The optional _PXM method evaluates to an integer that identifies the
proximity domain of a device object. On ACPI based kernel boot, the
field numa_node in 'struct device' is always set to -1 irrespective
of _PXM value that is specified in the ACPI device object. But in
case of device-tree based kernel boot the numa_node field is populated
and reflects to a DT property that is specified in DTS according to
the below document.
https://www.kernel.org/doc/Documentation/devicetree/bindings/numa.txt
http://lxr.free-electrons.com/source/drivers/of/device.c#L54
Without this patch dev_to_node() always returns a value of -1 for all
platform devices. This patch implements support for ACPI _PXM method
and updates the platform device numa node id using acpi_get_node(),
which provides the PXM to NUMA mapping information. The individual
platform device drivers should be able to use the NUMA aware memory
allocation functions kmalloc_node() and alloc_pages_node() to improve
the performance.
Signed-off-by: Shanker Donthineni <shankerd@xxxxxxxxxxxxxx>
---
Changes since v1: Edited commit text.
Not all the platform devices are attached to NUMA node 0 on Qualcomm
Datacenter Technologies server chips. Platform device drivers needs
to be aware of NUMA information to improve performance.
drivers/acpi/acpi_platform.c | 5 ++++-
1 file changed, 4 insertions(+), 1 deletion(-)
diff --git a/drivers/acpi/acpi_platform.c b/drivers/acpi/acpi_platform.c
index b4c1a6a..83d953e 100644
--- a/drivers/acpi/acpi_platform.c
+++ b/drivers/acpi/acpi_platform.c
@@ -119,11 +119,14 @@ struct platform_device *acpi_create_platform_device(struct acpi_device *adev,
if (IS_ERR(pdev))
dev_err(&adev->dev, "platform device creation failed: %ld\n",
PTR_ERR(pdev));
- else
+ else {
+ set_dev_node(&pdev->dev, acpi_get_node(adev->handle));
dev_dbg(&adev->dev, "created platform device %s\n",
dev_name(&pdev->dev));
+ }
kfree(resources);
+
return pdev;
}
EXPORT_SYMBOL_GPL(acpi_create_platform_device);