[<prev] [next>] [<thread-prev] [day] [month] [year] [list]
Message-ID: <Z7iz8I6QD9Amjfmw@Mac.home>
Date: Fri, 21 Feb 2025 09:12:16 -0800
From: Boqun Feng <boqun.feng@...il.com>
To: Sebastian Andrzej Siewior <bigeasy@...utronix.de>
Cc: linux-rt-devel@...ts.linux.dev, linux-kernel@...r.kernel.org,
Peter Zijlstra <peterz@...radead.org>,
Ingo Molnar <mingo@...hat.com>, Will Deacon <will@...nel.org>,
Waiman Long <longman@...hat.com>,
Guenter Roeck <linux@...ck-us.net>,
Clark Williams <clrkwllms@...nel.org>,
Steven Rostedt <rostedt@...dmis.org>
Subject: Re: [PATCH 1/2] lockdep: Don't disable interrupts on RT in
disable_irq_nosync_lockdep.*()
On Wed, Feb 12, 2025 at 11:36:18AM +0100, Sebastian Andrzej Siewior wrote:
> disable_irq_nosync_lockdep() disables interrupts with lockdep enabled to
> avoid false positive reports by lockdep that a certain lock has not been
> acquired with disabled interrupts. The user of this macros expects that
> a lock can be acquried without disabling interrupts because the IRQ line
> triggering the interrupt is disabled.
>
> This triggers a warning on PREEMPT_RT because after
> disable_irq_nosync_lockdep.*() the following spinlock_t now is acquired
> with disabled interrupts.
>
> On PREEMPT_RT there is no difference between spin_lock() and
> spin_lock_irq() so avoiding disabling interrupts in this case works for
> the two remaining callers as of today.
>
> Don't disable interrupts on PREEMPT_RT in disable_irq_nosync_lockdep.*().
>
> Reported-by: Guenter Roeck <linux@...ck-us.net>
> Closes: https://lore.kernel.org/760e34f9-6034-40e0-82a5-ee9becd24438@roeck-us.net
> Fixes: e8106b941ceab ("[PATCH] lockdep: core, add enable/disable_irq_irqsave/irqrestore() APIs")
Is this "Fixes" tag accurate? IIUC, the issue only exists after
CONFIG_PREEMPT_RT is introduced from mainline POV. Or we want this to be
backported in older stable kernels predating CONFIG_PREEMPT_RT?
Regards,
Boqun
> Signed-off-by: Sebastian Andrzej Siewior <bigeasy@...utronix.de>
> ---
> include/linux/interrupt.h | 8 ++++----
> 1 file changed, 4 insertions(+), 4 deletions(-)
>
> diff --git a/include/linux/interrupt.h b/include/linux/interrupt.h
> index 8cd9327e4e78d..a1b1be9bf73b2 100644
> --- a/include/linux/interrupt.h
> +++ b/include/linux/interrupt.h
> @@ -448,7 +448,7 @@ irq_calc_affinity_vectors(unsigned int minvec, unsigned int maxvec,
> static inline void disable_irq_nosync_lockdep(unsigned int irq)
> {
> disable_irq_nosync(irq);
> -#ifdef CONFIG_LOCKDEP
> +#if defined(CONFIG_LOCKDEP) && !defined(CONFIG_PREEMPT_RT)
> local_irq_disable();
> #endif
> }
> @@ -456,7 +456,7 @@ static inline void disable_irq_nosync_lockdep(unsigned int irq)
> static inline void disable_irq_nosync_lockdep_irqsave(unsigned int irq, unsigned long *flags)
> {
> disable_irq_nosync(irq);
> -#ifdef CONFIG_LOCKDEP
> +#if defined(CONFIG_LOCKDEP) && !defined(CONFIG_PREEMPT_RT)
> local_irq_save(*flags);
> #endif
> }
> @@ -471,7 +471,7 @@ static inline void disable_irq_lockdep(unsigned int irq)
>
> static inline void enable_irq_lockdep(unsigned int irq)
> {
> -#ifdef CONFIG_LOCKDEP
> +#if defined(CONFIG_LOCKDEP) && !defined(CONFIG_PREEMPT_RT)
> local_irq_enable();
> #endif
> enable_irq(irq);
> @@ -479,7 +479,7 @@ static inline void enable_irq_lockdep(unsigned int irq)
>
> static inline void enable_irq_lockdep_irqrestore(unsigned int irq, unsigned long *flags)
> {
> -#ifdef CONFIG_LOCKDEP
> +#if defined(CONFIG_LOCKDEP) && !defined(CONFIG_PREEMPT_RT)
> local_irq_restore(*flags);
> #endif
> enable_irq(irq);
> --
> 2.47.2
>
Powered by blists - more mailing lists