Re: [RFC PATCH v5 3/3] trace: add documentation, selftest and tooling for stackmap

[Pengfei Li]

On Wed, 2 Sep 2026 06:51:26 +0000 sashiko-bot@xxxxxxxxxx wrote:
> Is the documentation correct, or is the commit message correct?

The documentation and the selftests are correct. The commit message was
stale.

Reset clears the map and nothing else: it may run while tracing is
active and leaves the ring buffer untouched. That is what the code,
Documentation/trace/ftrace-stackmap.rst, stackmap-basic.tc and
stackmap-reset.tc all implement. The v5 commit message still described
the earlier design that required tracing to be stopped and cleared the
buffer, which was dropped after Steven's feedback on the v4 thread
(https://lore.kernel.org/all/20260821235129.078dd489@fedora/). v6
corrects the commit message on all three patches.

> Would it make sense to add a similar cleanup trap here

Yes. stackmap-instance-gate.tc now installs an EXIT cleanup trap
instead of relying on the manual rmdir in fail(). It also records
whether this test created instances/test_stackmap_gate, so cleanup
removes the instance only in that case and a pre-existing instance is
never removed on mkdir failure.

Pengfei