mirror of https://github.com/dynup/kpatch
create-diff-object: fix kpatch_replace_sections_syms() for non-text
It doesn't make sense to disassemble a data section. That just happened to work by accident. PC-relative offsets only need adjusting when associated with an instruction. Signed-off-by: Josh Poimboeuf <jpoimboe@redhat.com>
This commit is contained in:
parent
61e46094b5
commit
79f45d1b0a
|
@ -1575,17 +1575,18 @@ static void kpatch_replace_sections_syms(struct kpatch_elf *kelf)
|
|||
add_off = 0;
|
||||
break;
|
||||
case X86_64:
|
||||
if (rela->type == R_X86_64_PC32 ||
|
||||
if (!is_text_section(relasec->base) ||
|
||||
rela->type == R_X86_64_64 ||
|
||||
rela->type == R_X86_64_32S)
|
||||
add_off = 0;
|
||||
else if (rela->type == R_X86_64_PC32 ||
|
||||
rela->type == R_X86_64_PLT32) {
|
||||
struct insn insn;
|
||||
rela_insn(relasec, rela, &insn);
|
||||
add_off = (unsigned int)((long)insn.next_byte -
|
||||
(long)relasec->base->data->d_buf -
|
||||
rela->offset);
|
||||
} else if (rela->type == R_X86_64_64 ||
|
||||
rela->type == R_X86_64_32S)
|
||||
add_off = 0;
|
||||
else
|
||||
} else
|
||||
continue;
|
||||
break;
|
||||
default:
|
||||
|
|
Loading…
Reference in New Issue