CODEWITHSUNDEEP
assemblystructoffsetabistat
anon
anon

Reputation:

how to determine field offset in stat struct calling from assembly?

In order to get struct stat describing file one would call *stat family of functions that will fill memory that is pointed to by passed pointer with struct values.

In C we can call POSIX.1-2008 offsetof macro, but it is a macro, not available in assembly.

How to determine size of struct from assembly and how to determine offsets of fields, so that one could extract necessary fields like st_size, st_mode and so on?

Upvotes: 1

Views: 591

Answers (2)

David Foerster
David Foerster

Reputation: 1539

This might be a case where it’s easier to write a C module with a templated inline assembler section which uses references to struct members in its input and output arguments.

struct my_struct {
    int *a;
    int b;
};

int my_function(struct my_struct *p, int other) {
    int result;
    __asm__(
        "add %[pa], %[pb]\n"
        "add %[pb], %[other]\n"
        "some_command %[result], %[pb]"
        : [result] "=rm"(result), [pa] "+rm"(*p->a), [pb] "+r"(p->b)
        : [other] "g"(other)
        : "cc");
    return result;
}

(This inline assembler template syntax is a C extension supported by GCC and maybe Clang.)

Upvotes: 1

Peter Cordes
Peter Cordes

Reputation: 365372

The ABI for the platform determines the rules for C struct padding. For example, x86-64 Linux uses the x86-64 System V ABI.

Hard way: read the ABI and work it out yourself. Links in the tag wiki. (It's not actually that hard, the rules for alignment / padding within structs are quite simple for the x86-64 System V ABI: every member goes in order of declaration, and is naturally aligned. This can require padding. Windows x64 has its own struct-layout rules that don't always produce the same layout.)

Easy way: let the compiler calculate the offsets for you by writing a function that stores into the struct, and look at the asm. e.g.

#include <sys/stat.h>
void foo(struct stat *st) {
    st->st_size = 1;
    st->st_mode = 2;  // different numbers so you know which field is which in the output
}

    mov     QWORD PTR [rdi+48], 1
    mov     DWORD PTR [rdi+24], 2
    ret

(from the Godbolt compiler explorer)

Having the compiler output constants for use in separate .asm / .S

See some other Q&As with answers geared towards automating the process of generating definitions like #define or .equ, or NASM equ:

Upvotes: 3

Related Questions