On 2026-11-20 09:11, Ritesh Harjani wrote:
Jyoti Gupta <[email protected]> writes:

Several PowerPC code paths acquire device tree node references but fail
to drop them on error paths, early exits.

Add the missing of_node_put() calls across PowerPC platform to ensure
proper device node reference counting.

These changes prevent device node reference leaks and improve overall
resource management in PowerPC code.

Suggested-by: Sourabh Jain <[email protected]>
Signed-off-by: Jyoti Gupta <[email protected]>
---
Changelog:
v1 -> v2:
 - Rebased onto latest upstream.
 - Dropped the powermac/smp.c hunk: there is no leak there. The loop
  iterator drops cc on continue/advance and the existing code puts it
before the break, so the added of_node_put() calls were a double put.
 - Added Suggested-by tag.
- Link to v1 - https://lore.kernel.org/all/[email protected]/
---
 arch/powerpc/kexec/file_load_64.c            | 1 +
 arch/powerpc/platforms/44x/iss4xx.c          | 2 ++
 arch/powerpc/platforms/44x/warp.c            | 1 +
 arch/powerpc/platforms/85xx/mpc85xx_8259.c   | 1 +
 arch/powerpc/platforms/86xx/pic.c            | 1 +
 arch/powerpc/platforms/pasemi/setup.c        | 1 +
 arch/powerpc/platforms/powermac/pci.c        | 1 +
 arch/powerpc/platforms/pseries/hotplug-cpu.c | 4 ++--
 8 files changed, 10 insertions(+), 2 deletions(-)

Mostly minor nits. Otherwise largely looks ok to me.

Thankyou for the review.

diff --git a/arch/powerpc/kexec/file_load_64.c b/arch/powerpc/kexec/file_load_64.c
index bd80c5fb1b1f..8d4d6a51c046 100644
--- a/arch/powerpc/kexec/file_load_64.c
+++ b/arch/powerpc/kexec/file_load_64.c
@@ -204,6 +204,7 @@ static int add_usable_mem_property(void *fdt, struct device_node *dn, if (snprintf(path, NODE_PATH_LEN, "%pOF", dn) > (NODE_PATH_LEN - 1)) {
                pr_err("Buffer (%d) too small for memory node: %pOF\n",
                       NODE_PATH_LEN, dn);
+               of_node_put(dn);
                return -EOVERFLOW;

Instead can we use goto out and use the common cleanup defined in this function?

Agreed.Will fix in v3.
        }
        kexec_dprintk("Memory node path: %s\n", path);
diff --git a/arch/powerpc/platforms/44x/iss4xx.c b/arch/powerpc/platforms/44x/iss4xx.c
index ef883d97fe15..ef92b4656d17 100644
--- a/arch/powerpc/platforms/44x/iss4xx.c
+++ b/arch/powerpc/platforms/44x/iss4xx.c
@@ -74,6 +74,8 @@ static void __init iss4xx_init_irq(void)
 #endif
        } else
                panic("Unrecognized top level interrupt controller");
+
+       of_node_put(np);
 }

 #ifdef CONFIG_SMP
diff --git a/arch/powerpc/platforms/44x/warp.c b/arch/powerpc/platforms/44x/warp.c
index 6f674f86dc85..49afd60b1c30 100644
--- a/arch/powerpc/platforms/44x/warp.c
+++ b/arch/powerpc/platforms/44x/warp.c
@@ -183,6 +183,7 @@ static int pika_setup_leds(void)
                                printk(KERN_ERR __FILE__ ": Failed to get %s led 
gpio: %d\n",
                                       led->name, error);
                                of_node_put(child);
+                               of_node_put(np);
                                goto err_cleanup_pins;
                        }

diff --git a/arch/powerpc/platforms/85xx/mpc85xx_8259.c b/arch/powerpc/platforms/85xx/mpc85xx_8259.c
index cb00d596ad80..9b7283038892 100644
--- a/arch/powerpc/platforms/85xx/mpc85xx_8259.c
+++ b/arch/powerpc/platforms/85xx/mpc85xx_8259.c
@@ -52,6 +52,7 @@ void __init mpc85xx_8259_init(void)
        cascade_irq = irq_of_parse_and_map(cascade_node, 0);
        if (!cascade_irq) {
                pr_err("i8259: Failed to map cascade interrupt\n");
+               of_node_put(cascade_node);
                return;
        }

diff --git a/arch/powerpc/platforms/86xx/pic.c b/arch/powerpc/platforms/86xx/pic.c
index 9ca36de23532..4cbcc16b1009 100644
--- a/arch/powerpc/platforms/86xx/pic.c
+++ b/arch/powerpc/platforms/86xx/pic.c
@@ -58,6 +58,7 @@ void __init mpc86xx_init_irq(void)
        cascade_irq = irq_of_parse_and_map(cascade_node, 0);
        if (!cascade_irq) {
                printk(KERN_ERR "Failed to map cascade interrupt\n");
+               of_node_put(cascade_node);
                return;
        }

diff --git a/arch/powerpc/platforms/pasemi/setup.c b/arch/powerpc/platforms/pasemi/setup.c
index 7a961f2a228f..a82855752e58 100644
--- a/arch/powerpc/platforms/pasemi/setup.c
+++ b/arch/powerpc/platforms/pasemi/setup.c
@@ -272,6 +272,7 @@ static __init void pas_init_IRQ(void)
        opprop = of_get_property(root, "platform-open-pic", &opplen);
        if (!opprop) {
                pr_err("No platform-open-pic property.\n");
+               of_node_put(mpic_node);
                of_node_put(root);
                return;
        }
diff --git a/arch/powerpc/platforms/powermac/pci.c b/arch/powerpc/platforms/powermac/pci.c
index d71359b5331c..8bc56a70b4c8 100644
--- a/arch/powerpc/platforms/powermac/pci.c
+++ b/arch/powerpc/platforms/powermac/pci.c
@@ -938,6 +938,7 @@ void __init pmac_pci_init(void)
        /* pmac_check_ht_link(); */

 #else /* CONFIG_PPC64 */
+       of_node_put(ht);

After this reference to ht, I think we can drop __maybe_unused during ht
declaration. I guess it must be added since on PPC32 it won't get
used and so compiler could complain for an unused variable.

Agreed. With of_node_put(ht) now used on both PPC32 and PPC64, the
variable is no longer set-but-unused on PPC32, so the annotation is
redundant. I'll drop __maybe_unused in v3.

Thanks,
Jyoti Gupta

Reply via email to