Re: [PATCH] drivers: use __free attribute instead of of_node_put()

Re: [PATCH] drivers: use __free attribute instead of of_node_put()

[Shresth Prasad]

> Please fix the subject line to be "backlight: <driver>: ...". I came
> very close to deleting this patch without reading it ;-) .

Really sorry about that, I'll fix it.

> Do we need to get dev->of_node at all? The device, which we are
> borrowing, already owns a reference to the node so I don't see
> any point in this function taking an extra one.
>
> So why not simply make this:
>
>     struct device_node *np = dev->of_node;

Looking at it again, I'm not sure why the call to of_node_put is there in
the first place. I think removing it would be fine.

I'll fix both of these issues and send a patch v2.

Regards,
Shresth

Re: [PATCH] drivers: use __free attribute instead of of_node_put()

[Sudeep Holla]

On Fri, Apr 19, 2024 at 11:16:37PM +0530, Shresth Prasad wrote:
> > Please fix the subject line to be "backlight: <driver>: ...". I came
> > very close to deleting this patch without reading it ;-) .
> 
> Really sorry about that, I'll fix it.
> 
> > Do we need to get dev->of_node at all? The device, which we are
> > borrowing, already owns a reference to the node so I don't see
> > any point in this function taking an extra one.
> >
> > So why not simply make this:
> >
> >     struct device_node *np = dev->of_node;
> 
> Looking at it again, I'm not sure why the call to of_node_put is there in
> the first place. I think removing it would be fine.
> 
> I'll fix both of these issues and send a patch v2.

I assume you are using lore "Reply using the --to, --cc, and..." option
or something similar.

You seem to have mixed up 2 different message ID. I was not able to make
any sense of this email.

You have wrongly used [1] but you really want [2]. I think both have
very similar $subject and hence the confusion. Another reason
why getting subject right is very important on the mailing list.

--
Regards,
Sudeep

[1] 20240419131956.665769-1-vincenzo.mezzela@gmail.com
(https://lore.kernel.org/all/20240419131956.665769-1-vincenzo.mezzela@gmail.com/)
[2] 20240419111613.GA12884@aspen.lan
https://lore.kernel.org/all/20240419111613.GA12884@aspen.lan/

Re: [PATCH] drivers: use __free attribute instead of of_node_put()

[Shresth Prasad]

19 Apr 2024 11:56:47 pm Sudeep Holla <sudeep.holla@arm.com>:

> On Fri, Apr 19, 2024 at 11:16:37PM +0530, Shresth Prasad wrote:
>>> Please fix the subject line to be "backlight: <driver>: ...". I came
>>> very close to deleting this patch without reading it ;-) .
>> 
>> Really sorry about that, I'll fix it.
>> 
>>> Do we need to get dev->of_node at all? The device, which we are
>>> borrowing, already owns a reference to the node so I don't see
>>> any point in this function taking an extra one.
>>> 
>>> So why not simply make this:
>>> 
>>>     struct device_node *np = dev->of_node;
>> 
>> Looking at it again, I'm not sure why the call to of_node_put is there in
>> the first place. I think removing it would be fine.
>> 
>> I'll fix both of these issues and send a patch v2.
> 
> I assume you are using lore "Reply using the --to, --cc, and..." option
> or something similar.
> 
> You seem to have mixed up 2 different message ID. I was not able to make
> any sense of this email.
> 
> You have wrongly used [1] but you really want [2]. I think both have
> very similar $subject and hence the confusion. Another reason
> why getting subject right is very important on the mailing list.
> 
> -- 
> Regards,
> Sudeep
> 
> [1] 20240419131956.665769-1-vincenzo.mezzela@gmail.com
> (https://lore.kernel.org/all/20240419131956.665769-1-vincenzo.mezzela@gmail.com/)
> [2] 20240419111613.GA12884@aspen.lan
> https://lore.kernel.org/all/20240419111613.GA12884@aspen.lan/
I'm very new to using mailing lists, sorry for any confusion. I was indeed using the reply option on lore. I'll be more careful next.

Is there anyway to remove the incorrect reply from this thread?

Regards,
Shresth Prasad

[PATCH] drivers: use __free attribute instead of of_node_put()

[Vincenzo Mezzela]

Introduce the __free attribute for scope-based resource management.
Resources allocated with __free are automatically released at the end of
the scope. This enhancement aims to mitigate memory management issues
associated with forgetting to release resources by utilizing __free
instead of of_node_put().

The declaration of the device_node used within the do-while loops is
moved directly within the loop so that the resource is automatically
freed at the end of each iteration.

Suggested-by: Julia Lawall <julia.lawall@inria.fr>
Signed-off-by: Vincenzo Mezzela <vincenzo.mezzela@gmail.com>
---
 drivers/base/arch_topology.c | 41 ++++++++++++++----------------------
 1 file changed, 16 insertions(+), 25 deletions(-)

diff --git a/drivers/base/arch_topology.c b/drivers/base/arch_topology.c
index 024b78a0cfc1..58eeb8183747 100644
--- a/drivers/base/arch_topology.c
+++ b/drivers/base/arch_topology.c
@@ -513,10 +513,10 @@ core_initcall(free_raw_capacity);
  */
 static int __init get_cpu_for_node(struct device_node *node)
 {
-	struct device_node *cpu_node;
 	int cpu;
 
-	cpu_node = of_parse_phandle(node, "cpu", 0);
+	struct device_node *cpu_node __free(device_node) =
+		of_parse_phandle(node, "cpu", 0);
 	if (!cpu_node)
 		return -1;
 
@@ -527,7 +527,6 @@ static int __init get_cpu_for_node(struct device_node *node)
 		pr_info("CPU node for %pOF exist but the possible cpu range is :%*pbl\n",
 			cpu_node, cpumask_pr_args(cpu_possible_mask));
 
-	of_node_put(cpu_node);
 	return cpu;
 }
 
@@ -538,11 +537,11 @@ static int __init parse_core(struct device_node *core, int package_id,
 	bool leaf = true;
 	int i = 0;
 	int cpu;
-	struct device_node *t;
 
 	do {
 		snprintf(name, sizeof(name), "thread%d", i);
-		t = of_get_child_by_name(core, name);
+		struct device_node *t __free(device_node) =
+			of_get_child_by_name(core, name);
 		if (t) {
 			leaf = false;
 			cpu = get_cpu_for_node(t);
@@ -553,10 +552,8 @@ static int __init parse_core(struct device_node *core, int package_id,
 				cpu_topology[cpu].thread_id = i;
 			} else if (cpu != -ENODEV) {
 				pr_err("%pOF: Can't get CPU for thread\n", t);
-				of_node_put(t);
 				return -EINVAL;
 			}
-			of_node_put(t);
 		}
 		i++;
 	} while (t);
@@ -586,7 +583,6 @@ static int __init parse_cluster(struct device_node *cluster, int package_id,
 	char name[20];
 	bool leaf = true;
 	bool has_cores = false;
-	struct device_node *c;
 	int core_id = 0;
 	int i, ret;
 
@@ -598,13 +594,13 @@ static int __init parse_cluster(struct device_node *cluster, int package_id,
 	i = 0;
 	do {
 		snprintf(name, sizeof(name), "cluster%d", i);
-		c = of_get_child_by_name(cluster, name);
+		struct device_node *c __free(device_node) =
+			of_get_child_by_name(cluster, name);
 		if (c) {
 			leaf = false;
 			ret = parse_cluster(c, package_id, i, depth + 1);
 			if (depth > 0)
 				pr_warn("Topology for clusters of clusters not yet supported\n");
-			of_node_put(c);
 			if (ret != 0)
 				return ret;
 		}
@@ -615,14 +611,14 @@ static int __init parse_cluster(struct device_node *cluster, int package_id,
 	i = 0;
 	do {
 		snprintf(name, sizeof(name), "core%d", i);
-		c = of_get_child_by_name(cluster, name);
+		struct device_node *c __free(device_node) =
+			of_get_child_by_name(cluster, name);
 		if (c) {
 			has_cores = true;
 
 			if (depth == 0) {
 				pr_err("%pOF: cpu-map children should be clusters\n",
 				       c);
-				of_node_put(c);
 				return -EINVAL;
 			}
 
@@ -635,7 +631,6 @@ static int __init parse_cluster(struct device_node *cluster, int package_id,
 				ret = -EINVAL;
 			}
 
-			of_node_put(c);
 			if (ret != 0)
 				return ret;
 		}
@@ -651,17 +646,16 @@ static int __init parse_cluster(struct device_node *cluster, int package_id,
 static int __init parse_socket(struct device_node *socket)
 {
 	char name[20];
-	struct device_node *c;
 	bool has_socket = false;
 	int package_id = 0, ret;
 
 	do {
 		snprintf(name, sizeof(name), "socket%d", package_id);
-		c = of_get_child_by_name(socket, name);
+		struct device_node *c __free(device_node) =
+			of_get_child_by_name(socket, name);
 		if (c) {
 			has_socket = true;
 			ret = parse_cluster(c, package_id, -1, 0);
-			of_node_put(c);
 			if (ret != 0)
 				return ret;
 		}
@@ -676,11 +670,11 @@ static int __init parse_socket(struct device_node *socket)
 
 static int __init parse_dt_topology(void)
 {
-	struct device_node *cn, *map;
 	int ret = 0;
 	int cpu;
 
-	cn = of_find_node_by_path("/cpus");
+	struct device_node *cn __free(device_node) =
+		of_find_node_by_path("/cpus");
 	if (!cn) {
 		pr_err("No CPU information found in DT\n");
 		return 0;
@@ -690,13 +684,14 @@ static int __init parse_dt_topology(void)
 	 * When topology is provided cpu-map is essentially a root
 	 * cluster with restricted subnodes.
 	 */
-	map = of_get_child_by_name(cn, "cpu-map");
+	struct device_node *map __free(devide_node) =
+		of_get_child_by_name(cn, "cpu-map");
 	if (!map)
-		goto out;
+		return ret;
 
 	ret = parse_socket(map);
 	if (ret != 0)
-		goto out_map;
+		return ret;
 
 	topology_normalize_cpu_scale();
 
@@ -710,10 +705,6 @@ static int __init parse_dt_topology(void)
 			break;
 		}
 
-out_map:
-	of_node_put(map);
-out:
-	of_node_put(cn);
 	return ret;
 }
 #endif
-- 
2.34.1

Re: [PATCH] drivers: use __free attribute instead of of_node_put()

[Sudeep Holla]

On Fri, Apr 19, 2024 at 03:19:56PM +0200, Vincenzo Mezzela wrote:
> Introduce the __free attribute for scope-based resource management.
> Resources allocated with __free are automatically released at the end of
> the scope. This enhancement aims to mitigate memory management issues
> associated with forgetting to release resources by utilizing __free
> instead of of_node_put().
>
> The declaration of the device_node used within the do-while loops is
> moved directly within the loop so that the resource is automatically
> freed at the end of each iteration.
>
> Suggested-by: Julia Lawall <julia.lawall@inria.fr>
> Signed-off-by: Vincenzo Mezzela <vincenzo.mezzela@gmail.com>
> ---
>  drivers/base/arch_topology.c | 41 ++++++++++++++----------------------
>  1 file changed, 16 insertions(+), 25 deletions(-)
>
> diff --git a/drivers/base/arch_topology.c b/drivers/base/arch_topology.c
> index 024b78a0cfc1..58eeb8183747 100644
> --- a/drivers/base/arch_topology.c
> +++ b/drivers/base/arch_topology.c
> @@ -513,10 +513,10 @@ core_initcall(free_raw_capacity);
>   */
>  static int __init get_cpu_for_node(struct device_node *node)
>  {
> -	struct device_node *cpu_node;
>  	int cpu;
>
> -	cpu_node = of_parse_phandle(node, "cpu", 0);
> +	struct device_node *cpu_node __free(device_node) =

Missing include <linux/cleanup.h> for this ?

> +		of_parse_phandle(node, "cpu", 0);
>  	if (!cpu_node)
>  		return -1;
>
> @@ -527,7 +527,6 @@ static int __init get_cpu_for_node(struct device_node *node)
>  		pr_info("CPU node for %pOF exist but the possible cpu range is :%*pbl\n",
>  			cpu_node, cpumask_pr_args(cpu_possible_mask));
>
> -	of_node_put(cpu_node);
>  	return cpu;
>  }
>
> @@ -538,11 +537,11 @@ static int __init parse_core(struct device_node *core, int package_id,
>  	bool leaf = true;
>  	int i = 0;
>  	int cpu;
> -	struct device_node *t;
>
>  	do {
>  		snprintf(name, sizeof(name), "thread%d", i);
> -		t = of_get_child_by_name(core, name);
> +		struct device_node *t __free(device_node) =
> +			of_get_child_by_name(core, name);
>  		if (t) {
>  			leaf = false;
>  			cpu = get_cpu_for_node(t);
> @@ -553,10 +552,8 @@ static int __init parse_core(struct device_node *core, int package_id,
>  				cpu_topology[cpu].thread_id = i;
>  			} else if (cpu != -ENODEV) {
>  				pr_err("%pOF: Can't get CPU for thread\n", t);
> -				of_node_put(t);
>  				return -EINVAL;
>  			}
> -			of_node_put(t);

OK you moved 't' inside the loop and this must be taken care, but...

>  		}
>  		i++;
>  	} while (t);

....now, will it even compile if 't' is not in scope ? I think you might get
compilation here. If not, I still don't understand what is the value of
't' being checked there.

> @@ -586,7 +583,6 @@ static int __init parse_cluster(struct device_node *cluster, int package_id,
>  	char name[20];
>  	bool leaf = true;
>  	bool has_cores = false;
> -	struct device_node *c;
>  	int core_id = 0;
>  	int i, ret;
>
> @@ -598,13 +594,13 @@ static int __init parse_cluster(struct device_node *cluster, int package_id,
>  	i = 0;
>  	do {
>  		snprintf(name, sizeof(name), "cluster%d", i);
> -		c = of_get_child_by_name(cluster, name);
> +		struct device_node *c __free(device_node) =
> +			of_get_child_by_name(cluster, name);
>  		if (c) {
>  			leaf = false;
>  			ret = parse_cluster(c, package_id, i, depth + 1);
>  			if (depth > 0)
>  				pr_warn("Topology for clusters of clusters not yet supported\n");
> -			of_node_put(c);
>  			if (ret != 0)
>  				return ret;
>  		}
> @@ -615,14 +611,14 @@ static int __init parse_cluster(struct device_node *cluster, int package_id,
>  	i = 0;
>  	do {
>  		snprintf(name, sizeof(name), "core%d", i);
> -		c = of_get_child_by_name(cluster, name);
> +		struct device_node *c __free(device_node) =
> +			of_get_child_by_name(cluster, name);
>  		if (c) {
>  			has_cores = true;
>
>  			if (depth == 0) {
>  				pr_err("%pOF: cpu-map children should be clusters\n",
>  				       c);
> -				of_node_put(c);
>  				return -EINVAL;
>  			}
>
> @@ -635,7 +631,6 @@ static int __init parse_cluster(struct device_node *cluster, int package_id,
>  				ret = -EINVAL;
>  			}
>
> -			of_node_put(c);
>  			if (ret != 0)
>  				return ret;
>  		}
> @@ -651,17 +646,16 @@ static int __init parse_cluster(struct device_node *cluster, int package_id,
>  static int __init parse_socket(struct device_node *socket)
>  {
>  	char name[20];
> -	struct device_node *c;
>  	bool has_socket = false;
>  	int package_id = 0, ret;
>
>  	do {
>  		snprintf(name, sizeof(name), "socket%d", package_id);
> -		c = of_get_child_by_name(socket, name);
> +		struct device_node *c __free(device_node) =
> +			of_get_child_by_name(socket, name);
>  		if (c) {
>  			has_socket = true;
>  			ret = parse_cluster(c, package_id, -1, 0);
> -			of_node_put(c);
>  			if (ret != 0)
>  				return ret;
>  		}

Same thing applies to these while(c) loop. I don't understand how this
could work even if it is compiling fine which I doubt.

> @@ -676,11 +670,11 @@ static int __init parse_socket(struct device_node *socket)
>
>  static int __init parse_dt_topology(void)
>  {
> -	struct device_node *cn, *map;
>  	int ret = 0;
>  	int cpu;
>
> -	cn = of_find_node_by_path("/cpus");
> +	struct device_node *cn __free(device_node) =
> +		of_find_node_by_path("/cpus");
>  	if (!cn) {
>  		pr_err("No CPU information found in DT\n");
>  		return 0;
> @@ -690,13 +684,14 @@ static int __init parse_dt_topology(void)
>  	 * When topology is provided cpu-map is essentially a root
>  	 * cluster with restricted subnodes.
>  	 */
> -	map = of_get_child_by_name(cn, "cpu-map");
> +	struct device_node *map __free(devide_node) =

If not above ones, this must fail to compile. Perhaps s/devide_node/device_node/ ?
I now doubt if this patch is compile tested ?

--
Regards,
Sudeep

Re: [PATCH] drivers: use __free attribute instead of of_node_put()

[Vincenzo Mezzela]

On 19/04/24 16:01, Sudeep Holla wrote:
> On Fri, Apr 19, 2024 at 03:19:56PM +0200, Vincenzo Mezzela wrote:
>> Introduce the __free attribute for scope-based resource management.
>> Resources allocated with __free are automatically released at the end of
>> the scope. This enhancement aims to mitigate memory management issues
>> associated with forgetting to release resources by utilizing __free
>> instead of of_node_put().
>>
>> The declaration of the device_node used within the do-while loops is
>> moved directly within the loop so that the resource is automatically
>> freed at the end of each iteration.
>>
>> Suggested-by: Julia Lawall <julia.lawall@inria.fr>
>> Signed-off-by: Vincenzo Mezzela <vincenzo.mezzela@gmail.com>
>> ---
>>   drivers/base/arch_topology.c | 41 ++++++++++++++----------------------
>>   1 file changed, 16 insertions(+), 25 deletions(-)
>>
>> diff --git a/drivers/base/arch_topology.c b/drivers/base/arch_topology.c
>> index 024b78a0cfc1..58eeb8183747 100644
>> --- a/drivers/base/arch_topology.c
>> +++ b/drivers/base/arch_topology.c
>> @@ -513,10 +513,10 @@ core_initcall(free_raw_capacity);
>>    */
>>   static int __init get_cpu_for_node(struct device_node *node)
>>   {
>> -	struct device_node *cpu_node;
>>   	int cpu;
>>
>> -	cpu_node = of_parse_phandle(node, "cpu", 0);
>> +	struct device_node *cpu_node __free(device_node) =
> Missing include <linux/cleanup.h> for this ?
>
>> +		of_parse_phandle(node, "cpu", 0);
>>   	if (!cpu_node)
>>   		return -1;
>>
>> @@ -527,7 +527,6 @@ static int __init get_cpu_for_node(struct device_node *node)
>>   		pr_info("CPU node for %pOF exist but the possible cpu range is :%*pbl\n",
>>   			cpu_node, cpumask_pr_args(cpu_possible_mask));
>>
>> -	of_node_put(cpu_node);
>>   	return cpu;
>>   }
>>
>> @@ -538,11 +537,11 @@ static int __init parse_core(struct device_node *core, int package_id,
>>   	bool leaf = true;
>>   	int i = 0;
>>   	int cpu;
>> -	struct device_node *t;
>>
>>   	do {
>>   		snprintf(name, sizeof(name), "thread%d", i);
>> -		t = of_get_child_by_name(core, name);
>> +		struct device_node *t __free(device_node) =
>> +			of_get_child_by_name(core, name);
>>   		if (t) {
>>   			leaf = false;
>>   			cpu = get_cpu_for_node(t);
>> @@ -553,10 +552,8 @@ static int __init parse_core(struct device_node *core, int package_id,
>>   				cpu_topology[cpu].thread_id = i;
>>   			} else if (cpu != -ENODEV) {
>>   				pr_err("%pOF: Can't get CPU for thread\n", t);
>> -				of_node_put(t);
>>   				return -EINVAL;
>>   			}
>> -			of_node_put(t);
> OK you moved 't' inside the loop and this must be taken care, but...
>
>>   		}
>>   		i++;
>>   	} while (t);
> ....now, will it even compile if 't' is not in scope ? I think you might get
> compilation here. If not, I still don't understand what is the value of
> 't' being checked there.
>
>> @@ -586,7 +583,6 @@ static int __init parse_cluster(struct device_node *cluster, int package_id,
>>   	char name[20];
>>   	bool leaf = true;
>>   	bool has_cores = false;
>> -	struct device_node *c;
>>   	int core_id = 0;
>>   	int i, ret;
>>
>> @@ -598,13 +594,13 @@ static int __init parse_cluster(struct device_node *cluster, int package_id,
>>   	i = 0;
>>   	do {
>>   		snprintf(name, sizeof(name), "cluster%d", i);
>> -		c = of_get_child_by_name(cluster, name);
>> +		struct device_node *c __free(device_node) =
>> +			of_get_child_by_name(cluster, name);
>>   		if (c) {
>>   			leaf = false;
>>   			ret = parse_cluster(c, package_id, i, depth + 1);
>>   			if (depth > 0)
>>   				pr_warn("Topology for clusters of clusters not yet supported\n");
>> -			of_node_put(c);
>>   			if (ret != 0)
>>   				return ret;
>>   		}
>> @@ -615,14 +611,14 @@ static int __init parse_cluster(struct device_node *cluster, int package_id,
>>   	i = 0;
>>   	do {
>>   		snprintf(name, sizeof(name), "core%d", i);
>> -		c = of_get_child_by_name(cluster, name);
>> +		struct device_node *c __free(device_node) =
>> +			of_get_child_by_name(cluster, name);
>>   		if (c) {
>>   			has_cores = true;
>>
>>   			if (depth == 0) {
>>   				pr_err("%pOF: cpu-map children should be clusters\n",
>>   				       c);
>> -				of_node_put(c);
>>   				return -EINVAL;
>>   			}
>>
>> @@ -635,7 +631,6 @@ static int __init parse_cluster(struct device_node *cluster, int package_id,
>>   				ret = -EINVAL;
>>   			}
>>
>> -			of_node_put(c);
>>   			if (ret != 0)
>>   				return ret;
>>   		}
>> @@ -651,17 +646,16 @@ static int __init parse_cluster(struct device_node *cluster, int package_id,
>>   static int __init parse_socket(struct device_node *socket)
>>   {
>>   	char name[20];
>> -	struct device_node *c;
>>   	bool has_socket = false;
>>   	int package_id = 0, ret;
>>
>>   	do {
>>   		snprintf(name, sizeof(name), "socket%d", package_id);
>> -		c = of_get_child_by_name(socket, name);
>> +		struct device_node *c __free(device_node) =
>> +			of_get_child_by_name(socket, name);
>>   		if (c) {
>>   			has_socket = true;
>>   			ret = parse_cluster(c, package_id, -1, 0);
>> -			of_node_put(c);
>>   			if (ret != 0)
>>   				return ret;
>>   		}
> Same thing applies to these while(c) loop. I don't understand how this
> could work even if it is compiling fine which I doubt.
>
>> @@ -676,11 +670,11 @@ static int __init parse_socket(struct device_node *socket)
>>
>>   static int __init parse_dt_topology(void)
>>   {
>> -	struct device_node *cn, *map;
>>   	int ret = 0;
>>   	int cpu;
>>
>> -	cn = of_find_node_by_path("/cpus");
>> +	struct device_node *cn __free(device_node) =
>> +		of_find_node_by_path("/cpus");
>>   	if (!cn) {
>>   		pr_err("No CPU information found in DT\n");
>>   		return 0;
>> @@ -690,13 +684,14 @@ static int __init parse_dt_topology(void)
>>   	 * When topology is provided cpu-map is essentially a root
>>   	 * cluster with restricted subnodes.
>>   	 */
>> -	map = of_get_child_by_name(cn, "cpu-map");
>> +	struct device_node *map __free(devide_node) =
> If not above ones, this must fail to compile. Perhaps s/devide_node/device_node/ ?
> I now doubt if this patch is compile tested ?
>
> --
> Regards,
> Sudeep

Hi,

As you rightly pointed out, I inadvertently omitted to compile this file 
during the kernel build process. Consequently, certain errors remained 
undetected. I apologize for the oversight.

I'll send an updated version of this patch soon.

Regards,

Vincenzo