[PATCH v3] srcutree: Don't transition atomic SRCU to big in srcu_gp_end()

[PATCH v3] srcutree: Don't transition atomic SRCU to big in srcu_gp_end()

[Kunwu Chan]

Atomic SRCU must remain in the small size state. Warn if this
invariant is violated and avoid transitioning to big in that case.

Signed-off-by: Kunwu Chan <kunwu.chan@gmail.com>
---
Changes since v2:
- Also prevent the transition when the atomic SRCU size-state
  invariant is violated, per Paul McKenney's suggestion.
-v2: https://lore.kernel.org/rcu/20260910084629.3695383-1-kunwu.chan@gmail.com/

Changes since v1:
- Drop the !is_atomic guard from v1 and add WARN_ON_ONCE() instead,
  per Paul McKenney's suggestion.
- v1: https://lore.kernel.org/rcu/20260907075829.2073224-8-kunwu.chan@linux.dev/
---
 kernel/rcu/srcutree.c | 7 +++++--
 1 file changed, 5 insertions(+), 2 deletions(-)

diff --git a/kernel/rcu/srcutree.c b/kernel/rcu/srcutree.c
index 84c022ec8e09..07db4e459b30 100644
--- a/kernel/rcu/srcutree.c
+++ b/kernel/rcu/srcutree.c
@@ -1073,8 +1073,11 @@ static void srcu_gp_end(struct srcu_struct *ssp, bool is_atomic)
 		raw_spin_unlock_irq_rcu_node(sup);
 	}
 
-	/* Transition to big if needed. */
-	if (ss_state != SRCU_SIZE_SMALL && ss_state != SRCU_SIZE_BIG) {
+	/* Transition to big if needed, but never for atomic SRCU. */
+	if (ssp->srcu_reader_flavor == SRCU_READ_FLAVOR_ATOMIC &&
+	    ss_state != SRCU_SIZE_SMALL) {
+		WARN_ON_ONCE(1);
+	} else if (ss_state != SRCU_SIZE_SMALL && ss_state != SRCU_SIZE_BIG) {
 		if (ss_state == SRCU_SIZE_ALLOC)
 			init_srcu_struct_nodes(ssp, GFP_KERNEL);
 		else
-- 
2.43.0

Re: [PATCH v3] srcutree: Don't transition atomic SRCU to big in srcu_gp_end()

[Paul E. McKenney]

On Fri, Sep 11, 2026 at 03:21:25PM +0800, Kunwu Chan wrote:
> Atomic SRCU must remain in the small size state. Warn if this
> invariant is violated and avoid transitioning to big in that case.
> 
> Signed-off-by: Kunwu Chan <kunwu.chan@gmail.com>

Very good, queued for further testing and review, thank you!

I folded the new "if" statement onto one line because some years back the
Linux kernel liberated itself from the 80-column constraint imposed by IBM
back in 1928.  We can now have up to 100 characters in a line of code.  ;-)

							Thanx, Paul

> ---
> Changes since v2:
> - Also prevent the transition when the atomic SRCU size-state
>   invariant is violated, per Paul McKenney's suggestion.
> -v2: https://lore.kernel.org/rcu/20260910084629.3695383-1-kunwu.chan@gmail.com/
> 
> Changes since v1:
> - Drop the !is_atomic guard from v1 and add WARN_ON_ONCE() instead,
>   per Paul McKenney's suggestion.
> - v1: https://lore.kernel.org/rcu/20260907075829.2073224-8-kunwu.chan@linux.dev/
> ---
>  kernel/rcu/srcutree.c | 7 +++++--
>  1 file changed, 5 insertions(+), 2 deletions(-)
> 
> diff --git a/kernel/rcu/srcutree.c b/kernel/rcu/srcutree.c
> index 84c022ec8e09..07db4e459b30 100644
> --- a/kernel/rcu/srcutree.c
> +++ b/kernel/rcu/srcutree.c
> @@ -1073,8 +1073,11 @@ static void srcu_gp_end(struct srcu_struct *ssp, bool is_atomic)
>  		raw_spin_unlock_irq_rcu_node(sup);
>  	}
>  
> -	/* Transition to big if needed. */
> -	if (ss_state != SRCU_SIZE_SMALL && ss_state != SRCU_SIZE_BIG) {
> +	/* Transition to big if needed, but never for atomic SRCU. */
> +	if (ssp->srcu_reader_flavor == SRCU_READ_FLAVOR_ATOMIC &&
> +	    ss_state != SRCU_SIZE_SMALL) {
> +		WARN_ON_ONCE(1);
> +	} else if (ss_state != SRCU_SIZE_SMALL && ss_state != SRCU_SIZE_BIG) {
>  		if (ss_state == SRCU_SIZE_ALLOC)
>  			init_srcu_struct_nodes(ssp, GFP_KERNEL);
>  		else
> -- 
> 2.43.0
> 
> 

Re: [PATCH v3] srcutree: Don't transition atomic SRCU to big in srcu_gp_end()

[KunWu Chan]

On Sat, Sep 12, 2026 at 12:37 AM Paul E. McKenney <paulmck@kernel.org> wrote:
>
> On Fri, Sep 11, 2026 at 03:21:25PM +0800, Kunwu Chan wrote:
> > Atomic SRCU must remain in the small size state. Warn if this
> > invariant is violated and avoid transitioning to big in that case.
> >
> > Signed-off-by: Kunwu Chan <kunwu.chan@gmail.com>
>
> Very good, queued for further testing and review, thank you!
>
> I folded the new "if" statement onto one line because some years back the
> Linux kernel liberated itself from the 80-column constraint imposed by IBM
> back in 1928.  We can now have up to 100 characters in a line of code.  ;-)

Thanks, Paul. And thanks for sharing that interesting bit of kernel
history! I didn't know about the IBM 80-column constraint from 1928.
I had assumed that 80 columns was still the current convention, so
it's good to know we can use up to 100 now. ;-)

Thanks,
Kunwu

>
>                                                         Thanx, Paul
>
> > ---
> > Changes since v2:
> > - Also prevent the transition when the atomic SRCU size-state
> >   invariant is violated, per Paul McKenney's suggestion.
> > -v2: https://lore.kernel.org/rcu/20260910084629.3695383-1-kunwu.chan@gmail.com/
> >
> > Changes since v1:
> > - Drop the !is_atomic guard from v1 and add WARN_ON_ONCE() instead,
> >   per Paul McKenney's suggestion.
> > - v1: https://lore.kernel.org/rcu/20260907075829.2073224-8-kunwu.chan@linux.dev/
> > ---
> >  kernel/rcu/srcutree.c | 7 +++++--
> >  1 file changed, 5 insertions(+), 2 deletions(-)
> >
> > diff --git a/kernel/rcu/srcutree.c b/kernel/rcu/srcutree.c
> > index 84c022ec8e09..07db4e459b30 100644
> > --- a/kernel/rcu/srcutree.c
> > +++ b/kernel/rcu/srcutree.c
> > @@ -1073,8 +1073,11 @@ static void srcu_gp_end(struct srcu_struct *ssp, bool is_atomic)
> >               raw_spin_unlock_irq_rcu_node(sup);
> >       }
> >
> > -     /* Transition to big if needed. */
> > -     if (ss_state != SRCU_SIZE_SMALL && ss_state != SRCU_SIZE_BIG) {
> > +     /* Transition to big if needed, but never for atomic SRCU. */
> > +     if (ssp->srcu_reader_flavor == SRCU_READ_FLAVOR_ATOMIC &&
> > +         ss_state != SRCU_SIZE_SMALL) {
> > +             WARN_ON_ONCE(1);
> > +     } else if (ss_state != SRCU_SIZE_SMALL && ss_state != SRCU_SIZE_BIG) {
> >               if (ss_state == SRCU_SIZE_ALLOC)
> >                       init_srcu_struct_nodes(ssp, GFP_KERNEL);
> >               else
> > --
> > 2.43.0
> >
> >

Re: [PATCH v3] srcutree: Don't transition atomic SRCU to big in srcu_gp_end()

[Bradley Morgan]

On 11 September 2026 08:21:25 BST, Kunwu Chan <kunwu.chan@gmail.com> wrote:

Better late than never.

>Atomic SRCU must remain in the small size state. Warn if this
>invariant is violated and avoid transitioning to big in that case.

Fair enough.

Reviewed-by: Bradley Morgan <brads@mainlining.org>


>
>Signed-off-by: Kunwu Chan <kunwu.chan@gmail.com>
>---
>Changes since v2:
>- Also prevent the transition when the atomic SRCU size-state
>  invariant is violated, per Paul McKenney's suggestion.
>-v2: https://lore.kernel.org/rcu/20260910084629.3695383-1-kunwu.chan@gmail.com/
>
>Changes since v1:
>- Drop the !is_atomic guard from v1 and add WARN_ON_ONCE() instead,
>  per Paul McKenney's suggestion.
>- v1: https://lore.kernel.org/rcu/20260907075829.2073224-8-kunwu.chan@linux.dev/
>---
> kernel/rcu/srcutree.c | 7 +++++--
> 1 file changed, 5 insertions(+), 2 deletions(-)
>
>diff --git a/kernel/rcu/srcutree.c b/kernel/rcu/srcutree.c
>index 84c022ec8e09..07db4e459b30 100644
>--- a/kernel/rcu/srcutree.c
>+++ b/kernel/rcu/srcutree.c
>@@ -1073,8 +1073,11 @@ static void srcu_gp_end(struct srcu_struct *ssp, bool is_atomic)
> 		raw_spin_unlock_irq_rcu_node(sup);
> 	}
> 
>-	/* Transition to big if needed. */
>-	if (ss_state != SRCU_SIZE_SMALL && ss_state != SRCU_SIZE_BIG) {
>+	/* Transition to big if needed, but never for atomic SRCU. */
>+	if (ssp->srcu_reader_flavor == SRCU_READ_FLAVOR_ATOMIC &&
>+	    ss_state != SRCU_SIZE_SMALL) {
>+		WARN_ON_ONCE(1);
>+	} else if (ss_state != SRCU_SIZE_SMALL && ss_state != SRCU_SIZE_BIG) {

I like this fix.

> 		if (ss_state == SRCU_SIZE_ALLOC)
> 			init_srcu_struct_nodes(ssp, GFP_KERNEL);
> 		else
>

--- Thanks!
https://lore.kernel.org/all/EE579805-42F2-4C58-B752-F28779EEB717@grrlz.net/