mov指令后缀无效报错？内核内联汇编打印字符A问题求助

Let's dig into why you're getting that assembler error and fix it step by step. Your code has two main issues: mixing AT&T and Intel syntax, plus a small quirk with how the GNU assembler (GAS) handles segment registers in 32-bit Intel mode.

1. Fix the mov ds, ax instruction #

The "invalid instruction suffix for mov'" error on this line happens because in 32-bit Intel mode, GAS can't automatically infer the operand size when moving to a segment register (which is 16-bit). You need to explicitly mark it as a 16-bit operation with the word` keyword:

mov word ds, ax

2. Fix the movb 0,'A' instruction #

Here you're using AT&T syntax conventions (the movb suffix + source-first operand order) but told GCC to use Intel syntax with -masm=intel. In Intel syntax:

• We don't use suffixes like movb; instead, we use byte ptr to specify a byte-sized memory operation.
• The operand order is destination first, source second.

Rewrite that line to match Intel syntax:

mov byte ptr [0], 'A'

This tells the assembler to write the byte value 'A' to memory offset 0 relative to the ds segment (which we set to 0xb800, so the physical address is 0xb8000—the start of text-mode video memory).

Full corrected code #

Putting it all together, your print function should look like this:

void print(){
    asm volatile(
        "mov ax, 0xb800\n"
        "mov word ds, ax\n"
        "mov byte ptr [0], 'A'\n"
    );
}

Quick runtime correctness note #

If you're compiling with -m32, you're targeting 32-bit protected mode. Loading ds directly with 0xb800 won't work here—segment registers hold selectors (not raw base addresses) in protected mode. If you're writing a real-mode kernel, you should compile for 16-bit instead (use -m16 with GCC, though note 16-bit support is limited in modern versions). For protected mode, you'll need to set up a GDT entry for the video memory segment first, then load the corresponding selector into ds.

内容的提问来源于stack exchange，提问作者Clement Poon