On 11/19/21 09:21, Mark Cave-Ayland wrote: > On 18/11/2021 19:20, Philippe Mathieu-Daudé wrote: > >> Commit 7852a77f598 fixed creating multiple TYPE_ISA_VGA devices, >> generalize the fix to all VGA devices. >> >> See https://gitlab.com/qemu-project/qemu/-/issues/44 >> >> Philippe Mathieu-Daudé (2): >> hw/display: Add Error* handle to vga_common_init() >> hw/display: Do not allow multiple identical VGA devices >> >> hw/display/vga_int.h | 2 +- >> hw/display/ati.c | 4 +++- >> hw/display/cirrus_vga.c | 4 +++- >> hw/display/cirrus_vga_isa.c | 4 +++- >> hw/display/qxl.c | 4 +++- >> hw/display/vga-isa-mm.c | 3 ++- >> hw/display/vga-isa.c | 11 ++--------- >> hw/display/vga-pci.c | 8 ++++++-- >> hw/display/vga.c | 17 ++++++++++++++++- >> hw/display/virtio-vga.c | 4 +++- >> hw/display/vmware_vga.c | 2 +- >> 11 files changed, 43 insertions(+), 20 deletions(-) > > Hi Phil, > > I don't think this is correct for non-ISA devices: for example years ago > I had a PC running Windows 98SE with 2 identical PCI graphics cards > configured in dual-head mode. > > IIRC the BIOS would bring up the first graphics card and configure it to > use the legacy ISA VGA ioports for compatibility, and then once the main > OS drivers loaded both cards were switched to PCI mode and configured > using the BARs as normal.
The problem here is QEMU technical debt, not the hardware. When vga_common_init() calls memory_region_init_ram_nomigrate() with obj=NULL, "vga.vram" is registered as a QOM singleton. Updating it would 1/ require non-QOM devices to be QOM'ified and 2/ break migration unless using HPFM which I don't master.
