[PATCH v2 9/9] iommu/arm-smmu: add support for non-pci devices

[robin.murphy@arm.com (Robin Murphy)]

On 07/07/15 04:30, Zhen Lei wrote:
> This patch support a master with multiple stream IDs, but doesn't support a
> master behinds more than one SMMUs.

This should probably include a binding documentation update to make it 
clear what values of #iommu-cells the driver supports in what 
circumstances, and that the cells themselves should contain raw stream IDs.

> Signed-off-by: Zhen Lei <thunder.leizhen@huawei.com>
> ---
>   drivers/iommu/arm-smmu-v3.c | 39 +++++++++++++++++++++++++++++++++++++--
>   1 file changed, 37 insertions(+), 2 deletions(-)
>
> diff --git a/drivers/iommu/arm-smmu-v3.c b/drivers/iommu/arm-smmu-v3.c
> index 274d059..582cd1e 100644
> --- a/drivers/iommu/arm-smmu-v3.c
> +++ b/drivers/iommu/arm-smmu-v3.c
> @@ -30,6 +30,7 @@
>   #include <linux/of_address.h>
>   #include <linux/pci.h>
>   #include <linux/platform_device.h>
> +#include <linux/amba/bus.h>
>   #include <linux/of_iommu.h>
>   #include <linux/of_platform.h>
>
> @@ -1928,9 +1929,35 @@ static int arm_smmu_of_xlate(struct device *dev, struct of_phandle_args *args)
>   		return -EINVAL;
>   	}
>
> -	/* We only support PCI, for now */
>   	if (!dev_is_pci(dev)) {
> -		return -ENODEV;
> +		int i;
> +		struct iommu_group *group;
> +
> +		group = iommu_group_get(dev);
> +		if (!group) {
> +			group = iommu_group_alloc();
> +			if (IS_ERR(group)) {
> +				dev_err(dev, "failed to allocate IOMMU group\n");
> +				return PTR_ERR(group);
> +			}
> +
> +			ret = iommu_group_add_device(group, dev);
> +			if (ret) {
> +				dev_err(dev, "failed to add device into IOMMU group\n");
> +				return PTR_ERR(group);

This leaks the group.

> +			}
> +		}
> +		iommu_group_put(group);
> +
> +		for (i = 0; i < args->args_count; i++) {

I'm dubious of the value of looping here - having n>1 #iommu-cells per 
phandle means that every platform device behind one SMMU must have the 
same number of stream IDs, or you still have to have repeated phandles 
for every device with some greater multiple of n stream IDs each, but 
ruling out any device with <n. I'm not sure how realistic that is, and 
whether there's any real benefit beyond saving a handful of bytes in the 
DTB.

> +			ret = arm_smmu_add_device(dev, args->args[i]);
> +			if (ret) {
> +				dev_err(dev,
> +					"failed to add sid=%d into SMMU\n",
> +					args->args[i]);
> +				return ret;
> +			}
> +		}
>   	} else {
>   		u32 sid;
>   		struct device_node *of_root;
> @@ -2725,6 +2752,14 @@ static int __init arm_smmu_init(void)
>   	if (ret)
>   		return ret;
>
> +	if (!iommu_present(&platform_bus_type))
> +		bus_set_iommu(&platform_bus_type, &arm_smmu_ops);
> +
> +#ifdef CONFIG_ARM_AMBA
> +	if (!iommu_present(&amba_bustype))
> +		bus_set_iommu(&amba_bustype, &arm_smmu_ops);
> +#endif
> +
>   	return bus_set_iommu(&pci_bus_type, &arm_smmu_ops);
>   }
>
> --
> 1.8.0
>
>
> _______________________________________________
> iommu mailing list
> iommu at lists.linux-foundation.org
> https://lists.linuxfoundation.org/mailman/listinfo/iommu
>