On Tue, Mar 10, 2020 at 10:12:25AM +0100, Joerg Roedel wrote:
> From: Joerg Roedel <jroedel@xxxxxxx>
>
> Make use of dev_iommu_priv_set/get() functions.
>
> Signed-off-by: Joerg Roedel <jroedel@xxxxxxx>
> ---
> drivers/iommu/ipmmu-vmsa.c | 7 ++-----
> 1 file changed, 2 insertions(+), 5 deletions(-)
>
> diff --git a/drivers/iommu/ipmmu-vmsa.c b/drivers/iommu/ipmmu-vmsa.c
> index ecb3f9464dd5..310cf09feea3 100644
> --- a/drivers/iommu/ipmmu-vmsa.c
> +++ b/drivers/iommu/ipmmu-vmsa.c
> @@ -89,9 +89,7 @@ static struct ipmmu_vmsa_domain *to_vmsa_domain(struct iommu_domain *dom)
>
> static struct ipmmu_vmsa_device *to_ipmmu(struct device *dev)
> {
> - struct iommu_fwspec *fwspec = dev_iommu_fwspec_get(dev);
> -
> - return fwspec ? fwspec->iommu_priv : NULL;
> + return dev_iommu_priv_get(dev);
The removal of the fwspec NULL check was worrying me a little. Now any
user of to_ipmmu() directly dereferences dev->iommu->priv where they
previously tested first whether dev->fwspec was set. But I didn't find
anything that could go wrong, and the resulting code looks better.
Reviewed-by: Jean-Philippe Brucker <jean-philippe@xxxxxxxxxx>
> }
>
> #define TLB_LOOP_TIMEOUT 100 /* 100us */
> @@ -727,14 +725,13 @@ static phys_addr_t ipmmu_iova_to_phys(struct iommu_domain *io_domain,
> static int ipmmu_init_platform_device(struct device *dev,
> struct of_phandle_args *args)
> {
> - struct iommu_fwspec *fwspec = dev_iommu_fwspec_get(dev);
> struct platform_device *ipmmu_pdev;
>
> ipmmu_pdev = of_find_device_by_node(args->np);
> if (!ipmmu_pdev)
> return -ENODEV;
>
> - fwspec->iommu_priv = platform_get_drvdata(ipmmu_pdev);
> + dev_iommu_priv_set(dev, platform_get_drvdata(ipmmu_pdev));
>
> return 0;
> }
> --
> 2.17.1
>
