[PATCH V2] pmdomain: tegra: Add support for multi-socket platforms

[PATCH V2] pmdomain: tegra: Add support for multi-socket platforms

[Jon Hunter]

On multi-socket platforms each socket has its own BPMP that is
registered with the kernel. For such platforms prefix the NUMA ID for
each socket to the BPMP powergate name to ensure there is a unique name
for each power-domain. Note that we only add the NUMA ID for powergates
that return a valid name because an invalid name indicates that the
powergate ID is not supported.

Signed-off-by: Jon Hunter <jonathanh@nvidia.com>
---
Changes since V1:
- Use NUMA ID as a prefix and not a suffix for the powergate name.

 drivers/pmdomain/tegra/powergate-bpmp.c | 5 +++++
 1 file changed, 5 insertions(+)

diff --git a/drivers/pmdomain/tegra/powergate-bpmp.c b/drivers/pmdomain/tegra/powergate-bpmp.c
index 8cde4f384846..5e869b0eb96e 100644
--- a/drivers/pmdomain/tegra/powergate-bpmp.c
+++ b/drivers/pmdomain/tegra/powergate-bpmp.c
@@ -137,6 +137,11 @@ static char *tegra_bpmp_powergate_get_name(struct tegra_bpmp *bpmp,
 	if (err < 0 || msg.rx.ret < 0)
 		return NULL;
 
+	if (response.get_name.name[0] != '\0' &&
+	    dev_to_node(bpmp->dev) != NUMA_NO_NODE)
+		return kasprintf(GFP_KERNEL, "%d-%s", dev_to_node(bpmp->dev),
+				 response.get_name.name);
+
 	return kstrdup(response.get_name.name, GFP_KERNEL);
 }
 
-- 
2.43.0

Re: [PATCH V2] pmdomain: tegra: Add support for multi-socket platforms

[Thierry Reding]

[-- Attachment #1: Type: text/plain, Size: 1951 bytes --]

On Tue, Jun 02, 2026 at 09:39:59AM +0100, Jon Hunter wrote:
> On multi-socket platforms each socket has its own BPMP that is
> registered with the kernel. For such platforms prefix the NUMA ID for
> each socket to the BPMP powergate name to ensure there is a unique name
> for each power-domain. Note that we only add the NUMA ID for powergates
> that return a valid name because an invalid name indicates that the
> powergate ID is not supported.
> 
> Signed-off-by: Jon Hunter <jonathanh@nvidia.com>
> ---
> Changes since V1:
> - Use NUMA ID as a prefix and not a suffix for the powergate name.
> 
>  drivers/pmdomain/tegra/powergate-bpmp.c | 5 +++++
>  1 file changed, 5 insertions(+)
> 
> diff --git a/drivers/pmdomain/tegra/powergate-bpmp.c b/drivers/pmdomain/tegra/powergate-bpmp.c
> index 8cde4f384846..5e869b0eb96e 100644
> --- a/drivers/pmdomain/tegra/powergate-bpmp.c
> +++ b/drivers/pmdomain/tegra/powergate-bpmp.c
> @@ -137,6 +137,11 @@ static char *tegra_bpmp_powergate_get_name(struct tegra_bpmp *bpmp,
>  	if (err < 0 || msg.rx.ret < 0)
>  		return NULL;
>  
> +	if (response.get_name.name[0] != '\0' &&
> +	    dev_to_node(bpmp->dev) != NUMA_NO_NODE)
> +		return kasprintf(GFP_KERNEL, "%d-%s", dev_to_node(bpmp->dev),
> +				 response.get_name.name);
> +
>  	return kstrdup(response.get_name.name, GFP_KERNEL);

Why are we not checking response.get_name.name for the empty string in
the single socket case? Or the other way around: why is this important
for multi-socket if it isn't for single socket?

Looking at the caller of this function, it will check for this case
itself, so I'm thinking we should rework that as part of this patch and
simply return NULL early for both cases when name[0] == '\0'. That way
the caller is simplified and we get a bit more consistency in this
function. It is going to make for a slightly bigger diff, but I think
overall it'd be an improvement.

Thierry

[-- Attachment #2: signature.asc --]
[-- Type: application/pgp-signature, Size: 833 bytes --]

Re: [PATCH V2] pmdomain: tegra: Add support for multi-socket platforms

[Jon Hunter]

On 02/06/2026 12:59, Thierry Reding wrote:
> On Tue, Jun 02, 2026 at 09:39:59AM +0100, Jon Hunter wrote:
>> On multi-socket platforms each socket has its own BPMP that is
>> registered with the kernel. For such platforms prefix the NUMA ID for
>> each socket to the BPMP powergate name to ensure there is a unique name
>> for each power-domain. Note that we only add the NUMA ID for powergates
>> that return a valid name because an invalid name indicates that the
>> powergate ID is not supported.
>>
>> Signed-off-by: Jon Hunter <jonathanh@nvidia.com>
>> ---
>> Changes since V1:
>> - Use NUMA ID as a prefix and not a suffix for the powergate name.
>>
>>   drivers/pmdomain/tegra/powergate-bpmp.c | 5 +++++
>>   1 file changed, 5 insertions(+)
>>
>> diff --git a/drivers/pmdomain/tegra/powergate-bpmp.c b/drivers/pmdomain/tegra/powergate-bpmp.c
>> index 8cde4f384846..5e869b0eb96e 100644
>> --- a/drivers/pmdomain/tegra/powergate-bpmp.c
>> +++ b/drivers/pmdomain/tegra/powergate-bpmp.c
>> @@ -137,6 +137,11 @@ static char *tegra_bpmp_powergate_get_name(struct tegra_bpmp *bpmp,
>>   	if (err < 0 || msg.rx.ret < 0)
>>   		return NULL;
>>   
>> +	if (response.get_name.name[0] != '\0' &&
>> +	    dev_to_node(bpmp->dev) != NUMA_NO_NODE)
>> +		return kasprintf(GFP_KERNEL, "%d-%s", dev_to_node(bpmp->dev),
>> +				 response.get_name.name);
>> +
>>   	return kstrdup(response.get_name.name, GFP_KERNEL);
> 
> Why are we not checking response.get_name.name for the empty string in
> the single socket case? Or the other way around: why is this important
> for multi-socket if it isn't for single socket?

Simply because we don't want to return a valid string such as '0-' or 
'1-' for the case where we received a NULL terminated string. The caller 
would interpret the name '0-' being valid when it is not.

> Looking at the caller of this function, it will check for this case
> itself, so I'm thinking we should rework that as part of this patch and
> simply return NULL early for both cases when name[0] == '\0'. That way
> the caller is simplified and we get a bit more consistency in this
> function. It is going to make for a slightly bigger diff, but I think
> overall it'd be an improvement.

Yes I agree that would be better and then just check for '\0' in one 
place. OK I will send an update.

Jon

-- 
nvpublic