linux nasm打印多个字符

Question

I am trying to write a program that will allow me to print multiple characters (strings of characters or integers). The problem that I am having is that my code only prints one of the characters, and then newlines and stays in an infinite loop. Here is my code:

SECTION .data
len EQU 32
SECTION .bss
num resb len
output resb len

SECTION .text
GLOBAL _start
_start:
Read:
    mov eax, 3
    mov ebx, 1
    mov ecx, num
    mov edx, len
    int 80h

Point:
    mov ecx, num

Print:
    mov al, [ecx]
    inc ecx
    mov [output], al

    mov eax, 4
    mov ebx, 1
    mov ecx, output
    mov edx, len
    int 80h

    cmp al, 0
    jz Exit

Clear:
    mov eax, 0
    mov [output], eax
    jmp Print  

Exit:
    mov eax, 1
    mov ebx, 0
    int 80h 

Could someone point out what I am doing wrong?

Thanks,

Rileyh

Answer 1

In the first time you enter the Print section, ecx is pointing to the start of the string and you use it to copy a single character to the start of the output string. But a few more instructions down, you overwrite ecx with the pointer to the output string, and never restore it, therefore you never manage to copy and print the rest of the string.

Also, why are you calling write() with a single character string with the aim to loop over it to print the entire string? Why not just pass num directly in instead of copying a single character to output and passing that?

Answer 2

In your last question, you showed message as a zero-terminated string, so cmp al, 0 would indicate the end of the string. sys_read does NOT create a zero-terminated string! (we can stuff a zero in there if we need it - eg as a filename for sys_open) sys_read will read a maximum of edx characters. sys_read from stdin returns when, and only when, the "enter" key is hit. If fewer than edx characters were entered, the string is terminated with a linefeed character (10 decimal or 0xA or 0Ah hex) - you could look for that... But, if the pesky user types more than edx characters, only edx characters go into your buffer, the "excess" remains in the OS's buffer (and can cause trouble later!). In this case your string is NOT terminated with a linefeed, so looking for it will fail. sys_read returns the number of characters actually read - up to edx - including the linefeed - in eax. If you don't want to include the linefeed in the length, you can decrement eax.

As an experiment, do a sys_read with some small number (say 4) in edx, then exit the program. Type "abcdls"(enter) and watch the "ls" be executed. If some joker typed "abcdrm -rf ."... well, don't!!!

Safest thing is to flush the OS's input buffer.

  mov ecx, num mov edx, len mov ebx, 1 mov eax, 3 int 80h cmp byte [ecx + eax - 1], 10 ; got linefeed? push eax ; save read length - doesn't alter flags je good flush: mov ecx, dummy_buf mov edx, 1 mov ebx, 1 mov eax, 3 int 80h cmp byte [ecx], 10 jne flush good: pop eax ; restore length from first sys_read 

Instead of defining dummy_buf in .bss (or .data), we could put it on the stack - trying to keep it simple here. This is imperfect - we don't know if our string is linefeed-terminated or not, and we don't check for error (unlikely reading from stdin). You'll find you're writing much more code dealing with errors and "idiot user" input than "doing the work". Inevitable! (it's a low-level language - we've gotta tell the CPU Every Single Thing!)

sys_write doesn't know about zero-terminated strings, either! It'll print edx characters, regardless of how much garbage that might be. You want to figure out how many characters you actually want to print, and put that in edx (that's why I saved/restored the original length above).

You mention "integers" and use num as a variable name. Neither of these functions know about "numbers" except as ascii codes. You're reading and writing characters. Converting a single-digit number to and from a character is easy - add or subtract '0' (48 decimal or 30h). Multiple digits are more complicated - look around for an example, if that's what you need.

Best, Frank