Re: [Qemu-devel] [PATCH V3 0/3] hw/pcie: Multi-root support for Q35

[Laszlo Ersek <lersek@redhat.com>]

Hello Marcel,

On 11/26/15 17:00, Marcel Apfelbaum wrote:
> Note:
> I took the liberty to CC all the reviewers that took their time
> and had a look on the previous version, thanks!!
> 
> The PXB host bridge provides a way to have multiple PCI hierarchies (PCI root buses).
> This series introduces the pxb-pcie counterpart for PCI Express machines(Currently Q35).
> 
> This approach works because the Root Complexes are exposed to guest as regular
> (legacy) opaque PCI host bridges.
> 
> Tested on Fedora and Windows guests with both Root Ports and PCIe Switches.
> 
> v2 -> v3:
>  Addressed Eduardo Habkost comments:
>  - Added a bus property to PC machines and use it when querying bus 0.
>  Addressed comments from multiple reviewers (Paolo,Markus,Gerd,Michael)
>  - The issue was the backport compatibility when the PXB changes.
>  - Following all the comments I chose:
>    - Leave the PXB intact as it does the job and all its features
>      (including the internal pci bridge) makes sense.
>    - Add a new device that re-uses all the PXB code but is exposed as
>      a different device to guests.
>    - Once the functionality of the new device diverges we will have
>      no problem to separate the code.


I don't think I can productively contribute to the review of this
series, but at least I'll try to follow the comments of others.

Also, your first patch looks like it touches code that is shared by the
i440fx PXB. I think it should cause no change in observable behavior, right?

Should I regression test it with OVMF? If so, that might take... forever. :(

On the other hand, if you have ACPI table dumps from within an i440fx
SeaBIOS Linux guest, both from before and after your QEMU patches, and
those dumps are identical, then that's good evidence against
regressions. (I tend to do such acpidump-based comparisons when messing
with ACPI builder code.)

Thanks (and sorry I can't help more ATM)
Laszlo


> 
> v1 -> v2:
>  Addressed Gerd Hoffmann comments:
>  - Added x-enable-internal-bridge compat property to keep the PCI
>    bridge for older machine to avoid breaking migration.
> 
> Thanks,
> Marcel
> 
> Marcel Apfelbaum (3):
>   hw/acpi: merge pxb adjacent memory/IO ranges
>   hw/pxb: introduce pxb-pcie expander for PCIe machines
>   hw/i386: extend pxb query for all PC machines
> 
>  hw/i386/acpi-build.c                | 126 +++++++++++++++++++++---------------
>  hw/i386/pc.c                        |   2 +-
>  hw/i386/pc_piix.c                   |   1 +
>  hw/i386/pc_q35.c                    |   1 +
>  hw/pci-bridge/pci_expander_bridge.c |  98 +++++++++++++++++++++++-----
>  include/hw/i386/pc.h                |   1 +
>  include/hw/pci/pci.h                |   1 +
>  7 files changed, 163 insertions(+), 67 deletions(-)
>