Re: [RFC PATCH v4 0/7] vfio-pci: Allow to mmap sub-page MMIO BARs and MSI-X table

[Yongji Xie <xyjxie@linux.vnet.ibm.com>]

On 2016/3/16 22:10, Bjorn Helgaas wrote:
> On Wed, Mar 16, 2016 at 06:51:56PM +0800, Yongji Xie wrote:
>> Ping.
> This is mainly VFIO stuff, and Alex had some security concerns, so I'm
> not going to spend much time looking at this until he's satisfied.
>
> When I do, I'll be looking hard at the resource_alignment kernel
> parameter.  I'm opposed to kernel parameters in general because
> they're very difficult for users to use correctly, and they lead to
> kernel code paths that are rarely tested and hard to maintain.  So
> I'll be looking for an excuse to reject changes in that area.
>
> The changelog for 2/7 says it "replaces IORESOURCE_STARTALIGN with
> IORESOURCE_WINDOW."  But even a glance at the patch itself shows that
> IORESOURCE_WINDOW is *added* to some places, and it doesn't *replace*
> IORESOURCE_STARTALIGN.

There is a problem with my statement. I mean we can use
IORESOURCE_WINDOW to identify bridge resources instead of
IORESOURCE_STARTALIGN here.

> The changelog for 4/7 says:
>
>    This is because vfio will not allow to passthrough one BAR's mmio
>    page which may be shared with other BARs.  To solve this performance
>    issue ...
>
> with no mention at all of the actual *reason* vfio doesn't allow that
> passthrough.  If I understand correctly, that reason has to do with
> security, so your justification must be much stronger than "solving a
> performance issue."

OK. I will try to make my justification become stronger.

Thanks,
Yongji Xie