Re: [Qemu-devel] [PATCH] exec: fix regression by making system-memory region UINT64_MAX size

[Marcel Apfelbaum <marcel.a@redhat.com>]

On Thu, 2013-11-07 at 22:48 +0100, Laszlo Ersek wrote:
> On 11/07/13 22:24, Marcel Apfelbaum wrote:
> > On Thu, 2013-11-07 at 22:12 +0100, Laszlo Ersek wrote:
> 
> >>   adding subregion 'pci-hole' to region 'system' at offset 60000000
> >>   warning: subregion collision 60000000/a0000000 (pci-hole) vs ffe00000/200000 (system.flash)
> > Thank you Laszlo for the detailed info!
> > I think the problem is right above. Why pci-hole and system.flash collide?
> > IMHO we should not play with priorities here, better solve the collision.
> 
> pc_init1()
>   pc_memory_init()
>     pc_system_firmware_init()
>       pc_system_flash_init() <---- sets base address to
>                                    0x100000000ULL - flash_size
>         pflash_cfi01_register()
>           sysbus_mmio_map()
>             sysbus_mmio_map_common()
>               memory_region_add_subregion()
>   i440fx_init()
>     memory_region_init_alias("pci-hole")
> 
> pc_init1() passes
> 
>     0x100000000ULL - below_4g_mem_size
> 
> to i440fx_init() as "pci_hole_size", which is then used as the size of
> the "pci-hole" alias.
> 
> We should probably subtract the size of the flash from this, but I don't
> know how to do that "elegantly". Yet another (output) parameter for
> pc_memory_init()? Blech.
> 
> Or look up the end address of "system.flash" by name?
Both seems ugly to me...
As Peter Maydell pointed out, the pci-hole belongs to the pc model,
so we can actually change its priority without affecting other
architectures.
 
> 
> Thanks
> Laszlo
>