From: Andrea Righi <arighi@nvidia.com>
To: zhidao su <soolaugust@gmail.com>
Cc: sched-ext@lists.linux.dev, linux-kernel@vger.kernel.org,
tj@kernel.org, void@manifault.com, changwoo@igalia.com,
peterz@infradead.org, mingo@redhat.com,
zhidao su <suzhidao@xiaomi.com>
Subject: Re: [PATCH] sched_ext: Use WRITE_ONCE() for the write side of scx_enable helper pointer
Date: Fri, 6 Mar 2026 15:09:10 +0100 [thread overview]
Message-ID: <aargBvjGr_7lwOKD@gpd4> (raw)
In-Reply-To: <20260306105901.2543743-1-suzhidao@xiaomi.com>
Hi,
On Fri, Mar 06, 2026 at 06:59:01PM +0800, zhidao su wrote:
> scx_enable() uses double-checked locking to lazily initialize a static
> kthread_worker pointer:
>
> if (!READ_ONCE(helper)) {
> mutex_lock(&helper_mutex);
> if (!helper) {
> helper = kthread_run_worker(0, "scx_enable_helper");
> ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
> plain write -- KCSAN data race
>
> The outer READ_ONCE() annotates the lockless fast-path read, but the
> write side uses a plain assignment without the matching WRITE_ONCE().
> The KCSAN documentation requires that if one accessor uses READ_ONCE()
> or WRITE_ONCE() on a variable to annotate lock-free access, all other
> accesses must also use the appropriate accessor. A plain write leaves
> the pair incomplete and will trigger KCSAN warnings.
>
> The error path also has the same issue:
>
> helper = NULL;
> ^^^^^^^^^^
> plain write -- KCSAN data race
>
> Fix both plain writes by using WRITE_ONCE() to complete the concurrent
> access annotation and make the code KCSAN-clean.
>
There should be a:
Fixes: b06ccbabe2506 ("sched_ext: Fix starvation of scx_enable() under fair-class saturation")
Apart than that LGTM.
Acked-by: Andrea Righi <arighi@nvidia.com>
Thanks,
-Andrea
> Signed-off-by: zhidao su <suzhidao@xiaomi.com>
> ---
> kernel/sched/ext.c | 4 ++--
> 1 file changed, 2 insertions(+), 2 deletions(-)
>
> diff --git a/kernel/sched/ext.c b/kernel/sched/ext.c
> index 9a1471ad5ae7..c4ccd685259f 100644
> --- a/kernel/sched/ext.c
> +++ b/kernel/sched/ext.c
> @@ -5355,9 +5355,9 @@ static int scx_enable(struct sched_ext_ops *ops, struct bpf_link *link)
> if (!READ_ONCE(helper)) {
> mutex_lock(&helper_mutex);
> if (!helper) {
> - helper = kthread_run_worker(0, "scx_enable_helper");
> + WRITE_ONCE(helper, kthread_run_worker(0, "scx_enable_helper"));
> if (IS_ERR_OR_NULL(helper)) {
> - helper = NULL;
> + WRITE_ONCE(helper, NULL);
> mutex_unlock(&helper_mutex);
> return -ENOMEM;
> }
> --
> 2.43.0
>
next prev parent reply other threads:[~2026-03-06 14:09 UTC|newest]
Thread overview: 7+ messages / expand[flat|nested] mbox.gz Atom feed top
2026-03-06 10:59 [PATCH] sched_ext: Use WRITE_ONCE() for the write side of scx_enable helper pointer zhidao su
2026-03-06 14:09 ` Andrea Righi [this message]
2026-03-06 14:17 ` [PATCH v2] " zhidao su
2026-03-06 16:34 ` Tejun Heo
2026-03-09 2:46 ` [PATCH v3] " zhidao su
2026-03-09 4:34 ` zhidao su
2026-03-09 16:20 ` Tejun Heo
Reply instructions:
You may reply publicly to this message via plain-text email
using any one of the following methods:
* Save the following mbox file, import it into your mail client,
and reply-to-all from there: mbox
Avoid top-posting and favor interleaved quoting:
https://en.wikipedia.org/wiki/Posting_style#Interleaved_style
* Reply using the --to, --cc, and --in-reply-to
switches of git-send-email(1):
git send-email \
--in-reply-to=aargBvjGr_7lwOKD@gpd4 \
--to=arighi@nvidia.com \
--cc=changwoo@igalia.com \
--cc=linux-kernel@vger.kernel.org \
--cc=mingo@redhat.com \
--cc=peterz@infradead.org \
--cc=sched-ext@lists.linux.dev \
--cc=soolaugust@gmail.com \
--cc=suzhidao@xiaomi.com \
--cc=tj@kernel.org \
--cc=void@manifault.com \
/path/to/YOUR_REPLY
https://kernel.org/pub/software/scm/git/docs/git-send-email.html
* If your mail client supports setting the In-Reply-To header
via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line
before the message body.
This is an external index of several public inboxes,
see mirroring instructions on how to clone and mirror
all data and code used by this external index.