Linux-ARM-Kernel Archive on lore.kernel.org
 help / color / mirror / Atom feed
* [PATCH] bus: vexpress-config: fix device_node refcount leak in vexpress_syscfg_probe()
@ 2026-06-10  3:30 Weigang He
  2026-06-10 12:38 ` Liviu Dudau
  0 siblings, 1 reply; 2+ messages in thread
From: Weigang He @ 2026-06-10  3:30 UTC (permalink / raw)
  To: Liviu Dudau
  Cc: Sudeep Holla, Lorenzo Pieralisi, Rob Herring, linux-arm-kernel,
	linux-kernel, Weigang He

vexpress_syscfg_probe() iterates the "arm,vexpress,config-bus"
compatible nodes and, for each one, takes a reference to the bridge
phandle via of_parse_phandle():

	bridge_np = of_parse_phandle(node, "arm,vexpress,config-bridge", 0);

bridge_np is only compared against pdev->dev.parent->of_node and is
never released - neither on the "continue" path when it does not match,
nor on the path that calls of_platform_populate() and falls through to
the next loop iteration. Each matching iteration leaks one device_node
reference; the leak repeats on every probe (driver bind/unbind, module
reload, or EPROBE_DEFER retry).

This is a regression of commit 557e37c05f28 ("bus: vexpress-config: add
missing of_node_put after calling of_parse_phandle"), which fixed the
equivalent leak in the predecessor function vexpress_config_populate().
Commit a5a38765ac79 ("bus: vexpress-config: simplify config bus
probing") removed that function and inlined the loop into the probe
routine, but did not carry over the of_node_put().

Use the __free(device_node) cleanup attribute on bridge_np so the
reference is released automatically at the end of each loop iteration.

Found by static analysis tool CodeQL.

Fixes: a5a38765ac79 ("bus: vexpress-config: simplify config bus probing")
Signed-off-by: Weigang He <geoffreyhe2@gmail.com>
---
 drivers/bus/vexpress-config.c | 4 ++--
 1 file changed, 2 insertions(+), 2 deletions(-)

diff --git a/drivers/bus/vexpress-config.c b/drivers/bus/vexpress-config.c
index 64ee920721ee7..cc247483d3823 100644
--- a/drivers/bus/vexpress-config.c
+++ b/drivers/bus/vexpress-config.c
@@ -390,9 +390,9 @@ static int vexpress_syscfg_probe(struct platform_device *pdev)
 	}
 
 	for_each_compatible_node(node, NULL, "arm,vexpress,config-bus") {
-		struct device_node *bridge_np;
+		struct device_node *bridge_np __free(device_node) =
+			of_parse_phandle(node, "arm,vexpress,config-bridge", 0);
 
-		bridge_np = of_parse_phandle(node, "arm,vexpress,config-bridge", 0);
 		if (bridge_np != pdev->dev.parent->of_node)
 			continue;
 

base-commit: 0f61b1860cc3f52aef9036d7235ed1f017632193
-- 
2.43.0



^ permalink raw reply related	[flat|nested] 2+ messages in thread

* Re: [PATCH] bus: vexpress-config: fix device_node refcount leak in vexpress_syscfg_probe()
  2026-06-10  3:30 [PATCH] bus: vexpress-config: fix device_node refcount leak in vexpress_syscfg_probe() Weigang He
@ 2026-06-10 12:38 ` Liviu Dudau
  0 siblings, 0 replies; 2+ messages in thread
From: Liviu Dudau @ 2026-06-10 12:38 UTC (permalink / raw)
  To: Weigang He
  Cc: Sudeep Holla, Lorenzo Pieralisi, Rob Herring, linux-arm-kernel,
	linux-kernel

On Wed, Jun 10, 2026 at 01:30:54PM +1000, Weigang He wrote:
> vexpress_syscfg_probe() iterates the "arm,vexpress,config-bus"
> compatible nodes and, for each one, takes a reference to the bridge
> phandle via of_parse_phandle():
> 
> 	bridge_np = of_parse_phandle(node, "arm,vexpress,config-bridge", 0);
> 
> bridge_np is only compared against pdev->dev.parent->of_node and is
> never released - neither on the "continue" path when it does not match,
> nor on the path that calls of_platform_populate() and falls through to
> the next loop iteration. Each matching iteration leaks one device_node
> reference; the leak repeats on every probe (driver bind/unbind, module
> reload, or EPROBE_DEFER retry).
> 
> This is a regression of commit 557e37c05f28 ("bus: vexpress-config: add
> missing of_node_put after calling of_parse_phandle"), which fixed the
> equivalent leak in the predecessor function vexpress_config_populate().
> Commit a5a38765ac79 ("bus: vexpress-config: simplify config bus
> probing") removed that function and inlined the loop into the probe
> routine, but did not carry over the of_node_put().
> 
> Use the __free(device_node) cleanup attribute on bridge_np so the
> reference is released automatically at the end of each loop iteration.
> 
> Found by static analysis tool CodeQL.
> 
> Fixes: a5a38765ac79 ("bus: vexpress-config: simplify config bus probing")
> Signed-off-by: Weigang He <geoffreyhe2@gmail.com>

Reviewed-by: Liviu Dudau <liviu.dudau@arm.com>

Thanks for the fix!

Best regards,
Liviu

> ---
>  drivers/bus/vexpress-config.c | 4 ++--
>  1 file changed, 2 insertions(+), 2 deletions(-)
> 
> diff --git a/drivers/bus/vexpress-config.c b/drivers/bus/vexpress-config.c
> index 64ee920721ee7..cc247483d3823 100644
> --- a/drivers/bus/vexpress-config.c
> +++ b/drivers/bus/vexpress-config.c
> @@ -390,9 +390,9 @@ static int vexpress_syscfg_probe(struct platform_device *pdev)
>  	}
>  
>  	for_each_compatible_node(node, NULL, "arm,vexpress,config-bus") {
> -		struct device_node *bridge_np;
> +		struct device_node *bridge_np __free(device_node) =
> +			of_parse_phandle(node, "arm,vexpress,config-bridge", 0);
>  
> -		bridge_np = of_parse_phandle(node, "arm,vexpress,config-bridge", 0);
>  		if (bridge_np != pdev->dev.parent->of_node)
>  			continue;
>  
> 
> base-commit: 0f61b1860cc3f52aef9036d7235ed1f017632193
> -- 
> 2.43.0
> 

-- 
====================
| I would like to |
| fix the world,  |
| but they're not |
| giving me the   |
 \ source code!  /
  ---------------
    ¯\_(ツ)_/¯


^ permalink raw reply	[flat|nested] 2+ messages in thread

end of thread, other threads:[~2026-06-10 12:39 UTC | newest]

Thread overview: 2+ messages (download: mbox.gz follow: Atom feed
-- links below jump to the message on this page --
2026-06-10  3:30 [PATCH] bus: vexpress-config: fix device_node refcount leak in vexpress_syscfg_probe() Weigang He
2026-06-10 12:38 ` Liviu Dudau

This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox