Re: [PATCH 2/2] selftests/ftrace: Account for 8-byte aligned trace_marker_raw events

[Hui Wang]

On 6/8/26 17:17, Masami Hiramatsu (Google) wrote:
On Sun,  7 Jun 2026 15:24:31 +0800
Hui Wang <hui.w...@canonical.com> wrote:

[...]
+    for config_file in \
+        /boot/config-$uname_r \
+        /lib/modules/$uname_r/config \
+        /lib/modules/$uname_r/build/.config

Hmm, also I don't like this, because this highly depends on the environment.
Instead, we can add CONFIG_IKCONFIG_PROC=y in 
tools/testing/selftests/ftrace/config.

Thank you,

Thanks for the review. I'll address all other comments in v2.

I have a concern about this specific point. On Ubuntu kernels, both 
CONFIG_IKCONFIG and CONFIG_IKCONFIG_PROC are disabled by default, so 
/proc/config.gz does not exist. If we drop the /boot/config-$(uname -r) 
lookup and rely solely on /proc/config.gz, this test would become 
unresolved on every Ubuntu kernel — a regression, since it works on 
those kernels today.

There is also existing precedent for the /boot/config-$(uname -r) 
fallback: tools/testing/selftests/mm/va_high_addr_switch.sh checks 
/proc/config.gz first and falls back to /boot/config-$(uname -r).

So how about we keep /boot/config-$(uname -r) as a fallback, but drop 
the /lib/modules/... paths you objected to. And add ftrace/config as you 
suggested here.

Thanks,
Hui.
+    do
+        if [ -f "$config_file" ]; then
+            grep -Eq "^${config}=(y|m)$" "$config_file"
+            return $?
+        fi
+    done
+
+    return 2
+}
+
  LOCALHOST=127.0.0.1
  
  yield() {
--
2.43.0