Re: [PATCH v11 05/19] perf print_insn: Use e_machine for fallback IP length check

[sashiko-bot@kernel.org]

Thank you for your contribution! Sashiko AI review found 1 potential issue(s) to consider:
- [Medium] is64bitip() incorrectly falls back to the host architecture via uname() for guest machines, breaking instruction decoding bitness for 32-bit guest analysis on x86_64 hosts.
--

commit 57b90f6c4edae48460ac4aff9f63c175178b0004
Author: Ian Rogers <irogers@google.com>

perf print_insn: Use e_machine for fallback IP length check

This patch avoids string comparisons with perf_env arch, switching to
using the more precise ELF machine. It also sorts header files and fixes
missing definitions.

> diff --git a/tools/perf/util/print_insn.c b/tools/perf/util/print_insn.c
> index 02e6fbb8ca04e..4068436f26ea4 100644
> --- a/tools/perf/util/print_insn.c
> +++ b/tools/perf/util/print_insn.c

[ ... ]

> @@ -33,13 +38,13 @@ size_t sample__fprintf_insn_raw(struct perf_sample *sample, FILE *fp)
>  static bool is64bitip(struct machine *machine, struct addr_location *al)
>  {
>  	const struct dso *dso = al->map ? map__dso(al->map) : NULL;
> +	uint16_t e_machine;
>  
>  	if (dso)
>  		return dso__is_64_bit(dso);
>  
> -	return machine__is(machine, "x86_64") ||
> -		machine__normalized_is(machine, "arm64") ||
> -		machine__normalized_is(machine, "s390");
> +	e_machine = perf_env__e_machine(machine->env, /*e_flags=*/NULL);
> +	return e_machine == EM_X86_64 || e_machine == EM_AARCH64 || e_machine == EM_S390;

[Severity: Medium]
Does this change break instruction decoding bitness for guest machines?

For guest machines, machine->env is often NULL. The previous code
handled this gracefully because machine__is() checks perf_env__raw_arch(),
which returns "unknown" when the environment is NULL, safely returning
false for the 64-bit check on x86_64 hosts.

With the new logic, passing a NULL env to perf_env__e_machine() delegates
to perf_env__e_machine_nocache(), which calls uname() to retrieve the
local host architecture.

If analyzing a 32-bit guest trace on a 64-bit x86_64 host without a valid
DSO present, this fallback would return EM_X86_64. Could this cause
is64bitip() to return true, improperly decoding the 32-bit guest
instructions with 64-bit bitness?

>  }

-- 
Sashiko AI review · https://sashiko.dev/#/patchset/20260602062452.2583619-1-irogers@google.com?part=5