From: Zhu Yijun <zhuyi...@huawei.com> With kernel 4.11, iommu/smmu will populate the MSI IOVA reserved window and PCI reserved window which has to be excluded from Guest iova allocations.
And on certain HiSilicon platforms (hip06/hip07), the GIC ITS and PCIe RC deviates from the standard implementation will reserve the hw msi regions in the smmu-v3 driver which means these address regions will not be translated. https://www.spinics.net/lists/linux-pci/msg65125.html On such platforms, reserved memory regions will export like this: root:~# cat /sys/kernel/iommu_groups/7/reserved_regions 0x00000000a8800000 0x00000000affeffff reserved 0x00000000c6000000 0x00000000c601ffff msi However, it falls within the Qemu default virtual memory address space. Take a look at hw/arm/virt.c: [VIRT_MEM] = { 0x40000000, RAMLIMIT_BYTES }, . . . memory_region_allocate_system_memory(ram, NULL, "mach-virt.ram", machine->ram_size); memory_region_add_subregion(sysmem, vms->memmap[VIRT_MEM].base, ram); So suppose we allocate 4GB mem to a VM, there is a chance that the reserved regions will get allocated for a Guest VF DMA iova and it will fail. This patchset create holes in the Qemu RAM address space which exclude the sysfs exported regions. Zhu Yijun (5): hw/vfio: Add function for getting reserved_region of device iommu group hw/arm/virt: Enable dynamic generation of guest RAM memory regions hw/arm: add scattered RAM memory region support hw/arm/boot: set fdt size cell of memory node from mem_list hw/arm/virt-acpi-build: Build srat table according to mem_list hw/arm/boot.c | 155 +++++++++++++++++++++++++++++++----------- hw/arm/virt-acpi-build.c | 40 +++++++++-- hw/arm/virt.c | 120 ++++++++++++++++++++++++++++++-- hw/vfio/common.c | 67 ++++++++++++++++++ hw/vfio/pci.c | 2 + hw/vfio/platform.c | 2 + include/exec/memory.h | 7 ++ include/hw/arm/arm.h | 1 + include/hw/arm/virt.h | 1 + include/hw/vfio/vfio-common.h | 3 + 10 files changed, 347 insertions(+), 51 deletions(-) -- 1.8.3.1
