Re: [PATCH 1/1] virtio-pci: fix memory_region_find for VirtIOPCIRegion's MR

[David Hildenbrand <david@redhat.com>]

On 30.09.24 16:00, Michael S. Tsirkin wrote:
> On Tue, Sep 24, 2024 at 02:31:20PM +0200, David Hildenbrand wrote:
>> On 24.09.24 03:11, Gao Shiyuan wrote:
>>
>> Make sure to version your patch series. For example, via
>> 	$ git format-patch -v1 ...
>>
>>> As shown below, if a virtio PCI device is attached under a pci-bridge, the MR
>>> of VirtIOPCIRegion does not belong to any address space. So memory_region_find
>>> cannot be used to search for this MR.
>>
>> I'm starting to wonder if memory_region_find() is really the right fun
>>
>>>
>>> Introduce the virtio-pci and pci_bridge_pci address spaces to solve this problem.
>>>
>>> Before:
>>> memory-region: pci_bridge_pci
>>>     0000000000000000-ffffffffffffffff (prio 0, i/o): pci_bridge_pci
>>>       00000000fe200000-00000000fe200fff (prio 1, i/o): virtio-blk-pci-msix
>>>         00000000fe200000-00000000fe20016f (prio 0, i/o): msix-table
>>>         00000000fe200800-00000000fe200807 (prio 0, i/o): msix-pba
>>>       000000a000400000-000000a000403fff (prio 1, i/o): virtio-pci
>>>         000000a000400000-000000a000400fff (prio 0, i/o): virtio-pci-common-virtio-blk
>>>         000000a000401000-000000a000401fff (prio 0, i/o): virtio-pci-isr-virtio-blk
>>>         000000a000402000-000000a000402fff (prio 0, i/o): virtio-pci-device-virtio-blk
>>>         000000a000403000-000000a000403fff (prio 0, i/o): virtio-pci-notify-virtio-blk
>>>
>>> After:
>>> address-space: pci_bridge_pci
>>>     0000000000000000-ffffffffffffffff (prio 0, i/o): pci_bridge_pci
>>>       00000000fe200000-00000000fe200fff (prio 1, i/o): virtio-blk-pci-msix
>>>         00000000fe200000-00000000fe20016f (prio 0, i/o): msix-table
>>>         00000000fe200800-00000000fe200807 (prio 0, i/o): msix-pba
>>>       000000a000400000-000000a000403fff (prio 1, i/o): virtio-pci
>>>         000000a000400000-000000a000400fff (prio 0, i/o): virtio-pci-common-virtio-blk
>>>         000000a000401000-000000a000401fff (prio 0, i/o): virtio-pci-isr-virtio-blk
>>>         000000a000402000-000000a000402fff (prio 0, i/o): virtio-pci-device-virtio-blk
>>>         000000a000403000-000000a000403fff (prio 0, i/o): virtio-pci-notify-virtio-blk
>>>
>>> address-space: virtio-pci
>>>     000000a000400000-000000a000403fff (prio 1, i/o): virtio-pci
>>>       000000a000400000-000000a000400fff (prio 0, i/o): virtio-pci-common-virtio-blk
>>>       000000a000401000-000000a000401fff (prio 0, i/o): virtio-pci-isr-virtio-blk
>>>       000000a000402000-000000a000402fff (prio 0, i/o): virtio-pci-device-virtio-blk
>>>       000000a000403000-000000a000403fff (prio 0, i/o): virtio-pci-notify-virtio-blk
>>>
>>> Resolves: https://gitlab.com/qemu-project/qemu/-/issues/2576
>>> Fixes: ffa8a3e ("virtio-pci: Add lookup subregion of VirtIOPCIRegion MR")
>>
>> Commit id is not unique. Use 12 digits please.
>>
>> I'm still not quite sure if memory_region_find() is really the right thing
>> to use here, but I'm no expert on that so I'm hoping virtio/PCI people can
>> review.
> 
> I donnu, what would you use?

I assume there is currently not better alternative. Scanning all address 
spaces to find+walk a flatview sounds quite wasteful during each 
virtio_address_space_lookup ...

So this change here looks ok to me for the time being.

-- 
Cheers,

David / dhildenb