%3
pop eax
%endmacro
以上工作了很长时间。然而，当我的内核开始变大时，数据和代码段重叠了。虽然两者都从 0 开始，但数据段中的数据与代码段中的代码或类似内容重叠。因此，我无法完全打印消息。
有没有办法改变数据段和代码段的基址，以便在基址之间有一些空间来编写一个最大约为 1 MB 的二进制小内核？
我已将链接附加到内核并在下面详细描述问题：
我使用以下方法制作内核：
make clean; make
在 src 文件夹内
使用以下命令在 qemu 中启动它：
sudo qemu-system-i386 -net nic,vlan=0,model=pcnet -net tap,vlan=0,ifname=tap,script=no -fda ../flp/fileb.flp -boot a -m 128启动后，我运行以下命令来验证字符串操作：
testnum然后我运行以下命令，我看到消息在中间被剥离并且在某一行之后没有打印：
pcnetops如果我在 console.c 中注释以下行并运行 pcnetops，我会打印所有内容：
print( "sizeof(char) == "); print(htos(sizeof(char))); print(CRLF); print("htos(stoh(ffffffff, LEFT_TO_RIGHT)): "); print(htos(stoh((unsigned char *) "ffffffff", LEFT_TO_RIGHT))); print(CRLF);
这就是为什么我怀疑我是否应该将代码段和数据段基地址分开的原因（现在两者都是 0）
