Re: [PATCH 00/20] iommu: Refactoring domain allocation interface

From: Baolu Lu
Date: Thu May 30 2024 - 22:54:46 EST


On 5/31/24 1:59 AM, Robin Murphy wrote:
On 29/05/2024 6:32 am, Lu Baolu wrote:
The IOMMU subsystem has undergone some changes, including the removal
of iommu_ops from the bus structure. Consequently, the existing domain
allocation interface, which relies on a bus type argument, is no longer
relevant:

     struct iommu_domain *iommu_domain_alloc(struct bus_type *bus)

This series is designed to refactor the use of this interface. It
proposes two new interfaces to replace iommu_domain_alloc():

- iommu_user_domain_alloc(): This interface is intended for allocating
   iommu domains managed by userspace for device passthrough scenarios,
   such as those used by iommufd, vfio, and vdpa. It clearly indicates
   that the domain is for user-managed device DMA.

   If an IOMMU driver does not implement iommu_ops->domain_alloc_user,
   this interface will rollback to the generic paging domain allocation.

- iommu_paging_domain_alloc(): This interface is for allocating iommu
   domains managed by kernel drivers for kernel DMA purposes. It takes a
   device pointer as a parameter, which better reflects the current
   design of the IOMMU subsystem.

The majority of device drivers currently using iommu_domain_alloc() do
so to allocate a domain for a specific device and then attach that
domain to the device. These cases can be straightforwardly migrated to
the new interfaces.

Ooh, nice! This was rising back up my to-do list as well, but I concur it's rather more straightforward than my version that did devious things to keep the iommu_domain_alloc() name...

However, there are some drivers with more complex use cases that do
not fit neatly into this new scheme. For example:

$ git grep "= iommu_domain_alloc"
arch/arm/mm/dma-mapping.c:      mapping->domain = iommu_domain_alloc(bus);

This one's simple enough, the refactor just needs to go one step deeper. I've just rebased and pushed my old patch for that, if you'd like it [1].

Great! With this change, we can safely replace iommu_domain_alloc().

diff --git a/arch/arm/mm/dma-mapping.c b/arch/arm/mm/dma-mapping.c
index 52f9c56cc3cb..88c2d68a69c9 100644
--- a/arch/arm/mm/dma-mapping.c
+++ b/arch/arm/mm/dma-mapping.c
@@ -1585,9 +1585,11 @@ arm_iommu_create_mapping(struct device *dev, dma_addr_t base, u64 size)

spin_lock_init(&mapping->lock);

- mapping->domain = iommu_domain_alloc(dev->bus);
- if (!mapping->domain)
+ mapping->domain = iommu_paging_domain_alloc(dev);
+ if (IS_ERR(mapping->domain)) {
+ err = PTR_ERR(mapping->domain);
goto err4;
+ }

kref_init(&mapping->kref);
return mapping;


drivers/gpu/drm/rockchip/rockchip_drm_drv.c:    private->domain = iommu_domain_alloc(private->iommu_dev->bus);

Both this one and usnic_uiom_alloc_pd() should be OK - back when I did all the figuring out to clean up iommu_present(), I specifically reworked them into "dev->bus" style as a reminder that it *is* supposed to be the right device for doing this with, even if the attach is a bit more distant.

Yeah! I will cleanup these two in the next version.


drivers/gpu/drm/tegra/drm.c:            tegra->domain = iommu_domain_alloc(&platform_bus_type);

This is the tricky one, where the device to hand may *not* be the right device for IOMMU API use [2]. FWIW my plan was to pull the "walk the platform bus to find any IOMMU-mapped device" trick into this code and use it both to remove the final iommu_present() and for a device-based domain allocation.

I am not familiar with this driver, so the solution you mentioned above
is the best option I can think of for now. I will incorporate this into
the next version.


drivers/infiniband/hw/usnic/usnic_uiom.c:       pd->domain = domain = iommu_domain_alloc(dev->bus);

This series leave those cases unchanged and keep iommu_domain_alloc()
for their usage. But new drivers should not use it anymore.

I'd certainly be keen for it to be gone ASAP, since I'm seeing increasing demand for supporting multiple IOMMU drivers, and this is the last bus-based thing standing in the way of that.

Agreed. With all iommu_domain_alloc() removed, iommu_domain_alloc()
could be dropped.


Thanks,
Robin.

[1] https://gitlab.arm.com/linux-arm/linux-rm/-/commit/f048cc6a323d8641898025ca96071df7cbe8bd52
[2] https://lore.kernel.org/linux-iommu/add31812-50d5-6cb0-3908-143c523abd37@xxxxxxxxxxxxx/

Best regards,
baolu