[PATCH 0/4] Make cpumask_of_node() robust against NUMA_NO_NODE

[PATCH 0/4] Make cpumask_of_node() robust against NUMA_NO_NODE

[John Garry]

This series aims to remedy an issue that not all per-arch versions of
cpumask_of_node() are robust against NUMA_NO_NODE.

In my view, cpumask_of_node() should be able to handle NUMA_NO_NODE. This
is because NUMA_NO_NODE is a valid index from the following flow, where
the device NUMA node is not set (from default):

device_initialize(dev)
	set_dev_node(dev, NUMA_NO_NODE);

mask = cpumask_of_node(dev_to_node(dev));

The CONFIG_DEBUG_PER_CPU_MAPS=n x86 version cpumask_of_node() would
produce an array out-of-index issue (when passed NUMA_NO_NODE), which I
think is attempted to be worked around here:
https://lore.kernel.org/linux-scsi/cf0f9085-6c87-4dd5-9114-925723e68495@oracle.com/T/#mdedb68052e419b4bfca9ce45bb33b58988018945

I also see a CVE which also looks related:
https://nvd.nist.gov/vuln/detail/cve-2024-39277

Each per-arch version could be picked up separately, as can the
asm-generic change.

John Garry (4):
  include/asm-generic/topology.h: Remove unused definition of
    cpumask_of_node()
  LoongArch: Make cpumask_of_node() robust against NUMA_NO_NODE
  MIPS: Loongson: Make cpumask_of_node() robust against NUMA_NO_NODE
  x86/cpu/topology: Make cpumask_of_node() robust against NUMA_NO_NODE

 arch/loongarch/include/asm/topology.h            | 4 +++-
 arch/mips/include/asm/mach-loongson64/topology.h | 4 +++-
 arch/x86/include/asm/topology.h                  | 2 ++
 arch/x86/mm/numa.c                               | 2 ++
 include/asm-generic/topology.h                   | 8 ++------
 5 files changed, 12 insertions(+), 8 deletions(-)

-- 
2.43.5

[PATCH 1/4] include/asm-generic/topology.h: Remove unused definition of cpumask_of_node()

[John Garry]

The definition of cpumask_of_node() in question is guarded by conflicting
CONFIG_NUMA and !CONFIG_NUMA checks, so remove it.

Signed-off-by: John Garry <john.g.garry@oracle.com>
---
 include/asm-generic/topology.h | 8 ++------
 1 file changed, 2 insertions(+), 6 deletions(-)

diff --git a/include/asm-generic/topology.h b/include/asm-generic/topology.h
index 4dbe715be65b4..9865ba48c5b16 100644
--- a/include/asm-generic/topology.h
+++ b/include/asm-generic/topology.h
@@ -45,11 +45,7 @@
 #endif
 
 #ifndef cpumask_of_node
-  #ifdef CONFIG_NUMA
-    #define cpumask_of_node(node)	((node) == 0 ? cpu_online_mask : cpu_none_mask)
-  #else
-    #define cpumask_of_node(node)	((void)(node), cpu_online_mask)
-  #endif
+#define cpumask_of_node(node)	((void)(node), cpu_online_mask)
 #endif
 #ifndef pcibus_to_node
 #define pcibus_to_node(bus)	((void)(bus), -1)
@@ -61,7 +57,7 @@
 				 cpumask_of_node(pcibus_to_node(bus)))
 #endif
 
-#endif	/* CONFIG_NUMA */
+#endif	/* !CONFIG_NUMA */
 
 #if !defined(CONFIG_NUMA) || !defined(CONFIG_HAVE_MEMORYLESS_NODES)
 
-- 
2.43.5

[PATCH 2/4] LoongArch: Make cpumask_of_node() robust against NUMA_NO_NODE

[John Garry]

The arch definition of cpumask_of_node() cannot handle NUMA_NO_NODE - which
is a valid index - so add a check for this.

Signed-off-by: John Garry <john.g.garry@oracle.com>
---
 arch/loongarch/include/asm/topology.h | 4 +++-
 1 file changed, 3 insertions(+), 1 deletion(-)

diff --git a/arch/loongarch/include/asm/topology.h b/arch/loongarch/include/asm/topology.h
index f06e7ff25bb7c..9857e4c20023c 100644
--- a/arch/loongarch/include/asm/topology.h
+++ b/arch/loongarch/include/asm/topology.h
@@ -12,7 +12,9 @@
 
 extern cpumask_t cpus_on_node[];
 
-#define cpumask_of_node(node)  (&cpus_on_node[node])
+#define cpumask_of_node(node)  ((node) == NUMA_NO_NODE ?	\
+				cpu_all_mask :			\
+				&cpus_on_node[node])
 
 struct pci_bus;
 extern int pcibus_to_node(struct pci_bus *);
-- 
2.43.5

[PATCH 3/4] MIPS: Loongson: Make cpumask_of_node() robust against NUMA_NO_NODE

[John Garry]

The arch definition of cpumask_of_node() cannot handle NUMA_NO_NODE - which
is a valid index - so add a check for this.

Signed-off-by: John Garry <john.g.garry@oracle.com>
---
 arch/mips/include/asm/mach-loongson64/topology.h | 4 +++-
 1 file changed, 3 insertions(+), 1 deletion(-)

diff --git a/arch/mips/include/asm/mach-loongson64/topology.h b/arch/mips/include/asm/mach-loongson64/topology.h
index 3414a1fd17835..e897dddc1366e 100644
--- a/arch/mips/include/asm/mach-loongson64/topology.h
+++ b/arch/mips/include/asm/mach-loongson64/topology.h
@@ -7,7 +7,9 @@
 #define cpu_to_node(cpu)	(cpu_logical_map(cpu) >> 2)
 
 extern cpumask_t __node_cpumask[];
-#define cpumask_of_node(node)	(&__node_cpumask[node])
+#define cpumask_of_node(node)	  ((node) == NUMA_NO_NODE ?	\
+				cpu_all_mask :			\
+				&__node_cpumask[node])
 
 struct pci_bus;
 extern int pcibus_to_node(struct pci_bus *);
-- 
2.43.5

[PATCH 4/4] x86/cpu/topology: Make cpumask_of_node() robust against NUMA_NO_NODE

[John Garry]

The arch definitions of cpumask_of_node() cannot handle NUMA_NO_NODE -
which is a valid index - so add checks for this.

Signed-off-by: John Garry <john.g.garry@oracle.com>
---
 arch/x86/include/asm/topology.h | 2 ++
 arch/x86/mm/numa.c              | 2 ++
 2 files changed, 4 insertions(+)

diff --git a/arch/x86/include/asm/topology.h b/arch/x86/include/asm/topology.h
index 1fadf0cf520c5..b51f2e771a582 100644
--- a/arch/x86/include/asm/topology.h
+++ b/arch/x86/include/asm/topology.h
@@ -69,6 +69,8 @@ extern const struct cpumask *cpumask_of_node(int node);
 /* Returns a pointer to the cpumask of CPUs on Node 'node'. */
 static inline const struct cpumask *cpumask_of_node(int node)
 {
+	if (node == NUMA_NO_NODE)
+		return cpu_all_mask;
 	return node_to_cpumask_map[node];
 }
 #endif
diff --git a/arch/x86/mm/numa.c b/arch/x86/mm/numa.c
index 7a97327140df8..0be94f4682232 100644
--- a/arch/x86/mm/numa.c
+++ b/arch/x86/mm/numa.c
@@ -408,6 +408,8 @@ void numa_remove_cpu(unsigned int cpu)
  */
 const struct cpumask *cpumask_of_node(int node)
 {
+	if (node == NUMA_NO_NODE)
+		return cpu_all_mask;
 	if ((unsigned)node >= nr_node_ids) {
 		printk(KERN_WARNING
 			"cpumask_of_node(%d): (unsigned)node >= nr_node_ids(%u)\n",
-- 
2.43.5

Re: [PATCH 2/4] LoongArch: Make cpumask_of_node() robust against NUMA_NO_NODE

[Huacai Chen]

Hi, John,

On Mon, Jan 5, 2026 at 6:07 PM John Garry <john.g.garry@oracle.com> wrote:
>
> The arch definition of cpumask_of_node() cannot handle NUMA_NO_NODE - which
> is a valid index - so add a check for this.
>
> Signed-off-by: John Garry <john.g.garry@oracle.com>
> ---
>  arch/loongarch/include/asm/topology.h | 4 +++-
>  1 file changed, 3 insertions(+), 1 deletion(-)
>
> diff --git a/arch/loongarch/include/asm/topology.h b/arch/loongarch/include/asm/topology.h
> index f06e7ff25bb7c..9857e4c20023c 100644
> --- a/arch/loongarch/include/asm/topology.h
> +++ b/arch/loongarch/include/asm/topology.h
> @@ -12,7 +12,9 @@
>
>  extern cpumask_t cpus_on_node[];
>
> -#define cpumask_of_node(node)  (&cpus_on_node[node])
> +#define cpumask_of_node(node)  ((node) == NUMA_NO_NODE ?       \
> +                               cpu_all_mask :                  \
> +                               &cpus_on_node[node])
You can define it in one line, so does the MIPS version.

Huacai

>
>  struct pci_bus;
>  extern int pcibus_to_node(struct pci_bus *);
> --
> 2.43.5
>

Re: [PATCH 2/4] LoongArch: Make cpumask_of_node() robust against NUMA_NO_NODE

[John Garry]

On 05/01/2026 12:54, Huacai Chen wrote:
> Hi, John,
> 
> On Mon, Jan 5, 2026 at 6:07 PM John Garry <john.g.garry@oracle.com> wrote:
>>
>> The arch definition of cpumask_of_node() cannot handle NUMA_NO_NODE - which
>> is a valid index - so add a check for this.
>>
>> Signed-off-by: John Garry <john.g.garry@oracle.com>
>> ---
>>   arch/loongarch/include/asm/topology.h | 4 +++-
>>   1 file changed, 3 insertions(+), 1 deletion(-)
>>
>> diff --git a/arch/loongarch/include/asm/topology.h b/arch/loongarch/include/asm/topology.h
>> index f06e7ff25bb7c..9857e4c20023c 100644
>> --- a/arch/loongarch/include/asm/topology.h
>> +++ b/arch/loongarch/include/asm/topology.h
>> @@ -12,7 +12,9 @@
>>
>>   extern cpumask_t cpus_on_node[];
>>
>> -#define cpumask_of_node(node)  (&cpus_on_node[node])
>> +#define cpumask_of_node(node)  ((node) == NUMA_NO_NODE ?       \
>> +                               cpu_all_mask :                  \
>> +                               &cpus_on_node[node])
> You can define it in one line, so does the MIPS version.
> 

ok, I can do it that way if you prefer. The reason I had it in flow was:
- a single line will overflow 80 characters, which some people still 
prefer not doing
- following example of other archs and general coding style to split 
macros functions across multiple lines.

Thanks,
John