[PATCH 2/2] tools: perf: tests: Fix code reading for riscv

[Ian Rogers]

On Mon, Dec 16, 2024 at 3:13 PM Charlie Jenkins <charlie at rivosinc.com> wrote:
>
> After binutils commit e43d876 which was first included in binutils 2.41,
> riscv no longer supports dumping in the middle of instructions. Increase
> the objdump window by 2-bytes to ensure that any instruction that sits
> on the boundary of the specified stop-address is not cut in half.
>
> Signed-off-by: Charlie Jenkins <charlie at rivosinc.com>
> ---
>  arch/riscv/Kconfig              |  5 +++++
>  tools/perf/tests/code-reading.c | 17 ++++++++++++++++-

Files under tools use a different Build system than the kernel. The
Kconfig value won't have an effect. Check out Makefile.config:
https://git.kernel.org/pub/scm/linux/kernel/git/perf/perf-tools-next.git/tree/tools/perf/Makefile.config?h=perf-tools-next
which is included into the build here:
https://git.kernel.org/pub/scm/linux/kernel/git/perf/perf-tools-next.git/tree/tools/perf/Makefile.perf?h=perf-tools-next#n313

>  2 files changed, 21 insertions(+), 1 deletion(-)
>
> diff --git a/arch/riscv/Kconfig b/arch/riscv/Kconfig
> index d4a7ca0388c071b536df59c0eb11d55f9080c7cd..f164047471267936bc62389b7d7d9a7cbdca8f97 100644
> --- a/arch/riscv/Kconfig
> +++ b/arch/riscv/Kconfig
> @@ -229,6 +229,11 @@ config GCC_SUPPORTS_DYNAMIC_FTRACE
>         def_bool CC_IS_GCC
>         depends on $(cc-option,-fpatchable-function-entry=8)
>
> +config RISCV_OBJDUMP_SUPPORTS_SPLIT_INSTRUCTION
> +       # Some versions of objdump do not support dumping partial instructions
> +       def_bool y
> +       depends on !(OBJDUMP_IS_GNU && OBJDUMP_VERSION > 24100)
> +
>  config HAVE_SHADOW_CALL_STACK
>         def_bool $(cc-option,-fsanitize=shadow-call-stack)
>         # https://github.com/riscv-non-isa/riscv-elf-psabi-doc/commit/a484e843e6eeb51f0cb7b8819e50da6d2444d769
> diff --git a/tools/perf/tests/code-reading.c b/tools/perf/tests/code-reading.c
> index 27c82cfb7e7de42284bf5af9cf7594a3a963052e..605f4a8e1dbc00d8a572503f45053c2f30ad19e3 100644
> --- a/tools/perf/tests/code-reading.c
> +++ b/tools/perf/tests/code-reading.c
> @@ -1,5 +1,6 @@
>  // SPDX-License-Identifier: GPL-2.0
>  #include <errno.h>
> +#include <linux/kconfig.h>
>  #include <linux/kernel.h>
>  #include <linux/types.h>
>  #include <inttypes.h>
> @@ -183,9 +184,23 @@ static int read_via_objdump(const char *filename, u64 addr, void *buf,
>         const char *fmt;
>         FILE *f;
>         int ret;
> +       u64 stop_address = addr + len;
> +
> +       if (IS_ENABLED(__riscv) && !IS_ENABLED(CONFIG_RISCV_OBJDUMP_SUPPORTS_SPLIT_INSTRUCTION)) {

It would be nice if this could be a runtime rather than build time detected.

Thanks,
Ian

> +               /*
> +                * On some versions of riscv objdump, dumping in the middle of
> +                * instructions is not supported. riscv instructions are aligned along
> +                * 2-byte intervals and can be either 2-bytes or 4-bytes. This makes it
> +                * possible that the stop-address lands in the middle of a 4-byte
> +                * instruction. Increase the stop_address by two to ensure an
> +                * instruction is not cut in half, but leave the len as-is so only the
> +                * expected number of bytes are collected.
> +                */
> +               stop_address += 2;
> +       }
>
>         fmt = "%s -z -d --start-address=0x%"PRIx64" --stop-address=0x%"PRIx64" %s";
> -       ret = snprintf(cmd, sizeof(cmd), fmt, test_objdump_path, addr, addr + len,
> +       ret = snprintf(cmd, sizeof(cmd), fmt, test_objdump_path, addr, stop_address,
>                        filename);
>         if (ret <= 0 || (size_t)ret >= sizeof(cmd))
>                 return -1;
>
> --
> 2.34.1
>