linux-kernel - Re: [PATCH] locking: detect spin_lock_irq() call with disabled interrupts

lists.openwall.net		lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening linux-cve-announce PHC
Open Source and information security mailing list archives

Hash Suite: Windows password security audit tool. GUI, reports in PDF.

[<prev] [next>] [<thread-prev] [day] [month] [year] [list]

Message-ID: <xdfm634qqtchbaqfnnnmwqwmto5bllkspfixescyxltvjb4y2c@l4uiiarg5kjs>
Date: Fri, 26 Sep 2025 11:18:34 +0200
From: Joel Granados <joel.granados@...nel.org>
To: Pavel Tikhomirov <ptikhomirov@...tuozzo.com>
Cc: Peter Zijlstra <peterz@...radead.org>, Ingo Molnar <mingo@...hat.com>, 
	Will Deacon <will@...nel.org>, Boqun Feng <boqun.feng@...il.com>, 
	Waiman Long <longman@...hat.com>, Kees Cook <kees@...nel.org>, 
	Andrew Morton <akpm@...ux-foundation.org>, Konstantin Khorenko <khorenko@...tuozzo.com>, 
	Denis Lunev <den@...tuozzo.com>, Aleksandr Mikhalitsyn <aleksandr.mikhalitsyn@...onical.com>, 
	linux-kernel@...r.kernel.org, linux-fsdevel@...r.kernel.org, kernel@...nvz.org
Subject: Re: [PATCH] locking: detect spin_lock_irq() call with disabled
 interrupts

On Fri, Jun 06, 2025 at 05:57:23PM +0800, Pavel Tikhomirov wrote:
> This is intended to easily detect irq spinlock self-deadlocks like:
> 
>   spin_lock_irq(A);
>   spin_lock_irq(B);
>   spin_unlock_irq(B);
>     IRQ {
>       spin_lock(A); <- deadlocks
>       spin_unlock(A);
>     }
>   spin_unlock_irq(A);
...
> diff --git a/kernel/sysctl.c b/kernel/sysctl.c
> index 9b4f0cff76ea..1e3cca2e3c8f 100644
> --- a/kernel/sysctl.c
> +++ b/kernel/sysctl.c
> @@ -50,6 +50,7 @@
>  #include <linux/sched/sysctl.h>
>  #include <linux/mount.h>
>  #include <linux/pid.h>
> +#include <linux/spinlock.h>
>  
>  #include "../lib/kstrtox.h"
>  
> @@ -1758,6 +1759,14 @@ static const struct ctl_table kern_table[] = {
>  		.extra2		= SYSCTL_INT_MAX,
>  	},
>  #endif
> +#ifdef CONFIG_DEBUG_SPINLOCK
> +	{
> +		.procname	= "debug_spin_lock_irq_with_disabled_interrupts",
> +		.data		= &debug_spin_lock_irq_with_disabled_interrupts,
> +		.mode		= 0644,
> +		.proc_handler	= proc_do_static_key,
> +	},
> +#endif
Patchwork reminded me about this one and I see now that I missed a
comment. In case you are working on a V2, do *NOT* add more ctl_table
elements to the kern_table. If you need to add a sysctl file for
locking, do it under kernel/locking. You can take inspiration from how
rtmutex_sysctl_table and "max_lock_depth" is handled.

best

>  };
>  
>  int __init sysctl_init_bases(void)
> diff --git a/lib/Kconfig.debug b/lib/Kconfig.debug
> index ebe33181b6e6..c4834f4c9d51 100644
> --- a/lib/Kconfig.debug
> +++ b/lib/Kconfig.debug
> @@ -1465,6 +1465,18 @@ config DEBUG_SPINLOCK
>  	  best used in conjunction with the NMI watchdog so that spinlock
>  	  deadlocks are also debuggable.
>  
> +config DEBUG_SPINLOCK_IRQ_WITH_DISABLED_INTERRUPTS_BY_DEFAULT
> +	bool "Detect spin_(un)lock_irq() call with disabled(enabled) interrupts"
> +	depends on DEBUG_SPINLOCK
> +	help
> +	  Say Y here to detect spin_lock_irq() and spin_unlock_irq() calls
> +	  with disabled (enabled) interrupts. This helps detecting bugs
> +	  where the code is not using the right locking primitives. E.g.
> +	  using spin_lock_irq() twice in a row (on different locks). And thus
> +	  code can reenable interrupts where they should be disabled and lead
> +	  to deadlock.
> +	  Say N if you are unsure.
> +
>  config DEBUG_MUTEXES
>  	bool "Mutex debugging: basic checks"
>  	depends on DEBUG_KERNEL && !PREEMPT_RT
> -- 
> 2.49.0
> 

-- 

Joel Granados

Download attachment "signature.asc" of type "application/pgp-signature" (660 bytes)