[PATCH] riscv: panic from init_IRQ if IRQ handler stacks cannot be allocated

[PATCH] riscv: panic from init_IRQ if IRQ handler stacks cannot be allocated

[Osama Abdelkader]

init_irq_stacks() and init_irq_scs() may fail when arch_alloc_vmap_stack
or scs_alloc return NULL. Return -ENOMEM from both and call panic() once
from init_IRQ(), covering per-CPU IRQ stacks and shadow IRQ stacks
consistently.

Signed-off-by: Osama Abdelkader <osama.abdelkader@gmail.com>
---
 arch/riscv/kernel/irq.c | 43 ++++++++++++++++++++++++++++++-----------
 1 file changed, 32 insertions(+), 11 deletions(-)

diff --git a/arch/riscv/kernel/irq.c b/arch/riscv/kernel/irq.c
index b6af20bc300f..43392276f7f8 100644
--- a/arch/riscv/kernel/irq.c
+++ b/arch/riscv/kernel/irq.c
@@ -5,6 +5,7 @@
  * Copyright (C) 2018 Christoph Hellwig
  */
 
+#include <linux/errno.h>
 #include <linux/interrupt.h>
 #include <linux/irqchip.h>
 #include <linux/irqdomain.h>
@@ -75,41 +76,53 @@ DECLARE_PER_CPU(ulong *, irq_shadow_call_stack_ptr);
 DEFINE_PER_CPU(ulong *, irq_shadow_call_stack_ptr);
 #endif
 
-static void init_irq_scs(void)
+static int __init init_irq_scs(void)
 {
 	int cpu;
+	void *s;
 
 	if (!scs_is_enabled())
-		return;
+		return 0;
 
-	for_each_possible_cpu(cpu)
-		per_cpu(irq_shadow_call_stack_ptr, cpu) =
-			scs_alloc(cpu_to_node(cpu));
+	for_each_possible_cpu(cpu) {
+		s = scs_alloc(cpu_to_node(cpu));
+		if (!s)
+			return -ENOMEM;
+		per_cpu(irq_shadow_call_stack_ptr, cpu) = s;
+	}
+
+	return 0;
 }
 
 DEFINE_PER_CPU(ulong *, irq_stack_ptr);
 
 #ifdef CONFIG_VMAP_STACK
-static void init_irq_stacks(void)
+static int __init init_irq_stacks(void)
 {
 	int cpu;
 	ulong *p;
 
 	for_each_possible_cpu(cpu) {
 		p = arch_alloc_vmap_stack(IRQ_STACK_SIZE, cpu_to_node(cpu));
+		if (!p)
+			return -ENOMEM;
 		per_cpu(irq_stack_ptr, cpu) = p;
 	}
+
+	return 0;
 }
 #else
 /* irq stack only needs to be 16 byte aligned - not IRQ_STACK_SIZE aligned. */
 DEFINE_PER_CPU_ALIGNED(ulong [IRQ_STACK_SIZE/sizeof(ulong)], irq_stack);
 
-static void init_irq_stacks(void)
+static int __init init_irq_stacks(void)
 {
 	int cpu;
 
 	for_each_possible_cpu(cpu)
 		per_cpu(irq_stack_ptr, cpu) = per_cpu(irq_stack, cpu);
+
+	return 0;
 }
 #endif /* CONFIG_VMAP_STACK */
 
@@ -129,8 +142,15 @@ void do_softirq_own_stack(void)
 #endif /* CONFIG_SOFTIRQ_ON_OWN_STACK */
 
 #else
-static void init_irq_scs(void) {}
-static void init_irq_stacks(void) {}
+static int __init init_irq_scs(void)
+{
+	return 0;
+}
+
+static int __init init_irq_stacks(void)
+{
+	return 0;
+}
 #endif /* CONFIG_IRQ_STACKS */
 
 int arch_show_interrupts(struct seq_file *p, int prec)
@@ -141,8 +161,9 @@ int arch_show_interrupts(struct seq_file *p, int prec)
 
 void __init init_IRQ(void)
 {
-	init_irq_scs();
-	init_irq_stacks();
+	if (init_irq_stacks() || init_irq_scs())
+		panic("Failed to allocate IRQ stack resources\n");
+
 	irqchip_init();
 	if (!handle_arch_irq)
 		panic("No interrupt controller found.");
-- 
2.43.0

Re: [PATCH] riscv: panic from init_IRQ if IRQ handler stacks cannot be allocated

[Paul Walmsley]

Hi,

On Fri, 27 Mar 2026, Osama Abdelkader wrote:

> init_irq_stacks() and init_irq_scs() may fail when arch_alloc_vmap_stack
> or scs_alloc return NULL. Return -ENOMEM from both and call panic() once
> from init_IRQ(), covering per-CPU IRQ stacks and shadow IRQ stacks
> consistently.
> 
> Signed-off-by: Osama Abdelkader <osama.abdelkader@gmail.com>

This patch appears to be written with AI assistance.  If so, it should be 
noted in the description.  Please see

  https://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git/tree/Documentation/process/coding-assistants.rst

> ---
>  arch/riscv/kernel/irq.c | 43 ++++++++++++++++++++++++++++++-----------
>  1 file changed, 32 insertions(+), 11 deletions(-)
> 
> diff --git a/arch/riscv/kernel/irq.c b/arch/riscv/kernel/irq.c
> index b6af20bc300f..43392276f7f8 100644
> --- a/arch/riscv/kernel/irq.c
> +++ b/arch/riscv/kernel/irq.c
> @@ -5,6 +5,7 @@
>   * Copyright (C) 2018 Christoph Hellwig
>   */
>  
> +#include <linux/errno.h>
>  #include <linux/interrupt.h>
>  #include <linux/irqchip.h>
>  #include <linux/irqdomain.h>
> @@ -75,41 +76,53 @@ DECLARE_PER_CPU(ulong *, irq_shadow_call_stack_ptr);
>  DEFINE_PER_CPU(ulong *, irq_shadow_call_stack_ptr);
>  #endif
>  
> -static void init_irq_scs(void)
> +static int __init init_irq_scs(void)
>  {
>  	int cpu;
> +	void *s;
>  
>  	if (!scs_is_enabled())
> -		return;
> +		return 0;
>  
> -	for_each_possible_cpu(cpu)
> -		per_cpu(irq_shadow_call_stack_ptr, cpu) =
> -			scs_alloc(cpu_to_node(cpu));
> +	for_each_possible_cpu(cpu) {
> +		s = scs_alloc(cpu_to_node(cpu));
> +		if (!s)
> +			return -ENOMEM;

I don't see why this bothers to return an error value if the sole caller 
is just going to panic.  Best to panic here.  Then no one reviewing this 
patch needs to be concerned about memory leaks.

> +		per_cpu(irq_shadow_call_stack_ptr, cpu) = s;
> +	}
> +
> +	return 0;
>  }
>  
>  DEFINE_PER_CPU(ulong *, irq_stack_ptr);
>  
>  #ifdef CONFIG_VMAP_STACK
> -static void init_irq_stacks(void)
> +static int __init init_irq_stacks(void)
>  {
>  	int cpu;
>  	ulong *p;
>  
>  	for_each_possible_cpu(cpu) {
>  		p = arch_alloc_vmap_stack(IRQ_STACK_SIZE, cpu_to_node(cpu));
> +		if (!p)
> +			return -ENOMEM;

Same comment as the above.

>  		per_cpu(irq_stack_ptr, cpu) = p;
>  	}
> +
> +	return 0;
>  }
>  #else
>  /* irq stack only needs to be 16 byte aligned - not IRQ_STACK_SIZE aligned. */
>  DEFINE_PER_CPU_ALIGNED(ulong [IRQ_STACK_SIZE/sizeof(ulong)], irq_stack);
>  
> -static void init_irq_stacks(void)
> +static int __init init_irq_stacks(void)
>  {
>  	int cpu;
>  
>  	for_each_possible_cpu(cpu)
>  		per_cpu(irq_stack_ptr, cpu) = per_cpu(irq_stack, cpu);
> +
> +	return 0;

Why?

>  }
>  #endif /* CONFIG_VMAP_STACK */
>  
> @@ -129,8 +142,15 @@ void do_softirq_own_stack(void)
>  #endif /* CONFIG_SOFTIRQ_ON_OWN_STACK */
>  
>  #else
> -static void init_irq_scs(void) {}
> -static void init_irq_stacks(void) {}
> +static int __init init_irq_scs(void)
> +{
> +	return 0;
> +}
> +
> +static int __init init_irq_stacks(void)
> +{
> +	return 0;
> +}
>  #endif /* CONFIG_IRQ_STACKS */
>  
>  int arch_show_interrupts(struct seq_file *p, int prec)
> @@ -141,8 +161,9 @@ int arch_show_interrupts(struct seq_file *p, int prec)
>  
>  void __init init_IRQ(void)
>  {
> -	init_irq_scs();
> -	init_irq_stacks();
> +	if (init_irq_stacks() || init_irq_scs())
> +		panic("Failed to allocate IRQ stack resources\n");

Any reason why the initialization order is reversed?

> +
>  	irqchip_init();
>  	if (!handle_arch_irq)
>  		panic("No interrupt controller found.");
> -- 
> 2.43.0


- Paul

Re: [PATCH] riscv: panic from init_IRQ if IRQ handler stacks cannot be allocated

[Osama Abdelkader]

Hi Paul,

On Thu, Apr 02, 2026 at 04:11:59PM -0600, Paul Walmsley wrote:
> Hi,
> 
> On Fri, 27 Mar 2026, Osama Abdelkader wrote:
> 
> > init_irq_stacks() and init_irq_scs() may fail when arch_alloc_vmap_stack
> > or scs_alloc return NULL. Return -ENOMEM from both and call panic() once
> > from init_IRQ(), covering per-CPU IRQ stacks and shadow IRQ stacks
> > consistently.
> > 
> > Signed-off-by: Osama Abdelkader <osama.abdelkader@gmail.com>
> 
> This patch appears to be written with AI assistance.  If so, it should be 
> noted in the description.  Please see
> 
>   https://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git/tree/Documentation/process/coding-assistants.rst
> 

This is actually riscv version of arm64 patch:
[PATCH] arm64: panic if IRQ shadow call stack allocation fails
https://lore.kernel.org/lkml/20260324161545.5441-1-osama.abdelkader@gmail.com/
where I initially used panic, but in review, returning -ENOMEM was suggested,
and this patch also, so I submitted v2 there:
[PATCH v2] arm64: panic from init_IRQ if IRQ handler stacks cannot be allocated
https://lore.kernel.org/lkml/20260326225755.50297-1-osama.abdelkader@gmail.com/

> > ---
> >  arch/riscv/kernel/irq.c | 43 ++++++++++++++++++++++++++++++-----------
> >  1 file changed, 32 insertions(+), 11 deletions(-)
> > 
> > diff --git a/arch/riscv/kernel/irq.c b/arch/riscv/kernel/irq.c
> > index b6af20bc300f..43392276f7f8 100644
> > --- a/arch/riscv/kernel/irq.c
> > +++ b/arch/riscv/kernel/irq.c
> > @@ -5,6 +5,7 @@
> >   * Copyright (C) 2018 Christoph Hellwig
> >   */
> >  
> > +#include <linux/errno.h>
> >  #include <linux/interrupt.h>
> >  #include <linux/irqchip.h>
> >  #include <linux/irqdomain.h>
> > @@ -75,41 +76,53 @@ DECLARE_PER_CPU(ulong *, irq_shadow_call_stack_ptr);
> >  DEFINE_PER_CPU(ulong *, irq_shadow_call_stack_ptr);
> >  #endif
> >  
> > -static void init_irq_scs(void)
> > +static int __init init_irq_scs(void)
> >  {
> >  	int cpu;
> > +	void *s;
> >  
> >  	if (!scs_is_enabled())
> > -		return;
> > +		return 0;
> >  
> > -	for_each_possible_cpu(cpu)
> > -		per_cpu(irq_shadow_call_stack_ptr, cpu) =
> > -			scs_alloc(cpu_to_node(cpu));
> > +	for_each_possible_cpu(cpu) {
> > +		s = scs_alloc(cpu_to_node(cpu));
> > +		if (!s)
> > +			return -ENOMEM;
> 
> I don't see why this bothers to return an error value if the sole caller 
> is just going to panic.  Best to panic here.  Then no one reviewing this 
> patch needs to be concerned about memory leaks.
> 

Okay, I'm going to use panic in place instead.

> > +		per_cpu(irq_shadow_call_stack_ptr, cpu) = s;
> > +	}
> > +
> > +	return 0;
> >  }
> >  
> >  DEFINE_PER_CPU(ulong *, irq_stack_ptr);
> >  
> >  #ifdef CONFIG_VMAP_STACK
> > -static void init_irq_stacks(void)
> > +static int __init init_irq_stacks(void)
> >  {
> >  	int cpu;
> >  	ulong *p;
> >  
> >  	for_each_possible_cpu(cpu) {
> >  		p = arch_alloc_vmap_stack(IRQ_STACK_SIZE, cpu_to_node(cpu));
> > +		if (!p)
> > +			return -ENOMEM;
> 
> Same comment as the above.
> 
> >  		per_cpu(irq_stack_ptr, cpu) = p;
> >  	}
> > +
> > +	return 0;
> >  }
> >  #else
> >  /* irq stack only needs to be 16 byte aligned - not IRQ_STACK_SIZE aligned. */
> >  DEFINE_PER_CPU_ALIGNED(ulong [IRQ_STACK_SIZE/sizeof(ulong)], irq_stack);
> >  
> > -static void init_irq_stacks(void)
> > +static int __init init_irq_stacks(void)
> >  {
> >  	int cpu;
> >  
> >  	for_each_possible_cpu(cpu)
> >  		per_cpu(irq_stack_ptr, cpu) = per_cpu(irq_stack, cpu);
> > +
> > +	return 0;
> 
> Why?
> 
> >  }
> >  #endif /* CONFIG_VMAP_STACK */
> >  
> > @@ -129,8 +142,15 @@ void do_softirq_own_stack(void)
> >  #endif /* CONFIG_SOFTIRQ_ON_OWN_STACK */
> >  
> >  #else
> > -static void init_irq_scs(void) {}
> > -static void init_irq_stacks(void) {}
> > +static int __init init_irq_scs(void)
> > +{
> > +	return 0;
> > +}
> > +
> > +static int __init init_irq_stacks(void)
> > +{
> > +	return 0;
> > +}
> >  #endif /* CONFIG_IRQ_STACKS */
> >  
> >  int arch_show_interrupts(struct seq_file *p, int prec)
> > @@ -141,8 +161,9 @@ int arch_show_interrupts(struct seq_file *p, int prec)
> >  
> >  void __init init_IRQ(void)
> >  {
> > -	init_irq_scs();
> > -	init_irq_stacks();
> > +	if (init_irq_stacks() || init_irq_scs())
> > +		panic("Failed to allocate IRQ stack resources\n");
> 
> Any reason why the initialization order is reversed?
> 
It's a mistake, will reverse it back.

Thanks,
Osama

> > +
> >  	irqchip_init();
> >  	if (!handle_arch_irq)
> >  		panic("No interrupt controller found.");
> > -- 
> > 2.43.0
> 
> 
> - Paul