On 5/19/26 13:58, Alex Bennée wrote: > Lucas Amaral <[email protected]> writes: > >> Set blob_alignment to the host page size so the guest kernel can >> align blob BAR offsets accordingly, avoiding alignment mismatches >> with the VMM's memory mapping API (e.g., hv_vm_map on macOS). >> >> Guest-side kernel support (drm_mm alignment) is pending upstream >> Linux merge. >> >> Signed-off-by: Lucas Amaral <[email protected]> >> --- >> hw/display/virtio-gpu.c | 8 ++++++++ >> 1 file changed, 8 insertions(+) >> >> diff --git a/hw/display/virtio-gpu.c b/hw/display/virtio-gpu.c >> index de7a86a7..44ae5034 100644 >> --- a/hw/display/virtio-gpu.c >> +++ b/hw/display/virtio-gpu.c >> @@ -1532,6 +1532,14 @@ void virtio_gpu_device_realize(DeviceState *qdev, >> Error **errp) >> #endif >> } >> >> + /* >> + * TODO: guest-side F_BLOB_ALIGNMENT support pending upstream Linux >> merge >> + * (Sergio Lopez's patches, Nov 2025). Until merged, the guest won't >> + * negotiate this feature. The host advertises it per OASIS virtio spec. >> + */ >> + g->parent_obj.virtio_config.blob_alignment = >> + cpu_to_le32(qemu_real_host_page_size()); >> + > > Is it merged yet?
Not merged, I actually missed the v2 patches Sergio sent couple weeks ago [1], thanks for bringing it up. In general, it's enough to have the field defined in the virtio spec, there should be no need to wait for the kernel patch to land. [1] https://lore.kernel.org/dri-devel/[email protected]/ -- Best regards, Dmitry
