Re: [PATCH] docs: kernel-hacking: discourage from calling disable_irq() in atomic

[Manfred Spraul]

Hi Alexander,

On 12/12/22 17:37, A. Sverdlin wrote:
> From: Alexander Sverdlin <alexander.sverdlin@xxxxxxxxxxx>
>
> Correct the example in documentation so that disable_irq() is not being
> called in atomic context and remove the comment allowing to do so "with
> care" from the function header itself.
>
> disable_irq() calls sleeping synchronize_irq(), it's not allowed to call
> them in atomic context.
>
> Link: https://lore.kernel.org/lkml/87k02wbs2n.ffs@tglx/
> Signed-off-by: Alexander Sverdlin <alexander.sverdlin@xxxxxxxxxxx>

Reviewed-by: Manfred Spraul <manfred@xxxxxxxxxxxxxxxx>

(but check below, I would prefer if the change to kernel/irq/manage.c is 
dropped.

> ---
>   Documentation/kernel-hacking/locking.rst                    | 4 ++--
>   Documentation/translations/it_IT/kernel-hacking/locking.rst | 4 ++--
>   kernel/irq/manage.c                                         | 2 --
>   3 files changed, 4 insertions(+), 6 deletions(-)
>
> diff --git a/Documentation/kernel-hacking/locking.rst b/Documentation/kernel-hacking/locking.rst
> index 6805ae6e86e65..95fd6e0900d92 100644
> --- a/Documentation/kernel-hacking/locking.rst
> +++ b/Documentation/kernel-hacking/locking.rst
> @@ -1274,11 +1274,11 @@ Manfred Spraul points out that you can still do this, even if the data
>   is very occasionally accessed in user context or softirqs/tasklets. The
>   irq handler doesn't use a lock, and all other accesses are done as so::
>   
> -        spin_lock(&lock);
> +        mutex_lock(&lock);
>           disable_irq(irq);
>           ...
>           enable_irq(irq);
> -        spin_unlock(&lock);
> +        mutex_unlock(&lock);
>   
>   The disable_irq() prevents the irq handler from running
>   (and waits for it to finish if it's currently running on other CPUs).
> diff --git a/Documentation/translations/it_IT/kernel-hacking/locking.rst b/Documentation/translations/it_IT/kernel-hacking/locking.rst
> index 51af37f2d6210..bfbada56cf351 100644
> --- a/Documentation/translations/it_IT/kernel-hacking/locking.rst
> +++ b/Documentation/translations/it_IT/kernel-hacking/locking.rst
> @@ -1309,11 +1309,11 @@ se i dati vengono occasionalmente utilizzati da un contesto utente o
>   da un'interruzione software. Il gestore d'interruzione non utilizza alcun
>   *lock*, e tutti gli altri accessi verranno fatti così::
>   
> -        spin_lock(&lock);
> +        mutex_lock(&lock);
>           disable_irq(irq);
>           ...
>           enable_irq(irq);
> -        spin_unlock(&lock);
> +        mutex_unlock(&lock);
>   
>   La funzione disable_irq() impedisce al gestore d'interruzioni
>   d'essere eseguito (e aspetta che finisca nel caso fosse in esecuzione su

> diff --git a/kernel/irq/manage.c b/kernel/irq/manage.c
> index 40fe7806cc8c9..2054de5bf3c53 100644
> --- a/kernel/irq/manage.c
> +++ b/kernel/irq/manage.c
> @@ -722,8 +722,6 @@ EXPORT_SYMBOL(disable_irq_nosync);
>    *	This function waits for any pending IRQ handlers for this interrupt
>    *	to complete before returning. If you use this function while
>    *	holding a resource the IRQ handler may need you will deadlock.
> - *
> - *	This function may be called - with care - from IRQ context.
>    */
>   void disable_irq(unsigned int irq)
>   {

Can you drop this part?

I haven't noticed that you added this change into the patch, and thus I 
created a seperate patch.

https://git.kernel.org/pub/scm/linux/kernel/git/akpm/25-new.git/tree/patches/kernel-irq-managec-disable_irq-might-sleep.patch

As core difference: I've added a might_sleep() into disable_irq().


--

    Manfred