Re: [Qemu-devel] [PATCH] vga: print friendly error message in case multiple vga devices are added

[Peter Maydell]

On 7 May 2018 at 11:10, Gerd Hoffmann <address@hidden> wrote:
> ... to a virtual machine.  Without that we fail a ramblock register
> sanity check, leading to a abort(), which isn't exactly user friendly.
>
> https://bugzilla.redhat.com/show_bug.cgi?id=1206037
> Signed-off-by: Gerd Hoffmann <address@hidden>
> ---
>  hw/display/vga.c | 8 ++++++++
>  1 file changed, 8 insertions(+)
>
> diff --git a/hw/display/vga.c b/hw/display/vga.c
> index 72181330b8..328b6413ad 100644
> --- a/hw/display/vga.c
> +++ b/hw/display/vga.c
> @@ -2184,6 +2184,14 @@ void vga_common_init(VGACommonState *s, Object *obj, 
> bool global_vmstate)
>      }
>      s->vbe_size_mask = s->vbe_size - 1;
>
> +    if (global_vmstate) {
> +        static int have_vga;
> +        if (have_vga) {
> +            error_report("only one vga device is supported");
> +            exit(1);
> +        }
> +        have_vga = true;
> +    }
>      s->is_vbe_vmstate = 1;
>      memory_region_init_ram_nomigrate(&s->vram, obj, "vga.vram", s->vram_size,
>                             &error_fatal);
> --
> 2.9.3
>

This seems like a bit of an irritating reason to refuse to
allow multiple VGA devices though, since we could make
them work by having the vmstate be not-global. Which is
exactly what it ought to be -- we're only defaulting to
"use global" here for migration compatibility, I think.
Since multiple VGA devices didn't work before, there's
no old configs we need to migrate from, and so I think
we could just say "the second and subsequent VGA devices
get non-global ram vmstate" ?

thanks
-- PMM