Hi Xingang, On 3/25/21 8:22 AM, Wang Xingang wrote: > From: Xingang Wang <wangxinga...@huawei.com> > > The idmap of smmuv3 and root complex covers the whole RID space for now, > this patch add explicit idmap info according to root bus number range. > This add smmuv3 idmap for certain bus which has enabled the iommu property. > > Signed-off-by: Xingang Wang <wangxinga...@huawei.com> > Signed-off-by: Jiahui Cen <cenjia...@huawei.com> > --- > hw/arm/virt-acpi-build.c | 103 ++++++++++++++++++++++++++++++--------- > 1 file changed, 81 insertions(+), 22 deletions(-) > > diff --git a/hw/arm/virt-acpi-build.c b/hw/arm/virt-acpi-build.c > index f5a2b2d4cb..5491036c86 100644 > --- a/hw/arm/virt-acpi-build.c > +++ b/hw/arm/virt-acpi-build.c > @@ -44,6 +44,7 @@ > #include "hw/acpi/tpm.h" > #include "hw/pci/pcie_host.h" > #include "hw/pci/pci.h" > +#include "hw/pci/pci_bus.h" > #include "hw/pci-host/gpex.h" > #include "hw/arm/virt.h" > #include "hw/mem/nvdimm.h" > @@ -237,6 +238,41 @@ static void acpi_dsdt_add_tpm(Aml *scope, > VirtMachineState *vms) > aml_append(scope, dev); > } > > +typedef > +struct AcpiIortMapping { > + AcpiIortIdMapping idmap; > + bool iommu; > +} AcpiIortMapping; > + > +/* For all PCI host bridges, walk and insert DMAR scope */ this comment should rather be in the caller also DMAR is not the ARM vocable.
I would add the comment for this function: build the ID mapping for aa given PCI host bridge > +static int > +iort_host_bridges(Object *obj, void *opaque) > +{ > + GArray *map_blob = opaque; > + AcpiIortMapping map; > + AcpiIortIdMapping *idmap = &map.idmap; > + int bus_num, max_bus; > + > + if (object_dynamic_cast(obj, TYPE_PCI_HOST_BRIDGE)) { > + PCIBus *bus = PCI_HOST_BRIDGE(obj)->bus; > + > + if (bus) { > + bus_num = pci_bus_num(bus); > + max_bus = pci_root_bus_max_bus(bus); > + > + idmap->input_base = cpu_to_le32(bus_num << 8); > + idmap->id_count = cpu_to_le32((max_bus - bus_num + 1) << 8); > + idmap->output_base = cpu_to_le32(bus_num << 8); > + idmap->flags = cpu_to_le32(0); > + > + map.iommu = pci_root_bus_has_iommu(bus); if iommu is not set, we don't need to populate the idmap and we may even directly continue, ie. not add the element the map_bap_blob, no? > + g_array_append_val(map_blob, map); > + } > + } > + > + return 0; > +} > + > static void > build_iort(GArray *table_data, BIOSLinker *linker, VirtMachineState *vms) > { > @@ -247,6 +283,21 @@ build_iort(GArray *table_data, BIOSLinker *linker, > VirtMachineState *vms) > AcpiIortSmmu3 *smmu; > size_t node_size, iort_node_offset, iort_length, smmu_offset = 0; > AcpiIortRC *rc; > + int smmu_mapping_count; > + GArray *map_blob = g_array_new(false, true, sizeof(AcpiIortMapping)); > + AcpiIortMapping *map; > + > + /* pci_for_each_bus(vms->bus, insert_map, map_blob); */ comment to be removed > + object_child_foreach_recursive(object_get_root(), > + iort_host_bridges, map_blob); > + > + smmu_mapping_count = 0; > + for (int i = 0; i < map_blob->len; i++) { > + map = &g_array_index(map_blob, AcpiIortMapping, i); > + if (map->iommu) { > + smmu_mapping_count++; > + } > + } > > iort = acpi_data_push(table_data, sizeof(*iort)); > > @@ -280,13 +331,13 @@ build_iort(GArray *table_data, BIOSLinker *linker, > VirtMachineState *vms) > > /* SMMUv3 node */ > smmu_offset = iort_node_offset + node_size; > - node_size = sizeof(*smmu) + sizeof(*idmap); > + node_size = sizeof(*smmu) + sizeof(*idmap) * smmu_mapping_count; > iort_length += node_size; > smmu = acpi_data_push(table_data, node_size); > > smmu->type = ACPI_IORT_NODE_SMMU_V3; > smmu->length = cpu_to_le16(node_size); > - smmu->mapping_count = cpu_to_le32(1); > + smmu->mapping_count = cpu_to_le32(smmu_mapping_count); > smmu->mapping_offset = cpu_to_le32(sizeof(*smmu)); > smmu->base_address = cpu_to_le64(vms->memmap[VIRT_SMMU].base); > smmu->flags = cpu_to_le32(ACPI_IORT_SMMU_V3_COHACC_OVERRIDE); > @@ -295,23 +346,28 @@ build_iort(GArray *table_data, BIOSLinker *linker, > VirtMachineState *vms) > smmu->gerr_gsiv = cpu_to_le32(irq + 2); > smmu->sync_gsiv = cpu_to_le32(irq + 3); > > - /* Identity RID mapping covering the whole input RID range */ > - idmap = &smmu->id_mapping_array[0]; > - idmap->input_base = 0; > - idmap->id_count = cpu_to_le32(0xFFFF); > - idmap->output_base = 0; > - /* output IORT node is the ITS group node (the first node) */ > - idmap->output_reference = cpu_to_le32(iort_node_offset); > + for (int i = 0, j = 0; i < map_blob->len; i++) { > + map = &g_array_index(map_blob, AcpiIortMapping, i); > + > + if (!map->iommu) { > + continue; > + } > + > + idmap = &smmu->id_mapping_array[j++]; > + *idmap = map->idmap; > + /* output IORT node is the ITS group node (the first node) */ > + idmap->output_reference = cpu_to_le32(iort_node_offset); > + } > } > > /* Root Complex Node */ > - node_size = sizeof(*rc) + sizeof(*idmap); > + node_size = sizeof(*rc) + sizeof(*idmap) * map_blob->len; > iort_length += node_size; > rc = acpi_data_push(table_data, node_size); > > rc->type = ACPI_IORT_NODE_PCI_ROOT_COMPLEX; > rc->length = cpu_to_le16(node_size); > - rc->mapping_count = cpu_to_le32(1); > + rc->mapping_count = cpu_to_le32(map_blob->len); > rc->mapping_offset = cpu_to_le32(sizeof(*rc)); > > /* fully coherent device */ > @@ -319,20 +375,23 @@ build_iort(GArray *table_data, BIOSLinker *linker, > VirtMachineState *vms) > rc->memory_properties.memory_flags = 0x3; /* CCA = CPM = DCAS = 1 */ > rc->pci_segment_number = 0; /* MCFG pci_segment */ > > - /* Identity RID mapping covering the whole input RID range */ > - idmap = &rc->id_mapping_array[0]; > - idmap->input_base = 0; > - idmap->id_count = cpu_to_le32(0xFFFF); > - idmap->output_base = 0; > + for (int i = 0; i < map_blob->len; i++) { > + map = &g_array_index(map_blob, AcpiIortMapping, i); > + idmap = &rc->id_mapping_array[i]; > > - if (vms->iommu == VIRT_IOMMU_SMMUV3) { > - /* output IORT node is the smmuv3 node */ > - idmap->output_reference = cpu_to_le32(smmu_offset); > - } else { > - /* output IORT node is the ITS group node (the first node) */ > - idmap->output_reference = cpu_to_le32(iort_node_offset); > + *idmap = map->idmap; We we descibe the link between SMMU SID and ITS deviceids, shouldn't we leave the whole range? Even if an RID is not translated, it may use MSIs? > + > + if (vms->iommu == VIRT_IOMMU_SMMUV3 && map->iommu) { > + /* output IORT node is the smmuv3 node */ > + idmap->output_reference = cpu_to_le32(smmu_offset); > + } else { > + /* output IORT node is the ITS group node (the first node) */ > + idmap->output_reference = cpu_to_le32(iort_node_offset); is that case we have a direct mapping between RC and ITS (no instantiated SMMU). Shouldn't the whole RID range being used in that case? Thanks Eric > + } > } > > + g_array_free(map_blob, true); > + > /* > * Update the pointer address in case table_data->data moves during above > * acpi_data_push operations. >