[PATCH v1] ARM: npcm: Fix OF node refcount leaks in SMP setup

[PATCH v1] ARM: npcm: Fix OF node refcount leaks in SMP setup

[Yuho Choi]

npcm7xx_smp_boot_secondary() and npcm7xx_smp_prepare_cpus() look up
the GCR and SCU nodes with of_find_compatible_node(). The returned
nodes are used for of_iomap(), but the node references are never
released.

of_iomap() does not consume the device node reference, and iounmap()
only releases the MMIO mapping. Drop each node reference after the
corresponding mapping attempt.

Fixes: 7bffa14c9aed ("arm: npcm: add basic support for Nuvoton BMCs")
Signed-off-by: Yuho Choi <dbgh9129@gmail.com>
---
 arch/arm/mach-npcm/platsmp.c | 2 ++
 1 file changed, 2 insertions(+)

diff --git a/arch/arm/mach-npcm/platsmp.c b/arch/arm/mach-npcm/platsmp.c
index 41891d3aa124..4c1fc9983746 100644
--- a/arch/arm/mach-npcm/platsmp.c
+++ b/arch/arm/mach-npcm/platsmp.c
@@ -32,6 +32,7 @@ static int npcm7xx_smp_boot_secondary(unsigned int cpu,
 		goto out;
 	}
 	gcr_base = of_iomap(gcr_np, 0);
+	of_node_put(gcr_np);
 	if (!gcr_base) {
 		pr_err("could not iomap gcr");
 		ret = -ENOMEM;
@@ -60,6 +61,7 @@ static void __init npcm7xx_smp_prepare_cpus(unsigned int max_cpus)
 		return;
 	}
 	scu_base = of_iomap(scu_np, 0);
+	of_node_put(scu_np);
 	if (!scu_base) {
 		pr_err("could not iomap scu");
 		return;
-- 
2.43.0

Re: [PATCH v1] ARM: npcm: Fix OF node refcount leaks in SMP setup

[최유호]

Hi,

Just a gentle ping on this patch.

I would appreciate any feedback when you have a chance to review this.

Thanks

Re: [PATCH v1] ARM: npcm: Fix OF node refcount leaks in SMP setup

[Andrew Jeffery]

On Sun, 24 May 2026 23:38:46 -0400, Yuho Choi wrote:
> npcm7xx_smp_boot_secondary() and npcm7xx_smp_prepare_cpus() look up
> the GCR and SCU nodes with of_find_compatible_node(). The returned
> nodes are used for of_iomap(), but the node references are never
> released.
> 
> of_iomap() does not consume the device node reference, and iounmap()
> only releases the MMIO mapping. Drop each node reference after the
> corresponding mapping attempt.
> 
> [...]

Thanks, I've applied this to the BMC tree.

-- 
Andrew Jeffery <andrew@codeconstruct.com.au>