On Tue, Nov 26, 2013 at 9:52 AM, Matthew Leach <matthew.le...@arm.com> wrote:
> Hi,
>
> When unwinding the stack, the unwind code checks for two opcodes that
> denote a registrations of a signal handler. This is broken on BE as
> the opcodes will be in the wrong byte-order as insns are always LE.
>
> Add the correct checks when compiling for AArch64 big endian.
>
> This has been tested with the glibc test "tst-backtrace4" and passes
> with this patch applied.
>
> Please note that I don't have commit access, if this is OK could
> someone merge it for me?
>
> Thanks,
> Matt Leach
>
> libgcc/
> 2013-11-26 Matthew Leach <matthew.le...@arm.com>
>
> * config/aarch64/linux-unwind.h (aarch64_fallback_frame_state):
> Check for correct opcodes on BE.
>
> diff --git a/libgcc/config/aarch64/linux-unwind.h
> b/libgcc/config/aarch64/linux-unwind.h
> index fde4d14..9f30b87 100644
> --- a/libgcc/config/aarch64/linux-unwind.h
> +++ b/libgcc/config/aarch64/linux-unwind.h
> @@ -25,6 +25,14 @@
> #include <signal.h>
> #include <sys/ucontext.h>
>
> +#if __AARCH64EB__
> +#define MOVZ_X8_8B 0x681180d2
> +#define SVC_0 0x010000d4
> +#else
> +#define MOVZ_X8_8B 0xd2801168
> +#define SVC_0 0xd4000001
> +#endif
I think this should have a comment about the byte-order of
instructions so it is easier to understand why they are different
between big and little-endian.
Thanks,
Andrew Pinski
> +
> #define MD_FALLBACK_FRAME_STATE_FOR aarch64_fallback_frame_state
>
> static _Unwind_Reason_Code
> @@ -55,7 +63,7 @@ aarch64_fallback_frame_state (struct _Unwind_Context
> *context,
> 0xd2801168 movz x8, #0x8b
> 0xd4000001 svc 0x0
> */
> - if (pc[0] != 0xd2801168 || pc[1] != 0xd4000001)
> + if (pc[0] != MOVZ_X8_8B || pc[1] != SVC_0)
> {
> return _URC_END_OF_STACK;
> }
> --
> 1.7.9.5
>
>
