From: Michal Hocko <mhocko@suse.com>
To: Sebastian Andrzej Siewior <bigeasy@linutronix.de>
Cc: linux-kernel@vger.kernel.org, linux-rt-devel@lists.linux.dev,
Ben Segall <bsegall@google.com>, Boqun Feng <boqun@kernel.org>,
Clark Williams <clrkwllms@kernel.org>,
David Woodhouse <dwmw2@infradead.org>,
Dietmar Eggemann <dietmar.eggemann@arm.com>,
Ingo Molnar <mingo@redhat.com>,
Juri Lelli <juri.lelli@redhat.com>,
K Prateek Nayak <kprateek.nayak@amd.com>,
Mel Gorman <mgorman@suse.de>,
Peter Zijlstra <peterz@infradead.org>, Sean Young <sean@mess.org>,
Steven Rostedt <rostedt@goodmis.org>,
Thierry Reding <thierry.reding@gmail.com>,
Valentin Schneider <vschneid@redhat.com>,
Vincent Guittot <vincent.guittot@linaro.org>,
Waiman Long <longman@redhat.com>, Will Deacon <will@kernel.org>
Subject: Re: [PATCH] sched: Allow sleeping spinlocks on PREEMPT_RT within non_block_start()/end block.
Date: Fri, 21 Aug 2026 13:46:06 +0200 [thread overview]
Message-ID: <aog6fr80a2buniSZ@tiehlicka> (raw)
In-Reply-To: <20260821095755.am1-Segb@linutronix.de>
On Fri 21-08-26 11:57:55, Sebastian Andrzej Siewior wrote:
> Commit 312364f3534cc ("kernel.h: Add non_block_start/end()") added
> non_block_star()/end(). The intention was to catch any callbacks which
> depend on locks or sleep able conditionals in order to ensure forward
> progress. According to the commit message spinlocks were excluded
> "because spinlocks can't have an indirect dependency upon the page
> allocator…".
> On PREEMPT_RT this leads to a splat because spinlock_t and rwlock_t are
> turned into a sleeping spinlock which have a might_sleep() as they could
> schedule() on lock contention.
>
> Besids mm, the other used of non_block_start() is pwm. Here the hrtimer
> can acnqure a spinlock_t on RT and I can't tell other user but I assume
> it is not atomic on RT since it wouldn't be able to acquire its own
> (spinlock_t) lock.
>
> Since spinlock_t (and rwlock_t) are not problem, add a sleeping_lock
> argument to __might_resched() signaling if this is scheduling request
> is from a sleeping lock (true) or regular scheduling request (false).
> Add the `true' argument to rtlock_might_resched() which is used on
> PREEMPT_RT for both sleeping lock types.
This definitely makes a lot of sense to me. It looks like a
straightforward solution for false positives. The patch looks good to
me, I do not feel qualified to give you Acked-by though.
Thanks!
> Cc: Michal Hocko <mhocko@suse.com>
> Cc: David Woodhouse <dwmw2@infradead.org>
> Cc: Sean Young <sean@mess.org>
> Cc: Thierry Reding <thierry.reding@gmail.com>
> Signed-off-by: Sebastian Andrzej Siewior <bigeasy@linutronix.de>
> ---
> include/linux/kernel.h | 5 +++--
> include/linux/sched.h | 24 ++++++++++++------------
> kernel/locking/spinlock_rt.c | 2 +-
> kernel/sched/core.c | 7 ++++---
> 4 files changed, 20 insertions(+), 18 deletions(-)
>
> diff --git a/include/linux/kernel.h b/include/linux/kernel.h
> index e5570a16cbb1a..fbafdfe9bdb78 100644
> --- a/include/linux/kernel.h
> +++ b/include/linux/kernel.h
> @@ -70,7 +70,8 @@ extern int dynamic_might_resched(void);
> #endif /* CONFIG_PREEMPT_* */
>
> #ifdef CONFIG_DEBUG_ATOMIC_SLEEP
> -extern void __might_resched(const char *file, int line, unsigned int offsets);
> +extern void __might_resched(const char *file, int line, unsigned int offsets,
> + bool rt_sleeping_lock);
> extern void __might_sleep(const char *file, int line);
> extern void __cant_sleep(const char *file, int line, int preempt_offset);
> extern void __cant_migrate(const char *file, int line);
> @@ -128,7 +129,7 @@ extern void __cant_migrate(const char *file, int line);
> # define non_block_end() WARN_ON(current->non_block_count-- == 0)
> #else
> static inline void __might_resched(const char *file, int line,
> - unsigned int offsets) { }
> + unsigned int offsets, bool rt_sleeping_lock) { }
> static inline void __might_sleep(const char *file, int line) { }
> # define might_sleep() do { might_resched(); } while (0)
> # define cant_sleep() do { } while (0)
> diff --git a/include/linux/sched.h b/include/linux/sched.h
> index 373bcc0598d10..650e42d55e1e8 100644
> --- a/include/linux/sched.h
> +++ b/include/linux/sched.h
> @@ -2158,9 +2158,9 @@ static inline int _cond_resched(void)
>
> #endif /* !CONFIG_PREEMPTION || CONFIG_PREEMPT_DYNAMIC */
>
> -#define cond_resched() ({ \
> - __might_resched(__FILE__, __LINE__, 0); \
> - _cond_resched(); \
> +#define cond_resched() ({ \
> + __might_resched(__FILE__, __LINE__, 0, false); \
> + _cond_resched(); \
> })
>
> extern int __cond_resched_lock(spinlock_t *lock) __must_hold(lock);
> @@ -2186,19 +2186,19 @@ extern int __cond_resched_rwlock_write(rwlock_t *lock) __must_hold(lock);
> (PREEMPT_LOCK_OFFSET + (1U << MIGHT_RESCHED_RCU_SHIFT))
> #endif
>
> -#define cond_resched_lock(lock) ({ \
> - __might_resched(__FILE__, __LINE__, PREEMPT_LOCK_RESCHED_OFFSETS); \
> - __cond_resched_lock(lock); \
> +#define cond_resched_lock(lock) ({ \
> + __might_resched(__FILE__, __LINE__, PREEMPT_LOCK_RESCHED_OFFSETS, false); \
> + __cond_resched_lock(lock); \
> })
>
> -#define cond_resched_rwlock_read(lock) ({ \
> - __might_resched(__FILE__, __LINE__, PREEMPT_LOCK_RESCHED_OFFSETS); \
> - __cond_resched_rwlock_read(lock); \
> +#define cond_resched_rwlock_read(lock) ({ \
> + __might_resched(__FILE__, __LINE__, PREEMPT_LOCK_RESCHED_OFFSETS, false); \
> + __cond_resched_rwlock_read(lock); \
> })
>
> -#define cond_resched_rwlock_write(lock) ({ \
> - __might_resched(__FILE__, __LINE__, PREEMPT_LOCK_RESCHED_OFFSETS); \
> - __cond_resched_rwlock_write(lock); \
> +#define cond_resched_rwlock_write(lock) ({ \
> + __might_resched(__FILE__, __LINE__, PREEMPT_LOCK_RESCHED_OFFSETS, false); \
> + __cond_resched_rwlock_write(lock); \
> })
>
> #ifndef CONFIG_PREEMPT_RT
> diff --git a/kernel/locking/spinlock_rt.c b/kernel/locking/spinlock_rt.c
> index 1d5e1b3c60bfa..0a89e7f10af18 100644
> --- a/kernel/locking/spinlock_rt.c
> +++ b/kernel/locking/spinlock_rt.c
> @@ -33,7 +33,7 @@
> (rcu_preempt_depth() << MIGHT_RESCHED_RCU_SHIFT)
>
> #define rtlock_might_resched() \
> - __might_resched(__FILE__, __LINE__, RTLOCK_RESCHED_OFFSETS)
> + __might_resched(__FILE__, __LINE__, RTLOCK_RESCHED_OFFSETS, true)
>
> static __always_inline void rtlock_lock(struct rt_mutex_base *rtm)
> {
> diff --git a/kernel/sched/core.c b/kernel/sched/core.c
> index 96226707c2f61..ee0fbb41fc61c 100644
> --- a/kernel/sched/core.c
> +++ b/kernel/sched/core.c
> @@ -9123,7 +9123,7 @@ void __might_sleep(const char *file, int line)
> (void *)current->task_state_change,
> (void *)current->task_state_change);
>
> - __might_resched(file, line, 0);
> + __might_resched(file, line, 0, false);
> }
> EXPORT_SYMBOL(__might_sleep);
>
> @@ -9148,7 +9148,8 @@ static inline bool resched_offsets_ok(unsigned int offsets)
> return nested == offsets;
> }
>
> -void __might_resched(const char *file, int line, unsigned int offsets)
> +void __might_resched(const char *file, int line, unsigned int offsets,
> + bool rt_sleeping_lock)
> {
> /* Ratelimiting timestamp: */
> static unsigned long prev_jiffy;
> @@ -9159,7 +9160,7 @@ void __might_resched(const char *file, int line, unsigned int offsets)
> rcu_sleep_check();
>
> if ((resched_offsets_ok(offsets) && !irqs_disabled() &&
> - !is_idle_task(current) && !current->non_block_count) ||
> + !is_idle_task(current) && (rt_sleeping_lock || !current->non_block_count)) ||
> system_state == SYSTEM_BOOTING || system_state > SYSTEM_RUNNING ||
> oops_in_progress)
> return;
> --
> 2.55.0
--
Michal Hocko
SUSE Labs
next prev parent reply other threads:[~2026-08-21 11:46 UTC|newest]
Thread overview: 3+ messages / expand[flat|nested] mbox.gz Atom feed top
2026-08-21 9:57 [PATCH] sched: Allow sleeping spinlocks on PREEMPT_RT within non_block_start()/end block Sebastian Andrzej Siewior
2026-08-21 11:46 ` Michal Hocko [this message]
2026-08-21 12:13 ` [PATCH] sched: Allow sleeping spinlocks on PREEMPT_RT within non_block_start() section David Woodhouse
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=aog6fr80a2buniSZ@tiehlicka \
--to=mhocko@suse.com \
--cc=bigeasy@linutronix.de \
--cc=boqun@kernel.org \
--cc=bsegall@google.com \
--cc=clrkwllms@kernel.org \
--cc=dietmar.eggemann@arm.com \
--cc=dwmw2@infradead.org \
--cc=juri.lelli@redhat.com \
--cc=kprateek.nayak@amd.com \
--cc=linux-kernel@vger.kernel.org \
--cc=linux-rt-devel@lists.linux.dev \
--cc=longman@redhat.com \
--cc=mgorman@suse.de \
--cc=mingo@redhat.com \
--cc=peterz@infradead.org \
--cc=rostedt@goodmis.org \
--cc=sean@mess.org \
--cc=thierry.reding@gmail.com \
--cc=vincent.guittot@linaro.org \
--cc=vschneid@redhat.com \
--cc=will@kernel.org \
/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.