118 lines
2.7 KiB
ArmAsm
118 lines
2.7 KiB
ArmAsm
/*
|
|
# Segment registers
|
|
|
|
Show how most segment registers work in 16-bit real mode.
|
|
|
|
Expected outcome: 'A' character gets printed 6 times to screen.
|
|
|
|
On failure, trash is shows.
|
|
|
|
I think their goal was to implement process virtualization in the past.
|
|
|
|
The special semantics of other registers will be covered in other files.
|
|
|
|
Rationale of the registers:
|
|
|
|
- extend other registers. For e.g., http://stackoverflow.com/questions/17777146/what-is-the-purpose-of-cs-and-ip-registers-intel-8086
|
|
- rudimentary virtual process spaces
|
|
|
|
## ES
|
|
|
|
TODO: this does seem to have special properties as used by string instructions.
|
|
|
|
- BIOS calls: `int 13h` disk read, `int 15` memory detection
|
|
|
|
## FS
|
|
|
|
## GS
|
|
|
|
FS and GS are general purpose: they don't generate or are affected implicitly by instructions.
|
|
|
|
## Vs protected mode
|
|
|
|
In protected mode, the segment registers point to entries on the GDT:
|
|
|
|
- http://reverseengineering.stackexchange.com/questions/2006/how-are-the-segment-registers-fs-gs-cs-ss-ds-es-used-in-linux
|
|
- http://wiki.osdev.org/Segmentation
|
|
|
|
Those entries still contain an analogous offset to the real mode offset,
|
|
but also much more data, like segment width and permission.
|
|
|
|
## How they work
|
|
|
|
What they do is simple: the full addressing syntax is:
|
|
|
|
%segment:a(b, c, d)
|
|
|
|
and the final address is calculated at:
|
|
|
|
%segment * 16 + a + b * c + d
|
|
|
|
So if we set a segment to 1, it just adds 16 to addresses.
|
|
|
|
## Instruction encoding
|
|
|
|
The command:
|
|
|
|
objdump -D -b binary -m i8086 segment_registers_real.img
|
|
|
|
Shows that non ds encodings are achieved through a prefix, except for `ds`:
|
|
|
|
20: a0 63 7c mov 0x7c63,%al
|
|
34: 26 a0 63 7c mov %es:0x7c63,%al
|
|
40: 64 a0 63 7c mov %fs:0x7c63,%al
|
|
4c: 65 a0 63 7c mov %gs:0x7c63,%al
|
|
58: 36 a0 63 7c mov %ss:0x7c63,%al
|
|
|
|
This makes `ds` the most efficient one for data access, and thus a good default.
|
|
*/
|
|
|
|
#include "common.h"
|
|
BEGIN
|
|
CLEAR
|
|
|
|
/*
|
|
It is not possible to encode moving immediates
|
|
to segment registers: we must either:
|
|
|
|
- pass through a general register ax
|
|
- pop from the stack
|
|
*/
|
|
mov $1, %ax
|
|
mov %ax, %ds
|
|
mov %ds:msg, %al
|
|
PUTC <%al>
|
|
|
|
/*
|
|
%ds is the default segment for GAS memory operations
|
|
if we don't write it explicitly.
|
|
*/
|
|
mov msg, %al
|
|
PUTC <%al>
|
|
|
|
mov $1, %ax
|
|
mov %ax, %es
|
|
mov %es:msg, %al
|
|
PUTC <%al>
|
|
|
|
mov $1, %ax
|
|
mov %ax, %fs
|
|
mov %fs:msg, %al
|
|
PUTC <%al>
|
|
|
|
mov $1, %ax
|
|
mov %ax, %gs
|
|
mov %gs:msg, %al
|
|
PUTC <%al>
|
|
|
|
mov $1, %ax
|
|
mov %ax, %ss
|
|
mov %ss:msg, %al
|
|
PUTC <%al>
|
|
|
|
hlt
|
|
msg:
|
|
/* We push the correct A forward 16 bytes in memory to compensate for the segments. */
|
|
.fill 0x10
|
|
.byte 'A'
|