Re: [RFC PATCH] ACPI / scan: Introduce _CCA parsing logic and setting up device coherency

[Tom Lendacky <thomas.lendacky@amd.com>]

On 04/01/2015 10:20 AM, Suravee Suthikulpanit wrote:
> ACPI v5.1 introduced _CCA object for specifying cache coherency attribute
> for devices. This patch implements a logic, which traverses device namespace
> to parse the coherency information, and calling the corresponded
> arch_setup_dma_ops().
>
> It checks the _CCA during ACPI scan, and setup coherency during acpi_device creation.
> Then, this is propagted to the platform_device when it is created at later time.
>
> CC: Mark Salter <msalter@redhat.com>
> CC: Hanjun Guo <hanjun.guo@linaro.org>
> CC: Al Stone <al.stone@linaro.org>
> Signed-off-by: Suravee Suthikulpanit <Suravee.Suthikulpanit@amd.com>
> ---
>
> NOTE:
>
> The original patch has been submitted by Mark Salter, and discussed here:
>
> * http://comments.gmane.org/gmane.linux.acpi.devel/70424
>
> One of the reveiw comment was dissusing whether we need the do/while loop for
> travesing parents of the device.  AFAIK, there are no APIs that would do upsearch
> from a particular node.  The acpi_get_handle() is the closest one, but it is using
> ACPI_NS_NO_UPSEARCH.
>
> This has been rebased and tested with:
> 	* http://git.linaro.org/leg/acpi/acpi.git acpi-5.1-v11
> 	* [V8 PATCH 0/3] Introduce ACPI support for ahci_platform driver
> 	  (https://lkml.org/lkml/2015/3/30/729)
>
>   drivers/acpi/acpi_platform.c |  5 +++-
>   drivers/acpi/scan.c          | 56 ++++++++++++++++++++++++++++++++++++++++++++
>   2 files changed, 60 insertions(+), 1 deletion(-)
>
> diff --git a/drivers/acpi/acpi_platform.c b/drivers/acpi/acpi_platform.c
> index 1284138..47e37a8 100644
> --- a/drivers/acpi/acpi_platform.c
> +++ b/drivers/acpi/acpi_platform.c
> @@ -108,9 +108,12 @@ 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 {
> +		arch_setup_dma_ops(&pdev->dev, 0, 0, NULL,
> +				   is_device_dma_coherent(&adev->dev));
>   		dev_dbg(&adev->dev, "created platform device %s\n",
>   			dev_name(&pdev->dev));
> +	}
>
>   	kfree(resources);
>   	return pdev;
> diff --git a/drivers/acpi/scan.c b/drivers/acpi/scan.c
> index 230ec983..4d81c55 100644
> --- a/drivers/acpi/scan.c
> +++ b/drivers/acpi/scan.c
> @@ -11,6 +11,7 @@
>   #include <linux/kthread.h>
>   #include <linux/dmi.h>
>   #include <linux/nls.h>
> +#include <linux/dma-mapping.h>
>
>   #include <asm/pgtable.h>
>
> @@ -57,6 +58,9 @@ struct acpi_device_bus_id{
>   	struct list_head node;
>   };
>
> +static int acpi_bus_type_and_status(acpi_handle handle, int *type,
> +				    unsigned long long *sta);
> +
>   void acpi_scan_lock_acquire(void)
>   {
>   	mutex_lock(&acpi_scan_lock);
> @@ -1327,6 +1331,51 @@ void acpi_bus_put_acpi_device(struct acpi_device *adev)
>   	put_device(&adev->dev);
>   }
>
> +/**
> + * acpi_check_and_set_coherency - check and set cache coherency of a device
> + * @device: ACPI device
> + *
> + * Search a device and its parents for a _CCA method and calculate
> + * the effective coherency property of the device. If found, set up
> + * appropriate dma_ops for the device.
> + *
> + * Note From ACPIv5.1 spec:
> + * If used _CCA must be included under all bus-master-capable
> + * devices defined as children of \_SB. The value of _CCA is inherited
> + * by all descendants of these devices, so it need not be repeated for
> + * their children devices and will be ignored by OSPM if it is provided
> + * there.
> + */
> +static int acpi_check_and_set_coherency(struct acpi_device *device)
> +{
> +	acpi_status status;
> +	unsigned long long eff_cca = ~0ULL;
> +	acpi_handle tmp, handle = device->handle;
> +
> +	/* Calculate effective _CCA from all parents of this device */
> +	do {
> +		unsigned long cca;
> +
> +		status = acpi_get_handle(handle, "_CCA", &tmp);
> +		if (ACPI_SUCCESS(status)) {
> +			status = acpi_evaluate_integer(tmp, "_CCA", NULL, &cca);
> +			/* Other values besides 0 and 1 are reserved */
> +			if (ACPI_FAILURE(status) || (cca != 0 && cca != 1))
> +				return -EINVAL;
> +			eff_cca = cca;
> +		}
> +
> +		status = acpi_get_parent(handle, &handle);
> +		if (ACPI_FAILURE(status))
> +			break;
> +	} while (ACPI_SUCCESS(status));

Having the above portion/loop be a separate function that could be
callable by others (such as drivers) would be good so that both the
ACPI code and the driver code look for the _CCA attribute the same
way - similar to the of_dma_is_coherent function.

Possibly even work this into a a generic call like the device property
functions (even though this doesn't fall into the _DSD area for ACPI) -
where for device tree the of_dma_is_coherent call is made and for ACPI
an acpi_dma_is_coherent call is made - device_dma_is_coherent?

Thanks,
Tom

> +
> +	if (eff_cca != ~0ULL)
> +		arch_setup_dma_ops(&device->dev, 0, 0, NULL, eff_cca);
> +
> +	return 0;
> +}
> +
>   int acpi_device_add(struct acpi_device *device,
>   		    void (*release)(struct device *))
>   {
> @@ -1398,6 +1447,13 @@ int acpi_device_add(struct acpi_device *device,
>   		device->dev.parent = &device->parent->dev;
>   	device->dev.bus = &acpi_bus_type;
>   	device->dev.release = release;
> +
> +	result = acpi_check_and_set_coherency(device);
> +	if (result) {
> +		dev_err(&device->dev, "Error getting device _CCA information\n");
> +		goto err;
> +	}
> +
>   	result = device_add(&device->dev);
>   	if (result) {
>   		dev_err(&device->dev, "Error registering device\n");
>