search for: sysmmus

...ven master device. It is usually referenced by 'owner' - * pointer. -*/ -struct exynos_iommu_owner { - struct list_head controllers; /* list of sysmmu_drvdata.owner_node */ - struct iommu_domain *domain; /* domain this device is attached */ - struct mutex rpm_lock; /* for runtime pm of all sysmmus */ -}; - /* * This structure exynos specific generalization of struct iommu_domain. * It contains list of SYSMMU controllers from all master devices, which has @@ -271,13 +259,23 @@ struct sysmmu_drvdata { bool active; /* current status */ struct exynos_iommu_domain *domain; /* domain we...

...all-back, no? > I tried to move all the initialization from xlate() to device_probe(), > but such approach doesn't work. device_probe() is exynos_sysmmu_probe(), then yes, this is called before any of the xlate() calls are made. Would it work to keep the iommu_device structures in the sysmmus and also create them for the owners? This isn't really a nice solution but should work the the IOMMU driver until there is a better way to fix this. Regards, Joerg

Hi, here is a patch-set to remove all calls of iommu_group_get_for_dev() from the IOMMU drivers and move the per-device group setup and default domain allocation into the IOMMU core code. This eliminates some ugly back and forth between IOMMU core code and the IOMMU drivers, where the driver called iommu_group_get_for_dev() which itself called back into the driver. The patch-set started as a

Hi, here is a patch-set to remove all calls of iommu_group_get_for_dev() from the IOMMU drivers and move the per-device group setup and default domain allocation into the IOMMU core code. This eliminates some ugly back and forth between IOMMU core code and the IOMMU drivers, where the driver called iommu_group_get_for_dev() which itself called back into the driver. The patch-set started as a