Re: [PATCH 1/2] vfio/pci: Fix vfio-pci sub-page MMIO BAR mmaping in live migration

[Alex Williamson <alex.williamson@redhat.com>]

On Fri, 3 Sep 2021 17:36:10 +0800
Kunkun Jiang <jiangkunkun@huawei.com> wrote:

> We expand MemoryRegions of vfio-pci sub-page MMIO BARs to
> vfio_pci_write_config to improve IO performance.
> The MemoryRegions of destination VM will not be expanded
> successful in live migration, because their addresses have
> been updated in vmstate_load_state (vfio_pci_load_config).

What's the call path through vfio_pci_write_config() that you're
relying on to get triggered to enable this and why wouldn't we just
walk all sub-page BARs in vfio_pci_load_config() to resolve the issue
then?  It's my understanding that we do this update in write-config
because it's required that the VM sizes the BAR before using it, which
is not the case when we resume from migration.  Thanks,

Alex
 
> Remove the restriction on base address change in
> vfio_pci_write_config for correct mmapping sub-page MMIO
> BARs. Accroding to my analysis, the remaining parameter
> verification is enough.
> 
> Fixes: c5e2fb3ce4d (vfio: Add save and load functions for VFIO PCI devices)
> Reported-by: Nianyao Tang <tangnianyao@huawei.com>
> Reported-by: Qixin Gan <ganqixin@huawei.com>
> Signed-off-by: Kunkun Jiang <jiangkunkun@huawei.com>
> ---
>  hw/vfio/pci.c | 8 +-------
>  1 file changed, 1 insertion(+), 7 deletions(-)
> 
> diff --git a/hw/vfio/pci.c b/hw/vfio/pci.c
> index e1ea1d8a23..891b211ddf 100644
> --- a/hw/vfio/pci.c
> +++ b/hw/vfio/pci.c
> @@ -1189,18 +1189,12 @@ void vfio_pci_write_config(PCIDevice *pdev,
>          }
>      } else if (ranges_overlap(addr, len, PCI_BASE_ADDRESS_0, 24) ||
>          range_covers_byte(addr, len, PCI_COMMAND)) {
> -        pcibus_t old_addr[PCI_NUM_REGIONS - 1];
>          int bar;
>  
> -        for (bar = 0; bar < PCI_ROM_SLOT; bar++) {
> -            old_addr[bar] = pdev->io_regions[bar].addr;
> -        }
> -
>          pci_default_write_config(pdev, addr, val, len);
>  
>          for (bar = 0; bar < PCI_ROM_SLOT; bar++) {
> -            if (old_addr[bar] != pdev->io_regions[bar].addr &&
> -                vdev->bars[bar].region.size > 0 &&
> +            if (vdev->bars[bar].region.size > 0 &&
>                  vdev->bars[bar].region.size < qemu_real_host_page_size) {
>                  vfio_sub_page_bar_update_mapping(pdev, bar);
>              }