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 12:27 -0800, Jordan Justen wrote:
> On Sun, Nov 3, 2013 at 12:48 PM, Marcel Apfelbaum <marcel.a@redhat.com> wrote:
> > The commit:
> >
> > Commit: a53ae8e934cd54686875b5bcfc2f434244ee55d6
> > Author: Marcel Apfelbaum <marcel.a@redhat.com>
> > Date:   Mon Sep 16 11:21:16 2013 +0300
> >
> >     hw/pci: partially handle pci master abort
> >
> > introduced a regression on make check:
> 
> Laszlo pointed out that my OVMF flash support series was not working
> with QEMU master. It was working with QEMU 1.6.0. I then bisected the
> issue to this commit. It seems this commit regresses -pflash support
> for both KVM and non-KVM modes.
> 
> Can you reproduce the issue this with command?
> x86_64-softmmu/qemu-system-x86_64 -pflash pc-bios/bios.bin
> (with or without adding -enable-kvm)
Thanks Jordan for pointing this out.
This patch revealed a lot of hidden issues inside qemu ...
I succeeded to reproduce the issue.

I saw that get_page_addr_code associated address 0xfffffff0
(after calling io_tlb_to_region) with the new master abort region.
It breaks because memory_region_is_unassigned returns true for the new region.
It is interesting what happened before this region was added.

I will investigate this issue,
Thanks,
Marcel

 

> I tried adding this patch ("exec: fix regression by making
> system-memory region UINT64_MAX size") and it did not help the -pflash
> regression.
> 
> -Jordan