diff mbox series

usb: typec: anx7411: Use scope-based resource management in anx7411_typec_port_probe()

Message ID 889729ac-3fc5-4666-b9f5-ce6e588a341a@web.de (mailing list archive)
State Superseded
Headers show
Series usb: typec: anx7411: Use scope-based resource management in anx7411_typec_port_probe() | expand

Commit Message

Markus Elfring June 5, 2024, 5:11 p.m. UTC 
From: Markus Elfring <elfring@users.sourceforge.net>
Date: Wed, 5 Jun 2024 18:56:19 +0200

Scope-based resource management became supported also for another
programming interface by contributions of Jonathan Cameron on 2024-02-17.
See also the commit 59ed5e2d505bf5f9b4af64d0021cd0c96aec1f7c ("device
property: Add cleanup.h based fwnode_handle_put() scope based cleanup.").

* Thus use the attribute “__free(fwnode_handle)”.

* Reduce the scope for the local variable “fwnode”.

Fixes: fe6d8a9c8e64 ("usb: typec: anx7411: Add Analogix PD ANX7411 support")
Signed-off-by: Markus Elfring <elfring@users.sourceforge.net>
---
 drivers/usb/typec/anx7411.c | 7 ++++---
 1 file changed, 4 insertions(+), 3 deletions(-)

--
2.45.1

Comments

Heikki Krogerus June 6, 2024, 7:50 a.m. UTC | #1 
On Wed, Jun 05, 2024 at 07:11:04PM +0200, Markus Elfring wrote:
> From: Markus Elfring <elfring@users.sourceforge.net>
> Date: Wed, 5 Jun 2024 18:56:19 +0200
> 
> Scope-based resource management became supported also for another
> programming interface by contributions of Jonathan Cameron on 2024-02-17.
> See also the commit 59ed5e2d505bf5f9b4af64d0021cd0c96aec1f7c ("device
> property: Add cleanup.h based fwnode_handle_put() scope based cleanup.").
> 
> * Thus use the attribute “__free(fwnode_handle)”.
> 
> * Reduce the scope for the local variable “fwnode”.
> 
> Fixes: fe6d8a9c8e64 ("usb: typec: anx7411: Add Analogix PD ANX7411 support")
> Signed-off-by: Markus Elfring <elfring@users.sourceforge.net>

Was the fwnode leaked, or why else is this a "fix"? It's not clear
from the commit message. In any case:

Reviewed-by: Heikki Krogerus <heikki.krogerus@linux.intel.com>

> ---
>  drivers/usb/typec/anx7411.c | 7 ++++---
>  1 file changed, 4 insertions(+), 3 deletions(-)
> 
> diff --git a/drivers/usb/typec/anx7411.c b/drivers/usb/typec/anx7411.c
> index b12a07edc71b..9fb52f233a30 100644
> --- a/drivers/usb/typec/anx7411.c
> +++ b/drivers/usb/typec/anx7411.c
> @@ -16,6 +16,7 @@
>  #include <linux/of_graph.h>
>  #include <linux/of_platform.h>
>  #include <linux/pm_runtime.h>
> +#include <linux/property.h>
>  #include <linux/regulator/consumer.h>
>  #include <linux/slab.h>
>  #include <linux/types.h>
> @@ -1142,11 +1143,11 @@ static int anx7411_typec_port_probe(struct anx7411_data *ctx,
>  {
>  	struct typec_capability *cap = &ctx->typec.caps;
>  	struct typec_params *typecp = &ctx->typec;
> -	struct fwnode_handle *fwnode;
>  	const char *buf;
>  	int ret, i;
> 
> -	fwnode = device_get_named_child_node(dev, "connector");
> +	struct fwnode_handle *fwnode __free(fwnode_handle)
> +				     = device_get_named_child_node(dev, "connector");
>  	if (!fwnode)
>  		return -EINVAL;
> 
> @@ -1237,7 +1238,7 @@ static int anx7411_typec_port_probe(struct anx7411_data *ctx,
>  		typecp->caps_flags |= HAS_SINK_WATT;
>  	}
> 
> -	cap->fwnode = fwnode;
> +	cap->fwnode = no_free_ptr(fwnode);
> 
>  	ctx->typec.role_sw = usb_role_switch_get(dev);
>  	if (IS_ERR(ctx->typec.role_sw)) {
> --
> 2.45.1
Markus Elfring June 6, 2024, 8:40 a.m. UTC | #2 
>> Scope-based resource management became supported also for another
>> programming interface by contributions of Jonathan Cameron on 2024-02-17.
>> See also the commit 59ed5e2d505bf5f9b4af64d0021cd0c96aec1f7c ("device
>> property: Add cleanup.h based fwnode_handle_put() scope based cleanup.").
>>
>> * Thus use the attribute “__free(fwnode_handle)”.
>>
>> * Reduce the scope for the local variable “fwnode”.
>>
>> Fixes: fe6d8a9c8e64 ("usb: typec: anx7411: Add Analogix PD ANX7411 support")
>> Signed-off-by: Markus Elfring <elfring@users.sourceforge.net>
>
> Was the fwnode leaked, or why else is this a "fix"? It's not clear
> from the commit message.
Can you notice that fwnode_handle_put() calls were forgotten
in the mentioned function implementation?
https://elixir.bootlin.com/linux/v6.10-rc2/source/drivers/usb/typec/anx7411.c#L1140

I propose another code cleanup accordingly.


Will development attention grow anyhow for information in the available
API documentation?

Example:
device_get_named_child_node()
https://elixir.bootlin.com/linux/v6.10-rc2/source/drivers/base/property.c#L839

Regards,
Markus
Jonathan Cameron June 6, 2024, 9:17 a.m. UTC | #3 
On Wed, 5 Jun 2024 19:11:04 +0200
Markus Elfring <Markus.Elfring@web.de> wrote:

> From: Markus Elfring <elfring@users.sourceforge.net>
> Date: Wed, 5 Jun 2024 18:56:19 +0200
> 
> Scope-based resource management became supported also for another
> programming interface by contributions of Jonathan Cameron on 2024-02-17.
> See also the commit 59ed5e2d505bf5f9b4af64d0021cd0c96aec1f7c ("device
> property: Add cleanup.h based fwnode_handle_put() scope based cleanup.").
> 
> * Thus use the attribute “__free(fwnode_handle)”.
> 
> * Reduce the scope for the local variable “fwnode”.
> 
> Fixes: fe6d8a9c8e64 ("usb: typec: anx7411: Add Analogix PD ANX7411 support")
> Signed-off-by: Markus Elfring <elfring@users.sourceforge.net>

Hi Markus,

Good catch. However in this case I think this is insufficient.
Also your patch description should more clearly state the bug rather
and impacts (a potential resource leak).

I'm not 100% sure how this should work though.

If the expectation is that a reference to the fwnode is held when we
enter typec_register_port(), then if that errors out then we
need fwnode_handle_put().

If expectation is that the reference is not held, then we should
always call fwnode_handle_put() before that is called.
Internally it just uses this to fill in port->dev.fwnode.

Given typec_get_fw_cap() is called from there and doesn't get a reference
I think expectation is that the fwnode is held by the driver calling
typec_register_port() until that is unregistered.

Hence should be put in the error path of that call in the calling driver.

	ctx->typec.port = typec_register_port(dev, cap);
	if (IS_ERR(ctx->typec.port)) {
		// fwnode_handle_put() in here.
 		ret = PTR_ERR(ctx->typec.port);
		ctx->typec.port = NULL;
		dev_err(dev, "Failed to register type c port %d\n", ret);
		return ret;
	}

That makes it tricky to use no_free_ptr() so I wonder if this is
a case where the old fashioned fix of adding all the relevant
fwnode_handle_put() calls is the better option.  The __free()
approach doesn't always fit.

Jonathan

 
> ---
>  drivers/usb/typec/anx7411.c | 7 ++++---
>  1 file changed, 4 insertions(+), 3 deletions(-)
> 
> diff --git a/drivers/usb/typec/anx7411.c b/drivers/usb/typec/anx7411.c
> index b12a07edc71b..9fb52f233a30 100644
> --- a/drivers/usb/typec/anx7411.c
> +++ b/drivers/usb/typec/anx7411.c
> @@ -16,6 +16,7 @@
>  #include <linux/of_graph.h>
>  #include <linux/of_platform.h>
>  #include <linux/pm_runtime.h>
> +#include <linux/property.h>
>  #include <linux/regulator/consumer.h>
>  #include <linux/slab.h>
>  #include <linux/types.h>
> @@ -1142,11 +1143,11 @@ static int anx7411_typec_port_probe(struct anx7411_data *ctx,
>  {
>  	struct typec_capability *cap = &ctx->typec.caps;
>  	struct typec_params *typecp = &ctx->typec;
> -	struct fwnode_handle *fwnode;
>  	const char *buf;
>  	int ret, i;
> 
> -	fwnode = device_get_named_child_node(dev, "connector");
> +	struct fwnode_handle *fwnode __free(fwnode_handle)
> +				     = device_get_named_child_node(dev, "connector");
>  	if (!fwnode)
>  		return -EINVAL;
> 
> @@ -1237,7 +1238,7 @@ static int anx7411_typec_port_probe(struct anx7411_data *ctx,
>  		typecp->caps_flags |= HAS_SINK_WATT;
>  	}
> 
> -	cap->fwnode = fwnode;
> +	cap->fwnode = no_free_ptr(fwnode);
> 
>  	ctx->typec.role_sw = usb_role_switch_get(dev);
>  	if (IS_ERR(ctx->typec.role_sw)) {
> --
> 2.45.1
>
Markus Elfring June 6, 2024, 11:01 a.m. UTC | #4 
> That makes it tricky to use no_free_ptr() so I wonder if this is
> a case where the old fashioned fix of adding all the relevant
> fwnode_handle_put() calls is the better option.  The __free()
> approach doesn't always fit.

Can involved development concerns be clarified better anyhow?

Regards,
Markus
diff mbox series

Patch

diff --git a/drivers/usb/typec/anx7411.c b/drivers/usb/typec/anx7411.c
index b12a07edc71b..9fb52f233a30 100644
--- a/drivers/usb/typec/anx7411.c
+++ b/drivers/usb/typec/anx7411.c
@@ -16,6 +16,7 @@ 
 #include <linux/of_graph.h>
 #include <linux/of_platform.h>
 #include <linux/pm_runtime.h>
+#include <linux/property.h>
 #include <linux/regulator/consumer.h>
 #include <linux/slab.h>
 #include <linux/types.h>
@@ -1142,11 +1143,11 @@  static int anx7411_typec_port_probe(struct anx7411_data *ctx,
 {
 	struct typec_capability *cap = &ctx->typec.caps;
 	struct typec_params *typecp = &ctx->typec;
-	struct fwnode_handle *fwnode;
 	const char *buf;
 	int ret, i;

-	fwnode = device_get_named_child_node(dev, "connector");
+	struct fwnode_handle *fwnode __free(fwnode_handle)
+				     = device_get_named_child_node(dev, "connector");
 	if (!fwnode)
 		return -EINVAL;

@@ -1237,7 +1238,7 @@  static int anx7411_typec_port_probe(struct anx7411_data *ctx,
 		typecp->caps_flags |= HAS_SINK_WATT;
 	}

-	cap->fwnode = fwnode;
+	cap->fwnode = no_free_ptr(fwnode);

 	ctx->typec.role_sw = usb_role_switch_get(dev);
 	if (IS_ERR(ctx->typec.role_sw)) {