Re: [PATCH perf-tools-next v4 4/4] perf tests: Add shell test for kernel symbol beautifier

On Fri, Aug 21, 2026 at 09:01:55PM +0000, [email protected] wrote:
> --- Patch [4]: [PATCH perf-tools-next v4 4/4] perf tests: Add shell test for kernel symbol beautifier ---
> commit c9c4a1b0dcd918dc4702b41032200c74d4363a15
> Author: Aaron Tomlin <[email protected]>
> 
> perf tests: Add shell test for kernel symbol beautifier
> 
> This commit adds a dedicated shell test script to verify that perf trace
> properly symbolises kernel virtual addresses and function pointers using
> both the default kallsyms beautifier and BTF type routing.
> 
> > diff --git a/tools/perf/tests/shell/trace_ksym_beautifier.sh b/tools/perf/tests/shell/trace_ksym_beautifier.sh
> > new file mode 100755
> > index 0000000000000..c030b03e2b662
> > --- /dev/null
> > +++ b/tools/perf/tests/shell/trace_ksym_beautifier.sh
> > @@ -0,0 +1,43 @@
> [ ... ]
> > +test_ksym_kallsyms() {
> > +  echo "Testing perf trace kernel symbol beautifier (default kallsyms)"
> > +  output="$(perf trace -e kmem:kmalloc --max-events=1 2>&1)"
> 
> [Severity: Medium]
> Could this wait indefinitely on a completely quiescent system?
> 
> Because it waits system-wide for the kmem:kmalloc event without a timeout
> or a deterministic workload to trigger it, it might stall the test suite
> if the event takes too long to fire.

Fair. I will propose: 'perf trace -e kmem:kmalloc --max-events=1 -- true 2>&1'.

> > +test_ksym_btf() {
> > +  echo "Testing perf trace kernel symbol beautifier (BTF)"
> > +  if [ ! -f /sys/kernel/btf/vmlinux ]; then
> > +    echo "Skipping BTF test due to missing vmlinux BTF"
> > +    return
> > +  fi
> > +
> > +  output="$(perf trace -e csd:csd_function_entry --force-btf --max-events=1 2>&1)"
> 
> [Severity: Medium]
> Will this test fail erroneously on uniprocessor (UP) kernels?
> 
> The csd:csd_function_entry tracepoint is defined in kernel/smp.c, which is
> only available when CONFIG_SMP is enabled. On a UP kernel, perf trace will
> fail with an unknown tracepoint error, which the script would parse as a
> symbolization failure.
> 
> [Severity: Medium]
> Additionally, can this cause an indefinite hang on idle SMP systems?
> 
> The csd:csd_function_entry event relies on SMP cross-calls (IPIs). On an
> idle or isolated test VM, this event might not occur organically for
> extended periods, potentially causing perf trace to block forever without
> a timeout or a triggering workload.

Acknowledged. Instead: 'perf trace -e timer:hrtimer_start --force-btf --max-events=1 -- sleep 0.01 2>&1'

Kind regards,
-- 
Aaron Tomlin