Hi Joao, On 6/22/23 23:48, Joao Martins wrote: > From: Yi Liu <[email protected]> > > Add a pci_setup_iommu_ops() that uses a newly added structure > (PCIIOMMUOps) instead of using PCIIOMMUFunc. The old pci_setup_iommu() > that uses PCIIOMMUFunc is still kept for other IOMMUs to get an > an address space for a PCI device in vendor specific way. double 'an' > > In preparation to expand to supplying vIOMMU attributes, add a > alternate helper pci_setup_iommu_ops() to setup the PCI device IOMMU. > For now the PCIIOMMUOps just defines the address_space, but it will > be extended to have another callback. > > Signed-off-by: Yi Liu <[email protected]> > [joao: Massage commit message and subject, and make it a complementary > rather than changing every single consumer of pci_setup_iommu()] > Signed-off-by: Joao Martins <[email protected]> > --- > v1: https://lore.kernel.org/all/[email protected]/ > --- > include/hw/pci/pci.h | 7 +++++++ > include/hw/pci/pci_bus.h | 1 + > hw/pci/pci.c | 26 +++++++++++++++++++++++--- > 3 files changed, 31 insertions(+), 3 deletions(-) > > diff --git a/include/hw/pci/pci.h b/include/hw/pci/pci.h > index e6d0574a2999..f59aef5a329a 100644 > --- a/include/hw/pci/pci.h > +++ b/include/hw/pci/pci.h > @@ -368,6 +368,13 @@ typedef AddressSpace *(*PCIIOMMUFunc)(PCIBus *, void *, > int); > AddressSpace *pci_device_iommu_address_space(PCIDevice *dev); > void pci_setup_iommu(PCIBus *bus, PCIIOMMUFunc fn, void *opaque); > > +typedef struct PCIIOMMUOps PCIIOMMUOps; > +struct PCIIOMMUOps { > + AddressSpace * (*get_address_space)(PCIBus *bus, > + void *opaque, int32_t devfn); > +}; > +void pci_setup_iommu_ops(PCIBus *bus, const PCIIOMMUOps *iommu_ops, void > *opaque); > + > pcibus_t pci_bar_address(PCIDevice *d, > int reg, uint8_t type, pcibus_t size); > > diff --git a/include/hw/pci/pci_bus.h b/include/hw/pci/pci_bus.h > index 56531759578f..fb770b236d69 100644 > --- a/include/hw/pci/pci_bus.h > +++ b/include/hw/pci/pci_bus.h > @@ -35,6 +35,7 @@ struct PCIBus { > enum PCIBusFlags flags; > PCIIOMMUFunc iommu_fn; > void *iommu_opaque; > + const PCIIOMMUOps *iommu_ops; > uint8_t devfn_min; > uint32_t slot_reserved_mask; > pci_set_irq_fn set_irq; > diff --git a/hw/pci/pci.c b/hw/pci/pci.c > index bf38905b7dc0..4e32c09e81d6 100644 > --- a/hw/pci/pci.c > +++ b/hw/pci/pci.c > @@ -2639,7 +2639,15 @@ AddressSpace *pci_device_iommu_address_space(PCIDevice > *dev) > PCIBus *iommu_bus = bus; > uint8_t devfn = dev->devfn; > > - while (iommu_bus && !iommu_bus->iommu_fn && iommu_bus->parent_dev) { > + /* > + * get_address_space() callback is mandatory when iommu uses > + * pci_setup_iommu_ops(), so needs to ensure its presence in > + * the iommu_bus search. the fact that it is mandatory should also be documented along with the PCIIOMMUOps struct definition and enforced at pci_setup_iommu_ops() removing the need for checking iommu_bus->iommu_ops->get_address_space > + */ > + while (iommu_bus && > + !(iommu_bus->iommu_fn || > + (iommu_bus->iommu_ops && > iommu_bus->iommu_ops->get_address_space)) && > + iommu_bus->parent_dev) { > PCIBus *parent_bus = pci_get_bus(iommu_bus->parent_dev); > > /* > @@ -2678,8 +2686,14 @@ AddressSpace *pci_device_iommu_address_space(PCIDevice > *dev) > > iommu_bus = parent_bus; > } > - if (!pci_bus_bypass_iommu(bus) && iommu_bus && iommu_bus->iommu_fn) { > - return iommu_bus->iommu_fn(bus, iommu_bus->iommu_opaque, devfn); > + if (!pci_bus_bypass_iommu(bus) && iommu_bus) { > + if (iommu_bus->iommu_fn) { > + return iommu_bus->iommu_fn(bus, iommu_bus->iommu_opaque, devfn); > + } else if (iommu_bus->iommu_ops && > + iommu_bus->iommu_ops->get_address_space) { > + return iommu_bus->iommu_ops->get_address_space(bus, > + iommu_bus->iommu_opaque, devfn); > + } > } > return &address_space_memory; > } > @@ -2690,6 +2704,12 @@ void pci_setup_iommu(PCIBus *bus, PCIIOMMUFunc fn, > void *opaque) > bus->iommu_opaque = opaque; > } > > +void pci_setup_iommu_ops(PCIBus *bus, const PCIIOMMUOps *ops, void *opaque) > +{ > + bus->iommu_ops = ops; > + bus->iommu_opaque = opaque; maybe assert if both iommu_fn and iommu_ops are set to make sure the conversion is not done partially? If I understand it correctly either of those 2 ops should be set and not both. > +} > + > static void pci_dev_get_w64(PCIBus *b, PCIDevice *dev, void *opaque) > { > Range *range = opaque;
Thanks Eric
