Hi Peter, On 2/8/22 4:17 PM, Peter Maydell wrote: > On Tue, 8 Feb 2022 at 15:08, Eric Auger <eric.au...@redhat.com> wrote: >> Representing the CRB cmd/response buffer as a standard >> RAM region causes some trouble when the device is used >> with VFIO. Indeed VFIO attempts to DMA_MAP this region >> as usual RAM but this latter does not have a valid page >> size alignment causing such an error report: >> "vfio_listener_region_add received unaligned region". >> To allow VFIO to detect that failing dma mapping >> this region is not an issue, let's use a ram_device >> memory region type instead. > This seems like VFIO's problem to me. There's nothing > that guarantees alignment for memory regions at all, > whether they're RAM, IO or anything else.
VFIO dma maps all the guest RAM. I understand the cmd/response buffer is RAM but does not need to be dma mapped, all the more so it has a bad alignment. By the way the PPI region also has the ram_device type (tpm_ppi.c tpm_ppi_init). In that case, using the ram_device type allows VFIO to discriminate between critical mapping errors and non critical ones. We have no other mean atm. Thanks Eric > >> + s->crb_cmd_buf = qemu_memalign(qemu_real_host_page_size, >> + HOST_PAGE_ALIGN(CRB_CTRL_CMD_SIZE)); >> + >> memory_region_init_io(&s->mmio, OBJECT(s), &tpm_crb_memory_ops, s, >> "tpm-crb-mmio", sizeof(s->regs)); >> - memory_region_init_ram(&s->cmdmem, OBJECT(s), >> - "tpm-crb-cmd", CRB_CTRL_CMD_SIZE, errp); >> + memory_region_init_ram_device_ptr(&s->cmdmem, OBJECT(s), "tpm-crb-cmd", >> + CRB_CTRL_CMD_SIZE, s->crb_cmd_buf); >> + vmstate_register_ram(&s->cmdmem, dev); >> >> memory_region_add_subregion(get_system_memory(), >> TPM_CRB_ADDR_BASE, &s->mmio); >> @@ -309,12 +315,25 @@ static void tpm_crb_realize(DeviceState *dev, Error >> **errp) >> qemu_register_reset(tpm_crb_reset, dev); >> } > As QEMU code goes, this seems much worse than what it replaces. > To have a memory region backed by RAM and migrated in the > usual way, memory_region_init_ram() is the right thing. > > thanks > -- PMM >
