[Bug ld/27119] ld improperly relocates function address, creating an invalid pointer

[me at rhysre dot net]

https://sourceware.org/bugzilla/show_bug.cgi?id=27119

--- Comment #2 from Rhys Rustad-Elliott <me at rhysre dot net> ---
(In reply to H.J. Lu from comment #1)
> Please try
> 
> [hjl@gnu-cfl-2 pr27119]$ cat link.lds 
> OUTPUT_FORMAT("elf64-x86-64")
> OUTPUT_ARCH(i386:x86-64)
> ENTRY(_start)
> 
> SECTIONS
> {
>   . = 0x0000000001000000;
>   .text : {
>     *(.text)
>   }
>   .data : {
>     *(.data)
>   }
> }
> [hjl@gnu-cfl-2 pr27119]$ make
> gcc -B./ -fPIE   -c -o prog.o prog.c
> gcc -B./ -fPIE   -c -o func.o func.c
> ./ld -pie -o prog prog.o func.o -T link.lds
> objcopy -O binary prog binary
> [hjl@gnu-cfl-2 pr27119]$ radare2 -a x86 -b 64 binary
>  -- Setup dbg.fpregs to true to visualize the fpu registers in the debugger
> view.
> [0x00000000]> pd 12
>             0x00000000      55             push rbp
>             0x00000001      4889e5         mov rbp, rsp
>             0x00000004      488d05070000.  lea rax, [0x00000012]       ;
> "UH\x89\xe5\x90]\xc3"
>             0x0000000b      488945f8       mov qword [rbp - 8], rax
>             0x0000000f      90             nop
>             0x00000010      5d             pop rbp
>             0x00000011      c3             ret
>             0x00000012      55             push rbp
>             0x00000013      4889e5         mov rbp, rsp
>             0x00000016      90             nop
>             0x00000017      5d             pop rbp
>             0x00000018      c3             ret
> [0x00000000]>

Confirming that linking to an ELF and then using objcopy to convert it to a
flat binary works around this issue for me as described here.

I assume regardless that this behavior should still be considered a bug in ld
though right?

-- 
You are receiving this mail because:
You are on the CC list for the bug.