[<prev] [next>] [<thread-prev] [day] [month] [year] [list]
Message-ID: <20250421184524.215a65fc@kernel.org>
Date: Mon, 21 Apr 2025 18:45:24 -0700
From: Jakub Kicinski <kuba@...nel.org>
To: Justin Iurman <justin.iurman@...ege.be>
Cc: netdev@...r.kernel.org, davem@...emloft.net, edumazet@...gle.com,
pabeni@...hat.com, horms@...nel.org, shuah@...nel.org
Subject: Re: [PATCH net-next] selftests: net: kmemleak for lwt dst cache
tests
On Thu, 17 Apr 2025 15:08:30 +0200 Justin Iurman wrote:
> Force the use of kmemleak to check everything's OK and report results
> for each test case. Also, useless sleeps were removed, and the bash
> script was renamed to something that makes more sense. Due to kmemleak,
> some tests may be false negatives. To mitigate that (i.e., to have more
> stable results), the solution of a kmemleak scan at the end (vs. for
> each test) was preferred.
I'm open to other opinions but to me this is not a scalable direction.
Checking for splats and memleaks does not belong in any particular
test script, it belongs in whatever executes the test. We could add
some standard way to get ksft to scan for splats and leaks, maybe in
tools/testing/selftests/kselftest/runner.sh ?
> +++ b/tools/testing/selftests/net/config
> +CONFIG_DEBUG_KMEMLEAK=y
Enabling kmemleak for all netdev runners is too heavy, if others
feel like we should have the test case do the scanning - let's
detect if kmemleak is available. The expectation is that the user
added kernel/configs/debug.config to their build if they want
debug checks.
--
pw-bot: cr
Powered by blists - more mailing lists