[PATCH 1/3] crypto: nx: Hold the reference returned by of_find_compatible_node

[PATCH 1/3] crypto: nx: Hold the reference returned by of_find_compatible_node

[Liang He]

In nx842_pseries_init(), we should hold the reference returned by
of_find_compatible_node() and use it to call of_node_put to keep
refcount balance.

Signed-off-by: Liang He <windhl@126.com>
---
 drivers/crypto/nx/nx-common-pseries.c | 5 ++++-
 1 file changed, 4 insertions(+), 1 deletion(-)

diff --git a/drivers/crypto/nx/nx-common-pseries.c b/drivers/crypto/nx/nx-common-pseries.c
index 7584a34ba88c..3ea334b7f820 100644
--- a/drivers/crypto/nx/nx-common-pseries.c
+++ b/drivers/crypto/nx/nx-common-pseries.c
@@ -1208,10 +1208,13 @@ static struct vio_driver nx842_vio_driver = {
 static int __init nx842_pseries_init(void)
 {
 	struct nx842_devdata *new_devdata;
+	struct device_node *np;
 	int ret;
 
-	if (!of_find_compatible_node(NULL, NULL, "ibm,compression"))
+	np = of_find_compatible_node(NULL, NULL, "ibm,compression");
+	if (!np)
 		return -ENODEV;
+	of_node_put(np);
 
 	RCU_INIT_POINTER(devdata, NULL);
 	new_devdata = kzalloc(sizeof(*new_devdata), GFP_KERNEL);
-- 
2.25.1

[PATCH 2/3] arm: omap2: pm33xx-core: Fix missing of_node_put() in amx3_idle_init()

[Liang He]

To keep refcount balance, we need to call of_node_put() for the
reference 'state_node' returned by of_parse_phandle() in fail
path or when it is not used anymore.

Signed-off-by: Liang He <windhl@126.com>
---
 arch/arm/mach-omap2/pm33xx-core.c | 6 +++++-
 1 file changed, 5 insertions(+), 1 deletion(-)

diff --git a/arch/arm/mach-omap2/pm33xx-core.c b/arch/arm/mach-omap2/pm33xx-core.c
index bf0d25fd2cea..58b98a17746c 100644
--- a/arch/arm/mach-omap2/pm33xx-core.c
+++ b/arch/arm/mach-omap2/pm33xx-core.c
@@ -393,12 +393,15 @@ static int __init amx3_idle_init(struct device_node *cpu_node, int cpu)
 		if (!state_node)
 			break;
 
-		if (!of_device_is_available(state_node))
+		if (!of_device_is_available(state_node)) {
+			of_node_put(state_node);
 			continue;
+		}
 
 		if (i == CPUIDLE_STATE_MAX) {
 			pr_warn("%s: cpuidle states reached max possible\n",
 				__func__);
+			of_node_put(state_node);
 			break;
 		}
 
@@ -408,6 +411,7 @@ static int __init amx3_idle_init(struct device_node *cpu_node, int cpu)
 			states[state_count].wfi_flags |= WFI_FLAG_WAKE_M3 |
 							 WFI_FLAG_FLUSH_CACHE;
 
+		of_node_put(state_node);
 		state_count++;
 	}
 
-- 
2.25.1

[PATCH 3/3] arm: omap2: prm_common: Add the refcount for new reference creation

[Liang He]

In omap2_prm_base_init(), for_each_matching_node_and_match() can
automatically increase and decrease the refcounting. However, a
new reference is escaped out into 'data->np', so we need to use
of_node_put() for the old reference and use of_node_get() for the
new one.

Signed-off-by: Liang He <windhl@126.com>
---
 arch/arm/mach-omap2/prm_common.c | 3 ++-
 1 file changed, 2 insertions(+), 1 deletion(-)

diff --git a/arch/arm/mach-omap2/prm_common.c b/arch/arm/mach-omap2/prm_common.c
index fb2d48cfe756..bed151e3f165 100644
--- a/arch/arm/mach-omap2/prm_common.c
+++ b/arch/arm/mach-omap2/prm_common.c
@@ -764,7 +764,8 @@ int __init omap2_prm_base_init(void)
 			prm_base.pa = res.start + data->offset;
 		}
 
-		data->np = np;
+		of_node_put(data->np);
+		data->np = of_node_get(np);
 
 		if (data->init)
 			data->init(data);
-- 
2.25.1

Re: [PATCH 3/3] arm: omap2: prm_common: Add the refcount for new reference creation

[Tony Lindgren]

* Liang He <windhl@126.com> [220701 14:22]:
> In omap2_prm_base_init(), for_each_matching_node_and_match() can
> automatically increase and decrease the refcounting. However, a
> new reference is escaped out into 'data->np', so we need to use
> of_node_put() for the old reference and use of_node_get() for the
> new one.
> 
> Signed-off-by: Liang He <windhl@126.com>
> ---
>  arch/arm/mach-omap2/prm_common.c | 3 ++-
>  1 file changed, 2 insertions(+), 1 deletion(-)
> 
> diff --git a/arch/arm/mach-omap2/prm_common.c b/arch/arm/mach-omap2/prm_common.c
> index fb2d48cfe756..bed151e3f165 100644
> --- a/arch/arm/mach-omap2/prm_common.c
> +++ b/arch/arm/mach-omap2/prm_common.c
> @@ -764,7 +764,8 @@ int __init omap2_prm_base_init(void)
>  			prm_base.pa = res.start + data->offset;
>  		}
>  
> -		data->np = np;
> +		of_node_put(data->np);
> +		data->np = of_node_get(np);
>  
>  		if (data->init)
>  			data->init(data);

Hmm so I don't get how this improves things? Don't we enable with
the same refcounting?

Regards,

Tony

Re:Re: [PATCH 3/3] arm: omap2: prm_common: Add the refcount for new reference creation

[Liang He]

At 2022-07-04 15:47:01, "Tony Lindgren" <tony@atomide.com> wrote:
>* Liang He <windhl@126.com> [220701 14:22]:
>> In omap2_prm_base_init(), for_each_matching_node_and_match() can
>> automatically increase and decrease the refcounting. However, a
>> new reference is escaped out into 'data->np', so we need to use
>> of_node_put() for the old reference and use of_node_get() for the
>> new one.
>> 
>> Signed-off-by: Liang He <windhl@126.com>
>> ---
>>  arch/arm/mach-omap2/prm_common.c | 3 ++-
>>  1 file changed, 2 insertions(+), 1 deletion(-)
>> 
>> diff --git a/arch/arm/mach-omap2/prm_common.c b/arch/arm/mach-omap2/prm_common.c
>> index fb2d48cfe756..bed151e3f165 100644
>> --- a/arch/arm/mach-omap2/prm_common.c
>> +++ b/arch/arm/mach-omap2/prm_common.c
>> @@ -764,7 +764,8 @@ int __init omap2_prm_base_init(void)
>>  			prm_base.pa = res.start + data->offset;
>>  		}
>>  
>> -		data->np = np;
>> +		of_node_put(data->np);
>> +		data->np = of_node_get(np);
>>  
>>  		if (data->init)
>>  			data->init(data);
>
>Hmm so I don't get how this improves things? Don't we enable with
>the same refcounting?
>
>Regards,
>
>Tony

Hi, Tony.

Thanks very much for reviewing this patch.

I try to explain as following:

First, there is an old reference (pointer)  pointing to device_node A and stored in data->np.
Then, in the for_each_xxx iteration, 'np' is a new reference (pointer) pointing to device_node B.

After 'data->np = np' is done, the old reference is destroyed and its refcount should be decreased.
Similarly, a new reference to devce_node B is created in 'data->np', so its refcount should be increased.

Sorry if my understanding is wrong or if I do not catch your point.

Thans again,

Liang