Re: [PATCH v2 14/14] perf tests: Check CoreSight emulated instruction entry and exit

[sashiko-bot@kernel.org]

Thank you for your contribution! Sashiko AI review found 1 potential issue(s) to consider:
- [Medium] perf tests: test flakiness due to page faults on the emulated MRS instruction
--

commit 2d0bd9a9a4d9e3186e0f8fab3a726c18a86eb407
Author: Leo Yan <leo.yan@arm.com>

perf tests: Check CoreSight emulated instruction entry and exit

This commit adds a test to check CoreSight emulated instruction entry and
exit by recording traploop's EL0 read of ID_AA64ISAR0_EL1. It uses
instruction encoding to filter out irrelevant exceptions like page faults,
expecting a 4-byte PC advance for the emulated MRS instruction.

> diff --git a/tools/perf/tests/shell/coresight/trap_entry_exit.sh b/tools/perf/tests/shell/coresight/trap_entry_exit.sh
> new file mode 100755
> index 0000000000000..3ebd4033b247a
> --- /dev/null
> +++ b/tools/perf/tests/shell/coresight/trap_entry_exit.sh
> @@ -0,0 +1,24 @@
> +#!/bin/bash
> +# SPDX-License-Identifier: GPL-2.0
> +# CoreSight emulated instruction entry and exit (exclusive)
> +
> +# shellcheck source=../lib/coresight_exception.sh
> +. "$(dirname "$0")/../lib/coresight_exception.sh"
> +
> +test_trap_entry_exit()
> +{
[ ... ]
> +	decode_trace || return 1
> +	# Page faults also appear as "int", but retry the same PC. Select MRS
> +	# Xt, ID_AA64ISAR0_EL1 by its encoding; the low five bits select Xt.
> +	check_exception_pairs "int.*insn: [01][[:xdigit:]] 06 38 d5" "iret" 4 || return 1

[Severity: Medium]
Will this test become flaky if an instruction page fault occurs exactly on
the MRS instruction itself?

The test filters events using the MRS instruction encoding to skip
irrelevant page faults. However, if a page fault happens directly on the
MRS instruction, the exception type is still int and the fetched
instruction bytes will match the MRS encoding.

When the page fault is resolved, the kernel returns via iret to retry the
exact same PC, resulting in a PC difference of 0.

Since check_exception_pairs strictly enforces a PC difference of 4 for all
matched pairs, wouldn't this 0-byte advance cause the test to falsely
fail?

-- 
Sashiko AI review · https://sashiko.dev/#/patchset/20260923-perf_cs_etm_fix_non_taken-v2-0-6ab8c07a5455@arm.com?part=14